spa_misc.c revision 3ee8c80c747c4aa3f83351a6920f30c411236e1b
1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21/*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
24 * Copyright 2015 Nexenta Systems, Inc.  All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26 * Copyright 2013 Saso Kiselkov. All rights reserved.
27 * Copyright (c) 2014 Integros [integros.com]
28 * Copyright (c) 2017 Datto Inc.
29 */
30
31#include <sys/zfs_context.h>
32#include <sys/spa_impl.h>
33#include <sys/spa_boot.h>
34#include <sys/zio.h>
35#include <sys/zio_checksum.h>
36#include <sys/zio_compress.h>
37#include <sys/dmu.h>
38#include <sys/dmu_tx.h>
39#include <sys/zap.h>
40#include <sys/zil.h>
41#include <sys/vdev_impl.h>
42#include <sys/metaslab.h>
43#include <sys/uberblock_impl.h>
44#include <sys/txg.h>
45#include <sys/avl.h>
46#include <sys/unique.h>
47#include <sys/dsl_pool.h>
48#include <sys/dsl_dir.h>
49#include <sys/dsl_prop.h>
50#include <sys/dsl_scan.h>
51#include <sys/fs/zfs.h>
52#include <sys/metaslab_impl.h>
53#include <sys/arc.h>
54#include <sys/ddt.h>
55#include "zfs_prop.h"
56#include <sys/zfeature.h>
57
58/*
59 * SPA locking
60 *
61 * There are four basic locks for managing spa_t structures:
62 *
63 * spa_namespace_lock (global mutex)
64 *
65 *	This lock must be acquired to do any of the following:
66 *
67 *		- Lookup a spa_t by name
68 *		- Add or remove a spa_t from the namespace
69 *		- Increase spa_refcount from non-zero
70 *		- Check if spa_refcount is zero
71 *		- Rename a spa_t
72 *		- add/remove/attach/detach devices
73 *		- Held for the duration of create/destroy/import/export
74 *
75 *	It does not need to handle recursion.  A create or destroy may
76 *	reference objects (files or zvols) in other pools, but by
77 *	definition they must have an existing reference, and will never need
78 *	to lookup a spa_t by name.
79 *
80 * spa_refcount (per-spa refcount_t protected by mutex)
81 *
82 *	This reference count keep track of any active users of the spa_t.  The
83 *	spa_t cannot be destroyed or freed while this is non-zero.  Internally,
84 *	the refcount is never really 'zero' - opening a pool implicitly keeps
85 *	some references in the DMU.  Internally we check against spa_minref, but
86 *	present the image of a zero/non-zero value to consumers.
87 *
88 * spa_config_lock[] (per-spa array of rwlocks)
89 *
90 *	This protects the spa_t from config changes, and must be held in
91 *	the following circumstances:
92 *
93 *		- RW_READER to perform I/O to the spa
94 *		- RW_WRITER to change the vdev config
95 *
96 * The locking order is fairly straightforward:
97 *
98 *		spa_namespace_lock	->	spa_refcount
99 *
100 *	The namespace lock must be acquired to increase the refcount from 0
101 *	or to check if it is zero.
102 *
103 *		spa_refcount		->	spa_config_lock[]
104 *
105 *	There must be at least one valid reference on the spa_t to acquire
106 *	the config lock.
107 *
108 *		spa_namespace_lock	->	spa_config_lock[]
109 *
110 *	The namespace lock must always be taken before the config lock.
111 *
112 *
113 * The spa_namespace_lock can be acquired directly and is globally visible.
114 *
115 * The namespace is manipulated using the following functions, all of which
116 * require the spa_namespace_lock to be held.
117 *
118 *	spa_lookup()		Lookup a spa_t by name.
119 *
120 *	spa_add()		Create a new spa_t in the namespace.
121 *
122 *	spa_remove()		Remove a spa_t from the namespace.  This also
123 *				frees up any memory associated with the spa_t.
124 *
125 *	spa_next()		Returns the next spa_t in the system, or the
126 *				first if NULL is passed.
127 *
128 *	spa_evict_all()		Shutdown and remove all spa_t structures in
129 *				the system.
130 *
131 *	spa_guid_exists()	Determine whether a pool/device guid exists.
132 *
133 * The spa_refcount is manipulated using the following functions:
134 *
135 *	spa_open_ref()		Adds a reference to the given spa_t.  Must be
136 *				called with spa_namespace_lock held if the
137 *				refcount is currently zero.
138 *
139 *	spa_close()		Remove a reference from the spa_t.  This will
140 *				not free the spa_t or remove it from the
141 *				namespace.  No locking is required.
142 *
143 *	spa_refcount_zero()	Returns true if the refcount is currently
144 *				zero.  Must be called with spa_namespace_lock
145 *				held.
146 *
147 * The spa_config_lock[] is an array of rwlocks, ordered as follows:
148 * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV.
149 * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}().
150 *
151 * To read the configuration, it suffices to hold one of these locks as reader.
152 * To modify the configuration, you must hold all locks as writer.  To modify
153 * vdev state without altering the vdev tree's topology (e.g. online/offline),
154 * you must hold SCL_STATE and SCL_ZIO as writer.
155 *
156 * We use these distinct config locks to avoid recursive lock entry.
157 * For example, spa_sync() (which holds SCL_CONFIG as reader) induces
158 * block allocations (SCL_ALLOC), which may require reading space maps
159 * from disk (dmu_read() -> zio_read() -> SCL_ZIO).
160 *
161 * The spa config locks cannot be normal rwlocks because we need the
162 * ability to hand off ownership.  For example, SCL_ZIO is acquired
163 * by the issuing thread and later released by an interrupt thread.
164 * They do, however, obey the usual write-wanted semantics to prevent
165 * writer (i.e. system administrator) starvation.
166 *
167 * The lock acquisition rules are as follows:
168 *
169 * SCL_CONFIG
170 *	Protects changes to the vdev tree topology, such as vdev
171 *	add/remove/attach/detach.  Protects the dirty config list
172 *	(spa_config_dirty_list) and the set of spares and l2arc devices.
173 *
174 * SCL_STATE
175 *	Protects changes to pool state and vdev state, such as vdev
176 *	online/offline/fault/degrade/clear.  Protects the dirty state list
177 *	(spa_state_dirty_list) and global pool state (spa_state).
178 *
179 * SCL_ALLOC
180 *	Protects changes to metaslab groups and classes.
181 *	Held as reader by metaslab_alloc() and metaslab_claim().
182 *
183 * SCL_ZIO
184 *	Held by bp-level zios (those which have no io_vd upon entry)
185 *	to prevent changes to the vdev tree.  The bp-level zio implicitly
186 *	protects all of its vdev child zios, which do not hold SCL_ZIO.
187 *
188 * SCL_FREE
189 *	Protects changes to metaslab groups and classes.
190 *	Held as reader by metaslab_free().  SCL_FREE is distinct from
191 *	SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free
192 *	blocks in zio_done() while another i/o that holds either
193 *	SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete.
194 *
195 * SCL_VDEV
196 *	Held as reader to prevent changes to the vdev tree during trivial
197 *	inquiries such as bp_get_dsize().  SCL_VDEV is distinct from the
198 *	other locks, and lower than all of them, to ensure that it's safe
199 *	to acquire regardless of caller context.
200 *
201 * In addition, the following rules apply:
202 *
203 * (a)	spa_props_lock protects pool properties, spa_config and spa_config_list.
204 *	The lock ordering is SCL_CONFIG > spa_props_lock.
205 *
206 * (b)	I/O operations on leaf vdevs.  For any zio operation that takes
207 *	an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(),
208 *	or zio_write_phys() -- the caller must ensure that the config cannot
209 *	cannot change in the interim, and that the vdev cannot be reopened.
210 *	SCL_STATE as reader suffices for both.
211 *
212 * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit().
213 *
214 *	spa_vdev_enter()	Acquire the namespace lock and the config lock
215 *				for writing.
216 *
217 *	spa_vdev_exit()		Release the config lock, wait for all I/O
218 *				to complete, sync the updated configs to the
219 *				cache, and release the namespace lock.
220 *
221 * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit().
222 * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual
223 * locking is, always, based on spa_namespace_lock and spa_config_lock[].
224 *
225 * spa_rename() is also implemented within this file since it requires
226 * manipulation of the namespace.
227 */
228
229static avl_tree_t spa_namespace_avl;
230kmutex_t spa_namespace_lock;
231static kcondvar_t spa_namespace_cv;
232static int spa_active_count;
233int spa_max_replication_override = SPA_DVAS_PER_BP;
234
235static kmutex_t spa_spare_lock;
236static avl_tree_t spa_spare_avl;
237static kmutex_t spa_l2cache_lock;
238static avl_tree_t spa_l2cache_avl;
239
240kmem_cache_t *spa_buffer_pool;
241int spa_mode_global;
242
243#ifdef ZFS_DEBUG
244/*
245 * Everything except dprintf, spa, and indirect_remap is on by default
246 * in debug builds.
247 */
248int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_SPA | ZFS_DEBUG_INDIRECT_REMAP);
249#else
250int zfs_flags = 0;
251#endif
252
253/*
254 * zfs_recover can be set to nonzero to attempt to recover from
255 * otherwise-fatal errors, typically caused by on-disk corruption.  When
256 * set, calls to zfs_panic_recover() will turn into warning messages.
257 * This should only be used as a last resort, as it typically results
258 * in leaked space, or worse.
259 */
260boolean_t zfs_recover = B_FALSE;
261
262/*
263 * If destroy encounters an EIO while reading metadata (e.g. indirect
264 * blocks), space referenced by the missing metadata can not be freed.
265 * Normally this causes the background destroy to become "stalled", as
266 * it is unable to make forward progress.  While in this stalled state,
267 * all remaining space to free from the error-encountering filesystem is
268 * "temporarily leaked".  Set this flag to cause it to ignore the EIO,
269 * permanently leak the space from indirect blocks that can not be read,
270 * and continue to free everything else that it can.
271 *
272 * The default, "stalling" behavior is useful if the storage partially
273 * fails (i.e. some but not all i/os fail), and then later recovers.  In
274 * this case, we will be able to continue pool operations while it is
275 * partially failed, and when it recovers, we can continue to free the
276 * space, with no leaks.  However, note that this case is actually
277 * fairly rare.
278 *
279 * Typically pools either (a) fail completely (but perhaps temporarily,
280 * e.g. a top-level vdev going offline), or (b) have localized,
281 * permanent errors (e.g. disk returns the wrong data due to bit flip or
282 * firmware bug).  In case (a), this setting does not matter because the
283 * pool will be suspended and the sync thread will not be able to make
284 * forward progress regardless.  In case (b), because the error is
285 * permanent, the best we can do is leak the minimum amount of space,
286 * which is what setting this flag will do.  Therefore, it is reasonable
287 * for this flag to normally be set, but we chose the more conservative
288 * approach of not setting it, so that there is no possibility of
289 * leaking space in the "partial temporary" failure case.
290 */
291boolean_t zfs_free_leak_on_eio = B_FALSE;
292
293/*
294 * Expiration time in milliseconds. This value has two meanings. First it is
295 * used to determine when the spa_deadman() logic should fire. By default the
296 * spa_deadman() will fire if spa_sync() has not completed in 1000 seconds.
297 * Secondly, the value determines if an I/O is considered "hung". Any I/O that
298 * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting
299 * in a system panic.
300 */
301uint64_t zfs_deadman_synctime_ms = 1000000ULL;
302
303/*
304 * Check time in milliseconds. This defines the frequency at which we check
305 * for hung I/O.
306 */
307uint64_t zfs_deadman_checktime_ms = 5000ULL;
308
309/*
310 * Override the zfs deadman behavior via /etc/system. By default the
311 * deadman is enabled except on VMware and sparc deployments.
312 */
313int zfs_deadman_enabled = -1;
314
315/*
316 * The worst case is single-sector max-parity RAID-Z blocks, in which
317 * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1)
318 * times the size; so just assume that.  Add to this the fact that
319 * we can have up to 3 DVAs per bp, and one more factor of 2 because
320 * the block may be dittoed with up to 3 DVAs by ddt_sync().  All together,
321 * the worst case is:
322 *     (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24
323 */
324int spa_asize_inflation = 24;
325
326/*
327 * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in
328 * the pool to be consumed.  This ensures that we don't run the pool
329 * completely out of space, due to unaccounted changes (e.g. to the MOS).
330 * It also limits the worst-case time to allocate space.  If we have
331 * less than this amount of free space, most ZPL operations (e.g. write,
332 * create) will return ENOSPC.
333 *
334 * Certain operations (e.g. file removal, most administrative actions) can
335 * use half the slop space.  They will only return ENOSPC if less than half
336 * the slop space is free.  Typically, once the pool has less than the slop
337 * space free, the user will use these operations to free up space in the pool.
338 * These are the operations that call dsl_pool_adjustedsize() with the netfree
339 * argument set to TRUE.
340 *
341 * A very restricted set of operations are always permitted, regardless of
342 * the amount of free space.  These are the operations that call
343 * dsl_sync_task(ZFS_SPACE_CHECK_NONE), e.g. "zfs destroy".  If these
344 * operations result in a net increase in the amount of space used,
345 * it is possible to run the pool completely out of space, causing it to
346 * be permanently read-only.
347 *
348 * Note that on very small pools, the slop space will be larger than
349 * 3.2%, in an effort to have it be at least spa_min_slop (128MB),
350 * but we never allow it to be more than half the pool size.
351 *
352 * See also the comments in zfs_space_check_t.
353 */
354int spa_slop_shift = 5;
355uint64_t spa_min_slop = 128 * 1024 * 1024;
356
357/*PRINTFLIKE2*/
358void
359spa_load_failed(spa_t *spa, const char *fmt, ...)
360{
361	va_list adx;
362	char buf[256];
363
364	va_start(adx, fmt);
365	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
366	va_end(adx);
367
368	zfs_dbgmsg("spa_load(%s): FAILED: %s", spa->spa_name, buf);
369}
370
371/*PRINTFLIKE2*/
372void
373spa_load_note(spa_t *spa, const char *fmt, ...)
374{
375	va_list adx;
376	char buf[256];
377
378	va_start(adx, fmt);
379	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
380	va_end(adx);
381
382	zfs_dbgmsg("spa_load(%s): %s", spa->spa_name, buf);
383}
384
385/*
386 * ==========================================================================
387 * SPA config locking
388 * ==========================================================================
389 */
390static void
391spa_config_lock_init(spa_t *spa)
392{
393	for (int i = 0; i < SCL_LOCKS; i++) {
394		spa_config_lock_t *scl = &spa->spa_config_lock[i];
395		mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL);
396		cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL);
397		refcount_create_untracked(&scl->scl_count);
398		scl->scl_writer = NULL;
399		scl->scl_write_wanted = 0;
400	}
401}
402
403static void
404spa_config_lock_destroy(spa_t *spa)
405{
406	for (int i = 0; i < SCL_LOCKS; i++) {
407		spa_config_lock_t *scl = &spa->spa_config_lock[i];
408		mutex_destroy(&scl->scl_lock);
409		cv_destroy(&scl->scl_cv);
410		refcount_destroy(&scl->scl_count);
411		ASSERT(scl->scl_writer == NULL);
412		ASSERT(scl->scl_write_wanted == 0);
413	}
414}
415
416int
417spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw)
418{
419	for (int i = 0; i < SCL_LOCKS; i++) {
420		spa_config_lock_t *scl = &spa->spa_config_lock[i];
421		if (!(locks & (1 << i)))
422			continue;
423		mutex_enter(&scl->scl_lock);
424		if (rw == RW_READER) {
425			if (scl->scl_writer || scl->scl_write_wanted) {
426				mutex_exit(&scl->scl_lock);
427				spa_config_exit(spa, locks & ((1 << i) - 1),
428				    tag);
429				return (0);
430			}
431		} else {
432			ASSERT(scl->scl_writer != curthread);
433			if (!refcount_is_zero(&scl->scl_count)) {
434				mutex_exit(&scl->scl_lock);
435				spa_config_exit(spa, locks & ((1 << i) - 1),
436				    tag);
437				return (0);
438			}
439			scl->scl_writer = curthread;
440		}
441		(void) refcount_add(&scl->scl_count, tag);
442		mutex_exit(&scl->scl_lock);
443	}
444	return (1);
445}
446
447void
448spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw)
449{
450	int wlocks_held = 0;
451
452	ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY);
453
454	for (int i = 0; i < SCL_LOCKS; i++) {
455		spa_config_lock_t *scl = &spa->spa_config_lock[i];
456		if (scl->scl_writer == curthread)
457			wlocks_held |= (1 << i);
458		if (!(locks & (1 << i)))
459			continue;
460		mutex_enter(&scl->scl_lock);
461		if (rw == RW_READER) {
462			while (scl->scl_writer || scl->scl_write_wanted) {
463				cv_wait(&scl->scl_cv, &scl->scl_lock);
464			}
465		} else {
466			ASSERT(scl->scl_writer != curthread);
467			while (!refcount_is_zero(&scl->scl_count)) {
468				scl->scl_write_wanted++;
469				cv_wait(&scl->scl_cv, &scl->scl_lock);
470				scl->scl_write_wanted--;
471			}
472			scl->scl_writer = curthread;
473		}
474		(void) refcount_add(&scl->scl_count, tag);
475		mutex_exit(&scl->scl_lock);
476	}
477	ASSERT3U(wlocks_held, <=, locks);
478}
479
480void
481spa_config_exit(spa_t *spa, int locks, void *tag)
482{
483	for (int i = SCL_LOCKS - 1; i >= 0; i--) {
484		spa_config_lock_t *scl = &spa->spa_config_lock[i];
485		if (!(locks & (1 << i)))
486			continue;
487		mutex_enter(&scl->scl_lock);
488		ASSERT(!refcount_is_zero(&scl->scl_count));
489		if (refcount_remove(&scl->scl_count, tag) == 0) {
490			ASSERT(scl->scl_writer == NULL ||
491			    scl->scl_writer == curthread);
492			scl->scl_writer = NULL;	/* OK in either case */
493			cv_broadcast(&scl->scl_cv);
494		}
495		mutex_exit(&scl->scl_lock);
496	}
497}
498
499int
500spa_config_held(spa_t *spa, int locks, krw_t rw)
501{
502	int locks_held = 0;
503
504	for (int i = 0; i < SCL_LOCKS; i++) {
505		spa_config_lock_t *scl = &spa->spa_config_lock[i];
506		if (!(locks & (1 << i)))
507			continue;
508		if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) ||
509		    (rw == RW_WRITER && scl->scl_writer == curthread))
510			locks_held |= 1 << i;
511	}
512
513	return (locks_held);
514}
515
516/*
517 * ==========================================================================
518 * SPA namespace functions
519 * ==========================================================================
520 */
521
522/*
523 * Lookup the named spa_t in the AVL tree.  The spa_namespace_lock must be held.
524 * Returns NULL if no matching spa_t is found.
525 */
526spa_t *
527spa_lookup(const char *name)
528{
529	static spa_t search;	/* spa_t is large; don't allocate on stack */
530	spa_t *spa;
531	avl_index_t where;
532	char *cp;
533
534	ASSERT(MUTEX_HELD(&spa_namespace_lock));
535
536	(void) strlcpy(search.spa_name, name, sizeof (search.spa_name));
537
538	/*
539	 * If it's a full dataset name, figure out the pool name and
540	 * just use that.
541	 */
542	cp = strpbrk(search.spa_name, "/@#");
543	if (cp != NULL)
544		*cp = '\0';
545
546	spa = avl_find(&spa_namespace_avl, &search, &where);
547
548	return (spa);
549}
550
551/*
552 * Fires when spa_sync has not completed within zfs_deadman_synctime_ms.
553 * If the zfs_deadman_enabled flag is set then it inspects all vdev queues
554 * looking for potentially hung I/Os.
555 */
556void
557spa_deadman(void *arg)
558{
559	spa_t *spa = arg;
560
561	/*
562	 * Disable the deadman timer if the pool is suspended.
563	 */
564	if (spa_suspended(spa)) {
565		VERIFY(cyclic_reprogram(spa->spa_deadman_cycid, CY_INFINITY));
566		return;
567	}
568
569	zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu",
570	    (gethrtime() - spa->spa_sync_starttime) / NANOSEC,
571	    ++spa->spa_deadman_calls);
572	if (zfs_deadman_enabled)
573		vdev_deadman(spa->spa_root_vdev);
574}
575
576/*
577 * Create an uninitialized spa_t with the given name.  Requires
578 * spa_namespace_lock.  The caller must ensure that the spa_t doesn't already
579 * exist by calling spa_lookup() first.
580 */
581spa_t *
582spa_add(const char *name, nvlist_t *config, const char *altroot)
583{
584	spa_t *spa;
585	spa_config_dirent_t *dp;
586	cyc_handler_t hdlr;
587	cyc_time_t when;
588
589	ASSERT(MUTEX_HELD(&spa_namespace_lock));
590
591	spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP);
592
593	mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL);
594	mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL);
595	mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL);
596	mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL);
597	mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL);
598	mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL);
599	mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL);
600	mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL);
601	mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL);
602	mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL);
603	mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL);
604	mutex_init(&spa->spa_iokstat_lock, NULL, MUTEX_DEFAULT, NULL);
605	mutex_init(&spa->spa_alloc_lock, NULL, MUTEX_DEFAULT, NULL);
606
607	cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL);
608	cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL);
609	cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL);
610	cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL);
611	cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL);
612
613	for (int t = 0; t < TXG_SIZE; t++)
614		bplist_create(&spa->spa_free_bplist[t]);
615
616	(void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name));
617	spa->spa_state = POOL_STATE_UNINITIALIZED;
618	spa->spa_freeze_txg = UINT64_MAX;
619	spa->spa_final_txg = UINT64_MAX;
620	spa->spa_load_max_txg = UINT64_MAX;
621	spa->spa_proc = &p0;
622	spa->spa_proc_state = SPA_PROC_NONE;
623
624	hdlr.cyh_func = spa_deadman;
625	hdlr.cyh_arg = spa;
626	hdlr.cyh_level = CY_LOW_LEVEL;
627
628	spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms);
629
630	/*
631	 * This determines how often we need to check for hung I/Os after
632	 * the cyclic has already fired. Since checking for hung I/Os is
633	 * an expensive operation we don't want to check too frequently.
634	 * Instead wait for 5 seconds before checking again.
635	 */
636	when.cyt_interval = MSEC2NSEC(zfs_deadman_checktime_ms);
637	when.cyt_when = CY_INFINITY;
638	mutex_enter(&cpu_lock);
639	spa->spa_deadman_cycid = cyclic_add(&hdlr, &when);
640	mutex_exit(&cpu_lock);
641
642	refcount_create(&spa->spa_refcount);
643	spa_config_lock_init(spa);
644
645	avl_add(&spa_namespace_avl, spa);
646
647	/*
648	 * Set the alternate root, if there is one.
649	 */
650	if (altroot) {
651		spa->spa_root = spa_strdup(altroot);
652		spa_active_count++;
653	}
654
655	avl_create(&spa->spa_alloc_tree, zio_bookmark_compare,
656	    sizeof (zio_t), offsetof(zio_t, io_alloc_node));
657
658	/*
659	 * Every pool starts with the default cachefile
660	 */
661	list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t),
662	    offsetof(spa_config_dirent_t, scd_link));
663
664	dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP);
665	dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path);
666	list_insert_head(&spa->spa_config_list, dp);
667
668	VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME,
669	    KM_SLEEP) == 0);
670
671	if (config != NULL) {
672		nvlist_t *features;
673
674		if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
675		    &features) == 0) {
676			VERIFY(nvlist_dup(features, &spa->spa_label_features,
677			    0) == 0);
678		}
679
680		VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0);
681	}
682
683	if (spa->spa_label_features == NULL) {
684		VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME,
685		    KM_SLEEP) == 0);
686	}
687
688	spa->spa_iokstat = kstat_create("zfs", 0, name,
689	    "disk", KSTAT_TYPE_IO, 1, 0);
690	if (spa->spa_iokstat) {
691		spa->spa_iokstat->ks_lock = &spa->spa_iokstat_lock;
692		kstat_install(spa->spa_iokstat);
693	}
694
695	spa->spa_debug = ((zfs_flags & ZFS_DEBUG_SPA) != 0);
696
697	spa->spa_min_ashift = INT_MAX;
698	spa->spa_max_ashift = 0;
699
700	/*
701	 * As a pool is being created, treat all features as disabled by
702	 * setting SPA_FEATURE_DISABLED for all entries in the feature
703	 * refcount cache.
704	 */
705	for (int i = 0; i < SPA_FEATURES; i++) {
706		spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED;
707	}
708
709	return (spa);
710}
711
712/*
713 * Removes a spa_t from the namespace, freeing up any memory used.  Requires
714 * spa_namespace_lock.  This is called only after the spa_t has been closed and
715 * deactivated.
716 */
717void
718spa_remove(spa_t *spa)
719{
720	spa_config_dirent_t *dp;
721
722	ASSERT(MUTEX_HELD(&spa_namespace_lock));
723	ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);
724	ASSERT3U(refcount_count(&spa->spa_refcount), ==, 0);
725
726	nvlist_free(spa->spa_config_splitting);
727
728	avl_remove(&spa_namespace_avl, spa);
729	cv_broadcast(&spa_namespace_cv);
730
731	if (spa->spa_root) {
732		spa_strfree(spa->spa_root);
733		spa_active_count--;
734	}
735
736	while ((dp = list_head(&spa->spa_config_list)) != NULL) {
737		list_remove(&spa->spa_config_list, dp);
738		if (dp->scd_path != NULL)
739			spa_strfree(dp->scd_path);
740		kmem_free(dp, sizeof (spa_config_dirent_t));
741	}
742
743	avl_destroy(&spa->spa_alloc_tree);
744	list_destroy(&spa->spa_config_list);
745
746	nvlist_free(spa->spa_label_features);
747	nvlist_free(spa->spa_load_info);
748	spa_config_set(spa, NULL);
749
750	mutex_enter(&cpu_lock);
751	if (spa->spa_deadman_cycid != CYCLIC_NONE)
752		cyclic_remove(spa->spa_deadman_cycid);
753	mutex_exit(&cpu_lock);
754	spa->spa_deadman_cycid = CYCLIC_NONE;
755
756	refcount_destroy(&spa->spa_refcount);
757
758	spa_config_lock_destroy(spa);
759
760	kstat_delete(spa->spa_iokstat);
761	spa->spa_iokstat = NULL;
762
763	for (int t = 0; t < TXG_SIZE; t++)
764		bplist_destroy(&spa->spa_free_bplist[t]);
765
766	zio_checksum_templates_free(spa);
767
768	cv_destroy(&spa->spa_async_cv);
769	cv_destroy(&spa->spa_evicting_os_cv);
770	cv_destroy(&spa->spa_proc_cv);
771	cv_destroy(&spa->spa_scrub_io_cv);
772	cv_destroy(&spa->spa_suspend_cv);
773
774	mutex_destroy(&spa->spa_alloc_lock);
775	mutex_destroy(&spa->spa_async_lock);
776	mutex_destroy(&spa->spa_errlist_lock);
777	mutex_destroy(&spa->spa_errlog_lock);
778	mutex_destroy(&spa->spa_evicting_os_lock);
779	mutex_destroy(&spa->spa_history_lock);
780	mutex_destroy(&spa->spa_proc_lock);
781	mutex_destroy(&spa->spa_props_lock);
782	mutex_destroy(&spa->spa_cksum_tmpls_lock);
783	mutex_destroy(&spa->spa_scrub_lock);
784	mutex_destroy(&spa->spa_suspend_lock);
785	mutex_destroy(&spa->spa_vdev_top_lock);
786	mutex_destroy(&spa->spa_iokstat_lock);
787
788	kmem_free(spa, sizeof (spa_t));
789}
790
791/*
792 * Given a pool, return the next pool in the namespace, or NULL if there is
793 * none.  If 'prev' is NULL, return the first pool.
794 */
795spa_t *
796spa_next(spa_t *prev)
797{
798	ASSERT(MUTEX_HELD(&spa_namespace_lock));
799
800	if (prev)
801		return (AVL_NEXT(&spa_namespace_avl, prev));
802	else
803		return (avl_first(&spa_namespace_avl));
804}
805
806/*
807 * ==========================================================================
808 * SPA refcount functions
809 * ==========================================================================
810 */
811
812/*
813 * Add a reference to the given spa_t.  Must have at least one reference, or
814 * have the namespace lock held.
815 */
816void
817spa_open_ref(spa_t *spa, void *tag)
818{
819	ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref ||
820	    MUTEX_HELD(&spa_namespace_lock));
821	(void) refcount_add(&spa->spa_refcount, tag);
822}
823
824/*
825 * Remove a reference to the given spa_t.  Must have at least one reference, or
826 * have the namespace lock held.
827 */
828void
829spa_close(spa_t *spa, void *tag)
830{
831	ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref ||
832	    MUTEX_HELD(&spa_namespace_lock));
833	(void) refcount_remove(&spa->spa_refcount, tag);
834}
835
836/*
837 * Remove a reference to the given spa_t held by a dsl dir that is
838 * being asynchronously released.  Async releases occur from a taskq
839 * performing eviction of dsl datasets and dirs.  The namespace lock
840 * isn't held and the hold by the object being evicted may contribute to
841 * spa_minref (e.g. dataset or directory released during pool export),
842 * so the asserts in spa_close() do not apply.
843 */
844void
845spa_async_close(spa_t *spa, void *tag)
846{
847	(void) refcount_remove(&spa->spa_refcount, tag);
848}
849
850/*
851 * Check to see if the spa refcount is zero.  Must be called with
852 * spa_namespace_lock held.  We really compare against spa_minref, which is the
853 * number of references acquired when opening a pool
854 */
855boolean_t
856spa_refcount_zero(spa_t *spa)
857{
858	ASSERT(MUTEX_HELD(&spa_namespace_lock));
859
860	return (refcount_count(&spa->spa_refcount) == spa->spa_minref);
861}
862
863/*
864 * ==========================================================================
865 * SPA spare and l2cache tracking
866 * ==========================================================================
867 */
868
869/*
870 * Hot spares and cache devices are tracked using the same code below,
871 * for 'auxiliary' devices.
872 */
873
874typedef struct spa_aux {
875	uint64_t	aux_guid;
876	uint64_t	aux_pool;
877	avl_node_t	aux_avl;
878	int		aux_count;
879} spa_aux_t;
880
881static int
882spa_aux_compare(const void *a, const void *b)
883{
884	const spa_aux_t *sa = a;
885	const spa_aux_t *sb = b;
886
887	if (sa->aux_guid < sb->aux_guid)
888		return (-1);
889	else if (sa->aux_guid > sb->aux_guid)
890		return (1);
891	else
892		return (0);
893}
894
895void
896spa_aux_add(vdev_t *vd, avl_tree_t *avl)
897{
898	avl_index_t where;
899	spa_aux_t search;
900	spa_aux_t *aux;
901
902	search.aux_guid = vd->vdev_guid;
903	if ((aux = avl_find(avl, &search, &where)) != NULL) {
904		aux->aux_count++;
905	} else {
906		aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP);
907		aux->aux_guid = vd->vdev_guid;
908		aux->aux_count = 1;
909		avl_insert(avl, aux, where);
910	}
911}
912
913void
914spa_aux_remove(vdev_t *vd, avl_tree_t *avl)
915{
916	spa_aux_t search;
917	spa_aux_t *aux;
918	avl_index_t where;
919
920	search.aux_guid = vd->vdev_guid;
921	aux = avl_find(avl, &search, &where);
922
923	ASSERT(aux != NULL);
924
925	if (--aux->aux_count == 0) {
926		avl_remove(avl, aux);
927		kmem_free(aux, sizeof (spa_aux_t));
928	} else if (aux->aux_pool == spa_guid(vd->vdev_spa)) {
929		aux->aux_pool = 0ULL;
930	}
931}
932
933boolean_t
934spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl)
935{
936	spa_aux_t search, *found;
937
938	search.aux_guid = guid;
939	found = avl_find(avl, &search, NULL);
940
941	if (pool) {
942		if (found)
943			*pool = found->aux_pool;
944		else
945			*pool = 0ULL;
946	}
947
948	if (refcnt) {
949		if (found)
950			*refcnt = found->aux_count;
951		else
952			*refcnt = 0;
953	}
954
955	return (found != NULL);
956}
957
958void
959spa_aux_activate(vdev_t *vd, avl_tree_t *avl)
960{
961	spa_aux_t search, *found;
962	avl_index_t where;
963
964	search.aux_guid = vd->vdev_guid;
965	found = avl_find(avl, &search, &where);
966	ASSERT(found != NULL);
967	ASSERT(found->aux_pool == 0ULL);
968
969	found->aux_pool = spa_guid(vd->vdev_spa);
970}
971
972/*
973 * Spares are tracked globally due to the following constraints:
974 *
975 * 	- A spare may be part of multiple pools.
976 * 	- A spare may be added to a pool even if it's actively in use within
977 *	  another pool.
978 * 	- A spare in use in any pool can only be the source of a replacement if
979 *	  the target is a spare in the same pool.
980 *
981 * We keep track of all spares on the system through the use of a reference
982 * counted AVL tree.  When a vdev is added as a spare, or used as a replacement
983 * spare, then we bump the reference count in the AVL tree.  In addition, we set
984 * the 'vdev_isspare' member to indicate that the device is a spare (active or
985 * inactive).  When a spare is made active (used to replace a device in the
986 * pool), we also keep track of which pool its been made a part of.
987 *
988 * The 'spa_spare_lock' protects the AVL tree.  These functions are normally
989 * called under the spa_namespace lock as part of vdev reconfiguration.  The
990 * separate spare lock exists for the status query path, which does not need to
991 * be completely consistent with respect to other vdev configuration changes.
992 */
993
994static int
995spa_spare_compare(const void *a, const void *b)
996{
997	return (spa_aux_compare(a, b));
998}
999
1000void
1001spa_spare_add(vdev_t *vd)
1002{
1003	mutex_enter(&spa_spare_lock);
1004	ASSERT(!vd->vdev_isspare);
1005	spa_aux_add(vd, &spa_spare_avl);
1006	vd->vdev_isspare = B_TRUE;
1007	mutex_exit(&spa_spare_lock);
1008}
1009
1010void
1011spa_spare_remove(vdev_t *vd)
1012{
1013	mutex_enter(&spa_spare_lock);
1014	ASSERT(vd->vdev_isspare);
1015	spa_aux_remove(vd, &spa_spare_avl);
1016	vd->vdev_isspare = B_FALSE;
1017	mutex_exit(&spa_spare_lock);
1018}
1019
1020boolean_t
1021spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt)
1022{
1023	boolean_t found;
1024
1025	mutex_enter(&spa_spare_lock);
1026	found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl);
1027	mutex_exit(&spa_spare_lock);
1028
1029	return (found);
1030}
1031
1032void
1033spa_spare_activate(vdev_t *vd)
1034{
1035	mutex_enter(&spa_spare_lock);
1036	ASSERT(vd->vdev_isspare);
1037	spa_aux_activate(vd, &spa_spare_avl);
1038	mutex_exit(&spa_spare_lock);
1039}
1040
1041/*
1042 * Level 2 ARC devices are tracked globally for the same reasons as spares.
1043 * Cache devices currently only support one pool per cache device, and so
1044 * for these devices the aux reference count is currently unused beyond 1.
1045 */
1046
1047static int
1048spa_l2cache_compare(const void *a, const void *b)
1049{
1050	return (spa_aux_compare(a, b));
1051}
1052
1053void
1054spa_l2cache_add(vdev_t *vd)
1055{
1056	mutex_enter(&spa_l2cache_lock);
1057	ASSERT(!vd->vdev_isl2cache);
1058	spa_aux_add(vd, &spa_l2cache_avl);
1059	vd->vdev_isl2cache = B_TRUE;
1060	mutex_exit(&spa_l2cache_lock);
1061}
1062
1063void
1064spa_l2cache_remove(vdev_t *vd)
1065{
1066	mutex_enter(&spa_l2cache_lock);
1067	ASSERT(vd->vdev_isl2cache);
1068	spa_aux_remove(vd, &spa_l2cache_avl);
1069	vd->vdev_isl2cache = B_FALSE;
1070	mutex_exit(&spa_l2cache_lock);
1071}
1072
1073boolean_t
1074spa_l2cache_exists(uint64_t guid, uint64_t *pool)
1075{
1076	boolean_t found;
1077
1078	mutex_enter(&spa_l2cache_lock);
1079	found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl);
1080	mutex_exit(&spa_l2cache_lock);
1081
1082	return (found);
1083}
1084
1085void
1086spa_l2cache_activate(vdev_t *vd)
1087{
1088	mutex_enter(&spa_l2cache_lock);
1089	ASSERT(vd->vdev_isl2cache);
1090	spa_aux_activate(vd, &spa_l2cache_avl);
1091	mutex_exit(&spa_l2cache_lock);
1092}
1093
1094/*
1095 * ==========================================================================
1096 * SPA vdev locking
1097 * ==========================================================================
1098 */
1099
1100/*
1101 * Lock the given spa_t for the purpose of adding or removing a vdev.
1102 * Grabs the global spa_namespace_lock plus the spa config lock for writing.
1103 * It returns the next transaction group for the spa_t.
1104 */
1105uint64_t
1106spa_vdev_enter(spa_t *spa)
1107{
1108	mutex_enter(&spa->spa_vdev_top_lock);
1109	mutex_enter(&spa_namespace_lock);
1110	return (spa_vdev_config_enter(spa));
1111}
1112
1113/*
1114 * Internal implementation for spa_vdev_enter().  Used when a vdev
1115 * operation requires multiple syncs (i.e. removing a device) while
1116 * keeping the spa_namespace_lock held.
1117 */
1118uint64_t
1119spa_vdev_config_enter(spa_t *spa)
1120{
1121	ASSERT(MUTEX_HELD(&spa_namespace_lock));
1122
1123	spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);
1124
1125	return (spa_last_synced_txg(spa) + 1);
1126}
1127
1128/*
1129 * Used in combination with spa_vdev_config_enter() to allow the syncing
1130 * of multiple transactions without releasing the spa_namespace_lock.
1131 */
1132void
1133spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag)
1134{
1135	ASSERT(MUTEX_HELD(&spa_namespace_lock));
1136
1137	int config_changed = B_FALSE;
1138
1139	ASSERT(txg > spa_last_synced_txg(spa));
1140
1141	spa->spa_pending_vdev = NULL;
1142
1143	/*
1144	 * Reassess the DTLs.
1145	 */
1146	vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE);
1147
1148	if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) {
1149		config_changed = B_TRUE;
1150		spa->spa_config_generation++;
1151	}
1152
1153	/*
1154	 * Verify the metaslab classes.
1155	 */
1156	ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0);
1157	ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0);
1158
1159	spa_config_exit(spa, SCL_ALL, spa);
1160
1161	/*
1162	 * Panic the system if the specified tag requires it.  This
1163	 * is useful for ensuring that configurations are updated
1164	 * transactionally.
1165	 */
1166	if (zio_injection_enabled)
1167		zio_handle_panic_injection(spa, tag, 0);
1168
1169	/*
1170	 * Note: this txg_wait_synced() is important because it ensures
1171	 * that there won't be more than one config change per txg.
1172	 * This allows us to use the txg as the generation number.
1173	 */
1174	if (error == 0)
1175		txg_wait_synced(spa->spa_dsl_pool, txg);
1176
1177	if (vd != NULL) {
1178		ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL);
1179		spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);
1180		vdev_free(vd);
1181		spa_config_exit(spa, SCL_ALL, spa);
1182	}
1183
1184	/*
1185	 * If the config changed, update the config cache.
1186	 */
1187	if (config_changed)
1188		spa_write_cachefile(spa, B_FALSE, B_TRUE);
1189}
1190
1191/*
1192 * Unlock the spa_t after adding or removing a vdev.  Besides undoing the
1193 * locking of spa_vdev_enter(), we also want make sure the transactions have
1194 * synced to disk, and then update the global configuration cache with the new
1195 * information.
1196 */
1197int
1198spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error)
1199{
1200	spa_vdev_config_exit(spa, vd, txg, error, FTAG);
1201	mutex_exit(&spa_namespace_lock);
1202	mutex_exit(&spa->spa_vdev_top_lock);
1203
1204	return (error);
1205}
1206
1207/*
1208 * Lock the given spa_t for the purpose of changing vdev state.
1209 */
1210void
1211spa_vdev_state_enter(spa_t *spa, int oplocks)
1212{
1213	int locks = SCL_STATE_ALL | oplocks;
1214
1215	/*
1216	 * Root pools may need to read of the underlying devfs filesystem
1217	 * when opening up a vdev.  Unfortunately if we're holding the
1218	 * SCL_ZIO lock it will result in a deadlock when we try to issue
1219	 * the read from the root filesystem.  Instead we "prefetch"
1220	 * the associated vnodes that we need prior to opening the
1221	 * underlying devices and cache them so that we can prevent
1222	 * any I/O when we are doing the actual open.
1223	 */
1224	if (spa_is_root(spa)) {
1225		int low = locks & ~(SCL_ZIO - 1);
1226		int high = locks & ~low;
1227
1228		spa_config_enter(spa, high, spa, RW_WRITER);
1229		vdev_hold(spa->spa_root_vdev);
1230		spa_config_enter(spa, low, spa, RW_WRITER);
1231	} else {
1232		spa_config_enter(spa, locks, spa, RW_WRITER);
1233	}
1234	spa->spa_vdev_locks = locks;
1235}
1236
1237int
1238spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error)
1239{
1240	boolean_t config_changed = B_FALSE;
1241
1242	if (vd != NULL || error == 0)
1243		vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev,
1244		    0, 0, B_FALSE);
1245
1246	if (vd != NULL) {
1247		vdev_state_dirty(vd->vdev_top);
1248		config_changed = B_TRUE;
1249		spa->spa_config_generation++;
1250	}
1251
1252	if (spa_is_root(spa))
1253		vdev_rele(spa->spa_root_vdev);
1254
1255	ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL);
1256	spa_config_exit(spa, spa->spa_vdev_locks, spa);
1257
1258	/*
1259	 * If anything changed, wait for it to sync.  This ensures that,
1260	 * from the system administrator's perspective, zpool(1M) commands
1261	 * are synchronous.  This is important for things like zpool offline:
1262	 * when the command completes, you expect no further I/O from ZFS.
1263	 */
1264	if (vd != NULL)
1265		txg_wait_synced(spa->spa_dsl_pool, 0);
1266
1267	/*
1268	 * If the config changed, update the config cache.
1269	 */
1270	if (config_changed) {
1271		mutex_enter(&spa_namespace_lock);
1272		spa_write_cachefile(spa, B_FALSE, B_TRUE);
1273		mutex_exit(&spa_namespace_lock);
1274	}
1275
1276	return (error);
1277}
1278
1279/*
1280 * ==========================================================================
1281 * Miscellaneous functions
1282 * ==========================================================================
1283 */
1284
1285void
1286spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx)
1287{
1288	if (!nvlist_exists(spa->spa_label_features, feature)) {
1289		fnvlist_add_boolean(spa->spa_label_features, feature);
1290		/*
1291		 * When we are creating the pool (tx_txg==TXG_INITIAL), we can't
1292		 * dirty the vdev config because lock SCL_CONFIG is not held.
1293		 * Thankfully, in this case we don't need to dirty the config
1294		 * because it will be written out anyway when we finish
1295		 * creating the pool.
1296		 */
1297		if (tx->tx_txg != TXG_INITIAL)
1298			vdev_config_dirty(spa->spa_root_vdev);
1299	}
1300}
1301
1302void
1303spa_deactivate_mos_feature(spa_t *spa, const char *feature)
1304{
1305	if (nvlist_remove_all(spa->spa_label_features, feature) == 0)
1306		vdev_config_dirty(spa->spa_root_vdev);
1307}
1308
1309/*
1310 * Rename a spa_t.
1311 */
1312int
1313spa_rename(const char *name, const char *newname)
1314{
1315	spa_t *spa;
1316	int err;
1317
1318	/*
1319	 * Lookup the spa_t and grab the config lock for writing.  We need to
1320	 * actually open the pool so that we can sync out the necessary labels.
1321	 * It's OK to call spa_open() with the namespace lock held because we
1322	 * allow recursive calls for other reasons.
1323	 */
1324	mutex_enter(&spa_namespace_lock);
1325	if ((err = spa_open(name, &spa, FTAG)) != 0) {
1326		mutex_exit(&spa_namespace_lock);
1327		return (err);
1328	}
1329
1330	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
1331
1332	avl_remove(&spa_namespace_avl, spa);
1333	(void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name));
1334	avl_add(&spa_namespace_avl, spa);
1335
1336	/*
1337	 * Sync all labels to disk with the new names by marking the root vdev
1338	 * dirty and waiting for it to sync.  It will pick up the new pool name
1339	 * during the sync.
1340	 */
1341	vdev_config_dirty(spa->spa_root_vdev);
1342
1343	spa_config_exit(spa, SCL_ALL, FTAG);
1344
1345	txg_wait_synced(spa->spa_dsl_pool, 0);
1346
1347	/*
1348	 * Sync the updated config cache.
1349	 */
1350	spa_write_cachefile(spa, B_FALSE, B_TRUE);
1351
1352	spa_close(spa, FTAG);
1353
1354	mutex_exit(&spa_namespace_lock);
1355
1356	return (0);
1357}
1358
1359/*
1360 * Return the spa_t associated with given pool_guid, if it exists.  If
1361 * device_guid is non-zero, determine whether the pool exists *and* contains
1362 * a device with the specified device_guid.
1363 */
1364spa_t *
1365spa_by_guid(uint64_t pool_guid, uint64_t device_guid)
1366{
1367	spa_t *spa;
1368	avl_tree_t *t = &spa_namespace_avl;
1369
1370	ASSERT(MUTEX_HELD(&spa_namespace_lock));
1371
1372	for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) {
1373		if (spa->spa_state == POOL_STATE_UNINITIALIZED)
1374			continue;
1375		if (spa->spa_root_vdev == NULL)
1376			continue;
1377		if (spa_guid(spa) == pool_guid) {
1378			if (device_guid == 0)
1379				break;
1380
1381			if (vdev_lookup_by_guid(spa->spa_root_vdev,
1382			    device_guid) != NULL)
1383				break;
1384
1385			/*
1386			 * Check any devices we may be in the process of adding.
1387			 */
1388			if (spa->spa_pending_vdev) {
1389				if (vdev_lookup_by_guid(spa->spa_pending_vdev,
1390				    device_guid) != NULL)
1391					break;
1392			}
1393		}
1394	}
1395
1396	return (spa);
1397}
1398
1399/*
1400 * Determine whether a pool with the given pool_guid exists.
1401 */
1402boolean_t
1403spa_guid_exists(uint64_t pool_guid, uint64_t device_guid)
1404{
1405	return (spa_by_guid(pool_guid, device_guid) != NULL);
1406}
1407
1408char *
1409spa_strdup(const char *s)
1410{
1411	size_t len;
1412	char *new;
1413
1414	len = strlen(s);
1415	new = kmem_alloc(len + 1, KM_SLEEP);
1416	bcopy(s, new, len);
1417	new[len] = '\0';
1418
1419	return (new);
1420}
1421
1422void
1423spa_strfree(char *s)
1424{
1425	kmem_free(s, strlen(s) + 1);
1426}
1427
1428uint64_t
1429spa_get_random(uint64_t range)
1430{
1431	uint64_t r;
1432
1433	ASSERT(range != 0);
1434
1435	(void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t));
1436
1437	return (r % range);
1438}
1439
1440uint64_t
1441spa_generate_guid(spa_t *spa)
1442{
1443	uint64_t guid = spa_get_random(-1ULL);
1444
1445	if (spa != NULL) {
1446		while (guid == 0 || spa_guid_exists(spa_guid(spa), guid))
1447			guid = spa_get_random(-1ULL);
1448	} else {
1449		while (guid == 0 || spa_guid_exists(guid, 0))
1450			guid = spa_get_random(-1ULL);
1451	}
1452
1453	return (guid);
1454}
1455
1456void
1457snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp)
1458{
1459	char type[256];
1460	char *checksum = NULL;
1461	char *compress = NULL;
1462
1463	if (bp != NULL) {
1464		if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) {
1465			dmu_object_byteswap_t bswap =
1466			    DMU_OT_BYTESWAP(BP_GET_TYPE(bp));
1467			(void) snprintf(type, sizeof (type), "bswap %s %s",
1468			    DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ?
1469			    "metadata" : "data",
1470			    dmu_ot_byteswap[bswap].ob_name);
1471		} else {
1472			(void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name,
1473			    sizeof (type));
1474		}
1475		if (!BP_IS_EMBEDDED(bp)) {
1476			checksum =
1477			    zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name;
1478		}
1479		compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name;
1480	}
1481
1482	SNPRINTF_BLKPTR(snprintf, ' ', buf, buflen, bp, type, checksum,
1483	    compress);
1484}
1485
1486void
1487spa_freeze(spa_t *spa)
1488{
1489	uint64_t freeze_txg = 0;
1490
1491	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
1492	if (spa->spa_freeze_txg == UINT64_MAX) {
1493		freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE;
1494		spa->spa_freeze_txg = freeze_txg;
1495	}
1496	spa_config_exit(spa, SCL_ALL, FTAG);
1497	if (freeze_txg != 0)
1498		txg_wait_synced(spa_get_dsl(spa), freeze_txg);
1499}
1500
1501void
1502zfs_panic_recover(const char *fmt, ...)
1503{
1504	va_list adx;
1505
1506	va_start(adx, fmt);
1507	vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx);
1508	va_end(adx);
1509}
1510
1511/*
1512 * This is a stripped-down version of strtoull, suitable only for converting
1513 * lowercase hexadecimal numbers that don't overflow.
1514 */
1515uint64_t
1516zfs_strtonum(const char *str, char **nptr)
1517{
1518	uint64_t val = 0;
1519	char c;
1520	int digit;
1521
1522	while ((c = *str) != '\0') {
1523		if (c >= '0' && c <= '9')
1524			digit = c - '0';
1525		else if (c >= 'a' && c <= 'f')
1526			digit = 10 + c - 'a';
1527		else
1528			break;
1529
1530		val *= 16;
1531		val += digit;
1532
1533		str++;
1534	}
1535
1536	if (nptr)
1537		*nptr = (char *)str;
1538
1539	return (val);
1540}
1541
1542/*
1543 * ==========================================================================
1544 * Accessor functions
1545 * ==========================================================================
1546 */
1547
1548boolean_t
1549spa_shutting_down(spa_t *spa)
1550{
1551	return (spa->spa_async_suspended);
1552}
1553
1554dsl_pool_t *
1555spa_get_dsl(spa_t *spa)
1556{
1557	return (spa->spa_dsl_pool);
1558}
1559
1560boolean_t
1561spa_is_initializing(spa_t *spa)
1562{
1563	return (spa->spa_is_initializing);
1564}
1565
1566boolean_t
1567spa_indirect_vdevs_loaded(spa_t *spa)
1568{
1569	return (spa->spa_indirect_vdevs_loaded);
1570}
1571
1572blkptr_t *
1573spa_get_rootblkptr(spa_t *spa)
1574{
1575	return (&spa->spa_ubsync.ub_rootbp);
1576}
1577
1578void
1579spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp)
1580{
1581	spa->spa_uberblock.ub_rootbp = *bp;
1582}
1583
1584void
1585spa_altroot(spa_t *spa, char *buf, size_t buflen)
1586{
1587	if (spa->spa_root == NULL)
1588		buf[0] = '\0';
1589	else
1590		(void) strncpy(buf, spa->spa_root, buflen);
1591}
1592
1593int
1594spa_sync_pass(spa_t *spa)
1595{
1596	return (spa->spa_sync_pass);
1597}
1598
1599char *
1600spa_name(spa_t *spa)
1601{
1602	return (spa->spa_name);
1603}
1604
1605uint64_t
1606spa_guid(spa_t *spa)
1607{
1608	dsl_pool_t *dp = spa_get_dsl(spa);
1609	uint64_t guid;
1610
1611	/*
1612	 * If we fail to parse the config during spa_load(), we can go through
1613	 * the error path (which posts an ereport) and end up here with no root
1614	 * vdev.  We stash the original pool guid in 'spa_config_guid' to handle
1615	 * this case.
1616	 */
1617	if (spa->spa_root_vdev == NULL)
1618		return (spa->spa_config_guid);
1619
1620	guid = spa->spa_last_synced_guid != 0 ?
1621	    spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid;
1622
1623	/*
1624	 * Return the most recently synced out guid unless we're
1625	 * in syncing context.
1626	 */
1627	if (dp && dsl_pool_sync_context(dp))
1628		return (spa->spa_root_vdev->vdev_guid);
1629	else
1630		return (guid);
1631}
1632
1633uint64_t
1634spa_load_guid(spa_t *spa)
1635{
1636	/*
1637	 * This is a GUID that exists solely as a reference for the
1638	 * purposes of the arc.  It is generated at load time, and
1639	 * is never written to persistent storage.
1640	 */
1641	return (spa->spa_load_guid);
1642}
1643
1644uint64_t
1645spa_last_synced_txg(spa_t *spa)
1646{
1647	return (spa->spa_ubsync.ub_txg);
1648}
1649
1650uint64_t
1651spa_first_txg(spa_t *spa)
1652{
1653	return (spa->spa_first_txg);
1654}
1655
1656uint64_t
1657spa_syncing_txg(spa_t *spa)
1658{
1659	return (spa->spa_syncing_txg);
1660}
1661
1662/*
1663 * Return the last txg where data can be dirtied. The final txgs
1664 * will be used to just clear out any deferred frees that remain.
1665 */
1666uint64_t
1667spa_final_dirty_txg(spa_t *spa)
1668{
1669	return (spa->spa_final_txg - TXG_DEFER_SIZE);
1670}
1671
1672pool_state_t
1673spa_state(spa_t *spa)
1674{
1675	return (spa->spa_state);
1676}
1677
1678spa_load_state_t
1679spa_load_state(spa_t *spa)
1680{
1681	return (spa->spa_load_state);
1682}
1683
1684uint64_t
1685spa_freeze_txg(spa_t *spa)
1686{
1687	return (spa->spa_freeze_txg);
1688}
1689
1690/* ARGSUSED */
1691uint64_t
1692spa_get_worst_case_asize(spa_t *spa, uint64_t lsize)
1693{
1694	return (lsize * spa_asize_inflation);
1695}
1696
1697/*
1698 * Return the amount of slop space in bytes.  It is 1/32 of the pool (3.2%),
1699 * or at least 128MB, unless that would cause it to be more than half the
1700 * pool size.
1701 *
1702 * See the comment above spa_slop_shift for details.
1703 */
1704uint64_t
1705spa_get_slop_space(spa_t *spa)
1706{
1707	uint64_t space = spa_get_dspace(spa);
1708	return (MAX(space >> spa_slop_shift, MIN(space >> 1, spa_min_slop)));
1709}
1710
1711uint64_t
1712spa_get_dspace(spa_t *spa)
1713{
1714	return (spa->spa_dspace);
1715}
1716
1717void
1718spa_update_dspace(spa_t *spa)
1719{
1720	spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) +
1721	    ddt_get_dedup_dspace(spa);
1722	if (spa->spa_vdev_removal != NULL) {
1723		/*
1724		 * We can't allocate from the removing device, so
1725		 * subtract its size.  This prevents the DMU/DSL from
1726		 * filling up the (now smaller) pool while we are in the
1727		 * middle of removing the device.
1728		 *
1729		 * Note that the DMU/DSL doesn't actually know or care
1730		 * how much space is allocated (it does its own tracking
1731		 * of how much space has been logically used).  So it
1732		 * doesn't matter that the data we are moving may be
1733		 * allocated twice (on the old device and the new
1734		 * device).
1735		 */
1736		vdev_t *vd = spa->spa_vdev_removal->svr_vdev;
1737		spa->spa_dspace -= spa_deflate(spa) ?
1738		    vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;
1739	}
1740}
1741
1742/*
1743 * Return the failure mode that has been set to this pool. The default
1744 * behavior will be to block all I/Os when a complete failure occurs.
1745 */
1746uint8_t
1747spa_get_failmode(spa_t *spa)
1748{
1749	return (spa->spa_failmode);
1750}
1751
1752boolean_t
1753spa_suspended(spa_t *spa)
1754{
1755	return (spa->spa_suspended);
1756}
1757
1758uint64_t
1759spa_version(spa_t *spa)
1760{
1761	return (spa->spa_ubsync.ub_version);
1762}
1763
1764boolean_t
1765spa_deflate(spa_t *spa)
1766{
1767	return (spa->spa_deflate);
1768}
1769
1770metaslab_class_t *
1771spa_normal_class(spa_t *spa)
1772{
1773	return (spa->spa_normal_class);
1774}
1775
1776metaslab_class_t *
1777spa_log_class(spa_t *spa)
1778{
1779	return (spa->spa_log_class);
1780}
1781
1782void
1783spa_evicting_os_register(spa_t *spa, objset_t *os)
1784{
1785	mutex_enter(&spa->spa_evicting_os_lock);
1786	list_insert_head(&spa->spa_evicting_os_list, os);
1787	mutex_exit(&spa->spa_evicting_os_lock);
1788}
1789
1790void
1791spa_evicting_os_deregister(spa_t *spa, objset_t *os)
1792{
1793	mutex_enter(&spa->spa_evicting_os_lock);
1794	list_remove(&spa->spa_evicting_os_list, os);
1795	cv_broadcast(&spa->spa_evicting_os_cv);
1796	mutex_exit(&spa->spa_evicting_os_lock);
1797}
1798
1799void
1800spa_evicting_os_wait(spa_t *spa)
1801{
1802	mutex_enter(&spa->spa_evicting_os_lock);
1803	while (!list_is_empty(&spa->spa_evicting_os_list))
1804		cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock);
1805	mutex_exit(&spa->spa_evicting_os_lock);
1806
1807	dmu_buf_user_evict_wait();
1808}
1809
1810int
1811spa_max_replication(spa_t *spa)
1812{
1813	/*
1814	 * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to
1815	 * handle BPs with more than one DVA allocated.  Set our max
1816	 * replication level accordingly.
1817	 */
1818	if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS)
1819		return (1);
1820	return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override));
1821}
1822
1823int
1824spa_prev_software_version(spa_t *spa)
1825{
1826	return (spa->spa_prev_software_version);
1827}
1828
1829uint64_t
1830spa_deadman_synctime(spa_t *spa)
1831{
1832	return (spa->spa_deadman_synctime);
1833}
1834
1835uint64_t
1836dva_get_dsize_sync(spa_t *spa, const dva_t *dva)
1837{
1838	uint64_t asize = DVA_GET_ASIZE(dva);
1839	uint64_t dsize = asize;
1840
1841	ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
1842
1843	if (asize != 0 && spa->spa_deflate) {
1844		vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));
1845		dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio;
1846	}
1847
1848	return (dsize);
1849}
1850
1851uint64_t
1852bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp)
1853{
1854	uint64_t dsize = 0;
1855
1856	for (int d = 0; d < BP_GET_NDVAS(bp); d++)
1857		dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);
1858
1859	return (dsize);
1860}
1861
1862uint64_t
1863bp_get_dsize(spa_t *spa, const blkptr_t *bp)
1864{
1865	uint64_t dsize = 0;
1866
1867	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
1868
1869	for (int d = 0; d < BP_GET_NDVAS(bp); d++)
1870		dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);
1871
1872	spa_config_exit(spa, SCL_VDEV, FTAG);
1873
1874	return (dsize);
1875}
1876
1877/*
1878 * ==========================================================================
1879 * Initialization and Termination
1880 * ==========================================================================
1881 */
1882
1883static int
1884spa_name_compare(const void *a1, const void *a2)
1885{
1886	const spa_t *s1 = a1;
1887	const spa_t *s2 = a2;
1888	int s;
1889
1890	s = strcmp(s1->spa_name, s2->spa_name);
1891	if (s > 0)
1892		return (1);
1893	if (s < 0)
1894		return (-1);
1895	return (0);
1896}
1897
1898int
1899spa_busy(void)
1900{
1901	return (spa_active_count);
1902}
1903
1904void
1905spa_boot_init()
1906{
1907	spa_config_load();
1908}
1909
1910void
1911spa_init(int mode)
1912{
1913	mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL);
1914	mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL);
1915	mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL);
1916	cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL);
1917
1918	avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t),
1919	    offsetof(spa_t, spa_avl));
1920
1921	avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t),
1922	    offsetof(spa_aux_t, aux_avl));
1923
1924	avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t),
1925	    offsetof(spa_aux_t, aux_avl));
1926
1927	spa_mode_global = mode;
1928
1929#ifdef _KERNEL
1930	spa_arch_init();
1931#else
1932	if (spa_mode_global != FREAD && dprintf_find_string("watch")) {
1933		arc_procfd = open("/proc/self/ctl", O_WRONLY);
1934		if (arc_procfd == -1) {
1935			perror("could not enable watchpoints: "
1936			    "opening /proc/self/ctl failed: ");
1937		} else {
1938			arc_watch = B_TRUE;
1939		}
1940	}
1941#endif
1942
1943	refcount_init();
1944	unique_init();
1945	range_tree_init();
1946	metaslab_alloc_trace_init();
1947	zio_init();
1948	dmu_init();
1949	zil_init();
1950	vdev_cache_stat_init();
1951	zfs_prop_init();
1952	zpool_prop_init();
1953	zpool_feature_init();
1954	spa_config_load();
1955	l2arc_start();
1956}
1957
1958void
1959spa_fini(void)
1960{
1961	l2arc_stop();
1962
1963	spa_evict_all();
1964
1965	vdev_cache_stat_fini();
1966	zil_fini();
1967	dmu_fini();
1968	zio_fini();
1969	metaslab_alloc_trace_fini();
1970	range_tree_fini();
1971	unique_fini();
1972	refcount_fini();
1973
1974	avl_destroy(&spa_namespace_avl);
1975	avl_destroy(&spa_spare_avl);
1976	avl_destroy(&spa_l2cache_avl);
1977
1978	cv_destroy(&spa_namespace_cv);
1979	mutex_destroy(&spa_namespace_lock);
1980	mutex_destroy(&spa_spare_lock);
1981	mutex_destroy(&spa_l2cache_lock);
1982}
1983
1984/*
1985 * Return whether this pool has slogs. No locking needed.
1986 * It's not a problem if the wrong answer is returned as it's only for
1987 * performance and not correctness
1988 */
1989boolean_t
1990spa_has_slogs(spa_t *spa)
1991{
1992	return (spa->spa_log_class->mc_rotor != NULL);
1993}
1994
1995spa_log_state_t
1996spa_get_log_state(spa_t *spa)
1997{
1998	return (spa->spa_log_state);
1999}
2000
2001void
2002spa_set_log_state(spa_t *spa, spa_log_state_t state)
2003{
2004	spa->spa_log_state = state;
2005}
2006
2007boolean_t
2008spa_is_root(spa_t *spa)
2009{
2010	return (spa->spa_is_root);
2011}
2012
2013boolean_t
2014spa_writeable(spa_t *spa)
2015{
2016	return (!!(spa->spa_mode & FWRITE));
2017}
2018
2019/*
2020 * Returns true if there is a pending sync task in any of the current
2021 * syncing txg, the current quiescing txg, or the current open txg.
2022 */
2023boolean_t
2024spa_has_pending_synctask(spa_t *spa)
2025{
2026	return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks));
2027}
2028
2029int
2030spa_mode(spa_t *spa)
2031{
2032	return (spa->spa_mode);
2033}
2034
2035uint64_t
2036spa_bootfs(spa_t *spa)
2037{
2038	return (spa->spa_bootfs);
2039}
2040
2041uint64_t
2042spa_delegation(spa_t *spa)
2043{
2044	return (spa->spa_delegation);
2045}
2046
2047objset_t *
2048spa_meta_objset(spa_t *spa)
2049{
2050	return (spa->spa_meta_objset);
2051}
2052
2053enum zio_checksum
2054spa_dedup_checksum(spa_t *spa)
2055{
2056	return (spa->spa_dedup_checksum);
2057}
2058
2059/*
2060 * Reset pool scan stat per scan pass (or reboot).
2061 */
2062void
2063spa_scan_stat_init(spa_t *spa)
2064{
2065	/* data not stored on disk */
2066	spa->spa_scan_pass_start = gethrestime_sec();
2067	if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan))
2068		spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start;
2069	else
2070		spa->spa_scan_pass_scrub_pause = 0;
2071	spa->spa_scan_pass_scrub_spent_paused = 0;
2072	spa->spa_scan_pass_exam = 0;
2073	vdev_scan_stat_init(spa->spa_root_vdev);
2074}
2075
2076/*
2077 * Get scan stats for zpool status reports
2078 */
2079int
2080spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps)
2081{
2082	dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL;
2083
2084	if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE)
2085		return (SET_ERROR(ENOENT));
2086	bzero(ps, sizeof (pool_scan_stat_t));
2087
2088	/* data stored on disk */
2089	ps->pss_func = scn->scn_phys.scn_func;
2090	ps->pss_start_time = scn->scn_phys.scn_start_time;
2091	ps->pss_end_time = scn->scn_phys.scn_end_time;
2092	ps->pss_to_examine = scn->scn_phys.scn_to_examine;
2093	ps->pss_examined = scn->scn_phys.scn_examined;
2094	ps->pss_to_process = scn->scn_phys.scn_to_process;
2095	ps->pss_processed = scn->scn_phys.scn_processed;
2096	ps->pss_errors = scn->scn_phys.scn_errors;
2097	ps->pss_state = scn->scn_phys.scn_state;
2098
2099	/* data not stored on disk */
2100	ps->pss_pass_start = spa->spa_scan_pass_start;
2101	ps->pss_pass_exam = spa->spa_scan_pass_exam;
2102	ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause;
2103	ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused;
2104
2105	return (0);
2106}
2107
2108boolean_t
2109spa_debug_enabled(spa_t *spa)
2110{
2111	return (spa->spa_debug);
2112}
2113
2114int
2115spa_maxblocksize(spa_t *spa)
2116{
2117	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS))
2118		return (SPA_MAXBLOCKSIZE);
2119	else
2120		return (SPA_OLD_MAXBLOCKSIZE);
2121}
2122
2123/*
2124 * Returns the txg that the last device removal completed. No indirect mappings
2125 * have been added since this txg.
2126 */
2127uint64_t
2128spa_get_last_removal_txg(spa_t *spa)
2129{
2130	uint64_t vdevid;
2131	uint64_t ret = -1ULL;
2132
2133	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
2134	/*
2135	 * sr_prev_indirect_vdev is only modified while holding all the
2136	 * config locks, so it is sufficient to hold SCL_VDEV as reader when
2137	 * examining it.
2138	 */
2139	vdevid = spa->spa_removing_phys.sr_prev_indirect_vdev;
2140
2141	while (vdevid != -1ULL) {
2142		vdev_t *vd = vdev_lookup_top(spa, vdevid);
2143		vdev_indirect_births_t *vib = vd->vdev_indirect_births;
2144
2145		ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
2146
2147		/*
2148		 * If the removal did not remap any data, we don't care.
2149		 */
2150		if (vdev_indirect_births_count(vib) != 0) {
2151			ret = vdev_indirect_births_last_entry_txg(vib);
2152			break;
2153		}
2154
2155		vdevid = vd->vdev_indirect_config.vic_prev_indirect_vdev;
2156	}
2157	spa_config_exit(spa, SCL_VDEV, FTAG);
2158
2159	IMPLY(ret != -1ULL,
2160	    spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL));
2161
2162	return (ret);
2163}
2164