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. 23f78cdc34SPaul Dagnelie * Copyright (c) 2011, 2018 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] 281702cce7SAlek 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> 42094e47e9SGeorge Wilson #include <sys/vdev_initialize.h> 43fa9e4066Sahrens #include <sys/metaslab.h> 44fa9e4066Sahrens #include <sys/uberblock_impl.h> 45fa9e4066Sahrens #include <sys/txg.h> 46fa9e4066Sahrens #include <sys/avl.h> 47fa9e4066Sahrens #include <sys/unique.h> 48fa9e4066Sahrens #include <sys/dsl_pool.h> 49fa9e4066Sahrens #include <sys/dsl_dir.h> 50fa9e4066Sahrens #include <sys/dsl_prop.h> 513f9d6ad7SLin Ling #include <sys/dsl_scan.h> 52fa9e4066Sahrens #include <sys/fs/zfs.h> 536ce0521aSperrin #include <sys/metaslab_impl.h> 54e14bb325SJeff Bonwick #include <sys/arc.h> 55485bbbf5SGeorge Wilson #include <sys/ddt.h> 5691ebeef5Sahrens #include "zfs_prop.h" 5745818ee1SMatthew Ahrens #include <sys/zfeature.h> 58fa9e4066Sahrens 59fa9e4066Sahrens /* 60fa9e4066Sahrens * SPA locking 61fa9e4066Sahrens * 62fa9e4066Sahrens * There are four basic locks for managing spa_t structures: 63fa9e4066Sahrens * 64fa9e4066Sahrens * spa_namespace_lock (global mutex) 65fa9e4066Sahrens * 6644cd46caSbillm * This lock must be acquired to do any of the following: 67fa9e4066Sahrens * 6844cd46caSbillm * - Lookup a spa_t by name 6944cd46caSbillm * - Add or remove a spa_t from the namespace 7044cd46caSbillm * - Increase spa_refcount from non-zero 7144cd46caSbillm * - Check if spa_refcount is zero 7244cd46caSbillm * - Rename a spa_t 73ea8dc4b6Seschrock * - add/remove/attach/detach devices 7444cd46caSbillm * - Held for the duration of create/destroy/import/export 75fa9e4066Sahrens * 7644cd46caSbillm * It does not need to handle recursion. A create or destroy may 7744cd46caSbillm * reference objects (files or zvols) in other pools, but by 7844cd46caSbillm * definition they must have an existing reference, and will never need 7944cd46caSbillm * to lookup a spa_t by name. 80fa9e4066Sahrens * 81*e914ace2STim Schumacher * spa_refcount (per-spa zfs_refcount_t protected by mutex) 82fa9e4066Sahrens * 8344cd46caSbillm * This reference count keep track of any active users of the spa_t. The 8444cd46caSbillm * spa_t cannot be destroyed or freed while this is non-zero. Internally, 8544cd46caSbillm * the refcount is never really 'zero' - opening a pool implicitly keeps 86088f3894Sahrens * some references in the DMU. Internally we check against spa_minref, but 8744cd46caSbillm * present the image of a zero/non-zero value to consumers. 88fa9e4066Sahrens * 89e14bb325SJeff Bonwick * spa_config_lock[] (per-spa array of rwlocks) 90fa9e4066Sahrens * 9191ebeef5Sahrens * This protects the spa_t from config changes, and must be held in 9291ebeef5Sahrens * the following circumstances: 93fa9e4066Sahrens * 9444cd46caSbillm * - RW_READER to perform I/O to the spa 9544cd46caSbillm * - RW_WRITER to change the vdev config 96fa9e4066Sahrens * 97fa9e4066Sahrens * The locking order is fairly straightforward: 98fa9e4066Sahrens * 9944cd46caSbillm * spa_namespace_lock -> spa_refcount 100fa9e4066Sahrens * 10144cd46caSbillm * The namespace lock must be acquired to increase the refcount from 0 10244cd46caSbillm * or to check if it is zero. 103fa9e4066Sahrens * 104e14bb325SJeff Bonwick * spa_refcount -> spa_config_lock[] 105fa9e4066Sahrens * 10644cd46caSbillm * There must be at least one valid reference on the spa_t to acquire 10744cd46caSbillm * the config lock. 108fa9e4066Sahrens * 109e14bb325SJeff Bonwick * spa_namespace_lock -> spa_config_lock[] 110fa9e4066Sahrens * 11144cd46caSbillm * The namespace lock must always be taken before the config lock. 112fa9e4066Sahrens * 113fa9e4066Sahrens * 114e14bb325SJeff Bonwick * The spa_namespace_lock can be acquired directly and is globally visible. 115fa9e4066Sahrens * 116e14bb325SJeff Bonwick * The namespace is manipulated using the following functions, all of which 117e14bb325SJeff Bonwick * require the spa_namespace_lock to be held. 118fa9e4066Sahrens * 11944cd46caSbillm * spa_lookup() Lookup a spa_t by name. 120fa9e4066Sahrens * 12144cd46caSbillm * spa_add() Create a new spa_t in the namespace. 122fa9e4066Sahrens * 12344cd46caSbillm * spa_remove() Remove a spa_t from the namespace. This also 12444cd46caSbillm * frees up any memory associated with the spa_t. 125fa9e4066Sahrens * 12644cd46caSbillm * spa_next() Returns the next spa_t in the system, or the 12744cd46caSbillm * first if NULL is passed. 128fa9e4066Sahrens * 12944cd46caSbillm * spa_evict_all() Shutdown and remove all spa_t structures in 13044cd46caSbillm * the system. 131fa9e4066Sahrens * 132ea8dc4b6Seschrock * spa_guid_exists() Determine whether a pool/device guid exists. 133fa9e4066Sahrens * 134fa9e4066Sahrens * The spa_refcount is manipulated using the following functions: 135fa9e4066Sahrens * 13644cd46caSbillm * spa_open_ref() Adds a reference to the given spa_t. Must be 13744cd46caSbillm * called with spa_namespace_lock held if the 13844cd46caSbillm * refcount is currently zero. 139fa9e4066Sahrens * 14044cd46caSbillm * spa_close() Remove a reference from the spa_t. This will 14144cd46caSbillm * not free the spa_t or remove it from the 14244cd46caSbillm * namespace. No locking is required. 143fa9e4066Sahrens * 14444cd46caSbillm * spa_refcount_zero() Returns true if the refcount is currently 14544cd46caSbillm * zero. Must be called with spa_namespace_lock 14644cd46caSbillm * held. 147fa9e4066Sahrens * 148e14bb325SJeff Bonwick * The spa_config_lock[] is an array of rwlocks, ordered as follows: 149e14bb325SJeff Bonwick * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. 150e14bb325SJeff Bonwick * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). 151e14bb325SJeff Bonwick * 152e14bb325SJeff Bonwick * To read the configuration, it suffices to hold one of these locks as reader. 153e14bb325SJeff Bonwick * To modify the configuration, you must hold all locks as writer. To modify 154e14bb325SJeff Bonwick * vdev state without altering the vdev tree's topology (e.g. online/offline), 155e14bb325SJeff Bonwick * you must hold SCL_STATE and SCL_ZIO as writer. 156e14bb325SJeff Bonwick * 157e14bb325SJeff Bonwick * We use these distinct config locks to avoid recursive lock entry. 158e14bb325SJeff Bonwick * For example, spa_sync() (which holds SCL_CONFIG as reader) induces 159e14bb325SJeff Bonwick * block allocations (SCL_ALLOC), which may require reading space maps 160e14bb325SJeff Bonwick * from disk (dmu_read() -> zio_read() -> SCL_ZIO). 161e14bb325SJeff Bonwick * 162e14bb325SJeff Bonwick * The spa config locks cannot be normal rwlocks because we need the 163e14bb325SJeff Bonwick * ability to hand off ownership. For example, SCL_ZIO is acquired 164e14bb325SJeff Bonwick * by the issuing thread and later released by an interrupt thread. 165e14bb325SJeff Bonwick * They do, however, obey the usual write-wanted semantics to prevent 166e14bb325SJeff Bonwick * writer (i.e. system administrator) starvation. 167e14bb325SJeff Bonwick * 168e14bb325SJeff Bonwick * The lock acquisition rules are as follows: 169e14bb325SJeff Bonwick * 170e14bb325SJeff Bonwick * SCL_CONFIG 171e14bb325SJeff Bonwick * Protects changes to the vdev tree topology, such as vdev 172e14bb325SJeff Bonwick * add/remove/attach/detach. Protects the dirty config list 173e14bb325SJeff Bonwick * (spa_config_dirty_list) and the set of spares and l2arc devices. 174e14bb325SJeff Bonwick * 175e14bb325SJeff Bonwick * SCL_STATE 176e14bb325SJeff Bonwick * Protects changes to pool state and vdev state, such as vdev 177e14bb325SJeff Bonwick * online/offline/fault/degrade/clear. Protects the dirty state list 178e14bb325SJeff Bonwick * (spa_state_dirty_list) and global pool state (spa_state). 179e14bb325SJeff Bonwick * 180e14bb325SJeff Bonwick * SCL_ALLOC 181e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 182e14bb325SJeff Bonwick * Held as reader by metaslab_alloc() and metaslab_claim(). 183e14bb325SJeff Bonwick * 184e14bb325SJeff Bonwick * SCL_ZIO 185e14bb325SJeff Bonwick * Held by bp-level zios (those which have no io_vd upon entry) 186e14bb325SJeff Bonwick * to prevent changes to the vdev tree. The bp-level zio implicitly 187e14bb325SJeff Bonwick * protects all of its vdev child zios, which do not hold SCL_ZIO. 188e14bb325SJeff Bonwick * 189e14bb325SJeff Bonwick * SCL_FREE 190e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 191e14bb325SJeff Bonwick * Held as reader by metaslab_free(). SCL_FREE is distinct from 192e14bb325SJeff Bonwick * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free 193e14bb325SJeff Bonwick * blocks in zio_done() while another i/o that holds either 194e14bb325SJeff Bonwick * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. 195e14bb325SJeff Bonwick * 196e14bb325SJeff Bonwick * SCL_VDEV 197e14bb325SJeff Bonwick * Held as reader to prevent changes to the vdev tree during trivial 198b24ab676SJeff Bonwick * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the 199e14bb325SJeff Bonwick * other locks, and lower than all of them, to ensure that it's safe 200e14bb325SJeff Bonwick * to acquire regardless of caller context. 201e14bb325SJeff Bonwick * 202e14bb325SJeff Bonwick * In addition, the following rules apply: 203e14bb325SJeff Bonwick * 204e14bb325SJeff Bonwick * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. 205e14bb325SJeff Bonwick * The lock ordering is SCL_CONFIG > spa_props_lock. 206e14bb325SJeff Bonwick * 207e14bb325SJeff Bonwick * (b) I/O operations on leaf vdevs. For any zio operation that takes 208e14bb325SJeff Bonwick * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), 209e14bb325SJeff Bonwick * or zio_write_phys() -- the caller must ensure that the config cannot 210e14bb325SJeff Bonwick * cannot change in the interim, and that the vdev cannot be reopened. 211e14bb325SJeff Bonwick * SCL_STATE as reader suffices for both. 212fa9e4066Sahrens * 213ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 214fa9e4066Sahrens * 21544cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 216ea8dc4b6Seschrock * for writing. 217fa9e4066Sahrens * 21844cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 21944cd46caSbillm * to complete, sync the updated configs to the 220ea8dc4b6Seschrock * cache, and release the namespace lock. 221fa9e4066Sahrens * 222e14bb325SJeff Bonwick * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). 223e14bb325SJeff Bonwick * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual 224e14bb325SJeff Bonwick * locking is, always, based on spa_namespace_lock and spa_config_lock[]. 225fa9e4066Sahrens */ 226fa9e4066Sahrens 227fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 228fa9e4066Sahrens kmutex_t spa_namespace_lock; 229fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 2300373e76bSbonwick static int spa_active_count; 231416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 232fa9e4066Sahrens 23399653d4eSeschrock static kmutex_t spa_spare_lock; 23439c23413Seschrock static avl_tree_t spa_spare_avl; 235fa94a07fSbrendan static kmutex_t spa_l2cache_lock; 236fa94a07fSbrendan static avl_tree_t spa_l2cache_avl; 23799653d4eSeschrock 238fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 2398ad4d6ddSJeff Bonwick int spa_mode_global; 240fa9e4066Sahrens 241fa9e4066Sahrens #ifdef ZFS_DEBUG 2425cabbc6bSPrashanth Sreenivasa /* 2435cabbc6bSPrashanth Sreenivasa * Everything except dprintf, spa, and indirect_remap is on by default 2445cabbc6bSPrashanth Sreenivasa * in debug builds. 2455cabbc6bSPrashanth Sreenivasa */ 24621f7c81cSMatthew Ahrens int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_INDIRECT_REMAP); 247fa9e4066Sahrens #else 248fa9e4066Sahrens int zfs_flags = 0; 249fa9e4066Sahrens #endif 250fa9e4066Sahrens 2510125049cSahrens /* 2520125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 2530125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 2540125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 2558b36997aSMatthew Ahrens * This should only be used as a last resort, as it typically results 2568b36997aSMatthew Ahrens * in leaked space, or worse. 2570125049cSahrens */ 2587fd05ac4SMatthew Ahrens boolean_t zfs_recover = B_FALSE; 2597fd05ac4SMatthew Ahrens 2607fd05ac4SMatthew Ahrens /* 2617fd05ac4SMatthew Ahrens * If destroy encounters an EIO while reading metadata (e.g. indirect 2627fd05ac4SMatthew Ahrens * blocks), space referenced by the missing metadata can not be freed. 2637fd05ac4SMatthew Ahrens * Normally this causes the background destroy to become "stalled", as 2647fd05ac4SMatthew Ahrens * it is unable to make forward progress. While in this stalled state, 2657fd05ac4SMatthew Ahrens * all remaining space to free from the error-encountering filesystem is 2667fd05ac4SMatthew Ahrens * "temporarily leaked". Set this flag to cause it to ignore the EIO, 2677fd05ac4SMatthew Ahrens * permanently leak the space from indirect blocks that can not be read, 2687fd05ac4SMatthew Ahrens * and continue to free everything else that it can. 2697fd05ac4SMatthew Ahrens * 2707fd05ac4SMatthew Ahrens * The default, "stalling" behavior is useful if the storage partially 2717fd05ac4SMatthew Ahrens * fails (i.e. some but not all i/os fail), and then later recovers. In 2727fd05ac4SMatthew Ahrens * this case, we will be able to continue pool operations while it is 2737fd05ac4SMatthew Ahrens * partially failed, and when it recovers, we can continue to free the 2747fd05ac4SMatthew Ahrens * space, with no leaks. However, note that this case is actually 2757fd05ac4SMatthew Ahrens * fairly rare. 2767fd05ac4SMatthew Ahrens * 2777fd05ac4SMatthew Ahrens * Typically pools either (a) fail completely (but perhaps temporarily, 2787fd05ac4SMatthew Ahrens * e.g. a top-level vdev going offline), or (b) have localized, 2797fd05ac4SMatthew Ahrens * permanent errors (e.g. disk returns the wrong data due to bit flip or 2807fd05ac4SMatthew Ahrens * firmware bug). In case (a), this setting does not matter because the 2817fd05ac4SMatthew Ahrens * pool will be suspended and the sync thread will not be able to make 2827fd05ac4SMatthew Ahrens * forward progress regardless. In case (b), because the error is 2837fd05ac4SMatthew Ahrens * permanent, the best we can do is leak the minimum amount of space, 2847fd05ac4SMatthew Ahrens * which is what setting this flag will do. Therefore, it is reasonable 2857fd05ac4SMatthew Ahrens * for this flag to normally be set, but we chose the more conservative 2867fd05ac4SMatthew Ahrens * approach of not setting it, so that there is no possibility of 2877fd05ac4SMatthew Ahrens * leaking space in the "partial temporary" failure case. 2887fd05ac4SMatthew Ahrens */ 2897fd05ac4SMatthew Ahrens boolean_t zfs_free_leak_on_eio = B_FALSE; 2900125049cSahrens 29169962b56SMatthew Ahrens /* 29269962b56SMatthew Ahrens * Expiration time in milliseconds. This value has two meanings. First it is 29369962b56SMatthew Ahrens * used to determine when the spa_deadman() logic should fire. By default the 29469962b56SMatthew Ahrens * spa_deadman() will fire if spa_sync() has not completed in 1000 seconds. 29569962b56SMatthew Ahrens * Secondly, the value determines if an I/O is considered "hung". Any I/O that 29669962b56SMatthew Ahrens * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting 29769962b56SMatthew Ahrens * in a system panic. 29869962b56SMatthew Ahrens */ 29969962b56SMatthew Ahrens uint64_t zfs_deadman_synctime_ms = 1000000ULL; 300283b8460SGeorge.Wilson 301283b8460SGeorge.Wilson /* 30269962b56SMatthew Ahrens * Check time in milliseconds. This defines the frequency at which we check 30369962b56SMatthew Ahrens * for hung I/O. 304283b8460SGeorge.Wilson */ 30569962b56SMatthew Ahrens uint64_t zfs_deadman_checktime_ms = 5000ULL; 306283b8460SGeorge.Wilson 307283b8460SGeorge.Wilson /* 308283b8460SGeorge.Wilson * Override the zfs deadman behavior via /etc/system. By default the 309283b8460SGeorge.Wilson * deadman is enabled except on VMware and sparc deployments. 310283b8460SGeorge.Wilson */ 311283b8460SGeorge.Wilson int zfs_deadman_enabled = -1; 312283b8460SGeorge.Wilson 31369962b56SMatthew Ahrens /* 31469962b56SMatthew Ahrens * The worst case is single-sector max-parity RAID-Z blocks, in which 31569962b56SMatthew Ahrens * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) 31669962b56SMatthew Ahrens * times the size; so just assume that. Add to this the fact that 31769962b56SMatthew Ahrens * we can have up to 3 DVAs per bp, and one more factor of 2 because 31869962b56SMatthew Ahrens * the block may be dittoed with up to 3 DVAs by ddt_sync(). All together, 31969962b56SMatthew Ahrens * the worst case is: 32069962b56SMatthew Ahrens * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24 32169962b56SMatthew Ahrens */ 32269962b56SMatthew Ahrens int spa_asize_inflation = 24; 323fa9e4066Sahrens 3247d46dc6cSMatthew Ahrens /* 3257d46dc6cSMatthew Ahrens * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in 3267d46dc6cSMatthew Ahrens * the pool to be consumed. This ensures that we don't run the pool 3277d46dc6cSMatthew Ahrens * completely out of space, due to unaccounted changes (e.g. to the MOS). 3287d46dc6cSMatthew Ahrens * It also limits the worst-case time to allocate space. If we have 3297d46dc6cSMatthew Ahrens * less than this amount of free space, most ZPL operations (e.g. write, 3307d46dc6cSMatthew Ahrens * create) will return ENOSPC. 3317d46dc6cSMatthew Ahrens * 3327d46dc6cSMatthew Ahrens * Certain operations (e.g. file removal, most administrative actions) can 3337d46dc6cSMatthew Ahrens * use half the slop space. They will only return ENOSPC if less than half 3347d46dc6cSMatthew Ahrens * the slop space is free. Typically, once the pool has less than the slop 3357d46dc6cSMatthew Ahrens * space free, the user will use these operations to free up space in the pool. 3367d46dc6cSMatthew Ahrens * These are the operations that call dsl_pool_adjustedsize() with the netfree 3377d46dc6cSMatthew Ahrens * argument set to TRUE. 3387d46dc6cSMatthew Ahrens * 33986714001SSerapheim Dimitropoulos * Operations that are almost guaranteed to free up space in the absence of 34086714001SSerapheim Dimitropoulos * a pool checkpoint can use up to three quarters of the slop space 34186714001SSerapheim Dimitropoulos * (e.g zfs destroy). 34286714001SSerapheim Dimitropoulos * 3437d46dc6cSMatthew Ahrens * A very restricted set of operations are always permitted, regardless of 3447d46dc6cSMatthew Ahrens * the amount of free space. These are the operations that call 34586714001SSerapheim Dimitropoulos * dsl_sync_task(ZFS_SPACE_CHECK_NONE). If these operations result in a net 34686714001SSerapheim Dimitropoulos * increase in the amount of space used, it is possible to run the pool 34786714001SSerapheim Dimitropoulos * completely out of space, causing it to be permanently read-only. 3487d46dc6cSMatthew Ahrens * 3494b5c8e93SMatthew Ahrens * Note that on very small pools, the slop space will be larger than 3504b5c8e93SMatthew Ahrens * 3.2%, in an effort to have it be at least spa_min_slop (128MB), 3514b5c8e93SMatthew Ahrens * but we never allow it to be more than half the pool size. 3524b5c8e93SMatthew Ahrens * 3537d46dc6cSMatthew Ahrens * See also the comments in zfs_space_check_t. 3547d46dc6cSMatthew Ahrens */ 3557d46dc6cSMatthew Ahrens int spa_slop_shift = 5; 3564b5c8e93SMatthew Ahrens uint64_t spa_min_slop = 128 * 1024 * 1024; 3577d46dc6cSMatthew Ahrens 358f78cdc34SPaul Dagnelie int spa_allocators = 4; 359f78cdc34SPaul Dagnelie 3603ee8c80cSPavel Zakharov /*PRINTFLIKE2*/ 3613ee8c80cSPavel Zakharov void 3623ee8c80cSPavel Zakharov spa_load_failed(spa_t *spa, const char *fmt, ...) 3633ee8c80cSPavel Zakharov { 3643ee8c80cSPavel Zakharov va_list adx; 3653ee8c80cSPavel Zakharov char buf[256]; 3663ee8c80cSPavel Zakharov 3673ee8c80cSPavel Zakharov va_start(adx, fmt); 3683ee8c80cSPavel Zakharov (void) vsnprintf(buf, sizeof (buf), fmt, adx); 3693ee8c80cSPavel Zakharov va_end(adx); 3703ee8c80cSPavel Zakharov 3716f793812SPavel Zakharov zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa->spa_name, 3726f793812SPavel Zakharov spa->spa_trust_config ? "trusted" : "untrusted", buf); 3733ee8c80cSPavel Zakharov } 3743ee8c80cSPavel Zakharov 3753ee8c80cSPavel Zakharov /*PRINTFLIKE2*/ 3763ee8c80cSPavel Zakharov void 3773ee8c80cSPavel Zakharov spa_load_note(spa_t *spa, const char *fmt, ...) 3783ee8c80cSPavel Zakharov { 3793ee8c80cSPavel Zakharov va_list adx; 3803ee8c80cSPavel Zakharov char buf[256]; 3813ee8c80cSPavel Zakharov 3823ee8c80cSPavel Zakharov va_start(adx, fmt); 3833ee8c80cSPavel Zakharov (void) vsnprintf(buf, sizeof (buf), fmt, adx); 3843ee8c80cSPavel Zakharov va_end(adx); 3853ee8c80cSPavel Zakharov 3866f793812SPavel Zakharov zfs_dbgmsg("spa_load(%s, config %s): %s", spa->spa_name, 3876f793812SPavel Zakharov spa->spa_trust_config ? "trusted" : "untrusted", buf); 3883ee8c80cSPavel Zakharov } 3893ee8c80cSPavel Zakharov 390e05725b1Sbonwick /* 391e05725b1Sbonwick * ========================================================================== 392e05725b1Sbonwick * SPA config locking 393e05725b1Sbonwick * ========================================================================== 394e05725b1Sbonwick */ 395e05725b1Sbonwick static void 396e14bb325SJeff Bonwick spa_config_lock_init(spa_t *spa) 397e14bb325SJeff Bonwick { 398e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 399e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 400e14bb325SJeff Bonwick mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); 401e14bb325SJeff Bonwick cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); 402*e914ace2STim Schumacher zfs_refcount_create_untracked(&scl->scl_count); 403e14bb325SJeff Bonwick scl->scl_writer = NULL; 404e14bb325SJeff Bonwick scl->scl_write_wanted = 0; 405e14bb325SJeff Bonwick } 406e05725b1Sbonwick } 407e05725b1Sbonwick 408e05725b1Sbonwick static void 409e14bb325SJeff Bonwick spa_config_lock_destroy(spa_t *spa) 410e14bb325SJeff Bonwick { 411e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 412e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 413e14bb325SJeff Bonwick mutex_destroy(&scl->scl_lock); 414e14bb325SJeff Bonwick cv_destroy(&scl->scl_cv); 415*e914ace2STim Schumacher zfs_refcount_destroy(&scl->scl_count); 416e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL); 417e14bb325SJeff Bonwick ASSERT(scl->scl_write_wanted == 0); 418e14bb325SJeff Bonwick } 419e14bb325SJeff Bonwick } 420e14bb325SJeff Bonwick 421e14bb325SJeff Bonwick int 422e14bb325SJeff Bonwick spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) 423e05725b1Sbonwick { 424e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 425e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 426e14bb325SJeff Bonwick if (!(locks & (1 << i))) 427e14bb325SJeff Bonwick continue; 428e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 429e14bb325SJeff Bonwick if (rw == RW_READER) { 430e14bb325SJeff Bonwick if (scl->scl_writer || scl->scl_write_wanted) { 431e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 432e495b6e6SSaso Kiselkov spa_config_exit(spa, locks & ((1 << i) - 1), 433e495b6e6SSaso Kiselkov tag); 434e14bb325SJeff Bonwick return (0); 435e14bb325SJeff Bonwick } 436e14bb325SJeff Bonwick } else { 437e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 438*e914ace2STim Schumacher if (!zfs_refcount_is_zero(&scl->scl_count)) { 439e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 440e495b6e6SSaso Kiselkov spa_config_exit(spa, locks & ((1 << i) - 1), 441e495b6e6SSaso Kiselkov tag); 442e14bb325SJeff Bonwick return (0); 443e14bb325SJeff Bonwick } 444e14bb325SJeff Bonwick scl->scl_writer = curthread; 445e14bb325SJeff Bonwick } 446*e914ace2STim Schumacher (void) zfs_refcount_add(&scl->scl_count, tag); 447e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 448e14bb325SJeff Bonwick } 449e14bb325SJeff Bonwick return (1); 450e05725b1Sbonwick } 451e05725b1Sbonwick 452e05725b1Sbonwick void 453e14bb325SJeff Bonwick spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) 454e05725b1Sbonwick { 455f64c0e34SEric Taylor int wlocks_held = 0; 456f64c0e34SEric Taylor 4573b2aab18SMatthew Ahrens ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY); 4583b2aab18SMatthew Ahrens 459e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 460e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 461f64c0e34SEric Taylor if (scl->scl_writer == curthread) 462f64c0e34SEric Taylor wlocks_held |= (1 << i); 463e14bb325SJeff Bonwick if (!(locks & (1 << i))) 464e14bb325SJeff Bonwick continue; 465e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 466e14bb325SJeff Bonwick if (rw == RW_READER) { 467e14bb325SJeff Bonwick while (scl->scl_writer || scl->scl_write_wanted) { 468e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 469e14bb325SJeff Bonwick } 470e14bb325SJeff Bonwick } else { 471e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 472*e914ace2STim Schumacher while (!zfs_refcount_is_zero(&scl->scl_count)) { 473e14bb325SJeff Bonwick scl->scl_write_wanted++; 474e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 475e14bb325SJeff Bonwick scl->scl_write_wanted--; 476e14bb325SJeff Bonwick } 477e14bb325SJeff Bonwick scl->scl_writer = curthread; 478e14bb325SJeff Bonwick } 479*e914ace2STim Schumacher (void) zfs_refcount_add(&scl->scl_count, tag); 480e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 481e05725b1Sbonwick } 4825cabbc6bSPrashanth Sreenivasa ASSERT3U(wlocks_held, <=, locks); 483e05725b1Sbonwick } 484e05725b1Sbonwick 485e05725b1Sbonwick void 486e14bb325SJeff Bonwick spa_config_exit(spa_t *spa, int locks, void *tag) 487e05725b1Sbonwick { 488e14bb325SJeff Bonwick for (int i = SCL_LOCKS - 1; i >= 0; i--) { 489e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 490e14bb325SJeff Bonwick if (!(locks & (1 << i))) 491e14bb325SJeff Bonwick continue; 492e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 493*e914ace2STim Schumacher ASSERT(!zfs_refcount_is_zero(&scl->scl_count)); 494*e914ace2STim Schumacher if (zfs_refcount_remove(&scl->scl_count, tag) == 0) { 495e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL || 496e14bb325SJeff Bonwick scl->scl_writer == curthread); 497e14bb325SJeff Bonwick scl->scl_writer = NULL; /* OK in either case */ 498e14bb325SJeff Bonwick cv_broadcast(&scl->scl_cv); 499e14bb325SJeff Bonwick } 500e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 501e05725b1Sbonwick } 502e05725b1Sbonwick } 503e05725b1Sbonwick 504e14bb325SJeff Bonwick int 505e14bb325SJeff Bonwick spa_config_held(spa_t *spa, int locks, krw_t rw) 506e05725b1Sbonwick { 507e14bb325SJeff Bonwick int locks_held = 0; 508e05725b1Sbonwick 509e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 510e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 511e14bb325SJeff Bonwick if (!(locks & (1 << i))) 512e14bb325SJeff Bonwick continue; 513*e914ace2STim Schumacher if ((rw == RW_READER && 514*e914ace2STim Schumacher !zfs_refcount_is_zero(&scl->scl_count)) || 515e14bb325SJeff Bonwick (rw == RW_WRITER && scl->scl_writer == curthread)) 516e14bb325SJeff Bonwick locks_held |= 1 << i; 517e14bb325SJeff Bonwick } 518e14bb325SJeff Bonwick 519e14bb325SJeff Bonwick return (locks_held); 520e05725b1Sbonwick } 521e05725b1Sbonwick 522fa9e4066Sahrens /* 523fa9e4066Sahrens * ========================================================================== 524fa9e4066Sahrens * SPA namespace functions 525fa9e4066Sahrens * ========================================================================== 526fa9e4066Sahrens */ 527fa9e4066Sahrens 528fa9e4066Sahrens /* 529fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 530fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 531fa9e4066Sahrens */ 532fa9e4066Sahrens spa_t * 533fa9e4066Sahrens spa_lookup(const char *name) 534fa9e4066Sahrens { 535e14bb325SJeff Bonwick static spa_t search; /* spa_t is large; don't allocate on stack */ 536e14bb325SJeff Bonwick spa_t *spa; 537fa9e4066Sahrens avl_index_t where; 53840feaa91Sahrens char *cp; 539fa9e4066Sahrens 540fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 541fa9e4066Sahrens 5423b2aab18SMatthew Ahrens (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); 5433b2aab18SMatthew Ahrens 54440feaa91Sahrens /* 54540feaa91Sahrens * If it's a full dataset name, figure out the pool name and 54640feaa91Sahrens * just use that. 54740feaa91Sahrens */ 54878f17100SMatthew Ahrens cp = strpbrk(search.spa_name, "/@#"); 5493b2aab18SMatthew Ahrens if (cp != NULL) 55040feaa91Sahrens *cp = '\0'; 55140feaa91Sahrens 552fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 553fa9e4066Sahrens 554fa9e4066Sahrens return (spa); 555fa9e4066Sahrens } 556fa9e4066Sahrens 557283b8460SGeorge.Wilson /* 558283b8460SGeorge.Wilson * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. 559283b8460SGeorge.Wilson * If the zfs_deadman_enabled flag is set then it inspects all vdev queues 560283b8460SGeorge.Wilson * looking for potentially hung I/Os. 561283b8460SGeorge.Wilson */ 562283b8460SGeorge.Wilson void 563283b8460SGeorge.Wilson spa_deadman(void *arg) 564283b8460SGeorge.Wilson { 565283b8460SGeorge.Wilson spa_t *spa = arg; 566283b8460SGeorge.Wilson 5670713e232SGeorge Wilson /* 5680713e232SGeorge Wilson * Disable the deadman timer if the pool is suspended. 5690713e232SGeorge Wilson */ 5700713e232SGeorge Wilson if (spa_suspended(spa)) { 5710713e232SGeorge Wilson VERIFY(cyclic_reprogram(spa->spa_deadman_cycid, CY_INFINITY)); 5720713e232SGeorge Wilson return; 5730713e232SGeorge Wilson } 5740713e232SGeorge Wilson 575283b8460SGeorge.Wilson zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", 576283b8460SGeorge.Wilson (gethrtime() - spa->spa_sync_starttime) / NANOSEC, 577283b8460SGeorge.Wilson ++spa->spa_deadman_calls); 578283b8460SGeorge.Wilson if (zfs_deadman_enabled) 579283b8460SGeorge.Wilson vdev_deadman(spa->spa_root_vdev); 580283b8460SGeorge.Wilson } 581283b8460SGeorge.Wilson 582fa9e4066Sahrens /* 583fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 584fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 585fa9e4066Sahrens * exist by calling spa_lookup() first. 586fa9e4066Sahrens */ 587fa9e4066Sahrens spa_t * 588468c413aSTim Haley spa_add(const char *name, nvlist_t *config, const char *altroot) 589fa9e4066Sahrens { 590fa9e4066Sahrens spa_t *spa; 591c5904d13Seschrock spa_config_dirent_t *dp; 592283b8460SGeorge.Wilson cyc_handler_t hdlr; 593283b8460SGeorge.Wilson cyc_time_t when; 594fa9e4066Sahrens 595fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 596fa9e4066Sahrens 597fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 598fa9e4066Sahrens 599c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 600c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 60135a5a358SJonathan Adams mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 602bc9014e6SJustin Gibbs mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL); 603c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 60435a5a358SJonathan Adams mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); 605c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 60645818ee1SMatthew Ahrens mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL); 60735a5a358SJonathan Adams mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 608a1521560SJeff Bonwick mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); 609a1521560SJeff Bonwick mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); 610c3a66015SMatthew Ahrens mutex_init(&spa->spa_iokstat_lock, NULL, MUTEX_DEFAULT, NULL); 611c25056deSgw 612c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 613bc9014e6SJustin Gibbs cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL); 61435a5a358SJonathan Adams cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); 615c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 616e14bb325SJeff Bonwick cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); 617c25056deSgw 618b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 619cde58dbcSMatthew Ahrens bplist_create(&spa->spa_free_bplist[t]); 620b24ab676SJeff Bonwick 621e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); 622fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 623fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 6240373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 625468c413aSTim Haley spa->spa_load_max_txg = UINT64_MAX; 62635a5a358SJonathan Adams spa->spa_proc = &p0; 62735a5a358SJonathan Adams spa->spa_proc_state = SPA_PROC_NONE; 6286f793812SPavel Zakharov spa->spa_trust_config = B_TRUE; 629fa9e4066Sahrens 630283b8460SGeorge.Wilson hdlr.cyh_func = spa_deadman; 631283b8460SGeorge.Wilson hdlr.cyh_arg = spa; 632283b8460SGeorge.Wilson hdlr.cyh_level = CY_LOW_LEVEL; 633283b8460SGeorge.Wilson 63469962b56SMatthew Ahrens spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms); 635283b8460SGeorge.Wilson 636283b8460SGeorge.Wilson /* 637283b8460SGeorge.Wilson * This determines how often we need to check for hung I/Os after 638283b8460SGeorge.Wilson * the cyclic has already fired. Since checking for hung I/Os is 639283b8460SGeorge.Wilson * an expensive operation we don't want to check too frequently. 64069962b56SMatthew Ahrens * Instead wait for 5 seconds before checking again. 641283b8460SGeorge.Wilson */ 64269962b56SMatthew Ahrens when.cyt_interval = MSEC2NSEC(zfs_deadman_checktime_ms); 643283b8460SGeorge.Wilson when.cyt_when = CY_INFINITY; 644283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 645283b8460SGeorge.Wilson spa->spa_deadman_cycid = cyclic_add(&hdlr, &when); 646283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 647283b8460SGeorge.Wilson 648*e914ace2STim Schumacher zfs_refcount_create(&spa->spa_refcount); 649e14bb325SJeff Bonwick spa_config_lock_init(spa); 650fa9e4066Sahrens 651fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 652fa9e4066Sahrens 6530373e76bSbonwick /* 6540373e76bSbonwick * Set the alternate root, if there is one. 6550373e76bSbonwick */ 6560373e76bSbonwick if (altroot) { 6570373e76bSbonwick spa->spa_root = spa_strdup(altroot); 6580373e76bSbonwick spa_active_count++; 6590373e76bSbonwick } 6600373e76bSbonwick 661f78cdc34SPaul Dagnelie spa->spa_alloc_count = spa_allocators; 662f78cdc34SPaul Dagnelie spa->spa_alloc_locks = kmem_zalloc(spa->spa_alloc_count * 663f78cdc34SPaul Dagnelie sizeof (kmutex_t), KM_SLEEP); 664f78cdc34SPaul Dagnelie spa->spa_alloc_trees = kmem_zalloc(spa->spa_alloc_count * 665f78cdc34SPaul Dagnelie sizeof (avl_tree_t), KM_SLEEP); 666f78cdc34SPaul Dagnelie for (int i = 0; i < spa->spa_alloc_count; i++) { 667f78cdc34SPaul Dagnelie mutex_init(&spa->spa_alloc_locks[i], NULL, MUTEX_DEFAULT, NULL); 668f78cdc34SPaul Dagnelie avl_create(&spa->spa_alloc_trees[i], zio_bookmark_compare, 669f78cdc34SPaul Dagnelie sizeof (zio_t), offsetof(zio_t, io_alloc_node)); 670f78cdc34SPaul Dagnelie } 6710f7643c7SGeorge Wilson 672c5904d13Seschrock /* 673c5904d13Seschrock * Every pool starts with the default cachefile 674c5904d13Seschrock */ 675c5904d13Seschrock list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), 676c5904d13Seschrock offsetof(spa_config_dirent_t, scd_link)); 677c5904d13Seschrock 678c5904d13Seschrock dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); 679ef912c80STim Haley dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); 680c5904d13Seschrock list_insert_head(&spa->spa_config_list, dp); 681c5904d13Seschrock 6824b964adaSGeorge Wilson VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, 6834b964adaSGeorge Wilson KM_SLEEP) == 0); 6844b964adaSGeorge Wilson 685ad135b5dSChristopher Siden if (config != NULL) { 686ad135b5dSChristopher Siden nvlist_t *features; 687ad135b5dSChristopher Siden 688ad135b5dSChristopher Siden if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, 689ad135b5dSChristopher Siden &features) == 0) { 690ad135b5dSChristopher Siden VERIFY(nvlist_dup(features, &spa->spa_label_features, 691ad135b5dSChristopher Siden 0) == 0); 692ad135b5dSChristopher Siden } 693ad135b5dSChristopher Siden 694468c413aSTim Haley VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); 695ad135b5dSChristopher Siden } 696ad135b5dSChristopher Siden 697ad135b5dSChristopher Siden if (spa->spa_label_features == NULL) { 698ad135b5dSChristopher Siden VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, 699ad135b5dSChristopher Siden KM_SLEEP) == 0); 700ad135b5dSChristopher Siden } 701468c413aSTim Haley 702c3a66015SMatthew Ahrens spa->spa_iokstat = kstat_create("zfs", 0, name, 703c3a66015SMatthew Ahrens "disk", KSTAT_TYPE_IO, 1, 0); 704c3a66015SMatthew Ahrens if (spa->spa_iokstat) { 705c3a66015SMatthew Ahrens spa->spa_iokstat->ks_lock = &spa->spa_iokstat_lock; 706c3a66015SMatthew Ahrens kstat_install(spa->spa_iokstat); 707c3a66015SMatthew Ahrens } 708c3a66015SMatthew Ahrens 70981cd5c55SMatthew Ahrens spa->spa_min_ashift = INT_MAX; 71081cd5c55SMatthew Ahrens spa->spa_max_ashift = 0; 71181cd5c55SMatthew Ahrens 71243466aaeSMax Grossman /* 71343466aaeSMax Grossman * As a pool is being created, treat all features as disabled by 71443466aaeSMax Grossman * setting SPA_FEATURE_DISABLED for all entries in the feature 71543466aaeSMax Grossman * refcount cache. 71643466aaeSMax Grossman */ 71743466aaeSMax Grossman for (int i = 0; i < SPA_FEATURES; i++) { 71843466aaeSMax Grossman spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; 71943466aaeSMax Grossman } 72043466aaeSMax Grossman 721fa9e4066Sahrens return (spa); 722fa9e4066Sahrens } 723fa9e4066Sahrens 724fa9e4066Sahrens /* 725fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 726fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 727fa9e4066Sahrens * deactivated. 728fa9e4066Sahrens */ 729fa9e4066Sahrens void 730fa9e4066Sahrens spa_remove(spa_t *spa) 731fa9e4066Sahrens { 732c5904d13Seschrock spa_config_dirent_t *dp; 733c5904d13Seschrock 734fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 735fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 736*e914ace2STim Schumacher ASSERT3U(zfs_refcount_count(&spa->spa_refcount), ==, 0); 737fa9e4066Sahrens 7381195e687SMark J Musante nvlist_free(spa->spa_config_splitting); 7391195e687SMark J Musante 740fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 741fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 742fa9e4066Sahrens 7430373e76bSbonwick if (spa->spa_root) { 744fa9e4066Sahrens spa_strfree(spa->spa_root); 7450373e76bSbonwick spa_active_count--; 7460373e76bSbonwick } 747fa9e4066Sahrens 748c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 749c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 750c5904d13Seschrock if (dp->scd_path != NULL) 751c5904d13Seschrock spa_strfree(dp->scd_path); 752c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 753c5904d13Seschrock } 754c5904d13Seschrock 755f78cdc34SPaul Dagnelie for (int i = 0; i < spa->spa_alloc_count; i++) { 756f78cdc34SPaul Dagnelie avl_destroy(&spa->spa_alloc_trees[i]); 757f78cdc34SPaul Dagnelie mutex_destroy(&spa->spa_alloc_locks[i]); 758f78cdc34SPaul Dagnelie } 759f78cdc34SPaul Dagnelie kmem_free(spa->spa_alloc_locks, spa->spa_alloc_count * 760f78cdc34SPaul Dagnelie sizeof (kmutex_t)); 761f78cdc34SPaul Dagnelie kmem_free(spa->spa_alloc_trees, spa->spa_alloc_count * 762f78cdc34SPaul Dagnelie sizeof (avl_tree_t)); 763f78cdc34SPaul Dagnelie 764c5904d13Seschrock list_destroy(&spa->spa_config_list); 7652f8aaab3Seschrock 766ad135b5dSChristopher Siden nvlist_free(spa->spa_label_features); 7674b964adaSGeorge Wilson nvlist_free(spa->spa_load_info); 768fa9e4066Sahrens spa_config_set(spa, NULL); 769fa9e4066Sahrens 770283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 771283b8460SGeorge.Wilson if (spa->spa_deadman_cycid != CYCLIC_NONE) 772283b8460SGeorge.Wilson cyclic_remove(spa->spa_deadman_cycid); 773283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 774283b8460SGeorge.Wilson spa->spa_deadman_cycid = CYCLIC_NONE; 775283b8460SGeorge.Wilson 776*e914ace2STim Schumacher zfs_refcount_destroy(&spa->spa_refcount); 77791ebeef5Sahrens 778e14bb325SJeff Bonwick spa_config_lock_destroy(spa); 779fa9e4066Sahrens 780c3a66015SMatthew Ahrens kstat_delete(spa->spa_iokstat); 781c3a66015SMatthew Ahrens spa->spa_iokstat = NULL; 782c3a66015SMatthew Ahrens 783b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 784cde58dbcSMatthew Ahrens bplist_destroy(&spa->spa_free_bplist[t]); 785b24ab676SJeff Bonwick 78645818ee1SMatthew Ahrens zio_checksum_templates_free(spa); 78745818ee1SMatthew Ahrens 788c25056deSgw cv_destroy(&spa->spa_async_cv); 789bc9014e6SJustin Gibbs cv_destroy(&spa->spa_evicting_os_cv); 79035a5a358SJonathan Adams cv_destroy(&spa->spa_proc_cv); 791c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 792e14bb325SJeff Bonwick cv_destroy(&spa->spa_suspend_cv); 793c25056deSgw 7945ad82045Snd mutex_destroy(&spa->spa_async_lock); 795c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 79635a5a358SJonathan Adams mutex_destroy(&spa->spa_errlog_lock); 797bc9014e6SJustin Gibbs mutex_destroy(&spa->spa_evicting_os_lock); 79806eeb2adSek mutex_destroy(&spa->spa_history_lock); 79935a5a358SJonathan Adams mutex_destroy(&spa->spa_proc_lock); 800b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 80145818ee1SMatthew Ahrens mutex_destroy(&spa->spa_cksum_tmpls_lock); 80235a5a358SJonathan Adams mutex_destroy(&spa->spa_scrub_lock); 803e14bb325SJeff Bonwick mutex_destroy(&spa->spa_suspend_lock); 804a1521560SJeff Bonwick mutex_destroy(&spa->spa_vdev_top_lock); 805c3a66015SMatthew Ahrens mutex_destroy(&spa->spa_iokstat_lock); 8065ad82045Snd 807fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 808fa9e4066Sahrens } 809fa9e4066Sahrens 810fa9e4066Sahrens /* 811fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 812fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 813fa9e4066Sahrens */ 814fa9e4066Sahrens spa_t * 815fa9e4066Sahrens spa_next(spa_t *prev) 816fa9e4066Sahrens { 817fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 818fa9e4066Sahrens 819fa9e4066Sahrens if (prev) 820fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 821fa9e4066Sahrens else 822fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 823fa9e4066Sahrens } 824fa9e4066Sahrens 825fa9e4066Sahrens /* 826fa9e4066Sahrens * ========================================================================== 827fa9e4066Sahrens * SPA refcount functions 828fa9e4066Sahrens * ========================================================================== 829fa9e4066Sahrens */ 830fa9e4066Sahrens 831fa9e4066Sahrens /* 832fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 833fa9e4066Sahrens * have the namespace lock held. 834fa9e4066Sahrens */ 835fa9e4066Sahrens void 836fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 837fa9e4066Sahrens { 838*e914ace2STim Schumacher ASSERT(zfs_refcount_count(&spa->spa_refcount) >= spa->spa_minref || 839fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 840*e914ace2STim Schumacher (void) zfs_refcount_add(&spa->spa_refcount, tag); 841fa9e4066Sahrens } 842fa9e4066Sahrens 843fa9e4066Sahrens /* 844fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 845fa9e4066Sahrens * have the namespace lock held. 846fa9e4066Sahrens */ 847fa9e4066Sahrens void 848fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 849fa9e4066Sahrens { 850*e914ace2STim Schumacher ASSERT(zfs_refcount_count(&spa->spa_refcount) > spa->spa_minref || 851fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 852*e914ace2STim Schumacher (void) zfs_refcount_remove(&spa->spa_refcount, tag); 853fa9e4066Sahrens } 854fa9e4066Sahrens 855bc9014e6SJustin Gibbs /* 856bc9014e6SJustin Gibbs * Remove a reference to the given spa_t held by a dsl dir that is 857bc9014e6SJustin Gibbs * being asynchronously released. Async releases occur from a taskq 858bc9014e6SJustin Gibbs * performing eviction of dsl datasets and dirs. The namespace lock 859bc9014e6SJustin Gibbs * isn't held and the hold by the object being evicted may contribute to 860bc9014e6SJustin Gibbs * spa_minref (e.g. dataset or directory released during pool export), 861bc9014e6SJustin Gibbs * so the asserts in spa_close() do not apply. 862bc9014e6SJustin Gibbs */ 863bc9014e6SJustin Gibbs void 864bc9014e6SJustin Gibbs spa_async_close(spa_t *spa, void *tag) 865bc9014e6SJustin Gibbs { 866*e914ace2STim Schumacher (void) zfs_refcount_remove(&spa->spa_refcount, tag); 867bc9014e6SJustin Gibbs } 868bc9014e6SJustin Gibbs 869fa9e4066Sahrens /* 870fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 871088f3894Sahrens * spa_namespace_lock held. We really compare against spa_minref, which is the 872fa9e4066Sahrens * number of references acquired when opening a pool 873fa9e4066Sahrens */ 874fa9e4066Sahrens boolean_t 875fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 876fa9e4066Sahrens { 877fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 878fa9e4066Sahrens 879*e914ace2STim Schumacher return (zfs_refcount_count(&spa->spa_refcount) == spa->spa_minref); 880fa9e4066Sahrens } 881fa9e4066Sahrens 88299653d4eSeschrock /* 88399653d4eSeschrock * ========================================================================== 884fa94a07fSbrendan * SPA spare and l2cache tracking 88599653d4eSeschrock * ========================================================================== 88699653d4eSeschrock */ 88799653d4eSeschrock 888fa94a07fSbrendan /* 889fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 890fa94a07fSbrendan * for 'auxiliary' devices. 891fa94a07fSbrendan */ 892fa94a07fSbrendan 893fa94a07fSbrendan typedef struct spa_aux { 894fa94a07fSbrendan uint64_t aux_guid; 895fa94a07fSbrendan uint64_t aux_pool; 896fa94a07fSbrendan avl_node_t aux_avl; 897fa94a07fSbrendan int aux_count; 898fa94a07fSbrendan } spa_aux_t; 899fa94a07fSbrendan 900fa94a07fSbrendan static int 901fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 902fa94a07fSbrendan { 903fa94a07fSbrendan const spa_aux_t *sa = a; 904fa94a07fSbrendan const spa_aux_t *sb = b; 905fa94a07fSbrendan 906fa94a07fSbrendan if (sa->aux_guid < sb->aux_guid) 907fa94a07fSbrendan return (-1); 908fa94a07fSbrendan else if (sa->aux_guid > sb->aux_guid) 909fa94a07fSbrendan return (1); 910fa94a07fSbrendan else 911fa94a07fSbrendan return (0); 912fa94a07fSbrendan } 913fa94a07fSbrendan 914fa94a07fSbrendan void 915fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 916fa94a07fSbrendan { 917fa94a07fSbrendan avl_index_t where; 918fa94a07fSbrendan spa_aux_t search; 919fa94a07fSbrendan spa_aux_t *aux; 920fa94a07fSbrendan 921fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 922fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 923fa94a07fSbrendan aux->aux_count++; 924fa94a07fSbrendan } else { 925fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 926fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 927fa94a07fSbrendan aux->aux_count = 1; 928fa94a07fSbrendan avl_insert(avl, aux, where); 929fa94a07fSbrendan } 930fa94a07fSbrendan } 931fa94a07fSbrendan 932fa94a07fSbrendan void 933fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 934fa94a07fSbrendan { 935fa94a07fSbrendan spa_aux_t search; 936fa94a07fSbrendan spa_aux_t *aux; 937fa94a07fSbrendan avl_index_t where; 938fa94a07fSbrendan 939fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 940fa94a07fSbrendan aux = avl_find(avl, &search, &where); 941fa94a07fSbrendan 942fa94a07fSbrendan ASSERT(aux != NULL); 943fa94a07fSbrendan 944fa94a07fSbrendan if (--aux->aux_count == 0) { 945fa94a07fSbrendan avl_remove(avl, aux); 946fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 947fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 948fa94a07fSbrendan aux->aux_pool = 0ULL; 949fa94a07fSbrendan } 950fa94a07fSbrendan } 951fa94a07fSbrendan 952fa94a07fSbrendan boolean_t 95389a89ebfSlling spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) 954fa94a07fSbrendan { 955fa94a07fSbrendan spa_aux_t search, *found; 956fa94a07fSbrendan 957fa94a07fSbrendan search.aux_guid = guid; 95889a89ebfSlling found = avl_find(avl, &search, NULL); 959fa94a07fSbrendan 960fa94a07fSbrendan if (pool) { 961fa94a07fSbrendan if (found) 962fa94a07fSbrendan *pool = found->aux_pool; 963fa94a07fSbrendan else 964fa94a07fSbrendan *pool = 0ULL; 965fa94a07fSbrendan } 966fa94a07fSbrendan 96789a89ebfSlling if (refcnt) { 96889a89ebfSlling if (found) 96989a89ebfSlling *refcnt = found->aux_count; 97089a89ebfSlling else 97189a89ebfSlling *refcnt = 0; 97289a89ebfSlling } 97389a89ebfSlling 974fa94a07fSbrendan return (found != NULL); 975fa94a07fSbrendan } 976fa94a07fSbrendan 977fa94a07fSbrendan void 978fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 979fa94a07fSbrendan { 980fa94a07fSbrendan spa_aux_t search, *found; 981fa94a07fSbrendan avl_index_t where; 982fa94a07fSbrendan 983fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 984fa94a07fSbrendan found = avl_find(avl, &search, &where); 985fa94a07fSbrendan ASSERT(found != NULL); 986fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 987fa94a07fSbrendan 988fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 989fa94a07fSbrendan } 990fa94a07fSbrendan 99199653d4eSeschrock /* 99239c23413Seschrock * Spares are tracked globally due to the following constraints: 99339c23413Seschrock * 99454811da5SToomas Soome * - A spare may be part of multiple pools. 99554811da5SToomas Soome * - A spare may be added to a pool even if it's actively in use within 99639c23413Seschrock * another pool. 99754811da5SToomas Soome * - A spare in use in any pool can only be the source of a replacement if 99839c23413Seschrock * the target is a spare in the same pool. 99939c23413Seschrock * 100039c23413Seschrock * We keep track of all spares on the system through the use of a reference 100139c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 100239c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 100339c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 100439c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 100539c23413Seschrock * pool), we also keep track of which pool its been made a part of. 100639c23413Seschrock * 100739c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 100839c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 100939c23413Seschrock * separate spare lock exists for the status query path, which does not need to 101039c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 101199653d4eSeschrock */ 101239c23413Seschrock 101399653d4eSeschrock static int 101499653d4eSeschrock spa_spare_compare(const void *a, const void *b) 101599653d4eSeschrock { 1016fa94a07fSbrendan return (spa_aux_compare(a, b)); 101799653d4eSeschrock } 101899653d4eSeschrock 101999653d4eSeschrock void 102039c23413Seschrock spa_spare_add(vdev_t *vd) 102199653d4eSeschrock { 102299653d4eSeschrock mutex_enter(&spa_spare_lock); 102339c23413Seschrock ASSERT(!vd->vdev_isspare); 1024fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 102539c23413Seschrock vd->vdev_isspare = B_TRUE; 102699653d4eSeschrock mutex_exit(&spa_spare_lock); 102799653d4eSeschrock } 102899653d4eSeschrock 102999653d4eSeschrock void 103039c23413Seschrock spa_spare_remove(vdev_t *vd) 103199653d4eSeschrock { 103299653d4eSeschrock mutex_enter(&spa_spare_lock); 103339c23413Seschrock ASSERT(vd->vdev_isspare); 1034fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 103539c23413Seschrock vd->vdev_isspare = B_FALSE; 103699653d4eSeschrock mutex_exit(&spa_spare_lock); 103799653d4eSeschrock } 103899653d4eSeschrock 103999653d4eSeschrock boolean_t 104089a89ebfSlling spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) 104199653d4eSeschrock { 1042fa94a07fSbrendan boolean_t found; 104399653d4eSeschrock 104499653d4eSeschrock mutex_enter(&spa_spare_lock); 104589a89ebfSlling found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); 104699653d4eSeschrock mutex_exit(&spa_spare_lock); 104799653d4eSeschrock 1048fa94a07fSbrendan return (found); 104939c23413Seschrock } 105039c23413Seschrock 105139c23413Seschrock void 105239c23413Seschrock spa_spare_activate(vdev_t *vd) 105339c23413Seschrock { 105439c23413Seschrock mutex_enter(&spa_spare_lock); 105539c23413Seschrock ASSERT(vd->vdev_isspare); 1056fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 1057fa94a07fSbrendan mutex_exit(&spa_spare_lock); 1058fa94a07fSbrendan } 105939c23413Seschrock 1060fa94a07fSbrendan /* 1061fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 1062fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 1063fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 1064fa94a07fSbrendan */ 106539c23413Seschrock 1066fa94a07fSbrendan static int 1067fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 1068fa94a07fSbrendan { 1069fa94a07fSbrendan return (spa_aux_compare(a, b)); 1070fa94a07fSbrendan } 1071fa94a07fSbrendan 1072fa94a07fSbrendan void 1073fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 1074fa94a07fSbrendan { 1075fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1076fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 1077fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 1078fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 1079fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1080fa94a07fSbrendan } 1081fa94a07fSbrendan 1082fa94a07fSbrendan void 1083fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 1084fa94a07fSbrendan { 1085fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1086fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 1087fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 1088fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 1089fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1090fa94a07fSbrendan } 1091fa94a07fSbrendan 1092fa94a07fSbrendan boolean_t 1093fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 1094fa94a07fSbrendan { 1095fa94a07fSbrendan boolean_t found; 1096fa94a07fSbrendan 1097fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 109889a89ebfSlling found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); 1099fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1100fa94a07fSbrendan 1101fa94a07fSbrendan return (found); 1102fa94a07fSbrendan } 1103fa94a07fSbrendan 1104fa94a07fSbrendan void 1105fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 1106fa94a07fSbrendan { 1107fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1108fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 1109fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 1110fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1111fa94a07fSbrendan } 1112fa94a07fSbrendan 1113fa9e4066Sahrens /* 1114fa9e4066Sahrens * ========================================================================== 1115fa9e4066Sahrens * SPA vdev locking 1116fa9e4066Sahrens * ========================================================================== 1117fa9e4066Sahrens */ 1118fa9e4066Sahrens 1119fa9e4066Sahrens /* 1120ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 1121ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 1122fa9e4066Sahrens * It returns the next transaction group for the spa_t. 1123fa9e4066Sahrens */ 1124fa9e4066Sahrens uint64_t 1125fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 1126fa9e4066Sahrens { 1127a1521560SJeff Bonwick mutex_enter(&spa->spa_vdev_top_lock); 1128bbfd46c4SJeff Bonwick mutex_enter(&spa_namespace_lock); 112988ecc943SGeorge Wilson return (spa_vdev_config_enter(spa)); 113088ecc943SGeorge Wilson } 113188ecc943SGeorge Wilson 113288ecc943SGeorge Wilson /* 113388ecc943SGeorge Wilson * Internal implementation for spa_vdev_enter(). Used when a vdev 113488ecc943SGeorge Wilson * operation requires multiple syncs (i.e. removing a device) while 113588ecc943SGeorge Wilson * keeping the spa_namespace_lock held. 113688ecc943SGeorge Wilson */ 113788ecc943SGeorge Wilson uint64_t 113888ecc943SGeorge Wilson spa_vdev_config_enter(spa_t *spa) 113988ecc943SGeorge Wilson { 114088ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 11413d7072f8Seschrock 1142e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1143fa9e4066Sahrens 1144fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 1145fa9e4066Sahrens } 1146fa9e4066Sahrens 1147fa9e4066Sahrens /* 114888ecc943SGeorge Wilson * Used in combination with spa_vdev_config_enter() to allow the syncing 114988ecc943SGeorge Wilson * of multiple transactions without releasing the spa_namespace_lock. 1150fa9e4066Sahrens */ 115188ecc943SGeorge Wilson void 115288ecc943SGeorge Wilson spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) 1153fa9e4066Sahrens { 115488ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 115588ecc943SGeorge Wilson 11560e34b6a7Sbonwick int config_changed = B_FALSE; 1157ea8dc4b6Seschrock 11580373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 11590e34b6a7Sbonwick 1160e14bb325SJeff Bonwick spa->spa_pending_vdev = NULL; 1161e14bb325SJeff Bonwick 11620e34b6a7Sbonwick /* 11630e34b6a7Sbonwick * Reassess the DTLs. 11640e34b6a7Sbonwick */ 11650373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 11660e34b6a7Sbonwick 1167e14bb325SJeff Bonwick if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { 11680e34b6a7Sbonwick config_changed = B_TRUE; 11698f18d1faSGeorge Wilson spa->spa_config_generation++; 11700e34b6a7Sbonwick } 1171ea8dc4b6Seschrock 117288ecc943SGeorge Wilson /* 117388ecc943SGeorge Wilson * Verify the metaslab classes. 117488ecc943SGeorge Wilson */ 1175b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); 1176b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); 117788ecc943SGeorge Wilson 1178e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1179fa9e4066Sahrens 118088ecc943SGeorge Wilson /* 118188ecc943SGeorge Wilson * Panic the system if the specified tag requires it. This 118288ecc943SGeorge Wilson * is useful for ensuring that configurations are updated 118388ecc943SGeorge Wilson * transactionally. 118488ecc943SGeorge Wilson */ 118588ecc943SGeorge Wilson if (zio_injection_enabled) 11861195e687SMark J Musante zio_handle_panic_injection(spa, tag, 0); 118788ecc943SGeorge Wilson 1188fa9e4066Sahrens /* 1189fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 1190fa9e4066Sahrens * that there won't be more than one config change per txg. 1191fa9e4066Sahrens * This allows us to use the txg as the generation number. 1192fa9e4066Sahrens */ 1193fa9e4066Sahrens if (error == 0) 1194fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 1195fa9e4066Sahrens 1196fa9e4066Sahrens if (vd != NULL) { 11970713e232SGeorge Wilson ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL); 1198094e47e9SGeorge Wilson if (vd->vdev_ops->vdev_op_leaf) { 1199094e47e9SGeorge Wilson mutex_enter(&vd->vdev_initialize_lock); 1200094e47e9SGeorge Wilson vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED); 1201094e47e9SGeorge Wilson mutex_exit(&vd->vdev_initialize_lock); 1202094e47e9SGeorge Wilson } 1203094e47e9SGeorge Wilson 12048ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1205fa9e4066Sahrens vdev_free(vd); 12068ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1207fa9e4066Sahrens } 1208fa9e4066Sahrens 1209fa9e4066Sahrens /* 12100e34b6a7Sbonwick * If the config changed, update the config cache. 1211fa9e4066Sahrens */ 12120e34b6a7Sbonwick if (config_changed) 12135cabbc6bSPrashanth Sreenivasa spa_write_cachefile(spa, B_FALSE, B_TRUE); 121488ecc943SGeorge Wilson } 1215ea8dc4b6Seschrock 121688ecc943SGeorge Wilson /* 121788ecc943SGeorge Wilson * Unlock the spa_t after adding or removing a vdev. Besides undoing the 121888ecc943SGeorge Wilson * locking of spa_vdev_enter(), we also want make sure the transactions have 121988ecc943SGeorge Wilson * synced to disk, and then update the global configuration cache with the new 122088ecc943SGeorge Wilson * information. 122188ecc943SGeorge Wilson */ 122288ecc943SGeorge Wilson int 122388ecc943SGeorge Wilson spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 122488ecc943SGeorge Wilson { 122588ecc943SGeorge Wilson spa_vdev_config_exit(spa, vd, txg, error, FTAG); 1226ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1227bbfd46c4SJeff Bonwick mutex_exit(&spa->spa_vdev_top_lock); 1228fa9e4066Sahrens 1229fa9e4066Sahrens return (error); 1230fa9e4066Sahrens } 1231fa9e4066Sahrens 1232e14bb325SJeff Bonwick /* 1233e14bb325SJeff Bonwick * Lock the given spa_t for the purpose of changing vdev state. 1234e14bb325SJeff Bonwick */ 1235e14bb325SJeff Bonwick void 12368f18d1faSGeorge Wilson spa_vdev_state_enter(spa_t *spa, int oplocks) 1237e14bb325SJeff Bonwick { 12388f18d1faSGeorge Wilson int locks = SCL_STATE_ALL | oplocks; 12398f18d1faSGeorge Wilson 1240dcba9f3fSGeorge Wilson /* 1241dcba9f3fSGeorge Wilson * Root pools may need to read of the underlying devfs filesystem 1242dcba9f3fSGeorge Wilson * when opening up a vdev. Unfortunately if we're holding the 1243dcba9f3fSGeorge Wilson * SCL_ZIO lock it will result in a deadlock when we try to issue 1244dcba9f3fSGeorge Wilson * the read from the root filesystem. Instead we "prefetch" 1245dcba9f3fSGeorge Wilson * the associated vnodes that we need prior to opening the 1246dcba9f3fSGeorge Wilson * underlying devices and cache them so that we can prevent 1247dcba9f3fSGeorge Wilson * any I/O when we are doing the actual open. 1248dcba9f3fSGeorge Wilson */ 1249dcba9f3fSGeorge Wilson if (spa_is_root(spa)) { 12509842588bSGeorge Wilson int low = locks & ~(SCL_ZIO - 1); 12519842588bSGeorge Wilson int high = locks & ~low; 12529842588bSGeorge Wilson 12539842588bSGeorge Wilson spa_config_enter(spa, high, spa, RW_WRITER); 1254dcba9f3fSGeorge Wilson vdev_hold(spa->spa_root_vdev); 12559842588bSGeorge Wilson spa_config_enter(spa, low, spa, RW_WRITER); 1256dcba9f3fSGeorge Wilson } else { 1257dcba9f3fSGeorge Wilson spa_config_enter(spa, locks, spa, RW_WRITER); 1258dcba9f3fSGeorge Wilson } 12598f18d1faSGeorge Wilson spa->spa_vdev_locks = locks; 1260e14bb325SJeff Bonwick } 1261e14bb325SJeff Bonwick 1262e14bb325SJeff Bonwick int 1263e14bb325SJeff Bonwick spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) 1264e14bb325SJeff Bonwick { 1265c6065d0fSGeorge Wilson boolean_t config_changed = B_FALSE; 1266c6065d0fSGeorge Wilson 1267b24ab676SJeff Bonwick if (vd != NULL || error == 0) 1268b24ab676SJeff Bonwick vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 1269b24ab676SJeff Bonwick 0, 0, B_FALSE); 1270b24ab676SJeff Bonwick 12718f18d1faSGeorge Wilson if (vd != NULL) { 1272e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 1273c6065d0fSGeorge Wilson config_changed = B_TRUE; 12748f18d1faSGeorge Wilson spa->spa_config_generation++; 12758f18d1faSGeorge Wilson } 1276e14bb325SJeff Bonwick 1277dcba9f3fSGeorge Wilson if (spa_is_root(spa)) 1278dcba9f3fSGeorge Wilson vdev_rele(spa->spa_root_vdev); 1279dcba9f3fSGeorge Wilson 12808f18d1faSGeorge Wilson ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); 12818f18d1faSGeorge Wilson spa_config_exit(spa, spa->spa_vdev_locks, spa); 1282e14bb325SJeff Bonwick 12838ad4d6ddSJeff Bonwick /* 12848ad4d6ddSJeff Bonwick * If anything changed, wait for it to sync. This ensures that, 12858ad4d6ddSJeff Bonwick * from the system administrator's perspective, zpool(1M) commands 12868ad4d6ddSJeff Bonwick * are synchronous. This is important for things like zpool offline: 12878ad4d6ddSJeff Bonwick * when the command completes, you expect no further I/O from ZFS. 12888ad4d6ddSJeff Bonwick */ 12898ad4d6ddSJeff Bonwick if (vd != NULL) 12908ad4d6ddSJeff Bonwick txg_wait_synced(spa->spa_dsl_pool, 0); 12918ad4d6ddSJeff Bonwick 1292c6065d0fSGeorge Wilson /* 1293c6065d0fSGeorge Wilson * If the config changed, update the config cache. 1294c6065d0fSGeorge Wilson */ 1295c6065d0fSGeorge Wilson if (config_changed) { 1296c6065d0fSGeorge Wilson mutex_enter(&spa_namespace_lock); 12975cabbc6bSPrashanth Sreenivasa spa_write_cachefile(spa, B_FALSE, B_TRUE); 1298c6065d0fSGeorge Wilson mutex_exit(&spa_namespace_lock); 1299c6065d0fSGeorge Wilson } 1300c6065d0fSGeorge Wilson 1301e14bb325SJeff Bonwick return (error); 1302e14bb325SJeff Bonwick } 1303e14bb325SJeff Bonwick 1304fa9e4066Sahrens /* 1305fa9e4066Sahrens * ========================================================================== 1306fa9e4066Sahrens * Miscellaneous functions 1307fa9e4066Sahrens * ========================================================================== 1308fa9e4066Sahrens */ 1309fa9e4066Sahrens 1310ad135b5dSChristopher Siden void 131143466aaeSMax Grossman spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx) 1312ad135b5dSChristopher Siden { 13132acef22dSMatthew Ahrens if (!nvlist_exists(spa->spa_label_features, feature)) { 13142acef22dSMatthew Ahrens fnvlist_add_boolean(spa->spa_label_features, feature); 131543466aaeSMax Grossman /* 131643466aaeSMax Grossman * When we are creating the pool (tx_txg==TXG_INITIAL), we can't 131743466aaeSMax Grossman * dirty the vdev config because lock SCL_CONFIG is not held. 131843466aaeSMax Grossman * Thankfully, in this case we don't need to dirty the config 131943466aaeSMax Grossman * because it will be written out anyway when we finish 132043466aaeSMax Grossman * creating the pool. 132143466aaeSMax Grossman */ 132243466aaeSMax Grossman if (tx->tx_txg != TXG_INITIAL) 132343466aaeSMax Grossman vdev_config_dirty(spa->spa_root_vdev); 13242acef22dSMatthew Ahrens } 1325ad135b5dSChristopher Siden } 1326ad135b5dSChristopher Siden 1327ad135b5dSChristopher Siden void 1328ad135b5dSChristopher Siden spa_deactivate_mos_feature(spa_t *spa, const char *feature) 1329ad135b5dSChristopher Siden { 13302acef22dSMatthew Ahrens if (nvlist_remove_all(spa->spa_label_features, feature) == 0) 13312acef22dSMatthew Ahrens vdev_config_dirty(spa->spa_root_vdev); 1332ad135b5dSChristopher Siden } 1333ad135b5dSChristopher Siden 1334fa9e4066Sahrens /* 1335f9af39baSGeorge Wilson * Return the spa_t associated with given pool_guid, if it exists. If 1336f9af39baSGeorge Wilson * device_guid is non-zero, determine whether the pool exists *and* contains 1337f9af39baSGeorge Wilson * a device with the specified device_guid. 1338fa9e4066Sahrens */ 1339f9af39baSGeorge Wilson spa_t * 1340f9af39baSGeorge Wilson spa_by_guid(uint64_t pool_guid, uint64_t device_guid) 1341fa9e4066Sahrens { 1342fa9e4066Sahrens spa_t *spa; 1343fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 1344fa9e4066Sahrens 1345ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1346fa9e4066Sahrens 1347fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 1348fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 1349fa9e4066Sahrens continue; 1350fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 1351fa9e4066Sahrens continue; 135239c23413Seschrock if (spa_guid(spa) == pool_guid) { 135339c23413Seschrock if (device_guid == 0) 135439c23413Seschrock break; 135539c23413Seschrock 135639c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 135739c23413Seschrock device_guid) != NULL) 135839c23413Seschrock break; 135939c23413Seschrock 136039c23413Seschrock /* 13618654d025Sperrin * Check any devices we may be in the process of adding. 136239c23413Seschrock */ 136339c23413Seschrock if (spa->spa_pending_vdev) { 136439c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 136539c23413Seschrock device_guid) != NULL) 136639c23413Seschrock break; 136739c23413Seschrock } 136839c23413Seschrock } 1369fa9e4066Sahrens } 1370fa9e4066Sahrens 1371f9af39baSGeorge Wilson return (spa); 1372f9af39baSGeorge Wilson } 1373f9af39baSGeorge Wilson 1374f9af39baSGeorge Wilson /* 1375f9af39baSGeorge Wilson * Determine whether a pool with the given pool_guid exists. 1376f9af39baSGeorge Wilson */ 1377f9af39baSGeorge Wilson boolean_t 1378f9af39baSGeorge Wilson spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 1379f9af39baSGeorge Wilson { 1380f9af39baSGeorge Wilson return (spa_by_guid(pool_guid, device_guid) != NULL); 1381fa9e4066Sahrens } 1382fa9e4066Sahrens 1383fa9e4066Sahrens char * 1384fa9e4066Sahrens spa_strdup(const char *s) 1385fa9e4066Sahrens { 1386fa9e4066Sahrens size_t len; 1387fa9e4066Sahrens char *new; 1388fa9e4066Sahrens 1389fa9e4066Sahrens len = strlen(s); 1390fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 1391fa9e4066Sahrens bcopy(s, new, len); 1392fa9e4066Sahrens new[len] = '\0'; 1393fa9e4066Sahrens 1394fa9e4066Sahrens return (new); 1395fa9e4066Sahrens } 1396fa9e4066Sahrens 1397fa9e4066Sahrens void 1398fa9e4066Sahrens spa_strfree(char *s) 1399fa9e4066Sahrens { 1400fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 1401fa9e4066Sahrens } 1402fa9e4066Sahrens 1403fa9e4066Sahrens uint64_t 1404fa9e4066Sahrens spa_get_random(uint64_t range) 1405fa9e4066Sahrens { 1406fa9e4066Sahrens uint64_t r; 1407fa9e4066Sahrens 1408fa9e4066Sahrens ASSERT(range != 0); 1409fa9e4066Sahrens 1410fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 1411fa9e4066Sahrens 1412fa9e4066Sahrens return (r % range); 1413fa9e4066Sahrens } 1414fa9e4066Sahrens 14151195e687SMark J Musante uint64_t 14161195e687SMark J Musante spa_generate_guid(spa_t *spa) 14171195e687SMark J Musante { 14181195e687SMark J Musante uint64_t guid = spa_get_random(-1ULL); 14191195e687SMark J Musante 14201195e687SMark J Musante if (spa != NULL) { 14211195e687SMark J Musante while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) 14221195e687SMark J Musante guid = spa_get_random(-1ULL); 14231195e687SMark J Musante } else { 14241195e687SMark J Musante while (guid == 0 || spa_guid_exists(guid, 0)) 14251195e687SMark J Musante guid = spa_get_random(-1ULL); 14261195e687SMark J Musante } 14271195e687SMark J Musante 14281195e687SMark J Musante return (guid); 14291195e687SMark J Musante } 14301195e687SMark J Musante 1431fa9e4066Sahrens void 143243466aaeSMax Grossman snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp) 1433fa9e4066Sahrens { 1434ad135b5dSChristopher Siden char type[256]; 1435f0ba89beSJeff Bonwick char *checksum = NULL; 1436f0ba89beSJeff Bonwick char *compress = NULL; 1437f0ba89beSJeff Bonwick 1438f0ba89beSJeff Bonwick if (bp != NULL) { 1439ad135b5dSChristopher Siden if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { 1440ad135b5dSChristopher Siden dmu_object_byteswap_t bswap = 1441ad135b5dSChristopher Siden DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); 1442ad135b5dSChristopher Siden (void) snprintf(type, sizeof (type), "bswap %s %s", 1443ad135b5dSChristopher Siden DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? 1444ad135b5dSChristopher Siden "metadata" : "data", 1445ad135b5dSChristopher Siden dmu_ot_byteswap[bswap].ob_name); 1446ad135b5dSChristopher Siden } else { 1447ad135b5dSChristopher Siden (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, 1448ad135b5dSChristopher Siden sizeof (type)); 1449ad135b5dSChristopher Siden } 14505d7b4d43SMatthew Ahrens if (!BP_IS_EMBEDDED(bp)) { 14515d7b4d43SMatthew Ahrens checksum = 14525d7b4d43SMatthew Ahrens zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; 14535d7b4d43SMatthew Ahrens } 1454f0ba89beSJeff Bonwick compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; 1455f0ba89beSJeff Bonwick } 1456fa9e4066Sahrens 145743466aaeSMax Grossman SNPRINTF_BLKPTR(snprintf, ' ', buf, buflen, bp, type, checksum, 145843466aaeSMax Grossman compress); 1459fa9e4066Sahrens } 1460fa9e4066Sahrens 1461fa9e4066Sahrens void 1462fa9e4066Sahrens spa_freeze(spa_t *spa) 1463fa9e4066Sahrens { 1464fa9e4066Sahrens uint64_t freeze_txg = 0; 1465fa9e4066Sahrens 1466e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1467fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 1468fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 1469fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 1470fa9e4066Sahrens } 1471e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1472fa9e4066Sahrens if (freeze_txg != 0) 1473fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 1474fa9e4066Sahrens } 1475fa9e4066Sahrens 14760125049cSahrens void 14770125049cSahrens zfs_panic_recover(const char *fmt, ...) 14780125049cSahrens { 14790125049cSahrens va_list adx; 14800125049cSahrens 14810125049cSahrens va_start(adx, fmt); 14820125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 14830125049cSahrens va_end(adx); 14840125049cSahrens } 14850125049cSahrens 14863f9d6ad7SLin Ling /* 14873f9d6ad7SLin Ling * This is a stripped-down version of strtoull, suitable only for converting 1488f7170741SWill Andrews * lowercase hexadecimal numbers that don't overflow. 14893f9d6ad7SLin Ling */ 14903f9d6ad7SLin Ling uint64_t 14914585130bSYuri Pankov zfs_strtonum(const char *str, char **nptr) 14923f9d6ad7SLin Ling { 14933f9d6ad7SLin Ling uint64_t val = 0; 14943f9d6ad7SLin Ling char c; 14953f9d6ad7SLin Ling int digit; 14963f9d6ad7SLin Ling 14973f9d6ad7SLin Ling while ((c = *str) != '\0') { 14983f9d6ad7SLin Ling if (c >= '0' && c <= '9') 14993f9d6ad7SLin Ling digit = c - '0'; 15003f9d6ad7SLin Ling else if (c >= 'a' && c <= 'f') 15013f9d6ad7SLin Ling digit = 10 + c - 'a'; 15023f9d6ad7SLin Ling else 15033f9d6ad7SLin Ling break; 15043f9d6ad7SLin Ling 15053f9d6ad7SLin Ling val *= 16; 15063f9d6ad7SLin Ling val += digit; 15073f9d6ad7SLin Ling 15083f9d6ad7SLin Ling str++; 15093f9d6ad7SLin Ling } 15103f9d6ad7SLin Ling 15113f9d6ad7SLin Ling if (nptr) 15123f9d6ad7SLin Ling *nptr = (char *)str; 15133f9d6ad7SLin Ling 15143f9d6ad7SLin Ling return (val); 15153f9d6ad7SLin Ling } 15163f9d6ad7SLin Ling 1517fa9e4066Sahrens /* 1518fa9e4066Sahrens * ========================================================================== 1519fa9e4066Sahrens * Accessor functions 1520fa9e4066Sahrens * ========================================================================== 1521fa9e4066Sahrens */ 1522fa9e4066Sahrens 1523088f3894Sahrens boolean_t 152488b7b0f2SMatthew Ahrens spa_shutting_down(spa_t *spa) 1525fa9e4066Sahrens { 152688b7b0f2SMatthew Ahrens return (spa->spa_async_suspended); 1527fa9e4066Sahrens } 1528fa9e4066Sahrens 1529fa9e4066Sahrens dsl_pool_t * 1530fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1531fa9e4066Sahrens { 1532fa9e4066Sahrens return (spa->spa_dsl_pool); 1533fa9e4066Sahrens } 1534fa9e4066Sahrens 1535ad135b5dSChristopher Siden boolean_t 1536ad135b5dSChristopher Siden spa_is_initializing(spa_t *spa) 1537ad135b5dSChristopher Siden { 1538ad135b5dSChristopher Siden return (spa->spa_is_initializing); 1539ad135b5dSChristopher Siden } 1540ad135b5dSChristopher Siden 15415cabbc6bSPrashanth Sreenivasa boolean_t 15425cabbc6bSPrashanth Sreenivasa spa_indirect_vdevs_loaded(spa_t *spa) 15435cabbc6bSPrashanth Sreenivasa { 15445cabbc6bSPrashanth Sreenivasa return (spa->spa_indirect_vdevs_loaded); 15455cabbc6bSPrashanth Sreenivasa } 15465cabbc6bSPrashanth Sreenivasa 1547fa9e4066Sahrens blkptr_t * 1548fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1549fa9e4066Sahrens { 1550fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1551fa9e4066Sahrens } 1552fa9e4066Sahrens 1553fa9e4066Sahrens void 1554fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1555fa9e4066Sahrens { 1556fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1557fa9e4066Sahrens } 1558fa9e4066Sahrens 1559fa9e4066Sahrens void 1560fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1561fa9e4066Sahrens { 1562fa9e4066Sahrens if (spa->spa_root == NULL) 1563fa9e4066Sahrens buf[0] = '\0'; 1564fa9e4066Sahrens else 1565fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1566fa9e4066Sahrens } 1567fa9e4066Sahrens 1568fa9e4066Sahrens int 1569fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1570fa9e4066Sahrens { 1571fa9e4066Sahrens return (spa->spa_sync_pass); 1572fa9e4066Sahrens } 1573fa9e4066Sahrens 1574fa9e4066Sahrens char * 1575fa9e4066Sahrens spa_name(spa_t *spa) 1576fa9e4066Sahrens { 1577fa9e4066Sahrens return (spa->spa_name); 1578fa9e4066Sahrens } 1579fa9e4066Sahrens 1580fa9e4066Sahrens uint64_t 1581fa9e4066Sahrens spa_guid(spa_t *spa) 1582fa9e4066Sahrens { 1583dfbb9432SGeorge Wilson dsl_pool_t *dp = spa_get_dsl(spa); 1584dfbb9432SGeorge Wilson uint64_t guid; 1585dfbb9432SGeorge Wilson 1586b5989ec7Seschrock /* 1587b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1588b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1589e9103aaeSGarrett D'Amore * vdev. We stash the original pool guid in 'spa_config_guid' to handle 1590b5989ec7Seschrock * this case. 1591b5989ec7Seschrock */ 1592dfbb9432SGeorge Wilson if (spa->spa_root_vdev == NULL) 1593dfbb9432SGeorge Wilson return (spa->spa_config_guid); 1594dfbb9432SGeorge Wilson 1595dfbb9432SGeorge Wilson guid = spa->spa_last_synced_guid != 0 ? 1596dfbb9432SGeorge Wilson spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; 1597dfbb9432SGeorge Wilson 1598dfbb9432SGeorge Wilson /* 1599dfbb9432SGeorge Wilson * Return the most recently synced out guid unless we're 1600dfbb9432SGeorge Wilson * in syncing context. 1601dfbb9432SGeorge Wilson */ 1602dfbb9432SGeorge Wilson if (dp && dsl_pool_sync_context(dp)) 1603b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1604b5989ec7Seschrock else 1605dfbb9432SGeorge Wilson return (guid); 1606e9103aaeSGarrett D'Amore } 1607e9103aaeSGarrett D'Amore 1608e9103aaeSGarrett D'Amore uint64_t 1609e9103aaeSGarrett D'Amore spa_load_guid(spa_t *spa) 1610e9103aaeSGarrett D'Amore { 1611e9103aaeSGarrett D'Amore /* 1612e9103aaeSGarrett D'Amore * This is a GUID that exists solely as a reference for the 1613e9103aaeSGarrett D'Amore * purposes of the arc. It is generated at load time, and 1614e9103aaeSGarrett D'Amore * is never written to persistent storage. 1615e9103aaeSGarrett D'Amore */ 1616e9103aaeSGarrett D'Amore return (spa->spa_load_guid); 1617fa9e4066Sahrens } 1618fa9e4066Sahrens 1619fa9e4066Sahrens uint64_t 1620fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1621fa9e4066Sahrens { 1622fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1623fa9e4066Sahrens } 1624fa9e4066Sahrens 1625fa9e4066Sahrens uint64_t 1626fa9e4066Sahrens spa_first_txg(spa_t *spa) 1627fa9e4066Sahrens { 1628fa9e4066Sahrens return (spa->spa_first_txg); 1629fa9e4066Sahrens } 1630fa9e4066Sahrens 1631b24ab676SJeff Bonwick uint64_t 1632b24ab676SJeff Bonwick spa_syncing_txg(spa_t *spa) 1633b24ab676SJeff Bonwick { 1634b24ab676SJeff Bonwick return (spa->spa_syncing_txg); 1635b24ab676SJeff Bonwick } 1636b24ab676SJeff Bonwick 16373991b535SGeorge Wilson /* 16383991b535SGeorge Wilson * Return the last txg where data can be dirtied. The final txgs 16393991b535SGeorge Wilson * will be used to just clear out any deferred frees that remain. 16403991b535SGeorge Wilson */ 16413991b535SGeorge Wilson uint64_t 16423991b535SGeorge Wilson spa_final_dirty_txg(spa_t *spa) 16433991b535SGeorge Wilson { 16443991b535SGeorge Wilson return (spa->spa_final_txg - TXG_DEFER_SIZE); 16453991b535SGeorge Wilson } 16463991b535SGeorge Wilson 164788b7b0f2SMatthew Ahrens pool_state_t 1648fa9e4066Sahrens spa_state(spa_t *spa) 1649fa9e4066Sahrens { 1650fa9e4066Sahrens return (spa->spa_state); 1651fa9e4066Sahrens } 1652fa9e4066Sahrens 1653b16da2e2SGeorge Wilson spa_load_state_t 1654b16da2e2SGeorge Wilson spa_load_state(spa_t *spa) 1655b16da2e2SGeorge Wilson { 1656b16da2e2SGeorge Wilson return (spa->spa_load_state); 1657b16da2e2SGeorge Wilson } 1658b16da2e2SGeorge Wilson 1659fa9e4066Sahrens uint64_t 1660fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1661fa9e4066Sahrens { 1662fa9e4066Sahrens return (spa->spa_freeze_txg); 1663fa9e4066Sahrens } 1664fa9e4066Sahrens 1665fa9e4066Sahrens /* ARGSUSED */ 1666fa9e4066Sahrens uint64_t 166761e255ceSMatthew Ahrens spa_get_worst_case_asize(spa_t *spa, uint64_t lsize) 1668fa9e4066Sahrens { 166969962b56SMatthew Ahrens return (lsize * spa_asize_inflation); 167044cd46caSbillm } 167144cd46caSbillm 16727d46dc6cSMatthew Ahrens /* 16737d46dc6cSMatthew Ahrens * Return the amount of slop space in bytes. It is 1/32 of the pool (3.2%), 16744b5c8e93SMatthew Ahrens * or at least 128MB, unless that would cause it to be more than half the 16754b5c8e93SMatthew Ahrens * pool size. 16767d46dc6cSMatthew Ahrens * 16777d46dc6cSMatthew Ahrens * See the comment above spa_slop_shift for details. 16787d46dc6cSMatthew Ahrens */ 16797d46dc6cSMatthew Ahrens uint64_t 16804b5c8e93SMatthew Ahrens spa_get_slop_space(spa_t *spa) 16814b5c8e93SMatthew Ahrens { 16827d46dc6cSMatthew Ahrens uint64_t space = spa_get_dspace(spa); 16834b5c8e93SMatthew Ahrens return (MAX(space >> spa_slop_shift, MIN(space >> 1, spa_min_slop))); 16847d46dc6cSMatthew Ahrens } 16857d46dc6cSMatthew Ahrens 1686485bbbf5SGeorge Wilson uint64_t 1687485bbbf5SGeorge Wilson spa_get_dspace(spa_t *spa) 1688485bbbf5SGeorge Wilson { 1689485bbbf5SGeorge Wilson return (spa->spa_dspace); 1690485bbbf5SGeorge Wilson } 1691485bbbf5SGeorge Wilson 169286714001SSerapheim Dimitropoulos uint64_t 169386714001SSerapheim Dimitropoulos spa_get_checkpoint_space(spa_t *spa) 169486714001SSerapheim Dimitropoulos { 169586714001SSerapheim Dimitropoulos return (spa->spa_checkpoint_info.sci_dspace); 169686714001SSerapheim Dimitropoulos } 169786714001SSerapheim Dimitropoulos 1698485bbbf5SGeorge Wilson void 1699485bbbf5SGeorge Wilson spa_update_dspace(spa_t *spa) 1700485bbbf5SGeorge Wilson { 1701485bbbf5SGeorge Wilson spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + 1702485bbbf5SGeorge Wilson ddt_get_dedup_dspace(spa); 17035cabbc6bSPrashanth Sreenivasa if (spa->spa_vdev_removal != NULL) { 17045cabbc6bSPrashanth Sreenivasa /* 17055cabbc6bSPrashanth Sreenivasa * We can't allocate from the removing device, so 17065cabbc6bSPrashanth Sreenivasa * subtract its size. This prevents the DMU/DSL from 17075cabbc6bSPrashanth Sreenivasa * filling up the (now smaller) pool while we are in the 17085cabbc6bSPrashanth Sreenivasa * middle of removing the device. 17095cabbc6bSPrashanth Sreenivasa * 17105cabbc6bSPrashanth Sreenivasa * Note that the DMU/DSL doesn't actually know or care 17115cabbc6bSPrashanth Sreenivasa * how much space is allocated (it does its own tracking 17125cabbc6bSPrashanth Sreenivasa * of how much space has been logically used). So it 17135cabbc6bSPrashanth Sreenivasa * doesn't matter that the data we are moving may be 17145cabbc6bSPrashanth Sreenivasa * allocated twice (on the old device and the new 17155cabbc6bSPrashanth Sreenivasa * device). 17165cabbc6bSPrashanth Sreenivasa */ 17173a4b1be9SMatthew Ahrens spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 17183a4b1be9SMatthew Ahrens vdev_t *vd = 17193a4b1be9SMatthew Ahrens vdev_lookup_top(spa, spa->spa_vdev_removal->svr_vdev_id); 17205cabbc6bSPrashanth Sreenivasa spa->spa_dspace -= spa_deflate(spa) ? 17215cabbc6bSPrashanth Sreenivasa vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space; 17223a4b1be9SMatthew Ahrens spa_config_exit(spa, SCL_VDEV, FTAG); 17235cabbc6bSPrashanth Sreenivasa } 1724485bbbf5SGeorge Wilson } 1725485bbbf5SGeorge Wilson 17260a4e9518Sgw /* 17270a4e9518Sgw * Return the failure mode that has been set to this pool. The default 17280a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 17290a4e9518Sgw */ 17300a4e9518Sgw uint8_t 17310a4e9518Sgw spa_get_failmode(spa_t *spa) 17320a4e9518Sgw { 17330a4e9518Sgw return (spa->spa_failmode); 17340a4e9518Sgw } 17350a4e9518Sgw 1736e14bb325SJeff Bonwick boolean_t 1737e14bb325SJeff Bonwick spa_suspended(spa_t *spa) 1738e14bb325SJeff Bonwick { 1739e14bb325SJeff Bonwick return (spa->spa_suspended); 1740e14bb325SJeff Bonwick } 1741e14bb325SJeff Bonwick 174244cd46caSbillm uint64_t 174344cd46caSbillm spa_version(spa_t *spa) 174444cd46caSbillm { 174544cd46caSbillm return (spa->spa_ubsync.ub_version); 174644cd46caSbillm } 174744cd46caSbillm 1748b24ab676SJeff Bonwick boolean_t 1749b24ab676SJeff Bonwick spa_deflate(spa_t *spa) 1750b24ab676SJeff Bonwick { 1751b24ab676SJeff Bonwick return (spa->spa_deflate); 1752b24ab676SJeff Bonwick } 1753b24ab676SJeff Bonwick 1754b24ab676SJeff Bonwick metaslab_class_t * 1755b24ab676SJeff Bonwick spa_normal_class(spa_t *spa) 1756b24ab676SJeff Bonwick { 1757b24ab676SJeff Bonwick return (spa->spa_normal_class); 1758b24ab676SJeff Bonwick } 1759b24ab676SJeff Bonwick 1760b24ab676SJeff Bonwick metaslab_class_t * 1761b24ab676SJeff Bonwick spa_log_class(spa_t *spa) 1762b24ab676SJeff Bonwick { 1763b24ab676SJeff Bonwick return (spa->spa_log_class); 1764b24ab676SJeff Bonwick } 1765b24ab676SJeff Bonwick 1766bc9014e6SJustin Gibbs void 1767bc9014e6SJustin Gibbs spa_evicting_os_register(spa_t *spa, objset_t *os) 1768bc9014e6SJustin Gibbs { 1769bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1770bc9014e6SJustin Gibbs list_insert_head(&spa->spa_evicting_os_list, os); 1771bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1772bc9014e6SJustin Gibbs } 1773bc9014e6SJustin Gibbs 1774bc9014e6SJustin Gibbs void 1775bc9014e6SJustin Gibbs spa_evicting_os_deregister(spa_t *spa, objset_t *os) 1776bc9014e6SJustin Gibbs { 1777bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1778bc9014e6SJustin Gibbs list_remove(&spa->spa_evicting_os_list, os); 1779bc9014e6SJustin Gibbs cv_broadcast(&spa->spa_evicting_os_cv); 1780bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1781bc9014e6SJustin Gibbs } 1782bc9014e6SJustin Gibbs 1783bc9014e6SJustin Gibbs void 1784bc9014e6SJustin Gibbs spa_evicting_os_wait(spa_t *spa) 1785bc9014e6SJustin Gibbs { 1786bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1787bc9014e6SJustin Gibbs while (!list_is_empty(&spa->spa_evicting_os_list)) 1788bc9014e6SJustin Gibbs cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock); 1789bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1790bc9014e6SJustin Gibbs 1791bc9014e6SJustin Gibbs dmu_buf_user_evict_wait(); 1792bc9014e6SJustin Gibbs } 1793bc9014e6SJustin Gibbs 179444cd46caSbillm int 179544cd46caSbillm spa_max_replication(spa_t *spa) 179644cd46caSbillm { 179744cd46caSbillm /* 1798e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 179944cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 180044cd46caSbillm * replication level accordingly. 1801fa9e4066Sahrens */ 1802e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 180344cd46caSbillm return (1); 180444cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1805fa9e4066Sahrens } 1806fa9e4066Sahrens 18073f9d6ad7SLin Ling int 18083f9d6ad7SLin Ling spa_prev_software_version(spa_t *spa) 18093f9d6ad7SLin Ling { 18103f9d6ad7SLin Ling return (spa->spa_prev_software_version); 18113f9d6ad7SLin Ling } 18123f9d6ad7SLin Ling 1813283b8460SGeorge.Wilson uint64_t 1814283b8460SGeorge.Wilson spa_deadman_synctime(spa_t *spa) 1815283b8460SGeorge.Wilson { 1816283b8460SGeorge.Wilson return (spa->spa_deadman_synctime); 1817283b8460SGeorge.Wilson } 1818283b8460SGeorge.Wilson 181999653d4eSeschrock uint64_t 1820b24ab676SJeff Bonwick dva_get_dsize_sync(spa_t *spa, const dva_t *dva) 182199653d4eSeschrock { 1822b24ab676SJeff Bonwick uint64_t asize = DVA_GET_ASIZE(dva); 1823b24ab676SJeff Bonwick uint64_t dsize = asize; 182499653d4eSeschrock 1825b24ab676SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); 182699653d4eSeschrock 1827b24ab676SJeff Bonwick if (asize != 0 && spa->spa_deflate) { 1828b24ab676SJeff Bonwick vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); 1829b24ab676SJeff Bonwick dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 183099653d4eSeschrock } 1831b24ab676SJeff Bonwick 1832b24ab676SJeff Bonwick return (dsize); 1833b24ab676SJeff Bonwick } 1834b24ab676SJeff Bonwick 1835b24ab676SJeff Bonwick uint64_t 1836b24ab676SJeff Bonwick bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) 1837b24ab676SJeff Bonwick { 1838b24ab676SJeff Bonwick uint64_t dsize = 0; 1839b24ab676SJeff Bonwick 18405d7b4d43SMatthew Ahrens for (int d = 0; d < BP_GET_NDVAS(bp); d++) 1841b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1842b24ab676SJeff Bonwick 1843b24ab676SJeff Bonwick return (dsize); 1844b24ab676SJeff Bonwick } 1845b24ab676SJeff Bonwick 1846b24ab676SJeff Bonwick uint64_t 1847b24ab676SJeff Bonwick bp_get_dsize(spa_t *spa, const blkptr_t *bp) 1848b24ab676SJeff Bonwick { 1849b24ab676SJeff Bonwick uint64_t dsize = 0; 1850b24ab676SJeff Bonwick 1851b24ab676SJeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 1852b24ab676SJeff Bonwick 18535d7b4d43SMatthew Ahrens for (int d = 0; d < BP_GET_NDVAS(bp); d++) 1854b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1855b24ab676SJeff Bonwick 1856e14bb325SJeff Bonwick spa_config_exit(spa, SCL_VDEV, FTAG); 1857b24ab676SJeff Bonwick 1858b24ab676SJeff Bonwick return (dsize); 185999653d4eSeschrock } 186099653d4eSeschrock 1861abe1fd01SDon Brady uint64_t 1862abe1fd01SDon Brady spa_dirty_data(spa_t *spa) 1863abe1fd01SDon Brady { 1864abe1fd01SDon Brady return (spa->spa_dsl_pool->dp_dirty_total); 1865abe1fd01SDon Brady } 1866abe1fd01SDon Brady 1867fa9e4066Sahrens /* 1868fa9e4066Sahrens * ========================================================================== 1869fa9e4066Sahrens * Initialization and Termination 1870fa9e4066Sahrens * ========================================================================== 1871fa9e4066Sahrens */ 1872fa9e4066Sahrens 1873fa9e4066Sahrens static int 1874fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 1875fa9e4066Sahrens { 1876fa9e4066Sahrens const spa_t *s1 = a1; 1877fa9e4066Sahrens const spa_t *s2 = a2; 1878fa9e4066Sahrens int s; 1879fa9e4066Sahrens 1880fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 1881fa9e4066Sahrens if (s > 0) 1882fa9e4066Sahrens return (1); 1883fa9e4066Sahrens if (s < 0) 1884fa9e4066Sahrens return (-1); 1885fa9e4066Sahrens return (0); 1886fa9e4066Sahrens } 1887fa9e4066Sahrens 18880373e76bSbonwick int 18890373e76bSbonwick spa_busy(void) 18900373e76bSbonwick { 18910373e76bSbonwick return (spa_active_count); 18920373e76bSbonwick } 18930373e76bSbonwick 1894e7cbe64fSgw void 1895e7cbe64fSgw spa_boot_init() 1896e7cbe64fSgw { 1897e7cbe64fSgw spa_config_load(); 1898e7cbe64fSgw } 1899e7cbe64fSgw 1900fa9e4066Sahrens void 1901fa9e4066Sahrens spa_init(int mode) 1902fa9e4066Sahrens { 1903fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 1904c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 1905fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 1906fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 1907fa9e4066Sahrens 1908fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 1909fa9e4066Sahrens offsetof(spa_t, spa_avl)); 1910fa9e4066Sahrens 1911fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 1912fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 1913fa94a07fSbrendan 1914fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 1915fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 191699653d4eSeschrock 19178ad4d6ddSJeff Bonwick spa_mode_global = mode; 1918fa9e4066Sahrens 1919283b8460SGeorge.Wilson #ifdef _KERNEL 1920283b8460SGeorge.Wilson spa_arch_init(); 1921283b8460SGeorge.Wilson #else 1922cd1c8b85SMatthew Ahrens if (spa_mode_global != FREAD && dprintf_find_string("watch")) { 1923cd1c8b85SMatthew Ahrens arc_procfd = open("/proc/self/ctl", O_WRONLY); 1924cd1c8b85SMatthew Ahrens if (arc_procfd == -1) { 1925cd1c8b85SMatthew Ahrens perror("could not enable watchpoints: " 1926cd1c8b85SMatthew Ahrens "opening /proc/self/ctl failed: "); 1927cd1c8b85SMatthew Ahrens } else { 1928cd1c8b85SMatthew Ahrens arc_watch = B_TRUE; 1929cd1c8b85SMatthew Ahrens } 1930cd1c8b85SMatthew Ahrens } 1931cd1c8b85SMatthew Ahrens #endif 1932cd1c8b85SMatthew Ahrens 1933*e914ace2STim Schumacher zfs_refcount_init(); 1934fa9e4066Sahrens unique_init(); 19350713e232SGeorge Wilson range_tree_init(); 19368363e80aSGeorge Wilson metaslab_alloc_trace_init(); 1937fa9e4066Sahrens zio_init(); 1938fa9e4066Sahrens dmu_init(); 1939fa9e4066Sahrens zil_init(); 194087db74c1Sek vdev_cache_stat_init(); 194191ebeef5Sahrens zfs_prop_init(); 1942990b4856Slling zpool_prop_init(); 1943ad135b5dSChristopher Siden zpool_feature_init(); 1944fa9e4066Sahrens spa_config_load(); 1945e14bb325SJeff Bonwick l2arc_start(); 1946fa9e4066Sahrens } 1947fa9e4066Sahrens 1948fa9e4066Sahrens void 1949fa9e4066Sahrens spa_fini(void) 1950fa9e4066Sahrens { 1951e14bb325SJeff Bonwick l2arc_stop(); 1952e14bb325SJeff Bonwick 1953fa9e4066Sahrens spa_evict_all(); 1954fa9e4066Sahrens 195587db74c1Sek vdev_cache_stat_fini(); 1956fa9e4066Sahrens zil_fini(); 1957fa9e4066Sahrens dmu_fini(); 1958fa9e4066Sahrens zio_fini(); 19598363e80aSGeorge Wilson metaslab_alloc_trace_fini(); 19600713e232SGeorge Wilson range_tree_fini(); 196191ebeef5Sahrens unique_fini(); 1962*e914ace2STim Schumacher zfs_refcount_fini(); 1963fa9e4066Sahrens 1964fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 196599653d4eSeschrock avl_destroy(&spa_spare_avl); 1966fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 1967fa9e4066Sahrens 1968fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 1969fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 1970c25056deSgw mutex_destroy(&spa_spare_lock); 1971fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 1972fa9e4066Sahrens } 19736ce0521aSperrin 19746ce0521aSperrin /* 19756ce0521aSperrin * Return whether this pool has slogs. No locking needed. 19766ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 19776ce0521aSperrin * performance and not correctness 19786ce0521aSperrin */ 19796ce0521aSperrin boolean_t 19806ce0521aSperrin spa_has_slogs(spa_t *spa) 19816ce0521aSperrin { 19826ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 19836ce0521aSperrin } 1984bf82a41bSeschrock 1985b24ab676SJeff Bonwick spa_log_state_t 1986b24ab676SJeff Bonwick spa_get_log_state(spa_t *spa) 1987b24ab676SJeff Bonwick { 1988b24ab676SJeff Bonwick return (spa->spa_log_state); 1989b24ab676SJeff Bonwick } 1990b24ab676SJeff Bonwick 1991b24ab676SJeff Bonwick void 1992b24ab676SJeff Bonwick spa_set_log_state(spa_t *spa, spa_log_state_t state) 1993b24ab676SJeff Bonwick { 1994b24ab676SJeff Bonwick spa->spa_log_state = state; 1995b24ab676SJeff Bonwick } 1996b24ab676SJeff Bonwick 1997bf82a41bSeschrock boolean_t 1998bf82a41bSeschrock spa_is_root(spa_t *spa) 1999bf82a41bSeschrock { 2000bf82a41bSeschrock return (spa->spa_is_root); 2001bf82a41bSeschrock } 20028ad4d6ddSJeff Bonwick 20038ad4d6ddSJeff Bonwick boolean_t 20048ad4d6ddSJeff Bonwick spa_writeable(spa_t *spa) 20058ad4d6ddSJeff Bonwick { 20066f793812SPavel Zakharov return (!!(spa->spa_mode & FWRITE) && spa->spa_trust_config); 20078ad4d6ddSJeff Bonwick } 20088ad4d6ddSJeff Bonwick 200973527f44SAlex Reece /* 201073527f44SAlex Reece * Returns true if there is a pending sync task in any of the current 201173527f44SAlex Reece * syncing txg, the current quiescing txg, or the current open txg. 201273527f44SAlex Reece */ 201373527f44SAlex Reece boolean_t 201473527f44SAlex Reece spa_has_pending_synctask(spa_t *spa) 201573527f44SAlex Reece { 201686714001SSerapheim Dimitropoulos return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks) || 201786714001SSerapheim Dimitropoulos !txg_all_lists_empty(&spa->spa_dsl_pool->dp_early_sync_tasks)); 201873527f44SAlex Reece } 201973527f44SAlex Reece 20208ad4d6ddSJeff Bonwick int 20218ad4d6ddSJeff Bonwick spa_mode(spa_t *spa) 20228ad4d6ddSJeff Bonwick { 20238ad4d6ddSJeff Bonwick return (spa->spa_mode); 20248ad4d6ddSJeff Bonwick } 2025b24ab676SJeff Bonwick 2026b24ab676SJeff Bonwick uint64_t 2027b24ab676SJeff Bonwick spa_bootfs(spa_t *spa) 2028b24ab676SJeff Bonwick { 2029b24ab676SJeff Bonwick return (spa->spa_bootfs); 2030b24ab676SJeff Bonwick } 2031b24ab676SJeff Bonwick 2032b24ab676SJeff Bonwick uint64_t 2033b24ab676SJeff Bonwick spa_delegation(spa_t *spa) 2034b24ab676SJeff Bonwick { 2035b24ab676SJeff Bonwick return (spa->spa_delegation); 2036b24ab676SJeff Bonwick } 2037b24ab676SJeff Bonwick 2038b24ab676SJeff Bonwick objset_t * 2039b24ab676SJeff Bonwick spa_meta_objset(spa_t *spa) 2040b24ab676SJeff Bonwick { 2041b24ab676SJeff Bonwick return (spa->spa_meta_objset); 2042b24ab676SJeff Bonwick } 2043b24ab676SJeff Bonwick 2044b24ab676SJeff Bonwick enum zio_checksum 2045b24ab676SJeff Bonwick spa_dedup_checksum(spa_t *spa) 2046b24ab676SJeff Bonwick { 2047b24ab676SJeff Bonwick return (spa->spa_dedup_checksum); 2048b24ab676SJeff Bonwick } 20493f9d6ad7SLin Ling 20503f9d6ad7SLin Ling /* 20513f9d6ad7SLin Ling * Reset pool scan stat per scan pass (or reboot). 20523f9d6ad7SLin Ling */ 20533f9d6ad7SLin Ling void 20543f9d6ad7SLin Ling spa_scan_stat_init(spa_t *spa) 20553f9d6ad7SLin Ling { 20563f9d6ad7SLin Ling /* data not stored on disk */ 20573f9d6ad7SLin Ling spa->spa_scan_pass_start = gethrestime_sec(); 20581702cce7SAlek Pinchuk if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan)) 20591702cce7SAlek Pinchuk spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start; 20601702cce7SAlek Pinchuk else 20611702cce7SAlek Pinchuk spa->spa_scan_pass_scrub_pause = 0; 20621702cce7SAlek Pinchuk spa->spa_scan_pass_scrub_spent_paused = 0; 20633f9d6ad7SLin Ling spa->spa_scan_pass_exam = 0; 20643f9d6ad7SLin Ling vdev_scan_stat_init(spa->spa_root_vdev); 20653f9d6ad7SLin Ling } 20663f9d6ad7SLin Ling 20673f9d6ad7SLin Ling /* 20683f9d6ad7SLin Ling * Get scan stats for zpool status reports 20693f9d6ad7SLin Ling */ 20703f9d6ad7SLin Ling int 20713f9d6ad7SLin Ling spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) 20723f9d6ad7SLin Ling { 20733f9d6ad7SLin Ling dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; 20743f9d6ad7SLin Ling 20753f9d6ad7SLin Ling if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) 2076be6fd75aSMatthew Ahrens return (SET_ERROR(ENOENT)); 20773f9d6ad7SLin Ling bzero(ps, sizeof (pool_scan_stat_t)); 20783f9d6ad7SLin Ling 20793f9d6ad7SLin Ling /* data stored on disk */ 20803f9d6ad7SLin Ling ps->pss_func = scn->scn_phys.scn_func; 20813f9d6ad7SLin Ling ps->pss_start_time = scn->scn_phys.scn_start_time; 20823f9d6ad7SLin Ling ps->pss_end_time = scn->scn_phys.scn_end_time; 20833f9d6ad7SLin Ling ps->pss_to_examine = scn->scn_phys.scn_to_examine; 20843f9d6ad7SLin Ling ps->pss_examined = scn->scn_phys.scn_examined; 20853f9d6ad7SLin Ling ps->pss_to_process = scn->scn_phys.scn_to_process; 20863f9d6ad7SLin Ling ps->pss_processed = scn->scn_phys.scn_processed; 20873f9d6ad7SLin Ling ps->pss_errors = scn->scn_phys.scn_errors; 20883f9d6ad7SLin Ling ps->pss_state = scn->scn_phys.scn_state; 20893f9d6ad7SLin Ling 20903f9d6ad7SLin Ling /* data not stored on disk */ 20913f9d6ad7SLin Ling ps->pss_pass_start = spa->spa_scan_pass_start; 20923f9d6ad7SLin Ling ps->pss_pass_exam = spa->spa_scan_pass_exam; 20931702cce7SAlek Pinchuk ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause; 20941702cce7SAlek Pinchuk ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused; 20953f9d6ad7SLin Ling 20963f9d6ad7SLin Ling return (0); 20973f9d6ad7SLin Ling } 209809c9d376SGeorge Wilson 2099b5152584SMatthew Ahrens int 2100b5152584SMatthew Ahrens spa_maxblocksize(spa_t *spa) 2101b5152584SMatthew Ahrens { 2102b5152584SMatthew Ahrens if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) 2103b5152584SMatthew Ahrens return (SPA_MAXBLOCKSIZE); 2104b5152584SMatthew Ahrens else 2105b5152584SMatthew Ahrens return (SPA_OLD_MAXBLOCKSIZE); 2106b5152584SMatthew Ahrens } 21075cabbc6bSPrashanth Sreenivasa 210854811da5SToomas Soome int 210954811da5SToomas Soome spa_maxdnodesize(spa_t *spa) 211054811da5SToomas Soome { 211154811da5SToomas Soome if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) 211254811da5SToomas Soome return (DNODE_MAX_SIZE); 211354811da5SToomas Soome else 211454811da5SToomas Soome return (DNODE_MIN_SIZE); 211554811da5SToomas Soome } 211654811da5SToomas Soome 21175cabbc6bSPrashanth Sreenivasa /* 21185cabbc6bSPrashanth Sreenivasa * Returns the txg that the last device removal completed. No indirect mappings 21195cabbc6bSPrashanth Sreenivasa * have been added since this txg. 21205cabbc6bSPrashanth Sreenivasa */ 21215cabbc6bSPrashanth Sreenivasa uint64_t 21225cabbc6bSPrashanth Sreenivasa spa_get_last_removal_txg(spa_t *spa) 21235cabbc6bSPrashanth Sreenivasa { 21245cabbc6bSPrashanth Sreenivasa uint64_t vdevid; 21255cabbc6bSPrashanth Sreenivasa uint64_t ret = -1ULL; 21265cabbc6bSPrashanth Sreenivasa 21275cabbc6bSPrashanth Sreenivasa spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 21285cabbc6bSPrashanth Sreenivasa /* 21295cabbc6bSPrashanth Sreenivasa * sr_prev_indirect_vdev is only modified while holding all the 21305cabbc6bSPrashanth Sreenivasa * config locks, so it is sufficient to hold SCL_VDEV as reader when 21315cabbc6bSPrashanth Sreenivasa * examining it. 21325cabbc6bSPrashanth Sreenivasa */ 21335cabbc6bSPrashanth Sreenivasa vdevid = spa->spa_removing_phys.sr_prev_indirect_vdev; 21345cabbc6bSPrashanth Sreenivasa 21355cabbc6bSPrashanth Sreenivasa while (vdevid != -1ULL) { 21365cabbc6bSPrashanth Sreenivasa vdev_t *vd = vdev_lookup_top(spa, vdevid); 21375cabbc6bSPrashanth Sreenivasa vdev_indirect_births_t *vib = vd->vdev_indirect_births; 21385cabbc6bSPrashanth Sreenivasa 21395cabbc6bSPrashanth Sreenivasa ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops); 21405cabbc6bSPrashanth Sreenivasa 21415cabbc6bSPrashanth Sreenivasa /* 21425cabbc6bSPrashanth Sreenivasa * If the removal did not remap any data, we don't care. 21435cabbc6bSPrashanth Sreenivasa */ 21445cabbc6bSPrashanth Sreenivasa if (vdev_indirect_births_count(vib) != 0) { 21455cabbc6bSPrashanth Sreenivasa ret = vdev_indirect_births_last_entry_txg(vib); 21465cabbc6bSPrashanth Sreenivasa break; 21475cabbc6bSPrashanth Sreenivasa } 21485cabbc6bSPrashanth Sreenivasa 21495cabbc6bSPrashanth Sreenivasa vdevid = vd->vdev_indirect_config.vic_prev_indirect_vdev; 21505cabbc6bSPrashanth Sreenivasa } 21515cabbc6bSPrashanth Sreenivasa spa_config_exit(spa, SCL_VDEV, FTAG); 21525cabbc6bSPrashanth Sreenivasa 21535cabbc6bSPrashanth Sreenivasa IMPLY(ret != -1ULL, 21545cabbc6bSPrashanth Sreenivasa spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); 21555cabbc6bSPrashanth Sreenivasa 21565cabbc6bSPrashanth Sreenivasa return (ret); 21575cabbc6bSPrashanth Sreenivasa } 21586f793812SPavel Zakharov 21596f793812SPavel Zakharov boolean_t 21606f793812SPavel Zakharov spa_trust_config(spa_t *spa) 21616f793812SPavel Zakharov { 21626f793812SPavel Zakharov return (spa->spa_trust_config); 21636f793812SPavel Zakharov } 21646f793812SPavel Zakharov 21656f793812SPavel Zakharov uint64_t 21666f793812SPavel Zakharov spa_missing_tvds_allowed(spa_t *spa) 21676f793812SPavel Zakharov { 21686f793812SPavel Zakharov return (spa->spa_missing_tvds_allowed); 21696f793812SPavel Zakharov } 21706f793812SPavel Zakharov 21716f793812SPavel Zakharov void 21726f793812SPavel Zakharov spa_set_missing_tvds(spa_t *spa, uint64_t missing) 21736f793812SPavel Zakharov { 21746f793812SPavel Zakharov spa->spa_missing_tvds = missing; 21756f793812SPavel Zakharov } 217686714001SSerapheim Dimitropoulos 217786714001SSerapheim Dimitropoulos boolean_t 217886714001SSerapheim Dimitropoulos spa_top_vdevs_spacemap_addressable(spa_t *spa) 217986714001SSerapheim Dimitropoulos { 218086714001SSerapheim Dimitropoulos vdev_t *rvd = spa->spa_root_vdev; 218186714001SSerapheim Dimitropoulos for (uint64_t c = 0; c < rvd->vdev_children; c++) { 218286714001SSerapheim Dimitropoulos if (!vdev_is_spacemap_addressable(rvd->vdev_child[c])) 218386714001SSerapheim Dimitropoulos return (B_FALSE); 218486714001SSerapheim Dimitropoulos } 218586714001SSerapheim Dimitropoulos return (B_TRUE); 218686714001SSerapheim Dimitropoulos } 218786714001SSerapheim Dimitropoulos 218886714001SSerapheim Dimitropoulos boolean_t 218986714001SSerapheim Dimitropoulos spa_has_checkpoint(spa_t *spa) 219086714001SSerapheim Dimitropoulos { 219186714001SSerapheim Dimitropoulos return (spa->spa_checkpoint_txg != 0); 219286714001SSerapheim Dimitropoulos } 219386714001SSerapheim Dimitropoulos 219486714001SSerapheim Dimitropoulos boolean_t 219586714001SSerapheim Dimitropoulos spa_importing_readonly_checkpoint(spa_t *spa) 219686714001SSerapheim Dimitropoulos { 219786714001SSerapheim Dimitropoulos return ((spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT) && 219886714001SSerapheim Dimitropoulos spa->spa_mode == FREAD); 219986714001SSerapheim Dimitropoulos } 220086714001SSerapheim Dimitropoulos 220186714001SSerapheim Dimitropoulos uint64_t 220286714001SSerapheim Dimitropoulos spa_min_claim_txg(spa_t *spa) 220386714001SSerapheim Dimitropoulos { 220486714001SSerapheim Dimitropoulos uint64_t checkpoint_txg = spa->spa_uberblock.ub_checkpoint_txg; 220586714001SSerapheim Dimitropoulos 220686714001SSerapheim Dimitropoulos if (checkpoint_txg != 0) 220786714001SSerapheim Dimitropoulos return (checkpoint_txg + 1); 220886714001SSerapheim Dimitropoulos 220986714001SSerapheim Dimitropoulos return (spa->spa_first_txg); 221086714001SSerapheim Dimitropoulos } 221186714001SSerapheim Dimitropoulos 221286714001SSerapheim Dimitropoulos /* 221386714001SSerapheim Dimitropoulos * If there is a checkpoint, async destroys may consume more space from 221486714001SSerapheim Dimitropoulos * the pool instead of freeing it. In an attempt to save the pool from 221586714001SSerapheim Dimitropoulos * getting suspended when it is about to run out of space, we stop 221686714001SSerapheim Dimitropoulos * processing async destroys. 221786714001SSerapheim Dimitropoulos */ 221886714001SSerapheim Dimitropoulos boolean_t 221986714001SSerapheim Dimitropoulos spa_suspend_async_destroy(spa_t *spa) 222086714001SSerapheim Dimitropoulos { 222186714001SSerapheim Dimitropoulos dsl_pool_t *dp = spa_get_dsl(spa); 222286714001SSerapheim Dimitropoulos 222386714001SSerapheim Dimitropoulos uint64_t unreserved = dsl_pool_unreserved_space(dp, 222486714001SSerapheim Dimitropoulos ZFS_SPACE_CHECK_EXTRA_RESERVED); 222586714001SSerapheim Dimitropoulos uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes; 222686714001SSerapheim Dimitropoulos uint64_t avail = (unreserved > used) ? (unreserved - used) : 0; 222786714001SSerapheim Dimitropoulos 222886714001SSerapheim Dimitropoulos if (spa_has_checkpoint(spa) && avail == 0) 222986714001SSerapheim Dimitropoulos return (B_TRUE); 223086714001SSerapheim Dimitropoulos 223186714001SSerapheim Dimitropoulos return (B_FALSE); 223286714001SSerapheim Dimitropoulos } 2233