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. 23814dcd43SSerapheim Dimitropoulos * Copyright (c) 2011, 2019 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. 29663207adSDon Brady * Copyright (c) 2017, Intel Corporation. 30fa9e4066Sahrens */ 31fa9e4066Sahrens 32fa9e4066Sahrens #include <sys/zfs_context.h> 33fa9e4066Sahrens #include <sys/spa_impl.h> 34283b8460SGeorge.Wilson #include <sys/spa_boot.h> 35fa9e4066Sahrens #include <sys/zio.h> 36fa9e4066Sahrens #include <sys/zio_checksum.h> 37fa9e4066Sahrens #include <sys/zio_compress.h> 38fa9e4066Sahrens #include <sys/dmu.h> 39fa9e4066Sahrens #include <sys/dmu_tx.h> 40fa9e4066Sahrens #include <sys/zap.h> 41fa9e4066Sahrens #include <sys/zil.h> 42fa9e4066Sahrens #include <sys/vdev_impl.h> 43094e47e9SGeorge Wilson #include <sys/vdev_initialize.h> 44084fd14fSBrian Behlendorf #include <sys/vdev_trim.h> 45fa9e4066Sahrens #include <sys/metaslab.h> 46fa9e4066Sahrens #include <sys/uberblock_impl.h> 47fa9e4066Sahrens #include <sys/txg.h> 48fa9e4066Sahrens #include <sys/avl.h> 49fa9e4066Sahrens #include <sys/unique.h> 50fa9e4066Sahrens #include <sys/dsl_pool.h> 51fa9e4066Sahrens #include <sys/dsl_dir.h> 52fa9e4066Sahrens #include <sys/dsl_prop.h> 533f9d6ad7SLin Ling #include <sys/dsl_scan.h> 54fa9e4066Sahrens #include <sys/fs/zfs.h> 556ce0521aSperrin #include <sys/metaslab_impl.h> 56e14bb325SJeff Bonwick #include <sys/arc.h> 57485bbbf5SGeorge Wilson #include <sys/ddt.h> 5891ebeef5Sahrens #include "zfs_prop.h" 59*4d7988d6SPaul Dagnelie #include <sys/btree.h> 6045818ee1SMatthew Ahrens #include <sys/zfeature.h> 61fa9e4066Sahrens 62fa9e4066Sahrens /* 63fa9e4066Sahrens * SPA locking 64fa9e4066Sahrens * 65814dcd43SSerapheim Dimitropoulos * There are three basic locks for managing spa_t structures: 66fa9e4066Sahrens * 67fa9e4066Sahrens * spa_namespace_lock (global mutex) 68fa9e4066Sahrens * 6944cd46caSbillm * This lock must be acquired to do any of the following: 70fa9e4066Sahrens * 7144cd46caSbillm * - Lookup a spa_t by name 7244cd46caSbillm * - Add or remove a spa_t from the namespace 7344cd46caSbillm * - Increase spa_refcount from non-zero 7444cd46caSbillm * - Check if spa_refcount is zero 7544cd46caSbillm * - Rename a spa_t 76ea8dc4b6Seschrock * - add/remove/attach/detach devices 7744cd46caSbillm * - Held for the duration of create/destroy/import/export 78fa9e4066Sahrens * 7944cd46caSbillm * It does not need to handle recursion. A create or destroy may 8044cd46caSbillm * reference objects (files or zvols) in other pools, but by 8144cd46caSbillm * definition they must have an existing reference, and will never need 8244cd46caSbillm * to lookup a spa_t by name. 83fa9e4066Sahrens * 84e914ace2STim Schumacher * spa_refcount (per-spa zfs_refcount_t protected by mutex) 85fa9e4066Sahrens * 8644cd46caSbillm * This reference count keep track of any active users of the spa_t. The 8744cd46caSbillm * spa_t cannot be destroyed or freed while this is non-zero. Internally, 8844cd46caSbillm * the refcount is never really 'zero' - opening a pool implicitly keeps 89088f3894Sahrens * some references in the DMU. Internally we check against spa_minref, but 9044cd46caSbillm * present the image of a zero/non-zero value to consumers. 91fa9e4066Sahrens * 92e14bb325SJeff Bonwick * spa_config_lock[] (per-spa array of rwlocks) 93fa9e4066Sahrens * 9491ebeef5Sahrens * This protects the spa_t from config changes, and must be held in 9591ebeef5Sahrens * the following circumstances: 96fa9e4066Sahrens * 9744cd46caSbillm * - RW_READER to perform I/O to the spa 9844cd46caSbillm * - RW_WRITER to change the vdev config 99fa9e4066Sahrens * 100fa9e4066Sahrens * The locking order is fairly straightforward: 101fa9e4066Sahrens * 10244cd46caSbillm * spa_namespace_lock -> spa_refcount 103fa9e4066Sahrens * 10444cd46caSbillm * The namespace lock must be acquired to increase the refcount from 0 10544cd46caSbillm * or to check if it is zero. 106fa9e4066Sahrens * 107e14bb325SJeff Bonwick * spa_refcount -> spa_config_lock[] 108fa9e4066Sahrens * 10944cd46caSbillm * There must be at least one valid reference on the spa_t to acquire 11044cd46caSbillm * the config lock. 111fa9e4066Sahrens * 112e14bb325SJeff Bonwick * spa_namespace_lock -> spa_config_lock[] 113fa9e4066Sahrens * 11444cd46caSbillm * The namespace lock must always be taken before the config lock. 115fa9e4066Sahrens * 116fa9e4066Sahrens * 117e14bb325SJeff Bonwick * The spa_namespace_lock can be acquired directly and is globally visible. 118fa9e4066Sahrens * 119e14bb325SJeff Bonwick * The namespace is manipulated using the following functions, all of which 120e14bb325SJeff Bonwick * require the spa_namespace_lock to be held. 121fa9e4066Sahrens * 12244cd46caSbillm * spa_lookup() Lookup a spa_t by name. 123fa9e4066Sahrens * 12444cd46caSbillm * spa_add() Create a new spa_t in the namespace. 125fa9e4066Sahrens * 12644cd46caSbillm * spa_remove() Remove a spa_t from the namespace. This also 12744cd46caSbillm * frees up any memory associated with the spa_t. 128fa9e4066Sahrens * 12944cd46caSbillm * spa_next() Returns the next spa_t in the system, or the 13044cd46caSbillm * first if NULL is passed. 131fa9e4066Sahrens * 13244cd46caSbillm * spa_evict_all() Shutdown and remove all spa_t structures in 13344cd46caSbillm * the system. 134fa9e4066Sahrens * 135ea8dc4b6Seschrock * spa_guid_exists() Determine whether a pool/device guid exists. 136fa9e4066Sahrens * 137fa9e4066Sahrens * The spa_refcount is manipulated using the following functions: 138fa9e4066Sahrens * 13944cd46caSbillm * spa_open_ref() Adds a reference to the given spa_t. Must be 14044cd46caSbillm * called with spa_namespace_lock held if the 14144cd46caSbillm * refcount is currently zero. 142fa9e4066Sahrens * 14344cd46caSbillm * spa_close() Remove a reference from the spa_t. This will 14444cd46caSbillm * not free the spa_t or remove it from the 14544cd46caSbillm * namespace. No locking is required. 146fa9e4066Sahrens * 14744cd46caSbillm * spa_refcount_zero() Returns true if the refcount is currently 14844cd46caSbillm * zero. Must be called with spa_namespace_lock 14944cd46caSbillm * held. 150fa9e4066Sahrens * 151e14bb325SJeff Bonwick * The spa_config_lock[] is an array of rwlocks, ordered as follows: 152e14bb325SJeff Bonwick * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV. 153e14bb325SJeff Bonwick * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}(). 154e14bb325SJeff Bonwick * 155e14bb325SJeff Bonwick * To read the configuration, it suffices to hold one of these locks as reader. 156e14bb325SJeff Bonwick * To modify the configuration, you must hold all locks as writer. To modify 157e14bb325SJeff Bonwick * vdev state without altering the vdev tree's topology (e.g. online/offline), 158e14bb325SJeff Bonwick * you must hold SCL_STATE and SCL_ZIO as writer. 159e14bb325SJeff Bonwick * 160e14bb325SJeff Bonwick * We use these distinct config locks to avoid recursive lock entry. 161e14bb325SJeff Bonwick * For example, spa_sync() (which holds SCL_CONFIG as reader) induces 162e14bb325SJeff Bonwick * block allocations (SCL_ALLOC), which may require reading space maps 163e14bb325SJeff Bonwick * from disk (dmu_read() -> zio_read() -> SCL_ZIO). 164e14bb325SJeff Bonwick * 165e14bb325SJeff Bonwick * The spa config locks cannot be normal rwlocks because we need the 166e14bb325SJeff Bonwick * ability to hand off ownership. For example, SCL_ZIO is acquired 167e14bb325SJeff Bonwick * by the issuing thread and later released by an interrupt thread. 168e14bb325SJeff Bonwick * They do, however, obey the usual write-wanted semantics to prevent 169e14bb325SJeff Bonwick * writer (i.e. system administrator) starvation. 170e14bb325SJeff Bonwick * 171e14bb325SJeff Bonwick * The lock acquisition rules are as follows: 172e14bb325SJeff Bonwick * 173e14bb325SJeff Bonwick * SCL_CONFIG 174e14bb325SJeff Bonwick * Protects changes to the vdev tree topology, such as vdev 175e14bb325SJeff Bonwick * add/remove/attach/detach. Protects the dirty config list 176e14bb325SJeff Bonwick * (spa_config_dirty_list) and the set of spares and l2arc devices. 177e14bb325SJeff Bonwick * 178e14bb325SJeff Bonwick * SCL_STATE 179e14bb325SJeff Bonwick * Protects changes to pool state and vdev state, such as vdev 180e14bb325SJeff Bonwick * online/offline/fault/degrade/clear. Protects the dirty state list 181e14bb325SJeff Bonwick * (spa_state_dirty_list) and global pool state (spa_state). 182e14bb325SJeff Bonwick * 183e14bb325SJeff Bonwick * SCL_ALLOC 184e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 185e14bb325SJeff Bonwick * Held as reader by metaslab_alloc() and metaslab_claim(). 186e14bb325SJeff Bonwick * 187e14bb325SJeff Bonwick * SCL_ZIO 188e14bb325SJeff Bonwick * Held by bp-level zios (those which have no io_vd upon entry) 189e14bb325SJeff Bonwick * to prevent changes to the vdev tree. The bp-level zio implicitly 190e14bb325SJeff Bonwick * protects all of its vdev child zios, which do not hold SCL_ZIO. 191e14bb325SJeff Bonwick * 192e14bb325SJeff Bonwick * SCL_FREE 193e14bb325SJeff Bonwick * Protects changes to metaslab groups and classes. 194e14bb325SJeff Bonwick * Held as reader by metaslab_free(). SCL_FREE is distinct from 195e14bb325SJeff Bonwick * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free 196e14bb325SJeff Bonwick * blocks in zio_done() while another i/o that holds either 197e14bb325SJeff Bonwick * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete. 198e14bb325SJeff Bonwick * 199e14bb325SJeff Bonwick * SCL_VDEV 200e14bb325SJeff Bonwick * Held as reader to prevent changes to the vdev tree during trivial 201b24ab676SJeff Bonwick * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the 202e14bb325SJeff Bonwick * other locks, and lower than all of them, to ensure that it's safe 203e14bb325SJeff Bonwick * to acquire regardless of caller context. 204e14bb325SJeff Bonwick * 205e14bb325SJeff Bonwick * In addition, the following rules apply: 206e14bb325SJeff Bonwick * 207e14bb325SJeff Bonwick * (a) spa_props_lock protects pool properties, spa_config and spa_config_list. 208e14bb325SJeff Bonwick * The lock ordering is SCL_CONFIG > spa_props_lock. 209e14bb325SJeff Bonwick * 210e14bb325SJeff Bonwick * (b) I/O operations on leaf vdevs. For any zio operation that takes 211e14bb325SJeff Bonwick * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(), 212e14bb325SJeff Bonwick * or zio_write_phys() -- the caller must ensure that the config cannot 213e14bb325SJeff Bonwick * cannot change in the interim, and that the vdev cannot be reopened. 214e14bb325SJeff Bonwick * SCL_STATE as reader suffices for both. 215fa9e4066Sahrens * 216ea8dc4b6Seschrock * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit(). 217fa9e4066Sahrens * 21844cd46caSbillm * spa_vdev_enter() Acquire the namespace lock and the config lock 219ea8dc4b6Seschrock * for writing. 220fa9e4066Sahrens * 22144cd46caSbillm * spa_vdev_exit() Release the config lock, wait for all I/O 22244cd46caSbillm * to complete, sync the updated configs to the 223ea8dc4b6Seschrock * cache, and release the namespace lock. 224fa9e4066Sahrens * 225e14bb325SJeff Bonwick * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit(). 226e14bb325SJeff Bonwick * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual 227e14bb325SJeff Bonwick * locking is, always, based on spa_namespace_lock and spa_config_lock[]. 228fa9e4066Sahrens */ 229fa9e4066Sahrens 230fa9e4066Sahrens static avl_tree_t spa_namespace_avl; 231fa9e4066Sahrens kmutex_t spa_namespace_lock; 232fa9e4066Sahrens static kcondvar_t spa_namespace_cv; 2330373e76bSbonwick static int spa_active_count; 234416e0cd8Sek int spa_max_replication_override = SPA_DVAS_PER_BP; 235fa9e4066Sahrens 23699653d4eSeschrock static kmutex_t spa_spare_lock; 23739c23413Seschrock static avl_tree_t spa_spare_avl; 238fa94a07fSbrendan static kmutex_t spa_l2cache_lock; 239fa94a07fSbrendan static avl_tree_t spa_l2cache_avl; 24099653d4eSeschrock 241fa9e4066Sahrens kmem_cache_t *spa_buffer_pool; 2428ad4d6ddSJeff Bonwick int spa_mode_global; 243fa9e4066Sahrens 244fa9e4066Sahrens #ifdef ZFS_DEBUG 2455cabbc6bSPrashanth Sreenivasa /* 2465cabbc6bSPrashanth Sreenivasa * Everything except dprintf, spa, and indirect_remap is on by default 2475cabbc6bSPrashanth Sreenivasa * in debug builds. 2485cabbc6bSPrashanth Sreenivasa */ 24921f7c81cSMatthew Ahrens int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_INDIRECT_REMAP); 250fa9e4066Sahrens #else 251fa9e4066Sahrens int zfs_flags = 0; 252fa9e4066Sahrens #endif 253fa9e4066Sahrens 2540125049cSahrens /* 2550125049cSahrens * zfs_recover can be set to nonzero to attempt to recover from 2560125049cSahrens * otherwise-fatal errors, typically caused by on-disk corruption. When 2570125049cSahrens * set, calls to zfs_panic_recover() will turn into warning messages. 2588b36997aSMatthew Ahrens * This should only be used as a last resort, as it typically results 2598b36997aSMatthew Ahrens * in leaked space, or worse. 2600125049cSahrens */ 2617fd05ac4SMatthew Ahrens boolean_t zfs_recover = B_FALSE; 2627fd05ac4SMatthew Ahrens 2637fd05ac4SMatthew Ahrens /* 2647fd05ac4SMatthew Ahrens * If destroy encounters an EIO while reading metadata (e.g. indirect 2657fd05ac4SMatthew Ahrens * blocks), space referenced by the missing metadata can not be freed. 2667fd05ac4SMatthew Ahrens * Normally this causes the background destroy to become "stalled", as 2677fd05ac4SMatthew Ahrens * it is unable to make forward progress. While in this stalled state, 2687fd05ac4SMatthew Ahrens * all remaining space to free from the error-encountering filesystem is 2697fd05ac4SMatthew Ahrens * "temporarily leaked". Set this flag to cause it to ignore the EIO, 2707fd05ac4SMatthew Ahrens * permanently leak the space from indirect blocks that can not be read, 2717fd05ac4SMatthew Ahrens * and continue to free everything else that it can. 2727fd05ac4SMatthew Ahrens * 2737fd05ac4SMatthew Ahrens * The default, "stalling" behavior is useful if the storage partially 2747fd05ac4SMatthew Ahrens * fails (i.e. some but not all i/os fail), and then later recovers. In 2757fd05ac4SMatthew Ahrens * this case, we will be able to continue pool operations while it is 2767fd05ac4SMatthew Ahrens * partially failed, and when it recovers, we can continue to free the 2777fd05ac4SMatthew Ahrens * space, with no leaks. However, note that this case is actually 2787fd05ac4SMatthew Ahrens * fairly rare. 2797fd05ac4SMatthew Ahrens * 2807fd05ac4SMatthew Ahrens * Typically pools either (a) fail completely (but perhaps temporarily, 2817fd05ac4SMatthew Ahrens * e.g. a top-level vdev going offline), or (b) have localized, 2827fd05ac4SMatthew Ahrens * permanent errors (e.g. disk returns the wrong data due to bit flip or 2837fd05ac4SMatthew Ahrens * firmware bug). In case (a), this setting does not matter because the 2847fd05ac4SMatthew Ahrens * pool will be suspended and the sync thread will not be able to make 2857fd05ac4SMatthew Ahrens * forward progress regardless. In case (b), because the error is 2867fd05ac4SMatthew Ahrens * permanent, the best we can do is leak the minimum amount of space, 2877fd05ac4SMatthew Ahrens * which is what setting this flag will do. Therefore, it is reasonable 2887fd05ac4SMatthew Ahrens * for this flag to normally be set, but we chose the more conservative 2897fd05ac4SMatthew Ahrens * approach of not setting it, so that there is no possibility of 2907fd05ac4SMatthew Ahrens * leaking space in the "partial temporary" failure case. 2917fd05ac4SMatthew Ahrens */ 2927fd05ac4SMatthew Ahrens boolean_t zfs_free_leak_on_eio = B_FALSE; 2930125049cSahrens 29469962b56SMatthew Ahrens /* 29569962b56SMatthew Ahrens * Expiration time in milliseconds. This value has two meanings. First it is 29669962b56SMatthew Ahrens * used to determine when the spa_deadman() logic should fire. By default the 29769962b56SMatthew Ahrens * spa_deadman() will fire if spa_sync() has not completed in 1000 seconds. 29869962b56SMatthew Ahrens * Secondly, the value determines if an I/O is considered "hung". Any I/O that 29969962b56SMatthew Ahrens * has not completed in zfs_deadman_synctime_ms is considered "hung" resulting 30069962b56SMatthew Ahrens * in a system panic. 30169962b56SMatthew Ahrens */ 30269962b56SMatthew Ahrens uint64_t zfs_deadman_synctime_ms = 1000000ULL; 303283b8460SGeorge.Wilson 304283b8460SGeorge.Wilson /* 30569962b56SMatthew Ahrens * Check time in milliseconds. This defines the frequency at which we check 30669962b56SMatthew Ahrens * for hung I/O. 307283b8460SGeorge.Wilson */ 30869962b56SMatthew Ahrens uint64_t zfs_deadman_checktime_ms = 5000ULL; 309283b8460SGeorge.Wilson 310283b8460SGeorge.Wilson /* 311283b8460SGeorge.Wilson * Override the zfs deadman behavior via /etc/system. By default the 312283b8460SGeorge.Wilson * deadman is enabled except on VMware and sparc deployments. 313283b8460SGeorge.Wilson */ 314283b8460SGeorge.Wilson int zfs_deadman_enabled = -1; 315283b8460SGeorge.Wilson 31669962b56SMatthew Ahrens /* 31769962b56SMatthew Ahrens * The worst case is single-sector max-parity RAID-Z blocks, in which 31869962b56SMatthew Ahrens * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1) 31969962b56SMatthew Ahrens * times the size; so just assume that. Add to this the fact that 32069962b56SMatthew Ahrens * we can have up to 3 DVAs per bp, and one more factor of 2 because 32169962b56SMatthew Ahrens * the block may be dittoed with up to 3 DVAs by ddt_sync(). All together, 32269962b56SMatthew Ahrens * the worst case is: 32369962b56SMatthew Ahrens * (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24 32469962b56SMatthew Ahrens */ 32569962b56SMatthew Ahrens int spa_asize_inflation = 24; 326fa9e4066Sahrens 3277d46dc6cSMatthew Ahrens /* 3287d46dc6cSMatthew Ahrens * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in 3297d46dc6cSMatthew Ahrens * the pool to be consumed. This ensures that we don't run the pool 3307d46dc6cSMatthew Ahrens * completely out of space, due to unaccounted changes (e.g. to the MOS). 3317d46dc6cSMatthew Ahrens * It also limits the worst-case time to allocate space. If we have 3327d46dc6cSMatthew Ahrens * less than this amount of free space, most ZPL operations (e.g. write, 3337d46dc6cSMatthew Ahrens * create) will return ENOSPC. 3347d46dc6cSMatthew Ahrens * 3357d46dc6cSMatthew Ahrens * Certain operations (e.g. file removal, most administrative actions) can 3367d46dc6cSMatthew Ahrens * use half the slop space. They will only return ENOSPC if less than half 3377d46dc6cSMatthew Ahrens * the slop space is free. Typically, once the pool has less than the slop 3387d46dc6cSMatthew Ahrens * space free, the user will use these operations to free up space in the pool. 3397d46dc6cSMatthew Ahrens * These are the operations that call dsl_pool_adjustedsize() with the netfree 3407d46dc6cSMatthew Ahrens * argument set to TRUE. 3417d46dc6cSMatthew Ahrens * 34286714001SSerapheim Dimitropoulos * Operations that are almost guaranteed to free up space in the absence of 34386714001SSerapheim Dimitropoulos * a pool checkpoint can use up to three quarters of the slop space 34486714001SSerapheim Dimitropoulos * (e.g zfs destroy). 34586714001SSerapheim Dimitropoulos * 3467d46dc6cSMatthew Ahrens * A very restricted set of operations are always permitted, regardless of 3477d46dc6cSMatthew Ahrens * the amount of free space. These are the operations that call 34886714001SSerapheim Dimitropoulos * dsl_sync_task(ZFS_SPACE_CHECK_NONE). If these operations result in a net 34986714001SSerapheim Dimitropoulos * increase in the amount of space used, it is possible to run the pool 35086714001SSerapheim Dimitropoulos * completely out of space, causing it to be permanently read-only. 3517d46dc6cSMatthew Ahrens * 3524b5c8e93SMatthew Ahrens * Note that on very small pools, the slop space will be larger than 3534b5c8e93SMatthew Ahrens * 3.2%, in an effort to have it be at least spa_min_slop (128MB), 3544b5c8e93SMatthew Ahrens * but we never allow it to be more than half the pool size. 3554b5c8e93SMatthew Ahrens * 3567d46dc6cSMatthew Ahrens * See also the comments in zfs_space_check_t. 3577d46dc6cSMatthew Ahrens */ 3587d46dc6cSMatthew Ahrens int spa_slop_shift = 5; 3594b5c8e93SMatthew Ahrens uint64_t spa_min_slop = 128 * 1024 * 1024; 3607d46dc6cSMatthew Ahrens 361f78cdc34SPaul Dagnelie int spa_allocators = 4; 362f78cdc34SPaul Dagnelie 3633ee8c80cSPavel Zakharov /*PRINTFLIKE2*/ 3643ee8c80cSPavel Zakharov void 3653ee8c80cSPavel Zakharov spa_load_failed(spa_t *spa, const char *fmt, ...) 3663ee8c80cSPavel Zakharov { 3673ee8c80cSPavel Zakharov va_list adx; 3683ee8c80cSPavel Zakharov char buf[256]; 3693ee8c80cSPavel Zakharov 3703ee8c80cSPavel Zakharov va_start(adx, fmt); 3713ee8c80cSPavel Zakharov (void) vsnprintf(buf, sizeof (buf), fmt, adx); 3723ee8c80cSPavel Zakharov va_end(adx); 3733ee8c80cSPavel Zakharov 3746f793812SPavel Zakharov zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa->spa_name, 3756f793812SPavel Zakharov spa->spa_trust_config ? "trusted" : "untrusted", buf); 3763ee8c80cSPavel Zakharov } 3773ee8c80cSPavel Zakharov 3783ee8c80cSPavel Zakharov /*PRINTFLIKE2*/ 3793ee8c80cSPavel Zakharov void 3803ee8c80cSPavel Zakharov spa_load_note(spa_t *spa, const char *fmt, ...) 3813ee8c80cSPavel Zakharov { 3823ee8c80cSPavel Zakharov va_list adx; 3833ee8c80cSPavel Zakharov char buf[256]; 3843ee8c80cSPavel Zakharov 3853ee8c80cSPavel Zakharov va_start(adx, fmt); 3863ee8c80cSPavel Zakharov (void) vsnprintf(buf, sizeof (buf), fmt, adx); 3873ee8c80cSPavel Zakharov va_end(adx); 3883ee8c80cSPavel Zakharov 3896f793812SPavel Zakharov zfs_dbgmsg("spa_load(%s, config %s): %s", spa->spa_name, 3906f793812SPavel Zakharov spa->spa_trust_config ? "trusted" : "untrusted", buf); 3913ee8c80cSPavel Zakharov } 3923ee8c80cSPavel Zakharov 393663207adSDon Brady /* 394663207adSDon Brady * By default dedup and user data indirects land in the special class 395663207adSDon Brady */ 396663207adSDon Brady int zfs_ddt_data_is_special = B_TRUE; 397663207adSDon Brady int zfs_user_indirect_is_special = B_TRUE; 398663207adSDon Brady 399663207adSDon Brady /* 400663207adSDon Brady * The percentage of special class final space reserved for metadata only. 401663207adSDon Brady * Once we allocate 100 - zfs_special_class_metadata_reserve_pct we only 402663207adSDon Brady * let metadata into the class. 403663207adSDon Brady */ 404663207adSDon Brady int zfs_special_class_metadata_reserve_pct = 25; 405663207adSDon Brady 406e05725b1Sbonwick /* 407e05725b1Sbonwick * ========================================================================== 408e05725b1Sbonwick * SPA config locking 409e05725b1Sbonwick * ========================================================================== 410e05725b1Sbonwick */ 411e05725b1Sbonwick static void 412e14bb325SJeff Bonwick spa_config_lock_init(spa_t *spa) 413e14bb325SJeff Bonwick { 414e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 415e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 416e14bb325SJeff Bonwick mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL); 417e14bb325SJeff Bonwick cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL); 418e914ace2STim Schumacher zfs_refcount_create_untracked(&scl->scl_count); 419e14bb325SJeff Bonwick scl->scl_writer = NULL; 420e14bb325SJeff Bonwick scl->scl_write_wanted = 0; 421e14bb325SJeff Bonwick } 422e05725b1Sbonwick } 423e05725b1Sbonwick 424e05725b1Sbonwick static void 425e14bb325SJeff Bonwick spa_config_lock_destroy(spa_t *spa) 426e14bb325SJeff Bonwick { 427e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 428e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 429e14bb325SJeff Bonwick mutex_destroy(&scl->scl_lock); 430e14bb325SJeff Bonwick cv_destroy(&scl->scl_cv); 431e914ace2STim Schumacher zfs_refcount_destroy(&scl->scl_count); 432e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL); 433e14bb325SJeff Bonwick ASSERT(scl->scl_write_wanted == 0); 434e14bb325SJeff Bonwick } 435e14bb325SJeff Bonwick } 436e14bb325SJeff Bonwick 437e14bb325SJeff Bonwick int 438e14bb325SJeff Bonwick spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw) 439e05725b1Sbonwick { 440e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 441e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 442e14bb325SJeff Bonwick if (!(locks & (1 << i))) 443e14bb325SJeff Bonwick continue; 444e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 445e14bb325SJeff Bonwick if (rw == RW_READER) { 446e14bb325SJeff Bonwick if (scl->scl_writer || scl->scl_write_wanted) { 447e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 448e495b6e6SSaso Kiselkov spa_config_exit(spa, locks & ((1 << i) - 1), 449e495b6e6SSaso Kiselkov tag); 450e14bb325SJeff Bonwick return (0); 451e14bb325SJeff Bonwick } 452e14bb325SJeff Bonwick } else { 453e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 454e914ace2STim Schumacher if (!zfs_refcount_is_zero(&scl->scl_count)) { 455e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 456e495b6e6SSaso Kiselkov spa_config_exit(spa, locks & ((1 << i) - 1), 457e495b6e6SSaso Kiselkov tag); 458e14bb325SJeff Bonwick return (0); 459e14bb325SJeff Bonwick } 460e14bb325SJeff Bonwick scl->scl_writer = curthread; 461e14bb325SJeff Bonwick } 462e914ace2STim Schumacher (void) zfs_refcount_add(&scl->scl_count, tag); 463e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 464e14bb325SJeff Bonwick } 465e14bb325SJeff Bonwick return (1); 466e05725b1Sbonwick } 467e05725b1Sbonwick 468e05725b1Sbonwick void 469e14bb325SJeff Bonwick spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw) 470e05725b1Sbonwick { 471f64c0e34SEric Taylor int wlocks_held = 0; 472f64c0e34SEric Taylor 4733b2aab18SMatthew Ahrens ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY); 4743b2aab18SMatthew Ahrens 475e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 476e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 477f64c0e34SEric Taylor if (scl->scl_writer == curthread) 478f64c0e34SEric Taylor wlocks_held |= (1 << i); 479e14bb325SJeff Bonwick if (!(locks & (1 << i))) 480e14bb325SJeff Bonwick continue; 481e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 482e14bb325SJeff Bonwick if (rw == RW_READER) { 483e14bb325SJeff Bonwick while (scl->scl_writer || scl->scl_write_wanted) { 484e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 485e14bb325SJeff Bonwick } 486e14bb325SJeff Bonwick } else { 487e14bb325SJeff Bonwick ASSERT(scl->scl_writer != curthread); 488e914ace2STim Schumacher while (!zfs_refcount_is_zero(&scl->scl_count)) { 489e14bb325SJeff Bonwick scl->scl_write_wanted++; 490e14bb325SJeff Bonwick cv_wait(&scl->scl_cv, &scl->scl_lock); 491e14bb325SJeff Bonwick scl->scl_write_wanted--; 492e14bb325SJeff Bonwick } 493e14bb325SJeff Bonwick scl->scl_writer = curthread; 494e14bb325SJeff Bonwick } 495e914ace2STim Schumacher (void) zfs_refcount_add(&scl->scl_count, tag); 496e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 497e05725b1Sbonwick } 4985cabbc6bSPrashanth Sreenivasa ASSERT3U(wlocks_held, <=, locks); 499e05725b1Sbonwick } 500e05725b1Sbonwick 501e05725b1Sbonwick void 502e14bb325SJeff Bonwick spa_config_exit(spa_t *spa, int locks, void *tag) 503e05725b1Sbonwick { 504e14bb325SJeff Bonwick for (int i = SCL_LOCKS - 1; i >= 0; i--) { 505e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 506e14bb325SJeff Bonwick if (!(locks & (1 << i))) 507e14bb325SJeff Bonwick continue; 508e14bb325SJeff Bonwick mutex_enter(&scl->scl_lock); 509e914ace2STim Schumacher ASSERT(!zfs_refcount_is_zero(&scl->scl_count)); 510e914ace2STim Schumacher if (zfs_refcount_remove(&scl->scl_count, tag) == 0) { 511e14bb325SJeff Bonwick ASSERT(scl->scl_writer == NULL || 512e14bb325SJeff Bonwick scl->scl_writer == curthread); 513e14bb325SJeff Bonwick scl->scl_writer = NULL; /* OK in either case */ 514e14bb325SJeff Bonwick cv_broadcast(&scl->scl_cv); 515e14bb325SJeff Bonwick } 516e14bb325SJeff Bonwick mutex_exit(&scl->scl_lock); 517e05725b1Sbonwick } 518e05725b1Sbonwick } 519e05725b1Sbonwick 520e14bb325SJeff Bonwick int 521e14bb325SJeff Bonwick spa_config_held(spa_t *spa, int locks, krw_t rw) 522e05725b1Sbonwick { 523e14bb325SJeff Bonwick int locks_held = 0; 524e05725b1Sbonwick 525e14bb325SJeff Bonwick for (int i = 0; i < SCL_LOCKS; i++) { 526e14bb325SJeff Bonwick spa_config_lock_t *scl = &spa->spa_config_lock[i]; 527e14bb325SJeff Bonwick if (!(locks & (1 << i))) 528e14bb325SJeff Bonwick continue; 529e914ace2STim Schumacher if ((rw == RW_READER && 530e914ace2STim Schumacher !zfs_refcount_is_zero(&scl->scl_count)) || 531e14bb325SJeff Bonwick (rw == RW_WRITER && scl->scl_writer == curthread)) 532e14bb325SJeff Bonwick locks_held |= 1 << i; 533e14bb325SJeff Bonwick } 534e14bb325SJeff Bonwick 535e14bb325SJeff Bonwick return (locks_held); 536e05725b1Sbonwick } 537e05725b1Sbonwick 538fa9e4066Sahrens /* 539fa9e4066Sahrens * ========================================================================== 540fa9e4066Sahrens * SPA namespace functions 541fa9e4066Sahrens * ========================================================================== 542fa9e4066Sahrens */ 543fa9e4066Sahrens 544fa9e4066Sahrens /* 545fa9e4066Sahrens * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held. 546fa9e4066Sahrens * Returns NULL if no matching spa_t is found. 547fa9e4066Sahrens */ 548fa9e4066Sahrens spa_t * 549fa9e4066Sahrens spa_lookup(const char *name) 550fa9e4066Sahrens { 551e14bb325SJeff Bonwick static spa_t search; /* spa_t is large; don't allocate on stack */ 552e14bb325SJeff Bonwick spa_t *spa; 553fa9e4066Sahrens avl_index_t where; 55440feaa91Sahrens char *cp; 555fa9e4066Sahrens 556fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 557fa9e4066Sahrens 5583b2aab18SMatthew Ahrens (void) strlcpy(search.spa_name, name, sizeof (search.spa_name)); 5593b2aab18SMatthew Ahrens 56040feaa91Sahrens /* 56140feaa91Sahrens * If it's a full dataset name, figure out the pool name and 56240feaa91Sahrens * just use that. 56340feaa91Sahrens */ 56478f17100SMatthew Ahrens cp = strpbrk(search.spa_name, "/@#"); 5653b2aab18SMatthew Ahrens if (cp != NULL) 56640feaa91Sahrens *cp = '\0'; 56740feaa91Sahrens 568fa9e4066Sahrens spa = avl_find(&spa_namespace_avl, &search, &where); 569fa9e4066Sahrens 570fa9e4066Sahrens return (spa); 571fa9e4066Sahrens } 572fa9e4066Sahrens 573283b8460SGeorge.Wilson /* 574283b8460SGeorge.Wilson * Fires when spa_sync has not completed within zfs_deadman_synctime_ms. 575283b8460SGeorge.Wilson * If the zfs_deadman_enabled flag is set then it inspects all vdev queues 576283b8460SGeorge.Wilson * looking for potentially hung I/Os. 577283b8460SGeorge.Wilson */ 578283b8460SGeorge.Wilson void 579283b8460SGeorge.Wilson spa_deadman(void *arg) 580283b8460SGeorge.Wilson { 581283b8460SGeorge.Wilson spa_t *spa = arg; 582283b8460SGeorge.Wilson 5830713e232SGeorge Wilson /* 5840713e232SGeorge Wilson * Disable the deadman timer if the pool is suspended. 5850713e232SGeorge Wilson */ 5860713e232SGeorge Wilson if (spa_suspended(spa)) { 5870713e232SGeorge Wilson VERIFY(cyclic_reprogram(spa->spa_deadman_cycid, CY_INFINITY)); 5880713e232SGeorge Wilson return; 5890713e232SGeorge Wilson } 5900713e232SGeorge Wilson 591283b8460SGeorge.Wilson zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu", 592283b8460SGeorge.Wilson (gethrtime() - spa->spa_sync_starttime) / NANOSEC, 593283b8460SGeorge.Wilson ++spa->spa_deadman_calls); 594283b8460SGeorge.Wilson if (zfs_deadman_enabled) 595283b8460SGeorge.Wilson vdev_deadman(spa->spa_root_vdev); 596283b8460SGeorge.Wilson } 597283b8460SGeorge.Wilson 598814dcd43SSerapheim Dimitropoulos int 599814dcd43SSerapheim Dimitropoulos spa_log_sm_sort_by_txg(const void *va, const void *vb) 600814dcd43SSerapheim Dimitropoulos { 601814dcd43SSerapheim Dimitropoulos const spa_log_sm_t *a = va; 602814dcd43SSerapheim Dimitropoulos const spa_log_sm_t *b = vb; 603814dcd43SSerapheim Dimitropoulos 604*4d7988d6SPaul Dagnelie return (TREE_CMP(a->sls_txg, b->sls_txg)); 605814dcd43SSerapheim Dimitropoulos } 606814dcd43SSerapheim Dimitropoulos 607fa9e4066Sahrens /* 608fa9e4066Sahrens * Create an uninitialized spa_t with the given name. Requires 609fa9e4066Sahrens * spa_namespace_lock. The caller must ensure that the spa_t doesn't already 610fa9e4066Sahrens * exist by calling spa_lookup() first. 611fa9e4066Sahrens */ 612fa9e4066Sahrens spa_t * 613468c413aSTim Haley spa_add(const char *name, nvlist_t *config, const char *altroot) 614fa9e4066Sahrens { 615fa9e4066Sahrens spa_t *spa; 616c5904d13Seschrock spa_config_dirent_t *dp; 617283b8460SGeorge.Wilson cyc_handler_t hdlr; 618283b8460SGeorge.Wilson cyc_time_t when; 619fa9e4066Sahrens 620fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 621fa9e4066Sahrens 622fa9e4066Sahrens spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP); 623fa9e4066Sahrens 624c25056deSgw mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL); 625c25056deSgw mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL); 62635a5a358SJonathan Adams mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL); 627bc9014e6SJustin Gibbs mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL); 628c25056deSgw mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL); 62935a5a358SJonathan Adams mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL); 630c25056deSgw mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL); 63145818ee1SMatthew Ahrens mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL); 63235a5a358SJonathan Adams mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL); 633a1521560SJeff Bonwick mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL); 634a1521560SJeff Bonwick mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL); 635c3a66015SMatthew Ahrens mutex_init(&spa->spa_iokstat_lock, NULL, MUTEX_DEFAULT, NULL); 636814dcd43SSerapheim Dimitropoulos mutex_init(&spa->spa_flushed_ms_lock, NULL, MUTEX_DEFAULT, NULL); 637c25056deSgw 638c25056deSgw cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); 639bc9014e6SJustin Gibbs cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL); 64035a5a358SJonathan Adams cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL); 641c25056deSgw cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL); 642e14bb325SJeff Bonwick cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL); 643c25056deSgw 644b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 645cde58dbcSMatthew Ahrens bplist_create(&spa->spa_free_bplist[t]); 646b24ab676SJeff Bonwick 647e14bb325SJeff Bonwick (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name)); 648fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 649fa9e4066Sahrens spa->spa_freeze_txg = UINT64_MAX; 6500373e76bSbonwick spa->spa_final_txg = UINT64_MAX; 651468c413aSTim Haley spa->spa_load_max_txg = UINT64_MAX; 65235a5a358SJonathan Adams spa->spa_proc = &p0; 65335a5a358SJonathan Adams spa->spa_proc_state = SPA_PROC_NONE; 6546f793812SPavel Zakharov spa->spa_trust_config = B_TRUE; 655fa9e4066Sahrens 656283b8460SGeorge.Wilson hdlr.cyh_func = spa_deadman; 657283b8460SGeorge.Wilson hdlr.cyh_arg = spa; 658283b8460SGeorge.Wilson hdlr.cyh_level = CY_LOW_LEVEL; 659283b8460SGeorge.Wilson 66069962b56SMatthew Ahrens spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms); 661283b8460SGeorge.Wilson 662283b8460SGeorge.Wilson /* 663283b8460SGeorge.Wilson * This determines how often we need to check for hung I/Os after 664283b8460SGeorge.Wilson * the cyclic has already fired. Since checking for hung I/Os is 665283b8460SGeorge.Wilson * an expensive operation we don't want to check too frequently. 66669962b56SMatthew Ahrens * Instead wait for 5 seconds before checking again. 667283b8460SGeorge.Wilson */ 66869962b56SMatthew Ahrens when.cyt_interval = MSEC2NSEC(zfs_deadman_checktime_ms); 669283b8460SGeorge.Wilson when.cyt_when = CY_INFINITY; 670283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 671283b8460SGeorge.Wilson spa->spa_deadman_cycid = cyclic_add(&hdlr, &when); 672283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 673283b8460SGeorge.Wilson 674e914ace2STim Schumacher zfs_refcount_create(&spa->spa_refcount); 675e14bb325SJeff Bonwick spa_config_lock_init(spa); 676fa9e4066Sahrens 677fa9e4066Sahrens avl_add(&spa_namespace_avl, spa); 678fa9e4066Sahrens 6790373e76bSbonwick /* 6800373e76bSbonwick * Set the alternate root, if there is one. 6810373e76bSbonwick */ 6820373e76bSbonwick if (altroot) { 6830373e76bSbonwick spa->spa_root = spa_strdup(altroot); 6840373e76bSbonwick spa_active_count++; 6850373e76bSbonwick } 6860373e76bSbonwick 687f78cdc34SPaul Dagnelie spa->spa_alloc_count = spa_allocators; 688f78cdc34SPaul Dagnelie spa->spa_alloc_locks = kmem_zalloc(spa->spa_alloc_count * 689f78cdc34SPaul Dagnelie sizeof (kmutex_t), KM_SLEEP); 690f78cdc34SPaul Dagnelie spa->spa_alloc_trees = kmem_zalloc(spa->spa_alloc_count * 691f78cdc34SPaul Dagnelie sizeof (avl_tree_t), KM_SLEEP); 692f78cdc34SPaul Dagnelie for (int i = 0; i < spa->spa_alloc_count; i++) { 693f78cdc34SPaul Dagnelie mutex_init(&spa->spa_alloc_locks[i], NULL, MUTEX_DEFAULT, NULL); 694f78cdc34SPaul Dagnelie avl_create(&spa->spa_alloc_trees[i], zio_bookmark_compare, 695f78cdc34SPaul Dagnelie sizeof (zio_t), offsetof(zio_t, io_alloc_node)); 696f78cdc34SPaul Dagnelie } 697814dcd43SSerapheim Dimitropoulos avl_create(&spa->spa_metaslabs_by_flushed, metaslab_sort_by_flushed, 698814dcd43SSerapheim Dimitropoulos sizeof (metaslab_t), offsetof(metaslab_t, ms_spa_txg_node)); 699814dcd43SSerapheim Dimitropoulos avl_create(&spa->spa_sm_logs_by_txg, spa_log_sm_sort_by_txg, 700814dcd43SSerapheim Dimitropoulos sizeof (spa_log_sm_t), offsetof(spa_log_sm_t, sls_node)); 701814dcd43SSerapheim Dimitropoulos list_create(&spa->spa_log_summary, sizeof (log_summary_entry_t), 702814dcd43SSerapheim Dimitropoulos offsetof(log_summary_entry_t, lse_node)); 7030f7643c7SGeorge Wilson 704c5904d13Seschrock /* 705c5904d13Seschrock * Every pool starts with the default cachefile 706c5904d13Seschrock */ 707c5904d13Seschrock list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t), 708c5904d13Seschrock offsetof(spa_config_dirent_t, scd_link)); 709c5904d13Seschrock 710c5904d13Seschrock dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP); 711ef912c80STim Haley dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); 712c5904d13Seschrock list_insert_head(&spa->spa_config_list, dp); 713c5904d13Seschrock 7144b964adaSGeorge Wilson VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, 7154b964adaSGeorge Wilson KM_SLEEP) == 0); 7164b964adaSGeorge Wilson 717ad135b5dSChristopher Siden if (config != NULL) { 718ad135b5dSChristopher Siden nvlist_t *features; 719ad135b5dSChristopher Siden 720ad135b5dSChristopher Siden if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, 721ad135b5dSChristopher Siden &features) == 0) { 722ad135b5dSChristopher Siden VERIFY(nvlist_dup(features, &spa->spa_label_features, 723ad135b5dSChristopher Siden 0) == 0); 724ad135b5dSChristopher Siden } 725ad135b5dSChristopher Siden 726468c413aSTim Haley VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); 727ad135b5dSChristopher Siden } 728ad135b5dSChristopher Siden 729ad135b5dSChristopher Siden if (spa->spa_label_features == NULL) { 730ad135b5dSChristopher Siden VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, 731ad135b5dSChristopher Siden KM_SLEEP) == 0); 732ad135b5dSChristopher Siden } 733468c413aSTim Haley 734c3a66015SMatthew Ahrens spa->spa_iokstat = kstat_create("zfs", 0, name, 735c3a66015SMatthew Ahrens "disk", KSTAT_TYPE_IO, 1, 0); 736c3a66015SMatthew Ahrens if (spa->spa_iokstat) { 737c3a66015SMatthew Ahrens spa->spa_iokstat->ks_lock = &spa->spa_iokstat_lock; 738c3a66015SMatthew Ahrens kstat_install(spa->spa_iokstat); 739c3a66015SMatthew Ahrens } 740c3a66015SMatthew Ahrens 74181cd5c55SMatthew Ahrens spa->spa_min_ashift = INT_MAX; 74281cd5c55SMatthew Ahrens spa->spa_max_ashift = 0; 74381cd5c55SMatthew Ahrens 74443466aaeSMax Grossman /* 74543466aaeSMax Grossman * As a pool is being created, treat all features as disabled by 74643466aaeSMax Grossman * setting SPA_FEATURE_DISABLED for all entries in the feature 74743466aaeSMax Grossman * refcount cache. 74843466aaeSMax Grossman */ 74943466aaeSMax Grossman for (int i = 0; i < SPA_FEATURES; i++) { 75043466aaeSMax Grossman spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED; 75143466aaeSMax Grossman } 75243466aaeSMax Grossman 753e0f1c0afSOlaf Faaland list_create(&spa->spa_leaf_list, sizeof (vdev_t), 754e0f1c0afSOlaf Faaland offsetof(vdev_t, vdev_leaf_node)); 755e0f1c0afSOlaf Faaland 756fa9e4066Sahrens return (spa); 757fa9e4066Sahrens } 758fa9e4066Sahrens 759fa9e4066Sahrens /* 760fa9e4066Sahrens * Removes a spa_t from the namespace, freeing up any memory used. Requires 761fa9e4066Sahrens * spa_namespace_lock. This is called only after the spa_t has been closed and 762fa9e4066Sahrens * deactivated. 763fa9e4066Sahrens */ 764fa9e4066Sahrens void 765fa9e4066Sahrens spa_remove(spa_t *spa) 766fa9e4066Sahrens { 767c5904d13Seschrock spa_config_dirent_t *dp; 768c5904d13Seschrock 769fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 770814dcd43SSerapheim Dimitropoulos ASSERT(spa_state(spa) == POOL_STATE_UNINITIALIZED); 771e914ace2STim Schumacher ASSERT3U(zfs_refcount_count(&spa->spa_refcount), ==, 0); 772fa9e4066Sahrens 7731195e687SMark J Musante nvlist_free(spa->spa_config_splitting); 7741195e687SMark J Musante 775fa9e4066Sahrens avl_remove(&spa_namespace_avl, spa); 776fa9e4066Sahrens cv_broadcast(&spa_namespace_cv); 777fa9e4066Sahrens 7780373e76bSbonwick if (spa->spa_root) { 779fa9e4066Sahrens spa_strfree(spa->spa_root); 7800373e76bSbonwick spa_active_count--; 7810373e76bSbonwick } 782fa9e4066Sahrens 783c5904d13Seschrock while ((dp = list_head(&spa->spa_config_list)) != NULL) { 784c5904d13Seschrock list_remove(&spa->spa_config_list, dp); 785c5904d13Seschrock if (dp->scd_path != NULL) 786c5904d13Seschrock spa_strfree(dp->scd_path); 787c5904d13Seschrock kmem_free(dp, sizeof (spa_config_dirent_t)); 788c5904d13Seschrock } 789c5904d13Seschrock 790f78cdc34SPaul Dagnelie for (int i = 0; i < spa->spa_alloc_count; i++) { 791f78cdc34SPaul Dagnelie avl_destroy(&spa->spa_alloc_trees[i]); 792f78cdc34SPaul Dagnelie mutex_destroy(&spa->spa_alloc_locks[i]); 793f78cdc34SPaul Dagnelie } 794f78cdc34SPaul Dagnelie kmem_free(spa->spa_alloc_locks, spa->spa_alloc_count * 795f78cdc34SPaul Dagnelie sizeof (kmutex_t)); 796f78cdc34SPaul Dagnelie kmem_free(spa->spa_alloc_trees, spa->spa_alloc_count * 797f78cdc34SPaul Dagnelie sizeof (avl_tree_t)); 798f78cdc34SPaul Dagnelie 799814dcd43SSerapheim Dimitropoulos avl_destroy(&spa->spa_metaslabs_by_flushed); 800814dcd43SSerapheim Dimitropoulos avl_destroy(&spa->spa_sm_logs_by_txg); 801814dcd43SSerapheim Dimitropoulos list_destroy(&spa->spa_log_summary); 802c5904d13Seschrock list_destroy(&spa->spa_config_list); 803e0f1c0afSOlaf Faaland list_destroy(&spa->spa_leaf_list); 8042f8aaab3Seschrock 805ad135b5dSChristopher Siden nvlist_free(spa->spa_label_features); 8064b964adaSGeorge Wilson nvlist_free(spa->spa_load_info); 807fa9e4066Sahrens spa_config_set(spa, NULL); 808fa9e4066Sahrens 809283b8460SGeorge.Wilson mutex_enter(&cpu_lock); 810283b8460SGeorge.Wilson if (spa->spa_deadman_cycid != CYCLIC_NONE) 811283b8460SGeorge.Wilson cyclic_remove(spa->spa_deadman_cycid); 812283b8460SGeorge.Wilson mutex_exit(&cpu_lock); 813283b8460SGeorge.Wilson spa->spa_deadman_cycid = CYCLIC_NONE; 814283b8460SGeorge.Wilson 815e914ace2STim Schumacher zfs_refcount_destroy(&spa->spa_refcount); 81691ebeef5Sahrens 817e14bb325SJeff Bonwick spa_config_lock_destroy(spa); 818fa9e4066Sahrens 819c3a66015SMatthew Ahrens kstat_delete(spa->spa_iokstat); 820c3a66015SMatthew Ahrens spa->spa_iokstat = NULL; 821c3a66015SMatthew Ahrens 822b24ab676SJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 823cde58dbcSMatthew Ahrens bplist_destroy(&spa->spa_free_bplist[t]); 824b24ab676SJeff Bonwick 82545818ee1SMatthew Ahrens zio_checksum_templates_free(spa); 82645818ee1SMatthew Ahrens 827c25056deSgw cv_destroy(&spa->spa_async_cv); 828bc9014e6SJustin Gibbs cv_destroy(&spa->spa_evicting_os_cv); 82935a5a358SJonathan Adams cv_destroy(&spa->spa_proc_cv); 830c25056deSgw cv_destroy(&spa->spa_scrub_io_cv); 831e14bb325SJeff Bonwick cv_destroy(&spa->spa_suspend_cv); 832c25056deSgw 833814dcd43SSerapheim Dimitropoulos mutex_destroy(&spa->spa_flushed_ms_lock); 8345ad82045Snd mutex_destroy(&spa->spa_async_lock); 835c25056deSgw mutex_destroy(&spa->spa_errlist_lock); 83635a5a358SJonathan Adams mutex_destroy(&spa->spa_errlog_lock); 837bc9014e6SJustin Gibbs mutex_destroy(&spa->spa_evicting_os_lock); 83806eeb2adSek mutex_destroy(&spa->spa_history_lock); 83935a5a358SJonathan Adams mutex_destroy(&spa->spa_proc_lock); 840b1b8ab34Slling mutex_destroy(&spa->spa_props_lock); 84145818ee1SMatthew Ahrens mutex_destroy(&spa->spa_cksum_tmpls_lock); 84235a5a358SJonathan Adams mutex_destroy(&spa->spa_scrub_lock); 843e14bb325SJeff Bonwick mutex_destroy(&spa->spa_suspend_lock); 844a1521560SJeff Bonwick mutex_destroy(&spa->spa_vdev_top_lock); 845c3a66015SMatthew Ahrens mutex_destroy(&spa->spa_iokstat_lock); 8465ad82045Snd 847fa9e4066Sahrens kmem_free(spa, sizeof (spa_t)); 848fa9e4066Sahrens } 849fa9e4066Sahrens 850fa9e4066Sahrens /* 851fa9e4066Sahrens * Given a pool, return the next pool in the namespace, or NULL if there is 852fa9e4066Sahrens * none. If 'prev' is NULL, return the first pool. 853fa9e4066Sahrens */ 854fa9e4066Sahrens spa_t * 855fa9e4066Sahrens spa_next(spa_t *prev) 856fa9e4066Sahrens { 857fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 858fa9e4066Sahrens 859fa9e4066Sahrens if (prev) 860fa9e4066Sahrens return (AVL_NEXT(&spa_namespace_avl, prev)); 861fa9e4066Sahrens else 862fa9e4066Sahrens return (avl_first(&spa_namespace_avl)); 863fa9e4066Sahrens } 864fa9e4066Sahrens 865fa9e4066Sahrens /* 866fa9e4066Sahrens * ========================================================================== 867fa9e4066Sahrens * SPA refcount functions 868fa9e4066Sahrens * ========================================================================== 869fa9e4066Sahrens */ 870fa9e4066Sahrens 871fa9e4066Sahrens /* 872fa9e4066Sahrens * Add a reference to the given spa_t. Must have at least one reference, or 873fa9e4066Sahrens * have the namespace lock held. 874fa9e4066Sahrens */ 875fa9e4066Sahrens void 876fa9e4066Sahrens spa_open_ref(spa_t *spa, void *tag) 877fa9e4066Sahrens { 878e914ace2STim Schumacher ASSERT(zfs_refcount_count(&spa->spa_refcount) >= spa->spa_minref || 879fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 880e914ace2STim Schumacher (void) zfs_refcount_add(&spa->spa_refcount, tag); 881fa9e4066Sahrens } 882fa9e4066Sahrens 883fa9e4066Sahrens /* 884fa9e4066Sahrens * Remove a reference to the given spa_t. Must have at least one reference, or 885fa9e4066Sahrens * have the namespace lock held. 886fa9e4066Sahrens */ 887fa9e4066Sahrens void 888fa9e4066Sahrens spa_close(spa_t *spa, void *tag) 889fa9e4066Sahrens { 890e914ace2STim Schumacher ASSERT(zfs_refcount_count(&spa->spa_refcount) > spa->spa_minref || 891fa9e4066Sahrens MUTEX_HELD(&spa_namespace_lock)); 892e914ace2STim Schumacher (void) zfs_refcount_remove(&spa->spa_refcount, tag); 893fa9e4066Sahrens } 894fa9e4066Sahrens 895bc9014e6SJustin Gibbs /* 896bc9014e6SJustin Gibbs * Remove a reference to the given spa_t held by a dsl dir that is 897bc9014e6SJustin Gibbs * being asynchronously released. Async releases occur from a taskq 898bc9014e6SJustin Gibbs * performing eviction of dsl datasets and dirs. The namespace lock 899bc9014e6SJustin Gibbs * isn't held and the hold by the object being evicted may contribute to 900bc9014e6SJustin Gibbs * spa_minref (e.g. dataset or directory released during pool export), 901bc9014e6SJustin Gibbs * so the asserts in spa_close() do not apply. 902bc9014e6SJustin Gibbs */ 903bc9014e6SJustin Gibbs void 904bc9014e6SJustin Gibbs spa_async_close(spa_t *spa, void *tag) 905bc9014e6SJustin Gibbs { 906e914ace2STim Schumacher (void) zfs_refcount_remove(&spa->spa_refcount, tag); 907bc9014e6SJustin Gibbs } 908bc9014e6SJustin Gibbs 909fa9e4066Sahrens /* 910fa9e4066Sahrens * Check to see if the spa refcount is zero. Must be called with 911088f3894Sahrens * spa_namespace_lock held. We really compare against spa_minref, which is the 912fa9e4066Sahrens * number of references acquired when opening a pool 913fa9e4066Sahrens */ 914fa9e4066Sahrens boolean_t 915fa9e4066Sahrens spa_refcount_zero(spa_t *spa) 916fa9e4066Sahrens { 917fa9e4066Sahrens ASSERT(MUTEX_HELD(&spa_namespace_lock)); 918fa9e4066Sahrens 919e914ace2STim Schumacher return (zfs_refcount_count(&spa->spa_refcount) == spa->spa_minref); 920fa9e4066Sahrens } 921fa9e4066Sahrens 92299653d4eSeschrock /* 92399653d4eSeschrock * ========================================================================== 924fa94a07fSbrendan * SPA spare and l2cache tracking 92599653d4eSeschrock * ========================================================================== 92699653d4eSeschrock */ 92799653d4eSeschrock 928fa94a07fSbrendan /* 929fa94a07fSbrendan * Hot spares and cache devices are tracked using the same code below, 930fa94a07fSbrendan * for 'auxiliary' devices. 931fa94a07fSbrendan */ 932fa94a07fSbrendan 933fa94a07fSbrendan typedef struct spa_aux { 934fa94a07fSbrendan uint64_t aux_guid; 935fa94a07fSbrendan uint64_t aux_pool; 936fa94a07fSbrendan avl_node_t aux_avl; 937fa94a07fSbrendan int aux_count; 938fa94a07fSbrendan } spa_aux_t; 939fa94a07fSbrendan 940c4ab0d3fSGvozden Neskovic static inline int 941fa94a07fSbrendan spa_aux_compare(const void *a, const void *b) 942fa94a07fSbrendan { 943c4ab0d3fSGvozden Neskovic const spa_aux_t *sa = (const spa_aux_t *)a; 944c4ab0d3fSGvozden Neskovic const spa_aux_t *sb = (const spa_aux_t *)b; 945fa94a07fSbrendan 946*4d7988d6SPaul Dagnelie return (TREE_CMP(sa->aux_guid, sb->aux_guid)); 947fa94a07fSbrendan } 948fa94a07fSbrendan 949fa94a07fSbrendan void 950fa94a07fSbrendan spa_aux_add(vdev_t *vd, avl_tree_t *avl) 951fa94a07fSbrendan { 952fa94a07fSbrendan avl_index_t where; 953fa94a07fSbrendan spa_aux_t search; 954fa94a07fSbrendan spa_aux_t *aux; 955fa94a07fSbrendan 956fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 957fa94a07fSbrendan if ((aux = avl_find(avl, &search, &where)) != NULL) { 958fa94a07fSbrendan aux->aux_count++; 959fa94a07fSbrendan } else { 960fa94a07fSbrendan aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP); 961fa94a07fSbrendan aux->aux_guid = vd->vdev_guid; 962fa94a07fSbrendan aux->aux_count = 1; 963fa94a07fSbrendan avl_insert(avl, aux, where); 964fa94a07fSbrendan } 965fa94a07fSbrendan } 966fa94a07fSbrendan 967fa94a07fSbrendan void 968fa94a07fSbrendan spa_aux_remove(vdev_t *vd, avl_tree_t *avl) 969fa94a07fSbrendan { 970fa94a07fSbrendan spa_aux_t search; 971fa94a07fSbrendan spa_aux_t *aux; 972fa94a07fSbrendan avl_index_t where; 973fa94a07fSbrendan 974fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 975fa94a07fSbrendan aux = avl_find(avl, &search, &where); 976fa94a07fSbrendan 977fa94a07fSbrendan ASSERT(aux != NULL); 978fa94a07fSbrendan 979fa94a07fSbrendan if (--aux->aux_count == 0) { 980fa94a07fSbrendan avl_remove(avl, aux); 981fa94a07fSbrendan kmem_free(aux, sizeof (spa_aux_t)); 982fa94a07fSbrendan } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) { 983fa94a07fSbrendan aux->aux_pool = 0ULL; 984fa94a07fSbrendan } 985fa94a07fSbrendan } 986fa94a07fSbrendan 987fa94a07fSbrendan boolean_t 98889a89ebfSlling spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl) 989fa94a07fSbrendan { 990fa94a07fSbrendan spa_aux_t search, *found; 991fa94a07fSbrendan 992fa94a07fSbrendan search.aux_guid = guid; 99389a89ebfSlling found = avl_find(avl, &search, NULL); 994fa94a07fSbrendan 995fa94a07fSbrendan if (pool) { 996fa94a07fSbrendan if (found) 997fa94a07fSbrendan *pool = found->aux_pool; 998fa94a07fSbrendan else 999fa94a07fSbrendan *pool = 0ULL; 1000fa94a07fSbrendan } 1001fa94a07fSbrendan 100289a89ebfSlling if (refcnt) { 100389a89ebfSlling if (found) 100489a89ebfSlling *refcnt = found->aux_count; 100589a89ebfSlling else 100689a89ebfSlling *refcnt = 0; 100789a89ebfSlling } 100889a89ebfSlling 1009fa94a07fSbrendan return (found != NULL); 1010fa94a07fSbrendan } 1011fa94a07fSbrendan 1012fa94a07fSbrendan void 1013fa94a07fSbrendan spa_aux_activate(vdev_t *vd, avl_tree_t *avl) 1014fa94a07fSbrendan { 1015fa94a07fSbrendan spa_aux_t search, *found; 1016fa94a07fSbrendan avl_index_t where; 1017fa94a07fSbrendan 1018fa94a07fSbrendan search.aux_guid = vd->vdev_guid; 1019fa94a07fSbrendan found = avl_find(avl, &search, &where); 1020fa94a07fSbrendan ASSERT(found != NULL); 1021fa94a07fSbrendan ASSERT(found->aux_pool == 0ULL); 1022fa94a07fSbrendan 1023fa94a07fSbrendan found->aux_pool = spa_guid(vd->vdev_spa); 1024fa94a07fSbrendan } 1025fa94a07fSbrendan 102699653d4eSeschrock /* 102739c23413Seschrock * Spares are tracked globally due to the following constraints: 102839c23413Seschrock * 102954811da5SToomas Soome * - A spare may be part of multiple pools. 103054811da5SToomas Soome * - A spare may be added to a pool even if it's actively in use within 103139c23413Seschrock * another pool. 103254811da5SToomas Soome * - A spare in use in any pool can only be the source of a replacement if 103339c23413Seschrock * the target is a spare in the same pool. 103439c23413Seschrock * 103539c23413Seschrock * We keep track of all spares on the system through the use of a reference 103639c23413Seschrock * counted AVL tree. When a vdev is added as a spare, or used as a replacement 103739c23413Seschrock * spare, then we bump the reference count in the AVL tree. In addition, we set 103839c23413Seschrock * the 'vdev_isspare' member to indicate that the device is a spare (active or 103939c23413Seschrock * inactive). When a spare is made active (used to replace a device in the 104039c23413Seschrock * pool), we also keep track of which pool its been made a part of. 104139c23413Seschrock * 104239c23413Seschrock * The 'spa_spare_lock' protects the AVL tree. These functions are normally 104339c23413Seschrock * called under the spa_namespace lock as part of vdev reconfiguration. The 104439c23413Seschrock * separate spare lock exists for the status query path, which does not need to 104539c23413Seschrock * be completely consistent with respect to other vdev configuration changes. 104699653d4eSeschrock */ 104739c23413Seschrock 104899653d4eSeschrock static int 104999653d4eSeschrock spa_spare_compare(const void *a, const void *b) 105099653d4eSeschrock { 1051fa94a07fSbrendan return (spa_aux_compare(a, b)); 105299653d4eSeschrock } 105399653d4eSeschrock 105499653d4eSeschrock void 105539c23413Seschrock spa_spare_add(vdev_t *vd) 105699653d4eSeschrock { 105799653d4eSeschrock mutex_enter(&spa_spare_lock); 105839c23413Seschrock ASSERT(!vd->vdev_isspare); 1059fa94a07fSbrendan spa_aux_add(vd, &spa_spare_avl); 106039c23413Seschrock vd->vdev_isspare = B_TRUE; 106199653d4eSeschrock mutex_exit(&spa_spare_lock); 106299653d4eSeschrock } 106399653d4eSeschrock 106499653d4eSeschrock void 106539c23413Seschrock spa_spare_remove(vdev_t *vd) 106699653d4eSeschrock { 106799653d4eSeschrock mutex_enter(&spa_spare_lock); 106839c23413Seschrock ASSERT(vd->vdev_isspare); 1069fa94a07fSbrendan spa_aux_remove(vd, &spa_spare_avl); 107039c23413Seschrock vd->vdev_isspare = B_FALSE; 107199653d4eSeschrock mutex_exit(&spa_spare_lock); 107299653d4eSeschrock } 107399653d4eSeschrock 107499653d4eSeschrock boolean_t 107589a89ebfSlling spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt) 107699653d4eSeschrock { 1077fa94a07fSbrendan boolean_t found; 107899653d4eSeschrock 107999653d4eSeschrock mutex_enter(&spa_spare_lock); 108089a89ebfSlling found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl); 108199653d4eSeschrock mutex_exit(&spa_spare_lock); 108299653d4eSeschrock 1083fa94a07fSbrendan return (found); 108439c23413Seschrock } 108539c23413Seschrock 108639c23413Seschrock void 108739c23413Seschrock spa_spare_activate(vdev_t *vd) 108839c23413Seschrock { 108939c23413Seschrock mutex_enter(&spa_spare_lock); 109039c23413Seschrock ASSERT(vd->vdev_isspare); 1091fa94a07fSbrendan spa_aux_activate(vd, &spa_spare_avl); 1092fa94a07fSbrendan mutex_exit(&spa_spare_lock); 1093fa94a07fSbrendan } 109439c23413Seschrock 1095fa94a07fSbrendan /* 1096fa94a07fSbrendan * Level 2 ARC devices are tracked globally for the same reasons as spares. 1097fa94a07fSbrendan * Cache devices currently only support one pool per cache device, and so 1098fa94a07fSbrendan * for these devices the aux reference count is currently unused beyond 1. 1099fa94a07fSbrendan */ 110039c23413Seschrock 1101fa94a07fSbrendan static int 1102fa94a07fSbrendan spa_l2cache_compare(const void *a, const void *b) 1103fa94a07fSbrendan { 1104fa94a07fSbrendan return (spa_aux_compare(a, b)); 1105fa94a07fSbrendan } 1106fa94a07fSbrendan 1107fa94a07fSbrendan void 1108fa94a07fSbrendan spa_l2cache_add(vdev_t *vd) 1109fa94a07fSbrendan { 1110fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1111fa94a07fSbrendan ASSERT(!vd->vdev_isl2cache); 1112fa94a07fSbrendan spa_aux_add(vd, &spa_l2cache_avl); 1113fa94a07fSbrendan vd->vdev_isl2cache = B_TRUE; 1114fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1115fa94a07fSbrendan } 1116fa94a07fSbrendan 1117fa94a07fSbrendan void 1118fa94a07fSbrendan spa_l2cache_remove(vdev_t *vd) 1119fa94a07fSbrendan { 1120fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1121fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 1122fa94a07fSbrendan spa_aux_remove(vd, &spa_l2cache_avl); 1123fa94a07fSbrendan vd->vdev_isl2cache = B_FALSE; 1124fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1125fa94a07fSbrendan } 1126fa94a07fSbrendan 1127fa94a07fSbrendan boolean_t 1128fa94a07fSbrendan spa_l2cache_exists(uint64_t guid, uint64_t *pool) 1129fa94a07fSbrendan { 1130fa94a07fSbrendan boolean_t found; 1131fa94a07fSbrendan 1132fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 113389a89ebfSlling found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl); 1134fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1135fa94a07fSbrendan 1136fa94a07fSbrendan return (found); 1137fa94a07fSbrendan } 1138fa94a07fSbrendan 1139fa94a07fSbrendan void 1140fa94a07fSbrendan spa_l2cache_activate(vdev_t *vd) 1141fa94a07fSbrendan { 1142fa94a07fSbrendan mutex_enter(&spa_l2cache_lock); 1143fa94a07fSbrendan ASSERT(vd->vdev_isl2cache); 1144fa94a07fSbrendan spa_aux_activate(vd, &spa_l2cache_avl); 1145fa94a07fSbrendan mutex_exit(&spa_l2cache_lock); 1146fa94a07fSbrendan } 1147fa94a07fSbrendan 1148fa9e4066Sahrens /* 1149fa9e4066Sahrens * ========================================================================== 1150fa9e4066Sahrens * SPA vdev locking 1151fa9e4066Sahrens * ========================================================================== 1152fa9e4066Sahrens */ 1153fa9e4066Sahrens 1154fa9e4066Sahrens /* 1155ea8dc4b6Seschrock * Lock the given spa_t for the purpose of adding or removing a vdev. 1156ea8dc4b6Seschrock * Grabs the global spa_namespace_lock plus the spa config lock for writing. 1157fa9e4066Sahrens * It returns the next transaction group for the spa_t. 1158fa9e4066Sahrens */ 1159fa9e4066Sahrens uint64_t 1160fa9e4066Sahrens spa_vdev_enter(spa_t *spa) 1161fa9e4066Sahrens { 1162a1521560SJeff Bonwick mutex_enter(&spa->spa_vdev_top_lock); 1163bbfd46c4SJeff Bonwick mutex_enter(&spa_namespace_lock); 1164084fd14fSBrian Behlendorf 1165084fd14fSBrian Behlendorf vdev_autotrim_stop_all(spa); 1166084fd14fSBrian Behlendorf 116788ecc943SGeorge Wilson return (spa_vdev_config_enter(spa)); 116888ecc943SGeorge Wilson } 116988ecc943SGeorge Wilson 117088ecc943SGeorge Wilson /* 117188ecc943SGeorge Wilson * Internal implementation for spa_vdev_enter(). Used when a vdev 117288ecc943SGeorge Wilson * operation requires multiple syncs (i.e. removing a device) while 117388ecc943SGeorge Wilson * keeping the spa_namespace_lock held. 117488ecc943SGeorge Wilson */ 117588ecc943SGeorge Wilson uint64_t 117688ecc943SGeorge Wilson spa_vdev_config_enter(spa_t *spa) 117788ecc943SGeorge Wilson { 117888ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 11793d7072f8Seschrock 1180e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1181fa9e4066Sahrens 1182fa9e4066Sahrens return (spa_last_synced_txg(spa) + 1); 1183fa9e4066Sahrens } 1184fa9e4066Sahrens 1185fa9e4066Sahrens /* 118688ecc943SGeorge Wilson * Used in combination with spa_vdev_config_enter() to allow the syncing 118788ecc943SGeorge Wilson * of multiple transactions without releasing the spa_namespace_lock. 1188fa9e4066Sahrens */ 118988ecc943SGeorge Wilson void 119088ecc943SGeorge Wilson spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag) 1191fa9e4066Sahrens { 119288ecc943SGeorge Wilson ASSERT(MUTEX_HELD(&spa_namespace_lock)); 119388ecc943SGeorge Wilson 11940e34b6a7Sbonwick int config_changed = B_FALSE; 1195ea8dc4b6Seschrock 11960373e76bSbonwick ASSERT(txg > spa_last_synced_txg(spa)); 11970e34b6a7Sbonwick 1198e14bb325SJeff Bonwick spa->spa_pending_vdev = NULL; 1199e14bb325SJeff Bonwick 12000e34b6a7Sbonwick /* 12010e34b6a7Sbonwick * Reassess the DTLs. 12020e34b6a7Sbonwick */ 12030373e76bSbonwick vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE); 12040e34b6a7Sbonwick 1205e14bb325SJeff Bonwick if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) { 12060e34b6a7Sbonwick config_changed = B_TRUE; 12078f18d1faSGeorge Wilson spa->spa_config_generation++; 12080e34b6a7Sbonwick } 1209ea8dc4b6Seschrock 121088ecc943SGeorge Wilson /* 121188ecc943SGeorge Wilson * Verify the metaslab classes. 121288ecc943SGeorge Wilson */ 1213b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0); 1214b24ab676SJeff Bonwick ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0); 1215663207adSDon Brady ASSERT(metaslab_class_validate(spa_special_class(spa)) == 0); 1216663207adSDon Brady ASSERT(metaslab_class_validate(spa_dedup_class(spa)) == 0); 121788ecc943SGeorge Wilson 1218e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1219fa9e4066Sahrens 122088ecc943SGeorge Wilson /* 122188ecc943SGeorge Wilson * Panic the system if the specified tag requires it. This 122288ecc943SGeorge Wilson * is useful for ensuring that configurations are updated 122388ecc943SGeorge Wilson * transactionally. 122488ecc943SGeorge Wilson */ 122588ecc943SGeorge Wilson if (zio_injection_enabled) 12261195e687SMark J Musante zio_handle_panic_injection(spa, tag, 0); 122788ecc943SGeorge Wilson 1228fa9e4066Sahrens /* 1229fa9e4066Sahrens * Note: this txg_wait_synced() is important because it ensures 1230fa9e4066Sahrens * that there won't be more than one config change per txg. 1231fa9e4066Sahrens * This allows us to use the txg as the generation number. 1232fa9e4066Sahrens */ 1233fa9e4066Sahrens if (error == 0) 1234fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 1235fa9e4066Sahrens 1236fa9e4066Sahrens if (vd != NULL) { 12370713e232SGeorge Wilson ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL); 1238094e47e9SGeorge Wilson if (vd->vdev_ops->vdev_op_leaf) { 1239094e47e9SGeorge Wilson mutex_enter(&vd->vdev_initialize_lock); 1240084fd14fSBrian Behlendorf vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED, 1241084fd14fSBrian Behlendorf NULL); 1242094e47e9SGeorge Wilson mutex_exit(&vd->vdev_initialize_lock); 1243084fd14fSBrian Behlendorf 1244084fd14fSBrian Behlendorf mutex_enter(&vd->vdev_trim_lock); 1245084fd14fSBrian Behlendorf vdev_trim_stop(vd, VDEV_TRIM_CANCELED, NULL); 1246084fd14fSBrian Behlendorf mutex_exit(&vd->vdev_trim_lock); 1247094e47e9SGeorge Wilson } 1248094e47e9SGeorge Wilson 1249084fd14fSBrian Behlendorf /* 1250084fd14fSBrian Behlendorf * The vdev may be both a leaf and top-level device. 1251084fd14fSBrian Behlendorf */ 1252084fd14fSBrian Behlendorf vdev_autotrim_stop_wait(vd); 1253084fd14fSBrian Behlendorf 12548ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, spa, RW_WRITER); 1255fa9e4066Sahrens vdev_free(vd); 12568ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, spa); 1257fa9e4066Sahrens } 1258fa9e4066Sahrens 1259fa9e4066Sahrens /* 12600e34b6a7Sbonwick * If the config changed, update the config cache. 1261fa9e4066Sahrens */ 12620e34b6a7Sbonwick if (config_changed) 12635cabbc6bSPrashanth Sreenivasa spa_write_cachefile(spa, B_FALSE, B_TRUE); 126488ecc943SGeorge Wilson } 1265ea8dc4b6Seschrock 126688ecc943SGeorge Wilson /* 126788ecc943SGeorge Wilson * Unlock the spa_t after adding or removing a vdev. Besides undoing the 126888ecc943SGeorge Wilson * locking of spa_vdev_enter(), we also want make sure the transactions have 126988ecc943SGeorge Wilson * synced to disk, and then update the global configuration cache with the new 127088ecc943SGeorge Wilson * information. 127188ecc943SGeorge Wilson */ 127288ecc943SGeorge Wilson int 127388ecc943SGeorge Wilson spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error) 127488ecc943SGeorge Wilson { 1275084fd14fSBrian Behlendorf vdev_autotrim_restart(spa); 1276084fd14fSBrian Behlendorf 127788ecc943SGeorge Wilson spa_vdev_config_exit(spa, vd, txg, error, FTAG); 1278ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1279bbfd46c4SJeff Bonwick mutex_exit(&spa->spa_vdev_top_lock); 1280fa9e4066Sahrens 1281fa9e4066Sahrens return (error); 1282fa9e4066Sahrens } 1283fa9e4066Sahrens 1284e14bb325SJeff Bonwick /* 1285e14bb325SJeff Bonwick * Lock the given spa_t for the purpose of changing vdev state. 1286e14bb325SJeff Bonwick */ 1287e14bb325SJeff Bonwick void 12888f18d1faSGeorge Wilson spa_vdev_state_enter(spa_t *spa, int oplocks) 1289e14bb325SJeff Bonwick { 12908f18d1faSGeorge Wilson int locks = SCL_STATE_ALL | oplocks; 12918f18d1faSGeorge Wilson 1292dcba9f3fSGeorge Wilson /* 1293dcba9f3fSGeorge Wilson * Root pools may need to read of the underlying devfs filesystem 1294dcba9f3fSGeorge Wilson * when opening up a vdev. Unfortunately if we're holding the 1295dcba9f3fSGeorge Wilson * SCL_ZIO lock it will result in a deadlock when we try to issue 1296dcba9f3fSGeorge Wilson * the read from the root filesystem. Instead we "prefetch" 1297dcba9f3fSGeorge Wilson * the associated vnodes that we need prior to opening the 1298dcba9f3fSGeorge Wilson * underlying devices and cache them so that we can prevent 1299dcba9f3fSGeorge Wilson * any I/O when we are doing the actual open. 1300dcba9f3fSGeorge Wilson */ 1301dcba9f3fSGeorge Wilson if (spa_is_root(spa)) { 13029842588bSGeorge Wilson int low = locks & ~(SCL_ZIO - 1); 13039842588bSGeorge Wilson int high = locks & ~low; 13049842588bSGeorge Wilson 13059842588bSGeorge Wilson spa_config_enter(spa, high, spa, RW_WRITER); 1306dcba9f3fSGeorge Wilson vdev_hold(spa->spa_root_vdev); 13079842588bSGeorge Wilson spa_config_enter(spa, low, spa, RW_WRITER); 1308dcba9f3fSGeorge Wilson } else { 1309dcba9f3fSGeorge Wilson spa_config_enter(spa, locks, spa, RW_WRITER); 1310dcba9f3fSGeorge Wilson } 13118f18d1faSGeorge Wilson spa->spa_vdev_locks = locks; 1312e14bb325SJeff Bonwick } 1313e14bb325SJeff Bonwick 1314e14bb325SJeff Bonwick int 1315e14bb325SJeff Bonwick spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error) 1316e14bb325SJeff Bonwick { 1317c6065d0fSGeorge Wilson boolean_t config_changed = B_FALSE; 1318c6065d0fSGeorge Wilson 1319b24ab676SJeff Bonwick if (vd != NULL || error == 0) 1320b24ab676SJeff Bonwick vdev_dtl_reassess(vd ? vd->vdev_top : spa->spa_root_vdev, 1321b24ab676SJeff Bonwick 0, 0, B_FALSE); 1322b24ab676SJeff Bonwick 13238f18d1faSGeorge Wilson if (vd != NULL) { 1324e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 1325c6065d0fSGeorge Wilson config_changed = B_TRUE; 13268f18d1faSGeorge Wilson spa->spa_config_generation++; 13278f18d1faSGeorge Wilson } 1328e14bb325SJeff Bonwick 1329dcba9f3fSGeorge Wilson if (spa_is_root(spa)) 1330dcba9f3fSGeorge Wilson vdev_rele(spa->spa_root_vdev); 1331dcba9f3fSGeorge Wilson 13328f18d1faSGeorge Wilson ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL); 13338f18d1faSGeorge Wilson spa_config_exit(spa, spa->spa_vdev_locks, spa); 1334e14bb325SJeff Bonwick 13358ad4d6ddSJeff Bonwick /* 13368ad4d6ddSJeff Bonwick * If anything changed, wait for it to sync. This ensures that, 13378ad4d6ddSJeff Bonwick * from the system administrator's perspective, zpool(1M) commands 13388ad4d6ddSJeff Bonwick * are synchronous. This is important for things like zpool offline: 13398ad4d6ddSJeff Bonwick * when the command completes, you expect no further I/O from ZFS. 13408ad4d6ddSJeff Bonwick */ 13418ad4d6ddSJeff Bonwick if (vd != NULL) 13428ad4d6ddSJeff Bonwick txg_wait_synced(spa->spa_dsl_pool, 0); 13438ad4d6ddSJeff Bonwick 1344c6065d0fSGeorge Wilson /* 1345c6065d0fSGeorge Wilson * If the config changed, update the config cache. 1346c6065d0fSGeorge Wilson */ 1347c6065d0fSGeorge Wilson if (config_changed) { 1348c6065d0fSGeorge Wilson mutex_enter(&spa_namespace_lock); 13495cabbc6bSPrashanth Sreenivasa spa_write_cachefile(spa, B_FALSE, B_TRUE); 1350c6065d0fSGeorge Wilson mutex_exit(&spa_namespace_lock); 1351c6065d0fSGeorge Wilson } 1352c6065d0fSGeorge Wilson 1353e14bb325SJeff Bonwick return (error); 1354e14bb325SJeff Bonwick } 1355e14bb325SJeff Bonwick 1356fa9e4066Sahrens /* 1357fa9e4066Sahrens * ========================================================================== 1358fa9e4066Sahrens * Miscellaneous functions 1359fa9e4066Sahrens * ========================================================================== 1360fa9e4066Sahrens */ 1361fa9e4066Sahrens 1362ad135b5dSChristopher Siden void 136343466aaeSMax Grossman spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx) 1364ad135b5dSChristopher Siden { 13652acef22dSMatthew Ahrens if (!nvlist_exists(spa->spa_label_features, feature)) { 13662acef22dSMatthew Ahrens fnvlist_add_boolean(spa->spa_label_features, feature); 136743466aaeSMax Grossman /* 136843466aaeSMax Grossman * When we are creating the pool (tx_txg==TXG_INITIAL), we can't 136943466aaeSMax Grossman * dirty the vdev config because lock SCL_CONFIG is not held. 137043466aaeSMax Grossman * Thankfully, in this case we don't need to dirty the config 137143466aaeSMax Grossman * because it will be written out anyway when we finish 137243466aaeSMax Grossman * creating the pool. 137343466aaeSMax Grossman */ 137443466aaeSMax Grossman if (tx->tx_txg != TXG_INITIAL) 137543466aaeSMax Grossman vdev_config_dirty(spa->spa_root_vdev); 13762acef22dSMatthew Ahrens } 1377ad135b5dSChristopher Siden } 1378ad135b5dSChristopher Siden 1379ad135b5dSChristopher Siden void 1380ad135b5dSChristopher Siden spa_deactivate_mos_feature(spa_t *spa, const char *feature) 1381ad135b5dSChristopher Siden { 13822acef22dSMatthew Ahrens if (nvlist_remove_all(spa->spa_label_features, feature) == 0) 13832acef22dSMatthew Ahrens vdev_config_dirty(spa->spa_root_vdev); 1384ad135b5dSChristopher Siden } 1385ad135b5dSChristopher Siden 1386fa9e4066Sahrens /* 1387f9af39baSGeorge Wilson * Return the spa_t associated with given pool_guid, if it exists. If 1388f9af39baSGeorge Wilson * device_guid is non-zero, determine whether the pool exists *and* contains 1389f9af39baSGeorge Wilson * a device with the specified device_guid. 1390fa9e4066Sahrens */ 1391f9af39baSGeorge Wilson spa_t * 1392f9af39baSGeorge Wilson spa_by_guid(uint64_t pool_guid, uint64_t device_guid) 1393fa9e4066Sahrens { 1394fa9e4066Sahrens spa_t *spa; 1395fa9e4066Sahrens avl_tree_t *t = &spa_namespace_avl; 1396fa9e4066Sahrens 1397ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1398fa9e4066Sahrens 1399fa9e4066Sahrens for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) { 1400fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) 1401fa9e4066Sahrens continue; 1402fa9e4066Sahrens if (spa->spa_root_vdev == NULL) 1403fa9e4066Sahrens continue; 140439c23413Seschrock if (spa_guid(spa) == pool_guid) { 140539c23413Seschrock if (device_guid == 0) 140639c23413Seschrock break; 140739c23413Seschrock 140839c23413Seschrock if (vdev_lookup_by_guid(spa->spa_root_vdev, 140939c23413Seschrock device_guid) != NULL) 141039c23413Seschrock break; 141139c23413Seschrock 141239c23413Seschrock /* 14138654d025Sperrin * Check any devices we may be in the process of adding. 141439c23413Seschrock */ 141539c23413Seschrock if (spa->spa_pending_vdev) { 141639c23413Seschrock if (vdev_lookup_by_guid(spa->spa_pending_vdev, 141739c23413Seschrock device_guid) != NULL) 141839c23413Seschrock break; 141939c23413Seschrock } 142039c23413Seschrock } 1421fa9e4066Sahrens } 1422fa9e4066Sahrens 1423f9af39baSGeorge Wilson return (spa); 1424f9af39baSGeorge Wilson } 1425f9af39baSGeorge Wilson 1426f9af39baSGeorge Wilson /* 1427f9af39baSGeorge Wilson * Determine whether a pool with the given pool_guid exists. 1428f9af39baSGeorge Wilson */ 1429f9af39baSGeorge Wilson boolean_t 1430f9af39baSGeorge Wilson spa_guid_exists(uint64_t pool_guid, uint64_t device_guid) 1431f9af39baSGeorge Wilson { 1432f9af39baSGeorge Wilson return (spa_by_guid(pool_guid, device_guid) != NULL); 1433fa9e4066Sahrens } 1434fa9e4066Sahrens 1435fa9e4066Sahrens char * 1436fa9e4066Sahrens spa_strdup(const char *s) 1437fa9e4066Sahrens { 1438fa9e4066Sahrens size_t len; 1439fa9e4066Sahrens char *new; 1440fa9e4066Sahrens 1441fa9e4066Sahrens len = strlen(s); 1442fa9e4066Sahrens new = kmem_alloc(len + 1, KM_SLEEP); 1443fa9e4066Sahrens bcopy(s, new, len); 1444fa9e4066Sahrens new[len] = '\0'; 1445fa9e4066Sahrens 1446fa9e4066Sahrens return (new); 1447fa9e4066Sahrens } 1448fa9e4066Sahrens 1449fa9e4066Sahrens void 1450fa9e4066Sahrens spa_strfree(char *s) 1451fa9e4066Sahrens { 1452fa9e4066Sahrens kmem_free(s, strlen(s) + 1); 1453fa9e4066Sahrens } 1454fa9e4066Sahrens 1455fa9e4066Sahrens uint64_t 1456fa9e4066Sahrens spa_get_random(uint64_t range) 1457fa9e4066Sahrens { 1458fa9e4066Sahrens uint64_t r; 1459fa9e4066Sahrens 1460fa9e4066Sahrens ASSERT(range != 0); 1461fa9e4066Sahrens 1462e0f1c0afSOlaf Faaland if (range == 1) 1463e0f1c0afSOlaf Faaland return (0); 1464e0f1c0afSOlaf Faaland 1465fa9e4066Sahrens (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t)); 1466fa9e4066Sahrens 1467fa9e4066Sahrens return (r % range); 1468fa9e4066Sahrens } 1469fa9e4066Sahrens 14701195e687SMark J Musante uint64_t 14711195e687SMark J Musante spa_generate_guid(spa_t *spa) 14721195e687SMark J Musante { 14731195e687SMark J Musante uint64_t guid = spa_get_random(-1ULL); 14741195e687SMark J Musante 14751195e687SMark J Musante if (spa != NULL) { 14761195e687SMark J Musante while (guid == 0 || spa_guid_exists(spa_guid(spa), guid)) 14771195e687SMark J Musante guid = spa_get_random(-1ULL); 14781195e687SMark J Musante } else { 14791195e687SMark J Musante while (guid == 0 || spa_guid_exists(guid, 0)) 14801195e687SMark J Musante guid = spa_get_random(-1ULL); 14811195e687SMark J Musante } 14821195e687SMark J Musante 14831195e687SMark J Musante return (guid); 14841195e687SMark J Musante } 14851195e687SMark J Musante 1486fa9e4066Sahrens void 148743466aaeSMax Grossman snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp) 1488fa9e4066Sahrens { 1489ad135b5dSChristopher Siden char type[256]; 1490f0ba89beSJeff Bonwick char *checksum = NULL; 1491f0ba89beSJeff Bonwick char *compress = NULL; 1492f0ba89beSJeff Bonwick 1493f0ba89beSJeff Bonwick if (bp != NULL) { 1494ad135b5dSChristopher Siden if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) { 1495ad135b5dSChristopher Siden dmu_object_byteswap_t bswap = 1496ad135b5dSChristopher Siden DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); 1497ad135b5dSChristopher Siden (void) snprintf(type, sizeof (type), "bswap %s %s", 1498ad135b5dSChristopher Siden DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ? 1499ad135b5dSChristopher Siden "metadata" : "data", 1500ad135b5dSChristopher Siden dmu_ot_byteswap[bswap].ob_name); 1501ad135b5dSChristopher Siden } else { 1502ad135b5dSChristopher Siden (void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name, 1503ad135b5dSChristopher Siden sizeof (type)); 1504ad135b5dSChristopher Siden } 15055d7b4d43SMatthew Ahrens if (!BP_IS_EMBEDDED(bp)) { 15065d7b4d43SMatthew Ahrens checksum = 15075d7b4d43SMatthew Ahrens zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name; 15085d7b4d43SMatthew Ahrens } 1509f0ba89beSJeff Bonwick compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name; 1510f0ba89beSJeff Bonwick } 1511fa9e4066Sahrens 151243466aaeSMax Grossman SNPRINTF_BLKPTR(snprintf, ' ', buf, buflen, bp, type, checksum, 151343466aaeSMax Grossman compress); 1514fa9e4066Sahrens } 1515fa9e4066Sahrens 1516fa9e4066Sahrens void 1517fa9e4066Sahrens spa_freeze(spa_t *spa) 1518fa9e4066Sahrens { 1519fa9e4066Sahrens uint64_t freeze_txg = 0; 1520fa9e4066Sahrens 1521e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1522fa9e4066Sahrens if (spa->spa_freeze_txg == UINT64_MAX) { 1523fa9e4066Sahrens freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE; 1524fa9e4066Sahrens spa->spa_freeze_txg = freeze_txg; 1525fa9e4066Sahrens } 1526e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1527fa9e4066Sahrens if (freeze_txg != 0) 1528fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), freeze_txg); 1529fa9e4066Sahrens } 1530fa9e4066Sahrens 15310125049cSahrens void 15320125049cSahrens zfs_panic_recover(const char *fmt, ...) 15330125049cSahrens { 15340125049cSahrens va_list adx; 15350125049cSahrens 15360125049cSahrens va_start(adx, fmt); 15370125049cSahrens vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx); 15380125049cSahrens va_end(adx); 15390125049cSahrens } 15400125049cSahrens 15413f9d6ad7SLin Ling /* 15423f9d6ad7SLin Ling * This is a stripped-down version of strtoull, suitable only for converting 1543f7170741SWill Andrews * lowercase hexadecimal numbers that don't overflow. 15443f9d6ad7SLin Ling */ 15453f9d6ad7SLin Ling uint64_t 15464585130bSYuri Pankov zfs_strtonum(const char *str, char **nptr) 15473f9d6ad7SLin Ling { 15483f9d6ad7SLin Ling uint64_t val = 0; 15493f9d6ad7SLin Ling char c; 15503f9d6ad7SLin Ling int digit; 15513f9d6ad7SLin Ling 15523f9d6ad7SLin Ling while ((c = *str) != '\0') { 15533f9d6ad7SLin Ling if (c >= '0' && c <= '9') 15543f9d6ad7SLin Ling digit = c - '0'; 15553f9d6ad7SLin Ling else if (c >= 'a' && c <= 'f') 15563f9d6ad7SLin Ling digit = 10 + c - 'a'; 15573f9d6ad7SLin Ling else 15583f9d6ad7SLin Ling break; 15593f9d6ad7SLin Ling 15603f9d6ad7SLin Ling val *= 16; 15613f9d6ad7SLin Ling val += digit; 15623f9d6ad7SLin Ling 15633f9d6ad7SLin Ling str++; 15643f9d6ad7SLin Ling } 15653f9d6ad7SLin Ling 15663f9d6ad7SLin Ling if (nptr) 15673f9d6ad7SLin Ling *nptr = (char *)str; 15683f9d6ad7SLin Ling 15693f9d6ad7SLin Ling return (val); 15703f9d6ad7SLin Ling } 15713f9d6ad7SLin Ling 1572663207adSDon Brady void 1573663207adSDon Brady spa_activate_allocation_classes(spa_t *spa, dmu_tx_t *tx) 1574663207adSDon Brady { 1575663207adSDon Brady /* 1576663207adSDon Brady * We bump the feature refcount for each special vdev added to the pool 1577663207adSDon Brady */ 1578663207adSDon Brady ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)); 1579663207adSDon Brady spa_feature_incr(spa, SPA_FEATURE_ALLOCATION_CLASSES, tx); 1580663207adSDon Brady } 1581663207adSDon Brady 1582fa9e4066Sahrens /* 1583fa9e4066Sahrens * ========================================================================== 1584fa9e4066Sahrens * Accessor functions 1585fa9e4066Sahrens * ========================================================================== 1586fa9e4066Sahrens */ 1587fa9e4066Sahrens 1588088f3894Sahrens boolean_t 158988b7b0f2SMatthew Ahrens spa_shutting_down(spa_t *spa) 1590fa9e4066Sahrens { 159188b7b0f2SMatthew Ahrens return (spa->spa_async_suspended); 1592fa9e4066Sahrens } 1593fa9e4066Sahrens 1594fa9e4066Sahrens dsl_pool_t * 1595fa9e4066Sahrens spa_get_dsl(spa_t *spa) 1596fa9e4066Sahrens { 1597fa9e4066Sahrens return (spa->spa_dsl_pool); 1598fa9e4066Sahrens } 1599fa9e4066Sahrens 1600ad135b5dSChristopher Siden boolean_t 1601ad135b5dSChristopher Siden spa_is_initializing(spa_t *spa) 1602ad135b5dSChristopher Siden { 1603ad135b5dSChristopher Siden return (spa->spa_is_initializing); 1604ad135b5dSChristopher Siden } 1605ad135b5dSChristopher Siden 16065cabbc6bSPrashanth Sreenivasa boolean_t 16075cabbc6bSPrashanth Sreenivasa spa_indirect_vdevs_loaded(spa_t *spa) 16085cabbc6bSPrashanth Sreenivasa { 16095cabbc6bSPrashanth Sreenivasa return (spa->spa_indirect_vdevs_loaded); 16105cabbc6bSPrashanth Sreenivasa } 16115cabbc6bSPrashanth Sreenivasa 1612fa9e4066Sahrens blkptr_t * 1613fa9e4066Sahrens spa_get_rootblkptr(spa_t *spa) 1614fa9e4066Sahrens { 1615fa9e4066Sahrens return (&spa->spa_ubsync.ub_rootbp); 1616fa9e4066Sahrens } 1617fa9e4066Sahrens 1618fa9e4066Sahrens void 1619fa9e4066Sahrens spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp) 1620fa9e4066Sahrens { 1621fa9e4066Sahrens spa->spa_uberblock.ub_rootbp = *bp; 1622fa9e4066Sahrens } 1623fa9e4066Sahrens 1624fa9e4066Sahrens void 1625fa9e4066Sahrens spa_altroot(spa_t *spa, char *buf, size_t buflen) 1626fa9e4066Sahrens { 1627fa9e4066Sahrens if (spa->spa_root == NULL) 1628fa9e4066Sahrens buf[0] = '\0'; 1629fa9e4066Sahrens else 1630fa9e4066Sahrens (void) strncpy(buf, spa->spa_root, buflen); 1631fa9e4066Sahrens } 1632fa9e4066Sahrens 1633fa9e4066Sahrens int 1634fa9e4066Sahrens spa_sync_pass(spa_t *spa) 1635fa9e4066Sahrens { 1636fa9e4066Sahrens return (spa->spa_sync_pass); 1637fa9e4066Sahrens } 1638fa9e4066Sahrens 1639fa9e4066Sahrens char * 1640fa9e4066Sahrens spa_name(spa_t *spa) 1641fa9e4066Sahrens { 1642fa9e4066Sahrens return (spa->spa_name); 1643fa9e4066Sahrens } 1644fa9e4066Sahrens 1645fa9e4066Sahrens uint64_t 1646fa9e4066Sahrens spa_guid(spa_t *spa) 1647fa9e4066Sahrens { 1648dfbb9432SGeorge Wilson dsl_pool_t *dp = spa_get_dsl(spa); 1649dfbb9432SGeorge Wilson uint64_t guid; 1650dfbb9432SGeorge Wilson 1651b5989ec7Seschrock /* 1652b5989ec7Seschrock * If we fail to parse the config during spa_load(), we can go through 1653b5989ec7Seschrock * the error path (which posts an ereport) and end up here with no root 1654e9103aaeSGarrett D'Amore * vdev. We stash the original pool guid in 'spa_config_guid' to handle 1655b5989ec7Seschrock * this case. 1656b5989ec7Seschrock */ 1657dfbb9432SGeorge Wilson if (spa->spa_root_vdev == NULL) 1658dfbb9432SGeorge Wilson return (spa->spa_config_guid); 1659dfbb9432SGeorge Wilson 1660dfbb9432SGeorge Wilson guid = spa->spa_last_synced_guid != 0 ? 1661dfbb9432SGeorge Wilson spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid; 1662dfbb9432SGeorge Wilson 1663dfbb9432SGeorge Wilson /* 1664dfbb9432SGeorge Wilson * Return the most recently synced out guid unless we're 1665dfbb9432SGeorge Wilson * in syncing context. 1666dfbb9432SGeorge Wilson */ 1667dfbb9432SGeorge Wilson if (dp && dsl_pool_sync_context(dp)) 1668b5989ec7Seschrock return (spa->spa_root_vdev->vdev_guid); 1669b5989ec7Seschrock else 1670dfbb9432SGeorge Wilson return (guid); 1671e9103aaeSGarrett D'Amore } 1672e9103aaeSGarrett D'Amore 1673e9103aaeSGarrett D'Amore uint64_t 1674e9103aaeSGarrett D'Amore spa_load_guid(spa_t *spa) 1675e9103aaeSGarrett D'Amore { 1676e9103aaeSGarrett D'Amore /* 1677e9103aaeSGarrett D'Amore * This is a GUID that exists solely as a reference for the 1678e9103aaeSGarrett D'Amore * purposes of the arc. It is generated at load time, and 1679e9103aaeSGarrett D'Amore * is never written to persistent storage. 1680e9103aaeSGarrett D'Amore */ 1681e9103aaeSGarrett D'Amore return (spa->spa_load_guid); 1682fa9e4066Sahrens } 1683fa9e4066Sahrens 1684fa9e4066Sahrens uint64_t 1685fa9e4066Sahrens spa_last_synced_txg(spa_t *spa) 1686fa9e4066Sahrens { 1687fa9e4066Sahrens return (spa->spa_ubsync.ub_txg); 1688fa9e4066Sahrens } 1689fa9e4066Sahrens 1690fa9e4066Sahrens uint64_t 1691fa9e4066Sahrens spa_first_txg(spa_t *spa) 1692fa9e4066Sahrens { 1693fa9e4066Sahrens return (spa->spa_first_txg); 1694fa9e4066Sahrens } 1695fa9e4066Sahrens 1696b24ab676SJeff Bonwick uint64_t 1697b24ab676SJeff Bonwick spa_syncing_txg(spa_t *spa) 1698b24ab676SJeff Bonwick { 1699b24ab676SJeff Bonwick return (spa->spa_syncing_txg); 1700b24ab676SJeff Bonwick } 1701b24ab676SJeff Bonwick 17023991b535SGeorge Wilson /* 17033991b535SGeorge Wilson * Return the last txg where data can be dirtied. The final txgs 17043991b535SGeorge Wilson * will be used to just clear out any deferred frees that remain. 17053991b535SGeorge Wilson */ 17063991b535SGeorge Wilson uint64_t 17073991b535SGeorge Wilson spa_final_dirty_txg(spa_t *spa) 17083991b535SGeorge Wilson { 17093991b535SGeorge Wilson return (spa->spa_final_txg - TXG_DEFER_SIZE); 17103991b535SGeorge Wilson } 17113991b535SGeorge Wilson 171288b7b0f2SMatthew Ahrens pool_state_t 1713fa9e4066Sahrens spa_state(spa_t *spa) 1714fa9e4066Sahrens { 1715fa9e4066Sahrens return (spa->spa_state); 1716fa9e4066Sahrens } 1717fa9e4066Sahrens 1718b16da2e2SGeorge Wilson spa_load_state_t 1719b16da2e2SGeorge Wilson spa_load_state(spa_t *spa) 1720b16da2e2SGeorge Wilson { 1721b16da2e2SGeorge Wilson return (spa->spa_load_state); 1722b16da2e2SGeorge Wilson } 1723b16da2e2SGeorge Wilson 1724fa9e4066Sahrens uint64_t 1725fa9e4066Sahrens spa_freeze_txg(spa_t *spa) 1726fa9e4066Sahrens { 1727fa9e4066Sahrens return (spa->spa_freeze_txg); 1728fa9e4066Sahrens } 1729fa9e4066Sahrens 1730fa9e4066Sahrens /* ARGSUSED */ 1731fa9e4066Sahrens uint64_t 173261e255ceSMatthew Ahrens spa_get_worst_case_asize(spa_t *spa, uint64_t lsize) 1733fa9e4066Sahrens { 173469962b56SMatthew Ahrens return (lsize * spa_asize_inflation); 173544cd46caSbillm } 173644cd46caSbillm 17377d46dc6cSMatthew Ahrens /* 17387d46dc6cSMatthew Ahrens * Return the amount of slop space in bytes. It is 1/32 of the pool (3.2%), 17394b5c8e93SMatthew Ahrens * or at least 128MB, unless that would cause it to be more than half the 17404b5c8e93SMatthew Ahrens * pool size. 17417d46dc6cSMatthew Ahrens * 17427d46dc6cSMatthew Ahrens * See the comment above spa_slop_shift for details. 17437d46dc6cSMatthew Ahrens */ 17447d46dc6cSMatthew Ahrens uint64_t 17454b5c8e93SMatthew Ahrens spa_get_slop_space(spa_t *spa) 17464b5c8e93SMatthew Ahrens { 17477d46dc6cSMatthew Ahrens uint64_t space = spa_get_dspace(spa); 17484b5c8e93SMatthew Ahrens return (MAX(space >> spa_slop_shift, MIN(space >> 1, spa_min_slop))); 17497d46dc6cSMatthew Ahrens } 17507d46dc6cSMatthew Ahrens 1751485bbbf5SGeorge Wilson uint64_t 1752485bbbf5SGeorge Wilson spa_get_dspace(spa_t *spa) 1753485bbbf5SGeorge Wilson { 1754485bbbf5SGeorge Wilson return (spa->spa_dspace); 1755485bbbf5SGeorge Wilson } 1756485bbbf5SGeorge Wilson 175786714001SSerapheim Dimitropoulos uint64_t 175886714001SSerapheim Dimitropoulos spa_get_checkpoint_space(spa_t *spa) 175986714001SSerapheim Dimitropoulos { 176086714001SSerapheim Dimitropoulos return (spa->spa_checkpoint_info.sci_dspace); 176186714001SSerapheim Dimitropoulos } 176286714001SSerapheim Dimitropoulos 1763485bbbf5SGeorge Wilson void 1764485bbbf5SGeorge Wilson spa_update_dspace(spa_t *spa) 1765485bbbf5SGeorge Wilson { 1766485bbbf5SGeorge Wilson spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) + 1767485bbbf5SGeorge Wilson ddt_get_dedup_dspace(spa); 17685cabbc6bSPrashanth Sreenivasa if (spa->spa_vdev_removal != NULL) { 17695cabbc6bSPrashanth Sreenivasa /* 17705cabbc6bSPrashanth Sreenivasa * We can't allocate from the removing device, so 17715cabbc6bSPrashanth Sreenivasa * subtract its size. This prevents the DMU/DSL from 17725cabbc6bSPrashanth Sreenivasa * filling up the (now smaller) pool while we are in the 17735cabbc6bSPrashanth Sreenivasa * middle of removing the device. 17745cabbc6bSPrashanth Sreenivasa * 17755cabbc6bSPrashanth Sreenivasa * Note that the DMU/DSL doesn't actually know or care 17765cabbc6bSPrashanth Sreenivasa * how much space is allocated (it does its own tracking 17775cabbc6bSPrashanth Sreenivasa * of how much space has been logically used). So it 17785cabbc6bSPrashanth Sreenivasa * doesn't matter that the data we are moving may be 17795cabbc6bSPrashanth Sreenivasa * allocated twice (on the old device and the new 17805cabbc6bSPrashanth Sreenivasa * device). 17815cabbc6bSPrashanth Sreenivasa */ 17823a4b1be9SMatthew Ahrens spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 17833a4b1be9SMatthew Ahrens vdev_t *vd = 17843a4b1be9SMatthew Ahrens vdev_lookup_top(spa, spa->spa_vdev_removal->svr_vdev_id); 17855cabbc6bSPrashanth Sreenivasa spa->spa_dspace -= spa_deflate(spa) ? 17865cabbc6bSPrashanth Sreenivasa vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space; 17873a4b1be9SMatthew Ahrens spa_config_exit(spa, SCL_VDEV, FTAG); 17885cabbc6bSPrashanth Sreenivasa } 1789485bbbf5SGeorge Wilson } 1790485bbbf5SGeorge Wilson 17910a4e9518Sgw /* 17920a4e9518Sgw * Return the failure mode that has been set to this pool. The default 17930a4e9518Sgw * behavior will be to block all I/Os when a complete failure occurs. 17940a4e9518Sgw */ 17950a4e9518Sgw uint8_t 17960a4e9518Sgw spa_get_failmode(spa_t *spa) 17970a4e9518Sgw { 17980a4e9518Sgw return (spa->spa_failmode); 17990a4e9518Sgw } 18000a4e9518Sgw 1801e14bb325SJeff Bonwick boolean_t 1802e14bb325SJeff Bonwick spa_suspended(spa_t *spa) 1803e14bb325SJeff Bonwick { 1804e0f1c0afSOlaf Faaland return (spa->spa_suspended != ZIO_SUSPEND_NONE); 1805e14bb325SJeff Bonwick } 1806e14bb325SJeff Bonwick 180744cd46caSbillm uint64_t 180844cd46caSbillm spa_version(spa_t *spa) 180944cd46caSbillm { 181044cd46caSbillm return (spa->spa_ubsync.ub_version); 181144cd46caSbillm } 181244cd46caSbillm 1813b24ab676SJeff Bonwick boolean_t 1814b24ab676SJeff Bonwick spa_deflate(spa_t *spa) 1815b24ab676SJeff Bonwick { 1816b24ab676SJeff Bonwick return (spa->spa_deflate); 1817b24ab676SJeff Bonwick } 1818b24ab676SJeff Bonwick 1819b24ab676SJeff Bonwick metaslab_class_t * 1820b24ab676SJeff Bonwick spa_normal_class(spa_t *spa) 1821b24ab676SJeff Bonwick { 1822b24ab676SJeff Bonwick return (spa->spa_normal_class); 1823b24ab676SJeff Bonwick } 1824b24ab676SJeff Bonwick 1825b24ab676SJeff Bonwick metaslab_class_t * 1826b24ab676SJeff Bonwick spa_log_class(spa_t *spa) 1827b24ab676SJeff Bonwick { 1828b24ab676SJeff Bonwick return (spa->spa_log_class); 1829b24ab676SJeff Bonwick } 1830b24ab676SJeff Bonwick 1831663207adSDon Brady metaslab_class_t * 1832663207adSDon Brady spa_special_class(spa_t *spa) 1833663207adSDon Brady { 1834663207adSDon Brady return (spa->spa_special_class); 1835663207adSDon Brady } 1836663207adSDon Brady 1837663207adSDon Brady metaslab_class_t * 1838663207adSDon Brady spa_dedup_class(spa_t *spa) 1839663207adSDon Brady { 1840663207adSDon Brady return (spa->spa_dedup_class); 1841663207adSDon Brady } 1842663207adSDon Brady 1843663207adSDon Brady /* 1844663207adSDon Brady * Locate an appropriate allocation class 1845663207adSDon Brady */ 1846663207adSDon Brady metaslab_class_t * 1847663207adSDon Brady spa_preferred_class(spa_t *spa, uint64_t size, dmu_object_type_t objtype, 1848663207adSDon Brady uint_t level, uint_t special_smallblk) 1849663207adSDon Brady { 1850663207adSDon Brady if (DMU_OT_IS_ZIL(objtype)) { 1851663207adSDon Brady if (spa->spa_log_class->mc_groups != 0) 1852663207adSDon Brady return (spa_log_class(spa)); 1853663207adSDon Brady else 1854663207adSDon Brady return (spa_normal_class(spa)); 1855663207adSDon Brady } 1856663207adSDon Brady 1857663207adSDon Brady boolean_t has_special_class = spa->spa_special_class->mc_groups != 0; 1858663207adSDon Brady 1859663207adSDon Brady if (DMU_OT_IS_DDT(objtype)) { 1860663207adSDon Brady if (spa->spa_dedup_class->mc_groups != 0) 1861663207adSDon Brady return (spa_dedup_class(spa)); 1862663207adSDon Brady else if (has_special_class && zfs_ddt_data_is_special) 1863663207adSDon Brady return (spa_special_class(spa)); 1864663207adSDon Brady else 1865663207adSDon Brady return (spa_normal_class(spa)); 1866663207adSDon Brady } 1867663207adSDon Brady 1868663207adSDon Brady /* Indirect blocks for user data can land in special if allowed */ 1869663207adSDon Brady if (level > 0 && (DMU_OT_IS_FILE(objtype) || objtype == DMU_OT_ZVOL)) { 1870663207adSDon Brady if (has_special_class && zfs_user_indirect_is_special) 1871663207adSDon Brady return (spa_special_class(spa)); 1872663207adSDon Brady else 1873663207adSDon Brady return (spa_normal_class(spa)); 1874663207adSDon Brady } 1875663207adSDon Brady 1876663207adSDon Brady if (DMU_OT_IS_METADATA(objtype) || level > 0) { 1877663207adSDon Brady if (has_special_class) 1878663207adSDon Brady return (spa_special_class(spa)); 1879663207adSDon Brady else 1880663207adSDon Brady return (spa_normal_class(spa)); 1881663207adSDon Brady } 1882663207adSDon Brady 1883663207adSDon Brady /* 1884663207adSDon Brady * Allow small file blocks in special class in some cases (like 1885663207adSDon Brady * for the dRAID vdev feature). But always leave a reserve of 1886663207adSDon Brady * zfs_special_class_metadata_reserve_pct exclusively for metadata. 1887663207adSDon Brady */ 1888663207adSDon Brady if (DMU_OT_IS_FILE(objtype) && 1889663207adSDon Brady has_special_class && size <= special_smallblk) { 1890663207adSDon Brady metaslab_class_t *special = spa_special_class(spa); 1891663207adSDon Brady uint64_t alloc = metaslab_class_get_alloc(special); 1892663207adSDon Brady uint64_t space = metaslab_class_get_space(special); 1893663207adSDon Brady uint64_t limit = 1894663207adSDon Brady (space * (100 - zfs_special_class_metadata_reserve_pct)) 1895663207adSDon Brady / 100; 1896663207adSDon Brady 1897663207adSDon Brady if (alloc < limit) 1898663207adSDon Brady return (special); 1899663207adSDon Brady } 1900663207adSDon Brady 1901663207adSDon Brady return (spa_normal_class(spa)); 1902663207adSDon Brady } 1903663207adSDon Brady 1904bc9014e6SJustin Gibbs void 1905bc9014e6SJustin Gibbs spa_evicting_os_register(spa_t *spa, objset_t *os) 1906bc9014e6SJustin Gibbs { 1907bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1908bc9014e6SJustin Gibbs list_insert_head(&spa->spa_evicting_os_list, os); 1909bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1910bc9014e6SJustin Gibbs } 1911bc9014e6SJustin Gibbs 1912bc9014e6SJustin Gibbs void 1913bc9014e6SJustin Gibbs spa_evicting_os_deregister(spa_t *spa, objset_t *os) 1914bc9014e6SJustin Gibbs { 1915bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1916bc9014e6SJustin Gibbs list_remove(&spa->spa_evicting_os_list, os); 1917bc9014e6SJustin Gibbs cv_broadcast(&spa->spa_evicting_os_cv); 1918bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1919bc9014e6SJustin Gibbs } 1920bc9014e6SJustin Gibbs 1921bc9014e6SJustin Gibbs void 1922bc9014e6SJustin Gibbs spa_evicting_os_wait(spa_t *spa) 1923bc9014e6SJustin Gibbs { 1924bc9014e6SJustin Gibbs mutex_enter(&spa->spa_evicting_os_lock); 1925bc9014e6SJustin Gibbs while (!list_is_empty(&spa->spa_evicting_os_list)) 1926bc9014e6SJustin Gibbs cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock); 1927bc9014e6SJustin Gibbs mutex_exit(&spa->spa_evicting_os_lock); 1928bc9014e6SJustin Gibbs 1929bc9014e6SJustin Gibbs dmu_buf_user_evict_wait(); 1930bc9014e6SJustin Gibbs } 1931bc9014e6SJustin Gibbs 193244cd46caSbillm int 193344cd46caSbillm spa_max_replication(spa_t *spa) 193444cd46caSbillm { 193544cd46caSbillm /* 1936e7437265Sahrens * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to 193744cd46caSbillm * handle BPs with more than one DVA allocated. Set our max 193844cd46caSbillm * replication level accordingly. 1939fa9e4066Sahrens */ 1940e7437265Sahrens if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS) 194144cd46caSbillm return (1); 194244cd46caSbillm return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override)); 1943fa9e4066Sahrens } 1944fa9e4066Sahrens 19453f9d6ad7SLin Ling int 19463f9d6ad7SLin Ling spa_prev_software_version(spa_t *spa) 19473f9d6ad7SLin Ling { 19483f9d6ad7SLin Ling return (spa->spa_prev_software_version); 19493f9d6ad7SLin Ling } 19503f9d6ad7SLin Ling 1951283b8460SGeorge.Wilson uint64_t 1952283b8460SGeorge.Wilson spa_deadman_synctime(spa_t *spa) 1953283b8460SGeorge.Wilson { 1954283b8460SGeorge.Wilson return (spa->spa_deadman_synctime); 1955283b8460SGeorge.Wilson } 1956283b8460SGeorge.Wilson 1957084fd14fSBrian Behlendorf spa_autotrim_t 1958084fd14fSBrian Behlendorf spa_get_autotrim(spa_t *spa) 1959084fd14fSBrian Behlendorf { 1960084fd14fSBrian Behlendorf return (spa->spa_autotrim); 1961084fd14fSBrian Behlendorf } 1962084fd14fSBrian Behlendorf 196399653d4eSeschrock uint64_t 1964b24ab676SJeff Bonwick dva_get_dsize_sync(spa_t *spa, const dva_t *dva) 196599653d4eSeschrock { 1966b24ab676SJeff Bonwick uint64_t asize = DVA_GET_ASIZE(dva); 1967b24ab676SJeff Bonwick uint64_t dsize = asize; 196899653d4eSeschrock 1969b24ab676SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); 197099653d4eSeschrock 1971b24ab676SJeff Bonwick if (asize != 0 && spa->spa_deflate) { 1972b24ab676SJeff Bonwick vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva)); 1973b24ab676SJeff Bonwick dsize = (asize >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio; 197499653d4eSeschrock } 1975b24ab676SJeff Bonwick 1976b24ab676SJeff Bonwick return (dsize); 1977b24ab676SJeff Bonwick } 1978b24ab676SJeff Bonwick 1979b24ab676SJeff Bonwick uint64_t 1980b24ab676SJeff Bonwick bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp) 1981b24ab676SJeff Bonwick { 1982b24ab676SJeff Bonwick uint64_t dsize = 0; 1983b24ab676SJeff Bonwick 19845d7b4d43SMatthew Ahrens for (int d = 0; d < BP_GET_NDVAS(bp); d++) 1985b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1986b24ab676SJeff Bonwick 1987b24ab676SJeff Bonwick return (dsize); 1988b24ab676SJeff Bonwick } 1989b24ab676SJeff Bonwick 1990b24ab676SJeff Bonwick uint64_t 1991b24ab676SJeff Bonwick bp_get_dsize(spa_t *spa, const blkptr_t *bp) 1992b24ab676SJeff Bonwick { 1993b24ab676SJeff Bonwick uint64_t dsize = 0; 1994b24ab676SJeff Bonwick 1995b24ab676SJeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 1996b24ab676SJeff Bonwick 19975d7b4d43SMatthew Ahrens for (int d = 0; d < BP_GET_NDVAS(bp); d++) 1998b24ab676SJeff Bonwick dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]); 1999b24ab676SJeff Bonwick 2000e14bb325SJeff Bonwick spa_config_exit(spa, SCL_VDEV, FTAG); 2001b24ab676SJeff Bonwick 2002b24ab676SJeff Bonwick return (dsize); 200399653d4eSeschrock } 200499653d4eSeschrock 2005abe1fd01SDon Brady uint64_t 2006abe1fd01SDon Brady spa_dirty_data(spa_t *spa) 2007abe1fd01SDon Brady { 2008abe1fd01SDon Brady return (spa->spa_dsl_pool->dp_dirty_total); 2009abe1fd01SDon Brady } 2010abe1fd01SDon Brady 2011fa9e4066Sahrens /* 2012fa9e4066Sahrens * ========================================================================== 2013fa9e4066Sahrens * Initialization and Termination 2014fa9e4066Sahrens * ========================================================================== 2015fa9e4066Sahrens */ 2016fa9e4066Sahrens 2017fa9e4066Sahrens static int 2018fa9e4066Sahrens spa_name_compare(const void *a1, const void *a2) 2019fa9e4066Sahrens { 2020fa9e4066Sahrens const spa_t *s1 = a1; 2021fa9e4066Sahrens const spa_t *s2 = a2; 2022fa9e4066Sahrens int s; 2023fa9e4066Sahrens 2024fa9e4066Sahrens s = strcmp(s1->spa_name, s2->spa_name); 2025c4ab0d3fSGvozden Neskovic 2026*4d7988d6SPaul Dagnelie return (TREE_ISIGN(s)); 2027fa9e4066Sahrens } 2028fa9e4066Sahrens 20290373e76bSbonwick int 20300373e76bSbonwick spa_busy(void) 20310373e76bSbonwick { 20320373e76bSbonwick return (spa_active_count); 20330373e76bSbonwick } 20340373e76bSbonwick 2035e7cbe64fSgw void 2036e7cbe64fSgw spa_boot_init() 2037e7cbe64fSgw { 2038e7cbe64fSgw spa_config_load(); 2039e7cbe64fSgw } 2040e7cbe64fSgw 2041fa9e4066Sahrens void 2042fa9e4066Sahrens spa_init(int mode) 2043fa9e4066Sahrens { 2044fa9e4066Sahrens mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL); 2045c25056deSgw mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL); 2046fa94a07fSbrendan mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL); 2047fa9e4066Sahrens cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL); 2048fa9e4066Sahrens 2049fa9e4066Sahrens avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t), 2050fa9e4066Sahrens offsetof(spa_t, spa_avl)); 2051fa9e4066Sahrens 2052fa94a07fSbrendan avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t), 2053fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 2054fa94a07fSbrendan 2055fa94a07fSbrendan avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t), 2056fa94a07fSbrendan offsetof(spa_aux_t, aux_avl)); 205799653d4eSeschrock 20588ad4d6ddSJeff Bonwick spa_mode_global = mode; 2059fa9e4066Sahrens 2060283b8460SGeorge.Wilson #ifdef _KERNEL 2061283b8460SGeorge.Wilson spa_arch_init(); 2062283b8460SGeorge.Wilson #else 2063cd1c8b85SMatthew Ahrens if (spa_mode_global != FREAD && dprintf_find_string("watch")) { 2064cd1c8b85SMatthew Ahrens arc_procfd = open("/proc/self/ctl", O_WRONLY); 2065cd1c8b85SMatthew Ahrens if (arc_procfd == -1) { 2066cd1c8b85SMatthew Ahrens perror("could not enable watchpoints: " 2067cd1c8b85SMatthew Ahrens "opening /proc/self/ctl failed: "); 2068cd1c8b85SMatthew Ahrens } else { 2069cd1c8b85SMatthew Ahrens arc_watch = B_TRUE; 2070cd1c8b85SMatthew Ahrens } 2071cd1c8b85SMatthew Ahrens } 2072cd1c8b85SMatthew Ahrens #endif 2073cd1c8b85SMatthew Ahrens 2074e914ace2STim Schumacher zfs_refcount_init(); 2075fa9e4066Sahrens unique_init(); 2076*4d7988d6SPaul Dagnelie zfs_btree_init(); 2077*4d7988d6SPaul Dagnelie metaslab_stat_init(); 2078fa9e4066Sahrens zio_init(); 2079fa9e4066Sahrens dmu_init(); 2080fa9e4066Sahrens zil_init(); 208187db74c1Sek vdev_cache_stat_init(); 208212a8814cSTom Caputi vdev_mirror_stat_init(); 208391ebeef5Sahrens zfs_prop_init(); 2084990b4856Slling zpool_prop_init(); 2085ad135b5dSChristopher Siden zpool_feature_init(); 2086fa9e4066Sahrens spa_config_load(); 2087e14bb325SJeff Bonwick l2arc_start(); 2088a3874b8bSToomas Soome scan_init(); 2089fa9e4066Sahrens } 2090fa9e4066Sahrens 2091fa9e4066Sahrens void 2092fa9e4066Sahrens spa_fini(void) 2093fa9e4066Sahrens { 2094e14bb325SJeff Bonwick l2arc_stop(); 2095e14bb325SJeff Bonwick 2096fa9e4066Sahrens spa_evict_all(); 2097fa9e4066Sahrens 209887db74c1Sek vdev_cache_stat_fini(); 209912a8814cSTom Caputi vdev_mirror_stat_fini(); 2100fa9e4066Sahrens zil_fini(); 2101fa9e4066Sahrens dmu_fini(); 2102fa9e4066Sahrens zio_fini(); 2103*4d7988d6SPaul Dagnelie metaslab_stat_fini(); 2104*4d7988d6SPaul Dagnelie zfs_btree_fini(); 210591ebeef5Sahrens unique_fini(); 2106e914ace2STim Schumacher zfs_refcount_fini(); 2107a3874b8bSToomas Soome scan_fini(); 2108fa9e4066Sahrens 2109fa9e4066Sahrens avl_destroy(&spa_namespace_avl); 211099653d4eSeschrock avl_destroy(&spa_spare_avl); 2111fa94a07fSbrendan avl_destroy(&spa_l2cache_avl); 2112fa9e4066Sahrens 2113fa9e4066Sahrens cv_destroy(&spa_namespace_cv); 2114fa9e4066Sahrens mutex_destroy(&spa_namespace_lock); 2115c25056deSgw mutex_destroy(&spa_spare_lock); 2116fa94a07fSbrendan mutex_destroy(&spa_l2cache_lock); 2117fa9e4066Sahrens } 21186ce0521aSperrin 21196ce0521aSperrin /* 21206ce0521aSperrin * Return whether this pool has slogs. No locking needed. 21216ce0521aSperrin * It's not a problem if the wrong answer is returned as it's only for 21226ce0521aSperrin * performance and not correctness 21236ce0521aSperrin */ 21246ce0521aSperrin boolean_t 21256ce0521aSperrin spa_has_slogs(spa_t *spa) 21266ce0521aSperrin { 21276ce0521aSperrin return (spa->spa_log_class->mc_rotor != NULL); 21286ce0521aSperrin } 2129bf82a41bSeschrock 2130b24ab676SJeff Bonwick spa_log_state_t 2131b24ab676SJeff Bonwick spa_get_log_state(spa_t *spa) 2132b24ab676SJeff Bonwick { 2133b24ab676SJeff Bonwick return (spa->spa_log_state); 2134b24ab676SJeff Bonwick } 2135b24ab676SJeff Bonwick 2136b24ab676SJeff Bonwick void 2137b24ab676SJeff Bonwick spa_set_log_state(spa_t *spa, spa_log_state_t state) 2138b24ab676SJeff Bonwick { 2139b24ab676SJeff Bonwick spa->spa_log_state = state; 2140b24ab676SJeff Bonwick } 2141b24ab676SJeff Bonwick 2142bf82a41bSeschrock boolean_t 2143bf82a41bSeschrock spa_is_root(spa_t *spa) 2144bf82a41bSeschrock { 2145bf82a41bSeschrock return (spa->spa_is_root); 2146bf82a41bSeschrock } 21478ad4d6ddSJeff Bonwick 21488ad4d6ddSJeff Bonwick boolean_t 21498ad4d6ddSJeff Bonwick spa_writeable(spa_t *spa) 21508ad4d6ddSJeff Bonwick { 21516f793812SPavel Zakharov return (!!(spa->spa_mode & FWRITE) && spa->spa_trust_config); 21528ad4d6ddSJeff Bonwick } 21538ad4d6ddSJeff Bonwick 215473527f44SAlex Reece /* 215573527f44SAlex Reece * Returns true if there is a pending sync task in any of the current 215673527f44SAlex Reece * syncing txg, the current quiescing txg, or the current open txg. 215773527f44SAlex Reece */ 215873527f44SAlex Reece boolean_t 215973527f44SAlex Reece spa_has_pending_synctask(spa_t *spa) 216073527f44SAlex Reece { 216186714001SSerapheim Dimitropoulos return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks) || 216286714001SSerapheim Dimitropoulos !txg_all_lists_empty(&spa->spa_dsl_pool->dp_early_sync_tasks)); 216373527f44SAlex Reece } 216473527f44SAlex Reece 21658ad4d6ddSJeff Bonwick int 21668ad4d6ddSJeff Bonwick spa_mode(spa_t *spa) 21678ad4d6ddSJeff Bonwick { 21688ad4d6ddSJeff Bonwick return (spa->spa_mode); 21698ad4d6ddSJeff Bonwick } 2170b24ab676SJeff Bonwick 2171b24ab676SJeff Bonwick uint64_t 2172b24ab676SJeff Bonwick spa_bootfs(spa_t *spa) 2173b24ab676SJeff Bonwick { 2174b24ab676SJeff Bonwick return (spa->spa_bootfs); 2175b24ab676SJeff Bonwick } 2176b24ab676SJeff Bonwick 2177b24ab676SJeff Bonwick uint64_t 2178b24ab676SJeff Bonwick spa_delegation(spa_t *spa) 2179b24ab676SJeff Bonwick { 2180b24ab676SJeff Bonwick return (spa->spa_delegation); 2181b24ab676SJeff Bonwick } 2182b24ab676SJeff Bonwick 2183b24ab676SJeff Bonwick objset_t * 2184b24ab676SJeff Bonwick spa_meta_objset(spa_t *spa) 2185b24ab676SJeff Bonwick { 2186b24ab676SJeff Bonwick return (spa->spa_meta_objset); 2187b24ab676SJeff Bonwick } 2188b24ab676SJeff Bonwick 2189b24ab676SJeff Bonwick enum zio_checksum 2190b24ab676SJeff Bonwick spa_dedup_checksum(spa_t *spa) 2191b24ab676SJeff Bonwick { 2192b24ab676SJeff Bonwick return (spa->spa_dedup_checksum); 2193b24ab676SJeff Bonwick } 21943f9d6ad7SLin Ling 21953f9d6ad7SLin Ling /* 21963f9d6ad7SLin Ling * Reset pool scan stat per scan pass (or reboot). 21973f9d6ad7SLin Ling */ 21983f9d6ad7SLin Ling void 21993f9d6ad7SLin Ling spa_scan_stat_init(spa_t *spa) 22003f9d6ad7SLin Ling { 22013f9d6ad7SLin Ling /* data not stored on disk */ 22023f9d6ad7SLin Ling spa->spa_scan_pass_start = gethrestime_sec(); 22031702cce7SAlek Pinchuk if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan)) 22041702cce7SAlek Pinchuk spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start; 22051702cce7SAlek Pinchuk else 22061702cce7SAlek Pinchuk spa->spa_scan_pass_scrub_pause = 0; 22071702cce7SAlek Pinchuk spa->spa_scan_pass_scrub_spent_paused = 0; 22083f9d6ad7SLin Ling spa->spa_scan_pass_exam = 0; 2209a3874b8bSToomas Soome spa->spa_scan_pass_issued = 0; 22103f9d6ad7SLin Ling vdev_scan_stat_init(spa->spa_root_vdev); 22113f9d6ad7SLin Ling } 22123f9d6ad7SLin Ling 22133f9d6ad7SLin Ling /* 22143f9d6ad7SLin Ling * Get scan stats for zpool status reports 22153f9d6ad7SLin Ling */ 22163f9d6ad7SLin Ling int 22173f9d6ad7SLin Ling spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps) 22183f9d6ad7SLin Ling { 22193f9d6ad7SLin Ling dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL; 22203f9d6ad7SLin Ling 22213f9d6ad7SLin Ling if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE) 2222be6fd75aSMatthew Ahrens return (SET_ERROR(ENOENT)); 22233f9d6ad7SLin Ling bzero(ps, sizeof (pool_scan_stat_t)); 22243f9d6ad7SLin Ling 22253f9d6ad7SLin Ling /* data stored on disk */ 22263f9d6ad7SLin Ling ps->pss_func = scn->scn_phys.scn_func; 2227a3874b8bSToomas Soome ps->pss_state = scn->scn_phys.scn_state; 22283f9d6ad7SLin Ling ps->pss_start_time = scn->scn_phys.scn_start_time; 22293f9d6ad7SLin Ling ps->pss_end_time = scn->scn_phys.scn_end_time; 22303f9d6ad7SLin Ling ps->pss_to_examine = scn->scn_phys.scn_to_examine; 22313f9d6ad7SLin Ling ps->pss_to_process = scn->scn_phys.scn_to_process; 22323f9d6ad7SLin Ling ps->pss_processed = scn->scn_phys.scn_processed; 22333f9d6ad7SLin Ling ps->pss_errors = scn->scn_phys.scn_errors; 2234a3874b8bSToomas Soome ps->pss_examined = scn->scn_phys.scn_examined; 2235a3874b8bSToomas Soome ps->pss_issued = 2236a3874b8bSToomas Soome scn->scn_issued_before_pass + spa->spa_scan_pass_issued; 22373f9d6ad7SLin Ling ps->pss_state = scn->scn_phys.scn_state; 22383f9d6ad7SLin Ling 22393f9d6ad7SLin Ling /* data not stored on disk */ 22403f9d6ad7SLin Ling ps->pss_pass_start = spa->spa_scan_pass_start; 22413f9d6ad7SLin Ling ps->pss_pass_exam = spa->spa_scan_pass_exam; 2242a3874b8bSToomas Soome ps->pss_pass_issued = spa->spa_scan_pass_issued; 22431702cce7SAlek Pinchuk ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause; 22441702cce7SAlek Pinchuk ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused; 22453f9d6ad7SLin Ling 22463f9d6ad7SLin Ling return (0); 22473f9d6ad7SLin Ling } 224809c9d376SGeorge Wilson 2249b5152584SMatthew Ahrens int 2250b5152584SMatthew Ahrens spa_maxblocksize(spa_t *spa) 2251b5152584SMatthew Ahrens { 2252b5152584SMatthew Ahrens if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) 2253b5152584SMatthew Ahrens return (SPA_MAXBLOCKSIZE); 2254b5152584SMatthew Ahrens else 2255b5152584SMatthew Ahrens return (SPA_OLD_MAXBLOCKSIZE); 2256b5152584SMatthew Ahrens } 22575cabbc6bSPrashanth Sreenivasa 225854811da5SToomas Soome int 225954811da5SToomas Soome spa_maxdnodesize(spa_t *spa) 226054811da5SToomas Soome { 226154811da5SToomas Soome if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) 226254811da5SToomas Soome return (DNODE_MAX_SIZE); 226354811da5SToomas Soome else 226454811da5SToomas Soome return (DNODE_MIN_SIZE); 226554811da5SToomas Soome } 226654811da5SToomas Soome 2267e0f1c0afSOlaf Faaland boolean_t 2268e0f1c0afSOlaf Faaland spa_multihost(spa_t *spa) 2269e0f1c0afSOlaf Faaland { 2270e0f1c0afSOlaf Faaland return (spa->spa_multihost ? B_TRUE : B_FALSE); 2271e0f1c0afSOlaf Faaland } 2272e0f1c0afSOlaf Faaland 2273e0f1c0afSOlaf Faaland unsigned long 2274e0f1c0afSOlaf Faaland spa_get_hostid(void) 2275e0f1c0afSOlaf Faaland { 2276e0f1c0afSOlaf Faaland unsigned long myhostid; 2277e0f1c0afSOlaf Faaland 2278e0f1c0afSOlaf Faaland #ifdef _KERNEL 2279e0f1c0afSOlaf Faaland myhostid = zone_get_hostid(NULL); 2280e0f1c0afSOlaf Faaland #else /* _KERNEL */ 2281e0f1c0afSOlaf Faaland /* 2282e0f1c0afSOlaf Faaland * We're emulating the system's hostid in userland, so 2283e0f1c0afSOlaf Faaland * we can't use zone_get_hostid(). 2284e0f1c0afSOlaf Faaland */ 2285e0f1c0afSOlaf Faaland (void) ddi_strtoul(hw_serial, NULL, 10, &myhostid); 2286e0f1c0afSOlaf Faaland #endif /* _KERNEL */ 2287e0f1c0afSOlaf Faaland 2288e0f1c0afSOlaf Faaland return (myhostid); 2289e0f1c0afSOlaf Faaland } 2290e0f1c0afSOlaf Faaland 22915cabbc6bSPrashanth Sreenivasa /* 22925cabbc6bSPrashanth Sreenivasa * Returns the txg that the last device removal completed. No indirect mappings 22935cabbc6bSPrashanth Sreenivasa * have been added since this txg. 22945cabbc6bSPrashanth Sreenivasa */ 22955cabbc6bSPrashanth Sreenivasa uint64_t 22965cabbc6bSPrashanth Sreenivasa spa_get_last_removal_txg(spa_t *spa) 22975cabbc6bSPrashanth Sreenivasa { 22985cabbc6bSPrashanth Sreenivasa uint64_t vdevid; 22995cabbc6bSPrashanth Sreenivasa uint64_t ret = -1ULL; 23005cabbc6bSPrashanth Sreenivasa 23015cabbc6bSPrashanth Sreenivasa spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); 23025cabbc6bSPrashanth Sreenivasa /* 23035cabbc6bSPrashanth Sreenivasa * sr_prev_indirect_vdev is only modified while holding all the 23045cabbc6bSPrashanth Sreenivasa * config locks, so it is sufficient to hold SCL_VDEV as reader when 23055cabbc6bSPrashanth Sreenivasa * examining it. 23065cabbc6bSPrashanth Sreenivasa */ 23075cabbc6bSPrashanth Sreenivasa vdevid = spa->spa_removing_phys.sr_prev_indirect_vdev; 23085cabbc6bSPrashanth Sreenivasa 23095cabbc6bSPrashanth Sreenivasa while (vdevid != -1ULL) { 23105cabbc6bSPrashanth Sreenivasa vdev_t *vd = vdev_lookup_top(spa, vdevid); 23115cabbc6bSPrashanth Sreenivasa vdev_indirect_births_t *vib = vd->vdev_indirect_births; 23125cabbc6bSPrashanth Sreenivasa 23135cabbc6bSPrashanth Sreenivasa ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops); 23145cabbc6bSPrashanth Sreenivasa 23155cabbc6bSPrashanth Sreenivasa /* 23165cabbc6bSPrashanth Sreenivasa * If the removal did not remap any data, we don't care. 23175cabbc6bSPrashanth Sreenivasa */ 23185cabbc6bSPrashanth Sreenivasa if (vdev_indirect_births_count(vib) != 0) { 23195cabbc6bSPrashanth Sreenivasa ret = vdev_indirect_births_last_entry_txg(vib); 23205cabbc6bSPrashanth Sreenivasa break; 23215cabbc6bSPrashanth Sreenivasa } 23225cabbc6bSPrashanth Sreenivasa 23235cabbc6bSPrashanth Sreenivasa vdevid = vd->vdev_indirect_config.vic_prev_indirect_vdev; 23245cabbc6bSPrashanth Sreenivasa } 23255cabbc6bSPrashanth Sreenivasa spa_config_exit(spa, SCL_VDEV, FTAG); 23265cabbc6bSPrashanth Sreenivasa 23275cabbc6bSPrashanth Sreenivasa IMPLY(ret != -1ULL, 23285cabbc6bSPrashanth Sreenivasa spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); 23295cabbc6bSPrashanth Sreenivasa 23305cabbc6bSPrashanth Sreenivasa return (ret); 23315cabbc6bSPrashanth Sreenivasa } 23326f793812SPavel Zakharov 23336f793812SPavel Zakharov boolean_t 23346f793812SPavel Zakharov spa_trust_config(spa_t *spa) 23356f793812SPavel Zakharov { 23366f793812SPavel Zakharov return (spa->spa_trust_config); 23376f793812SPavel Zakharov } 23386f793812SPavel Zakharov 23396f793812SPavel Zakharov uint64_t 23406f793812SPavel Zakharov spa_missing_tvds_allowed(spa_t *spa) 23416f793812SPavel Zakharov { 23426f793812SPavel Zakharov return (spa->spa_missing_tvds_allowed); 23436f793812SPavel Zakharov } 23446f793812SPavel Zakharov 2345814dcd43SSerapheim Dimitropoulos space_map_t * 2346814dcd43SSerapheim Dimitropoulos spa_syncing_log_sm(spa_t *spa) 2347814dcd43SSerapheim Dimitropoulos { 2348814dcd43SSerapheim Dimitropoulos return (spa->spa_syncing_log_sm); 2349814dcd43SSerapheim Dimitropoulos } 2350814dcd43SSerapheim Dimitropoulos 23516f793812SPavel Zakharov void 23526f793812SPavel Zakharov spa_set_missing_tvds(spa_t *spa, uint64_t missing) 23536f793812SPavel Zakharov { 23546f793812SPavel Zakharov spa->spa_missing_tvds = missing; 23556f793812SPavel Zakharov } 235686714001SSerapheim Dimitropoulos 235786714001SSerapheim Dimitropoulos boolean_t 235886714001SSerapheim Dimitropoulos spa_top_vdevs_spacemap_addressable(spa_t *spa) 235986714001SSerapheim Dimitropoulos { 236086714001SSerapheim Dimitropoulos vdev_t *rvd = spa->spa_root_vdev; 236186714001SSerapheim Dimitropoulos for (uint64_t c = 0; c < rvd->vdev_children; c++) { 236286714001SSerapheim Dimitropoulos if (!vdev_is_spacemap_addressable(rvd->vdev_child[c])) 236386714001SSerapheim Dimitropoulos return (B_FALSE); 236486714001SSerapheim Dimitropoulos } 236586714001SSerapheim Dimitropoulos return (B_TRUE); 236686714001SSerapheim Dimitropoulos } 236786714001SSerapheim Dimitropoulos 236886714001SSerapheim Dimitropoulos boolean_t 236986714001SSerapheim Dimitropoulos spa_has_checkpoint(spa_t *spa) 237086714001SSerapheim Dimitropoulos { 237186714001SSerapheim Dimitropoulos return (spa->spa_checkpoint_txg != 0); 237286714001SSerapheim Dimitropoulos } 237386714001SSerapheim Dimitropoulos 237486714001SSerapheim Dimitropoulos boolean_t 237586714001SSerapheim Dimitropoulos spa_importing_readonly_checkpoint(spa_t *spa) 237686714001SSerapheim Dimitropoulos { 237786714001SSerapheim Dimitropoulos return ((spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT) && 237886714001SSerapheim Dimitropoulos spa->spa_mode == FREAD); 237986714001SSerapheim Dimitropoulos } 238086714001SSerapheim Dimitropoulos 238186714001SSerapheim Dimitropoulos uint64_t 238286714001SSerapheim Dimitropoulos spa_min_claim_txg(spa_t *spa) 238386714001SSerapheim Dimitropoulos { 238486714001SSerapheim Dimitropoulos uint64_t checkpoint_txg = spa->spa_uberblock.ub_checkpoint_txg; 238586714001SSerapheim Dimitropoulos 238686714001SSerapheim Dimitropoulos if (checkpoint_txg != 0) 238786714001SSerapheim Dimitropoulos return (checkpoint_txg + 1); 238886714001SSerapheim Dimitropoulos 238986714001SSerapheim Dimitropoulos return (spa->spa_first_txg); 239086714001SSerapheim Dimitropoulos } 239186714001SSerapheim Dimitropoulos 239286714001SSerapheim Dimitropoulos /* 239386714001SSerapheim Dimitropoulos * If there is a checkpoint, async destroys may consume more space from 239486714001SSerapheim Dimitropoulos * the pool instead of freeing it. In an attempt to save the pool from 239586714001SSerapheim Dimitropoulos * getting suspended when it is about to run out of space, we stop 239686714001SSerapheim Dimitropoulos * processing async destroys. 239786714001SSerapheim Dimitropoulos */ 239886714001SSerapheim Dimitropoulos boolean_t 239986714001SSerapheim Dimitropoulos spa_suspend_async_destroy(spa_t *spa) 240086714001SSerapheim Dimitropoulos { 240186714001SSerapheim Dimitropoulos dsl_pool_t *dp = spa_get_dsl(spa); 240286714001SSerapheim Dimitropoulos 240386714001SSerapheim Dimitropoulos uint64_t unreserved = dsl_pool_unreserved_space(dp, 240486714001SSerapheim Dimitropoulos ZFS_SPACE_CHECK_EXTRA_RESERVED); 240586714001SSerapheim Dimitropoulos uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes; 240686714001SSerapheim Dimitropoulos uint64_t avail = (unreserved > used) ? (unreserved - used) : 0; 240786714001SSerapheim Dimitropoulos 240886714001SSerapheim Dimitropoulos if (spa_has_checkpoint(spa) && avail == 0) 240986714001SSerapheim Dimitropoulos return (B_TRUE); 241086714001SSerapheim Dimitropoulos 241186714001SSerapheim Dimitropoulos return (B_FALSE); 241286714001SSerapheim Dimitropoulos } 2413