fs/btrfs/fs.h

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