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 /* 2254d692b7SGeorge Wilson * Copyright 2009 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/sunddi.h> 47e14bb325SJeff Bonwick #include <sys/arc.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 189*b24ab676SJeff 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_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 434c25056deSgw mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 435c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 436c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 437c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 438c25056deSgw 439c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 440c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 441e14bb325SJeff Bonwick cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); 442c25056deSgw 443*b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 444*b24ab676SJeff Bonwick bplist_init(&spa->spa_free_bplist[t]); 445*b24ab676SJeff Bonwick bplist_init(&spa->spa_deferred_bplist); 446*b24ab676SJeff Bonwick 447e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); 448fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 449fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 4500373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 451468c413aSTim Haley spa->spa_load_max_txg = UINT64_MAX; 452fa9e4066Sahrens 453fa9e4066Sahrens refcount_create(&spa->spa_refcount); 454e14bb325SJeff Bonwick spa_config_lock_init(spa); 455fa9e4066Sahrens 456fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 457fa9e4066Sahrens 458e14bb325SJeff Bonwick mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); 4590a4e9518Sgw 4600373e76bSbonwick /* 4610373e76bSbonwick * Set the alternate root, if there is one. 4620373e76bSbonwick */ 4630373e76bSbonwick if (altroot) { 4640373e76bSbonwick spa->spa_root = spa_strdup(altroot); 4650373e76bSbonwick spa_active_count++; 4660373e76bSbonwick } 4670373e76bSbonwick 468c5904d13Seschrock /* 469c5904d13Seschrock * Every pool starts with the default cachefile 470c5904d13Seschrock */ 471c5904d13Seschrock list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), 472c5904d13Seschrock offsetof(spa_config_dirent_t, scd_link)); 473c5904d13Seschrock 474c5904d13Seschrock dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); 475c5904d13Seschrock dp->scd_path = spa_strdup(spa_config_path); 476c5904d13Seschrock list_insert_head(&spa->spa_config_list, dp); 477c5904d13Seschrock 478468c413aSTim Haley if (config != NULL) 479468c413aSTim Haley VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); 480468c413aSTim Haley 481fa9e4066Sahrens return (spa); 482fa9e4066Sahrens } 483fa9e4066Sahrens 484fa9e4066Sahrens /* 485fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 486fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 487fa9e4066Sahrens * deactivated. 488fa9e4066Sahrens */ 489fa9e4066Sahrens void 490fa9e4066Sahrens spa_remove(spa_t *spa) 491fa9e4066Sahrens { 492c5904d13Seschrock spa_config_dirent_t *dp; 493c5904d13Seschrock 494fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 495fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 496fa9e4066Sahrens 497fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 498fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 499fa9e4066Sahrens 5000373e76bSbonwick if (spa->spa_root) { 501fa9e4066Sahrens spa_strfree(spa->spa_root); 5020373e76bSbonwick spa_active_count--; 5030373e76bSbonwick } 504fa9e4066Sahrens 505c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 506c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 507c5904d13Seschrock if (dp->scd_path != NULL) 508c5904d13Seschrock spa_strfree(dp->scd_path); 509c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 510c5904d13Seschrock } 511c5904d13Seschrock 512c5904d13Seschrock list_destroy(&spa->spa_config_list); 5132f8aaab3Seschrock 514fa9e4066Sahrens spa_config_set(spa, NULL); 515fa9e4066Sahrens 516fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 51791ebeef5Sahrens 518e14bb325SJeff Bonwick spa_config_lock_destroy(spa); 519fa9e4066Sahrens 520*b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 521*b24ab676SJeff Bonwick bplist_fini(&spa->spa_free_bplist[t]); 522*b24ab676SJeff Bonwick bplist_fini(&spa->spa_deferred_bplist); 523*b24ab676SJeff Bonwick 524c25056deSgw cv_destroy(&spa->spa_async_cv); 525c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 526e14bb325SJeff Bonwick cv_destroy(&spa->spa_suspend_cv); 527c25056deSgw 5285ad82045Snd mutex_destroy(&spa->spa_async_lock); 529c25056deSgw mutex_destroy(&spa->spa_scrub_lock); 530c25056deSgw mutex_destroy(&spa->spa_errlog_lock); 531c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 53206eeb2adSek mutex_destroy(&spa->spa_history_lock); 533b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 534e14bb325SJeff Bonwick mutex_destroy(&spa->spa_suspend_lock); 5355ad82045Snd 536fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 537fa9e4066Sahrens } 538fa9e4066Sahrens 539fa9e4066Sahrens /* 540fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 541fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 542fa9e4066Sahrens */ 543fa9e4066Sahrens spa_t * 544fa9e4066Sahrens spa_next(spa_t *prev) 545fa9e4066Sahrens { 546fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 547fa9e4066Sahrens 548fa9e4066Sahrens if (prev) 549fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 550fa9e4066Sahrens else 551fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 552fa9e4066Sahrens } 553fa9e4066Sahrens 554fa9e4066Sahrens /* 555fa9e4066Sahrens * ========================================================================== 556fa9e4066Sahrens * SPA refcount functions 557fa9e4066Sahrens * ========================================================================== 558fa9e4066Sahrens */ 559fa9e4066Sahrens 560fa9e4066Sahrens /* 561fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 562fa9e4066Sahrens * have the namespace lock held. 563fa9e4066Sahrens */ 564fa9e4066Sahrens void 565fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 566fa9e4066Sahrens { 567088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || 568fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 569fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 570fa9e4066Sahrens } 571fa9e4066Sahrens 572fa9e4066Sahrens /* 573fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 574fa9e4066Sahrens * have the namespace lock held. 575fa9e4066Sahrens */ 576fa9e4066Sahrens void 577fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 578fa9e4066Sahrens { 579088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || 580fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 581fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 582fa9e4066Sahrens } 583fa9e4066Sahrens 584fa9e4066Sahrens /* 585fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 586088f3894Sahrens * spa_namespace_lock held. We really compare against spa_minref, which is the 587fa9e4066Sahrens * number of references acquired when opening a pool 588fa9e4066Sahrens */ 589fa9e4066Sahrens boolean_t 590fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 591fa9e4066Sahrens { 592fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 593fa9e4066Sahrens 594088f3894Sahrens return (refcount_count(&spa->spa_refcount) == spa->spa_minref); 595fa9e4066Sahrens } 596fa9e4066Sahrens 59799653d4eSeschrock /* 59899653d4eSeschrock * ========================================================================== 599fa94a07fSbrendan * SPA spare and l2cache tracking 60099653d4eSeschrock * ========================================================================== 60199653d4eSeschrock */ 60299653d4eSeschrock 603fa94a07fSbrendan /* 604fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 605fa94a07fSbrendan * for 'auxiliary' devices. 606fa94a07fSbrendan */ 607fa94a07fSbrendan 608fa94a07fSbrendan typedef struct spa_aux { 609fa94a07fSbrendan uint64_t aux_guid; 610fa94a07fSbrendan uint64_t aux_pool; 611fa94a07fSbrendan avl_node_t aux_avl; 612fa94a07fSbrendan int aux_count; 613fa94a07fSbrendan } spa_aux_t; 614fa94a07fSbrendan 615fa94a07fSbrendan static int 616fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 617fa94a07fSbrendan { 618fa94a07fSbrendan const spa_aux_t *sa = a; 619fa94a07fSbrendan const spa_aux_t *sb = b; 620fa94a07fSbrendan 621fa94a07fSbrendan if (sa->aux_guid < sb->aux_guid) 622fa94a07fSbrendan return (-1); 623fa94a07fSbrendan else if (sa->aux_guid > sb->aux_guid) 624fa94a07fSbrendan return (1); 625fa94a07fSbrendan else 626fa94a07fSbrendan return (0); 627fa94a07fSbrendan } 628fa94a07fSbrendan 629fa94a07fSbrendan void 630fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 631fa94a07fSbrendan { 632fa94a07fSbrendan avl_index_t where; 633fa94a07fSbrendan spa_aux_t search; 634fa94a07fSbrendan spa_aux_t *aux; 635fa94a07fSbrendan 636fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 637fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 638fa94a07fSbrendan aux->aux_count++; 639fa94a07fSbrendan } else { 640fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 641fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 642fa94a07fSbrendan aux->aux_count = 1; 643fa94a07fSbrendan avl_insert(avl, aux, where); 644fa94a07fSbrendan } 645fa94a07fSbrendan } 646fa94a07fSbrendan 647fa94a07fSbrendan void 648fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 649fa94a07fSbrendan { 650fa94a07fSbrendan spa_aux_t search; 651fa94a07fSbrendan spa_aux_t *aux; 652fa94a07fSbrendan avl_index_t where; 653fa94a07fSbrendan 654fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 655fa94a07fSbrendan aux = avl_find(avl, &search, &where); 656fa94a07fSbrendan 657fa94a07fSbrendan ASSERT(aux != NULL); 658fa94a07fSbrendan 659fa94a07fSbrendan if (--aux->aux_count == 0) { 660fa94a07fSbrendan avl_remove(avl, aux); 661fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 662fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 663fa94a07fSbrendan aux->aux_pool = 0ULL; 664fa94a07fSbrendan } 665fa94a07fSbrendan } 666fa94a07fSbrendan 667fa94a07fSbrendan boolean_t 66889a89ebfSlling spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) 669fa94a07fSbrendan { 670fa94a07fSbrendan spa_aux_t search, *found; 671fa94a07fSbrendan 672fa94a07fSbrendan search.aux_guid = guid; 67389a89ebfSlling found = avl_find(avl, &search, NULL); 674fa94a07fSbrendan 675fa94a07fSbrendan if (pool) { 676fa94a07fSbrendan if (found) 677fa94a07fSbrendan *pool = found->aux_pool; 678fa94a07fSbrendan else 679fa94a07fSbrendan *pool = 0ULL; 680fa94a07fSbrendan } 681fa94a07fSbrendan 68289a89ebfSlling if (refcnt) { 68389a89ebfSlling if (found) 68489a89ebfSlling *refcnt = found->aux_count; 68589a89ebfSlling else 68689a89ebfSlling *refcnt = 0; 68789a89ebfSlling } 68889a89ebfSlling 689fa94a07fSbrendan return (found != NULL); 690fa94a07fSbrendan } 691fa94a07fSbrendan 692fa94a07fSbrendan void 693fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 694fa94a07fSbrendan { 695fa94a07fSbrendan spa_aux_t search, *found; 696fa94a07fSbrendan avl_index_t where; 697fa94a07fSbrendan 698fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 699fa94a07fSbrendan found = avl_find(avl, &search, &where); 700fa94a07fSbrendan ASSERT(found != NULL); 701fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 702fa94a07fSbrendan 703fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 704fa94a07fSbrendan } 705fa94a07fSbrendan 70699653d4eSeschrock /* 70739c23413Seschrock * Spares are tracked globally due to the following constraints: 70839c23413Seschrock * 70939c23413Seschrock * - A spare may be part of multiple pools. 71039c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 71139c23413Seschrock * another pool. 71239c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 71339c23413Seschrock * the target is a spare in the same pool. 71439c23413Seschrock * 71539c23413Seschrock * We keep track of all spares on the system through the use of a reference 71639c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 71739c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 71839c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 71939c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 72039c23413Seschrock * pool), we also keep track of which pool its been made a part of. 72139c23413Seschrock * 72239c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 72339c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 72439c23413Seschrock * separate spare lock exists for the status query path, which does not need to 72539c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 72699653d4eSeschrock */ 72739c23413Seschrock 72899653d4eSeschrock static int 72999653d4eSeschrock spa_spare_compare(const void *a, const void *b) 73099653d4eSeschrock { 731fa94a07fSbrendan return (spa_aux_compare(a, b)); 73299653d4eSeschrock } 73399653d4eSeschrock 73499653d4eSeschrock void 73539c23413Seschrock spa_spare_add(vdev_t *vd) 73699653d4eSeschrock { 73799653d4eSeschrock mutex_enter(&spa_spare_lock); 73839c23413Seschrock ASSERT(!vd->vdev_isspare); 739fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 74039c23413Seschrock vd->vdev_isspare = B_TRUE; 74199653d4eSeschrock mutex_exit(&spa_spare_lock); 74299653d4eSeschrock } 74399653d4eSeschrock 74499653d4eSeschrock void 74539c23413Seschrock spa_spare_remove(vdev_t *vd) 74699653d4eSeschrock { 74799653d4eSeschrock mutex_enter(&spa_spare_lock); 74839c23413Seschrock ASSERT(vd->vdev_isspare); 749fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 75039c23413Seschrock vd->vdev_isspare = B_FALSE; 75199653d4eSeschrock mutex_exit(&spa_spare_lock); 75299653d4eSeschrock } 75399653d4eSeschrock 75499653d4eSeschrock boolean_t 75589a89ebfSlling spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) 75699653d4eSeschrock { 757fa94a07fSbrendan boolean_t found; 75899653d4eSeschrock 75999653d4eSeschrock mutex_enter(&spa_spare_lock); 76089a89ebfSlling found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); 76199653d4eSeschrock mutex_exit(&spa_spare_lock); 76299653d4eSeschrock 763fa94a07fSbrendan return (found); 76439c23413Seschrock } 76539c23413Seschrock 76639c23413Seschrock void 76739c23413Seschrock spa_spare_activate(vdev_t *vd) 76839c23413Seschrock { 76939c23413Seschrock mutex_enter(&spa_spare_lock); 77039c23413Seschrock ASSERT(vd->vdev_isspare); 771fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 772fa94a07fSbrendan mutex_exit(&spa_spare_lock); 773fa94a07fSbrendan } 77439c23413Seschrock 775fa94a07fSbrendan /* 776fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 777fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 778fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 779fa94a07fSbrendan */ 78039c23413Seschrock 781fa94a07fSbrendan static int 782fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 783fa94a07fSbrendan { 784fa94a07fSbrendan return (spa_aux_compare(a, b)); 785fa94a07fSbrendan } 786fa94a07fSbrendan 787fa94a07fSbrendan void 788fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 789fa94a07fSbrendan { 790fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 791fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 792fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 793fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 794fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 795fa94a07fSbrendan } 796fa94a07fSbrendan 797fa94a07fSbrendan void 798fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 799fa94a07fSbrendan { 800fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 801fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 802fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 803fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 804fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 805fa94a07fSbrendan } 806fa94a07fSbrendan 807fa94a07fSbrendan boolean_t 808fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 809fa94a07fSbrendan { 810fa94a07fSbrendan boolean_t found; 811fa94a07fSbrendan 812fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 81389a89ebfSlling found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); 814fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 815fa94a07fSbrendan 816fa94a07fSbrendan return (found); 817fa94a07fSbrendan } 818fa94a07fSbrendan 819fa94a07fSbrendan void 820fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 821fa94a07fSbrendan { 822fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 823fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 824fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 825fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 826fa94a07fSbrendan } 827fa94a07fSbrendan 828fa9e4066Sahrens /* 829fa9e4066Sahrens * ========================================================================== 830fa9e4066Sahrens * SPA vdev locking 831fa9e4066Sahrens * ========================================================================== 832fa9e4066Sahrens */ 833fa9e4066Sahrens 834fa9e4066Sahrens /* 835ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 836ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 837fa9e4066Sahrens * It returns the next transaction group for the spa_t. 838fa9e4066Sahrens */ 839fa9e4066Sahrens uint64_t 840fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 841fa9e4066Sahrens { 8423d7072f8Seschrock mutex_enter(&spa_namespace_lock); 84388ecc943SGeorge Wilson return (spa_vdev_config_enter(spa)); 84488ecc943SGeorge Wilson } 84588ecc943SGeorge Wilson 84688ecc943SGeorge Wilson /* 84788ecc943SGeorge Wilson * Internal implementation for spa_vdev_enter(). Used when a vdev 84888ecc943SGeorge Wilson * operation requires multiple syncs (i.e. removing a device) while 84988ecc943SGeorge Wilson * keeping the spa_namespace_lock held. 85088ecc943SGeorge Wilson */ 85188ecc943SGeorge Wilson uint64_t 85288ecc943SGeorge Wilson spa_vdev_config_enter(spa_t *spa) 85388ecc943SGeorge Wilson { 85488ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 8553d7072f8Seschrock 856e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 857fa9e4066Sahrens 858fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 859fa9e4066Sahrens } 860fa9e4066Sahrens 861fa9e4066Sahrens /* 86288ecc943SGeorge Wilson * Used in combination with spa_vdev_config_enter() to allow the syncing 86388ecc943SGeorge Wilson * of multiple transactions without releasing the spa_namespace_lock. 864fa9e4066Sahrens */ 86588ecc943SGeorge Wilson void 86688ecc943SGeorge Wilson spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) 867fa9e4066Sahrens { 86888ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 86988ecc943SGeorge Wilson 8700e34b6a7Sbonwick int config_changed = B_FALSE; 871ea8dc4b6Seschrock 8720373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 8730e34b6a7Sbonwick 874e14bb325SJeff Bonwick spa->spa_pending_vdev = NULL; 875e14bb325SJeff Bonwick 8760e34b6a7Sbonwick /* 8770e34b6a7Sbonwick * Reassess the DTLs. 8780e34b6a7Sbonwick */ 8790373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 8800e34b6a7Sbonwick 8810e34b6a7Sbonwick /* 8820373e76bSbonwick * If the config changed, notify the scrub thread that it must restart. 8830e34b6a7Sbonwick */ 884e14bb325SJeff Bonwick if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { 885088f3894Sahrens dsl_pool_scrub_restart(spa->spa_dsl_pool); 8860e34b6a7Sbonwick config_changed = B_TRUE; 8878f18d1faSGeorge Wilson spa->spa_config_generation++; 8880e34b6a7Sbonwick } 889ea8dc4b6Seschrock 89088ecc943SGeorge Wilson /* 89188ecc943SGeorge Wilson * Verify the metaslab classes. 89288ecc943SGeorge Wilson */ 893*b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); 894*b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); 89588ecc943SGeorge Wilson 896e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 897fa9e4066Sahrens 89888ecc943SGeorge Wilson /* 89988ecc943SGeorge Wilson * Panic the system if the specified tag requires it. This 90088ecc943SGeorge Wilson * is useful for ensuring that configurations are updated 90188ecc943SGeorge Wilson * transactionally. 90288ecc943SGeorge Wilson */ 90388ecc943SGeorge Wilson if (zio_injection_enabled) 90488ecc943SGeorge Wilson zio_handle_panic_injection(spa, tag); 90588ecc943SGeorge Wilson 906fa9e4066Sahrens /* 907fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 908fa9e4066Sahrens * that there won't be more than one config change per txg. 909fa9e4066Sahrens * This allows us to use the txg as the generation number. 910fa9e4066Sahrens */ 911fa9e4066Sahrens if (error == 0) 912fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 913fa9e4066Sahrens 914fa9e4066Sahrens if (vd != NULL) { 9158ad4d6ddSJeff Bonwick ASSERT(!vd->vdev_detached || vd->vdev_dtl_smo.smo_object == 0); 9168ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 917fa9e4066Sahrens vdev_free(vd); 9188ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 919fa9e4066Sahrens } 920fa9e4066Sahrens 921fa9e4066Sahrens /* 9220e34b6a7Sbonwick * If the config changed, update the config cache. 923fa9e4066Sahrens */ 9240e34b6a7Sbonwick if (config_changed) 925c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 92688ecc943SGeorge Wilson } 927ea8dc4b6Seschrock 92888ecc943SGeorge Wilson /* 92988ecc943SGeorge Wilson * Unlock the spa_t after adding or removing a vdev. Besides undoing the 93088ecc943SGeorge Wilson * locking of spa_vdev_enter(), we also want make sure the transactions have 93188ecc943SGeorge Wilson * synced to disk, and then update the global configuration cache with the new 93288ecc943SGeorge Wilson * information. 93388ecc943SGeorge Wilson */ 93488ecc943SGeorge Wilson int 93588ecc943SGeorge Wilson spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 93688ecc943SGeorge Wilson { 93788ecc943SGeorge Wilson spa_vdev_config_exit(spa, vd, txg, error, FTAG); 938ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 939fa9e4066Sahrens 940fa9e4066Sahrens return (error); 941fa9e4066Sahrens } 942fa9e4066Sahrens 943e14bb325SJeff Bonwick /* 944e14bb325SJeff Bonwick * Lock the given spa_t for the purpose of changing vdev state. 945e14bb325SJeff Bonwick */ 946e14bb325SJeff Bonwick void 9478f18d1faSGeorge Wilson spa_vdev_state_enter(spa_t *spa, int oplocks) 948e14bb325SJeff Bonwick { 9498f18d1faSGeorge Wilson int locks = SCL_STATE_ALL | oplocks; 9508f18d1faSGeorge Wilson 9518f18d1faSGeorge Wilson spa_config_enter(spa, locks, spa, RW_WRITER); 9528f18d1faSGeorge Wilson spa->spa_vdev_locks = locks; 953e14bb325SJeff Bonwick } 954e14bb325SJeff Bonwick 955e14bb325SJeff Bonwick int 956e14bb325SJeff Bonwick spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) 957e14bb325SJeff Bonwick { 958*b24ab676SJeff Bonwick if (vd != NULL || error == 0) 959*b24ab676SJeff Bonwick vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 960*b24ab676SJeff Bonwick 0, 0, B_FALSE); 961*b24ab676SJeff Bonwick 9628f18d1faSGeorge Wilson if (vd != NULL) { 963e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 9648f18d1faSGeorge Wilson spa->spa_config_generation++; 9658f18d1faSGeorge Wilson } 966e14bb325SJeff Bonwick 9678f18d1faSGeorge Wilson ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); 9688f18d1faSGeorge Wilson spa_config_exit(spa, spa->spa_vdev_locks, spa); 969e14bb325SJeff Bonwick 9708ad4d6ddSJeff Bonwick /* 9718ad4d6ddSJeff Bonwick * If anything changed, wait for it to sync. This ensures that, 9728ad4d6ddSJeff Bonwick * from the system administrator's perspective, zpool(1M) commands 9738ad4d6ddSJeff Bonwick * are synchronous. This is important for things like zpool offline: 9748ad4d6ddSJeff Bonwick * when the command completes, you expect no further I/O from ZFS. 9758ad4d6ddSJeff Bonwick */ 9768ad4d6ddSJeff Bonwick if (vd != NULL) 9778ad4d6ddSJeff Bonwick txg_wait_synced(spa->spa_dsl_pool, 0); 9788ad4d6ddSJeff Bonwick 979e14bb325SJeff Bonwick return (error); 980e14bb325SJeff Bonwick } 981e14bb325SJeff Bonwick 982fa9e4066Sahrens /* 983fa9e4066Sahrens * ========================================================================== 984fa9e4066Sahrens * Miscellaneous functions 985fa9e4066Sahrens * ========================================================================== 986fa9e4066Sahrens */ 987fa9e4066Sahrens 988fa9e4066Sahrens /* 989fa9e4066Sahrens * Rename a spa_t. 990fa9e4066Sahrens */ 991fa9e4066Sahrens int 992fa9e4066Sahrens spa_rename(const char *name, const char *newname) 993fa9e4066Sahrens { 994fa9e4066Sahrens spa_t *spa; 995fa9e4066Sahrens int err; 996fa9e4066Sahrens 997fa9e4066Sahrens /* 998fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 999fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 1000fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 1001ea8dc4b6Seschrock * allow recursive calls for other reasons. 1002fa9e4066Sahrens */ 1003fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 1004fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 1005fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1006fa9e4066Sahrens return (err); 1007fa9e4066Sahrens } 1008fa9e4066Sahrens 1009e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1010fa9e4066Sahrens 1011fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 1012e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); 1013fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 1014fa9e4066Sahrens 1015fa9e4066Sahrens /* 1016fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 1017fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 1018fa9e4066Sahrens * during the sync. 1019fa9e4066Sahrens */ 1020fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 1021fa9e4066Sahrens 1022e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1023fa9e4066Sahrens 10240373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 1025fa9e4066Sahrens 1026fa9e4066Sahrens /* 1027fa9e4066Sahrens * Sync the updated config cache. 1028fa9e4066Sahrens */ 1029c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 1030fa9e4066Sahrens 1031fa9e4066Sahrens spa_close(spa, FTAG); 1032fa9e4066Sahrens 1033fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1034fa9e4066Sahrens 1035fa9e4066Sahrens return (0); 1036fa9e4066Sahrens } 1037fa9e4066Sahrens 1038fa9e4066Sahrens 1039fa9e4066Sahrens /* 1040fa9e4066Sahrens * Determine whether a pool with given pool_guid exists. If device_guid is 1041fa9e4066Sahrens * non-zero, determine whether the pool exists *and* contains a device with the 1042fa9e4066Sahrens * specified device_guid. 1043fa9e4066Sahrens */ 1044fa9e4066Sahrens boolean_t 1045fa9e4066Sahrens spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 1046fa9e4066Sahrens { 1047fa9e4066Sahrens spa_t *spa; 1048fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 1049fa9e4066Sahrens 1050ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1051fa9e4066Sahrens 1052fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 1053fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 1054fa9e4066Sahrens continue; 1055fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 1056fa9e4066Sahrens continue; 105739c23413Seschrock if (spa_guid(spa) == pool_guid) { 105839c23413Seschrock if (device_guid == 0) 105939c23413Seschrock break; 106039c23413Seschrock 106139c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 106239c23413Seschrock device_guid) != NULL) 106339c23413Seschrock break; 106439c23413Seschrock 106539c23413Seschrock /* 10668654d025Sperrin * Check any devices we may be in the process of adding. 106739c23413Seschrock */ 106839c23413Seschrock if (spa->spa_pending_vdev) { 106939c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 107039c23413Seschrock device_guid) != NULL) 107139c23413Seschrock break; 107239c23413Seschrock } 107339c23413Seschrock } 1074fa9e4066Sahrens } 1075fa9e4066Sahrens 1076fa9e4066Sahrens return (spa != NULL); 1077fa9e4066Sahrens } 1078fa9e4066Sahrens 1079fa9e4066Sahrens char * 1080fa9e4066Sahrens spa_strdup(const char *s) 1081fa9e4066Sahrens { 1082fa9e4066Sahrens size_t len; 1083fa9e4066Sahrens char *new; 1084fa9e4066Sahrens 1085fa9e4066Sahrens len = strlen(s); 1086fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 1087fa9e4066Sahrens bcopy(s, new, len); 1088fa9e4066Sahrens new[len] = '\0'; 1089fa9e4066Sahrens 1090fa9e4066Sahrens return (new); 1091fa9e4066Sahrens } 1092fa9e4066Sahrens 1093fa9e4066Sahrens void 1094fa9e4066Sahrens spa_strfree(char *s) 1095fa9e4066Sahrens { 1096fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 1097fa9e4066Sahrens } 1098fa9e4066Sahrens 1099fa9e4066Sahrens uint64_t 1100fa9e4066Sahrens spa_get_random(uint64_t range) 1101fa9e4066Sahrens { 1102fa9e4066Sahrens uint64_t r; 1103fa9e4066Sahrens 1104fa9e4066Sahrens ASSERT(range != 0); 1105fa9e4066Sahrens 1106fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 1107fa9e4066Sahrens 1108fa9e4066Sahrens return (r % range); 1109fa9e4066Sahrens } 1110fa9e4066Sahrens 1111fa9e4066Sahrens void 1112*b24ab676SJeff Bonwick sprintf_blkptr(char *buf, const blkptr_t *bp) 1113fa9e4066Sahrens { 1114*b24ab676SJeff Bonwick char *type = dmu_ot[BP_GET_TYPE(bp)].ot_name; 1115*b24ab676SJeff Bonwick char *checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; 1116*b24ab676SJeff Bonwick char *compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; 1117fa9e4066Sahrens 1118*b24ab676SJeff Bonwick SPRINTF_BLKPTR(snprintf, ' ', buf, bp, type, checksum, compress); 1119fa9e4066Sahrens } 1120fa9e4066Sahrens 1121fa9e4066Sahrens void 1122fa9e4066Sahrens spa_freeze(spa_t *spa) 1123fa9e4066Sahrens { 1124fa9e4066Sahrens uint64_t freeze_txg = 0; 1125fa9e4066Sahrens 1126e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1127fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 1128fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 1129fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 1130fa9e4066Sahrens } 1131e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1132fa9e4066Sahrens if (freeze_txg != 0) 1133fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 1134fa9e4066Sahrens } 1135fa9e4066Sahrens 11360125049cSahrens void 11370125049cSahrens zfs_panic_recover(const char *fmt, ...) 11380125049cSahrens { 11390125049cSahrens va_list adx; 11400125049cSahrens 11410125049cSahrens va_start(adx, fmt); 11420125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 11430125049cSahrens va_end(adx); 11440125049cSahrens } 11450125049cSahrens 1146fa9e4066Sahrens /* 1147fa9e4066Sahrens * ========================================================================== 1148fa9e4066Sahrens * Accessor functions 1149fa9e4066Sahrens * ========================================================================== 1150fa9e4066Sahrens */ 1151fa9e4066Sahrens 1152088f3894Sahrens boolean_t 115388b7b0f2SMatthew Ahrens spa_shutting_down(spa_t *spa) 1154fa9e4066Sahrens { 115588b7b0f2SMatthew Ahrens return (spa->spa_async_suspended); 1156fa9e4066Sahrens } 1157fa9e4066Sahrens 1158fa9e4066Sahrens dsl_pool_t * 1159fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1160fa9e4066Sahrens { 1161fa9e4066Sahrens return (spa->spa_dsl_pool); 1162fa9e4066Sahrens } 1163fa9e4066Sahrens 1164fa9e4066Sahrens blkptr_t * 1165fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1166fa9e4066Sahrens { 1167fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1168fa9e4066Sahrens } 1169fa9e4066Sahrens 1170fa9e4066Sahrens void 1171fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1172fa9e4066Sahrens { 1173fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1174fa9e4066Sahrens } 1175fa9e4066Sahrens 1176fa9e4066Sahrens void 1177fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1178fa9e4066Sahrens { 1179fa9e4066Sahrens if (spa->spa_root == NULL) 1180fa9e4066Sahrens buf[0] = '\0'; 1181fa9e4066Sahrens else 1182fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1183fa9e4066Sahrens } 1184fa9e4066Sahrens 1185fa9e4066Sahrens int 1186fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1187fa9e4066Sahrens { 1188fa9e4066Sahrens return (spa->spa_sync_pass); 1189fa9e4066Sahrens } 1190fa9e4066Sahrens 1191fa9e4066Sahrens char * 1192fa9e4066Sahrens spa_name(spa_t *spa) 1193fa9e4066Sahrens { 1194fa9e4066Sahrens return (spa->spa_name); 1195fa9e4066Sahrens } 1196fa9e4066Sahrens 1197fa9e4066Sahrens uint64_t 1198fa9e4066Sahrens spa_guid(spa_t *spa) 1199fa9e4066Sahrens { 1200b5989ec7Seschrock /* 1201b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1202b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1203b5989ec7Seschrock * vdev. We stash the original pool guid in 'spa_load_guid' to handle 1204b5989ec7Seschrock * this case. 1205b5989ec7Seschrock */ 1206b5989ec7Seschrock if (spa->spa_root_vdev != NULL) 1207b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1208b5989ec7Seschrock else 1209b5989ec7Seschrock return (spa->spa_load_guid); 1210fa9e4066Sahrens } 1211fa9e4066Sahrens 1212fa9e4066Sahrens uint64_t 1213fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1214fa9e4066Sahrens { 1215fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1216fa9e4066Sahrens } 1217fa9e4066Sahrens 1218fa9e4066Sahrens uint64_t 1219fa9e4066Sahrens spa_first_txg(spa_t *spa) 1220fa9e4066Sahrens { 1221fa9e4066Sahrens return (spa->spa_first_txg); 1222fa9e4066Sahrens } 1223fa9e4066Sahrens 1224*b24ab676SJeff Bonwick uint64_t 1225*b24ab676SJeff Bonwick spa_syncing_txg(spa_t *spa) 1226*b24ab676SJeff Bonwick { 1227*b24ab676SJeff Bonwick return (spa->spa_syncing_txg); 1228*b24ab676SJeff Bonwick } 1229*b24ab676SJeff Bonwick 123088b7b0f2SMatthew Ahrens pool_state_t 1231fa9e4066Sahrens spa_state(spa_t *spa) 1232fa9e4066Sahrens { 1233fa9e4066Sahrens return (spa->spa_state); 1234fa9e4066Sahrens } 1235fa9e4066Sahrens 1236fa9e4066Sahrens uint64_t 1237fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1238fa9e4066Sahrens { 1239fa9e4066Sahrens return (spa->spa_freeze_txg); 1240fa9e4066Sahrens } 1241fa9e4066Sahrens 1242fa9e4066Sahrens /* ARGSUSED */ 1243fa9e4066Sahrens uint64_t 1244fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 1245fa9e4066Sahrens { 1246fa9e4066Sahrens /* 1247*b24ab676SJeff Bonwick * The worst case is single-sector max-parity RAID-Z blocks, in which 1248*b24ab676SJeff Bonwick * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) 1249*b24ab676SJeff Bonwick * times the size; so just assume that. Add to this the fact that 1250*b24ab676SJeff Bonwick * we can have up to 3 DVAs per bp, and one more factor of 2 because 1251*b24ab676SJeff Bonwick * the block may be dittoed with up to 3 DVAs by ddt_sync(). 125244cd46caSbillm */ 1253*b24ab676SJeff Bonwick return (lsize * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2); 125444cd46caSbillm } 125544cd46caSbillm 12560a4e9518Sgw /* 12570a4e9518Sgw * Return the failure mode that has been set to this pool. The default 12580a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 12590a4e9518Sgw */ 12600a4e9518Sgw uint8_t 12610a4e9518Sgw spa_get_failmode(spa_t *spa) 12620a4e9518Sgw { 12630a4e9518Sgw return (spa->spa_failmode); 12640a4e9518Sgw } 12650a4e9518Sgw 1266e14bb325SJeff Bonwick boolean_t 1267e14bb325SJeff Bonwick spa_suspended(spa_t *spa) 1268e14bb325SJeff Bonwick { 1269e14bb325SJeff Bonwick return (spa->spa_suspended); 1270e14bb325SJeff Bonwick } 1271e14bb325SJeff Bonwick 127244cd46caSbillm uint64_t 127344cd46caSbillm spa_version(spa_t *spa) 127444cd46caSbillm { 127544cd46caSbillm return (spa->spa_ubsync.ub_version); 127644cd46caSbillm } 127744cd46caSbillm 1278*b24ab676SJeff Bonwick boolean_t 1279*b24ab676SJeff Bonwick spa_deflate(spa_t *spa) 1280*b24ab676SJeff Bonwick { 1281*b24ab676SJeff Bonwick return (spa->spa_deflate); 1282*b24ab676SJeff Bonwick } 1283*b24ab676SJeff Bonwick 1284*b24ab676SJeff Bonwick metaslab_class_t * 1285*b24ab676SJeff Bonwick spa_normal_class(spa_t *spa) 1286*b24ab676SJeff Bonwick { 1287*b24ab676SJeff Bonwick return (spa->spa_normal_class); 1288*b24ab676SJeff Bonwick } 1289*b24ab676SJeff Bonwick 1290*b24ab676SJeff Bonwick metaslab_class_t * 1291*b24ab676SJeff Bonwick spa_log_class(spa_t *spa) 1292*b24ab676SJeff Bonwick { 1293*b24ab676SJeff Bonwick return (spa->spa_log_class); 1294*b24ab676SJeff Bonwick } 1295*b24ab676SJeff Bonwick 129644cd46caSbillm int 129744cd46caSbillm spa_max_replication(spa_t *spa) 129844cd46caSbillm { 129944cd46caSbillm /* 1300e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 130144cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 130244cd46caSbillm * replication level accordingly. 1303fa9e4066Sahrens */ 1304e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 130544cd46caSbillm return (1); 130644cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1307fa9e4066Sahrens } 1308fa9e4066Sahrens 130999653d4eSeschrock uint64_t 1310*b24ab676SJeff Bonwick dva_get_dsize_sync(spa_t *spa, const dva_t *dva) 131199653d4eSeschrock { 1312*b24ab676SJeff Bonwick uint64_t asize = DVA_GET_ASIZE(dva); 1313*b24ab676SJeff Bonwick uint64_t dsize = asize; 131499653d4eSeschrock 1315*b24ab676SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); 131699653d4eSeschrock 1317*b24ab676SJeff Bonwick if (asize != 0 && spa->spa_deflate) { 1318*b24ab676SJeff Bonwick vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); 1319*b24ab676SJeff Bonwick dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 132099653d4eSeschrock } 1321*b24ab676SJeff Bonwick 1322*b24ab676SJeff Bonwick return (dsize); 1323*b24ab676SJeff Bonwick } 1324*b24ab676SJeff Bonwick 1325*b24ab676SJeff Bonwick uint64_t 1326*b24ab676SJeff Bonwick bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) 1327*b24ab676SJeff Bonwick { 1328*b24ab676SJeff Bonwick uint64_t dsize = 0; 1329*b24ab676SJeff Bonwick 1330*b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1331*b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1332*b24ab676SJeff Bonwick 1333*b24ab676SJeff Bonwick return (dsize); 1334*b24ab676SJeff Bonwick } 1335*b24ab676SJeff Bonwick 1336*b24ab676SJeff Bonwick uint64_t 1337*b24ab676SJeff Bonwick bp_get_dsize(spa_t *spa, const blkptr_t *bp) 1338*b24ab676SJeff Bonwick { 1339*b24ab676SJeff Bonwick uint64_t dsize = 0; 1340*b24ab676SJeff Bonwick 1341*b24ab676SJeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 1342*b24ab676SJeff Bonwick 1343*b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1344*b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1345*b24ab676SJeff Bonwick 1346e14bb325SJeff Bonwick spa_config_exit(spa, SCL_VDEV, FTAG); 1347*b24ab676SJeff Bonwick 1348*b24ab676SJeff Bonwick return (dsize); 134999653d4eSeschrock } 135099653d4eSeschrock 1351fa9e4066Sahrens /* 1352fa9e4066Sahrens * ========================================================================== 1353fa9e4066Sahrens * Initialization and Termination 1354fa9e4066Sahrens * ========================================================================== 1355fa9e4066Sahrens */ 1356fa9e4066Sahrens 1357fa9e4066Sahrens static int 1358fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1359fa9e4066Sahrens { 1360fa9e4066Sahrens const spa_t *s1 = a1; 1361fa9e4066Sahrens const spa_t *s2 = a2; 1362fa9e4066Sahrens int s; 1363fa9e4066Sahrens 1364fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1365fa9e4066Sahrens if (s > 0) 1366fa9e4066Sahrens return (1); 1367fa9e4066Sahrens if (s < 0) 1368fa9e4066Sahrens return (-1); 1369fa9e4066Sahrens return (0); 1370fa9e4066Sahrens } 1371fa9e4066Sahrens 13720373e76bSbonwick int 13730373e76bSbonwick spa_busy(void) 13740373e76bSbonwick { 13750373e76bSbonwick return (spa_active_count); 13760373e76bSbonwick } 13770373e76bSbonwick 1378e7cbe64fSgw void 1379e7cbe64fSgw spa_boot_init() 1380e7cbe64fSgw { 1381e7cbe64fSgw spa_config_load(); 1382e7cbe64fSgw } 1383e7cbe64fSgw 1384fa9e4066Sahrens void 1385fa9e4066Sahrens spa_init(int mode) 1386fa9e4066Sahrens { 1387fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1388c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1389fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 1390fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1391fa9e4066Sahrens 1392fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1393fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1394fa9e4066Sahrens 1395fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 1396fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 1397fa94a07fSbrendan 1398fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 1399fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 140099653d4eSeschrock 14018ad4d6ddSJeff Bonwick spa_mode_global = mode; 1402fa9e4066Sahrens 1403fa9e4066Sahrens refcount_init(); 1404fa9e4066Sahrens unique_init(); 1405fa9e4066Sahrens zio_init(); 1406fa9e4066Sahrens dmu_init(); 1407fa9e4066Sahrens zil_init(); 140887db74c1Sek vdev_cache_stat_init(); 140991ebeef5Sahrens zfs_prop_init(); 1410990b4856Slling zpool_prop_init(); 1411fa9e4066Sahrens spa_config_load(); 1412e14bb325SJeff Bonwick l2arc_start(); 1413fa9e4066Sahrens } 1414fa9e4066Sahrens 1415fa9e4066Sahrens void 1416fa9e4066Sahrens spa_fini(void) 1417fa9e4066Sahrens { 1418e14bb325SJeff Bonwick l2arc_stop(); 1419e14bb325SJeff Bonwick 1420fa9e4066Sahrens spa_evict_all(); 1421fa9e4066Sahrens 142287db74c1Sek vdev_cache_stat_fini(); 1423fa9e4066Sahrens zil_fini(); 1424fa9e4066Sahrens dmu_fini(); 1425fa9e4066Sahrens zio_fini(); 142691ebeef5Sahrens unique_fini(); 1427fa9e4066Sahrens refcount_fini(); 1428fa9e4066Sahrens 1429fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 143099653d4eSeschrock avl_destroy(&spa_spare_avl); 1431fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 1432fa9e4066Sahrens 1433fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1434fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1435c25056deSgw mutex_destroy(&spa_spare_lock); 1436fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 1437fa9e4066Sahrens } 14386ce0521aSperrin 14396ce0521aSperrin /* 14406ce0521aSperrin * Return whether this pool has slogs. No locking needed. 14416ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 14426ce0521aSperrin * performance and not correctness 14436ce0521aSperrin */ 14446ce0521aSperrin boolean_t 14456ce0521aSperrin spa_has_slogs(spa_t *spa) 14466ce0521aSperrin { 14476ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 14486ce0521aSperrin } 1449bf82a41bSeschrock 1450*b24ab676SJeff Bonwick spa_log_state_t 1451*b24ab676SJeff Bonwick spa_get_log_state(spa_t *spa) 1452*b24ab676SJeff Bonwick { 1453*b24ab676SJeff Bonwick return (spa->spa_log_state); 1454*b24ab676SJeff Bonwick } 1455*b24ab676SJeff Bonwick 1456*b24ab676SJeff Bonwick void 1457*b24ab676SJeff Bonwick spa_set_log_state(spa_t *spa, spa_log_state_t state) 1458*b24ab676SJeff Bonwick { 1459*b24ab676SJeff Bonwick spa->spa_log_state = state; 1460*b24ab676SJeff Bonwick } 1461*b24ab676SJeff Bonwick 1462bf82a41bSeschrock boolean_t 1463bf82a41bSeschrock spa_is_root(spa_t *spa) 1464bf82a41bSeschrock { 1465bf82a41bSeschrock return (spa->spa_is_root); 1466bf82a41bSeschrock } 14678ad4d6ddSJeff Bonwick 14688ad4d6ddSJeff Bonwick boolean_t 14698ad4d6ddSJeff Bonwick spa_writeable(spa_t *spa) 14708ad4d6ddSJeff Bonwick { 14718ad4d6ddSJeff Bonwick return (!!(spa->spa_mode & FWRITE)); 14728ad4d6ddSJeff Bonwick } 14738ad4d6ddSJeff Bonwick 14748ad4d6ddSJeff Bonwick int 14758ad4d6ddSJeff Bonwick spa_mode(spa_t *spa) 14768ad4d6ddSJeff Bonwick { 14778ad4d6ddSJeff Bonwick return (spa->spa_mode); 14788ad4d6ddSJeff Bonwick } 1479*b24ab676SJeff Bonwick 1480*b24ab676SJeff Bonwick uint64_t 1481*b24ab676SJeff Bonwick spa_bootfs(spa_t *spa) 1482*b24ab676SJeff Bonwick { 1483*b24ab676SJeff Bonwick return (spa->spa_bootfs); 1484*b24ab676SJeff Bonwick } 1485*b24ab676SJeff Bonwick 1486*b24ab676SJeff Bonwick uint64_t 1487*b24ab676SJeff Bonwick spa_delegation(spa_t *spa) 1488*b24ab676SJeff Bonwick { 1489*b24ab676SJeff Bonwick return (spa->spa_delegation); 1490*b24ab676SJeff Bonwick } 1491*b24ab676SJeff Bonwick 1492*b24ab676SJeff Bonwick objset_t * 1493*b24ab676SJeff Bonwick spa_meta_objset(spa_t *spa) 1494*b24ab676SJeff Bonwick { 1495*b24ab676SJeff Bonwick return (spa->spa_meta_objset); 1496*b24ab676SJeff Bonwick } 1497*b24ab676SJeff Bonwick 1498*b24ab676SJeff Bonwick enum zio_checksum 1499*b24ab676SJeff Bonwick spa_dedup_checksum(spa_t *spa) 1500*b24ab676SJeff Bonwick { 1501*b24ab676SJeff Bonwick return (spa->spa_dedup_checksum); 1502*b24ab676SJeff Bonwick } 1503