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 * 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; 17644cd46caSbillm static int spa_max_replication_override = SPA_DVAS_PER_BP; 177fa9e4066Sahrens 178*99653d4eSeschrock static avl_tree_t spa_spare_avl; 179*99653d4eSeschrock static kmutex_t spa_spare_lock; 180*99653d4eSeschrock 181fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 182fa9e4066Sahrens int spa_mode; 183fa9e4066Sahrens 184fa9e4066Sahrens #ifdef ZFS_DEBUG 185fa9e4066Sahrens int zfs_flags = ~0; 186fa9e4066Sahrens #else 187fa9e4066Sahrens int zfs_flags = 0; 188fa9e4066Sahrens #endif 189fa9e4066Sahrens 190fa9e4066Sahrens #define SPA_MINREF 5 /* spa_refcnt for an open-but-idle pool */ 191fa9e4066Sahrens 192fa9e4066Sahrens /* 193fa9e4066Sahrens * ========================================================================== 194fa9e4066Sahrens * SPA namespace functions 195fa9e4066Sahrens * ========================================================================== 196fa9e4066Sahrens */ 197fa9e4066Sahrens 198fa9e4066Sahrens /* 199fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 200fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 201fa9e4066Sahrens */ 202fa9e4066Sahrens spa_t * 203fa9e4066Sahrens spa_lookup(const char *name) 204fa9e4066Sahrens { 205fa9e4066Sahrens spa_t search, *spa; 206fa9e4066Sahrens avl_index_t where; 207fa9e4066Sahrens 208fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 209fa9e4066Sahrens 210fa9e4066Sahrens search.spa_name = (char *)name; 211fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 212fa9e4066Sahrens 213fa9e4066Sahrens return (spa); 214fa9e4066Sahrens } 215fa9e4066Sahrens 216fa9e4066Sahrens /* 217fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 218fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 219fa9e4066Sahrens * exist by calling spa_lookup() first. 220fa9e4066Sahrens */ 221fa9e4066Sahrens spa_t * 2220373e76bSbonwick spa_add(const char *name, const char *altroot) 223fa9e4066Sahrens { 224fa9e4066Sahrens spa_t *spa; 225fa9e4066Sahrens 226fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 227fa9e4066Sahrens 228fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 229fa9e4066Sahrens 230fa9e4066Sahrens spa->spa_name = spa_strdup(name); 231fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 232fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 2330373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 234fa9e4066Sahrens 235fa9e4066Sahrens refcount_create(&spa->spa_refcount); 236ea8dc4b6Seschrock refcount_create(&spa->spa_config_lock.scl_count); 237fa9e4066Sahrens 238fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 239fa9e4066Sahrens 2400373e76bSbonwick /* 2410373e76bSbonwick * Set the alternate root, if there is one. 2420373e76bSbonwick */ 2430373e76bSbonwick if (altroot) { 2440373e76bSbonwick spa->spa_root = spa_strdup(altroot); 2450373e76bSbonwick spa_active_count++; 2460373e76bSbonwick } 2470373e76bSbonwick 248fa9e4066Sahrens return (spa); 249fa9e4066Sahrens } 250fa9e4066Sahrens 251fa9e4066Sahrens /* 252fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 253fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 254fa9e4066Sahrens * deactivated. 255fa9e4066Sahrens */ 256fa9e4066Sahrens void 257fa9e4066Sahrens spa_remove(spa_t *spa) 258fa9e4066Sahrens { 259fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 260fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 261fa9e4066Sahrens ASSERT(spa->spa_scrub_thread == NULL); 262fa9e4066Sahrens 263fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 264fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 265fa9e4066Sahrens 2660373e76bSbonwick if (spa->spa_root) { 267fa9e4066Sahrens spa_strfree(spa->spa_root); 2680373e76bSbonwick spa_active_count--; 2690373e76bSbonwick } 270fa9e4066Sahrens 271fa9e4066Sahrens if (spa->spa_name) 272fa9e4066Sahrens spa_strfree(spa->spa_name); 273fa9e4066Sahrens 274fa9e4066Sahrens spa_config_set(spa, NULL); 275fa9e4066Sahrens 276fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 277ea8dc4b6Seschrock refcount_destroy(&spa->spa_config_lock.scl_count); 278fa9e4066Sahrens 279fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 280fa9e4066Sahrens } 281fa9e4066Sahrens 282fa9e4066Sahrens /* 283fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 284fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 285fa9e4066Sahrens */ 286fa9e4066Sahrens spa_t * 287fa9e4066Sahrens spa_next(spa_t *prev) 288fa9e4066Sahrens { 289fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 290fa9e4066Sahrens 291fa9e4066Sahrens if (prev) 292fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 293fa9e4066Sahrens else 294fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 295fa9e4066Sahrens } 296fa9e4066Sahrens 297fa9e4066Sahrens /* 298fa9e4066Sahrens * ========================================================================== 299fa9e4066Sahrens * SPA refcount functions 300fa9e4066Sahrens * ========================================================================== 301fa9e4066Sahrens */ 302fa9e4066Sahrens 303fa9e4066Sahrens /* 304fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 305fa9e4066Sahrens * have the namespace lock held. 306fa9e4066Sahrens */ 307fa9e4066Sahrens void 308fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 309fa9e4066Sahrens { 310fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 311fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 312fa9e4066Sahrens 313fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 314fa9e4066Sahrens } 315fa9e4066Sahrens 316fa9e4066Sahrens /* 317fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 318fa9e4066Sahrens * have the namespace lock held. 319fa9e4066Sahrens */ 320fa9e4066Sahrens void 321fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 322fa9e4066Sahrens { 323fa9e4066Sahrens ASSERT(refcount_count(&spa->spa_refcount) > SPA_MINREF || 324fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 325fa9e4066Sahrens 326fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 327fa9e4066Sahrens } 328fa9e4066Sahrens 329fa9e4066Sahrens /* 330fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 331fa9e4066Sahrens * spa_namespace_lock held. We really compare against SPA_MINREF, which is the 332fa9e4066Sahrens * number of references acquired when opening a pool 333fa9e4066Sahrens */ 334fa9e4066Sahrens boolean_t 335fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 336fa9e4066Sahrens { 337fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 338fa9e4066Sahrens 339fa9e4066Sahrens return (refcount_count(&spa->spa_refcount) == SPA_MINREF); 340fa9e4066Sahrens } 341fa9e4066Sahrens 342*99653d4eSeschrock /* 343*99653d4eSeschrock * ========================================================================== 344*99653d4eSeschrock * SPA spare tracking 345*99653d4eSeschrock * ========================================================================== 346*99653d4eSeschrock */ 347*99653d4eSeschrock 348*99653d4eSeschrock /* 349*99653d4eSeschrock * We track spare information on a global basis. This allows us to do two 350*99653d4eSeschrock * things: determine when a spare is no longer referenced by any active pool, 351*99653d4eSeschrock * and (quickly) determine if a spare is currently in use in another pool on the 352*99653d4eSeschrock * system. 353*99653d4eSeschrock */ 354*99653d4eSeschrock typedef struct spa_spare { 355*99653d4eSeschrock uint64_t spare_guid; 356*99653d4eSeschrock avl_node_t spare_avl; 357*99653d4eSeschrock int spare_count; 358*99653d4eSeschrock } spa_spare_t; 359*99653d4eSeschrock 360*99653d4eSeschrock static int 361*99653d4eSeschrock spa_spare_compare(const void *a, const void *b) 362*99653d4eSeschrock { 363*99653d4eSeschrock const spa_spare_t *sa = a; 364*99653d4eSeschrock const spa_spare_t *sb = b; 365*99653d4eSeschrock 366*99653d4eSeschrock if (sa->spare_guid < sb->spare_guid) 367*99653d4eSeschrock return (-1); 368*99653d4eSeschrock else if (sa->spare_guid > sb->spare_guid) 369*99653d4eSeschrock return (1); 370*99653d4eSeschrock else 371*99653d4eSeschrock return (0); 372*99653d4eSeschrock } 373*99653d4eSeschrock 374*99653d4eSeschrock void 375*99653d4eSeschrock spa_spare_add(uint64_t guid) 376*99653d4eSeschrock { 377*99653d4eSeschrock avl_index_t where; 378*99653d4eSeschrock spa_spare_t search; 379*99653d4eSeschrock spa_spare_t *spare; 380*99653d4eSeschrock 381*99653d4eSeschrock mutex_enter(&spa_spare_lock); 382*99653d4eSeschrock 383*99653d4eSeschrock search.spare_guid = guid; 384*99653d4eSeschrock if ((spare = avl_find(&spa_spare_avl, &search, &where)) != NULL) { 385*99653d4eSeschrock spare->spare_count++; 386*99653d4eSeschrock } else { 387*99653d4eSeschrock spare = kmem_alloc(sizeof (spa_spare_t), KM_SLEEP); 388*99653d4eSeschrock spare->spare_guid = guid; 389*99653d4eSeschrock spare->spare_count = 1; 390*99653d4eSeschrock avl_insert(&spa_spare_avl, spare, where); 391*99653d4eSeschrock } 392*99653d4eSeschrock 393*99653d4eSeschrock mutex_exit(&spa_spare_lock); 394*99653d4eSeschrock } 395*99653d4eSeschrock 396*99653d4eSeschrock void 397*99653d4eSeschrock spa_spare_remove(uint64_t guid) 398*99653d4eSeschrock { 399*99653d4eSeschrock spa_spare_t search; 400*99653d4eSeschrock spa_spare_t *spare; 401*99653d4eSeschrock avl_index_t where; 402*99653d4eSeschrock 403*99653d4eSeschrock mutex_enter(&spa_spare_lock); 404*99653d4eSeschrock 405*99653d4eSeschrock search.spare_guid = guid; 406*99653d4eSeschrock spare = avl_find(&spa_spare_avl, &search, &where); 407*99653d4eSeschrock 408*99653d4eSeschrock ASSERT(spare != NULL); 409*99653d4eSeschrock 410*99653d4eSeschrock if (--spare->spare_count == 0) { 411*99653d4eSeschrock avl_remove(&spa_spare_avl, spare); 412*99653d4eSeschrock kmem_free(spare, sizeof (spa_spare_t)); 413*99653d4eSeschrock } 414*99653d4eSeschrock 415*99653d4eSeschrock mutex_exit(&spa_spare_lock); 416*99653d4eSeschrock } 417*99653d4eSeschrock 418*99653d4eSeschrock boolean_t 419*99653d4eSeschrock spa_spare_inuse(uint64_t guid) 420*99653d4eSeschrock { 421*99653d4eSeschrock spa_spare_t search; 422*99653d4eSeschrock avl_index_t where; 423*99653d4eSeschrock boolean_t ret; 424*99653d4eSeschrock 425*99653d4eSeschrock mutex_enter(&spa_spare_lock); 426*99653d4eSeschrock 427*99653d4eSeschrock search.spare_guid = guid; 428*99653d4eSeschrock ret = (avl_find(&spa_spare_avl, &search, &where) != NULL); 429*99653d4eSeschrock 430*99653d4eSeschrock mutex_exit(&spa_spare_lock); 431*99653d4eSeschrock 432*99653d4eSeschrock return (ret); 433*99653d4eSeschrock } 434*99653d4eSeschrock 435fa9e4066Sahrens /* 436fa9e4066Sahrens * ========================================================================== 437fa9e4066Sahrens * SPA config locking 438fa9e4066Sahrens * ========================================================================== 439fa9e4066Sahrens */ 440fa9e4066Sahrens 441fa9e4066Sahrens /* 442fa9e4066Sahrens * Acquire the config lock. The config lock is a special rwlock that allows for 443fa9e4066Sahrens * recursive enters. Because these enters come from the same thread as well as 444fa9e4066Sahrens * asynchronous threads working on behalf of the owner, we must unilaterally 445fa9e4066Sahrens * allow all reads access as long at least one reader is held (even if a write 446fa9e4066Sahrens * is requested). This has the side effect of write starvation, but write locks 447fa9e4066Sahrens * are extremely rare, and a solution to this problem would be significantly 448fa9e4066Sahrens * more complex (if even possible). 449fa9e4066Sahrens * 450fa9e4066Sahrens * We would like to assert that the namespace lock isn't held, but this is a 451fa9e4066Sahrens * valid use during create. 452fa9e4066Sahrens */ 453fa9e4066Sahrens void 454ea8dc4b6Seschrock spa_config_enter(spa_t *spa, krw_t rw, void *tag) 455fa9e4066Sahrens { 456fa9e4066Sahrens spa_config_lock_t *scl = &spa->spa_config_lock; 457fa9e4066Sahrens 458fa9e4066Sahrens mutex_enter(&scl->scl_lock); 459fa9e4066Sahrens 460fa9e4066Sahrens if (scl->scl_writer != curthread) { 461fa9e4066Sahrens if (rw == RW_READER) { 462fa9e4066Sahrens while (scl->scl_writer != NULL) 463fa9e4066Sahrens cv_wait(&scl->scl_cv, &scl->scl_lock); 464fa9e4066Sahrens } else { 465ea8dc4b6Seschrock while (scl->scl_writer != NULL || 466ea8dc4b6Seschrock !refcount_is_zero(&scl->scl_count)) 467fa9e4066Sahrens cv_wait(&scl->scl_cv, &scl->scl_lock); 468fa9e4066Sahrens scl->scl_writer = curthread; 469fa9e4066Sahrens } 470fa9e4066Sahrens } 471fa9e4066Sahrens 472ea8dc4b6Seschrock (void) refcount_add(&scl->scl_count, tag); 473fa9e4066Sahrens 474fa9e4066Sahrens mutex_exit(&scl->scl_lock); 475fa9e4066Sahrens } 476fa9e4066Sahrens 477fa9e4066Sahrens /* 478fa9e4066Sahrens * Release the spa config lock, notifying any waiters in the process. 479fa9e4066Sahrens */ 480fa9e4066Sahrens void 481ea8dc4b6Seschrock spa_config_exit(spa_t *spa, void *tag) 482fa9e4066Sahrens { 483fa9e4066Sahrens spa_config_lock_t *scl = &spa->spa_config_lock; 484fa9e4066Sahrens 485fa9e4066Sahrens mutex_enter(&scl->scl_lock); 486fa9e4066Sahrens 487ea8dc4b6Seschrock ASSERT(!refcount_is_zero(&scl->scl_count)); 488ea8dc4b6Seschrock if (refcount_remove(&scl->scl_count, tag) == 0) { 489fa9e4066Sahrens cv_broadcast(&scl->scl_cv); 490fa9e4066Sahrens scl->scl_writer = NULL; /* OK in either case */ 491fa9e4066Sahrens } 492fa9e4066Sahrens 493fa9e4066Sahrens mutex_exit(&scl->scl_lock); 494fa9e4066Sahrens } 495fa9e4066Sahrens 496fa9e4066Sahrens /* 497fa9e4066Sahrens * Returns true if the config lock is held in the given manner. 498fa9e4066Sahrens */ 499fa9e4066Sahrens boolean_t 500fa9e4066Sahrens spa_config_held(spa_t *spa, krw_t rw) 501fa9e4066Sahrens { 502fa9e4066Sahrens spa_config_lock_t *scl = &spa->spa_config_lock; 503fa9e4066Sahrens boolean_t held; 504fa9e4066Sahrens 505fa9e4066Sahrens mutex_enter(&scl->scl_lock); 506fa9e4066Sahrens if (rw == RW_WRITER) 507fa9e4066Sahrens held = (scl->scl_writer == curthread); 508fa9e4066Sahrens else 509ea8dc4b6Seschrock held = !refcount_is_zero(&scl->scl_count); 510fa9e4066Sahrens mutex_exit(&scl->scl_lock); 511fa9e4066Sahrens 512fa9e4066Sahrens return (held); 513fa9e4066Sahrens } 514fa9e4066Sahrens 515fa9e4066Sahrens /* 516fa9e4066Sahrens * ========================================================================== 517fa9e4066Sahrens * SPA vdev locking 518fa9e4066Sahrens * ========================================================================== 519fa9e4066Sahrens */ 520fa9e4066Sahrens 521fa9e4066Sahrens /* 522ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 523ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 524fa9e4066Sahrens * It returns the next transaction group for the spa_t. 525fa9e4066Sahrens */ 526fa9e4066Sahrens uint64_t 527fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 528fa9e4066Sahrens { 529ea8dc4b6Seschrock /* 530ea8dc4b6Seschrock * Suspend scrub activity while we mess with the config. 531ea8dc4b6Seschrock */ 532ea8dc4b6Seschrock spa_scrub_suspend(spa); 533fa9e4066Sahrens 5340373e76bSbonwick mutex_enter(&spa_namespace_lock); 535ea8dc4b6Seschrock 536ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, spa); 537fa9e4066Sahrens 538fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 539fa9e4066Sahrens } 540fa9e4066Sahrens 541fa9e4066Sahrens /* 542fa9e4066Sahrens * Unlock the spa_t after adding or removing a vdev. Besides undoing the 543fa9e4066Sahrens * locking of spa_vdev_enter(), we also want make sure the transactions have 544fa9e4066Sahrens * synced to disk, and then update the global configuration cache with the new 545fa9e4066Sahrens * information. 546fa9e4066Sahrens */ 547fa9e4066Sahrens int 548fa9e4066Sahrens spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 549fa9e4066Sahrens { 5500e34b6a7Sbonwick int config_changed = B_FALSE; 551ea8dc4b6Seschrock 5520373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 5530e34b6a7Sbonwick 5540e34b6a7Sbonwick /* 5550e34b6a7Sbonwick * Reassess the DTLs. 5560e34b6a7Sbonwick */ 5570373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 5580e34b6a7Sbonwick 5590e34b6a7Sbonwick /* 5600373e76bSbonwick * If the config changed, notify the scrub thread that it must restart. 5610e34b6a7Sbonwick */ 5620e34b6a7Sbonwick if (error == 0 && !list_is_empty(&spa->spa_dirty_list)) { 5630e34b6a7Sbonwick config_changed = B_TRUE; 5640373e76bSbonwick spa_scrub_restart(spa, txg); 5650e34b6a7Sbonwick } 566ea8dc4b6Seschrock 567ea8dc4b6Seschrock spa_config_exit(spa, spa); 568fa9e4066Sahrens 569ea8dc4b6Seschrock /* 5705dabedeeSbonwick * Allow scrubbing to resume. 571ea8dc4b6Seschrock */ 572ea8dc4b6Seschrock spa_scrub_resume(spa); 573fa9e4066Sahrens 574fa9e4066Sahrens /* 575fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 576fa9e4066Sahrens * that there won't be more than one config change per txg. 577fa9e4066Sahrens * This allows us to use the txg as the generation number. 578fa9e4066Sahrens */ 579fa9e4066Sahrens if (error == 0) 580fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 581fa9e4066Sahrens 582fa9e4066Sahrens if (vd != NULL) { 583fa9e4066Sahrens ASSERT(!vd->vdev_detached || vd->vdev_dtl.smo_object == 0); 584fa9e4066Sahrens vdev_free(vd); 585fa9e4066Sahrens } 586fa9e4066Sahrens 587fa9e4066Sahrens /* 5880e34b6a7Sbonwick * If the config changed, update the config cache. 589fa9e4066Sahrens */ 5900e34b6a7Sbonwick if (config_changed) 591fa9e4066Sahrens spa_config_sync(); 592ea8dc4b6Seschrock 593ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 594fa9e4066Sahrens 595fa9e4066Sahrens return (error); 596fa9e4066Sahrens } 597fa9e4066Sahrens 598fa9e4066Sahrens /* 599fa9e4066Sahrens * ========================================================================== 600fa9e4066Sahrens * Miscellaneous functions 601fa9e4066Sahrens * ========================================================================== 602fa9e4066Sahrens */ 603fa9e4066Sahrens 604fa9e4066Sahrens /* 605fa9e4066Sahrens * Rename a spa_t. 606fa9e4066Sahrens */ 607fa9e4066Sahrens int 608fa9e4066Sahrens spa_rename(const char *name, const char *newname) 609fa9e4066Sahrens { 610fa9e4066Sahrens spa_t *spa; 611fa9e4066Sahrens int err; 612fa9e4066Sahrens 613fa9e4066Sahrens /* 614fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 615fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 616fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 617ea8dc4b6Seschrock * allow recursive calls for other reasons. 618fa9e4066Sahrens */ 619fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 620fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 621fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 622fa9e4066Sahrens return (err); 623fa9e4066Sahrens } 624fa9e4066Sahrens 625ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 626fa9e4066Sahrens 627fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 628fa9e4066Sahrens spa_strfree(spa->spa_name); 629fa9e4066Sahrens spa->spa_name = spa_strdup(newname); 630fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 631fa9e4066Sahrens 632fa9e4066Sahrens /* 633fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 634fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 635fa9e4066Sahrens * during the sync. 636fa9e4066Sahrens */ 637fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 638fa9e4066Sahrens 639ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 640fa9e4066Sahrens 6410373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 642fa9e4066Sahrens 643fa9e4066Sahrens /* 644fa9e4066Sahrens * Sync the updated config cache. 645fa9e4066Sahrens */ 646fa9e4066Sahrens spa_config_sync(); 647fa9e4066Sahrens 648fa9e4066Sahrens spa_close(spa, FTAG); 649fa9e4066Sahrens 650fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 651fa9e4066Sahrens 652fa9e4066Sahrens return (0); 653fa9e4066Sahrens } 654fa9e4066Sahrens 655fa9e4066Sahrens 656fa9e4066Sahrens /* 657fa9e4066Sahrens * Determine whether a pool with given pool_guid exists. If device_guid is 658fa9e4066Sahrens * non-zero, determine whether the pool exists *and* contains a device with the 659fa9e4066Sahrens * specified device_guid. 660fa9e4066Sahrens */ 661fa9e4066Sahrens boolean_t 662fa9e4066Sahrens spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 663fa9e4066Sahrens { 664fa9e4066Sahrens spa_t *spa; 665fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 666fa9e4066Sahrens 667ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 668fa9e4066Sahrens 669fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 670fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 671fa9e4066Sahrens continue; 672fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 673fa9e4066Sahrens continue; 674fa9e4066Sahrens if (spa_guid(spa) == pool_guid && (device_guid == 0 || 675fa9e4066Sahrens vdev_lookup_by_guid(spa->spa_root_vdev, device_guid))) 676fa9e4066Sahrens break; 677fa9e4066Sahrens } 678fa9e4066Sahrens 679fa9e4066Sahrens return (spa != NULL); 680fa9e4066Sahrens } 681fa9e4066Sahrens 682fa9e4066Sahrens char * 683fa9e4066Sahrens spa_strdup(const char *s) 684fa9e4066Sahrens { 685fa9e4066Sahrens size_t len; 686fa9e4066Sahrens char *new; 687fa9e4066Sahrens 688fa9e4066Sahrens len = strlen(s); 689fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 690fa9e4066Sahrens bcopy(s, new, len); 691fa9e4066Sahrens new[len] = '\0'; 692fa9e4066Sahrens 693fa9e4066Sahrens return (new); 694fa9e4066Sahrens } 695fa9e4066Sahrens 696fa9e4066Sahrens void 697fa9e4066Sahrens spa_strfree(char *s) 698fa9e4066Sahrens { 699fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 700fa9e4066Sahrens } 701fa9e4066Sahrens 702fa9e4066Sahrens uint64_t 703fa9e4066Sahrens spa_get_random(uint64_t range) 704fa9e4066Sahrens { 705fa9e4066Sahrens uint64_t r; 706fa9e4066Sahrens 707fa9e4066Sahrens ASSERT(range != 0); 708fa9e4066Sahrens 709fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 710fa9e4066Sahrens 711fa9e4066Sahrens return (r % range); 712fa9e4066Sahrens } 713fa9e4066Sahrens 714fa9e4066Sahrens void 715d80c45e0Sbonwick sprintf_blkptr(char *buf, int len, const blkptr_t *bp) 716fa9e4066Sahrens { 71744cd46caSbillm int d; 718fa9e4066Sahrens 719fa9e4066Sahrens if (bp == NULL) { 720fbabab8fSmaybee (void) snprintf(buf, len, "<NULL>"); 721fa9e4066Sahrens return; 722fa9e4066Sahrens } 723fa9e4066Sahrens 724fa9e4066Sahrens if (BP_IS_HOLE(bp)) { 725fbabab8fSmaybee (void) snprintf(buf, len, "<hole>"); 726fa9e4066Sahrens return; 727fa9e4066Sahrens } 728fa9e4066Sahrens 72944cd46caSbillm (void) snprintf(buf, len, "[L%llu %s] %llxL/%llxP ", 730fa9e4066Sahrens (u_longlong_t)BP_GET_LEVEL(bp), 731fa9e4066Sahrens dmu_ot[BP_GET_TYPE(bp)].ot_name, 732fa9e4066Sahrens (u_longlong_t)BP_GET_LSIZE(bp), 73344cd46caSbillm (u_longlong_t)BP_GET_PSIZE(bp)); 73444cd46caSbillm 73544cd46caSbillm for (d = 0; d < BP_GET_NDVAS(bp); d++) { 736d80c45e0Sbonwick const dva_t *dva = &bp->blk_dva[d]; 73744cd46caSbillm (void) snprintf(buf + strlen(buf), len - strlen(buf), 73844cd46caSbillm "DVA[%d]=<%llu:%llx:%llx> ", d, 73944cd46caSbillm (u_longlong_t)DVA_GET_VDEV(dva), 74044cd46caSbillm (u_longlong_t)DVA_GET_OFFSET(dva), 74144cd46caSbillm (u_longlong_t)DVA_GET_ASIZE(dva)); 74244cd46caSbillm } 74344cd46caSbillm 74444cd46caSbillm (void) snprintf(buf + strlen(buf), len - strlen(buf), 74544cd46caSbillm "%s %s %s %s birth=%llu fill=%llu cksum=%llx:%llx:%llx:%llx", 746fa9e4066Sahrens zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 747fa9e4066Sahrens zio_compress_table[BP_GET_COMPRESS(bp)].ci_name, 748fa9e4066Sahrens BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", 74944cd46caSbillm BP_IS_GANG(bp) ? "gang" : "contiguous", 750fa9e4066Sahrens (u_longlong_t)bp->blk_birth, 751fa9e4066Sahrens (u_longlong_t)bp->blk_fill, 752fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[0], 753fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[1], 754fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[2], 755fa9e4066Sahrens (u_longlong_t)bp->blk_cksum.zc_word[3]); 756fa9e4066Sahrens } 757fa9e4066Sahrens 758fa9e4066Sahrens void 759fa9e4066Sahrens spa_freeze(spa_t *spa) 760fa9e4066Sahrens { 761fa9e4066Sahrens uint64_t freeze_txg = 0; 762fa9e4066Sahrens 763ea8dc4b6Seschrock spa_config_enter(spa, RW_WRITER, FTAG); 764fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 765fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 766fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 767fa9e4066Sahrens } 768ea8dc4b6Seschrock spa_config_exit(spa, FTAG); 769fa9e4066Sahrens if (freeze_txg != 0) 770fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 771fa9e4066Sahrens } 772fa9e4066Sahrens 773fa9e4066Sahrens /* 774fa9e4066Sahrens * ========================================================================== 775fa9e4066Sahrens * Accessor functions 776fa9e4066Sahrens * ========================================================================== 777fa9e4066Sahrens */ 778fa9e4066Sahrens 779fa9e4066Sahrens krwlock_t * 780fa9e4066Sahrens spa_traverse_rwlock(spa_t *spa) 781fa9e4066Sahrens { 782fa9e4066Sahrens return (&spa->spa_traverse_lock); 783fa9e4066Sahrens } 784fa9e4066Sahrens 785fa9e4066Sahrens int 786fa9e4066Sahrens spa_traverse_wanted(spa_t *spa) 787fa9e4066Sahrens { 788fa9e4066Sahrens return (spa->spa_traverse_wanted); 789fa9e4066Sahrens } 790fa9e4066Sahrens 791fa9e4066Sahrens dsl_pool_t * 792fa9e4066Sahrens spa_get_dsl(spa_t *spa) 793fa9e4066Sahrens { 794fa9e4066Sahrens return (spa->spa_dsl_pool); 795fa9e4066Sahrens } 796fa9e4066Sahrens 797fa9e4066Sahrens blkptr_t * 798fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 799fa9e4066Sahrens { 800fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 801fa9e4066Sahrens } 802fa9e4066Sahrens 803fa9e4066Sahrens void 804fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 805fa9e4066Sahrens { 806fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 807fa9e4066Sahrens } 808fa9e4066Sahrens 809fa9e4066Sahrens void 810fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 811fa9e4066Sahrens { 812fa9e4066Sahrens if (spa->spa_root == NULL) 813fa9e4066Sahrens buf[0] = '\0'; 814fa9e4066Sahrens else 815fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 816fa9e4066Sahrens } 817fa9e4066Sahrens 818fa9e4066Sahrens int 819fa9e4066Sahrens spa_sync_pass(spa_t *spa) 820fa9e4066Sahrens { 821fa9e4066Sahrens return (spa->spa_sync_pass); 822fa9e4066Sahrens } 823fa9e4066Sahrens 824fa9e4066Sahrens char * 825fa9e4066Sahrens spa_name(spa_t *spa) 826fa9e4066Sahrens { 827fa9e4066Sahrens /* 828fa9e4066Sahrens * Accessing the name requires holding either the namespace lock or the 829fa9e4066Sahrens * config lock, both of which are required to do a rename. 830fa9e4066Sahrens */ 831fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock) || 832fa9e4066Sahrens spa_config_held(spa, RW_READER) || spa_config_held(spa, RW_WRITER)); 833fa9e4066Sahrens 834fa9e4066Sahrens return (spa->spa_name); 835fa9e4066Sahrens } 836fa9e4066Sahrens 837fa9e4066Sahrens uint64_t 838fa9e4066Sahrens spa_guid(spa_t *spa) 839fa9e4066Sahrens { 840fa9e4066Sahrens return (spa->spa_root_vdev->vdev_guid); 841fa9e4066Sahrens } 842fa9e4066Sahrens 843fa9e4066Sahrens uint64_t 844fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 845fa9e4066Sahrens { 846fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 847fa9e4066Sahrens } 848fa9e4066Sahrens 849fa9e4066Sahrens uint64_t 850fa9e4066Sahrens spa_first_txg(spa_t *spa) 851fa9e4066Sahrens { 852fa9e4066Sahrens return (spa->spa_first_txg); 853fa9e4066Sahrens } 854fa9e4066Sahrens 855fa9e4066Sahrens int 856fa9e4066Sahrens spa_state(spa_t *spa) 857fa9e4066Sahrens { 858fa9e4066Sahrens return (spa->spa_state); 859fa9e4066Sahrens } 860fa9e4066Sahrens 861fa9e4066Sahrens uint64_t 862fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 863fa9e4066Sahrens { 864fa9e4066Sahrens return (spa->spa_freeze_txg); 865fa9e4066Sahrens } 866fa9e4066Sahrens 867fa9e4066Sahrens /* 868fa9e4066Sahrens * In the future, this may select among different metaslab classes 869fa9e4066Sahrens * depending on the zdp. For now, there's no such distinction. 870fa9e4066Sahrens */ 871fa9e4066Sahrens metaslab_class_t * 872fa9e4066Sahrens spa_metaslab_class_select(spa_t *spa) 873fa9e4066Sahrens { 874fa9e4066Sahrens return (spa->spa_normal_class); 875fa9e4066Sahrens } 876fa9e4066Sahrens 877fa9e4066Sahrens /* 878*99653d4eSeschrock * Return how much space is allocated in the pool (ie. sum of all asize) 879fa9e4066Sahrens */ 880fa9e4066Sahrens uint64_t 881fa9e4066Sahrens spa_get_alloc(spa_t *spa) 882fa9e4066Sahrens { 883fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_alloc); 884fa9e4066Sahrens } 885fa9e4066Sahrens 886fa9e4066Sahrens /* 887*99653d4eSeschrock * Return how much (raid-z inflated) space there is in the pool. 888fa9e4066Sahrens */ 889fa9e4066Sahrens uint64_t 890fa9e4066Sahrens spa_get_space(spa_t *spa) 891fa9e4066Sahrens { 892fa9e4066Sahrens return (spa->spa_root_vdev->vdev_stat.vs_space); 893fa9e4066Sahrens } 894fa9e4066Sahrens 895*99653d4eSeschrock /* 896*99653d4eSeschrock * Return the amount of raid-z-deflated space in the pool. 897*99653d4eSeschrock */ 898*99653d4eSeschrock uint64_t 899*99653d4eSeschrock spa_get_dspace(spa_t *spa) 900*99653d4eSeschrock { 901*99653d4eSeschrock if (spa->spa_deflate) 902*99653d4eSeschrock return (spa->spa_root_vdev->vdev_stat.vs_dspace); 903*99653d4eSeschrock else 904*99653d4eSeschrock return (spa->spa_root_vdev->vdev_stat.vs_space); 905*99653d4eSeschrock } 906*99653d4eSeschrock 907fa9e4066Sahrens /* ARGSUSED */ 908fa9e4066Sahrens uint64_t 909fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 910fa9e4066Sahrens { 911fa9e4066Sahrens /* 912fa9e4066Sahrens * For now, the worst case is 512-byte RAID-Z blocks, in which 913fa9e4066Sahrens * case the space requirement is exactly 2x; so just assume that. 91444cd46caSbillm * Add to this the fact that we can have up to 3 DVAs per bp, and 91544cd46caSbillm * we have to multiply by a total of 6x. 91644cd46caSbillm */ 91744cd46caSbillm return (lsize * 6); 91844cd46caSbillm } 91944cd46caSbillm 92044cd46caSbillm uint64_t 92144cd46caSbillm spa_version(spa_t *spa) 92244cd46caSbillm { 92344cd46caSbillm return (spa->spa_ubsync.ub_version); 92444cd46caSbillm } 92544cd46caSbillm 92644cd46caSbillm int 92744cd46caSbillm spa_max_replication(spa_t *spa) 92844cd46caSbillm { 92944cd46caSbillm /* 93044cd46caSbillm * As of ZFS_VERSION == ZFS_VERSION_DITTO_BLOCKS, we are able to 93144cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 93244cd46caSbillm * replication level accordingly. 933fa9e4066Sahrens */ 93444cd46caSbillm if (spa_version(spa) < ZFS_VERSION_DITTO_BLOCKS) 93544cd46caSbillm return (1); 93644cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 937fa9e4066Sahrens } 938fa9e4066Sahrens 939*99653d4eSeschrock uint64_t 940*99653d4eSeschrock bp_get_dasize(spa_t *spa, const blkptr_t *bp) 941*99653d4eSeschrock { 942*99653d4eSeschrock int sz = 0, i; 943*99653d4eSeschrock 944*99653d4eSeschrock if (!spa->spa_deflate) 945*99653d4eSeschrock return (BP_GET_ASIZE(bp)); 946*99653d4eSeschrock 947*99653d4eSeschrock for (i = 0; i < SPA_DVAS_PER_BP; i++) { 948*99653d4eSeschrock vdev_t *vd = 949*99653d4eSeschrock vdev_lookup_top(spa, DVA_GET_VDEV(&bp->blk_dva[i])); 950*99653d4eSeschrock sz += (DVA_GET_ASIZE(&bp->blk_dva[i]) >> SPA_MINBLOCKSHIFT) * 951*99653d4eSeschrock vd->vdev_deflate_ratio; 952*99653d4eSeschrock } 953*99653d4eSeschrock return (sz); 954*99653d4eSeschrock } 955*99653d4eSeschrock 956fa9e4066Sahrens /* 957fa9e4066Sahrens * ========================================================================== 958fa9e4066Sahrens * Initialization and Termination 959fa9e4066Sahrens * ========================================================================== 960fa9e4066Sahrens */ 961fa9e4066Sahrens 962fa9e4066Sahrens static int 963fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 964fa9e4066Sahrens { 965fa9e4066Sahrens const spa_t *s1 = a1; 966fa9e4066Sahrens const spa_t *s2 = a2; 967fa9e4066Sahrens int s; 968fa9e4066Sahrens 969fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 970fa9e4066Sahrens if (s > 0) 971fa9e4066Sahrens return (1); 972fa9e4066Sahrens if (s < 0) 973fa9e4066Sahrens return (-1); 974fa9e4066Sahrens return (0); 975fa9e4066Sahrens } 976fa9e4066Sahrens 9770373e76bSbonwick int 9780373e76bSbonwick spa_busy(void) 9790373e76bSbonwick { 9800373e76bSbonwick return (spa_active_count); 9810373e76bSbonwick } 9820373e76bSbonwick 983fa9e4066Sahrens void 984fa9e4066Sahrens spa_init(int mode) 985fa9e4066Sahrens { 986fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 987fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 988fa9e4066Sahrens 989fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 990fa9e4066Sahrens offsetof(spa_t, spa_avl)); 991fa9e4066Sahrens 992*99653d4eSeschrock avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_spare_t), 993*99653d4eSeschrock offsetof(spa_spare_t, spare_avl)); 994*99653d4eSeschrock 995fa9e4066Sahrens spa_mode = mode; 996fa9e4066Sahrens 997fa9e4066Sahrens refcount_init(); 998fa9e4066Sahrens unique_init(); 999fa9e4066Sahrens zio_init(); 1000fa9e4066Sahrens dmu_init(); 1001fa9e4066Sahrens zil_init(); 1002fa9e4066Sahrens spa_config_load(); 1003fa9e4066Sahrens } 1004fa9e4066Sahrens 1005fa9e4066Sahrens void 1006fa9e4066Sahrens spa_fini(void) 1007fa9e4066Sahrens { 1008fa9e4066Sahrens spa_evict_all(); 1009fa9e4066Sahrens 1010fa9e4066Sahrens zil_fini(); 1011fa9e4066Sahrens dmu_fini(); 1012fa9e4066Sahrens zio_fini(); 1013fa9e4066Sahrens refcount_fini(); 1014fa9e4066Sahrens 1015fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 1016*99653d4eSeschrock avl_destroy(&spa_spare_avl); 1017fa9e4066Sahrens 1018fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1019fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1020fa9e4066Sahrens } 1021