mirror_zfs/lib/libzfs/libzfs_pool.c

4441 lines
110 KiB
C
Raw Normal View History

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
*/
/*
* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
* Copyright (c) 2018 Datto Inc.
* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
* Copyright (c) 2017, Intel Corporation.
* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>
2008-11-20 23:01:55 +03:00
*/
#include <errno.h>
#include <libintl.h>
#include <stdio.h>
#include <stdlib.h>
#include <strings.h>
#include <unistd.h>
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
#include <libgen.h>
#include <zone.h>
#include <sys/stat.h>
2008-11-20 23:01:55 +03:00
#include <sys/efi_partition.h>
Add libtpool (thread pools) OpenZFS provides a library called tpool which implements thread pools for user space applications. Porting this library means the zpool utility no longer needs to borrow the kernel mutex and taskq interfaces from libzpool. This code was updated to use the tpool library which behaves in a very similar fashion. Porting libtpool was relatively straight forward and minimal modifications were needed. The core changes were: * Fully convert the library to use pthreads. * Updated signal handling. * lmalloc/lfree converted to calloc/free * Implemented portable pthread_attr_clone() function. Finally, update the build system such that libzpool.so is no longer linked in to zfs(8), zpool(8), etc. All that is required is libzfs to which the zcommon soures were added (which is the way it always should have been). Removing the libzpool dependency resulted in several build issues which needed to be resolved. * Moved zfeature support to module/zcommon/zfeature_common.c * Moved ratelimiting to to module/zfs/zfs_ratelimit.c * Moved get_system_hostid() to lib/libspl/gethostid.c * Removed use of cmn_err() in zcommon source * Removed dprintf_setup() call from zpool_main.c and zfs_main.c * Removed highbit() and lowbit() * Removed unnecessary library dependencies from Makefiles * Removed fletcher-4 kstat in user space * Added sha2 support explicitly to libzfs * Added highbit64() and lowbit64() to zpool_util.c Reviewed-by: Tony Hutter <hutter2@llnl.gov> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #6442
2017-08-10 01:31:08 +03:00
#include <sys/systeminfo.h>
2008-11-20 23:01:55 +03:00
#include <sys/zfs_ioctl.h>
#include <sys/vdev_disk.h>
2009-07-03 02:44:48 +04:00
#include <dlfcn.h>
#include <libzutil.h>
2008-11-20 23:01:55 +03:00
#include "zfs_namecheck.h"
#include "zfs_prop.h"
#include "libzfs_impl.h"
#include "zfs_comutil.h"
#include "zfeature_common.h"
2008-11-20 23:01:55 +03:00
static boolean_t zpool_vdev_is_interior(const char *name);
typedef struct prop_flags {
int create:1; /* Validate property on creation */
int import:1; /* Validate property on import */
} prop_flags_t;
2008-11-20 23:01:55 +03:00
/*
* ====================================================================
* zpool property functions
* ====================================================================
*/
static int
zpool_get_all_props(zpool_handle_t *zhp)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
return (-1);
while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
if (errno == ENOMEM) {
if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
zcmd_free_nvlists(&zc);
return (-1);
}
} else {
zcmd_free_nvlists(&zc);
return (-1);
}
}
if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
zcmd_free_nvlists(&zc);
return (-1);
}
zcmd_free_nvlists(&zc);
return (0);
}
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
int
2008-11-20 23:01:55 +03:00
zpool_props_refresh(zpool_handle_t *zhp)
{
nvlist_t *old_props;
old_props = zhp->zpool_props;
if (zpool_get_all_props(zhp) != 0)
return (-1);
nvlist_free(old_props);
return (0);
}
static const char *
2008-11-20 23:01:55 +03:00
zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
zprop_source_t *src)
{
nvlist_t *nv, *nvl;
uint64_t ival;
char *value;
zprop_source_t source;
nvl = zhp->zpool_props;
if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
source = ival;
verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
} else {
source = ZPROP_SRC_DEFAULT;
if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
value = "-";
}
if (src)
*src = source;
return (value);
}
uint64_t
zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
{
nvlist_t *nv, *nvl;
uint64_t value;
zprop_source_t source;
if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
/*
* zpool_get_all_props() has most likely failed because
* the pool is faulted, but if all we need is the top level
* vdev's guid then get it from the zhp config nvlist.
*/
if ((prop == ZPOOL_PROP_GUID) &&
(nvlist_lookup_nvlist(zhp->zpool_config,
ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
== 0)) {
return (value);
}
2008-11-20 23:01:55 +03:00
return (zpool_prop_default_numeric(prop));
}
2008-11-20 23:01:55 +03:00
nvl = zhp->zpool_props;
if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
source = value;
verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
} else {
source = ZPROP_SRC_DEFAULT;
value = zpool_prop_default_numeric(prop);
}
if (src)
*src = source;
return (value);
}
/*
* Map VDEV STATE to printed strings.
*/
const char *
2008-11-20 23:01:55 +03:00
zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
{
switch (state) {
case VDEV_STATE_CLOSED:
case VDEV_STATE_OFFLINE:
return (gettext("OFFLINE"));
case VDEV_STATE_REMOVED:
return (gettext("REMOVED"));
case VDEV_STATE_CANT_OPEN:
if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
2008-11-20 23:01:55 +03:00
return (gettext("FAULTED"));
else if (aux == VDEV_AUX_SPLIT_POOL)
return (gettext("SPLIT"));
2008-11-20 23:01:55 +03:00
else
return (gettext("UNAVAIL"));
case VDEV_STATE_FAULTED:
return (gettext("FAULTED"));
case VDEV_STATE_DEGRADED:
return (gettext("DEGRADED"));
case VDEV_STATE_HEALTHY:
return (gettext("ONLINE"));
default:
break;
2008-11-20 23:01:55 +03:00
}
return (gettext("UNKNOWN"));
}
/*
* Map POOL STATE to printed strings.
*/
const char *
zpool_pool_state_to_name(pool_state_t state)
{
switch (state) {
default:
break;
case POOL_STATE_ACTIVE:
return (gettext("ACTIVE"));
case POOL_STATE_EXPORTED:
return (gettext("EXPORTED"));
case POOL_STATE_DESTROYED:
return (gettext("DESTROYED"));
case POOL_STATE_SPARE:
return (gettext("SPARE"));
case POOL_STATE_L2CACHE:
return (gettext("L2CACHE"));
case POOL_STATE_UNINITIALIZED:
return (gettext("UNINITIALIZED"));
case POOL_STATE_UNAVAIL:
return (gettext("UNAVAIL"));
case POOL_STATE_POTENTIALLY_ACTIVE:
return (gettext("POTENTIALLY_ACTIVE"));
}
return (gettext("UNKNOWN"));
}
/*
* Given a pool handle, return the pool health string ("ONLINE", "DEGRADED",
* "SUSPENDED", etc).
*/
const char *
zpool_get_state_str(zpool_handle_t *zhp)
{
zpool_errata_t errata;
zpool_status_t status;
nvlist_t *nvroot;
vdev_stat_t *vs;
uint_t vsc;
const char *str;
status = zpool_get_status(zhp, NULL, &errata);
if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
str = gettext("FAULTED");
} else if (status == ZPOOL_STATUS_IO_FAILURE_WAIT ||
status == ZPOOL_STATUS_IO_FAILURE_MMP) {
str = gettext("SUSPENDED");
} else {
verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
verify(nvlist_lookup_uint64_array(nvroot,
ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
== 0);
str = zpool_state_to_name(vs->vs_state, vs->vs_aux);
}
return (str);
}
/*
* Get a zpool property value for 'prop' and return the value in
* a pre-allocated buffer.
*/
int
zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf,
size_t len, zprop_source_t *srctype, boolean_t literal)
2008-11-20 23:01:55 +03:00
{
uint64_t intval;
const char *strval;
zprop_source_t src = ZPROP_SRC_NONE;
if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
2009-02-18 23:51:31 +03:00
switch (prop) {
case ZPOOL_PROP_NAME:
2008-11-20 23:01:55 +03:00
(void) strlcpy(buf, zpool_get_name(zhp), len);
2009-02-18 23:51:31 +03:00
break;
case ZPOOL_PROP_HEALTH:
(void) strlcpy(buf, zpool_get_state_str(zhp), len);
2009-02-18 23:51:31 +03:00
break;
case ZPOOL_PROP_GUID:
intval = zpool_get_prop_int(zhp, prop, &src);
(void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
2009-02-18 23:51:31 +03:00
break;
case ZPOOL_PROP_ALTROOT:
case ZPOOL_PROP_CACHEFILE:
case ZPOOL_PROP_COMMENT:
2009-02-18 23:51:31 +03:00
if (zhp->zpool_props != NULL ||
zpool_get_all_props(zhp) == 0) {
(void) strlcpy(buf,
zpool_get_prop_string(zhp, prop, &src),
len);
break;
2009-02-18 23:51:31 +03:00
}
/* FALLTHROUGH */
default:
2008-11-20 23:01:55 +03:00
(void) strlcpy(buf, "-", len);
2009-02-18 23:51:31 +03:00
break;
}
if (srctype != NULL)
*srctype = src;
2008-11-20 23:01:55 +03:00
return (0);
}
if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
prop != ZPOOL_PROP_NAME)
return (-1);
switch (zpool_prop_get_type(prop)) {
case PROP_TYPE_STRING:
(void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
len);
break;
case PROP_TYPE_NUMBER:
intval = zpool_get_prop_int(zhp, prop, &src);
switch (prop) {
case ZPOOL_PROP_SIZE:
case ZPOOL_PROP_ALLOCATED:
case ZPOOL_PROP_FREE:
case ZPOOL_PROP_FREEING:
case ZPOOL_PROP_LEAKED:
case ZPOOL_PROP_ASHIFT:
if (literal)
(void) snprintf(buf, len, "%llu",
(u_longlong_t)intval);
else
(void) zfs_nicenum(intval, buf, len);
2008-11-20 23:01:55 +03:00
break;
case ZPOOL_PROP_EXPANDSZ:
OpenZFS 9166 - zfs storage pool checkpoint Details about the motivation of this feature and its usage can be found in this blogpost: https://sdimitro.github.io/post/zpool-checkpoint/ A lightning talk of this feature can be found here: https://www.youtube.com/watch?v=fPQA8K40jAM Implementation details can be found in big block comment of spa_checkpoint.c Side-changes that are relevant to this commit but not explained elsewhere: * renames members of "struct metaslab trees to be shorter without losing meaning * space_map_{alloc,truncate}() accept a block size as a parameter. The reason is that in the current state all space maps that we allocate through the DMU use a global tunable (space_map_blksz) which defauls to 4KB. This is ok for metaslab space maps in terms of bandwirdth since they are scattered all over the disk. But for other space maps this default is probably not what we want. Examples are device removal's vdev_obsolete_sm or vdev_chedkpoint_sm from this review. Both of these have a 1:1 relationship with each vdev and could benefit from a bigger block size. Porting notes: * The part of dsl_scan_sync() which handles async destroys has been moved into the new dsl_process_async_destroys() function. * Remove "VERIFY(!(flags & FWRITE))" in "kernel.c" so zhack can write to block device backed pools. * ZTS: * Fix get_txg() in zpool_sync_001_pos due to "checkpoint_txg". * Don't use large dd block sizes on /dev/urandom under Linux in checkpoint_capacity. * Adopt Delphix-OS's setting of 4 (spa_asize_inflation = SPA_DVAS_PER_BP + 1) for the checkpoint_capacity test to speed its attempts to fill the pool * Create the base and nested pools with sync=disabled to speed up the "setup" phase. * Clear labels in test pool between checkpoint tests to avoid duplicate pool issues. * The import_rewind_device_replaced test has been marked as "known to fail" for the reasons listed in its DISCLAIMER. * New module parameters: zfs_spa_discard_memory_limit, zfs_remove_max_bytes_pause (not documented - debugging only) vdev_max_ms_count (formerly metaslabs_per_vdev) vdev_min_ms_count Authored by: Serapheim Dimitropoulos <serapheim.dimitro@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: John Kennedy <john.kennedy@delphix.com> Reviewed by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://illumos.org/issues/9166 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/7159fdb8 Closes #7570
2016-12-17 01:11:29 +03:00
case ZPOOL_PROP_CHECKPOINT:
if (intval == 0) {
(void) strlcpy(buf, "-", len);
} else if (literal) {
(void) snprintf(buf, len, "%llu",
(u_longlong_t)intval);
} else {
(void) zfs_nicebytes(intval, buf, len);
}
break;
2008-11-20 23:01:55 +03:00
case ZPOOL_PROP_CAPACITY:
if (literal) {
(void) snprintf(buf, len, "%llu",
(u_longlong_t)intval);
} else {
(void) snprintf(buf, len, "%llu%%",
(u_longlong_t)intval);
}
2008-11-20 23:01:55 +03:00
break;
case ZPOOL_PROP_FRAGMENTATION:
if (intval == UINT64_MAX) {
(void) strlcpy(buf, "-", len);
} else if (literal) {
(void) snprintf(buf, len, "%llu",
(u_longlong_t)intval);
} else {
(void) snprintf(buf, len, "%llu%%",
(u_longlong_t)intval);
}
break;
case ZPOOL_PROP_DEDUPRATIO:
if (literal)
(void) snprintf(buf, len, "%llu.%02llu",
(u_longlong_t)(intval / 100),
(u_longlong_t)(intval % 100));
else
(void) snprintf(buf, len, "%llu.%02llux",
(u_longlong_t)(intval / 100),
(u_longlong_t)(intval % 100));
break;
2008-11-20 23:01:55 +03:00
case ZPOOL_PROP_HEALTH:
(void) strlcpy(buf, zpool_get_state_str(zhp), len);
2008-11-20 23:01:55 +03:00
break;
case ZPOOL_PROP_VERSION:
if (intval >= SPA_VERSION_FEATURES) {
(void) snprintf(buf, len, "-");
break;
}
/* FALLTHROUGH */
2008-11-20 23:01:55 +03:00
default:
(void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
2008-11-20 23:01:55 +03:00
}
break;
case PROP_TYPE_INDEX:
intval = zpool_get_prop_int(zhp, prop, &src);
if (zpool_prop_index_to_string(prop, intval, &strval)
!= 0)
return (-1);
(void) strlcpy(buf, strval, len);
break;
default:
abort();
}
if (srctype)
*srctype = src;
return (0);
}
/*
* Check if the bootfs name has the same pool name as it is set to.
* Assuming bootfs is a valid dataset name.
*/
static boolean_t
bootfs_name_valid(const char *pool, char *bootfs)
{
int len = strlen(pool);
if (bootfs[0] == '\0')
return (B_TRUE);
2008-11-20 23:01:55 +03:00
if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
2008-11-20 23:01:55 +03:00
return (B_FALSE);
if (strncmp(pool, bootfs, len) == 0 &&
(bootfs[len] == '/' || bootfs[len] == '\0'))
return (B_TRUE);
return (B_FALSE);
}
boolean_t
zpool_is_bootable(zpool_handle_t *zhp)
{
char bootfs[ZFS_MAX_DATASET_NAME_LEN];
return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-",
sizeof (bootfs)) != 0);
}
2008-11-20 23:01:55 +03:00
/*
* Given an nvlist of zpool properties to be set, validate that they are
* correct, and parse any numeric properties (index, boolean, etc) if they are
* specified as strings.
*/
static nvlist_t *
zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
2008-11-20 23:01:55 +03:00
{
nvpair_t *elem;
nvlist_t *retprops;
zpool_prop_t prop;
char *strval;
uint64_t intval;
char *slash, *check;
2008-11-20 23:01:55 +03:00
struct stat64 statbuf;
zpool_handle_t *zhp;
2008-11-20 23:01:55 +03:00
if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
(void) no_memory(hdl);
return (NULL);
}
elem = NULL;
while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
const char *propname = nvpair_name(elem);
prop = zpool_name_to_prop(propname);
if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) {
int err;
char *fname = strchr(propname, '@') + 1;
err = zfeature_lookup_name(fname, NULL);
if (err != 0) {
ASSERT3U(err, ==, ENOENT);
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"invalid feature '%s'"), fname);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
if (nvpair_type(elem) != DATA_TYPE_STRING) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"'%s' must be a string"), propname);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
(void) nvpair_value_string(elem, &strval);
if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0 &&
strcmp(strval, ZFS_FEATURE_DISABLED) != 0) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s' can only be set to "
"'enabled' or 'disabled'"), propname);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
if (!flags.create &&
strcmp(strval, ZFS_FEATURE_DISABLED) == 0) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s' can only be set to "
"'disabled' at creation time"), propname);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
if (nvlist_add_uint64(retprops, propname, 0) != 0) {
(void) no_memory(hdl);
goto error;
}
continue;
}
2008-11-20 23:01:55 +03:00
/*
* Make sure this property is valid and applies to this type.
*/
if (prop == ZPOOL_PROP_INVAL) {
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"invalid property '%s'"), propname);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
if (zpool_prop_readonly(prop)) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
"is readonly"), propname);
(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
goto error;
}
if (!flags.create && zpool_prop_setonce(prop)) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s' can only be set at "
"creation time"), propname);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
2008-11-20 23:01:55 +03:00
if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
&strval, &intval, errbuf) != 0)
goto error;
/*
* Perform additional checking for specific properties.
*/
switch (prop) {
case ZPOOL_PROP_VERSION:
if (intval < version ||
!SPA_VERSION_IS_SUPPORTED(intval)) {
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s' number %d is invalid."),
propname, intval);
(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
goto error;
}
break;
case ZPOOL_PROP_ASHIFT:
if (intval != 0 &&
(intval < ASHIFT_MIN || intval > ASHIFT_MAX)) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s' number %d is invalid, only "
"values between %" PRId32 " and "
"%" PRId32 " are allowed."),
propname, intval, ASHIFT_MIN, ASHIFT_MAX);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
break;
2008-11-20 23:01:55 +03:00
case ZPOOL_PROP_BOOTFS:
if (flags.create || flags.import) {
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s' cannot be set at creation "
"or import time"), propname);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
if (version < SPA_VERSION_BOOTFS) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"pool must be upgraded to support "
"'%s' property"), propname);
(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
goto error;
}
/*
* bootfs property value has to be a dataset name and
* the dataset has to be in the same pool as it sets to.
*/
if (!bootfs_name_valid(poolname, strval)) {
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
"is an invalid name"), strval);
(void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
goto error;
}
if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"could not open pool '%s'"), poolname);
(void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
goto error;
}
zpool_close(zhp);
2008-11-20 23:01:55 +03:00
break;
case ZPOOL_PROP_ALTROOT:
if (!flags.create && !flags.import) {
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s' can only be set during pool "
"creation or import"), propname);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
if (strval[0] != '/') {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"bad alternate root '%s'"), strval);
(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
goto error;
}
break;
case ZPOOL_PROP_CACHEFILE:
if (strval[0] == '\0')
break;
if (strcmp(strval, "none") == 0)
break;
if (strval[0] != '/') {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s' must be empty, an "
"absolute path, or 'none'"), propname);
(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
goto error;
}
slash = strrchr(strval, '/');
if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
strcmp(slash, "/..") == 0) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"'%s' is not a valid file"), strval);
(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
goto error;
}
*slash = '\0';
if (strval[0] != '\0' &&
(stat64(strval, &statbuf) != 0 ||
!S_ISDIR(statbuf.st_mode))) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"'%s' is not a valid directory"),
strval);
(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
goto error;
}
*slash = '/';
break;
case ZPOOL_PROP_COMMENT:
for (check = strval; *check != '\0'; check++) {
if (!isprint(*check)) {
zfs_error_aux(hdl,
dgettext(TEXT_DOMAIN,
"comment may only have printable "
"characters"));
(void) zfs_error(hdl, EZFS_BADPROP,
errbuf);
goto error;
}
}
if (strlen(strval) > ZPROP_MAX_COMMENT) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"comment must not exceed %d characters"),
ZPROP_MAX_COMMENT);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
break;
case ZPOOL_PROP_READONLY:
if (!flags.import) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s' can only be set at "
"import time"), propname);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
break;
Multi-modifier protection (MMP) Add multihost=on|off pool property to control MMP. When enabled a new thread writes uberblocks to the last slot in each label, at a set frequency, to indicate to other hosts the pool is actively imported. These uberblocks are the last synced uberblock with an updated timestamp. Property defaults to off. During tryimport, find the "best" uberblock (newest txg and timestamp) repeatedly, checking for change in the found uberblock. Include the results of the activity test in the config returned by tryimport. These results are reported to user in "zpool import". Allow the user to control the period between MMP writes, and the duration of the activity test on import, via a new module parameter zfs_multihost_interval. The period is specified in milliseconds. The activity test duration is calculated from this value, and from the mmp_delay in the "best" uberblock found initially. Add a kstat interface to export statistics about Multiple Modifier Protection (MMP) updates. Include the last synced txg number, the timestamp, the delay since the last MMP update, the VDEV GUID, the VDEV label that received the last MMP update, and the VDEV path. Abbreviated output below. $ cat /proc/spl/kstat/zfs/mypool/multihost 31 0 0x01 10 880 105092382393521 105144180101111 txg timestamp mmp_delay vdev_guid vdev_label vdev_path 20468 261337 250274925 68396651780 3 /dev/sda 20468 261339 252023374 6267402363293 1 /dev/sdc 20468 261340 252000858 6698080955233 1 /dev/sdx 20468 261341 251980635 783892869810 2 /dev/sdy 20468 261342 253385953 8923255792467 3 /dev/sdd 20468 261344 253336622 042125143176 0 /dev/sdab 20468 261345 253310522 1200778101278 2 /dev/sde 20468 261346 253286429 0950576198362 2 /dev/sdt 20468 261347 253261545 96209817917 3 /dev/sds 20468 261349 253238188 8555725937673 3 /dev/sdb Add a new tunable zfs_multihost_history to specify the number of MMP updates to store history for. By default it is set to zero meaning that no MMP statistics are stored. When using ztest to generate activity, for automated tests of the MMP function, some test functions interfere with the test. For example, the pool is exported to run zdb and then imported again. Add a new ztest function, "-M", to alter ztest behavior to prevent this. Add new tests to verify the new functionality. Tests provided by Giuseppe Di Natale. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Ned Bass <bass6@llnl.gov> Reviewed-by: Andreas Dilger <andreas.dilger@intel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Olaf Faaland <faaland1@llnl.gov> Closes #745 Closes #6279
2017-07-08 06:20:35 +03:00
case ZPOOL_PROP_MULTIHOST:
if (get_system_hostid() == 0) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"requires a non-zero system hostid"));
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
break;
case ZPOOL_PROP_DEDUPDITTO:
Remove dedupditto functionality If dedup is in use, the `dedupditto` property can be set, causing ZFS to keep an extra copy of data that is referenced many times (>100x). The idea was that this data is more important than other data and thus we want to be really sure that it is not lost if the disk experiences a small amount of random corruption. ZFS (and system administrators) rely on the pool-level redundancy to protect their data (e.g. mirroring or RAIDZ). Since the user/sysadmin doesn't have control over what data will be offered extra redundancy by dedupditto, this extra redundancy is not very useful. The bulk of the data is still vulnerable to loss based on the pool-level redundancy. For example, if particle strikes corrupt 0.1% of blocks, you will either be saved by mirror/raidz, or you will be sad. This is true even if dedupditto saved another 0.01% of blocks from being corrupted. Therefore, the dedupditto functionality is rarely enabled (i.e. the property is rarely set), and it fulfills its promise of increased redundancy even more rarely. Additionally, this feature does not work as advertised (on existing releases), because scrub/resilver did not repair the extra (dedupditto) copy (see https://github.com/zfsonlinux/zfs/pull/8270). In summary, this seldom-used feature doesn't work, and even if it did it wouldn't provide useful data protection. It has a non-trivial maintenance burden (again see https://github.com/zfsonlinux/zfs/pull/8270). We should remove the dedupditto functionality. For backwards compatibility with the existing CLI, "zpool set dedupditto" will still "succeed" (exit code zero), but won't have any effect. For backwards compatibility with existing pools that had dedupditto enabled at some point, the code will still be able to understand dedupditto blocks and free them when appropriate. However, ZFS won't write any new dedupditto blocks. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Igor Kozhukhov <igor@dilos.org> Reviewed-by: Alek Pinchuk <apinchuk@datto.com> Issue #8270 Closes #8310
2019-06-20 00:54:02 +03:00
printf("Note: property '%s' no longer has "
"any effect\n", propname);
break;
default:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property '%s'(%d) not defined"), propname, prop);
break;
2008-11-20 23:01:55 +03:00
}
}
return (retprops);
error:
nvlist_free(retprops);
return (NULL);
}
/*
* Set zpool property : propname=propval.
*/
int
zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
int ret = -1;
char errbuf[1024];
nvlist_t *nvl = NULL;
nvlist_t *realprops;
uint64_t version;
prop_flags_t flags = { 0 };
2008-11-20 23:01:55 +03:00
(void) snprintf(errbuf, sizeof (errbuf),
dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
zhp->zpool_name);
if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
return (no_memory(zhp->zpool_hdl));
if (nvlist_add_string(nvl, propname, propval) != 0) {
nvlist_free(nvl);
return (no_memory(zhp->zpool_hdl));
}
version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
2008-11-20 23:01:55 +03:00
nvlist_free(nvl);
return (-1);
}
nvlist_free(nvl);
nvl = realprops;
/*
* Execute the corresponding ioctl() to set this property.
*/
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
nvlist_free(nvl);
return (-1);
}
ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
zcmd_free_nvlists(&zc);
nvlist_free(nvl);
if (ret)
(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
else
(void) zpool_props_refresh(zhp);
return (ret);
}
int
zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
{
libzfs_handle_t *hdl = zhp->zpool_hdl;
zprop_list_t *entry;
char buf[ZFS_MAXPROPLEN];
nvlist_t *features = NULL;
nvpair_t *nvp;
zprop_list_t **last;
boolean_t firstexpand = (NULL == *plp);
int i;
2008-11-20 23:01:55 +03:00
if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
return (-1);
last = plp;
while (*last != NULL)
last = &(*last)->pl_next;
if ((*plp)->pl_all)
features = zpool_get_features(zhp);
if ((*plp)->pl_all && firstexpand) {
for (i = 0; i < SPA_FEATURES; i++) {
zprop_list_t *entry = zfs_alloc(hdl,
sizeof (zprop_list_t));
entry->pl_prop = ZPROP_INVAL;
entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
spa_feature_table[i].fi_uname);
entry->pl_width = strlen(entry->pl_user_prop);
entry->pl_all = B_TRUE;
*last = entry;
last = &entry->pl_next;
}
}
/* add any unsupported features */
for (nvp = nvlist_next_nvpair(features, NULL);
nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
char *propname;
boolean_t found;
zprop_list_t *entry;
if (zfeature_is_supported(nvpair_name(nvp)))
continue;
propname = zfs_asprintf(hdl, "unsupported@%s",
nvpair_name(nvp));
/*
* Before adding the property to the list make sure that no
* other pool already added the same property.
*/
found = B_FALSE;
entry = *plp;
while (entry != NULL) {
if (entry->pl_user_prop != NULL &&
strcmp(propname, entry->pl_user_prop) == 0) {
found = B_TRUE;
break;
}
entry = entry->pl_next;
}
if (found) {
free(propname);
continue;
}
entry = zfs_alloc(hdl, sizeof (zprop_list_t));
entry->pl_prop = ZPROP_INVAL;
entry->pl_user_prop = propname;
entry->pl_width = strlen(entry->pl_user_prop);
entry->pl_all = B_TRUE;
*last = entry;
last = &entry->pl_next;
}
2008-11-20 23:01:55 +03:00
for (entry = *plp; entry != NULL; entry = entry->pl_next) {
if (entry->pl_fixed)
continue;
if (entry->pl_prop != ZPROP_INVAL &&
zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
NULL, B_FALSE) == 0) {
2008-11-20 23:01:55 +03:00
if (strlen(buf) > entry->pl_width)
entry->pl_width = strlen(buf);
}
}
return (0);
}
/*
* Get the state for the given feature on the given ZFS pool.
*/
int
zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
size_t len)
{
uint64_t refcount;
boolean_t found = B_FALSE;
nvlist_t *features = zpool_get_features(zhp);
boolean_t supported;
const char *feature = strchr(propname, '@') + 1;
supported = zpool_prop_feature(propname);
ASSERT(supported || zpool_prop_unsupported(propname));
/*
* Convert from feature name to feature guid. This conversion is
* unnecessary for unsupported@... properties because they already
* use guids.
*/
if (supported) {
int ret;
spa_feature_t fid;
ret = zfeature_lookup_name(feature, &fid);
if (ret != 0) {
(void) strlcpy(buf, "-", len);
return (ENOTSUP);
}
feature = spa_feature_table[fid].fi_guid;
}
if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
found = B_TRUE;
if (supported) {
if (!found) {
(void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
} else {
if (refcount == 0)
(void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
else
(void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
}
} else {
if (found) {
if (refcount == 0) {
(void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
} else {
(void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
}
} else {
(void) strlcpy(buf, "-", len);
return (ENOTSUP);
}
}
return (0);
}
2008-11-20 23:01:55 +03:00
/*
* Validate the given pool name, optionally putting an extended error message in
* 'buf'.
*/
boolean_t
zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
{
namecheck_err_t why;
char what;
int ret;
ret = pool_namecheck(pool, &why, &what);
/*
* The rules for reserved pool names were extended at a later point.
* But we need to support users with existing pools that may now be
* invalid. So we only check for this expanded set of names during a
* create (or import), and only in userland.
*/
if (ret == 0 && !isopen &&
(strncmp(pool, "mirror", 6) == 0 ||
strncmp(pool, "raidz", 5) == 0 ||
strncmp(pool, "spare", 5) == 0 ||
strcmp(pool, "log") == 0)) {
if (hdl != NULL)
zfs_error_aux(hdl,
dgettext(TEXT_DOMAIN, "name is reserved"));
return (B_FALSE);
}
if (ret != 0) {
if (hdl != NULL) {
switch (why) {
case NAME_ERR_TOOLONG:
zfs_error_aux(hdl,
dgettext(TEXT_DOMAIN, "name is too long"));
break;
case NAME_ERR_INVALCHAR:
zfs_error_aux(hdl,
dgettext(TEXT_DOMAIN, "invalid character "
"'%c' in pool name"), what);
break;
case NAME_ERR_NOLETTER:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"name must begin with a letter"));
break;
case NAME_ERR_RESERVED:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"name is reserved"));
break;
case NAME_ERR_DISKLIKE:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"pool name is reserved"));
break;
case NAME_ERR_LEADING_SLASH:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"leading slash in name"));
break;
case NAME_ERR_EMPTY_COMPONENT:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"empty component in name"));
break;
case NAME_ERR_TRAILING_SLASH:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"trailing slash in name"));
break;
case NAME_ERR_MULTIPLE_DELIMITERS:
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"multiple '@' and/or '#' delimiters in "
"name"));
2008-11-20 23:01:55 +03:00
break;
case NAME_ERR_NO_AT:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"permission set is missing '@'"));
break;
default:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"(%d) not defined"), why);
break;
2008-11-20 23:01:55 +03:00
}
}
return (B_FALSE);
}
return (B_TRUE);
}
/*
* Open a handle to the given pool, even if the pool is currently in the FAULTED
* state.
*/
zpool_handle_t *
zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
{
zpool_handle_t *zhp;
boolean_t missing;
/*
* Make sure the pool name is valid.
*/
if (!zpool_name_valid(hdl, B_TRUE, pool)) {
(void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
dgettext(TEXT_DOMAIN, "cannot open '%s'"),
pool);
return (NULL);
}
if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
return (NULL);
zhp->zpool_hdl = hdl;
(void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
if (zpool_refresh_stats(zhp, &missing) != 0) {
zpool_close(zhp);
return (NULL);
}
if (missing) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
(void) zfs_error_fmt(hdl, EZFS_NOENT,
dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
zpool_close(zhp);
return (NULL);
}
return (zhp);
}
/*
* Like the above, but silent on error. Used when iterating over pools (because
* the configuration cache may be out of date).
*/
int
zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
{
zpool_handle_t *zhp;
boolean_t missing;
if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
return (-1);
zhp->zpool_hdl = hdl;
(void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
if (zpool_refresh_stats(zhp, &missing) != 0) {
zpool_close(zhp);
return (-1);
}
if (missing) {
zpool_close(zhp);
*ret = NULL;
return (0);
}
*ret = zhp;
return (0);
}
/*
* Similar to zpool_open_canfail(), but refuses to open pools in the faulted
* state.
*/
zpool_handle_t *
zpool_open(libzfs_handle_t *hdl, const char *pool)
{
zpool_handle_t *zhp;
if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
return (NULL);
if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
(void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
zpool_close(zhp);
return (NULL);
}
return (zhp);
}
/*
* Close the handle. Simply frees the memory associated with the handle.
*/
void
zpool_close(zpool_handle_t *zhp)
{
nvlist_free(zhp->zpool_config);
nvlist_free(zhp->zpool_old_config);
nvlist_free(zhp->zpool_props);
2008-11-20 23:01:55 +03:00
free(zhp);
}
/*
* Return the name of the pool.
*/
const char *
zpool_get_name(zpool_handle_t *zhp)
{
return (zhp->zpool_name);
}
/*
* Return the state of the pool (ACTIVE or UNAVAILABLE)
*/
int
zpool_get_state(zpool_handle_t *zhp)
{
return (zhp->zpool_state);
}
/*
* Check if vdev list contains a special vdev
*/
static boolean_t
zpool_has_special_vdev(nvlist_t *nvroot)
{
nvlist_t **child;
uint_t children;
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child,
&children) == 0) {
for (uint_t c = 0; c < children; c++) {
char *bias;
if (nvlist_lookup_string(child[c],
ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0 &&
strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0) {
return (B_TRUE);
}
}
}
return (B_FALSE);
}
2008-11-20 23:01:55 +03:00
/*
* Create the named pool, using the provided vdev list. It is assumed
* that the consumer has already validated the contents of the nvlist, so we
* don't have to worry about error semantics.
*/
int
zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
nvlist_t *props, nvlist_t *fsprops)
2008-11-20 23:01:55 +03:00
{
zfs_cmd_t zc = {"\0"};
nvlist_t *zc_fsprops = NULL;
nvlist_t *zc_props = NULL;
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
nvlist_t *hidden_args = NULL;
uint8_t *wkeydata = NULL;
uint_t wkeylen = 0;
2008-11-20 23:01:55 +03:00
char msg[1024];
int ret = -1;
2008-11-20 23:01:55 +03:00
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot create '%s'"), pool);
if (!zpool_name_valid(hdl, B_FALSE, pool))
return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
return (-1);
if (props) {
prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
if ((zc_props = zpool_valid_proplist(hdl, pool, props,
SPA_VERSION_1, flags, msg)) == NULL) {
goto create_failed;
}
}
2008-11-20 23:01:55 +03:00
if (fsprops) {
uint64_t zoned;
char *zonestr;
zoned = ((nvlist_lookup_string(fsprops,
zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
strcmp(zonestr, "on") == 0);
if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
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
fsprops, zoned, NULL, NULL, B_TRUE, msg)) == NULL) {
goto create_failed;
}
if (nvlist_exists(zc_fsprops,
zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)) &&
!zpool_has_special_vdev(nvroot)) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"%s property requires a special vdev"),
zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS));
(void) zfs_error(hdl, EZFS_BADPROP, msg);
goto create_failed;
}
if (!zc_props &&
(nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
goto create_failed;
}
if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE,
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
&wkeydata, &wkeylen) != 0) {
zfs_error(hdl, EZFS_CRYPTOFAILED, msg);
goto create_failed;
}
if (nvlist_add_nvlist(zc_props,
ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
goto create_failed;
}
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
if (wkeydata != NULL) {
if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0)
goto create_failed;
if (nvlist_add_uint8_array(hidden_args, "wkeydata",
wkeydata, wkeylen) != 0)
goto create_failed;
if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS,
hidden_args) != 0)
goto create_failed;
}
2008-11-20 23:01:55 +03:00
}
if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
goto create_failed;
2008-11-20 23:01:55 +03:00
(void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
2008-11-20 23:01:55 +03:00
zcmd_free_nvlists(&zc);
nvlist_free(zc_props);
nvlist_free(zc_fsprops);
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
nvlist_free(hidden_args);
if (wkeydata != NULL)
free(wkeydata);
2008-11-20 23:01:55 +03:00
switch (errno) {
case EBUSY:
/*
* This can happen if the user has specified the same
* device multiple times. We can't reliably detect this
* until we try to add it and see we already have a
* label. This can also happen under if the device is
* part of an active md or lvm device.
2008-11-20 23:01:55 +03:00
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"one or more vdevs refer to the same device, or "
"one of\nthe devices is part of an active md or "
"lvm device"));
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_BADDEV, msg));
case ERANGE:
/*
* This happens if the record size is smaller or larger
* than the allowed size range, or not a power of 2.
*
* NOTE: although zfs_valid_proplist is called earlier,
* this case may have slipped through since the
* pool does not exist yet and it is therefore
* impossible to read properties e.g. max blocksize
* from the pool.
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"record size invalid"));
return (zfs_error(hdl, EZFS_BADPROP, msg));
2008-11-20 23:01:55 +03:00
case EOVERFLOW:
/*
* This occurs when one of the devices is below
* SPA_MINDEVSIZE. Unfortunately, we can't detect which
* device was the problem device since there's no
* reliable way to determine device size from userland.
*/
{
char buf[64];
zfs_nicebytes(SPA_MINDEVSIZE, buf,
sizeof (buf));
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"one or more devices is less than the "
"minimum size (%s)"), buf);
}
return (zfs_error(hdl, EZFS_BADDEV, msg));
case ENOSPC:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"one or more devices is out of space"));
return (zfs_error(hdl, EZFS_BADDEV, msg));
case ENOTBLK:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"cache device must be a disk or disk slice"));
return (zfs_error(hdl, EZFS_BADDEV, msg));
default:
return (zpool_standard_error(hdl, errno, msg));
}
}
create_failed:
2008-11-20 23:01:55 +03:00
zcmd_free_nvlists(&zc);
nvlist_free(zc_props);
nvlist_free(zc_fsprops);
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
nvlist_free(hidden_args);
if (wkeydata != NULL)
free(wkeydata);
return (ret);
2008-11-20 23:01:55 +03:00
}
/*
* Destroy the given pool. It is up to the caller to ensure that there are no
* datasets left in the pool.
*/
int
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
zpool_destroy(zpool_handle_t *zhp, const char *log_str)
2008-11-20 23:01:55 +03:00
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
zfs_handle_t *zfp = NULL;
libzfs_handle_t *hdl = zhp->zpool_hdl;
char msg[1024];
if (zhp->zpool_state == POOL_STATE_ACTIVE &&
(zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
2008-11-20 23:01:55 +03:00
return (-1);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
zc.zc_history = (uint64_t)(uintptr_t)log_str;
2008-11-20 23:01:55 +03:00
if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
2008-11-20 23:01:55 +03:00
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot destroy '%s'"), zhp->zpool_name);
if (errno == EROFS) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"one or more devices is read only"));
(void) zfs_error(hdl, EZFS_BADDEV, msg);
} else {
(void) zpool_standard_error(hdl, errno, msg);
}
if (zfp)
zfs_close(zfp);
return (-1);
}
if (zfp) {
remove_mountpoint(zfp);
zfs_close(zfp);
}
return (0);
}
OpenZFS 9166 - zfs storage pool checkpoint Details about the motivation of this feature and its usage can be found in this blogpost: https://sdimitro.github.io/post/zpool-checkpoint/ A lightning talk of this feature can be found here: https://www.youtube.com/watch?v=fPQA8K40jAM Implementation details can be found in big block comment of spa_checkpoint.c Side-changes that are relevant to this commit but not explained elsewhere: * renames members of "struct metaslab trees to be shorter without losing meaning * space_map_{alloc,truncate}() accept a block size as a parameter. The reason is that in the current state all space maps that we allocate through the DMU use a global tunable (space_map_blksz) which defauls to 4KB. This is ok for metaslab space maps in terms of bandwirdth since they are scattered all over the disk. But for other space maps this default is probably not what we want. Examples are device removal's vdev_obsolete_sm or vdev_chedkpoint_sm from this review. Both of these have a 1:1 relationship with each vdev and could benefit from a bigger block size. Porting notes: * The part of dsl_scan_sync() which handles async destroys has been moved into the new dsl_process_async_destroys() function. * Remove "VERIFY(!(flags & FWRITE))" in "kernel.c" so zhack can write to block device backed pools. * ZTS: * Fix get_txg() in zpool_sync_001_pos due to "checkpoint_txg". * Don't use large dd block sizes on /dev/urandom under Linux in checkpoint_capacity. * Adopt Delphix-OS's setting of 4 (spa_asize_inflation = SPA_DVAS_PER_BP + 1) for the checkpoint_capacity test to speed its attempts to fill the pool * Create the base and nested pools with sync=disabled to speed up the "setup" phase. * Clear labels in test pool between checkpoint tests to avoid duplicate pool issues. * The import_rewind_device_replaced test has been marked as "known to fail" for the reasons listed in its DISCLAIMER. * New module parameters: zfs_spa_discard_memory_limit, zfs_remove_max_bytes_pause (not documented - debugging only) vdev_max_ms_count (formerly metaslabs_per_vdev) vdev_min_ms_count Authored by: Serapheim Dimitropoulos <serapheim.dimitro@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: John Kennedy <john.kennedy@delphix.com> Reviewed by: Dan Kimmel <dan.kimmel@delphix.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://illumos.org/issues/9166 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/7159fdb8 Closes #7570
2016-12-17 01:11:29 +03:00
/*
* Create a checkpoint in the given pool.
*/
int
zpool_checkpoint(zpool_handle_t *zhp)
{
libzfs_handle_t *hdl = zhp->zpool_hdl;
char msg[1024];
int error;
error = lzc_pool_checkpoint(zhp->zpool_name);
if (error != 0) {
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot checkpoint '%s'"), zhp->zpool_name);
(void) zpool_standard_error(hdl, error, msg);
return (-1);
}
return (0);
}
/*
* Discard the checkpoint from the given pool.
*/
int
zpool_discard_checkpoint(zpool_handle_t *zhp)
{
libzfs_handle_t *hdl = zhp->zpool_hdl;
char msg[1024];
int error;
error = lzc_pool_checkpoint_discard(zhp->zpool_name);
if (error != 0) {
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot discard checkpoint in '%s'"), zhp->zpool_name);
(void) zpool_standard_error(hdl, error, msg);
return (-1);
}
return (0);
}
2008-11-20 23:01:55 +03:00
/*
* Add the given vdevs to the pool. The caller must have already performed the
* necessary verification to ensure that the vdev specification is well-formed.
*/
int
zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
int ret;
libzfs_handle_t *hdl = zhp->zpool_hdl;
char msg[1024];
nvlist_t **spares, **l2cache;
uint_t nspares, nl2cache;
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot add to '%s'"), zhp->zpool_name);
if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
SPA_VERSION_SPARES &&
nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
&spares, &nspares) == 0) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
"upgraded to add hot spares"));
return (zfs_error(hdl, EZFS_BADVERSION, msg));
}
if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
SPA_VERSION_L2CACHE &&
nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
&l2cache, &nl2cache) == 0) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
"upgraded to add cache devices"));
return (zfs_error(hdl, EZFS_BADVERSION, msg));
}
if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
return (-1);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
2008-11-20 23:01:55 +03:00
switch (errno) {
case EBUSY:
/*
* This can happen if the user has specified the same
* device multiple times. We can't reliably detect this
* until we try to add it and see we already have a
* label.
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"one or more vdevs refer to the same device"));
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
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
case EINVAL:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"invalid config; a pool with removing/removed "
"vdevs does not support adding raidz vdevs"));
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
2008-11-20 23:01:55 +03:00
case EOVERFLOW:
/*
* This occurs when one of the devices is below
2008-11-20 23:01:55 +03:00
* SPA_MINDEVSIZE. Unfortunately, we can't detect which
* device was the problem device since there's no
* reliable way to determine device size from userland.
*/
{
char buf[64];
zfs_nicebytes(SPA_MINDEVSIZE, buf,
sizeof (buf));
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"device is less than the minimum "
"size (%s)"), buf);
}
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
case ENOTSUP:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"pool must be upgraded to add these vdevs"));
(void) zfs_error(hdl, EZFS_BADVERSION, msg);
break;
case ENOTBLK:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"cache device must be a disk or disk slice"));
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
default:
(void) zpool_standard_error(hdl, errno, msg);
}
ret = -1;
} else {
ret = 0;
}
zcmd_free_nvlists(&zc);
return (ret);
}
/*
* Exports the pool from the system. The caller must ensure that there are no
* mounted datasets in the pool.
*/
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
static int
zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
const char *log_str)
2008-11-20 23:01:55 +03:00
{
zfs_cmd_t zc = {"\0"};
char msg[1024];
2008-11-20 23:01:55 +03:00
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot export '%s'"), zhp->zpool_name);
2008-11-20 23:01:55 +03:00
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
zc.zc_cookie = force;
2009-01-16 00:59:39 +03:00
zc.zc_guid = hardforce;
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
zc.zc_history = (uint64_t)(uintptr_t)log_str;
if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
switch (errno) {
case EXDEV:
zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
"use '-f' to override the following errors:\n"
"'%s' has an active shared spare which could be"
" used by other pools once '%s' is exported."),
zhp->zpool_name, zhp->zpool_name);
return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
msg));
default:
return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
msg));
}
}
2008-11-20 23:01:55 +03:00
return (0);
}
2009-01-16 00:59:39 +03:00
int
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
2009-01-16 00:59:39 +03:00
{
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
return (zpool_export_common(zhp, force, B_FALSE, log_str));
2009-01-16 00:59:39 +03:00
}
int
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
zpool_export_force(zpool_handle_t *zhp, const char *log_str)
2009-01-16 00:59:39 +03:00
{
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
2009-01-16 00:59:39 +03:00
}
static void
zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
nvlist_t *config)
{
nvlist_t *nv = NULL;
uint64_t rewindto;
int64_t loss = -1;
struct tm t;
char timestr[128];
if (!hdl->libzfs_printerr || config == NULL)
return;
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
return;
}
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
return;
(void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
if (localtime_r((time_t *)&rewindto, &t) != NULL &&
strftime(timestr, 128, "%c", &t) != 0) {
if (dryrun) {
(void) printf(dgettext(TEXT_DOMAIN,
"Would be able to return %s "
"to its state as of %s.\n"),
name, timestr);
} else {
(void) printf(dgettext(TEXT_DOMAIN,
"Pool %s returned to its state as of %s.\n"),
name, timestr);
}
if (loss > 120) {
(void) printf(dgettext(TEXT_DOMAIN,
"%s approximately %lld "),
dryrun ? "Would discard" : "Discarded",
((longlong_t)loss + 30) / 60);
(void) printf(dgettext(TEXT_DOMAIN,
"minutes of transactions.\n"));
} else if (loss > 0) {
(void) printf(dgettext(TEXT_DOMAIN,
"%s approximately %lld "),
dryrun ? "Would discard" : "Discarded",
(longlong_t)loss);
(void) printf(dgettext(TEXT_DOMAIN,
"seconds of transactions.\n"));
}
}
}
void
zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
nvlist_t *config)
{
nvlist_t *nv = NULL;
int64_t loss = -1;
uint64_t edata = UINT64_MAX;
uint64_t rewindto;
struct tm t;
char timestr[128];
if (!hdl->libzfs_printerr)
return;
if (reason >= 0)
(void) printf(dgettext(TEXT_DOMAIN, "action: "));
else
(void) printf(dgettext(TEXT_DOMAIN, "\t"));
/* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
goto no_info;
(void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
&edata);
(void) printf(dgettext(TEXT_DOMAIN,
"Recovery is possible, but will result in some data loss.\n"));
if (localtime_r((time_t *)&rewindto, &t) != NULL &&
strftime(timestr, 128, "%c", &t) != 0) {
(void) printf(dgettext(TEXT_DOMAIN,
"\tReturning the pool to its state as of %s\n"
"\tshould correct the problem. "),
timestr);
} else {
(void) printf(dgettext(TEXT_DOMAIN,
"\tReverting the pool to an earlier state "
"should correct the problem.\n\t"));
}
if (loss > 120) {
(void) printf(dgettext(TEXT_DOMAIN,
"Approximately %lld minutes of data\n"
"\tmust be discarded, irreversibly. "),
((longlong_t)loss + 30) / 60);
} else if (loss > 0) {
(void) printf(dgettext(TEXT_DOMAIN,
"Approximately %lld seconds of data\n"
"\tmust be discarded, irreversibly. "),
(longlong_t)loss);
}
if (edata != 0 && edata != UINT64_MAX) {
if (edata == 1) {
(void) printf(dgettext(TEXT_DOMAIN,
"After rewind, at least\n"
"\tone persistent user-data error will remain. "));
} else {
(void) printf(dgettext(TEXT_DOMAIN,
"After rewind, several\n"
"\tpersistent user-data errors will remain. "));
}
}
(void) printf(dgettext(TEXT_DOMAIN,
"Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
reason >= 0 ? "clear" : "import", name);
(void) printf(dgettext(TEXT_DOMAIN,
"A scrub of the pool\n"
"\tis strongly recommended after recovery.\n"));
return;
no_info:
(void) printf(dgettext(TEXT_DOMAIN,
"Destroy and re-create the pool from\n\ta backup source.\n"));
}
2008-11-20 23:01:55 +03:00
/*
* zpool_import() is a contracted interface. Should be kept the same
* if possible.
*
* Applications should use zpool_import_props() to import a pool with
* new properties value to be set.
*/
int
zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
char *altroot)
{
nvlist_t *props = NULL;
int ret;
if (altroot != NULL) {
if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
return (zfs_error_fmt(hdl, EZFS_NOMEM,
dgettext(TEXT_DOMAIN, "cannot import '%s'"),
newname));
}
if (nvlist_add_string(props,
2009-01-16 00:59:39 +03:00
zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
nvlist_add_string(props,
zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
2008-11-20 23:01:55 +03:00
nvlist_free(props);
return (zfs_error_fmt(hdl, EZFS_NOMEM,
dgettext(TEXT_DOMAIN, "cannot import '%s'"),
newname));
}
}
ret = zpool_import_props(hdl, config, newname, props,
ZFS_IMPORT_NORMAL);
nvlist_free(props);
2008-11-20 23:01:55 +03:00
return (ret);
}
static void
print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
int indent)
{
nvlist_t **child;
uint_t c, children;
char *vname;
uint64_t is_log = 0;
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
&is_log);
if (name != NULL)
(void) printf("\t%*s%s%s\n", indent, "", name,
is_log ? " [log]" : "");
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0)
return;
for (c = 0; c < children; c++) {
vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID);
print_vdev_tree(hdl, vname, child[c], indent + 2);
free(vname);
}
}
void
zpool_print_unsup_feat(nvlist_t *config)
{
nvlist_t *nvinfo, *unsup_feat;
nvpair_t *nvp;
verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
0);
verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
&unsup_feat) == 0);
for (nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
char *desc;
verify(nvpair_type(nvp) == DATA_TYPE_STRING);
verify(nvpair_value_string(nvp, &desc) == 0);
if (strlen(desc) > 0)
(void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
else
(void) printf("\t%s\n", nvpair_name(nvp));
}
}
2008-11-20 23:01:55 +03:00
/*
* Import the given pool using the known configuration and a list of
* properties to be set. The configuration should have come from
* zpool_find_import(). The 'newname' parameters control whether the pool
* is imported with a different name.
*/
int
zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
nvlist_t *props, int flags)
2008-11-20 23:01:55 +03:00
{
zfs_cmd_t zc = {"\0"};
zpool_load_policy_t policy;
nvlist_t *nv = NULL;
nvlist_t *nvinfo = NULL;
nvlist_t *missing = NULL;
2008-11-20 23:01:55 +03:00
char *thename;
char *origname;
int ret;
int error = 0;
2008-11-20 23:01:55 +03:00
char errbuf[1024];
verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
&origname) == 0);
(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
"cannot import pool '%s'"), origname);
if (newname != NULL) {
if (!zpool_name_valid(hdl, B_FALSE, newname))
return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
dgettext(TEXT_DOMAIN, "cannot import '%s'"),
newname));
thename = (char *)newname;
} else {
thename = origname;
}
if (props != NULL) {
2008-11-20 23:01:55 +03:00
uint64_t version;
prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2008-11-20 23:01:55 +03:00
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
&version) == 0);
if ((props = zpool_valid_proplist(hdl, origname,
props, version, flags, errbuf)) == NULL)
2008-11-20 23:01:55 +03:00
return (-1);
if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
2008-11-20 23:01:55 +03:00
nvlist_free(props);
return (-1);
}
nvlist_free(props);
2008-11-20 23:01:55 +03:00
}
(void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
&zc.zc_guid) == 0);
if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
zcmd_free_nvlists(&zc);
2008-11-20 23:01:55 +03:00
return (-1);
}
if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
zcmd_free_nvlists(&zc);
return (-1);
}
2008-11-20 23:01:55 +03:00
zc.zc_cookie = flags;
while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
errno == ENOMEM) {
if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
zcmd_free_nvlists(&zc);
return (-1);
}
}
if (ret != 0)
error = errno;
(void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
zcmd_free_nvlists(&zc);
zpool_get_load_policy(config, &policy);
if (error) {
2008-11-20 23:01:55 +03:00
char desc[1024];
Multi-modifier protection (MMP) Add multihost=on|off pool property to control MMP. When enabled a new thread writes uberblocks to the last slot in each label, at a set frequency, to indicate to other hosts the pool is actively imported. These uberblocks are the last synced uberblock with an updated timestamp. Property defaults to off. During tryimport, find the "best" uberblock (newest txg and timestamp) repeatedly, checking for change in the found uberblock. Include the results of the activity test in the config returned by tryimport. These results are reported to user in "zpool import". Allow the user to control the period between MMP writes, and the duration of the activity test on import, via a new module parameter zfs_multihost_interval. The period is specified in milliseconds. The activity test duration is calculated from this value, and from the mmp_delay in the "best" uberblock found initially. Add a kstat interface to export statistics about Multiple Modifier Protection (MMP) updates. Include the last synced txg number, the timestamp, the delay since the last MMP update, the VDEV GUID, the VDEV label that received the last MMP update, and the VDEV path. Abbreviated output below. $ cat /proc/spl/kstat/zfs/mypool/multihost 31 0 0x01 10 880 105092382393521 105144180101111 txg timestamp mmp_delay vdev_guid vdev_label vdev_path 20468 261337 250274925 68396651780 3 /dev/sda 20468 261339 252023374 6267402363293 1 /dev/sdc 20468 261340 252000858 6698080955233 1 /dev/sdx 20468 261341 251980635 783892869810 2 /dev/sdy 20468 261342 253385953 8923255792467 3 /dev/sdd 20468 261344 253336622 042125143176 0 /dev/sdab 20468 261345 253310522 1200778101278 2 /dev/sde 20468 261346 253286429 0950576198362 2 /dev/sdt 20468 261347 253261545 96209817917 3 /dev/sds 20468 261349 253238188 8555725937673 3 /dev/sdb Add a new tunable zfs_multihost_history to specify the number of MMP updates to store history for. By default it is set to zero meaning that no MMP statistics are stored. When using ztest to generate activity, for automated tests of the MMP function, some test functions interfere with the test. For example, the pool is exported to run zdb and then imported again. Add a new ztest function, "-M", to alter ztest behavior to prevent this. Add new tests to verify the new functionality. Tests provided by Giuseppe Di Natale. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Ned Bass <bass6@llnl.gov> Reviewed-by: Andreas Dilger <andreas.dilger@intel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Olaf Faaland <faaland1@llnl.gov> Closes #745 Closes #6279
2017-07-08 06:20:35 +03:00
char aux[256];
/*
* Dry-run failed, but we print out what success
* looks like if we found a best txg
*/
if (policy.zlp_rewind & ZPOOL_TRY_REWIND) {
zpool_rewind_exclaim(hdl, newname ? origname : thename,
B_TRUE, nv);
nvlist_free(nv);
return (-1);
}
2008-11-20 23:01:55 +03:00
if (newname == NULL)
(void) snprintf(desc, sizeof (desc),
dgettext(TEXT_DOMAIN, "cannot import '%s'"),
thename);
else
(void) snprintf(desc, sizeof (desc),
dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
origname, thename);
switch (error) {
2008-11-20 23:01:55 +03:00
case ENOTSUP:
if (nv != NULL && nvlist_lookup_nvlist(nv,
ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
(void) printf(dgettext(TEXT_DOMAIN, "This "
"pool uses the following feature(s) not "
"supported by this system:\n"));
zpool_print_unsup_feat(nv);
if (nvlist_exists(nvinfo,
ZPOOL_CONFIG_CAN_RDONLY)) {
(void) printf(dgettext(TEXT_DOMAIN,
"All unsupported features are only "
"required for writing to the pool."
"\nThe pool can be imported using "
"'-o readonly=on'.\n"));
}
}
2008-11-20 23:01:55 +03:00
/*
* Unsupported version.
*/
(void) zfs_error(hdl, EZFS_BADVERSION, desc);
break;
Multi-modifier protection (MMP) Add multihost=on|off pool property to control MMP. When enabled a new thread writes uberblocks to the last slot in each label, at a set frequency, to indicate to other hosts the pool is actively imported. These uberblocks are the last synced uberblock with an updated timestamp. Property defaults to off. During tryimport, find the "best" uberblock (newest txg and timestamp) repeatedly, checking for change in the found uberblock. Include the results of the activity test in the config returned by tryimport. These results are reported to user in "zpool import". Allow the user to control the period between MMP writes, and the duration of the activity test on import, via a new module parameter zfs_multihost_interval. The period is specified in milliseconds. The activity test duration is calculated from this value, and from the mmp_delay in the "best" uberblock found initially. Add a kstat interface to export statistics about Multiple Modifier Protection (MMP) updates. Include the last synced txg number, the timestamp, the delay since the last MMP update, the VDEV GUID, the VDEV label that received the last MMP update, and the VDEV path. Abbreviated output below. $ cat /proc/spl/kstat/zfs/mypool/multihost 31 0 0x01 10 880 105092382393521 105144180101111 txg timestamp mmp_delay vdev_guid vdev_label vdev_path 20468 261337 250274925 68396651780 3 /dev/sda 20468 261339 252023374 6267402363293 1 /dev/sdc 20468 261340 252000858 6698080955233 1 /dev/sdx 20468 261341 251980635 783892869810 2 /dev/sdy 20468 261342 253385953 8923255792467 3 /dev/sdd 20468 261344 253336622 042125143176 0 /dev/sdab 20468 261345 253310522 1200778101278 2 /dev/sde 20468 261346 253286429 0950576198362 2 /dev/sdt 20468 261347 253261545 96209817917 3 /dev/sds 20468 261349 253238188 8555725937673 3 /dev/sdb Add a new tunable zfs_multihost_history to specify the number of MMP updates to store history for. By default it is set to zero meaning that no MMP statistics are stored. When using ztest to generate activity, for automated tests of the MMP function, some test functions interfere with the test. For example, the pool is exported to run zdb and then imported again. Add a new ztest function, "-M", to alter ztest behavior to prevent this. Add new tests to verify the new functionality. Tests provided by Giuseppe Di Natale. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Ned Bass <bass6@llnl.gov> Reviewed-by: Andreas Dilger <andreas.dilger@intel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Olaf Faaland <faaland1@llnl.gov> Closes #745 Closes #6279
2017-07-08 06:20:35 +03:00
case EREMOTEIO:
if (nv != NULL && nvlist_lookup_nvlist(nv,
ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) {
char *hostname = "<unknown>";
uint64_t hostid = 0;
mmp_state_t mmp_state;
mmp_state = fnvlist_lookup_uint64(nvinfo,
ZPOOL_CONFIG_MMP_STATE);
if (nvlist_exists(nvinfo,
ZPOOL_CONFIG_MMP_HOSTNAME))
hostname = fnvlist_lookup_string(nvinfo,
ZPOOL_CONFIG_MMP_HOSTNAME);
if (nvlist_exists(nvinfo,
ZPOOL_CONFIG_MMP_HOSTID))
hostid = fnvlist_lookup_uint64(nvinfo,
ZPOOL_CONFIG_MMP_HOSTID);
if (mmp_state == MMP_STATE_ACTIVE) {
(void) snprintf(aux, sizeof (aux),
dgettext(TEXT_DOMAIN, "pool is imp"
"orted on host '%s' (hostid=%lx).\n"
"Export the pool on the other "
"system, then run 'zpool import'."),
hostname, (unsigned long) hostid);
} else if (mmp_state == MMP_STATE_NO_HOSTID) {
(void) snprintf(aux, sizeof (aux),
dgettext(TEXT_DOMAIN, "pool has "
"the multihost property on and "
"the\nsystem's hostid is not set. "
"Set a unique system hostid with "
"the zgenhostid(8) command.\n"));
Multi-modifier protection (MMP) Add multihost=on|off pool property to control MMP. When enabled a new thread writes uberblocks to the last slot in each label, at a set frequency, to indicate to other hosts the pool is actively imported. These uberblocks are the last synced uberblock with an updated timestamp. Property defaults to off. During tryimport, find the "best" uberblock (newest txg and timestamp) repeatedly, checking for change in the found uberblock. Include the results of the activity test in the config returned by tryimport. These results are reported to user in "zpool import". Allow the user to control the period between MMP writes, and the duration of the activity test on import, via a new module parameter zfs_multihost_interval. The period is specified in milliseconds. The activity test duration is calculated from this value, and from the mmp_delay in the "best" uberblock found initially. Add a kstat interface to export statistics about Multiple Modifier Protection (MMP) updates. Include the last synced txg number, the timestamp, the delay since the last MMP update, the VDEV GUID, the VDEV label that received the last MMP update, and the VDEV path. Abbreviated output below. $ cat /proc/spl/kstat/zfs/mypool/multihost 31 0 0x01 10 880 105092382393521 105144180101111 txg timestamp mmp_delay vdev_guid vdev_label vdev_path 20468 261337 250274925 68396651780 3 /dev/sda 20468 261339 252023374 6267402363293 1 /dev/sdc 20468 261340 252000858 6698080955233 1 /dev/sdx 20468 261341 251980635 783892869810 2 /dev/sdy 20468 261342 253385953 8923255792467 3 /dev/sdd 20468 261344 253336622 042125143176 0 /dev/sdab 20468 261345 253310522 1200778101278 2 /dev/sde 20468 261346 253286429 0950576198362 2 /dev/sdt 20468 261347 253261545 96209817917 3 /dev/sds 20468 261349 253238188 8555725937673 3 /dev/sdb Add a new tunable zfs_multihost_history to specify the number of MMP updates to store history for. By default it is set to zero meaning that no MMP statistics are stored. When using ztest to generate activity, for automated tests of the MMP function, some test functions interfere with the test. For example, the pool is exported to run zdb and then imported again. Add a new ztest function, "-M", to alter ztest behavior to prevent this. Add new tests to verify the new functionality. Tests provided by Giuseppe Di Natale. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Ned Bass <bass6@llnl.gov> Reviewed-by: Andreas Dilger <andreas.dilger@intel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Olaf Faaland <faaland1@llnl.gov> Closes #745 Closes #6279
2017-07-08 06:20:35 +03:00
}
(void) zfs_error_aux(hdl, aux);
}
(void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc);
break;
2008-11-20 23:01:55 +03:00
case EINVAL:
(void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
break;
case EROFS:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"one or more devices is read only"));
(void) zfs_error(hdl, EZFS_BADDEV, desc);
break;
case ENXIO:
if (nv && nvlist_lookup_nvlist(nv,
ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
nvlist_lookup_nvlist(nvinfo,
ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
(void) printf(dgettext(TEXT_DOMAIN,
OpenZFS 9075 - Improve ZFS pool import/load process and corrupted pool recovery Some work has been done lately to improve the debugability of the ZFS pool load (and import) process. This includes: 7638 Refactor spa_load_impl into several functions 8961 SPA load/import should tell us why it failed 7277 zdb should be able to print zfs_dbgmsg's To iterate on top of that, there's a few changes that were made to make the import process more resilient and crash free. One of the first tasks during the pool load process is to parse a config provided from userland that describes what devices the pool is composed of. A vdev tree is generated from that config, and then all the vdevs are opened. The Meta Object Set (MOS) of the pool is accessed, and several metadata objects that are necessary to load the pool are read. The exact configuration of the pool is also stored inside the MOS. Since the configuration provided from userland is external and might not accurately describe the vdev tree of the pool at the txg that is being loaded, it cannot be relied upon to safely operate the pool. For that reason, the configuration in the MOS is read early on. In the past, the two configurations were compared together and if there was a mismatch then the load process was aborted and an error was returned. The latter was a good way to ensure a pool does not get corrupted, however it made the pool load process needlessly fragile in cases where the vdev configuration changed or the userland configuration was outdated. Since the MOS is stored in 3 copies, the configuration provided by userland doesn't have to be perfect in order to read its contents. Hence, a new approach has been adopted: The pool is first opened with the untrusted userland configuration just so that the real configuration can be read from the MOS. The trusted MOS configuration is then used to generate a new vdev tree and the pool is re-opened. When the pool is opened with an untrusted configuration, writes are disabled to avoid accidentally damaging it. During reads, some sanity checks are performed on block pointers to see if each DVA points to a known vdev; when the configuration is untrusted, instead of panicking the system if those checks fail we simply avoid issuing reads to the invalid DVAs. This new two-step pool load process now allows rewinding pools accross vdev tree changes such as device replacement, addition, etc. Loading a pool from an external config file in a clustering environment also becomes much safer now since the pool will import even if the config is outdated and didn't, for instance, register a recent device addition. With this code in place, it became relatively easy to implement a long-sought-after feature: the ability to import a pool with missing top level (i.e. non-redundant) devices. Note that since this almost guarantees some loss of data, this feature is for now restricted to a read-only import. Porting notes (ZTS): * Fix 'make dist' target in zpool_import * The maximum path length allowed by tar is 99 characters. Several of the new test cases exceeded this limit resulting in them not being included in the tarball. Shorten the names slightly. * Set/get tunables using accessor functions. * Get last synced txg via the "zfs_txg_history" mechanism. * Clear zinject handlers in cleanup for import_cache_device_replaced and import_rewind_device_replaced in order that the zpool can be exported if there is an error. * Increase FILESIZE to 8G in zfs-test.sh to allow for a larger ext4 file system to be created on ZFS_DISK2. Also, there's no need to partition ZFS_DISK2 at all. The partitioning had already been disabled for multipath devices. Among other things, the partitioning steals some space from the ext4 file system, makes it difficult to accurately calculate the paramters to parted and can make some of the tests fail. * Increase FS_SIZE and FILE_SIZE in the zpool_import test configuration now that FILESIZE is larger. * Write more data in order that device evacuation take lonnger in a couple tests. * Use mkdir -p to avoid errors when the directory already exists. * Remove use of sudo in import_rewind_config_changed. Authored by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Andrew Stormont <andyjstormont@gmail.com> Approved by: Hans Rosenfeld <rosenfeld@grumpf.hope-2000.org> Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://illumos.org/issues/9075 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/619c0123 Closes #7459
2016-07-22 17:39:36 +03:00
"The devices below are missing or "
"corrupted, use '-m' to import the pool "
"anyway:\n"));
print_vdev_tree(hdl, NULL, missing, 2);
(void) printf("\n");
}
(void) zpool_standard_error(hdl, error, desc);
break;
case EEXIST:
(void) zpool_standard_error(hdl, error, desc);
break;
case EBUSY:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"one or more devices are already in use\n"));
(void) zfs_error(hdl, EZFS_BADDEV, desc);
break;
case ENAMETOOLONG:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"new name of at least one dataset is longer than "
"the maximum allowable length"));
(void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
break;
2008-11-20 23:01:55 +03:00
default:
(void) zpool_standard_error(hdl, error, desc);
zpool_explain_recover(hdl,
newname ? origname : thename, -error, nv);
break;
2008-11-20 23:01:55 +03:00
}
nvlist_free(nv);
2008-11-20 23:01:55 +03:00
ret = -1;
} else {
zpool_handle_t *zhp;
/*
* This should never fail, but play it safe anyway.
*/
if (zpool_open_silent(hdl, thename, &zhp) != 0)
2008-11-20 23:01:55 +03:00
ret = -1;
else if (zhp != NULL)
2008-11-20 23:01:55 +03:00
zpool_close(zhp);
if (policy.zlp_rewind &
(ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
zpool_rewind_exclaim(hdl, newname ? origname : thename,
((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv);
2008-11-20 23:01:55 +03:00
}
nvlist_free(nv);
return (0);
2008-11-20 23:01:55 +03:00
}
return (ret);
}
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
/*
* Translate vdev names to guids. If a vdev_path is determined to be
* unsuitable then a vd_errlist is allocated and the vdev path and errno
* are added to it.
*/
static int
zpool_translate_vdev_guids(zpool_handle_t *zhp, nvlist_t *vds,
nvlist_t *vdev_guids, nvlist_t *guids_to_paths, nvlist_t **vd_errlist)
{
nvlist_t *errlist = NULL;
int error = 0;
for (nvpair_t *elem = nvlist_next_nvpair(vds, NULL); elem != NULL;
elem = nvlist_next_nvpair(vds, elem)) {
boolean_t spare, cache;
char *vd_path = nvpair_name(elem);
nvlist_t *tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache,
NULL);
if ((tgt == NULL) || cache || spare) {
if (errlist == NULL) {
errlist = fnvlist_alloc();
error = EINVAL;
}
uint64_t err = (tgt == NULL) ? EZFS_NODEVICE :
(spare ? EZFS_ISSPARE : EZFS_ISL2CACHE);
fnvlist_add_int64(errlist, vd_path, err);
continue;
}
uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
fnvlist_add_uint64(vdev_guids, vd_path, guid);
char msg[MAXNAMELEN];
(void) snprintf(msg, sizeof (msg), "%llu", (u_longlong_t)guid);
fnvlist_add_string(guids_to_paths, msg, vd_path);
}
if (error != 0) {
verify(errlist != NULL);
if (vd_errlist != NULL)
*vd_errlist = errlist;
else
fnvlist_free(errlist);
}
return (error);
}
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
static int
xlate_init_err(int err)
{
switch (err) {
case ENODEV:
return (EZFS_NODEVICE);
case EINVAL:
case EROFS:
return (EZFS_BADDEV);
case EBUSY:
return (EZFS_INITIALIZING);
case ESRCH:
return (EZFS_NO_INITIALIZE);
}
return (err);
}
/*
* Begin, suspend, or cancel the initialization (initializing of all free
* blocks) for the given vdevs in the given pool.
*/
int
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
zpool_initialize_impl(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
nvlist_t *vds, boolean_t wait)
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
{
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
int err;
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
nvlist_t *vdev_guids = fnvlist_alloc();
nvlist_t *guids_to_paths = fnvlist_alloc();
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
nvlist_t *vd_errlist = NULL;
nvlist_t *errlist;
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
nvpair_t *elem;
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
guids_to_paths, &vd_errlist);
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
if (err != 0) {
verify(vd_errlist != NULL);
goto list_errors;
}
err = lzc_initialize(zhp->zpool_name, cmd_type,
vdev_guids, &errlist);
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
if (err != 0) {
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
if (errlist != NULL) {
vd_errlist = fnvlist_lookup_nvlist(errlist,
ZPOOL_INITIALIZE_VDEVS);
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
goto list_errors;
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
}
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
(void) zpool_standard_error(zhp->zpool_hdl, err,
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
dgettext(TEXT_DOMAIN, "operation failed"));
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
goto out;
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
}
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
if (wait) {
for (elem = nvlist_next_nvpair(vdev_guids, NULL); elem != NULL;
elem = nvlist_next_nvpair(vdev_guids, elem)) {
uint64_t guid = fnvpair_value_uint64(elem);
err = lzc_wait_tag(zhp->zpool_name,
ZPOOL_WAIT_INITIALIZE, guid, NULL);
if (err != 0) {
(void) zpool_standard_error_fmt(zhp->zpool_hdl,
err, dgettext(TEXT_DOMAIN, "error "
"waiting for '%s' to initialize"),
nvpair_name(elem));
goto out;
}
}
}
goto out;
list_errors:
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL;
elem = nvlist_next_nvpair(vd_errlist, elem)) {
int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem));
char *path;
if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
&path) != 0)
path = nvpair_name(elem);
(void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
"cannot initialize '%s'", path);
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
}
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
out:
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
fnvlist_free(vdev_guids);
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
fnvlist_free(guids_to_paths);
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
if (vd_errlist != NULL)
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
fnvlist_free(vd_errlist);
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
return (err == 0 ? 0 : -1);
}
int
zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
nvlist_t *vds)
{
return (zpool_initialize_impl(zhp, cmd_type, vds, B_FALSE));
}
int
zpool_initialize_wait(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
nvlist_t *vds)
{
return (zpool_initialize_impl(zhp, cmd_type, vds, B_TRUE));
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
}
static int
xlate_trim_err(int err)
{
switch (err) {
case ENODEV:
return (EZFS_NODEVICE);
case EINVAL:
case EROFS:
return (EZFS_BADDEV);
case EBUSY:
return (EZFS_TRIMMING);
case ESRCH:
return (EZFS_NO_TRIM);
case EOPNOTSUPP:
return (EZFS_TRIM_NOTSUP);
}
return (err);
}
/*
* Begin, suspend, or cancel the TRIM (discarding of all free blocks) for
* the given vdevs in the given pool.
*/
int
zpool_trim(zpool_handle_t *zhp, pool_trim_func_t cmd_type, nvlist_t *vds,
trimflags_t *trim_flags)
{
char msg[1024];
int err;
nvlist_t *vdev_guids = fnvlist_alloc();
nvlist_t *guids_to_paths = fnvlist_alloc();
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
nvlist_t *vd_errlist = NULL;
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
nvlist_t *errlist;
nvpair_t *elem;
err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
guids_to_paths, &vd_errlist);
if (err == 0) {
err = lzc_trim(zhp->zpool_name, cmd_type, trim_flags->rate,
trim_flags->secure, vdev_guids, &errlist);
if (err == 0) {
fnvlist_free(vdev_guids);
fnvlist_free(guids_to_paths);
return (0);
}
if (errlist != NULL) {
vd_errlist = fnvlist_lookup_nvlist(errlist,
ZPOOL_TRIM_VDEVS);
}
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "operation failed"));
} else {
verify(vd_errlist != NULL);
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
}
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
for (elem = nvlist_next_nvpair(vd_errlist, NULL);
elem != NULL; elem = nvlist_next_nvpair(vd_errlist, elem)) {
int64_t vd_error = xlate_trim_err(fnvpair_value_int64(elem));
char *path;
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
/*
* If only the pool was specified, and it was not a secure
* trim then suppress warnings for individual vdevs which
* do not support trimming.
*/
if (vd_error == EZFS_TRIM_NOTSUP &&
trim_flags->fullpool &&
!trim_flags->secure) {
continue;
}
if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
&path) != 0)
path = nvpair_name(elem);
(void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
"cannot trim '%s'", path);
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
}
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
fnvlist_free(vdev_guids);
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
fnvlist_free(guids_to_paths);
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
if (vd_errlist != NULL) {
fnvlist_free(vd_errlist);
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
return (-1);
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
}
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
Add TRIM support UNMAP/TRIM support is a frequently-requested feature to help prevent performance from degrading on SSDs and on various other SAN-like storage back-ends. By issuing UNMAP/TRIM commands for sectors which are no longer allocated the underlying device can often more efficiently manage itself. This TRIM implementation is modeled on the `zpool initialize` feature which writes a pattern to all unallocated space in the pool. The new `zpool trim` command uses the same vdev_xlate() code to calculate what sectors are unallocated, the same per- vdev TRIM thread model and locking, and the same basic CLI for a consistent user experience. The core difference is that instead of writing a pattern it will issue UNMAP/TRIM commands for those extents. The zio pipeline was updated to accommodate this by adding a new ZIO_TYPE_TRIM type and associated spa taskq. This new type makes is straight forward to add the platform specific TRIM/UNMAP calls to vdev_disk.c and vdev_file.c. These new ZIO_TYPE_TRIM zios are handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs. This makes it possible to largely avoid changing the pipieline, one exception is that TRIM zio's may exceed the 16M block size limit since they contain no data. In addition to the manual `zpool trim` command, a background automatic TRIM was added and is controlled by the 'autotrim' property. It relies on the exact same infrastructure as the manual TRIM. However, instead of relying on the extents in a metaslab's ms_allocatable range tree, a ms_trim tree is kept per metaslab. When 'autotrim=on', ranges added back to the ms_allocatable tree are also added to the ms_free tree. The ms_free tree is then periodically consumed by an autotrim thread which systematically walks a top level vdev's metaslabs. Since the automatic TRIM will skip ranges it considers too small there is value in occasionally running a full `zpool trim`. This may occur when the freed blocks are small and not enough time was allowed to aggregate them. An automatic TRIM and a manual `zpool trim` may be run concurrently, in which case the automatic TRIM will yield to the manual TRIM. Reviewed-by: Jorgen Lundman <lundman@lundman.net> Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com> Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com> Contributions-by: Tim Chase <tim@chase2k.com> Contributions-by: Chunwei Chen <tuxoko@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #8419 Closes #598
2019-03-29 19:13:20 +03:00
return (zpool_standard_error(zhp->zpool_hdl, err, msg));
OpenZFS 9102 - zfs should be able to initialize storage devices PROBLEM ======== The first access to a block incurs a performance penalty on some platforms (e.g. AWS's EBS, VMware VMDKs). Therefore we recommend that volumes are "thick provisioned", where supported by the platform (VMware). This can create a large delay in getting a new virtual machines up and running (or adding storage to an existing Engine). If the thick provision step is omitted, write performance will be suboptimal until all blocks on the LUN have been written. SOLUTION ========= This feature introduces a way to 'initialize' the disks at install or in the background to make sure we don't incur this first read penalty. When an entire LUN is added to ZFS, we make all space available immediately, and allow ZFS to find unallocated space and zero it out. This works with concurrent writes to arbitrary offsets, ensuring that we don't zero out something that has been (or is in the middle of being) written. This scheme can also be applied to existing pools (affecting only free regions on the vdev). Detailed design: - new subcommand:zpool initialize [-cs] <pool> [<vdev> ...] - start, suspend, or cancel initialization - Creates new open-context thread for each vdev - Thread iterates through all metaslabs in this vdev - Each metaslab: - select a metaslab - load the metaslab - mark the metaslab as being zeroed - walk all free ranges within that metaslab and translate them to ranges on the leaf vdev - issue a "zeroing" I/O on the leaf vdev that corresponds to a free range on the metaslab we're working on - continue until all free ranges for this metaslab have been "zeroed" - reset/unmark the metaslab being zeroed - if more metaslabs exist, then repeat above tasks. - if no more metaslabs, then we're done. - progress for the initialization is stored on-disk in the vdev’s leaf zap object. The following information is stored: - the last offset that has been initialized - the state of the initialization process (i.e. active, suspended, or canceled) - the start time for the initialization - progress is reported via the zpool status command and shows information for each of the vdevs that are initializing Porting notes: - Added zfs_initialize_value module parameter to set the pattern written by "zpool initialize". - Added zfs_vdev_{initializing,removal}_{min,max}_active module options. Authored by: George Wilson <george.wilson@delphix.com> Reviewed by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed by: Prakash Surya <prakash.surya@delphix.com> Reviewed by: loli10K <ezomori.nozomu@gmail.com> Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov> Approved by: Richard Lowe <richlowe@richlowe.net> Signed-off-by: Tim Chase <tim@chase2k.com> Ported-by: Tim Chase <tim@chase2k.com> OpenZFS-issue: https://www.illumos.org/issues/9102 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/c3963210eb Closes #8230
2018-12-19 17:54:59 +03:00
}
2008-11-20 23:01:55 +03:00
/*
* Scan the pool.
2008-11-20 23:01:55 +03:00
*/
int
zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
2008-11-20 23:01:55 +03:00
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
int err;
2008-11-20 23:01:55 +03:00
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
zc.zc_cookie = func;
zc.zc_flags = cmd;
2008-11-20 23:01:55 +03:00
if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
return (0);
err = errno;
/* ECANCELED on a scrub means we resumed a paused scrub */
if (err == ECANCELED && func == POOL_SCAN_SCRUB &&
cmd == POOL_SCRUB_NORMAL)
return (0);
if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL)
2008-11-20 23:01:55 +03:00
return (0);
if (func == POOL_SCAN_SCRUB) {
if (cmd == POOL_SCRUB_PAUSE) {
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot pause scrubbing %s"), zc.zc_name);
} else {
assert(cmd == POOL_SCRUB_NORMAL);
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot scrub %s"), zc.zc_name);
}
} else if (func == POOL_SCAN_RESILVER) {
assert(cmd == POOL_SCRUB_NORMAL);
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot restart resilver on %s"), zc.zc_name);
} else if (func == POOL_SCAN_NONE) {
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
zc.zc_name);
} else {
assert(!"unexpected result");
}
2008-11-20 23:01:55 +03:00
if (err == EBUSY) {
nvlist_t *nvroot;
pool_scan_stat_t *ps = NULL;
uint_t psc;
verify(nvlist_lookup_nvlist(zhp->zpool_config,
ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
(void) nvlist_lookup_uint64_array(nvroot,
ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
if (ps && ps->pss_func == POOL_SCAN_SCRUB) {
if (cmd == POOL_SCRUB_PAUSE)
return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg));
else
return (zfs_error(hdl, EZFS_SCRUBBING, msg));
} else {
return (zfs_error(hdl, EZFS_RESILVERING, msg));
}
} else if (err == ENOENT) {
return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
} else if (err == ENOTSUP && func == POOL_SCAN_RESILVER) {
return (zfs_error(hdl, EZFS_NO_RESILVER_DEFER, msg));
} else {
return (zpool_standard_error(hdl, err, msg));
}
}
2008-11-20 23:01:55 +03:00
/*
2009-07-03 02:44:48 +04:00
* Find a vdev that matches the search criteria specified. We use the
* the nvpair name to determine how we should look for the device.
2008-11-20 23:01:55 +03:00
* 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
* spare; but FALSE if its an INUSE spare.
*/
static nvlist_t *
2009-07-03 02:44:48 +04:00
vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
boolean_t *l2cache, boolean_t *log)
2008-11-20 23:01:55 +03:00
{
uint_t c, children;
nvlist_t **child;
nvlist_t *ret;
uint64_t is_log;
2009-07-03 02:44:48 +04:00
char *srchkey;
nvpair_t *pair = nvlist_next_nvpair(search, NULL);
/* Nothing to look for */
if (search == NULL || pair == NULL)
return (NULL);
/* Obtain the key we will use to search */
srchkey = nvpair_name(pair);
switch (nvpair_type(pair)) {
case DATA_TYPE_UINT64:
2009-07-03 02:44:48 +04:00
if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
uint64_t srchval, theguid;
verify(nvpair_value_uint64(pair, &srchval) == 0);
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
&theguid) == 0);
if (theguid == srchval)
return (nv);
2009-07-03 02:44:48 +04:00
}
break;
case DATA_TYPE_STRING: {
char *srchval, *val;
verify(nvpair_value_string(pair, &srchval) == 0);
if (nvlist_lookup_string(nv, srchkey, &val) != 0)
break;
2008-11-20 23:01:55 +03:00
2009-07-03 02:44:48 +04:00
/*
* Search for the requested value. Special cases:
*
2012-10-18 03:58:54 +04:00
* - ZPOOL_CONFIG_PATH for whole disk entries. These end in
* "-part1", or "p1". The suffix is hidden from the user,
* but included in the string, so this matches around it.
* - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
* is used to check all possible expanded paths.
* - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
*
* Otherwise, all other searches are simple string compares.
2009-07-03 02:44:48 +04:00
*/
if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
2009-07-03 02:44:48 +04:00
uint64_t wholedisk = 0;
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
&wholedisk);
2012-10-18 03:58:54 +04:00
if (zfs_strcmp_pathname(srchval, val, wholedisk) == 0)
return (nv);
} else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
char *type, *idx, *end, *p;
uint64_t id, vdev_id;
/*
* Determine our vdev type, keeping in mind
* that the srchval is composed of a type and
* vdev id pair (i.e. mirror-4).
*/
if ((type = strdup(srchval)) == NULL)
return (NULL);
if ((p = strrchr(type, '-')) == NULL) {
free(type);
break;
}
idx = p + 1;
*p = '\0';
/*
* If the types don't match then keep looking.
*/
if (strncmp(val, type, strlen(val)) != 0) {
free(type);
break;
}
verify(zpool_vdev_is_interior(type));
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
&id) == 0);
errno = 0;
vdev_id = strtoull(idx, &end, 10);
free(type);
if (errno != 0)
return (NULL);
/*
* Now verify that we have the correct vdev id.
*/
if (vdev_id == id)
return (nv);
2009-07-03 02:44:48 +04:00
}
2008-11-20 23:01:55 +03:00
/*
2009-07-03 02:44:48 +04:00
* Common case
2008-11-20 23:01:55 +03:00
*/
2009-07-03 02:44:48 +04:00
if (strcmp(srchval, val) == 0)
2008-11-20 23:01:55 +03:00
return (nv);
2009-07-03 02:44:48 +04:00
break;
}
default:
break;
2008-11-20 23:01:55 +03:00
}
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0)
return (NULL);
for (c = 0; c < children; c++) {
2009-07-03 02:44:48 +04:00
if ((ret = vdev_to_nvlist_iter(child[c], search,
avail_spare, l2cache, NULL)) != NULL) {
/*
* The 'is_log' value is only set for the toplevel
* vdev, not the leaf vdevs. So we always lookup the
* log device from the root of the vdev tree (where
* 'log' is non-NULL).
*/
if (log != NULL &&
nvlist_lookup_uint64(child[c],
ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
is_log) {
*log = B_TRUE;
}
2008-11-20 23:01:55 +03:00
return (ret);
}
}
2008-11-20 23:01:55 +03:00
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
&child, &children) == 0) {
for (c = 0; c < children; c++) {
2009-07-03 02:44:48 +04:00
if ((ret = vdev_to_nvlist_iter(child[c], search,
avail_spare, l2cache, NULL)) != NULL) {
2008-11-20 23:01:55 +03:00
*avail_spare = B_TRUE;
return (ret);
}
}
}
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
&child, &children) == 0) {
for (c = 0; c < children; c++) {
2009-07-03 02:44:48 +04:00
if ((ret = vdev_to_nvlist_iter(child[c], search,
avail_spare, l2cache, NULL)) != NULL) {
2008-11-20 23:01:55 +03:00
*l2cache = B_TRUE;
return (ret);
}
}
}
return (NULL);
}
2009-07-03 02:44:48 +04:00
/*
Add support for autoexpand property While the autoexpand property may seem like a small feature it depends on a significant amount of system infrastructure. Enough of that infrastructure is now in place that with a few modifications for Linux it can be supported. Auto-expand works as follows; when a block device is modified (re-sized, closed after being open r/w, etc) a change uevent is generated for udev. The ZED, which is monitoring udev events, passes the change event along to zfs_deliver_dle() if the disk or partition contains a zfs_member as identified by blkid. From here the device is matched against all imported pool vdevs using the vdev_guid which was read from the label by blkid. If a match is found the ZED reopens the pool vdev. This re-opening is important because it allows the vdev to be briefly closed so the disk partition table can be re-read. Otherwise, it wouldn't be possible to report the maximum possible expansion size. Finally, if the property autoexpand=on a vdev expansion will be attempted. After performing some sanity checks on the disk to verify that it is safe to expand, the primary partition (-part1) will be expanded and the partition table updated. The partition is then re-opened (again) to detect the updated size which allows the new capacity to be used. In order to make all of the above possible the following changes were required: * Updated the zpool_expand_001_pos and zpool_expand_003_pos tests. These tests now create a pool which is layered on a loopback, scsi_debug, and file vdev. This allows for testing of non- partitioned block device (loopback), a partition block device (scsi_debug), and a file which does not receive udev change events. This provided for better test coverage, and by removing the layering on ZFS volumes there issues surrounding layering one pool on another are avoided. * zpool_find_vdev_by_physpath() updated to accept a vdev guid. This allows for matching by guid rather than path which is a more reliable way for the ZED to reference a vdev. * Fixed zfs_zevent_wait() signal handling which could result in the ZED spinning when a signal was not handled. * Removed vdev_disk_rrpart() functionality which can be abandoned in favor of kernel provided blkdev_reread_part() function. * Added a rwlock which is held as a writer while a disk is being reopened. This is important to prevent errors from occurring for any configuration related IOs which bypass the SCL_ZIO lock. The zpool_reopen_007_pos.ksh test case was added to verify IO error are never observed when reopening. This is not expected to impact IO performance. Additional fixes which aren't critical but were discovered and resolved in the course of developing this functionality. * Added PHYS_PATH="/dev/zvol/dataset" to the vdev configuration for ZFS volumes. This is as good as a unique physical path, while the volumes are not used in the test cases anymore for other reasons this improvement was included. Reviewed by: Richard Elling <Richard.Elling@RichardElling.com> Signed-off-by: Sara Hartse <sara.hartse@delphix.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #120 Closes #2437 Closes #5771 Closes #7366 Closes #7582 Closes #7629
2018-07-24 01:40:15 +03:00
* Given a physical path or guid, find the associated vdev.
2009-07-03 02:44:48 +04:00
*/
nvlist_t *
zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
{
nvlist_t *search, *nvroot, *ret;
Add support for autoexpand property While the autoexpand property may seem like a small feature it depends on a significant amount of system infrastructure. Enough of that infrastructure is now in place that with a few modifications for Linux it can be supported. Auto-expand works as follows; when a block device is modified (re-sized, closed after being open r/w, etc) a change uevent is generated for udev. The ZED, which is monitoring udev events, passes the change event along to zfs_deliver_dle() if the disk or partition contains a zfs_member as identified by blkid. From here the device is matched against all imported pool vdevs using the vdev_guid which was read from the label by blkid. If a match is found the ZED reopens the pool vdev. This re-opening is important because it allows the vdev to be briefly closed so the disk partition table can be re-read. Otherwise, it wouldn't be possible to report the maximum possible expansion size. Finally, if the property autoexpand=on a vdev expansion will be attempted. After performing some sanity checks on the disk to verify that it is safe to expand, the primary partition (-part1) will be expanded and the partition table updated. The partition is then re-opened (again) to detect the updated size which allows the new capacity to be used. In order to make all of the above possible the following changes were required: * Updated the zpool_expand_001_pos and zpool_expand_003_pos tests. These tests now create a pool which is layered on a loopback, scsi_debug, and file vdev. This allows for testing of non- partitioned block device (loopback), a partition block device (scsi_debug), and a file which does not receive udev change events. This provided for better test coverage, and by removing the layering on ZFS volumes there issues surrounding layering one pool on another are avoided. * zpool_find_vdev_by_physpath() updated to accept a vdev guid. This allows for matching by guid rather than path which is a more reliable way for the ZED to reference a vdev. * Fixed zfs_zevent_wait() signal handling which could result in the ZED spinning when a signal was not handled. * Removed vdev_disk_rrpart() functionality which can be abandoned in favor of kernel provided blkdev_reread_part() function. * Added a rwlock which is held as a writer while a disk is being reopened. This is important to prevent errors from occurring for any configuration related IOs which bypass the SCL_ZIO lock. The zpool_reopen_007_pos.ksh test case was added to verify IO error are never observed when reopening. This is not expected to impact IO performance. Additional fixes which aren't critical but were discovered and resolved in the course of developing this functionality. * Added PHYS_PATH="/dev/zvol/dataset" to the vdev configuration for ZFS volumes. This is as good as a unique physical path, while the volumes are not used in the test cases anymore for other reasons this improvement was included. Reviewed by: Richard Elling <Richard.Elling@RichardElling.com> Signed-off-by: Sara Hartse <sara.hartse@delphix.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #120 Closes #2437 Closes #5771 Closes #7366 Closes #7582 Closes #7629
2018-07-24 01:40:15 +03:00
uint64_t guid;
char *end;
2009-07-03 02:44:48 +04:00
verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
Add support for autoexpand property While the autoexpand property may seem like a small feature it depends on a significant amount of system infrastructure. Enough of that infrastructure is now in place that with a few modifications for Linux it can be supported. Auto-expand works as follows; when a block device is modified (re-sized, closed after being open r/w, etc) a change uevent is generated for udev. The ZED, which is monitoring udev events, passes the change event along to zfs_deliver_dle() if the disk or partition contains a zfs_member as identified by blkid. From here the device is matched against all imported pool vdevs using the vdev_guid which was read from the label by blkid. If a match is found the ZED reopens the pool vdev. This re-opening is important because it allows the vdev to be briefly closed so the disk partition table can be re-read. Otherwise, it wouldn't be possible to report the maximum possible expansion size. Finally, if the property autoexpand=on a vdev expansion will be attempted. After performing some sanity checks on the disk to verify that it is safe to expand, the primary partition (-part1) will be expanded and the partition table updated. The partition is then re-opened (again) to detect the updated size which allows the new capacity to be used. In order to make all of the above possible the following changes were required: * Updated the zpool_expand_001_pos and zpool_expand_003_pos tests. These tests now create a pool which is layered on a loopback, scsi_debug, and file vdev. This allows for testing of non- partitioned block device (loopback), a partition block device (scsi_debug), and a file which does not receive udev change events. This provided for better test coverage, and by removing the layering on ZFS volumes there issues surrounding layering one pool on another are avoided. * zpool_find_vdev_by_physpath() updated to accept a vdev guid. This allows for matching by guid rather than path which is a more reliable way for the ZED to reference a vdev. * Fixed zfs_zevent_wait() signal handling which could result in the ZED spinning when a signal was not handled. * Removed vdev_disk_rrpart() functionality which can be abandoned in favor of kernel provided blkdev_reread_part() function. * Added a rwlock which is held as a writer while a disk is being reopened. This is important to prevent errors from occurring for any configuration related IOs which bypass the SCL_ZIO lock. The zpool_reopen_007_pos.ksh test case was added to verify IO error are never observed when reopening. This is not expected to impact IO performance. Additional fixes which aren't critical but were discovered and resolved in the course of developing this functionality. * Added PHYS_PATH="/dev/zvol/dataset" to the vdev configuration for ZFS volumes. This is as good as a unique physical path, while the volumes are not used in the test cases anymore for other reasons this improvement was included. Reviewed by: Richard Elling <Richard.Elling@RichardElling.com> Signed-off-by: Sara Hartse <sara.hartse@delphix.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #120 Closes #2437 Closes #5771 Closes #7366 Closes #7582 Closes #7629
2018-07-24 01:40:15 +03:00
guid = strtoull(ppath, &end, 0);
if (guid != 0 && *end == '\0') {
verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
} else {
verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH,
ppath) == 0);
}
2009-07-03 02:44:48 +04:00
verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0);
*avail_spare = B_FALSE;
*l2cache = B_FALSE;
if (log != NULL)
*log = B_FALSE;
2009-07-03 02:44:48 +04:00
ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
nvlist_free(search);
return (ret);
}
/*
* Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
*/
static boolean_t
zpool_vdev_is_interior(const char *name)
{
if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 ||
strncmp(name,
VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 ||
strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
return (B_TRUE);
return (B_FALSE);
}
2008-11-20 23:01:55 +03:00
nvlist_t *
zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
boolean_t *l2cache, boolean_t *log)
2008-11-20 23:01:55 +03:00
{
char *end;
2009-07-03 02:44:48 +04:00
nvlist_t *nvroot, *search, *ret;
2008-11-20 23:01:55 +03:00
uint64_t guid;
2009-07-03 02:44:48 +04:00
verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
guid = strtoull(path, &end, 0);
2008-11-20 23:01:55 +03:00
if (guid != 0 && *end == '\0') {
2009-07-03 02:44:48 +04:00
verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
} else if (zpool_vdev_is_interior(path)) {
verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2008-11-20 23:01:55 +03:00
} else {
2009-07-03 02:44:48 +04:00
verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2008-11-20 23:01:55 +03:00
}
verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0);
*avail_spare = B_FALSE;
*l2cache = B_FALSE;
if (log != NULL)
*log = B_FALSE;
2009-07-03 02:44:48 +04:00
ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
nvlist_free(search);
return (ret);
}
static int
Multi-modifier protection (MMP) Add multihost=on|off pool property to control MMP. When enabled a new thread writes uberblocks to the last slot in each label, at a set frequency, to indicate to other hosts the pool is actively imported. These uberblocks are the last synced uberblock with an updated timestamp. Property defaults to off. During tryimport, find the "best" uberblock (newest txg and timestamp) repeatedly, checking for change in the found uberblock. Include the results of the activity test in the config returned by tryimport. These results are reported to user in "zpool import". Allow the user to control the period between MMP writes, and the duration of the activity test on import, via a new module parameter zfs_multihost_interval. The period is specified in milliseconds. The activity test duration is calculated from this value, and from the mmp_delay in the "best" uberblock found initially. Add a kstat interface to export statistics about Multiple Modifier Protection (MMP) updates. Include the last synced txg number, the timestamp, the delay since the last MMP update, the VDEV GUID, the VDEV label that received the last MMP update, and the VDEV path. Abbreviated output below. $ cat /proc/spl/kstat/zfs/mypool/multihost 31 0 0x01 10 880 105092382393521 105144180101111 txg timestamp mmp_delay vdev_guid vdev_label vdev_path 20468 261337 250274925 68396651780 3 /dev/sda 20468 261339 252023374 6267402363293 1 /dev/sdc 20468 261340 252000858 6698080955233 1 /dev/sdx 20468 261341 251980635 783892869810 2 /dev/sdy 20468 261342 253385953 8923255792467 3 /dev/sdd 20468 261344 253336622 042125143176 0 /dev/sdab 20468 261345 253310522 1200778101278 2 /dev/sde 20468 261346 253286429 0950576198362 2 /dev/sdt 20468 261347 253261545 96209817917 3 /dev/sds 20468 261349 253238188 8555725937673 3 /dev/sdb Add a new tunable zfs_multihost_history to specify the number of MMP updates to store history for. By default it is set to zero meaning that no MMP statistics are stored. When using ztest to generate activity, for automated tests of the MMP function, some test functions interfere with the test. For example, the pool is exported to run zdb and then imported again. Add a new ztest function, "-M", to alter ztest behavior to prevent this. Add new tests to verify the new functionality. Tests provided by Giuseppe Di Natale. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Ned Bass <bass6@llnl.gov> Reviewed-by: Andreas Dilger <andreas.dilger@intel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Olaf Faaland <faaland1@llnl.gov> Closes #745 Closes #6279
2017-07-08 06:20:35 +03:00
vdev_is_online(nvlist_t *nv)
{
uint64_t ival;
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
return (0);
return (1);
}
/*
2009-07-03 02:44:48 +04:00
* Helper function for zpool_get_physpaths().
*/
2009-07-03 02:44:48 +04:00
static int
vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
size_t *bytes_written)
{
size_t bytes_left, pos, rsz;
char *tmppath;
const char *format;
if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
&tmppath) != 0)
return (EZFS_NODEVICE);
pos = *bytes_written;
bytes_left = physpath_size - pos;
format = (pos == 0) ? "%s" : " %s";
rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
*bytes_written += rsz;
if (rsz >= bytes_left) {
/* if physpath was not copied properly, clear it */
if (bytes_left != 0) {
physpath[pos] = 0;
}
return (EZFS_NOSPC);
}
return (0);
}
static int
vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
size_t *rsz, boolean_t is_spare)
{
char *type;
int ret;
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
return (EZFS_INVALCONFIG);
if (strcmp(type, VDEV_TYPE_DISK) == 0) {
/*
* An active spare device has ZPOOL_CONFIG_IS_SPARE set.
* For a spare vdev, we only want to boot from the active
* spare device.
*/
if (is_spare) {
uint64_t spare = 0;
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
&spare);
if (!spare)
return (EZFS_INVALCONFIG);
}
Multi-modifier protection (MMP) Add multihost=on|off pool property to control MMP. When enabled a new thread writes uberblocks to the last slot in each label, at a set frequency, to indicate to other hosts the pool is actively imported. These uberblocks are the last synced uberblock with an updated timestamp. Property defaults to off. During tryimport, find the "best" uberblock (newest txg and timestamp) repeatedly, checking for change in the found uberblock. Include the results of the activity test in the config returned by tryimport. These results are reported to user in "zpool import". Allow the user to control the period between MMP writes, and the duration of the activity test on import, via a new module parameter zfs_multihost_interval. The period is specified in milliseconds. The activity test duration is calculated from this value, and from the mmp_delay in the "best" uberblock found initially. Add a kstat interface to export statistics about Multiple Modifier Protection (MMP) updates. Include the last synced txg number, the timestamp, the delay since the last MMP update, the VDEV GUID, the VDEV label that received the last MMP update, and the VDEV path. Abbreviated output below. $ cat /proc/spl/kstat/zfs/mypool/multihost 31 0 0x01 10 880 105092382393521 105144180101111 txg timestamp mmp_delay vdev_guid vdev_label vdev_path 20468 261337 250274925 68396651780 3 /dev/sda 20468 261339 252023374 6267402363293 1 /dev/sdc 20468 261340 252000858 6698080955233 1 /dev/sdx 20468 261341 251980635 783892869810 2 /dev/sdy 20468 261342 253385953 8923255792467 3 /dev/sdd 20468 261344 253336622 042125143176 0 /dev/sdab 20468 261345 253310522 1200778101278 2 /dev/sde 20468 261346 253286429 0950576198362 2 /dev/sdt 20468 261347 253261545 96209817917 3 /dev/sds 20468 261349 253238188 8555725937673 3 /dev/sdb Add a new tunable zfs_multihost_history to specify the number of MMP updates to store history for. By default it is set to zero meaning that no MMP statistics are stored. When using ztest to generate activity, for automated tests of the MMP function, some test functions interfere with the test. For example, the pool is exported to run zdb and then imported again. Add a new ztest function, "-M", to alter ztest behavior to prevent this. Add new tests to verify the new functionality. Tests provided by Giuseppe Di Natale. Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Ned Bass <bass6@llnl.gov> Reviewed-by: Andreas Dilger <andreas.dilger@intel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Olaf Faaland <faaland1@llnl.gov> Closes #745 Closes #6279
2017-07-08 06:20:35 +03:00
if (vdev_is_online(nv)) {
2009-07-03 02:44:48 +04:00
if ((ret = vdev_get_one_physpath(nv, physpath,
phypath_size, rsz)) != 0)
return (ret);
}
} else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
2009-07-03 02:44:48 +04:00
strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
(is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
nvlist_t **child;
uint_t count;
int i, ret;
if (nvlist_lookup_nvlist_array(nv,
ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
return (EZFS_INVALCONFIG);
for (i = 0; i < count; i++) {
ret = vdev_get_physpaths(child[i], physpath,
phypath_size, rsz, is_spare);
if (ret == EZFS_NOSPC)
return (ret);
}
}
return (EZFS_POOL_INVALARG);
}
/*
* Get phys_path for a root pool config.
* Return 0 on success; non-zero on failure.
*/
static int
zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
{
2009-07-03 02:44:48 +04:00
size_t rsz;
nvlist_t *vdev_root;
nvlist_t **child;
uint_t count;
2009-07-03 02:44:48 +04:00
char *type;
2009-07-03 02:44:48 +04:00
rsz = 0;
2009-07-03 02:44:48 +04:00
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
&vdev_root) != 0)
return (EZFS_INVALCONFIG);
2009-07-03 02:44:48 +04:00
if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
&child, &count) != 0)
2009-07-03 02:44:48 +04:00
return (EZFS_INVALCONFIG);
2009-07-03 02:44:48 +04:00
/*
* root pool can only have a single top-level vdev.
2009-07-03 02:44:48 +04:00
*/
if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1)
2009-07-03 02:44:48 +04:00
return (EZFS_POOL_INVALARG);
2009-07-03 02:44:48 +04:00
(void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
B_FALSE);
/* No online devices */
if (rsz == 0)
return (EZFS_NODEVICE);
return (0);
2008-11-20 23:01:55 +03:00
}
2009-07-03 02:44:48 +04:00
/*
* Get phys_path for a root pool
* Return 0 on success; non-zero on failure.
*/
int
zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
{
return (zpool_get_config_physpath(zhp->zpool_config, physpath,
phypath_size));
}
/*
* Convert a vdev path to a GUID. Returns GUID or 0 on error.
*
* If is_spare, is_l2cache, or is_log is non-NULL, then store within it
* if the VDEV is a spare, l2cache, or log device. If they're NULL then
* ignore them.
*/
static uint64_t
zpool_vdev_path_to_guid_impl(zpool_handle_t *zhp, const char *path,
boolean_t *is_spare, boolean_t *is_l2cache, boolean_t *is_log)
{
uint64_t guid;
boolean_t spare = B_FALSE, l2cache = B_FALSE, log = B_FALSE;
nvlist_t *tgt;
if ((tgt = zpool_find_vdev(zhp, path, &spare, &l2cache,
&log)) == NULL)
return (0);
verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &guid) == 0);
if (is_spare != NULL)
*is_spare = spare;
if (is_l2cache != NULL)
*is_l2cache = l2cache;
if (is_log != NULL)
*is_log = log;
return (guid);
}
/* Convert a vdev path to a GUID. Returns GUID or 0 on error. */
uint64_t
zpool_vdev_path_to_guid(zpool_handle_t *zhp, const char *path)
{
return (zpool_vdev_path_to_guid_impl(zhp, path, NULL, NULL, NULL));
}
2008-11-20 23:01:55 +03:00
/*
* Bring the specified vdev online. The 'flags' parameter is a set of the
* ZFS_ONLINE_* flags.
*/
int
zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
vdev_state_t *newstate)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
char *pathname;
2008-11-20 23:01:55 +03:00
nvlist_t *tgt;
2009-07-03 02:44:48 +04:00
boolean_t avail_spare, l2cache, islog;
2008-11-20 23:01:55 +03:00
libzfs_handle_t *hdl = zhp->zpool_hdl;
int error;
2008-11-20 23:01:55 +03:00
2009-07-03 02:44:48 +04:00
if (flags & ZFS_ONLINE_EXPAND) {
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
} else {
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot online %s"), path);
}
2008-11-20 23:01:55 +03:00
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2009-07-03 02:44:48 +04:00
&islog)) == NULL)
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_NODEVICE, msg));
verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
if (avail_spare)
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_ISSPARE, msg));
if ((flags & ZFS_ONLINE_EXPAND ||
zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
2009-07-03 02:44:48 +04:00
uint64_t wholedisk = 0;
(void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
&wholedisk);
/*
* XXX - L2ARC 1.0 devices can't support expansion.
*/
if (l2cache) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"cannot expand cache devices"));
return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
}
if (wholedisk) {
const char *fullpath = path;
char buf[MAXPATHLEN];
if (path[0] != '/') {
error = zfs_resolve_shortname(path, buf,
sizeof (buf));
if (error != 0)
return (zfs_error(hdl, EZFS_NODEVICE,
msg));
fullpath = buf;
}
error = zpool_relabel_disk(hdl, fullpath, msg);
if (error != 0)
return (error);
2009-07-03 02:44:48 +04:00
}
}
2008-11-20 23:01:55 +03:00
zc.zc_cookie = VDEV_STATE_ONLINE;
zc.zc_obj = flags;
if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
if (errno == EINVAL) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
"from this pool into a new one. Use '%s' "
"instead"), "zpool detach");
return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
}
2008-11-20 23:01:55 +03:00
return (zpool_standard_error(hdl, errno, msg));
}
2008-11-20 23:01:55 +03:00
*newstate = zc.zc_cookie;
return (0);
}
/*
* Take the specified vdev offline
*/
int
zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
nvlist_t *tgt;
boolean_t avail_spare, l2cache;
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
NULL)) == NULL)
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_NODEVICE, msg));
verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
if (avail_spare)
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_ISSPARE, msg));
zc.zc_cookie = VDEV_STATE_OFFLINE;
zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2008-11-20 23:01:55 +03:00
return (0);
switch (errno) {
case EBUSY:
/*
* There are no other replicas of this device.
*/
return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2009-07-03 02:44:48 +04:00
case EEXIST:
/*
* The log device has unplayed logs
*/
return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2008-11-20 23:01:55 +03:00
default:
return (zpool_standard_error(hdl, errno, msg));
}
}
/*
* Mark the given vdev faulted.
*/
int
zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2008-11-20 23:01:55 +03:00
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);
2008-11-20 23:01:55 +03:00
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
zc.zc_guid = guid;
zc.zc_cookie = VDEV_STATE_FAULTED;
zc.zc_obj = aux;
2008-11-20 23:01:55 +03:00
if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2008-11-20 23:01:55 +03:00
return (0);
switch (errno) {
case EBUSY:
/*
* There are no other replicas of this device.
*/
return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
default:
return (zpool_standard_error(hdl, errno, msg));
}
}
/*
* Mark the given vdev degraded.
*/
int
zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2008-11-20 23:01:55 +03:00
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot degrade %llu"), (u_longlong_t)guid);
2008-11-20 23:01:55 +03:00
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
zc.zc_guid = guid;
zc.zc_cookie = VDEV_STATE_DEGRADED;
zc.zc_obj = aux;
2008-11-20 23:01:55 +03:00
if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2008-11-20 23:01:55 +03:00
return (0);
return (zpool_standard_error(hdl, errno, msg));
}
/*
* Returns TRUE if the given nvlist is a vdev that was originally swapped in as
* a hot spare.
*/
static boolean_t
is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
{
nvlist_t **child;
uint_t c, children;
char *type;
if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
&children) == 0) {
verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
&type) == 0);
if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
children == 2 && child[which] == tgt)
return (B_TRUE);
for (c = 0; c < children; c++)
if (is_replacing_spare(child[c], tgt, which))
return (B_TRUE);
}
return (B_FALSE);
}
/*
* Attach new_disk (fully described by nvroot) to old_disk.
* If 'replacing' is specified, the new disk will replace the old one.
*/
int
zpool_vdev_attach(zpool_handle_t *zhp,
const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
int ret;
nvlist_t *tgt;
boolean_t avail_spare, l2cache, islog;
uint64_t val;
char *newname;
2008-11-20 23:01:55 +03:00
nvlist_t **child;
uint_t children;
nvlist_t *config_root;
libzfs_handle_t *hdl = zhp->zpool_hdl;
boolean_t rootpool = zpool_is_bootable(zhp);
2008-11-20 23:01:55 +03:00
if (replacing)
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot replace %s with %s"), old_disk, new_disk);
else
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot attach %s to %s"), new_disk, old_disk);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
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
&islog)) == NULL)
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_NODEVICE, msg));
if (avail_spare)
return (zfs_error(hdl, EZFS_ISSPARE, msg));
if (l2cache)
return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
zc.zc_cookie = replacing;
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0 || children != 1) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"new device must be a single disk"));
return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
}
verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL)
return (-1);
2008-11-20 23:01:55 +03:00
/*
* If the target is a hot spare that has been swapped in, we can only
* replace it with another hot spare.
*/
if (replacing &&
nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
(zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
NULL) == NULL || !avail_spare) &&
is_replacing_spare(config_root, tgt, 1)) {
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"can only be replaced by another hot spare"));
free(newname);
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_BADTARGET, msg));
}
free(newname);
2008-11-20 23:01:55 +03:00
if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
return (-1);
ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2008-11-20 23:01:55 +03:00
zcmd_free_nvlists(&zc);
if (ret == 0) {
if (rootpool) {
2009-07-03 02:44:48 +04:00
/*
* XXX need a better way to prevent user from
* booting up a half-baked vdev.
*/
(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
"sure to wait until resilver is done "
"before rebooting.\n"));
}
2008-11-20 23:01:55 +03:00
return (0);
}
2008-11-20 23:01:55 +03:00
switch (errno) {
case ENOTSUP:
/*
* Can't attach to or replace this type of vdev.
*/
if (replacing) {
uint64_t version = zpool_get_prop_int(zhp,
ZPOOL_PROP_VERSION, NULL);
if (islog)
2008-11-20 23:01:55 +03:00
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"cannot replace a log with a spare"));
else if (version >= SPA_VERSION_MULTI_REPLACE)
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"already in replacing/spare config; wait "
"for completion or use 'zpool detach'"));
2008-11-20 23:01:55 +03:00
else
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"cannot replace a replacing device"));
} else {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"can only attach to mirrors and top-level "
"disks"));
}
(void) zfs_error(hdl, EZFS_BADTARGET, msg);
break;
case EINVAL:
/*
* The new device must be a single disk.
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"new device must be a single disk"));
(void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
break;
case EBUSY:
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
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, "
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
"or device removal is in progress"),
2008-11-20 23:01:55 +03:00
new_disk);
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
case EOVERFLOW:
/*
* The new device is too small.
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"device is too small"));
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
case EDOM:
/*
* The new device has a different optimal sector size.
2008-11-20 23:01:55 +03:00
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"new device has a different optimal sector size; use the "
"option '-o ashift=N' to override the optimal size"));
2008-11-20 23:01:55 +03:00
(void) zfs_error(hdl, EZFS_BADDEV, msg);
break;
case ENAMETOOLONG:
/*
* The resulting top-level vdev spec won't fit in the label.
*/
(void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
break;
default:
(void) zpool_standard_error(hdl, errno, msg);
}
return (-1);
}
/*
* Detach the specified device.
*/
int
zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
nvlist_t *tgt;
boolean_t avail_spare, l2cache;
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
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
NULL)) == NULL)
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_NODEVICE, msg));
if (avail_spare)
return (zfs_error(hdl, EZFS_ISSPARE, msg));
if (l2cache)
return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
return (0);
switch (errno) {
case ENOTSUP:
/*
* Can't detach from this type of vdev.
*/
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
"applicable to mirror and replacing vdevs"));
(void) zfs_error(hdl, EZFS_BADTARGET, msg);
2008-11-20 23:01:55 +03:00
break;
case EBUSY:
/*
* There are no other replicas of this device.
*/
(void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
break;
default:
(void) zpool_standard_error(hdl, errno, msg);
}
return (-1);
}
/*
* Find a mirror vdev in the source nvlist.
*
* The mchild array contains a list of disks in one of the top-level mirrors
* of the source pool. The schild array contains a list of disks that the
* user specified on the command line. We loop over the mchild array to
* see if any entry in the schild array matches.
*
* If a disk in the mchild array is found in the schild array, we return
* the index of that entry. Otherwise we return -1.
*/
static int
find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
nvlist_t **schild, uint_t schildren)
{
uint_t mc;
for (mc = 0; mc < mchildren; mc++) {
uint_t sc;
char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
mchild[mc], 0);
for (sc = 0; sc < schildren; sc++) {
char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
schild[sc], 0);
boolean_t result = (strcmp(mpath, spath) == 0);
free(spath);
if (result) {
free(mpath);
return (mc);
}
}
free(mpath);
}
return (-1);
}
/*
* Split a mirror pool. If newroot points to null, then a new nvlist
* is generated and it is the responsibility of the caller to free it.
*/
int
zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
nvlist_t *props, splitflags_t flags)
{
zfs_cmd_t zc = {"\0"};
char msg[1024];
nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
nvlist_t **varray = NULL, *zc_props = NULL;
uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
libzfs_handle_t *hdl = zhp->zpool_hdl;
uint64_t vers, readonly = B_FALSE;
boolean_t freelist = B_FALSE, memory_err = B_TRUE;
int retval = 0;
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
if (!zpool_name_valid(hdl, B_FALSE, newname))
return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
if ((config = zpool_get_config(zhp, NULL)) == NULL) {
(void) fprintf(stderr, gettext("Internal error: unable to "
"retrieve pool configuration\n"));
return (-1);
}
verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
== 0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
if (props) {
prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
props, vers, flags, msg)) == NULL)
return (-1);
(void) nvlist_lookup_uint64(zc_props,
zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
if (readonly) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property %s can only be set at import time"),
zpool_prop_to_name(ZPOOL_PROP_READONLY));
return (-1);
}
}
if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
&children) != 0) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"Source pool is missing vdev tree"));
nvlist_free(zc_props);
return (-1);
}
varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
vcount = 0;
if (*newroot == NULL ||
nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
&newchild, &newchildren) != 0)
newchildren = 0;
for (c = 0; c < children; c++) {
uint64_t is_log = B_FALSE, is_hole = B_FALSE;
char *type;
nvlist_t **mchild, *vdev;
uint_t mchildren;
int entry;
/*
* Unlike cache & spares, slogs are stored in the
* ZPOOL_CONFIG_CHILDREN array. We filter them out here.
*/
(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
&is_log);
(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
&is_hole);
if (is_log || is_hole) {
/*
* Create a hole vdev and put it in the config.
*/
if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
goto out;
if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
VDEV_TYPE_HOLE) != 0)
goto out;
if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
1) != 0)
goto out;
if (lastlog == 0)
lastlog = vcount;
varray[vcount++] = vdev;
continue;
}
lastlog = 0;
verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
== 0);
if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"Source pool must be composed only of mirrors\n"));
retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
goto out;
}
verify(nvlist_lookup_nvlist_array(child[c],
ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
/* find or add an entry for this top-level vdev */
if (newchildren > 0 &&
(entry = find_vdev_entry(zhp, mchild, mchildren,
newchild, newchildren)) >= 0) {
/* We found a disk that the user specified. */
vdev = mchild[entry];
++found;
} else {
/* User didn't specify a disk for this vdev. */
vdev = mchild[mchildren - 1];
}
if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
goto out;
}
/* did we find every disk the user specified? */
if (found != newchildren) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
"include at most one disk from each mirror"));
retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
goto out;
}
/* Prepare the nvlist for populating. */
if (*newroot == NULL) {
if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
goto out;
freelist = B_TRUE;
if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
VDEV_TYPE_ROOT) != 0)
goto out;
} else {
verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
}
/* Add all the children we found */
if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
lastlog == 0 ? vcount : lastlog) != 0)
goto out;
/*
* If we're just doing a dry run, exit now with success.
*/
if (flags.dryrun) {
memory_err = B_FALSE;
freelist = B_FALSE;
goto out;
}
/* now build up the config list & call the ioctl */
if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
goto out;
if (nvlist_add_nvlist(newconfig,
ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
nvlist_add_string(newconfig,
ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
goto out;
/*
* The new pool is automatically part of the namespace unless we
* explicitly export it.
*/
if (!flags.import)
zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
(void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
goto out;
if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
goto out;
if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
retval = zpool_standard_error(hdl, errno, msg);
goto out;
}
freelist = B_FALSE;
memory_err = B_FALSE;
out:
if (varray != NULL) {
int v;
for (v = 0; v < vcount; v++)
nvlist_free(varray[v]);
free(varray);
}
zcmd_free_nvlists(&zc);
nvlist_free(zc_props);
nvlist_free(newconfig);
if (freelist) {
nvlist_free(*newroot);
*newroot = NULL;
}
if (retval != 0)
return (retval);
if (memory_err)
return (no_memory(hdl));
return (0);
}
2008-11-20 23:01:55 +03:00
/*
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
* Remove the given device.
2008-11-20 23:01:55 +03:00
*/
int
zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
nvlist_t *tgt;
boolean_t avail_spare, l2cache, islog;
2008-11-20 23:01:55 +03:00
libzfs_handle_t *hdl = zhp->zpool_hdl;
uint64_t version;
2008-11-20 23:01:55 +03:00
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
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
&islog)) == NULL)
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_NODEVICE, msg));
version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
if (islog && version < SPA_VERSION_HOLES) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
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
"pool must be upgraded to support log removal"));
return (zfs_error(hdl, EZFS_BADVERSION, msg));
}
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
if (!islog && !avail_spare && !l2cache && zpool_is_bootable(zhp)) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"root pool can not have removed devices, "
"because GRUB does not understand them"));
return (zfs_error(hdl, EINVAL, msg));
}
zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
return (0);
switch (errno) {
case EINVAL:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"invalid config; all top-level vdevs must "
"have the same sector size and not be raidz."));
(void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
break;
case EBUSY:
if (islog) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"Mount encrypted datasets to replay logs."));
} else {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"Pool busy; removal may already be in progress"));
}
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
(void) zfs_error(hdl, EZFS_BUSY, msg);
break;
case EACCES:
if (islog) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"Mount encrypted datasets to replay logs."));
(void) zfs_error(hdl, EZFS_BUSY, msg);
} else {
(void) zpool_standard_error(hdl, errno, msg);
}
break;
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
default:
(void) zpool_standard_error(hdl, errno, msg);
}
return (-1);
}
int
zpool_vdev_remove_cancel(zpool_handle_t *zhp)
{
zfs_cmd_t zc;
char msg[1024];
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot cancel removal"));
bzero(&zc, sizeof (zc));
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
zc.zc_cookie = 1;
2008-11-20 23:01:55 +03:00
if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
return (0);
return (zpool_standard_error(hdl, errno, msg));
}
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
int
zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path,
uint64_t *sizep)
{
char msg[1024];
nvlist_t *tgt;
boolean_t avail_spare, l2cache, islog;
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"),
path);
if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
&islog)) == NULL)
return (zfs_error(hdl, EZFS_NODEVICE, msg));
if (avail_spare || l2cache || islog) {
*sizep = 0;
return (0);
}
if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"indirect size not available"));
return (zfs_error(hdl, EINVAL, msg));
}
return (0);
}
2008-11-20 23:01:55 +03:00
/*
* Clear the errors for the pool, or the particular device if specified.
*/
int
zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
2008-11-20 23:01:55 +03:00
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
nvlist_t *tgt;
zpool_load_policy_t policy;
2008-11-20 23:01:55 +03:00
boolean_t avail_spare, l2cache;
libzfs_handle_t *hdl = zhp->zpool_hdl;
nvlist_t *nvi = NULL;
int error;
2008-11-20 23:01:55 +03:00
if (path)
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
path);
else
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
zhp->zpool_name);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if (path) {
if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
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
&l2cache, NULL)) == NULL)
2008-11-20 23:01:55 +03:00
return (zfs_error(hdl, EZFS_NODEVICE, msg));
/*
* Don't allow error clearing for hot spares. Do allow
* error clearing for l2cache devices.
*/
if (avail_spare)
return (zfs_error(hdl, EZFS_ISSPARE, msg));
verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
&zc.zc_guid) == 0);
}
zpool_get_load_policy(rewindnvl, &policy);
zc.zc_cookie = policy.zlp_rewind;
if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
return (-1);
if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
return (-1);
while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
errno == ENOMEM) {
if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
zcmd_free_nvlists(&zc);
return (-1);
}
}
if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) &&
errno != EPERM && errno != EACCES)) {
if (policy.zlp_rewind &
(ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
(void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
zpool_rewind_exclaim(hdl, zc.zc_name,
((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0),
nvi);
nvlist_free(nvi);
}
zcmd_free_nvlists(&zc);
2008-11-20 23:01:55 +03:00
return (0);
}
2008-11-20 23:01:55 +03:00
zcmd_free_nvlists(&zc);
2008-11-20 23:01:55 +03:00
return (zpool_standard_error(hdl, errno, msg));
}
/*
* Similar to zpool_clear(), but takes a GUID (used by fmd).
*/
int
zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
char msg[1024];
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
(u_longlong_t)guid);
2008-11-20 23:01:55 +03:00
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
zc.zc_guid = guid;
zc.zc_cookie = ZPOOL_NO_REWIND;
2008-11-20 23:01:55 +03:00
if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
return (0);
return (zpool_standard_error(hdl, errno, msg));
}
/*
* Change the GUID for a pool.
*/
int
zpool_reguid(zpool_handle_t *zhp)
{
char msg[1024];
libzfs_handle_t *hdl = zhp->zpool_hdl;
zfs_cmd_t zc = {"\0"};
(void) snprintf(msg, sizeof (msg),
dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
return (0);
return (zpool_standard_error(hdl, errno, msg));
}
/*
* Reopen the pool.
*/
int
zpool_reopen_one(zpool_handle_t *zhp, void *data)
{
libzfs_handle_t *hdl = zpool_get_handle(zhp);
const char *pool_name = zpool_get_name(zhp);
boolean_t *scrub_restart = data;
int error;
error = lzc_reopen(pool_name, *scrub_restart);
if (error) {
return (zpool_standard_error_fmt(hdl, error,
dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), pool_name));
}
return (0);
}
/* call into libzfs_core to execute the sync IOCTL per pool */
int
zpool_sync_one(zpool_handle_t *zhp, void *data)
{
int ret;
libzfs_handle_t *hdl = zpool_get_handle(zhp);
const char *pool_name = zpool_get_name(zhp);
boolean_t *force = data;
nvlist_t *innvl = fnvlist_alloc();
fnvlist_add_boolean_value(innvl, "force", *force);
if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) {
nvlist_free(innvl);
return (zpool_standard_error_fmt(hdl, ret,
dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name));
}
nvlist_free(innvl);
return (0);
}
#define PATH_BUF_LEN 64
2008-11-20 23:01:55 +03:00
/*
* Given a vdev, return the name to display in iostat. If the vdev has a path,
* we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
* We also check if this is a whole disk, in which case we strip off the
* trailing 's0' slice name.
*
* This routine is also responsible for identifying when disks have been
* reconfigured in a new location. The kernel will have opened the device by
* devid, but the path will still refer to the old location. To catch this, we
* first do a path -> devid translation (which is fast for the common case). If
* the devid matches, we're done. If not, we do a reverse devid -> path
* translation and issue the appropriate ioctl() to update the path of the vdev.
* If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
* of these checks.
*/
char *
zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
int name_flags)
2008-11-20 23:01:55 +03:00
{
char *path, *type, *env;
2008-11-20 23:01:55 +03:00
uint64_t value;
char buf[PATH_BUF_LEN];
char tmpbuf[PATH_BUF_LEN];
2008-11-20 23:01:55 +03:00
/*
* vdev_name will be "root"/"root-0" for the root vdev, but it is the
* zpool name that will be displayed to the user.
*/
verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
if (zhp != NULL && strcmp(type, "root") == 0)
return (zfs_strdup(hdl, zpool_get_name(zhp)));
env = getenv("ZPOOL_VDEV_NAME_PATH");
if (env && (strtoul(env, NULL, 0) > 0 ||
!strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
name_flags |= VDEV_NAME_PATH;
env = getenv("ZPOOL_VDEV_NAME_GUID");
if (env && (strtoul(env, NULL, 0) > 0 ||
!strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
name_flags |= VDEV_NAME_GUID;
env = getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS");
if (env && (strtoul(env, NULL, 0) > 0 ||
!strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
name_flags |= VDEV_NAME_FOLLOW_LINKS;
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
name_flags & VDEV_NAME_GUID) {
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value);
(void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value);
2008-11-20 23:01:55 +03:00
path = buf;
} else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
char *rp = realpath(path, NULL);
if (rp) {
strlcpy(buf, rp, sizeof (buf));
path = buf;
free(rp);
}
}
/*
* For a block device only use the name.
*/
if ((strcmp(type, VDEV_TYPE_DISK) == 0) &&
!(name_flags & VDEV_NAME_PATH)) {
path = strrchr(path, '/');
path++;
}
2008-11-20 23:01:55 +03:00
/*
* Remove the partition from the path it this is a whole disk.
*/
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value)
== 0 && value && !(name_flags & VDEV_NAME_PATH)) {
return (zfs_strip_partition(path));
2008-11-20 23:01:55 +03:00
}
} else {
path = type;
2008-11-20 23:01:55 +03:00
/*
* If it's a raidz device, we need to stick in the parity level.
*/
if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
&value) == 0);
(void) snprintf(buf, sizeof (buf), "%s%llu", path,
2008-11-20 23:01:55 +03:00
(u_longlong_t)value);
path = buf;
2008-11-20 23:01:55 +03:00
}
/*
* We identify each top-level vdev by using a <type-id>
* naming convention.
*/
if (name_flags & VDEV_NAME_TYPE_ID) {
uint64_t id;
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
&id) == 0);
(void) snprintf(tmpbuf, sizeof (tmpbuf), "%s-%llu",
path, (u_longlong_t)id);
path = tmpbuf;
}
2008-11-20 23:01:55 +03:00
}
return (zfs_strdup(hdl, path));
}
static int
zbookmark_mem_compare(const void *a, const void *b)
2008-11-20 23:01:55 +03:00
{
return (memcmp(a, b, sizeof (zbookmark_phys_t)));
2008-11-20 23:01:55 +03:00
}
/*
* Retrieve the persistent error log, uniquify the members, and return to the
* caller.
*/
int
zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
{
zfs_cmd_t zc = {"\0"};
libzfs_handle_t *hdl = zhp->zpool_hdl;
2008-11-20 23:01:55 +03:00
uint64_t count;
zbookmark_phys_t *zb = NULL;
2008-11-20 23:01:55 +03:00
int i;
/*
* Retrieve the raw error list from the kernel. If the number of errors
* has increased, allocate more space and continue until we get the
* entire list.
*/
verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
&count) == 0);
if (count == 0)
return (0);
zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
count * sizeof (zbookmark_phys_t));
2008-11-20 23:01:55 +03:00
zc.zc_nvlist_dst_size = count;
(void) strcpy(zc.zc_name, zhp->zpool_name);
for (;;) {
if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
&zc) != 0) {
free((void *)(uintptr_t)zc.zc_nvlist_dst);
if (errno == ENOMEM) {
void *dst;
2008-11-20 23:01:55 +03:00
count = zc.zc_nvlist_dst_size;
dst = zfs_alloc(zhp->zpool_hdl, count *
sizeof (zbookmark_phys_t));
zc.zc_nvlist_dst = (uintptr_t)dst;
2008-11-20 23:01:55 +03:00
} else {
return (zpool_standard_error_fmt(hdl, errno,
dgettext(TEXT_DOMAIN, "errors: List of "
"errors unavailable")));
2008-11-20 23:01:55 +03:00
}
} else {
break;
}
}
/*
* Sort the resulting bookmarks. This is a little confusing due to the
* implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
* to first, and 'zc_nvlist_dst_size' indicates the number of bookmarks
2008-11-20 23:01:55 +03:00
* _not_ copied as part of the process. So we point the start of our
* array appropriate and decrement the total number of elements.
*/
zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
2008-11-20 23:01:55 +03:00
zc.zc_nvlist_dst_size;
count -= zc.zc_nvlist_dst_size;
qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
2008-11-20 23:01:55 +03:00
verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
/*
* Fill in the nverrlistp with nvlist's of dataset and object numbers.
*/
for (i = 0; i < count; i++) {
nvlist_t *nv;
/* ignoring zb_blkid and zb_level for now */
if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
zb[i-1].zb_object == zb[i].zb_object)
continue;
if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
goto nomem;
if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
zb[i].zb_objset) != 0) {
nvlist_free(nv);
goto nomem;
}
if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
zb[i].zb_object) != 0) {
nvlist_free(nv);
goto nomem;
}
if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
nvlist_free(nv);
goto nomem;
}
nvlist_free(nv);
}
free((void *)(uintptr_t)zc.zc_nvlist_dst);
return (0);
nomem:
free((void *)(uintptr_t)zc.zc_nvlist_dst);
return (no_memory(zhp->zpool_hdl));
}
/*
* Upgrade a ZFS pool to the latest on-disk version.
*/
int
zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) strcpy(zc.zc_name, zhp->zpool_name);
zc.zc_cookie = new_version;
if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
return (zpool_standard_error_fmt(hdl, errno,
dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
zhp->zpool_name));
return (0);
}
void
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
zfs_save_arguments(int argc, char **argv, char *string, int len)
2008-11-20 23:01:55 +03:00
{
int i;
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
(void) strlcpy(string, basename(argv[0]), len);
2008-11-20 23:01:55 +03:00
for (i = 1; i < argc; i++) {
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
(void) strlcat(string, " ", len);
(void) strlcat(string, argv[i], len);
2008-11-20 23:01:55 +03:00
}
}
int
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
zpool_log_history(libzfs_handle_t *hdl, const char *message)
{
zfs_cmd_t zc = {"\0"};
Illumos #2882, #2883, #2900 2882 implement libzfs_core 2883 changing "canmount" property to "on" should not always remount dataset 2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Chris Siden <christopher.siden@delphix.com> Reviewed by: Garrett D'Amore <garrett@damore.org> Reviewed by: Bill Pijewski <wdp@joyent.com> Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com> Approved by: Eric Schrock <Eric.Schrock@delphix.com> References: https://www.illumos.org/issues/2882 https://www.illumos.org/issues/2883 https://www.illumos.org/issues/2900 illumos/illumos-gate@4445fffbbb1ea25fd0e9ea68b9380dd7a6709025 Ported-by: Tim Chase <tim@chase2k.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1293 Porting notes: WARNING: This patch changes the user/kernel ABI. That means that the zfs/zpool utilities built from master are NOT compatible with the 0.6.2 kernel modules. Ensure you load the matching kernel modules from master after updating the utilities. Otherwise the zfs/zpool commands will be unable to interact with your pool and you will see errors similar to the following: $ zpool list failed to read pool configuration: bad address no pools available $ zfs list no datasets available Add zvol minor device creation to the new zfs_snapshot_nvl function. Remove the logging of the "release" operation in dsl_dataset_user_release_sync(). The logging caused a null dereference because ds->ds_dir is zeroed in dsl_dataset_destroy_sync() and the logging functions try to get the ds name via the dsl_dataset_name() function. I've got no idea why this particular code would have worked in Illumos. This code has subsequently been completely reworked in Illumos commit 3b2aab1 (3464 zfs synctask code needs restructuring). Squash some "may be used uninitialized" warning/erorrs. Fix some printf format warnings for %lld and %llu. Apply a few spa_writeable() changes that were made to Illumos in illumos/illumos-gate.git@cd1c8b8 as part of the 3112, 3113, 3114 and 3115 fixes. Add a missing call to fnvlist_free(nvl) in log_internal() that was added in Illumos to fix issue 3085 but couldn't be ported to ZoL at the time (zfsonlinux/zfs@9e11c73) because it depended on future work.
2013-08-28 15:45:09 +04:00
nvlist_t *args;
int err;
args = fnvlist_alloc();
fnvlist_add_string(args, "message", message);
err = zcmd_write_src_nvlist(hdl, &zc, args);
if (err == 0)
err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
nvlist_free(args);
zcmd_free_nvlists(&zc);
return (err);
2008-11-20 23:01:55 +03:00
}
/*
* Perform ioctl to get some command history of a pool.
*
* 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
* logical offset of the history buffer to start reading from.
*
* Upon return, 'off' is the next logical offset to read from and
* 'len' is the actual amount of bytes read into 'buf'.
*/
static int
get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
libzfs_handle_t *hdl = zhp->zpool_hdl;
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
zc.zc_history = (uint64_t)(uintptr_t)buf;
zc.zc_history_len = *len;
zc.zc_history_offset = *off;
if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
switch (errno) {
case EPERM:
return (zfs_error_fmt(hdl, EZFS_PERM,
dgettext(TEXT_DOMAIN,
"cannot show history for pool '%s'"),
zhp->zpool_name));
case ENOENT:
return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
dgettext(TEXT_DOMAIN, "cannot get history for pool "
"'%s'"), zhp->zpool_name));
case ENOTSUP:
return (zfs_error_fmt(hdl, EZFS_BADVERSION,
dgettext(TEXT_DOMAIN, "cannot get history for pool "
"'%s', pool must be upgraded"), zhp->zpool_name));
default:
return (zpool_standard_error_fmt(hdl, errno,
dgettext(TEXT_DOMAIN,
"cannot get history for '%s'"), zhp->zpool_name));
}
}
*len = zc.zc_history_len;
*off = zc.zc_history_offset;
return (0);
}
/*
* Retrieve the command history of a pool.
*/
int
zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
{
char *buf;
int buflen = 128 * 1024;
2008-11-20 23:01:55 +03:00
uint64_t off = 0;
nvlist_t **records = NULL;
uint_t numrecords = 0;
int err, i;
buf = malloc(buflen);
if (buf == NULL)
return (ENOMEM);
2008-11-20 23:01:55 +03:00
do {
uint64_t bytes_read = buflen;
2008-11-20 23:01:55 +03:00
uint64_t leftover;
if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
break;
/* if nothing else was read in, we're at EOF, just return */
if (!bytes_read)
break;
if ((err = zpool_history_unpack(buf, bytes_read,
&leftover, &records, &numrecords)) != 0)
break;
off -= leftover;
if (leftover == bytes_read) {
/*
* no progress made, because buffer is not big enough
* to hold this record; resize and retry.
*/
buflen *= 2;
free(buf);
buf = malloc(buflen);
if (buf == NULL)
return (ENOMEM);
}
2008-11-20 23:01:55 +03:00
/* CONSTCOND */
} while (1);
free(buf);
2008-11-20 23:01:55 +03:00
if (!err) {
verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
records, numrecords) == 0);
}
for (i = 0; i < numrecords; i++)
nvlist_free(records[i]);
free(records);
return (err);
}
Add linux events This topic branch leverages the Solaris style FMA call points in ZFS to create a user space visible event notification system under Linux. This new system is called zevent and it unifies all previous Solaris style ereports and sysevent notifications. Under this Linux specific scheme when a sysevent or ereport event occurs an nvlist describing the event is created which looks almost exactly like a Solaris ereport. These events are queued up in the kernel when they occur and conditionally logged to the console. It is then up to a user space application to consume the events and do whatever it likes with them. To make this possible the existing /dev/zfs ABI has been extended with two new ioctls which behave as follows. * ZFS_IOC_EVENTS_NEXT Get the next pending event. The kernel will keep track of the last event consumed by the file descriptor and provide the next one if available. If no new events are available the ioctl() will block waiting for the next event. This ioctl may also be called in a non-blocking mode by setting zc.zc_guid = ZEVENT_NONBLOCK. In the non-blocking case if no events are available ENOENT will be returned. It is possible that ESHUTDOWN will be returned if the ioctl() is called while module unloading is in progress. And finally ENOMEM may occur if the provided nvlist buffer is not large enough to contain the entire event. * ZFS_IOC_EVENTS_CLEAR Clear are events queued by the kernel. The kernel will keep a fairly large number of recent events queued, use this ioctl to clear the in kernel list. This will effect all user space processes consuming events. The zpool command has been extended to use this events ABI with the 'events' subcommand. You may run 'zpool events -v' to output a verbose log of all recent events. This is very similar to the Solaris 'fmdump -ev' command with the key difference being it also includes what would be considered sysevents under Solaris. You may also run in follow mode with the '-f' option. To clear the in kernel event queue use the '-c' option. $ sudo cmd/zpool/zpool events -fv TIME CLASS May 13 2010 16:31:15.777711000 ereport.fs.zfs.config.sync class = "ereport.fs.zfs.config.sync" ena = 0x40982b7897700001 detector = (embedded nvlist) version = 0x0 scheme = "zfs" pool = 0xed976600de75dfa6 (end detector) time = 0x4bec8bc3 0x2e5aed98 pool = "zpios" pool_guid = 0xed976600de75dfa6 pool_context = 0x0 While the 'zpool events' command is handy for interactive debugging it is not expected to be the primary consumer of zevents. This ABI was primarily added to facilitate the addition of a user space monitoring daemon. This daemon would consume all events posted by the kernel and based on the type of event perform an action. For most events simply forwarding them on to syslog is likely enough. But this interface also cleanly allows for more sophisticated actions to be taken such as generating an email for a failed drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2010-08-26 22:42:43 +04:00
/*
* Retrieve the next event given the passed 'zevent_fd' file descriptor.
* If there is a new event available 'nvp' will contain a newly allocated
* nvlist and 'dropped' will be set to the number of missed events since
* the last call to this function. When 'nvp' is set to NULL it indicates
* no new events are available. In either case the function returns 0 and
* it is up to the caller to free 'nvp'. In the case of a fatal error the
* function will return a non-zero value. When the function is called in
* blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
* it will not return until a new event is available.
Add linux events This topic branch leverages the Solaris style FMA call points in ZFS to create a user space visible event notification system under Linux. This new system is called zevent and it unifies all previous Solaris style ereports and sysevent notifications. Under this Linux specific scheme when a sysevent or ereport event occurs an nvlist describing the event is created which looks almost exactly like a Solaris ereport. These events are queued up in the kernel when they occur and conditionally logged to the console. It is then up to a user space application to consume the events and do whatever it likes with them. To make this possible the existing /dev/zfs ABI has been extended with two new ioctls which behave as follows. * ZFS_IOC_EVENTS_NEXT Get the next pending event. The kernel will keep track of the last event consumed by the file descriptor and provide the next one if available. If no new events are available the ioctl() will block waiting for the next event. This ioctl may also be called in a non-blocking mode by setting zc.zc_guid = ZEVENT_NONBLOCK. In the non-blocking case if no events are available ENOENT will be returned. It is possible that ESHUTDOWN will be returned if the ioctl() is called while module unloading is in progress. And finally ENOMEM may occur if the provided nvlist buffer is not large enough to contain the entire event. * ZFS_IOC_EVENTS_CLEAR Clear are events queued by the kernel. The kernel will keep a fairly large number of recent events queued, use this ioctl to clear the in kernel list. This will effect all user space processes consuming events. The zpool command has been extended to use this events ABI with the 'events' subcommand. You may run 'zpool events -v' to output a verbose log of all recent events. This is very similar to the Solaris 'fmdump -ev' command with the key difference being it also includes what would be considered sysevents under Solaris. You may also run in follow mode with the '-f' option. To clear the in kernel event queue use the '-c' option. $ sudo cmd/zpool/zpool events -fv TIME CLASS May 13 2010 16:31:15.777711000 ereport.fs.zfs.config.sync class = "ereport.fs.zfs.config.sync" ena = 0x40982b7897700001 detector = (embedded nvlist) version = 0x0 scheme = "zfs" pool = 0xed976600de75dfa6 (end detector) time = 0x4bec8bc3 0x2e5aed98 pool = "zpios" pool_guid = 0xed976600de75dfa6 pool_context = 0x0 While the 'zpool events' command is handy for interactive debugging it is not expected to be the primary consumer of zevents. This ABI was primarily added to facilitate the addition of a user space monitoring daemon. This daemon would consume all events posted by the kernel and based on the type of event perform an action. For most events simply forwarding them on to syslog is likely enough. But this interface also cleanly allows for more sophisticated actions to be taken such as generating an email for a failed drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2010-08-26 22:42:43 +04:00
*/
int
zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp,
int *dropped, unsigned flags, int zevent_fd)
Add linux events This topic branch leverages the Solaris style FMA call points in ZFS to create a user space visible event notification system under Linux. This new system is called zevent and it unifies all previous Solaris style ereports and sysevent notifications. Under this Linux specific scheme when a sysevent or ereport event occurs an nvlist describing the event is created which looks almost exactly like a Solaris ereport. These events are queued up in the kernel when they occur and conditionally logged to the console. It is then up to a user space application to consume the events and do whatever it likes with them. To make this possible the existing /dev/zfs ABI has been extended with two new ioctls which behave as follows. * ZFS_IOC_EVENTS_NEXT Get the next pending event. The kernel will keep track of the last event consumed by the file descriptor and provide the next one if available. If no new events are available the ioctl() will block waiting for the next event. This ioctl may also be called in a non-blocking mode by setting zc.zc_guid = ZEVENT_NONBLOCK. In the non-blocking case if no events are available ENOENT will be returned. It is possible that ESHUTDOWN will be returned if the ioctl() is called while module unloading is in progress. And finally ENOMEM may occur if the provided nvlist buffer is not large enough to contain the entire event. * ZFS_IOC_EVENTS_CLEAR Clear are events queued by the kernel. The kernel will keep a fairly large number of recent events queued, use this ioctl to clear the in kernel list. This will effect all user space processes consuming events. The zpool command has been extended to use this events ABI with the 'events' subcommand. You may run 'zpool events -v' to output a verbose log of all recent events. This is very similar to the Solaris 'fmdump -ev' command with the key difference being it also includes what would be considered sysevents under Solaris. You may also run in follow mode with the '-f' option. To clear the in kernel event queue use the '-c' option. $ sudo cmd/zpool/zpool events -fv TIME CLASS May 13 2010 16:31:15.777711000 ereport.fs.zfs.config.sync class = "ereport.fs.zfs.config.sync" ena = 0x40982b7897700001 detector = (embedded nvlist) version = 0x0 scheme = "zfs" pool = 0xed976600de75dfa6 (end detector) time = 0x4bec8bc3 0x2e5aed98 pool = "zpios" pool_guid = 0xed976600de75dfa6 pool_context = 0x0 While the 'zpool events' command is handy for interactive debugging it is not expected to be the primary consumer of zevents. This ABI was primarily added to facilitate the addition of a user space monitoring daemon. This daemon would consume all events posted by the kernel and based on the type of event perform an action. For most events simply forwarding them on to syslog is likely enough. But this interface also cleanly allows for more sophisticated actions to be taken such as generating an email for a failed drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2010-08-26 22:42:43 +04:00
{
zfs_cmd_t zc = {"\0"};
Add linux events This topic branch leverages the Solaris style FMA call points in ZFS to create a user space visible event notification system under Linux. This new system is called zevent and it unifies all previous Solaris style ereports and sysevent notifications. Under this Linux specific scheme when a sysevent or ereport event occurs an nvlist describing the event is created which looks almost exactly like a Solaris ereport. These events are queued up in the kernel when they occur and conditionally logged to the console. It is then up to a user space application to consume the events and do whatever it likes with them. To make this possible the existing /dev/zfs ABI has been extended with two new ioctls which behave as follows. * ZFS_IOC_EVENTS_NEXT Get the next pending event. The kernel will keep track of the last event consumed by the file descriptor and provide the next one if available. If no new events are available the ioctl() will block waiting for the next event. This ioctl may also be called in a non-blocking mode by setting zc.zc_guid = ZEVENT_NONBLOCK. In the non-blocking case if no events are available ENOENT will be returned. It is possible that ESHUTDOWN will be returned if the ioctl() is called while module unloading is in progress. And finally ENOMEM may occur if the provided nvlist buffer is not large enough to contain the entire event. * ZFS_IOC_EVENTS_CLEAR Clear are events queued by the kernel. The kernel will keep a fairly large number of recent events queued, use this ioctl to clear the in kernel list. This will effect all user space processes consuming events. The zpool command has been extended to use this events ABI with the 'events' subcommand. You may run 'zpool events -v' to output a verbose log of all recent events. This is very similar to the Solaris 'fmdump -ev' command with the key difference being it also includes what would be considered sysevents under Solaris. You may also run in follow mode with the '-f' option. To clear the in kernel event queue use the '-c' option. $ sudo cmd/zpool/zpool events -fv TIME CLASS May 13 2010 16:31:15.777711000 ereport.fs.zfs.config.sync class = "ereport.fs.zfs.config.sync" ena = 0x40982b7897700001 detector = (embedded nvlist) version = 0x0 scheme = "zfs" pool = 0xed976600de75dfa6 (end detector) time = 0x4bec8bc3 0x2e5aed98 pool = "zpios" pool_guid = 0xed976600de75dfa6 pool_context = 0x0 While the 'zpool events' command is handy for interactive debugging it is not expected to be the primary consumer of zevents. This ABI was primarily added to facilitate the addition of a user space monitoring daemon. This daemon would consume all events posted by the kernel and based on the type of event perform an action. For most events simply forwarding them on to syslog is likely enough. But this interface also cleanly allows for more sophisticated actions to be taken such as generating an email for a failed drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2010-08-26 22:42:43 +04:00
int error = 0;
*nvp = NULL;
*dropped = 0;
zc.zc_cleanup_fd = zevent_fd;
Add linux events This topic branch leverages the Solaris style FMA call points in ZFS to create a user space visible event notification system under Linux. This new system is called zevent and it unifies all previous Solaris style ereports and sysevent notifications. Under this Linux specific scheme when a sysevent or ereport event occurs an nvlist describing the event is created which looks almost exactly like a Solaris ereport. These events are queued up in the kernel when they occur and conditionally logged to the console. It is then up to a user space application to consume the events and do whatever it likes with them. To make this possible the existing /dev/zfs ABI has been extended with two new ioctls which behave as follows. * ZFS_IOC_EVENTS_NEXT Get the next pending event. The kernel will keep track of the last event consumed by the file descriptor and provide the next one if available. If no new events are available the ioctl() will block waiting for the next event. This ioctl may also be called in a non-blocking mode by setting zc.zc_guid = ZEVENT_NONBLOCK. In the non-blocking case if no events are available ENOENT will be returned. It is possible that ESHUTDOWN will be returned if the ioctl() is called while module unloading is in progress. And finally ENOMEM may occur if the provided nvlist buffer is not large enough to contain the entire event. * ZFS_IOC_EVENTS_CLEAR Clear are events queued by the kernel. The kernel will keep a fairly large number of recent events queued, use this ioctl to clear the in kernel list. This will effect all user space processes consuming events. The zpool command has been extended to use this events ABI with the 'events' subcommand. You may run 'zpool events -v' to output a verbose log of all recent events. This is very similar to the Solaris 'fmdump -ev' command with the key difference being it also includes what would be considered sysevents under Solaris. You may also run in follow mode with the '-f' option. To clear the in kernel event queue use the '-c' option. $ sudo cmd/zpool/zpool events -fv TIME CLASS May 13 2010 16:31:15.777711000 ereport.fs.zfs.config.sync class = "ereport.fs.zfs.config.sync" ena = 0x40982b7897700001 detector = (embedded nvlist) version = 0x0 scheme = "zfs" pool = 0xed976600de75dfa6 (end detector) time = 0x4bec8bc3 0x2e5aed98 pool = "zpios" pool_guid = 0xed976600de75dfa6 pool_context = 0x0 While the 'zpool events' command is handy for interactive debugging it is not expected to be the primary consumer of zevents. This ABI was primarily added to facilitate the addition of a user space monitoring daemon. This daemon would consume all events posted by the kernel and based on the type of event perform an action. For most events simply forwarding them on to syslog is likely enough. But this interface also cleanly allows for more sophisticated actions to be taken such as generating an email for a failed drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2010-08-26 22:42:43 +04:00
if (flags & ZEVENT_NONBLOCK)
Add linux events This topic branch leverages the Solaris style FMA call points in ZFS to create a user space visible event notification system under Linux. This new system is called zevent and it unifies all previous Solaris style ereports and sysevent notifications. Under this Linux specific scheme when a sysevent or ereport event occurs an nvlist describing the event is created which looks almost exactly like a Solaris ereport. These events are queued up in the kernel when they occur and conditionally logged to the console. It is then up to a user space application to consume the events and do whatever it likes with them. To make this possible the existing /dev/zfs ABI has been extended with two new ioctls which behave as follows. * ZFS_IOC_EVENTS_NEXT Get the next pending event. The kernel will keep track of the last event consumed by the file descriptor and provide the next one if available. If no new events are available the ioctl() will block waiting for the next event. This ioctl may also be called in a non-blocking mode by setting zc.zc_guid = ZEVENT_NONBLOCK. In the non-blocking case if no events are available ENOENT will be returned. It is possible that ESHUTDOWN will be returned if the ioctl() is called while module unloading is in progress. And finally ENOMEM may occur if the provided nvlist buffer is not large enough to contain the entire event. * ZFS_IOC_EVENTS_CLEAR Clear are events queued by the kernel. The kernel will keep a fairly large number of recent events queued, use this ioctl to clear the in kernel list. This will effect all user space processes consuming events. The zpool command has been extended to use this events ABI with the 'events' subcommand. You may run 'zpool events -v' to output a verbose log of all recent events. This is very similar to the Solaris 'fmdump -ev' command with the key difference being it also includes what would be considered sysevents under Solaris. You may also run in follow mode with the '-f' option. To clear the in kernel event queue use the '-c' option. $ sudo cmd/zpool/zpool events -fv TIME CLASS May 13 2010 16:31:15.777711000 ereport.fs.zfs.config.sync class = "ereport.fs.zfs.config.sync" ena = 0x40982b7897700001 detector = (embedded nvlist) version = 0x0 scheme = "zfs" pool = 0xed976600de75dfa6 (end detector) time = 0x4bec8bc3 0x2e5aed98 pool = "zpios" pool_guid = 0xed976600de75dfa6 pool_context = 0x0 While the 'zpool events' command is handy for interactive debugging it is not expected to be the primary consumer of zevents. This ABI was primarily added to facilitate the addition of a user space monitoring daemon. This daemon would consume all events posted by the kernel and based on the type of event perform an action. For most events simply forwarding them on to syslog is likely enough. But this interface also cleanly allows for more sophisticated actions to be taken such as generating an email for a failed drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2010-08-26 22:42:43 +04:00
zc.zc_guid = ZEVENT_NONBLOCK;
if (zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE) != 0)
return (-1);
retry:
if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
switch (errno) {
case ESHUTDOWN:
error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
dgettext(TEXT_DOMAIN, "zfs shutdown"));
goto out;
case ENOENT:
/* Blocking error case should not occur */
if (!(flags & ZEVENT_NONBLOCK))
Add linux events This topic branch leverages the Solaris style FMA call points in ZFS to create a user space visible event notification system under Linux. This new system is called zevent and it unifies all previous Solaris style ereports and sysevent notifications. Under this Linux specific scheme when a sysevent or ereport event occurs an nvlist describing the event is created which looks almost exactly like a Solaris ereport. These events are queued up in the kernel when they occur and conditionally logged to the console. It is then up to a user space application to consume the events and do whatever it likes with them. To make this possible the existing /dev/zfs ABI has been extended with two new ioctls which behave as follows. * ZFS_IOC_EVENTS_NEXT Get the next pending event. The kernel will keep track of the last event consumed by the file descriptor and provide the next one if available. If no new events are available the ioctl() will block waiting for the next event. This ioctl may also be called in a non-blocking mode by setting zc.zc_guid = ZEVENT_NONBLOCK. In the non-blocking case if no events are available ENOENT will be returned. It is possible that ESHUTDOWN will be returned if the ioctl() is called while module unloading is in progress. And finally ENOMEM may occur if the provided nvlist buffer is not large enough to contain the entire event. * ZFS_IOC_EVENTS_CLEAR Clear are events queued by the kernel. The kernel will keep a fairly large number of recent events queued, use this ioctl to clear the in kernel list. This will effect all user space processes consuming events. The zpool command has been extended to use this events ABI with the 'events' subcommand. You may run 'zpool events -v' to output a verbose log of all recent events. This is very similar to the Solaris 'fmdump -ev' command with the key difference being it also includes what would be considered sysevents under Solaris. You may also run in follow mode with the '-f' option. To clear the in kernel event queue use the '-c' option. $ sudo cmd/zpool/zpool events -fv TIME CLASS May 13 2010 16:31:15.777711000 ereport.fs.zfs.config.sync class = "ereport.fs.zfs.config.sync" ena = 0x40982b7897700001 detector = (embedded nvlist) version = 0x0 scheme = "zfs" pool = 0xed976600de75dfa6 (end detector) time = 0x4bec8bc3 0x2e5aed98 pool = "zpios" pool_guid = 0xed976600de75dfa6 pool_context = 0x0 While the 'zpool events' command is handy for interactive debugging it is not expected to be the primary consumer of zevents. This ABI was primarily added to facilitate the addition of a user space monitoring daemon. This daemon would consume all events posted by the kernel and based on the type of event perform an action. For most events simply forwarding them on to syslog is likely enough. But this interface also cleanly allows for more sophisticated actions to be taken such as generating an email for a failed drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2010-08-26 22:42:43 +04:00
error = zpool_standard_error_fmt(hdl, errno,
dgettext(TEXT_DOMAIN, "cannot get event"));
goto out;
case ENOMEM:
if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
error = zfs_error_fmt(hdl, EZFS_NOMEM,
dgettext(TEXT_DOMAIN, "cannot get event"));
goto out;
} else {
goto retry;
}
default:
error = zpool_standard_error_fmt(hdl, errno,
dgettext(TEXT_DOMAIN, "cannot get event"));
goto out;
}
}
error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
if (error != 0)
goto out;
*dropped = (int)zc.zc_cookie;
out:
zcmd_free_nvlists(&zc);
return (error);
}
/*
* Clear all events.
*/
int
zpool_events_clear(libzfs_handle_t *hdl, int *count)
{
zfs_cmd_t zc = {"\0"};
Add linux events This topic branch leverages the Solaris style FMA call points in ZFS to create a user space visible event notification system under Linux. This new system is called zevent and it unifies all previous Solaris style ereports and sysevent notifications. Under this Linux specific scheme when a sysevent or ereport event occurs an nvlist describing the event is created which looks almost exactly like a Solaris ereport. These events are queued up in the kernel when they occur and conditionally logged to the console. It is then up to a user space application to consume the events and do whatever it likes with them. To make this possible the existing /dev/zfs ABI has been extended with two new ioctls which behave as follows. * ZFS_IOC_EVENTS_NEXT Get the next pending event. The kernel will keep track of the last event consumed by the file descriptor and provide the next one if available. If no new events are available the ioctl() will block waiting for the next event. This ioctl may also be called in a non-blocking mode by setting zc.zc_guid = ZEVENT_NONBLOCK. In the non-blocking case if no events are available ENOENT will be returned. It is possible that ESHUTDOWN will be returned if the ioctl() is called while module unloading is in progress. And finally ENOMEM may occur if the provided nvlist buffer is not large enough to contain the entire event. * ZFS_IOC_EVENTS_CLEAR Clear are events queued by the kernel. The kernel will keep a fairly large number of recent events queued, use this ioctl to clear the in kernel list. This will effect all user space processes consuming events. The zpool command has been extended to use this events ABI with the 'events' subcommand. You may run 'zpool events -v' to output a verbose log of all recent events. This is very similar to the Solaris 'fmdump -ev' command with the key difference being it also includes what would be considered sysevents under Solaris. You may also run in follow mode with the '-f' option. To clear the in kernel event queue use the '-c' option. $ sudo cmd/zpool/zpool events -fv TIME CLASS May 13 2010 16:31:15.777711000 ereport.fs.zfs.config.sync class = "ereport.fs.zfs.config.sync" ena = 0x40982b7897700001 detector = (embedded nvlist) version = 0x0 scheme = "zfs" pool = 0xed976600de75dfa6 (end detector) time = 0x4bec8bc3 0x2e5aed98 pool = "zpios" pool_guid = 0xed976600de75dfa6 pool_context = 0x0 While the 'zpool events' command is handy for interactive debugging it is not expected to be the primary consumer of zevents. This ABI was primarily added to facilitate the addition of a user space monitoring daemon. This daemon would consume all events posted by the kernel and based on the type of event perform an action. For most events simply forwarding them on to syslog is likely enough. But this interface also cleanly allows for more sophisticated actions to be taken such as generating an email for a failed drive. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2010-08-26 22:42:43 +04:00
char msg[1024];
(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
"cannot clear events"));
if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
return (zpool_standard_error_fmt(hdl, errno, msg));
if (count != NULL)
*count = (int)zc.zc_cookie; /* # of events cleared */
return (0);
}
/*
* Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
* the passed zevent_fd file handle. On success zero is returned,
* otherwise -1 is returned and hdl->libzfs_error is set to the errno.
*/
int
zpool_events_seek(libzfs_handle_t *hdl, uint64_t eid, int zevent_fd)
{
zfs_cmd_t zc = {"\0"};
int error = 0;
zc.zc_guid = eid;
zc.zc_cleanup_fd = zevent_fd;
if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_SEEK, &zc) != 0) {
switch (errno) {
case ENOENT:
error = zfs_error_fmt(hdl, EZFS_NOENT,
dgettext(TEXT_DOMAIN, "cannot get event"));
break;
case ENOMEM:
error = zfs_error_fmt(hdl, EZFS_NOMEM,
dgettext(TEXT_DOMAIN, "cannot get event"));
break;
default:
error = zpool_standard_error_fmt(hdl, errno,
dgettext(TEXT_DOMAIN, "cannot get event"));
break;
}
}
return (error);
}
2008-11-20 23:01:55 +03:00
void
zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
char *pathname, size_t len)
{
zfs_cmd_t zc = {"\0"};
2008-11-20 23:01:55 +03:00
boolean_t mounted = B_FALSE;
char *mntpnt = NULL;
char dsname[ZFS_MAX_DATASET_NAME_LEN];
2008-11-20 23:01:55 +03:00
if (dsobj == 0) {
/* special case for the MOS */
(void) snprintf(pathname, len, "<metadata>:<0x%llx>",
(longlong_t)obj);
2008-11-20 23:01:55 +03:00
return;
}
/* get the dataset's name */
(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
zc.zc_obj = dsobj;
if (ioctl(zhp->zpool_hdl->libzfs_fd,
ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
/* just write out a path of two object numbers */
(void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
(longlong_t)dsobj, (longlong_t)obj);
2008-11-20 23:01:55 +03:00
return;
}
(void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
/* find out if the dataset is mounted */
mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
/* get the corrupted object's path */
(void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
zc.zc_obj = obj;
if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
&zc) == 0) {
if (mounted) {
(void) snprintf(pathname, len, "%s%s", mntpnt,
zc.zc_value);
} else {
(void) snprintf(pathname, len, "%s:%s",
dsname, zc.zc_value);
}
} else {
(void) snprintf(pathname, len, "%s:<0x%llx>", dsname,
(longlong_t)obj);
2008-11-20 23:01:55 +03:00
}
free(mntpnt);
}
Add subcommand to wait for background zfs activity to complete Currently the best way to wait for the completion of a long-running operation in a pool, like a scrub or device removal, is to poll 'zpool status' and parse its output, which is neither efficient nor convenient. This change adds a 'wait' subcommand to the zpool command. When invoked, 'zpool wait' will block until a specified type of background activity completes. Currently, this subcommand can wait for any of the following: - Scrubs or resilvers to complete - Devices to initialized - Devices to be replaced - Devices to be removed - Checkpoints to be discarded - Background freeing to complete For example, a scrub that is in progress could be waited for by running zpool wait -t scrub <pool> This also adds a -w flag to the attach, checkpoint, initialize, replace, remove, and scrub subcommands. When used, this flag makes the operations kicked off by these subcommands synchronous instead of asynchronous. This functionality is implemented using a new ioctl. The type of activity to wait for is provided as input to the ioctl, and the ioctl blocks until all activity of that type has completed. An ioctl was used over other methods of kernel-userspace communiction primarily for the sake of portability. Porting Notes: This is ported from Delphix OS change DLPX-44432. The following changes were made while porting: - Added ZoL-style ioctl input declaration. - Reorganized error handling in zpool_initialize in libzfs to integrate better with changes made for TRIM support. - Fixed check for whether a checkpoint discard is in progress. Previously it also waited if the pool had a checkpoint, instead of just if a checkpoint was being discarded. - Exposed zfs_initialize_chunk_size as a ZoL-style tunable. - Updated more existing tests to make use of new 'zpool wait' functionality, tests that don't exist in Delphix OS. - Used existing ZoL tunable zfs_scan_suspend_progress, together with zinject, in place of a new tunable zfs_scan_max_blks_per_txg. - Added support for a non-integral interval argument to zpool wait. Future work: ZoL has support for trimming devices, which Delphix OS does not. In the future, 'zpool wait' could be extended to add the ability to wait for trim operations to complete. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: John Kennedy <john.kennedy@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: John Gallagher <john.gallagher@delphix.com> Closes #9162
2019-09-14 04:09:06 +03:00
/*
* Wait while the specified activity is in progress in the pool.
*/
int
zpool_wait(zpool_handle_t *zhp, zpool_wait_activity_t activity)
{
boolean_t missing;
int error = zpool_wait_status(zhp, activity, &missing, NULL);
if (missing) {
(void) zpool_standard_error_fmt(zhp->zpool_hdl, ENOENT,
dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
zhp->zpool_name);
return (ENOENT);
} else {
return (error);
}
}
/*
* Wait for the given activity and return the status of the wait (whether or not
* any waiting was done) in the 'waited' parameter. Non-existent pools are
* reported via the 'missing' parameter, rather than by printing an error
* message. This is convenient when this function is called in a loop over a
* long period of time (as it is, for example, by zpool's wait cmd). In that
* scenario, a pool being exported or destroyed should be considered a normal
* event, so we don't want to print an error when we find that the pool doesn't
* exist.
*/
int
zpool_wait_status(zpool_handle_t *zhp, zpool_wait_activity_t activity,
boolean_t *missing, boolean_t *waited)
{
int error = lzc_wait(zhp->zpool_name, activity, waited);
*missing = (error == ENOENT);
if (*missing)
return (0);
if (error != 0) {
(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
zhp->zpool_name);
}
return (error);
}