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 #include "fs.h" 26 27 struct btrfs_block_group; 28 struct btrfs_trans_handle; 29 struct btrfs_transaction; 30 struct btrfs_zoned_device_info; 31 struct btrfs_space_info; 32 33 #define BTRFS_MAX_DATA_CHUNK_SIZE (10ULL * SZ_1G) 34 35 /* 36 * Arbitrary maximum size of one discard request to limit potentially long time 37 * spent in blkdev_issue_discard(). 38 */ 39 #define BTRFS_MAX_DISCARD_CHUNK_SIZE (SZ_1G) 40 41 extern struct mutex uuid_mutex; 42 43 #define BTRFS_STRIPE_LEN SZ_64K 44 #define BTRFS_STRIPE_LEN_SHIFT (16) 45 #define BTRFS_STRIPE_LEN_MASK (BTRFS_STRIPE_LEN - 1) 46 47 static_assert(ilog2(BTRFS_STRIPE_LEN) == BTRFS_STRIPE_LEN_SHIFT); 48 49 /* Used by sanity check for btrfs_raid_types. */ 50 #define const_ffs(n) (__builtin_ctzll(n) + 1) 51 52 /* 53 * The conversion from BTRFS_BLOCK_GROUP_* bits to btrfs_raid_type requires 54 * RAID0 always to be the lowest profile bit. 55 * Although it's part of on-disk format and should never change, do extra 56 * compile-time sanity checks. 57 */ 58 static_assert(const_ffs(BTRFS_BLOCK_GROUP_RAID0) < 59 const_ffs(BTRFS_BLOCK_GROUP_PROFILE_MASK & ~BTRFS_BLOCK_GROUP_RAID0)); 60 61 /* ilog2() can handle both constants and variables */ 62 #define BTRFS_BG_FLAG_TO_INDEX(profile) \ 63 ilog2((profile) >> (ilog2(BTRFS_BLOCK_GROUP_RAID0) - 1)) 64 65 enum btrfs_raid_types { 66 /* SINGLE is the special one as it doesn't have on-disk bit. */ 67 BTRFS_RAID_SINGLE = 0, 68 69 BTRFS_RAID_RAID0 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID0), 70 BTRFS_RAID_RAID1 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1), 71 BTRFS_RAID_DUP = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_DUP), 72 BTRFS_RAID_RAID10 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID10), 73 BTRFS_RAID_RAID5 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID5), 74 BTRFS_RAID_RAID6 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID6), 75 BTRFS_RAID_RAID1C3 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C3), 76 BTRFS_RAID_RAID1C4 = BTRFS_BG_FLAG_TO_INDEX(BTRFS_BLOCK_GROUP_RAID1C4), 77 78 BTRFS_NR_RAID_TYPES 79 }; 80 81 static_assert(BTRFS_RAID_RAID0 == 1); 82 static_assert(BTRFS_RAID_RAID1 == 2); 83 static_assert(BTRFS_RAID_DUP == 3); 84 static_assert(BTRFS_RAID_RAID10 == 4); 85 static_assert(BTRFS_RAID_RAID5 == 5); 86 static_assert(BTRFS_RAID_RAID6 == 6); 87 static_assert(BTRFS_RAID_RAID1C3 == 7); 88 static_assert(BTRFS_RAID_RAID1C4 == 8); 89 90 /* 91 * Use sequence counter to get consistent device stat data on 92 * 32-bit processors. 93 */ 94 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 95 #include <linux/seqlock.h> 96 #define __BTRFS_NEED_DEVICE_DATA_ORDERED 97 #define btrfs_device_data_ordered_init(device) \ 98 seqcount_init(&device->data_seqcount) 99 #else 100 #define btrfs_device_data_ordered_init(device) do { } while (0) 101 #endif 102 103 #define BTRFS_DEV_STATE_WRITEABLE (0) 104 #define BTRFS_DEV_STATE_IN_FS_METADATA (1) 105 #define BTRFS_DEV_STATE_MISSING (2) 106 #define BTRFS_DEV_STATE_REPLACE_TGT (3) 107 #define BTRFS_DEV_STATE_FLUSH_SENT (4) 108 #define BTRFS_DEV_STATE_NO_READA (5) 109 #define BTRFS_DEV_STATE_FLUSH_FAILED (6) 110 111 /* Set when the device item is found in chunk tree, used to catch unexpected registered device. */ 112 #define BTRFS_DEV_STATE_ITEM_FOUND (7) 113 114 /* Special value encoding failure to write primary super block. */ 115 #define BTRFS_SUPER_PRIMARY_WRITE_ERROR (INT_MAX / 2) 116 117 struct btrfs_fs_devices; 118 119 struct btrfs_device { 120 struct list_head dev_list; /* device_list_mutex */ 121 struct list_head dev_alloc_list; /* chunk mutex */ 122 struct list_head post_commit_list; /* chunk mutex */ 123 struct btrfs_fs_devices *fs_devices; 124 struct btrfs_fs_info *fs_info; 125 126 /* Device path or NULL if missing. */ 127 const char __rcu *name; 128 129 u64 generation; 130 131 struct file *bdev_file; 132 struct block_device *bdev; 133 134 struct btrfs_zoned_device_info *zone_info; 135 136 unsigned long dev_state; 137 138 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED 139 seqcount_t data_seqcount; 140 #endif 141 142 /* the internal btrfs device id */ 143 u64 devid; 144 145 /* size of the device in memory */ 146 u64 total_bytes; 147 148 /* size of the device on disk */ 149 u64 disk_total_bytes; 150 151 /* bytes used */ 152 u64 bytes_used; 153 154 /* optimal io alignment for this device */ 155 u32 io_align; 156 157 /* optimal io width for this device */ 158 u32 io_width; 159 /* type and info about this device */ 160 u64 type; 161 162 /* 163 * Counter of super block write errors, values larger than 164 * BTRFS_SUPER_PRIMARY_WRITE_ERROR encode primary super block write failure. 165 */ 166 atomic_t sb_write_errors; 167 168 /* minimal io size for this device */ 169 u32 sector_size; 170 171 /* physical drive uuid (or lvm uuid) */ 172 u8 uuid[BTRFS_UUID_SIZE]; 173 174 /* 175 * size of the device on the current transaction 176 * 177 * This variant is update when committing the transaction, 178 * and protected by chunk mutex 179 */ 180 u64 commit_total_bytes; 181 182 /* bytes used on the current transaction */ 183 u64 commit_bytes_used; 184 185 /* Bio used for flushing device barriers */ 186 struct bio flush_bio; 187 struct completion flush_wait; 188 189 /* per-device scrub information */ 190 struct scrub_ctx *scrub_ctx; 191 192 /* disk I/O failure stats. For detailed description refer to 193 * enum btrfs_dev_stat_values in ioctl.h */ 194 int dev_stats_valid; 195 196 /* Counter to record the change of device stats */ 197 atomic_t dev_stats_ccnt; 198 atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX]; 199 200 /* 201 * Device's major-minor number. Must be set even if the device is not 202 * opened (bdev == NULL), unless the device is missing. 203 */ 204 dev_t devt; 205 206 struct extent_io_tree alloc_state; 207 208 struct completion kobj_unregister; 209 /* For sysfs/FSID/devinfo/devid/ */ 210 struct kobject devid_kobj; 211 212 /* Bandwidth limit for scrub, in bytes */ 213 u64 scrub_speed_max; 214 215 /* 216 * A temporary number of allocated space during per-profile 217 * available space calculation. 218 */ 219 u64 per_profile_allocated; 220 }; 221 222 /* 223 * Block group or device which contains an active swapfile. Used for preventing 224 * unsafe operations while a swapfile is active. 225 * 226 * These are sorted on (ptr, inode) (note that a block group or device can 227 * contain more than one swapfile). We compare the pointer values because we 228 * don't actually care what the object is, we just need a quick check whether 229 * the object exists in the rbtree. 230 */ 231 struct btrfs_swapfile_pin { 232 struct rb_node node; 233 void *ptr; 234 struct inode *inode; 235 /* 236 * If true, ptr points to a struct btrfs_block_group. Otherwise, ptr 237 * points to a struct btrfs_device. 238 */ 239 bool is_block_group; 240 /* 241 * Only used when 'is_block_group' is true and it is the number of 242 * extents used by a swapfile for this block group ('ptr' field). 243 */ 244 int bg_extent_count; 245 }; 246 247 /* 248 * If we read those variants at the context of their own lock, we needn't 249 * use the following helpers, reading them directly is safe. 250 */ 251 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 252 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 253 static inline u64 \ 254 btrfs_device_get_##name(const struct btrfs_device *dev) \ 255 { \ 256 u64 size; \ 257 unsigned int seq; \ 258 \ 259 do { \ 260 seq = read_seqcount_begin(&dev->data_seqcount); \ 261 size = dev->name; \ 262 } while (read_seqcount_retry(&dev->data_seqcount, seq)); \ 263 return size; \ 264 } \ 265 \ 266 static inline void \ 267 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 268 { \ 269 preempt_disable(); \ 270 write_seqcount_begin(&dev->data_seqcount); \ 271 dev->name = size; \ 272 write_seqcount_end(&dev->data_seqcount); \ 273 preempt_enable(); \ 274 } 275 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 276 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 277 static inline u64 \ 278 btrfs_device_get_##name(const struct btrfs_device *dev) \ 279 { \ 280 u64 size; \ 281 \ 282 preempt_disable(); \ 283 size = dev->name; \ 284 preempt_enable(); \ 285 return size; \ 286 } \ 287 \ 288 static inline void \ 289 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 290 { \ 291 preempt_disable(); \ 292 dev->name = size; \ 293 preempt_enable(); \ 294 } 295 #else 296 #define BTRFS_DEVICE_GETSET_FUNCS(name) \ 297 static inline u64 \ 298 btrfs_device_get_##name(const struct btrfs_device *dev) \ 299 { \ 300 return dev->name; \ 301 } \ 302 \ 303 static inline void \ 304 btrfs_device_set_##name(struct btrfs_device *dev, u64 size) \ 305 { \ 306 dev->name = size; \ 307 } 308 #endif 309 310 BTRFS_DEVICE_GETSET_FUNCS(total_bytes); 311 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes); 312 BTRFS_DEVICE_GETSET_FUNCS(bytes_used); 313 314 enum btrfs_chunk_allocation_policy { 315 BTRFS_CHUNK_ALLOC_REGULAR, 316 BTRFS_CHUNK_ALLOC_ZONED, 317 }; 318 319 #define BTRFS_DEFAULT_RR_MIN_CONTIG_READ (SZ_256K) 320 /* Keep in sync with raid_attr table, current maximum is RAID1C4. */ 321 #define BTRFS_RAID1_MAX_MIRRORS (4) 322 /* 323 * Read policies for mirrored block group profiles, read picks the stripe based 324 * on these policies. 325 */ 326 enum btrfs_read_policy { 327 /* Use process PID to choose the stripe */ 328 BTRFS_READ_POLICY_PID, 329 #ifdef CONFIG_BTRFS_EXPERIMENTAL 330 /* Balancing RAID1 reads across all striped devices (round-robin). */ 331 BTRFS_READ_POLICY_RR, 332 /* Read from a specific device. */ 333 BTRFS_READ_POLICY_DEVID, 334 #endif 335 BTRFS_NR_READ_POLICY, 336 }; 337 338 struct btrfs_fs_devices { 339 u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */ 340 341 /* 342 * UUID written into the btree blocks: 343 * 344 * - If metadata_uuid != fsid then super block must have 345 * BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag set. 346 * 347 * - Following shall be true at all times: 348 * - metadata_uuid == btrfs_header::fsid 349 * - metadata_uuid == btrfs_dev_item::fsid 350 * 351 * - Relations between fsid and metadata_uuid in sb and fs_devices: 352 * - Normal: 353 * fs_devices->fsid == fs_devices->metadata_uuid == sb->fsid 354 * sb->metadata_uuid == 0 355 * 356 * - When the BTRFS_FEATURE_INCOMPAT_METADATA_UUID flag is set: 357 * fs_devices->fsid == sb->fsid 358 * fs_devices->metadata_uuid == sb->metadata_uuid 359 * 360 * - When in-memory fs_devices->temp_fsid is true 361 * fs_devices->fsid = random 362 * fs_devices->metadata_uuid == sb->fsid 363 */ 364 u8 metadata_uuid[BTRFS_FSID_SIZE]; 365 366 struct list_head fs_list; 367 368 /* 369 * Number of devices under this fsid including missing and 370 * replace-target device and excludes seed devices. 371 */ 372 u64 num_devices; 373 374 /* 375 * The number of devices that successfully opened, including 376 * replace-target, excludes seed devices. 377 */ 378 u64 open_devices; 379 380 /* The number of devices that are under the chunk allocation list. */ 381 u64 rw_devices; 382 383 /* Count of missing devices under this fsid excluding seed device. */ 384 u64 missing_devices; 385 u64 total_rw_bytes; 386 387 /* 388 * Count of devices from btrfs_super_block::num_devices for this fsid, 389 * which includes the seed device, excludes the transient replace-target 390 * device. 391 */ 392 u64 total_devices; 393 394 /* Highest generation number of seen devices */ 395 u64 latest_generation; 396 397 /* 398 * The mount device or a device with highest generation after removal 399 * or replace. 400 */ 401 struct btrfs_device *latest_dev; 402 403 /* 404 * All of the devices in the filesystem, protected by a mutex so we can 405 * safely walk it to write out the super blocks without worrying about 406 * adding/removing by the multi-device code. Scrubbing super block can 407 * kick off supers writing by holding this mutex lock. 408 */ 409 struct mutex device_list_mutex; 410 411 /* List of all devices, protected by device_list_mutex */ 412 struct list_head devices; 413 414 /* Devices which can satisfy space allocation. Protected by * chunk_mutex. */ 415 struct list_head alloc_list; 416 417 struct list_head seed_list; 418 419 /* Count fs-devices opened. */ 420 int opened; 421 422 /* 423 * Counter of the processes that are holding this fs_devices but not 424 * yet opened. 425 * This is for mounting handling, as we can only open the fs_devices 426 * after a super block is created. But we cannot take uuid_mutex 427 * during sget_fc(), thus we have to hold the fs_devices (meaning it 428 * cannot be released) until a super block is returned. 429 */ 430 int holding; 431 432 /* Set when we find or add a device that doesn't have the nonrot flag set. */ 433 bool rotating; 434 /* Devices support TRIM/discard commands. */ 435 bool discardable; 436 /* The filesystem is a seed filesystem. */ 437 bool seeding; 438 /* The mount needs to use a randomly generated fsid. */ 439 bool temp_fsid; 440 /* Enable/disable the filesystem stats tracking. */ 441 bool collect_fs_stats; 442 443 struct btrfs_fs_info *fs_info; 444 /* sysfs kobjects */ 445 struct kobject fsid_kobj; 446 struct kobject *devices_kobj; 447 struct kobject *devinfo_kobj; 448 struct completion kobj_unregister; 449 450 enum btrfs_chunk_allocation_policy chunk_alloc_policy; 451 452 /* Policy used to read the mirrored stripes. */ 453 enum btrfs_read_policy read_policy; 454 455 #ifdef CONFIG_BTRFS_EXPERIMENTAL 456 /* 457 * Minimum contiguous reads before switching to next device, the unit 458 * is one block/sectorsize. 459 */ 460 u32 rr_min_contig_read; 461 462 /* Device to be used for reading in case of RAID1. */ 463 u64 read_devid; 464 #endif 465 466 /* 467 * Each value indicates the available space for that profile. 468 * U64_MAX means the estimation is unavailable. 469 * 470 * Protected by per_profile_lock; 471 */ 472 u64 per_profile_avail[BTRFS_NR_RAID_TYPES]; 473 spinlock_t per_profile_lock; 474 }; 475 476 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info) \ 477 - sizeof(struct btrfs_chunk)) \ 478 / sizeof(struct btrfs_stripe) + 1) 479 480 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE \ 481 - 2 * sizeof(struct btrfs_disk_key) \ 482 - 2 * sizeof(struct btrfs_chunk)) \ 483 / sizeof(struct btrfs_stripe) + 1) 484 485 struct btrfs_io_stripe { 486 struct btrfs_device *dev; 487 /* Block mapping. */ 488 u64 physical; 489 bool rst_search_commit_root; 490 /* For the endio handler. */ 491 struct btrfs_io_context *bioc; 492 }; 493 494 struct btrfs_discard_stripe { 495 struct btrfs_device *dev; 496 u64 physical; 497 u64 length; 498 }; 499 500 /* 501 * Context for IO submission for device stripe. 502 * 503 * - Track the unfinished mirrors for mirror based profiles 504 * Mirror based profiles are SINGLE/DUP/RAID1/RAID10. 505 * 506 * - Contain the logical -> physical mapping info 507 * Used by submit_stripe_bio() for mapping logical bio 508 * into physical device address. 509 * 510 * - Contain device replace info 511 * Used by handle_ops_on_dev_replace() to copy logical bios 512 * into the new device. 513 * 514 * - Contain RAID56 full stripe logical bytenrs 515 */ 516 struct btrfs_io_context { 517 refcount_t refs; 518 struct btrfs_fs_info *fs_info; 519 /* Taken from struct btrfs_chunk_map::type. */ 520 u64 map_type; 521 struct bio *orig_bio; 522 atomic_t error; 523 u16 max_errors; 524 bool use_rst; 525 526 u64 logical; 527 u64 size; 528 /* Raid stripe tree ordered entry. */ 529 struct list_head rst_ordered_entry; 530 531 /* 532 * The total number of stripes, including the extra duplicated 533 * stripe for replace. 534 */ 535 u16 num_stripes; 536 537 /* 538 * The mirror_num of this bioc. 539 * 540 * This is for reads which use 0 as mirror_num, thus we should return a 541 * valid mirror_num (>0) for the reader. 542 */ 543 u16 mirror_num; 544 545 /* 546 * The following two members are for dev-replace case only. 547 * 548 * @replace_nr_stripes: Number of duplicated stripes which need to be 549 * written to replace target. 550 * Should be <= 2 (2 for DUP, otherwise <= 1). 551 * @replace_stripe_src: The array indicates where the duplicated stripes 552 * are from. 553 * 554 * The @replace_stripe_src[] array is mostly for RAID56 cases. 555 * As non-RAID56 stripes share the same contents of the mapped range, 556 * thus no need to bother where the duplicated ones are from. 557 * 558 * But for RAID56 case, all stripes contain different contents, thus 559 * we need a way to know the mapping. 560 * 561 * There is an example for the two members, using a RAID5 write: 562 * 563 * num_stripes: 4 (3 + 1 duplicated write) 564 * stripes[0]: dev = devid 1, physical = X 565 * stripes[1]: dev = devid 2, physical = Y 566 * stripes[2]: dev = devid 3, physical = Z 567 * stripes[3]: dev = devid 0, physical = Y 568 * 569 * replace_nr_stripes = 1 570 * replace_stripe_src = 1 <- Means stripes[1] is involved in replace. 571 * The duplicated stripe index would be 572 * (@num_stripes - 1). 573 * 574 * Note, that we can still have cases replace_nr_stripes = 2 for DUP. 575 * In that case, all stripes share the same content, thus we don't 576 * need to bother @replace_stripe_src value at all. 577 */ 578 u16 replace_nr_stripes; 579 s16 replace_stripe_src; 580 /* 581 * Logical bytenr of the full stripe start, only for RAID56 cases. 582 * 583 * When this value is set to other than (u64)-1, the stripes[] should 584 * follow this pattern: 585 * 586 * (real_stripes = num_stripes - replace_nr_stripes) 587 * (data_stripes = (is_raid6) ? (real_stripes - 2) : (real_stripes - 1)) 588 * 589 * stripes[0]: The first data stripe 590 * stripes[1]: The second data stripe 591 * ... 592 * stripes[data_stripes - 1]: The last data stripe 593 * stripes[data_stripes]: The P stripe 594 * stripes[data_stripes + 1]: The Q stripe (only for RAID6). 595 */ 596 u64 full_stripe_logical; 597 struct btrfs_io_stripe stripes[]; 598 }; 599 600 struct btrfs_device_info { 601 struct btrfs_device *dev; 602 u64 dev_offset; 603 u64 max_avail; 604 u64 total_avail; 605 }; 606 607 struct btrfs_raid_attr { 608 u8 sub_stripes; /* sub_stripes info for map */ 609 u8 dev_stripes; /* stripes per dev */ 610 u8 devs_max; /* max devs to use */ 611 u8 devs_min; /* min devs needed */ 612 u8 tolerated_failures; /* max tolerated fail devs */ 613 u8 devs_increment; /* ndevs has to be a multiple of this */ 614 u8 ncopies; /* how many copies to data has */ 615 u8 nparity; /* number of stripes worth of bytes to store 616 * parity information */ 617 u8 mindev_error; /* error code if min devs requisite is unmet */ 618 const char raid_name[8]; /* name of the raid */ 619 u64 bg_flag; /* block group flag of the raid */ 620 }; 621 622 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES]; 623 624 struct btrfs_chunk_map { 625 struct rb_node rb_node; 626 /* For mount time dev extent verification. */ 627 int verified_stripes; 628 refcount_t refs; 629 u64 start; 630 u64 chunk_len; 631 u64 stripe_size; 632 /* 633 * The real type that is utilized during logical address mapping. 634 * 635 * For most profiles it matches @on_disk_type, but for single-data-RAID56, 636 * the real type will be set to RAID1/RAID1C3, to avoid unsupported 637 * operations from raid56 lib. 638 */ 639 u64 type; 640 u64 on_disk_type; 641 int num_stripes; 642 int sub_stripes; 643 struct btrfs_io_stripe stripes[]; 644 }; 645 646 #define btrfs_chunk_map_size(n) (sizeof(struct btrfs_chunk_map) + \ 647 (sizeof(struct btrfs_io_stripe) * (n))) 648 649 static inline void btrfs_free_chunk_map(struct btrfs_chunk_map *map) 650 { 651 if (map && refcount_dec_and_test(&map->refs)) { 652 ASSERT(RB_EMPTY_NODE(&map->rb_node)); 653 kfree(map); 654 } 655 } 656 DEFINE_FREE(btrfs_free_chunk_map, struct btrfs_chunk_map *, btrfs_free_chunk_map(_T)) 657 658 struct btrfs_balance_control { 659 struct btrfs_balance_args data; 660 struct btrfs_balance_args meta; 661 struct btrfs_balance_args sys; 662 663 u64 flags; 664 665 struct btrfs_balance_progress stat; 666 }; 667 668 /* 669 * Search for a given device by the set parameters 670 */ 671 struct btrfs_dev_lookup_args { 672 u64 devid; 673 u8 *uuid; 674 u8 *fsid; 675 /* 676 * If devt is specified, all other members will be ignored as it is 677 * enough to uniquely locate a device. 678 */ 679 dev_t devt; 680 bool missing; 681 }; 682 683 /* We have to initialize to -1 because BTRFS_DEV_REPLACE_DEVID is 0 */ 684 #define BTRFS_DEV_LOOKUP_ARGS_INIT { .devid = (u64)-1 } 685 686 #define BTRFS_DEV_LOOKUP_ARGS(name) \ 687 struct btrfs_dev_lookup_args name = BTRFS_DEV_LOOKUP_ARGS_INIT 688 689 enum btrfs_map_op { 690 BTRFS_MAP_READ, 691 BTRFS_MAP_WRITE, 692 BTRFS_MAP_GET_READ_MIRRORS, 693 }; 694 695 static inline enum btrfs_map_op btrfs_op(const struct bio *bio) 696 { 697 switch (bio_op(bio)) { 698 case REQ_OP_WRITE: 699 case REQ_OP_ZONE_APPEND: 700 return BTRFS_MAP_WRITE; 701 default: 702 WARN_ON_ONCE(1); 703 fallthrough; 704 case REQ_OP_READ: 705 return BTRFS_MAP_READ; 706 } 707 } 708 709 static inline unsigned long btrfs_chunk_item_size(int num_stripes) 710 { 711 ASSERT(num_stripes); 712 return sizeof(struct btrfs_chunk) + 713 sizeof(struct btrfs_stripe) * (num_stripes - 1); 714 } 715 716 /* 717 * Do the type safe conversion from stripe_nr to offset inside the chunk. 718 * 719 * @stripe_nr is u32, with left shift it can overflow u32 for chunks larger 720 * than 4G. This does the proper type cast to avoid overflow. 721 */ 722 static inline u64 btrfs_stripe_nr_to_offset(u32 stripe_nr) 723 { 724 return (u64)stripe_nr << BTRFS_STRIPE_LEN_SHIFT; 725 } 726 727 void btrfs_get_bioc(struct btrfs_io_context *bioc); 728 void btrfs_put_bioc(struct btrfs_io_context *bioc); 729 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, 730 u64 logical, u64 *length, 731 struct btrfs_io_context **bioc_ret, 732 struct btrfs_io_stripe *smap, int *mirror_num_ret); 733 int btrfs_map_repair_block(struct btrfs_fs_info *fs_info, 734 struct btrfs_io_stripe *smap, u64 logical, 735 u32 length, int mirror_num); 736 struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info, 737 u64 logical, u64 *length_ret, 738 u32 *num_stripes, bool do_remap); 739 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info); 740 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info); 741 struct btrfs_block_group *btrfs_create_chunk(struct btrfs_trans_handle *trans, 742 struct btrfs_space_info *space_info, 743 u64 type); 744 void btrfs_mapping_tree_free(struct btrfs_fs_info *fs_info); 745 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, 746 blk_mode_t flags, void *holder); 747 struct btrfs_device *btrfs_scan_one_device(const char *path, bool mount_arg_dev); 748 int btrfs_forget_devices(dev_t devt); 749 void btrfs_close_devices(struct btrfs_fs_devices *fs_devices); 750 void btrfs_release_device_allow_freeze(struct file *bdev_file); 751 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices); 752 void btrfs_assign_next_active_device(struct btrfs_device *device, 753 struct btrfs_device *this_dev); 754 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info, 755 u64 devid, 756 const char *devpath); 757 int btrfs_get_dev_args_from_path(struct btrfs_fs_info *fs_info, 758 struct btrfs_dev_lookup_args *args, 759 const char *path); 760 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info, 761 const u64 *devid, const u8 *uuid, 762 const char *path); 763 void btrfs_put_dev_args_from_path(struct btrfs_dev_lookup_args *args); 764 int btrfs_rm_device(struct btrfs_fs_info *fs_info, 765 struct btrfs_dev_lookup_args *args, 766 struct file **bdev_file); 767 void __exit btrfs_cleanup_fs_uuids(void); 768 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len); 769 int btrfs_grow_device(struct btrfs_trans_handle *trans, 770 struct btrfs_device *device, u64 new_size); 771 struct btrfs_device *btrfs_find_device(const struct btrfs_fs_devices *fs_devices, 772 const struct btrfs_dev_lookup_args *args); 773 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); 774 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path); 775 struct file *btrfs_open_device_deny_freeze(const char *path, 776 struct super_block *sb); 777 int btrfs_balance(struct btrfs_fs_info *fs_info, 778 struct btrfs_balance_control *bctl, 779 struct btrfs_ioctl_balance_args *bargs); 780 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf); 781 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info); 782 int btrfs_recover_balance(struct btrfs_fs_info *fs_info); 783 int btrfs_pause_balance(struct btrfs_fs_info *fs_info); 784 int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset, 785 bool verbose); 786 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info); 787 bool btrfs_chunk_writeable(struct btrfs_fs_info *fs_info, u64 chunk_offset); 788 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index); 789 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info, 790 struct btrfs_ioctl_get_dev_stats *stats); 791 int btrfs_init_devices_late(struct btrfs_fs_info *fs_info); 792 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info); 793 int btrfs_init_writeback_bio_size(struct btrfs_fs_info *fs_info); 794 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans); 795 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev); 796 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev); 797 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev, 798 bool allow_freeze); 799 void btrfs_free_device(struct btrfs_device *device); 800 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info, 801 u64 logical); 802 u64 btrfs_calc_stripe_length(const struct btrfs_chunk_map *map); 803 int btrfs_nr_parity_stripes(u64 type); 804 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans, 805 struct btrfs_block_group *bg); 806 int btrfs_remove_dev_extents(struct btrfs_trans_handle *trans, struct btrfs_chunk_map *map); 807 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset); 808 809 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 810 struct btrfs_chunk_map *btrfs_alloc_chunk_map(int num_stripes, gfp_t gfp); 811 int btrfs_add_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map); 812 #endif 813 814 struct btrfs_chunk_map *btrfs_find_chunk_map(struct btrfs_fs_info *fs_info, 815 u64 logical, u64 length); 816 struct btrfs_chunk_map *btrfs_find_chunk_map_nolock(struct btrfs_fs_info *fs_info, 817 u64 logical, u64 length); 818 struct btrfs_chunk_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info, 819 u64 logical, u64 length); 820 void btrfs_remove_chunk_map(struct btrfs_fs_info *fs_info, struct btrfs_chunk_map *map); 821 struct btrfs_super_block *btrfs_read_disk_super(struct block_device *bdev, 822 int copy_num, bool drop_cache); 823 void btrfs_release_disk_super(struct btrfs_super_block *super); 824 825 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev, 826 int index) 827 { 828 atomic_inc(dev->dev_stat_values + index); 829 /* 830 * This memory barrier orders stores updating statistics before stores 831 * updating dev_stats_ccnt. 832 * 833 * It pairs with smp_rmb() in btrfs_run_dev_stats(). 834 */ 835 smp_mb__before_atomic(); 836 atomic_inc(&dev->dev_stats_ccnt); 837 } 838 839 static inline int btrfs_dev_stat_read(struct btrfs_device *dev, 840 int index) 841 { 842 return atomic_read(dev->dev_stat_values + index); 843 } 844 845 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev, 846 int index) 847 { 848 int ret; 849 850 ret = atomic_xchg(dev->dev_stat_values + index, 0); 851 /* 852 * atomic_xchg implies a full memory barriers as per atomic_t.txt: 853 * - RMW operations that have a return value are fully ordered; 854 * 855 * This implicit memory barriers is paired with the smp_rmb in 856 * btrfs_run_dev_stats 857 */ 858 atomic_inc(&dev->dev_stats_ccnt); 859 return ret; 860 } 861 862 static inline void btrfs_dev_stat_set(struct btrfs_device *dev, 863 int index, unsigned long val) 864 { 865 atomic_set(dev->dev_stat_values + index, val); 866 /* 867 * This memory barrier orders stores updating statistics before stores 868 * updating dev_stats_ccnt. 869 * 870 * It pairs with smp_rmb() in btrfs_run_dev_stats(). 871 */ 872 smp_mb__before_atomic(); 873 atomic_inc(&dev->dev_stats_ccnt); 874 } 875 876 static inline const char *btrfs_dev_name(const struct btrfs_device *device) 877 { 878 if (!device || test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state)) 879 return "<missing disk>"; 880 else 881 return rcu_dereference(device->name); 882 } 883 884 static inline void btrfs_warn_unknown_chunk_allocation(enum btrfs_chunk_allocation_policy pol) 885 { 886 WARN_ONCE(1, "unknown allocation policy %d, fallback to regular", pol); 887 } 888 889 static inline void btrfs_fs_devices_inc_holding(struct btrfs_fs_devices *fs_devices) 890 { 891 lockdep_assert_held(&uuid_mutex); 892 ASSERT(fs_devices->holding >= 0); 893 fs_devices->holding++; 894 } 895 896 static inline void btrfs_fs_devices_dec_holding(struct btrfs_fs_devices *fs_devices) 897 { 898 lockdep_assert_held(&uuid_mutex); 899 ASSERT(fs_devices->holding > 0); 900 fs_devices->holding--; 901 } 902 903 void btrfs_commit_device_sizes(struct btrfs_transaction *trans); 904 905 struct list_head * __attribute_const__ btrfs_get_fs_uuids(void); 906 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info, 907 struct btrfs_device *failing_dev); 908 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info, struct btrfs_device *device); 909 910 enum btrfs_raid_types __attribute_const__ btrfs_bg_flags_to_raid_index(u64 flags); 911 int btrfs_bg_type_to_factor(u64 flags); 912 const char *btrfs_bg_type_to_raid_name(u64 flags); 913 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info); 914 bool btrfs_verify_dev_items(const struct btrfs_fs_info *fs_info); 915 void btrfs_update_per_profile_avail(struct btrfs_fs_info *fs_info); 916 917 static inline bool btrfs_get_per_profile_avail(struct btrfs_fs_info *fs_info, 918 u64 profile, u64 *avail_ret) 919 { 920 enum btrfs_raid_types index = btrfs_bg_flags_to_raid_index(profile); 921 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 922 bool uptodate = false; 923 924 spin_lock(&fs_devices->per_profile_lock); 925 if (fs_devices->per_profile_avail[index] != U64_MAX) { 926 uptodate = true; 927 *avail_ret = fs_devices->per_profile_avail[index]; 928 } 929 spin_unlock(&fs_info->fs_devices->per_profile_lock); 930 return uptodate; 931 } 932 933 bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical); 934 935 bool btrfs_pinned_by_swapfile(struct btrfs_fs_info *fs_info, void *ptr); 936 const u8 *btrfs_sb_fsid_ptr(const struct btrfs_super_block *sb); 937 int btrfs_update_device(struct btrfs_trans_handle *trans, struct btrfs_device *device); 938 void btrfs_chunk_map_device_clear_bits(struct btrfs_chunk_map *map, unsigned int bits); 939 940 bool btrfs_first_pending_extent(struct btrfs_device *device, u64 start, u64 len, 941 u64 *pending_start, u64 *pending_end); 942 bool btrfs_find_hole_in_pending_extents(struct btrfs_device *device, 943 u64 *start, u64 *len, u64 min_hole_size); 944 int btrfs_remove_dev_stat_item(struct btrfs_trans_handle *trans, u64 devid); 945 946 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 947 struct btrfs_io_context *alloc_btrfs_io_context(struct btrfs_fs_info *fs_info, 948 u64 logical, u16 total_stripes); 949 #endif 950 951 #endif 952