c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2024-12-27 03:19:35 +03:00
Code Issues Packages Projects Releases Wiki Activity
a72751a342
mirror_zfs/module/zfs/dmu_recv.c

3810 lines
111 KiB
C
Raw Normal View History

Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
Replace dead opensolaris.org license link The commit replaces all findings of the link: http://www.opensolaris.org/os/licensing with this one: https://opensource.org/licenses/CDDL-1.0 Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tino Reichardt <milky-zfs@mcmilk.de> Closes #13619
2022-07-12 00:16:13 +03:00
* or https://opensource.org/licenses/CDDL-1.0.
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
Remove deduplicated send/receive code Deduplicated send streams (i.e. `zfs send -D` and `zfs receive` of such streams) are deprecated. Deduplicated send streams can be received by first converting them to non-deduplicated with the `zstream redup` command. This commit removes the code for sending and receiving deduplicated send streams. `zfs send -D` will now print a warning, ignore the `-D` flag, and generate a regular (non-deduplicated) send stream. `zfs receive` of a deduplicated send stream will print an error message and fail. The resulting code simplification (especially in the kernel's support for receiving dedup streams) should help enable future performance enhancements. Several new tests are added which leverage `zstream redup`. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Issue #7887 Issue #10117 Issue #10156 Closes #10212
2020-04-23 20:06:57 +03:00
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
* Copyright (c) 2014, Joyent, Inc. All rights reserved.
* Copyright 2014 HybridCluster. All rights reserved.
ZVOLs should not be allowed to have children zfs create, receive and rename can bypass this hierarchy rule. Update both userland and kernel module to prevent this issue and use pyzfs unit tests to exercise the ioctls directly. Note: this commit slightly changes zfs_ioc_create() ABI. This allow to differentiate a generic error (EINVAL) from the specific case where we tried to create a dataset below a ZVOL (ZFS_ERR_WRONG_PARENT). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
2019-02-09 02:44:15 +03:00
* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
Add zstd support to zfs This PR adds two new compression types, based on ZStandard: - zstd: A basic ZStandard compression algorithm Available compression. Levels for zstd are zstd-1 through zstd-19, where the compression increases with every level, but speed decreases. - zstd-fast: A faster version of the ZStandard compression algorithm zstd-fast is basically a "negative" level of zstd. The compression decreases with every level, but speed increases. Available compression levels for zstd-fast: - zstd-fast-1 through zstd-fast-10 - zstd-fast-20 through zstd-fast-100 (in increments of 10) - zstd-fast-500 and zstd-fast-1000 For more information check the man page. Implementation details: Rather than treat each level of zstd as a different algorithm (as was done historically with gzip), the block pointer `enum zio_compress` value is simply zstd for all levels, including zstd-fast, since they all use the same decompression function. The compress= property (a 64bit unsigned integer) uses the lower 7 bits to store the compression algorithm (matching the number of bits used in a block pointer, as the 8th bit was borrowed for embedded block pointers). The upper bits are used to store the compression level. It is necessary to be able to determine what compression level was used when later reading a block back, so the concept used in LZ4, where the first 32bits of the on-disk value are the size of the compressed data (since the allocation is rounded up to the nearest ashift), was extended, and we store the version of ZSTD and the level as well as the compressed size. This value is returned when decompressing a block, so that if the block needs to be recompressed (L2ARC, nop-write, etc), that the same parameters will be used to result in the matching checksum. All of the internal ZFS code ( `arc_buf_hdr_t`, `objset_t`, `zio_prop_t`, etc.) uses the separated _compress and _complevel variables. Only the properties ZAP contains the combined/bit-shifted value. The combined value is split when the compression_changed_cb() callback is called, and sets both objset members (os_compress and os_complevel). The userspace tools all use the combined/bit-shifted value. Additional notes: zdb can now also decode the ZSTD compression header (flag -Z) and inspect the size, version and compression level saved in that header. For each record, if it is ZSTD compressed, the parameters of the decoded compression header get printed. ZSTD is included with all current tests and new tests are added as-needed. Per-dataset feature flags now get activated when the property is set. If a compression algorithm requires a feature flag, zfs activates the feature when the property is set, rather than waiting for the first block to be born. This is currently only used by zstd but can be extended as needed. Portions-Sponsored-By: The FreeBSD Foundation Co-authored-by: Allan Jude <allanjude@freebsd.org> Co-authored-by: Brian Behlendorf <behlendorf1@llnl.gov> Co-authored-by: Sebastian Gottschall <s.gottschall@dd-wrt.com> Co-authored-by: Kjeld Schouten-Lebbing <kjeld@schouten-lebbing.nl> Co-authored-by: Michael Niewöhner <foss@mniewoehner.de> Signed-off-by: Allan Jude <allan@klarasystems.com> Signed-off-by: Allan Jude <allanjude@freebsd.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Sebastian Gottschall <s.gottschall@dd-wrt.com> Signed-off-by: Kjeld Schouten-Lebbing <kjeld@schouten-lebbing.nl> Signed-off-by: Michael Niewöhner <foss@mniewoehner.de> Closes #6247 Closes #9024 Closes #10277 Closes #10278
2020-08-18 20:10:17 +03:00
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
* Copyright (c) 2019 Datto Inc.
* Copyright (c) 2022 Axcient.
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
*/
Reject streams that set ->drr_payloadlen to unreasonably large values In the zstream code, Coverity reported: "The argument could be controlled by an attacker, who could invoke the function with arbitrary values (for example, a very high or negative buffer size)." It did not report this in the kernel. This is likely because the userspace code stored this in an int before passing it into the allocator, while the kernel code stored it in a uint32_t. However, this did reveal a potentially real problem. On 32-bit systems and systems with only 4GB of physical memory or less in general, it is possible to pass a large enough value that the system will hang. Even worse, on Linux systems, the kernel memory allocator is not able to support allocations up to the maximum 4GB allocation size that this allows. This had already been limited in userspace to 64MB by `ZFS_SENDRECV_MAX_NVLIST`, but we need a hard limit in the kernel to protect systems. After some discussion, we settle on 256MB as a hard upper limit. Attempting to receive a stream that requires more memory than that will result in E2BIG being returned to user space. Reported-by: Coverity (CID-1529836) Reported-by: Coverity (CID-1529837) Reported-by: Coverity (CID-1529838) Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Closes #14285
2023-01-24 00:16:22 +03:00
#include <sys/arc.h>
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
#include <sys/spa_impl.h>
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
#include <sys/dmu.h>
#include <sys/dmu_impl.h>
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
#include <sys/dmu_send.h>
#include <sys/dmu_recv.h>
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
#include <sys/dmu_tx.h>
#include <sys/dbuf.h>
#include <sys/dnode.h>
#include <sys/zfs_context.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_traverse.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_synctask.h>
#include <sys/zfs_ioctl.h>
#include <sys/zap.h>
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
#include <sys/zvol.h>
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
#include <sys/zio_checksum.h>
#include <sys/zfs_znode.h>
#include <zfs_fletcher.h>
#include <sys/avl.h>
#include <sys/ddt.h>
#include <sys/zfs_onexit.h>
#include <sys/dsl_destroy.h>
#include <sys/blkptr.h>
#include <sys/dsl_bookmark.h>
#include <sys/zfeature.h>
#include <sys/bqueue.h>
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
#include <sys/objlist.h>
#ifdef _KERNEL
#include <sys/zfs_vfsops.h>
#endif
Add zfs_file_* interface, remove vnodes Provide a common zfs_file_* interface which can be implemented on all platforms to perform normal file access from either the kernel module or the libzpool library. This allows all non-portable vnode_t usage in the common code to be replaced by the new portable zfs_file_t. The associated vnode and kobj compatibility functions, types, and macros have been removed from the SPL. Moving forward, vnodes should only be used in platform specific code when provided by the native operating system. Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Jorgen Lundman <lundman@lundman.net> Signed-off-by: Matt Macy <mmacy@FreeBSD.org> Closes #9556
2019-11-21 20:32:57 +03:00
#include <sys/zfs_file.h>
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Cleanup: Specify unsignedness on things that should not be signed In #13871, zfs_vdev_aggregation_limit_non_rotating and zfs_vdev_aggregation_limit being signed was pointed out as a possible reason not to eliminate an unnecessary MAX(unsigned, 0) since the unsigned value was assigned from them. There is no reason for these module parameters to be signed and upon inspection, it was found that there are a number of other module parameters that are signed, but should not be, so we make them unsigned. Making them unsigned made it clear that some other variables in the code should also be unsigned, so we also make those unsigned. This prevents users from setting negative values that could potentially cause bad behaviors. It also makes the code slightly easier to understand. Mostly module parameters that deal with timeouts, limits, bitshifts and percentages are made unsigned by this. Any that are boolean are left signed, since whether booleans should be considered signed or unsigned does not matter. Making zfs_arc_lotsfree_percent unsigned caused a `zfs_arc_lotsfree_percent >= 0` check to become redundant, so it was removed. Removing the check was also necessary to prevent a compiler error from -Werror=type-limits. Several end of line comments had to be moved to their own lines because replacing int with uint_t caused us to exceed the 80 character limit enforced by cstyle.pl. The following were kept signed because they are passed to taskq_create(), which expects signed values and modifying the OpenSolaris/Illumos DDI is out of scope of this patch: * metaslab_load_pct * zfs_sync_taskq_batch_pct * zfs_zil_clean_taskq_nthr_pct * zfs_zil_clean_taskq_minalloc * zfs_zil_clean_taskq_maxalloc * zfs_arc_prune_task_threads Also, negative values in those parameters was found to be harmless. The following were left signed because either negative values make sense, or more analysis was needed to determine whether negative values should be disallowed: * zfs_metaslab_switch_threshold * zfs_pd_bytes_max * zfs_livelist_min_percent_shared zfs_multihost_history was made static to be consistent with other parameters. A number of module parameters were marked as signed, but in reality referenced unsigned variables. upgrade_errlog_limit is one of the numerous examples. In the case of zfs_vdev_async_read_max_active, it was already uint32_t, but zdb had an extern int declaration for it. Interestingly, the documentation in zfs.4 was right for upgrade_errlog_limit despite the module parameter being wrongly marked, while the documentation for zfs_vdev_async_read_max_active (and friends) was wrong. It was also wrong for zstd_abort_size, which was unsigned, but was documented as signed. Also, the documentation in zfs.4 incorrectly described the following parameters as ulong when they were int: * zfs_arc_meta_adjust_restarts * zfs_override_estimate_recordsize They are now uint_t as of this patch and thus the man page has been updated to describe them as uint. dbuf_state_index was left alone since it does nothing and perhaps should be removed in another patch. If any module parameters were missed, they were not found by `grep -r 'ZFS_MODULE_PARAM' | grep ', INT'`. I did find a few that grep missed, but only because they were in files that had hits. This patch intentionally did not attempt to address whether some of these module parameters should be elevated to 64-bit parameters, because the length of a long on 32-bit is 32-bit. Lastly, it was pointed out during review that uint_t is a better match for these variables than uint32_t because FreeBSD kernel parameter definitions are designed for uint_t, whose bit width can change in future memory models. As a result, we change the existing parameters that are uint32_t to use uint_t. Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Neal Gompa <ngompa@datto.com> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Closes #13875
2022-09-28 02:42:41 +03:00
static uint_t zfs_recv_queue_length = SPA_MAXBLOCKSIZE;
static uint_t zfs_recv_queue_ff = 20;
static uint_t zfs_recv_write_batch_size = 1024 * 1024;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
static int zfs_recv_best_effort_corrective = 0;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Enable -Wwrite-strings Also, fix leak from ztest_global_vars_to_zdb_args() Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Closes #13348
2022-04-19 21:38:30 +03:00
static const void *const dmu_recv_tag = "dmu_recv_tag";
module/*.ko: prune .data, global .rodata Evaluated every variable that lives in .data (and globals in .rodata) in the kernel modules, and constified/eliminated/localised them appropriately. This means that all read-only data is now actually read-only data, and, if possible, at file scope. A lot of previously- global-symbols became inlinable (and inlined!) constants. Probably not in a big Wowee Performance Moment, but hey. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Closes #12899
2022-01-15 02:37:55 +03:00
const char *const recv_clone_name = "%recv";
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Wait for txg sync if the last DRR_FREEOBJECTS might result in a hole If we receive a DRR_FREEOBJECTS as the first entry in an object range, this might end up producing a hole if the freed objects were the only existing objects in the block. If the txg starts syncing before we've processed any following DRR_OBJECT records, this leads to a possible race where the backing arc_buf_t gets its psize set to 0 in the arc_write_ready() callback while still being referenced from a dirty record in the open txg. To prevent this, we insert a txg_wait_synced call if the first record in the range was a DRR_FREEOBJECTS that actually resulted in one or more freed objects. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: David Hedberg <david.hedberg@findity.com> Sponsored by: Findity AB Closes #11893 Closes #14358
2023-01-24 00:19:43 +03:00
typedef enum {
ORNS_NO,
ORNS_YES,
ORNS_MAYBE
} or_need_sync_t;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
static int receive_read_payload_and_next_header(dmu_recv_cookie_t *ra, int len,
void *buf);
struct receive_record_arg {
dmu_replay_record_t header;
void *payload; /* Pointer to a buffer containing the payload */
/*
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
* If the record is a WRITE or SPILL, pointer to the abd containing the
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
* payload.
*/
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_t *abd;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
int payload_size;
uint64_t bytes_read; /* bytes read from stream when record created */
boolean_t eos_marker; /* Marks the end of the stream */
bqueue_node_t node;
};
struct receive_writer_arg {
objset_t *os;
boolean_t byteswap;
bqueue_t q;
/*
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
* These three members are used to signal to the main thread when
* we're done.
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
*/
kmutex_t mutex;
kcondvar_t cv;
boolean_t done;
int err;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
const char *tofs;
boolean_t heal;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
boolean_t resumable;
boolean_t raw; /* DMU_BACKUP_FEATURE_RAW set */
boolean_t spill; /* DRR_FLAG_SPILL_BLOCK set */
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
boolean_t full; /* this is a full send stream */
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
uint64_t last_object;
uint64_t last_offset;
uint64_t max_object; /* highest object ID referenced in stream */
uint64_t bytes_read; /* bytes read when current record created */
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
list_t write_batch;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
/* Encryption parameters for the last received DRR_OBJECT_RANGE */
boolean_t or_crypt_params_present;
uint64_t or_firstobj;
uint64_t or_numslots;
uint8_t or_salt[ZIO_DATA_SALT_LEN];
uint8_t or_iv[ZIO_DATA_IV_LEN];
uint8_t or_mac[ZIO_DATA_MAC_LEN];
boolean_t or_byteorder;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
zio_t *heal_pio;
Wait for txg sync if the last DRR_FREEOBJECTS might result in a hole If we receive a DRR_FREEOBJECTS as the first entry in an object range, this might end up producing a hole if the freed objects were the only existing objects in the block. If the txg starts syncing before we've processed any following DRR_OBJECT records, this leads to a possible race where the backing arc_buf_t gets its psize set to 0 in the arc_write_ready() callback while still being referenced from a dirty record in the open txg. To prevent this, we insert a txg_wait_synced call if the first record in the range was a DRR_FREEOBJECTS that actually resulted in one or more freed objects. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: David Hedberg <david.hedberg@findity.com> Sponsored by: Findity AB Closes #11893 Closes #14358
2023-01-24 00:19:43 +03:00
/* Keep track of DRR_FREEOBJECTS right after DRR_OBJECT_RANGE */
or_need_sync_t or_need_sync;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
};
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
typedef struct dmu_recv_begin_arg {
const char *drba_origin;
dmu_recv_cookie_t *drba_cookie;
cred_t *drba_cred;
filesystem_limit/snapshot_limit is incorrectly enforced against root The filesystem_limit and snapshot_limit properties limit the number of filesystems or snapshots that can be created below this dataset. According to the manpage, "The limit is not enforced if the user is allowed to change the limit." Two types of users are allowed to change the limit: 1. Those that have been delegated the `filesystem_limit` or `snapshot_limit` permission, e.g. with `zfs allow USER filesystem_limit DATASET`. This works properly. 2. A user with elevated system privileges (e.g. root). This does not work - the root user will incorrectly get an error when trying to create a snapshot/filesystem, if it exceeds the `_limit` property. The problem is that `priv_policy_ns()` does not work if the `cred_t` is not that of the current process. This happens when `dsl_enforce_ds_ss_limits()` is called in syncing context (as part of a sync task's check func) to determine the permissions of the corresponding user process. This commit fixes the issue by passing the `task_struct` (typedef'ed as a `proc_t`) to syncing context, and then using `has_capability()` to determine if that process is privileged. Note that we still need to pass the `cred_t` to syncing context so that we can check if the user was delegated this permission with `zfs allow`. This problem only impacts Linux. Wrappers are added to FreeBSD but it continues to use `priv_check_cred()`, which works on arbitrary `cred_t`. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@ixsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #8226 Closes #10545
2020-07-12 03:18:02 +03:00
proc_t *drba_proc;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_crypto_params_t *drba_dcp;
} dmu_recv_begin_arg_t;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
static void
byteswap_record(dmu_replay_record_t *drr)
{
#define DO64(X) (drr->drr_u.X = BSWAP_64(drr->drr_u.X))
#define DO32(X) (drr->drr_u.X = BSWAP_32(drr->drr_u.X))
drr->drr_type = BSWAP_32(drr->drr_type);
drr->drr_payloadlen = BSWAP_32(drr->drr_payloadlen);
switch (drr->drr_type) {
case DRR_BEGIN:
DO64(drr_begin.drr_magic);
DO64(drr_begin.drr_versioninfo);
DO64(drr_begin.drr_creation_time);
DO32(drr_begin.drr_type);
DO32(drr_begin.drr_flags);
DO64(drr_begin.drr_toguid);
DO64(drr_begin.drr_fromguid);
break;
case DRR_OBJECT:
DO64(drr_object.drr_object);
DO32(drr_object.drr_type);
DO32(drr_object.drr_bonustype);
DO32(drr_object.drr_blksz);
DO32(drr_object.drr_bonuslen);
DO32(drr_object.drr_raw_bonuslen);
DO64(drr_object.drr_toguid);
DO64(drr_object.drr_maxblkid);
break;
case DRR_FREEOBJECTS:
DO64(drr_freeobjects.drr_firstobj);
DO64(drr_freeobjects.drr_numobjs);
DO64(drr_freeobjects.drr_toguid);
break;
case DRR_WRITE:
DO64(drr_write.drr_object);
DO32(drr_write.drr_type);
DO64(drr_write.drr_offset);
DO64(drr_write.drr_logical_size);
DO64(drr_write.drr_toguid);
ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_write.drr_key.ddk_cksum);
DO64(drr_write.drr_key.ddk_prop);
DO64(drr_write.drr_compressed_size);
break;
case DRR_WRITE_EMBEDDED:
DO64(drr_write_embedded.drr_object);
DO64(drr_write_embedded.drr_offset);
DO64(drr_write_embedded.drr_length);
DO64(drr_write_embedded.drr_toguid);
DO32(drr_write_embedded.drr_lsize);
DO32(drr_write_embedded.drr_psize);
break;
case DRR_FREE:
DO64(drr_free.drr_object);
DO64(drr_free.drr_offset);
DO64(drr_free.drr_length);
DO64(drr_free.drr_toguid);
break;
case DRR_SPILL:
DO64(drr_spill.drr_object);
DO64(drr_spill.drr_length);
DO64(drr_spill.drr_toguid);
DO64(drr_spill.drr_compressed_size);
DO32(drr_spill.drr_type);
break;
case DRR_OBJECT_RANGE:
DO64(drr_object_range.drr_firstobj);
DO64(drr_object_range.drr_numslots);
DO64(drr_object_range.drr_toguid);
break;
case DRR_REDACT:
DO64(drr_redact.drr_object);
DO64(drr_redact.drr_offset);
DO64(drr_redact.drr_length);
DO64(drr_redact.drr_toguid);
break;
case DRR_END:
DO64(drr_end.drr_toguid);
ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_end.drr_checksum);
break;
default:
break;
}
if (drr->drr_type != DRR_BEGIN) {
ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_checksum.drr_checksum);
}
#undef DO64
#undef DO32
}
static boolean_t
redact_snaps_contains(uint64_t *snaps, uint64_t num_snaps, uint64_t guid)
{
for (int i = 0; i < num_snaps; i++) {
if (snaps[i] == guid)
return (B_TRUE);
}
return (B_FALSE);
}
/*
* Check that the new stream we're trying to receive is redacted with respect to
* a subset of the snapshots that the origin was redacted with respect to. For
* the reasons behind this, see the man page on redacted zfs sends and receives.
*/
static boolean_t
compatible_redact_snaps(uint64_t *origin_snaps, uint64_t origin_num_snaps,
uint64_t *redact_snaps, uint64_t num_redact_snaps)
{
/*
* Short circuit the comparison; if we are redacted with respect to
* more snapshots than the origin, we can't be redacted with respect
* to a subset.
*/
if (num_redact_snaps > origin_num_snaps) {
return (B_FALSE);
}
for (int i = 0; i < num_redact_snaps; i++) {
if (!redact_snaps_contains(origin_snaps, origin_num_snaps,
redact_snaps[i])) {
return (B_FALSE);
}
}
return (B_TRUE);
}
static boolean_t
redact_check(dmu_recv_begin_arg_t *drba, dsl_dataset_t *origin)
{
uint64_t *origin_snaps;
uint64_t origin_num_snaps;
dmu_recv_cookie_t *drc = drba->drba_cookie;
struct drr_begin *drrb = drc->drc_drrb;
int featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
int err = 0;
boolean_t ret = B_TRUE;
uint64_t *redact_snaps;
uint_t numredactsnaps;
/*
* If this is a full send stream, we're safe no matter what.
*/
if (drrb->drr_fromguid == 0)
return (ret);
VERIFY(dsl_dataset_get_uint64_array_feature(origin,
SPA_FEATURE_REDACTED_DATASETS, &origin_num_snaps, &origin_snaps));
if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_FROM_SNAPS, &redact_snaps, &numredactsnaps) ==
0) {
/*
* If the send stream was sent from the redaction bookmark or
* the redacted version of the dataset, then we're safe. Verify
* that this is from the a compatible redaction bookmark or
* redacted dataset.
*/
if (!compatible_redact_snaps(origin_snaps, origin_num_snaps,
redact_snaps, numredactsnaps)) {
err = EINVAL;
}
} else if (featureflags & DMU_BACKUP_FEATURE_REDACTED) {
/*
* If the stream is redacted, it must be redacted with respect
* to a subset of what the origin is redacted with respect to.
* See case number 2 in the zfs man page section on redacted zfs
* send.
*/
err = nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_SNAPS, &redact_snaps, &numredactsnaps);
if (err != 0 || !compatible_redact_snaps(origin_snaps,
origin_num_snaps, redact_snaps, numredactsnaps)) {
err = EINVAL;
}
} else if (!redact_snaps_contains(origin_snaps, origin_num_snaps,
drrb->drr_toguid)) {
/*
* If the stream isn't redacted but the origin is, this must be
* one of the snapshots the origin is redacted with respect to.
* See case number 1 in the zfs man page section on redacted zfs
* send.
*/
err = EINVAL;
}
if (err != 0)
ret = B_FALSE;
return (ret);
}
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
/*
* If we previously received a stream with --large-block, we don't support
* receiving an incremental on top of it without --large-block. This avoids
* forcing a read-modify-write or trying to re-aggregate a string of WRITE
* records.
*/
static int
recv_check_large_blocks(dsl_dataset_t *ds, uint64_t featureflags)
{
if (dsl_dataset_feature_is_active(ds, SPA_FEATURE_LARGE_BLOCKS) &&
!(featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS))
return (SET_ERROR(ZFS_ERR_STREAM_LARGE_BLOCK_MISMATCH));
return (0);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
static int
recv_begin_check_existing_impl(dmu_recv_begin_arg_t *drba, dsl_dataset_t *ds,
uint64_t fromguid, uint64_t featureflags)
{
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
uint64_t obj;
ZVOLs should not be allowed to have children zfs create, receive and rename can bypass this hierarchy rule. Update both userland and kernel module to prevent this issue and use pyzfs unit tests to exercise the ioctls directly. Note: this commit slightly changes zfs_ioc_create() ABI. This allow to differentiate a generic error (EINVAL) from the specific case where we tried to create a dataset below a ZVOL (ZFS_ERR_WRONG_PARENT). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
2019-02-09 02:44:15 +03:00
uint64_t children;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
int error;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
dsl_dataset_t *snap;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_pool_t *dp = ds->ds_dir->dd_pool;
boolean_t encrypted = ds->ds_dir->dd_crypto_obj != 0;
boolean_t raw = (featureflags & DMU_BACKUP_FEATURE_RAW) != 0;
boolean_t embed = (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) != 0;
Always refuse receving non-resume stream when resume state exists This fixes a hole in the situation where the resume state is left from receiving a new dataset and, so, the state is set on the dataset itself (as opposed to %recv child). Additionally, distinguish incremental and resume streams in error messages. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Andriy Gapon <avg@FreeBSD.org> Closes #9252
2019-09-03 20:56:55 +03:00
/* Temporary clone name must not exist. */
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
error = zap_lookup(dp->dp_meta_objset,
dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, recv_clone_name,
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
8, 1, &obj);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (error != ENOENT)
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
return (error == 0 ? SET_ERROR(EBUSY) : error);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Always refuse receving non-resume stream when resume state exists This fixes a hole in the situation where the resume state is left from receiving a new dataset and, so, the state is set on the dataset itself (as opposed to %recv child). Additionally, distinguish incremental and resume streams in error messages. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Andriy Gapon <avg@FreeBSD.org> Closes #9252
2019-09-03 20:56:55 +03:00
/* Resume state must not be set. */
if (dsl_dataset_has_resume_receive_state(ds))
return (SET_ERROR(EBUSY));
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
/* New snapshot name must not exist if we're not healing it. */
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
error = zap_lookup(dp->dp_meta_objset,
dsl_dataset_phys(ds)->ds_snapnames_zapobj,
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
drba->drba_cookie->drc_tosnap, 8, 1, &obj);
if (drba->drba_cookie->drc_heal) {
if (error != 0)
return (error);
} else if (error != ENOENT) {
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
return (error == 0 ? SET_ERROR(EEXIST) : error);
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Always refuse receving non-resume stream when resume state exists This fixes a hole in the situation where the resume state is left from receiving a new dataset and, so, the state is set on the dataset itself (as opposed to %recv child). Additionally, distinguish incremental and resume streams in error messages. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Andriy Gapon <avg@FreeBSD.org> Closes #9252
2019-09-03 20:56:55 +03:00
/* Must not have children if receiving a ZVOL. */
ZVOLs should not be allowed to have children zfs create, receive and rename can bypass this hierarchy rule. Update both userland and kernel module to prevent this issue and use pyzfs unit tests to exercise the ioctls directly. Note: this commit slightly changes zfs_ioc_create() ABI. This allow to differentiate a generic error (EINVAL) from the specific case where we tried to create a dataset below a ZVOL (ZFS_ERR_WRONG_PARENT). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
2019-02-09 02:44:15 +03:00
error = zap_count(dp->dp_meta_objset,
dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, &children);
if (error != 0)
return (error);
if (drba->drba_cookie->drc_drrb->drr_type != DMU_OST_ZFS &&
children > 0)
return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* Check snapshot limit before receiving. We'll recheck again at the
* end, but might as well abort before receiving if we're already over
* the limit.
*
* Note that we do not check the file system limit with
* dsl_dir_fscount_check because the temporary %clones don't count
* against that limit.
*/
error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_SNAPSHOT_LIMIT,
filesystem_limit/snapshot_limit is incorrectly enforced against root The filesystem_limit and snapshot_limit properties limit the number of filesystems or snapshots that can be created below this dataset. According to the manpage, "The limit is not enforced if the user is allowed to change the limit." Two types of users are allowed to change the limit: 1. Those that have been delegated the `filesystem_limit` or `snapshot_limit` permission, e.g. with `zfs allow USER filesystem_limit DATASET`. This works properly. 2. A user with elevated system privileges (e.g. root). This does not work - the root user will incorrectly get an error when trying to create a snapshot/filesystem, if it exceeds the `_limit` property. The problem is that `priv_policy_ns()` does not work if the `cred_t` is not that of the current process. This happens when `dsl_enforce_ds_ss_limits()` is called in syncing context (as part of a sync task's check func) to determine the permissions of the corresponding user process. This commit fixes the issue by passing the `task_struct` (typedef'ed as a `proc_t`) to syncing context, and then using `has_capability()` to determine if that process is privileged. Note that we still need to pass the `cred_t` to syncing context so that we can check if the user was delegated this permission with `zfs allow`. This problem only impacts Linux. Wrappers are added to FreeBSD but it continues to use `priv_check_cred()`, which works on arbitrary `cred_t`. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@ixsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #8226 Closes #10545
2020-07-12 03:18:02 +03:00
NULL, drba->drba_cred, drba->drba_proc);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (error != 0)
return (error);
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (drba->drba_cookie->drc_heal) {
/* Encryption is incompatible with embedded data. */
if (encrypted && embed)
return (SET_ERROR(EINVAL));
/* Healing is not supported when in 'force' mode. */
if (drba->drba_cookie->drc_force)
return (SET_ERROR(EINVAL));
/* Must have keys loaded if doing encrypted non-raw recv. */
if (encrypted && !raw) {
if (spa_keystore_lookup_key(dp->dp_spa, ds->ds_object,
NULL, NULL) != 0)
return (SET_ERROR(EACCES));
}
error = dsl_dataset_hold_obj(dp, obj, FTAG, &snap);
if (error != 0)
return (error);
/*
* When not doing best effort corrective recv healing can only
* be done if the send stream is for the same snapshot as the
* one we are trying to heal.
*/
if (zfs_recv_best_effort_corrective == 0 &&
drba->drba_cookie->drc_drrb->drr_toguid !=
dsl_dataset_phys(snap)->ds_guid) {
dsl_dataset_rele(snap, FTAG);
return (SET_ERROR(ENOTSUP));
}
dsl_dataset_rele(snap, FTAG);
} else if (fromguid != 0) {
/* Sanity check the incremental recv */
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
uint64_t obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
/* Can't perform a raw receive on top of a non-raw receive */
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (!encrypted && raw)
return (SET_ERROR(EINVAL));
/* Encryption is incompatible with embedded data */
if (encrypted && embed)
return (SET_ERROR(EINVAL));
/* Find snapshot in this dir that matches fromguid. */
while (obj != 0) {
error = dsl_dataset_hold_obj(dp, obj, FTAG,
&snap);
if (error != 0)
return (SET_ERROR(ENODEV));
if (snap->ds_dir != ds->ds_dir) {
dsl_dataset_rele(snap, FTAG);
return (SET_ERROR(ENODEV));
}
if (dsl_dataset_phys(snap)->ds_guid == fromguid)
break;
obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
dsl_dataset_rele(snap, FTAG);
}
if (obj == 0)
return (SET_ERROR(ENODEV));
if (drba->drba_cookie->drc_force) {
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
drba->drba_cookie->drc_fromsnapobj = obj;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
} else {
/*
* If we are not forcing, there must be no
Fix incorrect error message for raw receive This patch fixes an incorrect error message that comes up when doing a non-forcing, raw, incremental receive into a dataset that has a newer snapshot than the "from" snapshot. In this case, the current code prints a confusing message about an IVset guid mismatch. This functionality is supported by non-raw receives as an undocumented feature, but was never supported by the raw receive code. If this is desired in the future, we can probably figure out a way to make it work. Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Issue #8758 Closes #8863
2019-06-10 19:45:08 +03:00
* changes since fromsnap. Raw sends have an
* additional constraint that requires that
* no "noop" snapshots exist between fromsnap
* and tosnap for the IVset checking code to
* work properly.
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
*/
Fix incorrect error message for raw receive This patch fixes an incorrect error message that comes up when doing a non-forcing, raw, incremental receive into a dataset that has a newer snapshot than the "from" snapshot. In this case, the current code prints a confusing message about an IVset guid mismatch. This functionality is supported by non-raw receives as an undocumented feature, but was never supported by the raw receive code. If this is desired in the future, we can probably figure out a way to make it work. Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Issue #8758 Closes #8863
2019-06-10 19:45:08 +03:00
if (dsl_dataset_modified_since_snap(ds, snap) ||
(raw &&
dsl_dataset_phys(ds)->ds_prev_snap_obj !=
snap->ds_object)) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_rele(snap, FTAG);
return (SET_ERROR(ETXTBSY));
}
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
drba->drba_cookie->drc_fromsnapobj =
ds->ds_prev->ds_object;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (dsl_dataset_feature_is_active(snap,
SPA_FEATURE_REDACTED_DATASETS) && !redact_check(drba,
snap)) {
dsl_dataset_rele(snap, FTAG);
return (SET_ERROR(EINVAL));
}
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
error = recv_check_large_blocks(snap, featureflags);
if (error != 0) {
dsl_dataset_rele(snap, FTAG);
return (error);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_rele(snap, FTAG);
} else {
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
/* If full and not healing then must be forced. */
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (!drba->drba_cookie->drc_force)
return (SET_ERROR(EEXIST));
/*
* We don't support using zfs recv -F to blow away
* encrypted filesystems. This would require the
* dsl dir to point to the old encryption key and
* the new one at the same time during the receive.
*/
if ((!encrypted && raw) || encrypted)
return (SET_ERROR(EINVAL));
/*
* Perform the same encryption checks we would if
* we were creating a new dataset from scratch.
*/
if (!raw) {
boolean_t will_encrypt;
error = dmu_objset_create_crypt_check(
ds->ds_dir->dd_parent, drba->drba_dcp,
&will_encrypt);
if (error != 0)
return (error);
if (will_encrypt && embed)
return (SET_ERROR(EINVAL));
}
}
return (0);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
/*
* Check that any feature flags used in the data stream we're receiving are
* supported by the pool we are receiving into.
*
* Note that some of the features we explicitly check here have additional
* (implicit) features they depend on, but those dependencies are enforced
* through the zfeature_register() calls declaring the features that we
* explicitly check.
*/
static int
recv_begin_check_feature_flags_impl(uint64_t featureflags, spa_t *spa)
{
/*
* Check if there are any unsupported feature flags.
*/
if (!DMU_STREAM_SUPPORTED(featureflags)) {
return (SET_ERROR(ZFS_ERR_UNKNOWN_SEND_STREAM_FEATURE));
}
/* Verify pool version supports SA if SA_SPILL feature set */
if ((featureflags & DMU_BACKUP_FEATURE_SA_SPILL) &&
spa_version(spa) < SPA_VERSION_SA)
return (SET_ERROR(ENOTSUP));
/*
Add zstd support to zfs This PR adds two new compression types, based on ZStandard: - zstd: A basic ZStandard compression algorithm Available compression. Levels for zstd are zstd-1 through zstd-19, where the compression increases with every level, but speed decreases. - zstd-fast: A faster version of the ZStandard compression algorithm zstd-fast is basically a "negative" level of zstd. The compression decreases with every level, but speed increases. Available compression levels for zstd-fast: - zstd-fast-1 through zstd-fast-10 - zstd-fast-20 through zstd-fast-100 (in increments of 10) - zstd-fast-500 and zstd-fast-1000 For more information check the man page. Implementation details: Rather than treat each level of zstd as a different algorithm (as was done historically with gzip), the block pointer `enum zio_compress` value is simply zstd for all levels, including zstd-fast, since they all use the same decompression function. The compress= property (a 64bit unsigned integer) uses the lower 7 bits to store the compression algorithm (matching the number of bits used in a block pointer, as the 8th bit was borrowed for embedded block pointers). The upper bits are used to store the compression level. It is necessary to be able to determine what compression level was used when later reading a block back, so the concept used in LZ4, where the first 32bits of the on-disk value are the size of the compressed data (since the allocation is rounded up to the nearest ashift), was extended, and we store the version of ZSTD and the level as well as the compressed size. This value is returned when decompressing a block, so that if the block needs to be recompressed (L2ARC, nop-write, etc), that the same parameters will be used to result in the matching checksum. All of the internal ZFS code ( `arc_buf_hdr_t`, `objset_t`, `zio_prop_t`, etc.) uses the separated _compress and _complevel variables. Only the properties ZAP contains the combined/bit-shifted value. The combined value is split when the compression_changed_cb() callback is called, and sets both objset members (os_compress and os_complevel). The userspace tools all use the combined/bit-shifted value. Additional notes: zdb can now also decode the ZSTD compression header (flag -Z) and inspect the size, version and compression level saved in that header. For each record, if it is ZSTD compressed, the parameters of the decoded compression header get printed. ZSTD is included with all current tests and new tests are added as-needed. Per-dataset feature flags now get activated when the property is set. If a compression algorithm requires a feature flag, zfs activates the feature when the property is set, rather than waiting for the first block to be born. This is currently only used by zstd but can be extended as needed. Portions-Sponsored-By: The FreeBSD Foundation Co-authored-by: Allan Jude <allanjude@freebsd.org> Co-authored-by: Brian Behlendorf <behlendorf1@llnl.gov> Co-authored-by: Sebastian Gottschall <s.gottschall@dd-wrt.com> Co-authored-by: Kjeld Schouten-Lebbing <kjeld@schouten-lebbing.nl> Co-authored-by: Michael Niewöhner <foss@mniewoehner.de> Signed-off-by: Allan Jude <allan@klarasystems.com> Signed-off-by: Allan Jude <allanjude@freebsd.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Sebastian Gottschall <s.gottschall@dd-wrt.com> Signed-off-by: Kjeld Schouten-Lebbing <kjeld@schouten-lebbing.nl> Signed-off-by: Michael Niewöhner <foss@mniewoehner.de> Closes #6247 Closes #9024 Closes #10277 Closes #10278
2020-08-18 20:10:17 +03:00
* LZ4 compressed, ZSTD compressed, embedded, mooched, large blocks,
* and large_dnodes in the stream can only be used if those pool
* features are enabled because we don't attempt to decompress /
* un-embed / un-mooch / split up the blocks / dnodes during the
* receive process.
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
*/
if ((featureflags & DMU_BACKUP_FEATURE_LZ4) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_LZ4_COMPRESS))
return (SET_ERROR(ENOTSUP));
Add zstd support to zfs This PR adds two new compression types, based on ZStandard: - zstd: A basic ZStandard compression algorithm Available compression. Levels for zstd are zstd-1 through zstd-19, where the compression increases with every level, but speed decreases. - zstd-fast: A faster version of the ZStandard compression algorithm zstd-fast is basically a "negative" level of zstd. The compression decreases with every level, but speed increases. Available compression levels for zstd-fast: - zstd-fast-1 through zstd-fast-10 - zstd-fast-20 through zstd-fast-100 (in increments of 10) - zstd-fast-500 and zstd-fast-1000 For more information check the man page. Implementation details: Rather than treat each level of zstd as a different algorithm (as was done historically with gzip), the block pointer `enum zio_compress` value is simply zstd for all levels, including zstd-fast, since they all use the same decompression function. The compress= property (a 64bit unsigned integer) uses the lower 7 bits to store the compression algorithm (matching the number of bits used in a block pointer, as the 8th bit was borrowed for embedded block pointers). The upper bits are used to store the compression level. It is necessary to be able to determine what compression level was used when later reading a block back, so the concept used in LZ4, where the first 32bits of the on-disk value are the size of the compressed data (since the allocation is rounded up to the nearest ashift), was extended, and we store the version of ZSTD and the level as well as the compressed size. This value is returned when decompressing a block, so that if the block needs to be recompressed (L2ARC, nop-write, etc), that the same parameters will be used to result in the matching checksum. All of the internal ZFS code ( `arc_buf_hdr_t`, `objset_t`, `zio_prop_t`, etc.) uses the separated _compress and _complevel variables. Only the properties ZAP contains the combined/bit-shifted value. The combined value is split when the compression_changed_cb() callback is called, and sets both objset members (os_compress and os_complevel). The userspace tools all use the combined/bit-shifted value. Additional notes: zdb can now also decode the ZSTD compression header (flag -Z) and inspect the size, version and compression level saved in that header. For each record, if it is ZSTD compressed, the parameters of the decoded compression header get printed. ZSTD is included with all current tests and new tests are added as-needed. Per-dataset feature flags now get activated when the property is set. If a compression algorithm requires a feature flag, zfs activates the feature when the property is set, rather than waiting for the first block to be born. This is currently only used by zstd but can be extended as needed. Portions-Sponsored-By: The FreeBSD Foundation Co-authored-by: Allan Jude <allanjude@freebsd.org> Co-authored-by: Brian Behlendorf <behlendorf1@llnl.gov> Co-authored-by: Sebastian Gottschall <s.gottschall@dd-wrt.com> Co-authored-by: Kjeld Schouten-Lebbing <kjeld@schouten-lebbing.nl> Co-authored-by: Michael Niewöhner <foss@mniewoehner.de> Signed-off-by: Allan Jude <allan@klarasystems.com> Signed-off-by: Allan Jude <allanjude@freebsd.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Sebastian Gottschall <s.gottschall@dd-wrt.com> Signed-off-by: Kjeld Schouten-Lebbing <kjeld@schouten-lebbing.nl> Signed-off-by: Michael Niewöhner <foss@mniewoehner.de> Closes #6247 Closes #9024 Closes #10277 Closes #10278
2020-08-18 20:10:17 +03:00
if ((featureflags & DMU_BACKUP_FEATURE_ZSTD) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_ZSTD_COMPRESS))
return (SET_ERROR(ENOTSUP));
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA))
return (SET_ERROR(ENOTSUP));
if ((featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS))
return (SET_ERROR(ENOTSUP));
if ((featureflags & DMU_BACKUP_FEATURE_LARGE_DNODE) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE))
return (SET_ERROR(ENOTSUP));
/*
* Receiving redacted streams requires that redacted datasets are
* enabled.
*/
if ((featureflags & DMU_BACKUP_FEATURE_REDACTED) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_REDACTED_DATASETS))
return (SET_ERROR(ENOTSUP));
return (0);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
static int
dmu_recv_begin_check(void *arg, dmu_tx_t *tx)
{
dmu_recv_begin_arg_t *drba = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
struct drr_begin *drrb = drba->drba_cookie->drc_drrb;
uint64_t fromguid = drrb->drr_fromguid;
int flags = drrb->drr_flags;
implicit conversion from 'boolean_t' to 'ds_hold_flags_t' Build error on illumos with gcc 10 did reveal: In function 'dmu_objset_refresh_ownership': ../../common/fs/zfs/dmu_objset.c:857:25: error: implicit conversion from 'boolean_t' to 'ds_hold_flags_t' {aka 'enum ds_hold_flags'} [-Werror=enum-conversion] 857 | dsl_dataset_disown(ds, decrypt, tag); | ^~~~~~~ cc1: all warnings being treated as errors libzfs_input_check.c: In function 'zfs_ioc_input_tests': libzfs_input_check.c:754:28: error: implicit conversion from 'enum dmu_objset_type' to 'enum lzc_dataset_type' [-Werror=enum-conversion] 754 | err = lzc_create(dataset, DMU_OST_ZFS, NULL, NULL, 0); | ^~~~~~~~~~~ cc1: all warnings being treated as errors The same issue is present in openzfs, and also the same issue about ds_hold_flags_t, which currently defines exactly one valid value. Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Toomas Soome <tsoome@me.com> Closes #11406
2020-12-28 03:31:02 +03:00
ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
int error;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
uint64_t featureflags = drba->drba_cookie->drc_featureflags;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_t *ds;
const char *tofs = drba->drba_cookie->drc_tofs;
/* already checked */
ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC);
ASSERT(!(featureflags & DMU_BACKUP_FEATURE_RESUMING));
if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
DMU_COMPOUNDSTREAM ||
drrb->drr_type >= DMU_OST_NUMTYPES ||
((flags & DRR_FLAG_CLONE) && drba->drba_origin == NULL))
return (SET_ERROR(EINVAL));
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
error = recv_begin_check_feature_flags_impl(featureflags, dp->dp_spa);
if (error != 0)
return (error);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
/* Resumable receives require extensible datasets */
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (drba->drba_cookie->drc_resumable &&
!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EXTENSIBLE_DATASET))
return (SET_ERROR(ENOTSUP));
if (featureflags & DMU_BACKUP_FEATURE_RAW) {
/* raw receives require the encryption feature */
if (!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_ENCRYPTION))
return (SET_ERROR(ENOTSUP));
/* embedded data is incompatible with encryption and raw recv */
if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)
return (SET_ERROR(EINVAL));
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
/* raw receives require spill block allocation flag */
if (!(flags & DRR_FLAG_SPILL_BLOCK))
return (SET_ERROR(ZFS_ERR_SPILL_BLOCK_FLAG_MISSING));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
} else {
Receive checks should allow unencrypted child datasets dmu_recv_begin_check() unconditionally sets the DS_HOLD_FLAG_DECRYPT flag before calling dsl_dataset_hold_flags(). If the key on the receiving side isn't loaded or the send stream contains embedded blocks, the receive check fails for a stream which is perfectly valid and could be received without any problem. This seems like a remnant of the initial design, where unencrypted datasets below encrypted ones weren't allowed. Add a condition to set `DS_HOLD_FLAG_DECRYPT` only for encrypted datasets, modify an existing test to detect this regression and add a test for raw replication streams. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Amanakis <gamanakis@gmail.com> Co-authored-by: George Amanakis <gamanakis@gmail.com> Signed-off-by: Attila Fülöp <attila@fueloep.org> Closes #13033 Closes #13076
2022-02-10 01:38:33 +03:00
/*
* We support unencrypted datasets below encrypted ones now,
* so add the DS_HOLD_FLAG_DECRYPT flag only if we are dealing
* with a dataset we may encrypt.
*/
Deny receiving into encrypted datasets if the keys are not loaded Commit 68ddc06b611854560fefa377437eb3c9480e084b introduced support for receiving unencrypted datasets as children of encrypted ones but unfortunately got the logic upside down. This resulted in failing to deny receives of incremental sends into encrypted datasets without their keys loaded. If receiving a filesystem, the receive was done into a newly created unencrypted child dataset of the target. In case of volumes the receive made the target volume undeletable since a dataset was created below it, which we obviously can't handle. Incremental streams with embedded blocks are affected as well. We fix the broken logic to properly deny receives in such cases. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Attila Fülöp <attila@fueloep.org> Closes #13598 Closes #14055 Closes #14119
2022-11-03 19:55:13 +03:00
if (drba->drba_dcp == NULL ||
Receive checks should allow unencrypted child datasets dmu_recv_begin_check() unconditionally sets the DS_HOLD_FLAG_DECRYPT flag before calling dsl_dataset_hold_flags(). If the key on the receiving side isn't loaded or the send stream contains embedded blocks, the receive check fails for a stream which is perfectly valid and could be received without any problem. This seems like a remnant of the initial design, where unencrypted datasets below encrypted ones weren't allowed. Add a condition to set `DS_HOLD_FLAG_DECRYPT` only for encrypted datasets, modify an existing test to detect this regression and add a test for raw replication streams. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Amanakis <gamanakis@gmail.com> Co-authored-by: George Amanakis <gamanakis@gmail.com> Signed-off-by: Attila Fülöp <attila@fueloep.org> Closes #13033 Closes #13076
2022-02-10 01:38:33 +03:00
drba->drba_dcp->cp_crypt != ZIO_CRYPT_OFF) {
dsflags |= DS_HOLD_FLAG_DECRYPT;
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
if (error == 0) {
/* target fs already exists; recv into temp clone */
/* Can't recv a clone into an existing fs */
if (flags & DRR_FLAG_CLONE || drba->drba_origin) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
error = recv_begin_check_existing_impl(drba, ds, fromguid,
featureflags);
dsl_dataset_rele_flags(ds, dsflags, FTAG);
} else if (error == ENOENT) {
/* target fs does not exist; must be a full backup or clone */
char buf[ZFS_MAX_DATASET_NAME_LEN];
ZVOLs should not be allowed to have children zfs create, receive and rename can bypass this hierarchy rule. Update both userland and kernel module to prevent this issue and use pyzfs unit tests to exercise the ioctls directly. Note: this commit slightly changes zfs_ioc_create() ABI. This allow to differentiate a generic error (EINVAL) from the specific case where we tried to create a dataset below a ZVOL (ZFS_ERR_WRONG_PARENT). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
2019-02-09 02:44:15 +03:00
objset_t *os;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
/* healing recv must be done "into" an existing snapshot */
if (drba->drba_cookie->drc_heal == B_TRUE)
return (SET_ERROR(ENOTSUP));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* If it's a non-clone incremental, we are missing the
* target fs, so fail the recv.
*/
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (fromguid != 0 && !((flags & DRR_FLAG_CLONE) ||
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drba->drba_origin))
return (SET_ERROR(ENOENT));
/*
* If we're receiving a full send as a clone, and it doesn't
* contain all the necessary free records and freeobject
* records, reject it.
*/
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (fromguid == 0 && drba->drba_origin != NULL &&
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
!(flags & DRR_FLAG_FREERECORDS))
return (SET_ERROR(EINVAL));
/* Open the parent of tofs */
ASSERT3U(strlen(tofs), <, sizeof (buf));
(void) strlcpy(buf, tofs, strrchr(tofs, '/') - tofs + 1);
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
error = dsl_dataset_hold(dp, buf, FTAG, &ds);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (error != 0)
return (error);
if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0 &&
drba->drba_origin == NULL) {
boolean_t will_encrypt;
/*
* Check that we aren't breaking any encryption rules
* and that we have all the parameters we need to
* create an encrypted dataset if necessary. If we are
* making an encrypted dataset the stream can't have
* embedded data.
*/
error = dmu_objset_create_crypt_check(ds->ds_dir,
drba->drba_dcp, &will_encrypt);
if (error != 0) {
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (error);
}
if (will_encrypt &&
(featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)) {
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (SET_ERROR(EINVAL));
}
}
/*
* Check filesystem and snapshot limits before receiving. We'll
* recheck snapshot limits again at the end (we create the
* filesystems and increment those counts during begin_sync).
*/
error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
filesystem_limit/snapshot_limit is incorrectly enforced against root The filesystem_limit and snapshot_limit properties limit the number of filesystems or snapshots that can be created below this dataset. According to the manpage, "The limit is not enforced if the user is allowed to change the limit." Two types of users are allowed to change the limit: 1. Those that have been delegated the `filesystem_limit` or `snapshot_limit` permission, e.g. with `zfs allow USER filesystem_limit DATASET`. This works properly. 2. A user with elevated system privileges (e.g. root). This does not work - the root user will incorrectly get an error when trying to create a snapshot/filesystem, if it exceeds the `_limit` property. The problem is that `priv_policy_ns()` does not work if the `cred_t` is not that of the current process. This happens when `dsl_enforce_ds_ss_limits()` is called in syncing context (as part of a sync task's check func) to determine the permissions of the corresponding user process. This commit fixes the issue by passing the `task_struct` (typedef'ed as a `proc_t`) to syncing context, and then using `has_capability()` to determine if that process is privileged. Note that we still need to pass the `cred_t` to syncing context so that we can check if the user was delegated this permission with `zfs allow`. This problem only impacts Linux. Wrappers are added to FreeBSD but it continues to use `priv_check_cred()`, which works on arbitrary `cred_t`. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@ixsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #8226 Closes #10545
2020-07-12 03:18:02 +03:00
ZFS_PROP_FILESYSTEM_LIMIT, NULL,
drba->drba_cred, drba->drba_proc);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (error != 0) {
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (error);
}
error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
filesystem_limit/snapshot_limit is incorrectly enforced against root The filesystem_limit and snapshot_limit properties limit the number of filesystems or snapshots that can be created below this dataset. According to the manpage, "The limit is not enforced if the user is allowed to change the limit." Two types of users are allowed to change the limit: 1. Those that have been delegated the `filesystem_limit` or `snapshot_limit` permission, e.g. with `zfs allow USER filesystem_limit DATASET`. This works properly. 2. A user with elevated system privileges (e.g. root). This does not work - the root user will incorrectly get an error when trying to create a snapshot/filesystem, if it exceeds the `_limit` property. The problem is that `priv_policy_ns()` does not work if the `cred_t` is not that of the current process. This happens when `dsl_enforce_ds_ss_limits()` is called in syncing context (as part of a sync task's check func) to determine the permissions of the corresponding user process. This commit fixes the issue by passing the `task_struct` (typedef'ed as a `proc_t`) to syncing context, and then using `has_capability()` to determine if that process is privileged. Note that we still need to pass the `cred_t` to syncing context so that we can check if the user was delegated this permission with `zfs allow`. This problem only impacts Linux. Wrappers are added to FreeBSD but it continues to use `priv_check_cred()`, which works on arbitrary `cred_t`. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@ixsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #8226 Closes #10545
2020-07-12 03:18:02 +03:00
ZFS_PROP_SNAPSHOT_LIMIT, NULL,
drba->drba_cred, drba->drba_proc);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (error != 0) {
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (error);
}
ZVOLs should not be allowed to have children zfs create, receive and rename can bypass this hierarchy rule. Update both userland and kernel module to prevent this issue and use pyzfs unit tests to exercise the ioctls directly. Note: this commit slightly changes zfs_ioc_create() ABI. This allow to differentiate a generic error (EINVAL) from the specific case where we tried to create a dataset below a ZVOL (ZFS_ERR_WRONG_PARENT). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
2019-02-09 02:44:15 +03:00
/* can't recv below anything but filesystems (eg. no ZVOLs) */
error = dmu_objset_from_ds(ds, &os);
if (error != 0) {
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
ZVOLs should not be allowed to have children zfs create, receive and rename can bypass this hierarchy rule. Update both userland and kernel module to prevent this issue and use pyzfs unit tests to exercise the ioctls directly. Note: this commit slightly changes zfs_ioc_create() ABI. This allow to differentiate a generic error (EINVAL) from the specific case where we tried to create a dataset below a ZVOL (ZFS_ERR_WRONG_PARENT). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
2019-02-09 02:44:15 +03:00
return (error);
}
if (dmu_objset_type(os) != DMU_OST_ZFS) {
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
ZVOLs should not be allowed to have children zfs create, receive and rename can bypass this hierarchy rule. Update both userland and kernel module to prevent this issue and use pyzfs unit tests to exercise the ioctls directly. Note: this commit slightly changes zfs_ioc_create() ABI. This allow to differentiate a generic error (EINVAL) from the specific case where we tried to create a dataset below a ZVOL (ZFS_ERR_WRONG_PARENT). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
2019-02-09 02:44:15 +03:00
return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (drba->drba_origin != NULL) {
dsl_dataset_t *origin;
error = dsl_dataset_hold_flags(dp, drba->drba_origin,
dsflags, FTAG, &origin);
if (error != 0) {
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (error);
}
if (!origin->ds_is_snapshot) {
dsl_dataset_rele_flags(origin, dsflags, FTAG);
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (SET_ERROR(EINVAL));
}
if (dsl_dataset_phys(origin)->ds_guid != fromguid &&
fromguid != 0) {
dsl_dataset_rele_flags(origin, dsflags, FTAG);
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (SET_ERROR(ENODEV));
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (origin->ds_dir->dd_crypto_obj != 0 &&
(featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)) {
dsl_dataset_rele_flags(origin, dsflags, FTAG);
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (SET_ERROR(EINVAL));
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
/*
* If the origin is redacted we need to verify that this
* send stream can safely be received on top of the
* origin.
*/
if (dsl_dataset_feature_is_active(origin,
SPA_FEATURE_REDACTED_DATASETS)) {
if (!redact_check(drba, origin)) {
dsl_dataset_rele_flags(origin, dsflags,
FTAG);
dsl_dataset_rele_flags(ds, dsflags,
FTAG);
return (SET_ERROR(EINVAL));
}
}
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
error = recv_check_large_blocks(ds, featureflags);
if (error != 0) {
dsl_dataset_rele_flags(origin, dsflags, FTAG);
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (error);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
dsl_dataset_rele_flags(origin, dsflags, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
ZVOLs should not be allowed to have children zfs create, receive and rename can bypass this hierarchy rule. Update both userland and kernel module to prevent this issue and use pyzfs unit tests to exercise the ioctls directly. Note: this commit slightly changes zfs_ioc_create() ABI. This allow to differentiate a generic error (EINVAL) from the specific case where we tried to create a dataset below a ZVOL (ZFS_ERR_WRONG_PARENT). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
2019-02-09 02:44:15 +03:00
Allow unencrypted children of encrypted datasets When encryption was first added to ZFS, we made a decision to prevent users from creating unencrypted children of encrypted datasets. The idea was to prevent users from inadvertently leaving some of their data unencrypted. However, since the release of 0.8.0, some legitimate reasons have been brought up for this behavior to be allowed. This patch simply removes this limitation from all code paths that had checks for it and updates the tests accordingly. Reviewed-by: Jason King <jason.king@joyent.com> Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Richard Laager <rlaager@wiktel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8737 Closes #8870
2019-06-20 22:29:51 +03:00
dsl_dataset_rele(ds, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
error = 0;
}
return (error);
}
static void
dmu_recv_begin_sync(void *arg, dmu_tx_t *tx)
{
dmu_recv_begin_arg_t *drba = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
objset_t *mos = dp->dp_meta_objset;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
dmu_recv_cookie_t *drc = drba->drba_cookie;
struct drr_begin *drrb = drc->drc_drrb;
const char *tofs = drc->drc_tofs;
uint64_t featureflags = drc->drc_featureflags;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_t *ds, *newds;
objset_t *os;
uint64_t dsobj;
implicit conversion from 'boolean_t' to 'ds_hold_flags_t' Build error on illumos with gcc 10 did reveal: In function 'dmu_objset_refresh_ownership': ../../common/fs/zfs/dmu_objset.c:857:25: error: implicit conversion from 'boolean_t' to 'ds_hold_flags_t' {aka 'enum ds_hold_flags'} [-Werror=enum-conversion] 857 | dsl_dataset_disown(ds, decrypt, tag); | ^~~~~~~ cc1: all warnings being treated as errors libzfs_input_check.c: In function 'zfs_ioc_input_tests': libzfs_input_check.c:754:28: error: implicit conversion from 'enum dmu_objset_type' to 'enum lzc_dataset_type' [-Werror=enum-conversion] 754 | err = lzc_create(dataset, DMU_OST_ZFS, NULL, NULL, 0); | ^~~~~~~~~~~ cc1: all warnings being treated as errors The same issue is present in openzfs, and also the same issue about ds_hold_flags_t, which currently defines exactly one valid value. Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Toomas Soome <tsoome@me.com> Closes #11406
2020-12-28 03:31:02 +03:00
ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
int error;
uint64_t crflags = 0;
dsl_crypto_params_t dummy_dcp = { 0 };
dsl_crypto_params_t *dcp = drba->drba_dcp;
if (drrb->drr_flags & DRR_FLAG_CI_DATA)
crflags |= DS_FLAG_CI_DATASET;
if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0)
dsflags |= DS_HOLD_FLAG_DECRYPT;
/*
* Raw, non-incremental recvs always use a dummy dcp with
* the raw cmd set. Raw incremental recvs do not use a dcp
* since the encryption parameters are already set in stone.
*/
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (dcp == NULL && drrb->drr_fromguid == 0 &&
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drba->drba_origin == NULL) {
ASSERT3P(dcp, ==, NULL);
dcp = &dummy_dcp;
if (featureflags & DMU_BACKUP_FEATURE_RAW)
dcp->cp_cmd = DCP_CMD_RAW_RECV;
}
error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
if (error == 0) {
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
/* Create temporary clone unless we're doing corrective recv */
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_t *snap = NULL;
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
if (drba->drba_cookie->drc_fromsnapobj != 0) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
VERIFY0(dsl_dataset_hold_obj(dp,
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
drba->drba_cookie->drc_fromsnapobj, FTAG, &snap));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
ASSERT3P(dcp, ==, NULL);
}
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (drc->drc_heal) {
/* When healing we want to use the provided snapshot */
VERIFY0(dsl_dataset_snap_lookup(ds, drc->drc_tosnap,
&dsobj));
} else {
dsobj = dsl_dataset_create_sync(ds->ds_dir,
recv_clone_name, snap, crflags, drba->drba_cred,
dcp, tx);
}
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
if (drba->drba_cookie->drc_fromsnapobj != 0)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_rele(snap, FTAG);
dsl_dataset_rele_flags(ds, dsflags, FTAG);
} else {
dsl_dir_t *dd;
const char *tail;
dsl_dataset_t *origin = NULL;
VERIFY0(dsl_dir_hold(dp, tofs, FTAG, &dd, &tail));
if (drba->drba_origin != NULL) {
VERIFY0(dsl_dataset_hold(dp, drba->drba_origin,
FTAG, &origin));
ASSERT3P(dcp, ==, NULL);
}
/* Create new dataset. */
dsobj = dsl_dataset_create_sync(dd, strrchr(tofs, '/') + 1,
origin, crflags, drba->drba_cred, dcp, tx);
if (origin != NULL)
dsl_dataset_rele(origin, FTAG);
dsl_dir_rele(dd, FTAG);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
drc->drc_newfs = B_TRUE;
}
VERIFY0(dsl_dataset_own_obj_force(dp, dsobj, dsflags, dmu_recv_tag,
&newds));
if (dsl_dataset_feature_is_active(newds,
SPA_FEATURE_REDACTED_DATASETS)) {
/*
* If the origin dataset is redacted, the child will be redacted
* when we create it. We clear the new dataset's
* redaction info; if it should be redacted, we'll fill
* in its information later.
*/
dsl_dataset_deactivate_feature(newds,
SPA_FEATURE_REDACTED_DATASETS, tx);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
VERIFY0(dmu_objset_from_ds(newds, &os));
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_resumable) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_zapify(newds, tx);
if (drrb->drr_fromguid != 0) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_FROMGUID,
8, 1, &drrb->drr_fromguid, tx));
}
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TOGUID,
8, 1, &drrb->drr_toguid, tx));
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TONAME,
1, strlen(drrb->drr_toname) + 1, drrb->drr_toname, tx));
uint64_t one = 1;
uint64_t zero = 0;
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OBJECT,
8, 1, &one, tx));
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OFFSET,
8, 1, &zero, tx));
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_BYTES,
8, 1, &zero, tx));
if (featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_LARGEBLOCK,
8, 1, &one, tx));
}
if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_EMBEDOK,
8, 1, &one, tx));
}
if (featureflags & DMU_BACKUP_FEATURE_COMPRESSED) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_COMPRESSOK,
8, 1, &one, tx));
}
if (featureflags & DMU_BACKUP_FEATURE_RAW) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_RAWOK,
8, 1, &one, tx));
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
uint64_t *redact_snaps;
uint_t numredactsnaps;
if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_FROM_SNAPS, &redact_snaps,
&numredactsnaps) == 0) {
VERIFY0(zap_add(mos, dsobj,
DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS,
sizeof (*redact_snaps), numredactsnaps,
redact_snaps, tx));
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
/*
* Usually the os->os_encrypted value is tied to the presence of a
* DSL Crypto Key object in the dd. However, that will not be received
* until dmu_recv_stream(), so we set the value manually for now.
*/
if (featureflags & DMU_BACKUP_FEATURE_RAW) {
os->os_encrypted = B_TRUE;
drba->drba_cookie->drc_raw = B_TRUE;
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (featureflags & DMU_BACKUP_FEATURE_REDACTED) {
uint64_t *redact_snaps;
uint_t numredactsnaps;
VERIFY0(nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_SNAPS, &redact_snaps, &numredactsnaps));
dsl_dataset_activate_redaction(newds, redact_snaps,
numredactsnaps, tx);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dmu_buf_will_dirty(newds->ds_dbuf, tx);
dsl_dataset_phys(newds)->ds_flags |= DS_FLAG_INCONSISTENT;
/*
* If we actually created a non-clone, we need to create the objset
* in our new dataset. If this is a raw send we postpone this until
* dmu_recv_stream() so that we can allocate the metadnode with the
* properties from the DRR_BEGIN payload.
*/
rrw_enter(&newds->ds_bp_rwlock, RW_READER, FTAG);
if (BP_IS_HOLE(dsl_dataset_get_blkptr(newds)) &&
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
(featureflags & DMU_BACKUP_FEATURE_RAW) == 0 &&
!drc->drc_heal) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
(void) dmu_objset_create_impl(dp->dp_spa,
newds, dsl_dataset_get_blkptr(newds), drrb->drr_type, tx);
}
rrw_exit(&newds->ds_bp_rwlock, FTAG);
drba->drba_cookie->drc_ds = newds;
dmu_objset_from_ds must be called with dp_config_rwlock held The normal lock order is that the dp_config_rwlock must be held before the ds_opening_lock. For example, dmu_objset_hold() does this. However, dmu_objset_open_impl() is called with the ds_opening_lock held, and if the dp_config_rwlock is not already held, it will attempt to acquire it. This may lead to deadlock, since the lock order is reversed. Looking at all the callers of dmu_objset_open_impl() (which is principally the callers of dmu_objset_from_ds()), almost all callers already have the dp_config_rwlock. However, there are a few places in the send and receive code paths that do not. For example: dsl_crypto_populate_key_nvlist, send_cb, dmu_recv_stream, receive_write_byref, redact_traverse_thread. This commit resolves the problem by requiring all callers ot dmu_objset_from_ds() to hold the dp_config_rwlock. In most cases, the code has been restructured such that we call dmu_objset_from_ds() earlier on in the send and receive processes, when we already have the dp_config_rwlock, and save the objset_t until we need it in the middle of the send or receive (similar to what we already do with the dsl_dataset_t). Thus we do not need to acquire the dp_config_rwlock in many new places. I also cleaned up code in dmu_redact_snap() and send_traverse_thread(). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #9662 Closes #10115
2020-03-12 20:55:02 +03:00
drba->drba_cookie->drc_os = os;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Enable compiler to typecheck logging Annotate spa logging declarations with printflike Workaround gcc bug (non disable-able warning) by replacing "" with " " Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matt Macy <mmacy@FreeBSD.org> Closes #9316
2019-09-12 23:28:26 +03:00
spa_history_log_internal_ds(newds, "receive", tx, " ");
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
static int
dmu_recv_resume_begin_check(void *arg, dmu_tx_t *tx)
{
dmu_recv_begin_arg_t *drba = arg;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
dmu_recv_cookie_t *drc = drba->drba_cookie;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_pool_t *dp = dmu_tx_pool(tx);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
struct drr_begin *drrb = drc->drc_drrb;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
int error;
implicit conversion from 'boolean_t' to 'ds_hold_flags_t' Build error on illumos with gcc 10 did reveal: In function 'dmu_objset_refresh_ownership': ../../common/fs/zfs/dmu_objset.c:857:25: error: implicit conversion from 'boolean_t' to 'ds_hold_flags_t' {aka 'enum ds_hold_flags'} [-Werror=enum-conversion] 857 | dsl_dataset_disown(ds, decrypt, tag); | ^~~~~~~ cc1: all warnings being treated as errors libzfs_input_check.c: In function 'zfs_ioc_input_tests': libzfs_input_check.c:754:28: error: implicit conversion from 'enum dmu_objset_type' to 'enum lzc_dataset_type' [-Werror=enum-conversion] 754 | err = lzc_create(dataset, DMU_OST_ZFS, NULL, NULL, 0); | ^~~~~~~~~~~ cc1: all warnings being treated as errors The same issue is present in openzfs, and also the same issue about ds_hold_flags_t, which currently defines exactly one valid value. Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Toomas Soome <tsoome@me.com> Closes #11406
2020-12-28 03:31:02 +03:00
ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_t *ds;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
const char *tofs = drc->drc_tofs;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* already checked */
ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
ASSERT(drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
DMU_COMPOUNDSTREAM ||
drrb->drr_type >= DMU_OST_NUMTYPES)
return (SET_ERROR(EINVAL));
/*
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
* This is mostly a sanity check since we should have already done these
* checks during a previous attempt to receive the data.
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
*/
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
error = recv_begin_check_feature_flags_impl(drc->drc_featureflags,
dp->dp_spa);
if (error != 0)
return (error);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* 6 extra bytes for /%recv */
char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
(void) snprintf(recvname, sizeof (recvname), "%s/%s",
tofs, recv_clone_name);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) {
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
/* raw receives require spill block allocation flag */
if (!(drrb->drr_flags & DRR_FLAG_SPILL_BLOCK))
return (SET_ERROR(ZFS_ERR_SPILL_BLOCK_FLAG_MISSING));
} else {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsflags |= DS_HOLD_FLAG_DECRYPT;
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Enforce "-F" flag on resuming recv of full/newfs on existing dataset When receiving full/newfs on existing dataset, then it should be done with "-F" flag. Its enforced for initial receive in checks done in zfs_receive_one function of libzfs. Similarly, on resuming full/newfs recv on existing dataset, it should be done with "-F" flag. When dataset doesn't exist, then full/new recv is done on newly created dataset and it's marked INCONSISTENT. But when receiving on existing dataset, recv is first done on %recv and its marked INCONSISTENT. Existing dataset is not marked INCONSISTENT. Resume of full/newfs receive with dataset not INCONSISTENT indicates that its resuming newfs on existing dataset. So, enforce "-F" flag in this case. Also return an error from dmu_recv_resume_begin_check() in zfs kernel, when its resuming full/newfs recv without force. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Chunwei Chen <david.chen@nutanix.com> Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com> Closes #13856 Closes #13857
2022-09-28 02:34:27 +03:00
boolean_t recvexist = B_TRUE;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (dsl_dataset_hold_flags(dp, recvname, dsflags, FTAG, &ds) != 0) {
/* %recv does not exist; continue in tofs */
Enforce "-F" flag on resuming recv of full/newfs on existing dataset When receiving full/newfs on existing dataset, then it should be done with "-F" flag. Its enforced for initial receive in checks done in zfs_receive_one function of libzfs. Similarly, on resuming full/newfs recv on existing dataset, it should be done with "-F" flag. When dataset doesn't exist, then full/new recv is done on newly created dataset and it's marked INCONSISTENT. But when receiving on existing dataset, recv is first done on %recv and its marked INCONSISTENT. Existing dataset is not marked INCONSISTENT. Resume of full/newfs receive with dataset not INCONSISTENT indicates that its resuming newfs on existing dataset. So, enforce "-F" flag in this case. Also return an error from dmu_recv_resume_begin_check() in zfs kernel, when its resuming full/newfs recv without force. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Chunwei Chen <david.chen@nutanix.com> Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com> Closes #13856 Closes #13857
2022-09-28 02:34:27 +03:00
recvexist = B_FALSE;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
if (error != 0)
return (error);
}
Enforce "-F" flag on resuming recv of full/newfs on existing dataset When receiving full/newfs on existing dataset, then it should be done with "-F" flag. Its enforced for initial receive in checks done in zfs_receive_one function of libzfs. Similarly, on resuming full/newfs recv on existing dataset, it should be done with "-F" flag. When dataset doesn't exist, then full/new recv is done on newly created dataset and it's marked INCONSISTENT. But when receiving on existing dataset, recv is first done on %recv and its marked INCONSISTENT. Existing dataset is not marked INCONSISTENT. Resume of full/newfs receive with dataset not INCONSISTENT indicates that its resuming newfs on existing dataset. So, enforce "-F" flag in this case. Also return an error from dmu_recv_resume_begin_check() in zfs kernel, when its resuming full/newfs recv without force. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Chunwei Chen <david.chen@nutanix.com> Signed-off-by: Jitendra Patidar <jitendra.patidar@nutanix.com> Closes #13856 Closes #13857
2022-09-28 02:34:27 +03:00
/*
* Resume of full/newfs recv on existing dataset should be done with
* force flag
*/
if (recvexist && drrb->drr_fromguid == 0 && !drc->drc_force) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(ZFS_ERR_RESUME_EXISTS));
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* check that ds is marked inconsistent */
if (!DS_IS_INCONSISTENT(ds)) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
/* check that there is resuming data, and that the toguid matches */
if (!dsl_dataset_is_zapified(ds)) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
uint64_t val;
error = zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val);
if (error != 0 || drrb->drr_toguid != val) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
/*
* Check if the receive is still running. If so, it will be owned.
* Note that nothing else can own the dataset (e.g. after the receive
* fails) because it will be marked inconsistent.
*/
if (dsl_dataset_has_owner(ds)) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EBUSY));
}
/* There should not be any snapshots of this fs yet. */
if (ds->ds_prev != NULL && ds->ds_prev->ds_dir == ds->ds_dir) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
/*
* Note: resume point will be checked when we process the first WRITE
* record.
*/
/* check that the origin matches */
val = 0;
(void) zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val);
if (drrb->drr_fromguid != val) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
Fix missing ivset guid with resumed raw base recv This patch corrects a bug introduced in 61152d1069. When resuming a raw base receive, the dmu_recv code always sets drc->drc_fromsnapobj to the object ID of the previous snapshot. For incrementals, this is correct, but for base sends, this should be left at 0. The presence of this ID eventually allows a check to run which determines whether or not the incoming stream and the previous snapshot have matching IVset guids. This check fails becuase it is not meant to run when there is no previous snapshot. When it does fail, the user receives an error stating that the incoming stream has the problem outlined in errata 4. This patch corrects this issue by simply ensuring drc->drc_fromsnapobj is left as 0 for base receives. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #10234 Closes #10239
2020-04-25 05:00:32 +03:00
if (ds->ds_prev != NULL && drrb->drr_fromguid != 0)
Fix errata #4 handling for resuming streams Currently, the handling for errata #4 has two issues which allow the checks for this issue to be bypassed using resumable sends. The first issue is that drc->drc_fromsnapobj is not set in the resuming code as it is in the non-resuming code. This causes dsl_crypto_recv_key_check() to skip its checks for the from_ivset_guid. The second issue is that resumable sends do not clean up their on-disk state if they fail the checks in dmu_recv_stream() that happen before any data is received. As a result of these two bugs, a user can attempt a resumable send of a dataset without a from_ivset_guid. This will fail the initial dmu_recv_stream() checks, leaving a valid resume state. The send can then be resumed, which skips those checks, allowing the receive to be completed. This commit fixes these issues by setting drc->drc_fromsnapobj in the resuming receive path and by ensuring that resumablereceives are properly cleaned up if they fail the initial dmu_recv_stream() checks. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #9818 Closes #9829
2020-01-14 23:25:20 +03:00
drc->drc_fromsnapobj = ds->ds_prev->ds_object;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
/*
* If we're resuming, and the send is redacted, then the original send
* must have been redacted, and must have been redacted with respect to
* the same snapshots.
*/
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_REDACTED) {
uint64_t num_ds_redact_snaps;
uint64_t *ds_redact_snaps;
uint_t num_stream_redact_snaps;
uint64_t *stream_redact_snaps;
if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_SNAPS, &stream_redact_snaps,
&num_stream_redact_snaps) != 0) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
if (!dsl_dataset_get_uint64_array_feature(ds,
SPA_FEATURE_REDACTED_DATASETS, &num_ds_redact_snaps,
&ds_redact_snaps)) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
for (int i = 0; i < num_ds_redact_snaps; i++) {
if (!redact_snaps_contains(ds_redact_snaps,
num_ds_redact_snaps, stream_redact_snaps[i])) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
}
}
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
error = recv_check_large_blocks(ds, drc->drc_featureflags);
if (error != 0) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (error);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (0);
}
static void
dmu_recv_resume_begin_sync(void *arg, dmu_tx_t *tx)
{
dmu_recv_begin_arg_t *drba = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
const char *tofs = drba->drba_cookie->drc_tofs;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
uint64_t featureflags = drba->drba_cookie->drc_featureflags;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_t *ds;
implicit conversion from 'boolean_t' to 'ds_hold_flags_t' Build error on illumos with gcc 10 did reveal: In function 'dmu_objset_refresh_ownership': ../../common/fs/zfs/dmu_objset.c:857:25: error: implicit conversion from 'boolean_t' to 'ds_hold_flags_t' {aka 'enum ds_hold_flags'} [-Werror=enum-conversion] 857 | dsl_dataset_disown(ds, decrypt, tag); | ^~~~~~~ cc1: all warnings being treated as errors libzfs_input_check.c: In function 'zfs_ioc_input_tests': libzfs_input_check.c:754:28: error: implicit conversion from 'enum dmu_objset_type' to 'enum lzc_dataset_type' [-Werror=enum-conversion] 754 | err = lzc_create(dataset, DMU_OST_ZFS, NULL, NULL, 0); | ^~~~~~~~~~~ cc1: all warnings being treated as errors The same issue is present in openzfs, and also the same issue about ds_hold_flags_t, which currently defines exactly one valid value. Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Toomas Soome <tsoome@me.com> Closes #11406
2020-12-28 03:31:02 +03:00
ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* 6 extra bytes for /%recv */
char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
(void) snprintf(recvname, sizeof (recvname), "%s/%s", tofs,
recv_clone_name);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (featureflags & DMU_BACKUP_FEATURE_RAW) {
drba->drba_cookie->drc_raw = B_TRUE;
} else {
dsflags |= DS_HOLD_FLAG_DECRYPT;
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (dsl_dataset_own_force(dp, recvname, dsflags, dmu_recv_tag, &ds)
!= 0) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* %recv does not exist; continue in tofs */
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
VERIFY0(dsl_dataset_own_force(dp, tofs, dsflags, dmu_recv_tag,
&ds));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drba->drba_cookie->drc_newfs = B_TRUE;
}
ASSERT(DS_IS_INCONSISTENT(ds));
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
ASSERT(!BP_IS_HOLE(dsl_dataset_get_blkptr(ds)) ||
drba->drba_cookie->drc_raw);
rrw_exit(&ds->ds_bp_rwlock, FTAG);
drba->drba_cookie->drc_ds = ds;
dmu_objset_from_ds must be called with dp_config_rwlock held The normal lock order is that the dp_config_rwlock must be held before the ds_opening_lock. For example, dmu_objset_hold() does this. However, dmu_objset_open_impl() is called with the ds_opening_lock held, and if the dp_config_rwlock is not already held, it will attempt to acquire it. This may lead to deadlock, since the lock order is reversed. Looking at all the callers of dmu_objset_open_impl() (which is principally the callers of dmu_objset_from_ds()), almost all callers already have the dp_config_rwlock. However, there are a few places in the send and receive code paths that do not. For example: dsl_crypto_populate_key_nvlist, send_cb, dmu_recv_stream, receive_write_byref, redact_traverse_thread. This commit resolves the problem by requiring all callers ot dmu_objset_from_ds() to hold the dp_config_rwlock. In most cases, the code has been restructured such that we call dmu_objset_from_ds() earlier on in the send and receive processes, when we already have the dp_config_rwlock, and save the objset_t until we need it in the middle of the send or receive (similar to what we already do with the dsl_dataset_t). Thus we do not need to acquire the dp_config_rwlock in many new places. I also cleaned up code in dmu_redact_snap() and send_traverse_thread(). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #9662 Closes #10115
2020-03-12 20:55:02 +03:00
VERIFY0(dmu_objset_from_ds(ds, &drba->drba_cookie->drc_os));
Fix errata #4 handling for resuming streams Currently, the handling for errata #4 has two issues which allow the checks for this issue to be bypassed using resumable sends. The first issue is that drc->drc_fromsnapobj is not set in the resuming code as it is in the non-resuming code. This causes dsl_crypto_recv_key_check() to skip its checks for the from_ivset_guid. The second issue is that resumable sends do not clean up their on-disk state if they fail the checks in dmu_recv_stream() that happen before any data is received. As a result of these two bugs, a user can attempt a resumable send of a dataset without a from_ivset_guid. This will fail the initial dmu_recv_stream() checks, leaving a valid resume state. The send can then be resumed, which skips those checks, allowing the receive to be completed. This commit fixes these issues by setting drc->drc_fromsnapobj in the resuming receive path and by ensuring that resumablereceives are properly cleaned up if they fail the initial dmu_recv_stream() checks. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #9818 Closes #9829
2020-01-14 23:25:20 +03:00
drba->drba_cookie->drc_should_save = B_TRUE;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Enable compiler to typecheck logging Annotate spa logging declarations with printflike Workaround gcc bug (non disable-able warning) by replacing "" with " " Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matt Macy <mmacy@FreeBSD.org> Closes #9316
2019-09-12 23:28:26 +03:00
spa_history_log_internal_ds(ds, "resume receive", tx, " ");
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
/*
* NB: callers *MUST* call dmu_recv_stream() if dmu_recv_begin()
* succeeds; otherwise we will leak the holds on the datasets.
*/
int
nvpair: Constify string functions After addressing coverity complaints involving `nvpair_name()`, the compiler started complaining about dropping const. This lead to a rabbit hole where not only `nvpair_name()` needed to be constified, but also `nvpair_value_string()`, `fnvpair_value_string()` and a few other static functions, plus variable pointers throughout the code. The result became a fairly big change, so it has been split out into its own patch. Reviewed-by: Tino Reichardt <milky-zfs@mcmilk.de> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Closes #14612
2023-03-11 21:39:24 +03:00
dmu_recv_begin(const char *tofs, const char *tosnap,
dmu_replay_record_t *drr_begin, boolean_t force, boolean_t heal,
boolean_t resumable, nvlist_t *localprops, nvlist_t *hidden_args,
const char *origin, dmu_recv_cookie_t *drc, zfs_file_t *fp,
offset_t *voffp)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
dmu_recv_begin_arg_t drba = { 0 };
Fix uninitialized scalar value read regression in dmu_recv_begin() da19d919a853ad05ef300fe000e6c96c4db84bcf changed this in a way that permits execution to reach `if (err == 0)` without initializing err. This could randomly cause the sync task to not execute. We fix that by initializing err to zero. Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Rob Norris <robn@despairlabs.com> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Reported-by: Coverity (CID-1535377) Closes #14607
2023-03-11 01:47:56 +03:00
int err = 0;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Remove bcopy(), bzero(), bcmp() bcopy() has a confusing argument order and is actually a move, not a copy; they're all deprecated since POSIX.1-2001 and removed in -2008, and we shim them out to mem*() on Linux anyway Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Closes #12996
2022-02-25 16:26:54 +03:00
memset(drc, 0, sizeof (dmu_recv_cookie_t));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drc->drc_drr_begin = drr_begin;
drc->drc_drrb = &drr_begin->drr_u.drr_begin;
drc->drc_tosnap = tosnap;
drc->drc_tofs = tofs;
drc->drc_force = force;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
drc->drc_heal = heal;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drc->drc_resumable = resumable;
drc->drc_cred = CRED();
filesystem_limit/snapshot_limit is incorrectly enforced against root The filesystem_limit and snapshot_limit properties limit the number of filesystems or snapshots that can be created below this dataset. According to the manpage, "The limit is not enforced if the user is allowed to change the limit." Two types of users are allowed to change the limit: 1. Those that have been delegated the `filesystem_limit` or `snapshot_limit` permission, e.g. with `zfs allow USER filesystem_limit DATASET`. This works properly. 2. A user with elevated system privileges (e.g. root). This does not work - the root user will incorrectly get an error when trying to create a snapshot/filesystem, if it exceeds the `_limit` property. The problem is that `priv_policy_ns()` does not work if the `cred_t` is not that of the current process. This happens when `dsl_enforce_ds_ss_limits()` is called in syncing context (as part of a sync task's check func) to determine the permissions of the corresponding user process. This commit fixes the issue by passing the `task_struct` (typedef'ed as a `proc_t`) to syncing context, and then using `has_capability()` to determine if that process is privileged. Note that we still need to pass the `cred_t` to syncing context so that we can check if the user was delegated this permission with `zfs allow`. This problem only impacts Linux. Wrappers are added to FreeBSD but it continues to use `priv_check_cred()`, which works on arbitrary `cred_t`. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@ixsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #8226 Closes #10545
2020-07-12 03:18:02 +03:00
drc->drc_proc = curproc;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drc->drc_clone = (origin != NULL);
if (drc->drc_drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) {
drc->drc_byteswap = B_TRUE;
(void) fletcher_4_incremental_byteswap(drr_begin,
sizeof (dmu_replay_record_t), &drc->drc_cksum);
byteswap_record(drr_begin);
} else if (drc->drc_drrb->drr_magic == DMU_BACKUP_MAGIC) {
(void) fletcher_4_incremental_native(drr_begin,
sizeof (dmu_replay_record_t), &drc->drc_cksum);
} else {
return (SET_ERROR(EINVAL));
}
Add zfs_file_* interface, remove vnodes Provide a common zfs_file_* interface which can be implemented on all platforms to perform normal file access from either the kernel module or the libzpool library. This allows all non-portable vnode_t usage in the common code to be replaced by the new portable zfs_file_t. The associated vnode and kobj compatibility functions, types, and macros have been removed from the SPL. Moving forward, vnodes should only be used in platform specific code when provided by the native operating system. Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Jorgen Lundman <lundman@lundman.net> Signed-off-by: Matt Macy <mmacy@FreeBSD.org> Closes #9556
2019-11-21 20:32:57 +03:00
drc->drc_fp = fp;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
drc->drc_voff = *voffp;
drc->drc_featureflags =
DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo);
uint32_t payloadlen = drc->drc_drr_begin->drr_payloadlen;
Reject streams that set ->drr_payloadlen to unreasonably large values In the zstream code, Coverity reported: "The argument could be controlled by an attacker, who could invoke the function with arbitrary values (for example, a very high or negative buffer size)." It did not report this in the kernel. This is likely because the userspace code stored this in an int before passing it into the allocator, while the kernel code stored it in a uint32_t. However, this did reveal a potentially real problem. On 32-bit systems and systems with only 4GB of physical memory or less in general, it is possible to pass a large enough value that the system will hang. Even worse, on Linux systems, the kernel memory allocator is not able to support allocations up to the maximum 4GB allocation size that this allows. This had already been limited in userspace to 64MB by `ZFS_SENDRECV_MAX_NVLIST`, but we need a hard limit in the kernel to protect systems. After some discussion, we settle on 256MB as a hard upper limit. Attempting to receive a stream that requires more memory than that will result in E2BIG being returned to user space. Reported-by: Coverity (CID-1529836) Reported-by: Coverity (CID-1529837) Reported-by: Coverity (CID-1529838) Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Closes #14285
2023-01-24 00:16:22 +03:00
/*
* Since OpenZFS 2.0.0, we have enforced a 64MB limit in userspace
* configurable via ZFS_SENDRECV_MAX_NVLIST. We enforce 256MB as a hard
* upper limit. Systems with less than 1GB of RAM will see a lower
* limit from `arc_all_memory() / 4`.
*/
if (payloadlen > (MIN((1U << 28), arc_all_memory() / 4)))
return (E2BIG);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (payloadlen != 0) {
Fix incremental receive silently failing for recursive sends The problem occurs because dmu_recv_begin pulls in the payload and next header from the input stream in order to use the contents of the begin record's nvlist. However, the change to do that before the other checks in dmu_recv_begin occur caused a regression where an empty send stream in a recursive send could have its END record consumed by this, which broke the logic of recv_skip. A test is also included to protect against this case in the future. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #12661 Closes #14568
2023-03-10 20:52:44 +03:00
void *payload = vmem_alloc(payloadlen, KM_SLEEP);
/*
* For compatibility with recursive send streams, we don't do
* this here if the stream could be part of a package. Instead,
* we'll do it in dmu_recv_stream. If we pull the next header
* too early, and it's the END record, we break the `recv_skip`
* logic.
*/
err = receive_read_payload_and_next_header(drc, payloadlen,
payload);
if (err != 0) {
vmem_free(payload, payloadlen);
return (err);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
err = nvlist_unpack(payload, payloadlen, &drc->drc_begin_nvl,
KM_SLEEP);
Reject streams that set ->drr_payloadlen to unreasonably large values In the zstream code, Coverity reported: "The argument could be controlled by an attacker, who could invoke the function with arbitrary values (for example, a very high or negative buffer size)." It did not report this in the kernel. This is likely because the userspace code stored this in an int before passing it into the allocator, while the kernel code stored it in a uint32_t. However, this did reveal a potentially real problem. On 32-bit systems and systems with only 4GB of physical memory or less in general, it is possible to pass a large enough value that the system will hang. Even worse, on Linux systems, the kernel memory allocator is not able to support allocations up to the maximum 4GB allocation size that this allows. This had already been limited in userspace to 64MB by `ZFS_SENDRECV_MAX_NVLIST`, but we need a hard limit in the kernel to protect systems. After some discussion, we settle on 256MB as a hard upper limit. Attempting to receive a stream that requires more memory than that will result in E2BIG being returned to user space. Reported-by: Coverity (CID-1529836) Reported-by: Coverity (CID-1529837) Reported-by: Coverity (CID-1529838) Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Closes #14285
2023-01-24 00:16:22 +03:00
vmem_free(payload, payloadlen);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (err != 0) {
kmem_free(drc->drc_next_rrd,
sizeof (*drc->drc_next_rrd));
return (err);
}
}
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
if (drc->drc_drrb->drr_flags & DRR_FLAG_SPILL_BLOCK)
drc->drc_spill = B_TRUE;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drba.drba_origin = origin;
drba.drba_cookie = drc;
drba.drba_cred = CRED();
filesystem_limit/snapshot_limit is incorrectly enforced against root The filesystem_limit and snapshot_limit properties limit the number of filesystems or snapshots that can be created below this dataset. According to the manpage, "The limit is not enforced if the user is allowed to change the limit." Two types of users are allowed to change the limit: 1. Those that have been delegated the `filesystem_limit` or `snapshot_limit` permission, e.g. with `zfs allow USER filesystem_limit DATASET`. This works properly. 2. A user with elevated system privileges (e.g. root). This does not work - the root user will incorrectly get an error when trying to create a snapshot/filesystem, if it exceeds the `_limit` property. The problem is that `priv_policy_ns()` does not work if the `cred_t` is not that of the current process. This happens when `dsl_enforce_ds_ss_limits()` is called in syncing context (as part of a sync task's check func) to determine the permissions of the corresponding user process. This commit fixes the issue by passing the `task_struct` (typedef'ed as a `proc_t`) to syncing context, and then using `has_capability()` to determine if that process is privileged. Note that we still need to pass the `cred_t` to syncing context so that we can check if the user was delegated this permission with `zfs allow`. This problem only impacts Linux. Wrappers are added to FreeBSD but it continues to use `priv_check_cred()`, which works on arbitrary `cred_t`. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@ixsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #8226 Closes #10545
2020-07-12 03:18:02 +03:00
drba.drba_proc = curproc;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING) {
err = dsl_sync_task(tofs,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dmu_recv_resume_begin_check, dmu_recv_resume_begin_sync,
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
&drba, 5, ZFS_SPACE_CHECK_NORMAL);
} else {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* For non-raw, non-incremental, non-resuming receives the
* user can specify encryption parameters on the command line
* with "zfs recv -o". For these receives we create a dcp and
* pass it to the sync task. Creating the dcp will implicitly
* remove the encryption params from the localprops nvlist,
* which avoids errors when trying to set these normally
* read-only properties. Any other kind of receive that
* attempts to set these properties will fail as a result.
*/
if ((DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo) &
DMU_BACKUP_FEATURE_RAW) == 0 &&
origin == NULL && drc->drc_drrb->drr_fromguid == 0) {
err = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
localprops, hidden_args, &drba.drba_dcp);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (err == 0) {
err = dsl_sync_task(tofs,
dmu_recv_begin_check, dmu_recv_begin_sync,
&drba, 5, ZFS_SPACE_CHECK_NORMAL);
dsl_crypto_params_free(drba.drba_dcp, !!err);
}
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (err != 0) {
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
nvlist_free(drc->drc_begin_nvl);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
return (err);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
/*
* Holds data need for corrective recv callback
*/
typedef struct cr_cb_data {
uint64_t size;
zbookmark_phys_t zb;
spa_t *spa;
} cr_cb_data_t;
static void
corrective_read_done(zio_t *zio)
{
cr_cb_data_t *data = zio->io_private;
/* Corruption corrected; update error log if needed */
Provide macros for setting and getting blkptr birth times There exist a couple of macros that are used to update the blkptr birth times but they can often be confusing. For example, the BP_PHYSICAL_BIRTH() macro will provide either the physical birth time if it is set or else return back the logical birth time. The complement to this macro is BP_SET_BIRTH() which will set the logical birth time and set the physical birth time if they are not the same. Consumers may get confused when they are trying to get the physical birth time and use the BP_PHYSICAL_BIRTH() macro only to find out that the logical birth time is what is actually returned. This change cleans up these macros and makes them symmetrical. The same functionally is preserved but the name is changed. Instead of calling BP_PHYSICAL_BIRTH(), consumer can now call BP_GET_BIRTH(). In additional to cleaning up this naming conventions, two new sets of macros are introduced -- BP_[SET|GET]_LOGICAL_BIRTH() and BP_[SET|GET]_PHYSICAL_BIRTH. These new macros allow the consumer to get and set the specific birth time. As part of the cleanup, the unused GRID macros have been removed and that portion of the blkptr are currently unused. Reviewed-by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Mark Maybee <mark.maybee@delphix.com> Signed-off-by: George Wilson <gwilson@delphix.com> Closes #15962
2024-03-26 01:01:54 +03:00
if (zio->io_error == 0) {
spa_remove_error(data->spa, &data->zb,
BP_GET_LOGICAL_BIRTH(zio->io_bp));
}
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
kmem_free(data, sizeof (cr_cb_data_t));
abd_free(zio->io_abd);
}
/*
* zio_rewrite the data pointed to by bp with the data from the rrd's abd.
*/
static int
do_corrective_recv(struct receive_writer_arg *rwa, struct drr_write *drrw,
struct receive_record_arg *rrd, blkptr_t *bp)
{
int err;
zio_t *io;
zbookmark_phys_t zb;
dnode_t *dn;
abd_t *abd = rrd->abd;
zio_cksum_t bp_cksum = bp->blk_cksum;
Finally drop long disabled vdev cache. It was a vdev level read cache, designed to aggregate many small reads by speculatively issuing bigger reads instead and caching the result. But since it has almost no idea about what is going on with exception of ZIO_FLAG_DONT_CACHE flag set by higher layers, it was found to make more harm than good, for which reason it was disabled for the past 12 years. These days we have much better instruments to enlarge the I/Os, such as speculative and prescient prefetches, I/O scheduler, I/O aggregation etc. Besides just the dead code removal this removes one extra mutex lock/unlock per write inside vdev_cache_write(), not otherwise disabled and trying to do some work. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Alexander Motin <mav@FreeBSD.org> Sponsored by: iXsystems, Inc. Closes #14953
2023-06-09 22:40:55 +03:00
zio_flag_t flags = ZIO_FLAG_SPECULATIVE | ZIO_FLAG_DONT_RETRY |
ZIO_FLAG_CANFAIL;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (rwa->raw)
flags |= ZIO_FLAG_RAW;
err = dnode_hold(rwa->os, drrw->drr_object, FTAG, &dn);
if (err != 0)
return (err);
SET_BOOKMARK(&zb, dmu_objset_id(rwa->os), drrw->drr_object, 0,
dbuf_whichblock(dn, 0, drrw->drr_offset));
dnode_rele(dn, FTAG);
if (!rwa->raw && DRR_WRITE_COMPRESSED(drrw)) {
/* Decompress the stream data */
abd_t *dabd = abd_alloc_linear(
drrw->drr_logical_size, B_FALSE);
err = zio_decompress_data(drrw->drr_compressiontype,
abd, abd_to_buf(dabd), abd_get_size(abd),
abd_get_size(dabd), NULL);
if (err != 0) {
abd_free(dabd);
return (err);
}
/* Swap in the newly decompressed data into the abd */
abd_free(abd);
abd = dabd;
}
if (!rwa->raw && BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) {
/* Recompress the data */
abd_t *cabd = abd_alloc_linear(BP_GET_PSIZE(bp),
B_FALSE);
Skip memory allocation when compressing holes Hole detection in the zio compression code allows us to opportunistically skip compression on holes. We can go a step further by not doing memory allocations on holes either. Reviewed-by: Brian Atkinson <batkinson@lanl.gov> Reviewed-by: Alexander Motin <mav@FreeBSD.org> Signed-off-by: Richard Yao <richard.yao@klarasystems.com> Sponsored-by: Wasabi Technology, Inc. Closes #14500
2023-02-28 01:41:02 +03:00
void *buf = abd_to_buf(cabd);
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
uint64_t csize = zio_compress_data(BP_GET_COMPRESS(bp),
Skip memory allocation when compressing holes Hole detection in the zio compression code allows us to opportunistically skip compression on holes. We can go a step further by not doing memory allocations on holes either. Reviewed-by: Brian Atkinson <batkinson@lanl.gov> Reviewed-by: Alexander Motin <mav@FreeBSD.org> Signed-off-by: Richard Yao <richard.yao@klarasystems.com> Sponsored-by: Wasabi Technology, Inc. Closes #14500
2023-02-28 01:41:02 +03:00
abd, &buf, abd_get_size(abd),
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
rwa->os->os_complevel);
abd_zero_off(cabd, csize, BP_GET_PSIZE(bp) - csize);
/* Swap in newly compressed data into the abd */
abd_free(abd);
abd = cabd;
flags |= ZIO_FLAG_RAW_COMPRESS;
}
/*
* The stream is not encrypted but the data on-disk is.
* We need to re-encrypt the buf using the same
* encryption type, salt, iv, and mac that was used to encrypt
* the block previosly.
*/
if (!rwa->raw && BP_USES_CRYPT(bp)) {
dsl_dataset_t *ds;
dsl_crypto_key_t *dck = NULL;
uint8_t salt[ZIO_DATA_SALT_LEN];
uint8_t iv[ZIO_DATA_IV_LEN];
uint8_t mac[ZIO_DATA_MAC_LEN];
boolean_t no_crypt = B_FALSE;
dsl_pool_t *dp = dmu_objset_pool(rwa->os);
abd_t *eabd = abd_alloc_linear(BP_GET_PSIZE(bp), B_FALSE);
zio_crypt_decode_params_bp(bp, salt, iv);
zio_crypt_decode_mac_bp(bp, mac);
dsl_pool_config_enter(dp, FTAG);
err = dsl_dataset_hold_flags(dp, rwa->tofs,
DS_HOLD_FLAG_DECRYPT, FTAG, &ds);
if (err != 0) {
dsl_pool_config_exit(dp, FTAG);
abd_free(eabd);
return (SET_ERROR(EACCES));
}
/* Look up the key from the spa's keystore */
err = spa_keystore_lookup_key(rwa->os->os_spa,
zb.zb_objset, FTAG, &dck);
if (err != 0) {
dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT,
FTAG);
dsl_pool_config_exit(dp, FTAG);
abd_free(eabd);
return (SET_ERROR(EACCES));
}
err = zio_do_crypt_abd(B_TRUE, &dck->dck_key,
BP_GET_TYPE(bp), BP_SHOULD_BYTESWAP(bp), salt, iv,
mac, abd_get_size(abd), abd, eabd, &no_crypt);
spa_keystore_dsl_key_rele(rwa->os->os_spa, dck, FTAG);
dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG);
dsl_pool_config_exit(dp, FTAG);
ASSERT0(no_crypt);
if (err != 0) {
abd_free(eabd);
return (err);
}
/* Swap in the newly encrypted data into the abd */
abd_free(abd);
abd = eabd;
/*
* We want to prevent zio_rewrite() from trying to
* encrypt the data again
*/
flags |= ZIO_FLAG_RAW_ENCRYPT;
}
rrd->abd = abd;
Provide macros for setting and getting blkptr birth times There exist a couple of macros that are used to update the blkptr birth times but they can often be confusing. For example, the BP_PHYSICAL_BIRTH() macro will provide either the physical birth time if it is set or else return back the logical birth time. The complement to this macro is BP_SET_BIRTH() which will set the logical birth time and set the physical birth time if they are not the same. Consumers may get confused when they are trying to get the physical birth time and use the BP_PHYSICAL_BIRTH() macro only to find out that the logical birth time is what is actually returned. This change cleans up these macros and makes them symmetrical. The same functionally is preserved but the name is changed. Instead of calling BP_PHYSICAL_BIRTH(), consumer can now call BP_GET_BIRTH(). In additional to cleaning up this naming conventions, two new sets of macros are introduced -- BP_[SET|GET]_LOGICAL_BIRTH() and BP_[SET|GET]_PHYSICAL_BIRTH. These new macros allow the consumer to get and set the specific birth time. As part of the cleanup, the unused GRID macros have been removed and that portion of the blkptr are currently unused. Reviewed-by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Mark Maybee <mark.maybee@delphix.com> Signed-off-by: George Wilson <gwilson@delphix.com> Closes #15962
2024-03-26 01:01:54 +03:00
io = zio_rewrite(NULL, rwa->os->os_spa, BP_GET_LOGICAL_BIRTH(bp), bp,
abd, BP_GET_PSIZE(bp), NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, flags,
&zb);
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
ASSERT(abd_get_size(abd) == BP_GET_LSIZE(bp) ||
abd_get_size(abd) == BP_GET_PSIZE(bp));
/* compute new bp checksum value and make sure it matches the old one */
zio_checksum_compute(io, BP_GET_CHECKSUM(bp), abd, abd_get_size(abd));
if (!ZIO_CHECKSUM_EQUAL(bp_cksum, io->io_bp->blk_cksum)) {
zio_destroy(io);
if (zfs_recv_best_effort_corrective != 0)
return (0);
return (SET_ERROR(ECKSUM));
}
/* Correct the corruption in place */
err = zio_wait(io);
if (err == 0) {
cr_cb_data_t *cb_data =
kmem_alloc(sizeof (cr_cb_data_t), KM_SLEEP);
cb_data->spa = rwa->os->os_spa;
cb_data->size = drrw->drr_logical_size;
cb_data->zb = zb;
/* Test if healing worked by re-reading the bp */
err = zio_wait(zio_read(rwa->heal_pio, rwa->os->os_spa, bp,
abd_alloc_for_io(drrw->drr_logical_size, B_FALSE),
drrw->drr_logical_size, corrective_read_done,
cb_data, ZIO_PRIORITY_ASYNC_READ, flags, NULL));
}
if (err != 0 && zfs_recv_best_effort_corrective != 0)
err = 0;
return (err);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
static int
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_read(dmu_recv_cookie_t *drc, int len, void *buf)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
int done = 0;
/*
* The code doesn't rely on this (lengths being multiples of 8). See
* comment in dump_bytes.
*/
ASSERT(len % 8 == 0 ||
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
(drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) != 0);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
while (done < len) {
Fix reading uninitialized variable in receive_read When zfs_file_read returns error, resid may be uninitialized. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Signed-off-by: Chunwei Chen <david.chen@nutanix.com> Closes #14404
2023-01-20 22:49:56 +03:00
ssize_t resid = len - done;
remove unneeded member drc_err of dmu_recv_cookie_t The member drc_err of dmu_recv_cookie_t is used only locally in receive_read, so we can replace it with a local variable. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10319
2020-05-14 22:10:29 +03:00
zfs_file_t *fp = drc->drc_fp;
int err = zfs_file_read(fp, (char *)buf + done,
Add zfs_file_* interface, remove vnodes Provide a common zfs_file_* interface which can be implemented on all platforms to perform normal file access from either the kernel module or the libzpool library. This allows all non-portable vnode_t usage in the common code to be replaced by the new portable zfs_file_t. The associated vnode and kobj compatibility functions, types, and macros have been removed from the SPL. Moving forward, vnodes should only be used in platform specific code when provided by the native operating system. Reviewed-by: Sean Eric Fagan <sef@ixsystems.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Jorgen Lundman <lundman@lundman.net> Signed-off-by: Matt Macy <mmacy@FreeBSD.org> Closes #9556
2019-11-21 20:32:57 +03:00
len - done, &resid);
Fix reading uninitialized variable in receive_read When zfs_file_read returns error, resid may be uninitialized. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Signed-off-by: Chunwei Chen <david.chen@nutanix.com> Closes #14404
2023-01-20 22:49:56 +03:00
if (err == 0 && resid == len - done) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
Separate warning for incomplete and corrupt streams This change adds a separate return code to zfs_ioc_recv that is used for incomplete streams, in addition to the existing return code for streams that contain corruption. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matthew Ahrens <mahrens@delphix.com> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #10122
2020-03-17 20:30:33 +03:00
* Note: ECKSUM or ZFS_ERR_STREAM_TRUNCATED indicates
* that the receive was interrupted and can
* potentially be resumed.
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
*/
remove unneeded member drc_err of dmu_recv_cookie_t The member drc_err of dmu_recv_cookie_t is used only locally in receive_read, so we can replace it with a local variable. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10319
2020-05-14 22:10:29 +03:00
err = SET_ERROR(ZFS_ERR_STREAM_TRUNCATED);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
drc->drc_voff += len - done - resid;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
done = len - resid;
remove unneeded member drc_err of dmu_recv_cookie_t The member drc_err of dmu_recv_cookie_t is used only locally in receive_read, so we can replace it with a local variable. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10319
2020-05-14 22:10:29 +03:00
if (err != 0)
return (err);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
drc->drc_bytes_read += len;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
ASSERT3U(done, ==, len);
return (0);
}
static inline uint8_t
deduce_nblkptr(dmu_object_type_t bonus_type, uint64_t bonus_size)
{
if (bonus_type == DMU_OT_SA) {
return (1);
} else {
return (1 +
((DN_OLD_MAX_BONUSLEN -
MIN(DN_OLD_MAX_BONUSLEN, bonus_size)) >> SPA_BLKPTRSHIFT));
}
}
static void
save_resume_state(struct receive_writer_arg *rwa,
uint64_t object, uint64_t offset, dmu_tx_t *tx)
{
int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;
if (!rwa->resumable)
return;
/*
* We use ds_resume_bytes[] != 0 to indicate that we need to
* update this on disk, so it must not be 0.
*/
ASSERT(rwa->bytes_read != 0);
/*
* We only resume from write records, which have a valid
* (non-meta-dnode) object number.
*/
ASSERT(object != 0);
/*
* For resuming to work correctly, we must receive records in order,
* sorted by object,offset. This is checked by the callers, but
* assert it here for good measure.
*/
ASSERT3U(object, >=, rwa->os->os_dsl_dataset->ds_resume_object[txgoff]);
ASSERT(object != rwa->os->os_dsl_dataset->ds_resume_object[txgoff] ||
offset >= rwa->os->os_dsl_dataset->ds_resume_offset[txgoff]);
ASSERT3U(rwa->bytes_read, >=,
rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff]);
rwa->os->os_dsl_dataset->ds_resume_object[txgoff] = object;
rwa->os->os_dsl_dataset->ds_resume_offset[txgoff] = offset;
rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff] = rwa->bytes_read;
}
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
static int
receive_object_is_same_generation(objset_t *os, uint64_t object,
dmu_object_type_t old_bonus_type, dmu_object_type_t new_bonus_type,
const void *new_bonus, boolean_t *samegenp)
{
zfs_file_info_t zoi;
int err;
dmu_buf_t *old_bonus_dbuf;
err = dmu_bonus_hold(os, object, FTAG, &old_bonus_dbuf);
if (err != 0)
return (err);
err = dmu_get_file_info(os, old_bonus_type, old_bonus_dbuf->db_data,
&zoi);
dmu_buf_rele(old_bonus_dbuf, FTAG);
if (err != 0)
return (err);
uint64_t old_gen = zoi.zfi_generation;
err = dmu_get_file_info(os, new_bonus_type, new_bonus, &zoi);
if (err != 0)
return (err);
uint64_t new_gen = zoi.zfi_generation;
*samegenp = (old_gen == new_gen);
return (0);
}
static int
receive_handle_existing_object(const struct receive_writer_arg *rwa,
const struct drr_object *drro, const dmu_object_info_t *doi,
const void *bonus_data,
uint64_t *object_to_hold, uint32_t *new_blksz)
{
uint32_t indblksz = drro->drr_indblkshift ?
1ULL << drro->drr_indblkshift : 0;
int nblkptr = deduce_nblkptr(drro->drr_bonustype,
drro->drr_bonuslen);
uint8_t dn_slots = drro->drr_dn_slots != 0 ?
drro->drr_dn_slots : DNODE_MIN_SLOTS;
boolean_t do_free_range = B_FALSE;
int err;
*object_to_hold = drro->drr_object;
/* nblkptr should be bounded by the bonus size and type */
if (rwa->raw && nblkptr != drro->drr_nblkptr)
return (SET_ERROR(EINVAL));
/*
* After the previous send stream, the sending system may
* have freed this object, and then happened to re-allocate
* this object number in a later txg. In this case, we are
* receiving a different logical file, and the block size may
* appear to be different. i.e. we may have a different
* block size for this object than what the send stream says.
* In this case we need to remove the object's contents,
* so that its structure can be changed and then its contents
* entirely replaced by subsequent WRITE records.
*
* If this is a -L (--large-block) incremental stream, and
* the previous stream was not -L, the block size may appear
* to increase. i.e. we may have a smaller block size for
* this object than what the send stream says. In this case
* we need to keep the object's contents and block size
* intact, so that we don't lose parts of the object's
* contents that are not changed by this incremental send
* stream.
*
* We can distinguish between the two above cases by using
* the ZPL's generation number (see
* receive_object_is_same_generation()). However, we only
* want to rely on the generation number when absolutely
* necessary, because with raw receives, the generation is
* encrypted. We also want to minimize dependence on the
* ZPL, so that other types of datasets can also be received
* (e.g. ZVOLs, although note that ZVOLS currently do not
* reallocate their objects or change their structure).
* Therefore, we check a number of different cases where we
* know it is safe to discard the object's contents, before
* using the ZPL's generation number to make the above
* distinction.
*/
if (drro->drr_blksz != doi->doi_data_block_size) {
if (rwa->raw) {
/*
* RAW streams always have large blocks, so
* we are sure that the data is not needed
* due to changing --large-block to be on.
* Which is fortunate since the bonus buffer
* (which contains the ZPL generation) is
* encrypted, and the key might not be
* loaded.
*/
do_free_range = B_TRUE;
} else if (rwa->full) {
/*
* This is a full send stream, so it always
* replaces what we have. Even if the
* generation numbers happen to match, this
* can not actually be the same logical file.
* This is relevant when receiving a full
* send as a clone.
*/
do_free_range = B_TRUE;
} else if (drro->drr_type !=
DMU_OT_PLAIN_FILE_CONTENTS ||
doi->doi_type != DMU_OT_PLAIN_FILE_CONTENTS) {
/*
* PLAIN_FILE_CONTENTS are the only type of
* objects that have ever been stored with
* large blocks, so we don't need the special
* logic below. ZAP blocks can shrink (when
* there's only one block), so we don't want
* to hit the error below about block size
* only increasing.
*/
do_free_range = B_TRUE;
} else if (doi->doi_max_offset <=
doi->doi_data_block_size) {
/*
* There is only one block. We can free it,
* because its contents will be replaced by a
* WRITE record. This can not be the no-L ->
* -L case, because the no-L case would have
* resulted in multiple blocks. If we
* supported -L -> no-L, it would not be safe
* to free the file's contents. Fortunately,
* that is not allowed (see
* recv_check_large_blocks()).
*/
do_free_range = B_TRUE;
} else {
boolean_t is_same_gen;
err = receive_object_is_same_generation(rwa->os,
drro->drr_object, doi->doi_bonus_type,
drro->drr_bonustype, bonus_data, &is_same_gen);
if (err != 0)
return (SET_ERROR(EINVAL));
if (is_same_gen) {
/*
* This is the same logical file, and
* the block size must be increasing.
* It could only decrease if
* --large-block was changed to be
* off, which is checked in
* recv_check_large_blocks().
*/
if (drro->drr_blksz <=
doi->doi_data_block_size)
return (SET_ERROR(EINVAL));
/*
* We keep the existing blocksize and
* contents.
*/
*new_blksz =
doi->doi_data_block_size;
} else {
do_free_range = B_TRUE;
}
}
}
/* nblkptr can only decrease if the object was reallocated */
if (nblkptr < doi->doi_nblkptr)
do_free_range = B_TRUE;
/* number of slots can only change on reallocation */
if (dn_slots != doi->doi_dnodesize >> DNODE_SHIFT)
do_free_range = B_TRUE;
/*
* For raw sends we also check a few other fields to
* ensure we are preserving the objset structure exactly
* as it was on the receive side:
* - A changed indirect block size
* - A smaller nlevels
*/
if (rwa->raw) {
if (indblksz != doi->doi_metadata_block_size)
do_free_range = B_TRUE;
if (drro->drr_nlevels < doi->doi_indirection)
do_free_range = B_TRUE;
}
if (do_free_range) {
err = dmu_free_long_range(rwa->os, drro->drr_object,
0, DMU_OBJECT_END);
if (err != 0)
return (SET_ERROR(EINVAL));
}
/*
Fix raw receive with different indirect block size. Unlike regular receive, raw receive require destination to have the same block structure as the source. In case of dnode reclaim this triggers two special cases, requiring special handling: - If dn_nlevels == 1, we can change the ibs, but dnode_set_blksz() should not dirty the data buffer if block size does not change, or durign receive dbuf_dirty_lightweight() will trigger assertion. - If dn_nlevels > 1, we just can't change the ibs, dnode_set_blksz() would fail and receive_object would trigger assertion, so we should destroy and recreate the dnode from scratch. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Signed-off-by: Alexander Motin <mav@FreeBSD.org> Sponsored by: iXsystems, Inc. Closes #15039
2023-07-15 02:16:40 +03:00
* The dmu does not currently support decreasing nlevels or changing
* indirect block size if there is already one, same as changing the
* number of of dnode slots on an object. For non-raw sends this
* does not matter and the new object can just use the previous one's
* parameters. For raw sends, however, the structure of the received
* dnode (including indirects and dnode slots) must match that of the
* send side. Therefore, instead of using dmu_object_reclaim(), we
* must free the object completely and call dmu_object_claim_dnsize()
* instead.
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
*/
Fix raw receive with different indirect block size. Unlike regular receive, raw receive require destination to have the same block structure as the source. In case of dnode reclaim this triggers two special cases, requiring special handling: - If dn_nlevels == 1, we can change the ibs, but dnode_set_blksz() should not dirty the data buffer if block size does not change, or durign receive dbuf_dirty_lightweight() will trigger assertion. - If dn_nlevels > 1, we just can't change the ibs, dnode_set_blksz() would fail and receive_object would trigger assertion, so we should destroy and recreate the dnode from scratch. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Signed-off-by: Alexander Motin <mav@FreeBSD.org> Sponsored by: iXsystems, Inc. Closes #15039
2023-07-15 02:16:40 +03:00
if ((rwa->raw && ((doi->doi_indirection > 1 &&
indblksz != doi->doi_metadata_block_size) ||
drro->drr_nlevels < doi->doi_indirection)) ||
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
dn_slots != doi->doi_dnodesize >> DNODE_SHIFT) {
err = dmu_free_long_object(rwa->os, drro->drr_object);
if (err != 0)
return (SET_ERROR(EINVAL));
txg_wait_synced(dmu_objset_pool(rwa->os), 0);
*object_to_hold = DMU_NEW_OBJECT;
}
/*
* For raw receives, free everything beyond the new incoming
* maxblkid. Normally this would be done with a DRR_FREE
* record that would come after this DRR_OBJECT record is
* processed. However, for raw receives we manually set the
* maxblkid from the drr_maxblkid and so we must first free
* everything above that blkid to ensure the DMU is always
* consistent with itself. We will never free the first block
* of the object here because a maxblkid of 0 could indicate
* an object with a single block or one with no blocks. This
* free may be skipped when dmu_free_long_range() was called
* above since it covers the entire object's contents.
*/
if (rwa->raw && *object_to_hold != DMU_NEW_OBJECT && !do_free_range) {
err = dmu_free_long_range(rwa->os, drro->drr_object,
(drro->drr_maxblkid + 1) * doi->doi_data_block_size,
DMU_OBJECT_END);
if (err != 0)
return (SET_ERROR(EINVAL));
}
return (0);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
noinline static int
receive_object(struct receive_writer_arg *rwa, struct drr_object *drro,
void *data)
{
dmu_object_info_t doi;
dmu_tx_t *tx;
int err;
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
uint32_t new_blksz = drro->drr_blksz;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
uint8_t dn_slots = drro->drr_dn_slots != 0 ?
drro->drr_dn_slots : DNODE_MIN_SLOTS;
if (drro->drr_type == DMU_OT_NONE ||
!DMU_OT_IS_VALID(drro->drr_type) ||
!DMU_OT_IS_VALID(drro->drr_bonustype) ||
drro->drr_checksumtype >= ZIO_CHECKSUM_FUNCTIONS ||
drro->drr_compress >= ZIO_COMPRESS_FUNCTIONS ||
P2PHASE(drro->drr_blksz, SPA_MINBLOCKSIZE) ||
drro->drr_blksz < SPA_MINBLOCKSIZE ||
drro->drr_blksz > spa_maxblocksize(dmu_objset_spa(rwa->os)) ||
drro->drr_bonuslen >
DN_BONUS_SIZE(spa_maxdnodesize(dmu_objset_spa(rwa->os))) ||
dn_slots >
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
(spa_maxdnodesize(dmu_objset_spa(rwa->os)) >> DNODE_SHIFT)) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (SET_ERROR(EINVAL));
}
if (rwa->raw) {
/*
* We should have received a DRR_OBJECT_RANGE record
* containing this block and stored it in rwa.
*/
if (drro->drr_object < rwa->or_firstobj ||
drro->drr_object >= rwa->or_firstobj + rwa->or_numslots ||
drro->drr_raw_bonuslen < drro->drr_bonuslen ||
drro->drr_indblkshift > SPA_MAXBLOCKSHIFT ||
drro->drr_nlevels > DN_MAX_LEVELS ||
drro->drr_nblkptr > DN_MAX_NBLKPTR ||
DN_SLOTS_TO_BONUSLEN(dn_slots) <
drro->drr_raw_bonuslen)
return (SET_ERROR(EINVAL));
} else {
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
/*
* The DRR_OBJECT_SPILL flag is valid when the DRR_BEGIN
* record indicates this by setting DRR_FLAG_SPILL_BLOCK.
*/
if (((drro->drr_flags & ~(DRR_OBJECT_SPILL))) ||
(!rwa->spill && DRR_OBJECT_HAS_SPILL(drro->drr_flags))) {
return (SET_ERROR(EINVAL));
}
if (drro->drr_raw_bonuslen != 0 || drro->drr_nblkptr != 0 ||
drro->drr_indblkshift != 0 || drro->drr_nlevels != 0) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (SET_ERROR(EINVAL));
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
err = dmu_object_info(rwa->os, drro->drr_object, &doi);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0 && err != ENOENT && err != EEXIST)
return (SET_ERROR(EINVAL));
if (drro->drr_object > rwa->max_object)
rwa->max_object = drro->drr_object;
/*
* If we are losing blkptrs or changing the block size this must
* be a new file instance. We must clear out the previous file
* contents before we can change this type of metadata in the dnode.
* Raw receives will also check that the indirect structure of the
* dnode hasn't changed.
*/
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
uint64_t object_to_hold;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err == 0) {
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
err = receive_handle_existing_object(rwa, drro, &doi, data,
&object_to_hold, &new_blksz);
Fix unchecked return values and unused return values Coverity complained about unchecked return values and unused values that turned out to be unused return values. Different approaches were used to handle the different cases of unchecked return values: * cmd/zdb/zdb.c: VERIFY0 was used in one place since the existing code had no error handling. An error message was printed in another to match the rest of the code. * cmd/zed/agents/zfs_retire.c: We dismiss the return value with `(void)` because the value is expected to be potentially unset. * cmd/zpool_influxdb/zpool_influxdb.c: We dismiss the return value with `(void)` because the values are expected to be potentially unset. * cmd/ztest.c: VERIFY0 was used since we want failures if something goes wrong in ztest. * module/zfs/dsl_dir.c: We dismiss the return value with `(void)` because there is no guarantee that the zap entry will always be there. For example, old pools imported readonly would not have it and we do not want to fail here because of that. * module/zfs/zfs_fm.c: `fnvlist_add_*()` was used since the allocations sleep and thus can never fail. * module/zfs/zvol.c: We dismiss the return value with `(void)` because we do not need it. This matches what is already done in the analogous `zfs_replay_write2()`. * tests/zfs-tests/cmd/draid.c: We suppress one return value with `(void)` since the code handles errors already. The other return value is handled by switching to `fnvlist_lookup_uint8_array()`. * tests/zfs-tests/cmd/file/file_fadvise.c: We add error handling. * tests/zfs-tests/cmd/mmap_sync.c: We add error handling for munmap, but ignore failures on remove() with (void) since it is expected to be able to fail. * tests/zfs-tests/cmd/mmapwrite.c: We add error handling. As for unused return values, they were all in places where there was error handling, so logic was added to handle the return values. Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Closes #13920
2022-09-24 02:52:03 +03:00
if (err != 0)
return (err);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
} else if (err == EEXIST) {
/*
* The object requested is currently an interior slot of a
* multi-slot dnode. This will be resolved when the next txg
* is synced out, since the send stream will have told us
* to free this slot when we freed the associated dnode
* earlier in the stream.
*/
txg_wait_synced(dmu_objset_pool(rwa->os), 0);
Fix issue in receive_object() during reallocation When receiving an object to a previously allocated interior slot the new object should be "allocated" by setting DMU_NEW_OBJECT, not "reallocated" with dnode_reallocate(). For resilience verify the slot is free as required in case the stream is malformed. Add a test case to generate more realistic incremental send streams that force reallocation to occur during the receive. Reviewed-by: Olaf Faaland <faaland1@llnl.gov> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8067 Closes #8614
2019-04-13 00:28:04 +03:00
if (dmu_object_info(rwa->os, drro->drr_object, NULL) != ENOENT)
return (SET_ERROR(EINVAL));
/* object was freed and we are about to allocate a new one */
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
object_to_hold = DMU_NEW_OBJECT;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
} else {
Wait for txg sync if the last DRR_FREEOBJECTS might result in a hole If we receive a DRR_FREEOBJECTS as the first entry in an object range, this might end up producing a hole if the freed objects were the only existing objects in the block. If the txg starts syncing before we've processed any following DRR_OBJECT records, this leads to a possible race where the backing arc_buf_t gets its psize set to 0 in the arc_write_ready() callback while still being referenced from a dirty record in the open txg. To prevent this, we insert a txg_wait_synced call if the first record in the range was a DRR_FREEOBJECTS that actually resulted in one or more freed objects. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: David Hedberg <david.hedberg@findity.com> Sponsored by: Findity AB Closes #11893 Closes #14358
2023-01-24 00:19:43 +03:00
/*
* If the only record in this range so far was DRR_FREEOBJECTS
* with at least one actually freed object, it's possible that
* the block will now be converted to a hole. We need to wait
* for the txg to sync to prevent races.
*/
if (rwa->or_need_sync == ORNS_YES)
txg_wait_synced(dmu_objset_pool(rwa->os), 0);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* object is free and we are about to allocate a new one */
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
object_to_hold = DMU_NEW_OBJECT;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Wait for txg sync if the last DRR_FREEOBJECTS might result in a hole If we receive a DRR_FREEOBJECTS as the first entry in an object range, this might end up producing a hole if the freed objects were the only existing objects in the block. If the txg starts syncing before we've processed any following DRR_OBJECT records, this leads to a possible race where the backing arc_buf_t gets its psize set to 0 in the arc_write_ready() callback while still being referenced from a dirty record in the open txg. To prevent this, we insert a txg_wait_synced call if the first record in the range was a DRR_FREEOBJECTS that actually resulted in one or more freed objects. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: David Hedberg <david.hedberg@findity.com> Sponsored by: Findity AB Closes #11893 Closes #14358
2023-01-24 00:19:43 +03:00
/* Only relevant for the first object in the range */
rwa->or_need_sync = ORNS_NO;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* If this is a multi-slot dnode there is a chance that this
* object will expand into a slot that is already used by
* another object from the previous snapshot. We must free
* these objects before we attempt to allocate the new dnode.
*/
if (dn_slots > 1) {
boolean_t need_sync = B_FALSE;
for (uint64_t slot = drro->drr_object + 1;
slot < drro->drr_object + dn_slots;
slot++) {
dmu_object_info_t slot_doi;
err = dmu_object_info(rwa->os, slot, &slot_doi);
if (err == ENOENT || err == EEXIST)
continue;
else if (err != 0)
return (err);
err = dmu_free_long_object(rwa->os, slot);
if (err != 0)
return (err);
need_sync = B_TRUE;
}
if (need_sync)
txg_wait_synced(dmu_objset_pool(rwa->os), 0);
}
tx = dmu_tx_create(rwa->os);
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
dmu_tx_hold_bonus(tx, object_to_hold);
dmu_tx_hold_write(tx, object_to_hold, 0, 0);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
err = dmu_tx_assign(tx, TXG_WAIT);
if (err != 0) {
dmu_tx_abort(tx);
return (err);
}
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
if (object_to_hold == DMU_NEW_OBJECT) {
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
/* Currently free, wants to be allocated */
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
err = dmu_object_claim_dnsize(rwa->os, drro->drr_object,
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
drro->drr_type, new_blksz,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drro->drr_bonustype, drro->drr_bonuslen,
dn_slots << DNODE_SHIFT, tx);
} else if (drro->drr_type != doi.doi_type ||
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
new_blksz != doi.doi_data_block_size ||
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drro->drr_bonustype != doi.doi_bonus_type ||
drro->drr_bonuslen != doi.doi_bonus_size) {
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
/* Currently allocated, but with different properties */
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
err = dmu_object_reclaim_dnsize(rwa->os, drro->drr_object,
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
drro->drr_type, new_blksz,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drro->drr_bonustype, drro->drr_bonuslen,
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
dn_slots << DNODE_SHIFT, rwa->spill ?
DRR_OBJECT_HAS_SPILL(drro->drr_flags) : B_FALSE, tx);
} else if (rwa->spill && !DRR_OBJECT_HAS_SPILL(drro->drr_flags)) {
/*
* Currently allocated, the existing version of this object
* may reference a spill block that is no longer allocated
* at the source and needs to be freed.
*/
err = dmu_object_rm_spill(rwa->os, drro->drr_object, tx);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Fix TXG_MASK cstyle Fix style issue for 'tx->tx_txg&TXG_MASK'. There should be white space around the '&' character. Split the dnode_reallocate() ASSERT to make it more readable to clearly separate the checks. Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8606
2019-04-12 21:30:59 +03:00
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0) {
dmu_tx_commit(tx);
return (SET_ERROR(EINVAL));
}
if (rwa->or_crypt_params_present) {
/*
* Set the crypt params for the buffer associated with this
* range of dnodes. This causes the blkptr_t to have the
* same crypt params (byteorder, salt, iv, mac) as on the
* sending side.
*
* Since we are committing this tx now, it is possible for
* the dnode block to end up on-disk with the incorrect MAC,
* if subsequent objects in this block are received in a
* different txg. However, since the dataset is marked as
* inconsistent, no code paths will do a non-raw read (or
* decrypt the block / verify the MAC). The receive code and
* scrub code can safely do raw reads and verify the
* checksum. They don't need to verify the MAC.
*/
dmu_buf_t *db = NULL;
uint64_t offset = rwa->or_firstobj * DNODE_MIN_SIZE;
err = dmu_buf_hold_by_dnode(DMU_META_DNODE(rwa->os),
offset, FTAG, &db, DMU_READ_PREFETCH | DMU_READ_NO_DECRYPT);
if (err != 0) {
dmu_tx_commit(tx);
return (SET_ERROR(EINVAL));
}
dmu_buf_set_crypt_params(db, rwa->or_byteorder,
rwa->or_salt, rwa->or_iv, rwa->or_mac, tx);
dmu_buf_rele(db, FTAG);
rwa->or_crypt_params_present = B_FALSE;
}
dmu_object_set_checksum(rwa->os, drro->drr_object,
drro->drr_checksumtype, tx);
dmu_object_set_compress(rwa->os, drro->drr_object,
drro->drr_compress, tx);
/* handle more restrictive dnode structuring for raw recvs */
if (rwa->raw) {
/*
Fix handling of maxblkid for raw sends Currently, the receive code can create an unreadable dataset from a correct raw send stream. This is because it is currently impossible to set maxblkid to a lower value without freeing the associated object. This means truncating files on the send side to a non-0 size could result in corruption. This patch solves this issue by adding a new 'force' flag to dnode_new_blkid() which will allow the raw receive code to force the DMU to accept the provided maxblkid even if it is a lower value than the existing one. For testing purposes the send_encrypted_files.ksh test has been extended to include a variety of truncated files and multiple snapshots. It also now leverages the xattrtest command to help ensure raw receives correctly handle xattrs. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8168 Closes #8487
2019-03-13 20:52:01 +03:00
* Set the indirect block size, block shift, nlevels.
* This will not fail because we ensured all of the
* blocks were freed earlier if this is a new object.
* For non-new objects block size and indirect block
* shift cannot change and nlevels can only increase.
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
*/
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
ASSERT3U(new_blksz, ==, drro->drr_blksz);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
VERIFY0(dmu_object_set_blocksize(rwa->os, drro->drr_object,
drro->drr_blksz, drro->drr_indblkshift, tx));
VERIFY0(dmu_object_set_nlevels(rwa->os, drro->drr_object,
drro->drr_nlevels, tx));
Fix handling of maxblkid for raw sends Currently, the receive code can create an unreadable dataset from a correct raw send stream. This is because it is currently impossible to set maxblkid to a lower value without freeing the associated object. This means truncating files on the send side to a non-0 size could result in corruption. This patch solves this issue by adding a new 'force' flag to dnode_new_blkid() which will allow the raw receive code to force the DMU to accept the provided maxblkid even if it is a lower value than the existing one. For testing purposes the send_encrypted_files.ksh test has been extended to include a variety of truncated files and multiple snapshots. It also now leverages the xattrtest command to help ensure raw receives correctly handle xattrs. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8168 Closes #8487
2019-03-13 20:52:01 +03:00
/*
Fix issues with truncated files in raw sends When receiving a raw send stream only reallocated objects whose contents were not freed by the standard indicators should call dmu_free_long_range(). Furthermore, if calling dmu_free_long_range() is required then the objects current block size must be used and not the new block size. Two additional test cases were added to provided realistic test coverage for processing reallocated objects which are part of a raw receive. Reviewed-by: Olaf Faaland <faaland1@llnl.gov> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8528 Closes #8607
2019-04-10 20:17:36 +03:00
* Set the maxblkid. This will always succeed because
* we freed all blocks beyond the new maxblkid above.
Fix handling of maxblkid for raw sends Currently, the receive code can create an unreadable dataset from a correct raw send stream. This is because it is currently impossible to set maxblkid to a lower value without freeing the associated object. This means truncating files on the send side to a non-0 size could result in corruption. This patch solves this issue by adding a new 'force' flag to dnode_new_blkid() which will allow the raw receive code to force the DMU to accept the provided maxblkid even if it is a lower value than the existing one. For testing purposes the send_encrypted_files.ksh test has been extended to include a variety of truncated files and multiple snapshots. It also now leverages the xattrtest command to help ensure raw receives correctly handle xattrs. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8168 Closes #8487
2019-03-13 20:52:01 +03:00
*/
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
VERIFY0(dmu_object_set_maxblkid(rwa->os, drro->drr_object,
drro->drr_maxblkid, tx));
}
if (data != NULL) {
dmu_buf_t *db;
Provide more flexible object allocation interface Object allocation performance can be improved for complex operations by providing an interface which returns the newly allocated dnode. This allows the caller to immediately use the dnode without incurring the expense of looking up the dnode by object number. The functions dmu_object_alloc_hold(), zap_create_hold(), and dmu_bonus_hold_by_dnode() were added for this purpose. The zap_create_* functions have been updated to take advantage of this new functionality. The dmu_bonus_hold_impl() function should really have never been included in sys/dmu.h and was removed. It's sole caller was converted to use dmu_bonus_hold_by_dnode(). The new symbols have been exported for use by Lustre. Reviewed-by: Tony Hutter <hutter2@llnl.gov> Reviewed by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8015
2019-01-11 01:37:43 +03:00
dnode_t *dn;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
uint32_t flags = DMU_READ_NO_PREFETCH;
if (rwa->raw)
flags |= DMU_READ_NO_DECRYPT;
Provide more flexible object allocation interface Object allocation performance can be improved for complex operations by providing an interface which returns the newly allocated dnode. This allows the caller to immediately use the dnode without incurring the expense of looking up the dnode by object number. The functions dmu_object_alloc_hold(), zap_create_hold(), and dmu_bonus_hold_by_dnode() were added for this purpose. The zap_create_* functions have been updated to take advantage of this new functionality. The dmu_bonus_hold_impl() function should really have never been included in sys/dmu.h and was removed. It's sole caller was converted to use dmu_bonus_hold_by_dnode(). The new symbols have been exported for use by Lustre. Reviewed-by: Tony Hutter <hutter2@llnl.gov> Reviewed by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8015
2019-01-11 01:37:43 +03:00
VERIFY0(dnode_hold(rwa->os, drro->drr_object, FTAG, &dn));
VERIFY0(dmu_bonus_hold_by_dnode(dn, FTAG, &db, flags));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dmu_buf_will_dirty(db, tx);
ASSERT3U(db->db_size, >=, drro->drr_bonuslen);
Remove bcopy(), bzero(), bcmp() bcopy() has a confusing argument order and is actually a move, not a copy; they're all deprecated since POSIX.1-2001 and removed in -2008, and we shim them out to mem*() on Linux anyway Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Closes #12996
2022-02-25 16:26:54 +03:00
memcpy(db->db_data, data, DRR_OBJECT_PAYLOAD_SIZE(drro));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* Raw bonus buffers have their byteorder determined by the
* DRR_OBJECT_RANGE record.
*/
if (rwa->byteswap && !rwa->raw) {
dmu_object_byteswap_t byteswap =
DMU_OT_BYTESWAP(drro->drr_bonustype);
dmu_ot_byteswap[byteswap].ob_func(db->db_data,
DRR_OBJECT_PAYLOAD_SIZE(drro));
}
dmu_buf_rele(db, FTAG);
Provide more flexible object allocation interface Object allocation performance can be improved for complex operations by providing an interface which returns the newly allocated dnode. This allows the caller to immediately use the dnode without incurring the expense of looking up the dnode by object number. The functions dmu_object_alloc_hold(), zap_create_hold(), and dmu_bonus_hold_by_dnode() were added for this purpose. The zap_create_* functions have been updated to take advantage of this new functionality. The dmu_bonus_hold_impl() function should really have never been included in sys/dmu.h and was removed. It's sole caller was converted to use dmu_bonus_hold_by_dnode(). The new symbols have been exported for use by Lustre. Reviewed-by: Tony Hutter <hutter2@llnl.gov> Reviewed by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8015
2019-01-11 01:37:43 +03:00
dnode_rele(dn, FTAG);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Update resume token at object receive. Before this change resume token was updated only on data receive. Usually it is enough to resume replication without much overlap. But we've got a report of a curios case, where replication source was traversed with recursive grep, which through enabled atime modified every object without modifying any data. It produced several gigabytes of replication traffic without a single data write and so without a single resume point. While the resume token was not designed to resume from an object, I've found that the send implementation always sends object before any data. So by requesting resume from offset 0 we are effectively resuming from the object, followed (or not) by the data at offset 0, just as we need it. Reviewed-by: Allan Jude <allan@klarasystems.com> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Signed-off-by: Alexander Motin <mav@FreeBSD.org> Sponsored by: iXsystems, Inc. Closes #15927
2024-03-21 03:22:36 +03:00
/*
* If the receive fails, we want the resume stream to start with the
* same record that we last successfully received. There is no way to
* request resume from the object record, but we can benefit from the
* fact that sender always sends object record before anything else,
* after which it will "resend" data at offset 0 and resume normally.
*/
save_resume_state(rwa, drro->drr_object, 0, tx);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dmu_tx_commit(tx);
return (0);
}
noinline static int
receive_freeobjects(struct receive_writer_arg *rwa,
struct drr_freeobjects *drrfo)
{
uint64_t obj;
int next_err = 0;
if (drrfo->drr_firstobj + drrfo->drr_numobjs < drrfo->drr_firstobj)
return (SET_ERROR(EINVAL));
for (obj = drrfo->drr_firstobj == 0 ? 1 : drrfo->drr_firstobj;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
obj < drrfo->drr_firstobj + drrfo->drr_numobjs &&
obj < DN_MAX_OBJECT && next_err == 0;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
next_err = dmu_object_next(rwa->os, &obj, FALSE, 0)) {
dmu_object_info_t doi;
int err;
err = dmu_object_info(rwa->os, obj, &doi);
if (err == ENOENT)
continue;
else if (err != 0)
return (err);
err = dmu_free_long_object(rwa->os, obj);
if (err != 0)
return (err);
Wait for txg sync if the last DRR_FREEOBJECTS might result in a hole If we receive a DRR_FREEOBJECTS as the first entry in an object range, this might end up producing a hole if the freed objects were the only existing objects in the block. If the txg starts syncing before we've processed any following DRR_OBJECT records, this leads to a possible race where the backing arc_buf_t gets its psize set to 0 in the arc_write_ready() callback while still being referenced from a dirty record in the open txg. To prevent this, we insert a txg_wait_synced call if the first record in the range was a DRR_FREEOBJECTS that actually resulted in one or more freed objects. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: David Hedberg <david.hedberg@findity.com> Sponsored by: Findity AB Closes #11893 Closes #14358
2023-01-24 00:19:43 +03:00
if (rwa->or_need_sync == ORNS_MAYBE)
rwa->or_need_sync = ORNS_YES;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
if (next_err != ESRCH)
return (next_err);
return (0);
}
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
/*
* Note: if this fails, the caller will clean up any records left on the
* rwa->write_batch list.
*/
static int
flush_write_batch_impl(struct receive_writer_arg *rwa)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
dnode_t *dn;
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
int err;
if (dnode_hold(rwa->os, rwa->last_object, FTAG, &dn) != 0)
return (SET_ERROR(EINVAL));
struct receive_record_arg *last_rrd = list_tail(&rwa->write_batch);
struct drr_write *last_drrw = &last_rrd->header.drr_u.drr_write;
struct receive_record_arg *first_rrd = list_head(&rwa->write_batch);
struct drr_write *first_drrw = &first_rrd->header.drr_u.drr_write;
ASSERT3U(rwa->last_object, ==, last_drrw->drr_object);
ASSERT3U(rwa->last_offset, ==, last_drrw->drr_offset);
dmu_tx_t *tx = dmu_tx_create(rwa->os);
dmu_tx_hold_write_by_dnode(tx, dn, first_drrw->drr_offset,
last_drrw->drr_offset - first_drrw->drr_offset +
last_drrw->drr_logical_size);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err != 0) {
dmu_tx_abort(tx);
dnode_rele(dn, FTAG);
return (err);
}
struct receive_record_arg *rrd;
while ((rrd = list_head(&rwa->write_batch)) != NULL) {
struct drr_write *drrw = &rrd->header.drr_u.drr_write;
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_t *abd = rrd->abd;
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
ASSERT3U(drrw->drr_object, ==, rwa->last_object);
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
if (drrw->drr_logical_size != dn->dn_datablksz) {
/*
* The WRITE record is larger than the object's block
* size. We must be receiving an incremental
* large-block stream into a dataset that previously did
* a non-large-block receive. Lightweight writes must
* be exactly one block, so we need to decompress the
* data (if compressed) and do a normal dmu_write().
*/
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
ASSERT3U(drrw->drr_logical_size, >, dn->dn_datablksz);
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
if (DRR_WRITE_COMPRESSED(drrw)) {
abd_t *decomp_abd =
abd_alloc_linear(drrw->drr_logical_size,
B_FALSE);
err = zio_decompress_data(
drrw->drr_compressiontype,
abd, abd_to_buf(decomp_abd),
abd_get_size(abd),
abd_get_size(decomp_abd), NULL);
if (err == 0) {
dmu_write_by_dnode(dn,
drrw->drr_offset,
drrw->drr_logical_size,
abd_to_buf(decomp_abd), tx);
}
abd_free(decomp_abd);
} else {
dmu_write_by_dnode(dn,
drrw->drr_offset,
drrw->drr_logical_size,
abd_to_buf(abd), tx);
}
if (err == 0)
abd_free(abd);
} else {
Zero zio_prop_t in flush_write_batch_impl() After 67a1b0379159c46bcd60a462a2790248046c8804 was merged, coverity started complaining about an uninitialized scalar variable in flush_write_batch_impl() due to the new field zp.zp_brtwrite. Upon inspection, it appears that uninitialized memory was being copied for non-raw streams, so this is a pre-existing issue. The addition of zp_brtwrite by the block cloning commit caused Coverity to begin to notice it. Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Rob Norris <robn@despairlabs.com> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Reported-by: Coverity (CID-1535378) Closes #14607
2023-03-11 01:42:09 +03:00
zio_prop_t zp = {0};
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
dmu_write_policy(rwa->os, dn, 0, 0, &zp);
Convert enum zio_flag to uint64_t We ran out of space in enum zio_flag for additional flags. Rather than introduce enum zio_flag2 and then modify a bunch of functions to take a second flags variable, we expand the type to 64 bits via `typedef uint64_t zio_flag_t`. Reviewed-by: Allan Jude <allan@klarasystems.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Richard Yao <richard.yao@klarasystems.com> Signed-off-by: Allan Jude <allan@klarasystems.com> Co-authored-by: Richard Yao <richard.yao@klarasystems.com> Closes #14086
2022-10-27 19:54:54 +03:00
zio_flag_t zio_flags = 0;
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
if (rwa->raw) {
zp.zp_encrypt = B_TRUE;
zp.zp_compress = drrw->drr_compressiontype;
zp.zp_byteorder = ZFS_HOST_BYTEORDER ^
!!DRR_IS_RAW_BYTESWAPPED(drrw->drr_flags) ^
rwa->byteswap;
Remove bcopy(), bzero(), bcmp() bcopy() has a confusing argument order and is actually a move, not a copy; they're all deprecated since POSIX.1-2001 and removed in -2008, and we shim them out to mem*() on Linux anyway Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Closes #12996
2022-02-25 16:26:54 +03:00
memcpy(zp.zp_salt, drrw->drr_salt,
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
ZIO_DATA_SALT_LEN);
Remove bcopy(), bzero(), bcmp() bcopy() has a confusing argument order and is actually a move, not a copy; they're all deprecated since POSIX.1-2001 and removed in -2008, and we shim them out to mem*() on Linux anyway Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Closes #12996
2022-02-25 16:26:54 +03:00
memcpy(zp.zp_iv, drrw->drr_iv,
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
ZIO_DATA_IV_LEN);
Remove bcopy(), bzero(), bcmp() bcopy() has a confusing argument order and is actually a move, not a copy; they're all deprecated since POSIX.1-2001 and removed in -2008, and we shim them out to mem*() on Linux anyway Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Closes #12996
2022-02-25 16:26:54 +03:00
memcpy(zp.zp_mac, drrw->drr_mac,
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
ZIO_DATA_MAC_LEN);
if (DMU_OT_IS_ENCRYPTED(zp.zp_type)) {
zp.zp_nopwrite = B_FALSE;
zp.zp_copies = MIN(zp.zp_copies,
SPA_DVAS_PER_BP - 1);
}
zio_flags |= ZIO_FLAG_RAW;
} else if (DRR_WRITE_COMPRESSED(drrw)) {
ASSERT3U(drrw->drr_compressed_size, >, 0);
ASSERT3U(drrw->drr_logical_size, >=,
drrw->drr_compressed_size);
zp.zp_compress = drrw->drr_compressiontype;
zio_flags |= ZIO_FLAG_RAW_COMPRESS;
} else if (rwa->byteswap) {
/*
* Note: compressed blocks never need to be
* byteswapped, because WRITE records for
* metadata blocks are never compressed. The
* exception is raw streams, which are written
* in the original byteorder, and the byteorder
* bit is preserved in the BP by setting
* zp_byteorder above.
*/
dmu_object_byteswap_t byteswap =
DMU_OT_BYTESWAP(drrw->drr_type);
dmu_ot_byteswap[byteswap].ob_func(
abd_to_buf(abd),
DRR_WRITE_PAYLOAD_SIZE(drrw));
}
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
/*
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
* Since this data can't be read until the receive
* completes, we can do a "lightweight" write for
* improved performance.
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
*/
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
err = dmu_lightweight_write_by_dnode(dn,
drrw->drr_offset, abd, &zp, zio_flags, tx);
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
}
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
if (err != 0) {
/*
* This rrd is left on the list, so the caller will
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
* free it (and the abd).
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
*/
break;
}
/*
* Note: If the receive fails, we want the resume stream to
* start with the same record that we last successfully
* received (as opposed to the next record), so that we can
* verify that we are resuming from the correct location.
*/
save_resume_state(rwa, drrw->drr_object, drrw->drr_offset, tx);
list_remove(&rwa->write_batch, rrd);
kmem_free(rrd, sizeof (*rrd));
}
dmu_tx_commit(tx);
dnode_rele(dn, FTAG);
return (err);
}
noinline static int
flush_write_batch(struct receive_writer_arg *rwa)
{
if (list_is_empty(&rwa->write_batch))
return (0);
int err = rwa->err;
if (err == 0)
err = flush_write_batch_impl(rwa);
if (err != 0) {
struct receive_record_arg *rrd;
while ((rrd = list_remove_head(&rwa->write_batch)) != NULL) {
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_free(rrd->abd);
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
kmem_free(rrd, sizeof (*rrd));
}
}
ASSERT(list_is_empty(&rwa->write_batch));
return (err);
}
noinline static int
receive_process_write_record(struct receive_writer_arg *rwa,
struct receive_record_arg *rrd)
{
int err = 0;
ASSERT3U(rrd->header.drr_type, ==, DRR_WRITE);
struct drr_write *drrw = &rrd->header.drr_u.drr_write;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (drrw->drr_offset + drrw->drr_logical_size < drrw->drr_offset ||
!DMU_OT_IS_VALID(drrw->drr_type))
return (SET_ERROR(EINVAL));
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (rwa->heal) {
blkptr_t *bp;
dmu_buf_t *dbp;
int flags = DB_RF_CANFAIL;
if (rwa->raw)
flags |= DB_RF_NO_DECRYPT;
if (rwa->byteswap) {
dmu_object_byteswap_t byteswap =
DMU_OT_BYTESWAP(drrw->drr_type);
dmu_ot_byteswap[byteswap].ob_func(abd_to_buf(rrd->abd),
DRR_WRITE_PAYLOAD_SIZE(drrw));
}
err = dmu_buf_hold_noread(rwa->os, drrw->drr_object,
drrw->drr_offset, FTAG, &dbp);
if (err != 0)
return (err);
/* Try to read the object to see if it needs healing */
err = dbuf_read((dmu_buf_impl_t *)dbp, NULL, flags);
/*
* We only try to heal when dbuf_read() returns a ECKSUMs.
* Other errors (even EIO) get returned to caller.
* EIO indicates that the device is not present/accessible,
* so writing to it will likely fail.
* If the block is healthy, we don't want to overwrite it
* unnecessarily.
*/
if (err != ECKSUM) {
dmu_buf_rele(dbp, FTAG);
return (err);
}
/* Make sure the on-disk block and recv record sizes match */
ZAP: Massively switch to _by_dnode() interfaces Before this change ZAP called dnode_hold() for almost every block access, that was clearly visible in profiler under heavy load, such as BRT. This patch makes it always hold the dnode reference between zap_lockdir() and zap_unlockdir(). It allows to avoid most of dnode operations between those. It also adds several new _by_dnode() APIs to ZAP and uses them in BRT code. Also adds dmu_prefetch_by_dnode() variant and uses it in the ZAP code. After this there remains only one call to dmu_buf_dnode_enter(), which seems to be unneeded. So remove the call and the functions. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Alexander Motin <mav@FreeBSD.org> Sponsored by: iXsystems, Inc. Closes #15951
2024-03-26 00:58:50 +03:00
if (drrw->drr_logical_size != dbp->db_size) {
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
err = ENOTSUP;
dmu_buf_rele(dbp, FTAG);
return (err);
}
/* Get the block pointer for the corrupted block */
bp = dmu_buf_get_blkptr(dbp);
err = do_corrective_recv(rwa, drrw, rrd, bp);
dmu_buf_rele(dbp, FTAG);
return (err);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* For resuming to work, records must be in increasing order
* by (object, offset).
*/
if (drrw->drr_object < rwa->last_object ||
(drrw->drr_object == rwa->last_object &&
drrw->drr_offset < rwa->last_offset)) {
return (SET_ERROR(EINVAL));
}
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
struct receive_record_arg *first_rrd = list_head(&rwa->write_batch);
struct drr_write *first_drrw = &first_rrd->header.drr_u.drr_write;
uint64_t batch_size =
MIN(zfs_recv_write_batch_size, DMU_MAX_ACCESS / 2);
if (first_rrd != NULL &&
(drrw->drr_object != first_drrw->drr_object ||
drrw->drr_offset >= first_drrw->drr_offset + batch_size)) {
err = flush_write_batch(rwa);
if (err != 0)
return (err);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
rwa->last_object = drrw->drr_object;
rwa->last_offset = drrw->drr_offset;
if (rwa->last_object > rwa->max_object)
rwa->max_object = rwa->last_object;
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
list_insert_tail(&rwa->write_batch, rrd);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
* Return EAGAIN to indicate that we will use this rrd again,
* so the caller should not free it
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
*/
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
return (EAGAIN);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
static int
receive_write_embedded(struct receive_writer_arg *rwa,
struct drr_write_embedded *drrwe, void *data)
{
dmu_tx_t *tx;
int err;
if (drrwe->drr_offset + drrwe->drr_length < drrwe->drr_offset)
return (SET_ERROR(EINVAL));
if (drrwe->drr_psize > BPE_PAYLOAD_SIZE)
return (SET_ERROR(EINVAL));
if (drrwe->drr_etype >= NUM_BP_EMBEDDED_TYPES)
return (SET_ERROR(EINVAL));
if (drrwe->drr_compression >= ZIO_COMPRESS_FUNCTIONS)
return (SET_ERROR(EINVAL));
if (rwa->raw)
return (SET_ERROR(EINVAL));
if (drrwe->drr_object > rwa->max_object)
rwa->max_object = drrwe->drr_object;
tx = dmu_tx_create(rwa->os);
dmu_tx_hold_write(tx, drrwe->drr_object,
drrwe->drr_offset, drrwe->drr_length);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err != 0) {
dmu_tx_abort(tx);
return (err);
}
dmu_write_embedded(rwa->os, drrwe->drr_object,
drrwe->drr_offset, data, drrwe->drr_etype,
drrwe->drr_compression, drrwe->drr_lsize, drrwe->drr_psize,
rwa->byteswap ^ ZFS_HOST_BYTEORDER, tx);
/* See comment in restore_write. */
save_resume_state(rwa, drrwe->drr_object, drrwe->drr_offset, tx);
dmu_tx_commit(tx);
return (0);
}
static int
receive_spill(struct receive_writer_arg *rwa, struct drr_spill *drrs,
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_t *abd)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
dmu_buf_t *db, *db_spill;
int err;
if (drrs->drr_length < SPA_MINBLOCKSIZE ||
drrs->drr_length > spa_maxblocksize(dmu_objset_spa(rwa->os)))
return (SET_ERROR(EINVAL));
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
/*
* This is an unmodified spill block which was added to the stream
* to resolve an issue with incorrectly removing spill blocks. It
* should be ignored by current versions of the code which support
* the DRR_FLAG_SPILL_BLOCK flag.
*/
if (rwa->spill && DRR_SPILL_IS_UNMODIFIED(drrs->drr_flags)) {
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_free(abd);
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
return (0);
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (rwa->raw) {
if (!DMU_OT_IS_VALID(drrs->drr_type) ||
drrs->drr_compressiontype >= ZIO_COMPRESS_FUNCTIONS ||
drrs->drr_compressed_size == 0)
return (SET_ERROR(EINVAL));
}
if (dmu_object_info(rwa->os, drrs->drr_object, NULL) != 0)
return (SET_ERROR(EINVAL));
if (drrs->drr_object > rwa->max_object)
rwa->max_object = drrs->drr_object;
VERIFY0(dmu_bonus_hold(rwa->os, drrs->drr_object, FTAG, &db));
if ((err = dmu_spill_hold_by_bonus(db, DMU_READ_NO_DECRYPT, FTAG,
&db_spill)) != 0) {
dmu_buf_rele(db, FTAG);
return (err);
}
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
dmu_tx_t *tx = dmu_tx_create(rwa->os);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dmu_tx_hold_spill(tx, db->db_object);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err != 0) {
dmu_buf_rele(db, FTAG);
dmu_buf_rele(db_spill, FTAG);
dmu_tx_abort(tx);
return (err);
}
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
/*
* Spill blocks may both grow and shrink. When a change in size
* occurs any existing dbuf must be updated to match the logical
* size of the provided arc_buf_t.
*/
if (db_spill->db_size != drrs->drr_length) {
dmu: Allow buffer fills to fail When ZFS overwrites a whole block, it does not bother to read the old content from disk. It is a good optimization, but if the buffer fill fails due to page fault or something else, the buffer ends up corrupted, neither keeping old content, nor getting the new one. On FreeBSD this is additionally complicated by page faults being blocked by VFS layer, always returning EFAULT on attempt to write from mmap()'ed but not yet cached address range. Normally it is not a big problem, since after original failure VFS will retry the write after reading the required data. The problem becomes worse in specific case when somebody tries to write into a file its own mmap()'ed content from the same location. In that situation the only copy of the data is getting corrupted on the page fault and the following retries only fixate the status quo. Block cloning makes this issue easier to reproduce, since it does not read the old data, unlike traditional file copy, that may work by chance. This patch provides the fill status to dmu_buf_fill_done(), that in case of error can destroy the corrupted buffer as if no write happened. One more complication in case of block cloning is that if error is possible during fill, dmu_buf_will_fill() must read the data via fall-back to dmu_buf_will_dirty(). It is required to allow in case of error restoring the buffer to a state after the cloning, not not before it, that would happen if we just call dbuf_undirty(). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Rob Norris <robn@despairlabs.com> Signed-off-by: Alexander Motin <mav@FreeBSD.org> Sponsored by: iXsystems, Inc. Closes #15665
2023-12-15 20:51:41 +03:00
dmu_buf_will_fill(db_spill, tx, B_FALSE);
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
VERIFY0(dbuf_spill_set_blksz(db_spill,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drrs->drr_length, tx));
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
arc_buf_t *abuf;
if (rwa->raw) {
boolean_t byteorder = ZFS_HOST_BYTEORDER ^
!!DRR_IS_RAW_BYTESWAPPED(drrs->drr_flags) ^
rwa->byteswap;
abuf = arc_loan_raw_buf(dmu_objset_spa(rwa->os),
drrs->drr_object, byteorder, drrs->drr_salt,
drrs->drr_iv, drrs->drr_mac, drrs->drr_type,
drrs->drr_compressed_size, drrs->drr_length,
drrs->drr_compressiontype, 0);
} else {
abuf = arc_loan_buf(dmu_objset_spa(rwa->os),
DMU_OT_IS_METADATA(drrs->drr_type),
drrs->drr_length);
if (rwa->byteswap) {
dmu_object_byteswap_t byteswap =
DMU_OT_BYTESWAP(drrs->drr_type);
dmu_ot_byteswap[byteswap].ob_func(abd_to_buf(abd),
DRR_SPILL_PAYLOAD_SIZE(drrs));
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Remove bcopy(), bzero(), bcmp() bcopy() has a confusing argument order and is actually a move, not a copy; they're all deprecated since POSIX.1-2001 and removed in -2008, and we shim them out to mem*() on Linux anyway Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Closes #12996
2022-02-25 16:26:54 +03:00
memcpy(abuf->b_data, abd_to_buf(abd), DRR_SPILL_PAYLOAD_SIZE(drrs));
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_free(abd);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dbuf_assign_arcbuf((dmu_buf_impl_t *)db_spill, abuf, tx);
dmu_buf_rele(db, FTAG);
dmu_buf_rele(db_spill, FTAG);
dmu_tx_commit(tx);
return (0);
}
noinline static int
receive_free(struct receive_writer_arg *rwa, struct drr_free *drrf)
{
int err;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drrf->drr_length != -1ULL &&
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drrf->drr_offset + drrf->drr_length < drrf->drr_offset)
return (SET_ERROR(EINVAL));
if (dmu_object_info(rwa->os, drrf->drr_object, NULL) != 0)
return (SET_ERROR(EINVAL));
if (drrf->drr_object > rwa->max_object)
rwa->max_object = drrf->drr_object;
err = dmu_free_long_range(rwa->os, drrf->drr_object,
drrf->drr_offset, drrf->drr_length);
return (err);
}
static int
receive_object_range(struct receive_writer_arg *rwa,
struct drr_object_range *drror)
{
/*
* By default, we assume this block is in our native format
* (ZFS_HOST_BYTEORDER). We then take into account whether
* the send stream is byteswapped (rwa->byteswap). Finally,
* we need to byteswap again if this particular block was
* in non-native format on the send side.
*/
boolean_t byteorder = ZFS_HOST_BYTEORDER ^ rwa->byteswap ^
!!DRR_IS_RAW_BYTESWAPPED(drror->drr_flags);
/*
* Since dnode block sizes are constant, we should not need to worry
* about making sure that the dnode block size is the same on the
* sending and receiving sides for the time being. For non-raw sends,
* this does not matter (and in fact we do not send a DRR_OBJECT_RANGE
* record at all). Raw sends require this record type because the
* encryption parameters are used to protect an entire block of bonus
* buffers. If the size of dnode blocks ever becomes variable,
* handling will need to be added to ensure that dnode block sizes
* match on the sending and receiving side.
*/
if (drror->drr_numslots != DNODES_PER_BLOCK ||
P2PHASE(drror->drr_firstobj, DNODES_PER_BLOCK) != 0 ||
!rwa->raw)
return (SET_ERROR(EINVAL));
if (drror->drr_firstobj > rwa->max_object)
rwa->max_object = drror->drr_firstobj;
/*
* The DRR_OBJECT_RANGE handling must be deferred to receive_object()
* so that the block of dnodes is not written out when it's empty,
* and converted to a HOLE BP.
*/
rwa->or_crypt_params_present = B_TRUE;
rwa->or_firstobj = drror->drr_firstobj;
rwa->or_numslots = drror->drr_numslots;
Remove bcopy(), bzero(), bcmp() bcopy() has a confusing argument order and is actually a move, not a copy; they're all deprecated since POSIX.1-2001 and removed in -2008, and we shim them out to mem*() on Linux anyway Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz> Closes #12996
2022-02-25 16:26:54 +03:00
memcpy(rwa->or_salt, drror->drr_salt, ZIO_DATA_SALT_LEN);
memcpy(rwa->or_iv, drror->drr_iv, ZIO_DATA_IV_LEN);
memcpy(rwa->or_mac, drror->drr_mac, ZIO_DATA_MAC_LEN);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
rwa->or_byteorder = byteorder;
Wait for txg sync if the last DRR_FREEOBJECTS might result in a hole If we receive a DRR_FREEOBJECTS as the first entry in an object range, this might end up producing a hole if the freed objects were the only existing objects in the block. If the txg starts syncing before we've processed any following DRR_OBJECT records, this leads to a possible race where the backing arc_buf_t gets its psize set to 0 in the arc_write_ready() callback while still being referenced from a dirty record in the open txg. To prevent this, we insert a txg_wait_synced call if the first record in the range was a DRR_FREEOBJECTS that actually resulted in one or more freed objects. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: David Hedberg <david.hedberg@findity.com> Sponsored by: Findity AB Closes #11893 Closes #14358
2023-01-24 00:19:43 +03:00
rwa->or_need_sync = ORNS_MAYBE;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (0);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
/*
* Until we have the ability to redact large ranges of data efficiently, we
* process these records as frees.
*/
noinline static int
receive_redact(struct receive_writer_arg *rwa, struct drr_redact *drrr)
{
struct drr_free drrf = {0};
drrf.drr_length = drrr->drr_length;
drrf.drr_object = drrr->drr_object;
drrf.drr_offset = drrr->drr_offset;
drrf.drr_toguid = drrr->drr_toguid;
return (receive_free(rwa, &drrf));
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* used to destroy the drc_ds on error */
static void
dmu_recv_cleanup_ds(dmu_recv_cookie_t *drc)
{
dsl_dataset_t *ds = drc->drc_ds;
implicit conversion from 'boolean_t' to 'ds_hold_flags_t' Build error on illumos with gcc 10 did reveal: In function 'dmu_objset_refresh_ownership': ../../common/fs/zfs/dmu_objset.c:857:25: error: implicit conversion from 'boolean_t' to 'ds_hold_flags_t' {aka 'enum ds_hold_flags'} [-Werror=enum-conversion] 857 | dsl_dataset_disown(ds, decrypt, tag); | ^~~~~~~ cc1: all warnings being treated as errors libzfs_input_check.c: In function 'zfs_ioc_input_tests': libzfs_input_check.c:754:28: error: implicit conversion from 'enum dmu_objset_type' to 'enum lzc_dataset_type' [-Werror=enum-conversion] 754 | err = lzc_create(dataset, DMU_OST_ZFS, NULL, NULL, 0); | ^~~~~~~~~~~ cc1: all warnings being treated as errors The same issue is present in openzfs, and also the same issue about ds_hold_flags_t, which currently defines exactly one valid value. Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Toomas Soome <tsoome@me.com> Closes #11406
2020-12-28 03:31:02 +03:00
ds_hold_flags_t dsflags;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
implicit conversion from 'boolean_t' to 'ds_hold_flags_t' Build error on illumos with gcc 10 did reveal: In function 'dmu_objset_refresh_ownership': ../../common/fs/zfs/dmu_objset.c:857:25: error: implicit conversion from 'boolean_t' to 'ds_hold_flags_t' {aka 'enum ds_hold_flags'} [-Werror=enum-conversion] 857 | dsl_dataset_disown(ds, decrypt, tag); | ^~~~~~~ cc1: all warnings being treated as errors libzfs_input_check.c: In function 'zfs_ioc_input_tests': libzfs_input_check.c:754:28: error: implicit conversion from 'enum dmu_objset_type' to 'enum lzc_dataset_type' [-Werror=enum-conversion] 754 | err = lzc_create(dataset, DMU_OST_ZFS, NULL, NULL, 0); | ^~~~~~~~~~~ cc1: all warnings being treated as errors The same issue is present in openzfs, and also the same issue about ds_hold_flags_t, which currently defines exactly one valid value. Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Toomas Soome <tsoome@me.com> Closes #11406
2020-12-28 03:31:02 +03:00
dsflags = (drc->drc_raw) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* Wait for the txg sync before cleaning up the receive. For
* resumable receives, this ensures that our resume state has
* been written out to disk. For raw receives, this ensures
* that the user accounting code will not attempt to do anything
* after we stopped receiving the dataset.
*/
txg_wait_synced(ds->ds_dir->dd_pool, 0);
ds->ds_objset->os_raw_receive = B_FALSE;
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
Fix errata #4 handling for resuming streams Currently, the handling for errata #4 has two issues which allow the checks for this issue to be bypassed using resumable sends. The first issue is that drc->drc_fromsnapobj is not set in the resuming code as it is in the non-resuming code. This causes dsl_crypto_recv_key_check() to skip its checks for the from_ivset_guid. The second issue is that resumable sends do not clean up their on-disk state if they fail the checks in dmu_recv_stream() that happen before any data is received. As a result of these two bugs, a user can attempt a resumable send of a dataset without a from_ivset_guid. This will fail the initial dmu_recv_stream() checks, leaving a valid resume state. The send can then be resumed, which skips those checks, allowing the receive to be completed. This commit fixes these issues by setting drc->drc_fromsnapobj in the resuming receive path and by ensuring that resumablereceives are properly cleaned up if they fail the initial dmu_recv_stream() checks. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #9818 Closes #9829
2020-01-14 23:25:20 +03:00
if (drc->drc_resumable && drc->drc_should_save &&
!BP_IS_HOLE(dsl_dataset_get_blkptr(ds))) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
rrw_exit(&ds->ds_bp_rwlock, FTAG);
dsl_dataset_disown(ds, dsflags, dmu_recv_tag);
} else {
char name[ZFS_MAX_DATASET_NAME_LEN];
rrw_exit(&ds->ds_bp_rwlock, FTAG);
dsl_dataset_name(ds, name);
dsl_dataset_disown(ds, dsflags, dmu_recv_tag);
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (!drc->drc_heal)
(void) dsl_destroy_head(name);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
}
static void
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_cksum(dmu_recv_cookie_t *drc, int len, void *buf)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_byteswap) {
(void) fletcher_4_incremental_byteswap(buf, len,
&drc->drc_cksum);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
} else {
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
(void) fletcher_4_incremental_native(buf, len, &drc->drc_cksum);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
}
/*
* Read the payload into a buffer of size len, and update the current record's
* payload field.
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
* Allocate drc->drc_next_rrd and read the next record's header into
* drc->drc_next_rrd->header.
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
* Verify checksum of payload and next record.
*/
static int
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_read_payload_and_next_header(dmu_recv_cookie_t *drc, int len, void *buf)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
int err;
if (len != 0) {
ASSERT3U(len, <=, SPA_MAXBLOCKSIZE);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
err = receive_read(drc, len, buf);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0)
return (err);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_cksum(drc, len, buf);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* note: rrd is NULL when reading the begin record's payload */
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_rrd != NULL) {
drc->drc_rrd->payload = buf;
drc->drc_rrd->payload_size = len;
drc->drc_rrd->bytes_read = drc->drc_bytes_read;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Fix 0 byte memory leak in zfs receive Currently, when a DRR_OBJECT record is read into memory in receive_read_record(), memory is allocated for the bonus buffer. However, if the object doesn't have a bonus buffer the code will still "allocate" the zero bytes, but the memory will not be passed to the processing thread for cleanup later. This causes the spl kmem tracking code to report a leak. This patch simply changes the code so that it only allocates this memory if it has a non-zero length. Reviewed by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8266
2019-01-18 22:06:48 +03:00
} else {
ASSERT3P(buf, ==, NULL);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
drc->drc_prev_cksum = drc->drc_cksum;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
drc->drc_next_rrd = kmem_zalloc(sizeof (*drc->drc_next_rrd), KM_SLEEP);
err = receive_read(drc, sizeof (drc->drc_next_rrd->header),
&drc->drc_next_rrd->header);
drc->drc_next_rrd->bytes_read = drc->drc_bytes_read;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0) {
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
drc->drc_next_rrd = NULL;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (err);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_next_rrd->header.drr_type == DRR_BEGIN) {
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
drc->drc_next_rrd = NULL;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (SET_ERROR(EINVAL));
}
/*
* Note: checksum is of everything up to but not including the
* checksum itself.
*/
ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_cksum(drc,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
&drc->drc_next_rrd->header);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
zio_cksum_t cksum_orig =
drc->drc_next_rrd->header.drr_u.drr_checksum.drr_checksum;
zio_cksum_t *cksump =
&drc->drc_next_rrd->header.drr_u.drr_checksum.drr_checksum;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_byteswap)
byteswap_record(&drc->drc_next_rrd->header);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if ((!ZIO_CHECKSUM_IS_ZERO(cksump)) &&
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
!ZIO_CHECKSUM_EQUAL(drc->drc_cksum, *cksump)) {
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
drc->drc_next_rrd = NULL;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (SET_ERROR(ECKSUM));
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_cksum(drc, sizeof (cksum_orig), &cksum_orig);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (0);
}
/*
* Issue the prefetch reads for any necessary indirect blocks.
*
* We use the object ignore list to tell us whether or not to issue prefetches
* for a given object. We do this for both correctness (in case the blocksize
* of an object has changed) and performance (if the object doesn't exist, don't
* needlessly try to issue prefetches). We also trim the list as we go through
* the stream to prevent it from growing to an unbounded size.
*
* The object numbers within will always be in sorted order, and any write
* records we see will also be in sorted order, but they're not sorted with
* respect to each other (i.e. we can get several object records before
* receiving each object's write records). As a result, once we've reached a
* given object number, we can safely remove any reference to lower object
* numbers in the ignore list. In practice, we receive up to 32 object records
* before receiving write records, so the list can have up to 32 nodes in it.
*/
static void
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_read_prefetch(dmu_recv_cookie_t *drc, uint64_t object, uint64_t offset,
uint64_t length)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (!objlist_exists(drc->drc_ignore_objlist, object)) {
dmu_prefetch(drc->drc_os, object, 1, offset, length,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
ZIO_PRIORITY_SYNC_READ);
}
}
/*
* Read records off the stream, issuing any necessary prefetches.
*/
static int
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_read_record(dmu_recv_cookie_t *drc)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
int err;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
switch (drc->drc_rrd->header.drr_type) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
case DRR_OBJECT:
{
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
struct drr_object *drro =
&drc->drc_rrd->header.drr_u.drr_object;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
uint32_t size = DRR_OBJECT_PAYLOAD_SIZE(drro);
Fix 0 byte memory leak in zfs receive Currently, when a DRR_OBJECT record is read into memory in receive_read_record(), memory is allocated for the bonus buffer. However, if the object doesn't have a bonus buffer the code will still "allocate" the zero bytes, but the memory will not be passed to the processing thread for cleanup later. This causes the spl kmem tracking code to report a leak. This patch simply changes the code so that it only allocates this memory if it has a non-zero length. Reviewed by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8266
2019-01-18 22:06:48 +03:00
void *buf = NULL;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dmu_object_info_t doi;
Fix 0 byte memory leak in zfs receive Currently, when a DRR_OBJECT record is read into memory in receive_read_record(), memory is allocated for the bonus buffer. However, if the object doesn't have a bonus buffer the code will still "allocate" the zero bytes, but the memory will not be passed to the processing thread for cleanup later. This causes the spl kmem tracking code to report a leak. This patch simply changes the code so that it only allocates this memory if it has a non-zero length. Reviewed by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8266
2019-01-18 22:06:48 +03:00
if (size != 0)
buf = kmem_zalloc(size, KM_SLEEP);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
err = receive_read_payload_and_next_header(drc, size, buf);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0) {
kmem_free(buf, size);
return (err);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
err = dmu_object_info(drc->drc_os, drro->drr_object, &doi);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* See receive_read_prefetch for an explanation why we're
* storing this object in the ignore_obj_list.
*/
if (err == ENOENT || err == EEXIST ||
(err == 0 && doi.doi_data_block_size != drro->drr_blksz)) {
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
objlist_insert(drc->drc_ignore_objlist,
drro->drr_object);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
err = 0;
}
return (err);
}
case DRR_FREEOBJECTS:
{
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
err = receive_read_payload_and_next_header(drc, 0, NULL);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (err);
}
case DRR_WRITE:
{
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
struct drr_write *drrw = &drc->drc_rrd->header.drr_u.drr_write;
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
int size = DRR_WRITE_PAYLOAD_SIZE(drrw);
abd_t *abd = abd_alloc_linear(size, B_FALSE);
err = receive_read_payload_and_next_header(drc, size,
abd_to_buf(abd));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0) {
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_free(abd);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (err);
}
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
drc->drc_rrd->abd = abd;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_read_prefetch(drc, drrw->drr_object, drrw->drr_offset,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drrw->drr_logical_size);
return (err);
}
case DRR_WRITE_EMBEDDED:
{
struct drr_write_embedded *drrwe =
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
&drc->drc_rrd->header.drr_u.drr_write_embedded;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
uint32_t size = P2ROUNDUP(drrwe->drr_psize, 8);
void *buf = kmem_zalloc(size, KM_SLEEP);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
err = receive_read_payload_and_next_header(drc, size, buf);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0) {
kmem_free(buf, size);
return (err);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
receive_read_prefetch(drc, drrwe->drr_object, drrwe->drr_offset,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drrwe->drr_length);
return (err);
}
case DRR_FREE:
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
case DRR_REDACT:
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
/*
* It might be beneficial to prefetch indirect blocks here, but
* we don't really have the data to decide for sure.
*/
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
err = receive_read_payload_and_next_header(drc, 0, NULL);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (err);
}
case DRR_END:
{
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
struct drr_end *drre = &drc->drc_rrd->header.drr_u.drr_end;
if (!ZIO_CHECKSUM_EQUAL(drc->drc_prev_cksum,
drre->drr_checksum))
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (SET_ERROR(ECKSUM));
return (0);
}
case DRR_SPILL:
{
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
struct drr_spill *drrs = &drc->drc_rrd->header.drr_u.drr_spill;
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
int size = DRR_SPILL_PAYLOAD_SIZE(drrs);
abd_t *abd = abd_alloc_linear(size, B_FALSE);
err = receive_read_payload_and_next_header(drc, size,
abd_to_buf(abd));
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (err != 0)
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_free(abd);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
else
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
drc->drc_rrd->abd = abd;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (err);
}
case DRR_OBJECT_RANGE:
{
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
err = receive_read_payload_and_next_header(drc, 0, NULL);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (err);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
default:
return (SET_ERROR(EINVAL));
}
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
static void
dprintf_drr(struct receive_record_arg *rrd, int err)
{
#ifdef ZFS_DEBUG
switch (rrd->header.drr_type) {
case DRR_OBJECT:
{
struct drr_object *drro = &rrd->header.drr_u.drr_object;
dprintf("drr_type = OBJECT obj = %llu type = %u "
"bonustype = %u blksz = %u bonuslen = %u cksumtype = %u "
"compress = %u dn_slots = %u err = %d\n",
Annotated dprintf as printf-like ZFS loves using %llu for uint64_t, but that requires a cast to not be noisy - which is even done in many, though not all, places. Also a couple places used %u for uint64_t, which were promoted to %llu. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Rich Ercolani <rincebrain@gmail.com> Closes #12233
2021-06-23 07:53:45 +03:00
(u_longlong_t)drro->drr_object, drro->drr_type,
drro->drr_bonustype, drro->drr_blksz, drro->drr_bonuslen,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drro->drr_checksumtype, drro->drr_compress,
drro->drr_dn_slots, err);
break;
}
case DRR_FREEOBJECTS:
{
struct drr_freeobjects *drrfo =
&rrd->header.drr_u.drr_freeobjects;
dprintf("drr_type = FREEOBJECTS firstobj = %llu "
"numobjs = %llu err = %d\n",
Annotated dprintf as printf-like ZFS loves using %llu for uint64_t, but that requires a cast to not be noisy - which is even done in many, though not all, places. Also a couple places used %u for uint64_t, which were promoted to %llu. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Rich Ercolani <rincebrain@gmail.com> Closes #12233
2021-06-23 07:53:45 +03:00
(u_longlong_t)drrfo->drr_firstobj,
(u_longlong_t)drrfo->drr_numobjs, err);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
break;
}
case DRR_WRITE:
{
struct drr_write *drrw = &rrd->header.drr_u.drr_write;
dprintf("drr_type = WRITE obj = %llu type = %u offset = %llu "
Fix issues with truncated files in raw sends This patch fixes a few issues with raw receives involving truncated files: * dnode_reallocate() now calls dnode_set_blksz() instead of dnode_setdblksz(). This ensures that any remaining dbufs with blkid 0 are resized along with their containing dnode upon reallocation. * One of the calls to dmu_free_long_range() in receive_object() needs to check that the object it is about to free some contents or hasn't been completely removed already by a previous call to dmu_free_long_object() in the same function. * The same call to dmu_free_long_range() in the previous point needs to ensure it uses the object's current block size and not the new block size. This ensures the blocks of the object that are supposed to be freed are completely removed and not simply partially zeroed out. This patch also adds handling for DRR_OBJECT_RANGE records to dprintf_drr() for debugging purposes. Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #7378 Closes #8528
2019-03-27 21:30:48 +03:00
"lsize = %llu cksumtype = %u flags = %u "
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
"compress = %u psize = %llu err = %d\n",
Annotated dprintf as printf-like ZFS loves using %llu for uint64_t, but that requires a cast to not be noisy - which is even done in many, though not all, places. Also a couple places used %u for uint64_t, which were promoted to %llu. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Rich Ercolani <rincebrain@gmail.com> Closes #12233
2021-06-23 07:53:45 +03:00
(u_longlong_t)drrw->drr_object, drrw->drr_type,
(u_longlong_t)drrw->drr_offset,
(u_longlong_t)drrw->drr_logical_size,
drrw->drr_checksumtype, drrw->drr_flags,
drrw->drr_compressiontype,
(u_longlong_t)drrw->drr_compressed_size, err);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
break;
}
case DRR_WRITE_BYREF:
{
struct drr_write_byref *drrwbr =
&rrd->header.drr_u.drr_write_byref;
dprintf("drr_type = WRITE_BYREF obj = %llu offset = %llu "
"length = %llu toguid = %llx refguid = %llx "
"refobject = %llu refoffset = %llu cksumtype = %u "
Fix issues with truncated files in raw sends This patch fixes a few issues with raw receives involving truncated files: * dnode_reallocate() now calls dnode_set_blksz() instead of dnode_setdblksz(). This ensures that any remaining dbufs with blkid 0 are resized along with their containing dnode upon reallocation. * One of the calls to dmu_free_long_range() in receive_object() needs to check that the object it is about to free some contents or hasn't been completely removed already by a previous call to dmu_free_long_object() in the same function. * The same call to dmu_free_long_range() in the previous point needs to ensure it uses the object's current block size and not the new block size. This ensures the blocks of the object that are supposed to be freed are completely removed and not simply partially zeroed out. This patch also adds handling for DRR_OBJECT_RANGE records to dprintf_drr() for debugging purposes. Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #7378 Closes #8528
2019-03-27 21:30:48 +03:00
"flags = %u err = %d\n",
Annotated dprintf as printf-like ZFS loves using %llu for uint64_t, but that requires a cast to not be noisy - which is even done in many, though not all, places. Also a couple places used %u for uint64_t, which were promoted to %llu. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Rich Ercolani <rincebrain@gmail.com> Closes #12233
2021-06-23 07:53:45 +03:00
(u_longlong_t)drrwbr->drr_object,
(u_longlong_t)drrwbr->drr_offset,
(u_longlong_t)drrwbr->drr_length,
(u_longlong_t)drrwbr->drr_toguid,
(u_longlong_t)drrwbr->drr_refguid,
(u_longlong_t)drrwbr->drr_refobject,
(u_longlong_t)drrwbr->drr_refoffset,
drrwbr->drr_checksumtype, drrwbr->drr_flags, err);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
break;
}
case DRR_WRITE_EMBEDDED:
{
struct drr_write_embedded *drrwe =
&rrd->header.drr_u.drr_write_embedded;
dprintf("drr_type = WRITE_EMBEDDED obj = %llu offset = %llu "
"length = %llu compress = %u etype = %u lsize = %u "
"psize = %u err = %d\n",
Annotated dprintf as printf-like ZFS loves using %llu for uint64_t, but that requires a cast to not be noisy - which is even done in many, though not all, places. Also a couple places used %u for uint64_t, which were promoted to %llu. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Rich Ercolani <rincebrain@gmail.com> Closes #12233
2021-06-23 07:53:45 +03:00
(u_longlong_t)drrwe->drr_object,
(u_longlong_t)drrwe->drr_offset,
(u_longlong_t)drrwe->drr_length,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
drrwe->drr_compression, drrwe->drr_etype,
drrwe->drr_lsize, drrwe->drr_psize, err);
break;
}
case DRR_FREE:
{
struct drr_free *drrf = &rrd->header.drr_u.drr_free;
dprintf("drr_type = FREE obj = %llu offset = %llu "
"length = %lld err = %d\n",
Annotated dprintf as printf-like ZFS loves using %llu for uint64_t, but that requires a cast to not be noisy - which is even done in many, though not all, places. Also a couple places used %u for uint64_t, which were promoted to %llu. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Rich Ercolani <rincebrain@gmail.com> Closes #12233
2021-06-23 07:53:45 +03:00
(u_longlong_t)drrf->drr_object,
(u_longlong_t)drrf->drr_offset,
(longlong_t)drrf->drr_length,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
err);
break;
}
case DRR_SPILL:
{
struct drr_spill *drrs = &rrd->header.drr_u.drr_spill;
dprintf("drr_type = SPILL obj = %llu length = %llu "
Annotated dprintf as printf-like ZFS loves using %llu for uint64_t, but that requires a cast to not be noisy - which is even done in many, though not all, places. Also a couple places used %u for uint64_t, which were promoted to %llu. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Rich Ercolani <rincebrain@gmail.com> Closes #12233
2021-06-23 07:53:45 +03:00
"err = %d\n", (u_longlong_t)drrs->drr_object,
(u_longlong_t)drrs->drr_length, err);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
break;
}
Fix issues with truncated files in raw sends This patch fixes a few issues with raw receives involving truncated files: * dnode_reallocate() now calls dnode_set_blksz() instead of dnode_setdblksz(). This ensures that any remaining dbufs with blkid 0 are resized along with their containing dnode upon reallocation. * One of the calls to dmu_free_long_range() in receive_object() needs to check that the object it is about to free some contents or hasn't been completely removed already by a previous call to dmu_free_long_object() in the same function. * The same call to dmu_free_long_range() in the previous point needs to ensure it uses the object's current block size and not the new block size. This ensures the blocks of the object that are supposed to be freed are completely removed and not simply partially zeroed out. This patch also adds handling for DRR_OBJECT_RANGE records to dprintf_drr() for debugging purposes. Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #7378 Closes #8528
2019-03-27 21:30:48 +03:00
case DRR_OBJECT_RANGE:
{
struct drr_object_range *drror =
&rrd->header.drr_u.drr_object_range;
dprintf("drr_type = OBJECT_RANGE firstobj = %llu "
"numslots = %llu flags = %u err = %d\n",
Annotated dprintf as printf-like ZFS loves using %llu for uint64_t, but that requires a cast to not be noisy - which is even done in many, though not all, places. Also a couple places used %u for uint64_t, which were promoted to %llu. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Rich Ercolani <rincebrain@gmail.com> Closes #12233
2021-06-23 07:53:45 +03:00
(u_longlong_t)drror->drr_firstobj,
(u_longlong_t)drror->drr_numslots,
Fix issues with truncated files in raw sends This patch fixes a few issues with raw receives involving truncated files: * dnode_reallocate() now calls dnode_set_blksz() instead of dnode_setdblksz(). This ensures that any remaining dbufs with blkid 0 are resized along with their containing dnode upon reallocation. * One of the calls to dmu_free_long_range() in receive_object() needs to check that the object it is about to free some contents or hasn't been completely removed already by a previous call to dmu_free_long_object() in the same function. * The same call to dmu_free_long_range() in the previous point needs to ensure it uses the object's current block size and not the new block size. This ensures the blocks of the object that are supposed to be freed are completely removed and not simply partially zeroed out. This patch also adds handling for DRR_OBJECT_RANGE records to dprintf_drr() for debugging purposes. Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #7378 Closes #8528
2019-03-27 21:30:48 +03:00
drror->drr_flags, err);
break;
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
default:
return;
}
#endif
}
/*
* Commit the records to the pool.
*/
static int
receive_process_record(struct receive_writer_arg *rwa,
struct receive_record_arg *rrd)
{
int err;
/* Processing in order, therefore bytes_read should be increasing. */
ASSERT3U(rrd->bytes_read, >=, rwa->bytes_read);
rwa->bytes_read = rrd->bytes_read;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
/* We can only heal write records; other ones get ignored */
if (rwa->heal && rrd->header.drr_type != DRR_WRITE) {
if (rrd->abd != NULL) {
abd_free(rrd->abd);
rrd->abd = NULL;
} else if (rrd->payload != NULL) {
kmem_free(rrd->payload, rrd->payload_size);
rrd->payload = NULL;
}
return (0);
}
if (!rwa->heal && rrd->header.drr_type != DRR_WRITE) {
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
err = flush_write_batch(rwa);
if (err != 0) {
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
if (rrd->abd != NULL) {
abd_free(rrd->abd);
rrd->abd = NULL;
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
rrd->payload = NULL;
} else if (rrd->payload != NULL) {
kmem_free(rrd->payload, rrd->payload_size);
rrd->payload = NULL;
}
return (err);
}
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
switch (rrd->header.drr_type) {
case DRR_OBJECT:
{
struct drr_object *drro = &rrd->header.drr_u.drr_object;
err = receive_object(rwa, drro, rrd->payload);
kmem_free(rrd->payload, rrd->payload_size);
rrd->payload = NULL;
break;
}
case DRR_FREEOBJECTS:
{
struct drr_freeobjects *drrfo =
&rrd->header.drr_u.drr_freeobjects;
err = receive_freeobjects(rwa, drrfo);
break;
}
case DRR_WRITE:
{
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
err = receive_process_write_record(rwa, rrd);
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (rwa->heal) {
/*
* If healing - always free the abd after processing
*/
abd_free(rrd->abd);
rrd->abd = NULL;
} else if (err != EAGAIN) {
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
/*
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
* On success, a non-healing
* receive_process_write_record() returns
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
* EAGAIN to indicate that we do not want to free
* the rrd or arc_buf.
*/
ASSERT(err != 0);
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_free(rrd->abd);
rrd->abd = NULL;
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
break;
}
case DRR_WRITE_EMBEDDED:
{
struct drr_write_embedded *drrwe =
&rrd->header.drr_u.drr_write_embedded;
err = receive_write_embedded(rwa, drrwe, rrd->payload);
kmem_free(rrd->payload, rrd->payload_size);
rrd->payload = NULL;
break;
}
case DRR_FREE:
{
struct drr_free *drrf = &rrd->header.drr_u.drr_free;
err = receive_free(rwa, drrf);
break;
}
case DRR_SPILL:
{
struct drr_spill *drrs = &rrd->header.drr_u.drr_spill;
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
err = receive_spill(rwa, drrs, rrd->abd);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0)
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
abd_free(rrd->abd);
rrd->abd = NULL;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
rrd->payload = NULL;
break;
}
case DRR_OBJECT_RANGE:
{
struct drr_object_range *drror =
&rrd->header.drr_u.drr_object_range;
Fix issues with truncated files in raw sends This patch fixes a few issues with raw receives involving truncated files: * dnode_reallocate() now calls dnode_set_blksz() instead of dnode_setdblksz(). This ensures that any remaining dbufs with blkid 0 are resized along with their containing dnode upon reallocation. * One of the calls to dmu_free_long_range() in receive_object() needs to check that the object it is about to free some contents or hasn't been completely removed already by a previous call to dmu_free_long_object() in the same function. * The same call to dmu_free_long_range() in the previous point needs to ensure it uses the object's current block size and not the new block size. This ensures the blocks of the object that are supposed to be freed are completely removed and not simply partially zeroed out. This patch also adds handling for DRR_OBJECT_RANGE records to dprintf_drr() for debugging purposes. Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #7378 Closes #8528
2019-03-27 21:30:48 +03:00
err = receive_object_range(rwa, drror);
break;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
case DRR_REDACT:
{
struct drr_redact *drrr = &rrd->header.drr_u.drr_redact;
err = receive_redact(rwa, drrr);
break;
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
default:
Fix issues with truncated files in raw sends This patch fixes a few issues with raw receives involving truncated files: * dnode_reallocate() now calls dnode_set_blksz() instead of dnode_setdblksz(). This ensures that any remaining dbufs with blkid 0 are resized along with their containing dnode upon reallocation. * One of the calls to dmu_free_long_range() in receive_object() needs to check that the object it is about to free some contents or hasn't been completely removed already by a previous call to dmu_free_long_object() in the same function. * The same call to dmu_free_long_range() in the previous point needs to ensure it uses the object's current block size and not the new block size. This ensures the blocks of the object that are supposed to be freed are completely removed and not simply partially zeroed out. This patch also adds handling for DRR_OBJECT_RANGE records to dprintf_drr() for debugging purposes. Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #7378 Closes #8528
2019-03-27 21:30:48 +03:00
err = (SET_ERROR(EINVAL));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
if (err != 0)
dprintf_drr(rrd, err);
return (err);
}
/*
* dmu_recv_stream's worker thread; pull records off the queue, and then call
* receive_process_record When we're done, signal the main thread and exit.
*/
Switch from _Noreturn to __attribute__((noreturn)) Parts of the Linux kernel build system struggle with _Noreturn. This results in the following warnings when building on RHEL 8.5, and likely other environments. Switch to using the __attribute__((noreturn)). warning: objtool: dbuf_free_range()+0x2b8: return with modified stack frame warning: objtool: dbuf_free_range()+0x0: stack state mismatch: cfa1=7+40 cfa2=7+8 ... WARNING: EXPORT symbol "arc_buf_size" [zfs.ko] version generation failed, symbol will not be versioned. WARNING: EXPORT symbol "spa_open" [zfs.ko] version generation failed, symbol will not be versioned. ... Additionally, __thread_exit() has been renamed spl_thread_exit() and made a static inline function. This was needed because the kernel will generate a warning for symbols which are __attribute__((noreturn)) and then exported with EXPORT_SYMBOL. While we could continue to use _Noreturn in user space I've also switched it to __attribute__((noreturn)) purely for consistency throughout the code base. Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org> Reviewed-by: Brian Atkinson <batkinson@lanl.gov> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #13238
2022-03-23 18:51:00 +03:00
static __attribute__((noreturn)) void
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
receive_writer_thread(void *arg)
{
struct receive_writer_arg *rwa = arg;
struct receive_record_arg *rrd;
fstrans_cookie_t cookie = spl_fstrans_mark();
for (rrd = bqueue_dequeue(&rwa->q); !rrd->eos_marker;
rrd = bqueue_dequeue(&rwa->q)) {
/*
* If there's an error, the main thread will stop putting things
* on the queue, but we need to clear everything in it before we
* can exit.
*/
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
int err = 0;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (rwa->err == 0) {
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
err = receive_process_record(rwa, rrd);
Improve zfs receive performance with lightweight write The performance of `zfs receive` can be bottlenecked on the CPU consumed by the `receive_writer` thread, especially when receiving streams with small compressed block sizes. Much of the CPU is spent creating and destroying dbuf's and arc buf's, one for each `WRITE` record in the send stream. This commit introduces the concept of "lightweight writes", which allows `zfs receive` to write to the DMU by providing an ABD, and instantiating only a new type of `dbuf_dirty_record_t`. The dbuf and arc buf for this "dirty leaf block" are not instantiated. Because there is no dbuf with the dirty data, this mechanism doesn't support reading from "lightweight-dirty" blocks (they would see the on-disk state rather than the dirty data). Since the dedup-receive code has been removed, `zfs receive` is write-only, so this works fine. Because there are no arc bufs for the received data, the received data is no longer cached in the ARC. Testing a receive of a stream with average compressed block size of 4KB, this commit improves performance by 50%, while also reducing CPU usage by 50% of a CPU. On a per-block basis, CPU consumed by receive_writer() and dbuf_evict() is now 1/7th (14%) of what it was. Baseline: 450MB/s, CPU in receive_writer() 40% + dbuf_evict() 35% New: 670MB/s, CPU in receive_writer() 17% + dbuf_evict() 0% The code is also restructured in a few ways: Added a `dr_dnode` field to the dbuf_dirty_record_t. This simplifies some existing code that no longer needs `DB_DNODE_ENTER()` and related routines. The new field is needed by the lightweight-type dirty record. To ensure that the `dr_dnode` field remains valid until the dirty record is freed, we have to ensure that the `dnode_move()` doesn't relocate the dnode_t. To do this we keep a hold on the dnode until it's zio's have completed. This is already done by the user-accounting code (`userquota_updates_task()`), this commit extends that so that it always keeps the dnode hold until zio completion (see `dnode_rele_task()`). `dn_dirty_txg` was previously zeroed when the dnode was synced. This was not necessary, since its meaning can be "when was this dnode last dirtied". This change simplifies the new `dnode_rele_task()` code. Removed some dead code related to `DRR_WRITE_BYREF` (dedup receive). Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: George Wilson <gwilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #11105
2020-12-11 21:26:02 +03:00
} else if (rrd->abd != NULL) {
abd_free(rrd->abd);
rrd->abd = NULL;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
rrd->payload = NULL;
} else if (rrd->payload != NULL) {
kmem_free(rrd->payload, rrd->payload_size);
rrd->payload = NULL;
}
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
/*
* EAGAIN indicates that this record has been saved (on
* raw->write_batch), and will be used again, so we don't
* free it.
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
* When healing data we always need to free the record.
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
*/
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (err != EAGAIN || rwa->heal) {
Fix error handling in receive_writer_thread() If `receive_writer_thread()` gets an error from `receive_process_record()`, it should be saved in `rwa->err` so that we will stop processing records, and the main thread will notice that the receive has failed. When an error is first encountered, this happens correctly. However, if there are more records to dequeue, the next time through the loop we will reset `rwa->err` to zero, allowing us to try to process the following record (2 after the failed record). Depending on what types of records remain, we may incorrectly complete the receive "successfully", but without actually having processed all the records. The fix is to only set `rwa->err` if we got a *non-zero* error. This bug was introduced by #10099 "Improve zfs receive performance by batching writes". Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10320
2020-05-15 06:48:29 +03:00
if (rwa->err == 0)
rwa->err = err;
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
kmem_free(rrd, sizeof (*rrd));
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
kmem_free(rrd, sizeof (*rrd));
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (rwa->heal) {
zio_wait(rwa->heal_pio);
} else {
int err = flush_write_batch(rwa);
if (rwa->err == 0)
rwa->err = err;
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
mutex_enter(&rwa->mutex);
rwa->done = B_TRUE;
cv_signal(&rwa->cv);
mutex_exit(&rwa->mutex);
spl_fstrans_unmark(cookie);
thread_exit();
}
static int
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
resume_check(dmu_recv_cookie_t *drc, nvlist_t *begin_nvl)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
uint64_t val;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
objset_t *mos = dmu_objset_pool(drc->drc_os)->dp_meta_objset;
uint64_t dsobj = dmu_objset_id(drc->drc_os);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
uint64_t resume_obj, resume_off;
if (nvlist_lookup_uint64(begin_nvl,
"resume_object", &resume_obj) != 0 ||
nvlist_lookup_uint64(begin_nvl,
"resume_offset", &resume_off) != 0) {
return (SET_ERROR(EINVAL));
}
VERIFY0(zap_lookup(mos, dsobj,
DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val));
if (resume_obj != val)
return (SET_ERROR(EINVAL));
VERIFY0(zap_lookup(mos, dsobj,
DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val));
if (resume_off != val)
return (SET_ERROR(EINVAL));
return (0);
}
/*
* Read in the stream's records, one by one, and apply them to the pool. There
* are two threads involved; the thread that calls this function will spin up a
* worker thread, read the records off the stream one by one, and issue
* prefetches for any necessary indirect blocks. It will then push the records
* onto an internal blocking queue. The worker thread will pull the records off
* the queue, and actually write the data into the DMU. This way, the worker
* thread doesn't have to wait for reads to complete, since everything it needs
* (the indirect blocks) will be prefetched.
*
* NB: callers *must* call dmu_recv_end() if this succeeds.
*/
int
Remove deduplicated send/receive code Deduplicated send streams (i.e. `zfs send -D` and `zfs receive` of such streams) are deprecated. Deduplicated send streams can be received by first converting them to non-deduplicated with the `zstream redup` command. This commit removes the code for sending and receiving deduplicated send streams. `zfs send -D` will now print a warning, ignore the `-D` flag, and generate a regular (non-deduplicated) send stream. `zfs receive` of a deduplicated send stream will print an error message and fail. The resulting code simplification (especially in the kernel's support for receiving dedup streams) should help enable future performance enhancements. Several new tests are added which leverage `zstream redup`. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Issue #7887 Issue #10117 Issue #10156 Closes #10212
2020-04-23 20:06:57 +03:00
dmu_recv_stream(dmu_recv_cookie_t *drc, offset_t *voffp)
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
{
int err = 0;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
struct receive_writer_arg *rwa = kmem_zalloc(sizeof (*rwa), KM_SLEEP);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Fix dmu_recv_stream test for resumable Use dsl_dataset_has_resume_receive_state() not dsl_dataset_is_zapified() to check if stream is resumable. Reviewed-by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Alek Pinchuk <apinchuk@axcient.com> Reviewed-by: Ryan Moeller <ryan@ixsystems.com> Signed-off-by: Paul Zuchowski <pzuchowski@datto.com> Closes #12034
2021-05-14 07:46:14 +03:00
if (dsl_dataset_has_resume_receive_state(drc->drc_ds)) {
uint64_t bytes = 0;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
(void) zap_lookup(drc->drc_ds->ds_dir->dd_pool->dp_meta_objset,
drc->drc_ds->ds_object, DS_FIELD_RESUME_BYTES,
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
sizeof (bytes), 1, &bytes);
drc->drc_bytes_read += bytes;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
drc->drc_ignore_objlist = objlist_create();
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* these were verified in dmu_recv_begin */
ASSERT3U(DMU_GET_STREAM_HDRTYPE(drc->drc_drrb->drr_versioninfo), ==,
DMU_SUBSTREAM);
ASSERT3U(drc->drc_drrb->drr_type, <, DMU_OST_NUMTYPES);
ASSERT(dsl_dataset_phys(drc->drc_ds)->ds_flags & DS_FLAG_INCONSISTENT);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
ASSERT0(drc->drc_os->os_encrypted &&
(drc->drc_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/* handle DSL encryption key payload */
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
nvlist_t *keynvl = NULL;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
ASSERT(drc->drc_os->os_encrypted);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
ASSERT(drc->drc_raw);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
err = nvlist_lookup_nvlist(drc->drc_begin_nvl, "crypt_keydata",
&keynvl);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0)
goto out;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (!drc->drc_heal) {
/*
* If this is a new dataset we set the key immediately.
* Otherwise we don't want to change the key until we
* are sure the rest of the receive succeeded so we
* stash the keynvl away until then.
*/
err = dsl_crypto_recv_raw(spa_name(drc->drc_os->os_spa),
drc->drc_ds->ds_object, drc->drc_fromsnapobj,
drc->drc_drrb->drr_type, keynvl, drc->drc_newfs);
if (err != 0)
goto out;
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
/* see comment in dmu_recv_end_sync() */
drc->drc_ivset_guid = 0;
(void) nvlist_lookup_uint64(keynvl, "to_ivset_guid",
&drc->drc_ivset_guid);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (!drc->drc_newfs)
drc->drc_keynvl = fnvlist_dup(keynvl);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING) {
err = resume_check(drc, drc->drc_begin_nvl);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0)
goto out;
}
Fix incremental receive silently failing for recursive sends The problem occurs because dmu_recv_begin pulls in the payload and next header from the input stream in order to use the contents of the begin record's nvlist. However, the change to do that before the other checks in dmu_recv_begin occur caused a regression where an empty send stream in a recursive send could have its END record consumed by this, which broke the logic of recv_skip. A test is also included to protect against this case in the future. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #12661 Closes #14568
2023-03-10 20:52:44 +03:00
/*
* For compatibility with recursive send streams, we do this here,
* rather than in dmu_recv_begin. If we pull the next header too
* early, and it's the END record, we break the `recv_skip` logic.
*/
if (drc->drc_drr_begin->drr_payloadlen == 0) {
err = receive_read_payload_and_next_header(drc, 0, NULL);
if (err != 0)
goto out;
}
Fix errata #4 handling for resuming streams Currently, the handling for errata #4 has two issues which allow the checks for this issue to be bypassed using resumable sends. The first issue is that drc->drc_fromsnapobj is not set in the resuming code as it is in the non-resuming code. This causes dsl_crypto_recv_key_check() to skip its checks for the from_ivset_guid. The second issue is that resumable sends do not clean up their on-disk state if they fail the checks in dmu_recv_stream() that happen before any data is received. As a result of these two bugs, a user can attempt a resumable send of a dataset without a from_ivset_guid. This will fail the initial dmu_recv_stream() checks, leaving a valid resume state. The send can then be resumed, which skips those checks, allowing the receive to be completed. This commit fixes these issues by setting drc->drc_fromsnapobj in the resuming receive path and by ensuring that resumablereceives are properly cleaned up if they fail the initial dmu_recv_stream() checks. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #9818 Closes #9829
2020-01-14 23:25:20 +03:00
/*
* If we failed before this point we will clean up any new resume
* state that was created. Now that we've gotten past the initial
* checks we are ok to retain that resume state.
*/
drc->drc_should_save = B_TRUE;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
(void) bqueue_init(&rwa->q, zfs_recv_queue_ff,
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
MAX(zfs_recv_queue_length, 2 * zfs_max_recordsize),
offsetof(struct receive_record_arg, node));
cv_init(&rwa->cv, NULL, CV_DEFAULT, NULL);
mutex_init(&rwa->mutex, NULL, MUTEX_DEFAULT, NULL);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
rwa->os = drc->drc_os;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
rwa->byteswap = drc->drc_byteswap;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
rwa->heal = drc->drc_heal;
rwa->tofs = drc->drc_tofs;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
rwa->resumable = drc->drc_resumable;
rwa->raw = drc->drc_raw;
Fix send/recv lost spill block When receiving a DRR_OBJECT record the receive_object() function needs to determine how to handle a spill block associated with the object. It may need to be removed or kept depending on how the object was modified at the source. This determination is currently accomplished using a heuristic which takes in to account the DRR_OBJECT record and the existing object properties. This is a problem because there isn't quite enough information available to do the right thing under all circumstances. For example, when only the block size changes the spill block is removed when it should be kept. What's needed to resolve this is an additional flag in the DRR_OBJECT which indicates if the object being received references a spill block. The DRR_OBJECT_SPILL flag was added for this purpose. When set then the object references a spill block and it must be kept. Either it is update to date, or it will be replaced by a subsequent DRR_SPILL record. Conversely, if the object being received doesn't reference a spill block then any existing spill block should always be removed. Since previous versions of ZFS do not understand this new flag additional DRR_SPILL records will be inserted in to the stream. This has the advantage of being fully backward compatible. Existing ZFS systems receiving this stream will recreate the spill block if it was incorrectly removed. Updated ZFS versions will correctly ignore the additional spill blocks which can be identified by checking for the DRR_SPILL_UNMODIFIED flag. The small downside to this approach is that is may increase the size of the stream and of the received snapshot on previous versions of ZFS. Additionally, when receiving streams generated by previous unpatched versions of ZFS spill blocks may still be lost. OpenZFS-issue: https://www.illumos.org/issues/9952 FreeBSD-issue: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233277 Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Tom Caputi <tcaputi@datto.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8668
2019-05-08 01:18:44 +03:00
rwa->spill = drc->drc_spill;
File incorrectly zeroed when receiving incremental stream that toggles -L Background: By increasing the recordsize property above the default of 128KB, a filesystem may have "large" blocks. By default, a send stream of such a filesystem does not contain large WRITE records, instead it decreases objects' block sizes to 128KB and splits the large blocks into 128KB blocks, allowing the large-block filesystem to be received by a system that does not support the `large_blocks` feature. A send stream generated by `zfs send -L` (or `--large-block`) preserves the large block size on the receiving system, by using large WRITE records. When receiving an incremental send stream for a filesystem with large blocks, if the send stream's -L flag was toggled, a bug is encountered in which the file's contents are incorrectly zeroed out. The contents of any blocks that were not modified by this send stream will be lost. "Toggled" means that the previous send used `-L`, but this incremental does not use `-L` (-L to no-L); or that the previous send did not use `-L`, but this incremental does use `-L` (no-L to -L). Changes: This commit addresses the problem with several changes to the semantics of zfs send/receive: 1. "-L to no-L" incrementals are rejected. If the previous send used `-L`, but this incremental does not use `-L`, the `zfs receive` will fail with this error message: incremental send stream requires -L (--large-block), to match previous receive. 2. "no-L to -L" incrementals are handled correctly, preserving the smaller (128KB) block size of any already-received files that used large blocks on the sending system but were split by `zfs send` without the `-L` flag. 3. A new send stream format flag is added, `SWITCH_TO_LARGE_BLOCKS`. This feature indicates that we can correctly handle "no-L to -L" incrementals. This flag is currently not set on any send streams. In the future, we intend for incremental send streams of snapshots that have large blocks to use `-L` by default, and these streams will also have the `SWITCH_TO_LARGE_BLOCKS` feature set. This ensures that streams from the default use of `zfs send` won't encounter the bug mentioned above, because they can't be received by software with the bug. Implementation notes: To facilitate accessing the ZPL's generation number, `zfs_space_delta_cb()` has been renamed to `zpl_get_file_info()` and restructured to fill in a struct with ZPL-specific info including owner and generation. In the "no-L to -L" case, if this is a compressed send stream (from `zfs send -cL`), large WRITE records that are being written to small (128KB) blocksize files need to be decompressed so that they can be written split up into multiple blocks. The zio pipeline will recompress each smaller block individually. A new test case, `send-L_toggle`, is added, which tests the "no-L to -L" case and verifies that we get an error for the "-L to no-L" case. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #6224 Closes #10383
2020-06-09 20:41:01 +03:00
rwa->full = (drc->drc_drr_begin->drr_u.drr_begin.drr_fromguid == 0);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
rwa->os->os_raw_receive = drc->drc_raw;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (drc->drc_heal) {
rwa->heal_pio = zio_root(drc->drc_os->os_spa, NULL, NULL,
ZIO_FLAG_GODFATHER);
}
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
list_create(&rwa->write_batch, sizeof (struct receive_record_arg),
offsetof(struct receive_record_arg, node.bqn_node));
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
(void) thread_create(NULL, 0, receive_writer_thread, rwa, 0, curproc,
TS_RUN, minclsyspri);
/*
* We're reading rwa->err without locks, which is safe since we are the
* only reader, and the worker thread is the only writer. It's ok if we
* miss a write for an iteration or two of the loop, since the writer
* thread will keep freeing records we send it until we send it an eos
* marker.
*
* We can leave this loop in 3 ways: First, if rwa->err is
* non-zero. In that case, the writer thread will free the rrd we just
* pushed. Second, if we're interrupted; in that case, either it's the
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
* first loop and drc->drc_rrd was never allocated, or it's later, and
* drc->drc_rrd has been handed off to the writer thread who will free
* it. Finally, if receive_read_record fails or we're at the end of the
* stream, then we free drc->drc_rrd and exit.
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
*/
while (rwa->err == 0) {
Simplify issig(). We always call it twice with JUSTLOOKING and then FORREAL. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Pawel Jakub Dawidek <pawel@dawidek.net> Closes #16225
2024-05-29 20:49:11 +03:00
if (issig()) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
err = SET_ERROR(EINTR);
break;
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
ASSERT3P(drc->drc_rrd, ==, NULL);
drc->drc_rrd = drc->drc_next_rrd;
drc->drc_next_rrd = NULL;
/* Allocates and loads header into drc->drc_next_rrd */
err = receive_read_record(drc);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_rrd->header.drr_type == DRR_END || err != 0) {
kmem_free(drc->drc_rrd, sizeof (*drc->drc_rrd));
drc->drc_rrd = NULL;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
break;
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
bqueue_enqueue(&rwa->q, drc->drc_rrd,
sizeof (struct receive_record_arg) +
drc->drc_rrd->payload_size);
drc->drc_rrd = NULL;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
ASSERT3P(drc->drc_rrd, ==, NULL);
drc->drc_rrd = kmem_zalloc(sizeof (*drc->drc_rrd), KM_SLEEP);
drc->drc_rrd->eos_marker = B_TRUE;
bqueue_enqueue_flush(&rwa->q, drc->drc_rrd, 1);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
mutex_enter(&rwa->mutex);
while (!rwa->done) {
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
/*
* We need to use cv_wait_sig() so that any process that may
* be sleeping here can still fork.
*/
(void) cv_wait_sig(&rwa->cv, &rwa->mutex);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
mutex_exit(&rwa->mutex);
/*
* If we are receiving a full stream as a clone, all object IDs which
* are greater than the maximum ID referenced in the stream are
* by definition unused and must be freed.
*/
if (drc->drc_clone && drc->drc_drrb->drr_fromguid == 0) {
uint64_t obj = rwa->max_object + 1;
int free_err = 0;
int next_err = 0;
while (next_err == 0) {
free_err = dmu_free_long_object(rwa->os, obj);
if (free_err != 0 && free_err != ENOENT)
break;
next_err = dmu_object_next(rwa->os, &obj, FALSE, 0);
}
if (err == 0) {
if (free_err != 0 && free_err != ENOENT)
err = free_err;
else if (next_err != ESRCH)
err = next_err;
}
}
cv_destroy(&rwa->cv);
mutex_destroy(&rwa->mutex);
bqueue_destroy(&rwa->q);
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
list_destroy(&rwa->write_batch);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err == 0)
err = rwa->err;
out:
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
/*
* If we hit an error before we started the receive_writer_thread
* we need to clean up the next_rrd we create by processing the
* DRR_BEGIN record.
*/
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
if (drc->drc_next_rrd != NULL)
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
async zvol minor node creation interferes with receive When we finish a zfs receive, dmu_recv_end_sync() calls zvol_create_minors(async=TRUE). This kicks off some other threads that create the minor device nodes (in /dev/zvol/poolname/...). These async threads call zvol_prefetch_minors_impl() and zvol_create_minor(), which both call dmu_objset_own(), which puts a "long hold" on the dataset. Since the zvol minor node creation is asynchronous, this can happen after the `ZFS_IOC_RECV[_NEW]` ioctl and `zfs receive` process have completed. After the first receive ioctl has completed, userland may attempt to do another receive into the same dataset (e.g. the next incremental stream). This second receive and the asynchronous minor node creation can interfere with one another in several different ways, because they both require exclusive access to the dataset: 1. When the second receive is finishing up, dmu_recv_end_check() does dsl_dataset_handoff_check(), which can fail with EBUSY if the async minor node creation already has a "long hold" on this dataset. This causes the 2nd receive to fail. 2. The async udev rule can fail if zvol_id and/or systemd-udevd try to open the device while the the second receive's async attempt at minor node creation owns the dataset (via zvol_prefetch_minors_impl). This causes the minor node (/dev/zd*) to exist, but the udev-generated /dev/zvol/... to not exist. 3. The async minor node creation can silently fail with EBUSY if the first receive's zvol_create_minor() trys to own the dataset while the second receive's zvol_prefetch_minors_impl already owns the dataset. To address these problems, this change synchronously creates the minor node. To avoid the lock ordering problems that the asynchrony was introduced to fix (see #3681), we create the minor nodes from open context, with no locks held, rather than from syncing contex as was originally done. Implementation notes: We generally do not need to traverse children or prefetch anything (e.g. when running the recv, snapshot, create, or clone subcommands of zfs). We only need recursion when importing/opening a pool and when loading encryption keys. The existing recursive, asynchronous, prefetching code is preserved for use in these cases. Channel programs may need to create zvol minor nodes, when creating a snapshot of a zvol with the snapdev property set. We figure out what snapshots are created when running the LUA program in syncing context. In this case we need to remember what snapshots were created, and then try to create their minor nodes from open context, after the LUA code has completed. There are additional zvol use cases that asynchronously own the dataset, which can cause similar problems. E.g. changing the volmode or snapdev properties. These are less problematic because they are not recursive and don't touch datasets that are not involved in the operation, there is still potential for interference with subsequent operations. In the future, these cases should be similarly converted to create the zvol minor node synchronously from open context. The async tasks of removing and renaming minors do not own the objset, so they do not have this problem. However, it may make sense to also convert these operations to happen synchronously from open context, in the future. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> External-issue: DLPX-65948 Closes #7863 Closes #9885
2020-02-03 20:33:14 +03:00
/*
* The objset will be invalidated by dmu_recv_end() when we do
* dsl_dataset_clone_swap_sync_impl().
*/
drc->drc_os = NULL;
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
kmem_free(rwa, sizeof (*rwa));
nvlist_free(drc->drc_begin_nvl);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
if (err != 0) {
/*
* Clean up references. If receive is not resumable,
* destroy what we created, so we don't leave it in
* the inconsistent state.
*/
dmu_recv_cleanup_ds(drc);
nvlist_free(drc->drc_keynvl);
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
objlist_destroy(drc->drc_ignore_objlist);
drc->drc_ignore_objlist = NULL;
*voffp = drc->drc_voff;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
return (err);
}
static int
dmu_recv_end_check(void *arg, dmu_tx_t *tx)
{
dmu_recv_cookie_t *drc = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
int error;
ASSERT3P(drc->drc_ds->ds_owner, ==, dmu_recv_tag);
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (drc->drc_heal) {
error = 0;
} else if (!drc->drc_newfs) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_t *origin_head;
error = dsl_dataset_hold(dp, drc->drc_tofs, FTAG, &origin_head);
if (error != 0)
return (error);
if (drc->drc_force) {
/*
* We will destroy any snapshots in tofs (i.e. before
* origin_head) that are after the origin (which is
* the snap before drc_ds, because drc_ds can not
* have any snaps of its own).
*/
uint64_t obj;
obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
while (obj !=
dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
dsl_dataset_t *snap;
error = dsl_dataset_hold_obj(dp, obj, FTAG,
&snap);
if (error != 0)
break;
if (snap->ds_dir != origin_head->ds_dir)
error = SET_ERROR(EINVAL);
if (error == 0) {
error = dsl_destroy_snapshot_check_impl(
snap, B_FALSE);
}
obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
dsl_dataset_rele(snap, FTAG);
if (error != 0)
break;
}
if (error != 0) {
dsl_dataset_rele(origin_head, FTAG);
return (error);
}
}
if (drc->drc_keynvl != NULL) {
error = dsl_crypto_recv_raw_key_check(drc->drc_ds,
drc->drc_keynvl, tx);
if (error != 0) {
dsl_dataset_rele(origin_head, FTAG);
return (error);
}
}
error = dsl_dataset_clone_swap_check_impl(drc->drc_ds,
origin_head, drc->drc_force, drc->drc_owner, tx);
if (error != 0) {
dsl_dataset_rele(origin_head, FTAG);
return (error);
}
error = dsl_dataset_snapshot_check_impl(origin_head,
filesystem_limit/snapshot_limit is incorrectly enforced against root The filesystem_limit and snapshot_limit properties limit the number of filesystems or snapshots that can be created below this dataset. According to the manpage, "The limit is not enforced if the user is allowed to change the limit." Two types of users are allowed to change the limit: 1. Those that have been delegated the `filesystem_limit` or `snapshot_limit` permission, e.g. with `zfs allow USER filesystem_limit DATASET`. This works properly. 2. A user with elevated system privileges (e.g. root). This does not work - the root user will incorrectly get an error when trying to create a snapshot/filesystem, if it exceeds the `_limit` property. The problem is that `priv_policy_ns()` does not work if the `cred_t` is not that of the current process. This happens when `dsl_enforce_ds_ss_limits()` is called in syncing context (as part of a sync task's check func) to determine the permissions of the corresponding user process. This commit fixes the issue by passing the `task_struct` (typedef'ed as a `proc_t`) to syncing context, and then using `has_capability()` to determine if that process is privileged. Note that we still need to pass the `cred_t` to syncing context so that we can check if the user was delegated this permission with `zfs allow`. This problem only impacts Linux. Wrappers are added to FreeBSD but it continues to use `priv_check_cred()`, which works on arbitrary `cred_t`. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@ixsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #8226 Closes #10545
2020-07-12 03:18:02 +03:00
drc->drc_tosnap, tx, B_TRUE, 1,
drc->drc_cred, drc->drc_proc);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_rele(origin_head, FTAG);
if (error != 0)
return (error);
error = dsl_destroy_head_check_impl(drc->drc_ds, 1);
} else {
error = dsl_dataset_snapshot_check_impl(drc->drc_ds,
filesystem_limit/snapshot_limit is incorrectly enforced against root The filesystem_limit and snapshot_limit properties limit the number of filesystems or snapshots that can be created below this dataset. According to the manpage, "The limit is not enforced if the user is allowed to change the limit." Two types of users are allowed to change the limit: 1. Those that have been delegated the `filesystem_limit` or `snapshot_limit` permission, e.g. with `zfs allow USER filesystem_limit DATASET`. This works properly. 2. A user with elevated system privileges (e.g. root). This does not work - the root user will incorrectly get an error when trying to create a snapshot/filesystem, if it exceeds the `_limit` property. The problem is that `priv_policy_ns()` does not work if the `cred_t` is not that of the current process. This happens when `dsl_enforce_ds_ss_limits()` is called in syncing context (as part of a sync task's check func) to determine the permissions of the corresponding user process. This commit fixes the issue by passing the `task_struct` (typedef'ed as a `proc_t`) to syncing context, and then using `has_capability()` to determine if that process is privileged. Note that we still need to pass the `cred_t` to syncing context so that we can check if the user was delegated this permission with `zfs allow`. This problem only impacts Linux. Wrappers are added to FreeBSD but it continues to use `priv_check_cred()`, which works on arbitrary `cred_t`. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@ixsystems.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #8226 Closes #10545
2020-07-12 03:18:02 +03:00
drc->drc_tosnap, tx, B_TRUE, 1,
drc->drc_cred, drc->drc_proc);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
return (error);
}
static void
dmu_recv_end_sync(void *arg, dmu_tx_t *tx)
{
dmu_recv_cookie_t *drc = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
boolean_t encrypted = drc->drc_ds->ds_dir->dd_crypto_obj != 0;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
uint64_t newsnapobj = 0;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
spa_history_log_internal_ds(drc->drc_ds, "finish receiving",
tx, "snap=%s", drc->drc_tosnap);
drc->drc_ds->ds_objset->os_raw_receive = B_FALSE;
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (drc->drc_heal) {
if (drc->drc_keynvl != NULL) {
nvlist_free(drc->drc_keynvl);
drc->drc_keynvl = NULL;
}
} else if (!drc->drc_newfs) {
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_t *origin_head;
VERIFY0(dsl_dataset_hold(dp, drc->drc_tofs, FTAG,
&origin_head));
if (drc->drc_force) {
/*
* Destroy any snapshots of drc_tofs (origin_head)
* after the origin (the snap before drc_ds).
*/
uint64_t obj;
obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
while (obj !=
dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
dsl_dataset_t *snap;
VERIFY0(dsl_dataset_hold_obj(dp, obj, FTAG,
&snap));
ASSERT3P(snap->ds_dir, ==, origin_head->ds_dir);
obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
dsl_destroy_snapshot_sync_impl(snap,
B_FALSE, tx);
dsl_dataset_rele(snap, FTAG);
}
}
if (drc->drc_keynvl != NULL) {
dsl_crypto_recv_raw_key_sync(drc->drc_ds,
drc->drc_keynvl, tx);
nvlist_free(drc->drc_keynvl);
drc->drc_keynvl = NULL;
}
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
VERIFY3P(drc->drc_ds->ds_prev, ==,
origin_head->ds_prev);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_clone_swap_sync_impl(drc->drc_ds,
origin_head, tx);
dmu_objset_from_ds must be called with dp_config_rwlock held The normal lock order is that the dp_config_rwlock must be held before the ds_opening_lock. For example, dmu_objset_hold() does this. However, dmu_objset_open_impl() is called with the ds_opening_lock held, and if the dp_config_rwlock is not already held, it will attempt to acquire it. This may lead to deadlock, since the lock order is reversed. Looking at all the callers of dmu_objset_open_impl() (which is principally the callers of dmu_objset_from_ds()), almost all callers already have the dp_config_rwlock. However, there are a few places in the send and receive code paths that do not. For example: dsl_crypto_populate_key_nvlist, send_cb, dmu_recv_stream, receive_write_byref, redact_traverse_thread. This commit resolves the problem by requiring all callers ot dmu_objset_from_ds() to hold the dp_config_rwlock. In most cases, the code has been restructured such that we call dmu_objset_from_ds() earlier on in the send and receive processes, when we already have the dp_config_rwlock, and save the objset_t until we need it in the middle of the send or receive (similar to what we already do with the dsl_dataset_t). Thus we do not need to acquire the dp_config_rwlock in many new places. I also cleaned up code in dmu_redact_snap() and send_traverse_thread(). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #9662 Closes #10115
2020-03-12 20:55:02 +03:00
/*
* The objset was evicted by dsl_dataset_clone_swap_sync_impl,
* so drc_os is no longer valid.
*/
drc->drc_os = NULL;
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_snapshot_sync_impl(origin_head,
drc->drc_tosnap, tx);
/* set snapshot's creation time and guid */
dmu_buf_will_dirty(origin_head->ds_prev->ds_dbuf, tx);
dsl_dataset_phys(origin_head->ds_prev)->ds_creation_time =
drc->drc_drrb->drr_creation_time;
dsl_dataset_phys(origin_head->ds_prev)->ds_guid =
drc->drc_drrb->drr_toguid;
dsl_dataset_phys(origin_head->ds_prev)->ds_flags &=
~DS_FLAG_INCONSISTENT;
dmu_buf_will_dirty(origin_head->ds_dbuf, tx);
dsl_dataset_phys(origin_head)->ds_flags &=
~DS_FLAG_INCONSISTENT;
Remove deduplicated send/receive code Deduplicated send streams (i.e. `zfs send -D` and `zfs receive` of such streams) are deprecated. Deduplicated send streams can be received by first converting them to non-deduplicated with the `zstream redup` command. This commit removes the code for sending and receiving deduplicated send streams. `zfs send -D` will now print a warning, ignore the `-D` flag, and generate a regular (non-deduplicated) send stream. `zfs receive` of a deduplicated send stream will print an error message and fail. The resulting code simplification (especially in the kernel's support for receiving dedup streams) should help enable future performance enhancements. Several new tests are added which leverage `zstream redup`. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Issue #7887 Issue #10117 Issue #10156 Closes #10212
2020-04-23 20:06:57 +03:00
newsnapobj =
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
dsl_dataset_rele(origin_head, FTAG);
dsl_destroy_head_sync_impl(drc->drc_ds, tx);
if (drc->drc_owner != NULL)
VERIFY3P(origin_head->ds_owner, ==, drc->drc_owner);
} else {
dsl_dataset_t *ds = drc->drc_ds;
dsl_dataset_snapshot_sync_impl(ds, drc->drc_tosnap, tx);
/* set snapshot's creation time and guid */
dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
dsl_dataset_phys(ds->ds_prev)->ds_creation_time =
drc->drc_drrb->drr_creation_time;
dsl_dataset_phys(ds->ds_prev)->ds_guid =
drc->drc_drrb->drr_toguid;
dsl_dataset_phys(ds->ds_prev)->ds_flags &=
~DS_FLAG_INCONSISTENT;
dmu_buf_will_dirty(ds->ds_dbuf, tx);
dsl_dataset_phys(ds)->ds_flags &= ~DS_FLAG_INCONSISTENT;
if (dsl_dataset_has_resume_receive_state(ds)) {
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_FROMGUID, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_OBJECT, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_OFFSET, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_BYTES, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_TOGUID, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_TONAME, tx);
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS, tx);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
Remove deduplicated send/receive code Deduplicated send streams (i.e. `zfs send -D` and `zfs receive` of such streams) are deprecated. Deduplicated send streams can be received by first converting them to non-deduplicated with the `zstream redup` command. This commit removes the code for sending and receiving deduplicated send streams. `zfs send -D` will now print a warning, ignore the `-D` flag, and generate a regular (non-deduplicated) send stream. `zfs receive` of a deduplicated send stream will print an error message and fail. The resulting code simplification (especially in the kernel's support for receiving dedup streams) should help enable future performance enhancements. Several new tests are added which leverage `zstream redup`. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Issue #7887 Issue #10117 Issue #10156 Closes #10212
2020-04-23 20:06:57 +03:00
newsnapobj =
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj;
}
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
/*
* If this is a raw receive, the crypt_keydata nvlist will include
* a to_ivset_guid for us to set on the new snapshot. This value
* will override the value generated by the snapshot code. However,
* this value may not be present, because older implementations of
* the raw send code did not include this value, and we are still
* allowed to receive them if the zfs_disable_ivset_guid_check
* tunable is set, in which case we will leave the newly-generated
* value.
*/
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
if (!drc->drc_heal && drc->drc_raw && drc->drc_ivset_guid != 0) {
Remove deduplicated send/receive code Deduplicated send streams (i.e. `zfs send -D` and `zfs receive` of such streams) are deprecated. Deduplicated send streams can be received by first converting them to non-deduplicated with the `zstream redup` command. This commit removes the code for sending and receiving deduplicated send streams. `zfs send -D` will now print a warning, ignore the `-D` flag, and generate a regular (non-deduplicated) send stream. `zfs receive` of a deduplicated send stream will print an error message and fail. The resulting code simplification (especially in the kernel's support for receiving dedup streams) should help enable future performance enhancements. Several new tests are added which leverage `zstream redup`. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Issue #7887 Issue #10117 Issue #10156 Closes #10212
2020-04-23 20:06:57 +03:00
dmu_object_zapify(dp->dp_meta_objset, newsnapobj,
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
DMU_OT_DSL_DATASET, tx);
Remove deduplicated send/receive code Deduplicated send streams (i.e. `zfs send -D` and `zfs receive` of such streams) are deprecated. Deduplicated send streams can be received by first converting them to non-deduplicated with the `zstream redup` command. This commit removes the code for sending and receiving deduplicated send streams. `zfs send -D` will now print a warning, ignore the `-D` flag, and generate a regular (non-deduplicated) send stream. `zfs receive` of a deduplicated send stream will print an error message and fail. The resulting code simplification (especially in the kernel's support for receiving dedup streams) should help enable future performance enhancements. Several new tests are added which leverage `zstream redup`. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Issue #7887 Issue #10117 Issue #10156 Closes #10212
2020-04-23 20:06:57 +03:00
VERIFY0(zap_update(dp->dp_meta_objset, newsnapobj,
Detect and prevent mixed raw and non-raw sends Currently, there is an issue in the raw receive code where raw receives are allowed to happen on top of previously non-raw received datasets. This is a problem because the source-side dataset doesn't know about how the blocks on the destination were encrypted. As a result, any MAC in the objset's checksum-of-MACs tree that is a parent of both blocks encrypted on the source and blocks encrypted by the destination will be incorrect. This will result in authentication errors when we decrypt the dataset. This patch fixes this issue by adding a new check to the raw receive code. The code now maintains an "IVset guid", which acts as an identifier for the set of IVs used to encrypt a given snapshot. When a snapshot is raw received, the destination snapshot will take this value from the DRR_BEGIN payload. Non-raw receives and normal "zfs snap" operations will cause ZFS to generate a new IVset guid. When a raw incremental stream is received, ZFS will check that the "from" IVset guid in the stream matches that of the "from" destination snapshot. If they do not match, the code will error out the receive, preventing the problem. This patch requires an on-disk format change to add the IVset guids to snapshots and bookmarks. As a result, this patch has errata handling and a tunable to help affected users resolve the issue with as little interruption as possible. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Signed-off-by: Tom Caputi <tcaputi@datto.com> Closes #8308
2019-02-04 22:24:55 +03:00
DS_FIELD_IVSET_GUID, sizeof (uint64_t), 1,
&drc->drc_ivset_guid, tx));
}
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
/*
* Release the hold from dmu_recv_begin. This must be done before
* we return to open context, so that when we free the dataset's dnode
* we can evict its bonus buffer. Since the dataset may be destroyed
* at this point (and therefore won't have a valid pointer to the spa)
* we release the key mapping manually here while we do have a valid
* pointer, if it exists.
*/
if (!drc->drc_raw && encrypted) {
(void) spa_keystore_remove_mapping(dmu_tx_pool(tx)->dp_spa,
drc->drc_ds->ds_object, drc->drc_ds);
}
dsl_dataset_disown(drc->drc_ds, 0, dmu_recv_tag);
drc->drc_ds = NULL;
}
static int dmu_recv_end_modified_blocks = 3;
static int
dmu_recv_existing_end(dmu_recv_cookie_t *drc)
{
#ifdef _KERNEL
/*
* We will be destroying the ds; make sure its origin is unmounted if
* necessary.
*/
char name[ZFS_MAX_DATASET_NAME_LEN];
dsl_dataset_name(drc->drc_ds, name);
zfs_destroy_unmount_origin(name);
#endif
return (dsl_sync_task(drc->drc_tofs,
dmu_recv_end_check, dmu_recv_end_sync, drc,
dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL));
}
static int
dmu_recv_new_end(dmu_recv_cookie_t *drc)
{
return (dsl_sync_task(drc->drc_tofs,
dmu_recv_end_check, dmu_recv_end_sync, drc,
dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL));
}
int
dmu_recv_end(dmu_recv_cookie_t *drc, void *owner)
{
int error;
drc->drc_owner = owner;
if (drc->drc_newfs)
error = dmu_recv_new_end(drc);
else
error = dmu_recv_existing_end(drc);
if (error != 0) {
dmu_recv_cleanup_ds(drc);
nvlist_free(drc->drc_keynvl);
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
} else if (!drc->drc_heal) {
async zvol minor node creation interferes with receive When we finish a zfs receive, dmu_recv_end_sync() calls zvol_create_minors(async=TRUE). This kicks off some other threads that create the minor device nodes (in /dev/zvol/poolname/...). These async threads call zvol_prefetch_minors_impl() and zvol_create_minor(), which both call dmu_objset_own(), which puts a "long hold" on the dataset. Since the zvol minor node creation is asynchronous, this can happen after the `ZFS_IOC_RECV[_NEW]` ioctl and `zfs receive` process have completed. After the first receive ioctl has completed, userland may attempt to do another receive into the same dataset (e.g. the next incremental stream). This second receive and the asynchronous minor node creation can interfere with one another in several different ways, because they both require exclusive access to the dataset: 1. When the second receive is finishing up, dmu_recv_end_check() does dsl_dataset_handoff_check(), which can fail with EBUSY if the async minor node creation already has a "long hold" on this dataset. This causes the 2nd receive to fail. 2. The async udev rule can fail if zvol_id and/or systemd-udevd try to open the device while the the second receive's async attempt at minor node creation owns the dataset (via zvol_prefetch_minors_impl). This causes the minor node (/dev/zd*) to exist, but the udev-generated /dev/zvol/... to not exist. 3. The async minor node creation can silently fail with EBUSY if the first receive's zvol_create_minor() trys to own the dataset while the second receive's zvol_prefetch_minors_impl already owns the dataset. To address these problems, this change synchronously creates the minor node. To avoid the lock ordering problems that the asynchrony was introduced to fix (see #3681), we create the minor nodes from open context, with no locks held, rather than from syncing contex as was originally done. Implementation notes: We generally do not need to traverse children or prefetch anything (e.g. when running the recv, snapshot, create, or clone subcommands of zfs). We only need recursion when importing/opening a pool and when loading encryption keys. The existing recursive, asynchronous, prefetching code is preserved for use in these cases. Channel programs may need to create zvol minor nodes, when creating a snapshot of a zvol with the snapdev property set. We figure out what snapshots are created when running the LUA program in syncing context. In this case we need to remember what snapshots were created, and then try to create their minor nodes from open context, after the LUA code has completed. There are additional zvol use cases that asynchronously own the dataset, which can cause similar problems. E.g. changing the volmode or snapdev properties. These are less problematic because they are not recursive and don't touch datasets that are not involved in the operation, there is still potential for interference with subsequent operations. In the future, these cases should be similarly converted to create the zvol minor node synchronously from open context. The async tasks of removing and renaming minors do not own the objset, so they do not have this problem. However, it may make sense to also convert these operations to happen synchronously from open context, in the future. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> External-issue: DLPX-65948 Closes #7863 Closes #9885
2020-02-03 20:33:14 +03:00
if (drc->drc_newfs) {
zvol_create_minor(drc->drc_tofs);
}
char *snapname = kmem_asprintf("%s@%s",
drc->drc_tofs, drc->drc_tosnap);
zvol_create_minor(snapname);
kmem_strfree(snapname);
Refactor dmu_recv into its own file This change moves the bottom half of dmu_send.c (where the receive logic is kept) into a new file, dmu_recv.c, and does similarly for receive-related changes in header files. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7982
2018-10-10 00:05:13 +03:00
}
return (error);
}
/*
* Return TRUE if this objset is currently being received into.
*/
boolean_t
dmu_objset_is_receiving(objset_t *os)
{
return (os->os_dsl_dataset != NULL &&
os->os_dsl_dataset->ds_owner == dmu_recv_tag);
}
Cleanup: Specify unsignedness on things that should not be signed In #13871, zfs_vdev_aggregation_limit_non_rotating and zfs_vdev_aggregation_limit being signed was pointed out as a possible reason not to eliminate an unnecessary MAX(unsigned, 0) since the unsigned value was assigned from them. There is no reason for these module parameters to be signed and upon inspection, it was found that there are a number of other module parameters that are signed, but should not be, so we make them unsigned. Making them unsigned made it clear that some other variables in the code should also be unsigned, so we also make those unsigned. This prevents users from setting negative values that could potentially cause bad behaviors. It also makes the code slightly easier to understand. Mostly module parameters that deal with timeouts, limits, bitshifts and percentages are made unsigned by this. Any that are boolean are left signed, since whether booleans should be considered signed or unsigned does not matter. Making zfs_arc_lotsfree_percent unsigned caused a `zfs_arc_lotsfree_percent >= 0` check to become redundant, so it was removed. Removing the check was also necessary to prevent a compiler error from -Werror=type-limits. Several end of line comments had to be moved to their own lines because replacing int with uint_t caused us to exceed the 80 character limit enforced by cstyle.pl. The following were kept signed because they are passed to taskq_create(), which expects signed values and modifying the OpenSolaris/Illumos DDI is out of scope of this patch: * metaslab_load_pct * zfs_sync_taskq_batch_pct * zfs_zil_clean_taskq_nthr_pct * zfs_zil_clean_taskq_minalloc * zfs_zil_clean_taskq_maxalloc * zfs_arc_prune_task_threads Also, negative values in those parameters was found to be harmless. The following were left signed because either negative values make sense, or more analysis was needed to determine whether negative values should be disallowed: * zfs_metaslab_switch_threshold * zfs_pd_bytes_max * zfs_livelist_min_percent_shared zfs_multihost_history was made static to be consistent with other parameters. A number of module parameters were marked as signed, but in reality referenced unsigned variables. upgrade_errlog_limit is one of the numerous examples. In the case of zfs_vdev_async_read_max_active, it was already uint32_t, but zdb had an extern int declaration for it. Interestingly, the documentation in zfs.4 was right for upgrade_errlog_limit despite the module parameter being wrongly marked, while the documentation for zfs_vdev_async_read_max_active (and friends) was wrong. It was also wrong for zstd_abort_size, which was unsigned, but was documented as signed. Also, the documentation in zfs.4 incorrectly described the following parameters as ulong when they were int: * zfs_arc_meta_adjust_restarts * zfs_override_estimate_recordsize They are now uint_t as of this patch and thus the man page has been updated to describe them as uint. dbuf_state_index was left alone since it does nothing and perhaps should be removed in another patch. If any module parameters were missed, they were not found by `grep -r 'ZFS_MODULE_PARAM' | grep ', INT'`. I did find a few that grep missed, but only because they were in files that had hits. This patch intentionally did not attempt to address whether some of these module parameters should be elevated to 64-bit parameters, because the length of a long on 32-bit is 32-bit. Lastly, it was pointed out during review that uint_t is a better match for these variables than uint32_t because FreeBSD kernel parameter definitions are designed for uint_t, whose bit width can change in future memory models. As a result, we change the existing parameters that are uint32_t to use uint_t. Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Neal Gompa <ngompa@datto.com> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Closes #13875
2022-09-28 02:42:41 +03:00
ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_length, UINT, ZMOD_RW,
Make module tunables cross platform Adds ZFS_MODULE_PARAM to abstract module parameter setting to operating systems other than Linux. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matt Macy <mmacy@FreeBSD.org> Signed-off-by: Ryan Moeller <ryan@ixsystems.com> Closes #9230
2019-09-06 00:49:49 +03:00
"Maximum receive queue length");
Implement Redacted Send/Receive Redacted send/receive allows users to send subsets of their data to a target system. One possible use case for this feature is to not transmit sensitive information to a data warehousing, test/dev, or analytics environment. Another is to save space by not replicating unimportant data within a given dataset, for example in backup tools like zrepl. Redacted send/receive is a three-stage process. First, a clone (or clones) is made of the snapshot to be sent to the target. In this clone (or clones), all unnecessary or unwanted data is removed or modified. This clone is then snapshotted to create the "redaction snapshot" (or snapshots). Second, the new zfs redact command is used to create a redaction bookmark. The redaction bookmark stores the list of blocks in a snapshot that were modified by the redaction snapshot(s). Finally, the redaction bookmark is passed as a parameter to zfs send. When sending to the snapshot that was redacted, the redaction bookmark is used to filter out blocks that contain sensitive or unwanted information, and those blocks are not included in the send stream. When sending from the redaction bookmark, the blocks it contains are considered as candidate blocks in addition to those blocks in the destination snapshot that were modified since the creation_txg of the redaction bookmark. This step is necessary to allow the target to rehydrate data in the case where some blocks are accidentally or unnecessarily modified in the redaction snapshot. The changes to bookmarks to enable fast space estimation involve adding deadlists to bookmarks. There is also logic to manage the life cycles of these deadlists. The new size estimation process operates in cases where previously an accurate estimate could not be provided. In those cases, a send is performed where no data blocks are read, reducing the runtime significantly and providing a byte-accurate size estimate. Reviewed-by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Prashanth Sreenivasa <pks@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Pavel Zhakarov <pavel.zakharov@delphix.com> Reviewed-by: Sebastien Roy <sebastien.roy@delphix.com> Reviewed-by: Prakash Surya <prakash.surya@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Paul Dagnelie <pcd@delphix.com> Closes #7958
2019-06-19 19:48:13 +03:00
Cleanup: Specify unsignedness on things that should not be signed In #13871, zfs_vdev_aggregation_limit_non_rotating and zfs_vdev_aggregation_limit being signed was pointed out as a possible reason not to eliminate an unnecessary MAX(unsigned, 0) since the unsigned value was assigned from them. There is no reason for these module parameters to be signed and upon inspection, it was found that there are a number of other module parameters that are signed, but should not be, so we make them unsigned. Making them unsigned made it clear that some other variables in the code should also be unsigned, so we also make those unsigned. This prevents users from setting negative values that could potentially cause bad behaviors. It also makes the code slightly easier to understand. Mostly module parameters that deal with timeouts, limits, bitshifts and percentages are made unsigned by this. Any that are boolean are left signed, since whether booleans should be considered signed or unsigned does not matter. Making zfs_arc_lotsfree_percent unsigned caused a `zfs_arc_lotsfree_percent >= 0` check to become redundant, so it was removed. Removing the check was also necessary to prevent a compiler error from -Werror=type-limits. Several end of line comments had to be moved to their own lines because replacing int with uint_t caused us to exceed the 80 character limit enforced by cstyle.pl. The following were kept signed because they are passed to taskq_create(), which expects signed values and modifying the OpenSolaris/Illumos DDI is out of scope of this patch: * metaslab_load_pct * zfs_sync_taskq_batch_pct * zfs_zil_clean_taskq_nthr_pct * zfs_zil_clean_taskq_minalloc * zfs_zil_clean_taskq_maxalloc * zfs_arc_prune_task_threads Also, negative values in those parameters was found to be harmless. The following were left signed because either negative values make sense, or more analysis was needed to determine whether negative values should be disallowed: * zfs_metaslab_switch_threshold * zfs_pd_bytes_max * zfs_livelist_min_percent_shared zfs_multihost_history was made static to be consistent with other parameters. A number of module parameters were marked as signed, but in reality referenced unsigned variables. upgrade_errlog_limit is one of the numerous examples. In the case of zfs_vdev_async_read_max_active, it was already uint32_t, but zdb had an extern int declaration for it. Interestingly, the documentation in zfs.4 was right for upgrade_errlog_limit despite the module parameter being wrongly marked, while the documentation for zfs_vdev_async_read_max_active (and friends) was wrong. It was also wrong for zstd_abort_size, which was unsigned, but was documented as signed. Also, the documentation in zfs.4 incorrectly described the following parameters as ulong when they were int: * zfs_arc_meta_adjust_restarts * zfs_override_estimate_recordsize They are now uint_t as of this patch and thus the man page has been updated to describe them as uint. dbuf_state_index was left alone since it does nothing and perhaps should be removed in another patch. If any module parameters were missed, they were not found by `grep -r 'ZFS_MODULE_PARAM' | grep ', INT'`. I did find a few that grep missed, but only because they were in files that had hits. This patch intentionally did not attempt to address whether some of these module parameters should be elevated to 64-bit parameters, because the length of a long on 32-bit is 32-bit. Lastly, it was pointed out during review that uint_t is a better match for these variables than uint32_t because FreeBSD kernel parameter definitions are designed for uint_t, whose bit width can change in future memory models. As a result, we change the existing parameters that are uint32_t to use uint_t. Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Neal Gompa <ngompa@datto.com> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Closes #13875
2022-09-28 02:42:41 +03:00
ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_ff, UINT, ZMOD_RW,
Make module tunables cross platform Adds ZFS_MODULE_PARAM to abstract module parameter setting to operating systems other than Linux. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matt Macy <mmacy@FreeBSD.org> Signed-off-by: Ryan Moeller <ryan@ixsystems.com> Closes #9230
2019-09-06 00:49:49 +03:00
"Receive queue fill fraction");
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
Cleanup: Specify unsignedness on things that should not be signed In #13871, zfs_vdev_aggregation_limit_non_rotating and zfs_vdev_aggregation_limit being signed was pointed out as a possible reason not to eliminate an unnecessary MAX(unsigned, 0) since the unsigned value was assigned from them. There is no reason for these module parameters to be signed and upon inspection, it was found that there are a number of other module parameters that are signed, but should not be, so we make them unsigned. Making them unsigned made it clear that some other variables in the code should also be unsigned, so we also make those unsigned. This prevents users from setting negative values that could potentially cause bad behaviors. It also makes the code slightly easier to understand. Mostly module parameters that deal with timeouts, limits, bitshifts and percentages are made unsigned by this. Any that are boolean are left signed, since whether booleans should be considered signed or unsigned does not matter. Making zfs_arc_lotsfree_percent unsigned caused a `zfs_arc_lotsfree_percent >= 0` check to become redundant, so it was removed. Removing the check was also necessary to prevent a compiler error from -Werror=type-limits. Several end of line comments had to be moved to their own lines because replacing int with uint_t caused us to exceed the 80 character limit enforced by cstyle.pl. The following were kept signed because they are passed to taskq_create(), which expects signed values and modifying the OpenSolaris/Illumos DDI is out of scope of this patch: * metaslab_load_pct * zfs_sync_taskq_batch_pct * zfs_zil_clean_taskq_nthr_pct * zfs_zil_clean_taskq_minalloc * zfs_zil_clean_taskq_maxalloc * zfs_arc_prune_task_threads Also, negative values in those parameters was found to be harmless. The following were left signed because either negative values make sense, or more analysis was needed to determine whether negative values should be disallowed: * zfs_metaslab_switch_threshold * zfs_pd_bytes_max * zfs_livelist_min_percent_shared zfs_multihost_history was made static to be consistent with other parameters. A number of module parameters were marked as signed, but in reality referenced unsigned variables. upgrade_errlog_limit is one of the numerous examples. In the case of zfs_vdev_async_read_max_active, it was already uint32_t, but zdb had an extern int declaration for it. Interestingly, the documentation in zfs.4 was right for upgrade_errlog_limit despite the module parameter being wrongly marked, while the documentation for zfs_vdev_async_read_max_active (and friends) was wrong. It was also wrong for zstd_abort_size, which was unsigned, but was documented as signed. Also, the documentation in zfs.4 incorrectly described the following parameters as ulong when they were int: * zfs_arc_meta_adjust_restarts * zfs_override_estimate_recordsize They are now uint_t as of this patch and thus the man page has been updated to describe them as uint. dbuf_state_index was left alone since it does nothing and perhaps should be removed in another patch. If any module parameters were missed, they were not found by `grep -r 'ZFS_MODULE_PARAM' | grep ', INT'`. I did find a few that grep missed, but only because they were in files that had hits. This patch intentionally did not attempt to address whether some of these module parameters should be elevated to 64-bit parameters, because the length of a long on 32-bit is 32-bit. Lastly, it was pointed out during review that uint_t is a better match for these variables than uint32_t because FreeBSD kernel parameter definitions are designed for uint_t, whose bit width can change in future memory models. As a result, we change the existing parameters that are uint32_t to use uint_t. Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Neal Gompa <ngompa@datto.com> Signed-off-by: Richard Yao <richard.yao@alumni.stonybrook.edu> Closes #13875
2022-09-28 02:42:41 +03:00
ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, write_batch_size, UINT, ZMOD_RW,
Improve zfs receive performance by batching writes For each WRITE record in the stream, `zfs receive` creates a DMU transaction (`dmu_tx_create()`) and writes this block's data into the object. If per-block overheads (as opposed to per-byte overheads) dominate performance (as is often the case with small recordsize), the per-dmu-transaction overheads can be significant. For example, in some workloads the `receieve_writer` thread is 100% on CPU, and more than half of its CPU time is in these per-tx routines (e.g. dmu_tx_hold_write, dmu_tx_assign, dmu_tx_commit). To improve performance of `zfs receive`, this commit batches WRITE records which are to nearby offsets of the same object, and uses one DMU transaction to write them all. By default the batch size is 1MB, which for recordsize=8K reduces the number of DMU transactions by 128x for full send streams (incrementals will depend on how "clumpy" the changed blocks are). This commit improves the performance of `dd if=stream | zfs recv` from 78,800 blocks/sec to 98,100 blocks/sec (25% improvement). Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com> Closes #10099
2020-03-16 21:51:56 +03:00
"Maximum amount of writes to batch into one transaction");
Implement a new type of zfs receive: corrective receive (-c) This type of recv is used to heal corrupted data when a replica of the data already exists (in the form of a send file for example). With the provided send stream, corrective receive will read from disk blocks described by the WRITE records. When any of the reads come back with ECKSUM we use the data from the corresponding WRITE record to rewrite the corrupted block. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Paul Zuchowski <pzuchowski@datto.com> Signed-off-by: Alek Pinchuk <apinchuk@axcient.com> Closes #9372
2022-07-29 01:52:46 +03:00
ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, best_effort_corrective, INT, ZMOD_RW,
"Ignore errors during corrective receive");
/* END CSTYLED */
Reference in New Issue Copy Permalink