2fa9e406ahrens * CDDL HEADER START
3fa9e406ahrens *
4fa9e406ahrens * The contents of this file are subject to the terms of the
5ea8dc4beschrock * Common Development and Distribution License (the "License").
6ea8dc4beschrock * You may not use this file except in compliance with the License.
7fa9e406ahrens *
8fa9e406ahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9fa9e406ahrens * or http://www.opensolaris.org/os/licensing.
10fa9e406ahrens * See the License for the specific language governing permissions
11fa9e406ahrens * and limitations under the License.
12fa9e406ahrens *
13fa9e406ahrens * When distributing Covered Code, include this CDDL HEADER in each
14fa9e406ahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15fa9e406ahrens * If applicable, add the following below this CDDL HEADER, with the
16fa9e406ahrens * fields enclosed by brackets "[]" replaced with your own identifying
17fa9e406ahrens * information: Portions Copyright [yyyy] [name of copyright owner]
18fa9e406ahrens *
19fa9e406ahrens * CDDL HEADER END
20fa9e406ahrens */
2398d1cbfGeorge Wilson * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
240ce4bbcMatthew Ahrens * Copyright (c) 2011, 2019 by Delphix. All rights reserved.
251437283Hans Rosenfeld * Copyright (c) 2015, Nexenta Systems, Inc.  All rights reserved.
26bc9014eJustin Gibbs * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
2745818eeMatthew Ahrens * Copyright 2013 Saso Kiselkov. All rights reserved.
28c3d26abMatthew Ahrens * Copyright (c) 2014 Integros [integros.com]
29c8811bdToomas Soome * Copyright 2016 Toomas Soome <tsoome@me.com>
30e830fb1Kody A Kantor * Copyright 2019 Joyent, Inc.
31663207aDon Brady * Copyright (c) 2017, Intel Corporation.
321702cceAlek Pinchuk * Copyright (c) 2017 Datto Inc.
330fb055eAndy Fiddaman * Copyright 2018 OmniOS Community Edition (OmniOSce) Association.
345aeb947Garrett D'Amore */
373e30c24Will Andrews * SPA: Storage Pool Allocator
383e30c24Will Andrews *
39fa9e406ahrens * This file contains all the routines used when modifying on-disk SPA state.
40fa9e406ahrens * This includes opening, importing, destroying, exporting a pool, and syncing a
41fa9e406ahrens * pool.
42fa9e406ahrens */
44fa9e406ahrens#include <sys/zfs_context.h>
45ea8dc4beschrock#include <sys/fm/fs/zfs.h>
46fa9e406ahrens#include <sys/spa_impl.h>
47fa9e406ahrens#include <sys/zio.h>
48fa9e406ahrens#include <sys/zio_checksum.h>
49fa9e406ahrens#include <sys/dmu.h>
50fa9e406ahrens#include <sys/dmu_tx.h>
51fa9e406ahrens#include <sys/zap.h>
52fa9e406ahrens#include <sys/zil.h>
53b24ab67Jeff Bonwick#include <sys/ddt.h>
54fa9e406ahrens#include <sys/vdev_impl.h>
555cabbc6Prashanth Sreenivasa#include <sys/vdev_removal.h>
565cabbc6Prashanth Sreenivasa#include <sys/vdev_indirect_mapping.h>
575cabbc6Prashanth Sreenivasa#include <sys/vdev_indirect_births.h>
58094e47eGeorge Wilson#include <sys/vdev_initialize.h>
59084fd14Brian Behlendorf#include <sys/vdev_trim.h>
60fa9e406ahrens#include <sys/metaslab.h>
6188ecc94George Wilson#include <sys/metaslab_impl.h>
62e0f1c0aOlaf Faaland#include <sys/mmp.h>
63fa9e406ahrens#include <sys/uberblock_impl.h>
64fa9e406ahrens#include <sys/txg.h>
65fa9e406ahrens#include <sys/avl.h>
665cabbc6Prashanth Sreenivasa#include <sys/bpobj.h>
67fa9e406ahrens#include <sys/dmu_traverse.h>
68b1b8ab3lling#include <sys/dmu_objset.h>
69fa9e406ahrens#include <sys/unique.h>
70fa9e406ahrens#include <sys/dsl_pool.h>
71b1b8ab3lling#include <sys/dsl_dataset.h>
72fa9e406ahrens#include <sys/dsl_dir.h>
73fa9e406ahrens#include <sys/dsl_prop.h>
74b1b8ab3lling#include <sys/dsl_synctask.h>
75fa9e406ahrens#include <sys/fs/zfs.h>
76fa94a07brendan#include <sys/arc.h>
77fa9e406ahrens#include <sys/callb.h>
789517395ek#include <sys/systeminfo.h>
79e7cbe64gw#include <sys/spa_boot.h>
80573ca77George Wilson#include <sys/zfs_ioctl.h>
813f9d6adLin Ling#include <sys/dsl_scan.h>
82ad135b5Christopher Siden#include <sys/zfeature.h>
833b2aab1Matthew Ahrens#include <sys/dsl_destroy.h>
84770499eDan Kimmel#include <sys/abd.h>
865679c89jv#ifdef	_KERNEL
87dedec47Jack Meng#include <sys/bootprops.h>
8835a5a35Jonathan Adams#include <sys/callb.h>
8935a5a35Jonathan Adams#include <sys/cpupart.h>
9035a5a35Jonathan Adams#include <sys/pool.h>
9135a5a35Jonathan Adams#include <sys/sysdc.h>
9235a5a35Jonathan Adams#include <sys/zone.h>
935679c89jv#endif	/* _KERNEL */
95990b485lling#include "zfs_prop.h"
96b7b9745perrin#include "zfs_comutil.h"
983cb69f7Will Andrews/*
993cb69f7Will Andrews * The interval, in seconds, at which failed configuration cache file writes
1003cb69f7Will Andrews * should be retried.
1013cb69f7Will Andrews */
1025cabbc6Prashanth Sreenivasaint zfs_ccw_retry_interval = 300;
1033cb69f7Will Andrews
10435a5a35Jonathan Adamstypedef enum zti_modes {
105ec94d32Adam Leventhal	ZTI_MODE_FIXED,			/* value is # of threads (min 1) */
106ec94d32Adam Leventhal	ZTI_MODE_BATCH,			/* cpu-intensive; value is ignored */
107ec94d32Adam Leventhal	ZTI_MODE_NULL,			/* don't create a taskq */
108ec94d32Adam Leventhal	ZTI_NMODES
10935a5a35Jonathan Adams} zti_modes_t;
111ec94d32Adam Leventhal#define	ZTI_P(n, q)	{ ZTI_MODE_FIXED, (n), (q) }
112ec94d32Adam Leventhal#define	ZTI_BATCH	{ ZTI_MODE_BATCH, 0, 1 }
113ec94d32Adam Leventhal#define	ZTI_NULL	{ ZTI_MODE_NULL, 0, 0 }
1142e0c549Jonathan Adams
115ec94d32Adam Leventhal#define	ZTI_N(n)	ZTI_P(n, 1)
116ec94d32Adam Leventhal#define	ZTI_ONE		ZTI_N(1)
1172e0c549Jonathan Adams
1182e0c549Jonathan Adamstypedef struct zio_taskq_info {
119ec94d32Adam Leventhal	zti_modes_t zti_mode;
12080eb36fGeorge Wilson	uint_t zti_value;
121ec94d32Adam Leventhal	uint_t zti_count;
1222e0c549Jonathan Adams} zio_taskq_info_t;
1232e0c549Jonathan Adams
1242e0c549Jonathan Adamsstatic const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = {
12535a5a35Jonathan Adams	"issue", "issue_high", "intr", "intr_high"
1262e0c549Jonathan Adams};
1272e0c549Jonathan Adams
12880eb36fGeorge Wilson/*
129ec94d32Adam Leventhal * This table defines the taskq settings for each ZFS I/O type. When
130ec94d32Adam Leventhal * initializing a pool, we use this table to create an appropriately sized
131ec94d32Adam Leventhal * taskq. Some operations are low volume and therefore have a small, static
132ec94d32Adam Leventhal * number of threads assigned to their taskqs using the ZTI_N(#) or ZTI_ONE
133ec94d32Adam Leventhal * macros. Other operations process a large amount of data; the ZTI_BATCH
134ec94d32Adam Leventhal * macro causes us to create a taskq oriented for throughput. Some operations
135084fd14Brian Behlendorf * are so high frequency and short-lived that the taskq itself can become a
136ec94d32Adam Leventhal * point of lock contention. The ZTI_P(#, #) macro indicates that we need an
137ec94d32Adam Leventhal * additional degree of parallelism specified by the number of threads per-
138ec94d32Adam Leventhal * taskq and the number of taskqs; when dispatching an event in this case, the
139ec94d32Adam Leventhal * particular taskq is chosen at random.
140ec94d32Adam Leventhal *
141ec94d32Adam Leventhal * The different taskq priorities are to handle the different contexts (issue
142ec94d32Adam Leventhal * and interrupt) and then to reserve threads for ZIO_PRIORITY_NOW I/Os that
143ec94d32Adam Leventhal * need to be handled with minimum delay.
14480eb36fGeorge Wilson */
14580eb36fGeorge Wilsonconst zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = {
14680eb36fGeorge Wilson	/* ISSUE	ISSUE_HIGH	INTR		INTR_HIGH */
147ec94d32Adam Leventhal	{ ZTI_ONE,	ZTI_NULL,	ZTI_ONE,	ZTI_NULL }, /* NULL */
1481b497abAdam H. Leventhal	{ ZTI_N(8),	ZTI_NULL,	ZTI_P(12, 8),	ZTI_NULL }, /* READ */
149ec94d32Adam Leventhal	{ ZTI_BATCH,	ZTI_N(5),	ZTI_N(8),	ZTI_N(5) }, /* WRITE */
150ec94d32Adam Leventhal	{ ZTI_P(12, 8),	ZTI_NULL,	ZTI_ONE,	ZTI_NULL }, /* FREE */
151ec94d32Adam Leventhal	{ ZTI_ONE,	ZTI_NULL,	ZTI_ONE,	ZTI_NULL }, /* CLAIM */
152ec94d32Adam Leventhal	{ ZTI_ONE,	ZTI_NULL,	ZTI_ONE,	ZTI_NULL }, /* IOCTL */
153084fd14Brian Behlendorf	{ ZTI_N(4),	ZTI_NULL,	ZTI_ONE,	ZTI_NULL }, /* TRIM */
1542e0c549Jonathan Adams};
1552e0c549Jonathan Adams
1563b2aab1Matthew Ahrensstatic void spa_sync_version(void *arg, dmu_tx_t *tx);
1573b2aab1Matthew Ahrensstatic void spa_sync_props(void *arg, dmu_tx_t *tx);
15889a89ebllingstatic boolean_t spa_has_active_shared_spare(spa_t *spa);
1598671400Serapheim Dimitropoulosstatic int spa_load_impl(spa_t *spa, spa_import_type_t type, char **ereport);
160cb04b87Mark J Musantestatic void spa_vdev_resilver_done(spa_t *spa);
16269962b5Matthew Ahrensuint_t		zio_taskq_batch_pct = 75;	/* 1 thread per cpu in pset */
16335a5a35Jonathan Adamsid_t		zio_taskq_psrset_bind = PS_NONE;
16435a5a35Jonathan Adamsboolean_t	zio_taskq_sysdc = B_TRUE;	/* use SDC scheduling class */
16535a5a35Jonathan Adamsuint_t		zio_taskq_basedc = 80;		/* base duty cycle */
16635a5a35Jonathan Adams
16735a5a35Jonathan Adamsboolean_t	spa_create_process = B_TRUE;	/* no process ==> no sysdc */
16801f55e4George Wilsonextern int	zfs_sync_pass_deferred_free;
16935a5a35Jonathan Adams
17035a5a35Jonathan Adams/*
171e144c4ePavel Zakharov * Report any spa_load_verify errors found, but do not fail spa_load.
172e144c4ePavel Zakharov * This is used by zdb to analyze non-idle pools.
173e144c4ePavel Zakharov */
174e144c4ePavel Zakharovboolean_t	spa_load_verify_dryrun = B_FALSE;
175e144c4ePavel Zakharov
176e144c4ePavel Zakharov/*
17735a5a35Jonathan Adams * This (illegal) pool name is used when temporarily importing a spa_t in order
17835a5a35Jonathan Adams * to get the vdev stats associated with the imported devices.
17935a5a35Jonathan Adams */
18035a5a35Jonathan Adams#define	TRYIMPORT_NAME	"$import"
18135a5a35Jonathan Adams
1836f79381Pavel Zakharov * For debugging purposes: print out vdev tree during pool import.
1846f79381Pavel Zakharov */
1856f79381Pavel Zakharovboolean_t	spa_load_print_vdev_tree = B_FALSE;
1866f79381Pavel Zakharov
1876f79381Pavel Zakharov/*
1886f79381Pavel Zakharov * A non-zero value for zfs_max_missing_tvds means that we allow importing
1896f79381Pavel Zakharov * pools with missing top-level vdevs. This is strictly intended for advanced
1906f79381Pavel Zakharov * pool recovery cases since missing data is almost inevitable. Pools with
1916f79381Pavel Zakharov * missing devices can only be imported read-only for safety reasons, and their
1926f79381Pavel Zakharov * fail-mode will be automatically set to "continue".
1936f79381Pavel Zakharov *
1946f79381Pavel Zakharov * With 1 missing vdev we should be able to import the pool and mount all
1956f79381Pavel Zakharov * datasets. User data that was not modified after the missing device has been
1966f79381Pavel Zakharov * added should be recoverable. This means that snapshots created prior to the
1976f79381Pavel Zakharov * addition of that device should be completely intact.
1986f79381Pavel Zakharov *
1996f79381Pavel Zakharov * With 2 missing vdevs, some datasets may fail to mount since there are
2006f79381Pavel Zakharov * dataset statistics that are stored as regular metadata. Some data might be
2016f79381Pavel Zakharov * recoverable if those vdevs were added recently.
2026f79381Pavel Zakharov *
2036f79381Pavel Zakharov * With 3 or more missing vdevs, the pool is severely damaged and MOS entries
2046f79381Pavel Zakharov * may be missing entirely. Chances of data recovery are very low. Note that
2056f79381Pavel Zakharov * there are also risks of performing an inadvertent rewind as we might be
2066f79381Pavel Zakharov * missing all the vdevs with the latest uberblocks.
2076f79381Pavel Zakharov */
2086f79381Pavel Zakharovuint64_t	zfs_max_missing_tvds = 0;
2096f79381Pavel Zakharov
2106f79381Pavel Zakharov/*
2116f79381Pavel Zakharov * The parameters below are similar to zfs_max_missing_tvds but are only
2126f79381Pavel Zakharov * intended for a preliminary open of the pool with an untrusted config which
2136f79381Pavel Zakharov * might be incomplete or out-dated.
2146f79381Pavel Zakharov *
2156f79381Pavel Zakharov * We are more tolerant for pools opened from a cachefile since we could have
2166f79381Pavel Zakharov * an out-dated cachefile where a device removal was not registered.
2176f79381Pavel Zakharov * We could have set the limit arbitrarily high but in the case where devices
2186f79381Pavel Zakharov * are really missing we would want to return the proper error codes; we chose
2196f79381Pavel Zakharov * SPA_DVAS_PER_BP - 1 so that some copies of the MOS would still be available
2206f79381Pavel Zakharov * and we get a chance to retrieve the trusted config.
2216f79381Pavel Zakharov */
2226f79381Pavel Zakharovuint64_t	zfs_max_missing_tvds_cachefile = SPA_DVAS_PER_BP - 1;
2238671400Serapheim Dimitropoulos
2246f79381Pavel Zakharov/*
2256f79381Pavel Zakharov * In the case where config was assembled by scanning device paths (/dev/dsks
2266f79381Pavel Zakharov * by default) we are less tolerant since all the existing devices should have
2276f79381Pavel Zakharov * been detected and we want spa_load to return the right error codes.
2286f79381Pavel Zakharov */
2296f79381Pavel Zakharovuint64_t	zfs_max_missing_tvds_scan = 0;
2306f79381Pavel Zakharov
2316f79381Pavel Zakharov/*
232e830fb1Kody A Kantor * Interval in seconds at which to poll spare vdevs for health.
233e830fb1Kody A Kantor * Setting this to zero disables spare polling.
234e830fb1Kody A Kantor * Set to three hours by default.
235e830fb1Kody A Kantor */
236e830fb1Kody A Kantoruint_t		spa_spare_poll_interval_seconds = 60 * 60 * 3;
237e830fb1Kody A Kantor
238e830fb1Kody A Kantor/*
2398671400Serapheim Dimitropoulos * Debugging aid that pauses spa_sync() towards the end.
2408671400Serapheim Dimitropoulos */
2418671400Serapheim Dimitropoulosboolean_t	zfs_pause_spa_sync = B_FALSE;
2428671400Serapheim Dimitropoulos
2438671400Serapheim Dimitropoulos/*
244990b485lling * ==========================================================================
245990b485lling * SPA properties routines
246990b485lling * ==========================================================================
247990b485lling */
250990b485lling * Add a (source=src, propname=propval) list to an nvlist.
251990b485lling */
2529d82f4fllingstatic void
253990b485llingspa_prop_add_list(nvlist_t *nvl, zpool_prop_t prop, char *strval,
254990b485lling    uint64_t intval, zprop_source_t src)
256990b485lling	const char *propname = zpool_prop_to_name(prop);
257990b485lling	nvlist_t *propval;
2599d82f4flling	VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2609d82f4flling	VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0);
2629d82f4flling	if (strval != NULL)
2639d82f4flling		VERIFY(nvlist_add_string(propval, ZPROP_VALUE, strval) == 0);
2649d82f4flling	else
2659d82f4flling		VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, intval) == 0);
2679d82f4flling	VERIFY(nvlist_add_nvlist(nvl, propname, propval) == 0);
268990b485lling	nvlist_free(propval);
272990b485lling * Get property values from the spa configuration.
273990b485lling */
2749d82f4fllingstatic void
275990b485llingspa_prop_get_config(spa_t *spa, nvlist_t **nvp)
2774263d13George Wilson	vdev_t *rvd = spa->spa_root_vdev;
278ad135b5Christopher Siden	dsl_pool_t *pool = spa->spa_dsl_pool;
2792e4c998George Wilson	uint64_t size, alloc, cap, version;
280990b485lling	zprop_source_t src = ZPROP_SRC_NONE;
281c5904d1eschrock	spa_config_dirent_t *dp;
2822e4c998George Wilson	metaslab_class_t *mc = spa_normal_class(spa);
284e14bb32Jeff Bonwick	ASSERT(MUTEX_HELD(&spa->spa_props_lock));
285e14bb32Jeff Bonwick
2864263d13George Wilson	if (rvd != NULL) {
287663207aDon Brady		alloc = metaslab_class_get_alloc(mc);
288663207aDon Brady		alloc += metaslab_class_get_alloc(spa_special_class(spa));
289663207aDon Brady		alloc += metaslab_class_get_alloc(spa_dedup_class(spa));
290663207aDon Brady
291663207aDon Brady		size = metaslab_class_get_space(mc);
292663207aDon Brady		size += metaslab_class_get_space(spa_special_class(spa));
293663207aDon Brady		size += metaslab_class_get_space(spa_dedup_class(spa));
294663207aDon Brady
295379c004Eric Schrock		spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src);
296379c004Eric Schrock		spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src);
297485bbbfGeorge Wilson		spa_prop_add_list(*nvp, ZPOOL_PROP_ALLOCATED, NULL, alloc, src);
298485bbbfGeorge Wilson		spa_prop_add_list(*nvp, ZPOOL_PROP_FREE, NULL,
299485bbbfGeorge Wilson		    size - alloc, src);
3008671400Serapheim Dimitropoulos		spa_prop_add_list(*nvp, ZPOOL_PROP_CHECKPOINT, NULL,
3018671400Serapheim Dimitropoulos		    spa->spa_checkpoint_info.sci_dspace, src);
3024263d13George Wilson
3032e4c998George Wilson		spa_prop_add_list(*nvp, ZPOOL_PROP_FRAGMENTATION, NULL,
3042e4c998George Wilson		    metaslab_class_fragmentation(mc), src);
3052e4c998George Wilson		spa_prop_add_list(*nvp, ZPOOL_PROP_EXPANDSZ, NULL,
3062e4c998George Wilson		    metaslab_class_expandable_space(mc), src);
307f9af39bGeorge Wilson		spa_prop_add_list(*nvp, ZPOOL_PROP_READONLY, NULL,
308f9af39bGeorge Wilson		    (spa_mode(spa) == FREAD), src);
309379c004Eric Schrock
310485bbbfGeorge Wilson		cap = (size == 0) ? 0 : (alloc * 100 / size);
311379c004Eric Schrock		spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src);
312379c004Eric Schrock
313b24ab67Jeff Bonwick		spa_prop_add_list(*nvp, ZPOOL_PROP_DEDUPRATIO, NULL,
314b24ab67Jeff Bonwick		    ddt_get_pool_dedup_ratio(spa), src);
315b24ab67Jeff Bonwick
316379c004Eric Schrock		spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL,
3174263d13George Wilson		    rvd->vdev_state, src);
318379c004Eric Schrock
319379c004Eric Schrock		version = spa_version(spa);
320379c004Eric Schrock		if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION))
321379c004Eric Schrock			src = ZPROP_SRC_DEFAULT;
322379c004Eric Schrock		else
323379c004Eric Schrock			src = ZPROP_SRC_LOCAL;
324379c004Eric Schrock		spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, src);
325379c004Eric Schrock	}
327ad135b5Christopher Siden	if (pool != NULL) {
328ad135b5Christopher Siden		/*
329ad135b5Christopher Siden		 * The $FREE directory was introduced in SPA_VERSION_DEADLISTS,
330ad135b5Christopher Siden		 * when opening pools before this version freedir will be NULL.
331ad135b5Christopher Siden		 */
3327fd05acMatthew Ahrens		if (pool->dp_free_dir != NULL) {
333ad135b5Christopher Siden			spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL,
334c137962Justin T. Gibbs			    dsl_dir_phys(pool->dp_free_dir)->dd_used_bytes,
335c137962Justin T. Gibbs			    src);
336ad135b5Christopher Siden		} else {
337ad135b5Christopher Siden			spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING,
338ad135b5Christopher Siden			    NULL, 0, src);
339ad135b5Christopher Siden		}
3407fd05acMatthew Ahrens
3417fd05acMatthew Ahrens		if (pool->dp_leak_dir != NULL) {
3427fd05acMatthew Ahrens			spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED, NULL,
343c137962Justin T. Gibbs			    dsl_dir_phys(pool->dp_leak_dir)->dd_used_bytes,
344c137962Justin T. Gibbs			    src);
3457fd05acMatthew Ahrens		} else {
3467fd05acMatthew Ahrens			spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED,
3477fd05acMatthew Ahrens			    NULL, 0, src);
3487fd05acMatthew Ahrens		}
349ad135b5Christopher Siden	}
350ad135b5Christopher Siden
3519d82f4flling	spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src);
3538704186Dan McDonald	if (spa->spa_comment != NULL) {
3548704186Dan McDonald		spa_prop_add_list(*nvp, ZPOOL_PROP_COMMENT, spa->spa_comment,
3558704186Dan McDonald		    0, ZPROP_SRC_LOCAL);
3568704186Dan McDonald	}
3578704186Dan McDonald
3589d82f4flling	if (spa->spa_root != NULL)
3599d82f4flling		spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root,
3609d82f4flling		    0, ZPROP_SRC_LOCAL);
362b515258Matthew Ahrens	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) {
363b515258Matthew Ahrens		spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
364b515258Matthew Ahrens		    MIN(zfs_max_recordsize, SPA_MAXBLOCKSIZE), ZPROP_SRC_NONE);
365b515258Matthew Ahrens	} else {
366b515258Matthew Ahrens		spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
367b515258Matthew Ahrens		    SPA_OLD_MAXBLOCKSIZE, ZPROP_SRC_NONE);
368b515258Matthew Ahrens	}
369b515258Matthew Ahrens
37054811daToomas Soome	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) {
37154811daToomas Soome		spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
37254811daToomas Soome		    DNODE_MAX_SIZE, ZPROP_SRC_NONE);
37354811daToomas Soome	} else {
37454811daToomas Soome		spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
37554811daToomas Soome		    DNODE_MIN_SIZE, ZPROP_SRC_NONE);
37654811daToomas Soome	}
37754811daToomas Soome
378c5904d1eschrock	if ((dp = list_head(&spa->spa_config_list)) != NULL) {
379c5904d1eschrock		if (dp->scd_path == NULL) {
3809d82f4flling			spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
381c5904d1eschrock			    "none", 0, ZPROP_SRC_LOCAL);
382c5904d1eschrock		} else if (strcmp(dp->scd_path, spa_config_path) != 0) {
3839d82f4flling			spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
384c5904d1eschrock			    dp->scd_path, 0, ZPROP_SRC_LOCAL);
3852f8aaabeschrock		}
3862f8aaabeschrock	}
390990b485lling * Get zpool property values.
391990b485lling */
393990b485llingspa_prop_get(spa_t *spa, nvlist_t **nvp)
395b24ab67Jeff Bonwick	objset_t *mos = spa->spa_meta_objset;
396990b485lling	zap_cursor_t zc;
397990b485lling	zap_attribute_t za;
398990b485lling	int err;
4009d82f4flling	VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
402e14bb32Jeff Bonwick	mutex_enter(&spa->spa_props_lock);
403e14bb32Jeff Bonwick
404990b485lling	/*
405990b485lling	 * Get properties from the spa config.
406990b485lling	 */
4079d82f4flling	spa_prop_get_config(spa, nvp);
409990b485lling	/* If no pool property object, no more prop to get. */
410afee20eGeorge Wilson	if (mos == NULL || spa->spa_pool_props_object == 0) {
411990b485lling		mutex_exit(&spa->spa_props_lock);
412990b485lling		return (0);
413990b485lling	}
415990b485lling	/*
416990b485lling	 * Get properties from the MOS pool property object.
417990b485lling	 */
418990b485lling	for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object);
419990b485lling	    (err = zap_cursor_retrieve(&zc, &za)) == 0;
420990b485lling	    zap_cursor_advance(&zc)) {
421990b485lling		uint64_t intval = 0;
422990b485lling		char *strval = NULL;
423990b485lling		zprop_source_t src = ZPROP_SRC_DEFAULT;
424990b485lling		zpool_prop_t prop;
4264ae5f5fAlan Somers		if ((prop = zpool_name_to_prop(za.za_name)) == ZPOOL_PROP_INVAL)
427990b485lling			continue;
429990b485lling		switch (za.za_integer_length) {
430990b485lling		case 8:
431990b485lling			/* integer property */
432990b485lling			if (za.za_first_integer !=
433990b485lling			    zpool_prop_default_numeric(prop))
434990b485lling				src = ZPROP_SRC_LOCAL;
436990b485lling			if (prop == ZPOOL_PROP_BOOTFS) {
437990b485lling				dsl_pool_t *dp;
438990b485lling				dsl_dataset_t *ds = NULL;
440990b485lling				dp = spa_get_dsl(spa);
4413b2aab1Matthew Ahrens				dsl_pool_config_enter(dp, FTAG);
442094e47eGeorge Wilson				err = dsl_dataset_hold_obj(dp,
443094e47eGeorge Wilson				    za.za_first_integer, FTAG, &ds);
444094e47eGeorge Wilson				if (err != 0) {
4453b2aab1Matthew Ahrens					dsl_pool_config_exit(dp, FTAG);
446990b485lling					break;
447990b485lling				}
4499adfa60Matthew Ahrens				strval = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN,
450990b485lling				    KM_SLEEP);
451990b485lling				dsl_dataset_name(ds, strval);
452745cd3cmaybee				dsl_dataset_rele(ds, FTAG);
4533b2aab1Matthew Ahrens				dsl_pool_config_exit(dp, FTAG);
454990b485lling			} else {
455990b485lling				strval = NULL;
456990b485lling				intval = za.za_first_integer;
457990b485lling			}
4599d82f4flling			spa_prop_add_list(*nvp, prop, strval, intval, src);
461990b485lling			if (strval != NULL)
4629adfa60Matthew Ahrens				kmem_free(strval, ZFS_MAX_DATASET_NAME_LEN);
464990b485lling			break;
466990b485lling		case 1:
467990b485lling			/* string property */
468990b485lling			strval = kmem_alloc(za.za_num_integers, KM_SLEEP);
469990b485lling			err = zap_lookup(mos, spa->spa_pool_props_object,
470990b485lling			    za.za_name, 1, za.za_num_integers, strval);
471990b485lling			if (err) {
472990b485lling				kmem_free(strval, za.za_num_integers);
473990b485lling				break;
474990b485lling			}
4759d82f4flling			spa_prop_add_list(*nvp, prop, strval, 0, src);
476990b485lling			kmem_free(strval, za.za_num_integers);
477990b485lling			break;
479990b485lling		default:
480990b485lling			break;
481990b485lling		}
482990b485lling	}
483990b485lling	zap_cursor_fini(&zc);
484990b485lling	mutex_exit(&spa->spa_props_lock);
486990b485lling	if (err && err != ENOENT) {
487990b485lling		nvlist_free(*nvp);
4889d82f4flling		*nvp = NULL;
489990b485lling		return (err);
490990b485lling	}
492990b485lling	return (0);
496990b485lling * Validate the given pool properties nvlist and modify the list
497990b485lling * for the property values to be set.
498990b485lling */
499990b485llingstatic int
500990b485llingspa_prop_validate(spa_t *spa, nvlist_t *props)
502990b485lling	nvpair_t *elem;
503990b485lling	int error = 0, reset_bootfs = 0;
504d5285caGeorge Wilson	uint64_t objnum = 0;
505ad135b5Christopher Siden	boolean_t has_feature = B_FALSE;
507990b485lling	elem = NULL;
508990b485lling	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
509990b485lling		uint64_t intval;
510ad135b5Christopher Siden		char *strval, *slash, *check, *fname;
511ad135b5Christopher Siden		const char *propname = nvpair_name(elem);
512ad135b5Christopher Siden		zpool_prop_t prop = zpool_name_to_prop(propname);
513ad135b5Christopher Siden
514ad135b5Christopher Siden		switch (prop) {
5154ae5f5fAlan Somers		case ZPOOL_PROP_INVAL:
516ad135b5Christopher Siden			if (!zpool_prop_feature(propname)) {
517be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
518ad135b5Christopher Siden				break;
519ad135b5Christopher Siden			}
521ad135b5Christopher Siden			/*
522ad135b5Christopher Siden			 * Sanitize the input.
523ad135b5Christopher Siden			 */
524ad135b5Christopher Siden			if (nvpair_type(elem) != DATA_TYPE_UINT64) {
525be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
526ad135b5Christopher Siden				break;
527ad135b5Christopher Siden			}
529ad135b5Christopher Siden			if (nvpair_value_uint64(elem, &intval) != 0) {
530be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
531ad135b5Christopher Siden				break;
532ad135b5Christopher Siden			}
533ad135b5Christopher Siden
534ad135b5Christopher Siden			if (intval != 0) {
535be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
536ad135b5Christopher Siden				break;
537ad135b5Christopher Siden			}
538ad135b5Christopher Siden
539ad135b5Christopher Siden			fname = strchr(propname, '@') + 1;
540ad135b5Christopher Siden			if (zfeature_lookup_name(fname, NULL) != 0) {
541be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
542ad135b5Christopher Siden				break;
543ad135b5Christopher Siden			}
544ad135b5Christopher Siden
545ad135b5Christopher Siden			has_feature = B_TRUE;
546ad135b5Christopher Siden			break;
548990b485lling		case ZPOOL_PROP_VERSION:
549990b485lling			error = nvpair_value_uint64(elem, &intval);
550990b485lling			if (!error &&
551ad135b5Christopher Siden			    (intval < spa_version(spa) ||
552ad135b5Christopher Siden			    intval > SPA_VERSION_BEFORE_FEATURES ||
553ad135b5Christopher Siden			    has_feature))
554be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
555990b485lling			break;
557990b485lling		case ZPOOL_PROP_DELEGATION:
558990b485lling		case ZPOOL_PROP_AUTOREPLACE:
559d5b5bb2Rich Morris		case ZPOOL_PROP_LISTSNAPS:
560573ca77George Wilson		case ZPOOL_PROP_AUTOEXPAND:
561084fd14Brian Behlendorf		case ZPOOL_PROP_AUTOTRIM:
562990b485lling			error = nvpair_value_uint64(elem, &intval);
563990b485lling			if (!error && intval > 1)
564be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
565990b485lling			break;
567e0f1c0aOlaf Faaland		case ZPOOL_PROP_MULTIHOST:
568e0f1c0aOlaf Faaland			error = nvpair_value_uint64(elem, &intval);
569e0f1c0aOlaf Faaland			if (!error && intval > 1)
570e0f1c0aOlaf Faaland				error = SET_ERROR(EINVAL);
571e0f1c0aOlaf Faaland
572e0f1c0aOlaf Faaland			if (!error && !spa_get_hostid())
573e0f1c0aOlaf Faaland				error = SET_ERROR(ENOTSUP);
574e0f1c0aOlaf Faaland
575e0f1c0aOlaf Faaland			break;
576e0f1c0aOlaf Faaland
577990b485lling		case ZPOOL_PROP_BOOTFS:
57825f89eeJeff Bonwick			/*
57925f89eeJeff Bonwick			 * If the pool version is less than SPA_VERSION_BOOTFS,
58025f89eeJeff Bonwick			 * or the pool is still being created (version == 0),
58125f89eeJeff Bonwick			 * the bootfs property cannot be set.
58225f89eeJeff Bonwick			 */
583990b485lling			if (spa_version(spa) < SPA_VERSION_BOOTFS) {
584be6fd75Matthew Ahrens				error = SET_ERROR(ENOTSUP);
585990b485lling				break;
586990b485lling			}
588990b485lling			/*
58915e6edfgw			 * Make sure the vdev config is bootable
590990b485lling			 */
59115e6edfgw			if (!vdev_is_bootable(spa->spa_root_vdev)) {
592be6fd75Matthew Ahrens				error = SET_ERROR(ENOTSUP);
593990b485lling				break;
594990b485lling			}
596990b485lling			reset_bootfs = 1;
598990b485lling			error = nvpair_value_string(elem, &strval);
600990b485lling			if (!error) {
601ad135b5Christopher Siden				objset_t *os;
602b515258Matthew Ahrens				uint64_t propval;
604990b485lling				if (strval == NULL || strval[0] == '\0') {
605990b485lling					objnum = zpool_prop_default_numeric(
606990b485lling					    ZPOOL_PROP_BOOTFS);
607990b485lling					break;
608990b485lling				}
610094e47eGeorge Wilson				error = dmu_objset_hold(strval, FTAG, &os);
611094e47eGeorge Wilson				if (error != 0)
612990b485lling					break;
614b515258Matthew Ahrens				/*
615b515258Matthew Ahrens				 * Must be ZPL, and its property settings
61654811daToomas Soome				 * must be supported.
617b515258Matthew Ahrens				 */
618503ad85Matthew Ahrens
619503ad85Matthew Ahrens				if (dmu_objset_type(os) != DMU_OST_ZFS) {
620be6fd75Matthew Ahrens					error = SET_ERROR(ENOTSUP);
6213b2aab1Matthew Ahrens				} else if ((error =
6223b2aab1Matthew Ahrens				    dsl_prop_get_int_ds(dmu_objset_ds(os),
62315e6edfgw				    zfs_prop_to_name(ZFS_PROP_COMPRESSION),
624b515258Matthew Ahrens				    &propval)) == 0 &&
625b515258Matthew Ahrens				    !BOOTFS_COMPRESS_VALID(propval)) {
626b515258Matthew Ahrens					error = SET_ERROR(ENOTSUP);
62715e6edfgw				} else {
62815e6edfgw					objnum = dmu_objset_id(os);
62915e6edfgw				}
630503ad85Matthew Ahrens				dmu_objset_rele(os, FTAG);
631990b485lling			}
632990b485lling			break;
633e14bb32Jeff Bonwick
6340a4e951gw		case ZPOOL_PROP_FAILUREMODE:
6350a4e951gw			error = nvpair_value_uint64(elem, &intval);
6360a4e951gw			if (!error && (intval < ZIO_FAILURE_MODE_WAIT ||
6370a4e951gw			    intval > ZIO_FAILURE_MODE_PANIC))
638be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
6400a4e951gw			/*
6410a4e951gw			 * This is a special case which only occurs when
6420a4e951gw			 * the pool has completely failed. This allows
6430a4e951gw			 * the user to change the in-core failmode property
6440a4e951gw			 * without syncing it out to disk (I/Os might
6450a4e951gw			 * currently be blocked). We do this by returning
6460a4e951gw			 * EIO to the caller (spa_prop_set) to trick it
6470a4e951gw			 * into thinking we encountered a property validation
6480a4e951gw			 * error.
6490a4e951gw			 */
650e14bb32Jeff Bonwick			if (!error && spa_suspended(spa)) {
6510a4e951gw				spa->spa_failmode = intval;
652be6fd75Matthew Ahrens				error = SET_ERROR(EIO);
6530a4e951gw			}
6540a4e951gw			break;
6562f8aaabeschrock		case ZPOOL_PROP_CACHEFILE:
6572f8aaabeschrock			if ((error = nvpair_value_string(elem, &strval)) != 0)
6582f8aaabeschrock				break;
6602f8aaabeschrock			if (strval[0] == '\0')
6612f8aaabeschrock				break;
6632f8aaabeschrock			if (strcmp(strval, "none") == 0)
6642f8aaabeschrock				break;
6662f8aaabeschrock			if (strval[0] != '/') {
667be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
6682f8aaabeschrock				break;
6692f8aaabeschrock			}
6712f8aaabeschrock			slash = strrchr(strval, '/');
6722f8aaabeschrock			ASSERT(slash != NULL);
6742f8aaabeschrock			if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
6752f8aaabeschrock			    strcmp(slash, "/..") == 0)
676be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
6772f8aaabeschrock			break;
678b24ab67Jeff Bonwick
6798704186Dan McDonald		case ZPOOL_PROP_COMMENT:
6808704186Dan McDonald			if ((error = nvpair_value_string(elem, &strval)) != 0)
6818704186Dan McDonald				break;
6828704186Dan McDonald			for (check = strval; *check != '\0'; check++) {
6838704186Dan McDonald				/*
6848704186Dan McDonald				 * The kernel doesn't have an easy isprint()
6858704186Dan McDonald				 * check.  For this kernel check, we merely
6868704186Dan McDonald				 * check ASCII apart from DEL.  Fix this if
6878704186Dan McDonald				 * there is an easy-to-use kernel isprint().
6888704186Dan McDonald				 */
6898704186Dan McDonald				if (*check >= 0x7f) {
690be6fd75Matthew Ahrens					error = SET_ERROR(EINVAL);
6918704186Dan McDonald					break;
6928704186Dan McDonald				}
6938704186Dan McDonald			}
6948704186Dan McDonald			if (strlen(strval) > ZPROP_MAX_COMMENT)
6958704186Dan McDonald				error = E2BIG;
6968704186Dan McDonald			break;
6978704186Dan McDonald
698b24ab67Jeff Bonwick		case ZPOOL_PROP_DEDUPDITTO:
699b24ab67Jeff Bonwick			if (spa_version(spa) < SPA_VERSION_DEDUP)
700be6fd75Matthew Ahrens				error = SET_ERROR(ENOTSUP);
701b24ab67Jeff Bonwick			else
702b24ab67Jeff Bonwick				error = nvpair_value_uint64(elem, &intval);
703b24ab67Jeff Bonwick			if (error == 0 &&
704b24ab67Jeff Bonwick			    intval != 0 && intval < ZIO_DEDUPDITTO_MIN)
705be6fd75Matthew Ahrens				error = SET_ERROR(EINVAL);
706b24ab67Jeff Bonwick			break;
707990b485lling		}
709990b485lling		if (error)
710990b485lling			break;
711990b485lling	}
713990b485lling	if (!error && reset_bootfs) {
714990b485lling		error = nvlist_remove(props,
715990b485lling		    zpool_prop_to_name(ZPOOL_PROP_BOOTFS), DATA_TYPE_STRING);
717990b485lling		if (!error) {
718990b485lling			error = nvlist_add_uint64(props,
719990b485lling			    zpool_prop_to_name(ZPOOL_PROP_BOOTFS), objnum);
720990b485lling		}
721990b485lling	}
723990b485lling	return (error);
726379c004Eric Schrockvoid
727379c004Eric Schrockspa_configfile_set(spa_t *spa, nvlist_t *nvp, boolean_t need_sync)
728379c004Eric Schrock{
729379c004Eric Schrock	char *cachefile;
730379c004Eric Schrock	spa_config_dirent_t *dp;
731379c004Eric Schrock
732379c004Eric Schrock	if (nvlist_lookup_string(nvp, zpool_prop_to_name(ZPOOL_PROP_CACHEFILE),
733379c004Eric Schrock	    &cachefile) != 0)
734379c004Eric Schrock		return;
735379c004Eric Schrock
736379c004Eric Schrock	dp = kmem_alloc(sizeof (spa_config_dirent_t),
737379c004Eric Schrock	    KM_SLEEP);
738379c004Eric Schrock
739379c004Eric Schrock	if (cachefile[0] == '\0')
740379c004Eric Schrock		dp->scd_path = spa_strdup(spa_config_path);
741379c004Eric Schrock	else if (strcmp(cachefile, "none") == 0)
742379c004Eric Schrock		dp->scd_path = NULL;
743379c004Eric Schrock	else
744379c004Eric Schrock		dp->scd_path = spa_strdup(cachefile);
745379c004Eric Schrock
746379c004Eric Schrock	list_insert_head(&spa->spa_config_list, dp);
747379c004Eric Schrock	if (need_sync)
748379c004Eric Schrock		spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
749379c004Eric Schrock}
750379c004Eric Schrock
752990b485llingspa_prop_set(spa_t *spa, nvlist_t *nvp)
754990b485lling	int error;
755ad135b5Christopher Siden	nvpair_t *elem = NULL;
756379c004Eric Schrock	boolean_t need_sync = B_FALSE;
758990b485lling	if ((error = spa_prop_validate(spa, nvp)) != 0)
759990b485lling		return (error);
761379c004Eric Schrock	while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) {
762ad135b5Christopher Siden		zpool_prop_t prop = zpool_name_to_prop(nvpair_name(elem));
763379c004Eric Schrock
764f9af39bGeorge Wilson		if (prop == ZPOOL_PROP_CACHEFILE ||
765f9af39bGeorge Wilson		    prop == ZPOOL_PROP_ALTROOT ||
766f9af39bGeorge Wilson		    prop == ZPOOL_PROP_READONLY)
767379c004Eric Schrock			continue;
768379c004Eric Schrock
7694ae5f5fAlan Somers		if (prop == ZPOOL_PROP_VERSION || prop == ZPOOL_PROP_INVAL) {
770ad135b5Christopher Siden			uint64_t ver;
771ad135b5Christopher Siden
772ad135b5Christopher Siden			if (prop == ZPOOL_PROP_VERSION) {
773ad135b5Christopher Siden				VERIFY(nvpair_value_uint64(elem, &ver) == 0);
774ad135b5Christopher Siden			} else {
775ad135b5Christopher Siden				<