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/blk_types.h> 10 #include <linux/blkdev.h> 11 #include <linux/sizes.h> 12 #include <linux/atomic.h> 13 #include <linux/sort.h> 14 #include <linux/list.h> 15 #include <linux/mutex.h> 16 #include <linux/log2.h> 17 #include <linux/kobject.h> 18 #include <linux/refcount.h> 19 #include <linux/completion.h> 20 #include <linux/rbtree.h> 21 #include <uapi/linux/btrfs.h> 22 #include <uapi/linux/btrfs_tree.h> 23 #include "messages.h" 24 #include "extent-io-tree.h" 25 26 struct block_device; 27 struct bdev_handle; 28 struct btrfs_fs_info; 29 struct btrfs_block_group; 30 struct btrfs_trans_handle; 31 struct btrfs_transaction; 32 struct btrfs_zoned_device_info; 33 34 #define BTRFS_MAX_DATA_CHUNK_SIZE (10ULL * SZ_1G) 35 36 /* 37 * Arbitrary maximum size of one discard request to limit potentially long time 38 * spent in blkdev_issue_discard(). 39 */ 40 #define BTRFS_MAX_DISCARD_CHUNK_SIZE (SZ_1G) 41 42 extern struct mutex uuid_mutex; 43 44 #define BTRFS_STRIPE_LEN SZ_64K 45 #define BTRFS_STRIPE_LEN_SHIFT (16) 46 #define BTRFS_STRIPE_LEN_MASK (BTRFS_STRIPE_LEN - 1) 47 48 static_assert(ilog2(BTRFS_STRIPE_LEN) == BTRFS_STRIPE_LEN_SHIFT); 49 50 /* Used by sanity check for btrfs_raid_types. */ 51 #define const_ffs(n) (__builtin_ctzll(n) + 1) 52 53 /* 54 * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires 55 * RAID0 always to be the lowest profile bit. 56 * Although it's part of on-disk format and should never change, do extra 57 * compile-time sanity checks. 58 */ 59 static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) < 60 const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0)); 61 static_assert(ilog2(BTRFS_BLOCK_GROUP_RAID0) > ilog2(BTRFS_BLOCK_GROUP_TYPE_MASK)); 62 63 /* ilog2() can handle both constants and variables */ 64 #define BTRFS_BG_FLAG_TO_INDEX(profile) \ 65 ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1)) 66 67 enum btrfs_raid_types { 68 /* SINGLE is the special one as it doesn't have on-disk bit. */ 69 BTRFS_RAID_SINGLE = 0, 70 71 BTRFS_RAID_RAID0 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0), 72 BTRFS_RAID_RAID1 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1), 73 BTRFS_RAID_DUP = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP), 74 BTRFS_RAID_RAID10 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10), 75 BTRFS_RAID_RAID5 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5), 76 BTRFS_RAID_RAID6 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6), 77 BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3), 78 BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4), 79 80 BTRFS_NR_RAID_TYPES 81 }; 82 83 /* 84 * Use sequence counter to get consistent device stat data on 85 * 32-bit processors. 86 */ 87 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 88 #include <linux/seqlock.h> 89 #define __BTRFS_NEED_DEVICE_DATA_ORDERED 90 #define btrfs_device_data_ordered_init(device) \ 91 seqcount_init(&device->data_seqcount) 92 #else 93 #define btrfs_device_data_ordered_init(device) do { } while (0) 94 #endif 95 96 #define BTRFS_DEV_STATE_WRITEABLE (0) 97 #define BTRFS_DEV_STATE_IN_FS_METADATA (1) 98 #define BTRFS_DEV_STATE_MISSING (2) 99 #define BTRFS_DEV_STATE_REPLACE_TGT (3) 100 #define BTRFS_DEV_STATE_FLUSH_SENT (4) 101 #define BTRFS_DEV_STATE_NO_READA (5) 102 103 /* Special value encoding failure to write primary super block. */ 104 #define BTRFS_SUPER_PRIMARY_WRITE_ERROR (INT_MAX / 2) 105 106 struct btrfs_fs_devices; 107 108 struct btrfs_device { 109 struct list_head dev_list; /* device_list_mutex */ 110 struct list_head dev_alloc_list; /* chunk mutex */ 111 struct list_head post_commit_list; /* chunk mutex */ 112 struct btrfs_fs_devices *fs_devices; 113 struct btrfs_fs_info *fs_info; 114 115 /* Device path or NULL if missing. */ 116 const char __rcu *name; 117 118 u64 generation; 119 120 struct file *bdev_file; 121 struct block_device *bdev; 122 123 struct btrfs_zoned_device_info *zone_info; 124 125 /* 126 * Device's major-minor number. Must be set even if the device is not 127 * opened (bdev == NULL), unless the device is missing. 128 */ 129 dev_t devt; 130 unsigned long dev_state; 131 blk_status_t last_flush_error; 132 133 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED 134 seqcount_t data_seqcount; 135 #endif 136 137 /* the internal btrfs device id */ 138 u64 devid; 139 140 /* size of the device in memory */ 141 u64 total_bytes; 142 143 /* size of the device on disk */ 144 u64 disk_total_bytes; 145 146 /* bytes used */ 147 u64 bytes_used; 148 149 /* optimal io alignment for this device */ 150 u32 io_align; 151 152 /* optimal io width for this device */ 153 u32 io_width; 154 /* type and info about this device */ 155 u64 type; 156 157 /* 158 * Counter of super block write errors, values larger than 159 * BTRFS_SUPER_PRIMARY_WRITE_ERROR encode primary super block write failure. 160 */ 161 atomic_t sb_write_errors; 162 163 /* minimal io size for this device */ 164 u32 sector_size; 165 166 /* physical drive uuid (or lvm uuid) */ 167 u8 uuid[BTRFS_UUID_SIZE]; 168 169 /* 170 * size of the device on the current transaction 171 * 172 * This variant is update when committing the transaction, 173 * and protected by chunk mutex 174 */ 175 u64 commit_total_bytes; 176 177 /* bytes used on the current transaction */ 178 u64 commit_bytes_used; 179 180 /* Bio used for flushing device barriers */ 181 struct bio flush_bio; 182 struct completion flush_wait; 183 184 /* per-device scrub information */ 185 struct scrub_ctx *scrub_ctx; 186 187 /* disk I/O failure stats. For detailed description refer to 188 * enum btrfs_dev_stat_values in ioctl.h */ 189 int dev_stats_valid; 190 191 /* Counter to record the change of device stats */ 192 atomic_t dev_stats_ccnt; 193 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX]; 194 195 struct extent_io_tree alloc_state; 196 197 struct completion kobj_unregister; 198 /* For sysfs/FSID/devinfo/devid/ */ 199 struct kobject devid_kobj; 200 201 /* Bandwidth limit for scrub, in bytes */ 202 u64 scrub_speed_max; 203 }; 204 205 /* 206 * Block group or device which contains an active swapfile. Used for preventing 207 * unsafe operations while a swapfile is active. 208 * 209 * These are sorted on (ptr, inode) (note that a block group or device can 210 * contain more than one swapfile). We compare the pointer values because we 211 * don't actually care what the object is, we just need a quick check whether 212 * the object exists in the rbtree. 213 */ 214 struct btrfs_swapfile_pin { 215 struct rb_node node; 216 void *ptr; 217 struct inode *inode; 218 /* 219 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr 220 * points to a struct btrfs_device. 221 */ 222 bool is_block_group; 223 /* 224 * Only used when 'is_block_group' is true and it is the number of 225 * extents used by a swapfile for this block group ('ptr' field). 226 */ 227 int bg_extent_count; 228 }; 229 230 /* 231 * If we read those variants at the context of their own lock, we needn't 232 * use the following helpers, reading them directly is safe. 233 */ 234 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 235 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 236 static inline u64 \ 237 btrfs_device_get_##name(const struct btrfs_device *dev) \ 238 { \ 239 u64 size; \ 240 unsigned int seq; \ 241 \ 242 do { \ 243 seq = read_seqcount_begin(&dev->data_seqcount); \ 244 size = dev->name; \ 245 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \ 246 return size; \ 247 } \ 248 \ 249 static inline void \ 250 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 251 { \ 252 preempt_disable(); \ 253 write_seqcount_begin(&dev->data_seqcount); \ 254 dev->name = size; \ 255 write_seqcount_end(&dev->data_seqcount); \ 256 preempt_enable(); \ 257 } 258 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 259 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 260 static inline u64 \ 261 btrfs_device_get_##name(const struct btrfs_device *dev) \ 262 { \ 263 u64 size; \ 264 \ 265 preempt_disable(); \ 266 size = dev->name; \ 267 preempt_enable(); \ 268 return size; \ 269 } \ 270 \ 271 static inline void \ 272 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 273 { \ 274 preempt_disable(); \ 275 dev->name = size; \ 276 preempt_enable(); \ 277 } 278 #else 279 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 280 static inline u64 \ 281 btrfs_device_get_##name(const struct btrfs_device *dev) \ 282 { \ 283 return dev->name; \ 284 } \ 285 \ 286 static inline void \ 287 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 288 { \ 289 dev->name = size; \ 290 } 291 #endif 292 293 BTRFS_DEVICE_GETSET_FUNCS(total_bytes); 294 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes); 295 BTRFS_DEVICE_GETSET_FUNCS(bytes_used); 296 297 enum btrfs_chunk_allocation_policy { 298 BTRFS_CHUNK_ALLOC_REGULAR, 299 BTRFS_CHUNK_ALLOC_ZONED, 300 }; 301 302 #define BTRFS_DEFAULT_RR_MIN_CONTIG_READ (SZ_256K) 303 /* Keep in sync with raid_attr table, current maximum is RAID1C4. */ 304 #define BTRFS_RAID1_MAX_MIRRORS (4) 305 /* 306 * Read policies for mirrored block group profiles, read picks the stripe based 307 * on these policies. 308 */ 309 enum btrfs_read_policy { 310 /* Use process PID to choose the stripe */ 311 BTRFS_READ_POLICY_PID, 312 #ifdef CONFIG_BTRFS_EXPERIMENTAL 313 /* Balancing RAID1 reads across all striped devices (round-robin). */ 314 BTRFS_READ_POLICY_RR, 315 /* Read from a specific device. */ 316 BTRFS_READ_POLICY_DEVID, 317 #endif 318 BTRFS_NR_READ_POLICY, 319 }; 320 321 #ifdef CONFIG_BTRFS_EXPERIMENTAL 322 /* 323 * Checksum mode - offload it to workqueues or do it synchronously in 324 * btrfs_submit_chunk(). 325 */ 326 enum btrfs_offload_csum_mode { 327 /* 328 * Choose offloading checksum or do it synchronously automatically. 329 * Do it synchronously if the checksum is fast, or offload to workqueues 330 * otherwise. 331 */ 332 BTRFS_OFFLOAD_CSUM_AUTO, 333 /* Always offload checksum to workqueues. */ 334 BTRFS_OFFLOAD_CSUM_FORCE_ON, 335 /* Never offload checksum to workqueues. */ 336 BTRFS_OFFLOAD_CSUM_FORCE_OFF, 337 }; 338 #endif 339 340 struct btrfs_fs_devices { 341 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */ 342 343 /* 344 * UUID written into the btree blocks: 345 * 346 * - If metadata_uuid != fsid then super block must have 347 * BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag set. 348 * 349 * - Following shall be true at all times: 350 * - metadata_uuid == btrfs_header::fsid 351 * - metadata_uuid == btrfs_dev_item::fsid 352 * 353 * - Relations between fsid and metadata_uuid in sb and fs_devices: 354 * - Normal: 355 * fs_devices->fsid == fs_devices->metadata_uuid == sb->fsid 356 * sb->metadata_uuid == 0 357 * 358 * - When the BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag is set: 359 * fs_devices->fsid == sb->fsid 360 * fs_devices->metadata_uuid == sb->metadata_uuid 361 * 362 * - When in-memory fs_devices->temp_fsid is true 363 * fs_devices->fsid = random 364 * fs_devices->metadata_uuid == sb->fsid 365 */ 366 u8 metadata_uuid[BTRFS_FSID_SIZE]; 367 368 struct list_head fs_list; 369 370 /* 371 * Number of devices under this fsid including missing and 372 * replace-target device and excludes seed devices. 373 */ 374 u64 num_devices; 375 376 /* 377 * The number of devices that successfully opened, including 378 * replace-target, excludes seed devices. 379 */ 380 u64 open_devices; 381 382 /* The number of devices that are under the chunk allocation list. */ 383 u64 rw_devices; 384 385 /* Count of missing devices under this fsid excluding seed device. */ 386 u64 missing_devices; 387 u64 total_rw_bytes; 388 389 /* 390 * Count of devices from btrfs_super_block::num_devices for this fsid, 391 * which includes the seed device, excludes the transient replace-target 392 * device. 393 */ 394 u64 total_devices; 395 396 /* Highest generation number of seen devices */ 397 u64 latest_generation; 398 399 /* 400 * The mount device or a device with highest generation after removal 401 * or replace. 402 */ 403 struct btrfs_device *latest_dev; 404 405 /* 406 * All of the devices in the filesystem, protected by a mutex so we can 407 * safely walk it to write out the super blocks without worrying about 408 * adding/removing by the multi-device code. Scrubbing super block can 409 * kick off supers writing by holding this mutex lock. 410 */ 411 struct mutex device_list_mutex; 412 413 /* List of all devices, protected by device_list_mutex */ 414 struct list_head devices; 415 416 /* Devices which can satisfy space allocation. Protected by * chunk_mutex. */ 417 struct list_head alloc_list; 418 419 struct list_head seed_list; 420 421 /* Count fs-devices opened. */ 422 int opened; 423 424 /* 425 * Counter of the processes that are holding this fs_devices but not 426 * yet opened. 427 * This is for mounting handling, as we can only open the fs_devices 428 * after a super block is created. But we cannot take uuid_mutex 429 * during sget_fc(), thus we have to hold the fs_devices (meaning it 430 * cannot be released) until a super block is returned. 431 */ 432 int holding; 433 434 /* Set when we find or add a device that doesn't have the nonrot flag set. */ 435 bool rotating; 436 /* Devices support TRIM/discard commands. */ 437 bool discardable; 438 /* The filesystem is a seed filesystem. */ 439 bool seeding; 440 /* The mount needs to use a randomly generated fsid. */ 441 bool temp_fsid; 442 /* Enable/disable the filesystem stats tracking. */ 443 bool collect_fs_stats; 444 445 struct btrfs_fs_info *fs_info; 446 /* sysfs kobjects */ 447 struct kobject fsid_kobj; 448 struct kobject *devices_kobj; 449 struct kobject *devinfo_kobj; 450 struct completion kobj_unregister; 451 452 enum btrfs_chunk_allocation_policy chunk_alloc_policy; 453 454 /* Policy used to read the mirrored stripes. */ 455 enum btrfs_read_policy read_policy; 456 457 #ifdef CONFIG_BTRFS_EXPERIMENTAL 458 /* 459 * Minimum contiguous reads before switching to next device, the unit 460 * is one block/sectorsize. 461 */ 462 u32 rr_min_contig_read; 463 464 /* Device to be used for reading in case of RAID1. */ 465 u64 read_devid; 466 467 /* Checksum mode - offload it or do it synchronously. */ 468 enum btrfs_offload_csum_mode offload_csum_mode; 469 #endif 470 }; 471 472 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \ 473 - sizeof(struct btrfs_chunk)) \ 474 / sizeof(struct btrfs_stripe) + 1) 475 476 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \ 477 - 2 * sizeof(struct btrfs_disk_key) \ 478 - 2 * sizeof(struct btrfs_chunk)) \ 479 / sizeof(struct btrfs_stripe) + 1) 480 481 struct btrfs_io_stripe { 482 struct btrfs_device *dev; 483 /* Block mapping. */ 484 u64 physical; 485 bool rst_search_commit_root; 486 /* For the endio handler. */ 487 struct btrfs_io_context *bioc; 488 }; 489 490 struct btrfs_discard_stripe { 491 struct btrfs_device *dev; 492 u64 physical; 493 u64 length; 494 }; 495 496 /* 497 * Context for IO submission for device stripe. 498 * 499 * - Track the unfinished mirrors for mirror based profiles 500 * Mirror based profiles are SINGLE/DUP/RAID1/RAID10. 501 * 502 * - Contain the logical -> physical mapping info 503 * Used by submit_stripe_bio() for mapping logical bio 504 * into physical device address. 505 * 506 * - Contain device replace info 507 * Used by handle_ops_on_dev_replace() to copy logical bios 508 * into the new device. 509 * 510 * - Contain RAID56 full stripe logical bytenrs 511 */ 512 struct btrfs_io_context { 513 refcount_t refs; 514 struct btrfs_fs_info *fs_info; 515 /* Taken from struct btrfs_chunk_map::type. */ 516 u64 map_type; 517 struct bio *orig_bio; 518 atomic_t error; 519 u16 max_errors; 520 bool use_rst; 521 522 u64 logical; 523 u64 size; 524 /* Raid stripe tree ordered entry. */ 525 struct list_head rst_ordered_entry; 526 527 /* 528 * The total number of stripes, including the extra duplicated 529 * stripe for replace. 530 */ 531 u16 num_stripes; 532 533 /* 534 * The mirror_num of this bioc. 535 * 536 * This is for reads which use 0 as mirror_num, thus we should return a 537 * valid mirror_num (>0) for the reader. 538 */ 539 u16 mirror_num; 540 541 /* 542 * The following two members are for dev-replace case only. 543 * 544 * @replace_nr_stripes: Number of duplicated stripes which need to be 545 * written to replace target. 546 * Should be <= 2 (2 for DUP, otherwise <= 1). 547 * @replace_stripe_src: The array indicates where the duplicated stripes 548 * are from. 549 * 550 * The @replace_stripe_src[] array is mostly for RAID56 cases. 551 * As non-RAID56 stripes share the same contents of the mapped range, 552 * thus no need to bother where the duplicated ones are from. 553 * 554 * But for RAID56 case, all stripes contain different contents, thus 555 * we need a way to know the mapping. 556 * 557 * There is an example for the two members, using a RAID5 write: 558 * 559 * num_stripes: 4 (3 + 1 duplicated write) 560 * stripes[0]: dev = devid 1, physical = X 561 * stripes[1]: dev = devid 2, physical = Y 562 * stripes[2]: dev = devid 3, physical = Z 563 * stripes[3]: dev = devid 0, physical = Y 564 * 565 * replace_nr_stripes = 1 566 * replace_stripe_src = 1 <- Means stripes[1] is involved in replace. 567 * The duplicated stripe index would be 568 * (@num_stripes - 1). 569 * 570 * Note, that we can still have cases replace_nr_stripes = 2 for DUP. 571 * In that case, all stripes share the same content, thus we don't 572 * need to bother @replace_stripe_src value at all. 573 */ 574 u16 replace_nr_stripes; 575 s16 replace_stripe_src; 576 /* 577 * Logical bytenr of the full stripe start, only for RAID56 cases. 578 * 579 * When this value is set to other than (u64)-1, the stripes[] should 580 * follow this pattern: 581 * 582 * (real_stripes = num_stripes - replace_nr_stripes) 583 * (data_stripes = (is_raid6) ? (real_stripes - 2) : (real_stripes - 1)) 584 * 585 * stripes[0]: The first data stripe 586 * stripes[1]: The second data stripe 587 * ... 588 * stripes[data_stripes - 1]: The last data stripe 589 * stripes[data_stripes]: The P stripe 590 * stripes[data_stripes + 1]: The Q stripe (only for RAID6). 591 */ 592 u64 full_stripe_logical; 593 struct btrfs_io_stripe stripes[]; 594 }; 595 596 struct btrfs_device_info { 597 struct btrfs_device *dev; 598 u64 dev_offset; 599 u64 max_avail; 600 u64 total_avail; 601 }; 602 603 struct btrfs_raid_attr { 604 u8 sub_stripes; /* sub_stripes info for map */ 605 u8 dev_stripes; /* stripes per dev */ 606 u8 devs_max; /* max devs to use */ 607 u8 devs_min; /* min devs needed */ 608 u8 tolerated_failures; /* max tolerated fail devs */ 609 u8 devs_increment; /* ndevs has to be a multiple of this */ 610 u8 ncopies; /* how many copies to data has */ 611 u8 nparity; /* number of stripes worth of bytes to store 612 * parity information */ 613 u8 mindev_error; /* error code if min devs requisite is unmet */ 614 const char raid_name[8]; /* name of the raid */ 615 u64 bg_flag; /* block group flag of the raid */ 616 }; 617 618 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES]; 619 620 struct btrfs_chunk_map { 621 struct rb_node rb_node; 622 /* For mount time dev extent verification. */ 623 int verified_stripes; 624 refcount_t refs; 625 u64 start; 626 u64 chunk_len; 627 u64 stripe_size; 628 u64 type; 629 int io_align; 630 int io_width; 631 int num_stripes; 632 int sub_stripes; 633 struct btrfs_io_stripe stripes[]; 634 }; 635 636 #define btrfs_chunk_map_size(n) (sizeof(struct btrfs_chunk_map) + \ 637 (sizeof(struct btrfs_io_stripe) * (n))) 638 639 static inline void btrfs_free_chunk_map(struct btrfs_chunk_map *map) 640 { 641 if (map && refcount_dec_and_test(&map->refs)) { 642 ASSERT(RB_EMPTY_NODE(&map->rb_node)); 643 kfree(map); 644 } 645 } 646 647 struct btrfs_balance_control { 648 struct btrfs_balance_args data; 649 struct btrfs_balance_args meta; 650 struct btrfs_balance_args sys; 651 652 u64 flags; 653 654 struct btrfs_balance_progress stat; 655 }; 656 657 /* 658 * Search for a given device by the set parameters 659 */ 660 struct btrfs_dev_lookup_args { 661 u64 devid; 662 u8 *uuid; 663 u8 *fsid; 664 /* 665 * If devt is specified, all other members will be ignored as it is 666 * enough to uniquely locate a device. 667 */ 668 dev_t devt; 669 bool missing; 670 }; 671 672 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */ 673 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 } 674 675 #define BTRFS_DEV_LOOKUP_ARGS(name) \ 676 struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT 677 678 enum btrfs_map_op { 679 BTRFS_MAP_READ, 680 BTRFS_MAP_WRITE, 681 BTRFS_MAP_GET_READ_MIRRORS, 682 }; 683 684 static inline enum btrfs_map_op btrfs_op(const struct bio *bio) 685 { 686 switch (bio_op(bio)) { 687 case REQ_OP_WRITE: 688 case REQ_OP_ZONE_APPEND: 689 return BTRFS_MAP_WRITE; 690 default: 691 WARN_ON_ONCE(1); 692 fallthrough; 693 case REQ_OP_READ: 694 return BTRFS_MAP_READ; 695 } 696 } 697 698 static inline unsigned long btrfs_chunk_item_size(int num_stripes) 699 { 700 ASSERT(num_stripes); 701 return sizeof(struct btrfs_chunk) + 702 sizeof(struct btrfs_stripe) * (num_stripes - 1); 703 } 704 705 /* 706 * Do the type safe conversion from stripe_nr to offset inside the chunk. 707 * 708 * @stripe_nr is u32, with left shift it can overflow u32 for chunks larger 709 * than 4G. This does the proper type cast to avoid overflow. 710 */ 711 static inline u64 btrfs_stripe_nr_to_offset(u32 stripe_nr) 712 { 713 return (u64)stripe_nr << BTRFS_STRIPE_LEN_SHIFT; 714 } 715 716 void btrfs_get_bioc(struct btrfs_io_context *bioc); 717 void btrfs_put_bioc(struct btrfs_io_context *bioc); 718 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, 719 u64 logical, u64 *length, 720 struct btrfs_io_context **bioc_ret, 721 struct btrfs_io_stripe *smap, int *mirror_num_ret); 722 int btrfs_map_repair_block(struct btrfs_fs_info *fs_info, 723 struct btrfs_io_stripe *smap, u64 logical, 724 u32 length, int mirror_num); 725 struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info, 726 u64 logical, u64 *length_ret, 727 u32 *num_stripes); 728 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info); 729 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info); 730 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans, 731 struct btrfs_space_info *space_info, 732 u64 type); 733 void btrfs_mapping_tree_free(struct btrfs_fs_info *fs_info); 734 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, 735 blk_mode_t flags, void *holder); 736 struct btrfs_device *btrfs_scan_one_device(const char *path, bool mount_arg_dev); 737 int btrfs_forget_devices(dev_t devt); 738 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices); 739 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices); 740 void btrfs_assign_next_active_device(struct btrfs_device *device, 741 struct btrfs_device *this_dev); 742 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info, 743 u64 devid, 744 const char *devpath); 745 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info, 746 struct btrfs_dev_lookup_args *args, 747 const char *path); 748 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info, 749 const u64 *devid, const u8 *uuid, 750 const char *path); 751 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args); 752 int btrfs_rm_device(struct btrfs_fs_info *fs_info, 753 struct btrfs_dev_lookup_args *args, 754 struct file **bdev_file); 755 void __exit btrfs_cleanup_fs_uuids(void); 756 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len); 757 int btrfs_grow_device(struct btrfs_trans_handle *trans, 758 struct btrfs_device *device, u64 new_size); 759 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices, 760 const struct btrfs_dev_lookup_args *args); 761 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); 762 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path); 763 int btrfs_balance(struct btrfs_fs_info *fs_info, 764 struct btrfs_balance_control *bctl, 765 struct btrfs_ioctl_balance_args *bargs); 766 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf); 767 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info); 768 int btrfs_recover_balance(struct btrfs_fs_info *fs_info); 769 int btrfs_pause_balance(struct btrfs_fs_info *fs_info); 770 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset, 771 bool verbose); 772 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info); 773 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset); 774 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index); 775 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info, 776 struct btrfs_ioctl_get_dev_stats *stats); 777 int btrfs_init_devices_late(struct btrfs_fs_info *fs_info); 778 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info); 779 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans); 780 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev); 781 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev); 782 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev); 783 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info, 784 u64 logical); 785 u64 btrfs_calc_stripe_length(const struct btrfs_chunk_map *map); 786 int btrfs_nr_parity_stripes(u64 type); 787 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans, 788 struct btrfs_block_group *bg); 789 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset); 790 791 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 792 struct btrfs_chunk_map *btrfs_alloc_chunk_map(int num_stripes, gfp_t gfp); 793 int btrfs_add_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map); 794 #endif 795 796 struct btrfs_chunk_map *btrfs_find_chunk_map(struct btrfs_fs_info *fs_info, 797 u64 logical, u64 length); 798 struct btrfs_chunk_map *btrfs_find_chunk_map_nolock(struct btrfs_fs_info *fs_info, 799 u64 logical, u64 length); 800 struct btrfs_chunk_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info, 801 u64 logical, u64 length); 802 void btrfs_remove_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map); 803 struct btrfs_super_block *btrfs_read_disk_super(struct block_device *bdev, 804 int copy_num, bool drop_cache); 805 void btrfs_release_disk_super(struct btrfs_super_block *super); 806 807 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev, 808 int index) 809 { 810 atomic_inc(dev->dev_stat_values + index); 811 /* 812 * This memory barrier orders stores updating statistics before stores 813 * updating dev_stats_ccnt. 814 * 815 * It pairs with smp_rmb() in btrfs_run_dev_stats(). 816 */ 817 smp_mb__before_atomic(); 818 atomic_inc(&dev->dev_stats_ccnt); 819 } 820 821 static inline int btrfs_dev_stat_read(struct btrfs_device *dev, 822 int index) 823 { 824 return atomic_read(dev->dev_stat_values + index); 825 } 826 827 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev, 828 int index) 829 { 830 int ret; 831 832 ret = atomic_xchg(dev->dev_stat_values + index, 0); 833 /* 834 * atomic_xchg implies a full memory barriers as per atomic_t.txt: 835 * - RMW operations that have a return value are fully ordered; 836 * 837 * This implicit memory barriers is paired with the smp_rmb in 838 * btrfs_run_dev_stats 839 */ 840 atomic_inc(&dev->dev_stats_ccnt); 841 return ret; 842 } 843 844 static inline void btrfs_dev_stat_set(struct btrfs_device *dev, 845 int index, unsigned long val) 846 { 847 atomic_set(dev->dev_stat_values + index, val); 848 /* 849 * This memory barrier orders stores updating statistics before stores 850 * updating dev_stats_ccnt. 851 * 852 * It pairs with smp_rmb() in btrfs_run_dev_stats(). 853 */ 854 smp_mb__before_atomic(); 855 atomic_inc(&dev->dev_stats_ccnt); 856 } 857 858 static inline const char *btrfs_dev_name(const struct btrfs_device *device) 859 { 860 if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state)) 861 return "<missing disk>"; 862 else 863 return rcu_dereference(device->name); 864 } 865 866 static inline void btrfs_warn_unknown_chunk_allocation(enum btrfs_chunk_allocation_policy pol) 867 { 868 WARN_ONCE(1, "unknown allocation policy %d, fallback to regular", pol); 869 } 870 871 static inline void btrfs_fs_devices_inc_holding(struct btrfs_fs_devices *fs_devices) 872 { 873 lockdep_assert_held(&uuid_mutex); 874 ASSERT(fs_devices->holding >= 0); 875 fs_devices->holding++; 876 } 877 878 static inline void btrfs_fs_devices_dec_holding(struct btrfs_fs_devices *fs_devices) 879 { 880 lockdep_assert_held(&uuid_mutex); 881 ASSERT(fs_devices->holding > 0); 882 fs_devices->holding--; 883 } 884 885 void btrfs_commit_device_sizes(struct btrfs_transaction *trans); 886 887 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void); 888 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info, 889 struct btrfs_device *failing_dev); 890 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, struct btrfs_device *device); 891 892 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags); 893 int btrfs_bg_type_to_factor(u64 flags); 894 const char *btrfs_bg_type_to_raid_name(u64 flags); 895 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info); 896 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical); 897 898 bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr); 899 const u8 *btrfs_sb_fsid_ptr(const struct btrfs_super_block *sb); 900 901 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 902 struct btrfs_io_context *alloc_btrfs_io_context(struct btrfs_fs_info *fs_info, 903 u64 logical, u16 total_stripes); 904 #endif 905 906 #endif 907