1fa9e4066Sahrens /* 2fa9e4066Sahrens * CDDL HEADER START 3fa9e4066Sahrens * 4fa9e4066Sahrens * The contents of this file are subject to the terms of the 5ea8dc4b6Seschrock * Common Development and Distribution License (the "License"). 6ea8dc4b6Seschrock * You may not use this file except in compliance with the License. 7fa9e4066Sahrens * 8fa9e4066Sahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9fa9e4066Sahrens * or http://www.opensolaris.org/os/licensing. 10fa9e4066Sahrens * See the License for the specific language governing permissions 11fa9e4066Sahrens * and limitations under the License. 12fa9e4066Sahrens * 13fa9e4066Sahrens * When distributing Covered Code, include this CDDL HEADER in each 14fa9e4066Sahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15fa9e4066Sahrens * If applicable, add the following below this CDDL HEADER, with the 16fa9e4066Sahrens * fields enclosed by brackets "[]" replaced with your own identifying 17fa9e4066Sahrens * information: Portions Copyright [yyyy] [name of copyright owner] 18fa9e4066Sahrens * 19fa9e4066Sahrens * CDDL HEADER END 20fa9e4066Sahrens */ 21fa9e4066Sahrens /* 229842588bSGeorge Wilson * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23*ad135b5dSChristopher Siden * Copyright (c) 2012 by Delphix. All rights reserved. 24e9103aaeSGarrett D'Amore * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 25fa9e4066Sahrens */ 26fa9e4066Sahrens 27fa9e4066Sahrens #include <sys/zfs_context.h> 28fa9e4066Sahrens #include <sys/spa_impl.h> 29fa9e4066Sahrens #include <sys/zio.h> 30fa9e4066Sahrens #include <sys/zio_checksum.h> 31fa9e4066Sahrens #include <sys/zio_compress.h> 32fa9e4066Sahrens #include <sys/dmu.h> 33fa9e4066Sahrens #include <sys/dmu_tx.h> 34fa9e4066Sahrens #include <sys/zap.h> 35fa9e4066Sahrens #include <sys/zil.h> 36fa9e4066Sahrens #include <sys/vdev_impl.h> 37fa9e4066Sahrens #include <sys/metaslab.h> 38fa9e4066Sahrens #include <sys/uberblock_impl.h> 39fa9e4066Sahrens #include <sys/txg.h> 40fa9e4066Sahrens #include <sys/avl.h> 41fa9e4066Sahrens #include <sys/unique.h> 42fa9e4066Sahrens #include <sys/dsl_pool.h> 43fa9e4066Sahrens #include <sys/dsl_dir.h> 44fa9e4066Sahrens #include <sys/dsl_prop.h> 453f9d6ad7SLin Ling #include <sys/dsl_scan.h> 46fa9e4066Sahrens #include <sys/fs/zfs.h> 476ce0521aSperrin #include <sys/metaslab_impl.h> 48e14bb325SJeff Bonwick #include <sys/arc.h> 49485bbbf5SGeorge Wilson #include <sys/ddt.h> 5091ebeef5Sahrens #include "zfs_prop.h" 51*ad135b5dSChristopher Siden #include "zfeature_common.h" 52fa9e4066Sahrens 53fa9e4066Sahrens /* 54fa9e4066Sahrens * SPA locking 55fa9e4066Sahrens * 56fa9e4066Sahrens * There are four basic locks for managing spa_t structures: 57fa9e4066Sahrens * 58fa9e4066Sahrens * spa_namespace_lock (global mutex) 59fa9e4066Sahrens * 6044cd46caSbillm * This lock must be acquired to do any of the following: 61fa9e4066Sahrens * 6244cd46caSbillm * - Lookup a spa_t by name 6344cd46caSbillm * - Add or remove a spa_t from the namespace 6444cd46caSbillm * - Increase spa_refcount from non-zero 6544cd46caSbillm * - Check if spa_refcount is zero 6644cd46caSbillm * - Rename a spa_t 67ea8dc4b6Seschrock * - add/remove/attach/detach devices 6844cd46caSbillm * - Held for the duration of create/destroy/import/export 69fa9e4066Sahrens * 7044cd46caSbillm * It does not need to handle recursion. A create or destroy may 7144cd46caSbillm * reference objects (files or zvols) in other pools, but by 7244cd46caSbillm * definition they must have an existing reference, and will never need 7344cd46caSbillm * to lookup a spa_t by name. 74fa9e4066Sahrens * 75fa9e4066Sahrens * spa_refcount (per-spa refcount_t protected by mutex) 76fa9e4066Sahrens * 7744cd46caSbillm * This reference count keep track of any active users of the spa_t. The 7844cd46caSbillm * spa_t cannot be destroyed or freed while this is non-zero. Internally, 7944cd46caSbillm * the refcount is never really 'zero' - opening a pool implicitly keeps 80088f3894Sahrens * some references in the DMU. Internally we check against spa_minref, but 8144cd46caSbillm * present the image of a zero/non-zero value to consumers. 82fa9e4066Sahrens * 83e14bb325SJeff Bonwick * spa_config_lock[] (per-spa array of rwlocks) 84fa9e4066Sahrens * 8591ebeef5Sahrens * This protects the spa_t from config changes, and must be held in 8691ebeef5Sahrens * the following circumstances: 87fa9e4066Sahrens * 8844cd46caSbillm * - RW_READER to perform I/O to the spa 8944cd46caSbillm * - RW_WRITER to change the vdev config 90fa9e4066Sahrens * 91fa9e4066Sahrens * The locking order is fairly straightforward: 92fa9e4066Sahrens * 9344cd46caSbillm * spa_namespace_lock -> spa_refcount 94fa9e4066Sahrens * 9544cd46caSbillm * The namespace lock must be acquired to increase the refcount from 0 9644cd46caSbillm * or to check if it is zero. 97fa9e4066Sahrens * 98e14bb325SJeff Bonwick * spa_refcount -> spa_config_lock[] 99fa9e4066Sahrens * 10044cd46caSbillm * There must be at least one valid reference on the spa_t to acquire 10144cd46caSbillm * the config lock. 102fa9e4066Sahrens * 103e14bb325SJeff Bonwick * spa_namespace_lock -> spa_config_lock[] 104fa9e4066Sahrens * 10544cd46caSbillm * The namespace lock must always be taken before the config lock. 106fa9e4066Sahrens * 107fa9e4066Sahrens * 108e14bb325SJeff Bonwick * The spa_namespace_lock can be acquired directly and is globally visible. 109fa9e4066Sahrens * 110e14bb325SJeff Bonwick * The namespace is manipulated using the following functions, all of which 111e14bb325SJeff Bonwick * require the spa_namespace_lock to be held. 112fa9e4066Sahrens * 11344cd46caSbillm * spa_lookup() Lookup a spa_t by name. 114fa9e4066Sahrens * 11544cd46caSbillm * spa_add() Create a new spa_t in the namespace. 116fa9e4066Sahrens * 11744cd46caSbillm * spa_remove() Remove a spa_t from the namespace. This also 11844cd46caSbillm * frees up any memory associated with the spa_t. 119fa9e4066Sahrens * 12044cd46caSbillm * spa_next() Returns the next spa_t in the system, or the 12144cd46caSbillm * first if NULL is passed. 122fa9e4066Sahrens * 12344cd46caSbillm * spa_evict_all() Shutdown and remove all spa_t structures in 12444cd46caSbillm * the system. 125fa9e4066Sahrens * 126ea8dc4b6Seschrock * spa_guid_exists() Determine whether a pool/device guid exists. 127fa9e4066Sahrens * 128fa9e4066Sahrens * The spa_refcount is manipulated using the following functions: 129fa9e4066Sahrens * 13044cd46caSbillm * spa_open_ref() Adds a reference to the given spa_t. Must be 13144cd46caSbillm * called with spa_namespace_lock held if the 13244cd46caSbillm * refcount is currently zero. 133fa9e4066Sahrens * 13444cd46caSbillm * spa_close() Remove a reference from the spa_t. This will 13544cd46caSbillm * not free the spa_t or remove it from the 13644cd46caSbillm * namespace. No locking is required. 137fa9e4066Sahrens * 13844cd46caSbillm * spa_refcount_zero() Returns true if the refcount is currently 13944cd46caSbillm * zero. Must be called with spa_namespace_lock 14044cd46caSbillm * held. 141fa9e4066Sahrens * 142e14bb325SJeff Bonwick * The spa_config_lock[] is an array of rwlocks, ordered as follows: 143e14bb325SJeff Bonwick * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. 144e14bb325SJeff Bonwick * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). 145e14bb325SJeff Bonwick * 146e14bb325SJeff Bonwick * To read the configuration, it suffices to hold one of these locks as reader. 147e14bb325SJeff Bonwick * To modify the configuration, you must hold all locks as writer. To modify 148e14bb325SJeff Bonwick * vdev state without altering the vdev tree's topology (e.g. online/offline), 149e14bb325SJeff Bonwick * you must hold SCL_STATE and SCL_ZIO as writer. 150e14bb325SJeff Bonwick * 151e14bb325SJeff Bonwick * We use these distinct config locks to avoid recursive lock entry. 152e14bb325SJeff Bonwick * For example, spa_sync() (which holds SCL_CONFIG as reader) induces 153e14bb325SJeff Bonwick * block allocations (SCL_ALLOC), which may require reading space maps 154e14bb325SJeff Bonwick * from disk (dmu_read() -> zio_read() -> SCL_ZIO). 155e14bb325SJeff Bonwick * 156e14bb325SJeff Bonwick * The spa config locks cannot be normal rwlocks because we need the 157e14bb325SJeff Bonwick * ability to hand off ownership. For example, SCL_ZIO is acquired 158e14bb325SJeff Bonwick * by the issuing thread and later released by an interrupt thread. 159e14bb325SJeff Bonwick * They do, however, obey the usual write-wanted semantics to prevent 160e14bb325SJeff Bonwick * writer (i.e. system administrator) starvation. 161e14bb325SJeff Bonwick * 162e14bb325SJeff Bonwick * The lock acquisition rules are as follows: 163e14bb325SJeff Bonwick * 164e14bb325SJeff Bonwick * SCL_CONFIG 165e14bb325SJeff Bonwick * Protects changes to the vdev tree topology, such as vdev 166e14bb325SJeff Bonwick * add/remove/attach/detach. Protects the dirty config list 167e14bb325SJeff Bonwick * (spa_config_dirty_list) and the set of spares and l2arc devices. 168e14bb325SJeff Bonwick * 169e14bb325SJeff Bonwick * SCL_STATE 170e14bb325SJeff Bonwick * Protects changes to pool state and vdev state, such as vdev 171e14bb325SJeff Bonwick * online/offline/fault/degrade/clear. Protects the dirty state list 172e14bb325SJeff Bonwick * (spa_state_dirty_list) and global pool state (spa_state). 173e14bb325SJeff Bonwick * 174e14bb325SJeff Bonwick * SCL_ALLOC 175e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 176e14bb325SJeff Bonwick * Held as reader by metaslab_alloc() and metaslab_claim(). 177e14bb325SJeff Bonwick * 178e14bb325SJeff Bonwick * SCL_ZIO 179e14bb325SJeff Bonwick * Held by bp-level zios (those which have no io_vd upon entry) 180e14bb325SJeff Bonwick * to prevent changes to the vdev tree. The bp-level zio implicitly 181e14bb325SJeff Bonwick * protects all of its vdev child zios, which do not hold SCL_ZIO. 182e14bb325SJeff Bonwick * 183e14bb325SJeff Bonwick * SCL_FREE 184e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 185e14bb325SJeff Bonwick * Held as reader by metaslab_free(). SCL_FREE is distinct from 186e14bb325SJeff Bonwick * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free 187e14bb325SJeff Bonwick * blocks in zio_done() while another i/o that holds either 188e14bb325SJeff Bonwick * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. 189e14bb325SJeff Bonwick * 190e14bb325SJeff Bonwick * SCL_VDEV 191e14bb325SJeff Bonwick * Held as reader to prevent changes to the vdev tree during trivial 192b24ab676SJeff Bonwick * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the 193e14bb325SJeff Bonwick * other locks, and lower than all of them, to ensure that it's safe 194e14bb325SJeff Bonwick * to acquire regardless of caller context. 195e14bb325SJeff Bonwick * 196e14bb325SJeff Bonwick * In addition, the following rules apply: 197e14bb325SJeff Bonwick * 198e14bb325SJeff Bonwick * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. 199e14bb325SJeff Bonwick * The lock ordering is SCL_CONFIG > spa_props_lock. 200e14bb325SJeff Bonwick * 201e14bb325SJeff Bonwick * (b) I/O operations on leaf vdevs. For any zio operation that takes 202e14bb325SJeff Bonwick * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), 203e14bb325SJeff Bonwick * or zio_write_phys() -- the caller must ensure that the config cannot 204e14bb325SJeff Bonwick * cannot change in the interim, and that the vdev cannot be reopened. 205e14bb325SJeff Bonwick * SCL_STATE as reader suffices for both. 206fa9e4066Sahrens * 207ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 208fa9e4066Sahrens * 20944cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 210ea8dc4b6Seschrock * for writing. 211fa9e4066Sahrens * 21244cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 21344cd46caSbillm * to complete, sync the updated configs to the 214ea8dc4b6Seschrock * cache, and release the namespace lock. 215fa9e4066Sahrens * 216e14bb325SJeff Bonwick * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). 217e14bb325SJeff Bonwick * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual 218e14bb325SJeff Bonwick * locking is, always, based on spa_namespace_lock and spa_config_lock[]. 219e14bb325SJeff Bonwick * 220*ad135b5dSChristopher Siden * spa_rename() is also implemented within this file since it requires 221e14bb325SJeff Bonwick * manipulation of the namespace. 222fa9e4066Sahrens */ 223fa9e4066Sahrens 224fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 225fa9e4066Sahrens kmutex_t spa_namespace_lock; 226fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 2270373e76bSbonwick static int spa_active_count; 228416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 229fa9e4066Sahrens 23099653d4eSeschrock static kmutex_t spa_spare_lock; 23139c23413Seschrock static avl_tree_t spa_spare_avl; 232fa94a07fSbrendan static kmutex_t spa_l2cache_lock; 233fa94a07fSbrendan static avl_tree_t spa_l2cache_avl; 23499653d4eSeschrock 235fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 2368ad4d6ddSJeff Bonwick int spa_mode_global; 237fa9e4066Sahrens 238fa9e4066Sahrens #ifdef ZFS_DEBUG 23940feaa91Sahrens /* Everything except dprintf is on by default in debug builds */ 24040feaa91Sahrens int zfs_flags = ~ZFS_DEBUG_DPRINTF; 241fa9e4066Sahrens #else 242fa9e4066Sahrens int zfs_flags = 0; 243fa9e4066Sahrens #endif 244fa9e4066Sahrens 2450125049cSahrens /* 2460125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 2470125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 2480125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 2490125049cSahrens */ 2500125049cSahrens int zfs_recover = 0; 2510125049cSahrens 252fa9e4066Sahrens 253e05725b1Sbonwick /* 254e05725b1Sbonwick * ========================================================================== 255e05725b1Sbonwick * SPA config locking 256e05725b1Sbonwick * ========================================================================== 257e05725b1Sbonwick */ 258e05725b1Sbonwick static void 259e14bb325SJeff Bonwick spa_config_lock_init(spa_t *spa) 260e14bb325SJeff Bonwick { 261e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 262e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 263e14bb325SJeff Bonwick mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); 264e14bb325SJeff Bonwick cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); 265e14bb325SJeff Bonwick refcount_create(&scl->scl_count); 266e14bb325SJeff Bonwick scl->scl_writer = NULL; 267e14bb325SJeff Bonwick scl->scl_write_wanted = 0; 268e14bb325SJeff Bonwick } 269e05725b1Sbonwick } 270e05725b1Sbonwick 271e05725b1Sbonwick static void 272e14bb325SJeff Bonwick spa_config_lock_destroy(spa_t *spa) 273e14bb325SJeff Bonwick { 274e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 275e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 276e14bb325SJeff Bonwick mutex_destroy(&scl->scl_lock); 277e14bb325SJeff Bonwick cv_destroy(&scl->scl_cv); 278e14bb325SJeff Bonwick refcount_destroy(&scl->scl_count); 279e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL); 280e14bb325SJeff Bonwick ASSERT(scl->scl_write_wanted == 0); 281e14bb325SJeff Bonwick } 282e14bb325SJeff Bonwick } 283e14bb325SJeff Bonwick 284e14bb325SJeff Bonwick int 285e14bb325SJeff Bonwick spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) 286e05725b1Sbonwick { 287e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 288e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 289e14bb325SJeff Bonwick if (!(locks & (1 << i))) 290e14bb325SJeff Bonwick continue; 291e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 292e14bb325SJeff Bonwick if (rw == RW_READER) { 293e14bb325SJeff Bonwick if (scl->scl_writer || scl->scl_write_wanted) { 294e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 295e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 296e14bb325SJeff Bonwick return (0); 297e14bb325SJeff Bonwick } 298e14bb325SJeff Bonwick } else { 299e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 300e14bb325SJeff Bonwick if (!refcount_is_zero(&scl->scl_count)) { 301e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 302e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 303e14bb325SJeff Bonwick return (0); 304e14bb325SJeff Bonwick } 305e14bb325SJeff Bonwick scl->scl_writer = curthread; 306e14bb325SJeff Bonwick } 307e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 308e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 309e14bb325SJeff Bonwick } 310e14bb325SJeff Bonwick return (1); 311e05725b1Sbonwick } 312e05725b1Sbonwick 313e05725b1Sbonwick void 314e14bb325SJeff Bonwick spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) 315e05725b1Sbonwick { 316f64c0e34SEric Taylor int wlocks_held = 0; 317f64c0e34SEric Taylor 318e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 319e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 320f64c0e34SEric Taylor if (scl->scl_writer == curthread) 321f64c0e34SEric Taylor wlocks_held |= (1 << i); 322e14bb325SJeff Bonwick if (!(locks & (1 << i))) 323e14bb325SJeff Bonwick continue; 324e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 325e14bb325SJeff Bonwick if (rw == RW_READER) { 326e14bb325SJeff Bonwick while (scl->scl_writer || scl->scl_write_wanted) { 327e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 328e14bb325SJeff Bonwick } 329e14bb325SJeff Bonwick } else { 330e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 331e14bb325SJeff Bonwick while (!refcount_is_zero(&scl->scl_count)) { 332e14bb325SJeff Bonwick scl->scl_write_wanted++; 333e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 334e14bb325SJeff Bonwick scl->scl_write_wanted--; 335e14bb325SJeff Bonwick } 336e14bb325SJeff Bonwick scl->scl_writer = curthread; 337e14bb325SJeff Bonwick } 338e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 339e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 340e05725b1Sbonwick } 341f64c0e34SEric Taylor ASSERT(wlocks_held <= locks); 342e05725b1Sbonwick } 343e05725b1Sbonwick 344e05725b1Sbonwick void 345e14bb325SJeff Bonwick spa_config_exit(spa_t *spa, int locks, void *tag) 346e05725b1Sbonwick { 347e14bb325SJeff Bonwick for (int i = SCL_LOCKS - 1; i >= 0; i--) { 348e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 349e14bb325SJeff Bonwick if (!(locks & (1 << i))) 350e14bb325SJeff Bonwick continue; 351e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 352e14bb325SJeff Bonwick ASSERT(!refcount_is_zero(&scl->scl_count)); 353e14bb325SJeff Bonwick if (refcount_remove(&scl->scl_count, tag) == 0) { 354e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL || 355e14bb325SJeff Bonwick scl->scl_writer == curthread); 356e14bb325SJeff Bonwick scl->scl_writer = NULL; /* OK in either case */ 357e14bb325SJeff Bonwick cv_broadcast(&scl->scl_cv); 358e14bb325SJeff Bonwick } 359e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 360e05725b1Sbonwick } 361e05725b1Sbonwick } 362e05725b1Sbonwick 363e14bb325SJeff Bonwick int 364e14bb325SJeff Bonwick spa_config_held(spa_t *spa, int locks, krw_t rw) 365e05725b1Sbonwick { 366e14bb325SJeff Bonwick int locks_held = 0; 367e05725b1Sbonwick 368e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 369e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 370e14bb325SJeff Bonwick if (!(locks & (1 << i))) 371e14bb325SJeff Bonwick continue; 372e14bb325SJeff Bonwick if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) || 373e14bb325SJeff Bonwick (rw == RW_WRITER && scl->scl_writer == curthread)) 374e14bb325SJeff Bonwick locks_held |= 1 << i; 375e14bb325SJeff Bonwick } 376e14bb325SJeff Bonwick 377e14bb325SJeff Bonwick return (locks_held); 378e05725b1Sbonwick } 379e05725b1Sbonwick 380fa9e4066Sahrens /* 381fa9e4066Sahrens * ========================================================================== 382fa9e4066Sahrens * SPA namespace functions 383fa9e4066Sahrens * ========================================================================== 384fa9e4066Sahrens */ 385fa9e4066Sahrens 386fa9e4066Sahrens /* 387fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 388fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 389fa9e4066Sahrens */ 390fa9e4066Sahrens spa_t * 391fa9e4066Sahrens spa_lookup(const char *name) 392fa9e4066Sahrens { 393e14bb325SJeff Bonwick static spa_t search; /* spa_t is large; don't allocate on stack */ 394e14bb325SJeff Bonwick spa_t *spa; 395fa9e4066Sahrens avl_index_t where; 39640feaa91Sahrens char c; 39740feaa91Sahrens char *cp; 398fa9e4066Sahrens 399fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 400fa9e4066Sahrens 40140feaa91Sahrens /* 40240feaa91Sahrens * If it's a full dataset name, figure out the pool name and 40340feaa91Sahrens * just use that. 40440feaa91Sahrens */ 40540feaa91Sahrens cp = strpbrk(name, "/@"); 40640feaa91Sahrens if (cp) { 40740feaa91Sahrens c = *cp; 40840feaa91Sahrens *cp = '\0'; 40940feaa91Sahrens } 41040feaa91Sahrens 411e14bb325SJeff Bonwick (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); 412fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 413fa9e4066Sahrens 41440feaa91Sahrens if (cp) 41540feaa91Sahrens *cp = c; 41640feaa91Sahrens 417fa9e4066Sahrens return (spa); 418fa9e4066Sahrens } 419fa9e4066Sahrens 420fa9e4066Sahrens /* 421fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 422fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 423fa9e4066Sahrens * exist by calling spa_lookup() first. 424fa9e4066Sahrens */ 425fa9e4066Sahrens spa_t * 426468c413aSTim Haley spa_add(const char *name, nvlist_t *config, const char *altroot) 427fa9e4066Sahrens { 428fa9e4066Sahrens spa_t *spa; 429c5904d13Seschrock spa_config_dirent_t *dp; 430fa9e4066Sahrens 431fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 432fa9e4066Sahrens 433fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 434fa9e4066Sahrens 435c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 436c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 43735a5a358SJonathan Adams mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 438c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 43935a5a358SJonathan Adams mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); 440c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 44135a5a358SJonathan Adams mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 442a1521560SJeff Bonwick mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); 443a1521560SJeff Bonwick mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); 444c25056deSgw 445c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 44635a5a358SJonathan Adams cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); 447c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 448e14bb325SJeff Bonwick cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); 449c25056deSgw 450b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 451cde58dbcSMatthew Ahrens bplist_create(&spa->spa_free_bplist[t]); 452b24ab676SJeff Bonwick 453e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); 454fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 455fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 4560373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 457468c413aSTim Haley spa->spa_load_max_txg = UINT64_MAX; 45835a5a358SJonathan Adams spa->spa_proc = &p0; 45935a5a358SJonathan Adams spa->spa_proc_state = SPA_PROC_NONE; 460fa9e4066Sahrens 461fa9e4066Sahrens refcount_create(&spa->spa_refcount); 462e14bb325SJeff Bonwick spa_config_lock_init(spa); 463fa9e4066Sahrens 464fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 465fa9e4066Sahrens 4660373e76bSbonwick /* 4670373e76bSbonwick * Set the alternate root, if there is one. 4680373e76bSbonwick */ 4690373e76bSbonwick if (altroot) { 4700373e76bSbonwick spa->spa_root = spa_strdup(altroot); 4710373e76bSbonwick spa_active_count++; 4720373e76bSbonwick } 4730373e76bSbonwick 474c5904d13Seschrock /* 475c5904d13Seschrock * Every pool starts with the default cachefile 476c5904d13Seschrock */ 477c5904d13Seschrock list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), 478c5904d13Seschrock offsetof(spa_config_dirent_t, scd_link)); 479c5904d13Seschrock 480c5904d13Seschrock dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); 481ef912c80STim Haley dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); 482c5904d13Seschrock list_insert_head(&spa->spa_config_list, dp); 483c5904d13Seschrock 4844b964adaSGeorge Wilson VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, 4854b964adaSGeorge Wilson KM_SLEEP) == 0); 4864b964adaSGeorge Wilson 487*ad135b5dSChristopher Siden if (config != NULL) { 488*ad135b5dSChristopher Siden nvlist_t *features; 489*ad135b5dSChristopher Siden 490*ad135b5dSChristopher Siden if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, 491*ad135b5dSChristopher Siden &features) == 0) { 492*ad135b5dSChristopher Siden VERIFY(nvlist_dup(features, &spa->spa_label_features, 493*ad135b5dSChristopher Siden 0) == 0); 494*ad135b5dSChristopher Siden } 495*ad135b5dSChristopher Siden 496468c413aSTim Haley VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); 497*ad135b5dSChristopher Siden } 498*ad135b5dSChristopher Siden 499*ad135b5dSChristopher Siden if (spa->spa_label_features == NULL) { 500*ad135b5dSChristopher Siden VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, 501*ad135b5dSChristopher Siden KM_SLEEP) == 0); 502*ad135b5dSChristopher Siden } 503468c413aSTim Haley 504fa9e4066Sahrens return (spa); 505fa9e4066Sahrens } 506fa9e4066Sahrens 507fa9e4066Sahrens /* 508fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 509fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 510fa9e4066Sahrens * deactivated. 511fa9e4066Sahrens */ 512fa9e4066Sahrens void 513fa9e4066Sahrens spa_remove(spa_t *spa) 514fa9e4066Sahrens { 515c5904d13Seschrock spa_config_dirent_t *dp; 516c5904d13Seschrock 517fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 518fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 519fa9e4066Sahrens 5201195e687SMark J Musante nvlist_free(spa->spa_config_splitting); 5211195e687SMark J Musante 522fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 523fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 524fa9e4066Sahrens 5250373e76bSbonwick if (spa->spa_root) { 526fa9e4066Sahrens spa_strfree(spa->spa_root); 5270373e76bSbonwick spa_active_count--; 5280373e76bSbonwick } 529fa9e4066Sahrens 530c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 531c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 532c5904d13Seschrock if (dp->scd_path != NULL) 533c5904d13Seschrock spa_strfree(dp->scd_path); 534c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 535c5904d13Seschrock } 536c5904d13Seschrock 537c5904d13Seschrock list_destroy(&spa->spa_config_list); 5382f8aaab3Seschrock 539*ad135b5dSChristopher Siden nvlist_free(spa->spa_label_features); 5404b964adaSGeorge Wilson nvlist_free(spa->spa_load_info); 541fa9e4066Sahrens spa_config_set(spa, NULL); 542fa9e4066Sahrens 543fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 54491ebeef5Sahrens 545e14bb325SJeff Bonwick spa_config_lock_destroy(spa); 546fa9e4066Sahrens 547b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 548cde58dbcSMatthew Ahrens bplist_destroy(&spa->spa_free_bplist[t]); 549b24ab676SJeff Bonwick 550c25056deSgw cv_destroy(&spa->spa_async_cv); 55135a5a358SJonathan Adams cv_destroy(&spa->spa_proc_cv); 552c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 553e14bb325SJeff Bonwick cv_destroy(&spa->spa_suspend_cv); 554c25056deSgw 5555ad82045Snd mutex_destroy(&spa->spa_async_lock); 556c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 55735a5a358SJonathan Adams mutex_destroy(&spa->spa_errlog_lock); 55806eeb2adSek mutex_destroy(&spa->spa_history_lock); 55935a5a358SJonathan Adams mutex_destroy(&spa->spa_proc_lock); 560b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 56135a5a358SJonathan Adams mutex_destroy(&spa->spa_scrub_lock); 562e14bb325SJeff Bonwick mutex_destroy(&spa->spa_suspend_lock); 563a1521560SJeff Bonwick mutex_destroy(&spa->spa_vdev_top_lock); 5645ad82045Snd 565fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 566fa9e4066Sahrens } 567fa9e4066Sahrens 568fa9e4066Sahrens /* 569fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 570fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 571fa9e4066Sahrens */ 572fa9e4066Sahrens spa_t * 573fa9e4066Sahrens spa_next(spa_t *prev) 574fa9e4066Sahrens { 575fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 576fa9e4066Sahrens 577fa9e4066Sahrens if (prev) 578fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 579fa9e4066Sahrens else 580fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 581fa9e4066Sahrens } 582fa9e4066Sahrens 583fa9e4066Sahrens /* 584fa9e4066Sahrens * ========================================================================== 585fa9e4066Sahrens * SPA refcount functions 586fa9e4066Sahrens * ========================================================================== 587fa9e4066Sahrens */ 588fa9e4066Sahrens 589fa9e4066Sahrens /* 590fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 591fa9e4066Sahrens * have the namespace lock held. 592fa9e4066Sahrens */ 593fa9e4066Sahrens void 594fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 595fa9e4066Sahrens { 596088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || 597fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 598fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 599fa9e4066Sahrens } 600fa9e4066Sahrens 601fa9e4066Sahrens /* 602fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 603fa9e4066Sahrens * have the namespace lock held. 604fa9e4066Sahrens */ 605fa9e4066Sahrens void 606fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 607fa9e4066Sahrens { 608088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || 609fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 610fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 611fa9e4066Sahrens } 612fa9e4066Sahrens 613fa9e4066Sahrens /* 614fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 615088f3894Sahrens * spa_namespace_lock held. We really compare against spa_minref, which is the 616fa9e4066Sahrens * number of references acquired when opening a pool 617fa9e4066Sahrens */ 618fa9e4066Sahrens boolean_t 619fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 620fa9e4066Sahrens { 621fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 622fa9e4066Sahrens 623088f3894Sahrens return (refcount_count(&spa->spa_refcount) == spa->spa_minref); 624fa9e4066Sahrens } 625fa9e4066Sahrens 62699653d4eSeschrock /* 62799653d4eSeschrock * ========================================================================== 628fa94a07fSbrendan * SPA spare and l2cache tracking 62999653d4eSeschrock * ========================================================================== 63099653d4eSeschrock */ 63199653d4eSeschrock 632fa94a07fSbrendan /* 633fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 634fa94a07fSbrendan * for 'auxiliary' devices. 635fa94a07fSbrendan */ 636fa94a07fSbrendan 637fa94a07fSbrendan typedef struct spa_aux { 638fa94a07fSbrendan uint64_t aux_guid; 639fa94a07fSbrendan uint64_t aux_pool; 640fa94a07fSbrendan avl_node_t aux_avl; 641fa94a07fSbrendan int aux_count; 642fa94a07fSbrendan } spa_aux_t; 643fa94a07fSbrendan 644fa94a07fSbrendan static int 645fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 646fa94a07fSbrendan { 647fa94a07fSbrendan const spa_aux_t *sa = a; 648fa94a07fSbrendan const spa_aux_t *sb = b; 649fa94a07fSbrendan 650fa94a07fSbrendan if (sa->aux_guid < sb->aux_guid) 651fa94a07fSbrendan return (-1); 652fa94a07fSbrendan else if (sa->aux_guid > sb->aux_guid) 653fa94a07fSbrendan return (1); 654fa94a07fSbrendan else 655fa94a07fSbrendan return (0); 656fa94a07fSbrendan } 657fa94a07fSbrendan 658fa94a07fSbrendan void 659fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 660fa94a07fSbrendan { 661fa94a07fSbrendan avl_index_t where; 662fa94a07fSbrendan spa_aux_t search; 663fa94a07fSbrendan spa_aux_t *aux; 664fa94a07fSbrendan 665fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 666fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 667fa94a07fSbrendan aux->aux_count++; 668fa94a07fSbrendan } else { 669fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 670fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 671fa94a07fSbrendan aux->aux_count = 1; 672fa94a07fSbrendan avl_insert(avl, aux, where); 673fa94a07fSbrendan } 674fa94a07fSbrendan } 675fa94a07fSbrendan 676fa94a07fSbrendan void 677fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 678fa94a07fSbrendan { 679fa94a07fSbrendan spa_aux_t search; 680fa94a07fSbrendan spa_aux_t *aux; 681fa94a07fSbrendan avl_index_t where; 682fa94a07fSbrendan 683fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 684fa94a07fSbrendan aux = avl_find(avl, &search, &where); 685fa94a07fSbrendan 686fa94a07fSbrendan ASSERT(aux != NULL); 687fa94a07fSbrendan 688fa94a07fSbrendan if (--aux->aux_count == 0) { 689fa94a07fSbrendan avl_remove(avl, aux); 690fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 691fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 692fa94a07fSbrendan aux->aux_pool = 0ULL; 693fa94a07fSbrendan } 694fa94a07fSbrendan } 695fa94a07fSbrendan 696fa94a07fSbrendan boolean_t 69789a89ebfSlling spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) 698fa94a07fSbrendan { 699fa94a07fSbrendan spa_aux_t search, *found; 700fa94a07fSbrendan 701fa94a07fSbrendan search.aux_guid = guid; 70289a89ebfSlling found = avl_find(avl, &search, NULL); 703fa94a07fSbrendan 704fa94a07fSbrendan if (pool) { 705fa94a07fSbrendan if (found) 706fa94a07fSbrendan *pool = found->aux_pool; 707fa94a07fSbrendan else 708fa94a07fSbrendan *pool = 0ULL; 709fa94a07fSbrendan } 710fa94a07fSbrendan 71189a89ebfSlling if (refcnt) { 71289a89ebfSlling if (found) 71389a89ebfSlling *refcnt = found->aux_count; 71489a89ebfSlling else 71589a89ebfSlling *refcnt = 0; 71689a89ebfSlling } 71789a89ebfSlling 718fa94a07fSbrendan return (found != NULL); 719fa94a07fSbrendan } 720fa94a07fSbrendan 721fa94a07fSbrendan void 722fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 723fa94a07fSbrendan { 724fa94a07fSbrendan spa_aux_t search, *found; 725fa94a07fSbrendan avl_index_t where; 726fa94a07fSbrendan 727fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 728fa94a07fSbrendan found = avl_find(avl, &search, &where); 729fa94a07fSbrendan ASSERT(found != NULL); 730fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 731fa94a07fSbrendan 732fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 733fa94a07fSbrendan } 734fa94a07fSbrendan 73599653d4eSeschrock /* 73639c23413Seschrock * Spares are tracked globally due to the following constraints: 73739c23413Seschrock * 73839c23413Seschrock * - A spare may be part of multiple pools. 73939c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 74039c23413Seschrock * another pool. 74139c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 74239c23413Seschrock * the target is a spare in the same pool. 74339c23413Seschrock * 74439c23413Seschrock * We keep track of all spares on the system through the use of a reference 74539c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 74639c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 74739c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 74839c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 74939c23413Seschrock * pool), we also keep track of which pool its been made a part of. 75039c23413Seschrock * 75139c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 75239c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 75339c23413Seschrock * separate spare lock exists for the status query path, which does not need to 75439c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 75599653d4eSeschrock */ 75639c23413Seschrock 75799653d4eSeschrock static int 75899653d4eSeschrock spa_spare_compare(const void *a, const void *b) 75999653d4eSeschrock { 760fa94a07fSbrendan return (spa_aux_compare(a, b)); 76199653d4eSeschrock } 76299653d4eSeschrock 76399653d4eSeschrock void 76439c23413Seschrock spa_spare_add(vdev_t *vd) 76599653d4eSeschrock { 76699653d4eSeschrock mutex_enter(&spa_spare_lock); 76739c23413Seschrock ASSERT(!vd->vdev_isspare); 768fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 76939c23413Seschrock vd->vdev_isspare = B_TRUE; 77099653d4eSeschrock mutex_exit(&spa_spare_lock); 77199653d4eSeschrock } 77299653d4eSeschrock 77399653d4eSeschrock void 77439c23413Seschrock spa_spare_remove(vdev_t *vd) 77599653d4eSeschrock { 77699653d4eSeschrock mutex_enter(&spa_spare_lock); 77739c23413Seschrock ASSERT(vd->vdev_isspare); 778fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 77939c23413Seschrock vd->vdev_isspare = B_FALSE; 78099653d4eSeschrock mutex_exit(&spa_spare_lock); 78199653d4eSeschrock } 78299653d4eSeschrock 78399653d4eSeschrock boolean_t 78489a89ebfSlling spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) 78599653d4eSeschrock { 786fa94a07fSbrendan boolean_t found; 78799653d4eSeschrock 78899653d4eSeschrock mutex_enter(&spa_spare_lock); 78989a89ebfSlling found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); 79099653d4eSeschrock mutex_exit(&spa_spare_lock); 79199653d4eSeschrock 792fa94a07fSbrendan return (found); 79339c23413Seschrock } 79439c23413Seschrock 79539c23413Seschrock void 79639c23413Seschrock spa_spare_activate(vdev_t *vd) 79739c23413Seschrock { 79839c23413Seschrock mutex_enter(&spa_spare_lock); 79939c23413Seschrock ASSERT(vd->vdev_isspare); 800fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 801fa94a07fSbrendan mutex_exit(&spa_spare_lock); 802fa94a07fSbrendan } 80339c23413Seschrock 804fa94a07fSbrendan /* 805fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 806fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 807fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 808fa94a07fSbrendan */ 80939c23413Seschrock 810fa94a07fSbrendan static int 811fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 812fa94a07fSbrendan { 813fa94a07fSbrendan return (spa_aux_compare(a, b)); 814fa94a07fSbrendan } 815fa94a07fSbrendan 816fa94a07fSbrendan void 817fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 818fa94a07fSbrendan { 819fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 820fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 821fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 822fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 823fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 824fa94a07fSbrendan } 825fa94a07fSbrendan 826fa94a07fSbrendan void 827fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 828fa94a07fSbrendan { 829fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 830fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 831fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 832fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 833fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 834fa94a07fSbrendan } 835fa94a07fSbrendan 836fa94a07fSbrendan boolean_t 837fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 838fa94a07fSbrendan { 839fa94a07fSbrendan boolean_t found; 840fa94a07fSbrendan 841fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 84289a89ebfSlling found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); 843fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 844fa94a07fSbrendan 845fa94a07fSbrendan return (found); 846fa94a07fSbrendan } 847fa94a07fSbrendan 848fa94a07fSbrendan void 849fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 850fa94a07fSbrendan { 851fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 852fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 853fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 854fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 855fa94a07fSbrendan } 856fa94a07fSbrendan 857fa9e4066Sahrens /* 858fa9e4066Sahrens * ========================================================================== 859fa9e4066Sahrens * SPA vdev locking 860fa9e4066Sahrens * ========================================================================== 861fa9e4066Sahrens */ 862fa9e4066Sahrens 863fa9e4066Sahrens /* 864ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 865ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 866fa9e4066Sahrens * It returns the next transaction group for the spa_t. 867fa9e4066Sahrens */ 868fa9e4066Sahrens uint64_t 869fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 870fa9e4066Sahrens { 871a1521560SJeff Bonwick mutex_enter(&spa->spa_vdev_top_lock); 872bbfd46c4SJeff Bonwick mutex_enter(&spa_namespace_lock); 87388ecc943SGeorge Wilson return (spa_vdev_config_enter(spa)); 87488ecc943SGeorge Wilson } 87588ecc943SGeorge Wilson 87688ecc943SGeorge Wilson /* 87788ecc943SGeorge Wilson * Internal implementation for spa_vdev_enter(). Used when a vdev 87888ecc943SGeorge Wilson * operation requires multiple syncs (i.e. removing a device) while 87988ecc943SGeorge Wilson * keeping the spa_namespace_lock held. 88088ecc943SGeorge Wilson */ 88188ecc943SGeorge Wilson uint64_t 88288ecc943SGeorge Wilson spa_vdev_config_enter(spa_t *spa) 88388ecc943SGeorge Wilson { 88488ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 8853d7072f8Seschrock 886e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 887fa9e4066Sahrens 888fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 889fa9e4066Sahrens } 890fa9e4066Sahrens 891fa9e4066Sahrens /* 89288ecc943SGeorge Wilson * Used in combination with spa_vdev_config_enter() to allow the syncing 89388ecc943SGeorge Wilson * of multiple transactions without releasing the spa_namespace_lock. 894fa9e4066Sahrens */ 89588ecc943SGeorge Wilson void 89688ecc943SGeorge Wilson spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) 897fa9e4066Sahrens { 89888ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 89988ecc943SGeorge Wilson 9000e34b6a7Sbonwick int config_changed = B_FALSE; 901ea8dc4b6Seschrock 9020373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 9030e34b6a7Sbonwick 904e14bb325SJeff Bonwick spa->spa_pending_vdev = NULL; 905e14bb325SJeff Bonwick 9060e34b6a7Sbonwick /* 9070e34b6a7Sbonwick * Reassess the DTLs. 9080e34b6a7Sbonwick */ 9090373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 9100e34b6a7Sbonwick 911e14bb325SJeff Bonwick if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { 9120e34b6a7Sbonwick config_changed = B_TRUE; 9138f18d1faSGeorge Wilson spa->spa_config_generation++; 9140e34b6a7Sbonwick } 915ea8dc4b6Seschrock 91688ecc943SGeorge Wilson /* 91788ecc943SGeorge Wilson * Verify the metaslab classes. 91888ecc943SGeorge Wilson */ 919b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); 920b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); 92188ecc943SGeorge Wilson 922e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 923fa9e4066Sahrens 92488ecc943SGeorge Wilson /* 92588ecc943SGeorge Wilson * Panic the system if the specified tag requires it. This 92688ecc943SGeorge Wilson * is useful for ensuring that configurations are updated 92788ecc943SGeorge Wilson * transactionally. 92888ecc943SGeorge Wilson */ 92988ecc943SGeorge Wilson if (zio_injection_enabled) 9301195e687SMark J Musante zio_handle_panic_injection(spa, tag, 0); 93188ecc943SGeorge Wilson 932fa9e4066Sahrens /* 933fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 934fa9e4066Sahrens * that there won't be more than one config change per txg. 935fa9e4066Sahrens * This allows us to use the txg as the generation number. 936fa9e4066Sahrens */ 937fa9e4066Sahrens if (error == 0) 938fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 939fa9e4066Sahrens 940fa9e4066Sahrens if (vd != NULL) { 9418ad4d6ddSJeff Bonwick ASSERT(!vd->vdev_detached || vd->vdev_dtl_smo.smo_object == 0); 9428ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 943fa9e4066Sahrens vdev_free(vd); 9448ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 945fa9e4066Sahrens } 946fa9e4066Sahrens 947fa9e4066Sahrens /* 9480e34b6a7Sbonwick * If the config changed, update the config cache. 949fa9e4066Sahrens */ 9500e34b6a7Sbonwick if (config_changed) 951c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 95288ecc943SGeorge Wilson } 953ea8dc4b6Seschrock 95488ecc943SGeorge Wilson /* 95588ecc943SGeorge Wilson * Unlock the spa_t after adding or removing a vdev. Besides undoing the 95688ecc943SGeorge Wilson * locking of spa_vdev_enter(), we also want make sure the transactions have 95788ecc943SGeorge Wilson * synced to disk, and then update the global configuration cache with the new 95888ecc943SGeorge Wilson * information. 95988ecc943SGeorge Wilson */ 96088ecc943SGeorge Wilson int 96188ecc943SGeorge Wilson spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 96288ecc943SGeorge Wilson { 96388ecc943SGeorge Wilson spa_vdev_config_exit(spa, vd, txg, error, FTAG); 964ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 965bbfd46c4SJeff Bonwick mutex_exit(&spa->spa_vdev_top_lock); 966fa9e4066Sahrens 967fa9e4066Sahrens return (error); 968fa9e4066Sahrens } 969fa9e4066Sahrens 970e14bb325SJeff Bonwick /* 971e14bb325SJeff Bonwick * Lock the given spa_t for the purpose of changing vdev state. 972e14bb325SJeff Bonwick */ 973e14bb325SJeff Bonwick void 9748f18d1faSGeorge Wilson spa_vdev_state_enter(spa_t *spa, int oplocks) 975e14bb325SJeff Bonwick { 9768f18d1faSGeorge Wilson int locks = SCL_STATE_ALL | oplocks; 9778f18d1faSGeorge Wilson 978dcba9f3fSGeorge Wilson /* 979dcba9f3fSGeorge Wilson * Root pools may need to read of the underlying devfs filesystem 980dcba9f3fSGeorge Wilson * when opening up a vdev. Unfortunately if we're holding the 981dcba9f3fSGeorge Wilson * SCL_ZIO lock it will result in a deadlock when we try to issue 982dcba9f3fSGeorge Wilson * the read from the root filesystem. Instead we "prefetch" 983dcba9f3fSGeorge Wilson * the associated vnodes that we need prior to opening the 984dcba9f3fSGeorge Wilson * underlying devices and cache them so that we can prevent 985dcba9f3fSGeorge Wilson * any I/O when we are doing the actual open. 986dcba9f3fSGeorge Wilson */ 987dcba9f3fSGeorge Wilson if (spa_is_root(spa)) { 9889842588bSGeorge Wilson int low = locks & ~(SCL_ZIO - 1); 9899842588bSGeorge Wilson int high = locks & ~low; 9909842588bSGeorge Wilson 9919842588bSGeorge Wilson spa_config_enter(spa, high, spa, RW_WRITER); 992dcba9f3fSGeorge Wilson vdev_hold(spa->spa_root_vdev); 9939842588bSGeorge Wilson spa_config_enter(spa, low, spa, RW_WRITER); 994dcba9f3fSGeorge Wilson } else { 995dcba9f3fSGeorge Wilson spa_config_enter(spa, locks, spa, RW_WRITER); 996dcba9f3fSGeorge Wilson } 9978f18d1faSGeorge Wilson spa->spa_vdev_locks = locks; 998e14bb325SJeff Bonwick } 999e14bb325SJeff Bonwick 1000e14bb325SJeff Bonwick int 1001e14bb325SJeff Bonwick spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) 1002e14bb325SJeff Bonwick { 1003c6065d0fSGeorge Wilson boolean_t config_changed = B_FALSE; 1004c6065d0fSGeorge Wilson 1005b24ab676SJeff Bonwick if (vd != NULL || error == 0) 1006b24ab676SJeff Bonwick vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 1007b24ab676SJeff Bonwick 0, 0, B_FALSE); 1008b24ab676SJeff Bonwick 10098f18d1faSGeorge Wilson if (vd != NULL) { 1010e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 1011c6065d0fSGeorge Wilson config_changed = B_TRUE; 10128f18d1faSGeorge Wilson spa->spa_config_generation++; 10138f18d1faSGeorge Wilson } 1014e14bb325SJeff Bonwick 1015dcba9f3fSGeorge Wilson if (spa_is_root(spa)) 1016dcba9f3fSGeorge Wilson vdev_rele(spa->spa_root_vdev); 1017dcba9f3fSGeorge Wilson 10188f18d1faSGeorge Wilson ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); 10198f18d1faSGeorge Wilson spa_config_exit(spa, spa->spa_vdev_locks, spa); 1020e14bb325SJeff Bonwick 10218ad4d6ddSJeff Bonwick /* 10228ad4d6ddSJeff Bonwick * If anything changed, wait for it to sync. This ensures that, 10238ad4d6ddSJeff Bonwick * from the system administrator's perspective, zpool(1M) commands 10248ad4d6ddSJeff Bonwick * are synchronous. This is important for things like zpool offline: 10258ad4d6ddSJeff Bonwick * when the command completes, you expect no further I/O from ZFS. 10268ad4d6ddSJeff Bonwick */ 10278ad4d6ddSJeff Bonwick if (vd != NULL) 10288ad4d6ddSJeff Bonwick txg_wait_synced(spa->spa_dsl_pool, 0); 10298ad4d6ddSJeff Bonwick 1030c6065d0fSGeorge Wilson /* 1031c6065d0fSGeorge Wilson * If the config changed, update the config cache. 1032c6065d0fSGeorge Wilson */ 1033c6065d0fSGeorge Wilson if (config_changed) { 1034c6065d0fSGeorge Wilson mutex_enter(&spa_namespace_lock); 1035c6065d0fSGeorge Wilson spa_config_sync(spa, B_FALSE, B_TRUE); 1036c6065d0fSGeorge Wilson mutex_exit(&spa_namespace_lock); 1037c6065d0fSGeorge Wilson } 1038c6065d0fSGeorge Wilson 1039e14bb325SJeff Bonwick return (error); 1040e14bb325SJeff Bonwick } 1041e14bb325SJeff Bonwick 1042fa9e4066Sahrens /* 1043fa9e4066Sahrens * ========================================================================== 1044fa9e4066Sahrens * Miscellaneous functions 1045fa9e4066Sahrens * ========================================================================== 1046fa9e4066Sahrens */ 1047fa9e4066Sahrens 1048*ad135b5dSChristopher Siden void 1049*ad135b5dSChristopher Siden spa_activate_mos_feature(spa_t *spa, const char *feature) 1050*ad135b5dSChristopher Siden { 1051*ad135b5dSChristopher Siden (void) nvlist_add_boolean(spa->spa_label_features, feature); 1052*ad135b5dSChristopher Siden vdev_config_dirty(spa->spa_root_vdev); 1053*ad135b5dSChristopher Siden } 1054*ad135b5dSChristopher Siden 1055*ad135b5dSChristopher Siden void 1056*ad135b5dSChristopher Siden spa_deactivate_mos_feature(spa_t *spa, const char *feature) 1057*ad135b5dSChristopher Siden { 1058*ad135b5dSChristopher Siden (void) nvlist_remove_all(spa->spa_label_features, feature); 1059*ad135b5dSChristopher Siden vdev_config_dirty(spa->spa_root_vdev); 1060*ad135b5dSChristopher Siden } 1061*ad135b5dSChristopher Siden 1062fa9e4066Sahrens /* 1063fa9e4066Sahrens * Rename a spa_t. 1064fa9e4066Sahrens */ 1065fa9e4066Sahrens int 1066fa9e4066Sahrens spa_rename(const char *name, const char *newname) 1067fa9e4066Sahrens { 1068fa9e4066Sahrens spa_t *spa; 1069fa9e4066Sahrens int err; 1070fa9e4066Sahrens 1071fa9e4066Sahrens /* 1072fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 1073fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 1074fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 1075ea8dc4b6Seschrock * allow recursive calls for other reasons. 1076fa9e4066Sahrens */ 1077fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 1078fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 1079fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1080fa9e4066Sahrens return (err); 1081fa9e4066Sahrens } 1082fa9e4066Sahrens 1083e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1084fa9e4066Sahrens 1085fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 1086e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); 1087fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 1088fa9e4066Sahrens 1089fa9e4066Sahrens /* 1090fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 1091fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 1092fa9e4066Sahrens * during the sync. 1093fa9e4066Sahrens */ 1094fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 1095fa9e4066Sahrens 1096e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1097fa9e4066Sahrens 10980373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 1099fa9e4066Sahrens 1100fa9e4066Sahrens /* 1101fa9e4066Sahrens * Sync the updated config cache. 1102fa9e4066Sahrens */ 1103c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 1104fa9e4066Sahrens 1105fa9e4066Sahrens spa_close(spa, FTAG); 1106fa9e4066Sahrens 1107fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1108fa9e4066Sahrens 1109fa9e4066Sahrens return (0); 1110fa9e4066Sahrens } 1111fa9e4066Sahrens 1112fa9e4066Sahrens /* 1113f9af39baSGeorge Wilson * Return the spa_t associated with given pool_guid, if it exists. If 1114f9af39baSGeorge Wilson * device_guid is non-zero, determine whether the pool exists *and* contains 1115f9af39baSGeorge Wilson * a device with the specified device_guid. 1116fa9e4066Sahrens */ 1117f9af39baSGeorge Wilson spa_t * 1118f9af39baSGeorge Wilson spa_by_guid(uint64_t pool_guid, uint64_t device_guid) 1119fa9e4066Sahrens { 1120fa9e4066Sahrens spa_t *spa; 1121fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 1122fa9e4066Sahrens 1123ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1124fa9e4066Sahrens 1125fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 1126fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 1127fa9e4066Sahrens continue; 1128fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 1129fa9e4066Sahrens continue; 113039c23413Seschrock if (spa_guid(spa) == pool_guid) { 113139c23413Seschrock if (device_guid == 0) 113239c23413Seschrock break; 113339c23413Seschrock 113439c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 113539c23413Seschrock device_guid) != NULL) 113639c23413Seschrock break; 113739c23413Seschrock 113839c23413Seschrock /* 11398654d025Sperrin * Check any devices we may be in the process of adding. 114039c23413Seschrock */ 114139c23413Seschrock if (spa->spa_pending_vdev) { 114239c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 114339c23413Seschrock device_guid) != NULL) 114439c23413Seschrock break; 114539c23413Seschrock } 114639c23413Seschrock } 1147fa9e4066Sahrens } 1148fa9e4066Sahrens 1149f9af39baSGeorge Wilson return (spa); 1150f9af39baSGeorge Wilson } 1151f9af39baSGeorge Wilson 1152f9af39baSGeorge Wilson /* 1153f9af39baSGeorge Wilson * Determine whether a pool with the given pool_guid exists. 1154f9af39baSGeorge Wilson */ 1155f9af39baSGeorge Wilson boolean_t 1156f9af39baSGeorge Wilson spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 1157f9af39baSGeorge Wilson { 1158f9af39baSGeorge Wilson return (spa_by_guid(pool_guid, device_guid) != NULL); 1159fa9e4066Sahrens } 1160fa9e4066Sahrens 1161fa9e4066Sahrens char * 1162fa9e4066Sahrens spa_strdup(const char *s) 1163fa9e4066Sahrens { 1164fa9e4066Sahrens size_t len; 1165fa9e4066Sahrens char *new; 1166fa9e4066Sahrens 1167fa9e4066Sahrens len = strlen(s); 1168fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 1169fa9e4066Sahrens bcopy(s, new, len); 1170fa9e4066Sahrens new[len] = '\0'; 1171fa9e4066Sahrens 1172fa9e4066Sahrens return (new); 1173fa9e4066Sahrens } 1174fa9e4066Sahrens 1175fa9e4066Sahrens void 1176fa9e4066Sahrens spa_strfree(char *s) 1177fa9e4066Sahrens { 1178fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 1179fa9e4066Sahrens } 1180fa9e4066Sahrens 1181fa9e4066Sahrens uint64_t 1182fa9e4066Sahrens spa_get_random(uint64_t range) 1183fa9e4066Sahrens { 1184fa9e4066Sahrens uint64_t r; 1185fa9e4066Sahrens 1186fa9e4066Sahrens ASSERT(range != 0); 1187fa9e4066Sahrens 1188fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 1189fa9e4066Sahrens 1190fa9e4066Sahrens return (r % range); 1191fa9e4066Sahrens } 1192fa9e4066Sahrens 11931195e687SMark J Musante uint64_t 11941195e687SMark J Musante spa_generate_guid(spa_t *spa) 11951195e687SMark J Musante { 11961195e687SMark J Musante uint64_t guid = spa_get_random(-1ULL); 11971195e687SMark J Musante 11981195e687SMark J Musante if (spa != NULL) { 11991195e687SMark J Musante while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) 12001195e687SMark J Musante guid = spa_get_random(-1ULL); 12011195e687SMark J Musante } else { 12021195e687SMark J Musante while (guid == 0 || spa_guid_exists(guid, 0)) 12031195e687SMark J Musante guid = spa_get_random(-1ULL); 12041195e687SMark J Musante } 12051195e687SMark J Musante 12061195e687SMark J Musante return (guid); 12071195e687SMark J Musante } 12081195e687SMark J Musante 1209fa9e4066Sahrens void 1210b24ab676SJeff Bonwick sprintf_blkptr(char *buf, const blkptr_t *bp) 1211fa9e4066Sahrens { 1212*ad135b5dSChristopher Siden char type[256]; 1213f0ba89beSJeff Bonwick char *checksum = NULL; 1214f0ba89beSJeff Bonwick char *compress = NULL; 1215f0ba89beSJeff Bonwick 1216f0ba89beSJeff Bonwick if (bp != NULL) { 1217*ad135b5dSChristopher Siden if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { 1218*ad135b5dSChristopher Siden dmu_object_byteswap_t bswap = 1219*ad135b5dSChristopher Siden DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); 1220*ad135b5dSChristopher Siden (void) snprintf(type, sizeof (type), "bswap %s %s", 1221*ad135b5dSChristopher Siden DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? 1222*ad135b5dSChristopher Siden "metadata" : "data", 1223*ad135b5dSChristopher Siden dmu_ot_byteswap[bswap].ob_name); 1224*ad135b5dSChristopher Siden } else { 1225*ad135b5dSChristopher Siden (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, 1226*ad135b5dSChristopher Siden sizeof (type)); 1227*ad135b5dSChristopher Siden } 1228f0ba89beSJeff Bonwick checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; 1229f0ba89beSJeff Bonwick compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; 1230f0ba89beSJeff Bonwick } 1231fa9e4066Sahrens 1232b24ab676SJeff Bonwick SPRINTF_BLKPTR(snprintf, ' ', buf, bp, type, checksum, compress); 1233fa9e4066Sahrens } 1234fa9e4066Sahrens 1235fa9e4066Sahrens void 1236fa9e4066Sahrens spa_freeze(spa_t *spa) 1237fa9e4066Sahrens { 1238fa9e4066Sahrens uint64_t freeze_txg = 0; 1239fa9e4066Sahrens 1240e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1241fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 1242fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 1243fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 1244fa9e4066Sahrens } 1245e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1246fa9e4066Sahrens if (freeze_txg != 0) 1247fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 1248fa9e4066Sahrens } 1249fa9e4066Sahrens 12500125049cSahrens void 12510125049cSahrens zfs_panic_recover(const char *fmt, ...) 12520125049cSahrens { 12530125049cSahrens va_list adx; 12540125049cSahrens 12550125049cSahrens va_start(adx, fmt); 12560125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 12570125049cSahrens va_end(adx); 12580125049cSahrens } 12590125049cSahrens 12603f9d6ad7SLin Ling /* 12613f9d6ad7SLin Ling * This is a stripped-down version of strtoull, suitable only for converting 12623f9d6ad7SLin Ling * lowercase hexidecimal numbers that don't overflow. 12633f9d6ad7SLin Ling */ 12643f9d6ad7SLin Ling uint64_t 12653f9d6ad7SLin Ling strtonum(const char *str, char **nptr) 12663f9d6ad7SLin Ling { 12673f9d6ad7SLin Ling uint64_t val = 0; 12683f9d6ad7SLin Ling char c; 12693f9d6ad7SLin Ling int digit; 12703f9d6ad7SLin Ling 12713f9d6ad7SLin Ling while ((c = *str) != '\0') { 12723f9d6ad7SLin Ling if (c >= '0' && c <= '9') 12733f9d6ad7SLin Ling digit = c - '0'; 12743f9d6ad7SLin Ling else if (c >= 'a' && c <= 'f') 12753f9d6ad7SLin Ling digit = 10 + c - 'a'; 12763f9d6ad7SLin Ling else 12773f9d6ad7SLin Ling break; 12783f9d6ad7SLin Ling 12793f9d6ad7SLin Ling val *= 16; 12803f9d6ad7SLin Ling val += digit; 12813f9d6ad7SLin Ling 12823f9d6ad7SLin Ling str++; 12833f9d6ad7SLin Ling } 12843f9d6ad7SLin Ling 12853f9d6ad7SLin Ling if (nptr) 12863f9d6ad7SLin Ling *nptr = (char *)str; 12873f9d6ad7SLin Ling 12883f9d6ad7SLin Ling return (val); 12893f9d6ad7SLin Ling } 12903f9d6ad7SLin Ling 1291fa9e4066Sahrens /* 1292fa9e4066Sahrens * ========================================================================== 1293fa9e4066Sahrens * Accessor functions 1294fa9e4066Sahrens * ========================================================================== 1295fa9e4066Sahrens */ 1296fa9e4066Sahrens 1297088f3894Sahrens boolean_t 129888b7b0f2SMatthew Ahrens spa_shutting_down(spa_t *spa) 1299fa9e4066Sahrens { 130088b7b0f2SMatthew Ahrens return (spa->spa_async_suspended); 1301fa9e4066Sahrens } 1302fa9e4066Sahrens 1303fa9e4066Sahrens dsl_pool_t * 1304fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1305fa9e4066Sahrens { 1306fa9e4066Sahrens return (spa->spa_dsl_pool); 1307fa9e4066Sahrens } 1308fa9e4066Sahrens 1309*ad135b5dSChristopher Siden boolean_t 1310*ad135b5dSChristopher Siden spa_is_initializing(spa_t *spa) 1311*ad135b5dSChristopher Siden { 1312*ad135b5dSChristopher Siden return (spa->spa_is_initializing); 1313*ad135b5dSChristopher Siden } 1314*ad135b5dSChristopher Siden 1315fa9e4066Sahrens blkptr_t * 1316fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1317fa9e4066Sahrens { 1318fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1319fa9e4066Sahrens } 1320fa9e4066Sahrens 1321fa9e4066Sahrens void 1322fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1323fa9e4066Sahrens { 1324fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1325fa9e4066Sahrens } 1326fa9e4066Sahrens 1327fa9e4066Sahrens void 1328fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1329fa9e4066Sahrens { 1330fa9e4066Sahrens if (spa->spa_root == NULL) 1331fa9e4066Sahrens buf[0] = '\0'; 1332fa9e4066Sahrens else 1333fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1334fa9e4066Sahrens } 1335fa9e4066Sahrens 1336fa9e4066Sahrens int 1337fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1338fa9e4066Sahrens { 1339fa9e4066Sahrens return (spa->spa_sync_pass); 1340fa9e4066Sahrens } 1341fa9e4066Sahrens 1342fa9e4066Sahrens char * 1343fa9e4066Sahrens spa_name(spa_t *spa) 1344fa9e4066Sahrens { 1345fa9e4066Sahrens return (spa->spa_name); 1346fa9e4066Sahrens } 1347fa9e4066Sahrens 1348fa9e4066Sahrens uint64_t 1349fa9e4066Sahrens spa_guid(spa_t *spa) 1350fa9e4066Sahrens { 1351b5989ec7Seschrock /* 1352b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1353b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1354e9103aaeSGarrett D'Amore * vdev. We stash the original pool guid in 'spa_config_guid' to handle 1355b5989ec7Seschrock * this case. 1356b5989ec7Seschrock */ 1357b5989ec7Seschrock if (spa->spa_root_vdev != NULL) 1358b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1359b5989ec7Seschrock else 1360e9103aaeSGarrett D'Amore return (spa->spa_config_guid); 1361e9103aaeSGarrett D'Amore } 1362e9103aaeSGarrett D'Amore 1363e9103aaeSGarrett D'Amore uint64_t 1364e9103aaeSGarrett D'Amore spa_load_guid(spa_t *spa) 1365e9103aaeSGarrett D'Amore { 1366e9103aaeSGarrett D'Amore /* 1367e9103aaeSGarrett D'Amore * This is a GUID that exists solely as a reference for the 1368e9103aaeSGarrett D'Amore * purposes of the arc. It is generated at load time, and 1369e9103aaeSGarrett D'Amore * is never written to persistent storage. 1370e9103aaeSGarrett D'Amore */ 1371e9103aaeSGarrett D'Amore return (spa->spa_load_guid); 1372fa9e4066Sahrens } 1373fa9e4066Sahrens 1374fa9e4066Sahrens uint64_t 1375fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1376fa9e4066Sahrens { 1377fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1378fa9e4066Sahrens } 1379fa9e4066Sahrens 1380fa9e4066Sahrens uint64_t 1381fa9e4066Sahrens spa_first_txg(spa_t *spa) 1382fa9e4066Sahrens { 1383fa9e4066Sahrens return (spa->spa_first_txg); 1384fa9e4066Sahrens } 1385fa9e4066Sahrens 1386b24ab676SJeff Bonwick uint64_t 1387b24ab676SJeff Bonwick spa_syncing_txg(spa_t *spa) 1388b24ab676SJeff Bonwick { 1389b24ab676SJeff Bonwick return (spa->spa_syncing_txg); 1390b24ab676SJeff Bonwick } 1391b24ab676SJeff Bonwick 139288b7b0f2SMatthew Ahrens pool_state_t 1393fa9e4066Sahrens spa_state(spa_t *spa) 1394fa9e4066Sahrens { 1395fa9e4066Sahrens return (spa->spa_state); 1396fa9e4066Sahrens } 1397fa9e4066Sahrens 1398b16da2e2SGeorge Wilson spa_load_state_t 1399b16da2e2SGeorge Wilson spa_load_state(spa_t *spa) 1400b16da2e2SGeorge Wilson { 1401b16da2e2SGeorge Wilson return (spa->spa_load_state); 1402b16da2e2SGeorge Wilson } 1403b16da2e2SGeorge Wilson 1404fa9e4066Sahrens uint64_t 1405fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1406fa9e4066Sahrens { 1407fa9e4066Sahrens return (spa->spa_freeze_txg); 1408fa9e4066Sahrens } 1409fa9e4066Sahrens 1410fa9e4066Sahrens /* ARGSUSED */ 1411fa9e4066Sahrens uint64_t 1412fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 1413fa9e4066Sahrens { 1414fa9e4066Sahrens /* 1415b24ab676SJeff Bonwick * The worst case is single-sector max-parity RAID-Z blocks, in which 1416b24ab676SJeff Bonwick * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) 1417b24ab676SJeff Bonwick * times the size; so just assume that. Add to this the fact that 1418b24ab676SJeff Bonwick * we can have up to 3 DVAs per bp, and one more factor of 2 because 1419b24ab676SJeff Bonwick * the block may be dittoed with up to 3 DVAs by ddt_sync(). 142044cd46caSbillm */ 1421b24ab676SJeff Bonwick return (lsize * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2); 142244cd46caSbillm } 142344cd46caSbillm 1424485bbbf5SGeorge Wilson uint64_t 1425485bbbf5SGeorge Wilson spa_get_dspace(spa_t *spa) 1426485bbbf5SGeorge Wilson { 1427485bbbf5SGeorge Wilson return (spa->spa_dspace); 1428485bbbf5SGeorge Wilson } 1429485bbbf5SGeorge Wilson 1430485bbbf5SGeorge Wilson void 1431485bbbf5SGeorge Wilson spa_update_dspace(spa_t *spa) 1432485bbbf5SGeorge Wilson { 1433485bbbf5SGeorge Wilson spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + 1434485bbbf5SGeorge Wilson ddt_get_dedup_dspace(spa); 1435485bbbf5SGeorge Wilson } 1436485bbbf5SGeorge Wilson 14370a4e9518Sgw /* 14380a4e9518Sgw * Return the failure mode that has been set to this pool. The default 14390a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 14400a4e9518Sgw */ 14410a4e9518Sgw uint8_t 14420a4e9518Sgw spa_get_failmode(spa_t *spa) 14430a4e9518Sgw { 14440a4e9518Sgw return (spa->spa_failmode); 14450a4e9518Sgw } 14460a4e9518Sgw 1447e14bb325SJeff Bonwick boolean_t 1448e14bb325SJeff Bonwick spa_suspended(spa_t *spa) 1449e14bb325SJeff Bonwick { 1450e14bb325SJeff Bonwick return (spa->spa_suspended); 1451e14bb325SJeff Bonwick } 1452e14bb325SJeff Bonwick 145344cd46caSbillm uint64_t 145444cd46caSbillm spa_version(spa_t *spa) 145544cd46caSbillm { 145644cd46caSbillm return (spa->spa_ubsync.ub_version); 145744cd46caSbillm } 145844cd46caSbillm 1459b24ab676SJeff Bonwick boolean_t 1460b24ab676SJeff Bonwick spa_deflate(spa_t *spa) 1461b24ab676SJeff Bonwick { 1462b24ab676SJeff Bonwick return (spa->spa_deflate); 1463b24ab676SJeff Bonwick } 1464b24ab676SJeff Bonwick 1465b24ab676SJeff Bonwick metaslab_class_t * 1466b24ab676SJeff Bonwick spa_normal_class(spa_t *spa) 1467b24ab676SJeff Bonwick { 1468b24ab676SJeff Bonwick return (spa->spa_normal_class); 1469b24ab676SJeff Bonwick } 1470b24ab676SJeff Bonwick 1471b24ab676SJeff Bonwick metaslab_class_t * 1472b24ab676SJeff Bonwick spa_log_class(spa_t *spa) 1473b24ab676SJeff Bonwick { 1474b24ab676SJeff Bonwick return (spa->spa_log_class); 1475b24ab676SJeff Bonwick } 1476b24ab676SJeff Bonwick 147744cd46caSbillm int 147844cd46caSbillm spa_max_replication(spa_t *spa) 147944cd46caSbillm { 148044cd46caSbillm /* 1481e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 148244cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 148344cd46caSbillm * replication level accordingly. 1484fa9e4066Sahrens */ 1485e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 148644cd46caSbillm return (1); 148744cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1488fa9e4066Sahrens } 1489fa9e4066Sahrens 14903f9d6ad7SLin Ling int 14913f9d6ad7SLin Ling spa_prev_software_version(spa_t *spa) 14923f9d6ad7SLin Ling { 14933f9d6ad7SLin Ling return (spa->spa_prev_software_version); 14943f9d6ad7SLin Ling } 14953f9d6ad7SLin Ling 149699653d4eSeschrock uint64_t 1497b24ab676SJeff Bonwick dva_get_dsize_sync(spa_t *spa, const dva_t *dva) 149899653d4eSeschrock { 1499b24ab676SJeff Bonwick uint64_t asize = DVA_GET_ASIZE(dva); 1500b24ab676SJeff Bonwick uint64_t dsize = asize; 150199653d4eSeschrock 1502b24ab676SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); 150399653d4eSeschrock 1504b24ab676SJeff Bonwick if (asize != 0 && spa->spa_deflate) { 1505b24ab676SJeff Bonwick vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); 1506b24ab676SJeff Bonwick dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 150799653d4eSeschrock } 1508b24ab676SJeff Bonwick 1509b24ab676SJeff Bonwick return (dsize); 1510b24ab676SJeff Bonwick } 1511b24ab676SJeff Bonwick 1512b24ab676SJeff Bonwick uint64_t 1513b24ab676SJeff Bonwick bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) 1514b24ab676SJeff Bonwick { 1515b24ab676SJeff Bonwick uint64_t dsize = 0; 1516b24ab676SJeff Bonwick 1517b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1518b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1519b24ab676SJeff Bonwick 1520b24ab676SJeff Bonwick return (dsize); 1521b24ab676SJeff Bonwick } 1522b24ab676SJeff Bonwick 1523b24ab676SJeff Bonwick uint64_t 1524b24ab676SJeff Bonwick bp_get_dsize(spa_t *spa, const blkptr_t *bp) 1525b24ab676SJeff Bonwick { 1526b24ab676SJeff Bonwick uint64_t dsize = 0; 1527b24ab676SJeff Bonwick 1528b24ab676SJeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 1529b24ab676SJeff Bonwick 1530b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1531b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1532b24ab676SJeff Bonwick 1533e14bb325SJeff Bonwick spa_config_exit(spa, SCL_VDEV, FTAG); 1534b24ab676SJeff Bonwick 1535b24ab676SJeff Bonwick return (dsize); 153699653d4eSeschrock } 153799653d4eSeschrock 1538fa9e4066Sahrens /* 1539fa9e4066Sahrens * ========================================================================== 1540fa9e4066Sahrens * Initialization and Termination 1541fa9e4066Sahrens * ========================================================================== 1542fa9e4066Sahrens */ 1543fa9e4066Sahrens 1544fa9e4066Sahrens static int 1545fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1546fa9e4066Sahrens { 1547fa9e4066Sahrens const spa_t *s1 = a1; 1548fa9e4066Sahrens const spa_t *s2 = a2; 1549fa9e4066Sahrens int s; 1550fa9e4066Sahrens 1551fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1552fa9e4066Sahrens if (s > 0) 1553fa9e4066Sahrens return (1); 1554fa9e4066Sahrens if (s < 0) 1555fa9e4066Sahrens return (-1); 1556fa9e4066Sahrens return (0); 1557fa9e4066Sahrens } 1558fa9e4066Sahrens 15590373e76bSbonwick int 15600373e76bSbonwick spa_busy(void) 15610373e76bSbonwick { 15620373e76bSbonwick return (spa_active_count); 15630373e76bSbonwick } 15640373e76bSbonwick 1565e7cbe64fSgw void 1566e7cbe64fSgw spa_boot_init() 1567e7cbe64fSgw { 1568e7cbe64fSgw spa_config_load(); 1569e7cbe64fSgw } 1570e7cbe64fSgw 1571fa9e4066Sahrens void 1572fa9e4066Sahrens spa_init(int mode) 1573fa9e4066Sahrens { 1574fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1575c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1576fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 1577fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1578fa9e4066Sahrens 1579fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1580fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1581fa9e4066Sahrens 1582fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 1583fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 1584fa94a07fSbrendan 1585fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 1586fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 158799653d4eSeschrock 15888ad4d6ddSJeff Bonwick spa_mode_global = mode; 1589fa9e4066Sahrens 1590fa9e4066Sahrens refcount_init(); 1591fa9e4066Sahrens unique_init(); 1592fa9e4066Sahrens zio_init(); 1593fa9e4066Sahrens dmu_init(); 1594fa9e4066Sahrens zil_init(); 159587db74c1Sek vdev_cache_stat_init(); 159691ebeef5Sahrens zfs_prop_init(); 1597990b4856Slling zpool_prop_init(); 1598*ad135b5dSChristopher Siden zpool_feature_init(); 1599fa9e4066Sahrens spa_config_load(); 1600e14bb325SJeff Bonwick l2arc_start(); 1601fa9e4066Sahrens } 1602fa9e4066Sahrens 1603fa9e4066Sahrens void 1604fa9e4066Sahrens spa_fini(void) 1605fa9e4066Sahrens { 1606e14bb325SJeff Bonwick l2arc_stop(); 1607e14bb325SJeff Bonwick 1608fa9e4066Sahrens spa_evict_all(); 1609fa9e4066Sahrens 161087db74c1Sek vdev_cache_stat_fini(); 1611fa9e4066Sahrens zil_fini(); 1612fa9e4066Sahrens dmu_fini(); 1613fa9e4066Sahrens zio_fini(); 161491ebeef5Sahrens unique_fini(); 1615fa9e4066Sahrens refcount_fini(); 1616fa9e4066Sahrens 1617fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 161899653d4eSeschrock avl_destroy(&spa_spare_avl); 1619fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 1620fa9e4066Sahrens 1621fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1622fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1623c25056deSgw mutex_destroy(&spa_spare_lock); 1624fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 1625fa9e4066Sahrens } 16266ce0521aSperrin 16276ce0521aSperrin /* 16286ce0521aSperrin * Return whether this pool has slogs. No locking needed. 16296ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 16306ce0521aSperrin * performance and not correctness 16316ce0521aSperrin */ 16326ce0521aSperrin boolean_t 16336ce0521aSperrin spa_has_slogs(spa_t *spa) 16346ce0521aSperrin { 16356ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 16366ce0521aSperrin } 1637bf82a41bSeschrock 1638b24ab676SJeff Bonwick spa_log_state_t 1639b24ab676SJeff Bonwick spa_get_log_state(spa_t *spa) 1640b24ab676SJeff Bonwick { 1641b24ab676SJeff Bonwick return (spa->spa_log_state); 1642b24ab676SJeff Bonwick } 1643b24ab676SJeff Bonwick 1644b24ab676SJeff Bonwick void 1645b24ab676SJeff Bonwick spa_set_log_state(spa_t *spa, spa_log_state_t state) 1646b24ab676SJeff Bonwick { 1647b24ab676SJeff Bonwick spa->spa_log_state = state; 1648b24ab676SJeff Bonwick } 1649b24ab676SJeff Bonwick 1650bf82a41bSeschrock boolean_t 1651bf82a41bSeschrock spa_is_root(spa_t *spa) 1652bf82a41bSeschrock { 1653bf82a41bSeschrock return (spa->spa_is_root); 1654bf82a41bSeschrock } 16558ad4d6ddSJeff Bonwick 16568ad4d6ddSJeff Bonwick boolean_t 16578ad4d6ddSJeff Bonwick spa_writeable(spa_t *spa) 16588ad4d6ddSJeff Bonwick { 16598ad4d6ddSJeff Bonwick return (!!(spa->spa_mode & FWRITE)); 16608ad4d6ddSJeff Bonwick } 16618ad4d6ddSJeff Bonwick 16628ad4d6ddSJeff Bonwick int 16638ad4d6ddSJeff Bonwick spa_mode(spa_t *spa) 16648ad4d6ddSJeff Bonwick { 16658ad4d6ddSJeff Bonwick return (spa->spa_mode); 16668ad4d6ddSJeff Bonwick } 1667b24ab676SJeff Bonwick 1668b24ab676SJeff Bonwick uint64_t 1669b24ab676SJeff Bonwick spa_bootfs(spa_t *spa) 1670b24ab676SJeff Bonwick { 1671b24ab676SJeff Bonwick return (spa->spa_bootfs); 1672b24ab676SJeff Bonwick } 1673b24ab676SJeff Bonwick 1674b24ab676SJeff Bonwick uint64_t 1675b24ab676SJeff Bonwick spa_delegation(spa_t *spa) 1676b24ab676SJeff Bonwick { 1677b24ab676SJeff Bonwick return (spa->spa_delegation); 1678b24ab676SJeff Bonwick } 1679b24ab676SJeff Bonwick 1680b24ab676SJeff Bonwick objset_t * 1681b24ab676SJeff Bonwick spa_meta_objset(spa_t *spa) 1682b24ab676SJeff Bonwick { 1683b24ab676SJeff Bonwick return (spa->spa_meta_objset); 1684b24ab676SJeff Bonwick } 1685b24ab676SJeff Bonwick 1686b24ab676SJeff Bonwick enum zio_checksum 1687b24ab676SJeff Bonwick spa_dedup_checksum(spa_t *spa) 1688b24ab676SJeff Bonwick { 1689b24ab676SJeff Bonwick return (spa->spa_dedup_checksum); 1690b24ab676SJeff Bonwick } 16913f9d6ad7SLin Ling 16923f9d6ad7SLin Ling /* 16933f9d6ad7SLin Ling * Reset pool scan stat per scan pass (or reboot). 16943f9d6ad7SLin Ling */ 16953f9d6ad7SLin Ling void 16963f9d6ad7SLin Ling spa_scan_stat_init(spa_t *spa) 16973f9d6ad7SLin Ling { 16983f9d6ad7SLin Ling /* data not stored on disk */ 16993f9d6ad7SLin Ling spa->spa_scan_pass_start = gethrestime_sec(); 17003f9d6ad7SLin Ling spa->spa_scan_pass_exam = 0; 17013f9d6ad7SLin Ling vdev_scan_stat_init(spa->spa_root_vdev); 17023f9d6ad7SLin Ling } 17033f9d6ad7SLin Ling 17043f9d6ad7SLin Ling /* 17053f9d6ad7SLin Ling * Get scan stats for zpool status reports 17063f9d6ad7SLin Ling */ 17073f9d6ad7SLin Ling int 17083f9d6ad7SLin Ling spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) 17093f9d6ad7SLin Ling { 17103f9d6ad7SLin Ling dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; 17113f9d6ad7SLin Ling 17123f9d6ad7SLin Ling if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) 17133f9d6ad7SLin Ling return (ENOENT); 17143f9d6ad7SLin Ling bzero(ps, sizeof (pool_scan_stat_t)); 17153f9d6ad7SLin Ling 17163f9d6ad7SLin Ling /* data stored on disk */ 17173f9d6ad7SLin Ling ps->pss_func = scn->scn_phys.scn_func; 17183f9d6ad7SLin Ling ps->pss_start_time = scn->scn_phys.scn_start_time; 17193f9d6ad7SLin Ling ps->pss_end_time = scn->scn_phys.scn_end_time; 17203f9d6ad7SLin Ling ps->pss_to_examine = scn->scn_phys.scn_to_examine; 17213f9d6ad7SLin Ling ps->pss_examined = scn->scn_phys.scn_examined; 17223f9d6ad7SLin Ling ps->pss_to_process = scn->scn_phys.scn_to_process; 17233f9d6ad7SLin Ling ps->pss_processed = scn->scn_phys.scn_processed; 17243f9d6ad7SLin Ling ps->pss_errors = scn->scn_phys.scn_errors; 17253f9d6ad7SLin Ling ps->pss_state = scn->scn_phys.scn_state; 17263f9d6ad7SLin Ling 17273f9d6ad7SLin Ling /* data not stored on disk */ 17283f9d6ad7SLin Ling ps->pss_pass_start = spa->spa_scan_pass_start; 17293f9d6ad7SLin Ling ps->pss_pass_exam = spa->spa_scan_pass_exam; 17303f9d6ad7SLin Ling 17313f9d6ad7SLin Ling return (0); 17323f9d6ad7SLin Ling } 173309c9d376SGeorge Wilson 173409c9d376SGeorge Wilson boolean_t 173509c9d376SGeorge Wilson spa_debug_enabled(spa_t *spa) 173609c9d376SGeorge Wilson { 173709c9d376SGeorge Wilson return (spa->spa_debug); 173809c9d376SGeorge Wilson } 1739