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 /* 229842588bSGeorge Wilson * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 232a104a52SAlex Reece * Copyright (c) 2011, 2014 by Delphix. All rights reserved. 24e9103aaeSGarrett D'Amore * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 25*bc9014e6SJustin Gibbs * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 26fa9e4066Sahrens */ 27fa9e4066Sahrens 28fa9e4066Sahrens #include <sys/zfs_context.h> 29fa9e4066Sahrens #include <sys/spa_impl.h> 30283b8460SGeorge.Wilson #include <sys/spa_boot.h> 31fa9e4066Sahrens #include <sys/zio.h> 32fa9e4066Sahrens #include <sys/zio_checksum.h> 33fa9e4066Sahrens #include <sys/zio_compress.h> 34fa9e4066Sahrens #include <sys/dmu.h> 35fa9e4066Sahrens #include <sys/dmu_tx.h> 36fa9e4066Sahrens #include <sys/zap.h> 37fa9e4066Sahrens #include <sys/zil.h> 38fa9e4066Sahrens #include <sys/vdev_impl.h> 39fa9e4066Sahrens #include <sys/metaslab.h> 40fa9e4066Sahrens #include <sys/uberblock_impl.h> 41fa9e4066Sahrens #include <sys/txg.h> 42fa9e4066Sahrens #include <sys/avl.h> 43fa9e4066Sahrens #include <sys/unique.h> 44fa9e4066Sahrens #include <sys/dsl_pool.h> 45fa9e4066Sahrens #include <sys/dsl_dir.h> 46fa9e4066Sahrens #include <sys/dsl_prop.h> 473f9d6ad7SLin Ling #include <sys/dsl_scan.h> 48fa9e4066Sahrens #include <sys/fs/zfs.h> 496ce0521aSperrin #include <sys/metaslab_impl.h> 50e14bb325SJeff Bonwick #include <sys/arc.h> 51485bbbf5SGeorge Wilson #include <sys/ddt.h> 5291ebeef5Sahrens #include "zfs_prop.h" 53ad135b5dSChristopher Siden #include "zfeature_common.h" 54fa9e4066Sahrens 55fa9e4066Sahrens /* 56fa9e4066Sahrens * SPA locking 57fa9e4066Sahrens * 58fa9e4066Sahrens * There are four basic locks for managing spa_t structures: 59fa9e4066Sahrens * 60fa9e4066Sahrens * spa_namespace_lock (global mutex) 61fa9e4066Sahrens * 6244cd46caSbillm * This lock must be acquired to do any of the following: 63fa9e4066Sahrens * 6444cd46caSbillm * - Lookup a spa_t by name 6544cd46caSbillm * - Add or remove a spa_t from the namespace 6644cd46caSbillm * - Increase spa_refcount from non-zero 6744cd46caSbillm * - Check if spa_refcount is zero 6844cd46caSbillm * - Rename a spa_t 69ea8dc4b6Seschrock * - add/remove/attach/detach devices 7044cd46caSbillm * - Held for the duration of create/destroy/import/export 71fa9e4066Sahrens * 7244cd46caSbillm * It does not need to handle recursion. A create or destroy may 7344cd46caSbillm * reference objects (files or zvols) in other pools, but by 7444cd46caSbillm * definition they must have an existing reference, and will never need 7544cd46caSbillm * to lookup a spa_t by name. 76fa9e4066Sahrens * 77fa9e4066Sahrens * spa_refcount (per-spa refcount_t protected by mutex) 78fa9e4066Sahrens * 7944cd46caSbillm * This reference count keep track of any active users of the spa_t. The 8044cd46caSbillm * spa_t cannot be destroyed or freed while this is non-zero. Internally, 8144cd46caSbillm * the refcount is never really 'zero' - opening a pool implicitly keeps 82088f3894Sahrens * some references in the DMU. Internally we check against spa_minref, but 8344cd46caSbillm * present the image of a zero/non-zero value to consumers. 84fa9e4066Sahrens * 85e14bb325SJeff Bonwick * spa_config_lock[] (per-spa array of rwlocks) 86fa9e4066Sahrens * 8791ebeef5Sahrens * This protects the spa_t from config changes, and must be held in 8891ebeef5Sahrens * the following circumstances: 89fa9e4066Sahrens * 9044cd46caSbillm * - RW_READER to perform I/O to the spa 9144cd46caSbillm * - RW_WRITER to change the vdev config 92fa9e4066Sahrens * 93fa9e4066Sahrens * The locking order is fairly straightforward: 94fa9e4066Sahrens * 9544cd46caSbillm * spa_namespace_lock -> spa_refcount 96fa9e4066Sahrens * 9744cd46caSbillm * The namespace lock must be acquired to increase the refcount from 0 9844cd46caSbillm * or to check if it is zero. 99fa9e4066Sahrens * 100e14bb325SJeff Bonwick * spa_refcount -> spa_config_lock[] 101fa9e4066Sahrens * 10244cd46caSbillm * There must be at least one valid reference on the spa_t to acquire 10344cd46caSbillm * the config lock. 104fa9e4066Sahrens * 105e14bb325SJeff Bonwick * spa_namespace_lock -> spa_config_lock[] 106fa9e4066Sahrens * 10744cd46caSbillm * The namespace lock must always be taken before the config lock. 108fa9e4066Sahrens * 109fa9e4066Sahrens * 110e14bb325SJeff Bonwick * The spa_namespace_lock can be acquired directly and is globally visible. 111fa9e4066Sahrens * 112e14bb325SJeff Bonwick * The namespace is manipulated using the following functions, all of which 113e14bb325SJeff Bonwick * require the spa_namespace_lock to be held. 114fa9e4066Sahrens * 11544cd46caSbillm * spa_lookup() Lookup a spa_t by name. 116fa9e4066Sahrens * 11744cd46caSbillm * spa_add() Create a new spa_t in the namespace. 118fa9e4066Sahrens * 11944cd46caSbillm * spa_remove() Remove a spa_t from the namespace. This also 12044cd46caSbillm * frees up any memory associated with the spa_t. 121fa9e4066Sahrens * 12244cd46caSbillm * spa_next() Returns the next spa_t in the system, or the 12344cd46caSbillm * first if NULL is passed. 124fa9e4066Sahrens * 12544cd46caSbillm * spa_evict_all() Shutdown and remove all spa_t structures in 12644cd46caSbillm * the system. 127fa9e4066Sahrens * 128ea8dc4b6Seschrock * spa_guid_exists() Determine whether a pool/device guid exists. 129fa9e4066Sahrens * 130fa9e4066Sahrens * The spa_refcount is manipulated using the following functions: 131fa9e4066Sahrens * 13244cd46caSbillm * spa_open_ref() Adds a reference to the given spa_t. Must be 13344cd46caSbillm * called with spa_namespace_lock held if the 13444cd46caSbillm * refcount is currently zero. 135fa9e4066Sahrens * 13644cd46caSbillm * spa_close() Remove a reference from the spa_t. This will 13744cd46caSbillm * not free the spa_t or remove it from the 13844cd46caSbillm * namespace. No locking is required. 139fa9e4066Sahrens * 14044cd46caSbillm * spa_refcount_zero() Returns true if the refcount is currently 14144cd46caSbillm * zero. Must be called with spa_namespace_lock 14244cd46caSbillm * held. 143fa9e4066Sahrens * 144e14bb325SJeff Bonwick * The spa_config_lock[] is an array of rwlocks, ordered as follows: 145e14bb325SJeff Bonwick * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. 146e14bb325SJeff Bonwick * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). 147e14bb325SJeff Bonwick * 148e14bb325SJeff Bonwick * To read the configuration, it suffices to hold one of these locks as reader. 149e14bb325SJeff Bonwick * To modify the configuration, you must hold all locks as writer. To modify 150e14bb325SJeff Bonwick * vdev state without altering the vdev tree's topology (e.g. online/offline), 151e14bb325SJeff Bonwick * you must hold SCL_STATE and SCL_ZIO as writer. 152e14bb325SJeff Bonwick * 153e14bb325SJeff Bonwick * We use these distinct config locks to avoid recursive lock entry. 154e14bb325SJeff Bonwick * For example, spa_sync() (which holds SCL_CONFIG as reader) induces 155e14bb325SJeff Bonwick * block allocations (SCL_ALLOC), which may require reading space maps 156e14bb325SJeff Bonwick * from disk (dmu_read() -> zio_read() -> SCL_ZIO). 157e14bb325SJeff Bonwick * 158e14bb325SJeff Bonwick * The spa config locks cannot be normal rwlocks because we need the 159e14bb325SJeff Bonwick * ability to hand off ownership. For example, SCL_ZIO is acquired 160e14bb325SJeff Bonwick * by the issuing thread and later released by an interrupt thread. 161e14bb325SJeff Bonwick * They do, however, obey the usual write-wanted semantics to prevent 162e14bb325SJeff Bonwick * writer (i.e. system administrator) starvation. 163e14bb325SJeff Bonwick * 164e14bb325SJeff Bonwick * The lock acquisition rules are as follows: 165e14bb325SJeff Bonwick * 166e14bb325SJeff Bonwick * SCL_CONFIG 167e14bb325SJeff Bonwick * Protects changes to the vdev tree topology, such as vdev 168e14bb325SJeff Bonwick * add/remove/attach/detach. Protects the dirty config list 169e14bb325SJeff Bonwick * (spa_config_dirty_list) and the set of spares and l2arc devices. 170e14bb325SJeff Bonwick * 171e14bb325SJeff Bonwick * SCL_STATE 172e14bb325SJeff Bonwick * Protects changes to pool state and vdev state, such as vdev 173e14bb325SJeff Bonwick * online/offline/fault/degrade/clear. Protects the dirty state list 174e14bb325SJeff Bonwick * (spa_state_dirty_list) and global pool state (spa_state). 175e14bb325SJeff Bonwick * 176e14bb325SJeff Bonwick * SCL_ALLOC 177e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 178e14bb325SJeff Bonwick * Held as reader by metaslab_alloc() and metaslab_claim(). 179e14bb325SJeff Bonwick * 180e14bb325SJeff Bonwick * SCL_ZIO 181e14bb325SJeff Bonwick * Held by bp-level zios (those which have no io_vd upon entry) 182e14bb325SJeff Bonwick * to prevent changes to the vdev tree. The bp-level zio implicitly 183e14bb325SJeff Bonwick * protects all of its vdev child zios, which do not hold SCL_ZIO. 184e14bb325SJeff Bonwick * 185e14bb325SJeff Bonwick * SCL_FREE 186e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 187e14bb325SJeff Bonwick * Held as reader by metaslab_free(). SCL_FREE is distinct from 188e14bb325SJeff Bonwick * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free 189e14bb325SJeff Bonwick * blocks in zio_done() while another i/o that holds either 190e14bb325SJeff Bonwick * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. 191e14bb325SJeff Bonwick * 192e14bb325SJeff Bonwick * SCL_VDEV 193e14bb325SJeff Bonwick * Held as reader to prevent changes to the vdev tree during trivial 194b24ab676SJeff Bonwick * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the 195e14bb325SJeff Bonwick * other locks, and lower than all of them, to ensure that it's safe 196e14bb325SJeff Bonwick * to acquire regardless of caller context. 197e14bb325SJeff Bonwick * 198e14bb325SJeff Bonwick * In addition, the following rules apply: 199e14bb325SJeff Bonwick * 200e14bb325SJeff Bonwick * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. 201e14bb325SJeff Bonwick * The lock ordering is SCL_CONFIG > spa_props_lock. 202e14bb325SJeff Bonwick * 203e14bb325SJeff Bonwick * (b) I/O operations on leaf vdevs. For any zio operation that takes 204e14bb325SJeff Bonwick * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), 205e14bb325SJeff Bonwick * or zio_write_phys() -- the caller must ensure that the config cannot 206e14bb325SJeff Bonwick * cannot change in the interim, and that the vdev cannot be reopened. 207e14bb325SJeff Bonwick * SCL_STATE as reader suffices for both. 208fa9e4066Sahrens * 209ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 210fa9e4066Sahrens * 21144cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 212ea8dc4b6Seschrock * for writing. 213fa9e4066Sahrens * 21444cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 21544cd46caSbillm * to complete, sync the updated configs to the 216ea8dc4b6Seschrock * cache, and release the namespace lock. 217fa9e4066Sahrens * 218e14bb325SJeff Bonwick * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). 219e14bb325SJeff Bonwick * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual 220e14bb325SJeff Bonwick * locking is, always, based on spa_namespace_lock and spa_config_lock[]. 221e14bb325SJeff Bonwick * 222ad135b5dSChristopher Siden * spa_rename() is also implemented within this file since it requires 223e14bb325SJeff Bonwick * manipulation of the namespace. 224fa9e4066Sahrens */ 225fa9e4066Sahrens 226fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 227fa9e4066Sahrens kmutex_t spa_namespace_lock; 228fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 2290373e76bSbonwick static int spa_active_count; 230416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 231fa9e4066Sahrens 23299653d4eSeschrock static kmutex_t spa_spare_lock; 23339c23413Seschrock static avl_tree_t spa_spare_avl; 234fa94a07fSbrendan static kmutex_t spa_l2cache_lock; 235fa94a07fSbrendan static avl_tree_t spa_l2cache_avl; 23699653d4eSeschrock 237fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 2388ad4d6ddSJeff Bonwick int spa_mode_global; 239fa9e4066Sahrens 240fa9e4066Sahrens #ifdef ZFS_DEBUG 2413b2aab18SMatthew Ahrens /* Everything except dprintf and spa is on by default in debug builds */ 2423b2aab18SMatthew Ahrens int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_SPA); 243fa9e4066Sahrens #else 244fa9e4066Sahrens int zfs_flags = 0; 245fa9e4066Sahrens #endif 246fa9e4066Sahrens 2470125049cSahrens /* 2480125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 2490125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 2500125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 2518b36997aSMatthew Ahrens * This should only be used as a last resort, as it typically results 2528b36997aSMatthew Ahrens * in leaked space, or worse. 2530125049cSahrens */ 2547fd05ac4SMatthew Ahrens boolean_t zfs_recover = B_FALSE; 2557fd05ac4SMatthew Ahrens 2567fd05ac4SMatthew Ahrens /* 2577fd05ac4SMatthew Ahrens * If destroy encounters an EIO while reading metadata (e.g. indirect 2587fd05ac4SMatthew Ahrens * blocks), space referenced by the missing metadata can not be freed. 2597fd05ac4SMatthew Ahrens * Normally this causes the background destroy to become "stalled", as 2607fd05ac4SMatthew Ahrens * it is unable to make forward progress. While in this stalled state, 2617fd05ac4SMatthew Ahrens * all remaining space to free from the error-encountering filesystem is 2627fd05ac4SMatthew Ahrens * "temporarily leaked". Set this flag to cause it to ignore the EIO, 2637fd05ac4SMatthew Ahrens * permanently leak the space from indirect blocks that can not be read, 2647fd05ac4SMatthew Ahrens * and continue to free everything else that it can. 2657fd05ac4SMatthew Ahrens * 2667fd05ac4SMatthew Ahrens * The default, "stalling" behavior is useful if the storage partially 2677fd05ac4SMatthew Ahrens * fails (i.e. some but not all i/os fail), and then later recovers. In 2687fd05ac4SMatthew Ahrens * this case, we will be able to continue pool operations while it is 2697fd05ac4SMatthew Ahrens * partially failed, and when it recovers, we can continue to free the 2707fd05ac4SMatthew Ahrens * space, with no leaks. However, note that this case is actually 2717fd05ac4SMatthew Ahrens * fairly rare. 2727fd05ac4SMatthew Ahrens * 2737fd05ac4SMatthew Ahrens * Typically pools either (a) fail completely (but perhaps temporarily, 2747fd05ac4SMatthew Ahrens * e.g. a top-level vdev going offline), or (b) have localized, 2757fd05ac4SMatthew Ahrens * permanent errors (e.g. disk returns the wrong data due to bit flip or 2767fd05ac4SMatthew Ahrens * firmware bug). In case (a), this setting does not matter because the 2777fd05ac4SMatthew Ahrens * pool will be suspended and the sync thread will not be able to make 2787fd05ac4SMatthew Ahrens * forward progress regardless. In case (b), because the error is 2797fd05ac4SMatthew Ahrens * permanent, the best we can do is leak the minimum amount of space, 2807fd05ac4SMatthew Ahrens * which is what setting this flag will do. Therefore, it is reasonable 2817fd05ac4SMatthew Ahrens * for this flag to normally be set, but we chose the more conservative 2827fd05ac4SMatthew Ahrens * approach of not setting it, so that there is no possibility of 2837fd05ac4SMatthew Ahrens * leaking space in the "partial temporary" failure case. 2847fd05ac4SMatthew Ahrens */ 2857fd05ac4SMatthew Ahrens boolean_t zfs_free_leak_on_eio = B_FALSE; 2860125049cSahrens 28769962b56SMatthew Ahrens /* 28869962b56SMatthew Ahrens * Expiration time in milliseconds. This value has two meanings. First it is 28969962b56SMatthew Ahrens * used to determine when the spa_deadman() logic should fire. By default the 29069962b56SMatthew Ahrens * spa_deadman() will fire if spa_sync() has not completed in 1000 seconds. 29169962b56SMatthew Ahrens * Secondly, the value determines if an I/O is considered "hung". Any I/O that 29269962b56SMatthew Ahrens * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting 29369962b56SMatthew Ahrens * in a system panic. 29469962b56SMatthew Ahrens */ 29569962b56SMatthew Ahrens uint64_t zfs_deadman_synctime_ms = 1000000ULL; 296283b8460SGeorge.Wilson 297283b8460SGeorge.Wilson /* 29869962b56SMatthew Ahrens * Check time in milliseconds. This defines the frequency at which we check 29969962b56SMatthew Ahrens * for hung I/O. 300283b8460SGeorge.Wilson */ 30169962b56SMatthew Ahrens uint64_t zfs_deadman_checktime_ms = 5000ULL; 302283b8460SGeorge.Wilson 303283b8460SGeorge.Wilson /* 304283b8460SGeorge.Wilson * Override the zfs deadman behavior via /etc/system. By default the 305283b8460SGeorge.Wilson * deadman is enabled except on VMware and sparc deployments. 306283b8460SGeorge.Wilson */ 307283b8460SGeorge.Wilson int zfs_deadman_enabled = -1; 308283b8460SGeorge.Wilson 30969962b56SMatthew Ahrens /* 31069962b56SMatthew Ahrens * The worst case is single-sector max-parity RAID-Z blocks, in which 31169962b56SMatthew Ahrens * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) 31269962b56SMatthew Ahrens * times the size; so just assume that. Add to this the fact that 31369962b56SMatthew Ahrens * we can have up to 3 DVAs per bp, and one more factor of 2 because 31469962b56SMatthew Ahrens * the block may be dittoed with up to 3 DVAs by ddt_sync(). All together, 31569962b56SMatthew Ahrens * the worst case is: 31669962b56SMatthew Ahrens * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24 31769962b56SMatthew Ahrens */ 31869962b56SMatthew Ahrens int spa_asize_inflation = 24; 319fa9e4066Sahrens 3207d46dc6cSMatthew Ahrens /* 3217d46dc6cSMatthew Ahrens * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in 3227d46dc6cSMatthew Ahrens * the pool to be consumed. This ensures that we don't run the pool 3237d46dc6cSMatthew Ahrens * completely out of space, due to unaccounted changes (e.g. to the MOS). 3247d46dc6cSMatthew Ahrens * It also limits the worst-case time to allocate space. If we have 3257d46dc6cSMatthew Ahrens * less than this amount of free space, most ZPL operations (e.g. write, 3267d46dc6cSMatthew Ahrens * create) will return ENOSPC. 3277d46dc6cSMatthew Ahrens * 3287d46dc6cSMatthew Ahrens * Certain operations (e.g. file removal, most administrative actions) can 3297d46dc6cSMatthew Ahrens * use half the slop space. They will only return ENOSPC if less than half 3307d46dc6cSMatthew Ahrens * the slop space is free. Typically, once the pool has less than the slop 3317d46dc6cSMatthew Ahrens * space free, the user will use these operations to free up space in the pool. 3327d46dc6cSMatthew Ahrens * These are the operations that call dsl_pool_adjustedsize() with the netfree 3337d46dc6cSMatthew Ahrens * argument set to TRUE. 3347d46dc6cSMatthew Ahrens * 3357d46dc6cSMatthew Ahrens * A very restricted set of operations are always permitted, regardless of 3367d46dc6cSMatthew Ahrens * the amount of free space. These are the operations that call 3377d46dc6cSMatthew Ahrens * dsl_sync_task(ZFS_SPACE_CHECK_NONE), e.g. "zfs destroy". If these 3387d46dc6cSMatthew Ahrens * operations result in a net increase in the amount of space used, 3397d46dc6cSMatthew Ahrens * it is possible to run the pool completely out of space, causing it to 3407d46dc6cSMatthew Ahrens * be permanently read-only. 3417d46dc6cSMatthew Ahrens * 3427d46dc6cSMatthew Ahrens * See also the comments in zfs_space_check_t. 3437d46dc6cSMatthew Ahrens */ 3447d46dc6cSMatthew Ahrens int spa_slop_shift = 5; 3457d46dc6cSMatthew Ahrens 346e05725b1Sbonwick /* 347e05725b1Sbonwick * ========================================================================== 348e05725b1Sbonwick * SPA config locking 349e05725b1Sbonwick * ========================================================================== 350e05725b1Sbonwick */ 351e05725b1Sbonwick static void 352e14bb325SJeff Bonwick spa_config_lock_init(spa_t *spa) 353e14bb325SJeff Bonwick { 354e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 355e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 356e14bb325SJeff Bonwick mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); 357e14bb325SJeff Bonwick cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); 3583b2aab18SMatthew Ahrens refcount_create_untracked(&scl->scl_count); 359e14bb325SJeff Bonwick scl->scl_writer = NULL; 360e14bb325SJeff Bonwick scl->scl_write_wanted = 0; 361e14bb325SJeff Bonwick } 362e05725b1Sbonwick } 363e05725b1Sbonwick 364e05725b1Sbonwick static void 365e14bb325SJeff Bonwick spa_config_lock_destroy(spa_t *spa) 366e14bb325SJeff Bonwick { 367e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 368e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 369e14bb325SJeff Bonwick mutex_destroy(&scl->scl_lock); 370e14bb325SJeff Bonwick cv_destroy(&scl->scl_cv); 371e14bb325SJeff Bonwick refcount_destroy(&scl->scl_count); 372e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL); 373e14bb325SJeff Bonwick ASSERT(scl->scl_write_wanted == 0); 374e14bb325SJeff Bonwick } 375e14bb325SJeff Bonwick } 376e14bb325SJeff Bonwick 377e14bb325SJeff Bonwick int 378e14bb325SJeff Bonwick spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) 379e05725b1Sbonwick { 380e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 381e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 382e14bb325SJeff Bonwick if (!(locks & (1 << i))) 383e14bb325SJeff Bonwick continue; 384e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 385e14bb325SJeff Bonwick if (rw == RW_READER) { 386e14bb325SJeff Bonwick if (scl->scl_writer || scl->scl_write_wanted) { 387e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 388e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 389e14bb325SJeff Bonwick return (0); 390e14bb325SJeff Bonwick } 391e14bb325SJeff Bonwick } else { 392e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 393e14bb325SJeff Bonwick if (!refcount_is_zero(&scl->scl_count)) { 394e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 395e14bb325SJeff Bonwick spa_config_exit(spa, locks ^ (1 << i), tag); 396e14bb325SJeff Bonwick return (0); 397e14bb325SJeff Bonwick } 398e14bb325SJeff Bonwick scl->scl_writer = curthread; 399e14bb325SJeff Bonwick } 400e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 401e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 402e14bb325SJeff Bonwick } 403e14bb325SJeff Bonwick return (1); 404e05725b1Sbonwick } 405e05725b1Sbonwick 406e05725b1Sbonwick void 407e14bb325SJeff Bonwick spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) 408e05725b1Sbonwick { 409f64c0e34SEric Taylor int wlocks_held = 0; 410f64c0e34SEric Taylor 4113b2aab18SMatthew Ahrens ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY); 4123b2aab18SMatthew Ahrens 413e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 414e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 415f64c0e34SEric Taylor if (scl->scl_writer == curthread) 416f64c0e34SEric Taylor wlocks_held |= (1 << i); 417e14bb325SJeff Bonwick if (!(locks & (1 << i))) 418e14bb325SJeff Bonwick continue; 419e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 420e14bb325SJeff Bonwick if (rw == RW_READER) { 421e14bb325SJeff Bonwick while (scl->scl_writer || scl->scl_write_wanted) { 422e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 423e14bb325SJeff Bonwick } 424e14bb325SJeff Bonwick } else { 425e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 426e14bb325SJeff Bonwick while (!refcount_is_zero(&scl->scl_count)) { 427e14bb325SJeff Bonwick scl->scl_write_wanted++; 428e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 429e14bb325SJeff Bonwick scl->scl_write_wanted--; 430e14bb325SJeff Bonwick } 431e14bb325SJeff Bonwick scl->scl_writer = curthread; 432e14bb325SJeff Bonwick } 433e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 434e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 435e05725b1Sbonwick } 436f64c0e34SEric Taylor ASSERT(wlocks_held <= locks); 437e05725b1Sbonwick } 438e05725b1Sbonwick 439e05725b1Sbonwick void 440e14bb325SJeff Bonwick spa_config_exit(spa_t *spa, int locks, void *tag) 441e05725b1Sbonwick { 442e14bb325SJeff Bonwick for (int i = SCL_LOCKS - 1; i >= 0; i--) { 443e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 444e14bb325SJeff Bonwick if (!(locks & (1 << i))) 445e14bb325SJeff Bonwick continue; 446e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 447e14bb325SJeff Bonwick ASSERT(!refcount_is_zero(&scl->scl_count)); 448e14bb325SJeff Bonwick if (refcount_remove(&scl->scl_count, tag) == 0) { 449e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL || 450e14bb325SJeff Bonwick scl->scl_writer == curthread); 451e14bb325SJeff Bonwick scl->scl_writer = NULL; /* OK in either case */ 452e14bb325SJeff Bonwick cv_broadcast(&scl->scl_cv); 453e14bb325SJeff Bonwick } 454e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 455e05725b1Sbonwick } 456e05725b1Sbonwick } 457e05725b1Sbonwick 458e14bb325SJeff Bonwick int 459e14bb325SJeff Bonwick spa_config_held(spa_t *spa, int locks, krw_t rw) 460e05725b1Sbonwick { 461e14bb325SJeff Bonwick int locks_held = 0; 462e05725b1Sbonwick 463e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 464e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 465e14bb325SJeff Bonwick if (!(locks & (1 << i))) 466e14bb325SJeff Bonwick continue; 467e14bb325SJeff Bonwick if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) || 468e14bb325SJeff Bonwick (rw == RW_WRITER && scl->scl_writer == curthread)) 469e14bb325SJeff Bonwick locks_held |= 1 << i; 470e14bb325SJeff Bonwick } 471e14bb325SJeff Bonwick 472e14bb325SJeff Bonwick return (locks_held); 473e05725b1Sbonwick } 474e05725b1Sbonwick 475fa9e4066Sahrens /* 476fa9e4066Sahrens * ========================================================================== 477fa9e4066Sahrens * SPA namespace functions 478fa9e4066Sahrens * ========================================================================== 479fa9e4066Sahrens */ 480fa9e4066Sahrens 481fa9e4066Sahrens /* 482fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 483fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 484fa9e4066Sahrens */ 485fa9e4066Sahrens spa_t * 486fa9e4066Sahrens spa_lookup(const char *name) 487fa9e4066Sahrens { 488e14bb325SJeff Bonwick static spa_t search; /* spa_t is large; don't allocate on stack */ 489e14bb325SJeff Bonwick spa_t *spa; 490fa9e4066Sahrens avl_index_t where; 49140feaa91Sahrens char *cp; 492fa9e4066Sahrens 493fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 494fa9e4066Sahrens 4953b2aab18SMatthew Ahrens (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); 4963b2aab18SMatthew Ahrens 49740feaa91Sahrens /* 49840feaa91Sahrens * If it's a full dataset name, figure out the pool name and 49940feaa91Sahrens * just use that. 50040feaa91Sahrens */ 50178f17100SMatthew Ahrens cp = strpbrk(search.spa_name, "/@#"); 5023b2aab18SMatthew Ahrens if (cp != NULL) 50340feaa91Sahrens *cp = '\0'; 50440feaa91Sahrens 505fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 506fa9e4066Sahrens 507fa9e4066Sahrens return (spa); 508fa9e4066Sahrens } 509fa9e4066Sahrens 510283b8460SGeorge.Wilson /* 511283b8460SGeorge.Wilson * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. 512283b8460SGeorge.Wilson * If the zfs_deadman_enabled flag is set then it inspects all vdev queues 513283b8460SGeorge.Wilson * looking for potentially hung I/Os. 514283b8460SGeorge.Wilson */ 515283b8460SGeorge.Wilson void 516283b8460SGeorge.Wilson spa_deadman(void *arg) 517283b8460SGeorge.Wilson { 518283b8460SGeorge.Wilson spa_t *spa = arg; 519283b8460SGeorge.Wilson 5200713e232SGeorge Wilson /* 5210713e232SGeorge Wilson * Disable the deadman timer if the pool is suspended. 5220713e232SGeorge Wilson */ 5230713e232SGeorge Wilson if (spa_suspended(spa)) { 5240713e232SGeorge Wilson VERIFY(cyclic_reprogram(spa->spa_deadman_cycid, CY_INFINITY)); 5250713e232SGeorge Wilson return; 5260713e232SGeorge Wilson } 5270713e232SGeorge Wilson 528283b8460SGeorge.Wilson zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", 529283b8460SGeorge.Wilson (gethrtime() - spa->spa_sync_starttime) / NANOSEC, 530283b8460SGeorge.Wilson ++spa->spa_deadman_calls); 531283b8460SGeorge.Wilson if (zfs_deadman_enabled) 532283b8460SGeorge.Wilson vdev_deadman(spa->spa_root_vdev); 533283b8460SGeorge.Wilson } 534283b8460SGeorge.Wilson 535fa9e4066Sahrens /* 536fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 537fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 538fa9e4066Sahrens * exist by calling spa_lookup() first. 539fa9e4066Sahrens */ 540fa9e4066Sahrens spa_t * 541468c413aSTim Haley spa_add(const char *name, nvlist_t *config, const char *altroot) 542fa9e4066Sahrens { 543fa9e4066Sahrens spa_t *spa; 544c5904d13Seschrock spa_config_dirent_t *dp; 545283b8460SGeorge.Wilson cyc_handler_t hdlr; 546283b8460SGeorge.Wilson cyc_time_t when; 547fa9e4066Sahrens 548fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 549fa9e4066Sahrens 550fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 551fa9e4066Sahrens 552c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 553c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 55435a5a358SJonathan Adams mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 555*bc9014e6SJustin Gibbs mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL); 556c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 55735a5a358SJonathan Adams mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); 558c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 55935a5a358SJonathan Adams mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 560a1521560SJeff Bonwick mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); 561a1521560SJeff Bonwick mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); 562c3a66015SMatthew Ahrens mutex_init(&spa->spa_iokstat_lock, NULL, MUTEX_DEFAULT, NULL); 563c25056deSgw 564c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 565*bc9014e6SJustin Gibbs cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL); 56635a5a358SJonathan Adams cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); 567c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 568e14bb325SJeff Bonwick cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); 569c25056deSgw 570b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 571cde58dbcSMatthew Ahrens bplist_create(&spa->spa_free_bplist[t]); 572b24ab676SJeff Bonwick 573e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); 574fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 575fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 5760373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 577468c413aSTim Haley spa->spa_load_max_txg = UINT64_MAX; 57835a5a358SJonathan Adams spa->spa_proc = &p0; 57935a5a358SJonathan Adams spa->spa_proc_state = SPA_PROC_NONE; 580fa9e4066Sahrens 581283b8460SGeorge.Wilson hdlr.cyh_func = spa_deadman; 582283b8460SGeorge.Wilson hdlr.cyh_arg = spa; 583283b8460SGeorge.Wilson hdlr.cyh_level = CY_LOW_LEVEL; 584283b8460SGeorge.Wilson 58569962b56SMatthew Ahrens spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms); 586283b8460SGeorge.Wilson 587283b8460SGeorge.Wilson /* 588283b8460SGeorge.Wilson * This determines how often we need to check for hung I/Os after 589283b8460SGeorge.Wilson * the cyclic has already fired. Since checking for hung I/Os is 590283b8460SGeorge.Wilson * an expensive operation we don't want to check too frequently. 59169962b56SMatthew Ahrens * Instead wait for 5 seconds before checking again. 592283b8460SGeorge.Wilson */ 59369962b56SMatthew Ahrens when.cyt_interval = MSEC2NSEC(zfs_deadman_checktime_ms); 594283b8460SGeorge.Wilson when.cyt_when = CY_INFINITY; 595283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 596283b8460SGeorge.Wilson spa->spa_deadman_cycid = cyclic_add(&hdlr, &when); 597283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 598283b8460SGeorge.Wilson 599fa9e4066Sahrens refcount_create(&spa->spa_refcount); 600e14bb325SJeff Bonwick spa_config_lock_init(spa); 601fa9e4066Sahrens 602fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 603fa9e4066Sahrens 6040373e76bSbonwick /* 6050373e76bSbonwick * Set the alternate root, if there is one. 6060373e76bSbonwick */ 6070373e76bSbonwick if (altroot) { 6080373e76bSbonwick spa->spa_root = spa_strdup(altroot); 6090373e76bSbonwick spa_active_count++; 6100373e76bSbonwick } 6110373e76bSbonwick 612c5904d13Seschrock /* 613c5904d13Seschrock * Every pool starts with the default cachefile 614c5904d13Seschrock */ 615c5904d13Seschrock list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), 616c5904d13Seschrock offsetof(spa_config_dirent_t, scd_link)); 617c5904d13Seschrock 618c5904d13Seschrock dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); 619ef912c80STim Haley dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); 620c5904d13Seschrock list_insert_head(&spa->spa_config_list, dp); 621c5904d13Seschrock 6224b964adaSGeorge Wilson VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, 6234b964adaSGeorge Wilson KM_SLEEP) == 0); 6244b964adaSGeorge Wilson 625ad135b5dSChristopher Siden if (config != NULL) { 626ad135b5dSChristopher Siden nvlist_t *features; 627ad135b5dSChristopher Siden 628ad135b5dSChristopher Siden if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, 629ad135b5dSChristopher Siden &features) == 0) { 630ad135b5dSChristopher Siden VERIFY(nvlist_dup(features, &spa->spa_label_features, 631ad135b5dSChristopher Siden 0) == 0); 632ad135b5dSChristopher Siden } 633ad135b5dSChristopher Siden 634468c413aSTim Haley VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); 635ad135b5dSChristopher Siden } 636ad135b5dSChristopher Siden 637ad135b5dSChristopher Siden if (spa->spa_label_features == NULL) { 638ad135b5dSChristopher Siden VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, 639ad135b5dSChristopher Siden KM_SLEEP) == 0); 640ad135b5dSChristopher Siden } 641468c413aSTim Haley 642c3a66015SMatthew Ahrens spa->spa_iokstat = kstat_create("zfs", 0, name, 643c3a66015SMatthew Ahrens "disk", KSTAT_TYPE_IO, 1, 0); 644c3a66015SMatthew Ahrens if (spa->spa_iokstat) { 645c3a66015SMatthew Ahrens spa->spa_iokstat->ks_lock = &spa->spa_iokstat_lock; 646c3a66015SMatthew Ahrens kstat_install(spa->spa_iokstat); 647c3a66015SMatthew Ahrens } 648c3a66015SMatthew Ahrens 6493b2aab18SMatthew Ahrens spa->spa_debug = ((zfs_flags & ZFS_DEBUG_SPA) != 0); 6503b2aab18SMatthew Ahrens 65143466aaeSMax Grossman /* 65243466aaeSMax Grossman * As a pool is being created, treat all features as disabled by 65343466aaeSMax Grossman * setting SPA_FEATURE_DISABLED for all entries in the feature 65443466aaeSMax Grossman * refcount cache. 65543466aaeSMax Grossman */ 65643466aaeSMax Grossman for (int i = 0; i < SPA_FEATURES; i++) { 65743466aaeSMax Grossman spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; 65843466aaeSMax Grossman } 65943466aaeSMax Grossman 660fa9e4066Sahrens return (spa); 661fa9e4066Sahrens } 662fa9e4066Sahrens 663fa9e4066Sahrens /* 664fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 665fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 666fa9e4066Sahrens * deactivated. 667fa9e4066Sahrens */ 668fa9e4066Sahrens void 669fa9e4066Sahrens spa_remove(spa_t *spa) 670fa9e4066Sahrens { 671c5904d13Seschrock spa_config_dirent_t *dp; 672c5904d13Seschrock 673fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 674fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 675*bc9014e6SJustin Gibbs ASSERT3U(refcount_count(&spa->spa_refcount), ==, 0); 676fa9e4066Sahrens 6771195e687SMark J Musante nvlist_free(spa->spa_config_splitting); 6781195e687SMark J Musante 679fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 680fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 681fa9e4066Sahrens 6820373e76bSbonwick if (spa->spa_root) { 683fa9e4066Sahrens spa_strfree(spa->spa_root); 6840373e76bSbonwick spa_active_count--; 6850373e76bSbonwick } 686fa9e4066Sahrens 687c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 688c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 689c5904d13Seschrock if (dp->scd_path != NULL) 690c5904d13Seschrock spa_strfree(dp->scd_path); 691c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 692c5904d13Seschrock } 693c5904d13Seschrock 694c5904d13Seschrock list_destroy(&spa->spa_config_list); 6952f8aaab3Seschrock 696ad135b5dSChristopher Siden nvlist_free(spa->spa_label_features); 6974b964adaSGeorge Wilson nvlist_free(spa->spa_load_info); 698fa9e4066Sahrens spa_config_set(spa, NULL); 699fa9e4066Sahrens 700283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 701283b8460SGeorge.Wilson if (spa->spa_deadman_cycid != CYCLIC_NONE) 702283b8460SGeorge.Wilson cyclic_remove(spa->spa_deadman_cycid); 703283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 704283b8460SGeorge.Wilson spa->spa_deadman_cycid = CYCLIC_NONE; 705283b8460SGeorge.Wilson 706fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 70791ebeef5Sahrens 708e14bb325SJeff Bonwick spa_config_lock_destroy(spa); 709fa9e4066Sahrens 710c3a66015SMatthew Ahrens kstat_delete(spa->spa_iokstat); 711c3a66015SMatthew Ahrens spa->spa_iokstat = NULL; 712c3a66015SMatthew Ahrens 713b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 714cde58dbcSMatthew Ahrens bplist_destroy(&spa->spa_free_bplist[t]); 715b24ab676SJeff Bonwick 716c25056deSgw cv_destroy(&spa->spa_async_cv); 717*bc9014e6SJustin Gibbs cv_destroy(&spa->spa_evicting_os_cv); 71835a5a358SJonathan Adams cv_destroy(&spa->spa_proc_cv); 719c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 720e14bb325SJeff Bonwick cv_destroy(&spa->spa_suspend_cv); 721c25056deSgw 7225ad82045Snd mutex_destroy(&spa->spa_async_lock); 723c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 72435a5a358SJonathan Adams mutex_destroy(&spa->spa_errlog_lock); 725*bc9014e6SJustin Gibbs mutex_destroy(&spa->spa_evicting_os_lock); 72606eeb2adSek mutex_destroy(&spa->spa_history_lock); 72735a5a358SJonathan Adams mutex_destroy(&spa->spa_proc_lock); 728b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 72935a5a358SJonathan Adams mutex_destroy(&spa->spa_scrub_lock); 730e14bb325SJeff Bonwick mutex_destroy(&spa->spa_suspend_lock); 731a1521560SJeff Bonwick mutex_destroy(&spa->spa_vdev_top_lock); 732c3a66015SMatthew Ahrens mutex_destroy(&spa->spa_iokstat_lock); 7335ad82045Snd 734fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 735fa9e4066Sahrens } 736fa9e4066Sahrens 737fa9e4066Sahrens /* 738fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 739fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 740fa9e4066Sahrens */ 741fa9e4066Sahrens spa_t * 742fa9e4066Sahrens spa_next(spa_t *prev) 743fa9e4066Sahrens { 744fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 745fa9e4066Sahrens 746fa9e4066Sahrens if (prev) 747fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 748fa9e4066Sahrens else 749fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 750fa9e4066Sahrens } 751fa9e4066Sahrens 752fa9e4066Sahrens /* 753fa9e4066Sahrens * ========================================================================== 754fa9e4066Sahrens * SPA refcount functions 755fa9e4066Sahrens * ========================================================================== 756fa9e4066Sahrens */ 757fa9e4066Sahrens 758fa9e4066Sahrens /* 759fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 760fa9e4066Sahrens * have the namespace lock held. 761fa9e4066Sahrens */ 762fa9e4066Sahrens void 763fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 764fa9e4066Sahrens { 765088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || 766fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 767fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 768fa9e4066Sahrens } 769fa9e4066Sahrens 770fa9e4066Sahrens /* 771fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 772fa9e4066Sahrens * have the namespace lock held. 773fa9e4066Sahrens */ 774fa9e4066Sahrens void 775fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 776fa9e4066Sahrens { 777088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || 778fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 779fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 780fa9e4066Sahrens } 781fa9e4066Sahrens 782*bc9014e6SJustin Gibbs /* 783*bc9014e6SJustin Gibbs * Remove a reference to the given spa_t held by a dsl dir that is 784*bc9014e6SJustin Gibbs * being asynchronously released. Async releases occur from a taskq 785*bc9014e6SJustin Gibbs * performing eviction of dsl datasets and dirs. The namespace lock 786*bc9014e6SJustin Gibbs * isn't held and the hold by the object being evicted may contribute to 787*bc9014e6SJustin Gibbs * spa_minref (e.g. dataset or directory released during pool export), 788*bc9014e6SJustin Gibbs * so the asserts in spa_close() do not apply. 789*bc9014e6SJustin Gibbs */ 790*bc9014e6SJustin Gibbs void 791*bc9014e6SJustin Gibbs spa_async_close(spa_t *spa, void *tag) 792*bc9014e6SJustin Gibbs { 793*bc9014e6SJustin Gibbs (void) refcount_remove(&spa->spa_refcount, tag); 794*bc9014e6SJustin Gibbs } 795*bc9014e6SJustin Gibbs 796fa9e4066Sahrens /* 797fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 798088f3894Sahrens * spa_namespace_lock held. We really compare against spa_minref, which is the 799fa9e4066Sahrens * number of references acquired when opening a pool 800fa9e4066Sahrens */ 801fa9e4066Sahrens boolean_t 802fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 803fa9e4066Sahrens { 804fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 805fa9e4066Sahrens 806088f3894Sahrens return (refcount_count(&spa->spa_refcount) == spa->spa_minref); 807fa9e4066Sahrens } 808fa9e4066Sahrens 80999653d4eSeschrock /* 81099653d4eSeschrock * ========================================================================== 811fa94a07fSbrendan * SPA spare and l2cache tracking 81299653d4eSeschrock * ========================================================================== 81399653d4eSeschrock */ 81499653d4eSeschrock 815fa94a07fSbrendan /* 816fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 817fa94a07fSbrendan * for 'auxiliary' devices. 818fa94a07fSbrendan */ 819fa94a07fSbrendan 820fa94a07fSbrendan typedef struct spa_aux { 821fa94a07fSbrendan uint64_t aux_guid; 822fa94a07fSbrendan uint64_t aux_pool; 823fa94a07fSbrendan avl_node_t aux_avl; 824fa94a07fSbrendan int aux_count; 825fa94a07fSbrendan } spa_aux_t; 826fa94a07fSbrendan 827fa94a07fSbrendan static int 828fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 829fa94a07fSbrendan { 830fa94a07fSbrendan const spa_aux_t *sa = a; 831fa94a07fSbrendan const spa_aux_t *sb = b; 832fa94a07fSbrendan 833fa94a07fSbrendan if (sa->aux_guid < sb->aux_guid) 834fa94a07fSbrendan return (-1); 835fa94a07fSbrendan else if (sa->aux_guid > sb->aux_guid) 836fa94a07fSbrendan return (1); 837fa94a07fSbrendan else 838fa94a07fSbrendan return (0); 839fa94a07fSbrendan } 840fa94a07fSbrendan 841fa94a07fSbrendan void 842fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 843fa94a07fSbrendan { 844fa94a07fSbrendan avl_index_t where; 845fa94a07fSbrendan spa_aux_t search; 846fa94a07fSbrendan spa_aux_t *aux; 847fa94a07fSbrendan 848fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 849fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 850fa94a07fSbrendan aux->aux_count++; 851fa94a07fSbrendan } else { 852fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 853fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 854fa94a07fSbrendan aux->aux_count = 1; 855fa94a07fSbrendan avl_insert(avl, aux, where); 856fa94a07fSbrendan } 857fa94a07fSbrendan } 858fa94a07fSbrendan 859fa94a07fSbrendan void 860fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 861fa94a07fSbrendan { 862fa94a07fSbrendan spa_aux_t search; 863fa94a07fSbrendan spa_aux_t *aux; 864fa94a07fSbrendan avl_index_t where; 865fa94a07fSbrendan 866fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 867fa94a07fSbrendan aux = avl_find(avl, &search, &where); 868fa94a07fSbrendan 869fa94a07fSbrendan ASSERT(aux != NULL); 870fa94a07fSbrendan 871fa94a07fSbrendan if (--aux->aux_count == 0) { 872fa94a07fSbrendan avl_remove(avl, aux); 873fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 874fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 875fa94a07fSbrendan aux->aux_pool = 0ULL; 876fa94a07fSbrendan } 877fa94a07fSbrendan } 878fa94a07fSbrendan 879fa94a07fSbrendan boolean_t 88089a89ebfSlling spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) 881fa94a07fSbrendan { 882fa94a07fSbrendan spa_aux_t search, *found; 883fa94a07fSbrendan 884fa94a07fSbrendan search.aux_guid = guid; 88589a89ebfSlling found = avl_find(avl, &search, NULL); 886fa94a07fSbrendan 887fa94a07fSbrendan if (pool) { 888fa94a07fSbrendan if (found) 889fa94a07fSbrendan *pool = found->aux_pool; 890fa94a07fSbrendan else 891fa94a07fSbrendan *pool = 0ULL; 892fa94a07fSbrendan } 893fa94a07fSbrendan 89489a89ebfSlling if (refcnt) { 89589a89ebfSlling if (found) 89689a89ebfSlling *refcnt = found->aux_count; 89789a89ebfSlling else 89889a89ebfSlling *refcnt = 0; 89989a89ebfSlling } 90089a89ebfSlling 901fa94a07fSbrendan return (found != NULL); 902fa94a07fSbrendan } 903fa94a07fSbrendan 904fa94a07fSbrendan void 905fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 906fa94a07fSbrendan { 907fa94a07fSbrendan spa_aux_t search, *found; 908fa94a07fSbrendan avl_index_t where; 909fa94a07fSbrendan 910fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 911fa94a07fSbrendan found = avl_find(avl, &search, &where); 912fa94a07fSbrendan ASSERT(found != NULL); 913fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 914fa94a07fSbrendan 915fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 916fa94a07fSbrendan } 917fa94a07fSbrendan 91899653d4eSeschrock /* 91939c23413Seschrock * Spares are tracked globally due to the following constraints: 92039c23413Seschrock * 92139c23413Seschrock * - A spare may be part of multiple pools. 92239c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 92339c23413Seschrock * another pool. 92439c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 92539c23413Seschrock * the target is a spare in the same pool. 92639c23413Seschrock * 92739c23413Seschrock * We keep track of all spares on the system through the use of a reference 92839c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 92939c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 93039c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 93139c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 93239c23413Seschrock * pool), we also keep track of which pool its been made a part of. 93339c23413Seschrock * 93439c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 93539c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 93639c23413Seschrock * separate spare lock exists for the status query path, which does not need to 93739c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 93899653d4eSeschrock */ 93939c23413Seschrock 94099653d4eSeschrock static int 94199653d4eSeschrock spa_spare_compare(const void *a, const void *b) 94299653d4eSeschrock { 943fa94a07fSbrendan return (spa_aux_compare(a, b)); 94499653d4eSeschrock } 94599653d4eSeschrock 94699653d4eSeschrock void 94739c23413Seschrock spa_spare_add(vdev_t *vd) 94899653d4eSeschrock { 94999653d4eSeschrock mutex_enter(&spa_spare_lock); 95039c23413Seschrock ASSERT(!vd->vdev_isspare); 951fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 95239c23413Seschrock vd->vdev_isspare = B_TRUE; 95399653d4eSeschrock mutex_exit(&spa_spare_lock); 95499653d4eSeschrock } 95599653d4eSeschrock 95699653d4eSeschrock void 95739c23413Seschrock spa_spare_remove(vdev_t *vd) 95899653d4eSeschrock { 95999653d4eSeschrock mutex_enter(&spa_spare_lock); 96039c23413Seschrock ASSERT(vd->vdev_isspare); 961fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 96239c23413Seschrock vd->vdev_isspare = B_FALSE; 96399653d4eSeschrock mutex_exit(&spa_spare_lock); 96499653d4eSeschrock } 96599653d4eSeschrock 96699653d4eSeschrock boolean_t 96789a89ebfSlling spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) 96899653d4eSeschrock { 969fa94a07fSbrendan boolean_t found; 97099653d4eSeschrock 97199653d4eSeschrock mutex_enter(&spa_spare_lock); 97289a89ebfSlling found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); 97399653d4eSeschrock mutex_exit(&spa_spare_lock); 97499653d4eSeschrock 975fa94a07fSbrendan return (found); 97639c23413Seschrock } 97739c23413Seschrock 97839c23413Seschrock void 97939c23413Seschrock spa_spare_activate(vdev_t *vd) 98039c23413Seschrock { 98139c23413Seschrock mutex_enter(&spa_spare_lock); 98239c23413Seschrock ASSERT(vd->vdev_isspare); 983fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 984fa94a07fSbrendan mutex_exit(&spa_spare_lock); 985fa94a07fSbrendan } 98639c23413Seschrock 987fa94a07fSbrendan /* 988fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 989fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 990fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 991fa94a07fSbrendan */ 99239c23413Seschrock 993fa94a07fSbrendan static int 994fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 995fa94a07fSbrendan { 996fa94a07fSbrendan return (spa_aux_compare(a, b)); 997fa94a07fSbrendan } 998fa94a07fSbrendan 999fa94a07fSbrendan void 1000fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 1001fa94a07fSbrendan { 1002fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1003fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 1004fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 1005fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 1006fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1007fa94a07fSbrendan } 1008fa94a07fSbrendan 1009fa94a07fSbrendan void 1010fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 1011fa94a07fSbrendan { 1012fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1013fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 1014fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 1015fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 1016fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1017fa94a07fSbrendan } 1018fa94a07fSbrendan 1019fa94a07fSbrendan boolean_t 1020fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 1021fa94a07fSbrendan { 1022fa94a07fSbrendan boolean_t found; 1023fa94a07fSbrendan 1024fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 102589a89ebfSlling found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); 1026fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1027fa94a07fSbrendan 1028fa94a07fSbrendan return (found); 1029fa94a07fSbrendan } 1030fa94a07fSbrendan 1031fa94a07fSbrendan void 1032fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 1033fa94a07fSbrendan { 1034fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1035fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 1036fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 1037fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1038fa94a07fSbrendan } 1039fa94a07fSbrendan 1040fa9e4066Sahrens /* 1041fa9e4066Sahrens * ========================================================================== 1042fa9e4066Sahrens * SPA vdev locking 1043fa9e4066Sahrens * ========================================================================== 1044fa9e4066Sahrens */ 1045fa9e4066Sahrens 1046fa9e4066Sahrens /* 1047ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 1048ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 1049fa9e4066Sahrens * It returns the next transaction group for the spa_t. 1050fa9e4066Sahrens */ 1051fa9e4066Sahrens uint64_t 1052fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 1053fa9e4066Sahrens { 1054a1521560SJeff Bonwick mutex_enter(&spa->spa_vdev_top_lock); 1055bbfd46c4SJeff Bonwick mutex_enter(&spa_namespace_lock); 105688ecc943SGeorge Wilson return (spa_vdev_config_enter(spa)); 105788ecc943SGeorge Wilson } 105888ecc943SGeorge Wilson 105988ecc943SGeorge Wilson /* 106088ecc943SGeorge Wilson * Internal implementation for spa_vdev_enter(). Used when a vdev 106188ecc943SGeorge Wilson * operation requires multiple syncs (i.e. removing a device) while 106288ecc943SGeorge Wilson * keeping the spa_namespace_lock held. 106388ecc943SGeorge Wilson */ 106488ecc943SGeorge Wilson uint64_t 106588ecc943SGeorge Wilson spa_vdev_config_enter(spa_t *spa) 106688ecc943SGeorge Wilson { 106788ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 10683d7072f8Seschrock 1069e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1070fa9e4066Sahrens 1071fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 1072fa9e4066Sahrens } 1073fa9e4066Sahrens 1074fa9e4066Sahrens /* 107588ecc943SGeorge Wilson * Used in combination with spa_vdev_config_enter() to allow the syncing 107688ecc943SGeorge Wilson * of multiple transactions without releasing the spa_namespace_lock. 1077fa9e4066Sahrens */ 107888ecc943SGeorge Wilson void 107988ecc943SGeorge Wilson spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) 1080fa9e4066Sahrens { 108188ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 108288ecc943SGeorge Wilson 10830e34b6a7Sbonwick int config_changed = B_FALSE; 1084ea8dc4b6Seschrock 10850373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 10860e34b6a7Sbonwick 1087e14bb325SJeff Bonwick spa->spa_pending_vdev = NULL; 1088e14bb325SJeff Bonwick 10890e34b6a7Sbonwick /* 10900e34b6a7Sbonwick * Reassess the DTLs. 10910e34b6a7Sbonwick */ 10920373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 10930e34b6a7Sbonwick 1094e14bb325SJeff Bonwick if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { 10950e34b6a7Sbonwick config_changed = B_TRUE; 10968f18d1faSGeorge Wilson spa->spa_config_generation++; 10970e34b6a7Sbonwick } 1098ea8dc4b6Seschrock 109988ecc943SGeorge Wilson /* 110088ecc943SGeorge Wilson * Verify the metaslab classes. 110188ecc943SGeorge Wilson */ 1102b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); 1103b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); 110488ecc943SGeorge Wilson 1105e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1106fa9e4066Sahrens 110788ecc943SGeorge Wilson /* 110888ecc943SGeorge Wilson * Panic the system if the specified tag requires it. This 110988ecc943SGeorge Wilson * is useful for ensuring that configurations are updated 111088ecc943SGeorge Wilson * transactionally. 111188ecc943SGeorge Wilson */ 111288ecc943SGeorge Wilson if (zio_injection_enabled) 11131195e687SMark J Musante zio_handle_panic_injection(spa, tag, 0); 111488ecc943SGeorge Wilson 1115fa9e4066Sahrens /* 1116fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 1117fa9e4066Sahrens * that there won't be more than one config change per txg. 1118fa9e4066Sahrens * This allows us to use the txg as the generation number. 1119fa9e4066Sahrens */ 1120fa9e4066Sahrens if (error == 0) 1121fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 1122fa9e4066Sahrens 1123fa9e4066Sahrens if (vd != NULL) { 11240713e232SGeorge Wilson ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL); 11258ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1126fa9e4066Sahrens vdev_free(vd); 11278ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1128fa9e4066Sahrens } 1129fa9e4066Sahrens 1130fa9e4066Sahrens /* 11310e34b6a7Sbonwick * If the config changed, update the config cache. 1132fa9e4066Sahrens */ 11330e34b6a7Sbonwick if (config_changed) 1134c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 113588ecc943SGeorge Wilson } 1136ea8dc4b6Seschrock 113788ecc943SGeorge Wilson /* 113888ecc943SGeorge Wilson * Unlock the spa_t after adding or removing a vdev. Besides undoing the 113988ecc943SGeorge Wilson * locking of spa_vdev_enter(), we also want make sure the transactions have 114088ecc943SGeorge Wilson * synced to disk, and then update the global configuration cache with the new 114188ecc943SGeorge Wilson * information. 114288ecc943SGeorge Wilson */ 114388ecc943SGeorge Wilson int 114488ecc943SGeorge Wilson spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 114588ecc943SGeorge Wilson { 114688ecc943SGeorge Wilson spa_vdev_config_exit(spa, vd, txg, error, FTAG); 1147ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1148bbfd46c4SJeff Bonwick mutex_exit(&spa->spa_vdev_top_lock); 1149fa9e4066Sahrens 1150fa9e4066Sahrens return (error); 1151fa9e4066Sahrens } 1152fa9e4066Sahrens 1153e14bb325SJeff Bonwick /* 1154e14bb325SJeff Bonwick * Lock the given spa_t for the purpose of changing vdev state. 1155e14bb325SJeff Bonwick */ 1156e14bb325SJeff Bonwick void 11578f18d1faSGeorge Wilson spa_vdev_state_enter(spa_t *spa, int oplocks) 1158e14bb325SJeff Bonwick { 11598f18d1faSGeorge Wilson int locks = SCL_STATE_ALL | oplocks; 11608f18d1faSGeorge Wilson 1161dcba9f3fSGeorge Wilson /* 1162dcba9f3fSGeorge Wilson * Root pools may need to read of the underlying devfs filesystem 1163dcba9f3fSGeorge Wilson * when opening up a vdev. Unfortunately if we're holding the 1164dcba9f3fSGeorge Wilson * SCL_ZIO lock it will result in a deadlock when we try to issue 1165dcba9f3fSGeorge Wilson * the read from the root filesystem. Instead we "prefetch" 1166dcba9f3fSGeorge Wilson * the associated vnodes that we need prior to opening the 1167dcba9f3fSGeorge Wilson * underlying devices and cache them so that we can prevent 1168dcba9f3fSGeorge Wilson * any I/O when we are doing the actual open. 1169dcba9f3fSGeorge Wilson */ 1170dcba9f3fSGeorge Wilson if (spa_is_root(spa)) { 11719842588bSGeorge Wilson int low = locks & ~(SCL_ZIO - 1); 11729842588bSGeorge Wilson int high = locks & ~low; 11739842588bSGeorge Wilson 11749842588bSGeorge Wilson spa_config_enter(spa, high, spa, RW_WRITER); 1175dcba9f3fSGeorge Wilson vdev_hold(spa->spa_root_vdev); 11769842588bSGeorge Wilson spa_config_enter(spa, low, spa, RW_WRITER); 1177dcba9f3fSGeorge Wilson } else { 1178dcba9f3fSGeorge Wilson spa_config_enter(spa, locks, spa, RW_WRITER); 1179dcba9f3fSGeorge Wilson } 11808f18d1faSGeorge Wilson spa->spa_vdev_locks = locks; 1181e14bb325SJeff Bonwick } 1182e14bb325SJeff Bonwick 1183e14bb325SJeff Bonwick int 1184e14bb325SJeff Bonwick spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) 1185e14bb325SJeff Bonwick { 1186c6065d0fSGeorge Wilson boolean_t config_changed = B_FALSE; 1187c6065d0fSGeorge Wilson 1188b24ab676SJeff Bonwick if (vd != NULL || error == 0) 1189b24ab676SJeff Bonwick vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 1190b24ab676SJeff Bonwick 0, 0, B_FALSE); 1191b24ab676SJeff Bonwick 11928f18d1faSGeorge Wilson if (vd != NULL) { 1193e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 1194c6065d0fSGeorge Wilson config_changed = B_TRUE; 11958f18d1faSGeorge Wilson spa->spa_config_generation++; 11968f18d1faSGeorge Wilson } 1197e14bb325SJeff Bonwick 1198dcba9f3fSGeorge Wilson if (spa_is_root(spa)) 1199dcba9f3fSGeorge Wilson vdev_rele(spa->spa_root_vdev); 1200dcba9f3fSGeorge Wilson 12018f18d1faSGeorge Wilson ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); 12028f18d1faSGeorge Wilson spa_config_exit(spa, spa->spa_vdev_locks, spa); 1203e14bb325SJeff Bonwick 12048ad4d6ddSJeff Bonwick /* 12058ad4d6ddSJeff Bonwick * If anything changed, wait for it to sync. This ensures that, 12068ad4d6ddSJeff Bonwick * from the system administrator's perspective, zpool(1M) commands 12078ad4d6ddSJeff Bonwick * are synchronous. This is important for things like zpool offline: 12088ad4d6ddSJeff Bonwick * when the command completes, you expect no further I/O from ZFS. 12098ad4d6ddSJeff Bonwick */ 12108ad4d6ddSJeff Bonwick if (vd != NULL) 12118ad4d6ddSJeff Bonwick txg_wait_synced(spa->spa_dsl_pool, 0); 12128ad4d6ddSJeff Bonwick 1213c6065d0fSGeorge Wilson /* 1214c6065d0fSGeorge Wilson * If the config changed, update the config cache. 1215c6065d0fSGeorge Wilson */ 1216c6065d0fSGeorge Wilson if (config_changed) { 1217c6065d0fSGeorge Wilson mutex_enter(&spa_namespace_lock); 1218c6065d0fSGeorge Wilson spa_config_sync(spa, B_FALSE, B_TRUE); 1219c6065d0fSGeorge Wilson mutex_exit(&spa_namespace_lock); 1220c6065d0fSGeorge Wilson } 1221c6065d0fSGeorge Wilson 1222e14bb325SJeff Bonwick return (error); 1223e14bb325SJeff Bonwick } 1224e14bb325SJeff Bonwick 1225fa9e4066Sahrens /* 1226fa9e4066Sahrens * ========================================================================== 1227fa9e4066Sahrens * Miscellaneous functions 1228fa9e4066Sahrens * ========================================================================== 1229fa9e4066Sahrens */ 1230fa9e4066Sahrens 1231ad135b5dSChristopher Siden void 123243466aaeSMax Grossman spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx) 1233ad135b5dSChristopher Siden { 12342acef22dSMatthew Ahrens if (!nvlist_exists(spa->spa_label_features, feature)) { 12352acef22dSMatthew Ahrens fnvlist_add_boolean(spa->spa_label_features, feature); 123643466aaeSMax Grossman /* 123743466aaeSMax Grossman * When we are creating the pool (tx_txg==TXG_INITIAL), we can't 123843466aaeSMax Grossman * dirty the vdev config because lock SCL_CONFIG is not held. 123943466aaeSMax Grossman * Thankfully, in this case we don't need to dirty the config 124043466aaeSMax Grossman * because it will be written out anyway when we finish 124143466aaeSMax Grossman * creating the pool. 124243466aaeSMax Grossman */ 124343466aaeSMax Grossman if (tx->tx_txg != TXG_INITIAL) 124443466aaeSMax Grossman vdev_config_dirty(spa->spa_root_vdev); 12452acef22dSMatthew Ahrens } 1246ad135b5dSChristopher Siden } 1247ad135b5dSChristopher Siden 1248ad135b5dSChristopher Siden void 1249ad135b5dSChristopher Siden spa_deactivate_mos_feature(spa_t *spa, const char *feature) 1250ad135b5dSChristopher Siden { 12512acef22dSMatthew Ahrens if (nvlist_remove_all(spa->spa_label_features, feature) == 0) 12522acef22dSMatthew Ahrens vdev_config_dirty(spa->spa_root_vdev); 1253ad135b5dSChristopher Siden } 1254ad135b5dSChristopher Siden 1255fa9e4066Sahrens /* 1256fa9e4066Sahrens * Rename a spa_t. 1257fa9e4066Sahrens */ 1258fa9e4066Sahrens int 1259fa9e4066Sahrens spa_rename(const char *name, const char *newname) 1260fa9e4066Sahrens { 1261fa9e4066Sahrens spa_t *spa; 1262fa9e4066Sahrens int err; 1263fa9e4066Sahrens 1264fa9e4066Sahrens /* 1265fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 1266fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 1267fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 1268ea8dc4b6Seschrock * allow recursive calls for other reasons. 1269fa9e4066Sahrens */ 1270fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 1271fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 1272fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1273fa9e4066Sahrens return (err); 1274fa9e4066Sahrens } 1275fa9e4066Sahrens 1276e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1277fa9e4066Sahrens 1278fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 1279e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); 1280fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 1281fa9e4066Sahrens 1282fa9e4066Sahrens /* 1283fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 1284fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 1285fa9e4066Sahrens * during the sync. 1286fa9e4066Sahrens */ 1287fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 1288fa9e4066Sahrens 1289e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1290fa9e4066Sahrens 12910373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 1292fa9e4066Sahrens 1293fa9e4066Sahrens /* 1294fa9e4066Sahrens * Sync the updated config cache. 1295fa9e4066Sahrens */ 1296c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 1297fa9e4066Sahrens 1298fa9e4066Sahrens spa_close(spa, FTAG); 1299fa9e4066Sahrens 1300fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1301fa9e4066Sahrens 1302fa9e4066Sahrens return (0); 1303fa9e4066Sahrens } 1304fa9e4066Sahrens 1305fa9e4066Sahrens /* 1306f9af39baSGeorge Wilson * Return the spa_t associated with given pool_guid, if it exists. If 1307f9af39baSGeorge Wilson * device_guid is non-zero, determine whether the pool exists *and* contains 1308f9af39baSGeorge Wilson * a device with the specified device_guid. 1309fa9e4066Sahrens */ 1310f9af39baSGeorge Wilson spa_t * 1311f9af39baSGeorge Wilson spa_by_guid(uint64_t pool_guid, uint64_t device_guid) 1312fa9e4066Sahrens { 1313fa9e4066Sahrens spa_t *spa; 1314fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 1315fa9e4066Sahrens 1316ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1317fa9e4066Sahrens 1318fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 1319fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 1320fa9e4066Sahrens continue; 1321fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 1322fa9e4066Sahrens continue; 132339c23413Seschrock if (spa_guid(spa) == pool_guid) { 132439c23413Seschrock if (device_guid == 0) 132539c23413Seschrock break; 132639c23413Seschrock 132739c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 132839c23413Seschrock device_guid) != NULL) 132939c23413Seschrock break; 133039c23413Seschrock 133139c23413Seschrock /* 13328654d025Sperrin * Check any devices we may be in the process of adding. 133339c23413Seschrock */ 133439c23413Seschrock if (spa->spa_pending_vdev) { 133539c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 133639c23413Seschrock device_guid) != NULL) 133739c23413Seschrock break; 133839c23413Seschrock } 133939c23413Seschrock } 1340fa9e4066Sahrens } 1341fa9e4066Sahrens 1342f9af39baSGeorge Wilson return (spa); 1343f9af39baSGeorge Wilson } 1344f9af39baSGeorge Wilson 1345f9af39baSGeorge Wilson /* 1346f9af39baSGeorge Wilson * Determine whether a pool with the given pool_guid exists. 1347f9af39baSGeorge Wilson */ 1348f9af39baSGeorge Wilson boolean_t 1349f9af39baSGeorge Wilson spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 1350f9af39baSGeorge Wilson { 1351f9af39baSGeorge Wilson return (spa_by_guid(pool_guid, device_guid) != NULL); 1352fa9e4066Sahrens } 1353fa9e4066Sahrens 1354fa9e4066Sahrens char * 1355fa9e4066Sahrens spa_strdup(const char *s) 1356fa9e4066Sahrens { 1357fa9e4066Sahrens size_t len; 1358fa9e4066Sahrens char *new; 1359fa9e4066Sahrens 1360fa9e4066Sahrens len = strlen(s); 1361fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 1362fa9e4066Sahrens bcopy(s, new, len); 1363fa9e4066Sahrens new[len] = '\0'; 1364fa9e4066Sahrens 1365fa9e4066Sahrens return (new); 1366fa9e4066Sahrens } 1367fa9e4066Sahrens 1368fa9e4066Sahrens void 1369fa9e4066Sahrens spa_strfree(char *s) 1370fa9e4066Sahrens { 1371fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 1372fa9e4066Sahrens } 1373fa9e4066Sahrens 1374fa9e4066Sahrens uint64_t 1375fa9e4066Sahrens spa_get_random(uint64_t range) 1376fa9e4066Sahrens { 1377fa9e4066Sahrens uint64_t r; 1378fa9e4066Sahrens 1379fa9e4066Sahrens ASSERT(range != 0); 1380fa9e4066Sahrens 1381fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 1382fa9e4066Sahrens 1383fa9e4066Sahrens return (r % range); 1384fa9e4066Sahrens } 1385fa9e4066Sahrens 13861195e687SMark J Musante uint64_t 13871195e687SMark J Musante spa_generate_guid(spa_t *spa) 13881195e687SMark J Musante { 13891195e687SMark J Musante uint64_t guid = spa_get_random(-1ULL); 13901195e687SMark J Musante 13911195e687SMark J Musante if (spa != NULL) { 13921195e687SMark J Musante while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) 13931195e687SMark J Musante guid = spa_get_random(-1ULL); 13941195e687SMark J Musante } else { 13951195e687SMark J Musante while (guid == 0 || spa_guid_exists(guid, 0)) 13961195e687SMark J Musante guid = spa_get_random(-1ULL); 13971195e687SMark J Musante } 13981195e687SMark J Musante 13991195e687SMark J Musante return (guid); 14001195e687SMark J Musante } 14011195e687SMark J Musante 1402fa9e4066Sahrens void 140343466aaeSMax Grossman snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp) 1404fa9e4066Sahrens { 1405ad135b5dSChristopher Siden char type[256]; 1406f0ba89beSJeff Bonwick char *checksum = NULL; 1407f0ba89beSJeff Bonwick char *compress = NULL; 1408f0ba89beSJeff Bonwick 1409f0ba89beSJeff Bonwick if (bp != NULL) { 1410ad135b5dSChristopher Siden if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { 1411ad135b5dSChristopher Siden dmu_object_byteswap_t bswap = 1412ad135b5dSChristopher Siden DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); 1413ad135b5dSChristopher Siden (void) snprintf(type, sizeof (type), "bswap %s %s", 1414ad135b5dSChristopher Siden DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? 1415ad135b5dSChristopher Siden "metadata" : "data", 1416ad135b5dSChristopher Siden dmu_ot_byteswap[bswap].ob_name); 1417ad135b5dSChristopher Siden } else { 1418ad135b5dSChristopher Siden (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, 1419ad135b5dSChristopher Siden sizeof (type)); 1420ad135b5dSChristopher Siden } 14215d7b4d43SMatthew Ahrens if (!BP_IS_EMBEDDED(bp)) { 14225d7b4d43SMatthew Ahrens checksum = 14235d7b4d43SMatthew Ahrens zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; 14245d7b4d43SMatthew Ahrens } 1425f0ba89beSJeff Bonwick compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; 1426f0ba89beSJeff Bonwick } 1427fa9e4066Sahrens 142843466aaeSMax Grossman SNPRINTF_BLKPTR(snprintf, ' ', buf, buflen, bp, type, checksum, 142943466aaeSMax Grossman compress); 1430fa9e4066Sahrens } 1431fa9e4066Sahrens 1432fa9e4066Sahrens void 1433fa9e4066Sahrens spa_freeze(spa_t *spa) 1434fa9e4066Sahrens { 1435fa9e4066Sahrens uint64_t freeze_txg = 0; 1436fa9e4066Sahrens 1437e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1438fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 1439fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 1440fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 1441fa9e4066Sahrens } 1442e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1443fa9e4066Sahrens if (freeze_txg != 0) 1444fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 1445fa9e4066Sahrens } 1446fa9e4066Sahrens 14470125049cSahrens void 14480125049cSahrens zfs_panic_recover(const char *fmt, ...) 14490125049cSahrens { 14500125049cSahrens va_list adx; 14510125049cSahrens 14520125049cSahrens va_start(adx, fmt); 14530125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 14540125049cSahrens va_end(adx); 14550125049cSahrens } 14560125049cSahrens 14573f9d6ad7SLin Ling /* 14583f9d6ad7SLin Ling * This is a stripped-down version of strtoull, suitable only for converting 1459f7170741SWill Andrews * lowercase hexadecimal numbers that don't overflow. 14603f9d6ad7SLin Ling */ 14613f9d6ad7SLin Ling uint64_t 14623f9d6ad7SLin Ling strtonum(const char *str, char **nptr) 14633f9d6ad7SLin Ling { 14643f9d6ad7SLin Ling uint64_t val = 0; 14653f9d6ad7SLin Ling char c; 14663f9d6ad7SLin Ling int digit; 14673f9d6ad7SLin Ling 14683f9d6ad7SLin Ling while ((c = *str) != '\0') { 14693f9d6ad7SLin Ling if (c >= '0' && c <= '9') 14703f9d6ad7SLin Ling digit = c - '0'; 14713f9d6ad7SLin Ling else if (c >= 'a' && c <= 'f') 14723f9d6ad7SLin Ling digit = 10 + c - 'a'; 14733f9d6ad7SLin Ling else 14743f9d6ad7SLin Ling break; 14753f9d6ad7SLin Ling 14763f9d6ad7SLin Ling val *= 16; 14773f9d6ad7SLin Ling val += digit; 14783f9d6ad7SLin Ling 14793f9d6ad7SLin Ling str++; 14803f9d6ad7SLin Ling } 14813f9d6ad7SLin Ling 14823f9d6ad7SLin Ling if (nptr) 14833f9d6ad7SLin Ling *nptr = (char *)str; 14843f9d6ad7SLin Ling 14853f9d6ad7SLin Ling return (val); 14863f9d6ad7SLin Ling } 14873f9d6ad7SLin Ling 1488fa9e4066Sahrens /* 1489fa9e4066Sahrens * ========================================================================== 1490fa9e4066Sahrens * Accessor functions 1491fa9e4066Sahrens * ========================================================================== 1492fa9e4066Sahrens */ 1493fa9e4066Sahrens 1494088f3894Sahrens boolean_t 149588b7b0f2SMatthew Ahrens spa_shutting_down(spa_t *spa) 1496fa9e4066Sahrens { 149788b7b0f2SMatthew Ahrens return (spa->spa_async_suspended); 1498fa9e4066Sahrens } 1499fa9e4066Sahrens 1500fa9e4066Sahrens dsl_pool_t * 1501fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1502fa9e4066Sahrens { 1503fa9e4066Sahrens return (spa->spa_dsl_pool); 1504fa9e4066Sahrens } 1505fa9e4066Sahrens 1506ad135b5dSChristopher Siden boolean_t 1507ad135b5dSChristopher Siden spa_is_initializing(spa_t *spa) 1508ad135b5dSChristopher Siden { 1509ad135b5dSChristopher Siden return (spa->spa_is_initializing); 1510ad135b5dSChristopher Siden } 1511ad135b5dSChristopher Siden 1512fa9e4066Sahrens blkptr_t * 1513fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1514fa9e4066Sahrens { 1515fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1516fa9e4066Sahrens } 1517fa9e4066Sahrens 1518fa9e4066Sahrens void 1519fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1520fa9e4066Sahrens { 1521fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1522fa9e4066Sahrens } 1523fa9e4066Sahrens 1524fa9e4066Sahrens void 1525fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1526fa9e4066Sahrens { 1527fa9e4066Sahrens if (spa->spa_root == NULL) 1528fa9e4066Sahrens buf[0] = '\0'; 1529fa9e4066Sahrens else 1530fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1531fa9e4066Sahrens } 1532fa9e4066Sahrens 1533fa9e4066Sahrens int 1534fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1535fa9e4066Sahrens { 1536fa9e4066Sahrens return (spa->spa_sync_pass); 1537fa9e4066Sahrens } 1538fa9e4066Sahrens 1539fa9e4066Sahrens char * 1540fa9e4066Sahrens spa_name(spa_t *spa) 1541fa9e4066Sahrens { 1542fa9e4066Sahrens return (spa->spa_name); 1543fa9e4066Sahrens } 1544fa9e4066Sahrens 1545fa9e4066Sahrens uint64_t 1546fa9e4066Sahrens spa_guid(spa_t *spa) 1547fa9e4066Sahrens { 1548dfbb9432SGeorge Wilson dsl_pool_t *dp = spa_get_dsl(spa); 1549dfbb9432SGeorge Wilson uint64_t guid; 1550dfbb9432SGeorge Wilson 1551b5989ec7Seschrock /* 1552b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1553b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1554e9103aaeSGarrett D'Amore * vdev. We stash the original pool guid in 'spa_config_guid' to handle 1555b5989ec7Seschrock * this case. 1556b5989ec7Seschrock */ 1557dfbb9432SGeorge Wilson if (spa->spa_root_vdev == NULL) 1558dfbb9432SGeorge Wilson return (spa->spa_config_guid); 1559dfbb9432SGeorge Wilson 1560dfbb9432SGeorge Wilson guid = spa->spa_last_synced_guid != 0 ? 1561dfbb9432SGeorge Wilson spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; 1562dfbb9432SGeorge Wilson 1563dfbb9432SGeorge Wilson /* 1564dfbb9432SGeorge Wilson * Return the most recently synced out guid unless we're 1565dfbb9432SGeorge Wilson * in syncing context. 1566dfbb9432SGeorge Wilson */ 1567dfbb9432SGeorge Wilson if (dp && dsl_pool_sync_context(dp)) 1568b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1569b5989ec7Seschrock else 1570dfbb9432SGeorge Wilson return (guid); 1571e9103aaeSGarrett D'Amore } 1572e9103aaeSGarrett D'Amore 1573e9103aaeSGarrett D'Amore uint64_t 1574e9103aaeSGarrett D'Amore spa_load_guid(spa_t *spa) 1575e9103aaeSGarrett D'Amore { 1576e9103aaeSGarrett D'Amore /* 1577e9103aaeSGarrett D'Amore * This is a GUID that exists solely as a reference for the 1578e9103aaeSGarrett D'Amore * purposes of the arc. It is generated at load time, and 1579e9103aaeSGarrett D'Amore * is never written to persistent storage. 1580e9103aaeSGarrett D'Amore */ 1581e9103aaeSGarrett D'Amore return (spa->spa_load_guid); 1582fa9e4066Sahrens } 1583fa9e4066Sahrens 1584fa9e4066Sahrens uint64_t 1585fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1586fa9e4066Sahrens { 1587fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1588fa9e4066Sahrens } 1589fa9e4066Sahrens 1590fa9e4066Sahrens uint64_t 1591fa9e4066Sahrens spa_first_txg(spa_t *spa) 1592fa9e4066Sahrens { 1593fa9e4066Sahrens return (spa->spa_first_txg); 1594fa9e4066Sahrens } 1595fa9e4066Sahrens 1596b24ab676SJeff Bonwick uint64_t 1597b24ab676SJeff Bonwick spa_syncing_txg(spa_t *spa) 1598b24ab676SJeff Bonwick { 1599b24ab676SJeff Bonwick return (spa->spa_syncing_txg); 1600b24ab676SJeff Bonwick } 1601b24ab676SJeff Bonwick 160288b7b0f2SMatthew Ahrens pool_state_t 1603fa9e4066Sahrens spa_state(spa_t *spa) 1604fa9e4066Sahrens { 1605fa9e4066Sahrens return (spa->spa_state); 1606fa9e4066Sahrens } 1607fa9e4066Sahrens 1608b16da2e2SGeorge Wilson spa_load_state_t 1609b16da2e2SGeorge Wilson spa_load_state(spa_t *spa) 1610b16da2e2SGeorge Wilson { 1611b16da2e2SGeorge Wilson return (spa->spa_load_state); 1612b16da2e2SGeorge Wilson } 1613b16da2e2SGeorge Wilson 1614fa9e4066Sahrens uint64_t 1615fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1616fa9e4066Sahrens { 1617fa9e4066Sahrens return (spa->spa_freeze_txg); 1618fa9e4066Sahrens } 1619fa9e4066Sahrens 1620fa9e4066Sahrens /* ARGSUSED */ 1621fa9e4066Sahrens uint64_t 1622fa9e4066Sahrens spa_get_asize(spa_t *spa, uint64_t lsize) 1623fa9e4066Sahrens { 162469962b56SMatthew Ahrens return (lsize * spa_asize_inflation); 162544cd46caSbillm } 162644cd46caSbillm 16277d46dc6cSMatthew Ahrens /* 16287d46dc6cSMatthew Ahrens * Return the amount of slop space in bytes. It is 1/32 of the pool (3.2%), 16297d46dc6cSMatthew Ahrens * or at least 32MB. 16307d46dc6cSMatthew Ahrens * 16317d46dc6cSMatthew Ahrens * See the comment above spa_slop_shift for details. 16327d46dc6cSMatthew Ahrens */ 16337d46dc6cSMatthew Ahrens uint64_t 16347d46dc6cSMatthew Ahrens spa_get_slop_space(spa_t *spa) { 16357d46dc6cSMatthew Ahrens uint64_t space = spa_get_dspace(spa); 16367d46dc6cSMatthew Ahrens return (MAX(space >> spa_slop_shift, SPA_MINDEVSIZE >> 1)); 16377d46dc6cSMatthew Ahrens } 16387d46dc6cSMatthew Ahrens 1639485bbbf5SGeorge Wilson uint64_t 1640485bbbf5SGeorge Wilson spa_get_dspace(spa_t *spa) 1641485bbbf5SGeorge Wilson { 1642485bbbf5SGeorge Wilson return (spa->spa_dspace); 1643485bbbf5SGeorge Wilson } 1644485bbbf5SGeorge Wilson 1645485bbbf5SGeorge Wilson void 1646485bbbf5SGeorge Wilson spa_update_dspace(spa_t *spa) 1647485bbbf5SGeorge Wilson { 1648485bbbf5SGeorge Wilson spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + 1649485bbbf5SGeorge Wilson ddt_get_dedup_dspace(spa); 1650485bbbf5SGeorge Wilson } 1651485bbbf5SGeorge Wilson 16520a4e9518Sgw /* 16530a4e9518Sgw * Return the failure mode that has been set to this pool. The default 16540a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 16550a4e9518Sgw */ 16560a4e9518Sgw uint8_t 16570a4e9518Sgw spa_get_failmode(spa_t *spa) 16580a4e9518Sgw { 16590a4e9518Sgw return (spa->spa_failmode); 16600a4e9518Sgw } 16610a4e9518Sgw 1662e14bb325SJeff Bonwick boolean_t 1663e14bb325SJeff Bonwick spa_suspended(spa_t *spa) 1664e14bb325SJeff Bonwick { 1665e14bb325SJeff Bonwick return (spa->spa_suspended); 1666e14bb325SJeff Bonwick } 1667e14bb325SJeff Bonwick 166844cd46caSbillm uint64_t 166944cd46caSbillm spa_version(spa_t *spa) 167044cd46caSbillm { 167144cd46caSbillm return (spa->spa_ubsync.ub_version); 167244cd46caSbillm } 167344cd46caSbillm 1674b24ab676SJeff Bonwick boolean_t 1675b24ab676SJeff Bonwick spa_deflate(spa_t *spa) 1676b24ab676SJeff Bonwick { 1677b24ab676SJeff Bonwick return (spa->spa_deflate); 1678b24ab676SJeff Bonwick } 1679b24ab676SJeff Bonwick 1680b24ab676SJeff Bonwick metaslab_class_t * 1681b24ab676SJeff Bonwick spa_normal_class(spa_t *spa) 1682b24ab676SJeff Bonwick { 1683b24ab676SJeff Bonwick return (spa->spa_normal_class); 1684b24ab676SJeff Bonwick } 1685b24ab676SJeff Bonwick 1686b24ab676SJeff Bonwick metaslab_class_t * 1687b24ab676SJeff Bonwick spa_log_class(spa_t *spa) 1688b24ab676SJeff Bonwick { 1689b24ab676SJeff Bonwick return (spa->spa_log_class); 1690b24ab676SJeff Bonwick } 1691b24ab676SJeff Bonwick 1692*bc9014e6SJustin Gibbs void 1693*bc9014e6SJustin Gibbs spa_evicting_os_register(spa_t *spa, objset_t *os) 1694*bc9014e6SJustin Gibbs { 1695*bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1696*bc9014e6SJustin Gibbs list_insert_head(&spa->spa_evicting_os_list, os); 1697*bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1698*bc9014e6SJustin Gibbs } 1699*bc9014e6SJustin Gibbs 1700*bc9014e6SJustin Gibbs void 1701*bc9014e6SJustin Gibbs spa_evicting_os_deregister(spa_t *spa, objset_t *os) 1702*bc9014e6SJustin Gibbs { 1703*bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1704*bc9014e6SJustin Gibbs list_remove(&spa->spa_evicting_os_list, os); 1705*bc9014e6SJustin Gibbs cv_broadcast(&spa->spa_evicting_os_cv); 1706*bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1707*bc9014e6SJustin Gibbs } 1708*bc9014e6SJustin Gibbs 1709*bc9014e6SJustin Gibbs void 1710*bc9014e6SJustin Gibbs spa_evicting_os_wait(spa_t *spa) 1711*bc9014e6SJustin Gibbs { 1712*bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1713*bc9014e6SJustin Gibbs while (!list_is_empty(&spa->spa_evicting_os_list)) 1714*bc9014e6SJustin Gibbs cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock); 1715*bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1716*bc9014e6SJustin Gibbs 1717*bc9014e6SJustin Gibbs dmu_buf_user_evict_wait(); 1718*bc9014e6SJustin Gibbs } 1719*bc9014e6SJustin Gibbs 172044cd46caSbillm int 172144cd46caSbillm spa_max_replication(spa_t *spa) 172244cd46caSbillm { 172344cd46caSbillm /* 1724e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 172544cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 172644cd46caSbillm * replication level accordingly. 1727fa9e4066Sahrens */ 1728e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 172944cd46caSbillm return (1); 173044cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1731fa9e4066Sahrens } 1732fa9e4066Sahrens 17333f9d6ad7SLin Ling int 17343f9d6ad7SLin Ling spa_prev_software_version(spa_t *spa) 17353f9d6ad7SLin Ling { 17363f9d6ad7SLin Ling return (spa->spa_prev_software_version); 17373f9d6ad7SLin Ling } 17383f9d6ad7SLin Ling 1739283b8460SGeorge.Wilson uint64_t 1740283b8460SGeorge.Wilson spa_deadman_synctime(spa_t *spa) 1741283b8460SGeorge.Wilson { 1742283b8460SGeorge.Wilson return (spa->spa_deadman_synctime); 1743283b8460SGeorge.Wilson } 1744283b8460SGeorge.Wilson 174599653d4eSeschrock uint64_t 1746b24ab676SJeff Bonwick dva_get_dsize_sync(spa_t *spa, const dva_t *dva) 174799653d4eSeschrock { 1748b24ab676SJeff Bonwick uint64_t asize = DVA_GET_ASIZE(dva); 1749b24ab676SJeff Bonwick uint64_t dsize = asize; 175099653d4eSeschrock 1751b24ab676SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); 175299653d4eSeschrock 1753b24ab676SJeff Bonwick if (asize != 0 && spa->spa_deflate) { 1754b24ab676SJeff Bonwick vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); 1755b24ab676SJeff Bonwick dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 175699653d4eSeschrock } 1757b24ab676SJeff Bonwick 1758b24ab676SJeff Bonwick return (dsize); 1759b24ab676SJeff Bonwick } 1760b24ab676SJeff Bonwick 1761b24ab676SJeff Bonwick uint64_t 1762b24ab676SJeff Bonwick bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) 1763b24ab676SJeff Bonwick { 1764b24ab676SJeff Bonwick uint64_t dsize = 0; 1765b24ab676SJeff Bonwick 17665d7b4d43SMatthew Ahrens for (int d = 0; d < BP_GET_NDVAS(bp); d++) 1767b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1768b24ab676SJeff Bonwick 1769b24ab676SJeff Bonwick return (dsize); 1770b24ab676SJeff Bonwick } 1771b24ab676SJeff Bonwick 1772b24ab676SJeff Bonwick uint64_t 1773b24ab676SJeff Bonwick bp_get_dsize(spa_t *spa, const blkptr_t *bp) 1774b24ab676SJeff Bonwick { 1775b24ab676SJeff Bonwick uint64_t dsize = 0; 1776b24ab676SJeff Bonwick 1777b24ab676SJeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 1778b24ab676SJeff Bonwick 17795d7b4d43SMatthew Ahrens for (int d = 0; d < BP_GET_NDVAS(bp); d++) 1780b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1781b24ab676SJeff Bonwick 1782e14bb325SJeff Bonwick spa_config_exit(spa, SCL_VDEV, FTAG); 1783b24ab676SJeff Bonwick 1784b24ab676SJeff Bonwick return (dsize); 178599653d4eSeschrock } 178699653d4eSeschrock 1787fa9e4066Sahrens /* 1788fa9e4066Sahrens * ========================================================================== 1789fa9e4066Sahrens * Initialization and Termination 1790fa9e4066Sahrens * ========================================================================== 1791fa9e4066Sahrens */ 1792fa9e4066Sahrens 1793fa9e4066Sahrens static int 1794fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1795fa9e4066Sahrens { 1796fa9e4066Sahrens const spa_t *s1 = a1; 1797fa9e4066Sahrens const spa_t *s2 = a2; 1798fa9e4066Sahrens int s; 1799fa9e4066Sahrens 1800fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1801fa9e4066Sahrens if (s > 0) 1802fa9e4066Sahrens return (1); 1803fa9e4066Sahrens if (s < 0) 1804fa9e4066Sahrens return (-1); 1805fa9e4066Sahrens return (0); 1806fa9e4066Sahrens } 1807fa9e4066Sahrens 18080373e76bSbonwick int 18090373e76bSbonwick spa_busy(void) 18100373e76bSbonwick { 18110373e76bSbonwick return (spa_active_count); 18120373e76bSbonwick } 18130373e76bSbonwick 1814e7cbe64fSgw void 1815e7cbe64fSgw spa_boot_init() 1816e7cbe64fSgw { 1817e7cbe64fSgw spa_config_load(); 1818e7cbe64fSgw } 1819e7cbe64fSgw 1820fa9e4066Sahrens void 1821fa9e4066Sahrens spa_init(int mode) 1822fa9e4066Sahrens { 1823fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1824c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1825fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 1826fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1827fa9e4066Sahrens 1828fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1829fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1830fa9e4066Sahrens 1831fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 1832fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 1833fa94a07fSbrendan 1834fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 1835fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 183699653d4eSeschrock 18378ad4d6ddSJeff Bonwick spa_mode_global = mode; 1838fa9e4066Sahrens 1839283b8460SGeorge.Wilson #ifdef _KERNEL 1840283b8460SGeorge.Wilson spa_arch_init(); 1841283b8460SGeorge.Wilson #else 1842cd1c8b85SMatthew Ahrens if (spa_mode_global != FREAD && dprintf_find_string("watch")) { 1843cd1c8b85SMatthew Ahrens arc_procfd = open("/proc/self/ctl", O_WRONLY); 1844cd1c8b85SMatthew Ahrens if (arc_procfd == -1) { 1845cd1c8b85SMatthew Ahrens perror("could not enable watchpoints: " 1846cd1c8b85SMatthew Ahrens "opening /proc/self/ctl failed: "); 1847cd1c8b85SMatthew Ahrens } else { 1848cd1c8b85SMatthew Ahrens arc_watch = B_TRUE; 1849cd1c8b85SMatthew Ahrens } 1850cd1c8b85SMatthew Ahrens } 1851cd1c8b85SMatthew Ahrens #endif 1852cd1c8b85SMatthew Ahrens 1853fa9e4066Sahrens refcount_init(); 1854fa9e4066Sahrens unique_init(); 18550713e232SGeorge Wilson range_tree_init(); 1856fa9e4066Sahrens zio_init(); 1857fa9e4066Sahrens dmu_init(); 1858fa9e4066Sahrens zil_init(); 185987db74c1Sek vdev_cache_stat_init(); 186091ebeef5Sahrens zfs_prop_init(); 1861990b4856Slling zpool_prop_init(); 1862ad135b5dSChristopher Siden zpool_feature_init(); 1863fa9e4066Sahrens spa_config_load(); 1864e14bb325SJeff Bonwick l2arc_start(); 1865fa9e4066Sahrens } 1866fa9e4066Sahrens 1867fa9e4066Sahrens void 1868fa9e4066Sahrens spa_fini(void) 1869fa9e4066Sahrens { 1870e14bb325SJeff Bonwick l2arc_stop(); 1871e14bb325SJeff Bonwick 1872fa9e4066Sahrens spa_evict_all(); 1873fa9e4066Sahrens 187487db74c1Sek vdev_cache_stat_fini(); 1875fa9e4066Sahrens zil_fini(); 1876fa9e4066Sahrens dmu_fini(); 1877fa9e4066Sahrens zio_fini(); 18780713e232SGeorge Wilson range_tree_fini(); 187991ebeef5Sahrens unique_fini(); 1880fa9e4066Sahrens refcount_fini(); 1881fa9e4066Sahrens 1882fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 188399653d4eSeschrock avl_destroy(&spa_spare_avl); 1884fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 1885fa9e4066Sahrens 1886fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1887fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1888c25056deSgw mutex_destroy(&spa_spare_lock); 1889fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 1890fa9e4066Sahrens } 18916ce0521aSperrin 18926ce0521aSperrin /* 18936ce0521aSperrin * Return whether this pool has slogs. No locking needed. 18946ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 18956ce0521aSperrin * performance and not correctness 18966ce0521aSperrin */ 18976ce0521aSperrin boolean_t 18986ce0521aSperrin spa_has_slogs(spa_t *spa) 18996ce0521aSperrin { 19006ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 19016ce0521aSperrin } 1902bf82a41bSeschrock 1903b24ab676SJeff Bonwick spa_log_state_t 1904b24ab676SJeff Bonwick spa_get_log_state(spa_t *spa) 1905b24ab676SJeff Bonwick { 1906b24ab676SJeff Bonwick return (spa->spa_log_state); 1907b24ab676SJeff Bonwick } 1908b24ab676SJeff Bonwick 1909b24ab676SJeff Bonwick void 1910b24ab676SJeff Bonwick spa_set_log_state(spa_t *spa, spa_log_state_t state) 1911b24ab676SJeff Bonwick { 1912b24ab676SJeff Bonwick spa->spa_log_state = state; 1913b24ab676SJeff Bonwick } 1914b24ab676SJeff Bonwick 1915bf82a41bSeschrock boolean_t 1916bf82a41bSeschrock spa_is_root(spa_t *spa) 1917bf82a41bSeschrock { 1918bf82a41bSeschrock return (spa->spa_is_root); 1919bf82a41bSeschrock } 19208ad4d6ddSJeff Bonwick 19218ad4d6ddSJeff Bonwick boolean_t 19228ad4d6ddSJeff Bonwick spa_writeable(spa_t *spa) 19238ad4d6ddSJeff Bonwick { 19248ad4d6ddSJeff Bonwick return (!!(spa->spa_mode & FWRITE)); 19258ad4d6ddSJeff Bonwick } 19268ad4d6ddSJeff Bonwick 192773527f44SAlex Reece /* 192873527f44SAlex Reece * Returns true if there is a pending sync task in any of the current 192973527f44SAlex Reece * syncing txg, the current quiescing txg, or the current open txg. 193073527f44SAlex Reece */ 193173527f44SAlex Reece boolean_t 193273527f44SAlex Reece spa_has_pending_synctask(spa_t *spa) 193373527f44SAlex Reece { 193473527f44SAlex Reece return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks)); 193573527f44SAlex Reece } 193673527f44SAlex Reece 19378ad4d6ddSJeff Bonwick int 19388ad4d6ddSJeff Bonwick spa_mode(spa_t *spa) 19398ad4d6ddSJeff Bonwick { 19408ad4d6ddSJeff Bonwick return (spa->spa_mode); 19418ad4d6ddSJeff Bonwick } 1942b24ab676SJeff Bonwick 1943b24ab676SJeff Bonwick uint64_t 1944b24ab676SJeff Bonwick spa_bootfs(spa_t *spa) 1945b24ab676SJeff Bonwick { 1946b24ab676SJeff Bonwick return (spa->spa_bootfs); 1947b24ab676SJeff Bonwick } 1948b24ab676SJeff Bonwick 1949b24ab676SJeff Bonwick uint64_t 1950b24ab676SJeff Bonwick spa_delegation(spa_t *spa) 1951b24ab676SJeff Bonwick { 1952b24ab676SJeff Bonwick return (spa->spa_delegation); 1953b24ab676SJeff Bonwick } 1954b24ab676SJeff Bonwick 1955b24ab676SJeff Bonwick objset_t * 1956b24ab676SJeff Bonwick spa_meta_objset(spa_t *spa) 1957b24ab676SJeff Bonwick { 1958b24ab676SJeff Bonwick return (spa->spa_meta_objset); 1959b24ab676SJeff Bonwick } 1960b24ab676SJeff Bonwick 1961b24ab676SJeff Bonwick enum zio_checksum 1962b24ab676SJeff Bonwick spa_dedup_checksum(spa_t *spa) 1963b24ab676SJeff Bonwick { 1964b24ab676SJeff Bonwick return (spa->spa_dedup_checksum); 1965b24ab676SJeff Bonwick } 19663f9d6ad7SLin Ling 19673f9d6ad7SLin Ling /* 19683f9d6ad7SLin Ling * Reset pool scan stat per scan pass (or reboot). 19693f9d6ad7SLin Ling */ 19703f9d6ad7SLin Ling void 19713f9d6ad7SLin Ling spa_scan_stat_init(spa_t *spa) 19723f9d6ad7SLin Ling { 19733f9d6ad7SLin Ling /* data not stored on disk */ 19743f9d6ad7SLin Ling spa->spa_scan_pass_start = gethrestime_sec(); 19753f9d6ad7SLin Ling spa->spa_scan_pass_exam = 0; 19763f9d6ad7SLin Ling vdev_scan_stat_init(spa->spa_root_vdev); 19773f9d6ad7SLin Ling } 19783f9d6ad7SLin Ling 19793f9d6ad7SLin Ling /* 19803f9d6ad7SLin Ling * Get scan stats for zpool status reports 19813f9d6ad7SLin Ling */ 19823f9d6ad7SLin Ling int 19833f9d6ad7SLin Ling spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) 19843f9d6ad7SLin Ling { 19853f9d6ad7SLin Ling dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; 19863f9d6ad7SLin Ling 19873f9d6ad7SLin Ling if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) 1988be6fd75aSMatthew Ahrens return (SET_ERROR(ENOENT)); 19893f9d6ad7SLin Ling bzero(ps, sizeof (pool_scan_stat_t)); 19903f9d6ad7SLin Ling 19913f9d6ad7SLin Ling /* data stored on disk */ 19923f9d6ad7SLin Ling ps->pss_func = scn->scn_phys.scn_func; 19933f9d6ad7SLin Ling ps->pss_start_time = scn->scn_phys.scn_start_time; 19943f9d6ad7SLin Ling ps->pss_end_time = scn->scn_phys.scn_end_time; 19953f9d6ad7SLin Ling ps->pss_to_examine = scn->scn_phys.scn_to_examine; 19963f9d6ad7SLin Ling ps->pss_examined = scn->scn_phys.scn_examined; 19973f9d6ad7SLin Ling ps->pss_to_process = scn->scn_phys.scn_to_process; 19983f9d6ad7SLin Ling ps->pss_processed = scn->scn_phys.scn_processed; 19993f9d6ad7SLin Ling ps->pss_errors = scn->scn_phys.scn_errors; 20003f9d6ad7SLin Ling ps->pss_state = scn->scn_phys.scn_state; 20013f9d6ad7SLin Ling 20023f9d6ad7SLin Ling /* data not stored on disk */ 20033f9d6ad7SLin Ling ps->pss_pass_start = spa->spa_scan_pass_start; 20043f9d6ad7SLin Ling ps->pss_pass_exam = spa->spa_scan_pass_exam; 20053f9d6ad7SLin Ling 20063f9d6ad7SLin Ling return (0); 20073f9d6ad7SLin Ling } 200809c9d376SGeorge Wilson 200909c9d376SGeorge Wilson boolean_t 201009c9d376SGeorge Wilson spa_debug_enabled(spa_t *spa) 201109c9d376SGeorge Wilson { 201209c9d376SGeorge Wilson return (spa->spa_debug); 201309c9d376SGeorge Wilson } 2014b5152584SMatthew Ahrens 2015b5152584SMatthew Ahrens int 2016b5152584SMatthew Ahrens spa_maxblocksize(spa_t *spa) 2017b5152584SMatthew Ahrens { 2018b5152584SMatthew Ahrens if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) 2019b5152584SMatthew Ahrens return (SPA_MAXBLOCKSIZE); 2020b5152584SMatthew Ahrens else 2021b5152584SMatthew Ahrens return (SPA_OLD_MAXBLOCKSIZE); 2022b5152584SMatthew Ahrens } 2023