1fa9e4066Sahrens /* 2fa9e4066Sahrens * CDDL HEADER START 3fa9e4066Sahrens * 4fa9e4066Sahrens * The contents of this file are subject to the terms of the 5ea8dc4b6Seschrock * Common Development and Distribution License (the "License"). 6ea8dc4b6Seschrock * You may not use this file except in compliance with the License. 7fa9e4066Sahrens * 8fa9e4066Sahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9fa9e4066Sahrens * or http://www.opensolaris.org/os/licensing. 10fa9e4066Sahrens * See the License for the specific language governing permissions 11fa9e4066Sahrens * and limitations under the License. 12fa9e4066Sahrens * 13fa9e4066Sahrens * When distributing Covered Code, include this CDDL HEADER in each 14fa9e4066Sahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15fa9e4066Sahrens * If applicable, add the following below this CDDL HEADER, with the 16fa9e4066Sahrens * fields enclosed by brackets "[]" replaced with your own identifying 17fa9e4066Sahrens * information: Portions Copyright [yyyy] [name of copyright owner] 18fa9e4066Sahrens * 19fa9e4066Sahrens * CDDL HEADER END 20fa9e4066Sahrens */ 21fa9e4066Sahrens /* 229842588bSGeorge Wilson * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23ad135b5dSChristopher Siden * Copyright (c) 2012 by Delphix. All rights reserved. 24e9103aaeSGarrett D'Amore * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 25fa9e4066Sahrens */ 26fa9e4066Sahrens 27fa9e4066Sahrens #include <sys/zfs_context.h> 28fa9e4066Sahrens #include <sys/spa_impl.h> 29283b8460SGeorge.Wilson #include <sys/spa_boot.h> 30fa9e4066Sahrens #include <sys/zio.h> 31fa9e4066Sahrens #include <sys/zio_checksum.h> 32fa9e4066Sahrens #include <sys/zio_compress.h> 33fa9e4066Sahrens #include <sys/dmu.h> 34fa9e4066Sahrens #include <sys/dmu_tx.h> 35fa9e4066Sahrens #include <sys/zap.h> 36fa9e4066Sahrens #include <sys/zil.h> 37fa9e4066Sahrens #include <sys/vdev_impl.h> 38fa9e4066Sahrens #include <sys/metaslab.h> 39fa9e4066Sahrens #include <sys/uberblock_impl.h> 40fa9e4066Sahrens #include <sys/txg.h> 41fa9e4066Sahrens #include <sys/avl.h> 42fa9e4066Sahrens #include <sys/unique.h> 43fa9e4066Sahrens #include <sys/dsl_pool.h> 44fa9e4066Sahrens #include <sys/dsl_dir.h> 45fa9e4066Sahrens #include <sys/dsl_prop.h> 463f9d6ad7SLin Ling #include <sys/dsl_scan.h> 47fa9e4066Sahrens #include <sys/fs/zfs.h> 486ce0521aSperrin #include <sys/metaslab_impl.h> 49e14bb325SJeff Bonwick #include <sys/arc.h> 50485bbbf5SGeorge Wilson #include <sys/ddt.h> 5191ebeef5Sahrens #include "zfs_prop.h" 52ad135b5dSChristopher Siden #include "zfeature_common.h" 53fa9e4066Sahrens 54fa9e4066Sahrens /* 55fa9e4066Sahrens * SPA locking 56fa9e4066Sahrens * 57fa9e4066Sahrens * There are four basic locks for managing spa_t structures: 58fa9e4066Sahrens * 59fa9e4066Sahrens * spa_namespace_lock (global mutex) 60fa9e4066Sahrens * 6144cd46caSbillm * This lock must be acquired to do any of the following: 62fa9e4066Sahrens * 6344cd46caSbillm * - Lookup a spa_t by name 6444cd46caSbillm * - Add or remove a spa_t from the namespace 6544cd46caSbillm * - Increase spa_refcount from non-zero 6644cd46caSbillm * - Check if spa_refcount is zero 6744cd46caSbillm * - Rename a spa_t 68ea8dc4b6Seschrock * - add/remove/attach/detach devices 6944cd46caSbillm * - Held for the duration of create/destroy/import/export 70fa9e4066Sahrens * 7144cd46caSbillm * It does not need to handle recursion. A create or destroy may 7244cd46caSbillm * reference objects (files or zvols) in other pools, but by 7344cd46caSbillm * definition they must have an existing reference, and will never need 7444cd46caSbillm * to lookup a spa_t by name. 75fa9e4066Sahrens * 76fa9e4066Sahrens * spa_refcount (per-spa refcount_t protected by mutex) 77fa9e4066Sahrens * 7844cd46caSbillm * This reference count keep track of any active users of the spa_t. The 7944cd46caSbillm * spa_t cannot be destroyed or freed while this is non-zero. Internally, 8044cd46caSbillm * the refcount is never really 'zero' - opening a pool implicitly keeps 81088f3894Sahrens * some references in the DMU. Internally we check against spa_minref, but 8244cd46caSbillm * present the image of a zero/non-zero value to consumers. 83fa9e4066Sahrens * 84e14bb325SJeff Bonwick * spa_config_lock[] (per-spa array of rwlocks) 85fa9e4066Sahrens * 8691ebeef5Sahrens * This protects the spa_t from config changes, and must be held in 8791ebeef5Sahrens * the following circumstances: 88fa9e4066Sahrens * 8944cd46caSbillm * - RW_READER to perform I/O to the spa 9044cd46caSbillm * - RW_WRITER to change the vdev config 91fa9e4066Sahrens * 92fa9e4066Sahrens * The locking order is fairly straightforward: 93fa9e4066Sahrens * 9444cd46caSbillm * spa_namespace_lock -> spa_refcount 95fa9e4066Sahrens * 9644cd46caSbillm * The namespace lock must be acquired to increase the refcount from 0 9744cd46caSbillm * or to check if it is zero. 98fa9e4066Sahrens * 99e14bb325SJeff Bonwick * spa_refcount -> spa_config_lock[] 100fa9e4066Sahrens * 10144cd46caSbillm * There must be at least one valid reference on the spa_t to acquire 10244cd46caSbillm * the config lock. 103fa9e4066Sahrens * 104e14bb325SJeff Bonwick * spa_namespace_lock -> spa_config_lock[] 105fa9e4066Sahrens * 10644cd46caSbillm * The namespace lock must always be taken before the config lock. 107fa9e4066Sahrens * 108fa9e4066Sahrens * 109e14bb325SJeff Bonwick * The spa_namespace_lock can be acquired directly and is globally visible. 110fa9e4066Sahrens * 111e14bb325SJeff Bonwick * The namespace is manipulated using the following functions, all of which 112e14bb325SJeff Bonwick * require the spa_namespace_lock to be held. 113fa9e4066Sahrens * 11444cd46caSbillm * spa_lookup() Lookup a spa_t by name. 115fa9e4066Sahrens * 11644cd46caSbillm * spa_add() Create a new spa_t in the namespace. 117fa9e4066Sahrens * 11844cd46caSbillm * spa_remove() Remove a spa_t from the namespace. This also 11944cd46caSbillm * frees up any memory associated with the spa_t. 120fa9e4066Sahrens * 12144cd46caSbillm * spa_next() Returns the next spa_t in the system, or the 12244cd46caSbillm * first if NULL is passed. 123fa9e4066Sahrens * 12444cd46caSbillm * spa_evict_all() Shutdown and remove all spa_t structures in 12544cd46caSbillm * the system. 126fa9e4066Sahrens * 127ea8dc4b6Seschrock * spa_guid_exists() Determine whether a pool/device guid exists. 128fa9e4066Sahrens * 129fa9e4066Sahrens * The spa_refcount is manipulated using the following functions: 130fa9e4066Sahrens * 13144cd46caSbillm * spa_open_ref() Adds a reference to the given spa_t. Must be 13244cd46caSbillm * called with spa_namespace_lock held if the 13344cd46caSbillm * refcount is currently zero. 134fa9e4066Sahrens * 13544cd46caSbillm * spa_close() Remove a reference from the spa_t. This will 13644cd46caSbillm * not free the spa_t or remove it from the 13744cd46caSbillm * namespace. No locking is required. 138fa9e4066Sahrens * 13944cd46caSbillm * spa_refcount_zero() Returns true if the refcount is currently 14044cd46caSbillm * zero. Must be called with spa_namespace_lock 14144cd46caSbillm * held. 142fa9e4066Sahrens * 143e14bb325SJeff Bonwick * The spa_config_lock[] is an array of rwlocks, ordered as follows: 144e14bb325SJeff Bonwick * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. 145e14bb325SJeff Bonwick * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). 146e14bb325SJeff Bonwick * 147e14bb325SJeff Bonwick * To read the configuration, it suffices to hold one of these locks as reader. 148e14bb325SJeff Bonwick * To modify the configuration, you must hold all locks as writer. To modify 149e14bb325SJeff Bonwick * vdev state without altering the vdev tree's topology (e.g. online/offline), 150e14bb325SJeff Bonwick * you must hold SCL_STATE and SCL_ZIO as writer. 151e14bb325SJeff Bonwick * 152e14bb325SJeff Bonwick * We use these distinct config locks to avoid recursive lock entry. 153e14bb325SJeff Bonwick * For example, spa_sync() (which holds SCL_CONFIG as reader) induces 154e14bb325SJeff Bonwick * block allocations (SCL_ALLOC), which may require reading space maps 155e14bb325SJeff Bonwick * from disk (dmu_read() -> zio_read() -> SCL_ZIO). 156e14bb325SJeff Bonwick * 157e14bb325SJeff Bonwick * The spa config locks cannot be normal rwlocks because we need the 158e14bb325SJeff Bonwick * ability to hand off ownership. For example, SCL_ZIO is acquired 159e14bb325SJeff Bonwick * by the issuing thread and later released by an interrupt thread. 160e14bb325SJeff Bonwick * They do, however, obey the usual write-wanted semantics to prevent 161e14bb325SJeff Bonwick * writer (i.e. system administrator) starvation. 162e14bb325SJeff Bonwick * 163e14bb325SJeff Bonwick * The lock acquisition rules are as follows: 164e14bb325SJeff Bonwick * 165e14bb325SJeff Bonwick * SCL_CONFIG 166e14bb325SJeff Bonwick * Protects changes to the vdev tree topology, such as vdev 167e14bb325SJeff Bonwick * add/remove/attach/detach. Protects the dirty config list 168e14bb325SJeff Bonwick * (spa_config_dirty_list) and the set of spares and l2arc devices. 169e14bb325SJeff Bonwick * 170e14bb325SJeff Bonwick * SCL_STATE 171e14bb325SJeff Bonwick * Protects changes to pool state and vdev state, such as vdev 172e14bb325SJeff Bonwick * online/offline/fault/degrade/clear. Protects the dirty state list 173e14bb325SJeff Bonwick * (spa_state_dirty_list) and global pool state (spa_state). 174e14bb325SJeff Bonwick * 175e14bb325SJeff Bonwick * SCL_ALLOC 176e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 177e14bb325SJeff Bonwick * Held as reader by metaslab_alloc() and metaslab_claim(). 178e14bb325SJeff Bonwick * 179e14bb325SJeff Bonwick * SCL_ZIO 180e14bb325SJeff Bonwick * Held by bp-level zios (those which have no io_vd upon entry) 181e14bb325SJeff Bonwick * to prevent changes to the vdev tree. The bp-level zio implicitly 182e14bb325SJeff Bonwick * protects all of its vdev child zios, which do not hold SCL_ZIO. 183e14bb325SJeff Bonwick * 184e14bb325SJeff Bonwick * SCL_FREE 185e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 186e14bb325SJeff Bonwick * Held as reader by metaslab_free(). SCL_FREE is distinct from 187e14bb325SJeff Bonwick * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free 188e14bb325SJeff Bonwick * blocks in zio_done() while another i/o that holds either 189e14bb325SJeff Bonwick * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. 190e14bb325SJeff Bonwick * 191e14bb325SJeff Bonwick * SCL_VDEV 192e14bb325SJeff Bonwick * Held as reader to prevent changes to the vdev tree during trivial 193b24ab676SJeff Bonwick * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the 194e14bb325SJeff Bonwick * other locks, and lower than all of them, to ensure that it's safe 195e14bb325SJeff Bonwick * to acquire regardless of caller context. 196e14bb325SJeff Bonwick * 197e14bb325SJeff Bonwick * In addition, the following rules apply: 198e14bb325SJeff Bonwick * 199e14bb325SJeff Bonwick * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. 200e14bb325SJeff Bonwick * The lock ordering is SCL_CONFIG > spa_props_lock. 201e14bb325SJeff Bonwick * 202e14bb325SJeff Bonwick * (b) I/O operations on leaf vdevs. For any zio operation that takes 203e14bb325SJeff Bonwick * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), 204e14bb325SJeff Bonwick * or zio_write_phys() -- the caller must ensure that the config cannot 205e14bb325SJeff Bonwick * cannot change in the interim, and that the vdev cannot be reopened. 206e14bb325SJeff Bonwick * SCL_STATE as reader suffices for both. 207fa9e4066Sahrens * 208ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 209fa9e4066Sahrens * 21044cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 211ea8dc4b6Seschrock * for writing. 212fa9e4066Sahrens * 21344cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 21444cd46caSbillm * to complete, sync the updated configs to the 215ea8dc4b6Seschrock * cache, and release the namespace lock. 216fa9e4066Sahrens * 217e14bb325SJeff Bonwick * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). 218e14bb325SJeff Bonwick * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual 219e14bb325SJeff Bonwick * locking is, always, based on spa_namespace_lock and spa_config_lock[]. 220e14bb325SJeff Bonwick * 221ad135b5dSChristopher Siden * spa_rename() is also implemented within this file since it requires 222e14bb325SJeff Bonwick * manipulation of the namespace. 223fa9e4066Sahrens */ 224fa9e4066Sahrens 225fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 226fa9e4066Sahrens kmutex_t spa_namespace_lock; 227fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 2280373e76bSbonwick static int spa_active_count; 229416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 230fa9e4066Sahrens 23199653d4eSeschrock static kmutex_t spa_spare_lock; 23239c23413Seschrock static avl_tree_t spa_spare_avl; 233fa94a07fSbrendan static kmutex_t spa_l2cache_lock; 234fa94a07fSbrendan static avl_tree_t spa_l2cache_avl; 23599653d4eSeschrock 236fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 2378ad4d6ddSJeff Bonwick int spa_mode_global; 238fa9e4066Sahrens 239fa9e4066Sahrens #ifdef ZFS_DEBUG 24040feaa91Sahrens /* Everything except dprintf is on by default in debug builds */ 24140feaa91Sahrens int zfs_flags = ~ZFS_DEBUG_DPRINTF; 242fa9e4066Sahrens #else 243fa9e4066Sahrens int zfs_flags = 0; 244fa9e4066Sahrens #endif 245fa9e4066Sahrens 2460125049cSahrens /* 2470125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 2480125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 2490125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 2500125049cSahrens */ 2510125049cSahrens int zfs_recover = 0; 2520125049cSahrens 253283b8460SGeorge.Wilson extern int zfs_txg_synctime_ms; 254283b8460SGeorge.Wilson 255283b8460SGeorge.Wilson /* 256283b8460SGeorge.Wilson * Expiration time in units of zfs_txg_synctime_ms. This value has two 257283b8460SGeorge.Wilson * meanings. First it is used to determine when the spa_deadman logic 258283b8460SGeorge.Wilson * should fire. By default the spa_deadman will fire if spa_sync has 259283b8460SGeorge.Wilson * not completed in 1000 * zfs_txg_synctime_ms (i.e. 1000 seconds). 260283b8460SGeorge.Wilson * Secondly, the value determines if an I/O is considered "hung". 261283b8460SGeorge.Wilson * Any I/O that has not completed in zfs_deadman_synctime is considered 262283b8460SGeorge.Wilson * "hung" resulting in a system panic. 263283b8460SGeorge.Wilson */ 264283b8460SGeorge.Wilson uint64_t zfs_deadman_synctime = 1000ULL; 265283b8460SGeorge.Wilson 266283b8460SGeorge.Wilson /* 267283b8460SGeorge.Wilson * Override the zfs deadman behavior via /etc/system. By default the 268283b8460SGeorge.Wilson * deadman is enabled except on VMware and sparc deployments. 269283b8460SGeorge.Wilson */ 270283b8460SGeorge.Wilson int zfs_deadman_enabled = -1; 271283b8460SGeorge.Wilson 272fa9e4066Sahrens 273e05725b1Sbonwick /* 274e05725b1Sbonwick * ========================================================================== 275e05725b1Sbonwick * SPA config locking 276e05725b1Sbonwick * ========================================================================== 277e05725b1Sbonwick */ 278e05725b1Sbonwick static void 279e14bb325SJeff Bonwick spa_config_lock_init(spa_t *spa) 280e14bb325SJeff Bonwick { 281e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 282e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 283e14bb325SJeff Bonwick mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); 284e14bb325SJeff Bonwick cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); 285e14bb325SJeff Bonwick refcount_create(&scl->scl_count); 286e14bb325SJeff Bonwick scl->scl_writer = NULL; 287e14bb325SJeff Bonwick scl->scl_write_wanted = 0; 288e14bb325SJeff Bonwick } 289e05725b1Sbonwick } 290e05725b1Sbonwick 291e05725b1Sbonwick static void 292e14bb325SJeff Bonwick spa_config_lock_destroy(spa_t *spa) 293e14bb325SJeff Bonwick { 294e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 295e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 296e14bb325SJeff Bonwick mutex_destroy(&scl->scl_lock); 297e14bb325SJeff Bonwick cv_destroy(&scl->scl_cv); 298e14bb325SJeff Bonwick refcount_destroy(&scl->scl_count); 299e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL); 300e14bb325SJeff Bonwick ASSERT(scl->scl_write_wanted == 0); 301e14bb325SJeff Bonwick } 302e14bb325SJeff Bonwick } 303e14bb325SJeff Bonwick 304e14bb325SJeff Bonwick int 305e14bb325SJeff Bonwick spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) 306e05725b1Sbonwick { 307e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 308e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 309e14bb325SJeff Bonwick if (!(locks & (1 << i))) 310e14bb325SJeff Bonwick continue; 311e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 312e14bb325SJeff Bonwick if (rw == RW_READER) { 313e14bb325SJeff Bonwick if (scl->scl_writer || scl->scl_write_wanted) { 314e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 315e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 316e14bb325SJeff Bonwick return (0); 317e14bb325SJeff Bonwick } 318e14bb325SJeff Bonwick } else { 319e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 320e14bb325SJeff Bonwick if (!refcount_is_zero(&scl->scl_count)) { 321e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 322e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 323e14bb325SJeff Bonwick return (0); 324e14bb325SJeff Bonwick } 325e14bb325SJeff Bonwick scl->scl_writer = curthread; 326e14bb325SJeff Bonwick } 327e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 328e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 329e14bb325SJeff Bonwick } 330e14bb325SJeff Bonwick return (1); 331e05725b1Sbonwick } 332e05725b1Sbonwick 333e05725b1Sbonwick void 334e14bb325SJeff Bonwick spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) 335e05725b1Sbonwick { 336f64c0e34SEric Taylor int wlocks_held = 0; 337f64c0e34SEric Taylor 338e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 339e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 340f64c0e34SEric Taylor if (scl->scl_writer == curthread) 341f64c0e34SEric Taylor wlocks_held |= (1 << i); 342e14bb325SJeff Bonwick if (!(locks & (1 << i))) 343e14bb325SJeff Bonwick continue; 344e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 345e14bb325SJeff Bonwick if (rw == RW_READER) { 346e14bb325SJeff Bonwick while (scl->scl_writer || scl->scl_write_wanted) { 347e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 348e14bb325SJeff Bonwick } 349e14bb325SJeff Bonwick } else { 350e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 351e14bb325SJeff Bonwick while (!refcount_is_zero(&scl->scl_count)) { 352e14bb325SJeff Bonwick scl->scl_write_wanted++; 353e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 354e14bb325SJeff Bonwick scl->scl_write_wanted--; 355e14bb325SJeff Bonwick } 356e14bb325SJeff Bonwick scl->scl_writer = curthread; 357e14bb325SJeff Bonwick } 358e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 359e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 360e05725b1Sbonwick } 361f64c0e34SEric Taylor ASSERT(wlocks_held <= locks); 362e05725b1Sbonwick } 363e05725b1Sbonwick 364e05725b1Sbonwick void 365e14bb325SJeff Bonwick spa_config_exit(spa_t *spa, int locks, void *tag) 366e05725b1Sbonwick { 367e14bb325SJeff Bonwick for (int i = SCL_LOCKS - 1; i >= 0; i--) { 368e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 369e14bb325SJeff Bonwick if (!(locks & (1 << i))) 370e14bb325SJeff Bonwick continue; 371e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 372e14bb325SJeff Bonwick ASSERT(!refcount_is_zero(&scl->scl_count)); 373e14bb325SJeff Bonwick if (refcount_remove(&scl->scl_count, tag) == 0) { 374e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL || 375e14bb325SJeff Bonwick scl->scl_writer == curthread); 376e14bb325SJeff Bonwick scl->scl_writer = NULL; /* OK in either case */ 377e14bb325SJeff Bonwick cv_broadcast(&scl->scl_cv); 378e14bb325SJeff Bonwick } 379e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 380e05725b1Sbonwick } 381e05725b1Sbonwick } 382e05725b1Sbonwick 383e14bb325SJeff Bonwick int 384e14bb325SJeff Bonwick spa_config_held(spa_t *spa, int locks, krw_t rw) 385e05725b1Sbonwick { 386e14bb325SJeff Bonwick int locks_held = 0; 387e05725b1Sbonwick 388e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 389e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 390e14bb325SJeff Bonwick if (!(locks & (1 << i))) 391e14bb325SJeff Bonwick continue; 392e14bb325SJeff Bonwick if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) || 393e14bb325SJeff Bonwick (rw == RW_WRITER && scl->scl_writer == curthread)) 394e14bb325SJeff Bonwick locks_held |= 1 << i; 395e14bb325SJeff Bonwick } 396e14bb325SJeff Bonwick 397e14bb325SJeff Bonwick return (locks_held); 398e05725b1Sbonwick } 399e05725b1Sbonwick 400fa9e4066Sahrens /* 401fa9e4066Sahrens * ========================================================================== 402fa9e4066Sahrens * SPA namespace functions 403fa9e4066Sahrens * ========================================================================== 404fa9e4066Sahrens */ 405fa9e4066Sahrens 406fa9e4066Sahrens /* 407fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 408fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 409fa9e4066Sahrens */ 410fa9e4066Sahrens spa_t * 411fa9e4066Sahrens spa_lookup(const char *name) 412fa9e4066Sahrens { 413e14bb325SJeff Bonwick static spa_t search; /* spa_t is large; don't allocate on stack */ 414e14bb325SJeff Bonwick spa_t *spa; 415fa9e4066Sahrens avl_index_t where; 41640feaa91Sahrens char c; 41740feaa91Sahrens char *cp; 418fa9e4066Sahrens 419fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 420fa9e4066Sahrens 42140feaa91Sahrens /* 42240feaa91Sahrens * If it's a full dataset name, figure out the pool name and 42340feaa91Sahrens * just use that. 42440feaa91Sahrens */ 42540feaa91Sahrens cp = strpbrk(name, "/@"); 42640feaa91Sahrens if (cp) { 42740feaa91Sahrens c = *cp; 42840feaa91Sahrens *cp = '\0'; 42940feaa91Sahrens } 43040feaa91Sahrens 431e14bb325SJeff Bonwick (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); 432fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 433fa9e4066Sahrens 43440feaa91Sahrens if (cp) 43540feaa91Sahrens *cp = c; 43640feaa91Sahrens 437fa9e4066Sahrens return (spa); 438fa9e4066Sahrens } 439fa9e4066Sahrens 440283b8460SGeorge.Wilson /* 441283b8460SGeorge.Wilson * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. 442283b8460SGeorge.Wilson * If the zfs_deadman_enabled flag is set then it inspects all vdev queues 443283b8460SGeorge.Wilson * looking for potentially hung I/Os. 444283b8460SGeorge.Wilson */ 445283b8460SGeorge.Wilson void 446283b8460SGeorge.Wilson spa_deadman(void *arg) 447283b8460SGeorge.Wilson { 448283b8460SGeorge.Wilson spa_t *spa = arg; 449283b8460SGeorge.Wilson 450283b8460SGeorge.Wilson zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", 451283b8460SGeorge.Wilson (gethrtime() - spa->spa_sync_starttime) / NANOSEC, 452283b8460SGeorge.Wilson ++spa->spa_deadman_calls); 453283b8460SGeorge.Wilson if (zfs_deadman_enabled) 454283b8460SGeorge.Wilson vdev_deadman(spa->spa_root_vdev); 455283b8460SGeorge.Wilson } 456283b8460SGeorge.Wilson 457fa9e4066Sahrens /* 458fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 459fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 460fa9e4066Sahrens * exist by calling spa_lookup() first. 461fa9e4066Sahrens */ 462fa9e4066Sahrens spa_t * 463468c413aSTim Haley spa_add(const char *name, nvlist_t *config, const char *altroot) 464fa9e4066Sahrens { 465fa9e4066Sahrens spa_t *spa; 466c5904d13Seschrock spa_config_dirent_t *dp; 467283b8460SGeorge.Wilson cyc_handler_t hdlr; 468283b8460SGeorge.Wilson cyc_time_t when; 469fa9e4066Sahrens 470fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 471fa9e4066Sahrens 472fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 473fa9e4066Sahrens 474c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 475c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 47635a5a358SJonathan Adams mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 477c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 47835a5a358SJonathan Adams mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); 479c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 48035a5a358SJonathan Adams mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 481a1521560SJeff Bonwick mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); 482a1521560SJeff Bonwick mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); 483*c3a66015SMatthew Ahrens mutex_init(&spa->spa_iokstat_lock, NULL, MUTEX_DEFAULT, NULL); 484c25056deSgw 485c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 48635a5a358SJonathan Adams cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); 487c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 488e14bb325SJeff Bonwick cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); 489c25056deSgw 490b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 491cde58dbcSMatthew Ahrens bplist_create(&spa->spa_free_bplist[t]); 492b24ab676SJeff Bonwick 493e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); 494fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 495fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 4960373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 497468c413aSTim Haley spa->spa_load_max_txg = UINT64_MAX; 49835a5a358SJonathan Adams spa->spa_proc = &p0; 49935a5a358SJonathan Adams spa->spa_proc_state = SPA_PROC_NONE; 500fa9e4066Sahrens 501283b8460SGeorge.Wilson hdlr.cyh_func = spa_deadman; 502283b8460SGeorge.Wilson hdlr.cyh_arg = spa; 503283b8460SGeorge.Wilson hdlr.cyh_level = CY_LOW_LEVEL; 504283b8460SGeorge.Wilson 505283b8460SGeorge.Wilson spa->spa_deadman_synctime = zfs_deadman_synctime * 506283b8460SGeorge.Wilson zfs_txg_synctime_ms * MICROSEC; 507283b8460SGeorge.Wilson 508283b8460SGeorge.Wilson /* 509283b8460SGeorge.Wilson * This determines how often we need to check for hung I/Os after 510283b8460SGeorge.Wilson * the cyclic has already fired. Since checking for hung I/Os is 511283b8460SGeorge.Wilson * an expensive operation we don't want to check too frequently. 512283b8460SGeorge.Wilson * Instead wait for 5 synctimes before checking again. 513283b8460SGeorge.Wilson */ 514283b8460SGeorge.Wilson when.cyt_interval = 5ULL * zfs_txg_synctime_ms * MICROSEC; 515283b8460SGeorge.Wilson when.cyt_when = CY_INFINITY; 516283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 517283b8460SGeorge.Wilson spa->spa_deadman_cycid = cyclic_add(&hdlr, &when); 518283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 519283b8460SGeorge.Wilson 520fa9e4066Sahrens refcount_create(&spa->spa_refcount); 521e14bb325SJeff Bonwick spa_config_lock_init(spa); 522fa9e4066Sahrens 523fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 524fa9e4066Sahrens 5250373e76bSbonwick /* 5260373e76bSbonwick * Set the alternate root, if there is one. 5270373e76bSbonwick */ 5280373e76bSbonwick if (altroot) { 5290373e76bSbonwick spa->spa_root = spa_strdup(altroot); 5300373e76bSbonwick spa_active_count++; 5310373e76bSbonwick } 5320373e76bSbonwick 533c5904d13Seschrock /* 534c5904d13Seschrock * Every pool starts with the default cachefile 535c5904d13Seschrock */ 536c5904d13Seschrock list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), 537c5904d13Seschrock offsetof(spa_config_dirent_t, scd_link)); 538c5904d13Seschrock 539c5904d13Seschrock dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); 540ef912c80STim Haley dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); 541c5904d13Seschrock list_insert_head(&spa->spa_config_list, dp); 542c5904d13Seschrock 5434b964adaSGeorge Wilson VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, 5444b964adaSGeorge Wilson KM_SLEEP) == 0); 5454b964adaSGeorge Wilson 546ad135b5dSChristopher Siden if (config != NULL) { 547ad135b5dSChristopher Siden nvlist_t *features; 548ad135b5dSChristopher Siden 549ad135b5dSChristopher Siden if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, 550ad135b5dSChristopher Siden &features) == 0) { 551ad135b5dSChristopher Siden VERIFY(nvlist_dup(features, &spa->spa_label_features, 552ad135b5dSChristopher Siden 0) == 0); 553ad135b5dSChristopher Siden } 554ad135b5dSChristopher Siden 555468c413aSTim Haley VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); 556ad135b5dSChristopher Siden } 557ad135b5dSChristopher Siden 558ad135b5dSChristopher Siden if (spa->spa_label_features == NULL) { 559ad135b5dSChristopher Siden VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, 560ad135b5dSChristopher Siden KM_SLEEP) == 0); 561ad135b5dSChristopher Siden } 562468c413aSTim Haley 563*c3a66015SMatthew Ahrens spa->spa_iokstat = kstat_create("zfs", 0, name, 564*c3a66015SMatthew Ahrens "disk", KSTAT_TYPE_IO, 1, 0); 565*c3a66015SMatthew Ahrens if (spa->spa_iokstat) { 566*c3a66015SMatthew Ahrens spa->spa_iokstat->ks_lock = &spa->spa_iokstat_lock; 567*c3a66015SMatthew Ahrens kstat_install(spa->spa_iokstat); 568*c3a66015SMatthew Ahrens } 569*c3a66015SMatthew Ahrens 570fa9e4066Sahrens return (spa); 571fa9e4066Sahrens } 572fa9e4066Sahrens 573fa9e4066Sahrens /* 574fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 575fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 576fa9e4066Sahrens * deactivated. 577fa9e4066Sahrens */ 578fa9e4066Sahrens void 579fa9e4066Sahrens spa_remove(spa_t *spa) 580fa9e4066Sahrens { 581c5904d13Seschrock spa_config_dirent_t *dp; 582c5904d13Seschrock 583fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 584fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 585fa9e4066Sahrens 5861195e687SMark J Musante nvlist_free(spa->spa_config_splitting); 5871195e687SMark J Musante 588fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 589fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 590fa9e4066Sahrens 5910373e76bSbonwick if (spa->spa_root) { 592fa9e4066Sahrens spa_strfree(spa->spa_root); 5930373e76bSbonwick spa_active_count--; 5940373e76bSbonwick } 595fa9e4066Sahrens 596c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 597c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 598c5904d13Seschrock if (dp->scd_path != NULL) 599c5904d13Seschrock spa_strfree(dp->scd_path); 600c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 601c5904d13Seschrock } 602c5904d13Seschrock 603c5904d13Seschrock list_destroy(&spa->spa_config_list); 6042f8aaab3Seschrock 605ad135b5dSChristopher Siden nvlist_free(spa->spa_label_features); 6064b964adaSGeorge Wilson nvlist_free(spa->spa_load_info); 607fa9e4066Sahrens spa_config_set(spa, NULL); 608fa9e4066Sahrens 609283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 610283b8460SGeorge.Wilson if (spa->spa_deadman_cycid != CYCLIC_NONE) 611283b8460SGeorge.Wilson cyclic_remove(spa->spa_deadman_cycid); 612283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 613283b8460SGeorge.Wilson spa->spa_deadman_cycid = CYCLIC_NONE; 614283b8460SGeorge.Wilson 615fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 61691ebeef5Sahrens 617e14bb325SJeff Bonwick spa_config_lock_destroy(spa); 618fa9e4066Sahrens 619*c3a66015SMatthew Ahrens kstat_delete(spa->spa_iokstat); 620*c3a66015SMatthew Ahrens spa->spa_iokstat = NULL; 621*c3a66015SMatthew Ahrens 622b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 623cde58dbcSMatthew Ahrens bplist_destroy(&spa->spa_free_bplist[t]); 624b24ab676SJeff Bonwick 625c25056deSgw cv_destroy(&spa->spa_async_cv); 62635a5a358SJonathan Adams cv_destroy(&spa->spa_proc_cv); 627c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 628e14bb325SJeff Bonwick cv_destroy(&spa->spa_suspend_cv); 629c25056deSgw 6305ad82045Snd mutex_destroy(&spa->spa_async_lock); 631c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 63235a5a358SJonathan Adams mutex_destroy(&spa->spa_errlog_lock); 63306eeb2adSek mutex_destroy(&spa->spa_history_lock); 63435a5a358SJonathan Adams mutex_destroy(&spa->spa_proc_lock); 635b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 63635a5a358SJonathan Adams mutex_destroy(&spa->spa_scrub_lock); 637e14bb325SJeff Bonwick mutex_destroy(&spa->spa_suspend_lock); 638a1521560SJeff Bonwick mutex_destroy(&spa->spa_vdev_top_lock); 639*c3a66015SMatthew Ahrens mutex_destroy(&spa->spa_iokstat_lock); 6405ad82045Snd 641fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 642fa9e4066Sahrens } 643fa9e4066Sahrens 644fa9e4066Sahrens /* 645fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 646fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 647fa9e4066Sahrens */ 648fa9e4066Sahrens spa_t * 649fa9e4066Sahrens spa_next(spa_t *prev) 650fa9e4066Sahrens { 651fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 652fa9e4066Sahrens 653fa9e4066Sahrens if (prev) 654fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 655fa9e4066Sahrens else 656fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 657fa9e4066Sahrens } 658fa9e4066Sahrens 659fa9e4066Sahrens /* 660fa9e4066Sahrens * ========================================================================== 661fa9e4066Sahrens * SPA refcount functions 662fa9e4066Sahrens * ========================================================================== 663fa9e4066Sahrens */ 664fa9e4066Sahrens 665fa9e4066Sahrens /* 666fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 667fa9e4066Sahrens * have the namespace lock held. 668fa9e4066Sahrens */ 669fa9e4066Sahrens void 670fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 671fa9e4066Sahrens { 672088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || 673fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 674fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 675fa9e4066Sahrens } 676fa9e4066Sahrens 677fa9e4066Sahrens /* 678fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 679fa9e4066Sahrens * have the namespace lock held. 680fa9e4066Sahrens */ 681fa9e4066Sahrens void 682fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 683fa9e4066Sahrens { 684088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || 685fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 686fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 687fa9e4066Sahrens } 688fa9e4066Sahrens 689fa9e4066Sahrens /* 690fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 691088f3894Sahrens * spa_namespace_lock held. We really compare against spa_minref, which is the 692fa9e4066Sahrens * number of references acquired when opening a pool 693fa9e4066Sahrens */ 694fa9e4066Sahrens boolean_t 695fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 696fa9e4066Sahrens { 697fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 698fa9e4066Sahrens 699088f3894Sahrens return (refcount_count(&spa->spa_refcount) == spa->spa_minref); 700fa9e4066Sahrens } 701fa9e4066Sahrens 70299653d4eSeschrock /* 70399653d4eSeschrock * ========================================================================== 704fa94a07fSbrendan * SPA spare and l2cache tracking 70599653d4eSeschrock * ========================================================================== 70699653d4eSeschrock */ 70799653d4eSeschrock 708fa94a07fSbrendan /* 709fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 710fa94a07fSbrendan * for 'auxiliary' devices. 711fa94a07fSbrendan */ 712fa94a07fSbrendan 713fa94a07fSbrendan typedef struct spa_aux { 714fa94a07fSbrendan uint64_t aux_guid; 715fa94a07fSbrendan uint64_t aux_pool; 716fa94a07fSbrendan avl_node_t aux_avl; 717fa94a07fSbrendan int aux_count; 718fa94a07fSbrendan } spa_aux_t; 719fa94a07fSbrendan 720fa94a07fSbrendan static int 721fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 722fa94a07fSbrendan { 723fa94a07fSbrendan const spa_aux_t *sa = a; 724fa94a07fSbrendan const spa_aux_t *sb = b; 725fa94a07fSbrendan 726fa94a07fSbrendan if (sa->aux_guid < sb->aux_guid) 727fa94a07fSbrendan return (-1); 728fa94a07fSbrendan else if (sa->aux_guid > sb->aux_guid) 729fa94a07fSbrendan return (1); 730fa94a07fSbrendan else 731fa94a07fSbrendan return (0); 732fa94a07fSbrendan } 733fa94a07fSbrendan 734fa94a07fSbrendan void 735fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 736fa94a07fSbrendan { 737fa94a07fSbrendan avl_index_t where; 738fa94a07fSbrendan spa_aux_t search; 739fa94a07fSbrendan spa_aux_t *aux; 740fa94a07fSbrendan 741fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 742fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 743fa94a07fSbrendan aux->aux_count++; 744fa94a07fSbrendan } else { 745fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 746fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 747fa94a07fSbrendan aux->aux_count = 1; 748fa94a07fSbrendan avl_insert(avl, aux, where); 749fa94a07fSbrendan } 750fa94a07fSbrendan } 751fa94a07fSbrendan 752fa94a07fSbrendan void 753fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 754fa94a07fSbrendan { 755fa94a07fSbrendan spa_aux_t search; 756fa94a07fSbrendan spa_aux_t *aux; 757fa94a07fSbrendan avl_index_t where; 758fa94a07fSbrendan 759fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 760fa94a07fSbrendan aux = avl_find(avl, &search, &where); 761fa94a07fSbrendan 762fa94a07fSbrendan ASSERT(aux != NULL); 763fa94a07fSbrendan 764fa94a07fSbrendan if (--aux->aux_count == 0) { 765fa94a07fSbrendan avl_remove(avl, aux); 766fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 767fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 768fa94a07fSbrendan aux->aux_pool = 0ULL; 769fa94a07fSbrendan } 770fa94a07fSbrendan } 771fa94a07fSbrendan 772fa94a07fSbrendan boolean_t 77389a89ebfSlling spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) 774fa94a07fSbrendan { 775fa94a07fSbrendan spa_aux_t search, *found; 776fa94a07fSbrendan 777fa94a07fSbrendan search.aux_guid = guid; 77889a89ebfSlling found = avl_find(avl, &search, NULL); 779fa94a07fSbrendan 780fa94a07fSbrendan if (pool) { 781fa94a07fSbrendan if (found) 782fa94a07fSbrendan *pool = found->aux_pool; 783fa94a07fSbrendan else 784fa94a07fSbrendan *pool = 0ULL; 785fa94a07fSbrendan } 786fa94a07fSbrendan 78789a89ebfSlling if (refcnt) { 78889a89ebfSlling if (found) 78989a89ebfSlling *refcnt = found->aux_count; 79089a89ebfSlling else 79189a89ebfSlling *refcnt = 0; 79289a89ebfSlling } 79389a89ebfSlling 794fa94a07fSbrendan return (found != NULL); 795fa94a07fSbrendan } 796fa94a07fSbrendan 797fa94a07fSbrendan void 798fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 799fa94a07fSbrendan { 800fa94a07fSbrendan spa_aux_t search, *found; 801fa94a07fSbrendan avl_index_t where; 802fa94a07fSbrendan 803fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 804fa94a07fSbrendan found = avl_find(avl, &search, &where); 805fa94a07fSbrendan ASSERT(found != NULL); 806fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 807fa94a07fSbrendan 808fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 809fa94a07fSbrendan } 810fa94a07fSbrendan 81199653d4eSeschrock /* 81239c23413Seschrock * Spares are tracked globally due to the following constraints: 81339c23413Seschrock * 81439c23413Seschrock * - A spare may be part of multiple pools. 81539c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 81639c23413Seschrock * another pool. 81739c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 81839c23413Seschrock * the target is a spare in the same pool. 81939c23413Seschrock * 82039c23413Seschrock * We keep track of all spares on the system through the use of a reference 82139c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 82239c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 82339c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 82439c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 82539c23413Seschrock * pool), we also keep track of which pool its been made a part of. 82639c23413Seschrock * 82739c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 82839c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 82939c23413Seschrock * separate spare lock exists for the status query path, which does not need to 83039c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 83199653d4eSeschrock */ 83239c23413Seschrock 83399653d4eSeschrock static int 83499653d4eSeschrock spa_spare_compare(const void *a, const void *b) 83599653d4eSeschrock { 836fa94a07fSbrendan return (spa_aux_compare(a, b)); 83799653d4eSeschrock } 83899653d4eSeschrock 83999653d4eSeschrock void 84039c23413Seschrock spa_spare_add(vdev_t *vd) 84199653d4eSeschrock { 84299653d4eSeschrock mutex_enter(&spa_spare_lock); 84339c23413Seschrock ASSERT(!vd->vdev_isspare); 844fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 84539c23413Seschrock vd->vdev_isspare = B_TRUE; 84699653d4eSeschrock mutex_exit(&spa_spare_lock); 84799653d4eSeschrock } 84899653d4eSeschrock 84999653d4eSeschrock void 85039c23413Seschrock spa_spare_remove(vdev_t *vd) 85199653d4eSeschrock { 85299653d4eSeschrock mutex_enter(&spa_spare_lock); 85339c23413Seschrock ASSERT(vd->vdev_isspare); 854fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 85539c23413Seschrock vd->vdev_isspare = B_FALSE; 85699653d4eSeschrock mutex_exit(&spa_spare_lock); 85799653d4eSeschrock } 85899653d4eSeschrock 85999653d4eSeschrock boolean_t 86089a89ebfSlling spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) 86199653d4eSeschrock { 862fa94a07fSbrendan boolean_t found; 86399653d4eSeschrock 86499653d4eSeschrock mutex_enter(&spa_spare_lock); 86589a89ebfSlling found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); 86699653d4eSeschrock mutex_exit(&spa_spare_lock); 86799653d4eSeschrock 868fa94a07fSbrendan return (found); 86939c23413Seschrock } 87039c23413Seschrock 87139c23413Seschrock void 87239c23413Seschrock spa_spare_activate(vdev_t *vd) 87339c23413Seschrock { 87439c23413Seschrock mutex_enter(&spa_spare_lock); 87539c23413Seschrock ASSERT(vd->vdev_isspare); 876fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 877fa94a07fSbrendan mutex_exit(&spa_spare_lock); 878fa94a07fSbrendan } 87939c23413Seschrock 880fa94a07fSbrendan /* 881fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 882fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 883fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 884fa94a07fSbrendan */ 88539c23413Seschrock 886fa94a07fSbrendan static int 887fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 888fa94a07fSbrendan { 889fa94a07fSbrendan return (spa_aux_compare(a, b)); 890fa94a07fSbrendan } 891fa94a07fSbrendan 892fa94a07fSbrendan void 893fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 894fa94a07fSbrendan { 895fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 896fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 897fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 898fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 899fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 900fa94a07fSbrendan } 901fa94a07fSbrendan 902fa94a07fSbrendan void 903fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 904fa94a07fSbrendan { 905fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 906fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 907fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 908fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 909fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 910fa94a07fSbrendan } 911fa94a07fSbrendan 912fa94a07fSbrendan boolean_t 913fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 914fa94a07fSbrendan { 915fa94a07fSbrendan boolean_t found; 916fa94a07fSbrendan 917fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 91889a89ebfSlling found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); 919fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 920fa94a07fSbrendan 921fa94a07fSbrendan return (found); 922fa94a07fSbrendan } 923fa94a07fSbrendan 924fa94a07fSbrendan void 925fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 926fa94a07fSbrendan { 927fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 928fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 929fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 930fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 931fa94a07fSbrendan } 932fa94a07fSbrendan 933fa9e4066Sahrens /* 934fa9e4066Sahrens * ========================================================================== 935fa9e4066Sahrens * SPA vdev locking 936fa9e4066Sahrens * ========================================================================== 937fa9e4066Sahrens */ 938fa9e4066Sahrens 939fa9e4066Sahrens /* 940ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 941ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 942fa9e4066Sahrens * It returns the next transaction group for the spa_t. 943fa9e4066Sahrens */ 944fa9e4066Sahrens uint64_t 945fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 946fa9e4066Sahrens { 947a1521560SJeff Bonwick mutex_enter(&spa->spa_vdev_top_lock); 948bbfd46c4SJeff Bonwick mutex_enter(&spa_namespace_lock); 94988ecc943SGeorge Wilson return (spa_vdev_config_enter(spa)); 95088ecc943SGeorge Wilson } 95188ecc943SGeorge Wilson 95288ecc943SGeorge Wilson /* 95388ecc943SGeorge Wilson * Internal implementation for spa_vdev_enter(). Used when a vdev 95488ecc943SGeorge Wilson * operation requires multiple syncs (i.e. removing a device) while 95588ecc943SGeorge Wilson * keeping the spa_namespace_lock held. 95688ecc943SGeorge Wilson */ 95788ecc943SGeorge Wilson uint64_t 95888ecc943SGeorge Wilson spa_vdev_config_enter(spa_t *spa) 95988ecc943SGeorge Wilson { 96088ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 9613d7072f8Seschrock 962e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 963fa9e4066Sahrens 964fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 965fa9e4066Sahrens } 966fa9e4066Sahrens 967fa9e4066Sahrens /* 96888ecc943SGeorge Wilson * Used in combination with spa_vdev_config_enter() to allow the syncing 96988ecc943SGeorge Wilson * of multiple transactions without releasing the spa_namespace_lock. 970fa9e4066Sahrens */ 97188ecc943SGeorge Wilson void 97288ecc943SGeorge Wilson spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) 973fa9e4066Sahrens { 97488ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 97588ecc943SGeorge Wilson 9760e34b6a7Sbonwick int config_changed = B_FALSE; 977ea8dc4b6Seschrock 9780373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 9790e34b6a7Sbonwick 980e14bb325SJeff Bonwick spa->spa_pending_vdev = NULL; 981e14bb325SJeff Bonwick 9820e34b6a7Sbonwick /* 9830e34b6a7Sbonwick * Reassess the DTLs. 9840e34b6a7Sbonwick */ 9850373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 9860e34b6a7Sbonwick 987e14bb325SJeff Bonwick if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { 9880e34b6a7Sbonwick config_changed = B_TRUE; 9898f18d1faSGeorge Wilson spa->spa_config_generation++; 9900e34b6a7Sbonwick } 991ea8dc4b6Seschrock 99288ecc943SGeorge Wilson /* 99388ecc943SGeorge Wilson * Verify the metaslab classes. 99488ecc943SGeorge Wilson */ 995b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); 996b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); 99788ecc943SGeorge Wilson 998e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 999fa9e4066Sahrens 100088ecc943SGeorge Wilson /* 100188ecc943SGeorge Wilson * Panic the system if the specified tag requires it. This 100288ecc943SGeorge Wilson * is useful for ensuring that configurations are updated 100388ecc943SGeorge Wilson * transactionally. 100488ecc943SGeorge Wilson */ 100588ecc943SGeorge Wilson if (zio_injection_enabled) 10061195e687SMark J Musante zio_handle_panic_injection(spa, tag, 0); 100788ecc943SGeorge Wilson 1008fa9e4066Sahrens /* 1009fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 1010fa9e4066Sahrens * that there won't be more than one config change per txg. 1011fa9e4066Sahrens * This allows us to use the txg as the generation number. 1012fa9e4066Sahrens */ 1013fa9e4066Sahrens if (error == 0) 1014fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 1015fa9e4066Sahrens 1016fa9e4066Sahrens if (vd != NULL) { 10178ad4d6ddSJeff Bonwick ASSERT(!vd->vdev_detached || vd->vdev_dtl_smo.smo_object == 0); 10188ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1019fa9e4066Sahrens vdev_free(vd); 10208ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1021fa9e4066Sahrens } 1022fa9e4066Sahrens 1023fa9e4066Sahrens /* 10240e34b6a7Sbonwick * If the config changed, update the config cache. 1025fa9e4066Sahrens */ 10260e34b6a7Sbonwick if (config_changed) 1027c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 102888ecc943SGeorge Wilson } 1029ea8dc4b6Seschrock 103088ecc943SGeorge Wilson /* 103188ecc943SGeorge Wilson * Unlock the spa_t after adding or removing a vdev. Besides undoing the 103288ecc943SGeorge Wilson * locking of spa_vdev_enter(), we also want make sure the transactions have 103388ecc943SGeorge Wilson * synced to disk, and then update the global configuration cache with the new 103488ecc943SGeorge Wilson * information. 103588ecc943SGeorge Wilson */ 103688ecc943SGeorge Wilson int 103788ecc943SGeorge Wilson spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 103888ecc943SGeorge Wilson { 103988ecc943SGeorge Wilson spa_vdev_config_exit(spa, vd, txg, error, FTAG); 1040ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1041bbfd46c4SJeff Bonwick mutex_exit(&spa->spa_vdev_top_lock); 1042fa9e4066Sahrens 1043fa9e4066Sahrens return (error); 1044fa9e4066Sahrens } 1045fa9e4066Sahrens 1046e14bb325SJeff Bonwick /* 1047e14bb325SJeff Bonwick * Lock the given spa_t for the purpose of changing vdev state. 1048e14bb325SJeff Bonwick */ 1049e14bb325SJeff Bonwick void 10508f18d1faSGeorge Wilson spa_vdev_state_enter(spa_t *spa, int oplocks) 1051e14bb325SJeff Bonwick { 10528f18d1faSGeorge Wilson int locks = SCL_STATE_ALL | oplocks; 10538f18d1faSGeorge Wilson 1054dcba9f3fSGeorge Wilson /* 1055dcba9f3fSGeorge Wilson * Root pools may need to read of the underlying devfs filesystem 1056dcba9f3fSGeorge Wilson * when opening up a vdev. Unfortunately if we're holding the 1057dcba9f3fSGeorge Wilson * SCL_ZIO lock it will result in a deadlock when we try to issue 1058dcba9f3fSGeorge Wilson * the read from the root filesystem. Instead we "prefetch" 1059dcba9f3fSGeorge Wilson * the associated vnodes that we need prior to opening the 1060dcba9f3fSGeorge Wilson * underlying devices and cache them so that we can prevent 1061dcba9f3fSGeorge Wilson * any I/O when we are doing the actual open. 1062dcba9f3fSGeorge Wilson */ 1063dcba9f3fSGeorge Wilson if (spa_is_root(spa)) { 10649842588bSGeorge Wilson int low = locks & ~(SCL_ZIO - 1); 10659842588bSGeorge Wilson int high = locks & ~low; 10669842588bSGeorge Wilson 10679842588bSGeorge Wilson spa_config_enter(spa, high, spa, RW_WRITER); 1068dcba9f3fSGeorge Wilson vdev_hold(spa->spa_root_vdev); 10699842588bSGeorge Wilson spa_config_enter(spa, low, spa, RW_WRITER); 1070dcba9f3fSGeorge Wilson } else { 1071dcba9f3fSGeorge Wilson spa_config_enter(spa, locks, spa, RW_WRITER); 1072dcba9f3fSGeorge Wilson } 10738f18d1faSGeorge Wilson spa->spa_vdev_locks = locks; 1074e14bb325SJeff Bonwick } 1075e14bb325SJeff Bonwick 1076e14bb325SJeff Bonwick int 1077e14bb325SJeff Bonwick spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) 1078e14bb325SJeff Bonwick { 1079c6065d0fSGeorge Wilson boolean_t config_changed = B_FALSE; 1080c6065d0fSGeorge Wilson 1081b24ab676SJeff Bonwick if (vd != NULL || error == 0) 1082b24ab676SJeff Bonwick vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 1083b24ab676SJeff Bonwick 0, 0, B_FALSE); 1084b24ab676SJeff Bonwick 10858f18d1faSGeorge Wilson if (vd != NULL) { 1086e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 1087c6065d0fSGeorge Wilson config_changed = B_TRUE; 10888f18d1faSGeorge Wilson spa->spa_config_generation++; 10898f18d1faSGeorge Wilson } 1090e14bb325SJeff Bonwick 1091dcba9f3fSGeorge Wilson if (spa_is_root(spa)) 1092dcba9f3fSGeorge Wilson vdev_rele(spa->spa_root_vdev); 1093dcba9f3fSGeorge Wilson 10948f18d1faSGeorge Wilson ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); 10958f18d1faSGeorge Wilson spa_config_exit(spa, spa->spa_vdev_locks, spa); 1096e14bb325SJeff Bonwick 10978ad4d6ddSJeff Bonwick /* 10988ad4d6ddSJeff Bonwick * If anything changed, wait for it to sync. This ensures that, 10998ad4d6ddSJeff Bonwick * from the system administrator's perspective, zpool(1M) commands 11008ad4d6ddSJeff Bonwick * are synchronous. This is important for things like zpool offline: 11018ad4d6ddSJeff Bonwick * when the command completes, you expect no further I/O from ZFS. 11028ad4d6ddSJeff Bonwick */ 11038ad4d6ddSJeff Bonwick if (vd != NULL) 11048ad4d6ddSJeff Bonwick txg_wait_synced(spa->spa_dsl_pool, 0); 11058ad4d6ddSJeff Bonwick 1106c6065d0fSGeorge Wilson /* 1107c6065d0fSGeorge Wilson * If the config changed, update the config cache. 1108c6065d0fSGeorge Wilson */ 1109c6065d0fSGeorge Wilson if (config_changed) { 1110c6065d0fSGeorge Wilson mutex_enter(&spa_namespace_lock); 1111c6065d0fSGeorge Wilson spa_config_sync(spa, B_FALSE, B_TRUE); 1112c6065d0fSGeorge Wilson mutex_exit(&spa_namespace_lock); 1113c6065d0fSGeorge Wilson } 1114c6065d0fSGeorge Wilson 1115e14bb325SJeff Bonwick return (error); 1116e14bb325SJeff Bonwick } 1117e14bb325SJeff Bonwick 1118fa9e4066Sahrens /* 1119fa9e4066Sahrens * ========================================================================== 1120fa9e4066Sahrens * Miscellaneous functions 1121fa9e4066Sahrens * ========================================================================== 1122fa9e4066Sahrens */ 1123fa9e4066Sahrens 1124ad135b5dSChristopher Siden void 1125ad135b5dSChristopher Siden spa_activate_mos_feature(spa_t *spa, const char *feature) 1126ad135b5dSChristopher Siden { 1127ad135b5dSChristopher Siden (void) nvlist_add_boolean(spa->spa_label_features, feature); 1128ad135b5dSChristopher Siden vdev_config_dirty(spa->spa_root_vdev); 1129ad135b5dSChristopher Siden } 1130ad135b5dSChristopher Siden 1131ad135b5dSChristopher Siden void 1132ad135b5dSChristopher Siden spa_deactivate_mos_feature(spa_t *spa, const char *feature) 1133ad135b5dSChristopher Siden { 1134ad135b5dSChristopher Siden (void) nvlist_remove_all(spa->spa_label_features, feature); 1135ad135b5dSChristopher Siden vdev_config_dirty(spa->spa_root_vdev); 1136ad135b5dSChristopher Siden } 1137ad135b5dSChristopher Siden 1138fa9e4066Sahrens /* 1139fa9e4066Sahrens * Rename a spa_t. 1140fa9e4066Sahrens */ 1141fa9e4066Sahrens int 1142fa9e4066Sahrens spa_rename(const char *name, const char *newname) 1143fa9e4066Sahrens { 1144fa9e4066Sahrens spa_t *spa; 1145fa9e4066Sahrens int err; 1146fa9e4066Sahrens 1147fa9e4066Sahrens /* 1148fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 1149fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 1150fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 1151ea8dc4b6Seschrock * allow recursive calls for other reasons. 1152fa9e4066Sahrens */ 1153fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 1154fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 1155fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1156fa9e4066Sahrens return (err); 1157fa9e4066Sahrens } 1158fa9e4066Sahrens 1159e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1160fa9e4066Sahrens 1161fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 1162e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); 1163fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 1164fa9e4066Sahrens 1165fa9e4066Sahrens /* 1166fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 1167fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 1168fa9e4066Sahrens * during the sync. 1169fa9e4066Sahrens */ 1170fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 1171fa9e4066Sahrens 1172e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1173fa9e4066Sahrens 11740373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 1175fa9e4066Sahrens 1176fa9e4066Sahrens /* 1177fa9e4066Sahrens * Sync the updated config cache. 1178fa9e4066Sahrens */ 1179c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 1180fa9e4066Sahrens 1181fa9e4066Sahrens spa_close(spa, FTAG); 1182fa9e4066Sahrens 1183fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1184fa9e4066Sahrens 1185fa9e4066Sahrens return (0); 1186fa9e4066Sahrens } 1187fa9e4066Sahrens 1188fa9e4066Sahrens /* 1189f9af39baSGeorge Wilson * Return the spa_t associated with given pool_guid, if it exists. If 1190f9af39baSGeorge Wilson * device_guid is non-zero, determine whether the pool exists *and* contains 1191f9af39baSGeorge Wilson * a device with the specified device_guid. 1192fa9e4066Sahrens */ 1193f9af39baSGeorge Wilson spa_t * 1194f9af39baSGeorge Wilson spa_by_guid(uint64_t pool_guid, uint64_t device_guid) 1195fa9e4066Sahrens { 1196fa9e4066Sahrens spa_t *spa; 1197fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 1198fa9e4066Sahrens 1199ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1200fa9e4066Sahrens 1201fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 1202fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 1203fa9e4066Sahrens continue; 1204fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 1205fa9e4066Sahrens continue; 120639c23413Seschrock if (spa_guid(spa) == pool_guid) { 120739c23413Seschrock if (device_guid == 0) 120839c23413Seschrock break; 120939c23413Seschrock 121039c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 121139c23413Seschrock device_guid) != NULL) 121239c23413Seschrock break; 121339c23413Seschrock 121439c23413Seschrock /* 12158654d025Sperrin * Check any devices we may be in the process of adding. 121639c23413Seschrock */ 121739c23413Seschrock if (spa->spa_pending_vdev) { 121839c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 121939c23413Seschrock device_guid) != NULL) 122039c23413Seschrock break; 122139c23413Seschrock } 122239c23413Seschrock } 1223fa9e4066Sahrens } 1224fa9e4066Sahrens 1225f9af39baSGeorge Wilson return (spa); 1226f9af39baSGeorge Wilson } 1227f9af39baSGeorge Wilson 1228f9af39baSGeorge Wilson /* 1229f9af39baSGeorge Wilson * Determine whether a pool with the given pool_guid exists. 1230f9af39baSGeorge Wilson */ 1231f9af39baSGeorge Wilson boolean_t 1232f9af39baSGeorge Wilson spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 1233f9af39baSGeorge Wilson { 1234f9af39baSGeorge Wilson return (spa_by_guid(pool_guid, device_guid) != NULL); 1235fa9e4066Sahrens } 1236fa9e4066Sahrens 1237fa9e4066Sahrens char * 1238fa9e4066Sahrens spa_strdup(const char *s) 1239fa9e4066Sahrens { 1240fa9e4066Sahrens size_t len; 1241fa9e4066Sahrens char *new; 1242fa9e4066Sahrens 1243fa9e4066Sahrens len = strlen(s); 1244fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 1245fa9e4066Sahrens bcopy(s, new, len); 1246fa9e4066Sahrens new[len] = '\0'; 1247fa9e4066Sahrens 1248fa9e4066Sahrens return (new); 1249fa9e4066Sahrens } 1250fa9e4066Sahrens 1251fa9e4066Sahrens void 1252fa9e4066Sahrens spa_strfree(char *s) 1253fa9e4066Sahrens { 1254fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 1255fa9e4066Sahrens } 1256fa9e4066Sahrens 1257fa9e4066Sahrens uint64_t 1258fa9e4066Sahrens spa_get_random(uint64_t range) 1259fa9e4066Sahrens { 1260fa9e4066Sahrens uint64_t r; 1261fa9e4066Sahrens 1262fa9e4066Sahrens ASSERT(range != 0); 1263fa9e4066Sahrens 1264fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 1265fa9e4066Sahrens 1266fa9e4066Sahrens return (r % range); 1267fa9e4066Sahrens } 1268fa9e4066Sahrens 12691195e687SMark J Musante uint64_t 12701195e687SMark J Musante spa_generate_guid(spa_t *spa) 12711195e687SMark J Musante { 12721195e687SMark J Musante uint64_t guid = spa_get_random(-1ULL); 12731195e687SMark J Musante 12741195e687SMark J Musante if (spa != NULL) { 12751195e687SMark J Musante while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) 12761195e687SMark J Musante guid = spa_get_random(-1ULL); 12771195e687SMark J Musante } else { 12781195e687SMark J Musante while (guid == 0 || spa_guid_exists(guid, 0)) 12791195e687SMark J Musante guid = spa_get_random(-1ULL); 12801195e687SMark J Musante } 12811195e687SMark J Musante 12821195e687SMark J Musante return (guid); 12831195e687SMark J Musante } 12841195e687SMark J Musante 1285fa9e4066Sahrens void 1286b24ab676SJeff Bonwick sprintf_blkptr(char *buf, const blkptr_t *bp) 1287fa9e4066Sahrens { 1288ad135b5dSChristopher Siden char type[256]; 1289f0ba89beSJeff Bonwick char *checksum = NULL; 1290f0ba89beSJeff Bonwick char *compress = NULL; 1291f0ba89beSJeff Bonwick 1292f0ba89beSJeff Bonwick if (bp != NULL) { 1293ad135b5dSChristopher Siden if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { 1294ad135b5dSChristopher Siden dmu_object_byteswap_t bswap = 1295ad135b5dSChristopher Siden DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); 1296ad135b5dSChristopher Siden (void) snprintf(type, sizeof (type), "bswap %s %s", 1297ad135b5dSChristopher Siden DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? 1298ad135b5dSChristopher Siden "metadata" : "data", 1299ad135b5dSChristopher Siden dmu_ot_byteswap[bswap].ob_name); 1300ad135b5dSChristopher Siden } else { 1301ad135b5dSChristopher Siden (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, 1302ad135b5dSChristopher Siden sizeof (type)); 1303ad135b5dSChristopher Siden } 1304f0ba89beSJeff Bonwick checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; 1305f0ba89beSJeff Bonwick compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; 1306f0ba89beSJeff Bonwick } 1307fa9e4066Sahrens 1308b24ab676SJeff Bonwick SPRINTF_BLKPTR(snprintf, ' ', buf, bp, type, checksum, compress); 1309fa9e4066Sahrens } 1310fa9e4066Sahrens 1311fa9e4066Sahrens void 1312fa9e4066Sahrens spa_freeze(spa_t *spa) 1313fa9e4066Sahrens { 1314fa9e4066Sahrens uint64_t freeze_txg = 0; 1315fa9e4066Sahrens 1316e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1317fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 1318fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 1319fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 1320fa9e4066Sahrens } 1321e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1322fa9e4066Sahrens if (freeze_txg != 0) 1323fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 1324fa9e4066Sahrens } 1325fa9e4066Sahrens 13260125049cSahrens void 13270125049cSahrens zfs_panic_recover(const char *fmt, ...) 13280125049cSahrens { 13290125049cSahrens va_list adx; 13300125049cSahrens 13310125049cSahrens va_start(adx, fmt); 13320125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 13330125049cSahrens va_end(adx); 13340125049cSahrens } 13350125049cSahrens 13363f9d6ad7SLin Ling /* 13373f9d6ad7SLin Ling * This is a stripped-down version of strtoull, suitable only for converting 13383f9d6ad7SLin Ling * lowercase hexidecimal numbers that don't overflow. 13393f9d6ad7SLin Ling */ 13403f9d6ad7SLin Ling uint64_t 13413f9d6ad7SLin Ling strtonum(const char *str, char **nptr) 13423f9d6ad7SLin Ling { 13433f9d6ad7SLin Ling uint64_t val = 0; 13443f9d6ad7SLin Ling char c; 13453f9d6ad7SLin Ling int digit; 13463f9d6ad7SLin Ling 13473f9d6ad7SLin Ling while ((c = *str) != '\0') { 13483f9d6ad7SLin Ling if (c >= '0' && c <= '9') 13493f9d6ad7SLin Ling digit = c - '0'; 13503f9d6ad7SLin Ling else if (c >= 'a' && c <= 'f') 13513f9d6ad7SLin Ling digit = 10 + c - 'a'; 13523f9d6ad7SLin Ling else 13533f9d6ad7SLin Ling break; 13543f9d6ad7SLin Ling 13553f9d6ad7SLin Ling val *= 16; 13563f9d6ad7SLin Ling val += digit; 13573f9d6ad7SLin Ling 13583f9d6ad7SLin Ling str++; 13593f9d6ad7SLin Ling } 13603f9d6ad7SLin Ling 13613f9d6ad7SLin Ling if (nptr) 13623f9d6ad7SLin Ling *nptr = (char *)str; 13633f9d6ad7SLin Ling 13643f9d6ad7SLin Ling return (val); 13653f9d6ad7SLin Ling } 13663f9d6ad7SLin Ling 1367fa9e4066Sahrens /* 1368fa9e4066Sahrens * ========================================================================== 1369fa9e4066Sahrens * Accessor functions 1370fa9e4066Sahrens * ========================================================================== 1371fa9e4066Sahrens */ 1372fa9e4066Sahrens 1373088f3894Sahrens boolean_t 137488b7b0f2SMatthew Ahrens spa_shutting_down(spa_t *spa) 1375fa9e4066Sahrens { 137688b7b0f2SMatthew Ahrens return (spa->spa_async_suspended); 1377fa9e4066Sahrens } 1378fa9e4066Sahrens 1379fa9e4066Sahrens dsl_pool_t * 1380fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1381fa9e4066Sahrens { 1382fa9e4066Sahrens return (spa->spa_dsl_pool); 1383fa9e4066Sahrens } 1384fa9e4066Sahrens 1385ad135b5dSChristopher Siden boolean_t 1386ad135b5dSChristopher Siden spa_is_initializing(spa_t *spa) 1387ad135b5dSChristopher Siden { 1388ad135b5dSChristopher Siden return (spa->spa_is_initializing); 1389ad135b5dSChristopher Siden } 1390ad135b5dSChristopher Siden 1391fa9e4066Sahrens blkptr_t * 1392fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1393fa9e4066Sahrens { 1394fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1395fa9e4066Sahrens } 1396fa9e4066Sahrens 1397fa9e4066Sahrens void 1398fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1399fa9e4066Sahrens { 1400fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1401fa9e4066Sahrens } 1402fa9e4066Sahrens 1403fa9e4066Sahrens void 1404fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1405fa9e4066Sahrens { 1406fa9e4066Sahrens if (spa->spa_root == NULL) 1407fa9e4066Sahrens buf[0] = '\0'; 1408fa9e4066Sahrens else 1409fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1410fa9e4066Sahrens } 1411fa9e4066Sahrens 1412fa9e4066Sahrens int 1413fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1414fa9e4066Sahrens { 1415fa9e4066Sahrens return (spa->spa_sync_pass); 1416fa9e4066Sahrens } 1417fa9e4066Sahrens 1418fa9e4066Sahrens char * 1419fa9e4066Sahrens spa_name(spa_t *spa) 1420fa9e4066Sahrens { 1421fa9e4066Sahrens return (spa->spa_name); 1422fa9e4066Sahrens } 1423fa9e4066Sahrens 1424fa9e4066Sahrens uint64_t 1425fa9e4066Sahrens spa_guid(spa_t *spa) 1426fa9e4066Sahrens { 1427dfbb9432SGeorge Wilson dsl_pool_t *dp = spa_get_dsl(spa); 1428dfbb9432SGeorge Wilson uint64_t guid; 1429dfbb9432SGeorge Wilson 1430b5989ec7Seschrock /* 1431b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1432b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1433e9103aaeSGarrett D'Amore * vdev. We stash the original pool guid in 'spa_config_guid' to handle 1434b5989ec7Seschrock * this case. 1435b5989ec7Seschrock */ 1436dfbb9432SGeorge Wilson if (spa->spa_root_vdev == NULL) 1437dfbb9432SGeorge Wilson return (spa->spa_config_guid); 1438dfbb9432SGeorge Wilson 1439dfbb9432SGeorge Wilson guid = spa->spa_last_synced_guid != 0 ? 1440dfbb9432SGeorge Wilson spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; 1441dfbb9432SGeorge Wilson 1442dfbb9432SGeorge Wilson /* 1443dfbb9432SGeorge Wilson * Return the most recently synced out guid unless we're 1444dfbb9432SGeorge Wilson * in syncing context. 1445dfbb9432SGeorge Wilson */ 1446dfbb9432SGeorge Wilson if (dp && dsl_pool_sync_context(dp)) 1447b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1448b5989ec7Seschrock else 1449dfbb9432SGeorge Wilson return (guid); 1450e9103aaeSGarrett D'Amore } 1451e9103aaeSGarrett D'Amore 1452e9103aaeSGarrett D'Amore uint64_t 1453e9103aaeSGarrett D'Amore spa_load_guid(spa_t *spa) 1454e9103aaeSGarrett D'Amore { 1455e9103aaeSGarrett D'Amore /* 1456e9103aaeSGarrett D'Amore * This is a GUID that exists solely as a reference for the 1457e9103aaeSGarrett D'Amore * purposes of the arc. It is generated at load time, and 1458e9103aaeSGarrett D'Amore * is never written to persistent storage. 1459e9103aaeSGarrett D'Amore */ 1460e9103aaeSGarrett D'Amore return (spa->spa_load_guid); 1461fa9e4066Sahrens } 1462fa9e4066Sahrens 1463fa9e4066Sahrens uint64_t 1464fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1465fa9e4066Sahrens { 1466fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1467fa9e4066Sahrens } 1468fa9e4066Sahrens 1469fa9e4066Sahrens uint64_t 1470fa9e4066Sahrens spa_first_txg(spa_t *spa) 1471fa9e4066Sahrens { 1472fa9e4066Sahrens return (spa->spa_first_txg); 1473fa9e4066Sahrens } 1474fa9e4066Sahrens 1475b24ab676SJeff Bonwick uint64_t 1476b24ab676SJeff Bonwick spa_syncing_txg(spa_t *spa) 1477b24ab676SJeff Bonwick { 1478b24ab676SJeff Bonwick return (spa->spa_syncing_txg); 1479b24ab676SJeff Bonwick } 1480b24ab676SJeff Bonwick 148188b7b0f2SMatthew Ahrens pool_state_t 1482fa9e4066Sahrens spa_state(spa_t *spa) 1483fa9e4066Sahrens { 1484fa9e4066Sahrens return (spa->spa_state); 1485fa9e4066Sahrens } 1486fa9e4066Sahrens 1487b16da2e2SGeorge Wilson spa_load_state_t 1488b16da2e2SGeorge Wilson spa_load_state(spa_t *spa) 1489b16da2e2SGeorge Wilson { 1490b16da2e2SGeorge Wilson return (spa->spa_load_state); 1491b16da2e2SGeorge Wilson } 1492b16da2e2SGeorge Wilson 1493fa9e4066Sahrens uint64_t 1494fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1495fa9e4066Sahrens { 1496fa9e4066Sahrens return (spa->spa_freeze_txg); 1497fa9e4066Sahrens } 1498fa9e4066Sahrens 1499fa9e4066Sahrens /* ARGSUSED */ 1500fa9e4066Sahrens uint64_t 1501fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 1502fa9e4066Sahrens { 1503fa9e4066Sahrens /* 1504b24ab676SJeff Bonwick * The worst case is single-sector max-parity RAID-Z blocks, in which 1505b24ab676SJeff Bonwick * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) 1506b24ab676SJeff Bonwick * times the size; so just assume that. Add to this the fact that 1507b24ab676SJeff Bonwick * we can have up to 3 DVAs per bp, and one more factor of 2 because 1508b24ab676SJeff Bonwick * the block may be dittoed with up to 3 DVAs by ddt_sync(). 150944cd46caSbillm */ 1510b24ab676SJeff Bonwick return (lsize * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2); 151144cd46caSbillm } 151244cd46caSbillm 1513485bbbf5SGeorge Wilson uint64_t 1514485bbbf5SGeorge Wilson spa_get_dspace(spa_t *spa) 1515485bbbf5SGeorge Wilson { 1516485bbbf5SGeorge Wilson return (spa->spa_dspace); 1517485bbbf5SGeorge Wilson } 1518485bbbf5SGeorge Wilson 1519485bbbf5SGeorge Wilson void 1520485bbbf5SGeorge Wilson spa_update_dspace(spa_t *spa) 1521485bbbf5SGeorge Wilson { 1522485bbbf5SGeorge Wilson spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + 1523485bbbf5SGeorge Wilson ddt_get_dedup_dspace(spa); 1524485bbbf5SGeorge Wilson } 1525485bbbf5SGeorge Wilson 15260a4e9518Sgw /* 15270a4e9518Sgw * Return the failure mode that has been set to this pool. The default 15280a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 15290a4e9518Sgw */ 15300a4e9518Sgw uint8_t 15310a4e9518Sgw spa_get_failmode(spa_t *spa) 15320a4e9518Sgw { 15330a4e9518Sgw return (spa->spa_failmode); 15340a4e9518Sgw } 15350a4e9518Sgw 1536e14bb325SJeff Bonwick boolean_t 1537e14bb325SJeff Bonwick spa_suspended(spa_t *spa) 1538e14bb325SJeff Bonwick { 1539e14bb325SJeff Bonwick return (spa->spa_suspended); 1540e14bb325SJeff Bonwick } 1541e14bb325SJeff Bonwick 154244cd46caSbillm uint64_t 154344cd46caSbillm spa_version(spa_t *spa) 154444cd46caSbillm { 154544cd46caSbillm return (spa->spa_ubsync.ub_version); 154644cd46caSbillm } 154744cd46caSbillm 1548b24ab676SJeff Bonwick boolean_t 1549b24ab676SJeff Bonwick spa_deflate(spa_t *spa) 1550b24ab676SJeff Bonwick { 1551b24ab676SJeff Bonwick return (spa->spa_deflate); 1552b24ab676SJeff Bonwick } 1553b24ab676SJeff Bonwick 1554b24ab676SJeff Bonwick metaslab_class_t * 1555b24ab676SJeff Bonwick spa_normal_class(spa_t *spa) 1556b24ab676SJeff Bonwick { 1557b24ab676SJeff Bonwick return (spa->spa_normal_class); 1558b24ab676SJeff Bonwick } 1559b24ab676SJeff Bonwick 1560b24ab676SJeff Bonwick metaslab_class_t * 1561b24ab676SJeff Bonwick spa_log_class(spa_t *spa) 1562b24ab676SJeff Bonwick { 1563b24ab676SJeff Bonwick return (spa->spa_log_class); 1564b24ab676SJeff Bonwick } 1565b24ab676SJeff Bonwick 156644cd46caSbillm int 156744cd46caSbillm spa_max_replication(spa_t *spa) 156844cd46caSbillm { 156944cd46caSbillm /* 1570e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 157144cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 157244cd46caSbillm * replication level accordingly. 1573fa9e4066Sahrens */ 1574e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 157544cd46caSbillm return (1); 157644cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1577fa9e4066Sahrens } 1578fa9e4066Sahrens 15793f9d6ad7SLin Ling int 15803f9d6ad7SLin Ling spa_prev_software_version(spa_t *spa) 15813f9d6ad7SLin Ling { 15823f9d6ad7SLin Ling return (spa->spa_prev_software_version); 15833f9d6ad7SLin Ling } 15843f9d6ad7SLin Ling 1585283b8460SGeorge.Wilson uint64_t 1586283b8460SGeorge.Wilson spa_deadman_synctime(spa_t *spa) 1587283b8460SGeorge.Wilson { 1588283b8460SGeorge.Wilson return (spa->spa_deadman_synctime); 1589283b8460SGeorge.Wilson } 1590283b8460SGeorge.Wilson 159199653d4eSeschrock uint64_t 1592b24ab676SJeff Bonwick dva_get_dsize_sync(spa_t *spa, const dva_t *dva) 159399653d4eSeschrock { 1594b24ab676SJeff Bonwick uint64_t asize = DVA_GET_ASIZE(dva); 1595b24ab676SJeff Bonwick uint64_t dsize = asize; 159699653d4eSeschrock 1597b24ab676SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); 159899653d4eSeschrock 1599b24ab676SJeff Bonwick if (asize != 0 && spa->spa_deflate) { 1600b24ab676SJeff Bonwick vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); 1601b24ab676SJeff Bonwick dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 160299653d4eSeschrock } 1603b24ab676SJeff Bonwick 1604b24ab676SJeff Bonwick return (dsize); 1605b24ab676SJeff Bonwick } 1606b24ab676SJeff Bonwick 1607b24ab676SJeff Bonwick uint64_t 1608b24ab676SJeff Bonwick bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) 1609b24ab676SJeff Bonwick { 1610b24ab676SJeff Bonwick uint64_t dsize = 0; 1611b24ab676SJeff Bonwick 1612b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1613b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1614b24ab676SJeff Bonwick 1615b24ab676SJeff Bonwick return (dsize); 1616b24ab676SJeff Bonwick } 1617b24ab676SJeff Bonwick 1618b24ab676SJeff Bonwick uint64_t 1619b24ab676SJeff Bonwick bp_get_dsize(spa_t *spa, const blkptr_t *bp) 1620b24ab676SJeff Bonwick { 1621b24ab676SJeff Bonwick uint64_t dsize = 0; 1622b24ab676SJeff Bonwick 1623b24ab676SJeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 1624b24ab676SJeff Bonwick 1625b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1626b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1627b24ab676SJeff Bonwick 1628e14bb325SJeff Bonwick spa_config_exit(spa, SCL_VDEV, FTAG); 1629b24ab676SJeff Bonwick 1630b24ab676SJeff Bonwick return (dsize); 163199653d4eSeschrock } 163299653d4eSeschrock 1633fa9e4066Sahrens /* 1634fa9e4066Sahrens * ========================================================================== 1635fa9e4066Sahrens * Initialization and Termination 1636fa9e4066Sahrens * ========================================================================== 1637fa9e4066Sahrens */ 1638fa9e4066Sahrens 1639fa9e4066Sahrens static int 1640fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1641fa9e4066Sahrens { 1642fa9e4066Sahrens const spa_t *s1 = a1; 1643fa9e4066Sahrens const spa_t *s2 = a2; 1644fa9e4066Sahrens int s; 1645fa9e4066Sahrens 1646fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1647fa9e4066Sahrens if (s > 0) 1648fa9e4066Sahrens return (1); 1649fa9e4066Sahrens if (s < 0) 1650fa9e4066Sahrens return (-1); 1651fa9e4066Sahrens return (0); 1652fa9e4066Sahrens } 1653fa9e4066Sahrens 16540373e76bSbonwick int 16550373e76bSbonwick spa_busy(void) 16560373e76bSbonwick { 16570373e76bSbonwick return (spa_active_count); 16580373e76bSbonwick } 16590373e76bSbonwick 1660e7cbe64fSgw void 1661e7cbe64fSgw spa_boot_init() 1662e7cbe64fSgw { 1663e7cbe64fSgw spa_config_load(); 1664e7cbe64fSgw } 1665e7cbe64fSgw 1666fa9e4066Sahrens void 1667fa9e4066Sahrens spa_init(int mode) 1668fa9e4066Sahrens { 1669fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1670c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1671fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 1672fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1673fa9e4066Sahrens 1674fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1675fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1676fa9e4066Sahrens 1677fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 1678fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 1679fa94a07fSbrendan 1680fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 1681fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 168299653d4eSeschrock 16838ad4d6ddSJeff Bonwick spa_mode_global = mode; 1684fa9e4066Sahrens 1685283b8460SGeorge.Wilson #ifdef _KERNEL 1686283b8460SGeorge.Wilson spa_arch_init(); 1687283b8460SGeorge.Wilson #else 1688cd1c8b85SMatthew Ahrens if (spa_mode_global != FREAD && dprintf_find_string("watch")) { 1689cd1c8b85SMatthew Ahrens arc_procfd = open("/proc/self/ctl", O_WRONLY); 1690cd1c8b85SMatthew Ahrens if (arc_procfd == -1) { 1691cd1c8b85SMatthew Ahrens perror("could not enable watchpoints: " 1692cd1c8b85SMatthew Ahrens "opening /proc/self/ctl failed: "); 1693cd1c8b85SMatthew Ahrens } else { 1694cd1c8b85SMatthew Ahrens arc_watch = B_TRUE; 1695cd1c8b85SMatthew Ahrens } 1696cd1c8b85SMatthew Ahrens } 1697cd1c8b85SMatthew Ahrens #endif 1698cd1c8b85SMatthew Ahrens 1699fa9e4066Sahrens refcount_init(); 1700fa9e4066Sahrens unique_init(); 170101f55e48SGeorge Wilson space_map_init(); 1702fa9e4066Sahrens zio_init(); 1703fa9e4066Sahrens dmu_init(); 1704fa9e4066Sahrens zil_init(); 170587db74c1Sek vdev_cache_stat_init(); 170691ebeef5Sahrens zfs_prop_init(); 1707990b4856Slling zpool_prop_init(); 1708ad135b5dSChristopher Siden zpool_feature_init(); 1709fa9e4066Sahrens spa_config_load(); 1710e14bb325SJeff Bonwick l2arc_start(); 1711fa9e4066Sahrens } 1712fa9e4066Sahrens 1713fa9e4066Sahrens void 1714fa9e4066Sahrens spa_fini(void) 1715fa9e4066Sahrens { 1716e14bb325SJeff Bonwick l2arc_stop(); 1717e14bb325SJeff Bonwick 1718fa9e4066Sahrens spa_evict_all(); 1719fa9e4066Sahrens 172087db74c1Sek vdev_cache_stat_fini(); 1721fa9e4066Sahrens zil_fini(); 1722fa9e4066Sahrens dmu_fini(); 1723fa9e4066Sahrens zio_fini(); 172401f55e48SGeorge Wilson space_map_fini(); 172591ebeef5Sahrens unique_fini(); 1726fa9e4066Sahrens refcount_fini(); 1727fa9e4066Sahrens 1728fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 172999653d4eSeschrock avl_destroy(&spa_spare_avl); 1730fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 1731fa9e4066Sahrens 1732fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1733fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1734c25056deSgw mutex_destroy(&spa_spare_lock); 1735fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 1736fa9e4066Sahrens } 17376ce0521aSperrin 17386ce0521aSperrin /* 17396ce0521aSperrin * Return whether this pool has slogs. No locking needed. 17406ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 17416ce0521aSperrin * performance and not correctness 17426ce0521aSperrin */ 17436ce0521aSperrin boolean_t 17446ce0521aSperrin spa_has_slogs(spa_t *spa) 17456ce0521aSperrin { 17466ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 17476ce0521aSperrin } 1748bf82a41bSeschrock 1749b24ab676SJeff Bonwick spa_log_state_t 1750b24ab676SJeff Bonwick spa_get_log_state(spa_t *spa) 1751b24ab676SJeff Bonwick { 1752b24ab676SJeff Bonwick return (spa->spa_log_state); 1753b24ab676SJeff Bonwick } 1754b24ab676SJeff Bonwick 1755b24ab676SJeff Bonwick void 1756b24ab676SJeff Bonwick spa_set_log_state(spa_t *spa, spa_log_state_t state) 1757b24ab676SJeff Bonwick { 1758b24ab676SJeff Bonwick spa->spa_log_state = state; 1759b24ab676SJeff Bonwick } 1760b24ab676SJeff Bonwick 1761bf82a41bSeschrock boolean_t 1762bf82a41bSeschrock spa_is_root(spa_t *spa) 1763bf82a41bSeschrock { 1764bf82a41bSeschrock return (spa->spa_is_root); 1765bf82a41bSeschrock } 17668ad4d6ddSJeff Bonwick 17678ad4d6ddSJeff Bonwick boolean_t 17688ad4d6ddSJeff Bonwick spa_writeable(spa_t *spa) 17698ad4d6ddSJeff Bonwick { 17708ad4d6ddSJeff Bonwick return (!!(spa->spa_mode & FWRITE)); 17718ad4d6ddSJeff Bonwick } 17728ad4d6ddSJeff Bonwick 17738ad4d6ddSJeff Bonwick int 17748ad4d6ddSJeff Bonwick spa_mode(spa_t *spa) 17758ad4d6ddSJeff Bonwick { 17768ad4d6ddSJeff Bonwick return (spa->spa_mode); 17778ad4d6ddSJeff Bonwick } 1778b24ab676SJeff Bonwick 1779b24ab676SJeff Bonwick uint64_t 1780b24ab676SJeff Bonwick spa_bootfs(spa_t *spa) 1781b24ab676SJeff Bonwick { 1782b24ab676SJeff Bonwick return (spa->spa_bootfs); 1783b24ab676SJeff Bonwick } 1784b24ab676SJeff Bonwick 1785b24ab676SJeff Bonwick uint64_t 1786b24ab676SJeff Bonwick spa_delegation(spa_t *spa) 1787b24ab676SJeff Bonwick { 1788b24ab676SJeff Bonwick return (spa->spa_delegation); 1789b24ab676SJeff Bonwick } 1790b24ab676SJeff Bonwick 1791b24ab676SJeff Bonwick objset_t * 1792b24ab676SJeff Bonwick spa_meta_objset(spa_t *spa) 1793b24ab676SJeff Bonwick { 1794b24ab676SJeff Bonwick return (spa->spa_meta_objset); 1795b24ab676SJeff Bonwick } 1796b24ab676SJeff Bonwick 1797b24ab676SJeff Bonwick enum zio_checksum 1798b24ab676SJeff Bonwick spa_dedup_checksum(spa_t *spa) 1799b24ab676SJeff Bonwick { 1800b24ab676SJeff Bonwick return (spa->spa_dedup_checksum); 1801b24ab676SJeff Bonwick } 18023f9d6ad7SLin Ling 18033f9d6ad7SLin Ling /* 18043f9d6ad7SLin Ling * Reset pool scan stat per scan pass (or reboot). 18053f9d6ad7SLin Ling */ 18063f9d6ad7SLin Ling void 18073f9d6ad7SLin Ling spa_scan_stat_init(spa_t *spa) 18083f9d6ad7SLin Ling { 18093f9d6ad7SLin Ling /* data not stored on disk */ 18103f9d6ad7SLin Ling spa->spa_scan_pass_start = gethrestime_sec(); 18113f9d6ad7SLin Ling spa->spa_scan_pass_exam = 0; 18123f9d6ad7SLin Ling vdev_scan_stat_init(spa->spa_root_vdev); 18133f9d6ad7SLin Ling } 18143f9d6ad7SLin Ling 18153f9d6ad7SLin Ling /* 18163f9d6ad7SLin Ling * Get scan stats for zpool status reports 18173f9d6ad7SLin Ling */ 18183f9d6ad7SLin Ling int 18193f9d6ad7SLin Ling spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) 18203f9d6ad7SLin Ling { 18213f9d6ad7SLin Ling dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; 18223f9d6ad7SLin Ling 18233f9d6ad7SLin Ling if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) 18243f9d6ad7SLin Ling return (ENOENT); 18253f9d6ad7SLin Ling bzero(ps, sizeof (pool_scan_stat_t)); 18263f9d6ad7SLin Ling 18273f9d6ad7SLin Ling /* data stored on disk */ 18283f9d6ad7SLin Ling ps->pss_func = scn->scn_phys.scn_func; 18293f9d6ad7SLin Ling ps->pss_start_time = scn->scn_phys.scn_start_time; 18303f9d6ad7SLin Ling ps->pss_end_time = scn->scn_phys.scn_end_time; 18313f9d6ad7SLin Ling ps->pss_to_examine = scn->scn_phys.scn_to_examine; 18323f9d6ad7SLin Ling ps->pss_examined = scn->scn_phys.scn_examined; 18333f9d6ad7SLin Ling ps->pss_to_process = scn->scn_phys.scn_to_process; 18343f9d6ad7SLin Ling ps->pss_processed = scn->scn_phys.scn_processed; 18353f9d6ad7SLin Ling ps->pss_errors = scn->scn_phys.scn_errors; 18363f9d6ad7SLin Ling ps->pss_state = scn->scn_phys.scn_state; 18373f9d6ad7SLin Ling 18383f9d6ad7SLin Ling /* data not stored on disk */ 18393f9d6ad7SLin Ling ps->pss_pass_start = spa->spa_scan_pass_start; 18403f9d6ad7SLin Ling ps->pss_pass_exam = spa->spa_scan_pass_exam; 18413f9d6ad7SLin Ling 18423f9d6ad7SLin Ling return (0); 18433f9d6ad7SLin Ling } 184409c9d376SGeorge Wilson 184509c9d376SGeorge Wilson boolean_t 184609c9d376SGeorge Wilson spa_debug_enabled(spa_t *spa) 184709c9d376SGeorge Wilson { 184809c9d376SGeorge Wilson return (spa->spa_debug); 184909c9d376SGeorge Wilson } 1850