1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or https://opensource.org/licenses/CDDL-1.0. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright (c) 2011, 2019 by Delphix. All rights reserved. 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 26 * Copyright 2013 Saso Kiselkov. All rights reserved. 27 * Copyright (c) 2016 Actifio, Inc. All rights reserved. 28 * Copyright (c) 2017, Intel Corporation. 29 * Copyright (c) 2019 Datto Inc. 30 */ 31 32 #ifndef _SYS_SPA_IMPL_H 33 #define _SYS_SPA_IMPL_H 34 35 #include <sys/spa.h> 36 #include <sys/spa_checkpoint.h> 37 #include <sys/spa_log_spacemap.h> 38 #include <sys/vdev.h> 39 #include <sys/vdev_rebuild.h> 40 #include <sys/vdev_removal.h> 41 #include <sys/metaslab.h> 42 #include <sys/dmu.h> 43 #include <sys/dsl_pool.h> 44 #include <sys/uberblock_impl.h> 45 #include <sys/zfs_context.h> 46 #include <sys/avl.h> 47 #include <sys/zfs_refcount.h> 48 #include <sys/bplist.h> 49 #include <sys/bpobj.h> 50 #include <sys/dsl_crypt.h> 51 #include <sys/zfeature.h> 52 #include <sys/zthr.h> 53 #include <sys/dsl_deadlist.h> 54 #include <zfeature_common.h> 55 56 #ifdef __cplusplus 57 extern "C" { 58 #endif 59 60 typedef struct spa_alloc { 61 kmutex_t spaa_lock; 62 avl_tree_t spaa_tree; 63 } ____cacheline_aligned spa_alloc_t; 64 65 typedef struct spa_error_entry { 66 zbookmark_phys_t se_bookmark; 67 char *se_name; 68 avl_node_t se_avl; 69 zbookmark_err_phys_t se_zep; /* not accounted in avl_find */ 70 } spa_error_entry_t; 71 72 typedef struct spa_history_phys { 73 uint64_t sh_pool_create_len; /* ending offset of zpool create */ 74 uint64_t sh_phys_max_off; /* physical EOF */ 75 uint64_t sh_bof; /* logical BOF */ 76 uint64_t sh_eof; /* logical EOF */ 77 uint64_t sh_records_lost; /* num of records overwritten */ 78 } spa_history_phys_t; 79 80 /* 81 * All members must be uint64_t, for byteswap purposes. 82 */ 83 typedef struct spa_removing_phys { 84 uint64_t sr_state; /* dsl_scan_state_t */ 85 86 /* 87 * The vdev ID that we most recently attempted to remove, 88 * or -1 if no removal has been attempted. 89 */ 90 uint64_t sr_removing_vdev; 91 92 /* 93 * The vdev ID that we most recently successfully removed, 94 * or -1 if no devices have been removed. 95 */ 96 uint64_t sr_prev_indirect_vdev; 97 98 uint64_t sr_start_time; 99 uint64_t sr_end_time; 100 101 /* 102 * Note that we can not use the space map's or indirect mapping's 103 * accounting as a substitute for these values, because we need to 104 * count frees of not-yet-copied data as though it did the copy. 105 * Otherwise, we could get into a situation where copied > to_copy, 106 * or we complete before copied == to_copy. 107 */ 108 uint64_t sr_to_copy; /* bytes that need to be copied */ 109 uint64_t sr_copied; /* bytes that have been copied or freed */ 110 } spa_removing_phys_t; 111 112 /* 113 * This struct is stored as an entry in the DMU_POOL_DIRECTORY_OBJECT 114 * (with key DMU_POOL_CONDENSING_INDIRECT). It is present if a condense 115 * of an indirect vdev's mapping object is in progress. 116 */ 117 typedef struct spa_condensing_indirect_phys { 118 /* 119 * The vdev ID of the indirect vdev whose indirect mapping is 120 * being condensed. 121 */ 122 uint64_t scip_vdev; 123 124 /* 125 * The vdev's old obsolete spacemap. This spacemap's contents are 126 * being integrated into the new mapping. 127 */ 128 uint64_t scip_prev_obsolete_sm_object; 129 130 /* 131 * The new mapping object that is being created. 132 */ 133 uint64_t scip_next_mapping_object; 134 } spa_condensing_indirect_phys_t; 135 136 struct spa_aux_vdev { 137 uint64_t sav_object; /* MOS object for device list */ 138 nvlist_t *sav_config; /* cached device config */ 139 vdev_t **sav_vdevs; /* devices */ 140 int sav_count; /* number devices */ 141 boolean_t sav_sync; /* sync the device list */ 142 nvlist_t **sav_pending; /* pending device additions */ 143 uint_t sav_npending; /* # pending devices */ 144 }; 145 146 typedef struct spa_config_lock { 147 kmutex_t scl_lock; 148 kthread_t *scl_writer; 149 int scl_write_wanted; 150 int scl_count; 151 kcondvar_t scl_cv; 152 } ____cacheline_aligned spa_config_lock_t; 153 154 typedef struct spa_config_dirent { 155 list_node_t scd_link; 156 char *scd_path; 157 } spa_config_dirent_t; 158 159 typedef enum zio_taskq_type { 160 ZIO_TASKQ_ISSUE = 0, 161 ZIO_TASKQ_ISSUE_HIGH, 162 ZIO_TASKQ_INTERRUPT, 163 ZIO_TASKQ_INTERRUPT_HIGH, 164 ZIO_TASKQ_TYPES 165 } zio_taskq_type_t; 166 167 /* 168 * State machine for the zpool-poolname process. The states transitions 169 * are done as follows: 170 * 171 * From To Routine 172 * PROC_NONE -> PROC_CREATED spa_activate() 173 * PROC_CREATED -> PROC_ACTIVE spa_thread() 174 * PROC_ACTIVE -> PROC_DEACTIVATE spa_deactivate() 175 * PROC_DEACTIVATE -> PROC_GONE spa_thread() 176 * PROC_GONE -> PROC_NONE spa_deactivate() 177 */ 178 typedef enum spa_proc_state { 179 SPA_PROC_NONE, /* spa_proc = &p0, no process created */ 180 SPA_PROC_CREATED, /* spa_activate() has proc, is waiting */ 181 SPA_PROC_ACTIVE, /* taskqs created, spa_proc set */ 182 SPA_PROC_DEACTIVATE, /* spa_deactivate() requests process exit */ 183 SPA_PROC_GONE /* spa_thread() is exiting, spa_proc = &p0 */ 184 } spa_proc_state_t; 185 186 typedef struct spa_taskqs { 187 uint_t stqs_count; 188 taskq_t **stqs_taskq; 189 } spa_taskqs_t; 190 191 typedef enum spa_all_vdev_zap_action { 192 AVZ_ACTION_NONE = 0, 193 AVZ_ACTION_DESTROY, /* Destroy all per-vdev ZAPs and the AVZ. */ 194 AVZ_ACTION_REBUILD, /* Populate the new AVZ, see spa_avz_rebuild */ 195 AVZ_ACTION_INITIALIZE 196 } spa_avz_action_t; 197 198 typedef enum spa_config_source { 199 SPA_CONFIG_SRC_NONE = 0, 200 SPA_CONFIG_SRC_SCAN, /* scan of path (default: /dev/dsk) */ 201 SPA_CONFIG_SRC_CACHEFILE, /* any cachefile */ 202 SPA_CONFIG_SRC_TRYIMPORT, /* returned from call to tryimport */ 203 SPA_CONFIG_SRC_SPLIT, /* new pool in a pool split */ 204 SPA_CONFIG_SRC_MOS /* MOS, but not always from right txg */ 205 } spa_config_source_t; 206 207 struct spa { 208 /* 209 * Fields protected by spa_namespace_lock. 210 */ 211 char spa_name[ZFS_MAX_DATASET_NAME_LEN]; /* pool name */ 212 char *spa_comment; /* comment */ 213 avl_node_t spa_avl; /* node in spa_namespace_avl */ 214 nvlist_t *spa_config; /* last synced config */ 215 nvlist_t *spa_config_syncing; /* currently syncing config */ 216 nvlist_t *spa_config_splitting; /* config for splitting */ 217 nvlist_t *spa_load_info; /* info and errors from load */ 218 uint64_t spa_config_txg; /* txg of last config change */ 219 uint32_t spa_sync_pass; /* iterate-to-convergence */ 220 pool_state_t spa_state; /* pool state */ 221 int spa_inject_ref; /* injection references */ 222 uint8_t spa_sync_on; /* sync threads are running */ 223 spa_load_state_t spa_load_state; /* current load operation */ 224 boolean_t spa_indirect_vdevs_loaded; /* mappings loaded? */ 225 boolean_t spa_trust_config; /* do we trust vdev tree? */ 226 boolean_t spa_is_splitting; /* in the middle of a split? */ 227 spa_config_source_t spa_config_source; /* where config comes from? */ 228 uint64_t spa_import_flags; /* import specific flags */ 229 spa_taskqs_t spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES]; 230 dsl_pool_t *spa_dsl_pool; 231 boolean_t spa_is_initializing; /* true while opening pool */ 232 boolean_t spa_is_exporting; /* true while exporting pool */ 233 metaslab_class_t *spa_normal_class; /* normal data class */ 234 metaslab_class_t *spa_log_class; /* intent log data class */ 235 metaslab_class_t *spa_embedded_log_class; /* log on normal vdevs */ 236 metaslab_class_t *spa_special_class; /* special allocation class */ 237 metaslab_class_t *spa_dedup_class; /* dedup allocation class */ 238 uint64_t spa_first_txg; /* first txg after spa_open() */ 239 uint64_t spa_final_txg; /* txg of export/destroy */ 240 uint64_t spa_freeze_txg; /* freeze pool at this txg */ 241 uint64_t spa_load_max_txg; /* best initial ub_txg */ 242 uint64_t spa_claim_max_txg; /* highest claimed birth txg */ 243 inode_timespec_t spa_loaded_ts; /* 1st successful open time */ 244 objset_t *spa_meta_objset; /* copy of dp->dp_meta_objset */ 245 kmutex_t spa_evicting_os_lock; /* Evicting objset list lock */ 246 list_t spa_evicting_os_list; /* Objsets being evicted. */ 247 kcondvar_t spa_evicting_os_cv; /* Objset Eviction Completion */ 248 txg_list_t spa_vdev_txg_list; /* per-txg dirty vdev list */ 249 vdev_t *spa_root_vdev; /* top-level vdev container */ 250 uint64_t spa_min_ashift; /* of vdevs in normal class */ 251 uint64_t spa_max_ashift; /* of vdevs in normal class */ 252 uint64_t spa_min_alloc; /* of vdevs in normal class */ 253 uint64_t spa_config_guid; /* config pool guid */ 254 uint64_t spa_load_guid; /* spa_load initialized guid */ 255 uint64_t spa_last_synced_guid; /* last synced guid */ 256 list_t spa_config_dirty_list; /* vdevs with dirty config */ 257 list_t spa_state_dirty_list; /* vdevs with dirty state */ 258 /* 259 * spa_allocs is an array, whose lengths is stored in spa_alloc_count. 260 * There is one tree and one lock for each allocator, to help improve 261 * allocation performance in write-heavy workloads. 262 */ 263 spa_alloc_t *spa_allocs; 264 int spa_alloc_count; 265 266 spa_aux_vdev_t spa_spares; /* hot spares */ 267 spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */ 268 nvlist_t *spa_label_features; /* Features for reading MOS */ 269 uint64_t spa_config_object; /* MOS object for pool config */ 270 uint64_t spa_config_generation; /* config generation number */ 271 uint64_t spa_syncing_txg; /* txg currently syncing */ 272 bpobj_t spa_deferred_bpobj; /* deferred-free bplist */ 273 bplist_t spa_free_bplist[TXG_SIZE]; /* bplist of stuff to free */ 274 zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ 275 /* checksum context templates */ 276 kmutex_t spa_cksum_tmpls_lock; 277 void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; 278 uberblock_t spa_ubsync; /* last synced uberblock */ 279 uberblock_t spa_uberblock; /* current uberblock */ 280 boolean_t spa_extreme_rewind; /* rewind past deferred frees */ 281 kmutex_t spa_scrub_lock; /* resilver/scrub lock */ 282 uint64_t spa_scrub_inflight; /* in-flight scrub bytes */ 283 284 /* in-flight verification bytes */ 285 uint64_t spa_load_verify_bytes; 286 kcondvar_t spa_scrub_io_cv; /* scrub I/O completion */ 287 uint8_t spa_scrub_active; /* active or suspended? */ 288 uint8_t spa_scrub_type; /* type of scrub we're doing */ 289 uint8_t spa_scrub_finished; /* indicator to rotate logs */ 290 uint8_t spa_scrub_started; /* started since last boot */ 291 uint8_t spa_scrub_reopen; /* scrub doing vdev_reopen */ 292 uint64_t spa_scan_pass_start; /* start time per pass/reboot */ 293 uint64_t spa_scan_pass_scrub_pause; /* scrub pause time */ 294 uint64_t spa_scan_pass_scrub_spent_paused; /* total paused */ 295 uint64_t spa_scan_pass_exam; /* examined bytes per pass */ 296 uint64_t spa_scan_pass_issued; /* issued bytes per pass */ 297 298 /* 299 * We are in the middle of a resilver, and another resilver 300 * is needed once this one completes. This is set iff any 301 * vdev_resilver_deferred is set. 302 */ 303 boolean_t spa_resilver_deferred; 304 kmutex_t spa_async_lock; /* protect async state */ 305 kthread_t *spa_async_thread; /* thread doing async task */ 306 int spa_async_suspended; /* async tasks suspended */ 307 kcondvar_t spa_async_cv; /* wait for thread_exit() */ 308 uint16_t spa_async_tasks; /* async task mask */ 309 uint64_t spa_missing_tvds; /* unopenable tvds on load */ 310 uint64_t spa_missing_tvds_allowed; /* allow loading spa? */ 311 312 uint64_t spa_nonallocating_dspace; 313 spa_removing_phys_t spa_removing_phys; 314 spa_vdev_removal_t *spa_vdev_removal; 315 316 spa_condensing_indirect_phys_t spa_condensing_indirect_phys; 317 spa_condensing_indirect_t *spa_condensing_indirect; 318 zthr_t *spa_condense_zthr; /* zthr doing condense. */ 319 320 uint64_t spa_checkpoint_txg; /* the txg of the checkpoint */ 321 spa_checkpoint_info_t spa_checkpoint_info; /* checkpoint accounting */ 322 zthr_t *spa_checkpoint_discard_zthr; 323 324 space_map_t *spa_syncing_log_sm; /* current log space map */ 325 avl_tree_t spa_sm_logs_by_txg; 326 kmutex_t spa_flushed_ms_lock; /* for metaslabs_by_flushed */ 327 avl_tree_t spa_metaslabs_by_flushed; 328 spa_unflushed_stats_t spa_unflushed_stats; 329 list_t spa_log_summary; 330 uint64_t spa_log_flushall_txg; 331 332 zthr_t *spa_livelist_delete_zthr; /* deleting livelists */ 333 zthr_t *spa_livelist_condense_zthr; /* condensing livelists */ 334 uint64_t spa_livelists_to_delete; /* set of livelists to free */ 335 livelist_condense_entry_t spa_to_condense; /* next to condense */ 336 337 char *spa_root; /* alternate root directory */ 338 uint64_t spa_ena; /* spa-wide ereport ENA */ 339 int spa_last_open_failed; /* error if last open failed */ 340 uint64_t spa_last_ubsync_txg; /* "best" uberblock txg */ 341 uint64_t spa_last_ubsync_txg_ts; /* timestamp from that ub */ 342 uint64_t spa_load_txg; /* ub txg that loaded */ 343 uint64_t spa_load_txg_ts; /* timestamp from that ub */ 344 uint64_t spa_load_meta_errors; /* verify metadata err count */ 345 uint64_t spa_load_data_errors; /* verify data err count */ 346 uint64_t spa_verify_min_txg; /* start txg of verify scrub */ 347 kmutex_t spa_errlog_lock; /* error log lock */ 348 uint64_t spa_errlog_last; /* last error log object */ 349 uint64_t spa_errlog_scrub; /* scrub error log object */ 350 kmutex_t spa_errlist_lock; /* error list/ereport lock */ 351 avl_tree_t spa_errlist_last; /* last error list */ 352 avl_tree_t spa_errlist_scrub; /* scrub error list */ 353 avl_tree_t spa_errlist_healed; /* list of healed blocks */ 354 uint64_t spa_deflate; /* should we deflate? */ 355 uint64_t spa_history; /* history object */ 356 kmutex_t spa_history_lock; /* history lock */ 357 vdev_t *spa_pending_vdev; /* pending vdev additions */ 358 kmutex_t spa_props_lock; /* property lock */ 359 uint64_t spa_pool_props_object; /* object for properties */ 360 uint64_t spa_bootfs; /* default boot filesystem */ 361 uint64_t spa_failmode; /* failure mode for the pool */ 362 uint64_t spa_deadman_failmode; /* failure mode for deadman */ 363 uint64_t spa_delegation; /* delegation on/off */ 364 list_t spa_config_list; /* previous cache file(s) */ 365 /* per-CPU array of root of async I/O: */ 366 zio_t **spa_async_zio_root; 367 zio_t *spa_suspend_zio_root; /* root of all suspended I/O */ 368 zio_t *spa_txg_zio[TXG_SIZE]; /* spa_sync() waits for this */ 369 kmutex_t spa_suspend_lock; /* protects suspend_zio_root */ 370 kcondvar_t spa_suspend_cv; /* notification of resume */ 371 zio_suspend_reason_t spa_suspended; /* pool is suspended */ 372 uint8_t spa_claiming; /* pool is doing zil_claim() */ 373 boolean_t spa_is_root; /* pool is root */ 374 int spa_minref; /* num refs when first opened */ 375 spa_mode_t spa_mode; /* SPA_MODE_{READ|WRITE} */ 376 boolean_t spa_read_spacemaps; /* spacemaps available if ro */ 377 spa_log_state_t spa_log_state; /* log state */ 378 uint64_t spa_autoexpand; /* lun expansion on/off */ 379 ddt_t *spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; /* in-core DDTs */ 380 uint64_t spa_ddt_stat_object; /* DDT statistics */ 381 uint64_t spa_dedup_dspace; /* Cache get_dedup_dspace() */ 382 uint64_t spa_dedup_checksum; /* default dedup checksum */ 383 uint64_t spa_dspace; /* dspace in normal class */ 384 struct brt *spa_brt; /* in-core BRT */ 385 kmutex_t spa_vdev_top_lock; /* dueling offline/remove */ 386 kmutex_t spa_proc_lock; /* protects spa_proc* */ 387 kcondvar_t spa_proc_cv; /* spa_proc_state transitions */ 388 spa_proc_state_t spa_proc_state; /* see definition */ 389 proc_t *spa_proc; /* "zpool-poolname" process */ 390 uintptr_t spa_did; /* if procp != p0, did of t1 */ 391 boolean_t spa_autoreplace; /* autoreplace set in open */ 392 int spa_vdev_locks; /* locks grabbed */ 393 uint64_t spa_creation_version; /* version at pool creation */ 394 uint64_t spa_prev_software_version; /* See ub_software_version */ 395 uint64_t spa_feat_for_write_obj; /* required to write to pool */ 396 uint64_t spa_feat_for_read_obj; /* required to read from pool */ 397 uint64_t spa_feat_desc_obj; /* Feature descriptions */ 398 uint64_t spa_feat_enabled_txg_obj; /* Feature enabled txg */ 399 kmutex_t spa_feat_stats_lock; /* protects spa_feat_stats */ 400 nvlist_t *spa_feat_stats; /* Cache of enabled features */ 401 /* cache feature refcounts */ 402 uint64_t spa_feat_refcount_cache[SPA_FEATURES]; 403 taskqid_t spa_deadman_tqid; /* Task id */ 404 uint64_t spa_deadman_calls; /* number of deadman calls */ 405 hrtime_t spa_sync_starttime; /* starting time of spa_sync */ 406 uint64_t spa_deadman_synctime; /* deadman sync expiration */ 407 uint64_t spa_deadman_ziotime; /* deadman zio expiration */ 408 uint64_t spa_all_vdev_zaps; /* ZAP of per-vd ZAP obj #s */ 409 spa_avz_action_t spa_avz_action; /* destroy/rebuild AVZ? */ 410 uint64_t spa_autotrim; /* automatic background trim? */ 411 uint64_t spa_errata; /* errata issues detected */ 412 spa_stats_t spa_stats; /* assorted spa statistics */ 413 spa_keystore_t spa_keystore; /* loaded crypto keys */ 414 415 /* arc_memory_throttle() parameters during low memory condition */ 416 uint64_t spa_lowmem_page_load; /* memory load during txg */ 417 uint64_t spa_lowmem_last_txg; /* txg window start */ 418 419 hrtime_t spa_ccw_fail_time; /* Conf cache write fail time */ 420 taskq_t *spa_zvol_taskq; /* Taskq for minor management */ 421 taskq_t *spa_prefetch_taskq; /* Taskq for prefetch threads */ 422 uint64_t spa_multihost; /* multihost aware (mmp) */ 423 mmp_thread_t spa_mmp; /* multihost mmp thread */ 424 list_t spa_leaf_list; /* list of leaf vdevs */ 425 uint64_t spa_leaf_list_gen; /* track leaf_list changes */ 426 uint32_t spa_hostid; /* cached system hostid */ 427 428 /* synchronization for threads in spa_wait */ 429 kmutex_t spa_activities_lock; 430 kcondvar_t spa_activities_cv; 431 kcondvar_t spa_waiters_cv; 432 int spa_waiters; /* number of waiting threads */ 433 boolean_t spa_waiters_cancel; /* waiters should return */ 434 435 char *spa_compatibility; /* compatibility file(s) */ 436 437 /* 438 * spa_refcount & spa_config_lock must be the last elements 439 * because zfs_refcount_t changes size based on compilation options. 440 * In order for the MDB module to function correctly, the other 441 * fields must remain in the same location. 442 */ 443 spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */ 444 zfs_refcount_t spa_refcount; /* number of opens */ 445 446 taskq_t *spa_upgrade_taskq; /* taskq for upgrade jobs */ 447 }; 448 449 extern char *spa_config_path; 450 extern const char *zfs_deadman_failmode; 451 extern uint_t spa_slop_shift; 452 extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q, 453 task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent); 454 extern void spa_taskq_dispatch_sync(spa_t *, zio_type_t t, zio_taskq_type_t q, 455 task_func_t *func, void *arg, uint_t flags); 456 extern void spa_load_spares(spa_t *spa); 457 extern void spa_load_l2cache(spa_t *spa); 458 extern sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, 459 const char *name); 460 extern void spa_event_post(sysevent_t *ev); 461 extern int param_set_deadman_failmode_common(const char *val); 462 extern void spa_set_deadman_synctime(hrtime_t ns); 463 extern void spa_set_deadman_ziotime(hrtime_t ns); 464 extern const char *spa_history_zone(void); 465 466 #ifdef __cplusplus 467 } 468 #endif 469 470 #endif /* _SYS_SPA_IMPL_H */ 471