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. 2394c2d0ebSMatthew Ahrens * Copyright (c) 2011, 2017 by Delphix. All rights reserved. 24e495b6e6SSaso Kiselkov * Copyright 2015 Nexenta Systems, Inc. All rights reserved. 25bc9014e6SJustin Gibbs * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 2645818ee1SMatthew Ahrens * Copyright 2013 Saso Kiselkov. All rights reserved. 27c3d26abcSMatthew Ahrens * Copyright (c) 2014 Integros [integros.com] 28*1702cce7SAlek Pinchuk * Copyright (c) 2017 Datto Inc. 29fa9e4066Sahrens */ 30fa9e4066Sahrens 31fa9e4066Sahrens #include <sys/zfs_context.h> 32fa9e4066Sahrens #include <sys/spa_impl.h> 33283b8460SGeorge.Wilson #include <sys/spa_boot.h> 34fa9e4066Sahrens #include <sys/zio.h> 35fa9e4066Sahrens #include <sys/zio_checksum.h> 36fa9e4066Sahrens #include <sys/zio_compress.h> 37fa9e4066Sahrens #include <sys/dmu.h> 38fa9e4066Sahrens #include <sys/dmu_tx.h> 39fa9e4066Sahrens #include <sys/zap.h> 40fa9e4066Sahrens #include <sys/zil.h> 41fa9e4066Sahrens #include <sys/vdev_impl.h> 42fa9e4066Sahrens #include <sys/metaslab.h> 43fa9e4066Sahrens #include <sys/uberblock_impl.h> 44fa9e4066Sahrens #include <sys/txg.h> 45fa9e4066Sahrens #include <sys/avl.h> 46fa9e4066Sahrens #include <sys/unique.h> 47fa9e4066Sahrens #include <sys/dsl_pool.h> 48fa9e4066Sahrens #include <sys/dsl_dir.h> 49fa9e4066Sahrens #include <sys/dsl_prop.h> 503f9d6ad7SLin Ling #include <sys/dsl_scan.h> 51fa9e4066Sahrens #include <sys/fs/zfs.h> 526ce0521aSperrin #include <sys/metaslab_impl.h> 53e14bb325SJeff Bonwick #include <sys/arc.h> 54485bbbf5SGeorge Wilson #include <sys/ddt.h> 5591ebeef5Sahrens #include "zfs_prop.h" 5645818ee1SMatthew Ahrens #include <sys/zfeature.h> 57fa9e4066Sahrens 58fa9e4066Sahrens /* 59fa9e4066Sahrens * SPA locking 60fa9e4066Sahrens * 61fa9e4066Sahrens * There are four basic locks for managing spa_t structures: 62fa9e4066Sahrens * 63fa9e4066Sahrens * spa_namespace_lock (global mutex) 64fa9e4066Sahrens * 6544cd46caSbillm * This lock must be acquired to do any of the following: 66fa9e4066Sahrens * 6744cd46caSbillm * - Lookup a spa_t by name 6844cd46caSbillm * - Add or remove a spa_t from the namespace 6944cd46caSbillm * - Increase spa_refcount from non-zero 7044cd46caSbillm * - Check if spa_refcount is zero 7144cd46caSbillm * - Rename a spa_t 72ea8dc4b6Seschrock * - add/remove/attach/detach devices 7344cd46caSbillm * - Held for the duration of create/destroy/import/export 74fa9e4066Sahrens * 7544cd46caSbillm * It does not need to handle recursion. A create or destroy may 7644cd46caSbillm * reference objects (files or zvols) in other pools, but by 7744cd46caSbillm * definition they must have an existing reference, and will never need 7844cd46caSbillm * to lookup a spa_t by name. 79fa9e4066Sahrens * 80fa9e4066Sahrens * spa_refcount (per-spa refcount_t protected by mutex) 81fa9e4066Sahrens * 8244cd46caSbillm * This reference count keep track of any active users of the spa_t. The 8344cd46caSbillm * spa_t cannot be destroyed or freed while this is non-zero. Internally, 8444cd46caSbillm * the refcount is never really 'zero' - opening a pool implicitly keeps 85088f3894Sahrens * some references in the DMU. Internally we check against spa_minref, but 8644cd46caSbillm * present the image of a zero/non-zero value to consumers. 87fa9e4066Sahrens * 88e14bb325SJeff Bonwick * spa_config_lock[] (per-spa array of rwlocks) 89fa9e4066Sahrens * 9091ebeef5Sahrens * This protects the spa_t from config changes, and must be held in 9191ebeef5Sahrens * the following circumstances: 92fa9e4066Sahrens * 9344cd46caSbillm * - RW_READER to perform I/O to the spa 9444cd46caSbillm * - RW_WRITER to change the vdev config 95fa9e4066Sahrens * 96fa9e4066Sahrens * The locking order is fairly straightforward: 97fa9e4066Sahrens * 9844cd46caSbillm * spa_namespace_lock -> spa_refcount 99fa9e4066Sahrens * 10044cd46caSbillm * The namespace lock must be acquired to increase the refcount from 0 10144cd46caSbillm * or to check if it is zero. 102fa9e4066Sahrens * 103e14bb325SJeff Bonwick * spa_refcount -> spa_config_lock[] 104fa9e4066Sahrens * 10544cd46caSbillm * There must be at least one valid reference on the spa_t to acquire 10644cd46caSbillm * the config lock. 107fa9e4066Sahrens * 108e14bb325SJeff Bonwick * spa_namespace_lock -> spa_config_lock[] 109fa9e4066Sahrens * 11044cd46caSbillm * The namespace lock must always be taken before the config lock. 111fa9e4066Sahrens * 112fa9e4066Sahrens * 113e14bb325SJeff Bonwick * The spa_namespace_lock can be acquired directly and is globally visible. 114fa9e4066Sahrens * 115e14bb325SJeff Bonwick * The namespace is manipulated using the following functions, all of which 116e14bb325SJeff Bonwick * require the spa_namespace_lock to be held. 117fa9e4066Sahrens * 11844cd46caSbillm * spa_lookup() Lookup a spa_t by name. 119fa9e4066Sahrens * 12044cd46caSbillm * spa_add() Create a new spa_t in the namespace. 121fa9e4066Sahrens * 12244cd46caSbillm * spa_remove() Remove a spa_t from the namespace. This also 12344cd46caSbillm * frees up any memory associated with the spa_t. 124fa9e4066Sahrens * 12544cd46caSbillm * spa_next() Returns the next spa_t in the system, or the 12644cd46caSbillm * first if NULL is passed. 127fa9e4066Sahrens * 12844cd46caSbillm * spa_evict_all() Shutdown and remove all spa_t structures in 12944cd46caSbillm * the system. 130fa9e4066Sahrens * 131ea8dc4b6Seschrock * spa_guid_exists() Determine whether a pool/device guid exists. 132fa9e4066Sahrens * 133fa9e4066Sahrens * The spa_refcount is manipulated using the following functions: 134fa9e4066Sahrens * 13544cd46caSbillm * spa_open_ref() Adds a reference to the given spa_t. Must be 13644cd46caSbillm * called with spa_namespace_lock held if the 13744cd46caSbillm * refcount is currently zero. 138fa9e4066Sahrens * 13944cd46caSbillm * spa_close() Remove a reference from the spa_t. This will 14044cd46caSbillm * not free the spa_t or remove it from the 14144cd46caSbillm * namespace. No locking is required. 142fa9e4066Sahrens * 14344cd46caSbillm * spa_refcount_zero() Returns true if the refcount is currently 14444cd46caSbillm * zero. Must be called with spa_namespace_lock 14544cd46caSbillm * held. 146fa9e4066Sahrens * 147e14bb325SJeff Bonwick * The spa_config_lock[] is an array of rwlocks, ordered as follows: 148e14bb325SJeff Bonwick * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. 149e14bb325SJeff Bonwick * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). 150e14bb325SJeff Bonwick * 151e14bb325SJeff Bonwick * To read the configuration, it suffices to hold one of these locks as reader. 152e14bb325SJeff Bonwick * To modify the configuration, you must hold all locks as writer. To modify 153e14bb325SJeff Bonwick * vdev state without altering the vdev tree's topology (e.g. online/offline), 154e14bb325SJeff Bonwick * you must hold SCL_STATE and SCL_ZIO as writer. 155e14bb325SJeff Bonwick * 156e14bb325SJeff Bonwick * We use these distinct config locks to avoid recursive lock entry. 157e14bb325SJeff Bonwick * For example, spa_sync() (which holds SCL_CONFIG as reader) induces 158e14bb325SJeff Bonwick * block allocations (SCL_ALLOC), which may require reading space maps 159e14bb325SJeff Bonwick * from disk (dmu_read() -> zio_read() -> SCL_ZIO). 160e14bb325SJeff Bonwick * 161e14bb325SJeff Bonwick * The spa config locks cannot be normal rwlocks because we need the 162e14bb325SJeff Bonwick * ability to hand off ownership. For example, SCL_ZIO is acquired 163e14bb325SJeff Bonwick * by the issuing thread and later released by an interrupt thread. 164e14bb325SJeff Bonwick * They do, however, obey the usual write-wanted semantics to prevent 165e14bb325SJeff Bonwick * writer (i.e. system administrator) starvation. 166e14bb325SJeff Bonwick * 167e14bb325SJeff Bonwick * The lock acquisition rules are as follows: 168e14bb325SJeff Bonwick * 169e14bb325SJeff Bonwick * SCL_CONFIG 170e14bb325SJeff Bonwick * Protects changes to the vdev tree topology, such as vdev 171e14bb325SJeff Bonwick * add/remove/attach/detach. Protects the dirty config list 172e14bb325SJeff Bonwick * (spa_config_dirty_list) and the set of spares and l2arc devices. 173e14bb325SJeff Bonwick * 174e14bb325SJeff Bonwick * SCL_STATE 175e14bb325SJeff Bonwick * Protects changes to pool state and vdev state, such as vdev 176e14bb325SJeff Bonwick * online/offline/fault/degrade/clear. Protects the dirty state list 177e14bb325SJeff Bonwick * (spa_state_dirty_list) and global pool state (spa_state). 178e14bb325SJeff Bonwick * 179e14bb325SJeff Bonwick * SCL_ALLOC 180e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 181e14bb325SJeff Bonwick * Held as reader by metaslab_alloc() and metaslab_claim(). 182e14bb325SJeff Bonwick * 183e14bb325SJeff Bonwick * SCL_ZIO 184e14bb325SJeff Bonwick * Held by bp-level zios (those which have no io_vd upon entry) 185e14bb325SJeff Bonwick * to prevent changes to the vdev tree. The bp-level zio implicitly 186e14bb325SJeff Bonwick * protects all of its vdev child zios, which do not hold SCL_ZIO. 187e14bb325SJeff Bonwick * 188e14bb325SJeff Bonwick * SCL_FREE 189e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 190e14bb325SJeff Bonwick * Held as reader by metaslab_free(). SCL_FREE is distinct from 191e14bb325SJeff Bonwick * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free 192e14bb325SJeff Bonwick * blocks in zio_done() while another i/o that holds either 193e14bb325SJeff Bonwick * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. 194e14bb325SJeff Bonwick * 195e14bb325SJeff Bonwick * SCL_VDEV 196e14bb325SJeff Bonwick * Held as reader to prevent changes to the vdev tree during trivial 197b24ab676SJeff Bonwick * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the 198e14bb325SJeff Bonwick * other locks, and lower than all of them, to ensure that it's safe 199e14bb325SJeff Bonwick * to acquire regardless of caller context. 200e14bb325SJeff Bonwick * 201e14bb325SJeff Bonwick * In addition, the following rules apply: 202e14bb325SJeff Bonwick * 203e14bb325SJeff Bonwick * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. 204e14bb325SJeff Bonwick * The lock ordering is SCL_CONFIG > spa_props_lock. 205e14bb325SJeff Bonwick * 206e14bb325SJeff Bonwick * (b) I/O operations on leaf vdevs. For any zio operation that takes 207e14bb325SJeff Bonwick * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), 208e14bb325SJeff Bonwick * or zio_write_phys() -- the caller must ensure that the config cannot 209e14bb325SJeff Bonwick * cannot change in the interim, and that the vdev cannot be reopened. 210e14bb325SJeff Bonwick * SCL_STATE as reader suffices for both. 211fa9e4066Sahrens * 212ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 213fa9e4066Sahrens * 21444cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 215ea8dc4b6Seschrock * for writing. 216fa9e4066Sahrens * 21744cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 21844cd46caSbillm * to complete, sync the updated configs to the 219ea8dc4b6Seschrock * cache, and release the namespace lock. 220fa9e4066Sahrens * 221e14bb325SJeff Bonwick * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). 222e14bb325SJeff Bonwick * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual 223e14bb325SJeff Bonwick * locking is, always, based on spa_namespace_lock and spa_config_lock[]. 224e14bb325SJeff Bonwick * 225ad135b5dSChristopher Siden * spa_rename() is also implemented within this file since it requires 226e14bb325SJeff Bonwick * manipulation of the namespace. 227fa9e4066Sahrens */ 228fa9e4066Sahrens 229fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 230fa9e4066Sahrens kmutex_t spa_namespace_lock; 231fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 2320373e76bSbonwick static int spa_active_count; 233416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 234fa9e4066Sahrens 23599653d4eSeschrock static kmutex_t spa_spare_lock; 23639c23413Seschrock static avl_tree_t spa_spare_avl; 237fa94a07fSbrendan static kmutex_t spa_l2cache_lock; 238fa94a07fSbrendan static avl_tree_t spa_l2cache_avl; 23999653d4eSeschrock 240fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 2418ad4d6ddSJeff Bonwick int spa_mode_global; 242fa9e4066Sahrens 243fa9e4066Sahrens #ifdef ZFS_DEBUG 2443b2aab18SMatthew Ahrens /* Everything except dprintf and spa is on by default in debug builds */ 2453b2aab18SMatthew Ahrens int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_SPA); 246fa9e4066Sahrens #else 247fa9e4066Sahrens int zfs_flags = 0; 248fa9e4066Sahrens #endif 249fa9e4066Sahrens 2500125049cSahrens /* 2510125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 2520125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 2530125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 2548b36997aSMatthew Ahrens * This should only be used as a last resort, as it typically results 2558b36997aSMatthew Ahrens * in leaked space, or worse. 2560125049cSahrens */ 2577fd05ac4SMatthew Ahrens boolean_t zfs_recover = B_FALSE; 2587fd05ac4SMatthew Ahrens 2597fd05ac4SMatthew Ahrens /* 2607fd05ac4SMatthew Ahrens * If destroy encounters an EIO while reading metadata (e.g. indirect 2617fd05ac4SMatthew Ahrens * blocks), space referenced by the missing metadata can not be freed. 2627fd05ac4SMatthew Ahrens * Normally this causes the background destroy to become "stalled", as 2637fd05ac4SMatthew Ahrens * it is unable to make forward progress. While in this stalled state, 2647fd05ac4SMatthew Ahrens * all remaining space to free from the error-encountering filesystem is 2657fd05ac4SMatthew Ahrens * "temporarily leaked". Set this flag to cause it to ignore the EIO, 2667fd05ac4SMatthew Ahrens * permanently leak the space from indirect blocks that can not be read, 2677fd05ac4SMatthew Ahrens * and continue to free everything else that it can. 2687fd05ac4SMatthew Ahrens * 2697fd05ac4SMatthew Ahrens * The default, "stalling" behavior is useful if the storage partially 2707fd05ac4SMatthew Ahrens * fails (i.e. some but not all i/os fail), and then later recovers. In 2717fd05ac4SMatthew Ahrens * this case, we will be able to continue pool operations while it is 2727fd05ac4SMatthew Ahrens * partially failed, and when it recovers, we can continue to free the 2737fd05ac4SMatthew Ahrens * space, with no leaks. However, note that this case is actually 2747fd05ac4SMatthew Ahrens * fairly rare. 2757fd05ac4SMatthew Ahrens * 2767fd05ac4SMatthew Ahrens * Typically pools either (a) fail completely (but perhaps temporarily, 2777fd05ac4SMatthew Ahrens * e.g. a top-level vdev going offline), or (b) have localized, 2787fd05ac4SMatthew Ahrens * permanent errors (e.g. disk returns the wrong data due to bit flip or 2797fd05ac4SMatthew Ahrens * firmware bug). In case (a), this setting does not matter because the 2807fd05ac4SMatthew Ahrens * pool will be suspended and the sync thread will not be able to make 2817fd05ac4SMatthew Ahrens * forward progress regardless. In case (b), because the error is 2827fd05ac4SMatthew Ahrens * permanent, the best we can do is leak the minimum amount of space, 2837fd05ac4SMatthew Ahrens * which is what setting this flag will do. Therefore, it is reasonable 2847fd05ac4SMatthew Ahrens * for this flag to normally be set, but we chose the more conservative 2857fd05ac4SMatthew Ahrens * approach of not setting it, so that there is no possibility of 2867fd05ac4SMatthew Ahrens * leaking space in the "partial temporary" failure case. 2877fd05ac4SMatthew Ahrens */ 2887fd05ac4SMatthew Ahrens boolean_t zfs_free_leak_on_eio = B_FALSE; 2890125049cSahrens 29069962b56SMatthew Ahrens /* 29169962b56SMatthew Ahrens * Expiration time in milliseconds. This value has two meanings. First it is 29269962b56SMatthew Ahrens * used to determine when the spa_deadman() logic should fire. By default the 29369962b56SMatthew Ahrens * spa_deadman() will fire if spa_sync() has not completed in 1000 seconds. 29469962b56SMatthew Ahrens * Secondly, the value determines if an I/O is considered "hung". Any I/O that 29569962b56SMatthew Ahrens * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting 29669962b56SMatthew Ahrens * in a system panic. 29769962b56SMatthew Ahrens */ 29869962b56SMatthew Ahrens uint64_t zfs_deadman_synctime_ms = 1000000ULL; 299283b8460SGeorge.Wilson 300283b8460SGeorge.Wilson /* 30169962b56SMatthew Ahrens * Check time in milliseconds. This defines the frequency at which we check 30269962b56SMatthew Ahrens * for hung I/O. 303283b8460SGeorge.Wilson */ 30469962b56SMatthew Ahrens uint64_t zfs_deadman_checktime_ms = 5000ULL; 305283b8460SGeorge.Wilson 306283b8460SGeorge.Wilson /* 307283b8460SGeorge.Wilson * Override the zfs deadman behavior via /etc/system. By default the 308283b8460SGeorge.Wilson * deadman is enabled except on VMware and sparc deployments. 309283b8460SGeorge.Wilson */ 310283b8460SGeorge.Wilson int zfs_deadman_enabled = -1; 311283b8460SGeorge.Wilson 31269962b56SMatthew Ahrens /* 31369962b56SMatthew Ahrens * The worst case is single-sector max-parity RAID-Z blocks, in which 31469962b56SMatthew Ahrens * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) 31569962b56SMatthew Ahrens * times the size; so just assume that. Add to this the fact that 31669962b56SMatthew Ahrens * we can have up to 3 DVAs per bp, and one more factor of 2 because 31769962b56SMatthew Ahrens * the block may be dittoed with up to 3 DVAs by ddt_sync(). All together, 31869962b56SMatthew Ahrens * the worst case is: 31969962b56SMatthew Ahrens * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24 32069962b56SMatthew Ahrens */ 32169962b56SMatthew Ahrens int spa_asize_inflation = 24; 322fa9e4066Sahrens 3237d46dc6cSMatthew Ahrens /* 3247d46dc6cSMatthew Ahrens * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in 3257d46dc6cSMatthew Ahrens * the pool to be consumed. This ensures that we don't run the pool 3267d46dc6cSMatthew Ahrens * completely out of space, due to unaccounted changes (e.g. to the MOS). 3277d46dc6cSMatthew Ahrens * It also limits the worst-case time to allocate space. If we have 3287d46dc6cSMatthew Ahrens * less than this amount of free space, most ZPL operations (e.g. write, 3297d46dc6cSMatthew Ahrens * create) will return ENOSPC. 3307d46dc6cSMatthew Ahrens * 3317d46dc6cSMatthew Ahrens * Certain operations (e.g. file removal, most administrative actions) can 3327d46dc6cSMatthew Ahrens * use half the slop space. They will only return ENOSPC if less than half 3337d46dc6cSMatthew Ahrens * the slop space is free. Typically, once the pool has less than the slop 3347d46dc6cSMatthew Ahrens * space free, the user will use these operations to free up space in the pool. 3357d46dc6cSMatthew Ahrens * These are the operations that call dsl_pool_adjustedsize() with the netfree 3367d46dc6cSMatthew Ahrens * argument set to TRUE. 3377d46dc6cSMatthew Ahrens * 3387d46dc6cSMatthew Ahrens * A very restricted set of operations are always permitted, regardless of 3397d46dc6cSMatthew Ahrens * the amount of free space. These are the operations that call 3407d46dc6cSMatthew Ahrens * dsl_sync_task(ZFS_SPACE_CHECK_NONE), e.g. "zfs destroy". If these 3417d46dc6cSMatthew Ahrens * operations result in a net increase in the amount of space used, 3427d46dc6cSMatthew Ahrens * it is possible to run the pool completely out of space, causing it to 3437d46dc6cSMatthew Ahrens * be permanently read-only. 3447d46dc6cSMatthew Ahrens * 3454b5c8e93SMatthew Ahrens * Note that on very small pools, the slop space will be larger than 3464b5c8e93SMatthew Ahrens * 3.2%, in an effort to have it be at least spa_min_slop (128MB), 3474b5c8e93SMatthew Ahrens * but we never allow it to be more than half the pool size. 3484b5c8e93SMatthew Ahrens * 3497d46dc6cSMatthew Ahrens * See also the comments in zfs_space_check_t. 3507d46dc6cSMatthew Ahrens */ 3517d46dc6cSMatthew Ahrens int spa_slop_shift = 5; 3524b5c8e93SMatthew Ahrens uint64_t spa_min_slop = 128 * 1024 * 1024; 3537d46dc6cSMatthew Ahrens 354e05725b1Sbonwick /* 355e05725b1Sbonwick * ========================================================================== 356e05725b1Sbonwick * SPA config locking 357e05725b1Sbonwick * ========================================================================== 358e05725b1Sbonwick */ 359e05725b1Sbonwick static void 360e14bb325SJeff Bonwick spa_config_lock_init(spa_t *spa) 361e14bb325SJeff Bonwick { 362e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 363e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 364e14bb325SJeff Bonwick mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); 365e14bb325SJeff Bonwick cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); 3663b2aab18SMatthew Ahrens refcount_create_untracked(&scl->scl_count); 367e14bb325SJeff Bonwick scl->scl_writer = NULL; 368e14bb325SJeff Bonwick scl->scl_write_wanted = 0; 369e14bb325SJeff Bonwick } 370e05725b1Sbonwick } 371e05725b1Sbonwick 372e05725b1Sbonwick static void 373e14bb325SJeff Bonwick spa_config_lock_destroy(spa_t *spa) 374e14bb325SJeff Bonwick { 375e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 376e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 377e14bb325SJeff Bonwick mutex_destroy(&scl->scl_lock); 378e14bb325SJeff Bonwick cv_destroy(&scl->scl_cv); 379e14bb325SJeff Bonwick refcount_destroy(&scl->scl_count); 380e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL); 381e14bb325SJeff Bonwick ASSERT(scl->scl_write_wanted == 0); 382e14bb325SJeff Bonwick } 383e14bb325SJeff Bonwick } 384e14bb325SJeff Bonwick 385e14bb325SJeff Bonwick int 386e14bb325SJeff Bonwick spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) 387e05725b1Sbonwick { 388e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 389e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 390e14bb325SJeff Bonwick if (!(locks & (1 << i))) 391e14bb325SJeff Bonwick continue; 392e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 393e14bb325SJeff Bonwick if (rw == RW_READER) { 394e14bb325SJeff Bonwick if (scl->scl_writer || scl->scl_write_wanted) { 395e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 396e495b6e6SSaso Kiselkov spa_config_exit(spa, locks & ((1 << i) - 1), 397e495b6e6SSaso Kiselkov tag); 398e14bb325SJeff Bonwick return (0); 399e14bb325SJeff Bonwick } 400e14bb325SJeff Bonwick } else { 401e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 402e14bb325SJeff Bonwick if (!refcount_is_zero(&scl->scl_count)) { 403e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 404e495b6e6SSaso Kiselkov spa_config_exit(spa, locks & ((1 << i) - 1), 405e495b6e6SSaso Kiselkov tag); 406e14bb325SJeff Bonwick return (0); 407e14bb325SJeff Bonwick } 408e14bb325SJeff Bonwick scl->scl_writer = curthread; 409e14bb325SJeff Bonwick } 410e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 411e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 412e14bb325SJeff Bonwick } 413e14bb325SJeff Bonwick return (1); 414e05725b1Sbonwick } 415e05725b1Sbonwick 416e05725b1Sbonwick void 417e14bb325SJeff Bonwick spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) 418e05725b1Sbonwick { 419f64c0e34SEric Taylor int wlocks_held = 0; 420f64c0e34SEric Taylor 4213b2aab18SMatthew Ahrens ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY); 4223b2aab18SMatthew Ahrens 423e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 424e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 425f64c0e34SEric Taylor if (scl->scl_writer == curthread) 426f64c0e34SEric Taylor wlocks_held |= (1 << i); 427e14bb325SJeff Bonwick if (!(locks & (1 << i))) 428e14bb325SJeff Bonwick continue; 429e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 430e14bb325SJeff Bonwick if (rw == RW_READER) { 431e14bb325SJeff Bonwick while (scl->scl_writer || scl->scl_write_wanted) { 432e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 433e14bb325SJeff Bonwick } 434e14bb325SJeff Bonwick } else { 435e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 436e14bb325SJeff Bonwick while (!refcount_is_zero(&scl->scl_count)) { 437e14bb325SJeff Bonwick scl->scl_write_wanted++; 438e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 439e14bb325SJeff Bonwick scl->scl_write_wanted--; 440e14bb325SJeff Bonwick } 441e14bb325SJeff Bonwick scl->scl_writer = curthread; 442e14bb325SJeff Bonwick } 443e14bb325SJeff Bonwick (void) refcount_add(&scl->scl_count, tag); 444e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 445e05725b1Sbonwick } 446f64c0e34SEric Taylor ASSERT(wlocks_held <= locks); 447e05725b1Sbonwick } 448e05725b1Sbonwick 449e05725b1Sbonwick void 450e14bb325SJeff Bonwick spa_config_exit(spa_t *spa, int locks, void *tag) 451e05725b1Sbonwick { 452e14bb325SJeff Bonwick for (int i = SCL_LOCKS - 1; i >= 0; i--) { 453e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 454e14bb325SJeff Bonwick if (!(locks & (1 << i))) 455e14bb325SJeff Bonwick continue; 456e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 457e14bb325SJeff Bonwick ASSERT(!refcount_is_zero(&scl->scl_count)); 458e14bb325SJeff Bonwick if (refcount_remove(&scl->scl_count, tag) == 0) { 459e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL || 460e14bb325SJeff Bonwick scl->scl_writer == curthread); 461e14bb325SJeff Bonwick scl->scl_writer = NULL; /* OK in either case */ 462e14bb325SJeff Bonwick cv_broadcast(&scl->scl_cv); 463e14bb325SJeff Bonwick } 464e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 465e05725b1Sbonwick } 466e05725b1Sbonwick } 467e05725b1Sbonwick 468e14bb325SJeff Bonwick int 469e14bb325SJeff Bonwick spa_config_held(spa_t *spa, int locks, krw_t rw) 470e05725b1Sbonwick { 471e14bb325SJeff Bonwick int locks_held = 0; 472e05725b1Sbonwick 473e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 474e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 475e14bb325SJeff Bonwick if (!(locks & (1 << i))) 476e14bb325SJeff Bonwick continue; 477e14bb325SJeff Bonwick if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) || 478e14bb325SJeff Bonwick (rw == RW_WRITER && scl->scl_writer == curthread)) 479e14bb325SJeff Bonwick locks_held |= 1 << i; 480e14bb325SJeff Bonwick } 481e14bb325SJeff Bonwick 482e14bb325SJeff Bonwick return (locks_held); 483e05725b1Sbonwick } 484e05725b1Sbonwick 485fa9e4066Sahrens /* 486fa9e4066Sahrens * ========================================================================== 487fa9e4066Sahrens * SPA namespace functions 488fa9e4066Sahrens * ========================================================================== 489fa9e4066Sahrens */ 490fa9e4066Sahrens 491fa9e4066Sahrens /* 492fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 493fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 494fa9e4066Sahrens */ 495fa9e4066Sahrens spa_t * 496fa9e4066Sahrens spa_lookup(const char *name) 497fa9e4066Sahrens { 498e14bb325SJeff Bonwick static spa_t search; /* spa_t is large; don't allocate on stack */ 499e14bb325SJeff Bonwick spa_t *spa; 500fa9e4066Sahrens avl_index_t where; 50140feaa91Sahrens char *cp; 502fa9e4066Sahrens 503fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 504fa9e4066Sahrens 5053b2aab18SMatthew Ahrens (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); 5063b2aab18SMatthew Ahrens 50740feaa91Sahrens /* 50840feaa91Sahrens * If it's a full dataset name, figure out the pool name and 50940feaa91Sahrens * just use that. 51040feaa91Sahrens */ 51178f17100SMatthew Ahrens cp = strpbrk(search.spa_name, "/@#"); 5123b2aab18SMatthew Ahrens if (cp != NULL) 51340feaa91Sahrens *cp = '\0'; 51440feaa91Sahrens 515fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 516fa9e4066Sahrens 517fa9e4066Sahrens return (spa); 518fa9e4066Sahrens } 519fa9e4066Sahrens 520283b8460SGeorge.Wilson /* 521283b8460SGeorge.Wilson * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. 522283b8460SGeorge.Wilson * If the zfs_deadman_enabled flag is set then it inspects all vdev queues 523283b8460SGeorge.Wilson * looking for potentially hung I/Os. 524283b8460SGeorge.Wilson */ 525283b8460SGeorge.Wilson void 526283b8460SGeorge.Wilson spa_deadman(void *arg) 527283b8460SGeorge.Wilson { 528283b8460SGeorge.Wilson spa_t *spa = arg; 529283b8460SGeorge.Wilson 5300713e232SGeorge Wilson /* 5310713e232SGeorge Wilson * Disable the deadman timer if the pool is suspended. 5320713e232SGeorge Wilson */ 5330713e232SGeorge Wilson if (spa_suspended(spa)) { 5340713e232SGeorge Wilson VERIFY(cyclic_reprogram(spa->spa_deadman_cycid, CY_INFINITY)); 5350713e232SGeorge Wilson return; 5360713e232SGeorge Wilson } 5370713e232SGeorge Wilson 538283b8460SGeorge.Wilson zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", 539283b8460SGeorge.Wilson (gethrtime() - spa->spa_sync_starttime) / NANOSEC, 540283b8460SGeorge.Wilson ++spa->spa_deadman_calls); 541283b8460SGeorge.Wilson if (zfs_deadman_enabled) 542283b8460SGeorge.Wilson vdev_deadman(spa->spa_root_vdev); 543283b8460SGeorge.Wilson } 544283b8460SGeorge.Wilson 545fa9e4066Sahrens /* 546fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 547fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 548fa9e4066Sahrens * exist by calling spa_lookup() first. 549fa9e4066Sahrens */ 550fa9e4066Sahrens spa_t * 551468c413aSTim Haley spa_add(const char *name, nvlist_t *config, const char *altroot) 552fa9e4066Sahrens { 553fa9e4066Sahrens spa_t *spa; 554c5904d13Seschrock spa_config_dirent_t *dp; 555283b8460SGeorge.Wilson cyc_handler_t hdlr; 556283b8460SGeorge.Wilson cyc_time_t when; 557fa9e4066Sahrens 558fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 559fa9e4066Sahrens 560fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 561fa9e4066Sahrens 562c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 563c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 56435a5a358SJonathan Adams mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 565bc9014e6SJustin Gibbs mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL); 566c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 56735a5a358SJonathan Adams mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); 568c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 56945818ee1SMatthew Ahrens mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL); 57035a5a358SJonathan Adams mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 571a1521560SJeff Bonwick mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); 572a1521560SJeff Bonwick mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); 573c3a66015SMatthew Ahrens mutex_init(&spa->spa_iokstat_lock, NULL, MUTEX_DEFAULT, NULL); 5740f7643c7SGeorge Wilson mutex_init(&spa->spa_alloc_lock, NULL, MUTEX_DEFAULT, NULL); 575c25056deSgw 576c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 577bc9014e6SJustin Gibbs cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL); 57835a5a358SJonathan Adams cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); 579c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 580e14bb325SJeff Bonwick cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); 581c25056deSgw 582b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 583cde58dbcSMatthew Ahrens bplist_create(&spa->spa_free_bplist[t]); 584b24ab676SJeff Bonwick 585e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); 586fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 587fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 5880373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 589468c413aSTim Haley spa->spa_load_max_txg = UINT64_MAX; 59035a5a358SJonathan Adams spa->spa_proc = &p0; 59135a5a358SJonathan Adams spa->spa_proc_state = SPA_PROC_NONE; 592fa9e4066Sahrens 593283b8460SGeorge.Wilson hdlr.cyh_func = spa_deadman; 594283b8460SGeorge.Wilson hdlr.cyh_arg = spa; 595283b8460SGeorge.Wilson hdlr.cyh_level = CY_LOW_LEVEL; 596283b8460SGeorge.Wilson 59769962b56SMatthew Ahrens spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms); 598283b8460SGeorge.Wilson 599283b8460SGeorge.Wilson /* 600283b8460SGeorge.Wilson * This determines how often we need to check for hung I/Os after 601283b8460SGeorge.Wilson * the cyclic has already fired. Since checking for hung I/Os is 602283b8460SGeorge.Wilson * an expensive operation we don't want to check too frequently. 60369962b56SMatthew Ahrens * Instead wait for 5 seconds before checking again. 604283b8460SGeorge.Wilson */ 60569962b56SMatthew Ahrens when.cyt_interval = MSEC2NSEC(zfs_deadman_checktime_ms); 606283b8460SGeorge.Wilson when.cyt_when = CY_INFINITY; 607283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 608283b8460SGeorge.Wilson spa->spa_deadman_cycid = cyclic_add(&hdlr, &when); 609283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 610283b8460SGeorge.Wilson 611fa9e4066Sahrens refcount_create(&spa->spa_refcount); 612e14bb325SJeff Bonwick spa_config_lock_init(spa); 613fa9e4066Sahrens 614fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 615fa9e4066Sahrens 6160373e76bSbonwick /* 6170373e76bSbonwick * Set the alternate root, if there is one. 6180373e76bSbonwick */ 6190373e76bSbonwick if (altroot) { 6200373e76bSbonwick spa->spa_root = spa_strdup(altroot); 6210373e76bSbonwick spa_active_count++; 6220373e76bSbonwick } 6230373e76bSbonwick 62494c2d0ebSMatthew Ahrens avl_create(&spa->spa_alloc_tree, zio_bookmark_compare, 6250f7643c7SGeorge Wilson sizeof (zio_t), offsetof(zio_t, io_alloc_node)); 6260f7643c7SGeorge Wilson 627c5904d13Seschrock /* 628c5904d13Seschrock * Every pool starts with the default cachefile 629c5904d13Seschrock */ 630c5904d13Seschrock list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), 631c5904d13Seschrock offsetof(spa_config_dirent_t, scd_link)); 632c5904d13Seschrock 633c5904d13Seschrock dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); 634ef912c80STim Haley dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); 635c5904d13Seschrock list_insert_head(&spa->spa_config_list, dp); 636c5904d13Seschrock 6374b964adaSGeorge Wilson VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, 6384b964adaSGeorge Wilson KM_SLEEP) == 0); 6394b964adaSGeorge Wilson 640ad135b5dSChristopher Siden if (config != NULL) { 641ad135b5dSChristopher Siden nvlist_t *features; 642ad135b5dSChristopher Siden 643ad135b5dSChristopher Siden if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, 644ad135b5dSChristopher Siden &features) == 0) { 645ad135b5dSChristopher Siden VERIFY(nvlist_dup(features, &spa->spa_label_features, 646ad135b5dSChristopher Siden 0) == 0); 647ad135b5dSChristopher Siden } 648ad135b5dSChristopher Siden 649468c413aSTim Haley VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); 650ad135b5dSChristopher Siden } 651ad135b5dSChristopher Siden 652ad135b5dSChristopher Siden if (spa->spa_label_features == NULL) { 653ad135b5dSChristopher Siden VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, 654ad135b5dSChristopher Siden KM_SLEEP) == 0); 655ad135b5dSChristopher Siden } 656468c413aSTim Haley 657c3a66015SMatthew Ahrens spa->spa_iokstat = kstat_create("zfs", 0, name, 658c3a66015SMatthew Ahrens "disk", KSTAT_TYPE_IO, 1, 0); 659c3a66015SMatthew Ahrens if (spa->spa_iokstat) { 660c3a66015SMatthew Ahrens spa->spa_iokstat->ks_lock = &spa->spa_iokstat_lock; 661c3a66015SMatthew Ahrens kstat_install(spa->spa_iokstat); 662c3a66015SMatthew Ahrens } 663c3a66015SMatthew Ahrens 6643b2aab18SMatthew Ahrens spa->spa_debug = ((zfs_flags & ZFS_DEBUG_SPA) != 0); 6653b2aab18SMatthew Ahrens 66681cd5c55SMatthew Ahrens spa->spa_min_ashift = INT_MAX; 66781cd5c55SMatthew Ahrens spa->spa_max_ashift = 0; 66881cd5c55SMatthew Ahrens 66943466aaeSMax Grossman /* 67043466aaeSMax Grossman * As a pool is being created, treat all features as disabled by 67143466aaeSMax Grossman * setting SPA_FEATURE_DISABLED for all entries in the feature 67243466aaeSMax Grossman * refcount cache. 67343466aaeSMax Grossman */ 67443466aaeSMax Grossman for (int i = 0; i < SPA_FEATURES; i++) { 67543466aaeSMax Grossman spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; 67643466aaeSMax Grossman } 67743466aaeSMax Grossman 678fa9e4066Sahrens return (spa); 679fa9e4066Sahrens } 680fa9e4066Sahrens 681fa9e4066Sahrens /* 682fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 683fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 684fa9e4066Sahrens * deactivated. 685fa9e4066Sahrens */ 686fa9e4066Sahrens void 687fa9e4066Sahrens spa_remove(spa_t *spa) 688fa9e4066Sahrens { 689c5904d13Seschrock spa_config_dirent_t *dp; 690c5904d13Seschrock 691fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 692fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 693bc9014e6SJustin Gibbs ASSERT3U(refcount_count(&spa->spa_refcount), ==, 0); 694fa9e4066Sahrens 6951195e687SMark J Musante nvlist_free(spa->spa_config_splitting); 6961195e687SMark J Musante 697fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 698fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 699fa9e4066Sahrens 7000373e76bSbonwick if (spa->spa_root) { 701fa9e4066Sahrens spa_strfree(spa->spa_root); 7020373e76bSbonwick spa_active_count--; 7030373e76bSbonwick } 704fa9e4066Sahrens 705c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 706c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 707c5904d13Seschrock if (dp->scd_path != NULL) 708c5904d13Seschrock spa_strfree(dp->scd_path); 709c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 710c5904d13Seschrock } 711c5904d13Seschrock 7120f7643c7SGeorge Wilson avl_destroy(&spa->spa_alloc_tree); 713c5904d13Seschrock list_destroy(&spa->spa_config_list); 7142f8aaab3Seschrock 715ad135b5dSChristopher Siden nvlist_free(spa->spa_label_features); 7164b964adaSGeorge Wilson nvlist_free(spa->spa_load_info); 717fa9e4066Sahrens spa_config_set(spa, NULL); 718fa9e4066Sahrens 719283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 720283b8460SGeorge.Wilson if (spa->spa_deadman_cycid != CYCLIC_NONE) 721283b8460SGeorge.Wilson cyclic_remove(spa->spa_deadman_cycid); 722283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 723283b8460SGeorge.Wilson spa->spa_deadman_cycid = CYCLIC_NONE; 724283b8460SGeorge.Wilson 725fa9e4066Sahrens refcount_destroy(&spa->spa_refcount); 72691ebeef5Sahrens 727e14bb325SJeff Bonwick spa_config_lock_destroy(spa); 728fa9e4066Sahrens 729c3a66015SMatthew Ahrens kstat_delete(spa->spa_iokstat); 730c3a66015SMatthew Ahrens spa->spa_iokstat = NULL; 731c3a66015SMatthew Ahrens 732b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 733cde58dbcSMatthew Ahrens bplist_destroy(&spa->spa_free_bplist[t]); 734b24ab676SJeff Bonwick 73545818ee1SMatthew Ahrens zio_checksum_templates_free(spa); 73645818ee1SMatthew Ahrens 737c25056deSgw cv_destroy(&spa->spa_async_cv); 738bc9014e6SJustin Gibbs cv_destroy(&spa->spa_evicting_os_cv); 73935a5a358SJonathan Adams cv_destroy(&spa->spa_proc_cv); 740c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 741e14bb325SJeff Bonwick cv_destroy(&spa->spa_suspend_cv); 742c25056deSgw 7430f7643c7SGeorge Wilson mutex_destroy(&spa->spa_alloc_lock); 7445ad82045Snd mutex_destroy(&spa->spa_async_lock); 745c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 74635a5a358SJonathan Adams mutex_destroy(&spa->spa_errlog_lock); 747bc9014e6SJustin Gibbs mutex_destroy(&spa->spa_evicting_os_lock); 74806eeb2adSek mutex_destroy(&spa->spa_history_lock); 74935a5a358SJonathan Adams mutex_destroy(&spa->spa_proc_lock); 750b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 75145818ee1SMatthew Ahrens mutex_destroy(&spa->spa_cksum_tmpls_lock); 75235a5a358SJonathan Adams mutex_destroy(&spa->spa_scrub_lock); 753e14bb325SJeff Bonwick mutex_destroy(&spa->spa_suspend_lock); 754a1521560SJeff Bonwick mutex_destroy(&spa->spa_vdev_top_lock); 755c3a66015SMatthew Ahrens mutex_destroy(&spa->spa_iokstat_lock); 7565ad82045Snd 757fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 758fa9e4066Sahrens } 759fa9e4066Sahrens 760fa9e4066Sahrens /* 761fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 762fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 763fa9e4066Sahrens */ 764fa9e4066Sahrens spa_t * 765fa9e4066Sahrens spa_next(spa_t *prev) 766fa9e4066Sahrens { 767fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 768fa9e4066Sahrens 769fa9e4066Sahrens if (prev) 770fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 771fa9e4066Sahrens else 772fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 773fa9e4066Sahrens } 774fa9e4066Sahrens 775fa9e4066Sahrens /* 776fa9e4066Sahrens * ========================================================================== 777fa9e4066Sahrens * SPA refcount functions 778fa9e4066Sahrens * ========================================================================== 779fa9e4066Sahrens */ 780fa9e4066Sahrens 781fa9e4066Sahrens /* 782fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 783fa9e4066Sahrens * have the namespace lock held. 784fa9e4066Sahrens */ 785fa9e4066Sahrens void 786fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 787fa9e4066Sahrens { 788088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref || 789fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 790fa9e4066Sahrens (void) refcount_add(&spa->spa_refcount, tag); 791fa9e4066Sahrens } 792fa9e4066Sahrens 793fa9e4066Sahrens /* 794fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 795fa9e4066Sahrens * have the namespace lock held. 796fa9e4066Sahrens */ 797fa9e4066Sahrens void 798fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 799fa9e4066Sahrens { 800088f3894Sahrens ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref || 801fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 802fa9e4066Sahrens (void) refcount_remove(&spa->spa_refcount, tag); 803fa9e4066Sahrens } 804fa9e4066Sahrens 805bc9014e6SJustin Gibbs /* 806bc9014e6SJustin Gibbs * Remove a reference to the given spa_t held by a dsl dir that is 807bc9014e6SJustin Gibbs * being asynchronously released. Async releases occur from a taskq 808bc9014e6SJustin Gibbs * performing eviction of dsl datasets and dirs. The namespace lock 809bc9014e6SJustin Gibbs * isn't held and the hold by the object being evicted may contribute to 810bc9014e6SJustin Gibbs * spa_minref (e.g. dataset or directory released during pool export), 811bc9014e6SJustin Gibbs * so the asserts in spa_close() do not apply. 812bc9014e6SJustin Gibbs */ 813bc9014e6SJustin Gibbs void 814bc9014e6SJustin Gibbs spa_async_close(spa_t *spa, void *tag) 815bc9014e6SJustin Gibbs { 816bc9014e6SJustin Gibbs (void) refcount_remove(&spa->spa_refcount, tag); 817bc9014e6SJustin Gibbs } 818bc9014e6SJustin Gibbs 819fa9e4066Sahrens /* 820fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 821088f3894Sahrens * spa_namespace_lock held. We really compare against spa_minref, which is the 822fa9e4066Sahrens * number of references acquired when opening a pool 823fa9e4066Sahrens */ 824fa9e4066Sahrens boolean_t 825fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 826fa9e4066Sahrens { 827fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 828fa9e4066Sahrens 829088f3894Sahrens return (refcount_count(&spa->spa_refcount) == spa->spa_minref); 830fa9e4066Sahrens } 831fa9e4066Sahrens 83299653d4eSeschrock /* 83399653d4eSeschrock * ========================================================================== 834fa94a07fSbrendan * SPA spare and l2cache tracking 83599653d4eSeschrock * ========================================================================== 83699653d4eSeschrock */ 83799653d4eSeschrock 838fa94a07fSbrendan /* 839fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 840fa94a07fSbrendan * for 'auxiliary' devices. 841fa94a07fSbrendan */ 842fa94a07fSbrendan 843fa94a07fSbrendan typedef struct spa_aux { 844fa94a07fSbrendan uint64_t aux_guid; 845fa94a07fSbrendan uint64_t aux_pool; 846fa94a07fSbrendan avl_node_t aux_avl; 847fa94a07fSbrendan int aux_count; 848fa94a07fSbrendan } spa_aux_t; 849fa94a07fSbrendan 850fa94a07fSbrendan static int 851fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 852fa94a07fSbrendan { 853fa94a07fSbrendan const spa_aux_t *sa = a; 854fa94a07fSbrendan const spa_aux_t *sb = b; 855fa94a07fSbrendan 856fa94a07fSbrendan if (sa->aux_guid < sb->aux_guid) 857fa94a07fSbrendan return (-1); 858fa94a07fSbrendan else if (sa->aux_guid > sb->aux_guid) 859fa94a07fSbrendan return (1); 860fa94a07fSbrendan else 861fa94a07fSbrendan return (0); 862fa94a07fSbrendan } 863fa94a07fSbrendan 864fa94a07fSbrendan void 865fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 866fa94a07fSbrendan { 867fa94a07fSbrendan avl_index_t where; 868fa94a07fSbrendan spa_aux_t search; 869fa94a07fSbrendan spa_aux_t *aux; 870fa94a07fSbrendan 871fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 872fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 873fa94a07fSbrendan aux->aux_count++; 874fa94a07fSbrendan } else { 875fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 876fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 877fa94a07fSbrendan aux->aux_count = 1; 878fa94a07fSbrendan avl_insert(avl, aux, where); 879fa94a07fSbrendan } 880fa94a07fSbrendan } 881fa94a07fSbrendan 882fa94a07fSbrendan void 883fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 884fa94a07fSbrendan { 885fa94a07fSbrendan spa_aux_t search; 886fa94a07fSbrendan spa_aux_t *aux; 887fa94a07fSbrendan avl_index_t where; 888fa94a07fSbrendan 889fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 890fa94a07fSbrendan aux = avl_find(avl, &search, &where); 891fa94a07fSbrendan 892fa94a07fSbrendan ASSERT(aux != NULL); 893fa94a07fSbrendan 894fa94a07fSbrendan if (--aux->aux_count == 0) { 895fa94a07fSbrendan avl_remove(avl, aux); 896fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 897fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 898fa94a07fSbrendan aux->aux_pool = 0ULL; 899fa94a07fSbrendan } 900fa94a07fSbrendan } 901fa94a07fSbrendan 902fa94a07fSbrendan boolean_t 90389a89ebfSlling spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) 904fa94a07fSbrendan { 905fa94a07fSbrendan spa_aux_t search, *found; 906fa94a07fSbrendan 907fa94a07fSbrendan search.aux_guid = guid; 90889a89ebfSlling found = avl_find(avl, &search, NULL); 909fa94a07fSbrendan 910fa94a07fSbrendan if (pool) { 911fa94a07fSbrendan if (found) 912fa94a07fSbrendan *pool = found->aux_pool; 913fa94a07fSbrendan else 914fa94a07fSbrendan *pool = 0ULL; 915fa94a07fSbrendan } 916fa94a07fSbrendan 91789a89ebfSlling if (refcnt) { 91889a89ebfSlling if (found) 91989a89ebfSlling *refcnt = found->aux_count; 92089a89ebfSlling else 92189a89ebfSlling *refcnt = 0; 92289a89ebfSlling } 92389a89ebfSlling 924fa94a07fSbrendan return (found != NULL); 925fa94a07fSbrendan } 926fa94a07fSbrendan 927fa94a07fSbrendan void 928fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 929fa94a07fSbrendan { 930fa94a07fSbrendan spa_aux_t search, *found; 931fa94a07fSbrendan avl_index_t where; 932fa94a07fSbrendan 933fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 934fa94a07fSbrendan found = avl_find(avl, &search, &where); 935fa94a07fSbrendan ASSERT(found != NULL); 936fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 937fa94a07fSbrendan 938fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 939fa94a07fSbrendan } 940fa94a07fSbrendan 94199653d4eSeschrock /* 94239c23413Seschrock * Spares are tracked globally due to the following constraints: 94339c23413Seschrock * 94439c23413Seschrock * - A spare may be part of multiple pools. 94539c23413Seschrock * - A spare may be added to a pool even if it's actively in use within 94639c23413Seschrock * another pool. 94739c23413Seschrock * - A spare in use in any pool can only be the source of a replacement if 94839c23413Seschrock * the target is a spare in the same pool. 94939c23413Seschrock * 95039c23413Seschrock * We keep track of all spares on the system through the use of a reference 95139c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 95239c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 95339c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 95439c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 95539c23413Seschrock * pool), we also keep track of which pool its been made a part of. 95639c23413Seschrock * 95739c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 95839c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 95939c23413Seschrock * separate spare lock exists for the status query path, which does not need to 96039c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 96199653d4eSeschrock */ 96239c23413Seschrock 96399653d4eSeschrock static int 96499653d4eSeschrock spa_spare_compare(const void *a, const void *b) 96599653d4eSeschrock { 966fa94a07fSbrendan return (spa_aux_compare(a, b)); 96799653d4eSeschrock } 96899653d4eSeschrock 96999653d4eSeschrock void 97039c23413Seschrock spa_spare_add(vdev_t *vd) 97199653d4eSeschrock { 97299653d4eSeschrock mutex_enter(&spa_spare_lock); 97339c23413Seschrock ASSERT(!vd->vdev_isspare); 974fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 97539c23413Seschrock vd->vdev_isspare = B_TRUE; 97699653d4eSeschrock mutex_exit(&spa_spare_lock); 97799653d4eSeschrock } 97899653d4eSeschrock 97999653d4eSeschrock void 98039c23413Seschrock spa_spare_remove(vdev_t *vd) 98199653d4eSeschrock { 98299653d4eSeschrock mutex_enter(&spa_spare_lock); 98339c23413Seschrock ASSERT(vd->vdev_isspare); 984fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 98539c23413Seschrock vd->vdev_isspare = B_FALSE; 98699653d4eSeschrock mutex_exit(&spa_spare_lock); 98799653d4eSeschrock } 98899653d4eSeschrock 98999653d4eSeschrock boolean_t 99089a89ebfSlling spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) 99199653d4eSeschrock { 992fa94a07fSbrendan boolean_t found; 99399653d4eSeschrock 99499653d4eSeschrock mutex_enter(&spa_spare_lock); 99589a89ebfSlling found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); 99699653d4eSeschrock mutex_exit(&spa_spare_lock); 99799653d4eSeschrock 998fa94a07fSbrendan return (found); 99939c23413Seschrock } 100039c23413Seschrock 100139c23413Seschrock void 100239c23413Seschrock spa_spare_activate(vdev_t *vd) 100339c23413Seschrock { 100439c23413Seschrock mutex_enter(&spa_spare_lock); 100539c23413Seschrock ASSERT(vd->vdev_isspare); 1006fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 1007fa94a07fSbrendan mutex_exit(&spa_spare_lock); 1008fa94a07fSbrendan } 100939c23413Seschrock 1010fa94a07fSbrendan /* 1011fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 1012fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 1013fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 1014fa94a07fSbrendan */ 101539c23413Seschrock 1016fa94a07fSbrendan static int 1017fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 1018fa94a07fSbrendan { 1019fa94a07fSbrendan return (spa_aux_compare(a, b)); 1020fa94a07fSbrendan } 1021fa94a07fSbrendan 1022fa94a07fSbrendan void 1023fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 1024fa94a07fSbrendan { 1025fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1026fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 1027fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 1028fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 1029fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1030fa94a07fSbrendan } 1031fa94a07fSbrendan 1032fa94a07fSbrendan void 1033fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 1034fa94a07fSbrendan { 1035fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1036fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 1037fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 1038fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 1039fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1040fa94a07fSbrendan } 1041fa94a07fSbrendan 1042fa94a07fSbrendan boolean_t 1043fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 1044fa94a07fSbrendan { 1045fa94a07fSbrendan boolean_t found; 1046fa94a07fSbrendan 1047fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 104889a89ebfSlling found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); 1049fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1050fa94a07fSbrendan 1051fa94a07fSbrendan return (found); 1052fa94a07fSbrendan } 1053fa94a07fSbrendan 1054fa94a07fSbrendan void 1055fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 1056fa94a07fSbrendan { 1057fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1058fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 1059fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 1060fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1061fa94a07fSbrendan } 1062fa94a07fSbrendan 1063fa9e4066Sahrens /* 1064fa9e4066Sahrens * ========================================================================== 1065fa9e4066Sahrens * SPA vdev locking 1066fa9e4066Sahrens * ========================================================================== 1067fa9e4066Sahrens */ 1068fa9e4066Sahrens 1069fa9e4066Sahrens /* 1070ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 1071ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 1072fa9e4066Sahrens * It returns the next transaction group for the spa_t. 1073fa9e4066Sahrens */ 1074fa9e4066Sahrens uint64_t 1075fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 1076fa9e4066Sahrens { 1077a1521560SJeff Bonwick mutex_enter(&spa->spa_vdev_top_lock); 1078bbfd46c4SJeff Bonwick mutex_enter(&spa_namespace_lock); 107988ecc943SGeorge Wilson return (spa_vdev_config_enter(spa)); 108088ecc943SGeorge Wilson } 108188ecc943SGeorge Wilson 108288ecc943SGeorge Wilson /* 108388ecc943SGeorge Wilson * Internal implementation for spa_vdev_enter(). Used when a vdev 108488ecc943SGeorge Wilson * operation requires multiple syncs (i.e. removing a device) while 108588ecc943SGeorge Wilson * keeping the spa_namespace_lock held. 108688ecc943SGeorge Wilson */ 108788ecc943SGeorge Wilson uint64_t 108888ecc943SGeorge Wilson spa_vdev_config_enter(spa_t *spa) 108988ecc943SGeorge Wilson { 109088ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 10913d7072f8Seschrock 1092e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1093fa9e4066Sahrens 1094fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 1095fa9e4066Sahrens } 1096fa9e4066Sahrens 1097fa9e4066Sahrens /* 109888ecc943SGeorge Wilson * Used in combination with spa_vdev_config_enter() to allow the syncing 109988ecc943SGeorge Wilson * of multiple transactions without releasing the spa_namespace_lock. 1100fa9e4066Sahrens */ 110188ecc943SGeorge Wilson void 110288ecc943SGeorge Wilson spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) 1103fa9e4066Sahrens { 110488ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 110588ecc943SGeorge Wilson 11060e34b6a7Sbonwick int config_changed = B_FALSE; 1107ea8dc4b6Seschrock 11080373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 11090e34b6a7Sbonwick 1110e14bb325SJeff Bonwick spa->spa_pending_vdev = NULL; 1111e14bb325SJeff Bonwick 11120e34b6a7Sbonwick /* 11130e34b6a7Sbonwick * Reassess the DTLs. 11140e34b6a7Sbonwick */ 11150373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 11160e34b6a7Sbonwick 1117e14bb325SJeff Bonwick if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { 11180e34b6a7Sbonwick config_changed = B_TRUE; 11198f18d1faSGeorge Wilson spa->spa_config_generation++; 11200e34b6a7Sbonwick } 1121ea8dc4b6Seschrock 112288ecc943SGeorge Wilson /* 112388ecc943SGeorge Wilson * Verify the metaslab classes. 112488ecc943SGeorge Wilson */ 1125b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); 1126b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); 112788ecc943SGeorge Wilson 1128e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1129fa9e4066Sahrens 113088ecc943SGeorge Wilson /* 113188ecc943SGeorge Wilson * Panic the system if the specified tag requires it. This 113288ecc943SGeorge Wilson * is useful for ensuring that configurations are updated 113388ecc943SGeorge Wilson * transactionally. 113488ecc943SGeorge Wilson */ 113588ecc943SGeorge Wilson if (zio_injection_enabled) 11361195e687SMark J Musante zio_handle_panic_injection(spa, tag, 0); 113788ecc943SGeorge Wilson 1138fa9e4066Sahrens /* 1139fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 1140fa9e4066Sahrens * that there won't be more than one config change per txg. 1141fa9e4066Sahrens * This allows us to use the txg as the generation number. 1142fa9e4066Sahrens */ 1143fa9e4066Sahrens if (error == 0) 1144fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 1145fa9e4066Sahrens 1146fa9e4066Sahrens if (vd != NULL) { 11470713e232SGeorge Wilson ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL); 11488ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1149fa9e4066Sahrens vdev_free(vd); 11508ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1151fa9e4066Sahrens } 1152fa9e4066Sahrens 1153fa9e4066Sahrens /* 11540e34b6a7Sbonwick * If the config changed, update the config cache. 1155fa9e4066Sahrens */ 11560e34b6a7Sbonwick if (config_changed) 1157c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 115888ecc943SGeorge Wilson } 1159ea8dc4b6Seschrock 116088ecc943SGeorge Wilson /* 116188ecc943SGeorge Wilson * Unlock the spa_t after adding or removing a vdev. Besides undoing the 116288ecc943SGeorge Wilson * locking of spa_vdev_enter(), we also want make sure the transactions have 116388ecc943SGeorge Wilson * synced to disk, and then update the global configuration cache with the new 116488ecc943SGeorge Wilson * information. 116588ecc943SGeorge Wilson */ 116688ecc943SGeorge Wilson int 116788ecc943SGeorge Wilson spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 116888ecc943SGeorge Wilson { 116988ecc943SGeorge Wilson spa_vdev_config_exit(spa, vd, txg, error, FTAG); 1170ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1171bbfd46c4SJeff Bonwick mutex_exit(&spa->spa_vdev_top_lock); 1172fa9e4066Sahrens 1173fa9e4066Sahrens return (error); 1174fa9e4066Sahrens } 1175fa9e4066Sahrens 1176e14bb325SJeff Bonwick /* 1177e14bb325SJeff Bonwick * Lock the given spa_t for the purpose of changing vdev state. 1178e14bb325SJeff Bonwick */ 1179e14bb325SJeff Bonwick void 11808f18d1faSGeorge Wilson spa_vdev_state_enter(spa_t *spa, int oplocks) 1181e14bb325SJeff Bonwick { 11828f18d1faSGeorge Wilson int locks = SCL_STATE_ALL | oplocks; 11838f18d1faSGeorge Wilson 1184dcba9f3fSGeorge Wilson /* 1185dcba9f3fSGeorge Wilson * Root pools may need to read of the underlying devfs filesystem 1186dcba9f3fSGeorge Wilson * when opening up a vdev. Unfortunately if we're holding the 1187dcba9f3fSGeorge Wilson * SCL_ZIO lock it will result in a deadlock when we try to issue 1188dcba9f3fSGeorge Wilson * the read from the root filesystem. Instead we "prefetch" 1189dcba9f3fSGeorge Wilson * the associated vnodes that we need prior to opening the 1190dcba9f3fSGeorge Wilson * underlying devices and cache them so that we can prevent 1191dcba9f3fSGeorge Wilson * any I/O when we are doing the actual open. 1192dcba9f3fSGeorge Wilson */ 1193dcba9f3fSGeorge Wilson if (spa_is_root(spa)) { 11949842588bSGeorge Wilson int low = locks & ~(SCL_ZIO - 1); 11959842588bSGeorge Wilson int high = locks & ~low; 11969842588bSGeorge Wilson 11979842588bSGeorge Wilson spa_config_enter(spa, high, spa, RW_WRITER); 1198dcba9f3fSGeorge Wilson vdev_hold(spa->spa_root_vdev); 11999842588bSGeorge Wilson spa_config_enter(spa, low, spa, RW_WRITER); 1200dcba9f3fSGeorge Wilson } else { 1201dcba9f3fSGeorge Wilson spa_config_enter(spa, locks, spa, RW_WRITER); 1202dcba9f3fSGeorge Wilson } 12038f18d1faSGeorge Wilson spa->spa_vdev_locks = locks; 1204e14bb325SJeff Bonwick } 1205e14bb325SJeff Bonwick 1206e14bb325SJeff Bonwick int 1207e14bb325SJeff Bonwick spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) 1208e14bb325SJeff Bonwick { 1209c6065d0fSGeorge Wilson boolean_t config_changed = B_FALSE; 1210c6065d0fSGeorge Wilson 1211b24ab676SJeff Bonwick if (vd != NULL || error == 0) 1212b24ab676SJeff Bonwick vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 1213b24ab676SJeff Bonwick 0, 0, B_FALSE); 1214b24ab676SJeff Bonwick 12158f18d1faSGeorge Wilson if (vd != NULL) { 1216e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 1217c6065d0fSGeorge Wilson config_changed = B_TRUE; 12188f18d1faSGeorge Wilson spa->spa_config_generation++; 12198f18d1faSGeorge Wilson } 1220e14bb325SJeff Bonwick 1221dcba9f3fSGeorge Wilson if (spa_is_root(spa)) 1222dcba9f3fSGeorge Wilson vdev_rele(spa->spa_root_vdev); 1223dcba9f3fSGeorge Wilson 12248f18d1faSGeorge Wilson ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); 12258f18d1faSGeorge Wilson spa_config_exit(spa, spa->spa_vdev_locks, spa); 1226e14bb325SJeff Bonwick 12278ad4d6ddSJeff Bonwick /* 12288ad4d6ddSJeff Bonwick * If anything changed, wait for it to sync. This ensures that, 12298ad4d6ddSJeff Bonwick * from the system administrator's perspective, zpool(1M) commands 12308ad4d6ddSJeff Bonwick * are synchronous. This is important for things like zpool offline: 12318ad4d6ddSJeff Bonwick * when the command completes, you expect no further I/O from ZFS. 12328ad4d6ddSJeff Bonwick */ 12338ad4d6ddSJeff Bonwick if (vd != NULL) 12348ad4d6ddSJeff Bonwick txg_wait_synced(spa->spa_dsl_pool, 0); 12358ad4d6ddSJeff Bonwick 1236c6065d0fSGeorge Wilson /* 1237c6065d0fSGeorge Wilson * If the config changed, update the config cache. 1238c6065d0fSGeorge Wilson */ 1239c6065d0fSGeorge Wilson if (config_changed) { 1240c6065d0fSGeorge Wilson mutex_enter(&spa_namespace_lock); 1241c6065d0fSGeorge Wilson spa_config_sync(spa, B_FALSE, B_TRUE); 1242c6065d0fSGeorge Wilson mutex_exit(&spa_namespace_lock); 1243c6065d0fSGeorge Wilson } 1244c6065d0fSGeorge Wilson 1245e14bb325SJeff Bonwick return (error); 1246e14bb325SJeff Bonwick } 1247e14bb325SJeff Bonwick 1248fa9e4066Sahrens /* 1249fa9e4066Sahrens * ========================================================================== 1250fa9e4066Sahrens * Miscellaneous functions 1251fa9e4066Sahrens * ========================================================================== 1252fa9e4066Sahrens */ 1253fa9e4066Sahrens 1254ad135b5dSChristopher Siden void 125543466aaeSMax Grossman spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx) 1256ad135b5dSChristopher Siden { 12572acef22dSMatthew Ahrens if (!nvlist_exists(spa->spa_label_features, feature)) { 12582acef22dSMatthew Ahrens fnvlist_add_boolean(spa->spa_label_features, feature); 125943466aaeSMax Grossman /* 126043466aaeSMax Grossman * When we are creating the pool (tx_txg==TXG_INITIAL), we can't 126143466aaeSMax Grossman * dirty the vdev config because lock SCL_CONFIG is not held. 126243466aaeSMax Grossman * Thankfully, in this case we don't need to dirty the config 126343466aaeSMax Grossman * because it will be written out anyway when we finish 126443466aaeSMax Grossman * creating the pool. 126543466aaeSMax Grossman */ 126643466aaeSMax Grossman if (tx->tx_txg != TXG_INITIAL) 126743466aaeSMax Grossman vdev_config_dirty(spa->spa_root_vdev); 12682acef22dSMatthew Ahrens } 1269ad135b5dSChristopher Siden } 1270ad135b5dSChristopher Siden 1271ad135b5dSChristopher Siden void 1272ad135b5dSChristopher Siden spa_deactivate_mos_feature(spa_t *spa, const char *feature) 1273ad135b5dSChristopher Siden { 12742acef22dSMatthew Ahrens if (nvlist_remove_all(spa->spa_label_features, feature) == 0) 12752acef22dSMatthew Ahrens vdev_config_dirty(spa->spa_root_vdev); 1276ad135b5dSChristopher Siden } 1277ad135b5dSChristopher Siden 1278fa9e4066Sahrens /* 1279fa9e4066Sahrens * Rename a spa_t. 1280fa9e4066Sahrens */ 1281fa9e4066Sahrens int 1282fa9e4066Sahrens spa_rename(const char *name, const char *newname) 1283fa9e4066Sahrens { 1284fa9e4066Sahrens spa_t *spa; 1285fa9e4066Sahrens int err; 1286fa9e4066Sahrens 1287fa9e4066Sahrens /* 1288fa9e4066Sahrens * Lookup the spa_t and grab the config lock for writing. We need to 1289fa9e4066Sahrens * actually open the pool so that we can sync out the necessary labels. 1290fa9e4066Sahrens * It's OK to call spa_open() with the namespace lock held because we 1291ea8dc4b6Seschrock * allow recursive calls for other reasons. 1292fa9e4066Sahrens */ 1293fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 1294fa9e4066Sahrens if ((err = spa_open(name, &spa, FTAG)) != 0) { 1295fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1296fa9e4066Sahrens return (err); 1297fa9e4066Sahrens } 1298fa9e4066Sahrens 1299e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1300fa9e4066Sahrens 1301fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 1302e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name)); 1303fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 1304fa9e4066Sahrens 1305fa9e4066Sahrens /* 1306fa9e4066Sahrens * Sync all labels to disk with the new names by marking the root vdev 1307fa9e4066Sahrens * dirty and waiting for it to sync. It will pick up the new pool name 1308fa9e4066Sahrens * during the sync. 1309fa9e4066Sahrens */ 1310fa9e4066Sahrens vdev_config_dirty(spa->spa_root_vdev); 1311fa9e4066Sahrens 1312e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1313fa9e4066Sahrens 13140373e76bSbonwick txg_wait_synced(spa->spa_dsl_pool, 0); 1315fa9e4066Sahrens 1316fa9e4066Sahrens /* 1317fa9e4066Sahrens * Sync the updated config cache. 1318fa9e4066Sahrens */ 1319c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 1320fa9e4066Sahrens 1321fa9e4066Sahrens spa_close(spa, FTAG); 1322fa9e4066Sahrens 1323fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1324fa9e4066Sahrens 1325fa9e4066Sahrens return (0); 1326fa9e4066Sahrens } 1327fa9e4066Sahrens 1328fa9e4066Sahrens /* 1329f9af39baSGeorge Wilson * Return the spa_t associated with given pool_guid, if it exists. If 1330f9af39baSGeorge Wilson * device_guid is non-zero, determine whether the pool exists *and* contains 1331f9af39baSGeorge Wilson * a device with the specified device_guid. 1332fa9e4066Sahrens */ 1333f9af39baSGeorge Wilson spa_t * 1334f9af39baSGeorge Wilson spa_by_guid(uint64_t pool_guid, uint64_t device_guid) 1335fa9e4066Sahrens { 1336fa9e4066Sahrens spa_t *spa; 1337fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 1338fa9e4066Sahrens 1339ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1340fa9e4066Sahrens 1341fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 1342fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 1343fa9e4066Sahrens continue; 1344fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 1345fa9e4066Sahrens continue; 134639c23413Seschrock if (spa_guid(spa) == pool_guid) { 134739c23413Seschrock if (device_guid == 0) 134839c23413Seschrock break; 134939c23413Seschrock 135039c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 135139c23413Seschrock device_guid) != NULL) 135239c23413Seschrock break; 135339c23413Seschrock 135439c23413Seschrock /* 13558654d025Sperrin * Check any devices we may be in the process of adding. 135639c23413Seschrock */ 135739c23413Seschrock if (spa->spa_pending_vdev) { 135839c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 135939c23413Seschrock device_guid) != NULL) 136039c23413Seschrock break; 136139c23413Seschrock } 136239c23413Seschrock } 1363fa9e4066Sahrens } 1364fa9e4066Sahrens 1365f9af39baSGeorge Wilson return (spa); 1366f9af39baSGeorge Wilson } 1367f9af39baSGeorge Wilson 1368f9af39baSGeorge Wilson /* 1369f9af39baSGeorge Wilson * Determine whether a pool with the given pool_guid exists. 1370f9af39baSGeorge Wilson */ 1371f9af39baSGeorge Wilson boolean_t 1372f9af39baSGeorge Wilson spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 1373f9af39baSGeorge Wilson { 1374f9af39baSGeorge Wilson return (spa_by_guid(pool_guid, device_guid) != NULL); 1375fa9e4066Sahrens } 1376fa9e4066Sahrens 1377fa9e4066Sahrens char * 1378fa9e4066Sahrens spa_strdup(const char *s) 1379fa9e4066Sahrens { 1380fa9e4066Sahrens size_t len; 1381fa9e4066Sahrens char *new; 1382fa9e4066Sahrens 1383fa9e4066Sahrens len = strlen(s); 1384fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 1385fa9e4066Sahrens bcopy(s, new, len); 1386fa9e4066Sahrens new[len] = '\0'; 1387fa9e4066Sahrens 1388fa9e4066Sahrens return (new); 1389fa9e4066Sahrens } 1390fa9e4066Sahrens 1391fa9e4066Sahrens void 1392fa9e4066Sahrens spa_strfree(char *s) 1393fa9e4066Sahrens { 1394fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 1395fa9e4066Sahrens } 1396fa9e4066Sahrens 1397fa9e4066Sahrens uint64_t 1398fa9e4066Sahrens spa_get_random(uint64_t range) 1399fa9e4066Sahrens { 1400fa9e4066Sahrens uint64_t r; 1401fa9e4066Sahrens 1402fa9e4066Sahrens ASSERT(range != 0); 1403fa9e4066Sahrens 1404fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 1405fa9e4066Sahrens 1406fa9e4066Sahrens return (r % range); 1407fa9e4066Sahrens } 1408fa9e4066Sahrens 14091195e687SMark J Musante uint64_t 14101195e687SMark J Musante spa_generate_guid(spa_t *spa) 14111195e687SMark J Musante { 14121195e687SMark J Musante uint64_t guid = spa_get_random(-1ULL); 14131195e687SMark J Musante 14141195e687SMark J Musante if (spa != NULL) { 14151195e687SMark J Musante while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) 14161195e687SMark J Musante guid = spa_get_random(-1ULL); 14171195e687SMark J Musante } else { 14181195e687SMark J Musante while (guid == 0 || spa_guid_exists(guid, 0)) 14191195e687SMark J Musante guid = spa_get_random(-1ULL); 14201195e687SMark J Musante } 14211195e687SMark J Musante 14221195e687SMark J Musante return (guid); 14231195e687SMark J Musante } 14241195e687SMark J Musante 1425fa9e4066Sahrens void 142643466aaeSMax Grossman snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp) 1427fa9e4066Sahrens { 1428ad135b5dSChristopher Siden char type[256]; 1429f0ba89beSJeff Bonwick char *checksum = NULL; 1430f0ba89beSJeff Bonwick char *compress = NULL; 1431f0ba89beSJeff Bonwick 1432f0ba89beSJeff Bonwick if (bp != NULL) { 1433ad135b5dSChristopher Siden if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { 1434ad135b5dSChristopher Siden dmu_object_byteswap_t bswap = 1435ad135b5dSChristopher Siden DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); 1436ad135b5dSChristopher Siden (void) snprintf(type, sizeof (type), "bswap %s %s", 1437ad135b5dSChristopher Siden DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? 1438ad135b5dSChristopher Siden "metadata" : "data", 1439ad135b5dSChristopher Siden dmu_ot_byteswap[bswap].ob_name); 1440ad135b5dSChristopher Siden } else { 1441ad135b5dSChristopher Siden (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, 1442ad135b5dSChristopher Siden sizeof (type)); 1443ad135b5dSChristopher Siden } 14445d7b4d43SMatthew Ahrens if (!BP_IS_EMBEDDED(bp)) { 14455d7b4d43SMatthew Ahrens checksum = 14465d7b4d43SMatthew Ahrens zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; 14475d7b4d43SMatthew Ahrens } 1448f0ba89beSJeff Bonwick compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; 1449f0ba89beSJeff Bonwick } 1450fa9e4066Sahrens 145143466aaeSMax Grossman SNPRINTF_BLKPTR(snprintf, ' ', buf, buflen, bp, type, checksum, 145243466aaeSMax Grossman compress); 1453fa9e4066Sahrens } 1454fa9e4066Sahrens 1455fa9e4066Sahrens void 1456fa9e4066Sahrens spa_freeze(spa_t *spa) 1457fa9e4066Sahrens { 1458fa9e4066Sahrens uint64_t freeze_txg = 0; 1459fa9e4066Sahrens 1460e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1461fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 1462fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 1463fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 1464fa9e4066Sahrens } 1465e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1466fa9e4066Sahrens if (freeze_txg != 0) 1467fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 1468fa9e4066Sahrens } 1469fa9e4066Sahrens 14700125049cSahrens void 14710125049cSahrens zfs_panic_recover(const char *fmt, ...) 14720125049cSahrens { 14730125049cSahrens va_list adx; 14740125049cSahrens 14750125049cSahrens va_start(adx, fmt); 14760125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 14770125049cSahrens va_end(adx); 14780125049cSahrens } 14790125049cSahrens 14803f9d6ad7SLin Ling /* 14813f9d6ad7SLin Ling * This is a stripped-down version of strtoull, suitable only for converting 1482f7170741SWill Andrews * lowercase hexadecimal numbers that don't overflow. 14833f9d6ad7SLin Ling */ 14843f9d6ad7SLin Ling uint64_t 14854585130bSYuri Pankov zfs_strtonum(const char *str, char **nptr) 14863f9d6ad7SLin Ling { 14873f9d6ad7SLin Ling uint64_t val = 0; 14883f9d6ad7SLin Ling char c; 14893f9d6ad7SLin Ling int digit; 14903f9d6ad7SLin Ling 14913f9d6ad7SLin Ling while ((c = *str) != '\0') { 14923f9d6ad7SLin Ling if (c >= '0' && c <= '9') 14933f9d6ad7SLin Ling digit = c - '0'; 14943f9d6ad7SLin Ling else if (c >= 'a' && c <= 'f') 14953f9d6ad7SLin Ling digit = 10 + c - 'a'; 14963f9d6ad7SLin Ling else 14973f9d6ad7SLin Ling break; 14983f9d6ad7SLin Ling 14993f9d6ad7SLin Ling val *= 16; 15003f9d6ad7SLin Ling val += digit; 15013f9d6ad7SLin Ling 15023f9d6ad7SLin Ling str++; 15033f9d6ad7SLin Ling } 15043f9d6ad7SLin Ling 15053f9d6ad7SLin Ling if (nptr) 15063f9d6ad7SLin Ling *nptr = (char *)str; 15073f9d6ad7SLin Ling 15083f9d6ad7SLin Ling return (val); 15093f9d6ad7SLin Ling } 15103f9d6ad7SLin Ling 1511fa9e4066Sahrens /* 1512fa9e4066Sahrens * ========================================================================== 1513fa9e4066Sahrens * Accessor functions 1514fa9e4066Sahrens * ========================================================================== 1515fa9e4066Sahrens */ 1516fa9e4066Sahrens 1517088f3894Sahrens boolean_t 151888b7b0f2SMatthew Ahrens spa_shutting_down(spa_t *spa) 1519fa9e4066Sahrens { 152088b7b0f2SMatthew Ahrens return (spa->spa_async_suspended); 1521fa9e4066Sahrens } 1522fa9e4066Sahrens 1523fa9e4066Sahrens dsl_pool_t * 1524fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1525fa9e4066Sahrens { 1526fa9e4066Sahrens return (spa->spa_dsl_pool); 1527fa9e4066Sahrens } 1528fa9e4066Sahrens 1529ad135b5dSChristopher Siden boolean_t 1530ad135b5dSChristopher Siden spa_is_initializing(spa_t *spa) 1531ad135b5dSChristopher Siden { 1532ad135b5dSChristopher Siden return (spa->spa_is_initializing); 1533ad135b5dSChristopher Siden } 1534ad135b5dSChristopher Siden 1535fa9e4066Sahrens blkptr_t * 1536fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1537fa9e4066Sahrens { 1538fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1539fa9e4066Sahrens } 1540fa9e4066Sahrens 1541fa9e4066Sahrens void 1542fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1543fa9e4066Sahrens { 1544fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1545fa9e4066Sahrens } 1546fa9e4066Sahrens 1547fa9e4066Sahrens void 1548fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1549fa9e4066Sahrens { 1550fa9e4066Sahrens if (spa->spa_root == NULL) 1551fa9e4066Sahrens buf[0] = '\0'; 1552fa9e4066Sahrens else 1553fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1554fa9e4066Sahrens } 1555fa9e4066Sahrens 1556fa9e4066Sahrens int 1557fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1558fa9e4066Sahrens { 1559fa9e4066Sahrens return (spa->spa_sync_pass); 1560fa9e4066Sahrens } 1561fa9e4066Sahrens 1562fa9e4066Sahrens char * 1563fa9e4066Sahrens spa_name(spa_t *spa) 1564fa9e4066Sahrens { 1565fa9e4066Sahrens return (spa->spa_name); 1566fa9e4066Sahrens } 1567fa9e4066Sahrens 1568fa9e4066Sahrens uint64_t 1569fa9e4066Sahrens spa_guid(spa_t *spa) 1570fa9e4066Sahrens { 1571dfbb9432SGeorge Wilson dsl_pool_t *dp = spa_get_dsl(spa); 1572dfbb9432SGeorge Wilson uint64_t guid; 1573dfbb9432SGeorge Wilson 1574b5989ec7Seschrock /* 1575b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1576b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1577e9103aaeSGarrett D'Amore * vdev. We stash the original pool guid in 'spa_config_guid' to handle 1578b5989ec7Seschrock * this case. 1579b5989ec7Seschrock */ 1580dfbb9432SGeorge Wilson if (spa->spa_root_vdev == NULL) 1581dfbb9432SGeorge Wilson return (spa->spa_config_guid); 1582dfbb9432SGeorge Wilson 1583dfbb9432SGeorge Wilson guid = spa->spa_last_synced_guid != 0 ? 1584dfbb9432SGeorge Wilson spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; 1585dfbb9432SGeorge Wilson 1586dfbb9432SGeorge Wilson /* 1587dfbb9432SGeorge Wilson * Return the most recently synced out guid unless we're 1588dfbb9432SGeorge Wilson * in syncing context. 1589dfbb9432SGeorge Wilson */ 1590dfbb9432SGeorge Wilson if (dp && dsl_pool_sync_context(dp)) 1591b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1592b5989ec7Seschrock else 1593dfbb9432SGeorge Wilson return (guid); 1594e9103aaeSGarrett D'Amore } 1595e9103aaeSGarrett D'Amore 1596e9103aaeSGarrett D'Amore uint64_t 1597e9103aaeSGarrett D'Amore spa_load_guid(spa_t *spa) 1598e9103aaeSGarrett D'Amore { 1599e9103aaeSGarrett D'Amore /* 1600e9103aaeSGarrett D'Amore * This is a GUID that exists solely as a reference for the 1601e9103aaeSGarrett D'Amore * purposes of the arc. It is generated at load time, and 1602e9103aaeSGarrett D'Amore * is never written to persistent storage. 1603e9103aaeSGarrett D'Amore */ 1604e9103aaeSGarrett D'Amore return (spa->spa_load_guid); 1605fa9e4066Sahrens } 1606fa9e4066Sahrens 1607fa9e4066Sahrens uint64_t 1608fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1609fa9e4066Sahrens { 1610fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1611fa9e4066Sahrens } 1612fa9e4066Sahrens 1613fa9e4066Sahrens uint64_t 1614fa9e4066Sahrens spa_first_txg(spa_t *spa) 1615fa9e4066Sahrens { 1616fa9e4066Sahrens return (spa->spa_first_txg); 1617fa9e4066Sahrens } 1618fa9e4066Sahrens 1619b24ab676SJeff Bonwick uint64_t 1620b24ab676SJeff Bonwick spa_syncing_txg(spa_t *spa) 1621b24ab676SJeff Bonwick { 1622b24ab676SJeff Bonwick return (spa->spa_syncing_txg); 1623b24ab676SJeff Bonwick } 1624b24ab676SJeff Bonwick 16253991b535SGeorge Wilson /* 16263991b535SGeorge Wilson * Return the last txg where data can be dirtied. The final txgs 16273991b535SGeorge Wilson * will be used to just clear out any deferred frees that remain. 16283991b535SGeorge Wilson */ 16293991b535SGeorge Wilson uint64_t 16303991b535SGeorge Wilson spa_final_dirty_txg(spa_t *spa) 16313991b535SGeorge Wilson { 16323991b535SGeorge Wilson return (spa->spa_final_txg - TXG_DEFER_SIZE); 16333991b535SGeorge Wilson } 16343991b535SGeorge Wilson 163588b7b0f2SMatthew Ahrens pool_state_t 1636fa9e4066Sahrens spa_state(spa_t *spa) 1637fa9e4066Sahrens { 1638fa9e4066Sahrens return (spa->spa_state); 1639fa9e4066Sahrens } 1640fa9e4066Sahrens 1641b16da2e2SGeorge Wilson spa_load_state_t 1642b16da2e2SGeorge Wilson spa_load_state(spa_t *spa) 1643b16da2e2SGeorge Wilson { 1644b16da2e2SGeorge Wilson return (spa->spa_load_state); 1645b16da2e2SGeorge Wilson } 1646b16da2e2SGeorge Wilson 1647fa9e4066Sahrens uint64_t 1648fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1649fa9e4066Sahrens { 1650fa9e4066Sahrens return (spa->spa_freeze_txg); 1651fa9e4066Sahrens } 1652fa9e4066Sahrens 1653fa9e4066Sahrens /* ARGSUSED */ 1654fa9e4066Sahrens uint64_t 165561e255ceSMatthew Ahrens spa_get_worst_case_asize(spa_t *spa, uint64_t lsize) 1656fa9e4066Sahrens { 165769962b56SMatthew Ahrens return (lsize * spa_asize_inflation); 165844cd46caSbillm } 165944cd46caSbillm 16607d46dc6cSMatthew Ahrens /* 16617d46dc6cSMatthew Ahrens * Return the amount of slop space in bytes. It is 1/32 of the pool (3.2%), 16624b5c8e93SMatthew Ahrens * or at least 128MB, unless that would cause it to be more than half the 16634b5c8e93SMatthew Ahrens * pool size. 16647d46dc6cSMatthew Ahrens * 16657d46dc6cSMatthew Ahrens * See the comment above spa_slop_shift for details. 16667d46dc6cSMatthew Ahrens */ 16677d46dc6cSMatthew Ahrens uint64_t 16684b5c8e93SMatthew Ahrens spa_get_slop_space(spa_t *spa) 16694b5c8e93SMatthew Ahrens { 16707d46dc6cSMatthew Ahrens uint64_t space = spa_get_dspace(spa); 16714b5c8e93SMatthew Ahrens return (MAX(space >> spa_slop_shift, MIN(space >> 1, spa_min_slop))); 16727d46dc6cSMatthew Ahrens } 16737d46dc6cSMatthew Ahrens 1674485bbbf5SGeorge Wilson uint64_t 1675485bbbf5SGeorge Wilson spa_get_dspace(spa_t *spa) 1676485bbbf5SGeorge Wilson { 1677485bbbf5SGeorge Wilson return (spa->spa_dspace); 1678485bbbf5SGeorge Wilson } 1679485bbbf5SGeorge Wilson 1680485bbbf5SGeorge Wilson void 1681485bbbf5SGeorge Wilson spa_update_dspace(spa_t *spa) 1682485bbbf5SGeorge Wilson { 1683485bbbf5SGeorge Wilson spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + 1684485bbbf5SGeorge Wilson ddt_get_dedup_dspace(spa); 1685485bbbf5SGeorge Wilson } 1686485bbbf5SGeorge Wilson 16870a4e9518Sgw /* 16880a4e9518Sgw * Return the failure mode that has been set to this pool. The default 16890a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 16900a4e9518Sgw */ 16910a4e9518Sgw uint8_t 16920a4e9518Sgw spa_get_failmode(spa_t *spa) 16930a4e9518Sgw { 16940a4e9518Sgw return (spa->spa_failmode); 16950a4e9518Sgw } 16960a4e9518Sgw 1697e14bb325SJeff Bonwick boolean_t 1698e14bb325SJeff Bonwick spa_suspended(spa_t *spa) 1699e14bb325SJeff Bonwick { 1700e14bb325SJeff Bonwick return (spa->spa_suspended); 1701e14bb325SJeff Bonwick } 1702e14bb325SJeff Bonwick 170344cd46caSbillm uint64_t 170444cd46caSbillm spa_version(spa_t *spa) 170544cd46caSbillm { 170644cd46caSbillm return (spa->spa_ubsync.ub_version); 170744cd46caSbillm } 170844cd46caSbillm 1709b24ab676SJeff Bonwick boolean_t 1710b24ab676SJeff Bonwick spa_deflate(spa_t *spa) 1711b24ab676SJeff Bonwick { 1712b24ab676SJeff Bonwick return (spa->spa_deflate); 1713b24ab676SJeff Bonwick } 1714b24ab676SJeff Bonwick 1715b24ab676SJeff Bonwick metaslab_class_t * 1716b24ab676SJeff Bonwick spa_normal_class(spa_t *spa) 1717b24ab676SJeff Bonwick { 1718b24ab676SJeff Bonwick return (spa->spa_normal_class); 1719b24ab676SJeff Bonwick } 1720b24ab676SJeff Bonwick 1721b24ab676SJeff Bonwick metaslab_class_t * 1722b24ab676SJeff Bonwick spa_log_class(spa_t *spa) 1723b24ab676SJeff Bonwick { 1724b24ab676SJeff Bonwick return (spa->spa_log_class); 1725b24ab676SJeff Bonwick } 1726b24ab676SJeff Bonwick 1727bc9014e6SJustin Gibbs void 1728bc9014e6SJustin Gibbs spa_evicting_os_register(spa_t *spa, objset_t *os) 1729bc9014e6SJustin Gibbs { 1730bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1731bc9014e6SJustin Gibbs list_insert_head(&spa->spa_evicting_os_list, os); 1732bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1733bc9014e6SJustin Gibbs } 1734bc9014e6SJustin Gibbs 1735bc9014e6SJustin Gibbs void 1736bc9014e6SJustin Gibbs spa_evicting_os_deregister(spa_t *spa, objset_t *os) 1737bc9014e6SJustin Gibbs { 1738bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1739bc9014e6SJustin Gibbs list_remove(&spa->spa_evicting_os_list, os); 1740bc9014e6SJustin Gibbs cv_broadcast(&spa->spa_evicting_os_cv); 1741bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1742bc9014e6SJustin Gibbs } 1743bc9014e6SJustin Gibbs 1744bc9014e6SJustin Gibbs void 1745bc9014e6SJustin Gibbs spa_evicting_os_wait(spa_t *spa) 1746bc9014e6SJustin Gibbs { 1747bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1748bc9014e6SJustin Gibbs while (!list_is_empty(&spa->spa_evicting_os_list)) 1749bc9014e6SJustin Gibbs cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock); 1750bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1751bc9014e6SJustin Gibbs 1752bc9014e6SJustin Gibbs dmu_buf_user_evict_wait(); 1753bc9014e6SJustin Gibbs } 1754bc9014e6SJustin Gibbs 175544cd46caSbillm int 175644cd46caSbillm spa_max_replication(spa_t *spa) 175744cd46caSbillm { 175844cd46caSbillm /* 1759e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 176044cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 176144cd46caSbillm * replication level accordingly. 1762fa9e4066Sahrens */ 1763e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 176444cd46caSbillm return (1); 176544cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1766fa9e4066Sahrens } 1767fa9e4066Sahrens 17683f9d6ad7SLin Ling int 17693f9d6ad7SLin Ling spa_prev_software_version(spa_t *spa) 17703f9d6ad7SLin Ling { 17713f9d6ad7SLin Ling return (spa->spa_prev_software_version); 17723f9d6ad7SLin Ling } 17733f9d6ad7SLin Ling 1774283b8460SGeorge.Wilson uint64_t 1775283b8460SGeorge.Wilson spa_deadman_synctime(spa_t *spa) 1776283b8460SGeorge.Wilson { 1777283b8460SGeorge.Wilson return (spa->spa_deadman_synctime); 1778283b8460SGeorge.Wilson } 1779283b8460SGeorge.Wilson 178099653d4eSeschrock uint64_t 1781b24ab676SJeff Bonwick dva_get_dsize_sync(spa_t *spa, const dva_t *dva) 178299653d4eSeschrock { 1783b24ab676SJeff Bonwick uint64_t asize = DVA_GET_ASIZE(dva); 1784b24ab676SJeff Bonwick uint64_t dsize = asize; 178599653d4eSeschrock 1786b24ab676SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); 178799653d4eSeschrock 1788b24ab676SJeff Bonwick if (asize != 0 && spa->spa_deflate) { 1789b24ab676SJeff Bonwick vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); 1790b24ab676SJeff Bonwick dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 179199653d4eSeschrock } 1792b24ab676SJeff Bonwick 1793b24ab676SJeff Bonwick return (dsize); 1794b24ab676SJeff Bonwick } 1795b24ab676SJeff Bonwick 1796b24ab676SJeff Bonwick uint64_t 1797b24ab676SJeff Bonwick bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) 1798b24ab676SJeff Bonwick { 1799b24ab676SJeff Bonwick uint64_t dsize = 0; 1800b24ab676SJeff Bonwick 18015d7b4d43SMatthew Ahrens for (int d = 0; d < BP_GET_NDVAS(bp); d++) 1802b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1803b24ab676SJeff Bonwick 1804b24ab676SJeff Bonwick return (dsize); 1805b24ab676SJeff Bonwick } 1806b24ab676SJeff Bonwick 1807b24ab676SJeff Bonwick uint64_t 1808b24ab676SJeff Bonwick bp_get_dsize(spa_t *spa, const blkptr_t *bp) 1809b24ab676SJeff Bonwick { 1810b24ab676SJeff Bonwick uint64_t dsize = 0; 1811b24ab676SJeff Bonwick 1812b24ab676SJeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 1813b24ab676SJeff Bonwick 18145d7b4d43SMatthew Ahrens for (int d = 0; d < BP_GET_NDVAS(bp); d++) 1815b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1816b24ab676SJeff Bonwick 1817e14bb325SJeff Bonwick spa_config_exit(spa, SCL_VDEV, FTAG); 1818b24ab676SJeff Bonwick 1819b24ab676SJeff Bonwick return (dsize); 182099653d4eSeschrock } 182199653d4eSeschrock 1822fa9e4066Sahrens /* 1823fa9e4066Sahrens * ========================================================================== 1824fa9e4066Sahrens * Initialization and Termination 1825fa9e4066Sahrens * ========================================================================== 1826fa9e4066Sahrens */ 1827fa9e4066Sahrens 1828fa9e4066Sahrens static int 1829fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1830fa9e4066Sahrens { 1831fa9e4066Sahrens const spa_t *s1 = a1; 1832fa9e4066Sahrens const spa_t *s2 = a2; 1833fa9e4066Sahrens int s; 1834fa9e4066Sahrens 1835fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1836fa9e4066Sahrens if (s > 0) 1837fa9e4066Sahrens return (1); 1838fa9e4066Sahrens if (s < 0) 1839fa9e4066Sahrens return (-1); 1840fa9e4066Sahrens return (0); 1841fa9e4066Sahrens } 1842fa9e4066Sahrens 18430373e76bSbonwick int 18440373e76bSbonwick spa_busy(void) 18450373e76bSbonwick { 18460373e76bSbonwick return (spa_active_count); 18470373e76bSbonwick } 18480373e76bSbonwick 1849e7cbe64fSgw void 1850e7cbe64fSgw spa_boot_init() 1851e7cbe64fSgw { 1852e7cbe64fSgw spa_config_load(); 1853e7cbe64fSgw } 1854e7cbe64fSgw 1855fa9e4066Sahrens void 1856fa9e4066Sahrens spa_init(int mode) 1857fa9e4066Sahrens { 1858fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1859c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1860fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 1861fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1862fa9e4066Sahrens 1863fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1864fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1865fa9e4066Sahrens 1866fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 1867fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 1868fa94a07fSbrendan 1869fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 1870fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 187199653d4eSeschrock 18728ad4d6ddSJeff Bonwick spa_mode_global = mode; 1873fa9e4066Sahrens 1874283b8460SGeorge.Wilson #ifdef _KERNEL 1875283b8460SGeorge.Wilson spa_arch_init(); 1876283b8460SGeorge.Wilson #else 1877cd1c8b85SMatthew Ahrens if (spa_mode_global != FREAD && dprintf_find_string("watch")) { 1878cd1c8b85SMatthew Ahrens arc_procfd = open("/proc/self/ctl", O_WRONLY); 1879cd1c8b85SMatthew Ahrens if (arc_procfd == -1) { 1880cd1c8b85SMatthew Ahrens perror("could not enable watchpoints: " 1881cd1c8b85SMatthew Ahrens "opening /proc/self/ctl failed: "); 1882cd1c8b85SMatthew Ahrens } else { 1883cd1c8b85SMatthew Ahrens arc_watch = B_TRUE; 1884cd1c8b85SMatthew Ahrens } 1885cd1c8b85SMatthew Ahrens } 1886cd1c8b85SMatthew Ahrens #endif 1887cd1c8b85SMatthew Ahrens 1888fa9e4066Sahrens refcount_init(); 1889fa9e4066Sahrens unique_init(); 18900713e232SGeorge Wilson range_tree_init(); 18918363e80aSGeorge Wilson metaslab_alloc_trace_init(); 1892fa9e4066Sahrens zio_init(); 1893fa9e4066Sahrens dmu_init(); 1894fa9e4066Sahrens zil_init(); 189587db74c1Sek vdev_cache_stat_init(); 189691ebeef5Sahrens zfs_prop_init(); 1897990b4856Slling zpool_prop_init(); 1898ad135b5dSChristopher Siden zpool_feature_init(); 1899fa9e4066Sahrens spa_config_load(); 1900e14bb325SJeff Bonwick l2arc_start(); 1901fa9e4066Sahrens } 1902fa9e4066Sahrens 1903fa9e4066Sahrens void 1904fa9e4066Sahrens spa_fini(void) 1905fa9e4066Sahrens { 1906e14bb325SJeff Bonwick l2arc_stop(); 1907e14bb325SJeff Bonwick 1908fa9e4066Sahrens spa_evict_all(); 1909fa9e4066Sahrens 191087db74c1Sek vdev_cache_stat_fini(); 1911fa9e4066Sahrens zil_fini(); 1912fa9e4066Sahrens dmu_fini(); 1913fa9e4066Sahrens zio_fini(); 19148363e80aSGeorge Wilson metaslab_alloc_trace_fini(); 19150713e232SGeorge Wilson range_tree_fini(); 191691ebeef5Sahrens unique_fini(); 1917fa9e4066Sahrens refcount_fini(); 1918fa9e4066Sahrens 1919fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 192099653d4eSeschrock avl_destroy(&spa_spare_avl); 1921fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 1922fa9e4066Sahrens 1923fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1924fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1925c25056deSgw mutex_destroy(&spa_spare_lock); 1926fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 1927fa9e4066Sahrens } 19286ce0521aSperrin 19296ce0521aSperrin /* 19306ce0521aSperrin * Return whether this pool has slogs. No locking needed. 19316ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 19326ce0521aSperrin * performance and not correctness 19336ce0521aSperrin */ 19346ce0521aSperrin boolean_t 19356ce0521aSperrin spa_has_slogs(spa_t *spa) 19366ce0521aSperrin { 19376ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 19386ce0521aSperrin } 1939bf82a41bSeschrock 1940b24ab676SJeff Bonwick spa_log_state_t 1941b24ab676SJeff Bonwick spa_get_log_state(spa_t *spa) 1942b24ab676SJeff Bonwick { 1943b24ab676SJeff Bonwick return (spa->spa_log_state); 1944b24ab676SJeff Bonwick } 1945b24ab676SJeff Bonwick 1946b24ab676SJeff Bonwick void 1947b24ab676SJeff Bonwick spa_set_log_state(spa_t *spa, spa_log_state_t state) 1948b24ab676SJeff Bonwick { 1949b24ab676SJeff Bonwick spa->spa_log_state = state; 1950b24ab676SJeff Bonwick } 1951b24ab676SJeff Bonwick 1952bf82a41bSeschrock boolean_t 1953bf82a41bSeschrock spa_is_root(spa_t *spa) 1954bf82a41bSeschrock { 1955bf82a41bSeschrock return (spa->spa_is_root); 1956bf82a41bSeschrock } 19578ad4d6ddSJeff Bonwick 19588ad4d6ddSJeff Bonwick boolean_t 19598ad4d6ddSJeff Bonwick spa_writeable(spa_t *spa) 19608ad4d6ddSJeff Bonwick { 19618ad4d6ddSJeff Bonwick return (!!(spa->spa_mode & FWRITE)); 19628ad4d6ddSJeff Bonwick } 19638ad4d6ddSJeff Bonwick 196473527f44SAlex Reece /* 196573527f44SAlex Reece * Returns true if there is a pending sync task in any of the current 196673527f44SAlex Reece * syncing txg, the current quiescing txg, or the current open txg. 196773527f44SAlex Reece */ 196873527f44SAlex Reece boolean_t 196973527f44SAlex Reece spa_has_pending_synctask(spa_t *spa) 197073527f44SAlex Reece { 197173527f44SAlex Reece return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks)); 197273527f44SAlex Reece } 197373527f44SAlex Reece 19748ad4d6ddSJeff Bonwick int 19758ad4d6ddSJeff Bonwick spa_mode(spa_t *spa) 19768ad4d6ddSJeff Bonwick { 19778ad4d6ddSJeff Bonwick return (spa->spa_mode); 19788ad4d6ddSJeff Bonwick } 1979b24ab676SJeff Bonwick 1980b24ab676SJeff Bonwick uint64_t 1981b24ab676SJeff Bonwick spa_bootfs(spa_t *spa) 1982b24ab676SJeff Bonwick { 1983b24ab676SJeff Bonwick return (spa->spa_bootfs); 1984b24ab676SJeff Bonwick } 1985b24ab676SJeff Bonwick 1986b24ab676SJeff Bonwick uint64_t 1987b24ab676SJeff Bonwick spa_delegation(spa_t *spa) 1988b24ab676SJeff Bonwick { 1989b24ab676SJeff Bonwick return (spa->spa_delegation); 1990b24ab676SJeff Bonwick } 1991b24ab676SJeff Bonwick 1992b24ab676SJeff Bonwick objset_t * 1993b24ab676SJeff Bonwick spa_meta_objset(spa_t *spa) 1994b24ab676SJeff Bonwick { 1995b24ab676SJeff Bonwick return (spa->spa_meta_objset); 1996b24ab676SJeff Bonwick } 1997b24ab676SJeff Bonwick 1998b24ab676SJeff Bonwick enum zio_checksum 1999b24ab676SJeff Bonwick spa_dedup_checksum(spa_t *spa) 2000b24ab676SJeff Bonwick { 2001b24ab676SJeff Bonwick return (spa->spa_dedup_checksum); 2002b24ab676SJeff Bonwick } 20033f9d6ad7SLin Ling 20043f9d6ad7SLin Ling /* 20053f9d6ad7SLin Ling * Reset pool scan stat per scan pass (or reboot). 20063f9d6ad7SLin Ling */ 20073f9d6ad7SLin Ling void 20083f9d6ad7SLin Ling spa_scan_stat_init(spa_t *spa) 20093f9d6ad7SLin Ling { 20103f9d6ad7SLin Ling /* data not stored on disk */ 20113f9d6ad7SLin Ling spa->spa_scan_pass_start = gethrestime_sec(); 2012*1702cce7SAlek Pinchuk if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan)) 2013*1702cce7SAlek Pinchuk spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start; 2014*1702cce7SAlek Pinchuk else 2015*1702cce7SAlek Pinchuk spa->spa_scan_pass_scrub_pause = 0; 2016*1702cce7SAlek Pinchuk spa->spa_scan_pass_scrub_spent_paused = 0; 20173f9d6ad7SLin Ling spa->spa_scan_pass_exam = 0; 20183f9d6ad7SLin Ling vdev_scan_stat_init(spa->spa_root_vdev); 20193f9d6ad7SLin Ling } 20203f9d6ad7SLin Ling 20213f9d6ad7SLin Ling /* 20223f9d6ad7SLin Ling * Get scan stats for zpool status reports 20233f9d6ad7SLin Ling */ 20243f9d6ad7SLin Ling int 20253f9d6ad7SLin Ling spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) 20263f9d6ad7SLin Ling { 20273f9d6ad7SLin Ling dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; 20283f9d6ad7SLin Ling 20293f9d6ad7SLin Ling if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) 2030be6fd75aSMatthew Ahrens return (SET_ERROR(ENOENT)); 20313f9d6ad7SLin Ling bzero(ps, sizeof (pool_scan_stat_t)); 20323f9d6ad7SLin Ling 20333f9d6ad7SLin Ling /* data stored on disk */ 20343f9d6ad7SLin Ling ps->pss_func = scn->scn_phys.scn_func; 20353f9d6ad7SLin Ling ps->pss_start_time = scn->scn_phys.scn_start_time; 20363f9d6ad7SLin Ling ps->pss_end_time = scn->scn_phys.scn_end_time; 20373f9d6ad7SLin Ling ps->pss_to_examine = scn->scn_phys.scn_to_examine; 20383f9d6ad7SLin Ling ps->pss_examined = scn->scn_phys.scn_examined; 20393f9d6ad7SLin Ling ps->pss_to_process = scn->scn_phys.scn_to_process; 20403f9d6ad7SLin Ling ps->pss_processed = scn->scn_phys.scn_processed; 20413f9d6ad7SLin Ling ps->pss_errors = scn->scn_phys.scn_errors; 20423f9d6ad7SLin Ling ps->pss_state = scn->scn_phys.scn_state; 20433f9d6ad7SLin Ling 20443f9d6ad7SLin Ling /* data not stored on disk */ 20453f9d6ad7SLin Ling ps->pss_pass_start = spa->spa_scan_pass_start; 20463f9d6ad7SLin Ling ps->pss_pass_exam = spa->spa_scan_pass_exam; 2047*1702cce7SAlek Pinchuk ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause; 2048*1702cce7SAlek Pinchuk ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused; 20493f9d6ad7SLin Ling 20503f9d6ad7SLin Ling return (0); 20513f9d6ad7SLin Ling } 205209c9d376SGeorge Wilson 205309c9d376SGeorge Wilson boolean_t 205409c9d376SGeorge Wilson spa_debug_enabled(spa_t *spa) 205509c9d376SGeorge Wilson { 205609c9d376SGeorge Wilson return (spa->spa_debug); 205709c9d376SGeorge Wilson } 2058b5152584SMatthew Ahrens 2059b5152584SMatthew Ahrens int 2060b5152584SMatthew Ahrens spa_maxblocksize(spa_t *spa) 2061b5152584SMatthew Ahrens { 2062b5152584SMatthew Ahrens if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) 2063b5152584SMatthew Ahrens return (SPA_MAXBLOCKSIZE); 2064b5152584SMatthew Ahrens else 2065b5152584SMatthew Ahrens return (SPA_OLD_MAXBLOCKSIZE); 2066b5152584SMatthew Ahrens } 2067