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