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> 47fa9e4066Sahrens 48fa9e4066Sahrens /* 49fa9e4066Sahrens * SPA locking 50fa9e4066Sahrens * 51fa9e4066Sahrens * There are four basic locks for managing spa_t structures: 52fa9e4066Sahrens * 53fa9e4066Sahrens * spa_namespace_lock (global mutex) 54fa9e4066Sahrens * 5544cd46caSbillm * This lock must be acquired to do any of the following: 56fa9e4066Sahrens * 5744cd46caSbillm * - Lookup a spa_t by name 5844cd46caSbillm * - Add or remove a spa_t from the namespace 5944cd46caSbillm * - Increase spa_refcount from non-zero 6044cd46caSbillm * - Check if spa_refcount is zero 6144cd46caSbillm * - Rename a spa_t 62ea8dc4b6Seschrock * - add/remove/attach/detach devices 6344cd46caSbillm * - Held for the duration of create/destroy/import/export 64fa9e4066Sahrens * 6544cd46caSbillm * It does not need to handle recursion. A create or destroy may 6644cd46caSbillm * reference objects (files or zvols) in other pools, but by 6744cd46caSbillm * definition they must have an existing reference, and will never need 6844cd46caSbillm * to lookup a spa_t by name. 69fa9e4066Sahrens * 70fa9e4066Sahrens * spa_refcount (per-spa refcount_t protected by mutex) 71fa9e4066Sahrens * 7244cd46caSbillm * This reference count keep track of any active users of the spa_t. The 7344cd46caSbillm * spa_t cannot be destroyed or freed while this is non-zero. Internally, 7444cd46caSbillm * the refcount is never really 'zero' - opening a pool implicitly keeps 7544cd46caSbillm * some references in the DMU. Internally we check against SPA_MINREF, but 7644cd46caSbillm * present the image of a zero/non-zero value to consumers. 77fa9e4066Sahrens * 78fa9e4066Sahrens * spa_config_lock (per-spa crazy rwlock) 79fa9e4066Sahrens * 8044cd46caSbillm * This SPA special is a recursive rwlock, capable of being acquired from 8144cd46caSbillm * asynchronous threads. It has protects the spa_t from config changes, 8244cd46caSbillm * and must be held in the following circumstances: 83fa9e4066Sahrens * 8444cd46caSbillm * - RW_READER to perform I/O to the spa 8544cd46caSbillm * - RW_WRITER to change the vdev config 86fa9e4066Sahrens * 87fa9e4066Sahrens * spa_config_cache_lock (per-spa mutex) 88fa9e4066Sahrens * 8944cd46caSbillm * This mutex prevents the spa_config nvlist from being updated. No 90fa9e4066Sahrens * other locks are required to obtain this lock, although implicitly you 91fa9e4066Sahrens * must have the namespace lock or non-zero refcount to have any kind 92fa9e4066Sahrens * of spa_t pointer at all. 93fa9e4066Sahrens * 94fa9e4066Sahrens * The locking order is fairly straightforward: 95fa9e4066Sahrens * 9644cd46caSbillm * spa_namespace_lock -> spa_refcount 97fa9e4066Sahrens * 9844cd46caSbillm * The namespace lock must be acquired to increase the refcount from 0 9944cd46caSbillm * or to check if it is zero. 100fa9e4066Sahrens * 10144cd46caSbillm * spa_refcount -> spa_config_lock 102fa9e4066Sahrens * 10344cd46caSbillm * There must be at least one valid reference on the spa_t to acquire 10444cd46caSbillm * the config lock. 105fa9e4066Sahrens * 10644cd46caSbillm * spa_namespace_lock -> spa_config_lock 107fa9e4066Sahrens * 10844cd46caSbillm * The namespace lock must always be taken before the config lock. 109fa9e4066Sahrens * 110fa9e4066Sahrens * 111fa9e4066Sahrens * The spa_namespace_lock and spa_config_cache_lock can be acquired directly and 112fa9e4066Sahrens * are globally visible. 113fa9e4066Sahrens * 114fa9e4066Sahrens * The namespace is manipulated using the following functions, all which require 115fa9e4066Sahrens * the spa_namespace_lock to be held. 116fa9e4066Sahrens * 11744cd46caSbillm * spa_lookup() Lookup a spa_t by name. 118fa9e4066Sahrens * 11944cd46caSbillm * spa_add() Create a new spa_t in the namespace. 120fa9e4066Sahrens * 12144cd46caSbillm * spa_remove() Remove a spa_t from the namespace. This also 12244cd46caSbillm * frees up any memory associated with the spa_t. 123fa9e4066Sahrens * 12444cd46caSbillm * spa_next() Returns the next spa_t in the system, or the 12544cd46caSbillm * first if NULL is passed. 126fa9e4066Sahrens * 12744cd46caSbillm * spa_evict_all() Shutdown and remove all spa_t structures in 12844cd46caSbillm * the system. 129fa9e4066Sahrens * 130ea8dc4b6Seschrock * spa_guid_exists() Determine whether a pool/device guid exists. 131fa9e4066Sahrens * 132fa9e4066Sahrens * The spa_refcount is manipulated using the following functions: 133fa9e4066Sahrens * 13444cd46caSbillm * spa_open_ref() Adds a reference to the given spa_t. Must be 13544cd46caSbillm * called with spa_namespace_lock held if the 13644cd46caSbillm * refcount is currently zero. 137fa9e4066Sahrens * 13844cd46caSbillm * spa_close() Remove a reference from the spa_t. This will 13944cd46caSbillm * not free the spa_t or remove it from the 14044cd46caSbillm * namespace. No locking is required. 141fa9e4066Sahrens * 14244cd46caSbillm * spa_refcount_zero() Returns true if the refcount is currently 14344cd46caSbillm * zero. Must be called with spa_namespace_lock 14444cd46caSbillm * held. 145fa9e4066Sahrens * 146fa9e4066Sahrens * The spa_config_lock is manipulated using the following functions: 147fa9e4066Sahrens * 14844cd46caSbillm * spa_config_enter() Acquire the config lock as RW_READER or 14944cd46caSbillm * RW_WRITER. At least one reference on the spa_t 15044cd46caSbillm * must exist. 151fa9e4066Sahrens * 15244cd46caSbillm * spa_config_exit() Release the config lock. 153fa9e4066Sahrens * 15444cd46caSbillm * spa_config_held() Returns true if the config lock is currently 15544cd46caSbillm * held in the given state. 156fa9e4066Sahrens * 157ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 158fa9e4066Sahrens * 15944cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 160ea8dc4b6Seschrock * for writing. 161fa9e4066Sahrens * 16244cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 16344cd46caSbillm * to complete, sync the updated configs to the 164ea8dc4b6Seschrock * cache, and release the namespace lock. 165fa9e4066Sahrens * 166fa9e4066Sahrens * The spa_name() function also requires either the spa_namespace_lock 167fa9e4066Sahrens * or the spa_config_lock, as both are needed to do a rename. spa_rename() is 168fa9e4066Sahrens * also implemented within this file since is requires manipulation of the 169fa9e4066Sahrens * namespace. 170fa9e4066Sahrens */ 171fa9e4066Sahrens 172fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 173fa9e4066Sahrens kmutex_t spa_namespace_lock; 174fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 1750373e76bSbonwick static int spa_active_count; 176416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 177fa9e4066Sahrens 17899653d4eSeschrock static kmutex_t spa_spare_lock; 17939c23413Seschrock static avl_tree_t spa_spare_avl; 18099653d4eSeschrock 181fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 182fa9e4066Sahrens int spa_mode; 183fa9e4066Sahrens 184fa9e4066Sahrens #ifdef ZFS_DEBUG 185*40feaa91Sahrens /* Everything except dprintf is on by default in debug builds */ 186*40feaa91Sahrens int zfs_flags = ~ZFS_DEBUG_DPRINTF; 187fa9e4066Sahrens #else 188fa9e4066Sahrens int zfs_flags = 0; 189fa9e4066Sahrens #endif 190fa9e4066Sahrens 1910125049cSahrens /* 1920125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 1930125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 1940125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 1950125049cSahrens */ 1960125049cSahrens int zfs_recover = 0; 1970125049cSahrens 198fa9e4066Sahrens #define SPA_MINREF 5 /* spa_refcnt for an open-but-idle pool */ 199fa9e4066Sahrens 200fa9e4066Sahrens /* 201fa9e4066Sahrens * ========================================================================== 202fa9e4066Sahrens * SPA namespace functions 203fa9e4066Sahrens * ========================================================================== 204fa9e4066Sahrens */ 205fa9e4066Sahrens 206fa9e4066Sahrens /* 207fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 208fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 209fa9e4066Sahrens */ 210fa9e4066Sahrens spa_t * 211fa9e4066Sahrens spa_lookup(const char *name) 212fa9e4066Sahrens { 213fa9e4066Sahrens spa_t search, *spa; 214fa9e4066Sahrens avl_index_t where; 215*40feaa91Sahrens char c; 216*40feaa91Sahrens char *cp; 217fa9e4066Sahrens 218fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 219fa9e4066Sahrens 220*40feaa91Sahrens /* 221*40feaa91Sahrens * If it's a full dataset name, figure out the pool name and 222*40feaa91Sahrens * just use that. 223*40feaa91Sahrens */ 224*40feaa91Sahrens cp = strpbrk(name, "/@"); 225*40feaa91Sahrens if (cp) { 226*40feaa91Sahrens c = *cp; 227*40feaa91Sahrens *cp = '\0'; 228*40feaa91Sahrens } 229*40feaa91Sahrens 230fa9e4066Sahrens search.spa_name = (char *)name; 231fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 232fa9e4066Sahrens 233*40feaa91Sahrens if (cp) 234*40feaa91Sahrens *cp = c; 235*40feaa91Sahrens 236fa9e4066Sahrens return (spa); 237fa9e4066Sahrens } 238fa9e4066Sahrens 239fa9e4066Sahrens /* 240fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 241fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 242fa9e4066Sahrens * exist by calling spa_lookup() first. 243fa9e4066Sahrens */ 244fa9e4066Sahrens spa_t * 2450373e76bSbonwick spa_add(const char *name, const char *altroot) 246fa9e4066Sahrens { 247fa9e4066Sahrens spa_t *spa; 248fa9e4066Sahrens 249fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 250fa9e4066Sahrens 251fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 252fa9e4066Sahrens 253fa9e4066Sahrens spa->spa_name = spa_strdup(name); 254fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 255fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 2560373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 257fa9e4066Sahrens 258fa9e4066Sahrens refcount_create(&spa->spa_refcount); 259ea8dc4b6Seschrock refcount_create(&spa->spa_config_lock.scl_count); 260fa9e4066Sahrens 261fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 262fa9e4066Sahrens 2630373e76bSbonwick /* 2640373e76bSbonwick * Set the alternate root, if there is one. 2650373e76bSbonwick */ 2660373e76bSbonwick if (altroot) { 2670373e76bSbonwick spa->spa_root = spa_strdup(altroot); 2680373e76bSbonwick spa_active_count++; 2690373e76bSbonwick } 2700373e76bSbonwick 271fa9e4066Sahrens return (spa); 272fa9e4066Sahrens } 273fa9e4066Sahrens 274fa9e4066Sahrens /* 275fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 276fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 277fa9e4066Sahrens * deactivated. 278fa9e4066Sahrens */ 279fa9e4066Sahrens void 280fa9e4066Sahrens spa_remove(spa_t *spa) 281fa9e4066Sahrens { 282fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 283fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 284fa9e4066Sahrens ASSERT(spa->spa_scrub_thread == NULL); 285fa9e4066Sahrens 286fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 287fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 288fa9e4066Sahrens 2890373e76bSbonwick if (spa->spa_root) { 290fa9e4066Sahrens spa_strfree(spa->spa_root); 2910373e76bSbonwick spa_active_count--; 2920373e76bSbonwick } 293fa9e4066Sahrens 294fa9e4066Sahrens if (spa->spa_name) 295fa9e4066Sahrens spa_strfree(spa->spa_name); 296fa9e4066Sahrens 297fa9e4066Sahrens spa_config_set(spa, NULL); 298fa9e4066Sahrens 299fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 300ea8dc4b6Seschrock refcount_destroy(&spa->spa_config_lock.scl_count); 301fa9e4066Sahrens 3025ad82045Snd mutex_destroy(&spa->spa_sync_bplist.bpl_lock); 3035ad82045Snd mutex_destroy(&spa->spa_config_lock.scl_lock); 3045ad82045Snd mutex_destroy(&spa->spa_errlist_lock); 3055ad82045Snd mutex_destroy(&spa->spa_errlog_lock); 3065ad82045Snd mutex_destroy(&spa->spa_scrub_lock); 3075ad82045Snd mutex_destroy(&spa->spa_config_cache_lock); 3085ad82045Snd mutex_destroy(&spa->spa_async_lock); 30906eeb2adSek mutex_destroy(&spa->spa_history_lock); 310b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 3115ad82045Snd 312fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 313fa9e4066Sahrens } 314fa9e4066Sahrens 315fa9e4066Sahrens /* 316fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 317fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 318fa9e4066Sahrens */ 319fa9e4066Sahrens spa_t * 320fa9e4066Sahrens spa_next(spa_t *prev) 321fa9e4066Sahrens { 322fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 323fa9e4066Sahrens 324fa9e4066Sahrens if (prev) 325fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 326fa9e4066Sahrens else 327fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 328fa9e4066Sahrens } 329fa9e4066Sahrens 330fa9e4066Sahrens /* 331fa9e4066Sahrens * ========================================================================== 332fa9e4066Sahrens * SPA refcount functions 333fa9e4066Sahrens * ========================================================================== 334fa9e4066Sahrens */ 335fa9e4066Sahrens 336fa9e4066Sahrens /* 337fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 338fa9e4066Sahrens * have the namespace lock held. 339fa9e4066Sahrens */ 340fa9e4066Sahrens void 341fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 342fa9e4066Sahrens { 343fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 344fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 345fa9e4066Sahrens 346fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 347fa9e4066Sahrens } 348fa9e4066Sahrens 349fa9e4066Sahrens /* 350fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 351fa9e4066Sahrens * have the namespace lock held. 352fa9e4066Sahrens */ 353fa9e4066Sahrens void 354fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 355fa9e4066Sahrens { 356fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 357fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 358fa9e4066Sahrens 359fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 360fa9e4066Sahrens } 361fa9e4066Sahrens 362fa9e4066Sahrens /* 363fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 364fa9e4066Sahrens * spa_namespace_lock held. We really compare against SPA_MINREF, which is the 365fa9e4066Sahrens * number of references acquired when opening a pool 366fa9e4066Sahrens */ 367fa9e4066Sahrens boolean_t 368fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 369fa9e4066Sahrens { 370fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 371fa9e4066Sahrens 372fa9e4066Sahrens return (refcount_count(&spa->spa_refcount) == SPA_MINREF); 373fa9e4066Sahrens } 374fa9e4066Sahrens 37599653d4eSeschrock /* 37699653d4eSeschrock * ========================================================================== 37799653d4eSeschrock * SPA spare tracking 37899653d4eSeschrock * ========================================================================== 37999653d4eSeschrock */ 38099653d4eSeschrock 38199653d4eSeschrock /* 38239c23413Seschrock * Spares are tracked globally due to the following constraints: 38339c23413Seschrock * 38439c23413Seschrock * - A spare may be part of multiple pools. 38539c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 38639c23413Seschrock * another pool. 38739c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 38839c23413Seschrock * the target is a spare in the same pool. 38939c23413Seschrock * 39039c23413Seschrock * We keep track of all spares on the system through the use of a reference 39139c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 39239c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 39339c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 39439c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 39539c23413Seschrock * pool), we also keep track of which pool its been made a part of. 39639c23413Seschrock * 39739c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 39839c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 39939c23413Seschrock * separate spare lock exists for the status query path, which does not need to 40039c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 40199653d4eSeschrock */ 40239c23413Seschrock 40399653d4eSeschrock typedef struct spa_spare { 40499653d4eSeschrock uint64_t spare_guid; 40539c23413Seschrock uint64_t spare_pool; 40699653d4eSeschrock avl_node_t spare_avl; 40799653d4eSeschrock int spare_count; 40899653d4eSeschrock } spa_spare_t; 40999653d4eSeschrock 41099653d4eSeschrock static int 41199653d4eSeschrock spa_spare_compare(const void *a, const void *b) 41299653d4eSeschrock { 41399653d4eSeschrock const spa_spare_t *sa = a; 41499653d4eSeschrock const spa_spare_t *sb = b; 41599653d4eSeschrock 41699653d4eSeschrock if (sa->spare_guid < sb->spare_guid) 41799653d4eSeschrock return (-1); 41899653d4eSeschrock else if (sa->spare_guid > sb->spare_guid) 41999653d4eSeschrock return (1); 42099653d4eSeschrock else 42199653d4eSeschrock return (0); 42299653d4eSeschrock } 42399653d4eSeschrock 42499653d4eSeschrock void 42539c23413Seschrock spa_spare_add(vdev_t *vd) 42699653d4eSeschrock { 42799653d4eSeschrock avl_index_t where; 42899653d4eSeschrock spa_spare_t search; 42999653d4eSeschrock spa_spare_t *spare; 43099653d4eSeschrock 43199653d4eSeschrock mutex_enter(&spa_spare_lock); 43239c23413Seschrock ASSERT(!vd->vdev_isspare); 43399653d4eSeschrock 43439c23413Seschrock search.spare_guid = vd->vdev_guid; 43599653d4eSeschrock if ((spare = avl_find(&spa_spare_avl, &search, &where)) != NULL) { 43699653d4eSeschrock spare->spare_count++; 43799653d4eSeschrock } else { 43839c23413Seschrock spare = kmem_zalloc(sizeof (spa_spare_t), KM_SLEEP); 43939c23413Seschrock spare->spare_guid = vd->vdev_guid; 44099653d4eSeschrock spare->spare_count = 1; 44199653d4eSeschrock avl_insert(&spa_spare_avl, spare, where); 44299653d4eSeschrock } 44339c23413Seschrock vd->vdev_isspare = B_TRUE; 44499653d4eSeschrock 44599653d4eSeschrock mutex_exit(&spa_spare_lock); 44699653d4eSeschrock } 44799653d4eSeschrock 44899653d4eSeschrock void 44939c23413Seschrock spa_spare_remove(vdev_t *vd) 45099653d4eSeschrock { 45199653d4eSeschrock spa_spare_t search; 45299653d4eSeschrock spa_spare_t *spare; 45399653d4eSeschrock avl_index_t where; 45499653d4eSeschrock 45599653d4eSeschrock mutex_enter(&spa_spare_lock); 45699653d4eSeschrock 45739c23413Seschrock search.spare_guid = vd->vdev_guid; 45899653d4eSeschrock spare = avl_find(&spa_spare_avl, &search, &where); 45999653d4eSeschrock 46039c23413Seschrock ASSERT(vd->vdev_isspare); 46199653d4eSeschrock ASSERT(spare != NULL); 46299653d4eSeschrock 46399653d4eSeschrock if (--spare->spare_count == 0) { 46499653d4eSeschrock avl_remove(&spa_spare_avl, spare); 46599653d4eSeschrock kmem_free(spare, sizeof (spa_spare_t)); 46639c23413Seschrock } else if (spare->spare_pool == spa_guid(vd->vdev_spa)) { 46739c23413Seschrock spare->spare_pool = 0ULL; 46899653d4eSeschrock } 46999653d4eSeschrock 47039c23413Seschrock vd->vdev_isspare = B_FALSE; 47199653d4eSeschrock mutex_exit(&spa_spare_lock); 47299653d4eSeschrock } 47399653d4eSeschrock 47499653d4eSeschrock boolean_t 47539c23413Seschrock spa_spare_exists(uint64_t guid, uint64_t *pool) 47699653d4eSeschrock { 47739c23413Seschrock spa_spare_t search, *found; 47899653d4eSeschrock avl_index_t where; 47999653d4eSeschrock 48099653d4eSeschrock mutex_enter(&spa_spare_lock); 48199653d4eSeschrock 48299653d4eSeschrock search.spare_guid = guid; 48339c23413Seschrock found = avl_find(&spa_spare_avl, &search, &where); 48439c23413Seschrock 48539c23413Seschrock if (pool) { 48639c23413Seschrock if (found) 48739c23413Seschrock *pool = found->spare_pool; 48839c23413Seschrock else 48939c23413Seschrock *pool = 0ULL; 49039c23413Seschrock } 49199653d4eSeschrock 49299653d4eSeschrock mutex_exit(&spa_spare_lock); 49399653d4eSeschrock 49439c23413Seschrock return (found != NULL); 49539c23413Seschrock } 49639c23413Seschrock 49739c23413Seschrock void 49839c23413Seschrock spa_spare_activate(vdev_t *vd) 49939c23413Seschrock { 50039c23413Seschrock spa_spare_t search, *found; 50139c23413Seschrock avl_index_t where; 50239c23413Seschrock 50339c23413Seschrock mutex_enter(&spa_spare_lock); 50439c23413Seschrock ASSERT(vd->vdev_isspare); 50539c23413Seschrock 50639c23413Seschrock search.spare_guid = vd->vdev_guid; 50739c23413Seschrock found = avl_find(&spa_spare_avl, &search, &where); 50839c23413Seschrock ASSERT(found != NULL); 50939c23413Seschrock ASSERT(found->spare_pool == 0ULL); 51039c23413Seschrock 51139c23413Seschrock found->spare_pool = spa_guid(vd->vdev_spa); 51239c23413Seschrock mutex_exit(&spa_spare_lock); 51399653d4eSeschrock } 51499653d4eSeschrock 515fa9e4066Sahrens /* 516fa9e4066Sahrens * ========================================================================== 517fa9e4066Sahrens * SPA config locking 518fa9e4066Sahrens * ========================================================================== 519fa9e4066Sahrens */ 520fa9e4066Sahrens 521fa9e4066Sahrens /* 522fa9e4066Sahrens * Acquire the config lock. The config lock is a special rwlock that allows for 523fa9e4066Sahrens * recursive enters. Because these enters come from the same thread as well as 524fa9e4066Sahrens * asynchronous threads working on behalf of the owner, we must unilaterally 525fa9e4066Sahrens * allow all reads access as long at least one reader is held (even if a write 526fa9e4066Sahrens * is requested). This has the side effect of write starvation, but write locks 527fa9e4066Sahrens * are extremely rare, and a solution to this problem would be significantly 528fa9e4066Sahrens * more complex (if even possible). 529fa9e4066Sahrens * 530fa9e4066Sahrens * We would like to assert that the namespace lock isn't held, but this is a 531fa9e4066Sahrens * valid use during create. 532fa9e4066Sahrens */ 533fa9e4066Sahrens void 534ea8dc4b6Seschrock spa_config_enter(spa_t *spa, krw_t rw, void *tag) 535fa9e4066Sahrens { 536fa9e4066Sahrens spa_config_lock_t *scl = &spa->spa_config_lock; 537fa9e4066Sahrens 538fa9e4066Sahrens mutex_enter(&scl->scl_lock); 539fa9e4066Sahrens 540fa9e4066Sahrens if (scl->scl_writer != curthread) { 541fa9e4066Sahrens if (rw == RW_READER) { 542fa9e4066Sahrens while (scl->scl_writer != NULL) 543fa9e4066Sahrens cv_wait(&scl->scl_cv, &scl->scl_lock); 544fa9e4066Sahrens } else { 545ea8dc4b6Seschrock while (scl->scl_writer != NULL || 546ea8dc4b6Seschrock !refcount_is_zero(&scl->scl_count)) 547fa9e4066Sahrens cv_wait(&scl->scl_cv, &scl->scl_lock); 548fa9e4066Sahrens scl->scl_writer = curthread; 549fa9e4066Sahrens } 550fa9e4066Sahrens } 551fa9e4066Sahrens 552ea8dc4b6Seschrock (void) refcount_add(&scl->scl_count, tag); 553fa9e4066Sahrens 554fa9e4066Sahrens mutex_exit(&scl->scl_lock); 555fa9e4066Sahrens } 556fa9e4066Sahrens 557fa9e4066Sahrens /* 558fa9e4066Sahrens * Release the spa config lock, notifying any waiters in the process. 559fa9e4066Sahrens */ 560fa9e4066Sahrens void 561ea8dc4b6Seschrock spa_config_exit(spa_t *spa, void *tag) 562fa9e4066Sahrens { 563fa9e4066Sahrens spa_config_lock_t *scl = &spa->spa_config_lock; 564fa9e4066Sahrens 565fa9e4066Sahrens mutex_enter(&scl->scl_lock); 566fa9e4066Sahrens 567ea8dc4b6Seschrock ASSERT(!refcount_is_zero(&scl->scl_count)); 568ea8dc4b6Seschrock if (refcount_remove(&scl->scl_count, tag) == 0) { 569fa9e4066Sahrens cv_broadcast(&scl->scl_cv); 570fa9e4066Sahrens scl->scl_writer = NULL; /* OK in either case */ 571fa9e4066Sahrens } 572fa9e4066Sahrens 573fa9e4066Sahrens mutex_exit(&scl->scl_lock); 574fa9e4066Sahrens } 575fa9e4066Sahrens 576fa9e4066Sahrens /* 577fa9e4066Sahrens * Returns true if the config lock is held in the given manner. 578fa9e4066Sahrens */ 579fa9e4066Sahrens boolean_t 580fa9e4066Sahrens spa_config_held(spa_t *spa, krw_t rw) 581fa9e4066Sahrens { 582fa9e4066Sahrens spa_config_lock_t *scl = &spa->spa_config_lock; 583fa9e4066Sahrens boolean_t held; 584fa9e4066Sahrens 585fa9e4066Sahrens mutex_enter(&scl->scl_lock); 586fa9e4066Sahrens if (rw == RW_WRITER) 587fa9e4066Sahrens held = (scl->scl_writer == curthread); 588fa9e4066Sahrens else 589ea8dc4b6Seschrock held = !refcount_is_zero(&scl->scl_count); 590fa9e4066Sahrens mutex_exit(&scl->scl_lock); 591fa9e4066Sahrens 592fa9e4066Sahrens return (held); 593fa9e4066Sahrens } 594fa9e4066Sahrens 595fa9e4066Sahrens /* 596fa9e4066Sahrens * ========================================================================== 597fa9e4066Sahrens * SPA vdev locking 598fa9e4066Sahrens * ========================================================================== 599fa9e4066Sahrens */ 600fa9e4066Sahrens 601fa9e4066Sahrens /* 602ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 603ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 604fa9e4066Sahrens * It returns the next transaction group for the spa_t. 605fa9e4066Sahrens */ 606fa9e4066Sahrens uint64_t 607fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 608fa9e4066Sahrens { 6093d7072f8Seschrock mutex_enter(&spa_namespace_lock); 6103d7072f8Seschrock 611ea8dc4b6Seschrock /* 6123d7072f8Seschrock * Suspend scrub activity while we mess with the config. We must do 6133d7072f8Seschrock * this after acquiring the namespace lock to avoid a 3-way deadlock 6143d7072f8Seschrock * with spa_scrub_stop() and the scrub thread. 615ea8dc4b6Seschrock */ 616ea8dc4b6Seschrock spa_scrub_suspend(spa); 617fa9e4066Sahrens 618ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, spa); 619fa9e4066Sahrens 620fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 621fa9e4066Sahrens } 622fa9e4066Sahrens 623fa9e4066Sahrens /* 624fa9e4066Sahrens * Unlock the spa_t after adding or removing a vdev. Besides undoing the 625fa9e4066Sahrens * locking of spa_vdev_enter(), we also want make sure the transactions have 626fa9e4066Sahrens * synced to disk, and then update the global configuration cache with the new 627fa9e4066Sahrens * information. 628fa9e4066Sahrens */ 629fa9e4066Sahrens int 630fa9e4066Sahrens spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 631fa9e4066Sahrens { 6320e34b6a7Sbonwick int config_changed = B_FALSE; 633ea8dc4b6Seschrock 6340373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 6350e34b6a7Sbonwick 6360e34b6a7Sbonwick /* 6370e34b6a7Sbonwick * Reassess the DTLs. 6380e34b6a7Sbonwick */ 6390373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 6400e34b6a7Sbonwick 6410e34b6a7Sbonwick /* 6420373e76bSbonwick * If the config changed, notify the scrub thread that it must restart. 6430e34b6a7Sbonwick */ 6440e34b6a7Sbonwick if (error == 0 && !list_is_empty(&spa->spa_dirty_list)) { 6450e34b6a7Sbonwick config_changed = B_TRUE; 6460373e76bSbonwick spa_scrub_restart(spa, txg); 6470e34b6a7Sbonwick } 648ea8dc4b6Seschrock 649ea8dc4b6Seschrock spa_config_exit(spa, spa); 650fa9e4066Sahrens 651ea8dc4b6Seschrock /* 6525dabedeeSbonwick * Allow scrubbing to resume. 653ea8dc4b6Seschrock */ 654ea8dc4b6Seschrock spa_scrub_resume(spa); 655fa9e4066Sahrens 656fa9e4066Sahrens /* 657fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 658fa9e4066Sahrens * that there won't be more than one config change per txg. 659fa9e4066Sahrens * This allows us to use the txg as the generation number. 660fa9e4066Sahrens */ 661fa9e4066Sahrens if (error == 0) 662fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 663fa9e4066Sahrens 664fa9e4066Sahrens if (vd != NULL) { 665fa9e4066Sahrens ASSERT(!vd->vdev_detached || vd->vdev_dtl.smo_object == 0); 666fa9e4066Sahrens vdev_free(vd); 667fa9e4066Sahrens } 668fa9e4066Sahrens 669fa9e4066Sahrens /* 6700e34b6a7Sbonwick * If the config changed, update the config cache. 671fa9e4066Sahrens */ 6720e34b6a7Sbonwick if (config_changed) 673fa9e4066Sahrens spa_config_sync(); 674ea8dc4b6Seschrock 675ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 676fa9e4066Sahrens 677fa9e4066Sahrens return (error); 678fa9e4066Sahrens } 679fa9e4066Sahrens 680fa9e4066Sahrens /* 681fa9e4066Sahrens * ========================================================================== 682fa9e4066Sahrens * Miscellaneous functions 683fa9e4066Sahrens * ========================================================================== 684fa9e4066Sahrens */ 685fa9e4066Sahrens 686fa9e4066Sahrens /* 687fa9e4066Sahrens * Rename a spa_t. 688fa9e4066Sahrens */ 689fa9e4066Sahrens int 690fa9e4066Sahrens spa_rename(const char *name, const char *newname) 691fa9e4066Sahrens { 692fa9e4066Sahrens spa_t *spa; 693fa9e4066Sahrens int err; 694fa9e4066Sahrens 695fa9e4066Sahrens /* 696fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 697fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 698fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 699ea8dc4b6Seschrock * allow recursive calls for other reasons. 700fa9e4066Sahrens */ 701fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 702fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 703fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 704fa9e4066Sahrens return (err); 705fa9e4066Sahrens } 706fa9e4066Sahrens 707ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 708fa9e4066Sahrens 709fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 710fa9e4066Sahrens spa_strfree(spa->spa_name); 711fa9e4066Sahrens spa->spa_name = spa_strdup(newname); 712fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 713fa9e4066Sahrens 714fa9e4066Sahrens /* 715fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 716fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 717fa9e4066Sahrens * during the sync. 718fa9e4066Sahrens */ 719fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 720fa9e4066Sahrens 721ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 722fa9e4066Sahrens 7230373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 724fa9e4066Sahrens 725fa9e4066Sahrens /* 726fa9e4066Sahrens * Sync the updated config cache. 727fa9e4066Sahrens */ 728fa9e4066Sahrens spa_config_sync(); 729fa9e4066Sahrens 730fa9e4066Sahrens spa_close(spa, FTAG); 731fa9e4066Sahrens 732fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 733fa9e4066Sahrens 734fa9e4066Sahrens return (0); 735fa9e4066Sahrens } 736fa9e4066Sahrens 737fa9e4066Sahrens 738fa9e4066Sahrens /* 739fa9e4066Sahrens * Determine whether a pool with given pool_guid exists. If device_guid is 740fa9e4066Sahrens * non-zero, determine whether the pool exists *and* contains a device with the 741fa9e4066Sahrens * specified device_guid. 742fa9e4066Sahrens */ 743fa9e4066Sahrens boolean_t 744fa9e4066Sahrens spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 745fa9e4066Sahrens { 746fa9e4066Sahrens spa_t *spa; 747fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 748fa9e4066Sahrens 749ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 750fa9e4066Sahrens 751fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 752fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 753fa9e4066Sahrens continue; 754fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 755fa9e4066Sahrens continue; 75639c23413Seschrock if (spa_guid(spa) == pool_guid) { 75739c23413Seschrock if (device_guid == 0) 75839c23413Seschrock break; 75939c23413Seschrock 76039c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 76139c23413Seschrock device_guid) != NULL) 76239c23413Seschrock break; 76339c23413Seschrock 76439c23413Seschrock /* 7658654d025Sperrin * Check any devices we may be in the process of adding. 76639c23413Seschrock */ 76739c23413Seschrock if (spa->spa_pending_vdev) { 76839c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 76939c23413Seschrock device_guid) != NULL) 77039c23413Seschrock break; 77139c23413Seschrock } 77239c23413Seschrock } 773fa9e4066Sahrens } 774fa9e4066Sahrens 775fa9e4066Sahrens return (spa != NULL); 776fa9e4066Sahrens } 777fa9e4066Sahrens 778fa9e4066Sahrens char * 779fa9e4066Sahrens spa_strdup(const char *s) 780fa9e4066Sahrens { 781fa9e4066Sahrens size_t len; 782fa9e4066Sahrens char *new; 783fa9e4066Sahrens 784fa9e4066Sahrens len = strlen(s); 785fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 786fa9e4066Sahrens bcopy(s, new, len); 787fa9e4066Sahrens new[len] = '\0'; 788fa9e4066Sahrens 789fa9e4066Sahrens return (new); 790fa9e4066Sahrens } 791fa9e4066Sahrens 792fa9e4066Sahrens void 793fa9e4066Sahrens spa_strfree(char *s) 794fa9e4066Sahrens { 795fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 796fa9e4066Sahrens } 797fa9e4066Sahrens 798fa9e4066Sahrens uint64_t 799fa9e4066Sahrens spa_get_random(uint64_t range) 800fa9e4066Sahrens { 801fa9e4066Sahrens uint64_t r; 802fa9e4066Sahrens 803fa9e4066Sahrens ASSERT(range != 0); 804fa9e4066Sahrens 805fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 806fa9e4066Sahrens 807fa9e4066Sahrens return (r % range); 808fa9e4066Sahrens } 809fa9e4066Sahrens 810fa9e4066Sahrens void 811d80c45e0Sbonwick sprintf_blkptr(char *buf, int len, const blkptr_t *bp) 812fa9e4066Sahrens { 81344cd46caSbillm int d; 814fa9e4066Sahrens 815fa9e4066Sahrens if (bp == NULL) { 816fbabab8fSmaybee (void) snprintf(buf, len, "<NULL>"); 817fa9e4066Sahrens return; 818fa9e4066Sahrens } 819fa9e4066Sahrens 820fa9e4066Sahrens if (BP_IS_HOLE(bp)) { 821fbabab8fSmaybee (void) snprintf(buf, len, "<hole>"); 822fa9e4066Sahrens return; 823fa9e4066Sahrens } 824fa9e4066Sahrens 82544cd46caSbillm (void) snprintf(buf, len, "[L%llu %s] %llxL/%llxP ", 826fa9e4066Sahrens (u_longlong_t)BP_GET_LEVEL(bp), 827fa9e4066Sahrens dmu_ot[BP_GET_TYPE(bp)].ot_name, 828fa9e4066Sahrens (u_longlong_t)BP_GET_LSIZE(bp), 82944cd46caSbillm (u_longlong_t)BP_GET_PSIZE(bp)); 83044cd46caSbillm 83144cd46caSbillm for (d = 0; d < BP_GET_NDVAS(bp); d++) { 832d80c45e0Sbonwick const dva_t *dva = &bp->blk_dva[d]; 83344cd46caSbillm (void) snprintf(buf + strlen(buf), len - strlen(buf), 83444cd46caSbillm "DVA[%d]=<%llu:%llx:%llx> ", d, 83544cd46caSbillm (u_longlong_t)DVA_GET_VDEV(dva), 83644cd46caSbillm (u_longlong_t)DVA_GET_OFFSET(dva), 83744cd46caSbillm (u_longlong_t)DVA_GET_ASIZE(dva)); 83844cd46caSbillm } 83944cd46caSbillm 84044cd46caSbillm (void) snprintf(buf + strlen(buf), len - strlen(buf), 84144cd46caSbillm "%s %s %s %s birth=%llu fill=%llu cksum=%llx:%llx:%llx:%llx", 842fa9e4066Sahrens zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 843fa9e4066Sahrens zio_compress_table[BP_GET_COMPRESS(bp)].ci_name, 844fa9e4066Sahrens BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", 84544cd46caSbillm BP_IS_GANG(bp) ? "gang" : "contiguous", 846fa9e4066Sahrens (u_longlong_t)bp->blk_birth, 847fa9e4066Sahrens (u_longlong_t)bp->blk_fill, 848fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[0], 849fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[1], 850fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[2], 851fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[3]); 852fa9e4066Sahrens } 853fa9e4066Sahrens 854fa9e4066Sahrens void 855fa9e4066Sahrens spa_freeze(spa_t *spa) 856fa9e4066Sahrens { 857fa9e4066Sahrens uint64_t freeze_txg = 0; 858fa9e4066Sahrens 859ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 860fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 861fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 862fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 863fa9e4066Sahrens } 864ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 865fa9e4066Sahrens if (freeze_txg != 0) 866fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 867fa9e4066Sahrens } 868fa9e4066Sahrens 8690125049cSahrens void 8700125049cSahrens zfs_panic_recover(const char *fmt, ...) 8710125049cSahrens { 8720125049cSahrens va_list adx; 8730125049cSahrens 8740125049cSahrens va_start(adx, fmt); 8750125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 8760125049cSahrens va_end(adx); 8770125049cSahrens } 8780125049cSahrens 879fa9e4066Sahrens /* 880fa9e4066Sahrens * ========================================================================== 881fa9e4066Sahrens * Accessor functions 882fa9e4066Sahrens * ========================================================================== 883fa9e4066Sahrens */ 884fa9e4066Sahrens 885fa9e4066Sahrens krwlock_t * 886fa9e4066Sahrens spa_traverse_rwlock(spa_t *spa) 887fa9e4066Sahrens { 888fa9e4066Sahrens return (&spa->spa_traverse_lock); 889fa9e4066Sahrens } 890fa9e4066Sahrens 891fa9e4066Sahrens int 892fa9e4066Sahrens spa_traverse_wanted(spa_t *spa) 893fa9e4066Sahrens { 894fa9e4066Sahrens return (spa->spa_traverse_wanted); 895fa9e4066Sahrens } 896fa9e4066Sahrens 897fa9e4066Sahrens dsl_pool_t * 898fa9e4066Sahrens spa_get_dsl(spa_t *spa) 899fa9e4066Sahrens { 900fa9e4066Sahrens return (spa->spa_dsl_pool); 901fa9e4066Sahrens } 902fa9e4066Sahrens 903fa9e4066Sahrens blkptr_t * 904fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 905fa9e4066Sahrens { 906fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 907fa9e4066Sahrens } 908fa9e4066Sahrens 909fa9e4066Sahrens void 910fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 911fa9e4066Sahrens { 912fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 913fa9e4066Sahrens } 914fa9e4066Sahrens 915fa9e4066Sahrens void 916fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 917fa9e4066Sahrens { 918fa9e4066Sahrens if (spa->spa_root == NULL) 919fa9e4066Sahrens buf[0] = '\0'; 920fa9e4066Sahrens else 921fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 922fa9e4066Sahrens } 923fa9e4066Sahrens 924fa9e4066Sahrens int 925fa9e4066Sahrens spa_sync_pass(spa_t *spa) 926fa9e4066Sahrens { 927fa9e4066Sahrens return (spa->spa_sync_pass); 928fa9e4066Sahrens } 929fa9e4066Sahrens 930fa9e4066Sahrens char * 931fa9e4066Sahrens spa_name(spa_t *spa) 932fa9e4066Sahrens { 933fa9e4066Sahrens /* 934fa9e4066Sahrens * Accessing the name requires holding either the namespace lock or the 935fa9e4066Sahrens * config lock, both of which are required to do a rename. 936fa9e4066Sahrens */ 937fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock) || 938fa9e4066Sahrens spa_config_held(spa, RW_READER) || spa_config_held(spa, RW_WRITER)); 939fa9e4066Sahrens 940fa9e4066Sahrens return (spa->spa_name); 941fa9e4066Sahrens } 942fa9e4066Sahrens 943fa9e4066Sahrens uint64_t 944fa9e4066Sahrens spa_guid(spa_t *spa) 945fa9e4066Sahrens { 946b5989ec7Seschrock /* 947b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 948b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 949b5989ec7Seschrock * vdev. We stash the original pool guid in 'spa_load_guid' to handle 950b5989ec7Seschrock * this case. 951b5989ec7Seschrock */ 952b5989ec7Seschrock if (spa->spa_root_vdev != NULL) 953b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 954b5989ec7Seschrock else 955b5989ec7Seschrock return (spa->spa_load_guid); 956fa9e4066Sahrens } 957fa9e4066Sahrens 958fa9e4066Sahrens uint64_t 959fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 960fa9e4066Sahrens { 961fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 962fa9e4066Sahrens } 963fa9e4066Sahrens 964fa9e4066Sahrens uint64_t 965fa9e4066Sahrens spa_first_txg(spa_t *spa) 966fa9e4066Sahrens { 967fa9e4066Sahrens return (spa->spa_first_txg); 968fa9e4066Sahrens } 969fa9e4066Sahrens 970fa9e4066Sahrens int 971fa9e4066Sahrens spa_state(spa_t *spa) 972fa9e4066Sahrens { 973fa9e4066Sahrens return (spa->spa_state); 974fa9e4066Sahrens } 975fa9e4066Sahrens 976fa9e4066Sahrens uint64_t 977fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 978fa9e4066Sahrens { 979fa9e4066Sahrens return (spa->spa_freeze_txg); 980fa9e4066Sahrens } 981fa9e4066Sahrens 982fa9e4066Sahrens /* 98399653d4eSeschrock * Return how much space is allocated in the pool (ie. sum of all asize) 984fa9e4066Sahrens */ 985fa9e4066Sahrens uint64_t 986fa9e4066Sahrens spa_get_alloc(spa_t *spa) 987fa9e4066Sahrens { 988fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_alloc); 989fa9e4066Sahrens } 990fa9e4066Sahrens 991fa9e4066Sahrens /* 99299653d4eSeschrock * Return how much (raid-z inflated) space there is in the pool. 993fa9e4066Sahrens */ 994fa9e4066Sahrens uint64_t 995fa9e4066Sahrens spa_get_space(spa_t *spa) 996fa9e4066Sahrens { 997fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_space); 998fa9e4066Sahrens } 999fa9e4066Sahrens 100099653d4eSeschrock /* 100199653d4eSeschrock * Return the amount of raid-z-deflated space in the pool. 100299653d4eSeschrock */ 100399653d4eSeschrock uint64_t 100499653d4eSeschrock spa_get_dspace(spa_t *spa) 100599653d4eSeschrock { 100699653d4eSeschrock if (spa->spa_deflate) 100799653d4eSeschrock return (spa->spa_root_vdev->vdev_stat.vs_dspace); 100899653d4eSeschrock else 100999653d4eSeschrock return (spa->spa_root_vdev->vdev_stat.vs_space); 101099653d4eSeschrock } 101199653d4eSeschrock 1012fa9e4066Sahrens /* ARGSUSED */ 1013fa9e4066Sahrens uint64_t 1014fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 1015fa9e4066Sahrens { 1016fa9e4066Sahrens /* 1017fa9e4066Sahrens * For now, the worst case is 512-byte RAID-Z blocks, in which 1018fa9e4066Sahrens * case the space requirement is exactly 2x; so just assume that. 101944cd46caSbillm * Add to this the fact that we can have up to 3 DVAs per bp, and 102044cd46caSbillm * we have to multiply by a total of 6x. 102144cd46caSbillm */ 102244cd46caSbillm return (lsize * 6); 102344cd46caSbillm } 102444cd46caSbillm 102544cd46caSbillm uint64_t 102644cd46caSbillm spa_version(spa_t *spa) 102744cd46caSbillm { 102844cd46caSbillm return (spa->spa_ubsync.ub_version); 102944cd46caSbillm } 103044cd46caSbillm 103144cd46caSbillm int 103244cd46caSbillm spa_max_replication(spa_t *spa) 103344cd46caSbillm { 103444cd46caSbillm /* 1035e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 103644cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 103744cd46caSbillm * replication level accordingly. 1038fa9e4066Sahrens */ 1039e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 104044cd46caSbillm return (1); 104144cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1042fa9e4066Sahrens } 1043fa9e4066Sahrens 104499653d4eSeschrock uint64_t 104599653d4eSeschrock bp_get_dasize(spa_t *spa, const blkptr_t *bp) 104699653d4eSeschrock { 104799653d4eSeschrock int sz = 0, i; 104899653d4eSeschrock 104999653d4eSeschrock if (!spa->spa_deflate) 105099653d4eSeschrock return (BP_GET_ASIZE(bp)); 105199653d4eSeschrock 105299653d4eSeschrock for (i = 0; i < SPA_DVAS_PER_BP; i++) { 105399653d4eSeschrock vdev_t *vd = 105499653d4eSeschrock vdev_lookup_top(spa, DVA_GET_VDEV(&bp->blk_dva[i])); 105599653d4eSeschrock sz += (DVA_GET_ASIZE(&bp->blk_dva[i]) >> SPA_MINBLOCKSHIFT) * 105699653d4eSeschrock vd->vdev_deflate_ratio; 105799653d4eSeschrock } 105899653d4eSeschrock return (sz); 105999653d4eSeschrock } 106099653d4eSeschrock 1061fa9e4066Sahrens /* 1062fa9e4066Sahrens * ========================================================================== 1063fa9e4066Sahrens * Initialization and Termination 1064fa9e4066Sahrens * ========================================================================== 1065fa9e4066Sahrens */ 1066fa9e4066Sahrens 1067fa9e4066Sahrens static int 1068fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1069fa9e4066Sahrens { 1070fa9e4066Sahrens const spa_t *s1 = a1; 1071fa9e4066Sahrens const spa_t *s2 = a2; 1072fa9e4066Sahrens int s; 1073fa9e4066Sahrens 1074fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1075fa9e4066Sahrens if (s > 0) 1076fa9e4066Sahrens return (1); 1077fa9e4066Sahrens if (s < 0) 1078fa9e4066Sahrens return (-1); 1079fa9e4066Sahrens return (0); 1080fa9e4066Sahrens } 1081fa9e4066Sahrens 10820373e76bSbonwick int 10830373e76bSbonwick spa_busy(void) 10840373e76bSbonwick { 10850373e76bSbonwick return (spa_active_count); 10860373e76bSbonwick } 10870373e76bSbonwick 1088fa9e4066Sahrens void 1089fa9e4066Sahrens spa_init(int mode) 1090fa9e4066Sahrens { 1091fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1092fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1093fa9e4066Sahrens 1094fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1095fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1096fa9e4066Sahrens 109799653d4eSeschrock avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_spare_t), 109899653d4eSeschrock offsetof(spa_spare_t, spare_avl)); 109999653d4eSeschrock 1100fa9e4066Sahrens spa_mode = mode; 1101fa9e4066Sahrens 1102fa9e4066Sahrens refcount_init(); 1103fa9e4066Sahrens unique_init(); 1104fa9e4066Sahrens zio_init(); 1105fa9e4066Sahrens dmu_init(); 1106fa9e4066Sahrens zil_init(); 1107fa9e4066Sahrens spa_config_load(); 1108fa9e4066Sahrens } 1109fa9e4066Sahrens 1110fa9e4066Sahrens void 1111fa9e4066Sahrens spa_fini(void) 1112fa9e4066Sahrens { 1113fa9e4066Sahrens spa_evict_all(); 1114fa9e4066Sahrens 1115fa9e4066Sahrens zil_fini(); 1116fa9e4066Sahrens dmu_fini(); 1117fa9e4066Sahrens zio_fini(); 1118fa9e4066Sahrens refcount_fini(); 1119fa9e4066Sahrens 1120fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 112199653d4eSeschrock avl_destroy(&spa_spare_avl); 1122fa9e4066Sahrens 1123fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1124fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1125fa9e4066Sahrens } 1126