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 /* 159 * On-disk indirect vdev state. 160 * 161 * An indirect vdev is described exclusively in the MOS config of a pool. 162 * The config for an indirect vdev includes several fields, which are 163 * accessed in memory by a vdev_indirect_config_t. 164 */ 165 typedef struct vdev_indirect_config { 166 /* 167 * Object (in MOS) which contains the indirect mapping. This object 168 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by 169 * vimep_src. The bonus buffer for this object is a 170 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev 171 * removal is initiated. 172 * 173 * Note that this object can be empty if none of the data on the vdev 174 * has been copied yet. 175 */ 176 uint64_t vic_mapping_object; 177 178 /* 179 * Object (in MOS) which contains the birth times for the mapping 180 * entries. This object contains an array of 181 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus 182 * buffer for this object is a vdev_indirect_birth_phys_t. This object 183 * is allocated when a vdev removal is initiated. 184 * 185 * Note that this object can be empty if none of the vdev has yet been 186 * copied. 187 */ 188 uint64_t vic_births_object; 189 190 /* 191 * This is the vdev ID which was removed previous to this vdev, or 192 * UINT64_MAX if there are no previously removed vdevs. 193 */ 194 uint64_t vic_prev_indirect_vdev; 195 } vdev_indirect_config_t; 196 197 /* 198 * Virtual device descriptor 199 */ 200 struct vdev { 201 /* 202 * Common to all vdev types. 203 */ 204 uint64_t vdev_id; /* child number in vdev parent */ 205 uint64_t vdev_guid; /* unique ID for this vdev */ 206 uint64_t vdev_guid_sum; /* self guid + all child guids */ 207 uint64_t vdev_orig_guid; /* orig. guid prior to remove */ 208 uint64_t vdev_asize; /* allocatable device capacity */ 209 uint64_t vdev_min_asize; /* min acceptable asize */ 210 uint64_t vdev_max_asize; /* max acceptable asize */ 211 uint64_t vdev_ashift; /* block alignment shift */ 212 213 /* 214 * Logical block alignment shift 215 * 216 * The smallest sized/aligned I/O supported by the device. 217 */ 218 uint64_t vdev_logical_ashift; 219 /* 220 * Physical block alignment shift 221 * 222 * The device supports logical I/Os with vdev_logical_ashift 223 * size/alignment, but optimum performance will be achieved by 224 * aligning/sizing requests to vdev_physical_ashift. Smaller 225 * requests may be inflated or incur device level read-modify-write 226 * operations. 227 * 228 * May be 0 to indicate no preference (i.e. use vdev_logical_ashift). 229 */ 230 uint64_t vdev_physical_ashift; 231 uint64_t vdev_state; /* see VDEV_STATE_* #defines */ 232 uint64_t vdev_prevstate; /* used when reopening a vdev */ 233 vdev_ops_t *vdev_ops; /* vdev operations */ 234 spa_t *vdev_spa; /* spa for this vdev */ 235 void *vdev_tsd; /* type-specific data */ 236 vdev_t *vdev_top; /* top-level vdev */ 237 vdev_t *vdev_parent; /* parent vdev */ 238 vdev_t **vdev_child; /* array of children */ 239 uint64_t vdev_children; /* number of children */ 240 vdev_stat_t vdev_stat; /* virtual device statistics */ 241 vdev_stat_ex_t vdev_stat_ex; /* extended statistics */ 242 boolean_t vdev_expanding; /* expand the vdev? */ 243 boolean_t vdev_reopening; /* reopen in progress? */ 244 boolean_t vdev_nonrot; /* true if solid state */ 245 int vdev_load_error; /* error on last load */ 246 int vdev_open_error; /* error on last open */ 247 int vdev_validate_error; /* error on last validate */ 248 kthread_t *vdev_open_thread; /* thread opening children */ 249 kthread_t *vdev_validate_thread; /* thread validating children */ 250 uint64_t vdev_crtxg; /* txg when top-level was added */ 251 uint64_t vdev_root_zap; 252 253 /* 254 * Top-level vdev state. 255 */ 256 uint64_t vdev_ms_array; /* metaslab array object */ 257 uint64_t vdev_ms_shift; /* metaslab size shift */ 258 uint64_t vdev_ms_count; /* number of metaslabs */ 259 metaslab_group_t *vdev_mg; /* metaslab group */ 260 metaslab_group_t *vdev_log_mg; /* embedded slog metaslab group */ 261 metaslab_t **vdev_ms; /* metaslab array */ 262 txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */ 263 txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */ 264 txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */ 265 boolean_t vdev_remove_wanted; /* async remove wanted? */ 266 boolean_t vdev_fault_wanted; /* async faulted wanted? */ 267 list_node_t vdev_config_dirty_node; /* config dirty list */ 268 list_node_t vdev_state_dirty_node; /* state dirty list */ 269 uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */ 270 uint64_t vdev_islog; /* is an intent log device */ 271 uint64_t vdev_noalloc; /* device is passivated? */ 272 uint64_t vdev_removing; /* device is being removed? */ 273 uint64_t vdev_failfast; /* device failfast setting */ 274 boolean_t vdev_autosit; /* automatic sitout management */ 275 boolean_t vdev_rz_expanding; /* raidz is being expanded? */ 276 boolean_t vdev_ishole; /* is a hole in the namespace */ 277 uint64_t vdev_top_zap; 278 vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias */ 279 uint64_t vdev_last_latency_check; 280 281 /* pool checkpoint related */ 282 space_map_t *vdev_checkpoint_sm; /* contains reserved blocks */ 283 284 /* Initialize related */ 285 boolean_t vdev_initialize_exit_wanted; 286 vdev_initializing_state_t vdev_initialize_state; 287 list_node_t vdev_initialize_node; 288 kthread_t *vdev_initialize_thread; 289 /* Protects vdev_initialize_thread and vdev_initialize_state. */ 290 kmutex_t vdev_initialize_lock; 291 kcondvar_t vdev_initialize_cv; 292 uint64_t vdev_initialize_offset[TXG_SIZE]; 293 uint64_t vdev_initialize_last_offset; 294 /* valid while initializing */ 295 zfs_range_tree_t *vdev_initialize_tree; 296 uint64_t vdev_initialize_bytes_est; 297 uint64_t vdev_initialize_bytes_done; 298 uint64_t vdev_initialize_action_time; /* start and end time */ 299 300 /* TRIM related */ 301 boolean_t vdev_trim_exit_wanted; 302 boolean_t vdev_autotrim_exit_wanted; 303 vdev_trim_state_t vdev_trim_state; 304 list_node_t vdev_trim_node; 305 kmutex_t vdev_autotrim_lock; 306 kcondvar_t vdev_autotrim_cv; 307 kcondvar_t vdev_autotrim_kick_cv; 308 kthread_t *vdev_autotrim_thread; 309 /* Protects vdev_trim_thread and vdev_trim_state. */ 310 kmutex_t vdev_trim_lock; 311 kcondvar_t vdev_trim_cv; 312 kthread_t *vdev_trim_thread; 313 uint64_t vdev_trim_offset[TXG_SIZE]; 314 uint64_t vdev_trim_last_offset; 315 uint64_t vdev_trim_bytes_est; 316 uint64_t vdev_trim_bytes_done; 317 uint64_t vdev_trim_rate; /* requested rate (bytes/sec) */ 318 uint64_t vdev_trim_partial; /* requested partial TRIM */ 319 uint64_t vdev_trim_secure; /* requested secure TRIM */ 320 uint64_t vdev_trim_action_time; /* start and end time */ 321 322 /* Rebuild related */ 323 boolean_t vdev_rebuilding; 324 boolean_t vdev_rebuild_exit_wanted; 325 boolean_t vdev_rebuild_cancel_wanted; 326 boolean_t vdev_rebuild_reset_wanted; 327 kmutex_t vdev_rebuild_lock; 328 kcondvar_t vdev_rebuild_cv; 329 kthread_t *vdev_rebuild_thread; 330 vdev_rebuild_t vdev_rebuild_config; 331 332 /* For limiting outstanding I/Os (initialize, TRIM) */ 333 kmutex_t vdev_initialize_io_lock; 334 kcondvar_t vdev_initialize_io_cv; 335 uint64_t vdev_initialize_inflight; 336 kmutex_t vdev_trim_io_lock; 337 kcondvar_t vdev_trim_io_cv; 338 uint64_t vdev_trim_inflight[3]; 339 340 /* 341 * Values stored in the config for an indirect or removing vdev. 342 */ 343 vdev_indirect_config_t vdev_indirect_config; 344 345 /* 346 * The vdev_indirect_rwlock protects the vdev_indirect_mapping 347 * pointer from changing on indirect vdevs (when it is condensed). 348 * Note that removing (not yet indirect) vdevs have different 349 * access patterns (the mapping is not accessed from open context, 350 * e.g. from zio_read) and locking strategy (e.g. svr_lock). 351 */ 352 krwlock_t vdev_indirect_rwlock; 353 vdev_indirect_mapping_t *vdev_indirect_mapping; 354 vdev_indirect_births_t *vdev_indirect_births; 355 356 /* 357 * In memory data structures used to manage the obsolete sm, for 358 * indirect or removing vdevs. 359 * 360 * The vdev_obsolete_segments is the in-core record of the segments 361 * that are no longer referenced anywhere in the pool (due to 362 * being freed or remapped and not referenced by any snapshots). 363 * During a sync, segments are added to vdev_obsolete_segments 364 * via vdev_indirect_mark_obsolete(); at the end of each sync 365 * pass, this is appended to vdev_obsolete_sm via 366 * vdev_indirect_sync_obsolete(). The vdev_obsolete_lock 367 * protects against concurrent modifications of vdev_obsolete_segments 368 * from multiple zio threads. 369 */ 370 kmutex_t vdev_obsolete_lock; 371 zfs_range_tree_t *vdev_obsolete_segments; 372 space_map_t *vdev_obsolete_sm; 373 374 /* 375 * Protects the vdev_scan_io_queue field itself as well as the 376 * structure's contents (when present). 377 */ 378 kmutex_t vdev_scan_io_queue_lock; 379 struct dsl_scan_io_queue *vdev_scan_io_queue; 380 381 /* 382 * Leaf vdev state. 383 */ 384 zfs_range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */ 385 space_map_t *vdev_dtl_sm; /* dirty time log space map */ 386 txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */ 387 uint64_t vdev_dtl_object; /* DTL object */ 388 uint64_t vdev_psize; /* physical device capacity */ 389 uint64_t vdev_wholedisk; /* true if this is a whole disk */ 390 uint64_t vdev_offline; /* persistent offline state */ 391 uint64_t vdev_faulted; /* persistent faulted state */ 392 uint64_t vdev_degraded; /* persistent degraded state */ 393 uint64_t vdev_removed; /* persistent removed state */ 394 uint64_t vdev_resilver_txg; /* persistent resilvering state */ 395 uint64_t vdev_rebuild_txg; /* persistent rebuilding state */ 396 char *vdev_path; /* vdev path (if any) */ 397 char *vdev_devid; /* vdev devid (if any) */ 398 char *vdev_physpath; /* vdev device path (if any) */ 399 char *vdev_enc_sysfs_path; /* enclosure sysfs path */ 400 char *vdev_fru; /* physical FRU location */ 401 uint64_t vdev_not_present; /* not present during import */ 402 uint64_t vdev_unspare; /* unspare when resilvering done */ 403 boolean_t vdev_nowritecache; /* true if flushwritecache failed */ 404 boolean_t vdev_has_trim; /* TRIM is supported */ 405 boolean_t vdev_has_securetrim; /* secure TRIM is supported */ 406 boolean_t vdev_checkremove; /* temporary online test */ 407 boolean_t vdev_forcefault; /* force online fault */ 408 boolean_t vdev_splitting; /* split or repair in progress */ 409 boolean_t vdev_delayed_close; /* delayed device close? */ 410 boolean_t vdev_tmpoffline; /* device taken offline temporarily? */ 411 boolean_t vdev_detached; /* device detached? */ 412 boolean_t vdev_cant_read; /* vdev is failing all reads */ 413 boolean_t vdev_cant_write; /* vdev is failing all writes */ 414 boolean_t vdev_isspare; /* was a hot spare */ 415 boolean_t vdev_isl2cache; /* was a l2cache device */ 416 boolean_t vdev_copy_uberblocks; /* post expand copy uberblocks */ 417 boolean_t vdev_resilver_deferred; /* resilver deferred */ 418 boolean_t vdev_kobj_flag; /* kobj event record */ 419 boolean_t vdev_attaching; /* vdev attach ashift handling */ 420 boolean_t vdev_is_blkdev; /* vdev is backed by block device */ 421 vdev_queue_t vdev_queue; /* I/O deadline schedule queue */ 422 spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */ 423 zio_t *vdev_probe_zio; /* root of current probe */ 424 vdev_aux_t vdev_label_aux; /* on-disk aux state */ 425 uint64_t vdev_leaf_zap; 426 hrtime_t vdev_mmp_pending; /* 0 if write finished */ 427 uint64_t vdev_mmp_kstat_id; /* to find kstat entry */ 428 uint64_t vdev_expansion_time; /* vdev's last expansion time */ 429 /* used to calculate average read latency */ 430 uint64_t *vdev_prev_histo; 431 int64_t vdev_outlier_count; /* read outlier amongst peers */ 432 hrtime_t vdev_read_sit_out_expire; /* end of sit out period */ 433 list_node_t vdev_leaf_node; /* leaf vdev list */ 434 435 /* 436 * For DTrace to work in userland (libzpool) context, these fields must 437 * remain at the end of the structure. DTrace will use the kernel's 438 * CTF definition for 'struct vdev', and since the size of a kmutex_t is 439 * larger in userland, the offsets for the rest of the fields would be 440 * incorrect. 441 */ 442 kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */ 443 kmutex_t vdev_stat_lock; /* vdev_stat */ 444 kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */ 445 446 /* 447 * We rate limit ZIO delay, deadman, and checksum events, since they 448 * can flood ZED with tons of events when a drive is acting up. 449 * 450 * We also rate limit Direct I/O write verify errors, since a user might 451 * be continually manipulating a buffer that can flood ZED with tons of 452 * events. 453 */ 454 zfs_ratelimit_t vdev_delay_rl; 455 zfs_ratelimit_t vdev_deadman_rl; 456 zfs_ratelimit_t vdev_dio_verify_rl; 457 zfs_ratelimit_t vdev_checksum_rl; 458 459 /* 460 * Vdev properties for tuning ZED or zfsd 461 */ 462 uint64_t vdev_checksum_n; 463 uint64_t vdev_checksum_t; 464 uint64_t vdev_io_n; 465 uint64_t vdev_io_t; 466 boolean_t vdev_slow_io_events; 467 uint64_t vdev_slow_io_n; 468 uint64_t vdev_slow_io_t; 469 uint64_t vdev_scheduler; /* control how I/Os are submitted */ 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_dispatch(vdev_t *vd, uint64_t txg); 596 extern void vdev_sync_done(vdev_t *vd, uint64_t txg); 597 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg); 598 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg); 599 600 /* 601 * Available vdev types. 602 */ 603 extern vdev_ops_t vdev_root_ops; 604 extern vdev_ops_t vdev_mirror_ops; 605 extern vdev_ops_t vdev_replacing_ops; 606 extern vdev_ops_t vdev_raidz_ops; 607 extern vdev_ops_t vdev_draid_ops; 608 extern vdev_ops_t vdev_draid_spare_ops; 609 extern vdev_ops_t vdev_disk_ops; 610 extern vdev_ops_t vdev_file_ops; 611 extern vdev_ops_t vdev_missing_ops; 612 extern vdev_ops_t vdev_hole_ops; 613 extern vdev_ops_t vdev_spare_ops; 614 extern vdev_ops_t vdev_indirect_ops; 615 616 /* 617 * Common size functions 618 */ 619 extern void vdev_default_xlate(vdev_t *vd, const zfs_range_seg64_t *logical_rs, 620 zfs_range_seg64_t *physical_rs, zfs_range_seg64_t *remain_rs); 621 extern uint64_t vdev_default_psize(vdev_t *vd, uint64_t asize, uint64_t txg); 622 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize, uint64_t txg); 623 extern uint64_t vdev_default_min_asize(vdev_t *vd); 624 extern uint64_t vdev_get_min_asize(vdev_t *vd); 625 extern void vdev_set_min_asize(vdev_t *vd); 626 extern uint64_t vdev_get_nparity(vdev_t *vd); 627 extern uint64_t vdev_get_ndisks(vdev_t *vd); 628 629 /* 630 * Global variables 631 */ 632 extern int zfs_vdev_standard_sm_blksz; 633 634 /* 635 * Functions from vdev_indirect.c 636 */ 637 extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx); 638 extern boolean_t vdev_indirect_should_condense(vdev_t *vd); 639 extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx); 640 extern int vdev_obsolete_sm_object(vdev_t *vd, uint64_t *sm_obj); 641 extern int vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise); 642 643 /* 644 * Other miscellaneous functions 645 */ 646 int vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj); 647 void vdev_metaslab_group_create(vdev_t *vd); 648 uint64_t vdev_best_ashift(uint64_t logical, uint64_t a, uint64_t b); 649 #if defined(__linux__) && defined(_KERNEL) 650 int param_get_raidz_impl(char *buf, zfs_kernel_param_t *kp); 651 #endif 652 int param_set_raidz_impl(ZFS_MODULE_PARAM_ARGS); 653 char *vdev_rt_name(vdev_t *vd, const char *name); 654 655 /* 656 * Vdev ashift optimization tunables 657 */ 658 extern uint_t zfs_vdev_min_auto_ashift; 659 extern uint_t zfs_vdev_max_auto_ashift; 660 int param_set_min_auto_ashift(ZFS_MODULE_PARAM_ARGS); 661 int param_set_max_auto_ashift(ZFS_MODULE_PARAM_ARGS); 662 663 /* 664 * VDEV checksum verification for Direct I/O writes 665 */ 666 extern uint_t zfs_vdev_direct_write_verify; 667 668 #ifdef __cplusplus 669 } 670 #endif 671 672 #endif /* _SYS_VDEV_IMPL_H */ 673