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 /* 221195e687SMark J Musante * Copyright 2010 Sun Microsystems, Inc. All rights reserved. 23fa9e4066Sahrens * Use is subject to license terms. 24fa9e4066Sahrens */ 25fa9e4066Sahrens 26fa9e4066Sahrens #include <sys/zfs_context.h> 27fa9e4066Sahrens #include <sys/spa_impl.h> 28fa9e4066Sahrens #include <sys/zio.h> 29fa9e4066Sahrens #include <sys/zio_checksum.h> 30fa9e4066Sahrens #include <sys/zio_compress.h> 31fa9e4066Sahrens #include <sys/dmu.h> 32fa9e4066Sahrens #include <sys/dmu_tx.h> 33fa9e4066Sahrens #include <sys/zap.h> 34fa9e4066Sahrens #include <sys/zil.h> 35fa9e4066Sahrens #include <sys/vdev_impl.h> 36fa9e4066Sahrens #include <sys/metaslab.h> 37fa9e4066Sahrens #include <sys/uberblock_impl.h> 38fa9e4066Sahrens #include <sys/txg.h> 39fa9e4066Sahrens #include <sys/avl.h> 40fa9e4066Sahrens #include <sys/unique.h> 41fa9e4066Sahrens #include <sys/dsl_pool.h> 42fa9e4066Sahrens #include <sys/dsl_dir.h> 43fa9e4066Sahrens #include <sys/dsl_prop.h> 44fa9e4066Sahrens #include <sys/fs/zfs.h> 456ce0521aSperrin #include <sys/metaslab_impl.h> 46e14bb325SJeff Bonwick #include <sys/arc.h> 47485bbbf5SGeorge Wilson #include <sys/ddt.h> 4891ebeef5Sahrens #include "zfs_prop.h" 49fa9e4066Sahrens 50fa9e4066Sahrens /* 51fa9e4066Sahrens * SPA locking 52fa9e4066Sahrens * 53fa9e4066Sahrens * There are four basic locks for managing spa_t structures: 54fa9e4066Sahrens * 55fa9e4066Sahrens * spa_namespace_lock (global mutex) 56fa9e4066Sahrens * 5744cd46caSbillm * This lock must be acquired to do any of the following: 58fa9e4066Sahrens * 5944cd46caSbillm * - Lookup a spa_t by name 6044cd46caSbillm * - Add or remove a spa_t from the namespace 6144cd46caSbillm * - Increase spa_refcount from non-zero 6244cd46caSbillm * - Check if spa_refcount is zero 6344cd46caSbillm * - Rename a spa_t 64ea8dc4b6Seschrock * - add/remove/attach/detach devices 6544cd46caSbillm * - Held for the duration of create/destroy/import/export 66fa9e4066Sahrens * 6744cd46caSbillm * It does not need to handle recursion. A create or destroy may 6844cd46caSbillm * reference objects (files or zvols) in other pools, but by 6944cd46caSbillm * definition they must have an existing reference, and will never need 7044cd46caSbillm * to lookup a spa_t by name. 71fa9e4066Sahrens * 72fa9e4066Sahrens * spa_refcount (per-spa refcount_t protected by mutex) 73fa9e4066Sahrens * 7444cd46caSbillm * This reference count keep track of any active users of the spa_t. The 7544cd46caSbillm * spa_t cannot be destroyed or freed while this is non-zero. Internally, 7644cd46caSbillm * the refcount is never really 'zero' - opening a pool implicitly keeps 77088f3894Sahrens * some references in the DMU. Internally we check against spa_minref, but 7844cd46caSbillm * present the image of a zero/non-zero value to consumers. 79fa9e4066Sahrens * 80e14bb325SJeff Bonwick * spa_config_lock[] (per-spa array of rwlocks) 81fa9e4066Sahrens * 8291ebeef5Sahrens * This protects the spa_t from config changes, and must be held in 8391ebeef5Sahrens * the following circumstances: 84fa9e4066Sahrens * 8544cd46caSbillm * - RW_READER to perform I/O to the spa 8644cd46caSbillm * - RW_WRITER to change the vdev config 87fa9e4066Sahrens * 88fa9e4066Sahrens * The locking order is fairly straightforward: 89fa9e4066Sahrens * 9044cd46caSbillm * spa_namespace_lock -> spa_refcount 91fa9e4066Sahrens * 9244cd46caSbillm * The namespace lock must be acquired to increase the refcount from 0 9344cd46caSbillm * or to check if it is zero. 94fa9e4066Sahrens * 95e14bb325SJeff Bonwick * spa_refcount -> spa_config_lock[] 96fa9e4066Sahrens * 9744cd46caSbillm * There must be at least one valid reference on the spa_t to acquire 9844cd46caSbillm * the config lock. 99fa9e4066Sahrens * 100e14bb325SJeff Bonwick * spa_namespace_lock -> spa_config_lock[] 101fa9e4066Sahrens * 10244cd46caSbillm * The namespace lock must always be taken before the config lock. 103fa9e4066Sahrens * 104fa9e4066Sahrens * 105e14bb325SJeff Bonwick * The spa_namespace_lock can be acquired directly and is globally visible. 106fa9e4066Sahrens * 107e14bb325SJeff Bonwick * The namespace is manipulated using the following functions, all of which 108e14bb325SJeff Bonwick * require the spa_namespace_lock to be held. 109fa9e4066Sahrens * 11044cd46caSbillm * spa_lookup() Lookup a spa_t by name. 111fa9e4066Sahrens * 11244cd46caSbillm * spa_add() Create a new spa_t in the namespace. 113fa9e4066Sahrens * 11444cd46caSbillm * spa_remove() Remove a spa_t from the namespace. This also 11544cd46caSbillm * frees up any memory associated with the spa_t. 116fa9e4066Sahrens * 11744cd46caSbillm * spa_next() Returns the next spa_t in the system, or the 11844cd46caSbillm * first if NULL is passed. 119fa9e4066Sahrens * 12044cd46caSbillm * spa_evict_all() Shutdown and remove all spa_t structures in 12144cd46caSbillm * the system. 122fa9e4066Sahrens * 123ea8dc4b6Seschrock * spa_guid_exists() Determine whether a pool/device guid exists. 124fa9e4066Sahrens * 125fa9e4066Sahrens * The spa_refcount is manipulated using the following functions: 126fa9e4066Sahrens * 12744cd46caSbillm * spa_open_ref() Adds a reference to the given spa_t. Must be 12844cd46caSbillm * called with spa_namespace_lock held if the 12944cd46caSbillm * refcount is currently zero. 130fa9e4066Sahrens * 13144cd46caSbillm * spa_close() Remove a reference from the spa_t. This will 13244cd46caSbillm * not free the spa_t or remove it from the 13344cd46caSbillm * namespace. No locking is required. 134fa9e4066Sahrens * 13544cd46caSbillm * spa_refcount_zero() Returns true if the refcount is currently 13644cd46caSbillm * zero. Must be called with spa_namespace_lock 13744cd46caSbillm * held. 138fa9e4066Sahrens * 139e14bb325SJeff Bonwick * The spa_config_lock[] is an array of rwlocks, ordered as follows: 140e14bb325SJeff Bonwick * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. 141e14bb325SJeff Bonwick * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). 142e14bb325SJeff Bonwick * 143e14bb325SJeff Bonwick * To read the configuration, it suffices to hold one of these locks as reader. 144e14bb325SJeff Bonwick * To modify the configuration, you must hold all locks as writer. To modify 145e14bb325SJeff Bonwick * vdev state without altering the vdev tree's topology (e.g. online/offline), 146e14bb325SJeff Bonwick * you must hold SCL_STATE and SCL_ZIO as writer. 147e14bb325SJeff Bonwick * 148e14bb325SJeff Bonwick * We use these distinct config locks to avoid recursive lock entry. 149e14bb325SJeff Bonwick * For example, spa_sync() (which holds SCL_CONFIG as reader) induces 150e14bb325SJeff Bonwick * block allocations (SCL_ALLOC), which may require reading space maps 151e14bb325SJeff Bonwick * from disk (dmu_read() -> zio_read() -> SCL_ZIO). 152e14bb325SJeff Bonwick * 153e14bb325SJeff Bonwick * The spa config locks cannot be normal rwlocks because we need the 154e14bb325SJeff Bonwick * ability to hand off ownership. For example, SCL_ZIO is acquired 155e14bb325SJeff Bonwick * by the issuing thread and later released by an interrupt thread. 156e14bb325SJeff Bonwick * They do, however, obey the usual write-wanted semantics to prevent 157e14bb325SJeff Bonwick * writer (i.e. system administrator) starvation. 158e14bb325SJeff Bonwick * 159e14bb325SJeff Bonwick * The lock acquisition rules are as follows: 160e14bb325SJeff Bonwick * 161e14bb325SJeff Bonwick * SCL_CONFIG 162e14bb325SJeff Bonwick * Protects changes to the vdev tree topology, such as vdev 163e14bb325SJeff Bonwick * add/remove/attach/detach. Protects the dirty config list 164e14bb325SJeff Bonwick * (spa_config_dirty_list) and the set of spares and l2arc devices. 165e14bb325SJeff Bonwick * 166e14bb325SJeff Bonwick * SCL_STATE 167e14bb325SJeff Bonwick * Protects changes to pool state and vdev state, such as vdev 168e14bb325SJeff Bonwick * online/offline/fault/degrade/clear. Protects the dirty state list 169e14bb325SJeff Bonwick * (spa_state_dirty_list) and global pool state (spa_state). 170e14bb325SJeff Bonwick * 171e14bb325SJeff Bonwick * SCL_ALLOC 172e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 173e14bb325SJeff Bonwick * Held as reader by metaslab_alloc() and metaslab_claim(). 174e14bb325SJeff Bonwick * 175e14bb325SJeff Bonwick * SCL_ZIO 176e14bb325SJeff Bonwick * Held by bp-level zios (those which have no io_vd upon entry) 177e14bb325SJeff Bonwick * to prevent changes to the vdev tree. The bp-level zio implicitly 178e14bb325SJeff Bonwick * protects all of its vdev child zios, which do not hold SCL_ZIO. 179e14bb325SJeff Bonwick * 180e14bb325SJeff Bonwick * SCL_FREE 181e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 182e14bb325SJeff Bonwick * Held as reader by metaslab_free(). SCL_FREE is distinct from 183e14bb325SJeff Bonwick * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free 184e14bb325SJeff Bonwick * blocks in zio_done() while another i/o that holds either 185e14bb325SJeff Bonwick * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. 186e14bb325SJeff Bonwick * 187e14bb325SJeff Bonwick * SCL_VDEV 188e14bb325SJeff Bonwick * Held as reader to prevent changes to the vdev tree during trivial 189b24ab676SJeff Bonwick * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the 190e14bb325SJeff Bonwick * other locks, and lower than all of them, to ensure that it's safe 191e14bb325SJeff Bonwick * to acquire regardless of caller context. 192e14bb325SJeff Bonwick * 193e14bb325SJeff Bonwick * In addition, the following rules apply: 194e14bb325SJeff Bonwick * 195e14bb325SJeff Bonwick * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. 196e14bb325SJeff Bonwick * The lock ordering is SCL_CONFIG > spa_props_lock. 197e14bb325SJeff Bonwick * 198e14bb325SJeff Bonwick * (b) I/O operations on leaf vdevs. For any zio operation that takes 199e14bb325SJeff Bonwick * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), 200e14bb325SJeff Bonwick * or zio_write_phys() -- the caller must ensure that the config cannot 201e14bb325SJeff Bonwick * cannot change in the interim, and that the vdev cannot be reopened. 202e14bb325SJeff Bonwick * SCL_STATE as reader suffices for both. 203fa9e4066Sahrens * 204ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 205fa9e4066Sahrens * 20644cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 207ea8dc4b6Seschrock * for writing. 208fa9e4066Sahrens * 20944cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 21044cd46caSbillm * to complete, sync the updated configs to the 211ea8dc4b6Seschrock * cache, and release the namespace lock. 212fa9e4066Sahrens * 213e14bb325SJeff Bonwick * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). 214e14bb325SJeff Bonwick * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual 215e14bb325SJeff Bonwick * locking is, always, based on spa_namespace_lock and spa_config_lock[]. 216e14bb325SJeff Bonwick * 217e14bb325SJeff Bonwick * spa_rename() is also implemented within this file since is requires 218e14bb325SJeff Bonwick * manipulation of the namespace. 219fa9e4066Sahrens */ 220fa9e4066Sahrens 221fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 222fa9e4066Sahrens kmutex_t spa_namespace_lock; 223fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 2240373e76bSbonwick static int spa_active_count; 225416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 226fa9e4066Sahrens 22799653d4eSeschrock static kmutex_t spa_spare_lock; 22839c23413Seschrock static avl_tree_t spa_spare_avl; 229fa94a07fSbrendan static kmutex_t spa_l2cache_lock; 230fa94a07fSbrendan static avl_tree_t spa_l2cache_avl; 23199653d4eSeschrock 232fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 2338ad4d6ddSJeff Bonwick int spa_mode_global; 234fa9e4066Sahrens 235fa9e4066Sahrens #ifdef ZFS_DEBUG 23640feaa91Sahrens /* Everything except dprintf is on by default in debug builds */ 23740feaa91Sahrens int zfs_flags = ~ZFS_DEBUG_DPRINTF; 238fa9e4066Sahrens #else 239fa9e4066Sahrens int zfs_flags = 0; 240fa9e4066Sahrens #endif 241fa9e4066Sahrens 2420125049cSahrens /* 2430125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 2440125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 2450125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 2460125049cSahrens */ 2470125049cSahrens int zfs_recover = 0; 2480125049cSahrens 249fa9e4066Sahrens 250e05725b1Sbonwick /* 251e05725b1Sbonwick * ========================================================================== 252e05725b1Sbonwick * SPA config locking 253e05725b1Sbonwick * ========================================================================== 254e05725b1Sbonwick */ 255e05725b1Sbonwick static void 256e14bb325SJeff Bonwick spa_config_lock_init(spa_t *spa) 257e14bb325SJeff Bonwick { 258e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 259e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 260e14bb325SJeff Bonwick mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); 261e14bb325SJeff Bonwick cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); 262e14bb325SJeff Bonwick refcount_create(&scl->scl_count); 263e14bb325SJeff Bonwick scl->scl_writer = NULL; 264e14bb325SJeff Bonwick scl->scl_write_wanted = 0; 265e14bb325SJeff Bonwick } 266e05725b1Sbonwick } 267e05725b1Sbonwick 268e05725b1Sbonwick static void 269e14bb325SJeff Bonwick spa_config_lock_destroy(spa_t *spa) 270e14bb325SJeff Bonwick { 271e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 272e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 273e14bb325SJeff Bonwick mutex_destroy(&scl->scl_lock); 274e14bb325SJeff Bonwick cv_destroy(&scl->scl_cv); 275e14bb325SJeff Bonwick refcount_destroy(&scl->scl_count); 276e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL); 277e14bb325SJeff Bonwick ASSERT(scl->scl_write_wanted == 0); 278e14bb325SJeff Bonwick } 279e14bb325SJeff Bonwick } 280e14bb325SJeff Bonwick 281e14bb325SJeff Bonwick int 282e14bb325SJeff Bonwick spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) 283e05725b1Sbonwick { 284e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 285e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 286e14bb325SJeff Bonwick if (!(locks & (1 << i))) 287e14bb325SJeff Bonwick continue; 288e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 289e14bb325SJeff Bonwick if (rw == RW_READER) { 290e14bb325SJeff Bonwick if (scl->scl_writer || scl->scl_write_wanted) { 291e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 292e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 293e14bb325SJeff Bonwick return (0); 294e14bb325SJeff Bonwick } 295e14bb325SJeff Bonwick } else { 296e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 297e14bb325SJeff Bonwick if (!refcount_is_zero(&scl->scl_count)) { 298e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 299e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 300e14bb325SJeff Bonwick return (0); 301e14bb325SJeff Bonwick } 302e14bb325SJeff Bonwick scl->scl_writer = curthread; 303e14bb325SJeff Bonwick } 304e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 305e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 306e14bb325SJeff Bonwick } 307e14bb325SJeff Bonwick return (1); 308e05725b1Sbonwick } 309e05725b1Sbonwick 310e05725b1Sbonwick void 311e14bb325SJeff Bonwick spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) 312e05725b1Sbonwick { 313f64c0e34SEric Taylor int wlocks_held = 0; 314f64c0e34SEric Taylor 315e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 316e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 317f64c0e34SEric Taylor if (scl->scl_writer == curthread) 318f64c0e34SEric Taylor wlocks_held |= (1 << i); 319e14bb325SJeff Bonwick if (!(locks & (1 << i))) 320e14bb325SJeff Bonwick continue; 321e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 322e14bb325SJeff Bonwick if (rw == RW_READER) { 323e14bb325SJeff Bonwick while (scl->scl_writer || scl->scl_write_wanted) { 324e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 325e14bb325SJeff Bonwick } 326e14bb325SJeff Bonwick } else { 327e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 328e14bb325SJeff Bonwick while (!refcount_is_zero(&scl->scl_count)) { 329e14bb325SJeff Bonwick scl->scl_write_wanted++; 330e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 331e14bb325SJeff Bonwick scl->scl_write_wanted--; 332e14bb325SJeff Bonwick } 333e14bb325SJeff Bonwick scl->scl_writer = curthread; 334e14bb325SJeff Bonwick } 335e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 336e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 337e05725b1Sbonwick } 338f64c0e34SEric Taylor ASSERT(wlocks_held <= locks); 339e05725b1Sbonwick } 340e05725b1Sbonwick 341e05725b1Sbonwick void 342e14bb325SJeff Bonwick spa_config_exit(spa_t *spa, int locks, void *tag) 343e05725b1Sbonwick { 344e14bb325SJeff Bonwick for (int i = SCL_LOCKS - 1; i >= 0; i--) { 345e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 346e14bb325SJeff Bonwick if (!(locks & (1 << i))) 347e14bb325SJeff Bonwick continue; 348e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 349e14bb325SJeff Bonwick ASSERT(!refcount_is_zero(&scl->scl_count)); 350e14bb325SJeff Bonwick if (refcount_remove(&scl->scl_count, tag) == 0) { 351e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL || 352e14bb325SJeff Bonwick scl->scl_writer == curthread); 353e14bb325SJeff Bonwick scl->scl_writer = NULL; /* OK in either case */ 354e14bb325SJeff Bonwick cv_broadcast(&scl->scl_cv); 355e14bb325SJeff Bonwick } 356e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 357e05725b1Sbonwick } 358e05725b1Sbonwick } 359e05725b1Sbonwick 360e14bb325SJeff Bonwick int 361e14bb325SJeff Bonwick spa_config_held(spa_t *spa, int locks, krw_t rw) 362e05725b1Sbonwick { 363e14bb325SJeff Bonwick int locks_held = 0; 364e05725b1Sbonwick 365e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 366e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 367e14bb325SJeff Bonwick if (!(locks & (1 << i))) 368e14bb325SJeff Bonwick continue; 369e14bb325SJeff Bonwick if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) || 370e14bb325SJeff Bonwick (rw == RW_WRITER && scl->scl_writer == curthread)) 371e14bb325SJeff Bonwick locks_held |= 1 << i; 372e14bb325SJeff Bonwick } 373e14bb325SJeff Bonwick 374e14bb325SJeff Bonwick return (locks_held); 375e05725b1Sbonwick } 376e05725b1Sbonwick 377fa9e4066Sahrens /* 378fa9e4066Sahrens * ========================================================================== 379fa9e4066Sahrens * SPA namespace functions 380fa9e4066Sahrens * ========================================================================== 381fa9e4066Sahrens */ 382fa9e4066Sahrens 383fa9e4066Sahrens /* 384fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 385fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 386fa9e4066Sahrens */ 387fa9e4066Sahrens spa_t * 388fa9e4066Sahrens spa_lookup(const char *name) 389fa9e4066Sahrens { 390e14bb325SJeff Bonwick static spa_t search; /* spa_t is large; don't allocate on stack */ 391e14bb325SJeff Bonwick spa_t *spa; 392fa9e4066Sahrens avl_index_t where; 39340feaa91Sahrens char c; 39440feaa91Sahrens char *cp; 395fa9e4066Sahrens 396fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 397fa9e4066Sahrens 39840feaa91Sahrens /* 39940feaa91Sahrens * If it's a full dataset name, figure out the pool name and 40040feaa91Sahrens * just use that. 40140feaa91Sahrens */ 40240feaa91Sahrens cp = strpbrk(name, "/@"); 40340feaa91Sahrens if (cp) { 40440feaa91Sahrens c = *cp; 40540feaa91Sahrens *cp = '\0'; 40640feaa91Sahrens } 40740feaa91Sahrens 408e14bb325SJeff Bonwick (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); 409fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 410fa9e4066Sahrens 41140feaa91Sahrens if (cp) 41240feaa91Sahrens *cp = c; 41340feaa91Sahrens 414fa9e4066Sahrens return (spa); 415fa9e4066Sahrens } 416fa9e4066Sahrens 417fa9e4066Sahrens /* 418fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 419fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 420fa9e4066Sahrens * exist by calling spa_lookup() first. 421fa9e4066Sahrens */ 422fa9e4066Sahrens spa_t * 423468c413aSTim Haley spa_add(const char *name, nvlist_t *config, const char *altroot) 424fa9e4066Sahrens { 425fa9e4066Sahrens spa_t *spa; 426c5904d13Seschrock spa_config_dirent_t *dp; 427fa9e4066Sahrens 428fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 429fa9e4066Sahrens 430fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 431fa9e4066Sahrens 432c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 433c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 43435a5a358SJonathan Adams mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 435c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 43635a5a358SJonathan Adams mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); 437c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 43835a5a358SJonathan Adams mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 439a1521560SJeff Bonwick mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); 440a1521560SJeff Bonwick mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); 441c25056deSgw 442c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 44335a5a358SJonathan Adams cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); 444c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 445e14bb325SJeff Bonwick cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); 446c25056deSgw 447b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 448b24ab676SJeff Bonwick bplist_init(&spa->spa_free_bplist[t]); 449b24ab676SJeff Bonwick bplist_init(&spa->spa_deferred_bplist); 450b24ab676SJeff Bonwick 451e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); 452fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 453fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 4540373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 455468c413aSTim Haley spa->spa_load_max_txg = UINT64_MAX; 45635a5a358SJonathan Adams spa->spa_proc = &p0; 45735a5a358SJonathan Adams spa->spa_proc_state = SPA_PROC_NONE; 458fa9e4066Sahrens 459fa9e4066Sahrens refcount_create(&spa->spa_refcount); 460e14bb325SJeff Bonwick spa_config_lock_init(spa); 461fa9e4066Sahrens 462fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 463fa9e4066Sahrens 4640373e76bSbonwick /* 4650373e76bSbonwick * Set the alternate root, if there is one. 4660373e76bSbonwick */ 4670373e76bSbonwick if (altroot) { 4680373e76bSbonwick spa->spa_root = spa_strdup(altroot); 4690373e76bSbonwick spa_active_count++; 4700373e76bSbonwick } 4710373e76bSbonwick 472c5904d13Seschrock /* 473c5904d13Seschrock * Every pool starts with the default cachefile 474c5904d13Seschrock */ 475c5904d13Seschrock list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), 476c5904d13Seschrock offsetof(spa_config_dirent_t, scd_link)); 477c5904d13Seschrock 478c5904d13Seschrock dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); 479ef912c80STim Haley dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); 480c5904d13Seschrock list_insert_head(&spa->spa_config_list, dp); 481c5904d13Seschrock 482468c413aSTim Haley if (config != NULL) 483468c413aSTim Haley VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); 484468c413aSTim Haley 485fa9e4066Sahrens return (spa); 486fa9e4066Sahrens } 487fa9e4066Sahrens 488fa9e4066Sahrens /* 489fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 490fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 491fa9e4066Sahrens * deactivated. 492fa9e4066Sahrens */ 493fa9e4066Sahrens void 494fa9e4066Sahrens spa_remove(spa_t *spa) 495fa9e4066Sahrens { 496c5904d13Seschrock spa_config_dirent_t *dp; 497c5904d13Seschrock 498fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 499fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 500fa9e4066Sahrens 5011195e687SMark J Musante nvlist_free(spa->spa_config_splitting); 5021195e687SMark J Musante 503fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 504fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 505fa9e4066Sahrens 5060373e76bSbonwick if (spa->spa_root) { 507fa9e4066Sahrens spa_strfree(spa->spa_root); 5080373e76bSbonwick spa_active_count--; 5090373e76bSbonwick } 510fa9e4066Sahrens 511c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 512c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 513c5904d13Seschrock if (dp->scd_path != NULL) 514c5904d13Seschrock spa_strfree(dp->scd_path); 515c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 516c5904d13Seschrock } 517c5904d13Seschrock 518c5904d13Seschrock list_destroy(&spa->spa_config_list); 5192f8aaab3Seschrock 520fa9e4066Sahrens spa_config_set(spa, NULL); 521fa9e4066Sahrens 522fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 52391ebeef5Sahrens 524e14bb325SJeff Bonwick spa_config_lock_destroy(spa); 525fa9e4066Sahrens 526b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 527b24ab676SJeff Bonwick bplist_fini(&spa->spa_free_bplist[t]); 528b24ab676SJeff Bonwick bplist_fini(&spa->spa_deferred_bplist); 529b24ab676SJeff Bonwick 530c25056deSgw cv_destroy(&spa->spa_async_cv); 53135a5a358SJonathan Adams cv_destroy(&spa->spa_proc_cv); 532c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 533e14bb325SJeff Bonwick cv_destroy(&spa->spa_suspend_cv); 534c25056deSgw 5355ad82045Snd mutex_destroy(&spa->spa_async_lock); 536c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 53735a5a358SJonathan Adams mutex_destroy(&spa->spa_errlog_lock); 53806eeb2adSek mutex_destroy(&spa->spa_history_lock); 53935a5a358SJonathan Adams mutex_destroy(&spa->spa_proc_lock); 540b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 54135a5a358SJonathan Adams mutex_destroy(&spa->spa_scrub_lock); 542e14bb325SJeff Bonwick mutex_destroy(&spa->spa_suspend_lock); 543a1521560SJeff Bonwick mutex_destroy(&spa->spa_vdev_top_lock); 5445ad82045Snd 545fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 546fa9e4066Sahrens } 547fa9e4066Sahrens 548fa9e4066Sahrens /* 549fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 550fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 551fa9e4066Sahrens */ 552fa9e4066Sahrens spa_t * 553fa9e4066Sahrens spa_next(spa_t *prev) 554fa9e4066Sahrens { 555fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 556fa9e4066Sahrens 557fa9e4066Sahrens if (prev) 558fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 559fa9e4066Sahrens else 560fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 561fa9e4066Sahrens } 562fa9e4066Sahrens 563fa9e4066Sahrens /* 564fa9e4066Sahrens * ========================================================================== 565fa9e4066Sahrens * SPA refcount functions 566fa9e4066Sahrens * ========================================================================== 567fa9e4066Sahrens */ 568fa9e4066Sahrens 569fa9e4066Sahrens /* 570fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 571fa9e4066Sahrens * have the namespace lock held. 572fa9e4066Sahrens */ 573fa9e4066Sahrens void 574fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 575fa9e4066Sahrens { 576088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || 577fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 578fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 579fa9e4066Sahrens } 580fa9e4066Sahrens 581fa9e4066Sahrens /* 582fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 583fa9e4066Sahrens * have the namespace lock held. 584fa9e4066Sahrens */ 585fa9e4066Sahrens void 586fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 587fa9e4066Sahrens { 588088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || 589fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 590fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 591fa9e4066Sahrens } 592fa9e4066Sahrens 593fa9e4066Sahrens /* 594fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 595088f3894Sahrens * spa_namespace_lock held. We really compare against spa_minref, which is the 596fa9e4066Sahrens * number of references acquired when opening a pool 597fa9e4066Sahrens */ 598fa9e4066Sahrens boolean_t 599fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 600fa9e4066Sahrens { 601fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 602fa9e4066Sahrens 603088f3894Sahrens return (refcount_count(&spa->spa_refcount) == spa->spa_minref); 604fa9e4066Sahrens } 605fa9e4066Sahrens 60699653d4eSeschrock /* 60799653d4eSeschrock * ========================================================================== 608fa94a07fSbrendan * SPA spare and l2cache tracking 60999653d4eSeschrock * ========================================================================== 61099653d4eSeschrock */ 61199653d4eSeschrock 612fa94a07fSbrendan /* 613fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 614fa94a07fSbrendan * for 'auxiliary' devices. 615fa94a07fSbrendan */ 616fa94a07fSbrendan 617fa94a07fSbrendan typedef struct spa_aux { 618fa94a07fSbrendan uint64_t aux_guid; 619fa94a07fSbrendan uint64_t aux_pool; 620fa94a07fSbrendan avl_node_t aux_avl; 621fa94a07fSbrendan int aux_count; 622fa94a07fSbrendan } spa_aux_t; 623fa94a07fSbrendan 624fa94a07fSbrendan static int 625fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 626fa94a07fSbrendan { 627fa94a07fSbrendan const spa_aux_t *sa = a; 628fa94a07fSbrendan const spa_aux_t *sb = b; 629fa94a07fSbrendan 630fa94a07fSbrendan if (sa->aux_guid < sb->aux_guid) 631fa94a07fSbrendan return (-1); 632fa94a07fSbrendan else if (sa->aux_guid > sb->aux_guid) 633fa94a07fSbrendan return (1); 634fa94a07fSbrendan else 635fa94a07fSbrendan return (0); 636fa94a07fSbrendan } 637fa94a07fSbrendan 638fa94a07fSbrendan void 639fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 640fa94a07fSbrendan { 641fa94a07fSbrendan avl_index_t where; 642fa94a07fSbrendan spa_aux_t search; 643fa94a07fSbrendan spa_aux_t *aux; 644fa94a07fSbrendan 645fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 646fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 647fa94a07fSbrendan aux->aux_count++; 648fa94a07fSbrendan } else { 649fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 650fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 651fa94a07fSbrendan aux->aux_count = 1; 652fa94a07fSbrendan avl_insert(avl, aux, where); 653fa94a07fSbrendan } 654fa94a07fSbrendan } 655fa94a07fSbrendan 656fa94a07fSbrendan void 657fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 658fa94a07fSbrendan { 659fa94a07fSbrendan spa_aux_t search; 660fa94a07fSbrendan spa_aux_t *aux; 661fa94a07fSbrendan avl_index_t where; 662fa94a07fSbrendan 663fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 664fa94a07fSbrendan aux = avl_find(avl, &search, &where); 665fa94a07fSbrendan 666fa94a07fSbrendan ASSERT(aux != NULL); 667fa94a07fSbrendan 668fa94a07fSbrendan if (--aux->aux_count == 0) { 669fa94a07fSbrendan avl_remove(avl, aux); 670fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 671fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 672fa94a07fSbrendan aux->aux_pool = 0ULL; 673fa94a07fSbrendan } 674fa94a07fSbrendan } 675fa94a07fSbrendan 676fa94a07fSbrendan boolean_t 67789a89ebfSlling spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) 678fa94a07fSbrendan { 679fa94a07fSbrendan spa_aux_t search, *found; 680fa94a07fSbrendan 681fa94a07fSbrendan search.aux_guid = guid; 68289a89ebfSlling found = avl_find(avl, &search, NULL); 683fa94a07fSbrendan 684fa94a07fSbrendan if (pool) { 685fa94a07fSbrendan if (found) 686fa94a07fSbrendan *pool = found->aux_pool; 687fa94a07fSbrendan else 688fa94a07fSbrendan *pool = 0ULL; 689fa94a07fSbrendan } 690fa94a07fSbrendan 69189a89ebfSlling if (refcnt) { 69289a89ebfSlling if (found) 69389a89ebfSlling *refcnt = found->aux_count; 69489a89ebfSlling else 69589a89ebfSlling *refcnt = 0; 69689a89ebfSlling } 69789a89ebfSlling 698fa94a07fSbrendan return (found != NULL); 699fa94a07fSbrendan } 700fa94a07fSbrendan 701fa94a07fSbrendan void 702fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 703fa94a07fSbrendan { 704fa94a07fSbrendan spa_aux_t search, *found; 705fa94a07fSbrendan avl_index_t where; 706fa94a07fSbrendan 707fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 708fa94a07fSbrendan found = avl_find(avl, &search, &where); 709fa94a07fSbrendan ASSERT(found != NULL); 710fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 711fa94a07fSbrendan 712fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 713fa94a07fSbrendan } 714fa94a07fSbrendan 71599653d4eSeschrock /* 71639c23413Seschrock * Spares are tracked globally due to the following constraints: 71739c23413Seschrock * 71839c23413Seschrock * - A spare may be part of multiple pools. 71939c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 72039c23413Seschrock * another pool. 72139c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 72239c23413Seschrock * the target is a spare in the same pool. 72339c23413Seschrock * 72439c23413Seschrock * We keep track of all spares on the system through the use of a reference 72539c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 72639c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 72739c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 72839c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 72939c23413Seschrock * pool), we also keep track of which pool its been made a part of. 73039c23413Seschrock * 73139c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 73239c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 73339c23413Seschrock * separate spare lock exists for the status query path, which does not need to 73439c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 73599653d4eSeschrock */ 73639c23413Seschrock 73799653d4eSeschrock static int 73899653d4eSeschrock spa_spare_compare(const void *a, const void *b) 73999653d4eSeschrock { 740fa94a07fSbrendan return (spa_aux_compare(a, b)); 74199653d4eSeschrock } 74299653d4eSeschrock 74399653d4eSeschrock void 74439c23413Seschrock spa_spare_add(vdev_t *vd) 74599653d4eSeschrock { 74699653d4eSeschrock mutex_enter(&spa_spare_lock); 74739c23413Seschrock ASSERT(!vd->vdev_isspare); 748fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 74939c23413Seschrock vd->vdev_isspare = B_TRUE; 75099653d4eSeschrock mutex_exit(&spa_spare_lock); 75199653d4eSeschrock } 75299653d4eSeschrock 75399653d4eSeschrock void 75439c23413Seschrock spa_spare_remove(vdev_t *vd) 75599653d4eSeschrock { 75699653d4eSeschrock mutex_enter(&spa_spare_lock); 75739c23413Seschrock ASSERT(vd->vdev_isspare); 758fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 75939c23413Seschrock vd->vdev_isspare = B_FALSE; 76099653d4eSeschrock mutex_exit(&spa_spare_lock); 76199653d4eSeschrock } 76299653d4eSeschrock 76399653d4eSeschrock boolean_t 76489a89ebfSlling spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) 76599653d4eSeschrock { 766fa94a07fSbrendan boolean_t found; 76799653d4eSeschrock 76899653d4eSeschrock mutex_enter(&spa_spare_lock); 76989a89ebfSlling found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); 77099653d4eSeschrock mutex_exit(&spa_spare_lock); 77199653d4eSeschrock 772fa94a07fSbrendan return (found); 77339c23413Seschrock } 77439c23413Seschrock 77539c23413Seschrock void 77639c23413Seschrock spa_spare_activate(vdev_t *vd) 77739c23413Seschrock { 77839c23413Seschrock mutex_enter(&spa_spare_lock); 77939c23413Seschrock ASSERT(vd->vdev_isspare); 780fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 781fa94a07fSbrendan mutex_exit(&spa_spare_lock); 782fa94a07fSbrendan } 78339c23413Seschrock 784fa94a07fSbrendan /* 785fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 786fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 787fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 788fa94a07fSbrendan */ 78939c23413Seschrock 790fa94a07fSbrendan static int 791fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 792fa94a07fSbrendan { 793fa94a07fSbrendan return (spa_aux_compare(a, b)); 794fa94a07fSbrendan } 795fa94a07fSbrendan 796fa94a07fSbrendan void 797fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 798fa94a07fSbrendan { 799fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 800fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 801fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 802fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 803fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 804fa94a07fSbrendan } 805fa94a07fSbrendan 806fa94a07fSbrendan void 807fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 808fa94a07fSbrendan { 809fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 810fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 811fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 812fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 813fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 814fa94a07fSbrendan } 815fa94a07fSbrendan 816fa94a07fSbrendan boolean_t 817fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 818fa94a07fSbrendan { 819fa94a07fSbrendan boolean_t found; 820fa94a07fSbrendan 821fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 82289a89ebfSlling found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); 823fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 824fa94a07fSbrendan 825fa94a07fSbrendan return (found); 826fa94a07fSbrendan } 827fa94a07fSbrendan 828fa94a07fSbrendan void 829fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 830fa94a07fSbrendan { 831fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 832fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 833fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 834fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 835fa94a07fSbrendan } 836fa94a07fSbrendan 837fa9e4066Sahrens /* 838fa9e4066Sahrens * ========================================================================== 839fa9e4066Sahrens * SPA vdev locking 840fa9e4066Sahrens * ========================================================================== 841fa9e4066Sahrens */ 842fa9e4066Sahrens 843fa9e4066Sahrens /* 844ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 845ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 846fa9e4066Sahrens * It returns the next transaction group for the spa_t. 847fa9e4066Sahrens */ 848fa9e4066Sahrens uint64_t 849fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 850fa9e4066Sahrens { 851a1521560SJeff Bonwick mutex_enter(&spa->spa_vdev_top_lock); 852bbfd46c4SJeff Bonwick mutex_enter(&spa_namespace_lock); 85388ecc943SGeorge Wilson return (spa_vdev_config_enter(spa)); 85488ecc943SGeorge Wilson } 85588ecc943SGeorge Wilson 85688ecc943SGeorge Wilson /* 85788ecc943SGeorge Wilson * Internal implementation for spa_vdev_enter(). Used when a vdev 85888ecc943SGeorge Wilson * operation requires multiple syncs (i.e. removing a device) while 85988ecc943SGeorge Wilson * keeping the spa_namespace_lock held. 86088ecc943SGeorge Wilson */ 86188ecc943SGeorge Wilson uint64_t 86288ecc943SGeorge Wilson spa_vdev_config_enter(spa_t *spa) 86388ecc943SGeorge Wilson { 86488ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 8653d7072f8Seschrock 866e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 867fa9e4066Sahrens 868fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 869fa9e4066Sahrens } 870fa9e4066Sahrens 871fa9e4066Sahrens /* 87288ecc943SGeorge Wilson * Used in combination with spa_vdev_config_enter() to allow the syncing 87388ecc943SGeorge Wilson * of multiple transactions without releasing the spa_namespace_lock. 874fa9e4066Sahrens */ 87588ecc943SGeorge Wilson void 87688ecc943SGeorge Wilson spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) 877fa9e4066Sahrens { 87888ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 87988ecc943SGeorge Wilson 8800e34b6a7Sbonwick int config_changed = B_FALSE; 881ea8dc4b6Seschrock 8820373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 8830e34b6a7Sbonwick 884e14bb325SJeff Bonwick spa->spa_pending_vdev = NULL; 885e14bb325SJeff Bonwick 8860e34b6a7Sbonwick /* 8870e34b6a7Sbonwick * Reassess the DTLs. 8880e34b6a7Sbonwick */ 8890373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 8900e34b6a7Sbonwick 8910e34b6a7Sbonwick /* 8920373e76bSbonwick * If the config changed, notify the scrub thread that it must restart. 8930e34b6a7Sbonwick */ 894e14bb325SJeff Bonwick if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { 895088f3894Sahrens dsl_pool_scrub_restart(spa->spa_dsl_pool); 8960e34b6a7Sbonwick config_changed = B_TRUE; 8978f18d1faSGeorge Wilson spa->spa_config_generation++; 8980e34b6a7Sbonwick } 899ea8dc4b6Seschrock 90088ecc943SGeorge Wilson /* 90188ecc943SGeorge Wilson * Verify the metaslab classes. 90288ecc943SGeorge Wilson */ 903b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); 904b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); 90588ecc943SGeorge Wilson 906e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 907fa9e4066Sahrens 90888ecc943SGeorge Wilson /* 90988ecc943SGeorge Wilson * Panic the system if the specified tag requires it. This 91088ecc943SGeorge Wilson * is useful for ensuring that configurations are updated 91188ecc943SGeorge Wilson * transactionally. 91288ecc943SGeorge Wilson */ 91388ecc943SGeorge Wilson if (zio_injection_enabled) 9141195e687SMark J Musante zio_handle_panic_injection(spa, tag, 0); 91588ecc943SGeorge Wilson 916fa9e4066Sahrens /* 917fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 918fa9e4066Sahrens * that there won't be more than one config change per txg. 919fa9e4066Sahrens * This allows us to use the txg as the generation number. 920fa9e4066Sahrens */ 921fa9e4066Sahrens if (error == 0) 922fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 923fa9e4066Sahrens 924fa9e4066Sahrens if (vd != NULL) { 9258ad4d6ddSJeff Bonwick ASSERT(!vd->vdev_detached || vd->vdev_dtl_smo.smo_object == 0); 9268ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 927fa9e4066Sahrens vdev_free(vd); 9288ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 929fa9e4066Sahrens } 930fa9e4066Sahrens 931fa9e4066Sahrens /* 9320e34b6a7Sbonwick * If the config changed, update the config cache. 933fa9e4066Sahrens */ 9340e34b6a7Sbonwick if (config_changed) 935c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 93688ecc943SGeorge Wilson } 937ea8dc4b6Seschrock 93888ecc943SGeorge Wilson /* 93988ecc943SGeorge Wilson * Unlock the spa_t after adding or removing a vdev. Besides undoing the 94088ecc943SGeorge Wilson * locking of spa_vdev_enter(), we also want make sure the transactions have 94188ecc943SGeorge Wilson * synced to disk, and then update the global configuration cache with the new 94288ecc943SGeorge Wilson * information. 94388ecc943SGeorge Wilson */ 94488ecc943SGeorge Wilson int 94588ecc943SGeorge Wilson spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 94688ecc943SGeorge Wilson { 94788ecc943SGeorge Wilson spa_vdev_config_exit(spa, vd, txg, error, FTAG); 948ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 949bbfd46c4SJeff Bonwick mutex_exit(&spa->spa_vdev_top_lock); 950fa9e4066Sahrens 951fa9e4066Sahrens return (error); 952fa9e4066Sahrens } 953fa9e4066Sahrens 954e14bb325SJeff Bonwick /* 955e14bb325SJeff Bonwick * Lock the given spa_t for the purpose of changing vdev state. 956e14bb325SJeff Bonwick */ 957e14bb325SJeff Bonwick void 9588f18d1faSGeorge Wilson spa_vdev_state_enter(spa_t *spa, int oplocks) 959e14bb325SJeff Bonwick { 9608f18d1faSGeorge Wilson int locks = SCL_STATE_ALL | oplocks; 9618f18d1faSGeorge Wilson 9628f18d1faSGeorge Wilson spa_config_enter(spa, locks, spa, RW_WRITER); 9638f18d1faSGeorge Wilson spa->spa_vdev_locks = locks; 964e14bb325SJeff Bonwick } 965e14bb325SJeff Bonwick 966e14bb325SJeff Bonwick int 967e14bb325SJeff Bonwick spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) 968e14bb325SJeff Bonwick { 969*c6065d0fSGeorge Wilson boolean_t config_changed = B_FALSE; 970*c6065d0fSGeorge Wilson 971b24ab676SJeff Bonwick if (vd != NULL || error == 0) 972b24ab676SJeff Bonwick vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 973b24ab676SJeff Bonwick 0, 0, B_FALSE); 974b24ab676SJeff Bonwick 9758f18d1faSGeorge Wilson if (vd != NULL) { 976e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 977*c6065d0fSGeorge Wilson config_changed = B_TRUE; 9788f18d1faSGeorge Wilson spa->spa_config_generation++; 9798f18d1faSGeorge Wilson } 980e14bb325SJeff Bonwick 9818f18d1faSGeorge Wilson ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); 9828f18d1faSGeorge Wilson spa_config_exit(spa, spa->spa_vdev_locks, spa); 983e14bb325SJeff Bonwick 9848ad4d6ddSJeff Bonwick /* 9858ad4d6ddSJeff Bonwick * If anything changed, wait for it to sync. This ensures that, 9868ad4d6ddSJeff Bonwick * from the system administrator's perspective, zpool(1M) commands 9878ad4d6ddSJeff Bonwick * are synchronous. This is important for things like zpool offline: 9888ad4d6ddSJeff Bonwick * when the command completes, you expect no further I/O from ZFS. 9898ad4d6ddSJeff Bonwick */ 9908ad4d6ddSJeff Bonwick if (vd != NULL) 9918ad4d6ddSJeff Bonwick txg_wait_synced(spa->spa_dsl_pool, 0); 9928ad4d6ddSJeff Bonwick 993*c6065d0fSGeorge Wilson /* 994*c6065d0fSGeorge Wilson * If the config changed, update the config cache. 995*c6065d0fSGeorge Wilson */ 996*c6065d0fSGeorge Wilson if (config_changed) { 997*c6065d0fSGeorge Wilson mutex_enter(&spa_namespace_lock); 998*c6065d0fSGeorge Wilson spa_config_sync(spa, B_FALSE, B_TRUE); 999*c6065d0fSGeorge Wilson mutex_exit(&spa_namespace_lock); 1000*c6065d0fSGeorge Wilson } 1001*c6065d0fSGeorge Wilson 1002e14bb325SJeff Bonwick return (error); 1003e14bb325SJeff Bonwick } 1004e14bb325SJeff Bonwick 1005fa9e4066Sahrens /* 1006fa9e4066Sahrens * ========================================================================== 1007fa9e4066Sahrens * Miscellaneous functions 1008fa9e4066Sahrens * ========================================================================== 1009fa9e4066Sahrens */ 1010fa9e4066Sahrens 1011fa9e4066Sahrens /* 1012fa9e4066Sahrens * Rename a spa_t. 1013fa9e4066Sahrens */ 1014fa9e4066Sahrens int 1015fa9e4066Sahrens spa_rename(const char *name, const char *newname) 1016fa9e4066Sahrens { 1017fa9e4066Sahrens spa_t *spa; 1018fa9e4066Sahrens int err; 1019fa9e4066Sahrens 1020fa9e4066Sahrens /* 1021fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 1022fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 1023fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 1024ea8dc4b6Seschrock * allow recursive calls for other reasons. 1025fa9e4066Sahrens */ 1026fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 1027fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 1028fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1029fa9e4066Sahrens return (err); 1030fa9e4066Sahrens } 1031fa9e4066Sahrens 1032e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1033fa9e4066Sahrens 1034fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 1035e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); 1036fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 1037fa9e4066Sahrens 1038fa9e4066Sahrens /* 1039fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 1040fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 1041fa9e4066Sahrens * during the sync. 1042fa9e4066Sahrens */ 1043fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 1044fa9e4066Sahrens 1045e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1046fa9e4066Sahrens 10470373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 1048fa9e4066Sahrens 1049fa9e4066Sahrens /* 1050fa9e4066Sahrens * Sync the updated config cache. 1051fa9e4066Sahrens */ 1052c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 1053fa9e4066Sahrens 1054fa9e4066Sahrens spa_close(spa, FTAG); 1055fa9e4066Sahrens 1056fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1057fa9e4066Sahrens 1058fa9e4066Sahrens return (0); 1059fa9e4066Sahrens } 1060fa9e4066Sahrens 1061fa9e4066Sahrens 1062fa9e4066Sahrens /* 1063fa9e4066Sahrens * Determine whether a pool with given pool_guid exists. If device_guid is 1064fa9e4066Sahrens * non-zero, determine whether the pool exists *and* contains a device with the 1065fa9e4066Sahrens * specified device_guid. 1066fa9e4066Sahrens */ 1067fa9e4066Sahrens boolean_t 1068fa9e4066Sahrens spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 1069fa9e4066Sahrens { 1070fa9e4066Sahrens spa_t *spa; 1071fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 1072fa9e4066Sahrens 1073ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1074fa9e4066Sahrens 1075fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 1076fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 1077fa9e4066Sahrens continue; 1078fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 1079fa9e4066Sahrens continue; 108039c23413Seschrock if (spa_guid(spa) == pool_guid) { 108139c23413Seschrock if (device_guid == 0) 108239c23413Seschrock break; 108339c23413Seschrock 108439c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 108539c23413Seschrock device_guid) != NULL) 108639c23413Seschrock break; 108739c23413Seschrock 108839c23413Seschrock /* 10898654d025Sperrin * Check any devices we may be in the process of adding. 109039c23413Seschrock */ 109139c23413Seschrock if (spa->spa_pending_vdev) { 109239c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 109339c23413Seschrock device_guid) != NULL) 109439c23413Seschrock break; 109539c23413Seschrock } 109639c23413Seschrock } 1097fa9e4066Sahrens } 1098fa9e4066Sahrens 1099fa9e4066Sahrens return (spa != NULL); 1100fa9e4066Sahrens } 1101fa9e4066Sahrens 1102fa9e4066Sahrens char * 1103fa9e4066Sahrens spa_strdup(const char *s) 1104fa9e4066Sahrens { 1105fa9e4066Sahrens size_t len; 1106fa9e4066Sahrens char *new; 1107fa9e4066Sahrens 1108fa9e4066Sahrens len = strlen(s); 1109fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 1110fa9e4066Sahrens bcopy(s, new, len); 1111fa9e4066Sahrens new[len] = '\0'; 1112fa9e4066Sahrens 1113fa9e4066Sahrens return (new); 1114fa9e4066Sahrens } 1115fa9e4066Sahrens 1116fa9e4066Sahrens void 1117fa9e4066Sahrens spa_strfree(char *s) 1118fa9e4066Sahrens { 1119fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 1120fa9e4066Sahrens } 1121fa9e4066Sahrens 1122fa9e4066Sahrens uint64_t 1123fa9e4066Sahrens spa_get_random(uint64_t range) 1124fa9e4066Sahrens { 1125fa9e4066Sahrens uint64_t r; 1126fa9e4066Sahrens 1127fa9e4066Sahrens ASSERT(range != 0); 1128fa9e4066Sahrens 1129fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 1130fa9e4066Sahrens 1131fa9e4066Sahrens return (r % range); 1132fa9e4066Sahrens } 1133fa9e4066Sahrens 11341195e687SMark J Musante uint64_t 11351195e687SMark J Musante spa_generate_guid(spa_t *spa) 11361195e687SMark J Musante { 11371195e687SMark J Musante uint64_t guid = spa_get_random(-1ULL); 11381195e687SMark J Musante 11391195e687SMark J Musante if (spa != NULL) { 11401195e687SMark J Musante while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) 11411195e687SMark J Musante guid = spa_get_random(-1ULL); 11421195e687SMark J Musante } else { 11431195e687SMark J Musante while (guid == 0 || spa_guid_exists(guid, 0)) 11441195e687SMark J Musante guid = spa_get_random(-1ULL); 11451195e687SMark J Musante } 11461195e687SMark J Musante 11471195e687SMark J Musante return (guid); 11481195e687SMark J Musante } 11491195e687SMark J Musante 1150fa9e4066Sahrens void 1151b24ab676SJeff Bonwick sprintf_blkptr(char *buf, const blkptr_t *bp) 1152fa9e4066Sahrens { 1153f0ba89beSJeff Bonwick char *type = NULL; 1154f0ba89beSJeff Bonwick char *checksum = NULL; 1155f0ba89beSJeff Bonwick char *compress = NULL; 1156f0ba89beSJeff Bonwick 1157f0ba89beSJeff Bonwick if (bp != NULL) { 1158f0ba89beSJeff Bonwick type = dmu_ot[BP_GET_TYPE(bp)].ot_name; 1159f0ba89beSJeff Bonwick checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; 1160f0ba89beSJeff Bonwick compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; 1161f0ba89beSJeff Bonwick } 1162fa9e4066Sahrens 1163b24ab676SJeff Bonwick SPRINTF_BLKPTR(snprintf, ' ', buf, bp, type, checksum, compress); 1164fa9e4066Sahrens } 1165fa9e4066Sahrens 1166fa9e4066Sahrens void 1167fa9e4066Sahrens spa_freeze(spa_t *spa) 1168fa9e4066Sahrens { 1169fa9e4066Sahrens uint64_t freeze_txg = 0; 1170fa9e4066Sahrens 1171e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1172fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 1173fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 1174fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 1175fa9e4066Sahrens } 1176e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1177fa9e4066Sahrens if (freeze_txg != 0) 1178fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 1179fa9e4066Sahrens } 1180fa9e4066Sahrens 11810125049cSahrens void 11820125049cSahrens zfs_panic_recover(const char *fmt, ...) 11830125049cSahrens { 11840125049cSahrens va_list adx; 11850125049cSahrens 11860125049cSahrens va_start(adx, fmt); 11870125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 11880125049cSahrens va_end(adx); 11890125049cSahrens } 11900125049cSahrens 1191fa9e4066Sahrens /* 1192fa9e4066Sahrens * ========================================================================== 1193fa9e4066Sahrens * Accessor functions 1194fa9e4066Sahrens * ========================================================================== 1195fa9e4066Sahrens */ 1196fa9e4066Sahrens 1197088f3894Sahrens boolean_t 119888b7b0f2SMatthew Ahrens spa_shutting_down(spa_t *spa) 1199fa9e4066Sahrens { 120088b7b0f2SMatthew Ahrens return (spa->spa_async_suspended); 1201fa9e4066Sahrens } 1202fa9e4066Sahrens 1203fa9e4066Sahrens dsl_pool_t * 1204fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1205fa9e4066Sahrens { 1206fa9e4066Sahrens return (spa->spa_dsl_pool); 1207fa9e4066Sahrens } 1208fa9e4066Sahrens 1209fa9e4066Sahrens blkptr_t * 1210fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1211fa9e4066Sahrens { 1212fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1213fa9e4066Sahrens } 1214fa9e4066Sahrens 1215fa9e4066Sahrens void 1216fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1217fa9e4066Sahrens { 1218fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1219fa9e4066Sahrens } 1220fa9e4066Sahrens 1221fa9e4066Sahrens void 1222fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1223fa9e4066Sahrens { 1224fa9e4066Sahrens if (spa->spa_root == NULL) 1225fa9e4066Sahrens buf[0] = '\0'; 1226fa9e4066Sahrens else 1227fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1228fa9e4066Sahrens } 1229fa9e4066Sahrens 1230fa9e4066Sahrens int 1231fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1232fa9e4066Sahrens { 1233fa9e4066Sahrens return (spa->spa_sync_pass); 1234fa9e4066Sahrens } 1235fa9e4066Sahrens 1236fa9e4066Sahrens char * 1237fa9e4066Sahrens spa_name(spa_t *spa) 1238fa9e4066Sahrens { 1239fa9e4066Sahrens return (spa->spa_name); 1240fa9e4066Sahrens } 1241fa9e4066Sahrens 1242fa9e4066Sahrens uint64_t 1243fa9e4066Sahrens spa_guid(spa_t *spa) 1244fa9e4066Sahrens { 1245b5989ec7Seschrock /* 1246b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1247b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1248b5989ec7Seschrock * vdev. We stash the original pool guid in 'spa_load_guid' to handle 1249b5989ec7Seschrock * this case. 1250b5989ec7Seschrock */ 1251b5989ec7Seschrock if (spa->spa_root_vdev != NULL) 1252b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1253b5989ec7Seschrock else 1254b5989ec7Seschrock return (spa->spa_load_guid); 1255fa9e4066Sahrens } 1256fa9e4066Sahrens 1257fa9e4066Sahrens uint64_t 1258fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1259fa9e4066Sahrens { 1260fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1261fa9e4066Sahrens } 1262fa9e4066Sahrens 1263fa9e4066Sahrens uint64_t 1264fa9e4066Sahrens spa_first_txg(spa_t *spa) 1265fa9e4066Sahrens { 1266fa9e4066Sahrens return (spa->spa_first_txg); 1267fa9e4066Sahrens } 1268fa9e4066Sahrens 1269b24ab676SJeff Bonwick uint64_t 1270b24ab676SJeff Bonwick spa_syncing_txg(spa_t *spa) 1271b24ab676SJeff Bonwick { 1272b24ab676SJeff Bonwick return (spa->spa_syncing_txg); 1273b24ab676SJeff Bonwick } 1274b24ab676SJeff Bonwick 127588b7b0f2SMatthew Ahrens pool_state_t 1276fa9e4066Sahrens spa_state(spa_t *spa) 1277fa9e4066Sahrens { 1278fa9e4066Sahrens return (spa->spa_state); 1279fa9e4066Sahrens } 1280fa9e4066Sahrens 1281b16da2e2SGeorge Wilson spa_load_state_t 1282b16da2e2SGeorge Wilson spa_load_state(spa_t *spa) 1283b16da2e2SGeorge Wilson { 1284b16da2e2SGeorge Wilson return (spa->spa_load_state); 1285b16da2e2SGeorge Wilson } 1286b16da2e2SGeorge Wilson 1287fa9e4066Sahrens uint64_t 1288fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1289fa9e4066Sahrens { 1290fa9e4066Sahrens return (spa->spa_freeze_txg); 1291fa9e4066Sahrens } 1292fa9e4066Sahrens 1293fa9e4066Sahrens /* ARGSUSED */ 1294fa9e4066Sahrens uint64_t 1295fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 1296fa9e4066Sahrens { 1297fa9e4066Sahrens /* 1298b24ab676SJeff Bonwick * The worst case is single-sector max-parity RAID-Z blocks, in which 1299b24ab676SJeff Bonwick * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) 1300b24ab676SJeff Bonwick * times the size; so just assume that. Add to this the fact that 1301b24ab676SJeff Bonwick * we can have up to 3 DVAs per bp, and one more factor of 2 because 1302b24ab676SJeff Bonwick * the block may be dittoed with up to 3 DVAs by ddt_sync(). 130344cd46caSbillm */ 1304b24ab676SJeff Bonwick return (lsize * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2); 130544cd46caSbillm } 130644cd46caSbillm 1307485bbbf5SGeorge Wilson uint64_t 1308485bbbf5SGeorge Wilson spa_get_dspace(spa_t *spa) 1309485bbbf5SGeorge Wilson { 1310485bbbf5SGeorge Wilson return (spa->spa_dspace); 1311485bbbf5SGeorge Wilson } 1312485bbbf5SGeorge Wilson 1313485bbbf5SGeorge Wilson void 1314485bbbf5SGeorge Wilson spa_update_dspace(spa_t *spa) 1315485bbbf5SGeorge Wilson { 1316485bbbf5SGeorge Wilson spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + 1317485bbbf5SGeorge Wilson ddt_get_dedup_dspace(spa); 1318485bbbf5SGeorge Wilson } 1319485bbbf5SGeorge Wilson 13200a4e9518Sgw /* 13210a4e9518Sgw * Return the failure mode that has been set to this pool. The default 13220a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 13230a4e9518Sgw */ 13240a4e9518Sgw uint8_t 13250a4e9518Sgw spa_get_failmode(spa_t *spa) 13260a4e9518Sgw { 13270a4e9518Sgw return (spa->spa_failmode); 13280a4e9518Sgw } 13290a4e9518Sgw 1330e14bb325SJeff Bonwick boolean_t 1331e14bb325SJeff Bonwick spa_suspended(spa_t *spa) 1332e14bb325SJeff Bonwick { 1333e14bb325SJeff Bonwick return (spa->spa_suspended); 1334e14bb325SJeff Bonwick } 1335e14bb325SJeff Bonwick 133644cd46caSbillm uint64_t 133744cd46caSbillm spa_version(spa_t *spa) 133844cd46caSbillm { 133944cd46caSbillm return (spa->spa_ubsync.ub_version); 134044cd46caSbillm } 134144cd46caSbillm 1342b24ab676SJeff Bonwick boolean_t 1343b24ab676SJeff Bonwick spa_deflate(spa_t *spa) 1344b24ab676SJeff Bonwick { 1345b24ab676SJeff Bonwick return (spa->spa_deflate); 1346b24ab676SJeff Bonwick } 1347b24ab676SJeff Bonwick 1348b24ab676SJeff Bonwick metaslab_class_t * 1349b24ab676SJeff Bonwick spa_normal_class(spa_t *spa) 1350b24ab676SJeff Bonwick { 1351b24ab676SJeff Bonwick return (spa->spa_normal_class); 1352b24ab676SJeff Bonwick } 1353b24ab676SJeff Bonwick 1354b24ab676SJeff Bonwick metaslab_class_t * 1355b24ab676SJeff Bonwick spa_log_class(spa_t *spa) 1356b24ab676SJeff Bonwick { 1357b24ab676SJeff Bonwick return (spa->spa_log_class); 1358b24ab676SJeff Bonwick } 1359b24ab676SJeff Bonwick 136044cd46caSbillm int 136144cd46caSbillm spa_max_replication(spa_t *spa) 136244cd46caSbillm { 136344cd46caSbillm /* 1364e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 136544cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 136644cd46caSbillm * replication level accordingly. 1367fa9e4066Sahrens */ 1368e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 136944cd46caSbillm return (1); 137044cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1371fa9e4066Sahrens } 1372fa9e4066Sahrens 137399653d4eSeschrock uint64_t 1374b24ab676SJeff Bonwick dva_get_dsize_sync(spa_t *spa, const dva_t *dva) 137599653d4eSeschrock { 1376b24ab676SJeff Bonwick uint64_t asize = DVA_GET_ASIZE(dva); 1377b24ab676SJeff Bonwick uint64_t dsize = asize; 137899653d4eSeschrock 1379b24ab676SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); 138099653d4eSeschrock 1381b24ab676SJeff Bonwick if (asize != 0 && spa->spa_deflate) { 1382b24ab676SJeff Bonwick vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); 1383b24ab676SJeff Bonwick dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 138499653d4eSeschrock } 1385b24ab676SJeff Bonwick 1386b24ab676SJeff Bonwick return (dsize); 1387b24ab676SJeff Bonwick } 1388b24ab676SJeff Bonwick 1389b24ab676SJeff Bonwick uint64_t 1390b24ab676SJeff Bonwick bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) 1391b24ab676SJeff Bonwick { 1392b24ab676SJeff Bonwick uint64_t dsize = 0; 1393b24ab676SJeff Bonwick 1394b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1395b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1396b24ab676SJeff Bonwick 1397b24ab676SJeff Bonwick return (dsize); 1398b24ab676SJeff Bonwick } 1399b24ab676SJeff Bonwick 1400b24ab676SJeff Bonwick uint64_t 1401b24ab676SJeff Bonwick bp_get_dsize(spa_t *spa, const blkptr_t *bp) 1402b24ab676SJeff Bonwick { 1403b24ab676SJeff Bonwick uint64_t dsize = 0; 1404b24ab676SJeff Bonwick 1405b24ab676SJeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 1406b24ab676SJeff Bonwick 1407b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1408b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1409b24ab676SJeff Bonwick 1410e14bb325SJeff Bonwick spa_config_exit(spa, SCL_VDEV, FTAG); 1411b24ab676SJeff Bonwick 1412b24ab676SJeff Bonwick return (dsize); 141399653d4eSeschrock } 141499653d4eSeschrock 1415fa9e4066Sahrens /* 1416fa9e4066Sahrens * ========================================================================== 1417fa9e4066Sahrens * Initialization and Termination 1418fa9e4066Sahrens * ========================================================================== 1419fa9e4066Sahrens */ 1420fa9e4066Sahrens 1421fa9e4066Sahrens static int 1422fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1423fa9e4066Sahrens { 1424fa9e4066Sahrens const spa_t *s1 = a1; 1425fa9e4066Sahrens const spa_t *s2 = a2; 1426fa9e4066Sahrens int s; 1427fa9e4066Sahrens 1428fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1429fa9e4066Sahrens if (s > 0) 1430fa9e4066Sahrens return (1); 1431fa9e4066Sahrens if (s < 0) 1432fa9e4066Sahrens return (-1); 1433fa9e4066Sahrens return (0); 1434fa9e4066Sahrens } 1435fa9e4066Sahrens 14360373e76bSbonwick int 14370373e76bSbonwick spa_busy(void) 14380373e76bSbonwick { 14390373e76bSbonwick return (spa_active_count); 14400373e76bSbonwick } 14410373e76bSbonwick 1442e7cbe64fSgw void 1443e7cbe64fSgw spa_boot_init() 1444e7cbe64fSgw { 1445e7cbe64fSgw spa_config_load(); 1446e7cbe64fSgw } 1447e7cbe64fSgw 1448fa9e4066Sahrens void 1449fa9e4066Sahrens spa_init(int mode) 1450fa9e4066Sahrens { 1451fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1452c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1453fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 1454fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1455fa9e4066Sahrens 1456fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1457fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1458fa9e4066Sahrens 1459fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 1460fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 1461fa94a07fSbrendan 1462fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 1463fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 146499653d4eSeschrock 14658ad4d6ddSJeff Bonwick spa_mode_global = mode; 1466fa9e4066Sahrens 1467fa9e4066Sahrens refcount_init(); 1468fa9e4066Sahrens unique_init(); 1469fa9e4066Sahrens zio_init(); 1470fa9e4066Sahrens dmu_init(); 1471fa9e4066Sahrens zil_init(); 147287db74c1Sek vdev_cache_stat_init(); 147391ebeef5Sahrens zfs_prop_init(); 1474990b4856Slling zpool_prop_init(); 1475fa9e4066Sahrens spa_config_load(); 1476e14bb325SJeff Bonwick l2arc_start(); 1477fa9e4066Sahrens } 1478fa9e4066Sahrens 1479fa9e4066Sahrens void 1480fa9e4066Sahrens spa_fini(void) 1481fa9e4066Sahrens { 1482e14bb325SJeff Bonwick l2arc_stop(); 1483e14bb325SJeff Bonwick 1484fa9e4066Sahrens spa_evict_all(); 1485fa9e4066Sahrens 148687db74c1Sek vdev_cache_stat_fini(); 1487fa9e4066Sahrens zil_fini(); 1488fa9e4066Sahrens dmu_fini(); 1489fa9e4066Sahrens zio_fini(); 149091ebeef5Sahrens unique_fini(); 1491fa9e4066Sahrens refcount_fini(); 1492fa9e4066Sahrens 1493fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 149499653d4eSeschrock avl_destroy(&spa_spare_avl); 1495fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 1496fa9e4066Sahrens 1497fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1498fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1499c25056deSgw mutex_destroy(&spa_spare_lock); 1500fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 1501fa9e4066Sahrens } 15026ce0521aSperrin 15036ce0521aSperrin /* 15046ce0521aSperrin * Return whether this pool has slogs. No locking needed. 15056ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 15066ce0521aSperrin * performance and not correctness 15076ce0521aSperrin */ 15086ce0521aSperrin boolean_t 15096ce0521aSperrin spa_has_slogs(spa_t *spa) 15106ce0521aSperrin { 15116ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 15126ce0521aSperrin } 1513bf82a41bSeschrock 1514b24ab676SJeff Bonwick spa_log_state_t 1515b24ab676SJeff Bonwick spa_get_log_state(spa_t *spa) 1516b24ab676SJeff Bonwick { 1517b24ab676SJeff Bonwick return (spa->spa_log_state); 1518b24ab676SJeff Bonwick } 1519b24ab676SJeff Bonwick 1520b24ab676SJeff Bonwick void 1521b24ab676SJeff Bonwick spa_set_log_state(spa_t *spa, spa_log_state_t state) 1522b24ab676SJeff Bonwick { 1523b24ab676SJeff Bonwick spa->spa_log_state = state; 1524b24ab676SJeff Bonwick } 1525b24ab676SJeff Bonwick 1526bf82a41bSeschrock boolean_t 1527bf82a41bSeschrock spa_is_root(spa_t *spa) 1528bf82a41bSeschrock { 1529bf82a41bSeschrock return (spa->spa_is_root); 1530bf82a41bSeschrock } 15318ad4d6ddSJeff Bonwick 15328ad4d6ddSJeff Bonwick boolean_t 15338ad4d6ddSJeff Bonwick spa_writeable(spa_t *spa) 15348ad4d6ddSJeff Bonwick { 15358ad4d6ddSJeff Bonwick return (!!(spa->spa_mode & FWRITE)); 15368ad4d6ddSJeff Bonwick } 15378ad4d6ddSJeff Bonwick 15388ad4d6ddSJeff Bonwick int 15398ad4d6ddSJeff Bonwick spa_mode(spa_t *spa) 15408ad4d6ddSJeff Bonwick { 15418ad4d6ddSJeff Bonwick return (spa->spa_mode); 15428ad4d6ddSJeff Bonwick } 1543b24ab676SJeff Bonwick 1544b24ab676SJeff Bonwick uint64_t 1545b24ab676SJeff Bonwick spa_bootfs(spa_t *spa) 1546b24ab676SJeff Bonwick { 1547b24ab676SJeff Bonwick return (spa->spa_bootfs); 1548b24ab676SJeff Bonwick } 1549b24ab676SJeff Bonwick 1550b24ab676SJeff Bonwick uint64_t 1551b24ab676SJeff Bonwick spa_delegation(spa_t *spa) 1552b24ab676SJeff Bonwick { 1553b24ab676SJeff Bonwick return (spa->spa_delegation); 1554b24ab676SJeff Bonwick } 1555b24ab676SJeff Bonwick 1556b24ab676SJeff Bonwick objset_t * 1557b24ab676SJeff Bonwick spa_meta_objset(spa_t *spa) 1558b24ab676SJeff Bonwick { 1559b24ab676SJeff Bonwick return (spa->spa_meta_objset); 1560b24ab676SJeff Bonwick } 1561b24ab676SJeff Bonwick 1562b24ab676SJeff Bonwick enum zio_checksum 1563b24ab676SJeff Bonwick spa_dedup_checksum(spa_t *spa) 1564b24ab676SJeff Bonwick { 1565b24ab676SJeff Bonwick return (spa->spa_dedup_checksum); 1566b24ab676SJeff Bonwick } 1567