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 /* 2287db74c1Sek * Copyright 2008 Sun Microsystems, Inc. All rights reserved. 23fa9e4066Sahrens * Use is subject to license terms. 24fa9e4066Sahrens */ 25fa9e4066Sahrens 26fa9e4066Sahrens #pragma ident "%Z%%M% %I% %E% SMI" 27fa9e4066Sahrens 28fa9e4066Sahrens #include <sys/zfs_context.h> 29fa9e4066Sahrens #include <sys/spa_impl.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> 46fa9e4066Sahrens #include <sys/fs/zfs.h> 476ce0521aSperrin #include <sys/metaslab_impl.h> 4891ebeef5Sahrens #include "zfs_prop.h" 49fa9e4066Sahrens 50fa9e4066Sahrens /* 51fa9e4066Sahrens * SPA locking 52fa9e4066Sahrens * 53fa9e4066Sahrens * There are four basic locks for managing spa_t structures: 54fa9e4066Sahrens * 55fa9e4066Sahrens * spa_namespace_lock (global mutex) 56fa9e4066Sahrens * 5744cd46caSbillm * This lock must be acquired to do any of the following: 58fa9e4066Sahrens * 5944cd46caSbillm * - Lookup a spa_t by name 6044cd46caSbillm * - Add or remove a spa_t from the namespace 6144cd46caSbillm * - Increase spa_refcount from non-zero 6244cd46caSbillm * - Check if spa_refcount is zero 6344cd46caSbillm * - Rename a spa_t 64ea8dc4b6Seschrock * - add/remove/attach/detach devices 6544cd46caSbillm * - Held for the duration of create/destroy/import/export 66fa9e4066Sahrens * 6744cd46caSbillm * It does not need to handle recursion. A create or destroy may 6844cd46caSbillm * reference objects (files or zvols) in other pools, but by 6944cd46caSbillm * definition they must have an existing reference, and will never need 7044cd46caSbillm * to lookup a spa_t by name. 71fa9e4066Sahrens * 72fa9e4066Sahrens * spa_refcount (per-spa refcount_t protected by mutex) 73fa9e4066Sahrens * 7444cd46caSbillm * This reference count keep track of any active users of the spa_t. The 7544cd46caSbillm * spa_t cannot be destroyed or freed while this is non-zero. Internally, 7644cd46caSbillm * the refcount is never really 'zero' - opening a pool implicitly keeps 7744cd46caSbillm * some references in the DMU. Internally we check against SPA_MINREF, but 7844cd46caSbillm * present the image of a zero/non-zero value to consumers. 79fa9e4066Sahrens * 8091ebeef5Sahrens * spa_config_lock (per-spa read-priority rwlock) 81fa9e4066Sahrens * 8291ebeef5Sahrens * This protects the spa_t from config changes, and must be held in 8391ebeef5Sahrens * the following circumstances: 84fa9e4066Sahrens * 8544cd46caSbillm * - RW_READER to perform I/O to the spa 8644cd46caSbillm * - RW_WRITER to change the vdev config 87fa9e4066Sahrens * 88fa9e4066Sahrens * spa_config_cache_lock (per-spa mutex) 89fa9e4066Sahrens * 9044cd46caSbillm * This mutex prevents the spa_config nvlist from being updated. No 91fa9e4066Sahrens * other locks are required to obtain this lock, although implicitly you 92fa9e4066Sahrens * must have the namespace lock or non-zero refcount to have any kind 93fa9e4066Sahrens * of spa_t pointer at all. 94fa9e4066Sahrens * 95fa9e4066Sahrens * The locking order is fairly straightforward: 96fa9e4066Sahrens * 9744cd46caSbillm * spa_namespace_lock -> spa_refcount 98fa9e4066Sahrens * 9944cd46caSbillm * The namespace lock must be acquired to increase the refcount from 0 10044cd46caSbillm * or to check if it is zero. 101fa9e4066Sahrens * 10244cd46caSbillm * spa_refcount -> spa_config_lock 103fa9e4066Sahrens * 10444cd46caSbillm * There must be at least one valid reference on the spa_t to acquire 10544cd46caSbillm * the config lock. 106fa9e4066Sahrens * 10744cd46caSbillm * spa_namespace_lock -> spa_config_lock 108fa9e4066Sahrens * 10944cd46caSbillm * The namespace lock must always be taken before the config lock. 110fa9e4066Sahrens * 111fa9e4066Sahrens * 112fa9e4066Sahrens * The spa_namespace_lock and spa_config_cache_lock can be acquired directly and 113fa9e4066Sahrens * are globally visible. 114fa9e4066Sahrens * 115fa9e4066Sahrens * The namespace is manipulated using the following functions, all which require 116fa9e4066Sahrens * the spa_namespace_lock to be held. 117fa9e4066Sahrens * 11844cd46caSbillm * spa_lookup() Lookup a spa_t by name. 119fa9e4066Sahrens * 12044cd46caSbillm * spa_add() Create a new spa_t in the namespace. 121fa9e4066Sahrens * 12244cd46caSbillm * spa_remove() Remove a spa_t from the namespace. This also 12344cd46caSbillm * frees up any memory associated with the spa_t. 124fa9e4066Sahrens * 12544cd46caSbillm * spa_next() Returns the next spa_t in the system, or the 12644cd46caSbillm * first if NULL is passed. 127fa9e4066Sahrens * 12844cd46caSbillm * spa_evict_all() Shutdown and remove all spa_t structures in 12944cd46caSbillm * the system. 130fa9e4066Sahrens * 131ea8dc4b6Seschrock * spa_guid_exists() Determine whether a pool/device guid exists. 132fa9e4066Sahrens * 133fa9e4066Sahrens * The spa_refcount is manipulated using the following functions: 134fa9e4066Sahrens * 13544cd46caSbillm * spa_open_ref() Adds a reference to the given spa_t. Must be 13644cd46caSbillm * called with spa_namespace_lock held if the 13744cd46caSbillm * refcount is currently zero. 138fa9e4066Sahrens * 13944cd46caSbillm * spa_close() Remove a reference from the spa_t. This will 14044cd46caSbillm * not free the spa_t or remove it from the 14144cd46caSbillm * namespace. No locking is required. 142fa9e4066Sahrens * 14344cd46caSbillm * spa_refcount_zero() Returns true if the refcount is currently 14444cd46caSbillm * zero. Must be called with spa_namespace_lock 14544cd46caSbillm * held. 146fa9e4066Sahrens * 147e05725b1Sbonwick * The spa_config_lock is a form of rwlock. It must be held as RW_READER 148e05725b1Sbonwick * to perform I/O to the pool, and as RW_WRITER to change the vdev config. 149e05725b1Sbonwick * The spa_config_lock is manipulated with spa_config_{enter,exit,held}(). 150fa9e4066Sahrens * 151ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 152fa9e4066Sahrens * 15344cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 154ea8dc4b6Seschrock * for writing. 155fa9e4066Sahrens * 15644cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 15744cd46caSbillm * to complete, sync the updated configs to the 158ea8dc4b6Seschrock * cache, and release the namespace lock. 159fa9e4066Sahrens * 160fa9e4066Sahrens * The spa_name() function also requires either the spa_namespace_lock 161fa9e4066Sahrens * or the spa_config_lock, as both are needed to do a rename. spa_rename() is 162fa9e4066Sahrens * also implemented within this file since is requires manipulation of the 163fa9e4066Sahrens * namespace. 164fa9e4066Sahrens */ 165fa9e4066Sahrens 166fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 167fa9e4066Sahrens kmutex_t spa_namespace_lock; 168fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 1690373e76bSbonwick static int spa_active_count; 170416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 171fa9e4066Sahrens 17299653d4eSeschrock static kmutex_t spa_spare_lock; 17339c23413Seschrock static avl_tree_t spa_spare_avl; 174fa94a07fSbrendan static kmutex_t spa_l2cache_lock; 175fa94a07fSbrendan static avl_tree_t spa_l2cache_avl; 17699653d4eSeschrock 177fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 178fa9e4066Sahrens int spa_mode; 179fa9e4066Sahrens 180fa9e4066Sahrens #ifdef ZFS_DEBUG 18140feaa91Sahrens /* Everything except dprintf is on by default in debug builds */ 18240feaa91Sahrens int zfs_flags = ~ZFS_DEBUG_DPRINTF; 183fa9e4066Sahrens #else 184fa9e4066Sahrens int zfs_flags = 0; 185fa9e4066Sahrens #endif 186fa9e4066Sahrens 1870125049cSahrens /* 1880125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 1890125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 1900125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 1910125049cSahrens */ 1920125049cSahrens int zfs_recover = 0; 1930125049cSahrens 194fa9e4066Sahrens #define SPA_MINREF 5 /* spa_refcnt for an open-but-idle pool */ 195fa9e4066Sahrens 196e05725b1Sbonwick /* 197e05725b1Sbonwick * ========================================================================== 198e05725b1Sbonwick * SPA config locking 199e05725b1Sbonwick * ========================================================================== 200e05725b1Sbonwick */ 201e05725b1Sbonwick static void 202e05725b1Sbonwick spa_config_lock_init(spa_config_lock_t *scl) 203e05725b1Sbonwick { 204e05725b1Sbonwick mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); 205e05725b1Sbonwick scl->scl_writer = NULL; 206e05725b1Sbonwick cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); 207e05725b1Sbonwick refcount_create(&scl->scl_count); 208e05725b1Sbonwick } 209e05725b1Sbonwick 210e05725b1Sbonwick static void 211e05725b1Sbonwick spa_config_lock_destroy(spa_config_lock_t *scl) 212e05725b1Sbonwick { 213e05725b1Sbonwick mutex_destroy(&scl->scl_lock); 214e05725b1Sbonwick ASSERT(scl->scl_writer == NULL); 215e05725b1Sbonwick cv_destroy(&scl->scl_cv); 216e05725b1Sbonwick refcount_destroy(&scl->scl_count); 217e05725b1Sbonwick } 218e05725b1Sbonwick 219e05725b1Sbonwick void 220e05725b1Sbonwick spa_config_enter(spa_t *spa, krw_t rw, void *tag) 221e05725b1Sbonwick { 222e05725b1Sbonwick spa_config_lock_t *scl = &spa->spa_config_lock; 223e05725b1Sbonwick 224e05725b1Sbonwick mutex_enter(&scl->scl_lock); 225e05725b1Sbonwick 226e05725b1Sbonwick if (rw == RW_READER) { 227e05725b1Sbonwick while (scl->scl_writer != NULL && scl->scl_writer != curthread) 228e05725b1Sbonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 229e05725b1Sbonwick } else { 230e05725b1Sbonwick while (!refcount_is_zero(&scl->scl_count) && 231e05725b1Sbonwick scl->scl_writer != curthread) 232e05725b1Sbonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 233e05725b1Sbonwick scl->scl_writer = curthread; 234e05725b1Sbonwick } 235e05725b1Sbonwick 236e05725b1Sbonwick (void) refcount_add(&scl->scl_count, tag); 237e05725b1Sbonwick 238e05725b1Sbonwick mutex_exit(&scl->scl_lock); 239e05725b1Sbonwick } 240e05725b1Sbonwick 241e05725b1Sbonwick void 242e05725b1Sbonwick spa_config_exit(spa_t *spa, void *tag) 243e05725b1Sbonwick { 244e05725b1Sbonwick spa_config_lock_t *scl = &spa->spa_config_lock; 245e05725b1Sbonwick 246e05725b1Sbonwick mutex_enter(&scl->scl_lock); 247e05725b1Sbonwick 248e05725b1Sbonwick ASSERT(!refcount_is_zero(&scl->scl_count)); 249e05725b1Sbonwick 250e05725b1Sbonwick if (refcount_remove(&scl->scl_count, tag) == 0) { 251e05725b1Sbonwick cv_broadcast(&scl->scl_cv); 252e05725b1Sbonwick ASSERT(scl->scl_writer == NULL || scl->scl_writer == curthread); 253e05725b1Sbonwick scl->scl_writer = NULL; /* OK in either case */ 254e05725b1Sbonwick } 255e05725b1Sbonwick 256e05725b1Sbonwick mutex_exit(&scl->scl_lock); 257e05725b1Sbonwick } 258e05725b1Sbonwick 259e05725b1Sbonwick boolean_t 260e05725b1Sbonwick spa_config_held(spa_t *spa, krw_t rw) 261e05725b1Sbonwick { 262e05725b1Sbonwick spa_config_lock_t *scl = &spa->spa_config_lock; 263e05725b1Sbonwick 264e05725b1Sbonwick if (rw == RW_READER) 265e05725b1Sbonwick return (!refcount_is_zero(&scl->scl_count)); 266e05725b1Sbonwick else 267e05725b1Sbonwick return (scl->scl_writer == curthread); 268e05725b1Sbonwick } 269e05725b1Sbonwick 270fa9e4066Sahrens /* 271fa9e4066Sahrens * ========================================================================== 272fa9e4066Sahrens * SPA namespace functions 273fa9e4066Sahrens * ========================================================================== 274fa9e4066Sahrens */ 275fa9e4066Sahrens 276fa9e4066Sahrens /* 277fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 278fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 279fa9e4066Sahrens */ 280fa9e4066Sahrens spa_t * 281fa9e4066Sahrens spa_lookup(const char *name) 282fa9e4066Sahrens { 283fa9e4066Sahrens spa_t search, *spa; 284fa9e4066Sahrens avl_index_t where; 28540feaa91Sahrens char c; 28640feaa91Sahrens char *cp; 287fa9e4066Sahrens 288fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 289fa9e4066Sahrens 29040feaa91Sahrens /* 29140feaa91Sahrens * If it's a full dataset name, figure out the pool name and 29240feaa91Sahrens * just use that. 29340feaa91Sahrens */ 29440feaa91Sahrens cp = strpbrk(name, "/@"); 29540feaa91Sahrens if (cp) { 29640feaa91Sahrens c = *cp; 29740feaa91Sahrens *cp = '\0'; 29840feaa91Sahrens } 29940feaa91Sahrens 300fa9e4066Sahrens search.spa_name = (char *)name; 301fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 302fa9e4066Sahrens 30340feaa91Sahrens if (cp) 30440feaa91Sahrens *cp = c; 30540feaa91Sahrens 306fa9e4066Sahrens return (spa); 307fa9e4066Sahrens } 308fa9e4066Sahrens 309fa9e4066Sahrens /* 310fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 311fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 312fa9e4066Sahrens * exist by calling spa_lookup() first. 313fa9e4066Sahrens */ 314fa9e4066Sahrens spa_t * 3150373e76bSbonwick spa_add(const char *name, const char *altroot) 316fa9e4066Sahrens { 317fa9e4066Sahrens spa_t *spa; 318*c5904d13Seschrock spa_config_dirent_t *dp; 319fa9e4066Sahrens 320fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 321fa9e4066Sahrens 322fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 323fa9e4066Sahrens 324c25056deSgw rw_init(&spa->spa_traverse_lock, NULL, RW_DEFAULT, NULL); 325c25056deSgw 326c25056deSgw mutex_init(&spa->spa_uberblock_lock, NULL, MUTEX_DEFAULT, NULL); 327c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 328c25056deSgw mutex_init(&spa->spa_config_cache_lock, NULL, MUTEX_DEFAULT, NULL); 329c25056deSgw mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 330c25056deSgw mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 331c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 332c25056deSgw mutex_init(&spa->spa_sync_bplist.bpl_lock, NULL, MUTEX_DEFAULT, NULL); 333c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 334c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 335c25056deSgw 336c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 337c25056deSgw cv_init(&spa->spa_scrub_cv, NULL, CV_DEFAULT, NULL); 338c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 339c25056deSgw 340fa9e4066Sahrens spa->spa_name = spa_strdup(name); 341fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 342fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 3430373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 344fa9e4066Sahrens 345fa9e4066Sahrens refcount_create(&spa->spa_refcount); 346e05725b1Sbonwick spa_config_lock_init(&spa->spa_config_lock); 347fa9e4066Sahrens 348fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 349fa9e4066Sahrens 3500a4e9518Sgw mutex_init(&spa->spa_zio_lock, NULL, MUTEX_DEFAULT, NULL); 3510a4e9518Sgw 3520373e76bSbonwick /* 3530373e76bSbonwick * Set the alternate root, if there is one. 3540373e76bSbonwick */ 3550373e76bSbonwick if (altroot) { 3560373e76bSbonwick spa->spa_root = spa_strdup(altroot); 3570373e76bSbonwick spa_active_count++; 3580373e76bSbonwick } 3590373e76bSbonwick 360*c5904d13Seschrock /* 361*c5904d13Seschrock * Every pool starts with the default cachefile 362*c5904d13Seschrock */ 363*c5904d13Seschrock list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), 364*c5904d13Seschrock offsetof(spa_config_dirent_t, scd_link)); 365*c5904d13Seschrock 366*c5904d13Seschrock dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); 367*c5904d13Seschrock dp->scd_path = spa_strdup(spa_config_path); 368*c5904d13Seschrock list_insert_head(&spa->spa_config_list, dp); 369*c5904d13Seschrock 370fa9e4066Sahrens return (spa); 371fa9e4066Sahrens } 372fa9e4066Sahrens 373fa9e4066Sahrens /* 374fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 375fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 376fa9e4066Sahrens * deactivated. 377fa9e4066Sahrens */ 378fa9e4066Sahrens void 379fa9e4066Sahrens spa_remove(spa_t *spa) 380fa9e4066Sahrens { 381*c5904d13Seschrock spa_config_dirent_t *dp; 382*c5904d13Seschrock 383fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 384fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 385fa9e4066Sahrens ASSERT(spa->spa_scrub_thread == NULL); 386fa9e4066Sahrens 387fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 388fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 389fa9e4066Sahrens 3900373e76bSbonwick if (spa->spa_root) { 391fa9e4066Sahrens spa_strfree(spa->spa_root); 3920373e76bSbonwick spa_active_count--; 3930373e76bSbonwick } 394fa9e4066Sahrens 395fa9e4066Sahrens if (spa->spa_name) 396fa9e4066Sahrens spa_strfree(spa->spa_name); 397fa9e4066Sahrens 398*c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 399*c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 400*c5904d13Seschrock if (dp->scd_path != NULL) 401*c5904d13Seschrock spa_strfree(dp->scd_path); 402*c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 403*c5904d13Seschrock } 404*c5904d13Seschrock 405*c5904d13Seschrock list_destroy(&spa->spa_config_list); 4062f8aaab3Seschrock 407fa9e4066Sahrens spa_config_set(spa, NULL); 408fa9e4066Sahrens 409fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 41091ebeef5Sahrens 411e05725b1Sbonwick spa_config_lock_destroy(&spa->spa_config_lock); 412fa9e4066Sahrens 413c25056deSgw rw_destroy(&spa->spa_traverse_lock); 414c25056deSgw 415c25056deSgw cv_destroy(&spa->spa_async_cv); 416c25056deSgw cv_destroy(&spa->spa_scrub_cv); 417c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 418c25056deSgw 419c25056deSgw mutex_destroy(&spa->spa_uberblock_lock); 4205ad82045Snd mutex_destroy(&spa->spa_async_lock); 421c25056deSgw mutex_destroy(&spa->spa_config_cache_lock); 422c25056deSgw mutex_destroy(&spa->spa_scrub_lock); 423c25056deSgw mutex_destroy(&spa->spa_errlog_lock); 424c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 425c25056deSgw mutex_destroy(&spa->spa_sync_bplist.bpl_lock); 42606eeb2adSek mutex_destroy(&spa->spa_history_lock); 427b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 4280a4e9518Sgw mutex_destroy(&spa->spa_zio_lock); 4295ad82045Snd 430fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 431fa9e4066Sahrens } 432fa9e4066Sahrens 433fa9e4066Sahrens /* 434fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 435fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 436fa9e4066Sahrens */ 437fa9e4066Sahrens spa_t * 438fa9e4066Sahrens spa_next(spa_t *prev) 439fa9e4066Sahrens { 440fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 441fa9e4066Sahrens 442fa9e4066Sahrens if (prev) 443fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 444fa9e4066Sahrens else 445fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 446fa9e4066Sahrens } 447fa9e4066Sahrens 448fa9e4066Sahrens /* 449fa9e4066Sahrens * ========================================================================== 450fa9e4066Sahrens * SPA refcount functions 451fa9e4066Sahrens * ========================================================================== 452fa9e4066Sahrens */ 453fa9e4066Sahrens 454fa9e4066Sahrens /* 455fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 456fa9e4066Sahrens * have the namespace lock held. 457fa9e4066Sahrens */ 458fa9e4066Sahrens void 459fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 460fa9e4066Sahrens { 461fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 462fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 463fa9e4066Sahrens 464fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 465fa9e4066Sahrens } 466fa9e4066Sahrens 467fa9e4066Sahrens /* 468fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 469fa9e4066Sahrens * have the namespace lock held. 470fa9e4066Sahrens */ 471fa9e4066Sahrens void 472fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 473fa9e4066Sahrens { 474fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 475fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 476fa9e4066Sahrens 477fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 478fa9e4066Sahrens } 479fa9e4066Sahrens 480fa9e4066Sahrens /* 481fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 482fa9e4066Sahrens * spa_namespace_lock held. We really compare against SPA_MINREF, which is the 483fa9e4066Sahrens * number of references acquired when opening a pool 484fa9e4066Sahrens */ 485fa9e4066Sahrens boolean_t 486fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 487fa9e4066Sahrens { 488fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 489fa9e4066Sahrens 490fa9e4066Sahrens return (refcount_count(&spa->spa_refcount) == SPA_MINREF); 491fa9e4066Sahrens } 492fa9e4066Sahrens 49399653d4eSeschrock /* 49499653d4eSeschrock * ========================================================================== 495fa94a07fSbrendan * SPA spare and l2cache tracking 49699653d4eSeschrock * ========================================================================== 49799653d4eSeschrock */ 49899653d4eSeschrock 499fa94a07fSbrendan /* 500fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 501fa94a07fSbrendan * for 'auxiliary' devices. 502fa94a07fSbrendan */ 503fa94a07fSbrendan 504fa94a07fSbrendan typedef struct spa_aux { 505fa94a07fSbrendan uint64_t aux_guid; 506fa94a07fSbrendan uint64_t aux_pool; 507fa94a07fSbrendan avl_node_t aux_avl; 508fa94a07fSbrendan int aux_count; 509fa94a07fSbrendan } spa_aux_t; 510fa94a07fSbrendan 511fa94a07fSbrendan static int 512fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 513fa94a07fSbrendan { 514fa94a07fSbrendan const spa_aux_t *sa = a; 515fa94a07fSbrendan const spa_aux_t *sb = b; 516fa94a07fSbrendan 517fa94a07fSbrendan if (sa->aux_guid < sb->aux_guid) 518fa94a07fSbrendan return (-1); 519fa94a07fSbrendan else if (sa->aux_guid > sb->aux_guid) 520fa94a07fSbrendan return (1); 521fa94a07fSbrendan else 522fa94a07fSbrendan return (0); 523fa94a07fSbrendan } 524fa94a07fSbrendan 525fa94a07fSbrendan void 526fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 527fa94a07fSbrendan { 528fa94a07fSbrendan avl_index_t where; 529fa94a07fSbrendan spa_aux_t search; 530fa94a07fSbrendan spa_aux_t *aux; 531fa94a07fSbrendan 532fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 533fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 534fa94a07fSbrendan aux->aux_count++; 535fa94a07fSbrendan } else { 536fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 537fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 538fa94a07fSbrendan aux->aux_count = 1; 539fa94a07fSbrendan avl_insert(avl, aux, where); 540fa94a07fSbrendan } 541fa94a07fSbrendan } 542fa94a07fSbrendan 543fa94a07fSbrendan void 544fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 545fa94a07fSbrendan { 546fa94a07fSbrendan spa_aux_t search; 547fa94a07fSbrendan spa_aux_t *aux; 548fa94a07fSbrendan avl_index_t where; 549fa94a07fSbrendan 550fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 551fa94a07fSbrendan aux = avl_find(avl, &search, &where); 552fa94a07fSbrendan 553fa94a07fSbrendan ASSERT(aux != NULL); 554fa94a07fSbrendan 555fa94a07fSbrendan if (--aux->aux_count == 0) { 556fa94a07fSbrendan avl_remove(avl, aux); 557fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 558fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 559fa94a07fSbrendan aux->aux_pool = 0ULL; 560fa94a07fSbrendan } 561fa94a07fSbrendan } 562fa94a07fSbrendan 563fa94a07fSbrendan boolean_t 564fa94a07fSbrendan spa_aux_exists(uint64_t guid, uint64_t *pool, avl_tree_t *avl) 565fa94a07fSbrendan { 566fa94a07fSbrendan spa_aux_t search, *found; 567fa94a07fSbrendan avl_index_t where; 568fa94a07fSbrendan 569fa94a07fSbrendan search.aux_guid = guid; 570fa94a07fSbrendan found = avl_find(avl, &search, &where); 571fa94a07fSbrendan 572fa94a07fSbrendan if (pool) { 573fa94a07fSbrendan if (found) 574fa94a07fSbrendan *pool = found->aux_pool; 575fa94a07fSbrendan else 576fa94a07fSbrendan *pool = 0ULL; 577fa94a07fSbrendan } 578fa94a07fSbrendan 579fa94a07fSbrendan return (found != NULL); 580fa94a07fSbrendan } 581fa94a07fSbrendan 582fa94a07fSbrendan void 583fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 584fa94a07fSbrendan { 585fa94a07fSbrendan spa_aux_t search, *found; 586fa94a07fSbrendan avl_index_t where; 587fa94a07fSbrendan 588fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 589fa94a07fSbrendan found = avl_find(avl, &search, &where); 590fa94a07fSbrendan ASSERT(found != NULL); 591fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 592fa94a07fSbrendan 593fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 594fa94a07fSbrendan } 595fa94a07fSbrendan 59699653d4eSeschrock /* 59739c23413Seschrock * Spares are tracked globally due to the following constraints: 59839c23413Seschrock * 59939c23413Seschrock * - A spare may be part of multiple pools. 60039c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 60139c23413Seschrock * another pool. 60239c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 60339c23413Seschrock * the target is a spare in the same pool. 60439c23413Seschrock * 60539c23413Seschrock * We keep track of all spares on the system through the use of a reference 60639c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 60739c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 60839c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 60939c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 61039c23413Seschrock * pool), we also keep track of which pool its been made a part of. 61139c23413Seschrock * 61239c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 61339c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 61439c23413Seschrock * separate spare lock exists for the status query path, which does not need to 61539c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 61699653d4eSeschrock */ 61739c23413Seschrock 61899653d4eSeschrock static int 61999653d4eSeschrock spa_spare_compare(const void *a, const void *b) 62099653d4eSeschrock { 621fa94a07fSbrendan return (spa_aux_compare(a, b)); 62299653d4eSeschrock } 62399653d4eSeschrock 62499653d4eSeschrock void 62539c23413Seschrock spa_spare_add(vdev_t *vd) 62699653d4eSeschrock { 62799653d4eSeschrock mutex_enter(&spa_spare_lock); 62839c23413Seschrock ASSERT(!vd->vdev_isspare); 629fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 63039c23413Seschrock vd->vdev_isspare = B_TRUE; 63199653d4eSeschrock mutex_exit(&spa_spare_lock); 63299653d4eSeschrock } 63399653d4eSeschrock 63499653d4eSeschrock void 63539c23413Seschrock spa_spare_remove(vdev_t *vd) 63699653d4eSeschrock { 63799653d4eSeschrock mutex_enter(&spa_spare_lock); 63839c23413Seschrock ASSERT(vd->vdev_isspare); 639fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 64039c23413Seschrock vd->vdev_isspare = B_FALSE; 64199653d4eSeschrock mutex_exit(&spa_spare_lock); 64299653d4eSeschrock } 64399653d4eSeschrock 64499653d4eSeschrock boolean_t 64539c23413Seschrock spa_spare_exists(uint64_t guid, uint64_t *pool) 64699653d4eSeschrock { 647fa94a07fSbrendan boolean_t found; 64899653d4eSeschrock 64999653d4eSeschrock mutex_enter(&spa_spare_lock); 650fa94a07fSbrendan found = spa_aux_exists(guid, pool, &spa_spare_avl); 65199653d4eSeschrock mutex_exit(&spa_spare_lock); 65299653d4eSeschrock 653fa94a07fSbrendan return (found); 65439c23413Seschrock } 65539c23413Seschrock 65639c23413Seschrock void 65739c23413Seschrock spa_spare_activate(vdev_t *vd) 65839c23413Seschrock { 65939c23413Seschrock mutex_enter(&spa_spare_lock); 66039c23413Seschrock ASSERT(vd->vdev_isspare); 661fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 662fa94a07fSbrendan mutex_exit(&spa_spare_lock); 663fa94a07fSbrendan } 66439c23413Seschrock 665fa94a07fSbrendan /* 666fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 667fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 668fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 669fa94a07fSbrendan */ 67039c23413Seschrock 671fa94a07fSbrendan static int 672fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 673fa94a07fSbrendan { 674fa94a07fSbrendan return (spa_aux_compare(a, b)); 675fa94a07fSbrendan } 676fa94a07fSbrendan 677fa94a07fSbrendan void 678fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 679fa94a07fSbrendan { 680fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 681fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 682fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 683fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 684fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 685fa94a07fSbrendan } 686fa94a07fSbrendan 687fa94a07fSbrendan void 688fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 689fa94a07fSbrendan { 690fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 691fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 692fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 693fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 694fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 695fa94a07fSbrendan } 696fa94a07fSbrendan 697fa94a07fSbrendan boolean_t 698fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 699fa94a07fSbrendan { 700fa94a07fSbrendan boolean_t found; 701fa94a07fSbrendan 702fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 703fa94a07fSbrendan found = spa_aux_exists(guid, pool, &spa_l2cache_avl); 704fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 705fa94a07fSbrendan 706fa94a07fSbrendan return (found); 707fa94a07fSbrendan } 708fa94a07fSbrendan 709fa94a07fSbrendan void 710fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 711fa94a07fSbrendan { 712fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 713fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 714fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 715fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 716fa94a07fSbrendan } 717fa94a07fSbrendan 718fa94a07fSbrendan void 719fa94a07fSbrendan spa_l2cache_space_update(vdev_t *vd, int64_t space, int64_t alloc) 720fa94a07fSbrendan { 721fa94a07fSbrendan vdev_space_update(vd, space, alloc, B_FALSE); 72299653d4eSeschrock } 72399653d4eSeschrock 724fa9e4066Sahrens /* 725fa9e4066Sahrens * ========================================================================== 726fa9e4066Sahrens * SPA vdev locking 727fa9e4066Sahrens * ========================================================================== 728fa9e4066Sahrens */ 729fa9e4066Sahrens 730fa9e4066Sahrens /* 731ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 732ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 733fa9e4066Sahrens * It returns the next transaction group for the spa_t. 734fa9e4066Sahrens */ 735fa9e4066Sahrens uint64_t 736fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 737fa9e4066Sahrens { 7383d7072f8Seschrock mutex_enter(&spa_namespace_lock); 7393d7072f8Seschrock 740ea8dc4b6Seschrock /* 7413d7072f8Seschrock * Suspend scrub activity while we mess with the config. We must do 7423d7072f8Seschrock * this after acquiring the namespace lock to avoid a 3-way deadlock 7433d7072f8Seschrock * with spa_scrub_stop() and the scrub thread. 744ea8dc4b6Seschrock */ 745ea8dc4b6Seschrock spa_scrub_suspend(spa); 746fa9e4066Sahrens 747ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, spa); 748fa9e4066Sahrens 749fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 750fa9e4066Sahrens } 751fa9e4066Sahrens 752fa9e4066Sahrens /* 753fa9e4066Sahrens * Unlock the spa_t after adding or removing a vdev. Besides undoing the 754fa9e4066Sahrens * locking of spa_vdev_enter(), we also want make sure the transactions have 755fa9e4066Sahrens * synced to disk, and then update the global configuration cache with the new 756fa9e4066Sahrens * information. 757fa9e4066Sahrens */ 758fa9e4066Sahrens int 759fa9e4066Sahrens spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 760fa9e4066Sahrens { 7610e34b6a7Sbonwick int config_changed = B_FALSE; 762ea8dc4b6Seschrock 7630373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 7640e34b6a7Sbonwick 7650e34b6a7Sbonwick /* 7660e34b6a7Sbonwick * Reassess the DTLs. 7670e34b6a7Sbonwick */ 7680373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 7690e34b6a7Sbonwick 7700e34b6a7Sbonwick /* 7710373e76bSbonwick * If the config changed, notify the scrub thread that it must restart. 7720e34b6a7Sbonwick */ 7730e34b6a7Sbonwick if (error == 0 && !list_is_empty(&spa->spa_dirty_list)) { 7740e34b6a7Sbonwick config_changed = B_TRUE; 7750373e76bSbonwick spa_scrub_restart(spa, txg); 7760e34b6a7Sbonwick } 777ea8dc4b6Seschrock 778ea8dc4b6Seschrock spa_config_exit(spa, spa); 779fa9e4066Sahrens 780ea8dc4b6Seschrock /* 7815dabedeeSbonwick * Allow scrubbing to resume. 782ea8dc4b6Seschrock */ 783ea8dc4b6Seschrock spa_scrub_resume(spa); 784fa9e4066Sahrens 785fa9e4066Sahrens /* 786fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 787fa9e4066Sahrens * that there won't be more than one config change per txg. 788fa9e4066Sahrens * This allows us to use the txg as the generation number. 789fa9e4066Sahrens */ 790fa9e4066Sahrens if (error == 0) 791fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 792fa9e4066Sahrens 793fa9e4066Sahrens if (vd != NULL) { 794fa9e4066Sahrens ASSERT(!vd->vdev_detached || vd->vdev_dtl.smo_object == 0); 795fa9e4066Sahrens vdev_free(vd); 796fa9e4066Sahrens } 797fa9e4066Sahrens 798fa9e4066Sahrens /* 7990e34b6a7Sbonwick * If the config changed, update the config cache. 800fa9e4066Sahrens */ 8010e34b6a7Sbonwick if (config_changed) 802*c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 803ea8dc4b6Seschrock 804ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 805fa9e4066Sahrens 806fa9e4066Sahrens return (error); 807fa9e4066Sahrens } 808fa9e4066Sahrens 809fa9e4066Sahrens /* 810fa9e4066Sahrens * ========================================================================== 811fa9e4066Sahrens * Miscellaneous functions 812fa9e4066Sahrens * ========================================================================== 813fa9e4066Sahrens */ 814fa9e4066Sahrens 815fa9e4066Sahrens /* 816fa9e4066Sahrens * Rename a spa_t. 817fa9e4066Sahrens */ 818fa9e4066Sahrens int 819fa9e4066Sahrens spa_rename(const char *name, const char *newname) 820fa9e4066Sahrens { 821fa9e4066Sahrens spa_t *spa; 822fa9e4066Sahrens int err; 823fa9e4066Sahrens 824fa9e4066Sahrens /* 825fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 826fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 827fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 828ea8dc4b6Seschrock * allow recursive calls for other reasons. 829fa9e4066Sahrens */ 830fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 831fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 832fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 833fa9e4066Sahrens return (err); 834fa9e4066Sahrens } 835fa9e4066Sahrens 836ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 837fa9e4066Sahrens 838fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 839fa9e4066Sahrens spa_strfree(spa->spa_name); 840fa9e4066Sahrens spa->spa_name = spa_strdup(newname); 841fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 842fa9e4066Sahrens 843fa9e4066Sahrens /* 844fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 845fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 846fa9e4066Sahrens * during the sync. 847fa9e4066Sahrens */ 848fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 849fa9e4066Sahrens 850ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 851fa9e4066Sahrens 8520373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 853fa9e4066Sahrens 854fa9e4066Sahrens /* 855fa9e4066Sahrens * Sync the updated config cache. 856fa9e4066Sahrens */ 857*c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 858fa9e4066Sahrens 859fa9e4066Sahrens spa_close(spa, FTAG); 860fa9e4066Sahrens 861fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 862fa9e4066Sahrens 863fa9e4066Sahrens return (0); 864fa9e4066Sahrens } 865fa9e4066Sahrens 866fa9e4066Sahrens 867fa9e4066Sahrens /* 868fa9e4066Sahrens * Determine whether a pool with given pool_guid exists. If device_guid is 869fa9e4066Sahrens * non-zero, determine whether the pool exists *and* contains a device with the 870fa9e4066Sahrens * specified device_guid. 871fa9e4066Sahrens */ 872fa9e4066Sahrens boolean_t 873fa9e4066Sahrens spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 874fa9e4066Sahrens { 875fa9e4066Sahrens spa_t *spa; 876fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 877fa9e4066Sahrens 878ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 879fa9e4066Sahrens 880fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 881fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 882fa9e4066Sahrens continue; 883fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 884fa9e4066Sahrens continue; 88539c23413Seschrock if (spa_guid(spa) == pool_guid) { 88639c23413Seschrock if (device_guid == 0) 88739c23413Seschrock break; 88839c23413Seschrock 88939c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 89039c23413Seschrock device_guid) != NULL) 89139c23413Seschrock break; 89239c23413Seschrock 89339c23413Seschrock /* 8948654d025Sperrin * Check any devices we may be in the process of adding. 89539c23413Seschrock */ 89639c23413Seschrock if (spa->spa_pending_vdev) { 89739c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 89839c23413Seschrock device_guid) != NULL) 89939c23413Seschrock break; 90039c23413Seschrock } 90139c23413Seschrock } 902fa9e4066Sahrens } 903fa9e4066Sahrens 904fa9e4066Sahrens return (spa != NULL); 905fa9e4066Sahrens } 906fa9e4066Sahrens 907fa9e4066Sahrens char * 908fa9e4066Sahrens spa_strdup(const char *s) 909fa9e4066Sahrens { 910fa9e4066Sahrens size_t len; 911fa9e4066Sahrens char *new; 912fa9e4066Sahrens 913fa9e4066Sahrens len = strlen(s); 914fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 915fa9e4066Sahrens bcopy(s, new, len); 916fa9e4066Sahrens new[len] = '\0'; 917fa9e4066Sahrens 918fa9e4066Sahrens return (new); 919fa9e4066Sahrens } 920fa9e4066Sahrens 921fa9e4066Sahrens void 922fa9e4066Sahrens spa_strfree(char *s) 923fa9e4066Sahrens { 924fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 925fa9e4066Sahrens } 926fa9e4066Sahrens 927fa9e4066Sahrens uint64_t 928fa9e4066Sahrens spa_get_random(uint64_t range) 929fa9e4066Sahrens { 930fa9e4066Sahrens uint64_t r; 931fa9e4066Sahrens 932fa9e4066Sahrens ASSERT(range != 0); 933fa9e4066Sahrens 934fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 935fa9e4066Sahrens 936fa9e4066Sahrens return (r % range); 937fa9e4066Sahrens } 938fa9e4066Sahrens 939fa9e4066Sahrens void 940d80c45e0Sbonwick sprintf_blkptr(char *buf, int len, const blkptr_t *bp) 941fa9e4066Sahrens { 94244cd46caSbillm int d; 943fa9e4066Sahrens 944fa9e4066Sahrens if (bp == NULL) { 945fbabab8fSmaybee (void) snprintf(buf, len, "<NULL>"); 946fa9e4066Sahrens return; 947fa9e4066Sahrens } 948fa9e4066Sahrens 949fa9e4066Sahrens if (BP_IS_HOLE(bp)) { 950fbabab8fSmaybee (void) snprintf(buf, len, "<hole>"); 951fa9e4066Sahrens return; 952fa9e4066Sahrens } 953fa9e4066Sahrens 95444cd46caSbillm (void) snprintf(buf, len, "[L%llu %s] %llxL/%llxP ", 955fa9e4066Sahrens (u_longlong_t)BP_GET_LEVEL(bp), 956fa9e4066Sahrens dmu_ot[BP_GET_TYPE(bp)].ot_name, 957fa9e4066Sahrens (u_longlong_t)BP_GET_LSIZE(bp), 95844cd46caSbillm (u_longlong_t)BP_GET_PSIZE(bp)); 95944cd46caSbillm 96044cd46caSbillm for (d = 0; d < BP_GET_NDVAS(bp); d++) { 961d80c45e0Sbonwick const dva_t *dva = &bp->blk_dva[d]; 96244cd46caSbillm (void) snprintf(buf + strlen(buf), len - strlen(buf), 96344cd46caSbillm "DVA[%d]=<%llu:%llx:%llx> ", d, 96444cd46caSbillm (u_longlong_t)DVA_GET_VDEV(dva), 96544cd46caSbillm (u_longlong_t)DVA_GET_OFFSET(dva), 96644cd46caSbillm (u_longlong_t)DVA_GET_ASIZE(dva)); 96744cd46caSbillm } 96844cd46caSbillm 96944cd46caSbillm (void) snprintf(buf + strlen(buf), len - strlen(buf), 97044cd46caSbillm "%s %s %s %s birth=%llu fill=%llu cksum=%llx:%llx:%llx:%llx", 971fa9e4066Sahrens zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 972fa9e4066Sahrens zio_compress_table[BP_GET_COMPRESS(bp)].ci_name, 973fa9e4066Sahrens BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", 97444cd46caSbillm BP_IS_GANG(bp) ? "gang" : "contiguous", 975fa9e4066Sahrens (u_longlong_t)bp->blk_birth, 976fa9e4066Sahrens (u_longlong_t)bp->blk_fill, 977fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[0], 978fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[1], 979fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[2], 980fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[3]); 981fa9e4066Sahrens } 982fa9e4066Sahrens 983fa9e4066Sahrens void 984fa9e4066Sahrens spa_freeze(spa_t *spa) 985fa9e4066Sahrens { 986fa9e4066Sahrens uint64_t freeze_txg = 0; 987fa9e4066Sahrens 988ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 989fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 990fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 991fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 992fa9e4066Sahrens } 993ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 994fa9e4066Sahrens if (freeze_txg != 0) 995fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 996fa9e4066Sahrens } 997fa9e4066Sahrens 9980125049cSahrens void 9990125049cSahrens zfs_panic_recover(const char *fmt, ...) 10000125049cSahrens { 10010125049cSahrens va_list adx; 10020125049cSahrens 10030125049cSahrens va_start(adx, fmt); 10040125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 10050125049cSahrens va_end(adx); 10060125049cSahrens } 10070125049cSahrens 1008fa9e4066Sahrens /* 1009fa9e4066Sahrens * ========================================================================== 1010fa9e4066Sahrens * Accessor functions 1011fa9e4066Sahrens * ========================================================================== 1012fa9e4066Sahrens */ 1013fa9e4066Sahrens 1014fa9e4066Sahrens krwlock_t * 1015fa9e4066Sahrens spa_traverse_rwlock(spa_t *spa) 1016fa9e4066Sahrens { 1017fa9e4066Sahrens return (&spa->spa_traverse_lock); 1018fa9e4066Sahrens } 1019fa9e4066Sahrens 1020fa9e4066Sahrens int 1021fa9e4066Sahrens spa_traverse_wanted(spa_t *spa) 1022fa9e4066Sahrens { 1023fa9e4066Sahrens return (spa->spa_traverse_wanted); 1024fa9e4066Sahrens } 1025fa9e4066Sahrens 1026fa9e4066Sahrens dsl_pool_t * 1027fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1028fa9e4066Sahrens { 1029fa9e4066Sahrens return (spa->spa_dsl_pool); 1030fa9e4066Sahrens } 1031fa9e4066Sahrens 1032fa9e4066Sahrens blkptr_t * 1033fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1034fa9e4066Sahrens { 1035fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1036fa9e4066Sahrens } 1037fa9e4066Sahrens 1038fa9e4066Sahrens void 1039fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1040fa9e4066Sahrens { 1041fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1042fa9e4066Sahrens } 1043fa9e4066Sahrens 1044fa9e4066Sahrens void 1045fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1046fa9e4066Sahrens { 1047fa9e4066Sahrens if (spa->spa_root == NULL) 1048fa9e4066Sahrens buf[0] = '\0'; 1049fa9e4066Sahrens else 1050fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1051fa9e4066Sahrens } 1052fa9e4066Sahrens 1053fa9e4066Sahrens int 1054fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1055fa9e4066Sahrens { 1056fa9e4066Sahrens return (spa->spa_sync_pass); 1057fa9e4066Sahrens } 1058fa9e4066Sahrens 1059fa9e4066Sahrens char * 1060fa9e4066Sahrens spa_name(spa_t *spa) 1061fa9e4066Sahrens { 1062fa9e4066Sahrens /* 1063fa9e4066Sahrens * Accessing the name requires holding either the namespace lock or the 1064fa9e4066Sahrens * config lock, both of which are required to do a rename. 1065fa9e4066Sahrens */ 1066fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock) || 1067e05725b1Sbonwick spa_config_held(spa, RW_READER)); 1068fa9e4066Sahrens 1069fa9e4066Sahrens return (spa->spa_name); 1070fa9e4066Sahrens } 1071fa9e4066Sahrens 1072fa9e4066Sahrens uint64_t 1073fa9e4066Sahrens spa_guid(spa_t *spa) 1074fa9e4066Sahrens { 1075b5989ec7Seschrock /* 1076b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1077b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1078b5989ec7Seschrock * vdev. We stash the original pool guid in 'spa_load_guid' to handle 1079b5989ec7Seschrock * this case. 1080b5989ec7Seschrock */ 1081b5989ec7Seschrock if (spa->spa_root_vdev != NULL) 1082b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1083b5989ec7Seschrock else 1084b5989ec7Seschrock return (spa->spa_load_guid); 1085fa9e4066Sahrens } 1086fa9e4066Sahrens 1087fa9e4066Sahrens uint64_t 1088fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1089fa9e4066Sahrens { 1090fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1091fa9e4066Sahrens } 1092fa9e4066Sahrens 1093fa9e4066Sahrens uint64_t 1094fa9e4066Sahrens spa_first_txg(spa_t *spa) 1095fa9e4066Sahrens { 1096fa9e4066Sahrens return (spa->spa_first_txg); 1097fa9e4066Sahrens } 1098fa9e4066Sahrens 1099fa9e4066Sahrens int 1100fa9e4066Sahrens spa_state(spa_t *spa) 1101fa9e4066Sahrens { 1102fa9e4066Sahrens return (spa->spa_state); 1103fa9e4066Sahrens } 1104fa9e4066Sahrens 1105fa9e4066Sahrens uint64_t 1106fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1107fa9e4066Sahrens { 1108fa9e4066Sahrens return (spa->spa_freeze_txg); 1109fa9e4066Sahrens } 1110fa9e4066Sahrens 1111fa9e4066Sahrens /* 111299653d4eSeschrock * Return how much space is allocated in the pool (ie. sum of all asize) 1113fa9e4066Sahrens */ 1114fa9e4066Sahrens uint64_t 1115fa9e4066Sahrens spa_get_alloc(spa_t *spa) 1116fa9e4066Sahrens { 1117fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_alloc); 1118fa9e4066Sahrens } 1119fa9e4066Sahrens 1120fa9e4066Sahrens /* 112199653d4eSeschrock * Return how much (raid-z inflated) space there is in the pool. 1122fa9e4066Sahrens */ 1123fa9e4066Sahrens uint64_t 1124fa9e4066Sahrens spa_get_space(spa_t *spa) 1125fa9e4066Sahrens { 1126fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_space); 1127fa9e4066Sahrens } 1128fa9e4066Sahrens 112999653d4eSeschrock /* 113099653d4eSeschrock * Return the amount of raid-z-deflated space in the pool. 113199653d4eSeschrock */ 113299653d4eSeschrock uint64_t 113399653d4eSeschrock spa_get_dspace(spa_t *spa) 113499653d4eSeschrock { 113599653d4eSeschrock if (spa->spa_deflate) 113699653d4eSeschrock return (spa->spa_root_vdev->vdev_stat.vs_dspace); 113799653d4eSeschrock else 113899653d4eSeschrock return (spa->spa_root_vdev->vdev_stat.vs_space); 113999653d4eSeschrock } 114099653d4eSeschrock 1141fa9e4066Sahrens /* ARGSUSED */ 1142fa9e4066Sahrens uint64_t 1143fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 1144fa9e4066Sahrens { 1145fa9e4066Sahrens /* 1146fa9e4066Sahrens * For now, the worst case is 512-byte RAID-Z blocks, in which 1147fa9e4066Sahrens * case the space requirement is exactly 2x; so just assume that. 114844cd46caSbillm * Add to this the fact that we can have up to 3 DVAs per bp, and 114944cd46caSbillm * we have to multiply by a total of 6x. 115044cd46caSbillm */ 115144cd46caSbillm return (lsize * 6); 115244cd46caSbillm } 115344cd46caSbillm 11540a4e9518Sgw /* 11550a4e9518Sgw * Return the failure mode that has been set to this pool. The default 11560a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 11570a4e9518Sgw */ 11580a4e9518Sgw uint8_t 11590a4e9518Sgw spa_get_failmode(spa_t *spa) 11600a4e9518Sgw { 11610a4e9518Sgw return (spa->spa_failmode); 11620a4e9518Sgw } 11630a4e9518Sgw 116444cd46caSbillm uint64_t 116544cd46caSbillm spa_version(spa_t *spa) 116644cd46caSbillm { 116744cd46caSbillm return (spa->spa_ubsync.ub_version); 116844cd46caSbillm } 116944cd46caSbillm 117044cd46caSbillm int 117144cd46caSbillm spa_max_replication(spa_t *spa) 117244cd46caSbillm { 117344cd46caSbillm /* 1174e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 117544cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 117644cd46caSbillm * replication level accordingly. 1177fa9e4066Sahrens */ 1178e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 117944cd46caSbillm return (1); 118044cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1181fa9e4066Sahrens } 1182fa9e4066Sahrens 118399653d4eSeschrock uint64_t 118499653d4eSeschrock bp_get_dasize(spa_t *spa, const blkptr_t *bp) 118599653d4eSeschrock { 118699653d4eSeschrock int sz = 0, i; 118799653d4eSeschrock 118899653d4eSeschrock if (!spa->spa_deflate) 118999653d4eSeschrock return (BP_GET_ASIZE(bp)); 119099653d4eSeschrock 11911934e92fSmaybee spa_config_enter(spa, RW_READER, FTAG); 119299653d4eSeschrock for (i = 0; i < SPA_DVAS_PER_BP; i++) { 119399653d4eSeschrock vdev_t *vd = 119499653d4eSeschrock vdev_lookup_top(spa, DVA_GET_VDEV(&bp->blk_dva[i])); 11951934e92fSmaybee if (vd) 11961934e92fSmaybee sz += (DVA_GET_ASIZE(&bp->blk_dva[i]) >> 11971934e92fSmaybee SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 119899653d4eSeschrock } 11991934e92fSmaybee spa_config_exit(spa, FTAG); 120099653d4eSeschrock return (sz); 120199653d4eSeschrock } 120299653d4eSeschrock 1203fa9e4066Sahrens /* 1204fa9e4066Sahrens * ========================================================================== 1205fa9e4066Sahrens * Initialization and Termination 1206fa9e4066Sahrens * ========================================================================== 1207fa9e4066Sahrens */ 1208fa9e4066Sahrens 1209fa9e4066Sahrens static int 1210fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1211fa9e4066Sahrens { 1212fa9e4066Sahrens const spa_t *s1 = a1; 1213fa9e4066Sahrens const spa_t *s2 = a2; 1214fa9e4066Sahrens int s; 1215fa9e4066Sahrens 1216fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1217fa9e4066Sahrens if (s > 0) 1218fa9e4066Sahrens return (1); 1219fa9e4066Sahrens if (s < 0) 1220fa9e4066Sahrens return (-1); 1221fa9e4066Sahrens return (0); 1222fa9e4066Sahrens } 1223fa9e4066Sahrens 12240373e76bSbonwick int 12250373e76bSbonwick spa_busy(void) 12260373e76bSbonwick { 12270373e76bSbonwick return (spa_active_count); 12280373e76bSbonwick } 12290373e76bSbonwick 1230e7cbe64fSgw void 1231e7cbe64fSgw spa_boot_init() 1232e7cbe64fSgw { 1233e7cbe64fSgw spa_config_load(); 1234e7cbe64fSgw } 1235e7cbe64fSgw 1236fa9e4066Sahrens void 1237fa9e4066Sahrens spa_init(int mode) 1238fa9e4066Sahrens { 1239fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1240c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1241fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 1242fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1243fa9e4066Sahrens 1244fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1245fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1246fa9e4066Sahrens 1247fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 1248fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 1249fa94a07fSbrendan 1250fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 1251fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 125299653d4eSeschrock 1253fa9e4066Sahrens spa_mode = mode; 1254fa9e4066Sahrens 1255fa9e4066Sahrens refcount_init(); 1256fa9e4066Sahrens unique_init(); 1257fa9e4066Sahrens zio_init(); 1258fa9e4066Sahrens dmu_init(); 1259fa9e4066Sahrens zil_init(); 126087db74c1Sek vdev_cache_stat_init(); 126191ebeef5Sahrens zfs_prop_init(); 1262990b4856Slling zpool_prop_init(); 1263fa9e4066Sahrens spa_config_load(); 1264fa9e4066Sahrens } 1265fa9e4066Sahrens 1266fa9e4066Sahrens void 1267fa9e4066Sahrens spa_fini(void) 1268fa9e4066Sahrens { 1269fa9e4066Sahrens spa_evict_all(); 1270fa9e4066Sahrens 127187db74c1Sek vdev_cache_stat_fini(); 1272fa9e4066Sahrens zil_fini(); 1273fa9e4066Sahrens dmu_fini(); 1274fa9e4066Sahrens zio_fini(); 127591ebeef5Sahrens unique_fini(); 1276fa9e4066Sahrens refcount_fini(); 1277fa9e4066Sahrens 1278fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 127999653d4eSeschrock avl_destroy(&spa_spare_avl); 1280fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 1281fa9e4066Sahrens 1282fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1283fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1284c25056deSgw mutex_destroy(&spa_spare_lock); 1285fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 1286fa9e4066Sahrens } 12876ce0521aSperrin 12886ce0521aSperrin /* 12896ce0521aSperrin * Return whether this pool has slogs. No locking needed. 12906ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 12916ce0521aSperrin * performance and not correctness 12926ce0521aSperrin */ 12936ce0521aSperrin boolean_t 12946ce0521aSperrin spa_has_slogs(spa_t *spa) 12956ce0521aSperrin { 12966ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 12976ce0521aSperrin } 1298