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 /* 2239c23413Seschrock * Copyright 2007 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 * 147fa9e4066Sahrens * The spa_config_lock is manipulated using the following functions: 148fa9e4066Sahrens * 14944cd46caSbillm * spa_config_enter() Acquire the config lock as RW_READER or 15044cd46caSbillm * RW_WRITER. At least one reference on the spa_t 15144cd46caSbillm * must exist. 152fa9e4066Sahrens * 15344cd46caSbillm * spa_config_exit() Release the config lock. 154fa9e4066Sahrens * 15544cd46caSbillm * spa_config_held() Returns true if the config lock is currently 15644cd46caSbillm * held in the given state. 157fa9e4066Sahrens * 158ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 159fa9e4066Sahrens * 16044cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 161ea8dc4b6Seschrock * for writing. 162fa9e4066Sahrens * 16344cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 16444cd46caSbillm * to complete, sync the updated configs to the 165ea8dc4b6Seschrock * cache, and release the namespace lock. 166fa9e4066Sahrens * 167fa9e4066Sahrens * The spa_name() function also requires either the spa_namespace_lock 168fa9e4066Sahrens * or the spa_config_lock, as both are needed to do a rename. spa_rename() is 169fa9e4066Sahrens * also implemented within this file since is requires manipulation of the 170fa9e4066Sahrens * namespace. 171fa9e4066Sahrens */ 172fa9e4066Sahrens 173fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 174fa9e4066Sahrens kmutex_t spa_namespace_lock; 175fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 1760373e76bSbonwick static int spa_active_count; 177416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 178fa9e4066Sahrens 17999653d4eSeschrock static kmutex_t spa_spare_lock; 18039c23413Seschrock static avl_tree_t spa_spare_avl; 18199653d4eSeschrock 182fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 183fa9e4066Sahrens int spa_mode; 184fa9e4066Sahrens 185fa9e4066Sahrens #ifdef ZFS_DEBUG 18640feaa91Sahrens /* Everything except dprintf is on by default in debug builds */ 18740feaa91Sahrens int zfs_flags = ~ZFS_DEBUG_DPRINTF; 188fa9e4066Sahrens #else 189fa9e4066Sahrens int zfs_flags = 0; 190fa9e4066Sahrens #endif 191fa9e4066Sahrens 1920125049cSahrens /* 1930125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 1940125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 1950125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 1960125049cSahrens */ 1970125049cSahrens int zfs_recover = 0; 1980125049cSahrens 199fa9e4066Sahrens #define SPA_MINREF 5 /* spa_refcnt for an open-but-idle pool */ 200fa9e4066Sahrens 201fa9e4066Sahrens /* 202fa9e4066Sahrens * ========================================================================== 203fa9e4066Sahrens * SPA namespace functions 204fa9e4066Sahrens * ========================================================================== 205fa9e4066Sahrens */ 206fa9e4066Sahrens 207fa9e4066Sahrens /* 208fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 209fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 210fa9e4066Sahrens */ 211fa9e4066Sahrens spa_t * 212fa9e4066Sahrens spa_lookup(const char *name) 213fa9e4066Sahrens { 214fa9e4066Sahrens spa_t search, *spa; 215fa9e4066Sahrens avl_index_t where; 21640feaa91Sahrens char c; 21740feaa91Sahrens char *cp; 218fa9e4066Sahrens 219fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 220fa9e4066Sahrens 22140feaa91Sahrens /* 22240feaa91Sahrens * If it's a full dataset name, figure out the pool name and 22340feaa91Sahrens * just use that. 22440feaa91Sahrens */ 22540feaa91Sahrens cp = strpbrk(name, "/@"); 22640feaa91Sahrens if (cp) { 22740feaa91Sahrens c = *cp; 22840feaa91Sahrens *cp = '\0'; 22940feaa91Sahrens } 23040feaa91Sahrens 231fa9e4066Sahrens search.spa_name = (char *)name; 232fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 233fa9e4066Sahrens 23440feaa91Sahrens if (cp) 23540feaa91Sahrens *cp = c; 23640feaa91Sahrens 237fa9e4066Sahrens return (spa); 238fa9e4066Sahrens } 239fa9e4066Sahrens 240fa9e4066Sahrens /* 241fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 242fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 243fa9e4066Sahrens * exist by calling spa_lookup() first. 244fa9e4066Sahrens */ 245fa9e4066Sahrens spa_t * 2460373e76bSbonwick spa_add(const char *name, const char *altroot) 247fa9e4066Sahrens { 248fa9e4066Sahrens spa_t *spa; 249fa9e4066Sahrens 250fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 251fa9e4066Sahrens 252fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 253fa9e4066Sahrens 254c25056deSgw rw_init(&spa->spa_traverse_lock, NULL, RW_DEFAULT, NULL); 255c25056deSgw 256c25056deSgw mutex_init(&spa->spa_uberblock_lock, NULL, MUTEX_DEFAULT, NULL); 257c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 258c25056deSgw mutex_init(&spa->spa_config_cache_lock, NULL, MUTEX_DEFAULT, NULL); 259c25056deSgw mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 260c25056deSgw mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 261c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 262c25056deSgw mutex_init(&spa->spa_sync_bplist.bpl_lock, NULL, MUTEX_DEFAULT, NULL); 263c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 264c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 265c25056deSgw 266c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 267c25056deSgw cv_init(&spa->spa_scrub_cv, NULL, CV_DEFAULT, NULL); 268c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 269c25056deSgw 270fa9e4066Sahrens spa->spa_name = spa_strdup(name); 271fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 272fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 2730373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 274fa9e4066Sahrens 275fa9e4066Sahrens refcount_create(&spa->spa_refcount); 27691ebeef5Sahrens rprw_init(&spa->spa_config_lock); 277fa9e4066Sahrens 278fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 279fa9e4066Sahrens 2800a4e9518Sgw mutex_init(&spa->spa_zio_lock, NULL, MUTEX_DEFAULT, NULL); 2810a4e9518Sgw 2820373e76bSbonwick /* 2830373e76bSbonwick * Set the alternate root, if there is one. 2840373e76bSbonwick */ 2850373e76bSbonwick if (altroot) { 2860373e76bSbonwick spa->spa_root = spa_strdup(altroot); 2870373e76bSbonwick spa_active_count++; 2880373e76bSbonwick } 2890373e76bSbonwick 290fa9e4066Sahrens return (spa); 291fa9e4066Sahrens } 292fa9e4066Sahrens 293fa9e4066Sahrens /* 294fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 295fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 296fa9e4066Sahrens * deactivated. 297fa9e4066Sahrens */ 298fa9e4066Sahrens void 299fa9e4066Sahrens spa_remove(spa_t *spa) 300fa9e4066Sahrens { 301fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 302fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 303fa9e4066Sahrens ASSERT(spa->spa_scrub_thread == NULL); 304fa9e4066Sahrens 305fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 306fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 307fa9e4066Sahrens 3080373e76bSbonwick if (spa->spa_root) { 309fa9e4066Sahrens spa_strfree(spa->spa_root); 3100373e76bSbonwick spa_active_count--; 3110373e76bSbonwick } 312fa9e4066Sahrens 313fa9e4066Sahrens if (spa->spa_name) 314fa9e4066Sahrens spa_strfree(spa->spa_name); 315fa9e4066Sahrens 316*2f8aaab3Seschrock if (spa->spa_config_dir) 317*2f8aaab3Seschrock spa_strfree(spa->spa_config_dir); 318*2f8aaab3Seschrock if (spa->spa_config_file) 319*2f8aaab3Seschrock spa_strfree(spa->spa_config_file); 320*2f8aaab3Seschrock 321fa9e4066Sahrens spa_config_set(spa, NULL); 322fa9e4066Sahrens 323fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 32491ebeef5Sahrens 32591ebeef5Sahrens rprw_destroy(&spa->spa_config_lock); 326fa9e4066Sahrens 327c25056deSgw rw_destroy(&spa->spa_traverse_lock); 328c25056deSgw 329c25056deSgw cv_destroy(&spa->spa_async_cv); 330c25056deSgw cv_destroy(&spa->spa_scrub_cv); 331c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 332c25056deSgw 333c25056deSgw mutex_destroy(&spa->spa_uberblock_lock); 3345ad82045Snd mutex_destroy(&spa->spa_async_lock); 335c25056deSgw mutex_destroy(&spa->spa_config_cache_lock); 336c25056deSgw mutex_destroy(&spa->spa_scrub_lock); 337c25056deSgw mutex_destroy(&spa->spa_errlog_lock); 338c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 339c25056deSgw mutex_destroy(&spa->spa_sync_bplist.bpl_lock); 34006eeb2adSek mutex_destroy(&spa->spa_history_lock); 341b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 3420a4e9518Sgw mutex_destroy(&spa->spa_zio_lock); 3435ad82045Snd 344fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 345fa9e4066Sahrens } 346fa9e4066Sahrens 347fa9e4066Sahrens /* 348fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 349fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 350fa9e4066Sahrens */ 351fa9e4066Sahrens spa_t * 352fa9e4066Sahrens spa_next(spa_t *prev) 353fa9e4066Sahrens { 354fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 355fa9e4066Sahrens 356fa9e4066Sahrens if (prev) 357fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 358fa9e4066Sahrens else 359fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 360fa9e4066Sahrens } 361fa9e4066Sahrens 362fa9e4066Sahrens /* 363fa9e4066Sahrens * ========================================================================== 364fa9e4066Sahrens * SPA refcount functions 365fa9e4066Sahrens * ========================================================================== 366fa9e4066Sahrens */ 367fa9e4066Sahrens 368fa9e4066Sahrens /* 369fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 370fa9e4066Sahrens * have the namespace lock held. 371fa9e4066Sahrens */ 372fa9e4066Sahrens void 373fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 374fa9e4066Sahrens { 375fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 376fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 377fa9e4066Sahrens 378fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 379fa9e4066Sahrens } 380fa9e4066Sahrens 381fa9e4066Sahrens /* 382fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 383fa9e4066Sahrens * have the namespace lock held. 384fa9e4066Sahrens */ 385fa9e4066Sahrens void 386fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 387fa9e4066Sahrens { 388fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 389fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 390fa9e4066Sahrens 391fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 392fa9e4066Sahrens } 393fa9e4066Sahrens 394fa9e4066Sahrens /* 395fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 396fa9e4066Sahrens * spa_namespace_lock held. We really compare against SPA_MINREF, which is the 397fa9e4066Sahrens * number of references acquired when opening a pool 398fa9e4066Sahrens */ 399fa9e4066Sahrens boolean_t 400fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 401fa9e4066Sahrens { 402fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 403fa9e4066Sahrens 404fa9e4066Sahrens return (refcount_count(&spa->spa_refcount) == SPA_MINREF); 405fa9e4066Sahrens } 406fa9e4066Sahrens 40799653d4eSeschrock /* 40899653d4eSeschrock * ========================================================================== 40999653d4eSeschrock * SPA spare tracking 41099653d4eSeschrock * ========================================================================== 41199653d4eSeschrock */ 41299653d4eSeschrock 41399653d4eSeschrock /* 41439c23413Seschrock * Spares are tracked globally due to the following constraints: 41539c23413Seschrock * 41639c23413Seschrock * - A spare may be part of multiple pools. 41739c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 41839c23413Seschrock * another pool. 41939c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 42039c23413Seschrock * the target is a spare in the same pool. 42139c23413Seschrock * 42239c23413Seschrock * We keep track of all spares on the system through the use of a reference 42339c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 42439c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 42539c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 42639c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 42739c23413Seschrock * pool), we also keep track of which pool its been made a part of. 42839c23413Seschrock * 42939c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 43039c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 43139c23413Seschrock * separate spare lock exists for the status query path, which does not need to 43239c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 43399653d4eSeschrock */ 43439c23413Seschrock 43599653d4eSeschrock typedef struct spa_spare { 43699653d4eSeschrock uint64_t spare_guid; 43739c23413Seschrock uint64_t spare_pool; 43899653d4eSeschrock avl_node_t spare_avl; 43999653d4eSeschrock int spare_count; 44099653d4eSeschrock } spa_spare_t; 44199653d4eSeschrock 44299653d4eSeschrock static int 44399653d4eSeschrock spa_spare_compare(const void *a, const void *b) 44499653d4eSeschrock { 44599653d4eSeschrock const spa_spare_t *sa = a; 44699653d4eSeschrock const spa_spare_t *sb = b; 44799653d4eSeschrock 44899653d4eSeschrock if (sa->spare_guid < sb->spare_guid) 44999653d4eSeschrock return (-1); 45099653d4eSeschrock else if (sa->spare_guid > sb->spare_guid) 45199653d4eSeschrock return (1); 45299653d4eSeschrock else 45399653d4eSeschrock return (0); 45499653d4eSeschrock } 45599653d4eSeschrock 45699653d4eSeschrock void 45739c23413Seschrock spa_spare_add(vdev_t *vd) 45899653d4eSeschrock { 45999653d4eSeschrock avl_index_t where; 46099653d4eSeschrock spa_spare_t search; 46199653d4eSeschrock spa_spare_t *spare; 46299653d4eSeschrock 46399653d4eSeschrock mutex_enter(&spa_spare_lock); 46439c23413Seschrock ASSERT(!vd->vdev_isspare); 46599653d4eSeschrock 46639c23413Seschrock search.spare_guid = vd->vdev_guid; 46799653d4eSeschrock if ((spare = avl_find(&spa_spare_avl, &search, &where)) != NULL) { 46899653d4eSeschrock spare->spare_count++; 46999653d4eSeschrock } else { 47039c23413Seschrock spare = kmem_zalloc(sizeof (spa_spare_t), KM_SLEEP); 47139c23413Seschrock spare->spare_guid = vd->vdev_guid; 47299653d4eSeschrock spare->spare_count = 1; 47399653d4eSeschrock avl_insert(&spa_spare_avl, spare, where); 47499653d4eSeschrock } 47539c23413Seschrock vd->vdev_isspare = B_TRUE; 47699653d4eSeschrock 47799653d4eSeschrock mutex_exit(&spa_spare_lock); 47899653d4eSeschrock } 47999653d4eSeschrock 48099653d4eSeschrock void 48139c23413Seschrock spa_spare_remove(vdev_t *vd) 48299653d4eSeschrock { 48399653d4eSeschrock spa_spare_t search; 48499653d4eSeschrock spa_spare_t *spare; 48599653d4eSeschrock avl_index_t where; 48699653d4eSeschrock 48799653d4eSeschrock mutex_enter(&spa_spare_lock); 48899653d4eSeschrock 48939c23413Seschrock search.spare_guid = vd->vdev_guid; 49099653d4eSeschrock spare = avl_find(&spa_spare_avl, &search, &where); 49199653d4eSeschrock 49239c23413Seschrock ASSERT(vd->vdev_isspare); 49399653d4eSeschrock ASSERT(spare != NULL); 49499653d4eSeschrock 49599653d4eSeschrock if (--spare->spare_count == 0) { 49699653d4eSeschrock avl_remove(&spa_spare_avl, spare); 49799653d4eSeschrock kmem_free(spare, sizeof (spa_spare_t)); 49839c23413Seschrock } else if (spare->spare_pool == spa_guid(vd->vdev_spa)) { 49939c23413Seschrock spare->spare_pool = 0ULL; 50099653d4eSeschrock } 50199653d4eSeschrock 50239c23413Seschrock vd->vdev_isspare = B_FALSE; 50399653d4eSeschrock mutex_exit(&spa_spare_lock); 50499653d4eSeschrock } 50599653d4eSeschrock 50699653d4eSeschrock boolean_t 50739c23413Seschrock spa_spare_exists(uint64_t guid, uint64_t *pool) 50899653d4eSeschrock { 50939c23413Seschrock spa_spare_t search, *found; 51099653d4eSeschrock avl_index_t where; 51199653d4eSeschrock 51299653d4eSeschrock mutex_enter(&spa_spare_lock); 51399653d4eSeschrock 51499653d4eSeschrock search.spare_guid = guid; 51539c23413Seschrock found = avl_find(&spa_spare_avl, &search, &where); 51639c23413Seschrock 51739c23413Seschrock if (pool) { 51839c23413Seschrock if (found) 51939c23413Seschrock *pool = found->spare_pool; 52039c23413Seschrock else 52139c23413Seschrock *pool = 0ULL; 52239c23413Seschrock } 52399653d4eSeschrock 52499653d4eSeschrock mutex_exit(&spa_spare_lock); 52599653d4eSeschrock 52639c23413Seschrock return (found != NULL); 52739c23413Seschrock } 52839c23413Seschrock 52939c23413Seschrock void 53039c23413Seschrock spa_spare_activate(vdev_t *vd) 53139c23413Seschrock { 53239c23413Seschrock spa_spare_t search, *found; 53339c23413Seschrock avl_index_t where; 53439c23413Seschrock 53539c23413Seschrock mutex_enter(&spa_spare_lock); 53639c23413Seschrock ASSERT(vd->vdev_isspare); 53739c23413Seschrock 53839c23413Seschrock search.spare_guid = vd->vdev_guid; 53939c23413Seschrock found = avl_find(&spa_spare_avl, &search, &where); 54039c23413Seschrock ASSERT(found != NULL); 54139c23413Seschrock ASSERT(found->spare_pool == 0ULL); 54239c23413Seschrock 54339c23413Seschrock found->spare_pool = spa_guid(vd->vdev_spa); 54439c23413Seschrock mutex_exit(&spa_spare_lock); 54599653d4eSeschrock } 54699653d4eSeschrock 547fa9e4066Sahrens /* 548fa9e4066Sahrens * ========================================================================== 549fa9e4066Sahrens * SPA config locking 550fa9e4066Sahrens * ========================================================================== 551fa9e4066Sahrens */ 552fa9e4066Sahrens void 553ea8dc4b6Seschrock spa_config_enter(spa_t *spa, krw_t rw, void *tag) 554fa9e4066Sahrens { 55591ebeef5Sahrens rprw_enter(&spa->spa_config_lock, rw, tag); 556fa9e4066Sahrens } 557fa9e4066Sahrens 558fa9e4066Sahrens void 559ea8dc4b6Seschrock spa_config_exit(spa_t *spa, void *tag) 560fa9e4066Sahrens { 56191ebeef5Sahrens rprw_exit(&spa->spa_config_lock, tag); 562fa9e4066Sahrens } 563fa9e4066Sahrens 564fa9e4066Sahrens boolean_t 565fa9e4066Sahrens spa_config_held(spa_t *spa, krw_t rw) 566fa9e4066Sahrens { 56791ebeef5Sahrens return (rprw_held(&spa->spa_config_lock, rw)); 568fa9e4066Sahrens } 569fa9e4066Sahrens 570fa9e4066Sahrens /* 571fa9e4066Sahrens * ========================================================================== 572fa9e4066Sahrens * SPA vdev locking 573fa9e4066Sahrens * ========================================================================== 574fa9e4066Sahrens */ 575fa9e4066Sahrens 576fa9e4066Sahrens /* 577ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 578ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 579fa9e4066Sahrens * It returns the next transaction group for the spa_t. 580fa9e4066Sahrens */ 581fa9e4066Sahrens uint64_t 582fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 583fa9e4066Sahrens { 5843d7072f8Seschrock mutex_enter(&spa_namespace_lock); 5853d7072f8Seschrock 586ea8dc4b6Seschrock /* 5873d7072f8Seschrock * Suspend scrub activity while we mess with the config. We must do 5883d7072f8Seschrock * this after acquiring the namespace lock to avoid a 3-way deadlock 5893d7072f8Seschrock * with spa_scrub_stop() and the scrub thread. 590ea8dc4b6Seschrock */ 591ea8dc4b6Seschrock spa_scrub_suspend(spa); 592fa9e4066Sahrens 593ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, spa); 594fa9e4066Sahrens 595fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 596fa9e4066Sahrens } 597fa9e4066Sahrens 598fa9e4066Sahrens /* 599fa9e4066Sahrens * Unlock the spa_t after adding or removing a vdev. Besides undoing the 600fa9e4066Sahrens * locking of spa_vdev_enter(), we also want make sure the transactions have 601fa9e4066Sahrens * synced to disk, and then update the global configuration cache with the new 602fa9e4066Sahrens * information. 603fa9e4066Sahrens */ 604fa9e4066Sahrens int 605fa9e4066Sahrens spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 606fa9e4066Sahrens { 6070e34b6a7Sbonwick int config_changed = B_FALSE; 608ea8dc4b6Seschrock 6090373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 6100e34b6a7Sbonwick 6110e34b6a7Sbonwick /* 6120e34b6a7Sbonwick * Reassess the DTLs. 6130e34b6a7Sbonwick */ 6140373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 6150e34b6a7Sbonwick 6160e34b6a7Sbonwick /* 6170373e76bSbonwick * If the config changed, notify the scrub thread that it must restart. 6180e34b6a7Sbonwick */ 6190e34b6a7Sbonwick if (error == 0 && !list_is_empty(&spa->spa_dirty_list)) { 6200e34b6a7Sbonwick config_changed = B_TRUE; 6210373e76bSbonwick spa_scrub_restart(spa, txg); 6220e34b6a7Sbonwick } 623ea8dc4b6Seschrock 624ea8dc4b6Seschrock spa_config_exit(spa, spa); 625fa9e4066Sahrens 626ea8dc4b6Seschrock /* 6275dabedeeSbonwick * Allow scrubbing to resume. 628ea8dc4b6Seschrock */ 629ea8dc4b6Seschrock spa_scrub_resume(spa); 630fa9e4066Sahrens 631fa9e4066Sahrens /* 632fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 633fa9e4066Sahrens * that there won't be more than one config change per txg. 634fa9e4066Sahrens * This allows us to use the txg as the generation number. 635fa9e4066Sahrens */ 636fa9e4066Sahrens if (error == 0) 637fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 638fa9e4066Sahrens 639fa9e4066Sahrens if (vd != NULL) { 640fa9e4066Sahrens ASSERT(!vd->vdev_detached || vd->vdev_dtl.smo_object == 0); 641fa9e4066Sahrens vdev_free(vd); 642fa9e4066Sahrens } 643fa9e4066Sahrens 644fa9e4066Sahrens /* 6450e34b6a7Sbonwick * If the config changed, update the config cache. 646fa9e4066Sahrens */ 6470e34b6a7Sbonwick if (config_changed) 648fa9e4066Sahrens spa_config_sync(); 649ea8dc4b6Seschrock 650ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 651fa9e4066Sahrens 652fa9e4066Sahrens return (error); 653fa9e4066Sahrens } 654fa9e4066Sahrens 655fa9e4066Sahrens /* 656fa9e4066Sahrens * ========================================================================== 657fa9e4066Sahrens * Miscellaneous functions 658fa9e4066Sahrens * ========================================================================== 659fa9e4066Sahrens */ 660fa9e4066Sahrens 661fa9e4066Sahrens /* 662fa9e4066Sahrens * Rename a spa_t. 663fa9e4066Sahrens */ 664fa9e4066Sahrens int 665fa9e4066Sahrens spa_rename(const char *name, const char *newname) 666fa9e4066Sahrens { 667fa9e4066Sahrens spa_t *spa; 668fa9e4066Sahrens int err; 669fa9e4066Sahrens 670fa9e4066Sahrens /* 671fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 672fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 673fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 674ea8dc4b6Seschrock * allow recursive calls for other reasons. 675fa9e4066Sahrens */ 676fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 677fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 678fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 679fa9e4066Sahrens return (err); 680fa9e4066Sahrens } 681fa9e4066Sahrens 682ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 683fa9e4066Sahrens 684fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 685fa9e4066Sahrens spa_strfree(spa->spa_name); 686fa9e4066Sahrens spa->spa_name = spa_strdup(newname); 687fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 688fa9e4066Sahrens 689fa9e4066Sahrens /* 690fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 691fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 692fa9e4066Sahrens * during the sync. 693fa9e4066Sahrens */ 694fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 695fa9e4066Sahrens 696ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 697fa9e4066Sahrens 6980373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 699fa9e4066Sahrens 700fa9e4066Sahrens /* 701fa9e4066Sahrens * Sync the updated config cache. 702fa9e4066Sahrens */ 703fa9e4066Sahrens spa_config_sync(); 704fa9e4066Sahrens 705fa9e4066Sahrens spa_close(spa, FTAG); 706fa9e4066Sahrens 707fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 708fa9e4066Sahrens 709fa9e4066Sahrens return (0); 710fa9e4066Sahrens } 711fa9e4066Sahrens 712fa9e4066Sahrens 713fa9e4066Sahrens /* 714fa9e4066Sahrens * Determine whether a pool with given pool_guid exists. If device_guid is 715fa9e4066Sahrens * non-zero, determine whether the pool exists *and* contains a device with the 716fa9e4066Sahrens * specified device_guid. 717fa9e4066Sahrens */ 718fa9e4066Sahrens boolean_t 719fa9e4066Sahrens spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 720fa9e4066Sahrens { 721fa9e4066Sahrens spa_t *spa; 722fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 723fa9e4066Sahrens 724ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 725fa9e4066Sahrens 726fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 727fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 728fa9e4066Sahrens continue; 729fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 730fa9e4066Sahrens continue; 73139c23413Seschrock if (spa_guid(spa) == pool_guid) { 73239c23413Seschrock if (device_guid == 0) 73339c23413Seschrock break; 73439c23413Seschrock 73539c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 73639c23413Seschrock device_guid) != NULL) 73739c23413Seschrock break; 73839c23413Seschrock 73939c23413Seschrock /* 7408654d025Sperrin * Check any devices we may be in the process of adding. 74139c23413Seschrock */ 74239c23413Seschrock if (spa->spa_pending_vdev) { 74339c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 74439c23413Seschrock device_guid) != NULL) 74539c23413Seschrock break; 74639c23413Seschrock } 74739c23413Seschrock } 748fa9e4066Sahrens } 749fa9e4066Sahrens 750fa9e4066Sahrens return (spa != NULL); 751fa9e4066Sahrens } 752fa9e4066Sahrens 753fa9e4066Sahrens char * 754fa9e4066Sahrens spa_strdup(const char *s) 755fa9e4066Sahrens { 756fa9e4066Sahrens size_t len; 757fa9e4066Sahrens char *new; 758fa9e4066Sahrens 759fa9e4066Sahrens len = strlen(s); 760fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 761fa9e4066Sahrens bcopy(s, new, len); 762fa9e4066Sahrens new[len] = '\0'; 763fa9e4066Sahrens 764fa9e4066Sahrens return (new); 765fa9e4066Sahrens } 766fa9e4066Sahrens 767fa9e4066Sahrens void 768fa9e4066Sahrens spa_strfree(char *s) 769fa9e4066Sahrens { 770fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 771fa9e4066Sahrens } 772fa9e4066Sahrens 773fa9e4066Sahrens uint64_t 774fa9e4066Sahrens spa_get_random(uint64_t range) 775fa9e4066Sahrens { 776fa9e4066Sahrens uint64_t r; 777fa9e4066Sahrens 778fa9e4066Sahrens ASSERT(range != 0); 779fa9e4066Sahrens 780fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 781fa9e4066Sahrens 782fa9e4066Sahrens return (r % range); 783fa9e4066Sahrens } 784fa9e4066Sahrens 785fa9e4066Sahrens void 786d80c45e0Sbonwick sprintf_blkptr(char *buf, int len, const blkptr_t *bp) 787fa9e4066Sahrens { 78844cd46caSbillm int d; 789fa9e4066Sahrens 790fa9e4066Sahrens if (bp == NULL) { 791fbabab8fSmaybee (void) snprintf(buf, len, "<NULL>"); 792fa9e4066Sahrens return; 793fa9e4066Sahrens } 794fa9e4066Sahrens 795fa9e4066Sahrens if (BP_IS_HOLE(bp)) { 796fbabab8fSmaybee (void) snprintf(buf, len, "<hole>"); 797fa9e4066Sahrens return; 798fa9e4066Sahrens } 799fa9e4066Sahrens 80044cd46caSbillm (void) snprintf(buf, len, "[L%llu %s] %llxL/%llxP ", 801fa9e4066Sahrens (u_longlong_t)BP_GET_LEVEL(bp), 802fa9e4066Sahrens dmu_ot[BP_GET_TYPE(bp)].ot_name, 803fa9e4066Sahrens (u_longlong_t)BP_GET_LSIZE(bp), 80444cd46caSbillm (u_longlong_t)BP_GET_PSIZE(bp)); 80544cd46caSbillm 80644cd46caSbillm for (d = 0; d < BP_GET_NDVAS(bp); d++) { 807d80c45e0Sbonwick const dva_t *dva = &bp->blk_dva[d]; 80844cd46caSbillm (void) snprintf(buf + strlen(buf), len - strlen(buf), 80944cd46caSbillm "DVA[%d]=<%llu:%llx:%llx> ", d, 81044cd46caSbillm (u_longlong_t)DVA_GET_VDEV(dva), 81144cd46caSbillm (u_longlong_t)DVA_GET_OFFSET(dva), 81244cd46caSbillm (u_longlong_t)DVA_GET_ASIZE(dva)); 81344cd46caSbillm } 81444cd46caSbillm 81544cd46caSbillm (void) snprintf(buf + strlen(buf), len - strlen(buf), 81644cd46caSbillm "%s %s %s %s birth=%llu fill=%llu cksum=%llx:%llx:%llx:%llx", 817fa9e4066Sahrens zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 818fa9e4066Sahrens zio_compress_table[BP_GET_COMPRESS(bp)].ci_name, 819fa9e4066Sahrens BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", 82044cd46caSbillm BP_IS_GANG(bp) ? "gang" : "contiguous", 821fa9e4066Sahrens (u_longlong_t)bp->blk_birth, 822fa9e4066Sahrens (u_longlong_t)bp->blk_fill, 823fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[0], 824fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[1], 825fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[2], 826fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[3]); 827fa9e4066Sahrens } 828fa9e4066Sahrens 829fa9e4066Sahrens void 830fa9e4066Sahrens spa_freeze(spa_t *spa) 831fa9e4066Sahrens { 832fa9e4066Sahrens uint64_t freeze_txg = 0; 833fa9e4066Sahrens 834ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 835fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 836fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 837fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 838fa9e4066Sahrens } 839ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 840fa9e4066Sahrens if (freeze_txg != 0) 841fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 842fa9e4066Sahrens } 843fa9e4066Sahrens 8440125049cSahrens void 8450125049cSahrens zfs_panic_recover(const char *fmt, ...) 8460125049cSahrens { 8470125049cSahrens va_list adx; 8480125049cSahrens 8490125049cSahrens va_start(adx, fmt); 8500125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 8510125049cSahrens va_end(adx); 8520125049cSahrens } 8530125049cSahrens 854fa9e4066Sahrens /* 855fa9e4066Sahrens * ========================================================================== 856fa9e4066Sahrens * Accessor functions 857fa9e4066Sahrens * ========================================================================== 858fa9e4066Sahrens */ 859fa9e4066Sahrens 860fa9e4066Sahrens krwlock_t * 861fa9e4066Sahrens spa_traverse_rwlock(spa_t *spa) 862fa9e4066Sahrens { 863fa9e4066Sahrens return (&spa->spa_traverse_lock); 864fa9e4066Sahrens } 865fa9e4066Sahrens 866fa9e4066Sahrens int 867fa9e4066Sahrens spa_traverse_wanted(spa_t *spa) 868fa9e4066Sahrens { 869fa9e4066Sahrens return (spa->spa_traverse_wanted); 870fa9e4066Sahrens } 871fa9e4066Sahrens 872fa9e4066Sahrens dsl_pool_t * 873fa9e4066Sahrens spa_get_dsl(spa_t *spa) 874fa9e4066Sahrens { 875fa9e4066Sahrens return (spa->spa_dsl_pool); 876fa9e4066Sahrens } 877fa9e4066Sahrens 878fa9e4066Sahrens blkptr_t * 879fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 880fa9e4066Sahrens { 881fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 882fa9e4066Sahrens } 883fa9e4066Sahrens 884fa9e4066Sahrens void 885fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 886fa9e4066Sahrens { 887fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 888fa9e4066Sahrens } 889fa9e4066Sahrens 890fa9e4066Sahrens void 891fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 892fa9e4066Sahrens { 893fa9e4066Sahrens if (spa->spa_root == NULL) 894fa9e4066Sahrens buf[0] = '\0'; 895fa9e4066Sahrens else 896fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 897fa9e4066Sahrens } 898fa9e4066Sahrens 899fa9e4066Sahrens int 900fa9e4066Sahrens spa_sync_pass(spa_t *spa) 901fa9e4066Sahrens { 902fa9e4066Sahrens return (spa->spa_sync_pass); 903fa9e4066Sahrens } 904fa9e4066Sahrens 905fa9e4066Sahrens char * 906fa9e4066Sahrens spa_name(spa_t *spa) 907fa9e4066Sahrens { 908fa9e4066Sahrens /* 909fa9e4066Sahrens * Accessing the name requires holding either the namespace lock or the 910fa9e4066Sahrens * config lock, both of which are required to do a rename. 911fa9e4066Sahrens */ 912fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock) || 913fa9e4066Sahrens spa_config_held(spa, RW_READER) || spa_config_held(spa, RW_WRITER)); 914fa9e4066Sahrens 915fa9e4066Sahrens return (spa->spa_name); 916fa9e4066Sahrens } 917fa9e4066Sahrens 918fa9e4066Sahrens uint64_t 919fa9e4066Sahrens spa_guid(spa_t *spa) 920fa9e4066Sahrens { 921b5989ec7Seschrock /* 922b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 923b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 924b5989ec7Seschrock * vdev. We stash the original pool guid in 'spa_load_guid' to handle 925b5989ec7Seschrock * this case. 926b5989ec7Seschrock */ 927b5989ec7Seschrock if (spa->spa_root_vdev != NULL) 928b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 929b5989ec7Seschrock else 930b5989ec7Seschrock return (spa->spa_load_guid); 931fa9e4066Sahrens } 932fa9e4066Sahrens 933fa9e4066Sahrens uint64_t 934fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 935fa9e4066Sahrens { 936fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 937fa9e4066Sahrens } 938fa9e4066Sahrens 939fa9e4066Sahrens uint64_t 940fa9e4066Sahrens spa_first_txg(spa_t *spa) 941fa9e4066Sahrens { 942fa9e4066Sahrens return (spa->spa_first_txg); 943fa9e4066Sahrens } 944fa9e4066Sahrens 945fa9e4066Sahrens int 946fa9e4066Sahrens spa_state(spa_t *spa) 947fa9e4066Sahrens { 948fa9e4066Sahrens return (spa->spa_state); 949fa9e4066Sahrens } 950fa9e4066Sahrens 951fa9e4066Sahrens uint64_t 952fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 953fa9e4066Sahrens { 954fa9e4066Sahrens return (spa->spa_freeze_txg); 955fa9e4066Sahrens } 956fa9e4066Sahrens 957fa9e4066Sahrens /* 95899653d4eSeschrock * Return how much space is allocated in the pool (ie. sum of all asize) 959fa9e4066Sahrens */ 960fa9e4066Sahrens uint64_t 961fa9e4066Sahrens spa_get_alloc(spa_t *spa) 962fa9e4066Sahrens { 963fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_alloc); 964fa9e4066Sahrens } 965fa9e4066Sahrens 966fa9e4066Sahrens /* 96799653d4eSeschrock * Return how much (raid-z inflated) space there is in the pool. 968fa9e4066Sahrens */ 969fa9e4066Sahrens uint64_t 970fa9e4066Sahrens spa_get_space(spa_t *spa) 971fa9e4066Sahrens { 972fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_space); 973fa9e4066Sahrens } 974fa9e4066Sahrens 97599653d4eSeschrock /* 97699653d4eSeschrock * Return the amount of raid-z-deflated space in the pool. 97799653d4eSeschrock */ 97899653d4eSeschrock uint64_t 97999653d4eSeschrock spa_get_dspace(spa_t *spa) 98099653d4eSeschrock { 98199653d4eSeschrock if (spa->spa_deflate) 98299653d4eSeschrock return (spa->spa_root_vdev->vdev_stat.vs_dspace); 98399653d4eSeschrock else 98499653d4eSeschrock return (spa->spa_root_vdev->vdev_stat.vs_space); 98599653d4eSeschrock } 98699653d4eSeschrock 987fa9e4066Sahrens /* ARGSUSED */ 988fa9e4066Sahrens uint64_t 989fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 990fa9e4066Sahrens { 991fa9e4066Sahrens /* 992fa9e4066Sahrens * For now, the worst case is 512-byte RAID-Z blocks, in which 993fa9e4066Sahrens * case the space requirement is exactly 2x; so just assume that. 99444cd46caSbillm * Add to this the fact that we can have up to 3 DVAs per bp, and 99544cd46caSbillm * we have to multiply by a total of 6x. 99644cd46caSbillm */ 99744cd46caSbillm return (lsize * 6); 99844cd46caSbillm } 99944cd46caSbillm 10000a4e9518Sgw /* 10010a4e9518Sgw * Return the failure mode that has been set to this pool. The default 10020a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 10030a4e9518Sgw */ 10040a4e9518Sgw uint8_t 10050a4e9518Sgw spa_get_failmode(spa_t *spa) 10060a4e9518Sgw { 10070a4e9518Sgw return (spa->spa_failmode); 10080a4e9518Sgw } 10090a4e9518Sgw 101044cd46caSbillm uint64_t 101144cd46caSbillm spa_version(spa_t *spa) 101244cd46caSbillm { 101344cd46caSbillm return (spa->spa_ubsync.ub_version); 101444cd46caSbillm } 101544cd46caSbillm 101644cd46caSbillm int 101744cd46caSbillm spa_max_replication(spa_t *spa) 101844cd46caSbillm { 101944cd46caSbillm /* 1020e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 102144cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 102244cd46caSbillm * replication level accordingly. 1023fa9e4066Sahrens */ 1024e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 102544cd46caSbillm return (1); 102644cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1027fa9e4066Sahrens } 1028fa9e4066Sahrens 102999653d4eSeschrock uint64_t 103099653d4eSeschrock bp_get_dasize(spa_t *spa, const blkptr_t *bp) 103199653d4eSeschrock { 103299653d4eSeschrock int sz = 0, i; 103399653d4eSeschrock 103499653d4eSeschrock if (!spa->spa_deflate) 103599653d4eSeschrock return (BP_GET_ASIZE(bp)); 103699653d4eSeschrock 10371934e92fSmaybee spa_config_enter(spa, RW_READER, FTAG); 103899653d4eSeschrock for (i = 0; i < SPA_DVAS_PER_BP; i++) { 103999653d4eSeschrock vdev_t *vd = 104099653d4eSeschrock vdev_lookup_top(spa, DVA_GET_VDEV(&bp->blk_dva[i])); 10411934e92fSmaybee if (vd) 10421934e92fSmaybee sz += (DVA_GET_ASIZE(&bp->blk_dva[i]) >> 10431934e92fSmaybee SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 104499653d4eSeschrock } 10451934e92fSmaybee spa_config_exit(spa, FTAG); 104699653d4eSeschrock return (sz); 104799653d4eSeschrock } 104899653d4eSeschrock 1049fa9e4066Sahrens /* 1050fa9e4066Sahrens * ========================================================================== 1051fa9e4066Sahrens * Initialization and Termination 1052fa9e4066Sahrens * ========================================================================== 1053fa9e4066Sahrens */ 1054fa9e4066Sahrens 1055fa9e4066Sahrens static int 1056fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1057fa9e4066Sahrens { 1058fa9e4066Sahrens const spa_t *s1 = a1; 1059fa9e4066Sahrens const spa_t *s2 = a2; 1060fa9e4066Sahrens int s; 1061fa9e4066Sahrens 1062fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1063fa9e4066Sahrens if (s > 0) 1064fa9e4066Sahrens return (1); 1065fa9e4066Sahrens if (s < 0) 1066fa9e4066Sahrens return (-1); 1067fa9e4066Sahrens return (0); 1068fa9e4066Sahrens } 1069fa9e4066Sahrens 10700373e76bSbonwick int 10710373e76bSbonwick spa_busy(void) 10720373e76bSbonwick { 10730373e76bSbonwick return (spa_active_count); 10740373e76bSbonwick } 10750373e76bSbonwick 1076fa9e4066Sahrens void 1077fa9e4066Sahrens spa_init(int mode) 1078fa9e4066Sahrens { 1079fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1080c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1081fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1082fa9e4066Sahrens 1083fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1084fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1085fa9e4066Sahrens 108699653d4eSeschrock avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_spare_t), 108799653d4eSeschrock offsetof(spa_spare_t, spare_avl)); 108899653d4eSeschrock 1089fa9e4066Sahrens spa_mode = mode; 1090fa9e4066Sahrens 1091fa9e4066Sahrens refcount_init(); 1092fa9e4066Sahrens unique_init(); 1093fa9e4066Sahrens zio_init(); 1094fa9e4066Sahrens dmu_init(); 1095fa9e4066Sahrens zil_init(); 109691ebeef5Sahrens zfs_prop_init(); 1097990b4856Slling zpool_prop_init(); 1098fa9e4066Sahrens spa_config_load(); 1099fa9e4066Sahrens } 1100fa9e4066Sahrens 1101fa9e4066Sahrens void 1102fa9e4066Sahrens spa_fini(void) 1103fa9e4066Sahrens { 1104fa9e4066Sahrens spa_evict_all(); 1105fa9e4066Sahrens 1106fa9e4066Sahrens zil_fini(); 1107fa9e4066Sahrens dmu_fini(); 1108fa9e4066Sahrens zio_fini(); 110991ebeef5Sahrens unique_fini(); 1110fa9e4066Sahrens refcount_fini(); 1111fa9e4066Sahrens 1112fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 111399653d4eSeschrock avl_destroy(&spa_spare_avl); 1114fa9e4066Sahrens 1115fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1116fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1117c25056deSgw mutex_destroy(&spa_spare_lock); 1118fa9e4066Sahrens } 11196ce0521aSperrin 11206ce0521aSperrin /* 11216ce0521aSperrin * Return whether this pool has slogs. No locking needed. 11226ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 11236ce0521aSperrin * performance and not correctness 11246ce0521aSperrin */ 11256ce0521aSperrin boolean_t 11266ce0521aSperrin spa_has_slogs(spa_t *spa) 11276ce0521aSperrin { 11286ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 11296ce0521aSperrin } 1130