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, 2018 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_removal.h> 42 43 #ifdef __cplusplus 44 extern "C" { 45 #endif 46 47 /* 48 * Virtual device descriptors. 49 * 50 * All storage pool operations go through the virtual device framework, 51 * which provides data replication and I/O scheduling. 52 */ 53 54 /* 55 * Forward declarations that lots of things need. 56 */ 57 typedef struct vdev_queue vdev_queue_t; 58 typedef struct vdev_cache vdev_cache_t; 59 typedef struct vdev_cache_entry vdev_cache_entry_t; 60 struct abd; 61 62 extern int zfs_vdev_queue_depth_pct; 63 extern int zfs_vdev_def_queue_depth; 64 extern uint32_t zfs_vdev_async_write_max_active; 65 66 /* 67 * Virtual device operations 68 */ 69 typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size, 70 uint64_t *ashift); 71 typedef void vdev_close_func_t(vdev_t *vd); 72 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize); 73 typedef void vdev_io_start_func_t(zio_t *zio); 74 typedef void vdev_io_done_func_t(zio_t *zio); 75 typedef void vdev_state_change_func_t(vdev_t *vd, int, int); 76 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, uint64_t, size_t); 77 typedef void vdev_hold_func_t(vdev_t *vd); 78 typedef void vdev_rele_func_t(vdev_t *vd); 79 80 typedef void vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd, 81 uint64_t offset, uint64_t size, void *arg); 82 typedef void vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size, 83 vdev_remap_cb_t callback, void *arg); 84 /* 85 * Given a target vdev, translates the logical range "in" to the physical 86 * range "res" 87 */ 88 typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg_t *in, 89 range_seg_t *res); 90 91 typedef struct vdev_ops { 92 vdev_open_func_t *vdev_op_open; 93 vdev_close_func_t *vdev_op_close; 94 vdev_asize_func_t *vdev_op_asize; 95 vdev_io_start_func_t *vdev_op_io_start; 96 vdev_io_done_func_t *vdev_op_io_done; 97 vdev_state_change_func_t *vdev_op_state_change; 98 vdev_need_resilver_func_t *vdev_op_need_resilver; 99 vdev_hold_func_t *vdev_op_hold; 100 vdev_rele_func_t *vdev_op_rele; 101 vdev_remap_func_t *vdev_op_remap; 102 /* 103 * For translating ranges from non-leaf vdevs (e.g. raidz) to leaves. 104 * Used when initializing vdevs. Isn't used by leaf ops. 105 */ 106 vdev_xlation_func_t *vdev_op_xlate; 107 char vdev_op_type[16]; 108 boolean_t vdev_op_leaf; 109 } vdev_ops_t; 110 111 /* 112 * Virtual device properties 113 */ 114 struct vdev_cache_entry { 115 struct abd *ve_abd; 116 uint64_t ve_offset; 117 uint64_t ve_lastused; 118 avl_node_t ve_offset_node; 119 avl_node_t ve_lastused_node; 120 uint32_t ve_hits; 121 uint16_t ve_missed_update; 122 zio_t *ve_fill_io; 123 }; 124 125 struct vdev_cache { 126 avl_tree_t vc_offset_tree; 127 avl_tree_t vc_lastused_tree; 128 kmutex_t vc_lock; 129 }; 130 131 typedef struct vdev_queue_class { 132 uint32_t vqc_active; 133 134 /* 135 * Sorted by offset or timestamp, depending on if the queue is 136 * LBA-ordered vs FIFO. 137 */ 138 avl_tree_t vqc_queued_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_active_tree; 145 avl_tree_t vq_read_offset_tree; 146 avl_tree_t vq_write_offset_tree; 147 uint64_t vq_last_offset; 148 hrtime_t vq_io_complete_ts; /* time last i/o completed */ 149 kmutex_t vq_lock; 150 }; 151 152 typedef enum vdev_alloc_bias { 153 VDEV_BIAS_NONE, 154 VDEV_BIAS_LOG, /* dedicated to ZIL data (SLOG) */ 155 VDEV_BIAS_SPECIAL, /* dedicated to ddt, metadata, and small blks */ 156 VDEV_BIAS_DEDUP /* dedicated to dedup metadata */ 157 } vdev_alloc_bias_t; 158 159 160 /* 161 * On-disk indirect vdev state. 162 * 163 * An indirect vdev is described exclusively in the MOS config of a pool. 164 * The config for an indirect vdev includes several fields, which are 165 * accessed in memory by a vdev_indirect_config_t. 166 */ 167 typedef struct vdev_indirect_config { 168 /* 169 * Object (in MOS) which contains the indirect mapping. This object 170 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by 171 * vimep_src. The bonus buffer for this object is a 172 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev 173 * removal is initiated. 174 * 175 * Note that this object can be empty if none of the data on the vdev 176 * has been copied yet. 177 */ 178 uint64_t vic_mapping_object; 179 180 /* 181 * Object (in MOS) which contains the birth times for the mapping 182 * entries. This object contains an array of 183 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus 184 * buffer for this object is a vdev_indirect_birth_phys_t. This object 185 * is allocated when a vdev removal is initiated. 186 * 187 * Note that this object can be empty if none of the vdev has yet been 188 * copied. 189 */ 190 uint64_t vic_births_object; 191 192 /* 193 * This is the vdev ID which was removed previous to this vdev, or 194 * UINT64_MAX if there are no previously removed vdevs. 195 */ 196 uint64_t vic_prev_indirect_vdev; 197 } vdev_indirect_config_t; 198 199 /* 200 * Virtual device descriptor 201 */ 202 struct vdev { 203 /* 204 * Common to all vdev types. 205 */ 206 uint64_t vdev_id; /* child number in vdev parent */ 207 uint64_t vdev_guid; /* unique ID for this vdev */ 208 uint64_t vdev_guid_sum; /* self guid + all child guids */ 209 uint64_t vdev_orig_guid; /* orig. guid prior to remove */ 210 uint64_t vdev_asize; /* allocatable device capacity */ 211 uint64_t vdev_min_asize; /* min acceptable asize */ 212 uint64_t vdev_max_asize; /* max acceptable asize */ 213 uint64_t vdev_ashift; /* block alignment shift */ 214 uint64_t vdev_state; /* see VDEV_STATE_* #defines */ 215 uint64_t vdev_prevstate; /* used when reopening a vdev */ 216 vdev_ops_t *vdev_ops; /* vdev operations */ 217 spa_t *vdev_spa; /* spa for this vdev */ 218 void *vdev_tsd; /* type-specific data */ 219 vnode_t *vdev_name_vp; /* vnode for pathname */ 220 vnode_t *vdev_devid_vp; /* vnode for devid */ 221 vdev_t *vdev_top; /* top-level vdev */ 222 vdev_t *vdev_parent; /* parent vdev */ 223 vdev_t **vdev_child; /* array of children */ 224 uint64_t vdev_children; /* number of children */ 225 vdev_stat_t vdev_stat; /* virtual device statistics */ 226 boolean_t vdev_expanding; /* expand the vdev? */ 227 boolean_t vdev_reopening; /* reopen in progress? */ 228 int vdev_open_error; /* error on last open */ 229 kthread_t *vdev_open_thread; /* thread opening children */ 230 uint64_t vdev_crtxg; /* txg when top-level was added */ 231 232 /* 233 * Top-level vdev state. 234 */ 235 uint64_t vdev_ms_array; /* metaslab array object */ 236 uint64_t vdev_ms_shift; /* metaslab size shift */ 237 uint64_t vdev_ms_count; /* number of metaslabs */ 238 metaslab_group_t *vdev_mg; /* metaslab group */ 239 metaslab_t **vdev_ms; /* metaslab array */ 240 txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */ 241 txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */ 242 txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */ 243 boolean_t vdev_remove_wanted; /* async remove wanted? */ 244 boolean_t vdev_probe_wanted; /* async probe wanted? */ 245 list_node_t vdev_config_dirty_node; /* config dirty list */ 246 list_node_t vdev_state_dirty_node; /* state dirty list */ 247 uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */ 248 uint64_t vdev_islog; /* is an intent log device */ 249 uint64_t vdev_removing; /* device is being removed? */ 250 boolean_t vdev_ishole; /* is a hole in the namespace */ 251 uint64_t vdev_top_zap; 252 vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias */ 253 254 /* pool checkpoint related */ 255 space_map_t *vdev_checkpoint_sm; /* contains reserved blocks */ 256 257 boolean_t vdev_initialize_exit_wanted; 258 vdev_initializing_state_t vdev_initialize_state; 259 kthread_t *vdev_initialize_thread; 260 /* Protects vdev_initialize_thread and vdev_initialize_state. */ 261 kmutex_t vdev_initialize_lock; 262 kcondvar_t vdev_initialize_cv; 263 uint64_t vdev_initialize_offset[TXG_SIZE]; 264 uint64_t vdev_initialize_last_offset; 265 range_tree_t *vdev_initialize_tree; /* valid while initializing */ 266 uint64_t vdev_initialize_bytes_est; 267 uint64_t vdev_initialize_bytes_done; 268 time_t vdev_initialize_action_time; /* start and end time */ 269 270 /* for limiting outstanding I/Os */ 271 kmutex_t vdev_initialize_io_lock; 272 kcondvar_t vdev_initialize_io_cv; 273 uint64_t vdev_initialize_inflight; 274 275 /* 276 * Values stored in the config for an indirect or removing vdev. 277 */ 278 vdev_indirect_config_t vdev_indirect_config; 279 280 /* 281 * The vdev_indirect_rwlock protects the vdev_indirect_mapping 282 * pointer from changing on indirect vdevs (when it is condensed). 283 * Note that removing (not yet indirect) vdevs have different 284 * access patterns (the mapping is not accessed from open context, 285 * e.g. from zio_read) and locking strategy (e.g. svr_lock). 286 */ 287 krwlock_t vdev_indirect_rwlock; 288 vdev_indirect_mapping_t *vdev_indirect_mapping; 289 vdev_indirect_births_t *vdev_indirect_births; 290 291 /* 292 * In memory data structures used to manage the obsolete sm, for 293 * indirect or removing vdevs. 294 * 295 * The vdev_obsolete_segments is the in-core record of the segments 296 * that are no longer referenced anywhere in the pool (due to 297 * being freed or remapped and not referenced by any snapshots). 298 * During a sync, segments are added to vdev_obsolete_segments 299 * via vdev_indirect_mark_obsolete(); at the end of each sync 300 * pass, this is appended to vdev_obsolete_sm via 301 * vdev_indirect_sync_obsolete(). The vdev_obsolete_lock 302 * protects against concurrent modifications of vdev_obsolete_segments 303 * from multiple zio threads. 304 */ 305 kmutex_t vdev_obsolete_lock; 306 range_tree_t *vdev_obsolete_segments; 307 space_map_t *vdev_obsolete_sm; 308 309 /* 310 * Protects the vdev_scan_io_queue field itself as well as the 311 * structure's contents (when present). 312 */ 313 kmutex_t vdev_scan_io_queue_lock; 314 struct dsl_scan_io_queue *vdev_scan_io_queue; 315 316 /* 317 * Leaf vdev state. 318 */ 319 range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */ 320 space_map_t *vdev_dtl_sm; /* dirty time log space map */ 321 txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */ 322 uint64_t vdev_dtl_object; /* DTL object */ 323 uint64_t vdev_psize; /* physical device capacity */ 324 uint64_t vdev_wholedisk; /* true if this is a whole disk */ 325 uint64_t vdev_offline; /* persistent offline state */ 326 uint64_t vdev_faulted; /* persistent faulted state */ 327 uint64_t vdev_degraded; /* persistent degraded state */ 328 uint64_t vdev_removed; /* persistent removed state */ 329 uint64_t vdev_resilver_txg; /* persistent resilvering state */ 330 uint64_t vdev_nparity; /* number of parity devices for raidz */ 331 char *vdev_path; /* vdev path (if any) */ 332 char *vdev_devid; /* vdev devid (if any) */ 333 char *vdev_physpath; /* vdev device path (if any) */ 334 char *vdev_fru; /* physical FRU location */ 335 uint64_t vdev_not_present; /* not present during import */ 336 uint64_t vdev_unspare; /* unspare when resilvering done */ 337 boolean_t vdev_nowritecache; /* true if flushwritecache failed */ 338 boolean_t vdev_checkremove; /* temporary online test */ 339 boolean_t vdev_forcefault; /* force online fault */ 340 boolean_t vdev_splitting; /* split or repair in progress */ 341 boolean_t vdev_delayed_close; /* delayed device close? */ 342 boolean_t vdev_tmpoffline; /* device taken offline temporarily? */ 343 boolean_t vdev_detached; /* device detached? */ 344 boolean_t vdev_cant_read; /* vdev is failing all reads */ 345 boolean_t vdev_cant_write; /* vdev is failing all writes */ 346 boolean_t vdev_isspare; /* was a hot spare */ 347 boolean_t vdev_isl2cache; /* was a l2cache device */ 348 vdev_queue_t vdev_queue; /* I/O deadline schedule queue */ 349 vdev_cache_t vdev_cache; /* physical block cache */ 350 spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */ 351 zio_t *vdev_probe_zio; /* root of current probe */ 352 vdev_aux_t vdev_label_aux; /* on-disk aux state */ 353 uint64_t vdev_leaf_zap; 354 hrtime_t vdev_mmp_pending; /* 0 if write finished */ 355 uint64_t vdev_mmp_kstat_id; /* to find kstat entry */ 356 list_node_t vdev_leaf_node; /* leaf vdev list */ 357 358 /* 359 * For DTrace to work in userland (libzpool) context, these fields must 360 * remain at the end of the structure. DTrace will use the kernel's 361 * CTF definition for 'struct vdev', and since the size of a kmutex_t is 362 * larger in userland, the offsets for the rest of the fields would be 363 * incorrect. 364 */ 365 kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */ 366 kmutex_t vdev_stat_lock; /* vdev_stat */ 367 kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */ 368 }; 369 370 #define VDEV_RAIDZ_MAXPARITY 3 371 372 #define VDEV_PAD_SIZE (8 << 10) 373 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */ 374 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 375 #define VDEV_PHYS_SIZE (112 << 10) 376 #define VDEV_UBERBLOCK_RING (128 << 10) 377 378 /* 379 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock 380 * ring when MMP is enabled. 381 */ 382 #define MMP_BLOCKS_PER_LABEL 1 383 384 /* The largest uberblock we support is 8k. */ 385 #define MAX_UBERBLOCK_SHIFT (13) 386 #define VDEV_UBERBLOCK_SHIFT(vd) \ 387 MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \ 388 MAX_UBERBLOCK_SHIFT) 389 #define VDEV_UBERBLOCK_COUNT(vd) \ 390 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) 391 #define VDEV_UBERBLOCK_OFFSET(vd, n) \ 392 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) 393 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) 394 395 typedef struct vdev_phys { 396 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; 397 zio_eck_t vp_zbt; 398 } vdev_phys_t; 399 400 typedef struct vdev_label { 401 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ 402 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */ 403 vdev_phys_t vl_vdev_phys; /* 112K */ 404 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ 405 } vdev_label_t; /* 256K total */ 406 407 /* 408 * vdev_dirty() flags 409 */ 410 #define VDD_METASLAB 0x01 411 #define VDD_DTL 0x02 412 413 /* Offset of embedded boot loader region on each label */ 414 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) 415 /* 416 * Size of embedded boot loader region on each label. 417 * The total size of the first two labels plus the boot area is 4MB. 418 */ 419 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ 420 421 /* 422 * Size of label regions at the start and end of each leaf device. 423 */ 424 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) 425 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) 426 #define VDEV_LABELS 4 427 #define VDEV_BEST_LABEL VDEV_LABELS 428 429 #define VDEV_ALLOC_LOAD 0 430 #define VDEV_ALLOC_ADD 1 431 #define VDEV_ALLOC_SPARE 2 432 #define VDEV_ALLOC_L2CACHE 3 433 #define VDEV_ALLOC_ROOTPOOL 4 434 #define VDEV_ALLOC_SPLIT 5 435 #define VDEV_ALLOC_ATTACH 6 436 437 /* 438 * Allocate or free a vdev 439 */ 440 extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, 441 vdev_ops_t *ops); 442 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config, 443 vdev_t *parent, uint_t id, int alloctype); 444 extern void vdev_free(vdev_t *vd); 445 446 /* 447 * Add or remove children and parents 448 */ 449 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd); 450 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd); 451 extern void vdev_compact_children(vdev_t *pvd); 452 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops); 453 extern void vdev_remove_parent(vdev_t *cvd); 454 455 /* 456 * vdev sync load and sync 457 */ 458 extern boolean_t vdev_log_state_valid(vdev_t *vd); 459 extern int vdev_load(vdev_t *vd); 460 extern int vdev_dtl_load(vdev_t *vd); 461 extern void vdev_sync(vdev_t *vd, uint64_t txg); 462 extern void vdev_sync_done(vdev_t *vd, uint64_t txg); 463 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg); 464 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg); 465 466 /* 467 * Available vdev types. 468 */ 469 extern vdev_ops_t vdev_root_ops; 470 extern vdev_ops_t vdev_mirror_ops; 471 extern vdev_ops_t vdev_replacing_ops; 472 extern vdev_ops_t vdev_raidz_ops; 473 extern vdev_ops_t vdev_disk_ops; 474 extern vdev_ops_t vdev_file_ops; 475 extern vdev_ops_t vdev_missing_ops; 476 extern vdev_ops_t vdev_hole_ops; 477 extern vdev_ops_t vdev_spare_ops; 478 extern vdev_ops_t vdev_indirect_ops; 479 480 /* 481 * Common size functions 482 */ 483 extern void vdev_default_xlate(vdev_t *vd, const range_seg_t *in, 484 range_seg_t *out); 485 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize); 486 extern uint64_t vdev_get_min_asize(vdev_t *vd); 487 extern void vdev_set_min_asize(vdev_t *vd); 488 489 /* 490 * Global variables 491 */ 492 extern int vdev_standard_sm_blksz; 493 /* zdb uses this tunable, so it must be declared here to make lint happy. */ 494 extern int zfs_vdev_cache_size; 495 496 /* 497 * Functions from vdev_indirect.c 498 */ 499 extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx); 500 extern boolean_t vdev_indirect_should_condense(vdev_t *vd); 501 extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx); 502 extern int vdev_obsolete_sm_object(vdev_t *vd); 503 extern boolean_t vdev_obsolete_counts_are_precise(vdev_t *vd); 504 505 /* 506 * Other miscellaneous functions 507 */ 508 int vdev_checkpoint_sm_object(vdev_t *vd); 509 510 /* 511 * The vdev_buf_t is used to translate between zio_t and buf_t, and back again. 512 */ 513 typedef struct vdev_buf { 514 buf_t vb_buf; /* buffer that describes the io */ 515 zio_t *vb_io; /* pointer back to the original zio_t */ 516 } vdev_buf_t; 517 518 #ifdef __cplusplus 519 } 520 #endif 521 522 #endif /* _SYS_VDEV_IMPL_H */ 523