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 http://www.opensolaris.org/os/licensing. 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 */ 26 27 #ifndef _SYS_VDEV_IMPL_H 28 #define _SYS_VDEV_IMPL_H 29 30 #include <sys/avl.h> 31 #include <sys/bpobj.h> 32 #include <sys/dmu.h> 33 #include <sys/metaslab.h> 34 #include <sys/nvpair.h> 35 #include <sys/space_map.h> 36 #include <sys/vdev.h> 37 #include <sys/dkio.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 typedef struct vdev_cache vdev_cache_t; 61 typedef struct vdev_cache_entry vdev_cache_entry_t; 62 struct abd; 63 64 extern int zfs_vdev_queue_depth_pct; 65 extern int zfs_vdev_def_queue_depth; 66 extern uint32_t zfs_vdev_async_write_max_active; 67 68 /* 69 * Virtual device operations 70 */ 71 typedef int vdev_init_func_t(spa_t *spa, nvlist_t *nv, void **tsd); 72 typedef void vdev_kobj_post_evt_func_t(vdev_t *vd); 73 typedef void vdev_fini_func_t(vdev_t *vd); 74 typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size, 75 uint64_t *ashift, uint64_t *pshift); 76 typedef void vdev_close_func_t(vdev_t *vd); 77 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize); 78 typedef uint64_t vdev_min_asize_func_t(vdev_t *vd); 79 typedef uint64_t vdev_min_alloc_func_t(vdev_t *vd); 80 typedef void vdev_io_start_func_t(zio_t *zio); 81 typedef void vdev_io_done_func_t(zio_t *zio); 82 typedef void vdev_state_change_func_t(vdev_t *vd, int, int); 83 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, const dva_t *dva, 84 size_t psize, uint64_t phys_birth); 85 typedef void vdev_hold_func_t(vdev_t *vd); 86 typedef void vdev_rele_func_t(vdev_t *vd); 87 88 typedef void vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd, 89 uint64_t offset, uint64_t size, void *arg); 90 typedef void vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size, 91 vdev_remap_cb_t callback, void *arg); 92 /* 93 * Given a target vdev, translates the logical range "in" to the physical 94 * range "res" 95 */ 96 typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg64_t *logical, 97 range_seg64_t *physical, range_seg64_t *remain); 98 typedef uint64_t vdev_rebuild_asize_func_t(vdev_t *vd, uint64_t start, 99 uint64_t size, uint64_t max_segment); 100 typedef void vdev_metaslab_init_func_t(vdev_t *vd, uint64_t *startp, 101 uint64_t *sizep); 102 typedef void vdev_config_generate_func_t(vdev_t *vd, nvlist_t *nv); 103 typedef uint64_t vdev_nparity_func_t(vdev_t *vd); 104 typedef uint64_t vdev_ndisks_func_t(vdev_t *vd); 105 106 typedef const struct vdev_ops { 107 vdev_init_func_t *vdev_op_init; 108 vdev_fini_func_t *vdev_op_fini; 109 vdev_open_func_t *vdev_op_open; 110 vdev_close_func_t *vdev_op_close; 111 vdev_asize_func_t *vdev_op_asize; 112 vdev_min_asize_func_t *vdev_op_min_asize; 113 vdev_min_alloc_func_t *vdev_op_min_alloc; 114 vdev_io_start_func_t *vdev_op_io_start; 115 vdev_io_done_func_t *vdev_op_io_done; 116 vdev_state_change_func_t *vdev_op_state_change; 117 vdev_need_resilver_func_t *vdev_op_need_resilver; 118 vdev_hold_func_t *vdev_op_hold; 119 vdev_rele_func_t *vdev_op_rele; 120 vdev_remap_func_t *vdev_op_remap; 121 vdev_xlation_func_t *vdev_op_xlate; 122 vdev_rebuild_asize_func_t *vdev_op_rebuild_asize; 123 vdev_metaslab_init_func_t *vdev_op_metaslab_init; 124 vdev_config_generate_func_t *vdev_op_config_generate; 125 vdev_nparity_func_t *vdev_op_nparity; 126 vdev_ndisks_func_t *vdev_op_ndisks; 127 vdev_kobj_post_evt_func_t *vdev_op_kobj_evt_post; 128 char vdev_op_type[16]; 129 boolean_t vdev_op_leaf; 130 } vdev_ops_t; 131 132 /* 133 * Virtual device properties 134 */ 135 struct vdev_cache_entry { 136 struct abd *ve_abd; 137 uint64_t ve_offset; 138 clock_t ve_lastused; 139 avl_node_t ve_offset_node; 140 avl_node_t ve_lastused_node; 141 uint32_t ve_hits; 142 uint16_t ve_missed_update; 143 zio_t *ve_fill_io; 144 }; 145 146 struct vdev_cache { 147 avl_tree_t vc_offset_tree; 148 avl_tree_t vc_lastused_tree; 149 kmutex_t vc_lock; 150 }; 151 152 typedef struct vdev_queue_class { 153 uint32_t vqc_active; 154 155 /* 156 * Sorted by offset or timestamp, depending on if the queue is 157 * LBA-ordered vs FIFO. 158 */ 159 avl_tree_t vqc_queued_tree; 160 } vdev_queue_class_t; 161 162 struct vdev_queue { 163 vdev_t *vq_vdev; 164 vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE]; 165 avl_tree_t vq_active_tree; 166 avl_tree_t vq_read_offset_tree; 167 avl_tree_t vq_write_offset_tree; 168 avl_tree_t vq_trim_offset_tree; 169 uint64_t vq_last_offset; 170 zio_priority_t vq_last_prio; /* Last sent I/O priority. */ 171 uint32_t vq_ia_active; /* Active interactive I/Os. */ 172 uint32_t vq_nia_credit; /* Non-interactive I/Os credit. */ 173 hrtime_t vq_io_complete_ts; /* time last i/o completed */ 174 hrtime_t vq_io_delta_ts; 175 zio_t vq_io_search; /* used as local for stack reduction */ 176 kmutex_t vq_lock; 177 }; 178 179 typedef enum vdev_alloc_bias { 180 VDEV_BIAS_NONE, 181 VDEV_BIAS_LOG, /* dedicated to ZIL data (SLOG) */ 182 VDEV_BIAS_SPECIAL, /* dedicated to ddt, metadata, and small blks */ 183 VDEV_BIAS_DEDUP /* dedicated to dedup metadata */ 184 } vdev_alloc_bias_t; 185 186 187 /* 188 * On-disk indirect vdev state. 189 * 190 * An indirect vdev is described exclusively in the MOS config of a pool. 191 * The config for an indirect vdev includes several fields, which are 192 * accessed in memory by a vdev_indirect_config_t. 193 */ 194 typedef struct vdev_indirect_config { 195 /* 196 * Object (in MOS) which contains the indirect mapping. This object 197 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by 198 * vimep_src. The bonus buffer for this object is a 199 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev 200 * removal is initiated. 201 * 202 * Note that this object can be empty if none of the data on the vdev 203 * has been copied yet. 204 */ 205 uint64_t vic_mapping_object; 206 207 /* 208 * Object (in MOS) which contains the birth times for the mapping 209 * entries. This object contains an array of 210 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus 211 * buffer for this object is a vdev_indirect_birth_phys_t. This object 212 * is allocated when a vdev removal is initiated. 213 * 214 * Note that this object can be empty if none of the vdev has yet been 215 * copied. 216 */ 217 uint64_t vic_births_object; 218 219 /* 220 * This is the vdev ID which was removed previous to this vdev, or 221 * UINT64_MAX if there are no previously removed vdevs. 222 */ 223 uint64_t vic_prev_indirect_vdev; 224 } vdev_indirect_config_t; 225 226 /* 227 * Virtual device descriptor 228 */ 229 struct vdev { 230 /* 231 * Common to all vdev types. 232 */ 233 uint64_t vdev_id; /* child number in vdev parent */ 234 uint64_t vdev_guid; /* unique ID for this vdev */ 235 uint64_t vdev_guid_sum; /* self guid + all child guids */ 236 uint64_t vdev_orig_guid; /* orig. guid prior to remove */ 237 uint64_t vdev_asize; /* allocatable device capacity */ 238 uint64_t vdev_min_asize; /* min acceptable asize */ 239 uint64_t vdev_max_asize; /* max acceptable asize */ 240 uint64_t vdev_ashift; /* block alignment shift */ 241 242 /* 243 * Logical block alignment shift 244 * 245 * The smallest sized/aligned I/O supported by the device. 246 */ 247 uint64_t vdev_logical_ashift; 248 /* 249 * Physical block alignment shift 250 * 251 * The device supports logical I/Os with vdev_logical_ashift 252 * size/alignment, but optimum performance will be achieved by 253 * aligning/sizing requests to vdev_physical_ashift. Smaller 254 * requests may be inflated or incur device level read-modify-write 255 * operations. 256 * 257 * May be 0 to indicate no preference (i.e. use vdev_logical_ashift). 258 */ 259 uint64_t vdev_physical_ashift; 260 uint64_t vdev_state; /* see VDEV_STATE_* #defines */ 261 uint64_t vdev_prevstate; /* used when reopening a vdev */ 262 vdev_ops_t *vdev_ops; /* vdev operations */ 263 spa_t *vdev_spa; /* spa for this vdev */ 264 void *vdev_tsd; /* type-specific data */ 265 vdev_t *vdev_top; /* top-level vdev */ 266 vdev_t *vdev_parent; /* parent vdev */ 267 vdev_t **vdev_child; /* array of children */ 268 uint64_t vdev_children; /* number of children */ 269 vdev_stat_t vdev_stat; /* virtual device statistics */ 270 vdev_stat_ex_t vdev_stat_ex; /* extended statistics */ 271 boolean_t vdev_expanding; /* expand the vdev? */ 272 boolean_t vdev_reopening; /* reopen in progress? */ 273 boolean_t vdev_nonrot; /* true if solid state */ 274 int vdev_load_error; /* error on last load */ 275 int vdev_open_error; /* error on last open */ 276 int vdev_validate_error; /* error on last validate */ 277 kthread_t *vdev_open_thread; /* thread opening children */ 278 kthread_t *vdev_validate_thread; /* thread validating children */ 279 uint64_t vdev_crtxg; /* txg when top-level was added */ 280 281 /* 282 * Top-level vdev state. 283 */ 284 uint64_t vdev_ms_array; /* metaslab array object */ 285 uint64_t vdev_ms_shift; /* metaslab size shift */ 286 uint64_t vdev_ms_count; /* number of metaslabs */ 287 metaslab_group_t *vdev_mg; /* metaslab group */ 288 metaslab_group_t *vdev_log_mg; /* embedded slog metaslab group */ 289 metaslab_t **vdev_ms; /* metaslab array */ 290 uint64_t vdev_pending_fastwrite; /* allocated fastwrites */ 291 txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */ 292 txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */ 293 txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */ 294 boolean_t vdev_remove_wanted; /* async remove wanted? */ 295 boolean_t vdev_probe_wanted; /* async probe wanted? */ 296 list_node_t vdev_config_dirty_node; /* config dirty list */ 297 list_node_t vdev_state_dirty_node; /* state dirty list */ 298 uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */ 299 uint64_t vdev_islog; /* is an intent log device */ 300 uint64_t vdev_removing; /* device is being removed? */ 301 boolean_t vdev_ishole; /* is a hole in the namespace */ 302 uint64_t vdev_top_zap; 303 vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias */ 304 305 /* pool checkpoint related */ 306 space_map_t *vdev_checkpoint_sm; /* contains reserved blocks */ 307 308 /* Initialize related */ 309 boolean_t vdev_initialize_exit_wanted; 310 vdev_initializing_state_t vdev_initialize_state; 311 list_node_t vdev_initialize_node; 312 kthread_t *vdev_initialize_thread; 313 /* Protects vdev_initialize_thread and vdev_initialize_state. */ 314 kmutex_t vdev_initialize_lock; 315 kcondvar_t vdev_initialize_cv; 316 uint64_t vdev_initialize_offset[TXG_SIZE]; 317 uint64_t vdev_initialize_last_offset; 318 range_tree_t *vdev_initialize_tree; /* valid while initializing */ 319 uint64_t vdev_initialize_bytes_est; 320 uint64_t vdev_initialize_bytes_done; 321 uint64_t vdev_initialize_action_time; /* start and end time */ 322 323 /* TRIM related */ 324 boolean_t vdev_trim_exit_wanted; 325 boolean_t vdev_autotrim_exit_wanted; 326 vdev_trim_state_t vdev_trim_state; 327 list_node_t vdev_trim_node; 328 kmutex_t vdev_autotrim_lock; 329 kcondvar_t vdev_autotrim_cv; 330 kthread_t *vdev_autotrim_thread; 331 /* Protects vdev_trim_thread and vdev_trim_state. */ 332 kmutex_t vdev_trim_lock; 333 kcondvar_t vdev_trim_cv; 334 kthread_t *vdev_trim_thread; 335 uint64_t vdev_trim_offset[TXG_SIZE]; 336 uint64_t vdev_trim_last_offset; 337 uint64_t vdev_trim_bytes_est; 338 uint64_t vdev_trim_bytes_done; 339 uint64_t vdev_trim_rate; /* requested rate (bytes/sec) */ 340 uint64_t vdev_trim_partial; /* requested partial TRIM */ 341 uint64_t vdev_trim_secure; /* requested secure TRIM */ 342 uint64_t vdev_trim_action_time; /* start and end time */ 343 344 /* Rebuild related */ 345 boolean_t vdev_rebuilding; 346 boolean_t vdev_rebuild_exit_wanted; 347 boolean_t vdev_rebuild_cancel_wanted; 348 boolean_t vdev_rebuild_reset_wanted; 349 kmutex_t vdev_rebuild_lock; 350 kcondvar_t vdev_rebuild_cv; 351 kthread_t *vdev_rebuild_thread; 352 vdev_rebuild_t vdev_rebuild_config; 353 354 /* For limiting outstanding I/Os (initialize, TRIM) */ 355 kmutex_t vdev_initialize_io_lock; 356 kcondvar_t vdev_initialize_io_cv; 357 uint64_t vdev_initialize_inflight; 358 kmutex_t vdev_trim_io_lock; 359 kcondvar_t vdev_trim_io_cv; 360 uint64_t vdev_trim_inflight[3]; 361 362 /* 363 * Values stored in the config for an indirect or removing vdev. 364 */ 365 vdev_indirect_config_t vdev_indirect_config; 366 367 /* 368 * The vdev_indirect_rwlock protects the vdev_indirect_mapping 369 * pointer from changing on indirect vdevs (when it is condensed). 370 * Note that removing (not yet indirect) vdevs have different 371 * access patterns (the mapping is not accessed from open context, 372 * e.g. from zio_read) and locking strategy (e.g. svr_lock). 373 */ 374 krwlock_t vdev_indirect_rwlock; 375 vdev_indirect_mapping_t *vdev_indirect_mapping; 376 vdev_indirect_births_t *vdev_indirect_births; 377 378 /* 379 * In memory data structures used to manage the obsolete sm, for 380 * indirect or removing vdevs. 381 * 382 * The vdev_obsolete_segments is the in-core record of the segments 383 * that are no longer referenced anywhere in the pool (due to 384 * being freed or remapped and not referenced by any snapshots). 385 * During a sync, segments are added to vdev_obsolete_segments 386 * via vdev_indirect_mark_obsolete(); at the end of each sync 387 * pass, this is appended to vdev_obsolete_sm via 388 * vdev_indirect_sync_obsolete(). The vdev_obsolete_lock 389 * protects against concurrent modifications of vdev_obsolete_segments 390 * from multiple zio threads. 391 */ 392 kmutex_t vdev_obsolete_lock; 393 range_tree_t *vdev_obsolete_segments; 394 space_map_t *vdev_obsolete_sm; 395 396 /* 397 * Protects the vdev_scan_io_queue field itself as well as the 398 * structure's contents (when present). 399 */ 400 kmutex_t vdev_scan_io_queue_lock; 401 struct dsl_scan_io_queue *vdev_scan_io_queue; 402 403 /* 404 * Leaf vdev state. 405 */ 406 range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */ 407 space_map_t *vdev_dtl_sm; /* dirty time log space map */ 408 txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */ 409 uint64_t vdev_dtl_object; /* DTL object */ 410 uint64_t vdev_psize; /* physical device capacity */ 411 uint64_t vdev_wholedisk; /* true if this is a whole disk */ 412 uint64_t vdev_offline; /* persistent offline state */ 413 uint64_t vdev_faulted; /* persistent faulted state */ 414 uint64_t vdev_degraded; /* persistent degraded state */ 415 uint64_t vdev_removed; /* persistent removed state */ 416 uint64_t vdev_resilver_txg; /* persistent resilvering state */ 417 uint64_t vdev_rebuild_txg; /* persistent rebuilding state */ 418 char *vdev_path; /* vdev path (if any) */ 419 char *vdev_devid; /* vdev devid (if any) */ 420 char *vdev_physpath; /* vdev device path (if any) */ 421 char *vdev_enc_sysfs_path; /* enclosure sysfs path */ 422 char *vdev_fru; /* physical FRU location */ 423 uint64_t vdev_not_present; /* not present during import */ 424 uint64_t vdev_unspare; /* unspare when resilvering done */ 425 boolean_t vdev_nowritecache; /* true if flushwritecache failed */ 426 boolean_t vdev_has_trim; /* TRIM is supported */ 427 boolean_t vdev_has_securetrim; /* secure TRIM is supported */ 428 boolean_t vdev_checkremove; /* temporary online test */ 429 boolean_t vdev_forcefault; /* force online fault */ 430 boolean_t vdev_splitting; /* split or repair in progress */ 431 boolean_t vdev_delayed_close; /* delayed device close? */ 432 boolean_t vdev_tmpoffline; /* device taken offline temporarily? */ 433 boolean_t vdev_detached; /* device detached? */ 434 boolean_t vdev_cant_read; /* vdev is failing all reads */ 435 boolean_t vdev_cant_write; /* vdev is failing all writes */ 436 boolean_t vdev_isspare; /* was a hot spare */ 437 boolean_t vdev_isl2cache; /* was a l2cache device */ 438 boolean_t vdev_copy_uberblocks; /* post expand copy uberblocks */ 439 boolean_t vdev_resilver_deferred; /* resilver deferred */ 440 boolean_t vdev_kobj_flag; /* kobj event record */ 441 vdev_queue_t vdev_queue; /* I/O deadline schedule queue */ 442 vdev_cache_t vdev_cache; /* physical block cache */ 443 spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */ 444 zio_t *vdev_probe_zio; /* root of current probe */ 445 vdev_aux_t vdev_label_aux; /* on-disk aux state */ 446 uint64_t vdev_leaf_zap; 447 hrtime_t vdev_mmp_pending; /* 0 if write finished */ 448 uint64_t vdev_mmp_kstat_id; /* to find kstat entry */ 449 uint64_t vdev_expansion_time; /* vdev's last expansion time */ 450 list_node_t vdev_leaf_node; /* leaf vdev list */ 451 452 /* 453 * For DTrace to work in userland (libzpool) context, these fields must 454 * remain at the end of the structure. DTrace will use the kernel's 455 * CTF definition for 'struct vdev', and since the size of a kmutex_t is 456 * larger in userland, the offsets for the rest of the fields would be 457 * incorrect. 458 */ 459 kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */ 460 kmutex_t vdev_stat_lock; /* vdev_stat */ 461 kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */ 462 463 /* 464 * We rate limit ZIO delay, deadman, and checksum events, since they 465 * can flood ZED with tons of events when a drive is acting up. 466 */ 467 zfs_ratelimit_t vdev_delay_rl; 468 zfs_ratelimit_t vdev_deadman_rl; 469 zfs_ratelimit_t vdev_checksum_rl; 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; 526 527 CTASSERT_GLOBAL(sizeof (vdev_boot_envblock_t) == VDEV_PAD_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 */ 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 range_seg64_t *logical_rs, 618 range_seg64_t *physical_rs, range_seg64_t *remain_rs); 619 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize); 620 extern uint64_t vdev_default_min_asize(vdev_t *vd); 621 extern uint64_t vdev_get_min_asize(vdev_t *vd); 622 extern void vdev_set_min_asize(vdev_t *vd); 623 extern uint64_t vdev_get_min_alloc(vdev_t *vd); 624 extern uint64_t vdev_get_nparity(vdev_t *vd); 625 extern uint64_t vdev_get_ndisks(vdev_t *vd); 626 627 /* 628 * Global variables 629 */ 630 extern int zfs_vdev_standard_sm_blksz; 631 /* zdb uses this tunable, so it must be declared here to make lint happy. */ 632 extern int zfs_vdev_cache_size; 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); 649 650 /* 651 * Vdev ashift optimization tunables 652 */ 653 extern uint64_t zfs_vdev_min_auto_ashift; 654 extern uint64_t zfs_vdev_max_auto_ashift; 655 int param_set_min_auto_ashift(ZFS_MODULE_PARAM_ARGS); 656 int param_set_max_auto_ashift(ZFS_MODULE_PARAM_ARGS); 657 658 #ifdef __cplusplus 659 } 660 #endif 661 662 #endif /* _SYS_VDEV_IMPL_H */ 663