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> 29*283b8460SGeorge.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 253*283b8460SGeorge.Wilson extern int zfs_txg_synctime_ms; 254*283b8460SGeorge.Wilson 255*283b8460SGeorge.Wilson /* 256*283b8460SGeorge.Wilson * Expiration time in units of zfs_txg_synctime_ms. This value has two 257*283b8460SGeorge.Wilson * meanings. First it is used to determine when the spa_deadman logic 258*283b8460SGeorge.Wilson * should fire. By default the spa_deadman will fire if spa_sync has 259*283b8460SGeorge.Wilson * not completed in 1000 * zfs_txg_synctime_ms (i.e. 1000 seconds). 260*283b8460SGeorge.Wilson * Secondly, the value determines if an I/O is considered "hung". 261*283b8460SGeorge.Wilson * Any I/O that has not completed in zfs_deadman_synctime is considered 262*283b8460SGeorge.Wilson * "hung" resulting in a system panic. 263*283b8460SGeorge.Wilson * 1000 zfs_txg_synctime_ms (i.e. 1000 seconds). 264*283b8460SGeorge.Wilson */ 265*283b8460SGeorge.Wilson uint64_t zfs_deadman_synctime = 1000ULL; 266*283b8460SGeorge.Wilson 267*283b8460SGeorge.Wilson /* 268*283b8460SGeorge.Wilson * Override the zfs deadman behavior via /etc/system. By default the 269*283b8460SGeorge.Wilson * deadman is enabled except on VMware and sparc deployments. 270*283b8460SGeorge.Wilson */ 271*283b8460SGeorge.Wilson int zfs_deadman_enabled = -1; 272*283b8460SGeorge.Wilson 273fa9e4066Sahrens 274e05725b1Sbonwick /* 275e05725b1Sbonwick * ========================================================================== 276e05725b1Sbonwick * SPA config locking 277e05725b1Sbonwick * ========================================================================== 278e05725b1Sbonwick */ 279e05725b1Sbonwick static void 280e14bb325SJeff Bonwick spa_config_lock_init(spa_t *spa) 281e14bb325SJeff Bonwick { 282e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 283e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 284e14bb325SJeff Bonwick mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); 285e14bb325SJeff Bonwick cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); 286e14bb325SJeff Bonwick refcount_create(&scl->scl_count); 287e14bb325SJeff Bonwick scl->scl_writer = NULL; 288e14bb325SJeff Bonwick scl->scl_write_wanted = 0; 289e14bb325SJeff Bonwick } 290e05725b1Sbonwick } 291e05725b1Sbonwick 292e05725b1Sbonwick static void 293e14bb325SJeff Bonwick spa_config_lock_destroy(spa_t *spa) 294e14bb325SJeff Bonwick { 295e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 296e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 297e14bb325SJeff Bonwick mutex_destroy(&scl->scl_lock); 298e14bb325SJeff Bonwick cv_destroy(&scl->scl_cv); 299e14bb325SJeff Bonwick refcount_destroy(&scl->scl_count); 300e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL); 301e14bb325SJeff Bonwick ASSERT(scl->scl_write_wanted == 0); 302e14bb325SJeff Bonwick } 303e14bb325SJeff Bonwick } 304e14bb325SJeff Bonwick 305e14bb325SJeff Bonwick int 306e14bb325SJeff Bonwick spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) 307e05725b1Sbonwick { 308e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 309e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 310e14bb325SJeff Bonwick if (!(locks & (1 << i))) 311e14bb325SJeff Bonwick continue; 312e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 313e14bb325SJeff Bonwick if (rw == RW_READER) { 314e14bb325SJeff Bonwick if (scl->scl_writer || scl->scl_write_wanted) { 315e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 316e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 317e14bb325SJeff Bonwick return (0); 318e14bb325SJeff Bonwick } 319e14bb325SJeff Bonwick } else { 320e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 321e14bb325SJeff Bonwick if (!refcount_is_zero(&scl->scl_count)) { 322e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 323e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 324e14bb325SJeff Bonwick return (0); 325e14bb325SJeff Bonwick } 326e14bb325SJeff Bonwick scl->scl_writer = curthread; 327e14bb325SJeff Bonwick } 328e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 329e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 330e14bb325SJeff Bonwick } 331e14bb325SJeff Bonwick return (1); 332e05725b1Sbonwick } 333e05725b1Sbonwick 334e05725b1Sbonwick void 335e14bb325SJeff Bonwick spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) 336e05725b1Sbonwick { 337f64c0e34SEric Taylor int wlocks_held = 0; 338f64c0e34SEric Taylor 339e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 340e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 341f64c0e34SEric Taylor if (scl->scl_writer == curthread) 342f64c0e34SEric Taylor wlocks_held |= (1 << i); 343e14bb325SJeff Bonwick if (!(locks & (1 << i))) 344e14bb325SJeff Bonwick continue; 345e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 346e14bb325SJeff Bonwick if (rw == RW_READER) { 347e14bb325SJeff Bonwick while (scl->scl_writer || scl->scl_write_wanted) { 348e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 349e14bb325SJeff Bonwick } 350e14bb325SJeff Bonwick } else { 351e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 352e14bb325SJeff Bonwick while (!refcount_is_zero(&scl->scl_count)) { 353e14bb325SJeff Bonwick scl->scl_write_wanted++; 354e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 355e14bb325SJeff Bonwick scl->scl_write_wanted--; 356e14bb325SJeff Bonwick } 357e14bb325SJeff Bonwick scl->scl_writer = curthread; 358e14bb325SJeff Bonwick } 359e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 360e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 361e05725b1Sbonwick } 362f64c0e34SEric Taylor ASSERT(wlocks_held <= locks); 363e05725b1Sbonwick } 364e05725b1Sbonwick 365e05725b1Sbonwick void 366e14bb325SJeff Bonwick spa_config_exit(spa_t *spa, int locks, void *tag) 367e05725b1Sbonwick { 368e14bb325SJeff Bonwick for (int i = SCL_LOCKS - 1; i >= 0; i--) { 369e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 370e14bb325SJeff Bonwick if (!(locks & (1 << i))) 371e14bb325SJeff Bonwick continue; 372e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 373e14bb325SJeff Bonwick ASSERT(!refcount_is_zero(&scl->scl_count)); 374e14bb325SJeff Bonwick if (refcount_remove(&scl->scl_count, tag) == 0) { 375e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL || 376e14bb325SJeff Bonwick scl->scl_writer == curthread); 377e14bb325SJeff Bonwick scl->scl_writer = NULL; /* OK in either case */ 378e14bb325SJeff Bonwick cv_broadcast(&scl->scl_cv); 379e14bb325SJeff Bonwick } 380e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 381e05725b1Sbonwick } 382e05725b1Sbonwick } 383e05725b1Sbonwick 384e14bb325SJeff Bonwick int 385e14bb325SJeff Bonwick spa_config_held(spa_t *spa, int locks, krw_t rw) 386e05725b1Sbonwick { 387e14bb325SJeff Bonwick int locks_held = 0; 388e05725b1Sbonwick 389e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 390e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 391e14bb325SJeff Bonwick if (!(locks & (1 << i))) 392e14bb325SJeff Bonwick continue; 393e14bb325SJeff Bonwick if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) || 394e14bb325SJeff Bonwick (rw == RW_WRITER && scl->scl_writer == curthread)) 395e14bb325SJeff Bonwick locks_held |= 1 << i; 396e14bb325SJeff Bonwick } 397e14bb325SJeff Bonwick 398e14bb325SJeff Bonwick return (locks_held); 399e05725b1Sbonwick } 400e05725b1Sbonwick 401fa9e4066Sahrens /* 402fa9e4066Sahrens * ========================================================================== 403fa9e4066Sahrens * SPA namespace functions 404fa9e4066Sahrens * ========================================================================== 405fa9e4066Sahrens */ 406fa9e4066Sahrens 407fa9e4066Sahrens /* 408fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 409fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 410fa9e4066Sahrens */ 411fa9e4066Sahrens spa_t * 412fa9e4066Sahrens spa_lookup(const char *name) 413fa9e4066Sahrens { 414e14bb325SJeff Bonwick static spa_t search; /* spa_t is large; don't allocate on stack */ 415e14bb325SJeff Bonwick spa_t *spa; 416fa9e4066Sahrens avl_index_t where; 41740feaa91Sahrens char c; 41840feaa91Sahrens char *cp; 419fa9e4066Sahrens 420fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 421fa9e4066Sahrens 42240feaa91Sahrens /* 42340feaa91Sahrens * If it's a full dataset name, figure out the pool name and 42440feaa91Sahrens * just use that. 42540feaa91Sahrens */ 42640feaa91Sahrens cp = strpbrk(name, "/@"); 42740feaa91Sahrens if (cp) { 42840feaa91Sahrens c = *cp; 42940feaa91Sahrens *cp = '\0'; 43040feaa91Sahrens } 43140feaa91Sahrens 432e14bb325SJeff Bonwick (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); 433fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 434fa9e4066Sahrens 43540feaa91Sahrens if (cp) 43640feaa91Sahrens *cp = c; 43740feaa91Sahrens 438fa9e4066Sahrens return (spa); 439fa9e4066Sahrens } 440fa9e4066Sahrens 441*283b8460SGeorge.Wilson /* 442*283b8460SGeorge.Wilson * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. 443*283b8460SGeorge.Wilson * If the zfs_deadman_enabled flag is set then it inspects all vdev queues 444*283b8460SGeorge.Wilson * looking for potentially hung I/Os. 445*283b8460SGeorge.Wilson */ 446*283b8460SGeorge.Wilson void 447*283b8460SGeorge.Wilson spa_deadman(void *arg) 448*283b8460SGeorge.Wilson { 449*283b8460SGeorge.Wilson spa_t *spa = arg; 450*283b8460SGeorge.Wilson 451*283b8460SGeorge.Wilson zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", 452*283b8460SGeorge.Wilson (gethrtime() - spa->spa_sync_starttime) / NANOSEC, 453*283b8460SGeorge.Wilson ++spa->spa_deadman_calls); 454*283b8460SGeorge.Wilson if (zfs_deadman_enabled) 455*283b8460SGeorge.Wilson vdev_deadman(spa->spa_root_vdev); 456*283b8460SGeorge.Wilson } 457*283b8460SGeorge.Wilson 458fa9e4066Sahrens /* 459fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 460fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 461fa9e4066Sahrens * exist by calling spa_lookup() first. 462fa9e4066Sahrens */ 463fa9e4066Sahrens spa_t * 464468c413aSTim Haley spa_add(const char *name, nvlist_t *config, const char *altroot) 465fa9e4066Sahrens { 466fa9e4066Sahrens spa_t *spa; 467c5904d13Seschrock spa_config_dirent_t *dp; 468*283b8460SGeorge.Wilson cyc_handler_t hdlr; 469*283b8460SGeorge.Wilson cyc_time_t when; 470fa9e4066Sahrens 471fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 472fa9e4066Sahrens 473fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 474fa9e4066Sahrens 475c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 476c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 47735a5a358SJonathan Adams mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 478c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 47935a5a358SJonathan Adams mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); 480c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 48135a5a358SJonathan Adams mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 482a1521560SJeff Bonwick mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); 483a1521560SJeff Bonwick mutex_init(&spa->spa_vdev_top_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 501*283b8460SGeorge.Wilson hdlr.cyh_func = spa_deadman; 502*283b8460SGeorge.Wilson hdlr.cyh_arg = spa; 503*283b8460SGeorge.Wilson hdlr.cyh_level = CY_LOW_LEVEL; 504*283b8460SGeorge.Wilson 505*283b8460SGeorge.Wilson spa->spa_deadman_synctime = zfs_deadman_synctime * 506*283b8460SGeorge.Wilson zfs_txg_synctime_ms * MICROSEC; 507*283b8460SGeorge.Wilson 508*283b8460SGeorge.Wilson /* 509*283b8460SGeorge.Wilson * This determines how often we need to check for hung I/Os after 510*283b8460SGeorge.Wilson * the cyclic has already fired. Since checking for hung I/Os is 511*283b8460SGeorge.Wilson * an expensive operation we don't want to check too frequently. 512*283b8460SGeorge.Wilson * Instead wait for 5 synctimes before checking again. 513*283b8460SGeorge.Wilson */ 514*283b8460SGeorge.Wilson when.cyt_interval = 5ULL * zfs_txg_synctime_ms * MICROSEC; 515*283b8460SGeorge.Wilson when.cyt_when = CY_INFINITY; 516*283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 517*283b8460SGeorge.Wilson spa->spa_deadman_cycid = cyclic_add(&hdlr, &when); 518*283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 519*283b8460SGeorge.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 563fa9e4066Sahrens return (spa); 564fa9e4066Sahrens } 565fa9e4066Sahrens 566fa9e4066Sahrens /* 567fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 568fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 569fa9e4066Sahrens * deactivated. 570fa9e4066Sahrens */ 571fa9e4066Sahrens void 572fa9e4066Sahrens spa_remove(spa_t *spa) 573fa9e4066Sahrens { 574c5904d13Seschrock spa_config_dirent_t *dp; 575c5904d13Seschrock 576fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 577fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 578fa9e4066Sahrens 5791195e687SMark J Musante nvlist_free(spa->spa_config_splitting); 5801195e687SMark J Musante 581fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 582fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 583fa9e4066Sahrens 5840373e76bSbonwick if (spa->spa_root) { 585fa9e4066Sahrens spa_strfree(spa->spa_root); 5860373e76bSbonwick spa_active_count--; 5870373e76bSbonwick } 588fa9e4066Sahrens 589c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 590c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 591c5904d13Seschrock if (dp->scd_path != NULL) 592c5904d13Seschrock spa_strfree(dp->scd_path); 593c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 594c5904d13Seschrock } 595c5904d13Seschrock 596c5904d13Seschrock list_destroy(&spa->spa_config_list); 5972f8aaab3Seschrock 598ad135b5dSChristopher Siden nvlist_free(spa->spa_label_features); 5994b964adaSGeorge Wilson nvlist_free(spa->spa_load_info); 600fa9e4066Sahrens spa_config_set(spa, NULL); 601fa9e4066Sahrens 602*283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 603*283b8460SGeorge.Wilson if (spa->spa_deadman_cycid != CYCLIC_NONE) 604*283b8460SGeorge.Wilson cyclic_remove(spa->spa_deadman_cycid); 605*283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 606*283b8460SGeorge.Wilson spa->spa_deadman_cycid = CYCLIC_NONE; 607*283b8460SGeorge.Wilson 608fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 60991ebeef5Sahrens 610e14bb325SJeff Bonwick spa_config_lock_destroy(spa); 611fa9e4066Sahrens 612b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 613cde58dbcSMatthew Ahrens bplist_destroy(&spa->spa_free_bplist[t]); 614b24ab676SJeff Bonwick 615c25056deSgw cv_destroy(&spa->spa_async_cv); 61635a5a358SJonathan Adams cv_destroy(&spa->spa_proc_cv); 617c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 618e14bb325SJeff Bonwick cv_destroy(&spa->spa_suspend_cv); 619c25056deSgw 6205ad82045Snd mutex_destroy(&spa->spa_async_lock); 621c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 62235a5a358SJonathan Adams mutex_destroy(&spa->spa_errlog_lock); 62306eeb2adSek mutex_destroy(&spa->spa_history_lock); 62435a5a358SJonathan Adams mutex_destroy(&spa->spa_proc_lock); 625b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 62635a5a358SJonathan Adams mutex_destroy(&spa->spa_scrub_lock); 627e14bb325SJeff Bonwick mutex_destroy(&spa->spa_suspend_lock); 628a1521560SJeff Bonwick mutex_destroy(&spa->spa_vdev_top_lock); 6295ad82045Snd 630fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 631fa9e4066Sahrens } 632fa9e4066Sahrens 633fa9e4066Sahrens /* 634fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 635fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 636fa9e4066Sahrens */ 637fa9e4066Sahrens spa_t * 638fa9e4066Sahrens spa_next(spa_t *prev) 639fa9e4066Sahrens { 640fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 641fa9e4066Sahrens 642fa9e4066Sahrens if (prev) 643fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 644fa9e4066Sahrens else 645fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 646fa9e4066Sahrens } 647fa9e4066Sahrens 648fa9e4066Sahrens /* 649fa9e4066Sahrens * ========================================================================== 650fa9e4066Sahrens * SPA refcount functions 651fa9e4066Sahrens * ========================================================================== 652fa9e4066Sahrens */ 653fa9e4066Sahrens 654fa9e4066Sahrens /* 655fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 656fa9e4066Sahrens * have the namespace lock held. 657fa9e4066Sahrens */ 658fa9e4066Sahrens void 659fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 660fa9e4066Sahrens { 661088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || 662fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 663fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 664fa9e4066Sahrens } 665fa9e4066Sahrens 666fa9e4066Sahrens /* 667fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 668fa9e4066Sahrens * have the namespace lock held. 669fa9e4066Sahrens */ 670fa9e4066Sahrens void 671fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 672fa9e4066Sahrens { 673088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || 674fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 675fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 676fa9e4066Sahrens } 677fa9e4066Sahrens 678fa9e4066Sahrens /* 679fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 680088f3894Sahrens * spa_namespace_lock held. We really compare against spa_minref, which is the 681fa9e4066Sahrens * number of references acquired when opening a pool 682fa9e4066Sahrens */ 683fa9e4066Sahrens boolean_t 684fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 685fa9e4066Sahrens { 686fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 687fa9e4066Sahrens 688088f3894Sahrens return (refcount_count(&spa->spa_refcount) == spa->spa_minref); 689fa9e4066Sahrens } 690fa9e4066Sahrens 69199653d4eSeschrock /* 69299653d4eSeschrock * ========================================================================== 693fa94a07fSbrendan * SPA spare and l2cache tracking 69499653d4eSeschrock * ========================================================================== 69599653d4eSeschrock */ 69699653d4eSeschrock 697fa94a07fSbrendan /* 698fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 699fa94a07fSbrendan * for 'auxiliary' devices. 700fa94a07fSbrendan */ 701fa94a07fSbrendan 702fa94a07fSbrendan typedef struct spa_aux { 703fa94a07fSbrendan uint64_t aux_guid; 704fa94a07fSbrendan uint64_t aux_pool; 705fa94a07fSbrendan avl_node_t aux_avl; 706fa94a07fSbrendan int aux_count; 707fa94a07fSbrendan } spa_aux_t; 708fa94a07fSbrendan 709fa94a07fSbrendan static int 710fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 711fa94a07fSbrendan { 712fa94a07fSbrendan const spa_aux_t *sa = a; 713fa94a07fSbrendan const spa_aux_t *sb = b; 714fa94a07fSbrendan 715fa94a07fSbrendan if (sa->aux_guid < sb->aux_guid) 716fa94a07fSbrendan return (-1); 717fa94a07fSbrendan else if (sa->aux_guid > sb->aux_guid) 718fa94a07fSbrendan return (1); 719fa94a07fSbrendan else 720fa94a07fSbrendan return (0); 721fa94a07fSbrendan } 722fa94a07fSbrendan 723fa94a07fSbrendan void 724fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 725fa94a07fSbrendan { 726fa94a07fSbrendan avl_index_t where; 727fa94a07fSbrendan spa_aux_t search; 728fa94a07fSbrendan spa_aux_t *aux; 729fa94a07fSbrendan 730fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 731fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 732fa94a07fSbrendan aux->aux_count++; 733fa94a07fSbrendan } else { 734fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 735fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 736fa94a07fSbrendan aux->aux_count = 1; 737fa94a07fSbrendan avl_insert(avl, aux, where); 738fa94a07fSbrendan } 739fa94a07fSbrendan } 740fa94a07fSbrendan 741fa94a07fSbrendan void 742fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 743fa94a07fSbrendan { 744fa94a07fSbrendan spa_aux_t search; 745fa94a07fSbrendan spa_aux_t *aux; 746fa94a07fSbrendan avl_index_t where; 747fa94a07fSbrendan 748fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 749fa94a07fSbrendan aux = avl_find(avl, &search, &where); 750fa94a07fSbrendan 751fa94a07fSbrendan ASSERT(aux != NULL); 752fa94a07fSbrendan 753fa94a07fSbrendan if (--aux->aux_count == 0) { 754fa94a07fSbrendan avl_remove(avl, aux); 755fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 756fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 757fa94a07fSbrendan aux->aux_pool = 0ULL; 758fa94a07fSbrendan } 759fa94a07fSbrendan } 760fa94a07fSbrendan 761fa94a07fSbrendan boolean_t 76289a89ebfSlling spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) 763fa94a07fSbrendan { 764fa94a07fSbrendan spa_aux_t search, *found; 765fa94a07fSbrendan 766fa94a07fSbrendan search.aux_guid = guid; 76789a89ebfSlling found = avl_find(avl, &search, NULL); 768fa94a07fSbrendan 769fa94a07fSbrendan if (pool) { 770fa94a07fSbrendan if (found) 771fa94a07fSbrendan *pool = found->aux_pool; 772fa94a07fSbrendan else 773fa94a07fSbrendan *pool = 0ULL; 774fa94a07fSbrendan } 775fa94a07fSbrendan 77689a89ebfSlling if (refcnt) { 77789a89ebfSlling if (found) 77889a89ebfSlling *refcnt = found->aux_count; 77989a89ebfSlling else 78089a89ebfSlling *refcnt = 0; 78189a89ebfSlling } 78289a89ebfSlling 783fa94a07fSbrendan return (found != NULL); 784fa94a07fSbrendan } 785fa94a07fSbrendan 786fa94a07fSbrendan void 787fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 788fa94a07fSbrendan { 789fa94a07fSbrendan spa_aux_t search, *found; 790fa94a07fSbrendan avl_index_t where; 791fa94a07fSbrendan 792fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 793fa94a07fSbrendan found = avl_find(avl, &search, &where); 794fa94a07fSbrendan ASSERT(found != NULL); 795fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 796fa94a07fSbrendan 797fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 798fa94a07fSbrendan } 799fa94a07fSbrendan 80099653d4eSeschrock /* 80139c23413Seschrock * Spares are tracked globally due to the following constraints: 80239c23413Seschrock * 80339c23413Seschrock * - A spare may be part of multiple pools. 80439c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 80539c23413Seschrock * another pool. 80639c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 80739c23413Seschrock * the target is a spare in the same pool. 80839c23413Seschrock * 80939c23413Seschrock * We keep track of all spares on the system through the use of a reference 81039c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 81139c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 81239c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 81339c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 81439c23413Seschrock * pool), we also keep track of which pool its been made a part of. 81539c23413Seschrock * 81639c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 81739c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 81839c23413Seschrock * separate spare lock exists for the status query path, which does not need to 81939c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 82099653d4eSeschrock */ 82139c23413Seschrock 82299653d4eSeschrock static int 82399653d4eSeschrock spa_spare_compare(const void *a, const void *b) 82499653d4eSeschrock { 825fa94a07fSbrendan return (spa_aux_compare(a, b)); 82699653d4eSeschrock } 82799653d4eSeschrock 82899653d4eSeschrock void 82939c23413Seschrock spa_spare_add(vdev_t *vd) 83099653d4eSeschrock { 83199653d4eSeschrock mutex_enter(&spa_spare_lock); 83239c23413Seschrock ASSERT(!vd->vdev_isspare); 833fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 83439c23413Seschrock vd->vdev_isspare = B_TRUE; 83599653d4eSeschrock mutex_exit(&spa_spare_lock); 83699653d4eSeschrock } 83799653d4eSeschrock 83899653d4eSeschrock void 83939c23413Seschrock spa_spare_remove(vdev_t *vd) 84099653d4eSeschrock { 84199653d4eSeschrock mutex_enter(&spa_spare_lock); 84239c23413Seschrock ASSERT(vd->vdev_isspare); 843fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 84439c23413Seschrock vd->vdev_isspare = B_FALSE; 84599653d4eSeschrock mutex_exit(&spa_spare_lock); 84699653d4eSeschrock } 84799653d4eSeschrock 84899653d4eSeschrock boolean_t 84989a89ebfSlling spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) 85099653d4eSeschrock { 851fa94a07fSbrendan boolean_t found; 85299653d4eSeschrock 85399653d4eSeschrock mutex_enter(&spa_spare_lock); 85489a89ebfSlling found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); 85599653d4eSeschrock mutex_exit(&spa_spare_lock); 85699653d4eSeschrock 857fa94a07fSbrendan return (found); 85839c23413Seschrock } 85939c23413Seschrock 86039c23413Seschrock void 86139c23413Seschrock spa_spare_activate(vdev_t *vd) 86239c23413Seschrock { 86339c23413Seschrock mutex_enter(&spa_spare_lock); 86439c23413Seschrock ASSERT(vd->vdev_isspare); 865fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 866fa94a07fSbrendan mutex_exit(&spa_spare_lock); 867fa94a07fSbrendan } 86839c23413Seschrock 869fa94a07fSbrendan /* 870fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 871fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 872fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 873fa94a07fSbrendan */ 87439c23413Seschrock 875fa94a07fSbrendan static int 876fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 877fa94a07fSbrendan { 878fa94a07fSbrendan return (spa_aux_compare(a, b)); 879fa94a07fSbrendan } 880fa94a07fSbrendan 881fa94a07fSbrendan void 882fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 883fa94a07fSbrendan { 884fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 885fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 886fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 887fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 888fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 889fa94a07fSbrendan } 890fa94a07fSbrendan 891fa94a07fSbrendan void 892fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 893fa94a07fSbrendan { 894fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 895fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 896fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 897fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 898fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 899fa94a07fSbrendan } 900fa94a07fSbrendan 901fa94a07fSbrendan boolean_t 902fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 903fa94a07fSbrendan { 904fa94a07fSbrendan boolean_t found; 905fa94a07fSbrendan 906fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 90789a89ebfSlling found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); 908fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 909fa94a07fSbrendan 910fa94a07fSbrendan return (found); 911fa94a07fSbrendan } 912fa94a07fSbrendan 913fa94a07fSbrendan void 914fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 915fa94a07fSbrendan { 916fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 917fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 918fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 919fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 920fa94a07fSbrendan } 921fa94a07fSbrendan 922fa9e4066Sahrens /* 923fa9e4066Sahrens * ========================================================================== 924fa9e4066Sahrens * SPA vdev locking 925fa9e4066Sahrens * ========================================================================== 926fa9e4066Sahrens */ 927fa9e4066Sahrens 928fa9e4066Sahrens /* 929ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 930ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 931fa9e4066Sahrens * It returns the next transaction group for the spa_t. 932fa9e4066Sahrens */ 933fa9e4066Sahrens uint64_t 934fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 935fa9e4066Sahrens { 936a1521560SJeff Bonwick mutex_enter(&spa->spa_vdev_top_lock); 937bbfd46c4SJeff Bonwick mutex_enter(&spa_namespace_lock); 93888ecc943SGeorge Wilson return (spa_vdev_config_enter(spa)); 93988ecc943SGeorge Wilson } 94088ecc943SGeorge Wilson 94188ecc943SGeorge Wilson /* 94288ecc943SGeorge Wilson * Internal implementation for spa_vdev_enter(). Used when a vdev 94388ecc943SGeorge Wilson * operation requires multiple syncs (i.e. removing a device) while 94488ecc943SGeorge Wilson * keeping the spa_namespace_lock held. 94588ecc943SGeorge Wilson */ 94688ecc943SGeorge Wilson uint64_t 94788ecc943SGeorge Wilson spa_vdev_config_enter(spa_t *spa) 94888ecc943SGeorge Wilson { 94988ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 9503d7072f8Seschrock 951e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 952fa9e4066Sahrens 953fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 954fa9e4066Sahrens } 955fa9e4066Sahrens 956fa9e4066Sahrens /* 95788ecc943SGeorge Wilson * Used in combination with spa_vdev_config_enter() to allow the syncing 95888ecc943SGeorge Wilson * of multiple transactions without releasing the spa_namespace_lock. 959fa9e4066Sahrens */ 96088ecc943SGeorge Wilson void 96188ecc943SGeorge Wilson spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) 962fa9e4066Sahrens { 96388ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 96488ecc943SGeorge Wilson 9650e34b6a7Sbonwick int config_changed = B_FALSE; 966ea8dc4b6Seschrock 9670373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 9680e34b6a7Sbonwick 969e14bb325SJeff Bonwick spa->spa_pending_vdev = NULL; 970e14bb325SJeff Bonwick 9710e34b6a7Sbonwick /* 9720e34b6a7Sbonwick * Reassess the DTLs. 9730e34b6a7Sbonwick */ 9740373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 9750e34b6a7Sbonwick 976e14bb325SJeff Bonwick if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { 9770e34b6a7Sbonwick config_changed = B_TRUE; 9788f18d1faSGeorge Wilson spa->spa_config_generation++; 9790e34b6a7Sbonwick } 980ea8dc4b6Seschrock 98188ecc943SGeorge Wilson /* 98288ecc943SGeorge Wilson * Verify the metaslab classes. 98388ecc943SGeorge Wilson */ 984b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); 985b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); 98688ecc943SGeorge Wilson 987e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 988fa9e4066Sahrens 98988ecc943SGeorge Wilson /* 99088ecc943SGeorge Wilson * Panic the system if the specified tag requires it. This 99188ecc943SGeorge Wilson * is useful for ensuring that configurations are updated 99288ecc943SGeorge Wilson * transactionally. 99388ecc943SGeorge Wilson */ 99488ecc943SGeorge Wilson if (zio_injection_enabled) 9951195e687SMark J Musante zio_handle_panic_injection(spa, tag, 0); 99688ecc943SGeorge Wilson 997fa9e4066Sahrens /* 998fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 999fa9e4066Sahrens * that there won't be more than one config change per txg. 1000fa9e4066Sahrens * This allows us to use the txg as the generation number. 1001fa9e4066Sahrens */ 1002fa9e4066Sahrens if (error == 0) 1003fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 1004fa9e4066Sahrens 1005fa9e4066Sahrens if (vd != NULL) { 10068ad4d6ddSJeff Bonwick ASSERT(!vd->vdev_detached || vd->vdev_dtl_smo.smo_object == 0); 10078ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1008fa9e4066Sahrens vdev_free(vd); 10098ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1010fa9e4066Sahrens } 1011fa9e4066Sahrens 1012fa9e4066Sahrens /* 10130e34b6a7Sbonwick * If the config changed, update the config cache. 1014fa9e4066Sahrens */ 10150e34b6a7Sbonwick if (config_changed) 1016c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 101788ecc943SGeorge Wilson } 1018ea8dc4b6Seschrock 101988ecc943SGeorge Wilson /* 102088ecc943SGeorge Wilson * Unlock the spa_t after adding or removing a vdev. Besides undoing the 102188ecc943SGeorge Wilson * locking of spa_vdev_enter(), we also want make sure the transactions have 102288ecc943SGeorge Wilson * synced to disk, and then update the global configuration cache with the new 102388ecc943SGeorge Wilson * information. 102488ecc943SGeorge Wilson */ 102588ecc943SGeorge Wilson int 102688ecc943SGeorge Wilson spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 102788ecc943SGeorge Wilson { 102888ecc943SGeorge Wilson spa_vdev_config_exit(spa, vd, txg, error, FTAG); 1029ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1030bbfd46c4SJeff Bonwick mutex_exit(&spa->spa_vdev_top_lock); 1031fa9e4066Sahrens 1032fa9e4066Sahrens return (error); 1033fa9e4066Sahrens } 1034fa9e4066Sahrens 1035e14bb325SJeff Bonwick /* 1036e14bb325SJeff Bonwick * Lock the given spa_t for the purpose of changing vdev state. 1037e14bb325SJeff Bonwick */ 1038e14bb325SJeff Bonwick void 10398f18d1faSGeorge Wilson spa_vdev_state_enter(spa_t *spa, int oplocks) 1040e14bb325SJeff Bonwick { 10418f18d1faSGeorge Wilson int locks = SCL_STATE_ALL | oplocks; 10428f18d1faSGeorge Wilson 1043dcba9f3fSGeorge Wilson /* 1044dcba9f3fSGeorge Wilson * Root pools may need to read of the underlying devfs filesystem 1045dcba9f3fSGeorge Wilson * when opening up a vdev. Unfortunately if we're holding the 1046dcba9f3fSGeorge Wilson * SCL_ZIO lock it will result in a deadlock when we try to issue 1047dcba9f3fSGeorge Wilson * the read from the root filesystem. Instead we "prefetch" 1048dcba9f3fSGeorge Wilson * the associated vnodes that we need prior to opening the 1049dcba9f3fSGeorge Wilson * underlying devices and cache them so that we can prevent 1050dcba9f3fSGeorge Wilson * any I/O when we are doing the actual open. 1051dcba9f3fSGeorge Wilson */ 1052dcba9f3fSGeorge Wilson if (spa_is_root(spa)) { 10539842588bSGeorge Wilson int low = locks & ~(SCL_ZIO - 1); 10549842588bSGeorge Wilson int high = locks & ~low; 10559842588bSGeorge Wilson 10569842588bSGeorge Wilson spa_config_enter(spa, high, spa, RW_WRITER); 1057dcba9f3fSGeorge Wilson vdev_hold(spa->spa_root_vdev); 10589842588bSGeorge Wilson spa_config_enter(spa, low, spa, RW_WRITER); 1059dcba9f3fSGeorge Wilson } else { 1060dcba9f3fSGeorge Wilson spa_config_enter(spa, locks, spa, RW_WRITER); 1061dcba9f3fSGeorge Wilson } 10628f18d1faSGeorge Wilson spa->spa_vdev_locks = locks; 1063e14bb325SJeff Bonwick } 1064e14bb325SJeff Bonwick 1065e14bb325SJeff Bonwick int 1066e14bb325SJeff Bonwick spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) 1067e14bb325SJeff Bonwick { 1068c6065d0fSGeorge Wilson boolean_t config_changed = B_FALSE; 1069c6065d0fSGeorge Wilson 1070b24ab676SJeff Bonwick if (vd != NULL || error == 0) 1071b24ab676SJeff Bonwick vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 1072b24ab676SJeff Bonwick 0, 0, B_FALSE); 1073b24ab676SJeff Bonwick 10748f18d1faSGeorge Wilson if (vd != NULL) { 1075e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 1076c6065d0fSGeorge Wilson config_changed = B_TRUE; 10778f18d1faSGeorge Wilson spa->spa_config_generation++; 10788f18d1faSGeorge Wilson } 1079e14bb325SJeff Bonwick 1080dcba9f3fSGeorge Wilson if (spa_is_root(spa)) 1081dcba9f3fSGeorge Wilson vdev_rele(spa->spa_root_vdev); 1082dcba9f3fSGeorge Wilson 10838f18d1faSGeorge Wilson ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); 10848f18d1faSGeorge Wilson spa_config_exit(spa, spa->spa_vdev_locks, spa); 1085e14bb325SJeff Bonwick 10868ad4d6ddSJeff Bonwick /* 10878ad4d6ddSJeff Bonwick * If anything changed, wait for it to sync. This ensures that, 10888ad4d6ddSJeff Bonwick * from the system administrator's perspective, zpool(1M) commands 10898ad4d6ddSJeff Bonwick * are synchronous. This is important for things like zpool offline: 10908ad4d6ddSJeff Bonwick * when the command completes, you expect no further I/O from ZFS. 10918ad4d6ddSJeff Bonwick */ 10928ad4d6ddSJeff Bonwick if (vd != NULL) 10938ad4d6ddSJeff Bonwick txg_wait_synced(spa->spa_dsl_pool, 0); 10948ad4d6ddSJeff Bonwick 1095c6065d0fSGeorge Wilson /* 1096c6065d0fSGeorge Wilson * If the config changed, update the config cache. 1097c6065d0fSGeorge Wilson */ 1098c6065d0fSGeorge Wilson if (config_changed) { 1099c6065d0fSGeorge Wilson mutex_enter(&spa_namespace_lock); 1100c6065d0fSGeorge Wilson spa_config_sync(spa, B_FALSE, B_TRUE); 1101c6065d0fSGeorge Wilson mutex_exit(&spa_namespace_lock); 1102c6065d0fSGeorge Wilson } 1103c6065d0fSGeorge Wilson 1104e14bb325SJeff Bonwick return (error); 1105e14bb325SJeff Bonwick } 1106e14bb325SJeff Bonwick 1107fa9e4066Sahrens /* 1108fa9e4066Sahrens * ========================================================================== 1109fa9e4066Sahrens * Miscellaneous functions 1110fa9e4066Sahrens * ========================================================================== 1111fa9e4066Sahrens */ 1112fa9e4066Sahrens 1113ad135b5dSChristopher Siden void 1114ad135b5dSChristopher Siden spa_activate_mos_feature(spa_t *spa, const char *feature) 1115ad135b5dSChristopher Siden { 1116ad135b5dSChristopher Siden (void) nvlist_add_boolean(spa->spa_label_features, feature); 1117ad135b5dSChristopher Siden vdev_config_dirty(spa->spa_root_vdev); 1118ad135b5dSChristopher Siden } 1119ad135b5dSChristopher Siden 1120ad135b5dSChristopher Siden void 1121ad135b5dSChristopher Siden spa_deactivate_mos_feature(spa_t *spa, const char *feature) 1122ad135b5dSChristopher Siden { 1123ad135b5dSChristopher Siden (void) nvlist_remove_all(spa->spa_label_features, feature); 1124ad135b5dSChristopher Siden vdev_config_dirty(spa->spa_root_vdev); 1125ad135b5dSChristopher Siden } 1126ad135b5dSChristopher Siden 1127fa9e4066Sahrens /* 1128fa9e4066Sahrens * Rename a spa_t. 1129fa9e4066Sahrens */ 1130fa9e4066Sahrens int 1131fa9e4066Sahrens spa_rename(const char *name, const char *newname) 1132fa9e4066Sahrens { 1133fa9e4066Sahrens spa_t *spa; 1134fa9e4066Sahrens int err; 1135fa9e4066Sahrens 1136fa9e4066Sahrens /* 1137fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 1138fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 1139fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 1140ea8dc4b6Seschrock * allow recursive calls for other reasons. 1141fa9e4066Sahrens */ 1142fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 1143fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 1144fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1145fa9e4066Sahrens return (err); 1146fa9e4066Sahrens } 1147fa9e4066Sahrens 1148e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1149fa9e4066Sahrens 1150fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 1151e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); 1152fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 1153fa9e4066Sahrens 1154fa9e4066Sahrens /* 1155fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 1156fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 1157fa9e4066Sahrens * during the sync. 1158fa9e4066Sahrens */ 1159fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 1160fa9e4066Sahrens 1161e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1162fa9e4066Sahrens 11630373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 1164fa9e4066Sahrens 1165fa9e4066Sahrens /* 1166fa9e4066Sahrens * Sync the updated config cache. 1167fa9e4066Sahrens */ 1168c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 1169fa9e4066Sahrens 1170fa9e4066Sahrens spa_close(spa, FTAG); 1171fa9e4066Sahrens 1172fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1173fa9e4066Sahrens 1174fa9e4066Sahrens return (0); 1175fa9e4066Sahrens } 1176fa9e4066Sahrens 1177fa9e4066Sahrens /* 1178f9af39baSGeorge Wilson * Return the spa_t associated with given pool_guid, if it exists. If 1179f9af39baSGeorge Wilson * device_guid is non-zero, determine whether the pool exists *and* contains 1180f9af39baSGeorge Wilson * a device with the specified device_guid. 1181fa9e4066Sahrens */ 1182f9af39baSGeorge Wilson spa_t * 1183f9af39baSGeorge Wilson spa_by_guid(uint64_t pool_guid, uint64_t device_guid) 1184fa9e4066Sahrens { 1185fa9e4066Sahrens spa_t *spa; 1186fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 1187fa9e4066Sahrens 1188ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1189fa9e4066Sahrens 1190fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 1191fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 1192fa9e4066Sahrens continue; 1193fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 1194fa9e4066Sahrens continue; 119539c23413Seschrock if (spa_guid(spa) == pool_guid) { 119639c23413Seschrock if (device_guid == 0) 119739c23413Seschrock break; 119839c23413Seschrock 119939c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 120039c23413Seschrock device_guid) != NULL) 120139c23413Seschrock break; 120239c23413Seschrock 120339c23413Seschrock /* 12048654d025Sperrin * Check any devices we may be in the process of adding. 120539c23413Seschrock */ 120639c23413Seschrock if (spa->spa_pending_vdev) { 120739c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 120839c23413Seschrock device_guid) != NULL) 120939c23413Seschrock break; 121039c23413Seschrock } 121139c23413Seschrock } 1212fa9e4066Sahrens } 1213fa9e4066Sahrens 1214f9af39baSGeorge Wilson return (spa); 1215f9af39baSGeorge Wilson } 1216f9af39baSGeorge Wilson 1217f9af39baSGeorge Wilson /* 1218f9af39baSGeorge Wilson * Determine whether a pool with the given pool_guid exists. 1219f9af39baSGeorge Wilson */ 1220f9af39baSGeorge Wilson boolean_t 1221f9af39baSGeorge Wilson spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 1222f9af39baSGeorge Wilson { 1223f9af39baSGeorge Wilson return (spa_by_guid(pool_guid, device_guid) != NULL); 1224fa9e4066Sahrens } 1225fa9e4066Sahrens 1226fa9e4066Sahrens char * 1227fa9e4066Sahrens spa_strdup(const char *s) 1228fa9e4066Sahrens { 1229fa9e4066Sahrens size_t len; 1230fa9e4066Sahrens char *new; 1231fa9e4066Sahrens 1232fa9e4066Sahrens len = strlen(s); 1233fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 1234fa9e4066Sahrens bcopy(s, new, len); 1235fa9e4066Sahrens new[len] = '\0'; 1236fa9e4066Sahrens 1237fa9e4066Sahrens return (new); 1238fa9e4066Sahrens } 1239fa9e4066Sahrens 1240fa9e4066Sahrens void 1241fa9e4066Sahrens spa_strfree(char *s) 1242fa9e4066Sahrens { 1243fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 1244fa9e4066Sahrens } 1245fa9e4066Sahrens 1246fa9e4066Sahrens uint64_t 1247fa9e4066Sahrens spa_get_random(uint64_t range) 1248fa9e4066Sahrens { 1249fa9e4066Sahrens uint64_t r; 1250fa9e4066Sahrens 1251fa9e4066Sahrens ASSERT(range != 0); 1252fa9e4066Sahrens 1253fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 1254fa9e4066Sahrens 1255fa9e4066Sahrens return (r % range); 1256fa9e4066Sahrens } 1257fa9e4066Sahrens 12581195e687SMark J Musante uint64_t 12591195e687SMark J Musante spa_generate_guid(spa_t *spa) 12601195e687SMark J Musante { 12611195e687SMark J Musante uint64_t guid = spa_get_random(-1ULL); 12621195e687SMark J Musante 12631195e687SMark J Musante if (spa != NULL) { 12641195e687SMark J Musante while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) 12651195e687SMark J Musante guid = spa_get_random(-1ULL); 12661195e687SMark J Musante } else { 12671195e687SMark J Musante while (guid == 0 || spa_guid_exists(guid, 0)) 12681195e687SMark J Musante guid = spa_get_random(-1ULL); 12691195e687SMark J Musante } 12701195e687SMark J Musante 12711195e687SMark J Musante return (guid); 12721195e687SMark J Musante } 12731195e687SMark J Musante 1274fa9e4066Sahrens void 1275b24ab676SJeff Bonwick sprintf_blkptr(char *buf, const blkptr_t *bp) 1276fa9e4066Sahrens { 1277ad135b5dSChristopher Siden char type[256]; 1278f0ba89beSJeff Bonwick char *checksum = NULL; 1279f0ba89beSJeff Bonwick char *compress = NULL; 1280f0ba89beSJeff Bonwick 1281f0ba89beSJeff Bonwick if (bp != NULL) { 1282ad135b5dSChristopher Siden if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { 1283ad135b5dSChristopher Siden dmu_object_byteswap_t bswap = 1284ad135b5dSChristopher Siden DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); 1285ad135b5dSChristopher Siden (void) snprintf(type, sizeof (type), "bswap %s %s", 1286ad135b5dSChristopher Siden DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? 1287ad135b5dSChristopher Siden "metadata" : "data", 1288ad135b5dSChristopher Siden dmu_ot_byteswap[bswap].ob_name); 1289ad135b5dSChristopher Siden } else { 1290ad135b5dSChristopher Siden (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, 1291ad135b5dSChristopher Siden sizeof (type)); 1292ad135b5dSChristopher Siden } 1293f0ba89beSJeff Bonwick checksum = zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; 1294f0ba89beSJeff Bonwick compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; 1295f0ba89beSJeff Bonwick } 1296fa9e4066Sahrens 1297b24ab676SJeff Bonwick SPRINTF_BLKPTR(snprintf, ' ', buf, bp, type, checksum, compress); 1298fa9e4066Sahrens } 1299fa9e4066Sahrens 1300fa9e4066Sahrens void 1301fa9e4066Sahrens spa_freeze(spa_t *spa) 1302fa9e4066Sahrens { 1303fa9e4066Sahrens uint64_t freeze_txg = 0; 1304fa9e4066Sahrens 1305e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1306fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 1307fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 1308fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 1309fa9e4066Sahrens } 1310e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1311fa9e4066Sahrens if (freeze_txg != 0) 1312fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 1313fa9e4066Sahrens } 1314fa9e4066Sahrens 13150125049cSahrens void 13160125049cSahrens zfs_panic_recover(const char *fmt, ...) 13170125049cSahrens { 13180125049cSahrens va_list adx; 13190125049cSahrens 13200125049cSahrens va_start(adx, fmt); 13210125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 13220125049cSahrens va_end(adx); 13230125049cSahrens } 13240125049cSahrens 13253f9d6ad7SLin Ling /* 13263f9d6ad7SLin Ling * This is a stripped-down version of strtoull, suitable only for converting 13273f9d6ad7SLin Ling * lowercase hexidecimal numbers that don't overflow. 13283f9d6ad7SLin Ling */ 13293f9d6ad7SLin Ling uint64_t 13303f9d6ad7SLin Ling strtonum(const char *str, char **nptr) 13313f9d6ad7SLin Ling { 13323f9d6ad7SLin Ling uint64_t val = 0; 13333f9d6ad7SLin Ling char c; 13343f9d6ad7SLin Ling int digit; 13353f9d6ad7SLin Ling 13363f9d6ad7SLin Ling while ((c = *str) != '\0') { 13373f9d6ad7SLin Ling if (c >= '0' && c <= '9') 13383f9d6ad7SLin Ling digit = c - '0'; 13393f9d6ad7SLin Ling else if (c >= 'a' && c <= 'f') 13403f9d6ad7SLin Ling digit = 10 + c - 'a'; 13413f9d6ad7SLin Ling else 13423f9d6ad7SLin Ling break; 13433f9d6ad7SLin Ling 13443f9d6ad7SLin Ling val *= 16; 13453f9d6ad7SLin Ling val += digit; 13463f9d6ad7SLin Ling 13473f9d6ad7SLin Ling str++; 13483f9d6ad7SLin Ling } 13493f9d6ad7SLin Ling 13503f9d6ad7SLin Ling if (nptr) 13513f9d6ad7SLin Ling *nptr = (char *)str; 13523f9d6ad7SLin Ling 13533f9d6ad7SLin Ling return (val); 13543f9d6ad7SLin Ling } 13553f9d6ad7SLin Ling 1356fa9e4066Sahrens /* 1357fa9e4066Sahrens * ========================================================================== 1358fa9e4066Sahrens * Accessor functions 1359fa9e4066Sahrens * ========================================================================== 1360fa9e4066Sahrens */ 1361fa9e4066Sahrens 1362088f3894Sahrens boolean_t 136388b7b0f2SMatthew Ahrens spa_shutting_down(spa_t *spa) 1364fa9e4066Sahrens { 136588b7b0f2SMatthew Ahrens return (spa->spa_async_suspended); 1366fa9e4066Sahrens } 1367fa9e4066Sahrens 1368fa9e4066Sahrens dsl_pool_t * 1369fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1370fa9e4066Sahrens { 1371fa9e4066Sahrens return (spa->spa_dsl_pool); 1372fa9e4066Sahrens } 1373fa9e4066Sahrens 1374ad135b5dSChristopher Siden boolean_t 1375ad135b5dSChristopher Siden spa_is_initializing(spa_t *spa) 1376ad135b5dSChristopher Siden { 1377ad135b5dSChristopher Siden return (spa->spa_is_initializing); 1378ad135b5dSChristopher Siden } 1379ad135b5dSChristopher Siden 1380fa9e4066Sahrens blkptr_t * 1381fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1382fa9e4066Sahrens { 1383fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1384fa9e4066Sahrens } 1385fa9e4066Sahrens 1386fa9e4066Sahrens void 1387fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1388fa9e4066Sahrens { 1389fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1390fa9e4066Sahrens } 1391fa9e4066Sahrens 1392fa9e4066Sahrens void 1393fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1394fa9e4066Sahrens { 1395fa9e4066Sahrens if (spa->spa_root == NULL) 1396fa9e4066Sahrens buf[0] = '\0'; 1397fa9e4066Sahrens else 1398fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1399fa9e4066Sahrens } 1400fa9e4066Sahrens 1401fa9e4066Sahrens int 1402fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1403fa9e4066Sahrens { 1404fa9e4066Sahrens return (spa->spa_sync_pass); 1405fa9e4066Sahrens } 1406fa9e4066Sahrens 1407fa9e4066Sahrens char * 1408fa9e4066Sahrens spa_name(spa_t *spa) 1409fa9e4066Sahrens { 1410fa9e4066Sahrens return (spa->spa_name); 1411fa9e4066Sahrens } 1412fa9e4066Sahrens 1413fa9e4066Sahrens uint64_t 1414fa9e4066Sahrens spa_guid(spa_t *spa) 1415fa9e4066Sahrens { 1416dfbb9432SGeorge Wilson dsl_pool_t *dp = spa_get_dsl(spa); 1417dfbb9432SGeorge Wilson uint64_t guid; 1418dfbb9432SGeorge Wilson 1419b5989ec7Seschrock /* 1420b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1421b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1422e9103aaeSGarrett D'Amore * vdev. We stash the original pool guid in 'spa_config_guid' to handle 1423b5989ec7Seschrock * this case. 1424b5989ec7Seschrock */ 1425dfbb9432SGeorge Wilson if (spa->spa_root_vdev == NULL) 1426dfbb9432SGeorge Wilson return (spa->spa_config_guid); 1427dfbb9432SGeorge Wilson 1428dfbb9432SGeorge Wilson guid = spa->spa_last_synced_guid != 0 ? 1429dfbb9432SGeorge Wilson spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; 1430dfbb9432SGeorge Wilson 1431dfbb9432SGeorge Wilson /* 1432dfbb9432SGeorge Wilson * Return the most recently synced out guid unless we're 1433dfbb9432SGeorge Wilson * in syncing context. 1434dfbb9432SGeorge Wilson */ 1435dfbb9432SGeorge Wilson if (dp && dsl_pool_sync_context(dp)) 1436b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1437b5989ec7Seschrock else 1438dfbb9432SGeorge Wilson return (guid); 1439e9103aaeSGarrett D'Amore } 1440e9103aaeSGarrett D'Amore 1441e9103aaeSGarrett D'Amore uint64_t 1442e9103aaeSGarrett D'Amore spa_load_guid(spa_t *spa) 1443e9103aaeSGarrett D'Amore { 1444e9103aaeSGarrett D'Amore /* 1445e9103aaeSGarrett D'Amore * This is a GUID that exists solely as a reference for the 1446e9103aaeSGarrett D'Amore * purposes of the arc. It is generated at load time, and 1447e9103aaeSGarrett D'Amore * is never written to persistent storage. 1448e9103aaeSGarrett D'Amore */ 1449e9103aaeSGarrett D'Amore return (spa->spa_load_guid); 1450fa9e4066Sahrens } 1451fa9e4066Sahrens 1452fa9e4066Sahrens uint64_t 1453fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1454fa9e4066Sahrens { 1455fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1456fa9e4066Sahrens } 1457fa9e4066Sahrens 1458fa9e4066Sahrens uint64_t 1459fa9e4066Sahrens spa_first_txg(spa_t *spa) 1460fa9e4066Sahrens { 1461fa9e4066Sahrens return (spa->spa_first_txg); 1462fa9e4066Sahrens } 1463fa9e4066Sahrens 1464b24ab676SJeff Bonwick uint64_t 1465b24ab676SJeff Bonwick spa_syncing_txg(spa_t *spa) 1466b24ab676SJeff Bonwick { 1467b24ab676SJeff Bonwick return (spa->spa_syncing_txg); 1468b24ab676SJeff Bonwick } 1469b24ab676SJeff Bonwick 147088b7b0f2SMatthew Ahrens pool_state_t 1471fa9e4066Sahrens spa_state(spa_t *spa) 1472fa9e4066Sahrens { 1473fa9e4066Sahrens return (spa->spa_state); 1474fa9e4066Sahrens } 1475fa9e4066Sahrens 1476b16da2e2SGeorge Wilson spa_load_state_t 1477b16da2e2SGeorge Wilson spa_load_state(spa_t *spa) 1478b16da2e2SGeorge Wilson { 1479b16da2e2SGeorge Wilson return (spa->spa_load_state); 1480b16da2e2SGeorge Wilson } 1481b16da2e2SGeorge Wilson 1482fa9e4066Sahrens uint64_t 1483fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1484fa9e4066Sahrens { 1485fa9e4066Sahrens return (spa->spa_freeze_txg); 1486fa9e4066Sahrens } 1487fa9e4066Sahrens 1488fa9e4066Sahrens /* ARGSUSED */ 1489fa9e4066Sahrens uint64_t 1490fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 1491fa9e4066Sahrens { 1492fa9e4066Sahrens /* 1493b24ab676SJeff Bonwick * The worst case is single-sector max-parity RAID-Z blocks, in which 1494b24ab676SJeff Bonwick * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) 1495b24ab676SJeff Bonwick * times the size; so just assume that. Add to this the fact that 1496b24ab676SJeff Bonwick * we can have up to 3 DVAs per bp, and one more factor of 2 because 1497b24ab676SJeff Bonwick * the block may be dittoed with up to 3 DVAs by ddt_sync(). 149844cd46caSbillm */ 1499b24ab676SJeff Bonwick return (lsize * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2); 150044cd46caSbillm } 150144cd46caSbillm 1502485bbbf5SGeorge Wilson uint64_t 1503485bbbf5SGeorge Wilson spa_get_dspace(spa_t *spa) 1504485bbbf5SGeorge Wilson { 1505485bbbf5SGeorge Wilson return (spa->spa_dspace); 1506485bbbf5SGeorge Wilson } 1507485bbbf5SGeorge Wilson 1508485bbbf5SGeorge Wilson void 1509485bbbf5SGeorge Wilson spa_update_dspace(spa_t *spa) 1510485bbbf5SGeorge Wilson { 1511485bbbf5SGeorge Wilson spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + 1512485bbbf5SGeorge Wilson ddt_get_dedup_dspace(spa); 1513485bbbf5SGeorge Wilson } 1514485bbbf5SGeorge Wilson 15150a4e9518Sgw /* 15160a4e9518Sgw * Return the failure mode that has been set to this pool. The default 15170a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 15180a4e9518Sgw */ 15190a4e9518Sgw uint8_t 15200a4e9518Sgw spa_get_failmode(spa_t *spa) 15210a4e9518Sgw { 15220a4e9518Sgw return (spa->spa_failmode); 15230a4e9518Sgw } 15240a4e9518Sgw 1525e14bb325SJeff Bonwick boolean_t 1526e14bb325SJeff Bonwick spa_suspended(spa_t *spa) 1527e14bb325SJeff Bonwick { 1528e14bb325SJeff Bonwick return (spa->spa_suspended); 1529e14bb325SJeff Bonwick } 1530e14bb325SJeff Bonwick 153144cd46caSbillm uint64_t 153244cd46caSbillm spa_version(spa_t *spa) 153344cd46caSbillm { 153444cd46caSbillm return (spa->spa_ubsync.ub_version); 153544cd46caSbillm } 153644cd46caSbillm 1537b24ab676SJeff Bonwick boolean_t 1538b24ab676SJeff Bonwick spa_deflate(spa_t *spa) 1539b24ab676SJeff Bonwick { 1540b24ab676SJeff Bonwick return (spa->spa_deflate); 1541b24ab676SJeff Bonwick } 1542b24ab676SJeff Bonwick 1543b24ab676SJeff Bonwick metaslab_class_t * 1544b24ab676SJeff Bonwick spa_normal_class(spa_t *spa) 1545b24ab676SJeff Bonwick { 1546b24ab676SJeff Bonwick return (spa->spa_normal_class); 1547b24ab676SJeff Bonwick } 1548b24ab676SJeff Bonwick 1549b24ab676SJeff Bonwick metaslab_class_t * 1550b24ab676SJeff Bonwick spa_log_class(spa_t *spa) 1551b24ab676SJeff Bonwick { 1552b24ab676SJeff Bonwick return (spa->spa_log_class); 1553b24ab676SJeff Bonwick } 1554b24ab676SJeff Bonwick 155544cd46caSbillm int 155644cd46caSbillm spa_max_replication(spa_t *spa) 155744cd46caSbillm { 155844cd46caSbillm /* 1559e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 156044cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 156144cd46caSbillm * replication level accordingly. 1562fa9e4066Sahrens */ 1563e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 156444cd46caSbillm return (1); 156544cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1566fa9e4066Sahrens } 1567fa9e4066Sahrens 15683f9d6ad7SLin Ling int 15693f9d6ad7SLin Ling spa_prev_software_version(spa_t *spa) 15703f9d6ad7SLin Ling { 15713f9d6ad7SLin Ling return (spa->spa_prev_software_version); 15723f9d6ad7SLin Ling } 15733f9d6ad7SLin Ling 1574*283b8460SGeorge.Wilson uint64_t 1575*283b8460SGeorge.Wilson spa_deadman_synctime(spa_t *spa) 1576*283b8460SGeorge.Wilson { 1577*283b8460SGeorge.Wilson return (spa->spa_deadman_synctime); 1578*283b8460SGeorge.Wilson } 1579*283b8460SGeorge.Wilson 158099653d4eSeschrock uint64_t 1581b24ab676SJeff Bonwick dva_get_dsize_sync(spa_t *spa, const dva_t *dva) 158299653d4eSeschrock { 1583b24ab676SJeff Bonwick uint64_t asize = DVA_GET_ASIZE(dva); 1584b24ab676SJeff Bonwick uint64_t dsize = asize; 158599653d4eSeschrock 1586b24ab676SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); 158799653d4eSeschrock 1588b24ab676SJeff Bonwick if (asize != 0 && spa->spa_deflate) { 1589b24ab676SJeff Bonwick vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); 1590b24ab676SJeff Bonwick dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 159199653d4eSeschrock } 1592b24ab676SJeff Bonwick 1593b24ab676SJeff Bonwick return (dsize); 1594b24ab676SJeff Bonwick } 1595b24ab676SJeff Bonwick 1596b24ab676SJeff Bonwick uint64_t 1597b24ab676SJeff Bonwick bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) 1598b24ab676SJeff Bonwick { 1599b24ab676SJeff Bonwick uint64_t dsize = 0; 1600b24ab676SJeff Bonwick 1601b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1602b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1603b24ab676SJeff Bonwick 1604b24ab676SJeff Bonwick return (dsize); 1605b24ab676SJeff Bonwick } 1606b24ab676SJeff Bonwick 1607b24ab676SJeff Bonwick uint64_t 1608b24ab676SJeff Bonwick bp_get_dsize(spa_t *spa, const blkptr_t *bp) 1609b24ab676SJeff Bonwick { 1610b24ab676SJeff Bonwick uint64_t dsize = 0; 1611b24ab676SJeff Bonwick 1612b24ab676SJeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 1613b24ab676SJeff Bonwick 1614b24ab676SJeff Bonwick for (int d = 0; d < SPA_DVAS_PER_BP; d++) 1615b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1616b24ab676SJeff Bonwick 1617e14bb325SJeff Bonwick spa_config_exit(spa, SCL_VDEV, FTAG); 1618b24ab676SJeff Bonwick 1619b24ab676SJeff Bonwick return (dsize); 162099653d4eSeschrock } 162199653d4eSeschrock 1622fa9e4066Sahrens /* 1623fa9e4066Sahrens * ========================================================================== 1624fa9e4066Sahrens * Initialization and Termination 1625fa9e4066Sahrens * ========================================================================== 1626fa9e4066Sahrens */ 1627fa9e4066Sahrens 1628fa9e4066Sahrens static int 1629fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1630fa9e4066Sahrens { 1631fa9e4066Sahrens const spa_t *s1 = a1; 1632fa9e4066Sahrens const spa_t *s2 = a2; 1633fa9e4066Sahrens int s; 1634fa9e4066Sahrens 1635fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1636fa9e4066Sahrens if (s > 0) 1637fa9e4066Sahrens return (1); 1638fa9e4066Sahrens if (s < 0) 1639fa9e4066Sahrens return (-1); 1640fa9e4066Sahrens return (0); 1641fa9e4066Sahrens } 1642fa9e4066Sahrens 16430373e76bSbonwick int 16440373e76bSbonwick spa_busy(void) 16450373e76bSbonwick { 16460373e76bSbonwick return (spa_active_count); 16470373e76bSbonwick } 16480373e76bSbonwick 1649e7cbe64fSgw void 1650e7cbe64fSgw spa_boot_init() 1651e7cbe64fSgw { 1652e7cbe64fSgw spa_config_load(); 1653e7cbe64fSgw } 1654e7cbe64fSgw 1655fa9e4066Sahrens void 1656fa9e4066Sahrens spa_init(int mode) 1657fa9e4066Sahrens { 1658fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1659c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1660fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 1661fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1662fa9e4066Sahrens 1663fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1664fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1665fa9e4066Sahrens 1666fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 1667fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 1668fa94a07fSbrendan 1669fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 1670fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 167199653d4eSeschrock 16728ad4d6ddSJeff Bonwick spa_mode_global = mode; 1673fa9e4066Sahrens 1674*283b8460SGeorge.Wilson #ifdef _KERNEL 1675*283b8460SGeorge.Wilson spa_arch_init(); 1676*283b8460SGeorge.Wilson #else 1677cd1c8b85SMatthew Ahrens if (spa_mode_global != FREAD && dprintf_find_string("watch")) { 1678cd1c8b85SMatthew Ahrens arc_procfd = open("/proc/self/ctl", O_WRONLY); 1679cd1c8b85SMatthew Ahrens if (arc_procfd == -1) { 1680cd1c8b85SMatthew Ahrens perror("could not enable watchpoints: " 1681cd1c8b85SMatthew Ahrens "opening /proc/self/ctl failed: "); 1682cd1c8b85SMatthew Ahrens } else { 1683cd1c8b85SMatthew Ahrens arc_watch = B_TRUE; 1684cd1c8b85SMatthew Ahrens } 1685cd1c8b85SMatthew Ahrens } 1686cd1c8b85SMatthew Ahrens #endif 1687cd1c8b85SMatthew Ahrens 1688fa9e4066Sahrens refcount_init(); 1689fa9e4066Sahrens unique_init(); 1690fa9e4066Sahrens zio_init(); 1691fa9e4066Sahrens dmu_init(); 1692fa9e4066Sahrens zil_init(); 169387db74c1Sek vdev_cache_stat_init(); 169491ebeef5Sahrens zfs_prop_init(); 1695990b4856Slling zpool_prop_init(); 1696ad135b5dSChristopher Siden zpool_feature_init(); 1697fa9e4066Sahrens spa_config_load(); 1698e14bb325SJeff Bonwick l2arc_start(); 1699fa9e4066Sahrens } 1700fa9e4066Sahrens 1701fa9e4066Sahrens void 1702fa9e4066Sahrens spa_fini(void) 1703fa9e4066Sahrens { 1704e14bb325SJeff Bonwick l2arc_stop(); 1705e14bb325SJeff Bonwick 1706fa9e4066Sahrens spa_evict_all(); 1707fa9e4066Sahrens 170887db74c1Sek vdev_cache_stat_fini(); 1709fa9e4066Sahrens zil_fini(); 1710fa9e4066Sahrens dmu_fini(); 1711fa9e4066Sahrens zio_fini(); 171291ebeef5Sahrens unique_fini(); 1713fa9e4066Sahrens refcount_fini(); 1714fa9e4066Sahrens 1715fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 171699653d4eSeschrock avl_destroy(&spa_spare_avl); 1717fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 1718fa9e4066Sahrens 1719fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1720fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1721c25056deSgw mutex_destroy(&spa_spare_lock); 1722fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 1723fa9e4066Sahrens } 17246ce0521aSperrin 17256ce0521aSperrin /* 17266ce0521aSperrin * Return whether this pool has slogs. No locking needed. 17276ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 17286ce0521aSperrin * performance and not correctness 17296ce0521aSperrin */ 17306ce0521aSperrin boolean_t 17316ce0521aSperrin spa_has_slogs(spa_t *spa) 17326ce0521aSperrin { 17336ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 17346ce0521aSperrin } 1735bf82a41bSeschrock 1736b24ab676SJeff Bonwick spa_log_state_t 1737b24ab676SJeff Bonwick spa_get_log_state(spa_t *spa) 1738b24ab676SJeff Bonwick { 1739b24ab676SJeff Bonwick return (spa->spa_log_state); 1740b24ab676SJeff Bonwick } 1741b24ab676SJeff Bonwick 1742b24ab676SJeff Bonwick void 1743b24ab676SJeff Bonwick spa_set_log_state(spa_t *spa, spa_log_state_t state) 1744b24ab676SJeff Bonwick { 1745b24ab676SJeff Bonwick spa->spa_log_state = state; 1746b24ab676SJeff Bonwick } 1747b24ab676SJeff Bonwick 1748bf82a41bSeschrock boolean_t 1749bf82a41bSeschrock spa_is_root(spa_t *spa) 1750bf82a41bSeschrock { 1751bf82a41bSeschrock return (spa->spa_is_root); 1752bf82a41bSeschrock } 17538ad4d6ddSJeff Bonwick 17548ad4d6ddSJeff Bonwick boolean_t 17558ad4d6ddSJeff Bonwick spa_writeable(spa_t *spa) 17568ad4d6ddSJeff Bonwick { 17578ad4d6ddSJeff Bonwick return (!!(spa->spa_mode & FWRITE)); 17588ad4d6ddSJeff Bonwick } 17598ad4d6ddSJeff Bonwick 17608ad4d6ddSJeff Bonwick int 17618ad4d6ddSJeff Bonwick spa_mode(spa_t *spa) 17628ad4d6ddSJeff Bonwick { 17638ad4d6ddSJeff Bonwick return (spa->spa_mode); 17648ad4d6ddSJeff Bonwick } 1765b24ab676SJeff Bonwick 1766b24ab676SJeff Bonwick uint64_t 1767b24ab676SJeff Bonwick spa_bootfs(spa_t *spa) 1768b24ab676SJeff Bonwick { 1769b24ab676SJeff Bonwick return (spa->spa_bootfs); 1770b24ab676SJeff Bonwick } 1771b24ab676SJeff Bonwick 1772b24ab676SJeff Bonwick uint64_t 1773b24ab676SJeff Bonwick spa_delegation(spa_t *spa) 1774b24ab676SJeff Bonwick { 1775b24ab676SJeff Bonwick return (spa->spa_delegation); 1776b24ab676SJeff Bonwick } 1777b24ab676SJeff Bonwick 1778b24ab676SJeff Bonwick objset_t * 1779b24ab676SJeff Bonwick spa_meta_objset(spa_t *spa) 1780b24ab676SJeff Bonwick { 1781b24ab676SJeff Bonwick return (spa->spa_meta_objset); 1782b24ab676SJeff Bonwick } 1783b24ab676SJeff Bonwick 1784b24ab676SJeff Bonwick enum zio_checksum 1785b24ab676SJeff Bonwick spa_dedup_checksum(spa_t *spa) 1786b24ab676SJeff Bonwick { 1787b24ab676SJeff Bonwick return (spa->spa_dedup_checksum); 1788b24ab676SJeff Bonwick } 17893f9d6ad7SLin Ling 17903f9d6ad7SLin Ling /* 17913f9d6ad7SLin Ling * Reset pool scan stat per scan pass (or reboot). 17923f9d6ad7SLin Ling */ 17933f9d6ad7SLin Ling void 17943f9d6ad7SLin Ling spa_scan_stat_init(spa_t *spa) 17953f9d6ad7SLin Ling { 17963f9d6ad7SLin Ling /* data not stored on disk */ 17973f9d6ad7SLin Ling spa->spa_scan_pass_start = gethrestime_sec(); 17983f9d6ad7SLin Ling spa->spa_scan_pass_exam = 0; 17993f9d6ad7SLin Ling vdev_scan_stat_init(spa->spa_root_vdev); 18003f9d6ad7SLin Ling } 18013f9d6ad7SLin Ling 18023f9d6ad7SLin Ling /* 18033f9d6ad7SLin Ling * Get scan stats for zpool status reports 18043f9d6ad7SLin Ling */ 18053f9d6ad7SLin Ling int 18063f9d6ad7SLin Ling spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) 18073f9d6ad7SLin Ling { 18083f9d6ad7SLin Ling dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; 18093f9d6ad7SLin Ling 18103f9d6ad7SLin Ling if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) 18113f9d6ad7SLin Ling return (ENOENT); 18123f9d6ad7SLin Ling bzero(ps, sizeof (pool_scan_stat_t)); 18133f9d6ad7SLin Ling 18143f9d6ad7SLin Ling /* data stored on disk */ 18153f9d6ad7SLin Ling ps->pss_func = scn->scn_phys.scn_func; 18163f9d6ad7SLin Ling ps->pss_start_time = scn->scn_phys.scn_start_time; 18173f9d6ad7SLin Ling ps->pss_end_time = scn->scn_phys.scn_end_time; 18183f9d6ad7SLin Ling ps->pss_to_examine = scn->scn_phys.scn_to_examine; 18193f9d6ad7SLin Ling ps->pss_examined = scn->scn_phys.scn_examined; 18203f9d6ad7SLin Ling ps->pss_to_process = scn->scn_phys.scn_to_process; 18213f9d6ad7SLin Ling ps->pss_processed = scn->scn_phys.scn_processed; 18223f9d6ad7SLin Ling ps->pss_errors = scn->scn_phys.scn_errors; 18233f9d6ad7SLin Ling ps->pss_state = scn->scn_phys.scn_state; 18243f9d6ad7SLin Ling 18253f9d6ad7SLin Ling /* data not stored on disk */ 18263f9d6ad7SLin Ling ps->pss_pass_start = spa->spa_scan_pass_start; 18273f9d6ad7SLin Ling ps->pss_pass_exam = spa->spa_scan_pass_exam; 18283f9d6ad7SLin Ling 18293f9d6ad7SLin Ling return (0); 18303f9d6ad7SLin Ling } 183109c9d376SGeorge Wilson 183209c9d376SGeorge Wilson boolean_t 183309c9d376SGeorge Wilson spa_debug_enabled(spa_t *spa) 183409c9d376SGeorge Wilson { 183509c9d376SGeorge Wilson return (spa->spa_debug); 183609c9d376SGeorge Wilson } 1837