1 // SPDX-License-Identifier: CDDL-1.0 2 /* 3 * CDDL HEADER START 4 * 5 * The contents of this file are subject to the terms of the 6 * Common Development and Distribution License (the "License"). 7 * You may not use this file except in compliance with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or https://opensource.org/licenses/CDDL-1.0. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 /* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011, 2020 by Delphix. All rights reserved. 25 * Copyright (c) 2017, Intel Corporation. 26 * Copyright (c) 2023, Klara Inc. 27 */ 28 29 #ifndef _SYS_VDEV_IMPL_H 30 #define _SYS_VDEV_IMPL_H 31 32 #include <sys/avl.h> 33 #include <sys/bpobj.h> 34 #include <sys/dmu.h> 35 #include <sys/metaslab.h> 36 #include <sys/nvpair.h> 37 #include <sys/space_map.h> 38 #include <sys/vdev.h> 39 #include <sys/uberblock_impl.h> 40 #include <sys/vdev_indirect_mapping.h> 41 #include <sys/vdev_indirect_births.h> 42 #include <sys/vdev_rebuild.h> 43 #include <sys/vdev_removal.h> 44 #include <sys/zfs_ratelimit.h> 45 46 #ifdef __cplusplus 47 extern "C" { 48 #endif 49 50 /* 51 * Virtual device descriptors. 52 * 53 * All storage pool operations go through the virtual device framework, 54 * which provides data replication and I/O scheduling. 55 */ 56 57 /* 58 * Forward declarations that lots of things need. 59 */ 60 typedef struct vdev_queue vdev_queue_t; 61 struct abd; 62 63 /* 64 * Virtual device operations 65 */ 66 typedef int vdev_init_func_t(spa_t *spa, nvlist_t *nv, void **tsd); 67 typedef void vdev_kobj_post_evt_func_t(vdev_t *vd); 68 typedef void vdev_fini_func_t(vdev_t *vd); 69 typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size, 70 uint64_t *ashift, uint64_t *pshift); 71 typedef void vdev_close_func_t(vdev_t *vd); 72 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize, uint64_t txg); 73 typedef uint64_t vdev_min_asize_func_t(vdev_t *vd); 74 typedef uint64_t vdev_min_alloc_func_t(vdev_t *vd); 75 typedef void vdev_io_start_func_t(zio_t *zio); 76 typedef void vdev_io_done_func_t(zio_t *zio); 77 typedef void vdev_state_change_func_t(vdev_t *vd, int, int); 78 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, const dva_t *dva, 79 size_t psize, uint64_t phys_birth); 80 typedef void vdev_hold_func_t(vdev_t *vd); 81 typedef void vdev_rele_func_t(vdev_t *vd); 82 83 typedef void vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd, 84 uint64_t offset, uint64_t size, void *arg); 85 typedef void vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size, 86 vdev_remap_cb_t callback, void *arg); 87 /* 88 * Given a target vdev, translates the logical range "in" to the physical 89 * range "res" 90 */ 91 typedef void vdev_xlation_func_t(vdev_t *cvd, const zfs_range_seg64_t *logical, 92 zfs_range_seg64_t *physical, zfs_range_seg64_t *remain); 93 typedef uint64_t vdev_rebuild_asize_func_t(vdev_t *vd, uint64_t start, 94 uint64_t size, uint64_t max_segment); 95 typedef void vdev_metaslab_init_func_t(vdev_t *vd, uint64_t *startp, 96 uint64_t *sizep); 97 typedef void vdev_config_generate_func_t(vdev_t *vd, nvlist_t *nv); 98 typedef uint64_t vdev_nparity_func_t(vdev_t *vd); 99 typedef uint64_t vdev_ndisks_func_t(vdev_t *vd); 100 101 typedef const struct vdev_ops { 102 vdev_init_func_t *vdev_op_init; 103 vdev_fini_func_t *vdev_op_fini; 104 vdev_open_func_t *vdev_op_open; 105 vdev_close_func_t *vdev_op_close; 106 vdev_asize_func_t *vdev_op_psize_to_asize; 107 vdev_asize_func_t *vdev_op_asize_to_psize; 108 vdev_min_asize_func_t *vdev_op_min_asize; 109 vdev_min_alloc_func_t *vdev_op_min_alloc; 110 vdev_io_start_func_t *vdev_op_io_start; 111 vdev_io_done_func_t *vdev_op_io_done; 112 vdev_state_change_func_t *vdev_op_state_change; 113 vdev_need_resilver_func_t *vdev_op_need_resilver; 114 vdev_hold_func_t *vdev_op_hold; 115 vdev_rele_func_t *vdev_op_rele; 116 vdev_remap_func_t *vdev_op_remap; 117 vdev_xlation_func_t *vdev_op_xlate; 118 vdev_rebuild_asize_func_t *vdev_op_rebuild_asize; 119 vdev_metaslab_init_func_t *vdev_op_metaslab_init; 120 vdev_config_generate_func_t *vdev_op_config_generate; 121 vdev_nparity_func_t *vdev_op_nparity; 122 vdev_ndisks_func_t *vdev_op_ndisks; 123 vdev_kobj_post_evt_func_t *vdev_op_kobj_evt_post; 124 char vdev_op_type[16]; 125 boolean_t vdev_op_leaf; 126 } vdev_ops_t; 127 128 /* 129 * Virtual device properties 130 */ 131 typedef union vdev_queue_class { 132 struct { 133 ulong_t vqc_list_numnodes; 134 list_t vqc_list; 135 }; 136 avl_tree_t vqc_tree; 137 } vdev_queue_class_t; 138 139 struct vdev_queue { 140 vdev_t *vq_vdev; 141 vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE]; 142 avl_tree_t vq_read_offset_tree; 143 avl_tree_t vq_write_offset_tree; 144 uint64_t vq_last_offset; 145 zio_priority_t vq_last_prio; /* Last sent I/O priority. */ 146 uint32_t vq_cqueued; /* Classes with queued I/Os. */ 147 uint32_t vq_cactive[ZIO_PRIORITY_NUM_QUEUEABLE]; 148 uint32_t vq_active; /* Number of active I/Os. */ 149 uint32_t vq_ia_active; /* Active interactive I/Os. */ 150 uint32_t vq_nia_credit; /* Non-interactive I/Os credit. */ 151 list_t vq_active_list; /* List of active I/Os. */ 152 hrtime_t vq_io_complete_ts; /* time last i/o completed */ 153 hrtime_t vq_io_delta_ts; 154 zio_t vq_io_search; /* used as local for stack reduction */ 155 kmutex_t vq_lock; 156 }; 157 158 typedef enum vdev_alloc_bias { 159 VDEV_BIAS_NONE, 160 VDEV_BIAS_LOG, /* dedicated to ZIL data (SLOG) */ 161 VDEV_BIAS_SPECIAL, /* dedicated to ddt, metadata, and small blks */ 162 VDEV_BIAS_DEDUP /* dedicated to dedup metadata */ 163 } vdev_alloc_bias_t; 164 165 166 /* 167 * On-disk indirect vdev state. 168 * 169 * An indirect vdev is described exclusively in the MOS config of a pool. 170 * The config for an indirect vdev includes several fields, which are 171 * accessed in memory by a vdev_indirect_config_t. 172 */ 173 typedef struct vdev_indirect_config { 174 /* 175 * Object (in MOS) which contains the indirect mapping. This object 176 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by 177 * vimep_src. The bonus buffer for this object is a 178 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev 179 * removal is initiated. 180 * 181 * Note that this object can be empty if none of the data on the vdev 182 * has been copied yet. 183 */ 184 uint64_t vic_mapping_object; 185 186 /* 187 * Object (in MOS) which contains the birth times for the mapping 188 * entries. This object contains an array of 189 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus 190 * buffer for this object is a vdev_indirect_birth_phys_t. This object 191 * is allocated when a vdev removal is initiated. 192 * 193 * Note that this object can be empty if none of the vdev has yet been 194 * copied. 195 */ 196 uint64_t vic_births_object; 197 198 /* 199 * This is the vdev ID which was removed previous to this vdev, or 200 * UINT64_MAX if there are no previously removed vdevs. 201 */ 202 uint64_t vic_prev_indirect_vdev; 203 } vdev_indirect_config_t; 204 205 /* 206 * Virtual device descriptor 207 */ 208 struct vdev { 209 /* 210 * Common to all vdev types. 211 */ 212 uint64_t vdev_id; /* child number in vdev parent */ 213 uint64_t vdev_guid; /* unique ID for this vdev */ 214 uint64_t vdev_guid_sum; /* self guid + all child guids */ 215 uint64_t vdev_orig_guid; /* orig. guid prior to remove */ 216 uint64_t vdev_asize; /* allocatable device capacity */ 217 uint64_t vdev_min_asize; /* min acceptable asize */ 218 uint64_t vdev_max_asize; /* max acceptable asize */ 219 uint64_t vdev_ashift; /* block alignment shift */ 220 221 /* 222 * Logical block alignment shift 223 * 224 * The smallest sized/aligned I/O supported by the device. 225 */ 226 uint64_t vdev_logical_ashift; 227 /* 228 * Physical block alignment shift 229 * 230 * The device supports logical I/Os with vdev_logical_ashift 231 * size/alignment, but optimum performance will be achieved by 232 * aligning/sizing requests to vdev_physical_ashift. Smaller 233 * requests may be inflated or incur device level read-modify-write 234 * operations. 235 * 236 * May be 0 to indicate no preference (i.e. use vdev_logical_ashift). 237 */ 238 uint64_t vdev_physical_ashift; 239 uint64_t vdev_state; /* see VDEV_STATE_* #defines */ 240 uint64_t vdev_prevstate; /* used when reopening a vdev */ 241 vdev_ops_t *vdev_ops; /* vdev operations */ 242 spa_t *vdev_spa; /* spa for this vdev */ 243 void *vdev_tsd; /* type-specific data */ 244 vdev_t *vdev_top; /* top-level vdev */ 245 vdev_t *vdev_parent; /* parent vdev */ 246 vdev_t **vdev_child; /* array of children */ 247 uint64_t vdev_children; /* number of children */ 248 vdev_stat_t vdev_stat; /* virtual device statistics */ 249 vdev_stat_ex_t vdev_stat_ex; /* extended statistics */ 250 boolean_t vdev_expanding; /* expand the vdev? */ 251 boolean_t vdev_reopening; /* reopen in progress? */ 252 boolean_t vdev_nonrot; /* true if solid state */ 253 int vdev_load_error; /* error on last load */ 254 int vdev_open_error; /* error on last open */ 255 int vdev_validate_error; /* error on last validate */ 256 kthread_t *vdev_open_thread; /* thread opening children */ 257 kthread_t *vdev_validate_thread; /* thread validating children */ 258 uint64_t vdev_crtxg; /* txg when top-level was added */ 259 uint64_t vdev_root_zap; 260 261 /* 262 * Top-level vdev state. 263 */ 264 uint64_t vdev_ms_array; /* metaslab array object */ 265 uint64_t vdev_ms_shift; /* metaslab size shift */ 266 uint64_t vdev_ms_count; /* number of metaslabs */ 267 metaslab_group_t *vdev_mg; /* metaslab group */ 268 metaslab_group_t *vdev_log_mg; /* embedded slog metaslab group */ 269 metaslab_t **vdev_ms; /* metaslab array */ 270 txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */ 271 txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */ 272 txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */ 273 boolean_t vdev_remove_wanted; /* async remove wanted? */ 274 boolean_t vdev_fault_wanted; /* async faulted wanted? */ 275 list_node_t vdev_config_dirty_node; /* config dirty list */ 276 list_node_t vdev_state_dirty_node; /* state dirty list */ 277 uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */ 278 uint64_t vdev_islog; /* is an intent log device */ 279 uint64_t vdev_noalloc; /* device is passivated? */ 280 uint64_t vdev_removing; /* device is being removed? */ 281 uint64_t vdev_failfast; /* device failfast setting */ 282 boolean_t vdev_rz_expanding; /* raidz is being expanded? */ 283 boolean_t vdev_ishole; /* is a hole in the namespace */ 284 uint64_t vdev_top_zap; 285 vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias */ 286 287 /* pool checkpoint related */ 288 space_map_t *vdev_checkpoint_sm; /* contains reserved blocks */ 289 290 /* Initialize related */ 291 boolean_t vdev_initialize_exit_wanted; 292 vdev_initializing_state_t vdev_initialize_state; 293 list_node_t vdev_initialize_node; 294 kthread_t *vdev_initialize_thread; 295 /* Protects vdev_initialize_thread and vdev_initialize_state. */ 296 kmutex_t vdev_initialize_lock; 297 kcondvar_t vdev_initialize_cv; 298 uint64_t vdev_initialize_offset[TXG_SIZE]; 299 uint64_t vdev_initialize_last_offset; 300 /* valid while initializing */ 301 zfs_range_tree_t *vdev_initialize_tree; 302 uint64_t vdev_initialize_bytes_est; 303 uint64_t vdev_initialize_bytes_done; 304 uint64_t vdev_initialize_action_time; /* start and end time */ 305 306 /* TRIM related */ 307 boolean_t vdev_trim_exit_wanted; 308 boolean_t vdev_autotrim_exit_wanted; 309 vdev_trim_state_t vdev_trim_state; 310 list_node_t vdev_trim_node; 311 kmutex_t vdev_autotrim_lock; 312 kcondvar_t vdev_autotrim_cv; 313 kcondvar_t vdev_autotrim_kick_cv; 314 kthread_t *vdev_autotrim_thread; 315 /* Protects vdev_trim_thread and vdev_trim_state. */ 316 kmutex_t vdev_trim_lock; 317 kcondvar_t vdev_trim_cv; 318 kthread_t *vdev_trim_thread; 319 uint64_t vdev_trim_offset[TXG_SIZE]; 320 uint64_t vdev_trim_last_offset; 321 uint64_t vdev_trim_bytes_est; 322 uint64_t vdev_trim_bytes_done; 323 uint64_t vdev_trim_rate; /* requested rate (bytes/sec) */ 324 uint64_t vdev_trim_partial; /* requested partial TRIM */ 325 uint64_t vdev_trim_secure; /* requested secure TRIM */ 326 uint64_t vdev_trim_action_time; /* start and end time */ 327 328 /* Rebuild related */ 329 boolean_t vdev_rebuilding; 330 boolean_t vdev_rebuild_exit_wanted; 331 boolean_t vdev_rebuild_cancel_wanted; 332 boolean_t vdev_rebuild_reset_wanted; 333 kmutex_t vdev_rebuild_lock; 334 kcondvar_t vdev_rebuild_cv; 335 kthread_t *vdev_rebuild_thread; 336 vdev_rebuild_t vdev_rebuild_config; 337 338 /* For limiting outstanding I/Os (initialize, TRIM) */ 339 kmutex_t vdev_initialize_io_lock; 340 kcondvar_t vdev_initialize_io_cv; 341 uint64_t vdev_initialize_inflight; 342 kmutex_t vdev_trim_io_lock; 343 kcondvar_t vdev_trim_io_cv; 344 uint64_t vdev_trim_inflight[3]; 345 346 /* 347 * Values stored in the config for an indirect or removing vdev. 348 */ 349 vdev_indirect_config_t vdev_indirect_config; 350 351 /* 352 * The vdev_indirect_rwlock protects the vdev_indirect_mapping 353 * pointer from changing on indirect vdevs (when it is condensed). 354 * Note that removing (not yet indirect) vdevs have different 355 * access patterns (the mapping is not accessed from open context, 356 * e.g. from zio_read) and locking strategy (e.g. svr_lock). 357 */ 358 krwlock_t vdev_indirect_rwlock; 359 vdev_indirect_mapping_t *vdev_indirect_mapping; 360 vdev_indirect_births_t *vdev_indirect_births; 361 362 /* 363 * In memory data structures used to manage the obsolete sm, for 364 * indirect or removing vdevs. 365 * 366 * The vdev_obsolete_segments is the in-core record of the segments 367 * that are no longer referenced anywhere in the pool (due to 368 * being freed or remapped and not referenced by any snapshots). 369 * During a sync, segments are added to vdev_obsolete_segments 370 * via vdev_indirect_mark_obsolete(); at the end of each sync 371 * pass, this is appended to vdev_obsolete_sm via 372 * vdev_indirect_sync_obsolete(). The vdev_obsolete_lock 373 * protects against concurrent modifications of vdev_obsolete_segments 374 * from multiple zio threads. 375 */ 376 kmutex_t vdev_obsolete_lock; 377 zfs_range_tree_t *vdev_obsolete_segments; 378 space_map_t *vdev_obsolete_sm; 379 380 /* 381 * Protects the vdev_scan_io_queue field itself as well as the 382 * structure's contents (when present). 383 */ 384 kmutex_t vdev_scan_io_queue_lock; 385 struct dsl_scan_io_queue *vdev_scan_io_queue; 386 387 /* 388 * Leaf vdev state. 389 */ 390 zfs_range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */ 391 space_map_t *vdev_dtl_sm; /* dirty time log space map */ 392 txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */ 393 uint64_t vdev_dtl_object; /* DTL object */ 394 uint64_t vdev_psize; /* physical device capacity */ 395 uint64_t vdev_wholedisk; /* true if this is a whole disk */ 396 uint64_t vdev_offline; /* persistent offline state */ 397 uint64_t vdev_faulted; /* persistent faulted state */ 398 uint64_t vdev_degraded; /* persistent degraded state */ 399 uint64_t vdev_removed; /* persistent removed state */ 400 uint64_t vdev_resilver_txg; /* persistent resilvering state */ 401 uint64_t vdev_rebuild_txg; /* persistent rebuilding state */ 402 char *vdev_path; /* vdev path (if any) */ 403 char *vdev_devid; /* vdev devid (if any) */ 404 char *vdev_physpath; /* vdev device path (if any) */ 405 char *vdev_enc_sysfs_path; /* enclosure sysfs path */ 406 char *vdev_fru; /* physical FRU location */ 407 uint64_t vdev_not_present; /* not present during import */ 408 uint64_t vdev_unspare; /* unspare when resilvering done */ 409 boolean_t vdev_nowritecache; /* true if flushwritecache failed */ 410 boolean_t vdev_has_trim; /* TRIM is supported */ 411 boolean_t vdev_has_securetrim; /* secure TRIM is supported */ 412 boolean_t vdev_checkremove; /* temporary online test */ 413 boolean_t vdev_forcefault; /* force online fault */ 414 boolean_t vdev_splitting; /* split or repair in progress */ 415 boolean_t vdev_delayed_close; /* delayed device close? */ 416 boolean_t vdev_tmpoffline; /* device taken offline temporarily? */ 417 boolean_t vdev_detached; /* device detached? */ 418 boolean_t vdev_cant_read; /* vdev is failing all reads */ 419 boolean_t vdev_cant_write; /* vdev is failing all writes */ 420 boolean_t vdev_isspare; /* was a hot spare */ 421 boolean_t vdev_isl2cache; /* was a l2cache device */ 422 boolean_t vdev_copy_uberblocks; /* post expand copy uberblocks */ 423 boolean_t vdev_resilver_deferred; /* resilver deferred */ 424 boolean_t vdev_kobj_flag; /* kobj event record */ 425 boolean_t vdev_attaching; /* vdev attach ashift handling */ 426 vdev_queue_t vdev_queue; /* I/O deadline schedule queue */ 427 spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */ 428 zio_t *vdev_probe_zio; /* root of current probe */ 429 vdev_aux_t vdev_label_aux; /* on-disk aux state */ 430 uint64_t vdev_leaf_zap; 431 hrtime_t vdev_mmp_pending; /* 0 if write finished */ 432 uint64_t vdev_mmp_kstat_id; /* to find kstat entry */ 433 uint64_t vdev_expansion_time; /* vdev's last expansion time */ 434 list_node_t vdev_leaf_node; /* leaf vdev list */ 435 436 /* 437 * For DTrace to work in userland (libzpool) context, these fields must 438 * remain at the end of the structure. DTrace will use the kernel's 439 * CTF definition for 'struct vdev', and since the size of a kmutex_t is 440 * larger in userland, the offsets for the rest of the fields would be 441 * incorrect. 442 */ 443 kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */ 444 kmutex_t vdev_stat_lock; /* vdev_stat */ 445 kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */ 446 447 /* 448 * We rate limit ZIO delay, deadman, and checksum events, since they 449 * can flood ZED with tons of events when a drive is acting up. 450 * 451 * We also rate limit Direct I/O write verify errors, since a user might 452 * be continually manipulating a buffer that can flood ZED with tons of 453 * events. 454 */ 455 zfs_ratelimit_t vdev_delay_rl; 456 zfs_ratelimit_t vdev_deadman_rl; 457 zfs_ratelimit_t vdev_dio_verify_rl; 458 zfs_ratelimit_t vdev_checksum_rl; 459 460 /* 461 * Vdev properties for tuning ZED or zfsd 462 */ 463 uint64_t vdev_checksum_n; 464 uint64_t vdev_checksum_t; 465 uint64_t vdev_io_n; 466 uint64_t vdev_io_t; 467 uint64_t vdev_slow_io_n; 468 uint64_t vdev_slow_io_t; 469 }; 470 471 #define VDEV_PAD_SIZE (8 << 10) 472 /* 2 padding areas (vl_pad1 and vl_be) to skip */ 473 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 474 #define VDEV_PHYS_SIZE (112 << 10) 475 #define VDEV_UBERBLOCK_RING (128 << 10) 476 477 /* 478 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock 479 * ring when MMP is enabled. 480 */ 481 #define MMP_BLOCKS_PER_LABEL 1 482 483 /* The largest uberblock we support is 8k. */ 484 #define MAX_UBERBLOCK_SHIFT (13) 485 #define VDEV_UBERBLOCK_SHIFT(vd) \ 486 MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \ 487 MAX_UBERBLOCK_SHIFT) 488 #define VDEV_UBERBLOCK_COUNT(vd) \ 489 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 490 #define VDEV_UBERBLOCK_OFFSET(vd, n) \ 491 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 492 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 493 494 typedef struct vdev_phys { 495 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 496 zio_eck_t vp_zbt; 497 } vdev_phys_t; 498 499 typedef enum vbe_vers { 500 /* 501 * The bootenv file is stored as ascii text in the envblock. 502 * It is used by the GRUB bootloader used on Linux to store the 503 * contents of the grubenv file. The file is stored as raw ASCII, 504 * and is protected by an embedded checksum. By default, GRUB will 505 * check if the boot filesystem supports storing the environment data 506 * in a special location, and if so, will invoke filesystem specific 507 * logic to retrieve it. This can be overridden by a variable, should 508 * the user so desire. 509 */ 510 VB_RAW = 0, 511 512 /* 513 * The bootenv file is converted to an nvlist and then packed into the 514 * envblock. 515 */ 516 VB_NVLIST = 1 517 } vbe_vers_t; 518 519 typedef struct vdev_boot_envblock { 520 uint64_t vbe_version; 521 char vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) - 522 sizeof (zio_eck_t)]; 523 zio_eck_t vbe_zbt; 524 } vdev_boot_envblock_t; 525 _Static_assert(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE, 526 "vdev_boot_envblock_t wrong size"); 527 528 typedef struct vdev_label { 529 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 530 vdev_boot_envblock_t vl_be; /* 8K */ 531 vdev_phys_t vl_vdev_phys; /* 112K */ 532 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 533 } vdev_label_t; /* 256K total */ 534 535 /* 536 * vdev_dirty() flags 537 */ 538 #define VDD_METASLAB 0x01 539 #define VDD_DTL 0x02 540 541 /* Offset of embedded boot loader region on each label */ 542 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 543 /* 544 * Size of embedded boot loader region on each label. 545 * The total size of the first two labels plus the boot area is 4MB. 546 * On RAIDZ, this space is overwritten during RAIDZ expansion. 547 */ 548 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 549 550 /* 551 * Size of label regions at the start and end of each leaf device. 552 */ 553 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 554 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 555 #define VDEV_LABELS 4 556 #define VDEV_BEST_LABEL VDEV_LABELS 557 #define VDEV_OFFSET_IS_LABEL(vd, off) \ 558 (((off) < VDEV_LABEL_START_SIZE) || \ 559 ((off) >= ((vd)->vdev_psize - VDEV_LABEL_END_SIZE))) 560 561 #define VDEV_ALLOC_LOAD 0 562 #define VDEV_ALLOC_ADD 1 563 #define VDEV_ALLOC_SPARE 2 564 #define VDEV_ALLOC_L2CACHE 3 565 #define VDEV_ALLOC_ROOTPOOL 4 566 #define VDEV_ALLOC_SPLIT 5 567 #define VDEV_ALLOC_ATTACH 6 568 569 /* 570 * Allocate or free a vdev 571 */ 572 extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, 573 vdev_ops_t *ops); 574 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config, 575 vdev_t *parent, uint_t id, int alloctype); 576 extern void vdev_free(vdev_t *vd); 577 578 /* 579 * Add or remove children and parents 580 */ 581 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd); 582 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd); 583 extern void vdev_compact_children(vdev_t *pvd); 584 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops); 585 extern void vdev_remove_parent(vdev_t *cvd); 586 587 /* 588 * vdev sync load and sync 589 */ 590 extern boolean_t vdev_log_state_valid(vdev_t *vd); 591 extern int vdev_load(vdev_t *vd); 592 extern int vdev_dtl_load(vdev_t *vd); 593 extern void vdev_sync(vdev_t *vd, uint64_t txg); 594 extern void vdev_sync_done(vdev_t *vd, uint64_t txg); 595 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg); 596 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg); 597 598 /* 599 * Available vdev types. 600 */ 601 extern vdev_ops_t vdev_root_ops; 602 extern vdev_ops_t vdev_mirror_ops; 603 extern vdev_ops_t vdev_replacing_ops; 604 extern vdev_ops_t vdev_raidz_ops; 605 extern vdev_ops_t vdev_draid_ops; 606 extern vdev_ops_t vdev_draid_spare_ops; 607 extern vdev_ops_t vdev_disk_ops; 608 extern vdev_ops_t vdev_file_ops; 609 extern vdev_ops_t vdev_missing_ops; 610 extern vdev_ops_t vdev_hole_ops; 611 extern vdev_ops_t vdev_spare_ops; 612 extern vdev_ops_t vdev_indirect_ops; 613 614 /* 615 * Common size functions 616 */ 617 extern void vdev_default_xlate(vdev_t *vd, const zfs_range_seg64_t *logical_rs, 618 zfs_range_seg64_t *physical_rs, zfs_range_seg64_t *remain_rs); 619 extern uint64_t vdev_default_psize(vdev_t *vd, uint64_t asize, uint64_t txg); 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 #if defined(__linux__) 649 int param_get_raidz_impl(char *buf, zfs_kernel_param_t *kp); 650 #endif 651 int param_set_raidz_impl(ZFS_MODULE_PARAM_ARGS); 652 653 /* 654 * Vdev ashift optimization tunables 655 */ 656 extern uint_t zfs_vdev_min_auto_ashift; 657 extern uint_t zfs_vdev_max_auto_ashift; 658 int param_set_min_auto_ashift(ZFS_MODULE_PARAM_ARGS); 659 int param_set_max_auto_ashift(ZFS_MODULE_PARAM_ARGS); 660 661 /* 662 * VDEV checksum verification for Direct I/O writes 663 */ 664 extern uint_t zfs_vdev_direct_write_verify; 665 666 #ifdef __cplusplus 667 } 668 #endif 669 670 #endif /* _SYS_VDEV_IMPL_H */ 671