2008-11-20 23:01:55 +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
|
|
|
|
* or http://www.opensolaris.org/os/licensing.
|
|
|
|
* 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
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
2010-05-29 00:45:14 +04:00
|
|
|
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
|
2011-11-08 04:26:52 +04:00
|
|
|
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
|
OpenZFS 9112 - Improve allocation performance on high-end systems
Overview
========
We parallelize the allocation process by creating the concept of
"allocators". There are a certain number of allocators per metaslab
group, defined by the value of a tunable at pool open time. Each
allocator for a given metaslab group has up to 2 active metaslabs; one
"primary", and one "secondary". The primary and secondary weight mean
the same thing they did in in the pre-allocator world; primary metaslabs
are used for most allocations, secondary metaslabs are used for ditto
blocks being allocated in the same metaslab group. There is also the
CLAIM weight, which has been separated out from the other weights, but
that is less important to understanding the patch. The active metaslabs
for each allocator are moved from their normal place in the metaslab
tree for the group to the back of the tree. This way, they will not be
selected for use by other allocators searching for new metaslabs unless
all the passive metaslabs are unsuitable for allocations. If that does
happen, the allocators will "steal" from each other to ensure that IOs
don't fail until there is truly no space left to perform allocations.
In addition, the alloc queue for each metaslab group has been broken
into a separate queue for each allocator. We don't want to dramatically
increase the number of inflight IOs on low-end systems, because it can
significantly increase txg times. On the other hand, we want to ensure
that there are enough IOs for each allocator to allow for good
coalescing before sending the IOs to the disk. As a result, we take a
compromise path; each allocator's alloc queue max depth starts at a
certain value for every txg. Every time an IO completes, we increase the
max depth. This should hopefully provide a good balance between the two
failure modes, while not dramatically increasing complexity.
We also parallelize the spa_alloc_tree and spa_alloc_lock, which cause
very similar contention when selecting IOs to allocate. This
parallelization uses the same allocator scheme as metaslab selection.
Performance Results
===================
Performance improvements from this change can vary significantly based
on the number of CPUs in the system, whether or not the system has a
NUMA architecture, the speed of the drives, the values for the various
tunables, and the workload being performed. For an fio async sequential
write workload on a 24 core NUMA system with 256 GB of RAM and 8 128 GB
SSDs, there is a roughly 25% performance improvement.
Future Work
===========
Analysis of the performance of the system with this patch applied shows
that a significant new bottleneck is the vdev disk queues, which also
need to be parallelized. Prototyping of this change has occurred, and
there was a performance improvement, but more work needs to be done
before its stability has been verified and it is ready to be upstreamed.
Authored by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Serapheim Dimitropoulos <serapheim.dimitro@delphix.com>
Reviewed by: Alexander Motin <mav@FreeBSD.org>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Gordon Ross <gwr@nexenta.com>
Ported-by: Paul Dagnelie <pcd@delphix.com>
Signed-off-by: Paul Dagnelie <pcd@delphix.com>
Porting Notes:
* Fix reservation test failures by increasing tolerance.
OpenZFS-issue: https://illumos.org/issues/9112
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/3f3cc3c3
Closes #7682
2018-02-12 23:56:06 +03:00
|
|
|
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
|
2013-01-23 13:54:30 +04:00
|
|
|
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
|
2017-04-13 19:40:00 +03:00
|
|
|
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
|
|
|
|
* Copyright 2016 Toomas Soome <tsoome@me.com>
|
2011-11-08 04:26:52 +04:00
|
|
|
*/
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
#ifndef _ZIO_H
|
|
|
|
#define _ZIO_H
|
|
|
|
|
2015-12-22 04:31:57 +03:00
|
|
|
#include <sys/zio_priority.h>
|
2008-11-20 23:01:55 +03:00
|
|
|
#include <sys/zfs_context.h>
|
|
|
|
#include <sys/spa.h>
|
|
|
|
#include <sys/txg.h>
|
|
|
|
#include <sys/avl.h>
|
|
|
|
#include <sys/fs/zfs.h>
|
|
|
|
#include <sys/zio_impl.h>
|
|
|
|
|
|
|
|
#ifdef __cplusplus
|
|
|
|
extern "C" {
|
|
|
|
#endif
|
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
/*
|
|
|
|
* Embedded checksum
|
|
|
|
*/
|
|
|
|
#define ZEC_MAGIC 0x210da7ab10c7a11ULL
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
typedef struct zio_eck {
|
|
|
|
uint64_t zec_magic; /* for validation, endianness */
|
|
|
|
zio_cksum_t zec_cksum; /* 256-bit checksum */
|
|
|
|
} zio_eck_t;
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Gang block headers are self-checksumming and contain an array
|
|
|
|
* of block pointers.
|
|
|
|
*/
|
|
|
|
#define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
|
|
|
|
#define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
|
2010-05-29 00:45:14 +04:00
|
|
|
sizeof (zio_eck_t)) / sizeof (blkptr_t))
|
2008-11-20 23:01:55 +03:00
|
|
|
#define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
|
2010-05-29 00:45:14 +04:00
|
|
|
sizeof (zio_eck_t) - \
|
2008-11-20 23:01:55 +03:00
|
|
|
(SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
|
|
|
|
sizeof (uint64_t))
|
|
|
|
|
|
|
|
typedef struct zio_gbh {
|
|
|
|
blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
|
|
|
|
uint64_t zg_filler[SPA_GBH_FILLER];
|
2010-05-29 00:45:14 +04:00
|
|
|
zio_eck_t zg_tail;
|
2008-11-20 23:01:55 +03:00
|
|
|
} zio_gbh_phys_t;
|
|
|
|
|
|
|
|
enum zio_checksum {
|
|
|
|
ZIO_CHECKSUM_INHERIT = 0,
|
|
|
|
ZIO_CHECKSUM_ON,
|
|
|
|
ZIO_CHECKSUM_OFF,
|
|
|
|
ZIO_CHECKSUM_LABEL,
|
|
|
|
ZIO_CHECKSUM_GANG_HEADER,
|
|
|
|
ZIO_CHECKSUM_ZILOG,
|
|
|
|
ZIO_CHECKSUM_FLETCHER_2,
|
|
|
|
ZIO_CHECKSUM_FLETCHER_4,
|
|
|
|
ZIO_CHECKSUM_SHA256,
|
2010-05-29 00:45:14 +04:00
|
|
|
ZIO_CHECKSUM_ZILOG2,
|
2016-06-16 01:47:05 +03:00
|
|
|
ZIO_CHECKSUM_NOPARITY,
|
|
|
|
ZIO_CHECKSUM_SHA512,
|
|
|
|
ZIO_CHECKSUM_SKEIN,
|
|
|
|
ZIO_CHECKSUM_EDONR,
|
2008-11-20 23:01:55 +03:00
|
|
|
ZIO_CHECKSUM_FUNCTIONS
|
|
|
|
};
|
|
|
|
|
2014-06-06 01:19:08 +04:00
|
|
|
/*
|
|
|
|
* The number of "legacy" compression functions which can be set on individual
|
|
|
|
* objects.
|
|
|
|
*/
|
|
|
|
#define ZIO_CHECKSUM_LEGACY_FUNCTIONS ZIO_CHECKSUM_ZILOG2
|
|
|
|
|
2009-07-03 02:44:48 +04:00
|
|
|
#define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4
|
2008-11-20 23:01:55 +03:00
|
|
|
#define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
|
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
#define ZIO_CHECKSUM_MASK 0xffULL
|
|
|
|
#define ZIO_CHECKSUM_VERIFY (1 << 8)
|
|
|
|
|
|
|
|
#define ZIO_DEDUPCHECKSUM ZIO_CHECKSUM_SHA256
|
|
|
|
#define ZIO_DEDUPDITTO_MIN 100
|
|
|
|
|
Native Encryption for ZFS on Linux
This change incorporates three major pieces:
The first change is a keystore that manages wrapping
and encryption keys for encrypted datasets. These
commands mostly involve manipulating the new
DSL Crypto Key ZAP Objects that live in the MOS. Each
encrypted dataset has its own DSL Crypto Key that is
protected with a user's key. This level of indirection
allows users to change their keys without re-encrypting
their entire datasets. The change implements the new
subcommands "zfs load-key", "zfs unload-key" and
"zfs change-key" which allow the user to manage their
encryption keys and settings. In addition, several new
flags and properties have been added to allow dataset
creation and to make mounting and unmounting more
convenient.
The second piece of this patch provides the ability to
encrypt, decyrpt, and authenticate protected datasets.
Each object set maintains a Merkel tree of Message
Authentication Codes that protect the lower layers,
similarly to how checksums are maintained. This part
impacts the zio layer, which handles the actual
encryption and generation of MACs, as well as the ARC
and DMU, which need to be able to handle encrypted
buffers and protected data.
The last addition is the ability to do raw, encrypted
sends and receives. The idea here is to send raw
encrypted and compressed data and receive it exactly
as is on a backup system. This means that the dataset
on the receiving system is protected using the same
user key that is in use on the sending side. By doing
so, datasets can be efficiently backed up to an
untrusted system without fear of data being
compromised.
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #494
Closes #5769
2017-08-14 20:36:48 +03:00
|
|
|
/* supported encryption algorithms */
|
|
|
|
enum zio_encrypt {
|
|
|
|
ZIO_CRYPT_INHERIT = 0,
|
|
|
|
ZIO_CRYPT_ON,
|
|
|
|
ZIO_CRYPT_OFF,
|
|
|
|
ZIO_CRYPT_AES_128_CCM,
|
|
|
|
ZIO_CRYPT_AES_192_CCM,
|
|
|
|
ZIO_CRYPT_AES_256_CCM,
|
|
|
|
ZIO_CRYPT_AES_128_GCM,
|
|
|
|
ZIO_CRYPT_AES_192_GCM,
|
|
|
|
ZIO_CRYPT_AES_256_GCM,
|
|
|
|
ZIO_CRYPT_FUNCTIONS
|
|
|
|
};
|
|
|
|
|
|
|
|
#define ZIO_CRYPT_ON_VALUE ZIO_CRYPT_AES_256_CCM
|
|
|
|
#define ZIO_CRYPT_DEFAULT ZIO_CRYPT_OFF
|
|
|
|
|
|
|
|
/* macros defining encryption lengths */
|
|
|
|
#define ZIO_OBJSET_MAC_LEN 32
|
|
|
|
#define ZIO_DATA_IV_LEN 12
|
|
|
|
#define ZIO_DATA_SALT_LEN 8
|
|
|
|
#define ZIO_DATA_MAC_LEN 16
|
|
|
|
|
2014-06-06 01:19:08 +04:00
|
|
|
/*
|
|
|
|
* The number of "legacy" compression functions which can be set on individual
|
|
|
|
* objects.
|
|
|
|
*/
|
|
|
|
#define ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4
|
|
|
|
|
2015-07-06 04:55:32 +03:00
|
|
|
/*
|
|
|
|
* The meaning of "compress = on" selected by the compression features enabled
|
|
|
|
* on a given pool.
|
|
|
|
*/
|
|
|
|
#define ZIO_COMPRESS_LEGACY_ON_VALUE ZIO_COMPRESS_LZJB
|
|
|
|
#define ZIO_COMPRESS_LZ4_ON_VALUE ZIO_COMPRESS_LZ4
|
|
|
|
|
|
|
|
#define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
#define BOOTFS_COMPRESS_VALID(compress) \
|
|
|
|
((compress) == ZIO_COMPRESS_LZJB || \
|
2013-01-23 13:54:30 +04:00
|
|
|
(compress) == ZIO_COMPRESS_LZ4 || \
|
2016-12-03 10:13:44 +03:00
|
|
|
(compress) == ZIO_COMPRESS_GZIP_1 || \
|
|
|
|
(compress) == ZIO_COMPRESS_GZIP_2 || \
|
|
|
|
(compress) == ZIO_COMPRESS_GZIP_3 || \
|
|
|
|
(compress) == ZIO_COMPRESS_GZIP_4 || \
|
|
|
|
(compress) == ZIO_COMPRESS_GZIP_5 || \
|
|
|
|
(compress) == ZIO_COMPRESS_GZIP_6 || \
|
|
|
|
(compress) == ZIO_COMPRESS_GZIP_7 || \
|
|
|
|
(compress) == ZIO_COMPRESS_GZIP_8 || \
|
|
|
|
(compress) == ZIO_COMPRESS_GZIP_9 || \
|
|
|
|
(compress) == ZIO_COMPRESS_ZLE || \
|
2015-07-06 04:55:32 +03:00
|
|
|
(compress) == ZIO_COMPRESS_ON || \
|
2010-05-29 00:45:14 +04:00
|
|
|
(compress) == ZIO_COMPRESS_OFF)
|
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
#define ZIO_FAILURE_MODE_WAIT 0
|
|
|
|
#define ZIO_FAILURE_MODE_CONTINUE 1
|
|
|
|
#define ZIO_FAILURE_MODE_PANIC 2
|
|
|
|
|
2018-03-15 20:56:55 +03:00
|
|
|
typedef enum zio_suspend_reason {
|
|
|
|
ZIO_SUSPEND_NONE = 0,
|
|
|
|
ZIO_SUSPEND_IOERR,
|
|
|
|
ZIO_SUSPEND_MMP,
|
|
|
|
} zio_suspend_reason_t;
|
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
enum zio_flag {
|
|
|
|
/*
|
|
|
|
* Flags inherited by gang, ddt, and vdev children,
|
|
|
|
* and that must be equal for two zios to aggregate
|
|
|
|
*/
|
|
|
|
ZIO_FLAG_DONT_AGGREGATE = 1 << 0,
|
|
|
|
ZIO_FLAG_IO_REPAIR = 1 << 1,
|
|
|
|
ZIO_FLAG_SELF_HEAL = 1 << 2,
|
|
|
|
ZIO_FLAG_RESILVER = 1 << 3,
|
|
|
|
ZIO_FLAG_SCRUB = 1 << 4,
|
2010-08-27 01:24:34 +04:00
|
|
|
ZIO_FLAG_SCAN_THREAD = 1 << 5,
|
2014-09-23 03:42:03 +04:00
|
|
|
ZIO_FLAG_PHYSICAL = 1 << 6,
|
2010-05-29 00:45:14 +04:00
|
|
|
|
|
|
|
#define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Flags inherited by ddt, gang, and vdev children.
|
|
|
|
*/
|
2014-09-23 03:42:03 +04:00
|
|
|
ZIO_FLAG_CANFAIL = 1 << 7, /* must be first for INHERIT */
|
|
|
|
ZIO_FLAG_SPECULATIVE = 1 << 8,
|
|
|
|
ZIO_FLAG_CONFIG_WRITER = 1 << 9,
|
|
|
|
ZIO_FLAG_DONT_RETRY = 1 << 10,
|
|
|
|
ZIO_FLAG_DONT_CACHE = 1 << 11,
|
|
|
|
ZIO_FLAG_NODATA = 1 << 12,
|
|
|
|
ZIO_FLAG_INDUCE_DAMAGE = 1 << 13,
|
2016-10-14 03:59:18 +03:00
|
|
|
ZIO_FLAG_IO_ALLOCATING = 1 << 14,
|
2010-05-29 00:45:14 +04:00
|
|
|
|
|
|
|
#define ZIO_FLAG_DDT_INHERIT (ZIO_FLAG_IO_RETRY - 1)
|
|
|
|
#define ZIO_FLAG_GANG_INHERIT (ZIO_FLAG_IO_RETRY - 1)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Flags inherited by vdev children.
|
|
|
|
*/
|
2016-10-14 03:59:18 +03:00
|
|
|
ZIO_FLAG_IO_RETRY = 1 << 15, /* must be first for INHERIT */
|
|
|
|
ZIO_FLAG_PROBE = 1 << 16,
|
|
|
|
ZIO_FLAG_TRYHARD = 1 << 17,
|
|
|
|
ZIO_FLAG_OPTIONAL = 1 << 18,
|
2010-05-29 00:45:14 +04:00
|
|
|
|
|
|
|
#define ZIO_FLAG_VDEV_INHERIT (ZIO_FLAG_DONT_QUEUE - 1)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Flags not inherited by any children.
|
|
|
|
*/
|
2016-10-14 03:59:18 +03:00
|
|
|
ZIO_FLAG_DONT_QUEUE = 1 << 19, /* must be first for INHERIT */
|
|
|
|
ZIO_FLAG_DONT_PROPAGATE = 1 << 20,
|
|
|
|
ZIO_FLAG_IO_BYPASS = 1 << 21,
|
|
|
|
ZIO_FLAG_IO_REWRITE = 1 << 22,
|
Native Encryption for ZFS on Linux
This change incorporates three major pieces:
The first change is a keystore that manages wrapping
and encryption keys for encrypted datasets. These
commands mostly involve manipulating the new
DSL Crypto Key ZAP Objects that live in the MOS. Each
encrypted dataset has its own DSL Crypto Key that is
protected with a user's key. This level of indirection
allows users to change their keys without re-encrypting
their entire datasets. The change implements the new
subcommands "zfs load-key", "zfs unload-key" and
"zfs change-key" which allow the user to manage their
encryption keys and settings. In addition, several new
flags and properties have been added to allow dataset
creation and to make mounting and unmounting more
convenient.
The second piece of this patch provides the ability to
encrypt, decyrpt, and authenticate protected datasets.
Each object set maintains a Merkel tree of Message
Authentication Codes that protect the lower layers,
similarly to how checksums are maintained. This part
impacts the zio layer, which handles the actual
encryption and generation of MACs, as well as the ARC
and DMU, which need to be able to handle encrypted
buffers and protected data.
The last addition is the ability to do raw, encrypted
sends and receives. The idea here is to send raw
encrypted and compressed data and receive it exactly
as is on a backup system. This means that the dataset
on the receiving system is protected using the same
user key that is in use on the sending side. By doing
so, datasets can be efficiently backed up to an
untrusted system without fear of data being
compromised.
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #494
Closes #5769
2017-08-14 20:36:48 +03:00
|
|
|
ZIO_FLAG_RAW_COMPRESS = 1 << 23,
|
|
|
|
ZIO_FLAG_RAW_ENCRYPT = 1 << 24,
|
|
|
|
ZIO_FLAG_GANG_CHILD = 1 << 25,
|
|
|
|
ZIO_FLAG_DDT_CHILD = 1 << 26,
|
|
|
|
ZIO_FLAG_GODFATHER = 1 << 27,
|
|
|
|
ZIO_FLAG_NOPWRITE = 1 << 28,
|
|
|
|
ZIO_FLAG_REEXECUTED = 1 << 29,
|
|
|
|
ZIO_FLAG_DELEGATED = 1 << 30,
|
|
|
|
ZIO_FLAG_FASTWRITE = 1 << 31,
|
2010-05-29 00:45:14 +04:00
|
|
|
};
|
|
|
|
|
|
|
|
#define ZIO_FLAG_MUSTSUCCEED 0
|
Native Encryption for ZFS on Linux
This change incorporates three major pieces:
The first change is a keystore that manages wrapping
and encryption keys for encrypted datasets. These
commands mostly involve manipulating the new
DSL Crypto Key ZAP Objects that live in the MOS. Each
encrypted dataset has its own DSL Crypto Key that is
protected with a user's key. This level of indirection
allows users to change their keys without re-encrypting
their entire datasets. The change implements the new
subcommands "zfs load-key", "zfs unload-key" and
"zfs change-key" which allow the user to manage their
encryption keys and settings. In addition, several new
flags and properties have been added to allow dataset
creation and to make mounting and unmounting more
convenient.
The second piece of this patch provides the ability to
encrypt, decyrpt, and authenticate protected datasets.
Each object set maintains a Merkel tree of Message
Authentication Codes that protect the lower layers,
similarly to how checksums are maintained. This part
impacts the zio layer, which handles the actual
encryption and generation of MACs, as well as the ARC
and DMU, which need to be able to handle encrypted
buffers and protected data.
The last addition is the ability to do raw, encrypted
sends and receives. The idea here is to send raw
encrypted and compressed data and receive it exactly
as is on a backup system. This means that the dataset
on the receiving system is protected using the same
user key that is in use on the sending side. By doing
so, datasets can be efficiently backed up to an
untrusted system without fear of data being
compromised.
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #494
Closes #5769
2017-08-14 20:36:48 +03:00
|
|
|
#define ZIO_FLAG_RAW (ZIO_FLAG_RAW_COMPRESS | ZIO_FLAG_RAW_ENCRYPT)
|
2010-05-29 00:45:14 +04:00
|
|
|
|
|
|
|
#define ZIO_DDT_CHILD_FLAGS(zio) \
|
|
|
|
(((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \
|
|
|
|
ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
|
|
|
|
|
2008-12-03 23:09:06 +03:00
|
|
|
#define ZIO_GANG_CHILD_FLAGS(zio) \
|
|
|
|
(((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \
|
|
|
|
ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
|
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
#define ZIO_VDEV_CHILD_FLAGS(zio) \
|
|
|
|
(((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \
|
OpenZFS 7614, 9064 - zfs device evacuation/removal
OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete
This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk. The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.
The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool. An entry becomes obsolete when all the blocks that use
it are freed. An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones). Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible. This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.
Note that when a device is removed, we do not verify the checksum of
the data that is copied. This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.
At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.
Porting Notes:
* Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
* Remove comment regarding "mpt" driver where zfs_remove_max_segment
is initialized to SPA_MAXBLOCKSIZE.
Change zfs_condense_indirect_commit_entry_delay_ticks to
zfs_condense_indirect_commit_entry_delay_ms for consistency with
most other tunables in which delays are specified in ms.
* ZTS changes:
Use set_tunable rather than mdb
Use zpool sync as appropriate
Use sync_pool instead of sync
Kill jobs during test_removal_with_operation to allow unmount/export
Don't add non-disk names such as "mirror" or "raidz" to $DISKS
Use $TEST_BASE_DIR instead of /tmp
Increase HZ from 100 to 1000 which is more common on Linux
removal_multiple_indirection.ksh
Reduce iterations in order to not time out on the code
coverage builders.
removal_resume_export:
Functionally, the test case is correct but there exists a race
where the kernel thread hasn't been fully started yet and is
not visible. Wait for up to 1 second for the removal thread
to be started before giving up on it. Also, increase the
amount of data copied in order that the removal not finish
before the export has a chance to fail.
* MMP compatibility, the concept of concrete versus non-concrete devices
has slightly changed the semantics of vdev_writeable(). Update
mmp_random_leaf_impl() accordingly.
* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
feature which is not supported by OpenZFS.
* Added support for new vdev removal tracepoints.
* Test cases removal_with_zdb and removal_condense_export have been
intentionally disabled. When run manually they pass as intended,
but when running in the automated test environment they produce
unreliable results on the latest Fedora release.
They may work better once the upstream pool import refectoring is
merged into ZoL at which point they will be re-enabled.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alex Reece <alex@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Richard Laager <rlaager@wiktel.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Garrett D'Amore <garrett@damore.org>
Ported-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900
2016-09-22 19:30:13 +03:00
|
|
|
ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_CANFAIL)
|
2010-05-29 00:45:14 +04:00
|
|
|
|
OpenZFS 8857 - zio_remove_child() panic due to already destroyed parent zio
PROBLEM
=======
It's possible for a parent zio to complete even though it has children
which have not completed. This can result in the following panic:
> $C
ffffff01809128c0 vpanic()
ffffff01809128e0 mutex_panic+0x58(fffffffffb94c904, ffffff597dde7f80)
ffffff0180912950 mutex_vector_enter+0x347(ffffff597dde7f80)
ffffff01809129b0 zio_remove_child+0x50(ffffff597dde7c58, ffffff32bd901ac0,
ffffff3373370908)
ffffff0180912a40 zio_done+0x390(ffffff32bd901ac0)
ffffff0180912a70 zio_execute+0x78(ffffff32bd901ac0)
ffffff0180912b30 taskq_thread+0x2d0(ffffff33bae44140)
ffffff0180912b40 thread_start+8()
> ::status
debugging crash dump vmcore.2 (64-bit) from batfs0390
operating system: 5.11 joyent_20170911T171900Z (i86pc)
image uuid: (not set)
panic message: mutex_enter: bad mutex, lp=ffffff597dde7f80
owner=ffffff3c59b39480 thread=ffffff0180912c40
dump content: kernel pages only
The problem is that dbuf_prefetch along with l2arc can create a zio tree
which confuses the parent zio and allows it to complete with while children
still exist. Here's the scenario:
zio tree:
pio
|--- lio
The parent zio, pio, has entered the zio_done stage and begins to check its
children to see there are still some that have not completed. In zio_done(),
the children are checked in the following order:
zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE)
zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_DONE)
zio_wait_for_children(zio, ZIO_CHILD_DDT, ZIO_WAIT_DONE)
zio_wait_for_children(zio, ZIO_CHILD_LOGICAL, ZIO_WAIT_DONE)
If pio, finds any child which has not completed then it stops executing and
goes to sleep. Each call to zio_wait_for_children() will grab the io_lock
while checking the particular child.
In this scenario, the pio has completed the first call to
zio_wait_for_children() to check for any ZIO_CHILD_VDEV children. Since
the only zio in the zio tree right now is the logical zio, lio, then it
completes that call and prepares to check the next child type.
In the meantime, the lio completes and in its callback creates a child vdev
zio, cio. The zio tree looks like this:
zio tree:
pio
|--- lio
|--- cio
The lio then grabs the parent's io_lock and removes itself.
zio tree:
pio
|--- cio
The pio continues to run but has already completed its check for ZIO_CHILD_VDEV
and will erroneously complete. When the child zio, cio, completes it will panic
the system trying to reference the parent zio which has been destroyed.
SOLUTION
========
The fix is to rework the zio_wait_for_children() logic to accept a bitfield
for all the children types that it's interested in checking. The
io_lock will is held the entire time we check all the children types. Since
the function now accepts a bitfield, a simple ZIO_CHILD_BIT() macro is provided
to allow for the conversion between a ZIO_CHILD type and the bitfield used by
the zio_wiat_for_children logic.
Authored by: George Wilson <george.wilson@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Andriy Gapon <avg@FreeBSD.org>
Reviewed by: Youzhong Yang <youzhong@gmail.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Dan McDonald <danmcd@omniti.com>
Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
OpenZFS-issue: https://www.illumos.org/issues/8857
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/862ff6d99c
Issue #5918
Closes #7168
2018-02-08 23:04:14 +03:00
|
|
|
#define ZIO_CHILD_BIT(x) (1 << (x))
|
|
|
|
#define ZIO_CHILD_BIT_IS_SET(val, x) ((val) & (1 << (x)))
|
|
|
|
|
2008-12-03 23:09:06 +03:00
|
|
|
enum zio_child {
|
|
|
|
ZIO_CHILD_VDEV = 0,
|
|
|
|
ZIO_CHILD_GANG,
|
2010-05-29 00:45:14 +04:00
|
|
|
ZIO_CHILD_DDT,
|
2008-12-03 23:09:06 +03:00
|
|
|
ZIO_CHILD_LOGICAL,
|
|
|
|
ZIO_CHILD_TYPES
|
|
|
|
};
|
|
|
|
|
OpenZFS 8857 - zio_remove_child() panic due to already destroyed parent zio
PROBLEM
=======
It's possible for a parent zio to complete even though it has children
which have not completed. This can result in the following panic:
> $C
ffffff01809128c0 vpanic()
ffffff01809128e0 mutex_panic+0x58(fffffffffb94c904, ffffff597dde7f80)
ffffff0180912950 mutex_vector_enter+0x347(ffffff597dde7f80)
ffffff01809129b0 zio_remove_child+0x50(ffffff597dde7c58, ffffff32bd901ac0,
ffffff3373370908)
ffffff0180912a40 zio_done+0x390(ffffff32bd901ac0)
ffffff0180912a70 zio_execute+0x78(ffffff32bd901ac0)
ffffff0180912b30 taskq_thread+0x2d0(ffffff33bae44140)
ffffff0180912b40 thread_start+8()
> ::status
debugging crash dump vmcore.2 (64-bit) from batfs0390
operating system: 5.11 joyent_20170911T171900Z (i86pc)
image uuid: (not set)
panic message: mutex_enter: bad mutex, lp=ffffff597dde7f80
owner=ffffff3c59b39480 thread=ffffff0180912c40
dump content: kernel pages only
The problem is that dbuf_prefetch along with l2arc can create a zio tree
which confuses the parent zio and allows it to complete with while children
still exist. Here's the scenario:
zio tree:
pio
|--- lio
The parent zio, pio, has entered the zio_done stage and begins to check its
children to see there are still some that have not completed. In zio_done(),
the children are checked in the following order:
zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE)
zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_DONE)
zio_wait_for_children(zio, ZIO_CHILD_DDT, ZIO_WAIT_DONE)
zio_wait_for_children(zio, ZIO_CHILD_LOGICAL, ZIO_WAIT_DONE)
If pio, finds any child which has not completed then it stops executing and
goes to sleep. Each call to zio_wait_for_children() will grab the io_lock
while checking the particular child.
In this scenario, the pio has completed the first call to
zio_wait_for_children() to check for any ZIO_CHILD_VDEV children. Since
the only zio in the zio tree right now is the logical zio, lio, then it
completes that call and prepares to check the next child type.
In the meantime, the lio completes and in its callback creates a child vdev
zio, cio. The zio tree looks like this:
zio tree:
pio
|--- lio
|--- cio
The lio then grabs the parent's io_lock and removes itself.
zio tree:
pio
|--- cio
The pio continues to run but has already completed its check for ZIO_CHILD_VDEV
and will erroneously complete. When the child zio, cio, completes it will panic
the system trying to reference the parent zio which has been destroyed.
SOLUTION
========
The fix is to rework the zio_wait_for_children() logic to accept a bitfield
for all the children types that it's interested in checking. The
io_lock will is held the entire time we check all the children types. Since
the function now accepts a bitfield, a simple ZIO_CHILD_BIT() macro is provided
to allow for the conversion between a ZIO_CHILD type and the bitfield used by
the zio_wiat_for_children logic.
Authored by: George Wilson <george.wilson@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Andriy Gapon <avg@FreeBSD.org>
Reviewed by: Youzhong Yang <youzhong@gmail.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Dan McDonald <danmcd@omniti.com>
Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
OpenZFS-issue: https://www.illumos.org/issues/8857
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/862ff6d99c
Issue #5918
Closes #7168
2018-02-08 23:04:14 +03:00
|
|
|
#define ZIO_CHILD_VDEV_BIT ZIO_CHILD_BIT(ZIO_CHILD_VDEV)
|
|
|
|
#define ZIO_CHILD_GANG_BIT ZIO_CHILD_BIT(ZIO_CHILD_GANG)
|
|
|
|
#define ZIO_CHILD_DDT_BIT ZIO_CHILD_BIT(ZIO_CHILD_DDT)
|
|
|
|
#define ZIO_CHILD_LOGICAL_BIT ZIO_CHILD_BIT(ZIO_CHILD_LOGICAL)
|
|
|
|
#define ZIO_CHILD_ALL_BITS \
|
Reduce taskq and context-switch cost of zio pipe
When doing a read from disk, ZFS creates 3 ZIO's: a zio_null(), the
logical zio_read(), and then a physical zio. Currently, each of these
results in a separate taskq_dispatch(zio_execute).
On high-read-iops workloads, this causes a significant performance
impact. By processing all 3 ZIO's in a single taskq entry, we reduce the
overhead on taskq locking and context switching. We accomplish this by
allowing zio_done() to return a "next zio to execute" to zio_execute().
This results in a ~12% performance increase for random reads, from
96,000 iops to 108,000 iops (with recordsize=8k, on SSD's).
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: George Wilson <george.wilson@delphix.com>
Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
External-issue: DLPX-59292
Closes #7736
2018-08-03 01:51:45 +03:00
|
|
|
(ZIO_CHILD_VDEV_BIT | ZIO_CHILD_GANG_BIT | \
|
OpenZFS 8857 - zio_remove_child() panic due to already destroyed parent zio
PROBLEM
=======
It's possible for a parent zio to complete even though it has children
which have not completed. This can result in the following panic:
> $C
ffffff01809128c0 vpanic()
ffffff01809128e0 mutex_panic+0x58(fffffffffb94c904, ffffff597dde7f80)
ffffff0180912950 mutex_vector_enter+0x347(ffffff597dde7f80)
ffffff01809129b0 zio_remove_child+0x50(ffffff597dde7c58, ffffff32bd901ac0,
ffffff3373370908)
ffffff0180912a40 zio_done+0x390(ffffff32bd901ac0)
ffffff0180912a70 zio_execute+0x78(ffffff32bd901ac0)
ffffff0180912b30 taskq_thread+0x2d0(ffffff33bae44140)
ffffff0180912b40 thread_start+8()
> ::status
debugging crash dump vmcore.2 (64-bit) from batfs0390
operating system: 5.11 joyent_20170911T171900Z (i86pc)
image uuid: (not set)
panic message: mutex_enter: bad mutex, lp=ffffff597dde7f80
owner=ffffff3c59b39480 thread=ffffff0180912c40
dump content: kernel pages only
The problem is that dbuf_prefetch along with l2arc can create a zio tree
which confuses the parent zio and allows it to complete with while children
still exist. Here's the scenario:
zio tree:
pio
|--- lio
The parent zio, pio, has entered the zio_done stage and begins to check its
children to see there are still some that have not completed. In zio_done(),
the children are checked in the following order:
zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE)
zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_DONE)
zio_wait_for_children(zio, ZIO_CHILD_DDT, ZIO_WAIT_DONE)
zio_wait_for_children(zio, ZIO_CHILD_LOGICAL, ZIO_WAIT_DONE)
If pio, finds any child which has not completed then it stops executing and
goes to sleep. Each call to zio_wait_for_children() will grab the io_lock
while checking the particular child.
In this scenario, the pio has completed the first call to
zio_wait_for_children() to check for any ZIO_CHILD_VDEV children. Since
the only zio in the zio tree right now is the logical zio, lio, then it
completes that call and prepares to check the next child type.
In the meantime, the lio completes and in its callback creates a child vdev
zio, cio. The zio tree looks like this:
zio tree:
pio
|--- lio
|--- cio
The lio then grabs the parent's io_lock and removes itself.
zio tree:
pio
|--- cio
The pio continues to run but has already completed its check for ZIO_CHILD_VDEV
and will erroneously complete. When the child zio, cio, completes it will panic
the system trying to reference the parent zio which has been destroyed.
SOLUTION
========
The fix is to rework the zio_wait_for_children() logic to accept a bitfield
for all the children types that it's interested in checking. The
io_lock will is held the entire time we check all the children types. Since
the function now accepts a bitfield, a simple ZIO_CHILD_BIT() macro is provided
to allow for the conversion between a ZIO_CHILD type and the bitfield used by
the zio_wiat_for_children logic.
Authored by: George Wilson <george.wilson@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Andriy Gapon <avg@FreeBSD.org>
Reviewed by: Youzhong Yang <youzhong@gmail.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Dan McDonald <danmcd@omniti.com>
Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
OpenZFS-issue: https://www.illumos.org/issues/8857
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/862ff6d99c
Issue #5918
Closes #7168
2018-02-08 23:04:14 +03:00
|
|
|
ZIO_CHILD_DDT_BIT | ZIO_CHILD_LOGICAL_BIT)
|
|
|
|
|
2008-12-03 23:09:06 +03:00
|
|
|
enum zio_wait_type {
|
|
|
|
ZIO_WAIT_READY = 0,
|
|
|
|
ZIO_WAIT_DONE,
|
|
|
|
ZIO_WAIT_TYPES
|
|
|
|
};
|
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
/*
|
|
|
|
* We'll take the unused errnos, 'EBADE' and 'EBADR' (from the Convergent
|
|
|
|
* graveyard) to indicate checksum errors and fragmentation.
|
|
|
|
*/
|
|
|
|
#define ECKSUM EBADE
|
|
|
|
#define EFRAGS EBADR
|
|
|
|
|
OpenZFS 7614, 9064 - zfs device evacuation/removal
OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete
This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk. The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.
The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool. An entry becomes obsolete when all the blocks that use
it are freed. An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones). Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible. This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.
Note that when a device is removed, we do not verify the checksum of
the data that is copied. This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.
At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.
Porting Notes:
* Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
* Remove comment regarding "mpt" driver where zfs_remove_max_segment
is initialized to SPA_MAXBLOCKSIZE.
Change zfs_condense_indirect_commit_entry_delay_ticks to
zfs_condense_indirect_commit_entry_delay_ms for consistency with
most other tunables in which delays are specified in ms.
* ZTS changes:
Use set_tunable rather than mdb
Use zpool sync as appropriate
Use sync_pool instead of sync
Kill jobs during test_removal_with_operation to allow unmount/export
Don't add non-disk names such as "mirror" or "raidz" to $DISKS
Use $TEST_BASE_DIR instead of /tmp
Increase HZ from 100 to 1000 which is more common on Linux
removal_multiple_indirection.ksh
Reduce iterations in order to not time out on the code
coverage builders.
removal_resume_export:
Functionally, the test case is correct but there exists a race
where the kernel thread hasn't been fully started yet and is
not visible. Wait for up to 1 second for the removal thread
to be started before giving up on it. Also, increase the
amount of data copied in order that the removal not finish
before the export has a chance to fail.
* MMP compatibility, the concept of concrete versus non-concrete devices
has slightly changed the semantics of vdev_writeable(). Update
mmp_random_leaf_impl() accordingly.
* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
feature which is not supported by OpenZFS.
* Added support for new vdev removal tracepoints.
* Test cases removal_with_zdb and removal_condense_export have been
intentionally disabled. When run manually they pass as intended,
but when running in the automated test environment they produce
unreliable results on the latest Fedora release.
They may work better once the upstream pool import refectoring is
merged into ZoL at which point they will be re-enabled.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alex Reece <alex@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Richard Laager <rlaager@wiktel.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Garrett D'Amore <garrett@damore.org>
Ported-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900
2016-09-22 19:30:13 +03:00
|
|
|
/* Similar for ENOACTIVE */
|
|
|
|
#define ENOTACTIVE ENOANO
|
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
typedef void zio_done_func_t(zio_t *zio);
|
|
|
|
|
2016-10-14 03:59:18 +03:00
|
|
|
extern int zio_dva_throttle_enabled;
|
Illumos #4045 write throttle & i/o scheduler performance work
4045 zfs write throttle & i/o scheduler performance work
1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver. The scheduler
issues a number of concurrent i/os from each class to the device. Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes). The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is. See the block comment in vdev_queue.c (reproduced
below) for more details.
2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load. The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system. When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount. This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens. One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync(). Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes. See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.
This diff has several other effects, including:
* the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.
* the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently. There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.
* zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc. This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).
--matt
APPENDIX: problems with the current i/o scheduler
The current ZFS i/o scheduler (vdev_queue.c) is deadline based. The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.
For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due". One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).
If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os. This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future. If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due. Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).
Notes on porting to ZFS on Linux:
- zio_t gained new members io_physdone and io_phys_children. Because
object caches in the Linux port call the constructor only once at
allocation time, objects may contain residual data when retrieved
from the cache. Therefore zio_create() was updated to zero out the two
new fields.
- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
(vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
This tree has been replaced by vq->vq_active_tree which is now used
for the same purpose.
- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
the number of vdev I/O buffers to pre-allocate. That global no longer
exists, so we instead use the sum of the *_max_active values for each of
the five I/O classes described above.
- The Illumos implementation of dmu_tx_delay() delays a transaction by
sleeping in condition variable embedded in the thread
(curthread->t_delay_cv). We do not have an equivalent CV to use in
Linux, so this change replaced the delay logic with a wrapper called
zfs_sleep_until(). This wrapper could be adopted upstream and in other
downstream ports to abstract away operating system-specific delay logic.
- These tunables are added as module parameters, and descriptions added
to the zfs-module-parameters.5 man page.
spa_asize_inflation
zfs_deadman_synctime_ms
zfs_vdev_max_active
zfs_vdev_async_write_active_min_dirty_percent
zfs_vdev_async_write_active_max_dirty_percent
zfs_vdev_async_read_max_active
zfs_vdev_async_read_min_active
zfs_vdev_async_write_max_active
zfs_vdev_async_write_min_active
zfs_vdev_scrub_max_active
zfs_vdev_scrub_min_active
zfs_vdev_sync_read_max_active
zfs_vdev_sync_read_min_active
zfs_vdev_sync_write_max_active
zfs_vdev_sync_write_min_active
zfs_dirty_data_max_percent
zfs_delay_min_dirty_percent
zfs_dirty_data_max_max_percent
zfs_dirty_data_max
zfs_dirty_data_max_max
zfs_dirty_data_sync
zfs_delay_scale
The latter four have type unsigned long, whereas they are uint64_t in
Illumos. This accommodates Linux's module_param() supported types, but
means they may overflow on 32-bit architectures.
The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
likely to overflow on 32-bit systems, since they express physical RAM
sizes in bytes. In fact, Illumos initializes zfs_dirty_data_max_max to
2^32 which does overflow. To resolve that, this port instead initializes
it in arc_init() to 25% of physical RAM, and adds the tunable
zfs_dirty_data_max_max_percent to override that percentage. While this
solution doesn't completely avoid the overflow issue, it should be a
reasonable default for most systems, and the minority of affected
systems can work around the issue by overriding the defaults.
- Fixed reversed logic in comment above zfs_delay_scale declaration.
- Clarified comments in vdev_queue.c regarding when per-queue minimums take
effect.
- Replaced dmu_tx_write_limit in the dmu_tx kstat file
with dmu_tx_dirty_delay and dmu_tx_dirty_over_max. The first counts
how many times a transaction has been delayed because the pool dirty
data has exceeded zfs_delay_min_dirty_percent. The latter counts how
many times the pool dirty data has exceeded zfs_dirty_data_max (which
we expect to never happen).
- The original patch would have regressed the bug fixed in
zfsonlinux/zfs@c418410, which prevented users from setting the
zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
A similar fix is added to vdev_queue_aggregate().
- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
heap instead of the stack. In Linux we can't afford such large
structures on the stack.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>
References:
http://www.illumos.org/issues/4045
illumos/illumos-gate@69962b5647e4a8b9b14998733b765925381b727e
Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-08-29 07:01:20 +04:00
|
|
|
extern const char *zio_type_name[ZIO_TYPES];
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
/*
|
|
|
|
* A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
|
|
|
|
* identifies any block in the pool. By convention, the meta-objset (MOS)
|
2010-05-29 00:45:14 +04:00
|
|
|
* is objset 0, and the meta-dnode is object 0. This covers all blocks
|
|
|
|
* except root blocks and ZIL blocks, which are defined as follows:
|
2008-11-20 23:01:55 +03:00
|
|
|
*
|
2010-05-29 00:45:14 +04:00
|
|
|
* Root blocks (objset_phys_t) are object 0, level -1: <objset, 0, -1, 0>.
|
|
|
|
* ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
|
|
|
|
* dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
|
2015-12-22 04:31:57 +03:00
|
|
|
* dnode visit bookmarks are <objset, object id of dnode, -3, 0>.
|
2008-11-20 23:01:55 +03:00
|
|
|
*
|
2010-05-29 00:45:14 +04:00
|
|
|
* Note: this structure is called a bookmark because its original purpose
|
|
|
|
* was to remember where to resume a pool-wide traverse.
|
2008-11-20 23:01:55 +03:00
|
|
|
*
|
2014-06-25 22:37:59 +04:00
|
|
|
* Note: this structure is passed between userland and the kernel, and is
|
|
|
|
* stored on disk (by virtue of being incorporated into other on-disk
|
|
|
|
* structures, e.g. dsl_scan_phys_t).
|
2008-11-20 23:01:55 +03:00
|
|
|
*/
|
2014-06-25 22:37:59 +04:00
|
|
|
struct zbookmark_phys {
|
2008-11-20 23:01:55 +03:00
|
|
|
uint64_t zb_objset;
|
|
|
|
uint64_t zb_object;
|
|
|
|
int64_t zb_level;
|
|
|
|
uint64_t zb_blkid;
|
Add visibility in to arc_read
This change is an attempt to add visibility into the arc_read calls
occurring on a system, in real time. To do this, a list was added to the
in memory SPA data structure for a pool, with each element on the list
corresponding to a call to arc_read. These entries are then exported
through the kstat interface, which can then be interpreted in userspace.
For each arc_read call, the following information is exported:
* A unique identifier (uint64_t)
* The time the entry was added to the list (hrtime_t)
(*not* wall clock time; relative to the other entries on the list)
* The objset ID (uint64_t)
* The object number (uint64_t)
* The indirection level (uint64_t)
* The block ID (uint64_t)
* The name of the function originating the arc_read call (char[24])
* The arc_flags from the arc_read call (uint32_t)
* The PID of the reading thread (pid_t)
* The command or name of thread originating read (char[16])
From this exported information one can see, in real time, exactly what
is being read, what function is generating the read, and whether or not
the read was found to be already cached.
There is still some work to be done, but this should serve as a good
starting point.
Specifically, dbuf_read's are not accounted for in the currently
exported information. Thus, a follow up patch should probably be added
to export these calls that never call into arc_read (they only hit the
dbuf hash table). In addition, it might be nice to create a utility
similar to "arcstat.py" to digest the exported information and display
it in a more readable format. Or perhaps, log the information and allow
for it to be "replayed" at a later time.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2013-09-07 03:09:05 +04:00
|
|
|
};
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
#define SET_BOOKMARK(zb, objset, object, level, blkid) \
|
|
|
|
{ \
|
|
|
|
(zb)->zb_objset = objset; \
|
|
|
|
(zb)->zb_object = object; \
|
|
|
|
(zb)->zb_level = level; \
|
|
|
|
(zb)->zb_blkid = blkid; \
|
|
|
|
}
|
|
|
|
|
|
|
|
#define ZB_DESTROYED_OBJSET (-1ULL)
|
|
|
|
|
|
|
|
#define ZB_ROOT_OBJECT (0ULL)
|
|
|
|
#define ZB_ROOT_LEVEL (-1LL)
|
|
|
|
#define ZB_ROOT_BLKID (0ULL)
|
|
|
|
|
|
|
|
#define ZB_ZIL_OBJECT (0ULL)
|
|
|
|
#define ZB_ZIL_LEVEL (-2LL)
|
|
|
|
|
2015-12-22 04:31:57 +03:00
|
|
|
#define ZB_DNODE_LEVEL (-3LL)
|
|
|
|
#define ZB_DNODE_BLKID (0ULL)
|
|
|
|
|
2012-12-14 03:24:15 +04:00
|
|
|
#define ZB_IS_ZERO(zb) \
|
|
|
|
((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \
|
|
|
|
(zb)->zb_level == 0 && (zb)->zb_blkid == 0)
|
|
|
|
#define ZB_IS_ROOT(zb) \
|
|
|
|
((zb)->zb_object == ZB_ROOT_OBJECT && \
|
|
|
|
(zb)->zb_level == ZB_ROOT_LEVEL && \
|
|
|
|
(zb)->zb_blkid == ZB_ROOT_BLKID)
|
|
|
|
|
2008-12-03 23:09:06 +03:00
|
|
|
typedef struct zio_prop {
|
|
|
|
enum zio_checksum zp_checksum;
|
|
|
|
enum zio_compress zp_compress;
|
|
|
|
dmu_object_type_t zp_type;
|
|
|
|
uint8_t zp_level;
|
2010-05-29 00:45:14 +04:00
|
|
|
uint8_t zp_copies;
|
2013-05-10 23:47:54 +04:00
|
|
|
boolean_t zp_dedup;
|
|
|
|
boolean_t zp_dedup_verify;
|
|
|
|
boolean_t zp_nopwrite;
|
Native Encryption for ZFS on Linux
This change incorporates three major pieces:
The first change is a keystore that manages wrapping
and encryption keys for encrypted datasets. These
commands mostly involve manipulating the new
DSL Crypto Key ZAP Objects that live in the MOS. Each
encrypted dataset has its own DSL Crypto Key that is
protected with a user's key. This level of indirection
allows users to change their keys without re-encrypting
their entire datasets. The change implements the new
subcommands "zfs load-key", "zfs unload-key" and
"zfs change-key" which allow the user to manage their
encryption keys and settings. In addition, several new
flags and properties have been added to allow dataset
creation and to make mounting and unmounting more
convenient.
The second piece of this patch provides the ability to
encrypt, decyrpt, and authenticate protected datasets.
Each object set maintains a Merkel tree of Message
Authentication Codes that protect the lower layers,
similarly to how checksums are maintained. This part
impacts the zio layer, which handles the actual
encryption and generation of MACs, as well as the ARC
and DMU, which need to be able to handle encrypted
buffers and protected data.
The last addition is the ability to do raw, encrypted
sends and receives. The idea here is to send raw
encrypted and compressed data and receive it exactly
as is on a backup system. This means that the dataset
on the receiving system is protected using the same
user key that is in use on the sending side. By doing
so, datasets can be efficiently backed up to an
untrusted system without fear of data being
compromised.
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #494
Closes #5769
2017-08-14 20:36:48 +03:00
|
|
|
boolean_t zp_encrypt;
|
|
|
|
boolean_t zp_byteorder;
|
|
|
|
uint8_t zp_salt[ZIO_DATA_SALT_LEN];
|
|
|
|
uint8_t zp_iv[ZIO_DATA_IV_LEN];
|
|
|
|
uint8_t zp_mac[ZIO_DATA_MAC_LEN];
|
2018-09-06 04:33:36 +03:00
|
|
|
uint32_t zp_zpl_smallblk;
|
2008-12-03 23:09:06 +03:00
|
|
|
} zio_prop_t;
|
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
typedef struct zio_cksum_report zio_cksum_report_t;
|
|
|
|
|
|
|
|
typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
|
2017-01-05 22:10:07 +03:00
|
|
|
const abd_t *good_data);
|
2010-05-29 00:45:14 +04:00
|
|
|
typedef void zio_cksum_free_f(void *cbdata, size_t size);
|
|
|
|
|
|
|
|
struct zio_bad_cksum; /* defined in zio_checksum.h */
|
2012-12-14 03:24:15 +04:00
|
|
|
struct dnode_phys;
|
2016-07-22 18:52:49 +03:00
|
|
|
struct abd;
|
2010-05-29 00:45:14 +04:00
|
|
|
|
|
|
|
struct zio_cksum_report {
|
|
|
|
struct zio_cksum_report *zcr_next;
|
|
|
|
nvlist_t *zcr_ereport;
|
|
|
|
nvlist_t *zcr_detector;
|
|
|
|
void *zcr_cbdata;
|
|
|
|
size_t zcr_cbinfo; /* passed to zcr_free() */
|
|
|
|
uint64_t zcr_align;
|
|
|
|
uint64_t zcr_length;
|
|
|
|
zio_cksum_finish_f *zcr_finish;
|
|
|
|
zio_cksum_free_f *zcr_free;
|
|
|
|
|
|
|
|
/* internal use only */
|
|
|
|
struct zio_bad_cksum *zcr_ckinfo; /* information from failure */
|
|
|
|
};
|
|
|
|
|
|
|
|
typedef void zio_vsd_cksum_report_f(zio_t *zio, zio_cksum_report_t *zcr,
|
|
|
|
void *arg);
|
|
|
|
|
|
|
|
zio_vsd_cksum_report_f zio_vsd_default_cksum_report;
|
|
|
|
|
|
|
|
typedef struct zio_vsd_ops {
|
|
|
|
zio_done_func_t *vsd_free;
|
|
|
|
zio_vsd_cksum_report_f *vsd_cksum_report;
|
|
|
|
} zio_vsd_ops_t;
|
|
|
|
|
2008-12-03 23:09:06 +03:00
|
|
|
typedef struct zio_gang_node {
|
|
|
|
zio_gbh_phys_t *gn_gbh;
|
|
|
|
struct zio_gang_node *gn_child[SPA_GBH_NBLKPTRS];
|
|
|
|
} zio_gang_node_t;
|
|
|
|
|
|
|
|
typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp,
|
2016-07-22 18:52:49 +03:00
|
|
|
zio_gang_node_t *gn, struct abd *data, uint64_t offset);
|
2008-12-03 23:09:06 +03:00
|
|
|
|
2016-07-22 18:52:49 +03:00
|
|
|
typedef void zio_transform_func_t(zio_t *zio, struct abd *data, uint64_t size);
|
2008-12-03 23:09:06 +03:00
|
|
|
|
|
|
|
typedef struct zio_transform {
|
2016-07-22 18:52:49 +03:00
|
|
|
struct abd *zt_orig_abd;
|
2008-12-03 23:09:06 +03:00
|
|
|
uint64_t zt_orig_size;
|
|
|
|
uint64_t zt_bufsize;
|
|
|
|
zio_transform_func_t *zt_transform;
|
|
|
|
struct zio_transform *zt_next;
|
|
|
|
} zio_transform_t;
|
|
|
|
|
Reduce taskq and context-switch cost of zio pipe
When doing a read from disk, ZFS creates 3 ZIO's: a zio_null(), the
logical zio_read(), and then a physical zio. Currently, each of these
results in a separate taskq_dispatch(zio_execute).
On high-read-iops workloads, this causes a significant performance
impact. By processing all 3 ZIO's in a single taskq entry, we reduce the
overhead on taskq locking and context switching. We accomplish this by
allowing zio_done() to return a "next zio to execute" to zio_execute().
This results in a ~12% performance increase for random reads, from
96,000 iops to 108,000 iops (with recordsize=8k, on SSD's).
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: George Wilson <george.wilson@delphix.com>
Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
External-issue: DLPX-59292
Closes #7736
2018-08-03 01:51:45 +03:00
|
|
|
typedef zio_t *zio_pipe_stage_t(zio_t *zio);
|
2008-12-03 23:09:06 +03:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The io_reexecute flags are distinct from io_flags because the child must
|
|
|
|
* be able to propagate them to the parent. The normal io_flags are local
|
|
|
|
* to the zio, not protected by any lock, and not modifiable by children;
|
|
|
|
* the reexecute flags are protected by io_lock, modifiable by children,
|
|
|
|
* and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
|
|
|
|
*/
|
|
|
|
#define ZIO_REEXECUTE_NOW 0x01
|
|
|
|
#define ZIO_REEXECUTE_SUSPEND 0x02
|
|
|
|
|
2017-01-12 22:52:56 +03:00
|
|
|
typedef struct zio_alloc_list {
|
|
|
|
list_t zal_list;
|
|
|
|
uint64_t zal_size;
|
|
|
|
} zio_alloc_list_t;
|
|
|
|
|
2009-02-18 23:51:31 +03:00
|
|
|
typedef struct zio_link {
|
|
|
|
zio_t *zl_parent;
|
|
|
|
zio_t *zl_child;
|
|
|
|
list_node_t zl_parent_node;
|
|
|
|
list_node_t zl_child_node;
|
|
|
|
} zio_link_t;
|
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
struct zio {
|
|
|
|
/* Core information about this I/O */
|
2014-06-25 22:37:59 +04:00
|
|
|
zbookmark_phys_t io_bookmark;
|
2008-12-03 23:09:06 +03:00
|
|
|
zio_prop_t io_prop;
|
|
|
|
zio_type_t io_type;
|
|
|
|
enum zio_child io_child_type;
|
|
|
|
int io_cmd;
|
Illumos #4045 write throttle & i/o scheduler performance work
4045 zfs write throttle & i/o scheduler performance work
1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver. The scheduler
issues a number of concurrent i/os from each class to the device. Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes). The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is. See the block comment in vdev_queue.c (reproduced
below) for more details.
2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load. The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system. When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount. This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens. One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync(). Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes. See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.
This diff has several other effects, including:
* the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.
* the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently. There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.
* zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc. This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).
--matt
APPENDIX: problems with the current i/o scheduler
The current ZFS i/o scheduler (vdev_queue.c) is deadline based. The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.
For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due". One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).
If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os. This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future. If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due. Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).
Notes on porting to ZFS on Linux:
- zio_t gained new members io_physdone and io_phys_children. Because
object caches in the Linux port call the constructor only once at
allocation time, objects may contain residual data when retrieved
from the cache. Therefore zio_create() was updated to zero out the two
new fields.
- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
(vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
This tree has been replaced by vq->vq_active_tree which is now used
for the same purpose.
- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
the number of vdev I/O buffers to pre-allocate. That global no longer
exists, so we instead use the sum of the *_max_active values for each of
the five I/O classes described above.
- The Illumos implementation of dmu_tx_delay() delays a transaction by
sleeping in condition variable embedded in the thread
(curthread->t_delay_cv). We do not have an equivalent CV to use in
Linux, so this change replaced the delay logic with a wrapper called
zfs_sleep_until(). This wrapper could be adopted upstream and in other
downstream ports to abstract away operating system-specific delay logic.
- These tunables are added as module parameters, and descriptions added
to the zfs-module-parameters.5 man page.
spa_asize_inflation
zfs_deadman_synctime_ms
zfs_vdev_max_active
zfs_vdev_async_write_active_min_dirty_percent
zfs_vdev_async_write_active_max_dirty_percent
zfs_vdev_async_read_max_active
zfs_vdev_async_read_min_active
zfs_vdev_async_write_max_active
zfs_vdev_async_write_min_active
zfs_vdev_scrub_max_active
zfs_vdev_scrub_min_active
zfs_vdev_sync_read_max_active
zfs_vdev_sync_read_min_active
zfs_vdev_sync_write_max_active
zfs_vdev_sync_write_min_active
zfs_dirty_data_max_percent
zfs_delay_min_dirty_percent
zfs_dirty_data_max_max_percent
zfs_dirty_data_max
zfs_dirty_data_max_max
zfs_dirty_data_sync
zfs_delay_scale
The latter four have type unsigned long, whereas they are uint64_t in
Illumos. This accommodates Linux's module_param() supported types, but
means they may overflow on 32-bit architectures.
The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
likely to overflow on 32-bit systems, since they express physical RAM
sizes in bytes. In fact, Illumos initializes zfs_dirty_data_max_max to
2^32 which does overflow. To resolve that, this port instead initializes
it in arc_init() to 25% of physical RAM, and adds the tunable
zfs_dirty_data_max_max_percent to override that percentage. While this
solution doesn't completely avoid the overflow issue, it should be a
reasonable default for most systems, and the minority of affected
systems can work around the issue by overriding the defaults.
- Fixed reversed logic in comment above zfs_delay_scale declaration.
- Clarified comments in vdev_queue.c regarding when per-queue minimums take
effect.
- Replaced dmu_tx_write_limit in the dmu_tx kstat file
with dmu_tx_dirty_delay and dmu_tx_dirty_over_max. The first counts
how many times a transaction has been delayed because the pool dirty
data has exceeded zfs_delay_min_dirty_percent. The latter counts how
many times the pool dirty data has exceeded zfs_dirty_data_max (which
we expect to never happen).
- The original patch would have regressed the bug fixed in
zfsonlinux/zfs@c418410, which prevented users from setting the
zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
A similar fix is added to vdev_queue_aggregate().
- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
heap instead of the stack. In Linux we can't afford such large
structures on the stack.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>
References:
http://www.illumos.org/issues/4045
illumos/illumos-gate@69962b5647e4a8b9b14998733b765925381b727e
Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-08-29 07:01:20 +04:00
|
|
|
zio_priority_t io_priority;
|
2008-12-03 23:09:06 +03:00
|
|
|
uint8_t io_reexecute;
|
2009-02-18 23:51:31 +03:00
|
|
|
uint8_t io_state[ZIO_WAIT_TYPES];
|
2008-11-20 23:01:55 +03:00
|
|
|
uint64_t io_txg;
|
2008-12-03 23:09:06 +03:00
|
|
|
spa_t *io_spa;
|
2008-11-20 23:01:55 +03:00
|
|
|
blkptr_t *io_bp;
|
2010-05-29 00:45:14 +04:00
|
|
|
blkptr_t *io_bp_override;
|
2008-11-20 23:01:55 +03:00
|
|
|
blkptr_t io_bp_copy;
|
2009-02-18 23:51:31 +03:00
|
|
|
list_t io_parent_list;
|
|
|
|
list_t io_child_list;
|
2008-11-20 23:01:55 +03:00
|
|
|
zio_t *io_logical;
|
2008-12-03 23:09:06 +03:00
|
|
|
zio_transform_t *io_transform_stack;
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
/* Callback info */
|
2016-05-15 18:02:28 +03:00
|
|
|
zio_done_func_t *io_ready;
|
|
|
|
zio_done_func_t *io_children_ready;
|
Illumos #4045 write throttle & i/o scheduler performance work
4045 zfs write throttle & i/o scheduler performance work
1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver. The scheduler
issues a number of concurrent i/os from each class to the device. Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes). The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is. See the block comment in vdev_queue.c (reproduced
below) for more details.
2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load. The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system. When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount. This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens. One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync(). Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes. See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.
This diff has several other effects, including:
* the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.
* the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently. There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.
* zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc. This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).
--matt
APPENDIX: problems with the current i/o scheduler
The current ZFS i/o scheduler (vdev_queue.c) is deadline based. The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.
For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due". One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).
If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os. This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future. If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due. Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).
Notes on porting to ZFS on Linux:
- zio_t gained new members io_physdone and io_phys_children. Because
object caches in the Linux port call the constructor only once at
allocation time, objects may contain residual data when retrieved
from the cache. Therefore zio_create() was updated to zero out the two
new fields.
- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
(vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
This tree has been replaced by vq->vq_active_tree which is now used
for the same purpose.
- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
the number of vdev I/O buffers to pre-allocate. That global no longer
exists, so we instead use the sum of the *_max_active values for each of
the five I/O classes described above.
- The Illumos implementation of dmu_tx_delay() delays a transaction by
sleeping in condition variable embedded in the thread
(curthread->t_delay_cv). We do not have an equivalent CV to use in
Linux, so this change replaced the delay logic with a wrapper called
zfs_sleep_until(). This wrapper could be adopted upstream and in other
downstream ports to abstract away operating system-specific delay logic.
- These tunables are added as module parameters, and descriptions added
to the zfs-module-parameters.5 man page.
spa_asize_inflation
zfs_deadman_synctime_ms
zfs_vdev_max_active
zfs_vdev_async_write_active_min_dirty_percent
zfs_vdev_async_write_active_max_dirty_percent
zfs_vdev_async_read_max_active
zfs_vdev_async_read_min_active
zfs_vdev_async_write_max_active
zfs_vdev_async_write_min_active
zfs_vdev_scrub_max_active
zfs_vdev_scrub_min_active
zfs_vdev_sync_read_max_active
zfs_vdev_sync_read_min_active
zfs_vdev_sync_write_max_active
zfs_vdev_sync_write_min_active
zfs_dirty_data_max_percent
zfs_delay_min_dirty_percent
zfs_dirty_data_max_max_percent
zfs_dirty_data_max
zfs_dirty_data_max_max
zfs_dirty_data_sync
zfs_delay_scale
The latter four have type unsigned long, whereas they are uint64_t in
Illumos. This accommodates Linux's module_param() supported types, but
means they may overflow on 32-bit architectures.
The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
likely to overflow on 32-bit systems, since they express physical RAM
sizes in bytes. In fact, Illumos initializes zfs_dirty_data_max_max to
2^32 which does overflow. To resolve that, this port instead initializes
it in arc_init() to 25% of physical RAM, and adds the tunable
zfs_dirty_data_max_max_percent to override that percentage. While this
solution doesn't completely avoid the overflow issue, it should be a
reasonable default for most systems, and the minority of affected
systems can work around the issue by overriding the defaults.
- Fixed reversed logic in comment above zfs_delay_scale declaration.
- Clarified comments in vdev_queue.c regarding when per-queue minimums take
effect.
- Replaced dmu_tx_write_limit in the dmu_tx kstat file
with dmu_tx_dirty_delay and dmu_tx_dirty_over_max. The first counts
how many times a transaction has been delayed because the pool dirty
data has exceeded zfs_delay_min_dirty_percent. The latter counts how
many times the pool dirty data has exceeded zfs_dirty_data_max (which
we expect to never happen).
- The original patch would have regressed the bug fixed in
zfsonlinux/zfs@c418410, which prevented users from setting the
zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
A similar fix is added to vdev_queue_aggregate().
- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
heap instead of the stack. In Linux we can't afford such large
structures on the stack.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>
References:
http://www.illumos.org/issues/4045
illumos/illumos-gate@69962b5647e4a8b9b14998733b765925381b727e
Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-08-29 07:01:20 +04:00
|
|
|
zio_done_func_t *io_physdone;
|
2008-11-20 23:01:55 +03:00
|
|
|
zio_done_func_t *io_done;
|
|
|
|
void *io_private;
|
2010-05-29 00:45:14 +04:00
|
|
|
int64_t io_prev_space_delta; /* DMU private */
|
2008-11-20 23:01:55 +03:00
|
|
|
blkptr_t io_bp_orig;
|
2016-07-11 20:45:52 +03:00
|
|
|
/* io_lsize != io_orig_size iff this is a raw write */
|
|
|
|
uint64_t io_lsize;
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
/* Data represented by this I/O */
|
2016-07-22 18:52:49 +03:00
|
|
|
struct abd *io_abd;
|
|
|
|
struct abd *io_orig_abd;
|
2008-11-20 23:01:55 +03:00
|
|
|
uint64_t io_size;
|
2010-05-29 00:45:14 +04:00
|
|
|
uint64_t io_orig_size;
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
/* Stuff for the vdev stack */
|
|
|
|
vdev_t *io_vd;
|
|
|
|
void *io_vsd;
|
2010-05-29 00:45:14 +04:00
|
|
|
const zio_vsd_ops_t *io_vsd_ops;
|
2018-09-06 04:33:36 +03:00
|
|
|
metaslab_class_t *io_metaslab_class; /* dva throttle class */
|
2010-05-29 00:45:14 +04:00
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
uint64_t io_offset;
|
2013-03-22 02:47:36 +04:00
|
|
|
hrtime_t io_timestamp; /* submitted at */
|
2016-10-14 03:59:18 +03:00
|
|
|
hrtime_t io_queued_timestamp;
|
2016-05-23 20:41:29 +03:00
|
|
|
hrtime_t io_target_timestamp;
|
2013-03-22 02:47:36 +04:00
|
|
|
hrtime_t io_delta; /* vdev queue service delta */
|
2016-02-29 21:05:23 +03:00
|
|
|
hrtime_t io_delay; /* Device access time (disk or */
|
|
|
|
/* file). */
|
Illumos #4045 write throttle & i/o scheduler performance work
4045 zfs write throttle & i/o scheduler performance work
1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver. The scheduler
issues a number of concurrent i/os from each class to the device. Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes). The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is. See the block comment in vdev_queue.c (reproduced
below) for more details.
2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load. The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system. When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount. This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens. One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync(). Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes. See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.
This diff has several other effects, including:
* the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.
* the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently. There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.
* zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc. This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).
--matt
APPENDIX: problems with the current i/o scheduler
The current ZFS i/o scheduler (vdev_queue.c) is deadline based. The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.
For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due". One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).
If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os. This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future. If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due. Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).
Notes on porting to ZFS on Linux:
- zio_t gained new members io_physdone and io_phys_children. Because
object caches in the Linux port call the constructor only once at
allocation time, objects may contain residual data when retrieved
from the cache. Therefore zio_create() was updated to zero out the two
new fields.
- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
(vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
This tree has been replaced by vq->vq_active_tree which is now used
for the same purpose.
- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
the number of vdev I/O buffers to pre-allocate. That global no longer
exists, so we instead use the sum of the *_max_active values for each of
the five I/O classes described above.
- The Illumos implementation of dmu_tx_delay() delays a transaction by
sleeping in condition variable embedded in the thread
(curthread->t_delay_cv). We do not have an equivalent CV to use in
Linux, so this change replaced the delay logic with a wrapper called
zfs_sleep_until(). This wrapper could be adopted upstream and in other
downstream ports to abstract away operating system-specific delay logic.
- These tunables are added as module parameters, and descriptions added
to the zfs-module-parameters.5 man page.
spa_asize_inflation
zfs_deadman_synctime_ms
zfs_vdev_max_active
zfs_vdev_async_write_active_min_dirty_percent
zfs_vdev_async_write_active_max_dirty_percent
zfs_vdev_async_read_max_active
zfs_vdev_async_read_min_active
zfs_vdev_async_write_max_active
zfs_vdev_async_write_min_active
zfs_vdev_scrub_max_active
zfs_vdev_scrub_min_active
zfs_vdev_sync_read_max_active
zfs_vdev_sync_read_min_active
zfs_vdev_sync_write_max_active
zfs_vdev_sync_write_min_active
zfs_dirty_data_max_percent
zfs_delay_min_dirty_percent
zfs_dirty_data_max_max_percent
zfs_dirty_data_max
zfs_dirty_data_max_max
zfs_dirty_data_sync
zfs_delay_scale
The latter four have type unsigned long, whereas they are uint64_t in
Illumos. This accommodates Linux's module_param() supported types, but
means they may overflow on 32-bit architectures.
The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
likely to overflow on 32-bit systems, since they express physical RAM
sizes in bytes. In fact, Illumos initializes zfs_dirty_data_max_max to
2^32 which does overflow. To resolve that, this port instead initializes
it in arc_init() to 25% of physical RAM, and adds the tunable
zfs_dirty_data_max_max_percent to override that percentage. While this
solution doesn't completely avoid the overflow issue, it should be a
reasonable default for most systems, and the minority of affected
systems can work around the issue by overriding the defaults.
- Fixed reversed logic in comment above zfs_delay_scale declaration.
- Clarified comments in vdev_queue.c regarding when per-queue minimums take
effect.
- Replaced dmu_tx_write_limit in the dmu_tx kstat file
with dmu_tx_dirty_delay and dmu_tx_dirty_over_max. The first counts
how many times a transaction has been delayed because the pool dirty
data has exceeded zfs_delay_min_dirty_percent. The latter counts how
many times the pool dirty data has exceeded zfs_dirty_data_max (which
we expect to never happen).
- The original patch would have regressed the bug fixed in
zfsonlinux/zfs@c418410, which prevented users from setting the
zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
A similar fix is added to vdev_queue_aggregate().
- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
heap instead of the stack. In Linux we can't afford such large
structures on the stack.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>
References:
http://www.illumos.org/issues/4045
illumos/illumos-gate@69962b5647e4a8b9b14998733b765925381b727e
Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-08-29 07:01:20 +04:00
|
|
|
avl_node_t io_queue_node;
|
2015-04-11 21:51:06 +03:00
|
|
|
avl_node_t io_offset_node;
|
2016-10-14 03:59:18 +03:00
|
|
|
avl_node_t io_alloc_node;
|
2017-01-12 22:52:56 +03:00
|
|
|
zio_alloc_list_t io_alloc_list;
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
/* Internal pipeline state */
|
2010-05-29 00:45:14 +04:00
|
|
|
enum zio_flag io_flags;
|
|
|
|
enum zio_stage io_stage;
|
|
|
|
enum zio_stage io_pipeline;
|
|
|
|
enum zio_flag io_orig_flags;
|
|
|
|
enum zio_stage io_orig_stage;
|
|
|
|
enum zio_stage io_orig_pipeline;
|
2016-10-14 03:59:18 +03:00
|
|
|
enum zio_stage io_pipeline_trace;
|
2008-12-03 23:09:06 +03:00
|
|
|
int io_error;
|
|
|
|
int io_child_error[ZIO_CHILD_TYPES];
|
|
|
|
uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
|
2010-05-29 00:45:14 +04:00
|
|
|
uint64_t io_child_count;
|
Illumos #4045 write throttle & i/o scheduler performance work
4045 zfs write throttle & i/o scheduler performance work
1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver. The scheduler
issues a number of concurrent i/os from each class to the device. Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes). The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is. See the block comment in vdev_queue.c (reproduced
below) for more details.
2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load. The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system. When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount. This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens. One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync(). Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes. See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.
This diff has several other effects, including:
* the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.
* the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently. There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.
* zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc. This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).
--matt
APPENDIX: problems with the current i/o scheduler
The current ZFS i/o scheduler (vdev_queue.c) is deadline based. The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.
For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due". One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).
If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os. This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future. If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due. Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).
Notes on porting to ZFS on Linux:
- zio_t gained new members io_physdone and io_phys_children. Because
object caches in the Linux port call the constructor only once at
allocation time, objects may contain residual data when retrieved
from the cache. Therefore zio_create() was updated to zero out the two
new fields.
- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
(vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
This tree has been replaced by vq->vq_active_tree which is now used
for the same purpose.
- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
the number of vdev I/O buffers to pre-allocate. That global no longer
exists, so we instead use the sum of the *_max_active values for each of
the five I/O classes described above.
- The Illumos implementation of dmu_tx_delay() delays a transaction by
sleeping in condition variable embedded in the thread
(curthread->t_delay_cv). We do not have an equivalent CV to use in
Linux, so this change replaced the delay logic with a wrapper called
zfs_sleep_until(). This wrapper could be adopted upstream and in other
downstream ports to abstract away operating system-specific delay logic.
- These tunables are added as module parameters, and descriptions added
to the zfs-module-parameters.5 man page.
spa_asize_inflation
zfs_deadman_synctime_ms
zfs_vdev_max_active
zfs_vdev_async_write_active_min_dirty_percent
zfs_vdev_async_write_active_max_dirty_percent
zfs_vdev_async_read_max_active
zfs_vdev_async_read_min_active
zfs_vdev_async_write_max_active
zfs_vdev_async_write_min_active
zfs_vdev_scrub_max_active
zfs_vdev_scrub_min_active
zfs_vdev_sync_read_max_active
zfs_vdev_sync_read_min_active
zfs_vdev_sync_write_max_active
zfs_vdev_sync_write_min_active
zfs_dirty_data_max_percent
zfs_delay_min_dirty_percent
zfs_dirty_data_max_max_percent
zfs_dirty_data_max
zfs_dirty_data_max_max
zfs_dirty_data_sync
zfs_delay_scale
The latter four have type unsigned long, whereas they are uint64_t in
Illumos. This accommodates Linux's module_param() supported types, but
means they may overflow on 32-bit architectures.
The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
likely to overflow on 32-bit systems, since they express physical RAM
sizes in bytes. In fact, Illumos initializes zfs_dirty_data_max_max to
2^32 which does overflow. To resolve that, this port instead initializes
it in arc_init() to 25% of physical RAM, and adds the tunable
zfs_dirty_data_max_max_percent to override that percentage. While this
solution doesn't completely avoid the overflow issue, it should be a
reasonable default for most systems, and the minority of affected
systems can work around the issue by overriding the defaults.
- Fixed reversed logic in comment above zfs_delay_scale declaration.
- Clarified comments in vdev_queue.c regarding when per-queue minimums take
effect.
- Replaced dmu_tx_write_limit in the dmu_tx kstat file
with dmu_tx_dirty_delay and dmu_tx_dirty_over_max. The first counts
how many times a transaction has been delayed because the pool dirty
data has exceeded zfs_delay_min_dirty_percent. The latter counts how
many times the pool dirty data has exceeded zfs_dirty_data_max (which
we expect to never happen).
- The original patch would have regressed the bug fixed in
zfsonlinux/zfs@c418410, which prevented users from setting the
zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
A similar fix is added to vdev_queue_aggregate().
- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
heap instead of the stack. In Linux we can't afford such large
structures on the stack.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>
References:
http://www.illumos.org/issues/4045
illumos/illumos-gate@69962b5647e4a8b9b14998733b765925381b727e
Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-08-29 07:01:20 +04:00
|
|
|
uint64_t io_phys_children;
|
2010-05-29 00:45:14 +04:00
|
|
|
uint64_t io_parent_count;
|
2008-12-03 23:09:06 +03:00
|
|
|
uint64_t *io_stall;
|
2009-07-03 02:44:48 +04:00
|
|
|
zio_t *io_gang_leader;
|
2008-12-03 23:09:06 +03:00
|
|
|
zio_gang_node_t *io_gang_tree;
|
|
|
|
void *io_executor;
|
2008-11-20 23:01:55 +03:00
|
|
|
void *io_waiter;
|
|
|
|
kmutex_t io_lock;
|
|
|
|
kcondvar_t io_cv;
|
OpenZFS 9112 - Improve allocation performance on high-end systems
Overview
========
We parallelize the allocation process by creating the concept of
"allocators". There are a certain number of allocators per metaslab
group, defined by the value of a tunable at pool open time. Each
allocator for a given metaslab group has up to 2 active metaslabs; one
"primary", and one "secondary". The primary and secondary weight mean
the same thing they did in in the pre-allocator world; primary metaslabs
are used for most allocations, secondary metaslabs are used for ditto
blocks being allocated in the same metaslab group. There is also the
CLAIM weight, which has been separated out from the other weights, but
that is less important to understanding the patch. The active metaslabs
for each allocator are moved from their normal place in the metaslab
tree for the group to the back of the tree. This way, they will not be
selected for use by other allocators searching for new metaslabs unless
all the passive metaslabs are unsuitable for allocations. If that does
happen, the allocators will "steal" from each other to ensure that IOs
don't fail until there is truly no space left to perform allocations.
In addition, the alloc queue for each metaslab group has been broken
into a separate queue for each allocator. We don't want to dramatically
increase the number of inflight IOs on low-end systems, because it can
significantly increase txg times. On the other hand, we want to ensure
that there are enough IOs for each allocator to allow for good
coalescing before sending the IOs to the disk. As a result, we take a
compromise path; each allocator's alloc queue max depth starts at a
certain value for every txg. Every time an IO completes, we increase the
max depth. This should hopefully provide a good balance between the two
failure modes, while not dramatically increasing complexity.
We also parallelize the spa_alloc_tree and spa_alloc_lock, which cause
very similar contention when selecting IOs to allocate. This
parallelization uses the same allocator scheme as metaslab selection.
Performance Results
===================
Performance improvements from this change can vary significantly based
on the number of CPUs in the system, whether or not the system has a
NUMA architecture, the speed of the drives, the values for the various
tunables, and the workload being performed. For an fio async sequential
write workload on a 24 core NUMA system with 256 GB of RAM and 8 128 GB
SSDs, there is a roughly 25% performance improvement.
Future Work
===========
Analysis of the performance of the system with this patch applied shows
that a significant new bottleneck is the vdev disk queues, which also
need to be parallelized. Prototyping of this change has occurred, and
there was a performance improvement, but more work needs to be done
before its stability has been verified and it is ready to be upstreamed.
Authored by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Serapheim Dimitropoulos <serapheim.dimitro@delphix.com>
Reviewed by: Alexander Motin <mav@FreeBSD.org>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Approved by: Gordon Ross <gwr@nexenta.com>
Ported-by: Paul Dagnelie <pcd@delphix.com>
Signed-off-by: Paul Dagnelie <pcd@delphix.com>
Porting Notes:
* Fix reservation test failures by increasing tolerance.
OpenZFS-issue: https://illumos.org/issues/9112
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/3f3cc3c3
Closes #7682
2018-02-12 23:56:06 +03:00
|
|
|
int io_allocator;
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
/* FMA state */
|
2010-05-29 00:45:14 +04:00
|
|
|
zio_cksum_report_t *io_cksum_report;
|
2008-11-20 23:01:55 +03:00
|
|
|
uint64_t io_ena;
|
2011-11-08 04:26:52 +04:00
|
|
|
|
|
|
|
/* Taskq dispatching state */
|
|
|
|
taskq_ent_t io_tqent;
|
2008-11-20 23:01:55 +03:00
|
|
|
};
|
|
|
|
|
2017-03-21 04:36:00 +03:00
|
|
|
extern int zio_bookmark_compare(const void *, const void *);
|
2016-10-14 03:59:18 +03:00
|
|
|
|
2009-02-18 23:51:31 +03:00
|
|
|
extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
|
2010-05-29 00:45:14 +04:00
|
|
|
zio_done_func_t *done, void *private, enum zio_flag flags);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
extern zio_t *zio_root(spa_t *spa,
|
2010-05-29 00:45:14 +04:00
|
|
|
zio_done_func_t *done, void *private, enum zio_flag flags);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2016-07-22 18:52:49 +03:00
|
|
|
extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
|
|
|
|
struct abd *data, uint64_t lsize, zio_done_func_t *done, void *private,
|
2014-06-25 22:37:59 +04:00
|
|
|
zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2008-12-03 23:09:06 +03:00
|
|
|
extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
|
2016-07-22 18:52:49 +03:00
|
|
|
struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp,
|
2016-05-15 18:02:28 +03:00
|
|
|
zio_done_func_t *ready, zio_done_func_t *children_ready,
|
|
|
|
zio_done_func_t *physdone, zio_done_func_t *done,
|
|
|
|
void *private, zio_priority_t priority, enum zio_flag flags,
|
|
|
|
const zbookmark_phys_t *zb);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2008-12-03 23:09:06 +03:00
|
|
|
extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
|
2016-07-22 18:52:49 +03:00
|
|
|
struct abd *data, uint64_t size, zio_done_func_t *done, void *private,
|
2014-06-25 22:37:59 +04:00
|
|
|
zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb);
|
2008-12-03 23:09:06 +03:00
|
|
|
|
2013-05-10 23:47:54 +04:00
|
|
|
extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
|
|
|
|
boolean_t nopwrite);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
|
|
|
|
const blkptr_t *bp,
|
|
|
|
zio_done_func_t *done, void *private, enum zio_flag flags);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
|
Illumos #4045 write throttle & i/o scheduler performance work
4045 zfs write throttle & i/o scheduler performance work
1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver. The scheduler
issues a number of concurrent i/os from each class to the device. Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes). The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is. See the block comment in vdev_queue.c (reproduced
below) for more details.
2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load. The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system. When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount. This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens. One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync(). Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes. See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.
This diff has several other effects, including:
* the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.
* the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently. There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.
* zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc. This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).
--matt
APPENDIX: problems with the current i/o scheduler
The current ZFS i/o scheduler (vdev_queue.c) is deadline based. The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.
For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due". One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).
If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os. This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future. If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due. Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).
Notes on porting to ZFS on Linux:
- zio_t gained new members io_physdone and io_phys_children. Because
object caches in the Linux port call the constructor only once at
allocation time, objects may contain residual data when retrieved
from the cache. Therefore zio_create() was updated to zero out the two
new fields.
- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
(vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
This tree has been replaced by vq->vq_active_tree which is now used
for the same purpose.
- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
the number of vdev I/O buffers to pre-allocate. That global no longer
exists, so we instead use the sum of the *_max_active values for each of
the five I/O classes described above.
- The Illumos implementation of dmu_tx_delay() delays a transaction by
sleeping in condition variable embedded in the thread
(curthread->t_delay_cv). We do not have an equivalent CV to use in
Linux, so this change replaced the delay logic with a wrapper called
zfs_sleep_until(). This wrapper could be adopted upstream and in other
downstream ports to abstract away operating system-specific delay logic.
- These tunables are added as module parameters, and descriptions added
to the zfs-module-parameters.5 man page.
spa_asize_inflation
zfs_deadman_synctime_ms
zfs_vdev_max_active
zfs_vdev_async_write_active_min_dirty_percent
zfs_vdev_async_write_active_max_dirty_percent
zfs_vdev_async_read_max_active
zfs_vdev_async_read_min_active
zfs_vdev_async_write_max_active
zfs_vdev_async_write_min_active
zfs_vdev_scrub_max_active
zfs_vdev_scrub_min_active
zfs_vdev_sync_read_max_active
zfs_vdev_sync_read_min_active
zfs_vdev_sync_write_max_active
zfs_vdev_sync_write_min_active
zfs_dirty_data_max_percent
zfs_delay_min_dirty_percent
zfs_dirty_data_max_max_percent
zfs_dirty_data_max
zfs_dirty_data_max_max
zfs_dirty_data_sync
zfs_delay_scale
The latter four have type unsigned long, whereas they are uint64_t in
Illumos. This accommodates Linux's module_param() supported types, but
means they may overflow on 32-bit architectures.
The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
likely to overflow on 32-bit systems, since they express physical RAM
sizes in bytes. In fact, Illumos initializes zfs_dirty_data_max_max to
2^32 which does overflow. To resolve that, this port instead initializes
it in arc_init() to 25% of physical RAM, and adds the tunable
zfs_dirty_data_max_max_percent to override that percentage. While this
solution doesn't completely avoid the overflow issue, it should be a
reasonable default for most systems, and the minority of affected
systems can work around the issue by overriding the defaults.
- Fixed reversed logic in comment above zfs_delay_scale declaration.
- Clarified comments in vdev_queue.c regarding when per-queue minimums take
effect.
- Replaced dmu_tx_write_limit in the dmu_tx kstat file
with dmu_tx_dirty_delay and dmu_tx_dirty_over_max. The first counts
how many times a transaction has been delayed because the pool dirty
data has exceeded zfs_delay_min_dirty_percent. The latter counts how
many times the pool dirty data has exceeded zfs_dirty_data_max (which
we expect to never happen).
- The original patch would have regressed the bug fixed in
zfsonlinux/zfs@c418410, which prevented users from setting the
zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
A similar fix is added to vdev_queue_aggregate().
- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
heap instead of the stack. In Linux we can't afford such large
structures on the stack.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>
References:
http://www.illumos.org/issues/4045
illumos/illumos-gate@69962b5647e4a8b9b14998733b765925381b727e
Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-08-29 07:01:20 +04:00
|
|
|
zio_done_func_t *done, void *private, enum zio_flag flags);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
|
2016-07-22 18:52:49 +03:00
|
|
|
uint64_t size, struct abd *data, int checksum,
|
Illumos #4045 write throttle & i/o scheduler performance work
4045 zfs write throttle & i/o scheduler performance work
1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver. The scheduler
issues a number of concurrent i/os from each class to the device. Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes). The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is. See the block comment in vdev_queue.c (reproduced
below) for more details.
2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load. The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system. When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount. This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens. One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync(). Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes. See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.
This diff has several other effects, including:
* the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.
* the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently. There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.
* zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc. This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).
--matt
APPENDIX: problems with the current i/o scheduler
The current ZFS i/o scheduler (vdev_queue.c) is deadline based. The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.
For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due". One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).
If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os. This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future. If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due. Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).
Notes on porting to ZFS on Linux:
- zio_t gained new members io_physdone and io_phys_children. Because
object caches in the Linux port call the constructor only once at
allocation time, objects may contain residual data when retrieved
from the cache. Therefore zio_create() was updated to zero out the two
new fields.
- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
(vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
This tree has been replaced by vq->vq_active_tree which is now used
for the same purpose.
- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
the number of vdev I/O buffers to pre-allocate. That global no longer
exists, so we instead use the sum of the *_max_active values for each of
the five I/O classes described above.
- The Illumos implementation of dmu_tx_delay() delays a transaction by
sleeping in condition variable embedded in the thread
(curthread->t_delay_cv). We do not have an equivalent CV to use in
Linux, so this change replaced the delay logic with a wrapper called
zfs_sleep_until(). This wrapper could be adopted upstream and in other
downstream ports to abstract away operating system-specific delay logic.
- These tunables are added as module parameters, and descriptions added
to the zfs-module-parameters.5 man page.
spa_asize_inflation
zfs_deadman_synctime_ms
zfs_vdev_max_active
zfs_vdev_async_write_active_min_dirty_percent
zfs_vdev_async_write_active_max_dirty_percent
zfs_vdev_async_read_max_active
zfs_vdev_async_read_min_active
zfs_vdev_async_write_max_active
zfs_vdev_async_write_min_active
zfs_vdev_scrub_max_active
zfs_vdev_scrub_min_active
zfs_vdev_sync_read_max_active
zfs_vdev_sync_read_min_active
zfs_vdev_sync_write_max_active
zfs_vdev_sync_write_min_active
zfs_dirty_data_max_percent
zfs_delay_min_dirty_percent
zfs_dirty_data_max_max_percent
zfs_dirty_data_max
zfs_dirty_data_max_max
zfs_dirty_data_sync
zfs_delay_scale
The latter four have type unsigned long, whereas they are uint64_t in
Illumos. This accommodates Linux's module_param() supported types, but
means they may overflow on 32-bit architectures.
The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
likely to overflow on 32-bit systems, since they express physical RAM
sizes in bytes. In fact, Illumos initializes zfs_dirty_data_max_max to
2^32 which does overflow. To resolve that, this port instead initializes
it in arc_init() to 25% of physical RAM, and adds the tunable
zfs_dirty_data_max_max_percent to override that percentage. While this
solution doesn't completely avoid the overflow issue, it should be a
reasonable default for most systems, and the minority of affected
systems can work around the issue by overriding the defaults.
- Fixed reversed logic in comment above zfs_delay_scale declaration.
- Clarified comments in vdev_queue.c regarding when per-queue minimums take
effect.
- Replaced dmu_tx_write_limit in the dmu_tx kstat file
with dmu_tx_dirty_delay and dmu_tx_dirty_over_max. The first counts
how many times a transaction has been delayed because the pool dirty
data has exceeded zfs_delay_min_dirty_percent. The latter counts how
many times the pool dirty data has exceeded zfs_dirty_data_max (which
we expect to never happen).
- The original patch would have regressed the bug fixed in
zfsonlinux/zfs@c418410, which prevented users from setting the
zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
A similar fix is added to vdev_queue_aggregate().
- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
heap instead of the stack. In Linux we can't afford such large
structures on the stack.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>
References:
http://www.illumos.org/issues/4045
illumos/illumos-gate@69962b5647e4a8b9b14998733b765925381b727e
Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-08-29 07:01:20 +04:00
|
|
|
zio_done_func_t *done, void *private, zio_priority_t priority,
|
|
|
|
enum zio_flag flags, boolean_t labels);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
|
2016-07-22 18:52:49 +03:00
|
|
|
uint64_t size, struct abd *data, int checksum,
|
Illumos #4045 write throttle & i/o scheduler performance work
4045 zfs write throttle & i/o scheduler performance work
1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver. The scheduler
issues a number of concurrent i/os from each class to the device. Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes). The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is. See the block comment in vdev_queue.c (reproduced
below) for more details.
2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load. The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system. When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount. This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens. One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync(). Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes. See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.
This diff has several other effects, including:
* the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.
* the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently. There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.
* zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc. This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).
--matt
APPENDIX: problems with the current i/o scheduler
The current ZFS i/o scheduler (vdev_queue.c) is deadline based. The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.
For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due". One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).
If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os. This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future. If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due. Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).
Notes on porting to ZFS on Linux:
- zio_t gained new members io_physdone and io_phys_children. Because
object caches in the Linux port call the constructor only once at
allocation time, objects may contain residual data when retrieved
from the cache. Therefore zio_create() was updated to zero out the two
new fields.
- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
(vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
This tree has been replaced by vq->vq_active_tree which is now used
for the same purpose.
- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
the number of vdev I/O buffers to pre-allocate. That global no longer
exists, so we instead use the sum of the *_max_active values for each of
the five I/O classes described above.
- The Illumos implementation of dmu_tx_delay() delays a transaction by
sleeping in condition variable embedded in the thread
(curthread->t_delay_cv). We do not have an equivalent CV to use in
Linux, so this change replaced the delay logic with a wrapper called
zfs_sleep_until(). This wrapper could be adopted upstream and in other
downstream ports to abstract away operating system-specific delay logic.
- These tunables are added as module parameters, and descriptions added
to the zfs-module-parameters.5 man page.
spa_asize_inflation
zfs_deadman_synctime_ms
zfs_vdev_max_active
zfs_vdev_async_write_active_min_dirty_percent
zfs_vdev_async_write_active_max_dirty_percent
zfs_vdev_async_read_max_active
zfs_vdev_async_read_min_active
zfs_vdev_async_write_max_active
zfs_vdev_async_write_min_active
zfs_vdev_scrub_max_active
zfs_vdev_scrub_min_active
zfs_vdev_sync_read_max_active
zfs_vdev_sync_read_min_active
zfs_vdev_sync_write_max_active
zfs_vdev_sync_write_min_active
zfs_dirty_data_max_percent
zfs_delay_min_dirty_percent
zfs_dirty_data_max_max_percent
zfs_dirty_data_max
zfs_dirty_data_max_max
zfs_dirty_data_sync
zfs_delay_scale
The latter four have type unsigned long, whereas they are uint64_t in
Illumos. This accommodates Linux's module_param() supported types, but
means they may overflow on 32-bit architectures.
The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
likely to overflow on 32-bit systems, since they express physical RAM
sizes in bytes. In fact, Illumos initializes zfs_dirty_data_max_max to
2^32 which does overflow. To resolve that, this port instead initializes
it in arc_init() to 25% of physical RAM, and adds the tunable
zfs_dirty_data_max_max_percent to override that percentage. While this
solution doesn't completely avoid the overflow issue, it should be a
reasonable default for most systems, and the minority of affected
systems can work around the issue by overriding the defaults.
- Fixed reversed logic in comment above zfs_delay_scale declaration.
- Clarified comments in vdev_queue.c regarding when per-queue minimums take
effect.
- Replaced dmu_tx_write_limit in the dmu_tx kstat file
with dmu_tx_dirty_delay and dmu_tx_dirty_over_max. The first counts
how many times a transaction has been delayed because the pool dirty
data has exceeded zfs_delay_min_dirty_percent. The latter counts how
many times the pool dirty data has exceeded zfs_dirty_data_max (which
we expect to never happen).
- The original patch would have regressed the bug fixed in
zfsonlinux/zfs@c418410, which prevented users from setting the
zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
A similar fix is added to vdev_queue_aggregate().
- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
heap instead of the stack. In Linux we can't afford such large
structures on the stack.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>
References:
http://www.illumos.org/issues/4045
illumos/illumos-gate@69962b5647e4a8b9b14998733b765925381b727e
Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-08-29 07:01:20 +04:00
|
|
|
zio_done_func_t *done, void *private, zio_priority_t priority,
|
|
|
|
enum zio_flag flags, boolean_t labels);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
|
|
|
|
const blkptr_t *bp, enum zio_flag flags);
|
|
|
|
|
Native Encryption for ZFS on Linux
This change incorporates three major pieces:
The first change is a keystore that manages wrapping
and encryption keys for encrypted datasets. These
commands mostly involve manipulating the new
DSL Crypto Key ZAP Objects that live in the MOS. Each
encrypted dataset has its own DSL Crypto Key that is
protected with a user's key. This level of indirection
allows users to change their keys without re-encrypting
their entire datasets. The change implements the new
subcommands "zfs load-key", "zfs unload-key" and
"zfs change-key" which allow the user to manage their
encryption keys and settings. In addition, several new
flags and properties have been added to allow dataset
creation and to make mounting and unmounting more
convenient.
The second piece of this patch provides the ability to
encrypt, decyrpt, and authenticate protected datasets.
Each object set maintains a Merkel tree of Message
Authentication Codes that protect the lower layers,
similarly to how checksums are maintained. This part
impacts the zio layer, which handles the actual
encryption and generation of MACs, as well as the ARC
and DMU, which need to be able to handle encrypted
buffers and protected data.
The last addition is the ability to do raw, encrypted
sends and receives. The idea here is to send raw
encrypted and compressed data and receive it exactly
as is on a backup system. This means that the dataset
on the receiving system is protected using the same
user key that is in use on the sending side. By doing
so, datasets can be efficiently backed up to an
untrusted system without fear of data being
compromised.
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #494
Closes #5769
2017-08-14 20:36:48 +03:00
|
|
|
extern int zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg,
|
|
|
|
blkptr_t *new_bp, uint64_t size, boolean_t *slog);
|
2008-11-20 23:01:55 +03:00
|
|
|
extern void zio_flush(zio_t *zio, vdev_t *vd);
|
2010-05-29 00:45:14 +04:00
|
|
|
extern void zio_shrink(zio_t *zio, uint64_t size);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
extern int zio_wait(zio_t *zio);
|
|
|
|
extern void zio_nowait(zio_t *zio);
|
|
|
|
extern void zio_execute(zio_t *zio);
|
|
|
|
extern void zio_interrupt(zio_t *zio);
|
2016-05-23 20:41:29 +03:00
|
|
|
extern void zio_delay_init(zio_t *zio);
|
|
|
|
extern void zio_delay_interrupt(zio_t *zio);
|
2017-12-19 01:06:07 +03:00
|
|
|
extern void zio_deadman(zio_t *zio, char *tag);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2016-10-14 03:59:18 +03:00
|
|
|
extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **);
|
|
|
|
extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **);
|
2009-02-18 23:51:31 +03:00
|
|
|
extern zio_t *zio_unique_parent(zio_t *cio);
|
|
|
|
extern void zio_add_child(zio_t *pio, zio_t *cio);
|
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
extern void *zio_buf_alloc(size_t size);
|
|
|
|
extern void zio_buf_free(void *buf, size_t size);
|
|
|
|
extern void *zio_data_buf_alloc(size_t size);
|
|
|
|
extern void zio_data_buf_free(void *buf, size_t size);
|
|
|
|
|
2016-07-22 18:52:49 +03:00
|
|
|
extern void zio_push_transform(zio_t *zio, struct abd *abd, uint64_t size,
|
2016-06-02 07:04:53 +03:00
|
|
|
uint64_t bufsize, zio_transform_func_t *transform);
|
|
|
|
extern void zio_pop_transforms(zio_t *zio);
|
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
extern void zio_resubmit_stage_async(void *);
|
|
|
|
|
|
|
|
extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
|
2016-07-22 18:52:49 +03:00
|
|
|
uint64_t offset, struct abd *data, uint64_t size, int type,
|
Illumos #4045 write throttle & i/o scheduler performance work
4045 zfs write throttle & i/o scheduler performance work
1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver. The scheduler
issues a number of concurrent i/os from each class to the device. Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes). The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is. See the block comment in vdev_queue.c (reproduced
below) for more details.
2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load. The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system. When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount. This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens. One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync(). Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes. See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.
This diff has several other effects, including:
* the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.
* the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently. There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.
* zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc. This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).
--matt
APPENDIX: problems with the current i/o scheduler
The current ZFS i/o scheduler (vdev_queue.c) is deadline based. The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.
For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due". One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).
If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os. This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future. If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due. Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).
Notes on porting to ZFS on Linux:
- zio_t gained new members io_physdone and io_phys_children. Because
object caches in the Linux port call the constructor only once at
allocation time, objects may contain residual data when retrieved
from the cache. Therefore zio_create() was updated to zero out the two
new fields.
- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
(vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
This tree has been replaced by vq->vq_active_tree which is now used
for the same purpose.
- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
the number of vdev I/O buffers to pre-allocate. That global no longer
exists, so we instead use the sum of the *_max_active values for each of
the five I/O classes described above.
- The Illumos implementation of dmu_tx_delay() delays a transaction by
sleeping in condition variable embedded in the thread
(curthread->t_delay_cv). We do not have an equivalent CV to use in
Linux, so this change replaced the delay logic with a wrapper called
zfs_sleep_until(). This wrapper could be adopted upstream and in other
downstream ports to abstract away operating system-specific delay logic.
- These tunables are added as module parameters, and descriptions added
to the zfs-module-parameters.5 man page.
spa_asize_inflation
zfs_deadman_synctime_ms
zfs_vdev_max_active
zfs_vdev_async_write_active_min_dirty_percent
zfs_vdev_async_write_active_max_dirty_percent
zfs_vdev_async_read_max_active
zfs_vdev_async_read_min_active
zfs_vdev_async_write_max_active
zfs_vdev_async_write_min_active
zfs_vdev_scrub_max_active
zfs_vdev_scrub_min_active
zfs_vdev_sync_read_max_active
zfs_vdev_sync_read_min_active
zfs_vdev_sync_write_max_active
zfs_vdev_sync_write_min_active
zfs_dirty_data_max_percent
zfs_delay_min_dirty_percent
zfs_dirty_data_max_max_percent
zfs_dirty_data_max
zfs_dirty_data_max_max
zfs_dirty_data_sync
zfs_delay_scale
The latter four have type unsigned long, whereas they are uint64_t in
Illumos. This accommodates Linux's module_param() supported types, but
means they may overflow on 32-bit architectures.
The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
likely to overflow on 32-bit systems, since they express physical RAM
sizes in bytes. In fact, Illumos initializes zfs_dirty_data_max_max to
2^32 which does overflow. To resolve that, this port instead initializes
it in arc_init() to 25% of physical RAM, and adds the tunable
zfs_dirty_data_max_max_percent to override that percentage. While this
solution doesn't completely avoid the overflow issue, it should be a
reasonable default for most systems, and the minority of affected
systems can work around the issue by overriding the defaults.
- Fixed reversed logic in comment above zfs_delay_scale declaration.
- Clarified comments in vdev_queue.c regarding when per-queue minimums take
effect.
- Replaced dmu_tx_write_limit in the dmu_tx kstat file
with dmu_tx_dirty_delay and dmu_tx_dirty_over_max. The first counts
how many times a transaction has been delayed because the pool dirty
data has exceeded zfs_delay_min_dirty_percent. The latter counts how
many times the pool dirty data has exceeded zfs_dirty_data_max (which
we expect to never happen).
- The original patch would have regressed the bug fixed in
zfsonlinux/zfs@c418410, which prevented users from setting the
zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
A similar fix is added to vdev_queue_aggregate().
- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
heap instead of the stack. In Linux we can't afford such large
structures on the stack.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>
References:
http://www.illumos.org/issues/4045
illumos/illumos-gate@69962b5647e4a8b9b14998733b765925381b727e
Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-08-29 07:01:20 +04:00
|
|
|
zio_priority_t priority, enum zio_flag flags,
|
|
|
|
zio_done_func_t *done, void *private);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2008-12-03 23:09:06 +03:00
|
|
|
extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset,
|
OpenZFS 9290 - device removal reduces redundancy of mirrors
Mirrors are supposed to provide redundancy in the face of whole-disk
failure and silent damage (e.g. some data on disk is not right, but ZFS
hasn't detected the whole device as being broken). However, the current
device removal implementation bypasses some of the mirror's redundancy.
Note that in no case is incorrect data returned, but we might get a
checksum error when we should have been able to find the right data.
There are two underlying problems:
1. When we remove a mirror device, we only read one side of the mirror.
Since we can't verify the checksum, this side may be silently bad, but
the good data is on the other side of the mirror (which we didn't read).
This can cause the removal to "bake in" the busted data – all copies of
the data in the new location are the same, busted version, while we left
the good version behind.
The fix for this is to read and copy both sides of the mirror. If the
old and new vdevs are mirrors, we will read both sides of the old
mirror, and write each copy to the corresponding side of the new mirror.
(If the old and new vdevs have a different number of children, we will
do this as best as possible.) Even though we aren't verifying checksums,
this ensures that as long as there's a good copy of the data, we'll have
a good copy after the removal, even if there's silent damage to one side
of the mirror. If we're removing a mirror that has some silent damage,
we'll have exactly the same damage in the new location (assuming that
the new location is also a mirror).
2. When we read from an indirect vdev that points to a mirror vdev, we
only consider one copy of the data. This can lead to reduced effective
redundancy, because we might read a bad copy of the data from one side
of the mirror, and not retry the other, good side of the mirror.
Note that the problem is not with the removal process, but rather after
the removal has completed (having copied correct data to both sides of
the mirror), if one side of the new mirror is silently damaged, we
encounter the problem when reading the relocated data via the indirect
vdev. Also note that the problem doesn't occur when ZFS knows that one
side of the mirror is bad, e.g. when a disk entirely fails or is
offlined.
The impact is that reads (from indirect vdevs that point to mirrors) may
return a checksum error even though the good data exists on one side of
the mirror, and scrub doesn't repair all data on the mirror (if some of
it is pointed to via an indirect vdev).
The fix for this is complicated by "split blocks" - one logical block
may be split into two (or more) pieces with each piece moved to a
different new location. In this case we need to read all versions of
each split (one from each side of the mirror), and figure out which
combination of versions results in the correct checksum, and then repair
the incorrect versions.
This ensures that we supply the same redundancy whether you use device
removal or not. For example, if a mirror has small silent errors on all
of its children, we can still reconstruct the correct data, as long as
those errors are at sufficiently-separated offsets (specifically,
separated by the largest block size - default of 128KB, but up to 16MB).
Porting notes:
* A new indirect vdev check was moved from dsl_scan_needs_resilver_cb()
to dsl_scan_needs_resilver(), which was added to ZoL as part of the
sequential scrub work.
* Passed NULL for zfs_ereport_post_checksum()'s zbookmark_phys_t
parameter. The extra parameter is unique to ZoL.
* When posting indirect checksum errors the ABD can be passed directly,
zfs_ereport_post_checksum() is not yet ABD-aware in OpenZFS.
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Tim Chase <tim@chase2k.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Ported-by: Tim Chase <tim@chase2k.com>
OpenZFS-issue: https://illumos.org/issues/9290
OpenZFS-commit: https://github.com/openzfs/openzfs/pull/591
Closes #6900
2018-02-13 22:37:56 +03:00
|
|
|
struct abd *data, uint64_t size, zio_type_t type, zio_priority_t priority,
|
2010-05-29 00:45:14 +04:00
|
|
|
enum zio_flag flags, zio_done_func_t *done, void *private);
|
2008-12-03 23:09:06 +03:00
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
extern void zio_vdev_io_bypass(zio_t *zio);
|
|
|
|
extern void zio_vdev_io_reissue(zio_t *zio);
|
|
|
|
extern void zio_vdev_io_redone(zio_t *zio);
|
|
|
|
|
2017-12-21 20:13:06 +03:00
|
|
|
extern void zio_change_priority(zio_t *pio, zio_priority_t priority);
|
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
extern void zio_checksum_verified(zio_t *zio);
|
2008-12-03 23:09:06 +03:00
|
|
|
extern int zio_worst_error(int e1, int e2);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2010-05-29 00:45:14 +04:00
|
|
|
extern enum zio_checksum zio_checksum_select(enum zio_checksum child,
|
|
|
|
enum zio_checksum parent);
|
|
|
|
extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa,
|
|
|
|
enum zio_checksum child, enum zio_checksum parent);
|
2015-07-06 04:55:32 +03:00
|
|
|
extern enum zio_compress zio_compress_select(spa_t *spa,
|
|
|
|
enum zio_compress child, enum zio_compress parent);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2018-03-15 20:56:55 +03:00
|
|
|
extern void zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t);
|
2009-07-03 02:44:48 +04:00
|
|
|
extern int zio_resume(spa_t *spa);
|
2008-12-03 23:09:06 +03:00
|
|
|
extern void zio_resume_wait(spa_t *spa);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Initial setup and teardown.
|
|
|
|
*/
|
|
|
|
extern void zio_init(void);
|
|
|
|
extern void zio_fini(void);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Fault injection
|
|
|
|
*/
|
|
|
|
struct zinject_record;
|
|
|
|
extern uint32_t zio_injection_enabled;
|
|
|
|
extern int zio_inject_fault(char *name, int flags, int *id,
|
|
|
|
struct zinject_record *record);
|
|
|
|
extern int zio_inject_list_next(int *id, char *name, size_t buflen,
|
|
|
|
struct zinject_record *record);
|
|
|
|
extern int zio_clear_fault(int id);
|
2010-05-29 00:45:14 +04:00
|
|
|
extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type);
|
2018-05-03 01:36:20 +03:00
|
|
|
extern int zio_handle_decrypt_injection(spa_t *spa, const zbookmark_phys_t *zb,
|
|
|
|
uint64_t type, int error);
|
2008-11-20 23:01:55 +03:00
|
|
|
extern int zio_handle_fault_injection(zio_t *zio, int error);
|
2009-07-03 02:44:48 +04:00
|
|
|
extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
|
2017-08-15 01:17:15 +03:00
|
|
|
extern int zio_handle_device_injections(vdev_t *vd, zio_t *zio, int err1,
|
|
|
|
int err2);
|
2008-12-03 23:09:06 +03:00
|
|
|
extern int zio_handle_label_injection(zio_t *zio, int error);
|
2010-05-29 00:45:14 +04:00
|
|
|
extern void zio_handle_ignored_writes(zio_t *zio);
|
2016-05-23 20:41:29 +03:00
|
|
|
extern hrtime_t zio_handle_io_delay(zio_t *zio);
|
2010-05-29 00:45:14 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Checksum ereport functions
|
|
|
|
*/
|
Native Encryption for ZFS on Linux
This change incorporates three major pieces:
The first change is a keystore that manages wrapping
and encryption keys for encrypted datasets. These
commands mostly involve manipulating the new
DSL Crypto Key ZAP Objects that live in the MOS. Each
encrypted dataset has its own DSL Crypto Key that is
protected with a user's key. This level of indirection
allows users to change their keys without re-encrypting
their entire datasets. The change implements the new
subcommands "zfs load-key", "zfs unload-key" and
"zfs change-key" which allow the user to manage their
encryption keys and settings. In addition, several new
flags and properties have been added to allow dataset
creation and to make mounting and unmounting more
convenient.
The second piece of this patch provides the ability to
encrypt, decyrpt, and authenticate protected datasets.
Each object set maintains a Merkel tree of Message
Authentication Codes that protect the lower layers,
similarly to how checksums are maintained. This part
impacts the zio layer, which handles the actual
encryption and generation of MACs, as well as the ARC
and DMU, which need to be able to handle encrypted
buffers and protected data.
The last addition is the ability to do raw, encrypted
sends and receives. The idea here is to send raw
encrypted and compressed data and receive it exactly
as is on a backup system. This means that the dataset
on the receiving system is protected using the same
user key that is in use on the sending side. By doing
so, datasets can be efficiently backed up to an
untrusted system without fear of data being
compromised.
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #494
Closes #5769
2017-08-14 20:36:48 +03:00
|
|
|
extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd,
|
2018-03-31 21:12:51 +03:00
|
|
|
const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset,
|
|
|
|
uint64_t length, void *arg, struct zio_bad_cksum *info);
|
2010-05-29 00:45:14 +04:00
|
|
|
extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
|
2017-01-05 22:10:07 +03:00
|
|
|
const abd_t *good_data, const abd_t *bad_data, boolean_t drop_if_identical);
|
2010-05-29 00:45:14 +04:00
|
|
|
|
|
|
|
extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
|
|
|
|
|
|
|
|
/* If we have the good data in hand, this function can be used */
|
2018-11-09 03:47:24 +03:00
|
|
|
extern int zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
|
2018-03-31 21:12:51 +03:00
|
|
|
const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset,
|
|
|
|
uint64_t length, const abd_t *good_data, const abd_t *bad_data,
|
|
|
|
struct zio_bad_cksum *info);
|
2010-05-29 00:45:14 +04:00
|
|
|
|
|
|
|
/* Called from spa_sync(), but primarily an injection handler */
|
|
|
|
extern void spa_handle_ignored_writes(spa_t *spa);
|
2008-11-20 23:01:55 +03:00
|
|
|
|
2014-06-25 22:37:59 +04:00
|
|
|
/* zbookmark_phys functions */
|
2015-12-22 04:31:57 +03:00
|
|
|
boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp,
|
|
|
|
const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
|
|
|
|
int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2,
|
|
|
|
uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);
|
2012-12-14 03:24:15 +04:00
|
|
|
|
2008-11-20 23:01:55 +03:00
|
|
|
#ifdef __cplusplus
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#endif /* _ZIO_H */
|