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 /* 22ea8dc4b6Seschrock * Copyright 2006 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 * 55fa9e4066Sahrens * This lock must be acquired to do any of the following: 56fa9e4066Sahrens * 57fa9e4066Sahrens * - Lookup a spa_t by name 58fa9e4066Sahrens * - Add or remove a spa_t from the namespace 59fa9e4066Sahrens * - Increase spa_refcount from non-zero 60fa9e4066Sahrens * - Check if spa_refcount is zero 61fa9e4066Sahrens * - Rename a spa_t 62ea8dc4b6Seschrock * - add/remove/attach/detach devices 63fa9e4066Sahrens * - Held for the duration of create/destroy/import/export 64fa9e4066Sahrens * 65fa9e4066Sahrens * It does not need to handle recursion. A create or destroy may 66fa9e4066Sahrens * reference objects (files or zvols) in other pools, but by 67fa9e4066Sahrens * definition they must have an existing reference, and will never need 68fa9e4066Sahrens * to lookup a spa_t by name. 69fa9e4066Sahrens * 70fa9e4066Sahrens * spa_refcount (per-spa refcount_t protected by mutex) 71fa9e4066Sahrens * 72fa9e4066Sahrens * This reference count keep track of any active users of the spa_t. The 73fa9e4066Sahrens * spa_t cannot be destroyed or freed while this is non-zero. Internally, 74fa9e4066Sahrens * the refcount is never really 'zero' - opening a pool implicitly keeps 75fa9e4066Sahrens * some references in the DMU. Internally we check against SPA_MINREF, but 76fa9e4066Sahrens * present the image of a zero/non-zero value to consumers. 77fa9e4066Sahrens * 78fa9e4066Sahrens * spa_config_lock (per-spa crazy rwlock) 79fa9e4066Sahrens * 80fa9e4066Sahrens * This SPA special is a recursive rwlock, capable of being acquired from 81fa9e4066Sahrens * asynchronous threads. It has protects the spa_t from config changes, 82fa9e4066Sahrens * and must be held in the following circumstances: 83fa9e4066Sahrens * 84fa9e4066Sahrens * - RW_READER to perform I/O to the spa 85fa9e4066Sahrens * - RW_WRITER to change the vdev config 86fa9e4066Sahrens * 87fa9e4066Sahrens * spa_config_cache_lock (per-spa mutex) 88fa9e4066Sahrens * 89fa9e4066Sahrens * 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 * 96fa9e4066Sahrens * spa_namespace_lock -> spa_refcount 97fa9e4066Sahrens * 98fa9e4066Sahrens * The namespace lock must be acquired to increase the refcount from 0 99fa9e4066Sahrens * or to check if it is zero. 100fa9e4066Sahrens * 101fa9e4066Sahrens * spa_refcount -> spa_config_lock 102fa9e4066Sahrens * 103fa9e4066Sahrens * There must be at least one valid reference on the spa_t to acquire 104fa9e4066Sahrens * the config lock. 105fa9e4066Sahrens * 106ea8dc4b6Seschrock * spa_namespace_lock -> spa_config_lock 107fa9e4066Sahrens * 108ea8dc4b6Seschrock * 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 * 117fa9e4066Sahrens * spa_lookup() Lookup a spa_t by name. 118fa9e4066Sahrens * 119fa9e4066Sahrens * spa_add() Create a new spa_t in the namespace. 120fa9e4066Sahrens * 121fa9e4066Sahrens * spa_remove() Remove a spa_t from the namespace. This also 122fa9e4066Sahrens * frees up any memory associated with the spa_t. 123fa9e4066Sahrens * 124fa9e4066Sahrens * spa_next() Returns the next spa_t in the system, or the 125fa9e4066Sahrens * first if NULL is passed. 126fa9e4066Sahrens * 127fa9e4066Sahrens * spa_evict_all() Shutdown and remove all spa_t structures in 128fa9e4066Sahrens * 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 * 134fa9e4066Sahrens * spa_open_ref() Adds a reference to the given spa_t. Must be 135fa9e4066Sahrens * called with spa_namespace_lock held if the 136fa9e4066Sahrens * refcount is currently zero. 137fa9e4066Sahrens * 138fa9e4066Sahrens * spa_close() Remove a reference from the spa_t. This will 139fa9e4066Sahrens * not free the spa_t or remove it from the 140fa9e4066Sahrens * namespace. No locking is required. 141fa9e4066Sahrens * 142fa9e4066Sahrens * spa_refcount_zero() Returns true if the refcount is currently 143fa9e4066Sahrens * zero. Must be called with spa_namespace_lock 144fa9e4066Sahrens * held. 145fa9e4066Sahrens * 146fa9e4066Sahrens * The spa_config_lock is manipulated using the following functions: 147fa9e4066Sahrens * 148fa9e4066Sahrens * spa_config_enter() Acquire the config lock as RW_READER or 149fa9e4066Sahrens * RW_WRITER. At least one reference on the spa_t 150fa9e4066Sahrens * must exist. 151fa9e4066Sahrens * 152fa9e4066Sahrens * spa_config_exit() Release the config lock. 153fa9e4066Sahrens * 154fa9e4066Sahrens * spa_config_held() Returns true if the config lock is currently 155fa9e4066Sahrens * held in the given state. 156fa9e4066Sahrens * 157ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 158fa9e4066Sahrens * 159ea8dc4b6Seschrock * spa_vdev_enter() Acquire the namespace lock and the config lock 160ea8dc4b6Seschrock * for writing. 161fa9e4066Sahrens * 162fa9e4066Sahrens * spa_vdev_exit() Release the config lock, wait for all I/O 163ea8dc4b6Seschrock * 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; 175*0373e76bSbonwick static int spa_active_count; 176fa9e4066Sahrens 177fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 178fa9e4066Sahrens int spa_mode; 179fa9e4066Sahrens 180fa9e4066Sahrens #ifdef ZFS_DEBUG 181fa9e4066Sahrens int zfs_flags = ~0; 182fa9e4066Sahrens #else 183fa9e4066Sahrens int zfs_flags = 0; 184fa9e4066Sahrens #endif 185fa9e4066Sahrens 186fa9e4066Sahrens #define SPA_MINREF 5 /* spa_refcnt for an open-but-idle pool */ 187fa9e4066Sahrens 188fa9e4066Sahrens /* 189fa9e4066Sahrens * ========================================================================== 190fa9e4066Sahrens * SPA namespace functions 191fa9e4066Sahrens * ========================================================================== 192fa9e4066Sahrens */ 193fa9e4066Sahrens 194fa9e4066Sahrens /* 195fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 196fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 197fa9e4066Sahrens */ 198fa9e4066Sahrens spa_t * 199fa9e4066Sahrens spa_lookup(const char *name) 200fa9e4066Sahrens { 201fa9e4066Sahrens spa_t search, *spa; 202fa9e4066Sahrens avl_index_t where; 203fa9e4066Sahrens 204fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 205fa9e4066Sahrens 206fa9e4066Sahrens search.spa_name = (char *)name; 207fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 208fa9e4066Sahrens 209fa9e4066Sahrens return (spa); 210fa9e4066Sahrens } 211fa9e4066Sahrens 212fa9e4066Sahrens /* 213fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 214fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 215fa9e4066Sahrens * exist by calling spa_lookup() first. 216fa9e4066Sahrens */ 217fa9e4066Sahrens spa_t * 218*0373e76bSbonwick spa_add(const char *name, const char *altroot) 219fa9e4066Sahrens { 220fa9e4066Sahrens spa_t *spa; 221fa9e4066Sahrens 222fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 223fa9e4066Sahrens 224fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 225fa9e4066Sahrens 226fa9e4066Sahrens spa->spa_name = spa_strdup(name); 227fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 228fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 229*0373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 230fa9e4066Sahrens 231fa9e4066Sahrens refcount_create(&spa->spa_refcount); 232ea8dc4b6Seschrock refcount_create(&spa->spa_config_lock.scl_count); 233fa9e4066Sahrens 234fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 235fa9e4066Sahrens 236*0373e76bSbonwick /* 237*0373e76bSbonwick * Set the alternate root, if there is one. 238*0373e76bSbonwick */ 239*0373e76bSbonwick if (altroot) { 240*0373e76bSbonwick spa->spa_root = spa_strdup(altroot); 241*0373e76bSbonwick spa_active_count++; 242*0373e76bSbonwick } 243*0373e76bSbonwick 244fa9e4066Sahrens return (spa); 245fa9e4066Sahrens } 246fa9e4066Sahrens 247fa9e4066Sahrens /* 248fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 249fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 250fa9e4066Sahrens * deactivated. 251fa9e4066Sahrens */ 252fa9e4066Sahrens void 253fa9e4066Sahrens spa_remove(spa_t *spa) 254fa9e4066Sahrens { 255fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 256fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 257fa9e4066Sahrens ASSERT(spa->spa_scrub_thread == NULL); 258fa9e4066Sahrens 259fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 260fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 261fa9e4066Sahrens 262*0373e76bSbonwick if (spa->spa_root) { 263fa9e4066Sahrens spa_strfree(spa->spa_root); 264*0373e76bSbonwick spa_active_count--; 265*0373e76bSbonwick } 266fa9e4066Sahrens 267fa9e4066Sahrens if (spa->spa_name) 268fa9e4066Sahrens spa_strfree(spa->spa_name); 269fa9e4066Sahrens 270fa9e4066Sahrens spa_config_set(spa, NULL); 271fa9e4066Sahrens 272fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 273ea8dc4b6Seschrock refcount_destroy(&spa->spa_config_lock.scl_count); 274fa9e4066Sahrens 275fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 276fa9e4066Sahrens } 277fa9e4066Sahrens 278fa9e4066Sahrens /* 279fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 280fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 281fa9e4066Sahrens */ 282fa9e4066Sahrens spa_t * 283fa9e4066Sahrens spa_next(spa_t *prev) 284fa9e4066Sahrens { 285fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 286fa9e4066Sahrens 287fa9e4066Sahrens if (prev) 288fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 289fa9e4066Sahrens else 290fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 291fa9e4066Sahrens } 292fa9e4066Sahrens 293fa9e4066Sahrens /* 294fa9e4066Sahrens * ========================================================================== 295fa9e4066Sahrens * SPA refcount functions 296fa9e4066Sahrens * ========================================================================== 297fa9e4066Sahrens */ 298fa9e4066Sahrens 299fa9e4066Sahrens /* 300fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 301fa9e4066Sahrens * have the namespace lock held. 302fa9e4066Sahrens */ 303fa9e4066Sahrens void 304fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 305fa9e4066Sahrens { 306fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 307fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 308fa9e4066Sahrens 309fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 310fa9e4066Sahrens } 311fa9e4066Sahrens 312fa9e4066Sahrens /* 313fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 314fa9e4066Sahrens * have the namespace lock held. 315fa9e4066Sahrens */ 316fa9e4066Sahrens void 317fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 318fa9e4066Sahrens { 319fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 320fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 321fa9e4066Sahrens 322fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 323fa9e4066Sahrens } 324fa9e4066Sahrens 325fa9e4066Sahrens /* 326fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 327fa9e4066Sahrens * spa_namespace_lock held. We really compare against SPA_MINREF, which is the 328fa9e4066Sahrens * number of references acquired when opening a pool 329fa9e4066Sahrens */ 330fa9e4066Sahrens boolean_t 331fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 332fa9e4066Sahrens { 333fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 334fa9e4066Sahrens 335fa9e4066Sahrens return (refcount_count(&spa->spa_refcount) == SPA_MINREF); 336fa9e4066Sahrens } 337fa9e4066Sahrens 338fa9e4066Sahrens /* 339fa9e4066Sahrens * ========================================================================== 340fa9e4066Sahrens * SPA config locking 341fa9e4066Sahrens * ========================================================================== 342fa9e4066Sahrens */ 343fa9e4066Sahrens 344fa9e4066Sahrens /* 345fa9e4066Sahrens * Acquire the config lock. The config lock is a special rwlock that allows for 346fa9e4066Sahrens * recursive enters. Because these enters come from the same thread as well as 347fa9e4066Sahrens * asynchronous threads working on behalf of the owner, we must unilaterally 348fa9e4066Sahrens * allow all reads access as long at least one reader is held (even if a write 349fa9e4066Sahrens * is requested). This has the side effect of write starvation, but write locks 350fa9e4066Sahrens * are extremely rare, and a solution to this problem would be significantly 351fa9e4066Sahrens * more complex (if even possible). 352fa9e4066Sahrens * 353fa9e4066Sahrens * We would like to assert that the namespace lock isn't held, but this is a 354fa9e4066Sahrens * valid use during create. 355fa9e4066Sahrens */ 356fa9e4066Sahrens void 357ea8dc4b6Seschrock spa_config_enter(spa_t *spa, krw_t rw, void *tag) 358fa9e4066Sahrens { 359fa9e4066Sahrens spa_config_lock_t *scl = &spa->spa_config_lock; 360fa9e4066Sahrens 361fa9e4066Sahrens mutex_enter(&scl->scl_lock); 362fa9e4066Sahrens 363fa9e4066Sahrens if (scl->scl_writer != curthread) { 364fa9e4066Sahrens if (rw == RW_READER) { 365fa9e4066Sahrens while (scl->scl_writer != NULL) 366fa9e4066Sahrens cv_wait(&scl->scl_cv, &scl->scl_lock); 367fa9e4066Sahrens } else { 368ea8dc4b6Seschrock while (scl->scl_writer != NULL || 369ea8dc4b6Seschrock !refcount_is_zero(&scl->scl_count)) 370fa9e4066Sahrens cv_wait(&scl->scl_cv, &scl->scl_lock); 371fa9e4066Sahrens scl->scl_writer = curthread; 372fa9e4066Sahrens } 373fa9e4066Sahrens } 374fa9e4066Sahrens 375ea8dc4b6Seschrock (void) refcount_add(&scl->scl_count, tag); 376fa9e4066Sahrens 377fa9e4066Sahrens mutex_exit(&scl->scl_lock); 378fa9e4066Sahrens } 379fa9e4066Sahrens 380fa9e4066Sahrens /* 381fa9e4066Sahrens * Release the spa config lock, notifying any waiters in the process. 382fa9e4066Sahrens */ 383fa9e4066Sahrens void 384ea8dc4b6Seschrock spa_config_exit(spa_t *spa, void *tag) 385fa9e4066Sahrens { 386fa9e4066Sahrens spa_config_lock_t *scl = &spa->spa_config_lock; 387fa9e4066Sahrens 388fa9e4066Sahrens mutex_enter(&scl->scl_lock); 389fa9e4066Sahrens 390ea8dc4b6Seschrock ASSERT(!refcount_is_zero(&scl->scl_count)); 391ea8dc4b6Seschrock if (refcount_remove(&scl->scl_count, tag) == 0) { 392fa9e4066Sahrens cv_broadcast(&scl->scl_cv); 393fa9e4066Sahrens scl->scl_writer = NULL; /* OK in either case */ 394fa9e4066Sahrens } 395fa9e4066Sahrens 396fa9e4066Sahrens mutex_exit(&scl->scl_lock); 397fa9e4066Sahrens } 398fa9e4066Sahrens 399fa9e4066Sahrens /* 400fa9e4066Sahrens * Returns true if the config lock is held in the given manner. 401fa9e4066Sahrens */ 402fa9e4066Sahrens boolean_t 403fa9e4066Sahrens spa_config_held(spa_t *spa, krw_t rw) 404fa9e4066Sahrens { 405fa9e4066Sahrens spa_config_lock_t *scl = &spa->spa_config_lock; 406fa9e4066Sahrens boolean_t held; 407fa9e4066Sahrens 408fa9e4066Sahrens mutex_enter(&scl->scl_lock); 409fa9e4066Sahrens if (rw == RW_WRITER) 410fa9e4066Sahrens held = (scl->scl_writer == curthread); 411fa9e4066Sahrens else 412ea8dc4b6Seschrock held = !refcount_is_zero(&scl->scl_count); 413fa9e4066Sahrens mutex_exit(&scl->scl_lock); 414fa9e4066Sahrens 415fa9e4066Sahrens return (held); 416fa9e4066Sahrens } 417fa9e4066Sahrens 418fa9e4066Sahrens /* 419fa9e4066Sahrens * ========================================================================== 420fa9e4066Sahrens * SPA vdev locking 421fa9e4066Sahrens * ========================================================================== 422fa9e4066Sahrens */ 423fa9e4066Sahrens 424fa9e4066Sahrens /* 425ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 426ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 427fa9e4066Sahrens * It returns the next transaction group for the spa_t. 428fa9e4066Sahrens */ 429fa9e4066Sahrens uint64_t 430fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 431fa9e4066Sahrens { 432ea8dc4b6Seschrock /* 433ea8dc4b6Seschrock * Suspend scrub activity while we mess with the config. 434ea8dc4b6Seschrock */ 435ea8dc4b6Seschrock spa_scrub_suspend(spa); 436fa9e4066Sahrens 437*0373e76bSbonwick mutex_enter(&spa_namespace_lock); 438ea8dc4b6Seschrock 439ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, spa); 440fa9e4066Sahrens 441fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 442fa9e4066Sahrens } 443fa9e4066Sahrens 444fa9e4066Sahrens /* 445fa9e4066Sahrens * Unlock the spa_t after adding or removing a vdev. Besides undoing the 446fa9e4066Sahrens * locking of spa_vdev_enter(), we also want make sure the transactions have 447fa9e4066Sahrens * synced to disk, and then update the global configuration cache with the new 448fa9e4066Sahrens * information. 449fa9e4066Sahrens */ 450fa9e4066Sahrens int 451fa9e4066Sahrens spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 452fa9e4066Sahrens { 4530e34b6a7Sbonwick int config_changed = B_FALSE; 454ea8dc4b6Seschrock 455*0373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 4560e34b6a7Sbonwick 4570e34b6a7Sbonwick /* 4580e34b6a7Sbonwick * Reassess the DTLs. 4590e34b6a7Sbonwick */ 460*0373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 4610e34b6a7Sbonwick 4620e34b6a7Sbonwick /* 463*0373e76bSbonwick * If the config changed, notify the scrub thread that it must restart. 4640e34b6a7Sbonwick */ 4650e34b6a7Sbonwick if (error == 0 && !list_is_empty(&spa->spa_dirty_list)) { 4660e34b6a7Sbonwick config_changed = B_TRUE; 467*0373e76bSbonwick spa_scrub_restart(spa, txg); 4680e34b6a7Sbonwick } 469ea8dc4b6Seschrock 470ea8dc4b6Seschrock spa_config_exit(spa, spa); 471fa9e4066Sahrens 472ea8dc4b6Seschrock /* 4735dabedeeSbonwick * Allow scrubbing to resume. 474ea8dc4b6Seschrock */ 475ea8dc4b6Seschrock spa_scrub_resume(spa); 476fa9e4066Sahrens 477fa9e4066Sahrens /* 478fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 479fa9e4066Sahrens * that there won't be more than one config change per txg. 480fa9e4066Sahrens * This allows us to use the txg as the generation number. 481fa9e4066Sahrens */ 482fa9e4066Sahrens if (error == 0) 483fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 484fa9e4066Sahrens 485fa9e4066Sahrens if (vd != NULL) { 486fa9e4066Sahrens ASSERT(!vd->vdev_detached || vd->vdev_dtl.smo_object == 0); 487fa9e4066Sahrens vdev_free(vd); 488fa9e4066Sahrens } 489fa9e4066Sahrens 490fa9e4066Sahrens /* 4910e34b6a7Sbonwick * If the config changed, update the config cache. 492fa9e4066Sahrens */ 4930e34b6a7Sbonwick if (config_changed) 494fa9e4066Sahrens spa_config_sync(); 495ea8dc4b6Seschrock 496ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 497fa9e4066Sahrens 498fa9e4066Sahrens return (error); 499fa9e4066Sahrens } 500fa9e4066Sahrens 501fa9e4066Sahrens /* 502fa9e4066Sahrens * ========================================================================== 503fa9e4066Sahrens * Miscellaneous functions 504fa9e4066Sahrens * ========================================================================== 505fa9e4066Sahrens */ 506fa9e4066Sahrens 507fa9e4066Sahrens /* 508fa9e4066Sahrens * Rename a spa_t. 509fa9e4066Sahrens */ 510fa9e4066Sahrens int 511fa9e4066Sahrens spa_rename(const char *name, const char *newname) 512fa9e4066Sahrens { 513fa9e4066Sahrens spa_t *spa; 514fa9e4066Sahrens int err; 515fa9e4066Sahrens 516fa9e4066Sahrens /* 517fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 518fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 519fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 520ea8dc4b6Seschrock * allow recursive calls for other reasons. 521fa9e4066Sahrens */ 522fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 523fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 524fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 525fa9e4066Sahrens return (err); 526fa9e4066Sahrens } 527fa9e4066Sahrens 528ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 529fa9e4066Sahrens 530fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 531fa9e4066Sahrens spa_strfree(spa->spa_name); 532fa9e4066Sahrens spa->spa_name = spa_strdup(newname); 533fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 534fa9e4066Sahrens 535fa9e4066Sahrens /* 536fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 537fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 538fa9e4066Sahrens * during the sync. 539fa9e4066Sahrens */ 540fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 541fa9e4066Sahrens 542ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 543fa9e4066Sahrens 544*0373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 545fa9e4066Sahrens 546fa9e4066Sahrens /* 547fa9e4066Sahrens * Sync the updated config cache. 548fa9e4066Sahrens */ 549fa9e4066Sahrens spa_config_sync(); 550fa9e4066Sahrens 551fa9e4066Sahrens spa_close(spa, FTAG); 552fa9e4066Sahrens 553fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 554fa9e4066Sahrens 555fa9e4066Sahrens return (0); 556fa9e4066Sahrens } 557fa9e4066Sahrens 558fa9e4066Sahrens 559fa9e4066Sahrens /* 560fa9e4066Sahrens * Determine whether a pool with given pool_guid exists. If device_guid is 561fa9e4066Sahrens * non-zero, determine whether the pool exists *and* contains a device with the 562fa9e4066Sahrens * specified device_guid. 563fa9e4066Sahrens */ 564fa9e4066Sahrens boolean_t 565fa9e4066Sahrens spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 566fa9e4066Sahrens { 567fa9e4066Sahrens spa_t *spa; 568fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 569fa9e4066Sahrens 570ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 571fa9e4066Sahrens 572fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 573fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 574fa9e4066Sahrens continue; 575fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 576fa9e4066Sahrens continue; 577fa9e4066Sahrens if (spa_guid(spa) == pool_guid && (device_guid == 0 || 578fa9e4066Sahrens vdev_lookup_by_guid(spa->spa_root_vdev, device_guid))) 579fa9e4066Sahrens break; 580fa9e4066Sahrens } 581fa9e4066Sahrens 582fa9e4066Sahrens return (spa != NULL); 583fa9e4066Sahrens } 584fa9e4066Sahrens 585fa9e4066Sahrens char * 586fa9e4066Sahrens spa_strdup(const char *s) 587fa9e4066Sahrens { 588fa9e4066Sahrens size_t len; 589fa9e4066Sahrens char *new; 590fa9e4066Sahrens 591fa9e4066Sahrens len = strlen(s); 592fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 593fa9e4066Sahrens bcopy(s, new, len); 594fa9e4066Sahrens new[len] = '\0'; 595fa9e4066Sahrens 596fa9e4066Sahrens return (new); 597fa9e4066Sahrens } 598fa9e4066Sahrens 599fa9e4066Sahrens void 600fa9e4066Sahrens spa_strfree(char *s) 601fa9e4066Sahrens { 602fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 603fa9e4066Sahrens } 604fa9e4066Sahrens 605fa9e4066Sahrens uint64_t 606fa9e4066Sahrens spa_get_random(uint64_t range) 607fa9e4066Sahrens { 608fa9e4066Sahrens uint64_t r; 609fa9e4066Sahrens 610fa9e4066Sahrens ASSERT(range != 0); 611fa9e4066Sahrens 612fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 613fa9e4066Sahrens 614fa9e4066Sahrens return (r % range); 615fa9e4066Sahrens } 616fa9e4066Sahrens 617fa9e4066Sahrens void 618fbabab8fSmaybee sprintf_blkptr(char *buf, int len, blkptr_t *bp) 619fa9e4066Sahrens { 620fa9e4066Sahrens /* XXBP - Need to see if we want all DVAs or not */ 621fa9e4066Sahrens dva_t *dva = BP_IDENTITY(bp); 622fa9e4066Sahrens 623fa9e4066Sahrens if (bp == NULL) { 624fbabab8fSmaybee (void) snprintf(buf, len, "<NULL>"); 625fa9e4066Sahrens return; 626fa9e4066Sahrens } 627fa9e4066Sahrens 628fa9e4066Sahrens if (BP_IS_HOLE(bp)) { 629fbabab8fSmaybee (void) snprintf(buf, len, "<hole>"); 630fa9e4066Sahrens return; 631fa9e4066Sahrens } 632fa9e4066Sahrens 633fbabab8fSmaybee (void) snprintf(buf, len, "[L%llu %s] vdev=%llu offset=%llx " 634fbabab8fSmaybee "size=%llxL/%llxP/%llxA %s %s %s %s " 635fbabab8fSmaybee "birth=%llu fill=%llu cksum=%llx:%llx:%llx:%llx", 636fa9e4066Sahrens (u_longlong_t)BP_GET_LEVEL(bp), 637fa9e4066Sahrens dmu_ot[BP_GET_TYPE(bp)].ot_name, 638fa9e4066Sahrens (u_longlong_t)DVA_GET_VDEV(dva), 639fa9e4066Sahrens (u_longlong_t)DVA_GET_OFFSET(dva), 640fa9e4066Sahrens (u_longlong_t)BP_GET_LSIZE(bp), 641fa9e4066Sahrens (u_longlong_t)BP_GET_PSIZE(bp), 642fa9e4066Sahrens (u_longlong_t)DVA_GET_ASIZE(dva), 643fa9e4066Sahrens zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 644fa9e4066Sahrens zio_compress_table[BP_GET_COMPRESS(bp)].ci_name, 645fa9e4066Sahrens BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", 646fbabab8fSmaybee DVA_GET_GANG(dva) == 0 ? "contiguous" : "gang", 647fa9e4066Sahrens (u_longlong_t)bp->blk_birth, 648fa9e4066Sahrens (u_longlong_t)bp->blk_fill, 649fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[0], 650fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[1], 651fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[2], 652fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[3]); 653fa9e4066Sahrens } 654fa9e4066Sahrens 655fa9e4066Sahrens void 656fa9e4066Sahrens spa_freeze(spa_t *spa) 657fa9e4066Sahrens { 658fa9e4066Sahrens uint64_t freeze_txg = 0; 659fa9e4066Sahrens 660ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 661fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 662fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 663fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 664fa9e4066Sahrens } 665ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 666fa9e4066Sahrens if (freeze_txg != 0) 667fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 668fa9e4066Sahrens } 669fa9e4066Sahrens 670fa9e4066Sahrens /* 671fa9e4066Sahrens * ========================================================================== 672fa9e4066Sahrens * Accessor functions 673fa9e4066Sahrens * ========================================================================== 674fa9e4066Sahrens */ 675fa9e4066Sahrens 676fa9e4066Sahrens krwlock_t * 677fa9e4066Sahrens spa_traverse_rwlock(spa_t *spa) 678fa9e4066Sahrens { 679fa9e4066Sahrens return (&spa->spa_traverse_lock); 680fa9e4066Sahrens } 681fa9e4066Sahrens 682fa9e4066Sahrens int 683fa9e4066Sahrens spa_traverse_wanted(spa_t *spa) 684fa9e4066Sahrens { 685fa9e4066Sahrens return (spa->spa_traverse_wanted); 686fa9e4066Sahrens } 687fa9e4066Sahrens 688fa9e4066Sahrens dsl_pool_t * 689fa9e4066Sahrens spa_get_dsl(spa_t *spa) 690fa9e4066Sahrens { 691fa9e4066Sahrens return (spa->spa_dsl_pool); 692fa9e4066Sahrens } 693fa9e4066Sahrens 694fa9e4066Sahrens blkptr_t * 695fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 696fa9e4066Sahrens { 697fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 698fa9e4066Sahrens } 699fa9e4066Sahrens 700fa9e4066Sahrens void 701fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 702fa9e4066Sahrens { 703fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 704fa9e4066Sahrens } 705fa9e4066Sahrens 706fa9e4066Sahrens void 707fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 708fa9e4066Sahrens { 709fa9e4066Sahrens if (spa->spa_root == NULL) 710fa9e4066Sahrens buf[0] = '\0'; 711fa9e4066Sahrens else 712fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 713fa9e4066Sahrens } 714fa9e4066Sahrens 715fa9e4066Sahrens int 716fa9e4066Sahrens spa_sync_pass(spa_t *spa) 717fa9e4066Sahrens { 718fa9e4066Sahrens return (spa->spa_sync_pass); 719fa9e4066Sahrens } 720fa9e4066Sahrens 721fa9e4066Sahrens char * 722fa9e4066Sahrens spa_name(spa_t *spa) 723fa9e4066Sahrens { 724fa9e4066Sahrens /* 725fa9e4066Sahrens * Accessing the name requires holding either the namespace lock or the 726fa9e4066Sahrens * config lock, both of which are required to do a rename. 727fa9e4066Sahrens */ 728fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock) || 729fa9e4066Sahrens spa_config_held(spa, RW_READER) || spa_config_held(spa, RW_WRITER)); 730fa9e4066Sahrens 731fa9e4066Sahrens return (spa->spa_name); 732fa9e4066Sahrens } 733fa9e4066Sahrens 734fa9e4066Sahrens uint64_t 735fa9e4066Sahrens spa_guid(spa_t *spa) 736fa9e4066Sahrens { 737fa9e4066Sahrens return (spa->spa_root_vdev->vdev_guid); 738fa9e4066Sahrens } 739fa9e4066Sahrens 740fa9e4066Sahrens uint64_t 741fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 742fa9e4066Sahrens { 743fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 744fa9e4066Sahrens } 745fa9e4066Sahrens 746fa9e4066Sahrens uint64_t 747fa9e4066Sahrens spa_first_txg(spa_t *spa) 748fa9e4066Sahrens { 749fa9e4066Sahrens return (spa->spa_first_txg); 750fa9e4066Sahrens } 751fa9e4066Sahrens 752fa9e4066Sahrens int 753fa9e4066Sahrens spa_state(spa_t *spa) 754fa9e4066Sahrens { 755fa9e4066Sahrens return (spa->spa_state); 756fa9e4066Sahrens } 757fa9e4066Sahrens 758fa9e4066Sahrens uint64_t 759fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 760fa9e4066Sahrens { 761fa9e4066Sahrens return (spa->spa_freeze_txg); 762fa9e4066Sahrens } 763fa9e4066Sahrens 764fa9e4066Sahrens /* 765fa9e4066Sahrens * In the future, this may select among different metaslab classes 766fa9e4066Sahrens * depending on the zdp. For now, there's no such distinction. 767fa9e4066Sahrens */ 768fa9e4066Sahrens metaslab_class_t * 769fa9e4066Sahrens spa_metaslab_class_select(spa_t *spa) 770fa9e4066Sahrens { 771fa9e4066Sahrens return (spa->spa_normal_class); 772fa9e4066Sahrens } 773fa9e4066Sahrens 774fa9e4066Sahrens /* 775fa9e4066Sahrens * Return pool-wide allocated space. 776fa9e4066Sahrens */ 777fa9e4066Sahrens uint64_t 778fa9e4066Sahrens spa_get_alloc(spa_t *spa) 779fa9e4066Sahrens { 780fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_alloc); 781fa9e4066Sahrens } 782fa9e4066Sahrens 783fa9e4066Sahrens /* 784fa9e4066Sahrens * Return pool-wide allocated space. 785fa9e4066Sahrens */ 786fa9e4066Sahrens uint64_t 787fa9e4066Sahrens spa_get_space(spa_t *spa) 788fa9e4066Sahrens { 789fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_space); 790fa9e4066Sahrens } 791fa9e4066Sahrens 792fa9e4066Sahrens /* ARGSUSED */ 793fa9e4066Sahrens uint64_t 794fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 795fa9e4066Sahrens { 796fa9e4066Sahrens /* 797fa9e4066Sahrens * For now, the worst case is 512-byte RAID-Z blocks, in which 798fa9e4066Sahrens * case the space requirement is exactly 2x; so just assume that. 799fa9e4066Sahrens */ 800fa9e4066Sahrens return (lsize << 1); 801fa9e4066Sahrens } 802fa9e4066Sahrens 803fa9e4066Sahrens /* 804fa9e4066Sahrens * ========================================================================== 805fa9e4066Sahrens * Initialization and Termination 806fa9e4066Sahrens * ========================================================================== 807fa9e4066Sahrens */ 808fa9e4066Sahrens 809fa9e4066Sahrens static int 810fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 811fa9e4066Sahrens { 812fa9e4066Sahrens const spa_t *s1 = a1; 813fa9e4066Sahrens const spa_t *s2 = a2; 814fa9e4066Sahrens int s; 815fa9e4066Sahrens 816fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 817fa9e4066Sahrens if (s > 0) 818fa9e4066Sahrens return (1); 819fa9e4066Sahrens if (s < 0) 820fa9e4066Sahrens return (-1); 821fa9e4066Sahrens return (0); 822fa9e4066Sahrens } 823fa9e4066Sahrens 824*0373e76bSbonwick int 825*0373e76bSbonwick spa_busy(void) 826*0373e76bSbonwick { 827*0373e76bSbonwick return (spa_active_count); 828*0373e76bSbonwick } 829*0373e76bSbonwick 830fa9e4066Sahrens void 831fa9e4066Sahrens spa_init(int mode) 832fa9e4066Sahrens { 833fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 834fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 835fa9e4066Sahrens 836fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 837fa9e4066Sahrens offsetof(spa_t, spa_avl)); 838fa9e4066Sahrens 839fa9e4066Sahrens spa_mode = mode; 840fa9e4066Sahrens 841fa9e4066Sahrens refcount_init(); 842fa9e4066Sahrens unique_init(); 843fa9e4066Sahrens zio_init(); 844fa9e4066Sahrens dmu_init(); 845fa9e4066Sahrens zil_init(); 846fa9e4066Sahrens spa_config_load(); 847fa9e4066Sahrens } 848fa9e4066Sahrens 849fa9e4066Sahrens void 850fa9e4066Sahrens spa_fini(void) 851fa9e4066Sahrens { 852fa9e4066Sahrens spa_evict_all(); 853fa9e4066Sahrens 854fa9e4066Sahrens zil_fini(); 855fa9e4066Sahrens dmu_fini(); 856fa9e4066Sahrens zio_fini(); 857fa9e4066Sahrens refcount_fini(); 858fa9e4066Sahrens 859fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 860fa9e4066Sahrens 861fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 862fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 863fa9e4066Sahrens } 864