1 /* SPDX-License-Identifier: GPL-2.0 */ 2 3 #ifndef BTRFS_FS_H 4 #define BTRFS_FS_H 5 6 #include <linux/blkdev.h> 7 #include <linux/sizes.h> 8 #include <linux/time64.h> 9 #include <linux/compiler.h> 10 #include <linux/math.h> 11 #include <linux/atomic.h> 12 #include <linux/percpu_counter.h> 13 #include <linux/completion.h> 14 #include <linux/lockdep.h> 15 #include <linux/spinlock.h> 16 #include <linux/mutex.h> 17 #include <linux/rwsem.h> 18 #include <linux/semaphore.h> 19 #include <linux/list.h> 20 #include <linux/pagemap.h> 21 #include <linux/radix-tree.h> 22 #include <linux/workqueue.h> 23 #include <linux/wait.h> 24 #include <linux/wait_bit.h> 25 #include <linux/sched.h> 26 #include <linux/rbtree.h> 27 #include <uapi/linux/btrfs.h> 28 #include <uapi/linux/btrfs_tree.h> 29 #include "extent-io-tree.h" 30 #include "async-thread.h" 31 #include "block-rsv.h" 32 33 struct inode; 34 struct super_block; 35 struct kobject; 36 struct reloc_control; 37 struct crypto_shash; 38 struct ulist; 39 struct btrfs_device; 40 struct btrfs_block_group; 41 struct btrfs_root; 42 struct btrfs_fs_devices; 43 struct btrfs_transaction; 44 struct btrfs_delayed_root; 45 struct btrfs_balance_control; 46 struct btrfs_subpage_info; 47 struct btrfs_stripe_hash_table; 48 struct btrfs_space_info; 49 50 /* 51 * Minimum data and metadata block size. 52 * 53 * Normally it's 4K, but for testing subpage block size on 4K page systems, we 54 * allow DEBUG builds to accept 2K page size. 55 */ 56 #ifdef CONFIG_BTRFS_DEBUG 57 #define BTRFS_MIN_BLOCKSIZE (SZ_2K) 58 #else 59 #define BTRFS_MIN_BLOCKSIZE (SZ_4K) 60 #endif 61 62 #define BTRFS_MAX_BLOCKSIZE (SZ_64K) 63 64 #define BTRFS_MAX_EXTENT_SIZE SZ_128M 65 66 #define BTRFS_OLDEST_GENERATION 0ULL 67 68 #define BTRFS_EMPTY_DIR_SIZE 0 69 70 #define BTRFS_DIRTY_METADATA_THRESH SZ_32M 71 72 #define BTRFS_SUPER_INFO_OFFSET SZ_64K 73 #define BTRFS_SUPER_INFO_SIZE 4096 74 static_assert(sizeof(struct btrfs_super_block) == BTRFS_SUPER_INFO_SIZE); 75 76 /* 77 * Number of metadata items necessary for an unlink operation: 78 * 79 * 1 for the possible orphan item 80 * 1 for the dir item 81 * 1 for the dir index 82 * 1 for the inode ref 83 * 1 for the inode 84 * 1 for the parent inode 85 */ 86 #define BTRFS_UNLINK_METADATA_UNITS 6 87 88 /* 89 * The reserved space at the beginning of each device. It covers the primary 90 * super block and leaves space for potential use by other tools like 91 * bootloaders or to lower potential damage of accidental overwrite. 92 */ 93 #define BTRFS_DEVICE_RANGE_RESERVED (SZ_1M) 94 /* 95 * Runtime (in-memory) states of filesystem 96 */ 97 enum { 98 /* 99 * Filesystem is being remounted, allow to skip some operations, like 100 * defrag 101 */ 102 BTRFS_FS_STATE_REMOUNTING, 103 /* Filesystem in RO mode */ 104 BTRFS_FS_STATE_RO, 105 /* Track if a transaction abort has been reported on this filesystem */ 106 BTRFS_FS_STATE_TRANS_ABORTED, 107 /* Track if log replay has failed. */ 108 BTRFS_FS_STATE_LOG_REPLAY_ABORTED, 109 /* 110 * Bio operations should be blocked on this filesystem because a source 111 * or target device is being destroyed as part of a device replace 112 */ 113 BTRFS_FS_STATE_DEV_REPLACING, 114 /* The btrfs_fs_info created for self-tests */ 115 BTRFS_FS_STATE_DUMMY_FS_INFO, 116 117 /* Checksum errors are ignored. */ 118 BTRFS_FS_STATE_NO_DATA_CSUMS, 119 BTRFS_FS_STATE_SKIP_META_CSUMS, 120 121 /* Indicates there was an error cleaning up a log tree. */ 122 BTRFS_FS_STATE_LOG_CLEANUP_ERROR, 123 124 /* No more delayed iput can be queued. */ 125 BTRFS_FS_STATE_NO_DELAYED_IPUT, 126 127 BTRFS_FS_STATE_COUNT 128 }; 129 130 enum { 131 BTRFS_FS_CLOSING_START, 132 BTRFS_FS_CLOSING_DONE, 133 BTRFS_FS_LOG_RECOVERING, 134 BTRFS_FS_OPEN, 135 BTRFS_FS_QUOTA_ENABLED, 136 BTRFS_FS_UPDATE_UUID_TREE_GEN, 137 BTRFS_FS_CREATING_FREE_SPACE_TREE, 138 BTRFS_FS_BTREE_ERR, 139 BTRFS_FS_LOG1_ERR, 140 BTRFS_FS_LOG2_ERR, 141 BTRFS_FS_QUOTA_OVERRIDE, 142 /* Used to record internally whether fs has been frozen */ 143 BTRFS_FS_FROZEN, 144 /* 145 * Indicate that balance has been set up from the ioctl and is in the 146 * main phase. The fs_info::balance_ctl is initialized. 147 */ 148 BTRFS_FS_BALANCE_RUNNING, 149 150 /* 151 * Indicate that relocation of a chunk has started, it's set per chunk 152 * and is toggled between chunks. 153 */ 154 BTRFS_FS_RELOC_RUNNING, 155 156 /* Indicate that the cleaner thread is awake and doing something. */ 157 BTRFS_FS_CLEANER_RUNNING, 158 159 /* 160 * The checksumming has an optimized version and is considered fast, 161 * so we don't need to offload checksums to workqueues. 162 */ 163 BTRFS_FS_CSUM_IMPL_FAST, 164 165 /* Indicate that the discard workqueue can service discards. */ 166 BTRFS_FS_DISCARD_RUNNING, 167 168 /* Indicate that we need to cleanup space cache v1 */ 169 BTRFS_FS_CLEANUP_SPACE_CACHE_V1, 170 171 /* Indicate that we can't trust the free space tree for caching yet */ 172 BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED, 173 174 /* Indicate whether there are any tree modification log users */ 175 BTRFS_FS_TREE_MOD_LOG_USERS, 176 177 /* Indicate that we want the transaction kthread to commit right now. */ 178 BTRFS_FS_COMMIT_TRANS, 179 180 /* Indicate we have half completed snapshot deletions pending. */ 181 BTRFS_FS_UNFINISHED_DROPS, 182 183 /* Indicate we have to finish a zone to do next allocation. */ 184 BTRFS_FS_NEED_ZONE_FINISH, 185 186 /* Indicate that we want to commit the transaction. */ 187 BTRFS_FS_NEED_TRANS_COMMIT, 188 189 /* This is set when active zone tracking is needed. */ 190 BTRFS_FS_ACTIVE_ZONE_TRACKING, 191 192 /* 193 * Indicate if we have some features changed, this is mostly for 194 * cleaner thread to update the sysfs interface. 195 */ 196 BTRFS_FS_FEATURE_CHANGED, 197 198 /* 199 * Indicate that we have found a tree block which is only aligned to 200 * sectorsize, but not to nodesize. This should be rare nowadays. 201 */ 202 BTRFS_FS_UNALIGNED_TREE_BLOCK, 203 204 #if BITS_PER_LONG == 32 205 /* Indicate if we have error/warn message printed on 32bit systems */ 206 BTRFS_FS_32BIT_ERROR, 207 BTRFS_FS_32BIT_WARN, 208 #endif 209 }; 210 211 /* 212 * Flags for mount options. 213 * 214 * Note: don't forget to add new options to btrfs_show_options() 215 */ 216 enum { 217 BTRFS_MOUNT_NODATASUM = (1ULL << 0), 218 BTRFS_MOUNT_NODATACOW = (1ULL << 1), 219 BTRFS_MOUNT_NOBARRIER = (1ULL << 2), 220 BTRFS_MOUNT_SSD = (1ULL << 3), 221 BTRFS_MOUNT_DEGRADED = (1ULL << 4), 222 BTRFS_MOUNT_COMPRESS = (1ULL << 5), 223 BTRFS_MOUNT_NOTREELOG = (1ULL << 6), 224 BTRFS_MOUNT_FLUSHONCOMMIT = (1ULL << 7), 225 BTRFS_MOUNT_SSD_SPREAD = (1ULL << 8), 226 BTRFS_MOUNT_NOSSD = (1ULL << 9), 227 BTRFS_MOUNT_DISCARD_SYNC = (1ULL << 10), 228 BTRFS_MOUNT_FORCE_COMPRESS = (1ULL << 11), 229 BTRFS_MOUNT_SPACE_CACHE = (1ULL << 12), 230 BTRFS_MOUNT_CLEAR_CACHE = (1ULL << 13), 231 BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED = (1ULL << 14), 232 BTRFS_MOUNT_ENOSPC_DEBUG = (1ULL << 15), 233 BTRFS_MOUNT_AUTO_DEFRAG = (1ULL << 16), 234 BTRFS_MOUNT_USEBACKUPROOT = (1ULL << 17), 235 BTRFS_MOUNT_SKIP_BALANCE = (1ULL << 18), 236 BTRFS_MOUNT_PANIC_ON_FATAL_ERROR = (1ULL << 19), 237 BTRFS_MOUNT_RESCAN_UUID_TREE = (1ULL << 20), 238 BTRFS_MOUNT_FRAGMENT_DATA = (1ULL << 21), 239 BTRFS_MOUNT_FRAGMENT_METADATA = (1ULL << 22), 240 BTRFS_MOUNT_FREE_SPACE_TREE = (1ULL << 23), 241 BTRFS_MOUNT_NOLOGREPLAY = (1ULL << 24), 242 BTRFS_MOUNT_REF_VERIFY = (1ULL << 25), 243 BTRFS_MOUNT_DISCARD_ASYNC = (1ULL << 26), 244 BTRFS_MOUNT_IGNOREBADROOTS = (1ULL << 27), 245 BTRFS_MOUNT_IGNOREDATACSUMS = (1ULL << 28), 246 BTRFS_MOUNT_NODISCARD = (1ULL << 29), 247 BTRFS_MOUNT_NOSPACECACHE = (1ULL << 30), 248 BTRFS_MOUNT_IGNOREMETACSUMS = (1ULL << 31), 249 BTRFS_MOUNT_IGNORESUPERFLAGS = (1ULL << 32), 250 BTRFS_MOUNT_REF_TRACKER = (1ULL << 33), 251 }; 252 253 /* 254 * Compat flags that we support. If any incompat flags are set other than the 255 * ones specified below then we will fail to mount 256 */ 257 #define BTRFS_FEATURE_COMPAT_SUPP 0ULL 258 #define BTRFS_FEATURE_COMPAT_SAFE_SET 0ULL 259 #define BTRFS_FEATURE_COMPAT_SAFE_CLEAR 0ULL 260 261 #define BTRFS_FEATURE_COMPAT_RO_SUPP \ 262 (BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE | \ 263 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID | \ 264 BTRFS_FEATURE_COMPAT_RO_VERITY | \ 265 BTRFS_FEATURE_COMPAT_RO_BLOCK_GROUP_TREE) 266 267 #define BTRFS_FEATURE_COMPAT_RO_SAFE_SET 0ULL 268 #define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR 0ULL 269 270 #define BTRFS_FEATURE_INCOMPAT_SUPP_STABLE \ 271 (BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF | \ 272 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL | \ 273 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS | \ 274 BTRFS_FEATURE_INCOMPAT_BIG_METADATA | \ 275 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO | \ 276 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD | \ 277 BTRFS_FEATURE_INCOMPAT_RAID56 | \ 278 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF | \ 279 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA | \ 280 BTRFS_FEATURE_INCOMPAT_NO_HOLES | \ 281 BTRFS_FEATURE_INCOMPAT_METADATA_UUID | \ 282 BTRFS_FEATURE_INCOMPAT_RAID1C34 | \ 283 BTRFS_FEATURE_INCOMPAT_ZONED | \ 284 BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA) 285 286 #ifdef CONFIG_BTRFS_EXPERIMENTAL 287 /* 288 * Features under development like Extent tree v2 support is enabled 289 * only under CONFIG_BTRFS_EXPERIMENTAL 290 */ 291 #define BTRFS_FEATURE_INCOMPAT_SUPP \ 292 (BTRFS_FEATURE_INCOMPAT_SUPP_STABLE | \ 293 BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE | \ 294 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2) 295 296 #else 297 298 #define BTRFS_FEATURE_INCOMPAT_SUPP \ 299 (BTRFS_FEATURE_INCOMPAT_SUPP_STABLE) 300 301 #endif 302 303 #define BTRFS_FEATURE_INCOMPAT_SAFE_SET \ 304 (BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF) 305 #define BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR 0ULL 306 307 #define BTRFS_DEFAULT_COMMIT_INTERVAL (30) 308 #define BTRFS_WARNING_COMMIT_INTERVAL (300) 309 #define BTRFS_DEFAULT_MAX_INLINE (2048) 310 311 enum btrfs_compression_type { 312 BTRFS_COMPRESS_NONE = 0, 313 BTRFS_COMPRESS_ZLIB = 1, 314 BTRFS_COMPRESS_LZO = 2, 315 BTRFS_COMPRESS_ZSTD = 3, 316 BTRFS_NR_COMPRESS_TYPES = 4, 317 318 BTRFS_DEFRAG_DONT_COMPRESS, 319 }; 320 321 struct btrfs_dev_replace { 322 /* See #define above */ 323 u64 replace_state; 324 /* Seconds since 1-Jan-1970 */ 325 time64_t time_started; 326 /* Seconds since 1-Jan-1970 */ 327 time64_t time_stopped; 328 atomic64_t num_write_errors; 329 atomic64_t num_uncorrectable_read_errors; 330 331 u64 cursor_left; 332 u64 committed_cursor_left; 333 u64 cursor_left_last_write_of_item; 334 u64 cursor_right; 335 336 /* See #define above */ 337 u64 cont_reading_from_srcdev_mode; 338 339 int is_valid; 340 int item_needs_writeback; 341 struct btrfs_device *srcdev; 342 struct btrfs_device *tgtdev; 343 344 struct mutex lock_finishing_cancel_unmount; 345 struct rw_semaphore rwsem; 346 347 struct btrfs_scrub_progress scrub_progress; 348 349 struct percpu_counter bio_counter; 350 wait_queue_head_t replace_wait; 351 352 struct task_struct *replace_task; 353 }; 354 355 /* 356 * Free clusters are used to claim free space in relatively large chunks, 357 * allowing us to do less seeky writes. They are used for all metadata 358 * allocations. In ssd_spread mode they are also used for data allocations. 359 */ 360 struct btrfs_free_cluster { 361 spinlock_t lock; 362 spinlock_t refill_lock; 363 struct rb_root root; 364 365 /* Largest extent in this cluster */ 366 u64 max_size; 367 368 /* First extent starting offset */ 369 u64 window_start; 370 371 /* We did a full search and couldn't create a cluster */ 372 bool fragmented; 373 374 struct btrfs_block_group *block_group; 375 /* 376 * When a cluster is allocated from a block group, we put the cluster 377 * onto a list in the block group so that it can be freed before the 378 * block group is freed. 379 */ 380 struct list_head block_group_list; 381 }; 382 383 /* Discard control. */ 384 /* 385 * Async discard uses multiple lists to differentiate the discard filter 386 * parameters. Index 0 is for completely free block groups where we need to 387 * ensure the entire block group is trimmed without being lossy. Indices 388 * afterwards represent monotonically decreasing discard filter sizes to 389 * prioritize what should be discarded next. 390 */ 391 #define BTRFS_NR_DISCARD_LISTS 3 392 #define BTRFS_DISCARD_INDEX_UNUSED 0 393 #define BTRFS_DISCARD_INDEX_START 1 394 395 struct btrfs_discard_ctl { 396 struct workqueue_struct *discard_workers; 397 struct delayed_work work; 398 spinlock_t lock; 399 struct btrfs_block_group *block_group; 400 struct list_head discard_list[BTRFS_NR_DISCARD_LISTS]; 401 u64 prev_discard; 402 u64 prev_discard_time; 403 atomic_t discardable_extents; 404 atomic64_t discardable_bytes; 405 u64 max_discard_size; 406 u64 delay_ms; 407 u32 iops_limit; 408 u32 kbps_limit; 409 u64 discard_extent_bytes; 410 u64 discard_bitmap_bytes; 411 atomic64_t discard_bytes_saved; 412 }; 413 414 /* 415 * Exclusive operations (device replace, resize, device add/remove, balance) 416 */ 417 enum btrfs_exclusive_operation { 418 BTRFS_EXCLOP_NONE, 419 BTRFS_EXCLOP_BALANCE_PAUSED, 420 BTRFS_EXCLOP_BALANCE, 421 BTRFS_EXCLOP_DEV_ADD, 422 BTRFS_EXCLOP_DEV_REMOVE, 423 BTRFS_EXCLOP_DEV_REPLACE, 424 BTRFS_EXCLOP_RESIZE, 425 BTRFS_EXCLOP_SWAP_ACTIVATE, 426 }; 427 428 /* Store data about transaction commits, exported via sysfs. */ 429 struct btrfs_commit_stats { 430 /* Total number of commits */ 431 u64 commit_count; 432 /* The maximum commit duration so far in ns */ 433 u64 max_commit_dur; 434 /* The last commit duration in ns */ 435 u64 last_commit_dur; 436 /* The total commit duration in ns */ 437 u64 total_commit_dur; 438 /* Start of the last critical section in ns. */ 439 u64 critical_section_start_time; 440 }; 441 442 struct btrfs_fs_info { 443 u8 chunk_tree_uuid[BTRFS_UUID_SIZE]; 444 unsigned long flags; 445 struct btrfs_root *tree_root; 446 struct btrfs_root *chunk_root; 447 struct btrfs_root *dev_root; 448 struct btrfs_root *fs_root; 449 struct btrfs_root *quota_root; 450 struct btrfs_root *uuid_root; 451 struct btrfs_root *data_reloc_root; 452 struct btrfs_root *block_group_root; 453 struct btrfs_root *stripe_root; 454 455 /* The log root tree is a directory of all the other log roots */ 456 struct btrfs_root *log_root_tree; 457 458 /* The tree that holds the global roots (csum, extent, etc) */ 459 rwlock_t global_root_lock; 460 struct rb_root global_root_tree; 461 462 spinlock_t fs_roots_radix_lock; 463 struct radix_tree_root fs_roots_radix; 464 465 /* Block group cache stuff */ 466 rwlock_t block_group_cache_lock; 467 struct rb_root_cached block_group_cache_tree; 468 469 /* Keep track of unallocated space */ 470 atomic64_t free_chunk_space; 471 472 /* Track ranges which are used by log trees blocks/logged data extents */ 473 struct extent_io_tree excluded_extents; 474 475 /* logical->physical extent mapping */ 476 struct rb_root_cached mapping_tree; 477 rwlock_t mapping_tree_lock; 478 479 /* 480 * Block reservation for extent, checksum, root tree and delayed dir 481 * index item. 482 */ 483 struct btrfs_block_rsv global_block_rsv; 484 /* Block reservation for metadata operations */ 485 struct btrfs_block_rsv trans_block_rsv; 486 /* Block reservation for chunk tree */ 487 struct btrfs_block_rsv chunk_block_rsv; 488 /* Block reservation for delayed operations */ 489 struct btrfs_block_rsv delayed_block_rsv; 490 /* Block reservation for delayed refs */ 491 struct btrfs_block_rsv delayed_refs_rsv; 492 /* Block reservation for treelog tree */ 493 struct btrfs_block_rsv treelog_rsv; 494 495 struct btrfs_block_rsv empty_block_rsv; 496 497 /* 498 * Updated while holding the lock 'trans_lock'. Due to the life cycle of 499 * a transaction, it can be directly read while holding a transaction 500 * handle, everywhere else must be read with btrfs_get_fs_generation(). 501 * Should always be updated using btrfs_set_fs_generation(). 502 */ 503 u64 generation; 504 /* 505 * Always use btrfs_get_last_trans_committed() and 506 * btrfs_set_last_trans_committed() to read and update this field. 507 */ 508 u64 last_trans_committed; 509 /* 510 * Generation of the last transaction used for block group relocation 511 * since the filesystem was last mounted (or 0 if none happened yet). 512 * Must be written and read while holding btrfs_fs_info::commit_root_sem. 513 */ 514 u64 last_reloc_trans; 515 516 /* 517 * This is updated to the current trans every time a full commit is 518 * required instead of the faster short fsync log commits 519 */ 520 u64 last_trans_log_full_commit; 521 unsigned long long mount_opt; 522 523 /* Compress related structures. */ 524 void *compr_wsm[BTRFS_NR_COMPRESS_TYPES]; 525 526 int compress_type; 527 int compress_level; 528 u32 commit_interval; 529 /* 530 * It is a suggestive number, the read side is safe even it gets a 531 * wrong number because we will write out the data into a regular 532 * extent. The write side(mount/remount) is under ->s_umount lock, 533 * so it is also safe. 534 */ 535 u64 max_inline; 536 537 struct btrfs_transaction *running_transaction; 538 wait_queue_head_t transaction_throttle; 539 wait_queue_head_t transaction_wait; 540 wait_queue_head_t transaction_blocked_wait; 541 wait_queue_head_t async_submit_wait; 542 543 /* 544 * Used to protect the incompat_flags, compat_flags, compat_ro_flags 545 * when they are updated. 546 * 547 * Because we do not clear the flags for ever, so we needn't use 548 * the lock on the read side. 549 * 550 * We also needn't use the lock when we mount the fs, because 551 * there is no other task which will update the flag. 552 */ 553 spinlock_t super_lock; 554 struct btrfs_super_block *super_copy; 555 struct btrfs_super_block *super_for_commit; 556 struct super_block *sb; 557 struct inode *btree_inode; 558 struct mutex tree_log_mutex; 559 struct mutex transaction_kthread_mutex; 560 struct mutex cleaner_mutex; 561 struct mutex chunk_mutex; 562 563 /* 564 * This is taken to make sure we don't set block groups ro after the 565 * free space cache has been allocated on them. 566 */ 567 struct mutex ro_block_group_mutex; 568 569 /* 570 * This is used during read/modify/write to make sure no two ios are 571 * trying to mod the same stripe at the same time. 572 */ 573 struct btrfs_stripe_hash_table *stripe_hash_table; 574 575 /* 576 * This protects the ordered operations list only while we are 577 * processing all of the entries on it. This way we make sure the 578 * commit code doesn't find the list temporarily empty because another 579 * function happens to be doing non-waiting preflush before jumping 580 * into the main commit. 581 */ 582 struct mutex ordered_operations_mutex; 583 584 struct rw_semaphore commit_root_sem; 585 586 struct rw_semaphore cleanup_work_sem; 587 588 struct rw_semaphore subvol_sem; 589 590 spinlock_t trans_lock; 591 /* 592 * The reloc mutex goes with the trans lock, it is taken during commit 593 * to protect us from the relocation code. 594 */ 595 struct mutex reloc_mutex; 596 597 struct list_head trans_list; 598 struct list_head dead_roots; 599 struct list_head caching_block_groups; 600 601 spinlock_t delayed_iput_lock; 602 struct list_head delayed_iputs; 603 atomic_t nr_delayed_iputs; 604 wait_queue_head_t delayed_iputs_wait; 605 606 atomic64_t tree_mod_seq; 607 608 /* This protects tree_mod_log and tree_mod_seq_list */ 609 rwlock_t tree_mod_log_lock; 610 struct rb_root tree_mod_log; 611 struct list_head tree_mod_seq_list; 612 613 atomic_t async_delalloc_pages; 614 615 /* This is used to protect the following list -- ordered_roots. */ 616 spinlock_t ordered_root_lock; 617 618 /* 619 * All fs/file tree roots in which there are data=ordered extents 620 * pending writeback are added into this list. 621 * 622 * These can span multiple transactions and basically include every 623 * dirty data page that isn't from nodatacow. 624 */ 625 struct list_head ordered_roots; 626 627 struct mutex delalloc_root_mutex; 628 spinlock_t delalloc_root_lock; 629 /* All fs/file tree roots that have delalloc inodes. */ 630 struct list_head delalloc_roots; 631 632 /* 633 * There is a pool of worker threads for checksumming during writes and 634 * a pool for checksumming after reads. This is because readers can 635 * run with FS locks held, and the writers may be waiting for those 636 * locks. We don't want ordering in the pending list to cause 637 * deadlocks, and so the two are serviced separately. 638 * 639 * A third pool does submit_bio to avoid deadlocking with the other two. 640 */ 641 struct btrfs_workqueue *workers; 642 struct btrfs_workqueue *delalloc_workers; 643 struct btrfs_workqueue *flush_workers; 644 struct workqueue_struct *endio_workers; 645 struct workqueue_struct *endio_meta_workers; 646 struct workqueue_struct *rmw_workers; 647 struct workqueue_struct *compressed_write_workers; 648 struct btrfs_workqueue *endio_write_workers; 649 struct btrfs_workqueue *endio_freespace_worker; 650 struct btrfs_workqueue *caching_workers; 651 652 /* 653 * Fixup workers take dirty pages that didn't properly go through the 654 * cow mechanism and make them safe to write. It happens for the 655 * sys_munmap function call path. 656 */ 657 struct btrfs_workqueue *fixup_workers; 658 struct btrfs_workqueue *delayed_workers; 659 660 struct task_struct *transaction_kthread; 661 struct task_struct *cleaner_kthread; 662 u32 thread_pool_size; 663 664 struct kobject *space_info_kobj; 665 struct kobject *qgroups_kobj; 666 struct kobject *discard_kobj; 667 668 /* Track the number of blocks (sectors) read by the filesystem. */ 669 struct percpu_counter stats_read_blocks; 670 671 /* Used to keep from writing metadata until there is a nice batch */ 672 struct percpu_counter dirty_metadata_bytes; 673 struct percpu_counter delalloc_bytes; 674 struct percpu_counter ordered_bytes; 675 s32 dirty_metadata_batch; 676 s32 delalloc_batch; 677 678 struct percpu_counter evictable_extent_maps; 679 u64 em_shrinker_last_root; 680 u64 em_shrinker_last_ino; 681 atomic64_t em_shrinker_nr_to_scan; 682 struct work_struct em_shrinker_work; 683 684 /* Protected by 'trans_lock'. */ 685 struct list_head dirty_cowonly_roots; 686 687 struct btrfs_fs_devices *fs_devices; 688 689 /* 690 * The space_info list is effectively read only after initial setup. 691 * It is populated at mount time and cleaned up after all block groups 692 * are removed. RCU is used to protect it. 693 */ 694 struct list_head space_info; 695 696 struct btrfs_space_info *data_sinfo; 697 698 struct reloc_control *reloc_ctl; 699 700 /* data_alloc_cluster is only used in ssd_spread mode */ 701 struct btrfs_free_cluster data_alloc_cluster; 702 703 /* All metadata allocations go through this cluster. */ 704 struct btrfs_free_cluster meta_alloc_cluster; 705 706 /* Auto defrag inodes go here. */ 707 spinlock_t defrag_inodes_lock; 708 struct rb_root defrag_inodes; 709 atomic_t defrag_running; 710 711 /* Used to protect avail_{data, metadata, system}_alloc_bits */ 712 seqlock_t profiles_lock; 713 /* 714 * These three are in extended format (availability of single chunks is 715 * denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other types are denoted 716 * by corresponding BTRFS_BLOCK_GROUP_* bits) 717 */ 718 u64 avail_data_alloc_bits; 719 u64 avail_metadata_alloc_bits; 720 u64 avail_system_alloc_bits; 721 722 /* Balance state */ 723 spinlock_t balance_lock; 724 struct mutex balance_mutex; 725 atomic_t balance_pause_req; 726 atomic_t balance_cancel_req; 727 struct btrfs_balance_control *balance_ctl; 728 wait_queue_head_t balance_wait_q; 729 730 /* Cancellation requests for chunk relocation */ 731 atomic_t reloc_cancel_req; 732 733 u32 data_chunk_allocations; 734 u32 metadata_ratio; 735 736 /* Private scrub information */ 737 struct mutex scrub_lock; 738 atomic_t scrubs_running; 739 atomic_t scrub_pause_req; 740 atomic_t scrubs_paused; 741 atomic_t scrub_cancel_req; 742 wait_queue_head_t scrub_pause_wait; 743 /* 744 * The worker pointers are NULL iff the refcount is 0, ie. scrub is not 745 * running. 746 */ 747 refcount_t scrub_workers_refcnt; 748 struct workqueue_struct *scrub_workers; 749 750 struct btrfs_discard_ctl discard_ctl; 751 752 /* Is qgroup tracking in a consistent state? */ 753 u64 qgroup_flags; 754 755 /* Holds configuration and tracking. Protected by qgroup_lock. */ 756 struct rb_root qgroup_tree; 757 spinlock_t qgroup_lock; 758 759 /* 760 * Protect user change for quota operations. If a transaction is needed, 761 * it must be started before locking this lock. 762 */ 763 struct mutex qgroup_ioctl_lock; 764 765 /* List of dirty qgroups to be written at next commit. */ 766 struct list_head dirty_qgroups; 767 768 /* Used by qgroup for an efficient tree traversal. */ 769 u64 qgroup_seq; 770 771 /* Qgroup rescan items. */ 772 /* Protects the progress item */ 773 struct mutex qgroup_rescan_lock; 774 struct btrfs_key qgroup_rescan_progress; 775 struct btrfs_workqueue *qgroup_rescan_workers; 776 struct completion qgroup_rescan_completion; 777 struct btrfs_work qgroup_rescan_work; 778 /* Protected by qgroup_rescan_lock */ 779 bool qgroup_rescan_running; 780 u8 qgroup_drop_subtree_thres; 781 u64 qgroup_enable_gen; 782 783 /* 784 * If this is not 0, then it indicates a serious filesystem error has 785 * happened and it contains that error (negative errno value). 786 */ 787 int fs_error; 788 789 /* Filesystem state */ 790 unsigned long fs_state; 791 792 struct btrfs_delayed_root *delayed_root; 793 794 /* Entries are eb->start >> nodesize_bits */ 795 struct xarray buffer_tree; 796 797 /* Next backup root to be overwritten */ 798 int backup_root_index; 799 800 /* Device replace state */ 801 struct btrfs_dev_replace dev_replace; 802 803 struct semaphore uuid_tree_rescan_sem; 804 805 /* Used to reclaim the metadata space in the background. */ 806 struct work_struct async_reclaim_work; 807 struct work_struct async_data_reclaim_work; 808 struct work_struct preempt_reclaim_work; 809 810 /* Reclaim partially filled block groups in the background */ 811 struct work_struct reclaim_bgs_work; 812 /* Protected by unused_bgs_lock. */ 813 struct list_head reclaim_bgs; 814 int bg_reclaim_threshold; 815 816 /* Protects the lists unused_bgs and reclaim_bgs. */ 817 spinlock_t unused_bgs_lock; 818 /* Protected by unused_bgs_lock. */ 819 struct list_head unused_bgs; 820 struct mutex unused_bg_unpin_mutex; 821 /* Protect block groups that are going to be deleted */ 822 struct mutex reclaim_bgs_lock; 823 824 /* Cached block sizes */ 825 u32 nodesize; 826 u32 nodesize_bits; 827 u32 sectorsize; 828 /* ilog2 of sectorsize, use to avoid 64bit division */ 829 u32 sectorsize_bits; 830 u32 block_min_order; 831 u32 block_max_order; 832 u32 csum_size; 833 u32 csums_per_leaf; 834 u32 stripesize; 835 836 /* 837 * Maximum size of an extent. BTRFS_MAX_EXTENT_SIZE on regular 838 * filesystem, on zoned it depends on the device constraints. 839 */ 840 u64 max_extent_size; 841 842 /* Block groups and devices containing active swapfiles. */ 843 spinlock_t swapfile_pins_lock; 844 struct rb_root swapfile_pins; 845 846 struct crypto_shash *csum_shash; 847 848 /* Type of exclusive operation running, protected by super_lock */ 849 enum btrfs_exclusive_operation exclusive_operation; 850 851 /* 852 * Zone size > 0 when in ZONED mode, otherwise it's used for a check 853 * if the mode is enabled 854 */ 855 u64 zone_size; 856 857 /* Constraints for ZONE_APPEND commands: */ 858 struct queue_limits limits; 859 u64 max_zone_append_size; 860 861 struct mutex zoned_meta_io_lock; 862 spinlock_t treelog_bg_lock; 863 u64 treelog_bg; 864 865 /* 866 * Start of the dedicated data relocation block group, protected by 867 * relocation_bg_lock. 868 */ 869 spinlock_t relocation_bg_lock; 870 u64 data_reloc_bg; 871 struct mutex zoned_data_reloc_io_lock; 872 873 struct btrfs_block_group *active_meta_bg; 874 struct btrfs_block_group *active_system_bg; 875 876 u64 nr_global_roots; 877 878 spinlock_t zone_active_bgs_lock; 879 struct list_head zone_active_bgs; 880 881 /* Updates are not protected by any lock */ 882 struct btrfs_commit_stats commit_stats; 883 884 /* 885 * Last generation where we dropped a non-relocation root. 886 * Use btrfs_set_last_root_drop_gen() and btrfs_get_last_root_drop_gen() 887 * to change it and to read it, respectively. 888 */ 889 u64 last_root_drop_gen; 890 891 /* 892 * Annotations for transaction events (structures are empty when 893 * compiled without lockdep). 894 */ 895 struct lockdep_map btrfs_trans_num_writers_map; 896 struct lockdep_map btrfs_trans_num_extwriters_map; 897 struct lockdep_map btrfs_state_change_map[4]; 898 struct lockdep_map btrfs_trans_pending_ordered_map; 899 struct lockdep_map btrfs_ordered_extent_map; 900 901 #ifdef CONFIG_BTRFS_DEBUG 902 spinlock_t ref_verify_lock; 903 struct rb_root block_tree; 904 905 struct kobject *debug_kobj; 906 struct list_head allocated_roots; 907 908 spinlock_t eb_leak_lock; 909 struct list_head allocated_ebs; 910 #endif 911 }; 912 913 #define folio_to_inode(_folio) (BTRFS_I(_Generic((_folio), \ 914 struct folio *: (_folio))->mapping->host)) 915 916 #define folio_to_fs_info(_folio) (folio_to_inode(_folio)->root->fs_info) 917 918 #define inode_to_fs_info(_inode) (BTRFS_I(_Generic((_inode), \ 919 struct inode *: (_inode)))->root->fs_info) 920 921 static inline gfp_t btrfs_alloc_write_mask(struct address_space *mapping) 922 { 923 return mapping_gfp_constraint(mapping, ~__GFP_FS); 924 } 925 926 /* Return the minimal folio size of the fs. */ 927 static inline unsigned int btrfs_min_folio_size(struct btrfs_fs_info *fs_info) 928 { 929 return 1U << (PAGE_SHIFT + fs_info->block_min_order); 930 } 931 932 static inline u64 btrfs_get_fs_generation(const struct btrfs_fs_info *fs_info) 933 { 934 return READ_ONCE(fs_info->generation); 935 } 936 937 static inline void btrfs_set_fs_generation(struct btrfs_fs_info *fs_info, u64 gen) 938 { 939 WRITE_ONCE(fs_info->generation, gen); 940 } 941 942 static inline u64 btrfs_get_last_trans_committed(const struct btrfs_fs_info *fs_info) 943 { 944 return READ_ONCE(fs_info->last_trans_committed); 945 } 946 947 static inline void btrfs_set_last_trans_committed(struct btrfs_fs_info *fs_info, u64 gen) 948 { 949 WRITE_ONCE(fs_info->last_trans_committed, gen); 950 } 951 952 static inline void btrfs_set_last_root_drop_gen(struct btrfs_fs_info *fs_info, 953 u64 gen) 954 { 955 WRITE_ONCE(fs_info->last_root_drop_gen, gen); 956 } 957 958 static inline u64 btrfs_get_last_root_drop_gen(const struct btrfs_fs_info *fs_info) 959 { 960 return READ_ONCE(fs_info->last_root_drop_gen); 961 } 962 963 /* 964 * Take the number of bytes to be checksummed and figure out how many leaves 965 * it would require to store the csums for that many bytes. 966 */ 967 static inline u64 btrfs_csum_bytes_to_leaves( 968 const struct btrfs_fs_info *fs_info, u64 csum_bytes) 969 { 970 const u64 num_csums = csum_bytes >> fs_info->sectorsize_bits; 971 972 return DIV_ROUND_UP_ULL(num_csums, fs_info->csums_per_leaf); 973 } 974 975 /* 976 * Use this if we would be adding new items, as we could split nodes as we cow 977 * down the tree. 978 */ 979 static inline u64 btrfs_calc_insert_metadata_size(const struct btrfs_fs_info *fs_info, 980 unsigned num_items) 981 { 982 return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * 2 * num_items; 983 } 984 985 /* 986 * Doing a truncate or a modification won't result in new nodes or leaves, just 987 * what we need for COW. 988 */ 989 static inline u64 btrfs_calc_metadata_size(const struct btrfs_fs_info *fs_info, 990 unsigned num_items) 991 { 992 return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * num_items; 993 } 994 995 #define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r->fs_info) >> 4) - \ 996 sizeof(struct btrfs_item)) 997 998 #define BTRFS_BYTES_TO_BLKS(fs_info, bytes) ((bytes) >> (fs_info)->sectorsize_bits) 999 1000 static inline bool btrfs_is_zoned(const struct btrfs_fs_info *fs_info) 1001 { 1002 return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && fs_info->zone_size > 0; 1003 } 1004 1005 /* 1006 * Count how many fs_info->max_extent_size cover the @size 1007 */ 1008 static inline u32 count_max_extents(const struct btrfs_fs_info *fs_info, u64 size) 1009 { 1010 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 1011 if (!fs_info) 1012 return div_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, BTRFS_MAX_EXTENT_SIZE); 1013 #endif 1014 1015 return div_u64(size + fs_info->max_extent_size - 1, fs_info->max_extent_size); 1016 } 1017 1018 static inline unsigned int btrfs_blocks_per_folio(const struct btrfs_fs_info *fs_info, 1019 const struct folio *folio) 1020 { 1021 return folio_size(folio) >> fs_info->sectorsize_bits; 1022 } 1023 1024 bool __attribute_const__ btrfs_supported_blocksize(u32 blocksize); 1025 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info, 1026 enum btrfs_exclusive_operation type); 1027 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info, 1028 enum btrfs_exclusive_operation type); 1029 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info); 1030 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info); 1031 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info, 1032 enum btrfs_exclusive_operation op); 1033 1034 int btrfs_check_ioctl_vol_args_path(const struct btrfs_ioctl_vol_args *vol_args); 1035 1036 u16 btrfs_csum_type_size(u16 type); 1037 int btrfs_super_csum_size(const struct btrfs_super_block *s); 1038 const char *btrfs_super_csum_name(u16 csum_type); 1039 const char *btrfs_super_csum_driver(u16 csum_type); 1040 size_t __attribute_const__ btrfs_get_num_csums(void); 1041 1042 static inline bool btrfs_is_empty_uuid(const u8 *uuid) 1043 { 1044 return uuid_is_null((const uuid_t *)uuid); 1045 } 1046 1047 /* Compatibility and incompatibility defines */ 1048 void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag, 1049 const char *name); 1050 void __btrfs_clear_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag, 1051 const char *name); 1052 void __btrfs_set_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag, 1053 const char *name); 1054 void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag, 1055 const char *name); 1056 1057 #define __btrfs_fs_incompat(fs_info, flags) \ 1058 (!!(btrfs_super_incompat_flags((fs_info)->super_copy) & (flags))) 1059 1060 #define __btrfs_fs_compat_ro(fs_info, flags) \ 1061 (!!(btrfs_super_compat_ro_flags((fs_info)->super_copy) & (flags))) 1062 1063 #define btrfs_set_fs_incompat(__fs_info, opt) \ 1064 __btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt) 1065 1066 #define btrfs_clear_fs_incompat(__fs_info, opt) \ 1067 __btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt) 1068 1069 #define btrfs_fs_incompat(fs_info, opt) \ 1070 __btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt) 1071 1072 #define btrfs_set_fs_compat_ro(__fs_info, opt) \ 1073 __btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt) 1074 1075 #define btrfs_clear_fs_compat_ro(__fs_info, opt) \ 1076 __btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt) 1077 1078 #define btrfs_fs_compat_ro(fs_info, opt) \ 1079 __btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt) 1080 1081 #define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) 1082 #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) 1083 #define btrfs_raw_test_opt(o, opt) ((o) & BTRFS_MOUNT_##opt) 1084 #define btrfs_test_opt(fs_info, opt) ((fs_info)->mount_opt & \ 1085 BTRFS_MOUNT_##opt) 1086 1087 static inline int btrfs_fs_closing(const struct btrfs_fs_info *fs_info) 1088 { 1089 /* Do it this way so we only ever do one test_bit in the normal case. */ 1090 if (test_bit(BTRFS_FS_CLOSING_START, &fs_info->flags)) { 1091 if (test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)) 1092 return 2; 1093 return 1; 1094 } 1095 return 0; 1096 } 1097 1098 /* 1099 * If we remount the fs to be R/O or umount the fs, the cleaner needn't do 1100 * anything except sleeping. This function is used to check the status of 1101 * the fs. 1102 * We check for BTRFS_FS_STATE_RO to avoid races with a concurrent remount, 1103 * since setting and checking for SB_RDONLY in the superblock's flags is not 1104 * atomic. 1105 */ 1106 static inline int btrfs_need_cleaner_sleep(const struct btrfs_fs_info *fs_info) 1107 { 1108 return test_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state) || 1109 btrfs_fs_closing(fs_info); 1110 } 1111 1112 static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info *fs_info) 1113 { 1114 clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags); 1115 } 1116 1117 #define BTRFS_FS_ERROR(fs_info) (READ_ONCE((fs_info)->fs_error)) 1118 1119 #define BTRFS_FS_LOG_CLEANUP_ERROR(fs_info) \ 1120 (unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR, \ 1121 &(fs_info)->fs_state))) 1122 1123 /* 1124 * We use folio flag owner_2 to indicate there is an ordered extent with 1125 * unfinished IO. 1126 */ 1127 #define folio_test_ordered(folio) folio_test_owner_2(folio) 1128 #define folio_set_ordered(folio) folio_set_owner_2(folio) 1129 #define folio_clear_ordered(folio) folio_clear_owner_2(folio) 1130 1131 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 1132 1133 #define EXPORT_FOR_TESTS 1134 1135 static inline bool btrfs_is_testing(const struct btrfs_fs_info *fs_info) 1136 { 1137 return unlikely(test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state)); 1138 } 1139 1140 void btrfs_test_destroy_inode(struct inode *inode); 1141 1142 #else 1143 1144 #define EXPORT_FOR_TESTS static 1145 1146 static inline bool btrfs_is_testing(const struct btrfs_fs_info *fs_info) 1147 { 1148 return false; 1149 } 1150 #endif 1151 1152 #endif 1153