1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * Copyright (C) 2007 Oracle. All rights reserved. 4 */ 5 6 #ifndef BTRFS_VOLUMES_H 7 #define BTRFS_VOLUMES_H 8 9 #include <linux/bio.h> 10 #include <linux/sort.h> 11 #include <linux/btrfs.h> 12 #include "async-thread.h" 13 14 #define BTRFS_MAX_DATA_CHUNK_SIZE (10ULL * SZ_1G) 15 16 extern struct mutex uuid_mutex; 17 18 #define BTRFS_STRIPE_LEN SZ_64K 19 20 struct btrfs_io_geometry { 21 /* remaining bytes before crossing a stripe */ 22 u64 len; 23 /* offset of logical address in chunk */ 24 u64 offset; 25 /* length of single IO stripe */ 26 u64 stripe_len; 27 /* number of stripe where address falls */ 28 u64 stripe_nr; 29 /* offset of address in stripe */ 30 u64 stripe_offset; 31 /* offset of raid56 stripe into the chunk */ 32 u64 raid56_stripe_offset; 33 }; 34 35 /* 36 * Use sequence counter to get consistent device stat data on 37 * 32-bit processors. 38 */ 39 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 40 #include <linux/seqlock.h> 41 #define __BTRFS_NEED_DEVICE_DATA_ORDERED 42 #define btrfs_device_data_ordered_init(device, info) \ 43 seqcount_mutex_init(&device->data_seqcount, &info->chunk_mutex) 44 #else 45 #define btrfs_device_data_ordered_init(device, info) do { } while (0) 46 #endif 47 48 #define BTRFS_DEV_STATE_WRITEABLE (0) 49 #define BTRFS_DEV_STATE_IN_FS_METADATA (1) 50 #define BTRFS_DEV_STATE_MISSING (2) 51 #define BTRFS_DEV_STATE_REPLACE_TGT (3) 52 #define BTRFS_DEV_STATE_FLUSH_SENT (4) 53 #define BTRFS_DEV_STATE_NO_READA (5) 54 55 struct btrfs_zoned_device_info; 56 57 struct btrfs_device { 58 struct list_head dev_list; /* device_list_mutex */ 59 struct list_head dev_alloc_list; /* chunk mutex */ 60 struct list_head post_commit_list; /* chunk mutex */ 61 struct btrfs_fs_devices *fs_devices; 62 struct btrfs_fs_info *fs_info; 63 64 struct rcu_string __rcu *name; 65 66 u64 generation; 67 68 struct block_device *bdev; 69 70 struct btrfs_zoned_device_info *zone_info; 71 72 /* the mode sent to blkdev_get */ 73 fmode_t mode; 74 75 unsigned long dev_state; 76 blk_status_t last_flush_error; 77 78 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED 79 /* A seqcount_t with associated chunk_mutex (for lockdep) */ 80 seqcount_mutex_t data_seqcount; 81 #endif 82 83 /* the internal btrfs device id */ 84 u64 devid; 85 86 /* size of the device in memory */ 87 u64 total_bytes; 88 89 /* size of the device on disk */ 90 u64 disk_total_bytes; 91 92 /* bytes used */ 93 u64 bytes_used; 94 95 /* optimal io alignment for this device */ 96 u32 io_align; 97 98 /* optimal io width for this device */ 99 u32 io_width; 100 /* type and info about this device */ 101 u64 type; 102 103 /* minimal io size for this device */ 104 u32 sector_size; 105 106 /* physical drive uuid (or lvm uuid) */ 107 u8 uuid[BTRFS_UUID_SIZE]; 108 109 /* 110 * size of the device on the current transaction 111 * 112 * This variant is update when committing the transaction, 113 * and protected by chunk mutex 114 */ 115 u64 commit_total_bytes; 116 117 /* bytes used on the current transaction */ 118 u64 commit_bytes_used; 119 120 /* for sending down flush barriers */ 121 struct bio *flush_bio; 122 struct completion flush_wait; 123 124 /* per-device scrub information */ 125 struct scrub_ctx *scrub_ctx; 126 127 /* readahead state */ 128 atomic_t reada_in_flight; 129 u64 reada_next; 130 struct reada_zone *reada_curr_zone; 131 struct radix_tree_root reada_zones; 132 struct radix_tree_root reada_extents; 133 134 /* disk I/O failure stats. For detailed description refer to 135 * enum btrfs_dev_stat_values in ioctl.h */ 136 int dev_stats_valid; 137 138 /* Counter to record the change of device stats */ 139 atomic_t dev_stats_ccnt; 140 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX]; 141 142 struct extent_io_tree alloc_state; 143 144 struct completion kobj_unregister; 145 /* For sysfs/FSID/devinfo/devid/ */ 146 struct kobject devid_kobj; 147 }; 148 149 /* 150 * If we read those variants at the context of their own lock, we needn't 151 * use the following helpers, reading them directly is safe. 152 */ 153 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 154 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 155 static inline u64 \ 156 btrfs_device_get_##name(const struct btrfs_device *dev) \ 157 { \ 158 u64 size; \ 159 unsigned int seq; \ 160 \ 161 do { \ 162 seq = read_seqcount_begin(&dev->data_seqcount); \ 163 size = dev->name; \ 164 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \ 165 return size; \ 166 } \ 167 \ 168 static inline void \ 169 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 170 { \ 171 write_seqcount_begin(&dev->data_seqcount); \ 172 dev->name = size; \ 173 write_seqcount_end(&dev->data_seqcount); \ 174 } 175 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 176 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 177 static inline u64 \ 178 btrfs_device_get_##name(const struct btrfs_device *dev) \ 179 { \ 180 u64 size; \ 181 \ 182 preempt_disable(); \ 183 size = dev->name; \ 184 preempt_enable(); \ 185 return size; \ 186 } \ 187 \ 188 static inline void \ 189 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 190 { \ 191 preempt_disable(); \ 192 dev->name = size; \ 193 preempt_enable(); \ 194 } 195 #else 196 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 197 static inline u64 \ 198 btrfs_device_get_##name(const struct btrfs_device *dev) \ 199 { \ 200 return dev->name; \ 201 } \ 202 \ 203 static inline void \ 204 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 205 { \ 206 dev->name = size; \ 207 } 208 #endif 209 210 BTRFS_DEVICE_GETSET_FUNCS(total_bytes); 211 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes); 212 BTRFS_DEVICE_GETSET_FUNCS(bytes_used); 213 214 enum btrfs_chunk_allocation_policy { 215 BTRFS_CHUNK_ALLOC_REGULAR, 216 }; 217 218 /* 219 * Read policies for mirrored block group profiles, read picks the stripe based 220 * on these policies. 221 */ 222 enum btrfs_read_policy { 223 /* Use process PID to choose the stripe */ 224 BTRFS_READ_POLICY_PID, 225 BTRFS_NR_READ_POLICY, 226 }; 227 228 struct btrfs_fs_devices { 229 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */ 230 u8 metadata_uuid[BTRFS_FSID_SIZE]; 231 bool fsid_change; 232 struct list_head fs_list; 233 234 u64 num_devices; 235 u64 open_devices; 236 u64 rw_devices; 237 u64 missing_devices; 238 u64 total_rw_bytes; 239 u64 total_devices; 240 241 /* Highest generation number of seen devices */ 242 u64 latest_generation; 243 244 struct block_device *latest_bdev; 245 246 /* all of the devices in the FS, protected by a mutex 247 * so we can safely walk it to write out the supers without 248 * worrying about add/remove by the multi-device code. 249 * Scrubbing super can kick off supers writing by holding 250 * this mutex lock. 251 */ 252 struct mutex device_list_mutex; 253 254 /* List of all devices, protected by device_list_mutex */ 255 struct list_head devices; 256 257 /* 258 * Devices which can satisfy space allocation. Protected by 259 * chunk_mutex 260 */ 261 struct list_head alloc_list; 262 263 struct list_head seed_list; 264 bool seeding; 265 266 int opened; 267 268 /* set when we find or add a device that doesn't have the 269 * nonrot flag set 270 */ 271 bool rotating; 272 273 struct btrfs_fs_info *fs_info; 274 /* sysfs kobjects */ 275 struct kobject fsid_kobj; 276 struct kobject *devices_kobj; 277 struct kobject *devinfo_kobj; 278 struct completion kobj_unregister; 279 280 enum btrfs_chunk_allocation_policy chunk_alloc_policy; 281 282 /* Policy used to read the mirrored stripes */ 283 enum btrfs_read_policy read_policy; 284 }; 285 286 #define BTRFS_BIO_INLINE_CSUM_SIZE 64 287 288 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \ 289 - sizeof(struct btrfs_chunk)) \ 290 / sizeof(struct btrfs_stripe) + 1) 291 292 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \ 293 - 2 * sizeof(struct btrfs_disk_key) \ 294 - 2 * sizeof(struct btrfs_chunk)) \ 295 / sizeof(struct btrfs_stripe) + 1) 296 297 /* 298 * we need the mirror number and stripe index to be passed around 299 * the call chain while we are processing end_io (especially errors). 300 * Really, what we need is a btrfs_bio structure that has this info 301 * and is properly sized with its stripe array, but we're not there 302 * quite yet. We have our own btrfs bioset, and all of the bios 303 * we allocate are actually btrfs_io_bios. We'll cram as much of 304 * struct btrfs_bio as we can into this over time. 305 */ 306 struct btrfs_io_bio { 307 unsigned int mirror_num; 308 struct btrfs_device *device; 309 u64 logical; 310 u8 *csum; 311 u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE]; 312 struct bvec_iter iter; 313 /* 314 * This member must come last, bio_alloc_bioset will allocate enough 315 * bytes for entire btrfs_io_bio but relies on bio being last. 316 */ 317 struct bio bio; 318 }; 319 320 static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio) 321 { 322 return container_of(bio, struct btrfs_io_bio, bio); 323 } 324 325 static inline void btrfs_io_bio_free_csum(struct btrfs_io_bio *io_bio) 326 { 327 if (io_bio->csum != io_bio->csum_inline) { 328 kfree(io_bio->csum); 329 io_bio->csum = NULL; 330 } 331 } 332 333 struct btrfs_bio_stripe { 334 struct btrfs_device *dev; 335 u64 physical; 336 u64 length; /* only used for discard mappings */ 337 }; 338 339 struct btrfs_bio { 340 refcount_t refs; 341 atomic_t stripes_pending; 342 struct btrfs_fs_info *fs_info; 343 u64 map_type; /* get from map_lookup->type */ 344 bio_end_io_t *end_io; 345 struct bio *orig_bio; 346 void *private; 347 atomic_t error; 348 int max_errors; 349 int num_stripes; 350 int mirror_num; 351 int num_tgtdevs; 352 int *tgtdev_map; 353 /* 354 * logical block numbers for the start of each stripe 355 * The last one or two are p/q. These are sorted, 356 * so raid_map[0] is the start of our full stripe 357 */ 358 u64 *raid_map; 359 struct btrfs_bio_stripe stripes[]; 360 }; 361 362 struct btrfs_device_info { 363 struct btrfs_device *dev; 364 u64 dev_offset; 365 u64 max_avail; 366 u64 total_avail; 367 }; 368 369 struct btrfs_raid_attr { 370 u8 sub_stripes; /* sub_stripes info for map */ 371 u8 dev_stripes; /* stripes per dev */ 372 u8 devs_max; /* max devs to use */ 373 u8 devs_min; /* min devs needed */ 374 u8 tolerated_failures; /* max tolerated fail devs */ 375 u8 devs_increment; /* ndevs has to be a multiple of this */ 376 u8 ncopies; /* how many copies to data has */ 377 u8 nparity; /* number of stripes worth of bytes to store 378 * parity information */ 379 u8 mindev_error; /* error code if min devs requisite is unmet */ 380 const char raid_name[8]; /* name of the raid */ 381 u64 bg_flag; /* block group flag of the raid */ 382 }; 383 384 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES]; 385 386 struct map_lookup { 387 u64 type; 388 int io_align; 389 int io_width; 390 u64 stripe_len; 391 int num_stripes; 392 int sub_stripes; 393 int verified_stripes; /* For mount time dev extent verification */ 394 struct btrfs_bio_stripe stripes[]; 395 }; 396 397 #define map_lookup_size(n) (sizeof(struct map_lookup) + \ 398 (sizeof(struct btrfs_bio_stripe) * (n))) 399 400 struct btrfs_balance_args; 401 struct btrfs_balance_progress; 402 struct btrfs_balance_control { 403 struct btrfs_balance_args data; 404 struct btrfs_balance_args meta; 405 struct btrfs_balance_args sys; 406 407 u64 flags; 408 409 struct btrfs_balance_progress stat; 410 }; 411 412 enum btrfs_map_op { 413 BTRFS_MAP_READ, 414 BTRFS_MAP_WRITE, 415 BTRFS_MAP_DISCARD, 416 BTRFS_MAP_GET_READ_MIRRORS, 417 }; 418 419 static inline enum btrfs_map_op btrfs_op(struct bio *bio) 420 { 421 switch (bio_op(bio)) { 422 case REQ_OP_DISCARD: 423 return BTRFS_MAP_DISCARD; 424 case REQ_OP_WRITE: 425 return BTRFS_MAP_WRITE; 426 default: 427 WARN_ON_ONCE(1); 428 fallthrough; 429 case REQ_OP_READ: 430 return BTRFS_MAP_READ; 431 } 432 } 433 434 void btrfs_get_bbio(struct btrfs_bio *bbio); 435 void btrfs_put_bbio(struct btrfs_bio *bbio); 436 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, 437 u64 logical, u64 *length, 438 struct btrfs_bio **bbio_ret, int mirror_num); 439 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, 440 u64 logical, u64 *length, 441 struct btrfs_bio **bbio_ret); 442 int btrfs_get_io_geometry(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, 443 u64 logical, u64 len, struct btrfs_io_geometry *io_geom); 444 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info); 445 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info); 446 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 type); 447 void btrfs_mapping_tree_free(struct extent_map_tree *tree); 448 blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio, 449 int mirror_num); 450 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, 451 fmode_t flags, void *holder); 452 struct btrfs_device *btrfs_scan_one_device(const char *path, 453 fmode_t flags, void *holder); 454 int btrfs_forget_devices(const char *path); 455 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices); 456 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices); 457 void btrfs_assign_next_active_device(struct btrfs_device *device, 458 struct btrfs_device *this_dev); 459 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info, 460 u64 devid, 461 const char *devpath); 462 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info, 463 const u64 *devid, 464 const u8 *uuid); 465 void btrfs_free_device(struct btrfs_device *device); 466 int btrfs_rm_device(struct btrfs_fs_info *fs_info, 467 const char *device_path, u64 devid); 468 void __exit btrfs_cleanup_fs_uuids(void); 469 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len); 470 int btrfs_grow_device(struct btrfs_trans_handle *trans, 471 struct btrfs_device *device, u64 new_size); 472 struct btrfs_device *btrfs_find_device(struct btrfs_fs_devices *fs_devices, 473 u64 devid, u8 *uuid, u8 *fsid); 474 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); 475 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path); 476 int btrfs_balance(struct btrfs_fs_info *fs_info, 477 struct btrfs_balance_control *bctl, 478 struct btrfs_ioctl_balance_args *bargs); 479 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf); 480 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info); 481 int btrfs_recover_balance(struct btrfs_fs_info *fs_info); 482 int btrfs_pause_balance(struct btrfs_fs_info *fs_info); 483 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info); 484 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info); 485 int btrfs_uuid_scan_kthread(void *data); 486 int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset); 487 int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes, 488 u64 *start, u64 *max_avail); 489 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index); 490 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info, 491 struct btrfs_ioctl_get_dev_stats *stats); 492 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info); 493 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info); 494 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans); 495 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev); 496 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev); 497 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev); 498 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info, 499 u64 logical, u64 len); 500 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info, 501 u64 logical); 502 int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans, 503 u64 chunk_offset, u64 chunk_size); 504 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset); 505 struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info, 506 u64 logical, u64 length); 507 void btrfs_release_disk_super(struct btrfs_super_block *super); 508 509 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev, 510 int index) 511 { 512 atomic_inc(dev->dev_stat_values + index); 513 /* 514 * This memory barrier orders stores updating statistics before stores 515 * updating dev_stats_ccnt. 516 * 517 * It pairs with smp_rmb() in btrfs_run_dev_stats(). 518 */ 519 smp_mb__before_atomic(); 520 atomic_inc(&dev->dev_stats_ccnt); 521 } 522 523 static inline int btrfs_dev_stat_read(struct btrfs_device *dev, 524 int index) 525 { 526 return atomic_read(dev->dev_stat_values + index); 527 } 528 529 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev, 530 int index) 531 { 532 int ret; 533 534 ret = atomic_xchg(dev->dev_stat_values + index, 0); 535 /* 536 * atomic_xchg implies a full memory barriers as per atomic_t.txt: 537 * - RMW operations that have a return value are fully ordered; 538 * 539 * This implicit memory barriers is paired with the smp_rmb in 540 * btrfs_run_dev_stats 541 */ 542 atomic_inc(&dev->dev_stats_ccnt); 543 return ret; 544 } 545 546 static inline void btrfs_dev_stat_set(struct btrfs_device *dev, 547 int index, unsigned long val) 548 { 549 atomic_set(dev->dev_stat_values + index, val); 550 /* 551 * This memory barrier orders stores updating statistics before stores 552 * updating dev_stats_ccnt. 553 * 554 * It pairs with smp_rmb() in btrfs_run_dev_stats(). 555 */ 556 smp_mb__before_atomic(); 557 atomic_inc(&dev->dev_stats_ccnt); 558 } 559 560 /* 561 * Convert block group flags (BTRFS_BLOCK_GROUP_*) to btrfs_raid_types, which 562 * can be used as index to access btrfs_raid_array[]. 563 */ 564 static inline enum btrfs_raid_types btrfs_bg_flags_to_raid_index(u64 flags) 565 { 566 if (flags & BTRFS_BLOCK_GROUP_RAID10) 567 return BTRFS_RAID_RAID10; 568 else if (flags & BTRFS_BLOCK_GROUP_RAID1) 569 return BTRFS_RAID_RAID1; 570 else if (flags & BTRFS_BLOCK_GROUP_RAID1C3) 571 return BTRFS_RAID_RAID1C3; 572 else if (flags & BTRFS_BLOCK_GROUP_RAID1C4) 573 return BTRFS_RAID_RAID1C4; 574 else if (flags & BTRFS_BLOCK_GROUP_DUP) 575 return BTRFS_RAID_DUP; 576 else if (flags & BTRFS_BLOCK_GROUP_RAID0) 577 return BTRFS_RAID_RAID0; 578 else if (flags & BTRFS_BLOCK_GROUP_RAID5) 579 return BTRFS_RAID_RAID5; 580 else if (flags & BTRFS_BLOCK_GROUP_RAID6) 581 return BTRFS_RAID_RAID6; 582 583 return BTRFS_RAID_SINGLE; /* BTRFS_BLOCK_GROUP_SINGLE */ 584 } 585 586 void btrfs_commit_device_sizes(struct btrfs_transaction *trans); 587 588 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void); 589 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info, 590 struct btrfs_device *failing_dev); 591 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, 592 struct block_device *bdev, 593 const char *device_path); 594 595 int btrfs_bg_type_to_factor(u64 flags); 596 const char *btrfs_bg_type_to_raid_name(u64 flags); 597 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info); 598 599 #endif 600