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, 2020 by Delphix. All rights reserved. 24 * Copyright (c) 2017, Intel Corporation. 25 * Copyright (c) 2023, Klara Inc. 26 */ 27 28 #ifndef _SYS_VDEV_IMPL_H 29 #define _SYS_VDEV_IMPL_H 30 31 #include <sys/avl.h> 32 #include <sys/bpobj.h> 33 #include <sys/dmu.h> 34 #include <sys/metaslab.h> 35 #include <sys/nvpair.h> 36 #include <sys/space_map.h> 37 #include <sys/vdev.h> 38 #include <sys/uberblock_impl.h> 39 #include <sys/vdev_indirect_mapping.h> 40 #include <sys/vdev_indirect_births.h> 41 #include <sys/vdev_rebuild.h> 42 #include <sys/vdev_removal.h> 43 #include <sys/zfs_ratelimit.h> 44 45 #ifdef __cplusplus 46 extern "C" { 47 #endif 48 49 /* 50 * Virtual device descriptors. 51 * 52 * All storage pool operations go through the virtual device framework, 53 * which provides data replication and I/O scheduling. 54 */ 55 56 /* 57 * Forward declarations that lots of things need. 58 */ 59 typedef struct vdev_queue vdev_queue_t; 60 struct abd; 61 62 extern uint_t zfs_vdev_queue_depth_pct; 63 extern uint_t zfs_vdev_def_queue_depth; 64 extern uint_t zfs_vdev_async_write_max_active; 65 66 /* 67 * Virtual device operations 68 */ 69 typedef int vdev_init_func_t(spa_t *spa, nvlist_t *nv, void **tsd); 70 typedef void vdev_kobj_post_evt_func_t(vdev_t *vd); 71 typedef void vdev_fini_func_t(vdev_t *vd); 72 typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size, 73 uint64_t *ashift, uint64_t *pshift); 74 typedef void vdev_close_func_t(vdev_t *vd); 75 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize, uint64_t txg); 76 typedef uint64_t vdev_min_asize_func_t(vdev_t *vd); 77 typedef uint64_t vdev_min_alloc_func_t(vdev_t *vd); 78 typedef void vdev_io_start_func_t(zio_t *zio); 79 typedef void vdev_io_done_func_t(zio_t *zio); 80 typedef void vdev_state_change_func_t(vdev_t *vd, int, int); 81 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, const dva_t *dva, 82 size_t psize, uint64_t phys_birth); 83 typedef void vdev_hold_func_t(vdev_t *vd); 84 typedef void vdev_rele_func_t(vdev_t *vd); 85 86 typedef void vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd, 87 uint64_t offset, uint64_t size, void *arg); 88 typedef void vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size, 89 vdev_remap_cb_t callback, void *arg); 90 /* 91 * Given a target vdev, translates the logical range "in" to the physical 92 * range "res" 93 */ 94 typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg64_t *logical, 95 range_seg64_t *physical, range_seg64_t *remain); 96 typedef uint64_t vdev_rebuild_asize_func_t(vdev_t *vd, uint64_t start, 97 uint64_t size, uint64_t max_segment); 98 typedef void vdev_metaslab_init_func_t(vdev_t *vd, uint64_t *startp, 99 uint64_t *sizep); 100 typedef void vdev_config_generate_func_t(vdev_t *vd, nvlist_t *nv); 101 typedef uint64_t vdev_nparity_func_t(vdev_t *vd); 102 typedef uint64_t vdev_ndisks_func_t(vdev_t *vd); 103 104 typedef const struct vdev_ops { 105 vdev_init_func_t *vdev_op_init; 106 vdev_fini_func_t *vdev_op_fini; 107 vdev_open_func_t *vdev_op_open; 108 vdev_close_func_t *vdev_op_close; 109 vdev_asize_func_t *vdev_op_asize; 110 vdev_min_asize_func_t *vdev_op_min_asize; 111 vdev_min_alloc_func_t *vdev_op_min_alloc; 112 vdev_io_start_func_t *vdev_op_io_start; 113 vdev_io_done_func_t *vdev_op_io_done; 114 vdev_state_change_func_t *vdev_op_state_change; 115 vdev_need_resilver_func_t *vdev_op_need_resilver; 116 vdev_hold_func_t *vdev_op_hold; 117 vdev_rele_func_t *vdev_op_rele; 118 vdev_remap_func_t *vdev_op_remap; 119 vdev_xlation_func_t *vdev_op_xlate; 120 vdev_rebuild_asize_func_t *vdev_op_rebuild_asize; 121 vdev_metaslab_init_func_t *vdev_op_metaslab_init; 122 vdev_config_generate_func_t *vdev_op_config_generate; 123 vdev_nparity_func_t *vdev_op_nparity; 124 vdev_ndisks_func_t *vdev_op_ndisks; 125 vdev_kobj_post_evt_func_t *vdev_op_kobj_evt_post; 126 char vdev_op_type[16]; 127 boolean_t vdev_op_leaf; 128 } vdev_ops_t; 129 130 /* 131 * Virtual device properties 132 */ 133 typedef union vdev_queue_class { 134 struct { 135 ulong_t vqc_list_numnodes; 136 list_t vqc_list; 137 }; 138 avl_tree_t vqc_tree; 139 } vdev_queue_class_t; 140 141 struct vdev_queue { 142 vdev_t *vq_vdev; 143 vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE]; 144 avl_tree_t vq_read_offset_tree; 145 avl_tree_t vq_write_offset_tree; 146 uint64_t vq_last_offset; 147 zio_priority_t vq_last_prio; /* Last sent I/O priority. */ 148 uint32_t vq_cqueued; /* Classes with queued I/Os. */ 149 uint32_t vq_cactive[ZIO_PRIORITY_NUM_QUEUEABLE]; 150 uint32_t vq_active; /* Number of active I/Os. */ 151 uint32_t vq_ia_active; /* Active interactive I/Os. */ 152 uint32_t vq_nia_credit; /* Non-interactive I/Os credit. */ 153 list_t vq_active_list; /* List of active I/Os. */ 154 hrtime_t vq_io_complete_ts; /* time last i/o completed */ 155 hrtime_t vq_io_delta_ts; 156 zio_t vq_io_search; /* used as local for stack reduction */ 157 kmutex_t vq_lock; 158 }; 159 160 typedef enum vdev_alloc_bias { 161 VDEV_BIAS_NONE, 162 VDEV_BIAS_LOG, /* dedicated to ZIL data (SLOG) */ 163 VDEV_BIAS_SPECIAL, /* dedicated to ddt, metadata, and small blks */ 164 VDEV_BIAS_DEDUP /* dedicated to dedup metadata */ 165 } vdev_alloc_bias_t; 166 167 168 /* 169 * On-disk indirect vdev state. 170 * 171 * An indirect vdev is described exclusively in the MOS config of a pool. 172 * The config for an indirect vdev includes several fields, which are 173 * accessed in memory by a vdev_indirect_config_t. 174 */ 175 typedef struct vdev_indirect_config { 176 /* 177 * Object (in MOS) which contains the indirect mapping. This object 178 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by 179 * vimep_src. The bonus buffer for this object is a 180 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev 181 * removal is initiated. 182 * 183 * Note that this object can be empty if none of the data on the vdev 184 * has been copied yet. 185 */ 186 uint64_t vic_mapping_object; 187 188 /* 189 * Object (in MOS) which contains the birth times for the mapping 190 * entries. This object contains an array of 191 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus 192 * buffer for this object is a vdev_indirect_birth_phys_t. This object 193 * is allocated when a vdev removal is initiated. 194 * 195 * Note that this object can be empty if none of the vdev has yet been 196 * copied. 197 */ 198 uint64_t vic_births_object; 199 200 /* 201 * This is the vdev ID which was removed previous to this vdev, or 202 * UINT64_MAX if there are no previously removed vdevs. 203 */ 204 uint64_t vic_prev_indirect_vdev; 205 } vdev_indirect_config_t; 206 207 /* 208 * Virtual device descriptor 209 */ 210 struct vdev { 211 /* 212 * Common to all vdev types. 213 */ 214 uint64_t vdev_id; /* child number in vdev parent */ 215 uint64_t vdev_guid; /* unique ID for this vdev */ 216 uint64_t vdev_guid_sum; /* self guid + all child guids */ 217 uint64_t vdev_orig_guid; /* orig. guid prior to remove */ 218 uint64_t vdev_asize; /* allocatable device capacity */ 219 uint64_t vdev_min_asize; /* min acceptable asize */ 220 uint64_t vdev_max_asize; /* max acceptable asize */ 221 uint64_t vdev_ashift; /* block alignment shift */ 222 223 /* 224 * Logical block alignment shift 225 * 226 * The smallest sized/aligned I/O supported by the device. 227 */ 228 uint64_t vdev_logical_ashift; 229 /* 230 * Physical block alignment shift 231 * 232 * The device supports logical I/Os with vdev_logical_ashift 233 * size/alignment, but optimum performance will be achieved by 234 * aligning/sizing requests to vdev_physical_ashift. Smaller 235 * requests may be inflated or incur device level read-modify-write 236 * operations. 237 * 238 * May be 0 to indicate no preference (i.e. use vdev_logical_ashift). 239 */ 240 uint64_t vdev_physical_ashift; 241 uint64_t vdev_state; /* see VDEV_STATE_* #defines */ 242 uint64_t vdev_prevstate; /* used when reopening a vdev */ 243 vdev_ops_t *vdev_ops; /* vdev operations */ 244 spa_t *vdev_spa; /* spa for this vdev */ 245 void *vdev_tsd; /* type-specific data */ 246 vdev_t *vdev_top; /* top-level vdev */ 247 vdev_t *vdev_parent; /* parent vdev */ 248 vdev_t **vdev_child; /* array of children */ 249 uint64_t vdev_children; /* number of children */ 250 vdev_stat_t vdev_stat; /* virtual device statistics */ 251 vdev_stat_ex_t vdev_stat_ex; /* extended statistics */ 252 boolean_t vdev_expanding; /* expand the vdev? */ 253 boolean_t vdev_reopening; /* reopen in progress? */ 254 boolean_t vdev_nonrot; /* true if solid state */ 255 int vdev_load_error; /* error on last load */ 256 int vdev_open_error; /* error on last open */ 257 int vdev_validate_error; /* error on last validate */ 258 kthread_t *vdev_open_thread; /* thread opening children */ 259 kthread_t *vdev_validate_thread; /* thread validating children */ 260 uint64_t vdev_crtxg; /* txg when top-level was added */ 261 uint64_t vdev_root_zap; 262 263 /* 264 * Top-level vdev state. 265 */ 266 uint64_t vdev_ms_array; /* metaslab array object */ 267 uint64_t vdev_ms_shift; /* metaslab size shift */ 268 uint64_t vdev_ms_count; /* number of metaslabs */ 269 metaslab_group_t *vdev_mg; /* metaslab group */ 270 metaslab_group_t *vdev_log_mg; /* embedded slog metaslab group */ 271 metaslab_t **vdev_ms; /* metaslab array */ 272 txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */ 273 txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */ 274 txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */ 275 boolean_t vdev_remove_wanted; /* async remove wanted? */ 276 boolean_t vdev_fault_wanted; /* async faulted wanted? */ 277 list_node_t vdev_config_dirty_node; /* config dirty list */ 278 list_node_t vdev_state_dirty_node; /* state dirty list */ 279 uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */ 280 uint64_t vdev_islog; /* is an intent log device */ 281 uint64_t vdev_noalloc; /* device is passivated? */ 282 uint64_t vdev_removing; /* device is being removed? */ 283 uint64_t vdev_failfast; /* device failfast setting */ 284 boolean_t vdev_rz_expanding; /* raidz is being expanded? */ 285 boolean_t vdev_ishole; /* is a hole in the namespace */ 286 uint64_t vdev_top_zap; 287 vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias */ 288 289 /* pool checkpoint related */ 290 space_map_t *vdev_checkpoint_sm; /* contains reserved blocks */ 291 292 /* Initialize related */ 293 boolean_t vdev_initialize_exit_wanted; 294 vdev_initializing_state_t vdev_initialize_state; 295 list_node_t vdev_initialize_node; 296 kthread_t *vdev_initialize_thread; 297 /* Protects vdev_initialize_thread and vdev_initialize_state. */ 298 kmutex_t vdev_initialize_lock; 299 kcondvar_t vdev_initialize_cv; 300 uint64_t vdev_initialize_offset[TXG_SIZE]; 301 uint64_t vdev_initialize_last_offset; 302 range_tree_t *vdev_initialize_tree; /* valid while initializing */ 303 uint64_t vdev_initialize_bytes_est; 304 uint64_t vdev_initialize_bytes_done; 305 uint64_t vdev_initialize_action_time; /* start and end time */ 306 307 /* TRIM related */ 308 boolean_t vdev_trim_exit_wanted; 309 boolean_t vdev_autotrim_exit_wanted; 310 vdev_trim_state_t vdev_trim_state; 311 list_node_t vdev_trim_node; 312 kmutex_t vdev_autotrim_lock; 313 kcondvar_t vdev_autotrim_cv; 314 kcondvar_t vdev_autotrim_kick_cv; 315 kthread_t *vdev_autotrim_thread; 316 /* Protects vdev_trim_thread and vdev_trim_state. */ 317 kmutex_t vdev_trim_lock; 318 kcondvar_t vdev_trim_cv; 319 kthread_t *vdev_trim_thread; 320 uint64_t vdev_trim_offset[TXG_SIZE]; 321 uint64_t vdev_trim_last_offset; 322 uint64_t vdev_trim_bytes_est; 323 uint64_t vdev_trim_bytes_done; 324 uint64_t vdev_trim_rate; /* requested rate (bytes/sec) */ 325 uint64_t vdev_trim_partial; /* requested partial TRIM */ 326 uint64_t vdev_trim_secure; /* requested secure TRIM */ 327 uint64_t vdev_trim_action_time; /* start and end time */ 328 329 /* Rebuild related */ 330 boolean_t vdev_rebuilding; 331 boolean_t vdev_rebuild_exit_wanted; 332 boolean_t vdev_rebuild_cancel_wanted; 333 boolean_t vdev_rebuild_reset_wanted; 334 kmutex_t vdev_rebuild_lock; 335 kcondvar_t vdev_rebuild_cv; 336 kthread_t *vdev_rebuild_thread; 337 vdev_rebuild_t vdev_rebuild_config; 338 339 /* For limiting outstanding I/Os (initialize, TRIM) */ 340 kmutex_t vdev_initialize_io_lock; 341 kcondvar_t vdev_initialize_io_cv; 342 uint64_t vdev_initialize_inflight; 343 kmutex_t vdev_trim_io_lock; 344 kcondvar_t vdev_trim_io_cv; 345 uint64_t vdev_trim_inflight[3]; 346 347 /* 348 * Values stored in the config for an indirect or removing vdev. 349 */ 350 vdev_indirect_config_t vdev_indirect_config; 351 352 /* 353 * The vdev_indirect_rwlock protects the vdev_indirect_mapping 354 * pointer from changing on indirect vdevs (when it is condensed). 355 * Note that removing (not yet indirect) vdevs have different 356 * access patterns (the mapping is not accessed from open context, 357 * e.g. from zio_read) and locking strategy (e.g. svr_lock). 358 */ 359 krwlock_t vdev_indirect_rwlock; 360 vdev_indirect_mapping_t *vdev_indirect_mapping; 361 vdev_indirect_births_t *vdev_indirect_births; 362 363 /* 364 * In memory data structures used to manage the obsolete sm, for 365 * indirect or removing vdevs. 366 * 367 * The vdev_obsolete_segments is the in-core record of the segments 368 * that are no longer referenced anywhere in the pool (due to 369 * being freed or remapped and not referenced by any snapshots). 370 * During a sync, segments are added to vdev_obsolete_segments 371 * via vdev_indirect_mark_obsolete(); at the end of each sync 372 * pass, this is appended to vdev_obsolete_sm via 373 * vdev_indirect_sync_obsolete(). The vdev_obsolete_lock 374 * protects against concurrent modifications of vdev_obsolete_segments 375 * from multiple zio threads. 376 */ 377 kmutex_t vdev_obsolete_lock; 378 range_tree_t *vdev_obsolete_segments; 379 space_map_t *vdev_obsolete_sm; 380 381 /* 382 * Protects the vdev_scan_io_queue field itself as well as the 383 * structure's contents (when present). 384 */ 385 kmutex_t vdev_scan_io_queue_lock; 386 struct dsl_scan_io_queue *vdev_scan_io_queue; 387 388 /* 389 * Leaf vdev state. 390 */ 391 range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */ 392 space_map_t *vdev_dtl_sm; /* dirty time log space map */ 393 txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */ 394 uint64_t vdev_dtl_object; /* DTL object */ 395 uint64_t vdev_psize; /* physical device capacity */ 396 uint64_t vdev_wholedisk; /* true if this is a whole disk */ 397 uint64_t vdev_offline; /* persistent offline state */ 398 uint64_t vdev_faulted; /* persistent faulted state */ 399 uint64_t vdev_degraded; /* persistent degraded state */ 400 uint64_t vdev_removed; /* persistent removed state */ 401 uint64_t vdev_resilver_txg; /* persistent resilvering state */ 402 uint64_t vdev_rebuild_txg; /* persistent rebuilding state */ 403 char *vdev_path; /* vdev path (if any) */ 404 char *vdev_devid; /* vdev devid (if any) */ 405 char *vdev_physpath; /* vdev device path (if any) */ 406 char *vdev_enc_sysfs_path; /* enclosure sysfs path */ 407 char *vdev_fru; /* physical FRU location */ 408 uint64_t vdev_not_present; /* not present during import */ 409 uint64_t vdev_unspare; /* unspare when resilvering done */ 410 boolean_t vdev_nowritecache; /* true if flushwritecache failed */ 411 boolean_t vdev_has_trim; /* TRIM is supported */ 412 boolean_t vdev_has_securetrim; /* secure TRIM is supported */ 413 boolean_t vdev_checkremove; /* temporary online test */ 414 boolean_t vdev_forcefault; /* force online fault */ 415 boolean_t vdev_splitting; /* split or repair in progress */ 416 boolean_t vdev_delayed_close; /* delayed device close? */ 417 boolean_t vdev_tmpoffline; /* device taken offline temporarily? */ 418 boolean_t vdev_detached; /* device detached? */ 419 boolean_t vdev_cant_read; /* vdev is failing all reads */ 420 boolean_t vdev_cant_write; /* vdev is failing all writes */ 421 boolean_t vdev_isspare; /* was a hot spare */ 422 boolean_t vdev_isl2cache; /* was a l2cache device */ 423 boolean_t vdev_copy_uberblocks; /* post expand copy uberblocks */ 424 boolean_t vdev_resilver_deferred; /* resilver deferred */ 425 boolean_t vdev_kobj_flag; /* kobj event record */ 426 boolean_t vdev_attaching; /* vdev attach ashift handling */ 427 vdev_queue_t vdev_queue; /* I/O deadline schedule queue */ 428 spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */ 429 zio_t *vdev_probe_zio; /* root of current probe */ 430 vdev_aux_t vdev_label_aux; /* on-disk aux state */ 431 uint64_t vdev_leaf_zap; 432 hrtime_t vdev_mmp_pending; /* 0 if write finished */ 433 uint64_t vdev_mmp_kstat_id; /* to find kstat entry */ 434 uint64_t vdev_expansion_time; /* vdev's last expansion time */ 435 list_node_t vdev_leaf_node; /* leaf vdev list */ 436 437 /* 438 * For DTrace to work in userland (libzpool) context, these fields must 439 * remain at the end of the structure. DTrace will use the kernel's 440 * CTF definition for 'struct vdev', and since the size of a kmutex_t is 441 * larger in userland, the offsets for the rest of the fields would be 442 * incorrect. 443 */ 444 kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */ 445 kmutex_t vdev_stat_lock; /* vdev_stat */ 446 kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */ 447 448 /* 449 * We rate limit ZIO delay, deadman, and checksum events, since they 450 * can flood ZED with tons of events when a drive is acting up. 451 * 452 * We also rate limit Direct I/O write verify errors, since a user might 453 * be continually manipulating a buffer that can flood ZED with tons of 454 * events. 455 */ 456 zfs_ratelimit_t vdev_delay_rl; 457 zfs_ratelimit_t vdev_deadman_rl; 458 zfs_ratelimit_t vdev_dio_verify_rl; 459 zfs_ratelimit_t vdev_checksum_rl; 460 461 /* 462 * Vdev properties for tuning ZED or zfsd 463 */ 464 uint64_t vdev_checksum_n; 465 uint64_t vdev_checksum_t; 466 uint64_t vdev_io_n; 467 uint64_t vdev_io_t; 468 uint64_t vdev_slow_io_n; 469 uint64_t vdev_slow_io_t; 470 }; 471 472 #define VDEV_PAD_SIZE (8 << 10) 473 /* 2 padding areas (vl_pad1 and vl_be) to skip */ 474 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 475 #define VDEV_PHYS_SIZE (112 << 10) 476 #define VDEV_UBERBLOCK_RING (128 << 10) 477 478 /* 479 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock 480 * ring when MMP is enabled. 481 */ 482 #define MMP_BLOCKS_PER_LABEL 1 483 484 /* The largest uberblock we support is 8k. */ 485 #define MAX_UBERBLOCK_SHIFT (13) 486 #define VDEV_UBERBLOCK_SHIFT(vd) \ 487 MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \ 488 MAX_UBERBLOCK_SHIFT) 489 #define VDEV_UBERBLOCK_COUNT(vd) \ 490 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 491 #define VDEV_UBERBLOCK_OFFSET(vd, n) \ 492 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 493 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 494 495 typedef struct vdev_phys { 496 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 497 zio_eck_t vp_zbt; 498 } vdev_phys_t; 499 500 typedef enum vbe_vers { 501 /* 502 * The bootenv file is stored as ascii text in the envblock. 503 * It is used by the GRUB bootloader used on Linux to store the 504 * contents of the grubenv file. The file is stored as raw ASCII, 505 * and is protected by an embedded checksum. By default, GRUB will 506 * check if the boot filesystem supports storing the environment data 507 * in a special location, and if so, will invoke filesystem specific 508 * logic to retrieve it. This can be overridden by a variable, should 509 * the user so desire. 510 */ 511 VB_RAW = 0, 512 513 /* 514 * The bootenv file is converted to an nvlist and then packed into the 515 * envblock. 516 */ 517 VB_NVLIST = 1 518 } vbe_vers_t; 519 520 typedef struct vdev_boot_envblock { 521 uint64_t vbe_version; 522 char vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) - 523 sizeof (zio_eck_t)]; 524 zio_eck_t vbe_zbt; 525 } vdev_boot_envblock_t; 526 _Static_assert(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE, 527 "vdev_boot_envblock_t wrong size"); 528 529 typedef struct vdev_label { 530 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 531 vdev_boot_envblock_t vl_be; /* 8K */ 532 vdev_phys_t vl_vdev_phys; /* 112K */ 533 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 534 } vdev_label_t; /* 256K total */ 535 536 /* 537 * vdev_dirty() flags 538 */ 539 #define VDD_METASLAB 0x01 540 #define VDD_DTL 0x02 541 542 /* Offset of embedded boot loader region on each label */ 543 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 544 /* 545 * Size of embedded boot loader region on each label. 546 * The total size of the first two labels plus the boot area is 4MB. 547 * On RAIDZ, this space is overwritten during RAIDZ expansion. 548 */ 549 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 550 551 /* 552 * Size of label regions at the start and end of each leaf device. 553 */ 554 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 555 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 556 #define VDEV_LABELS 4 557 #define VDEV_BEST_LABEL VDEV_LABELS 558 #define VDEV_OFFSET_IS_LABEL(vd, off) \ 559 (((off) < VDEV_LABEL_START_SIZE) || \ 560 ((off) >= ((vd)->vdev_psize - VDEV_LABEL_END_SIZE))) 561 562 #define VDEV_ALLOC_LOAD 0 563 #define VDEV_ALLOC_ADD 1 564 #define VDEV_ALLOC_SPARE 2 565 #define VDEV_ALLOC_L2CACHE 3 566 #define VDEV_ALLOC_ROOTPOOL 4 567 #define VDEV_ALLOC_SPLIT 5 568 #define VDEV_ALLOC_ATTACH 6 569 570 /* 571 * Allocate or free a vdev 572 */ 573 extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, 574 vdev_ops_t *ops); 575 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config, 576 vdev_t *parent, uint_t id, int alloctype); 577 extern void vdev_free(vdev_t *vd); 578 579 /* 580 * Add or remove children and parents 581 */ 582 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd); 583 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd); 584 extern void vdev_compact_children(vdev_t *pvd); 585 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops); 586 extern void vdev_remove_parent(vdev_t *cvd); 587 588 /* 589 * vdev sync load and sync 590 */ 591 extern boolean_t vdev_log_state_valid(vdev_t *vd); 592 extern int vdev_load(vdev_t *vd); 593 extern int vdev_dtl_load(vdev_t *vd); 594 extern void vdev_sync(vdev_t *vd, uint64_t txg); 595 extern void vdev_sync_done(vdev_t *vd, uint64_t txg); 596 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg); 597 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg); 598 599 /* 600 * Available vdev types. 601 */ 602 extern vdev_ops_t vdev_root_ops; 603 extern vdev_ops_t vdev_mirror_ops; 604 extern vdev_ops_t vdev_replacing_ops; 605 extern vdev_ops_t vdev_raidz_ops; 606 extern vdev_ops_t vdev_draid_ops; 607 extern vdev_ops_t vdev_draid_spare_ops; 608 extern vdev_ops_t vdev_disk_ops; 609 extern vdev_ops_t vdev_file_ops; 610 extern vdev_ops_t vdev_missing_ops; 611 extern vdev_ops_t vdev_hole_ops; 612 extern vdev_ops_t vdev_spare_ops; 613 extern vdev_ops_t vdev_indirect_ops; 614 615 /* 616 * Common size functions 617 */ 618 extern void vdev_default_xlate(vdev_t *vd, const range_seg64_t *logical_rs, 619 range_seg64_t *physical_rs, range_seg64_t *remain_rs); 620 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize, uint64_t txg); 621 extern uint64_t vdev_default_min_asize(vdev_t *vd); 622 extern uint64_t vdev_get_min_asize(vdev_t *vd); 623 extern void vdev_set_min_asize(vdev_t *vd); 624 extern uint64_t vdev_get_min_alloc(vdev_t *vd); 625 extern uint64_t vdev_get_nparity(vdev_t *vd); 626 extern uint64_t vdev_get_ndisks(vdev_t *vd); 627 628 /* 629 * Global variables 630 */ 631 extern int zfs_vdev_standard_sm_blksz; 632 633 /* 634 * Functions from vdev_indirect.c 635 */ 636 extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx); 637 extern boolean_t vdev_indirect_should_condense(vdev_t *vd); 638 extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx); 639 extern int vdev_obsolete_sm_object(vdev_t *vd, uint64_t *sm_obj); 640 extern int vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise); 641 642 /* 643 * Other miscellaneous functions 644 */ 645 int vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj); 646 void vdev_metaslab_group_create(vdev_t *vd); 647 uint64_t vdev_best_ashift(uint64_t logical, uint64_t a, uint64_t b); 648 649 /* 650 * Vdev ashift optimization tunables 651 */ 652 extern uint_t zfs_vdev_min_auto_ashift; 653 extern uint_t zfs_vdev_max_auto_ashift; 654 int param_set_min_auto_ashift(ZFS_MODULE_PARAM_ARGS); 655 int param_set_max_auto_ashift(ZFS_MODULE_PARAM_ARGS); 656 657 /* 658 * VDEV checksum verification for Direct I/O writes 659 */ 660 extern uint_t zfs_vdev_direct_write_verify; 661 662 #ifdef __cplusplus 663 } 664 #endif 665 666 #endif /* _SYS_VDEV_IMPL_H */ 667