1 /* SPDX-License-Identifier: GPL-2.0 */
6 #include <linux/blkdev.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"
15 #define BTRFS_MAX_EXTENT_SIZE SZ_128M
17 #define BTRFS_OLDEST_GENERATION 0ULL
19 #define BTRFS_EMPTY_DIR_SIZE 0
21 #define BTRFS_DIRTY_METADATA_THRESH SZ_32M
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
);
28 * Number of metadata items necessary for an unlink operation:
30 * 1 for the possible orphan item
35 * 1 for the parent inode
37 #define BTRFS_UNLINK_METADATA_UNITS 6
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.
44 #define BTRFS_DEVICE_RANGE_RESERVED (SZ_1M)
46 * Runtime (in-memory) states of filesystem
50 * Filesystem is being remounted, allow to skip some operations, like
53 BTRFS_FS_STATE_REMOUNTING
,
54 /* Filesystem in RO mode */
56 /* Track if a transaction abort has been reported on this filesystem */
57 BTRFS_FS_STATE_TRANS_ABORTED
,
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
62 BTRFS_FS_STATE_DEV_REPLACING
,
63 /* The btrfs_fs_info created for self-tests */
64 BTRFS_FS_STATE_DUMMY_FS_INFO
,
66 BTRFS_FS_STATE_NO_CSUMS
,
68 /* Indicates there was an error cleaning up a log tree. */
69 BTRFS_FS_STATE_LOG_CLEANUP_ERROR
,
75 BTRFS_FS_CLOSING_START
,
76 BTRFS_FS_CLOSING_DONE
,
77 BTRFS_FS_LOG_RECOVERING
,
79 BTRFS_FS_QUOTA_ENABLED
,
80 BTRFS_FS_UPDATE_UUID_TREE_GEN
,
81 BTRFS_FS_CREATING_FREE_SPACE_TREE
,
85 BTRFS_FS_QUOTA_OVERRIDE
,
86 /* Used to record internally whether fs has been frozen */
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.
92 BTRFS_FS_BALANCE_RUNNING
,
95 * Indicate that relocation of a chunk has started, it's set per chunk
96 * and is toggled between chunks.
98 BTRFS_FS_RELOC_RUNNING
,
100 /* Indicate that the cleaner thread is awake and doing something. */
101 BTRFS_FS_CLEANER_RUNNING
,
104 * The checksumming has an optimized version and is considered fast,
105 * so we don't need to offload checksums to workqueues.
107 BTRFS_FS_CSUM_IMPL_FAST
,
109 /* Indicate that the discard workqueue can service discards. */
110 BTRFS_FS_DISCARD_RUNNING
,
112 /* Indicate that we need to cleanup space cache v1 */
113 BTRFS_FS_CLEANUP_SPACE_CACHE_V1
,
115 /* Indicate that we can't trust the free space tree for caching yet */
116 BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED
,
118 /* Indicate whether there are any tree modification log users */
119 BTRFS_FS_TREE_MOD_LOG_USERS
,
121 /* Indicate that we want the transaction kthread to commit right now. */
122 BTRFS_FS_COMMIT_TRANS
,
124 /* Indicate we have half completed snapshot deletions pending. */
125 BTRFS_FS_UNFINISHED_DROPS
,
127 /* Indicate we have to finish a zone to do next allocation. */
128 BTRFS_FS_NEED_ZONE_FINISH
,
130 /* Indicate that we want to commit the transaction. */
131 BTRFS_FS_NEED_TRANS_COMMIT
,
133 /* This is set when active zone tracking is needed. */
134 BTRFS_FS_ACTIVE_ZONE_TRACKING
,
137 * Indicate if we have some features changed, this is mostly for
138 * cleaner thread to update the sysfs interface.
140 BTRFS_FS_FEATURE_CHANGED
,
142 #if BITS_PER_LONG == 32
143 /* Indicate if we have error/warn message printed on 32bit systems */
144 BTRFS_FS_32BIT_ERROR
,
150 * Flags for mount options.
152 * Note: don't forget to add new options to btrfs_show_options()
155 BTRFS_MOUNT_NODATASUM
= (1UL << 0),
156 BTRFS_MOUNT_NODATACOW
= (1UL << 1),
157 BTRFS_MOUNT_NOBARRIER
= (1UL << 2),
158 BTRFS_MOUNT_SSD
= (1UL << 3),
159 BTRFS_MOUNT_DEGRADED
= (1UL << 4),
160 BTRFS_MOUNT_COMPRESS
= (1UL << 5),
161 BTRFS_MOUNT_NOTREELOG
= (1UL << 6),
162 BTRFS_MOUNT_FLUSHONCOMMIT
= (1UL << 7),
163 BTRFS_MOUNT_SSD_SPREAD
= (1UL << 8),
164 BTRFS_MOUNT_NOSSD
= (1UL << 9),
165 BTRFS_MOUNT_DISCARD_SYNC
= (1UL << 10),
166 BTRFS_MOUNT_FORCE_COMPRESS
= (1UL << 11),
167 BTRFS_MOUNT_SPACE_CACHE
= (1UL << 12),
168 BTRFS_MOUNT_CLEAR_CACHE
= (1UL << 13),
169 BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED
= (1UL << 14),
170 BTRFS_MOUNT_ENOSPC_DEBUG
= (1UL << 15),
171 BTRFS_MOUNT_AUTO_DEFRAG
= (1UL << 16),
172 BTRFS_MOUNT_USEBACKUPROOT
= (1UL << 17),
173 BTRFS_MOUNT_SKIP_BALANCE
= (1UL << 18),
174 BTRFS_MOUNT_CHECK_INTEGRITY
= (1UL << 19),
175 BTRFS_MOUNT_CHECK_INTEGRITY_DATA
= (1UL << 20),
176 BTRFS_MOUNT_PANIC_ON_FATAL_ERROR
= (1UL << 21),
177 BTRFS_MOUNT_RESCAN_UUID_TREE
= (1UL << 22),
178 BTRFS_MOUNT_FRAGMENT_DATA
= (1UL << 23),
179 BTRFS_MOUNT_FRAGMENT_METADATA
= (1UL << 24),
180 BTRFS_MOUNT_FREE_SPACE_TREE
= (1UL << 25),
181 BTRFS_MOUNT_NOLOGREPLAY
= (1UL << 26),
182 BTRFS_MOUNT_REF_VERIFY
= (1UL << 27),
183 BTRFS_MOUNT_DISCARD_ASYNC
= (1UL << 28),
184 BTRFS_MOUNT_IGNOREBADROOTS
= (1UL << 29),
185 BTRFS_MOUNT_IGNOREDATACSUMS
= (1UL << 30),
186 BTRFS_MOUNT_NODISCARD
= (1UL << 31),
190 * Compat flags that we support. If any incompat flags are set other than the
191 * ones specified below then we will fail to mount
193 #define BTRFS_FEATURE_COMPAT_SUPP 0ULL
194 #define BTRFS_FEATURE_COMPAT_SAFE_SET 0ULL
195 #define BTRFS_FEATURE_COMPAT_SAFE_CLEAR 0ULL
197 #define BTRFS_FEATURE_COMPAT_RO_SUPP \
198 (BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE | \
199 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID | \
200 BTRFS_FEATURE_COMPAT_RO_VERITY | \
201 BTRFS_FEATURE_COMPAT_RO_BLOCK_GROUP_TREE)
203 #define BTRFS_FEATURE_COMPAT_RO_SAFE_SET 0ULL
204 #define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR 0ULL
206 #define BTRFS_FEATURE_INCOMPAT_SUPP_STABLE \
207 (BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF | \
208 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL | \
209 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS | \
210 BTRFS_FEATURE_INCOMPAT_BIG_METADATA | \
211 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO | \
212 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD | \
213 BTRFS_FEATURE_INCOMPAT_RAID56 | \
214 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF | \
215 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA | \
216 BTRFS_FEATURE_INCOMPAT_NO_HOLES | \
217 BTRFS_FEATURE_INCOMPAT_METADATA_UUID | \
218 BTRFS_FEATURE_INCOMPAT_RAID1C34 | \
219 BTRFS_FEATURE_INCOMPAT_ZONED)
221 #ifdef CONFIG_BTRFS_DEBUG
223 * Features under developmen like Extent tree v2 support is enabled
224 * only under CONFIG_BTRFS_DEBUG.
226 #define BTRFS_FEATURE_INCOMPAT_SUPP \
227 (BTRFS_FEATURE_INCOMPAT_SUPP_STABLE | \
228 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2)
232 #define BTRFS_FEATURE_INCOMPAT_SUPP \
233 (BTRFS_FEATURE_INCOMPAT_SUPP_STABLE)
237 #define BTRFS_FEATURE_INCOMPAT_SAFE_SET \
238 (BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
239 #define BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR 0ULL
241 #define BTRFS_DEFAULT_COMMIT_INTERVAL (30)
242 #define BTRFS_DEFAULT_MAX_INLINE (2048)
244 struct btrfs_dev_replace
{
245 /* See #define above */
247 /* Seconds since 1-Jan-1970 */
248 time64_t time_started
;
249 /* Seconds since 1-Jan-1970 */
250 time64_t time_stopped
;
251 atomic64_t num_write_errors
;
252 atomic64_t num_uncorrectable_read_errors
;
255 u64 committed_cursor_left
;
256 u64 cursor_left_last_write_of_item
;
259 /* See #define above */
260 u64 cont_reading_from_srcdev_mode
;
263 int item_needs_writeback
;
264 struct btrfs_device
*srcdev
;
265 struct btrfs_device
*tgtdev
;
267 struct mutex lock_finishing_cancel_unmount
;
268 struct rw_semaphore rwsem
;
270 struct btrfs_scrub_progress scrub_progress
;
272 struct percpu_counter bio_counter
;
273 wait_queue_head_t replace_wait
;
277 * Free clusters are used to claim free space in relatively large chunks,
278 * allowing us to do less seeky writes. They are used for all metadata
279 * allocations. In ssd_spread mode they are also used for data allocations.
281 struct btrfs_free_cluster
{
283 spinlock_t refill_lock
;
286 /* Largest extent in this cluster */
289 /* First extent starting offset */
292 /* We did a full search and couldn't create a cluster */
295 struct btrfs_block_group
*block_group
;
297 * When a cluster is allocated from a block group, we put the cluster
298 * onto a list in the block group so that it can be freed before the
299 * block group is freed.
301 struct list_head block_group_list
;
304 /* Discard control. */
306 * Async discard uses multiple lists to differentiate the discard filter
307 * parameters. Index 0 is for completely free block groups where we need to
308 * ensure the entire block group is trimmed without being lossy. Indices
309 * afterwards represent monotonically decreasing discard filter sizes to
310 * prioritize what should be discarded next.
312 #define BTRFS_NR_DISCARD_LISTS 3
313 #define BTRFS_DISCARD_INDEX_UNUSED 0
314 #define BTRFS_DISCARD_INDEX_START 1
316 struct btrfs_discard_ctl
{
317 struct workqueue_struct
*discard_workers
;
318 struct delayed_work work
;
320 struct btrfs_block_group
*block_group
;
321 struct list_head discard_list
[BTRFS_NR_DISCARD_LISTS
];
323 u64 prev_discard_time
;
324 atomic_t discardable_extents
;
325 atomic64_t discardable_bytes
;
326 u64 max_discard_size
;
330 u64 discard_extent_bytes
;
331 u64 discard_bitmap_bytes
;
332 atomic64_t discard_bytes_saved
;
336 * Exclusive operations (device replace, resize, device add/remove, balance)
338 enum btrfs_exclusive_operation
{
340 BTRFS_EXCLOP_BALANCE_PAUSED
,
341 BTRFS_EXCLOP_BALANCE
,
342 BTRFS_EXCLOP_DEV_ADD
,
343 BTRFS_EXCLOP_DEV_REMOVE
,
344 BTRFS_EXCLOP_DEV_REPLACE
,
346 BTRFS_EXCLOP_SWAP_ACTIVATE
,
349 /* Store data about transaction commits, exported via sysfs. */
350 struct btrfs_commit_stats
{
351 /* Total number of commits */
353 /* The maximum commit duration so far in ns */
355 /* The last commit duration in ns */
357 /* The total commit duration in ns */
358 u64 total_commit_dur
;
361 struct btrfs_fs_info
{
362 u8 chunk_tree_uuid
[BTRFS_UUID_SIZE
];
364 struct btrfs_root
*tree_root
;
365 struct btrfs_root
*chunk_root
;
366 struct btrfs_root
*dev_root
;
367 struct btrfs_root
*fs_root
;
368 struct btrfs_root
*quota_root
;
369 struct btrfs_root
*uuid_root
;
370 struct btrfs_root
*data_reloc_root
;
371 struct btrfs_root
*block_group_root
;
373 /* The log root tree is a directory of all the other log roots */
374 struct btrfs_root
*log_root_tree
;
376 /* The tree that holds the global roots (csum, extent, etc) */
377 rwlock_t global_root_lock
;
378 struct rb_root global_root_tree
;
380 spinlock_t fs_roots_radix_lock
;
381 struct radix_tree_root fs_roots_radix
;
383 /* Block group cache stuff */
384 rwlock_t block_group_cache_lock
;
385 struct rb_root_cached block_group_cache_tree
;
387 /* Keep track of unallocated space */
388 atomic64_t free_chunk_space
;
390 /* Track ranges which are used by log trees blocks/logged data extents */
391 struct extent_io_tree excluded_extents
;
393 /* logical->physical extent mapping */
394 struct extent_map_tree mapping_tree
;
397 * Block reservation for extent, checksum, root tree and delayed dir
400 struct btrfs_block_rsv global_block_rsv
;
401 /* Block reservation for metadata operations */
402 struct btrfs_block_rsv trans_block_rsv
;
403 /* Block reservation for chunk tree */
404 struct btrfs_block_rsv chunk_block_rsv
;
405 /* Block reservation for delayed operations */
406 struct btrfs_block_rsv delayed_block_rsv
;
407 /* Block reservation for delayed refs */
408 struct btrfs_block_rsv delayed_refs_rsv
;
410 struct btrfs_block_rsv empty_block_rsv
;
413 u64 last_trans_committed
;
415 * Generation of the last transaction used for block group relocation
416 * since the filesystem was last mounted (or 0 if none happened yet).
417 * Must be written and read while holding btrfs_fs_info::commit_root_sem.
419 u64 last_reloc_trans
;
422 * This is updated to the current trans every time a full commit is
423 * required instead of the faster short fsync log commits
425 u64 last_trans_log_full_commit
;
426 unsigned long mount_opt
;
428 unsigned long compress_type
:4;
429 unsigned int compress_level
;
432 * It is a suggestive number, the read side is safe even it gets a
433 * wrong number because we will write out the data into a regular
434 * extent. The write side(mount/remount) is under ->s_umount lock,
435 * so it is also safe.
439 struct btrfs_transaction
*running_transaction
;
440 wait_queue_head_t transaction_throttle
;
441 wait_queue_head_t transaction_wait
;
442 wait_queue_head_t transaction_blocked_wait
;
443 wait_queue_head_t async_submit_wait
;
446 * Used to protect the incompat_flags, compat_flags, compat_ro_flags
447 * when they are updated.
449 * Because we do not clear the flags for ever, so we needn't use
450 * the lock on the read side.
452 * We also needn't use the lock when we mount the fs, because
453 * there is no other task which will update the flag.
455 spinlock_t super_lock
;
456 struct btrfs_super_block
*super_copy
;
457 struct btrfs_super_block
*super_for_commit
;
458 struct super_block
*sb
;
459 struct inode
*btree_inode
;
460 struct mutex tree_log_mutex
;
461 struct mutex transaction_kthread_mutex
;
462 struct mutex cleaner_mutex
;
463 struct mutex chunk_mutex
;
466 * This is taken to make sure we don't set block groups ro after the
467 * free space cache has been allocated on them.
469 struct mutex ro_block_group_mutex
;
472 * This is used during read/modify/write to make sure no two ios are
473 * trying to mod the same stripe at the same time.
475 struct btrfs_stripe_hash_table
*stripe_hash_table
;
478 * This protects the ordered operations list only while we are
479 * processing all of the entries on it. This way we make sure the
480 * commit code doesn't find the list temporarily empty because another
481 * function happens to be doing non-waiting preflush before jumping
482 * into the main commit.
484 struct mutex ordered_operations_mutex
;
486 struct rw_semaphore commit_root_sem
;
488 struct rw_semaphore cleanup_work_sem
;
490 struct rw_semaphore subvol_sem
;
492 spinlock_t trans_lock
;
494 * The reloc mutex goes with the trans lock, it is taken during commit
495 * to protect us from the relocation code.
497 struct mutex reloc_mutex
;
499 struct list_head trans_list
;
500 struct list_head dead_roots
;
501 struct list_head caching_block_groups
;
503 spinlock_t delayed_iput_lock
;
504 struct list_head delayed_iputs
;
505 atomic_t nr_delayed_iputs
;
506 wait_queue_head_t delayed_iputs_wait
;
508 atomic64_t tree_mod_seq
;
510 /* This protects tree_mod_log and tree_mod_seq_list */
511 rwlock_t tree_mod_log_lock
;
512 struct rb_root tree_mod_log
;
513 struct list_head tree_mod_seq_list
;
515 atomic_t async_delalloc_pages
;
517 /* This is used to protect the following list -- ordered_roots. */
518 spinlock_t ordered_root_lock
;
521 * All fs/file tree roots in which there are data=ordered extents
522 * pending writeback are added into this list.
524 * These can span multiple transactions and basically include every
525 * dirty data page that isn't from nodatacow.
527 struct list_head ordered_roots
;
529 struct mutex delalloc_root_mutex
;
530 spinlock_t delalloc_root_lock
;
531 /* All fs/file tree roots that have delalloc inodes. */
532 struct list_head delalloc_roots
;
535 * There is a pool of worker threads for checksumming during writes and
536 * a pool for checksumming after reads. This is because readers can
537 * run with FS locks held, and the writers may be waiting for those
538 * locks. We don't want ordering in the pending list to cause
539 * deadlocks, and so the two are serviced separately.
541 * A third pool does submit_bio to avoid deadlocking with the other two.
543 struct btrfs_workqueue
*workers
;
544 struct btrfs_workqueue
*delalloc_workers
;
545 struct btrfs_workqueue
*flush_workers
;
546 struct workqueue_struct
*endio_workers
;
547 struct workqueue_struct
*endio_meta_workers
;
548 struct workqueue_struct
*rmw_workers
;
549 struct workqueue_struct
*compressed_write_workers
;
550 struct btrfs_workqueue
*endio_write_workers
;
551 struct btrfs_workqueue
*endio_freespace_worker
;
552 struct btrfs_workqueue
*caching_workers
;
555 * Fixup workers take dirty pages that didn't properly go through the
556 * cow mechanism and make them safe to write. It happens for the
557 * sys_munmap function call path.
559 struct btrfs_workqueue
*fixup_workers
;
560 struct btrfs_workqueue
*delayed_workers
;
562 struct task_struct
*transaction_kthread
;
563 struct task_struct
*cleaner_kthread
;
564 u32 thread_pool_size
;
566 struct kobject
*space_info_kobj
;
567 struct kobject
*qgroups_kobj
;
568 struct kobject
*discard_kobj
;
570 /* Used to keep from writing metadata until there is a nice batch */
571 struct percpu_counter dirty_metadata_bytes
;
572 struct percpu_counter delalloc_bytes
;
573 struct percpu_counter ordered_bytes
;
574 s32 dirty_metadata_batch
;
577 /* Protected by 'trans_lock'. */
578 struct list_head dirty_cowonly_roots
;
580 struct btrfs_fs_devices
*fs_devices
;
583 * The space_info list is effectively read only after initial setup.
584 * It is populated at mount time and cleaned up after all block groups
585 * are removed. RCU is used to protect it.
587 struct list_head space_info
;
589 struct btrfs_space_info
*data_sinfo
;
591 struct reloc_control
*reloc_ctl
;
593 /* data_alloc_cluster is only used in ssd_spread mode */
594 struct btrfs_free_cluster data_alloc_cluster
;
596 /* All metadata allocations go through this cluster. */
597 struct btrfs_free_cluster meta_alloc_cluster
;
599 /* Auto defrag inodes go here. */
600 spinlock_t defrag_inodes_lock
;
601 struct rb_root defrag_inodes
;
602 atomic_t defrag_running
;
604 /* Used to protect avail_{data, metadata, system}_alloc_bits */
605 seqlock_t profiles_lock
;
607 * These three are in extended format (availability of single chunks is
608 * denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other types are denoted
609 * by corresponding BTRFS_BLOCK_GROUP_* bits)
611 u64 avail_data_alloc_bits
;
612 u64 avail_metadata_alloc_bits
;
613 u64 avail_system_alloc_bits
;
616 spinlock_t balance_lock
;
617 struct mutex balance_mutex
;
618 atomic_t balance_pause_req
;
619 atomic_t balance_cancel_req
;
620 struct btrfs_balance_control
*balance_ctl
;
621 wait_queue_head_t balance_wait_q
;
623 /* Cancellation requests for chunk relocation */
624 atomic_t reloc_cancel_req
;
626 u32 data_chunk_allocations
;
631 /* Private scrub information */
632 struct mutex scrub_lock
;
633 atomic_t scrubs_running
;
634 atomic_t scrub_pause_req
;
635 atomic_t scrubs_paused
;
636 atomic_t scrub_cancel_req
;
637 wait_queue_head_t scrub_pause_wait
;
639 * The worker pointers are NULL iff the refcount is 0, ie. scrub is not
642 refcount_t scrub_workers_refcnt
;
643 struct workqueue_struct
*scrub_workers
;
644 struct btrfs_subpage_info
*subpage_info
;
646 struct btrfs_discard_ctl discard_ctl
;
648 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
649 u32 check_integrity_print_mask
;
651 /* Is qgroup tracking in a consistent state? */
654 /* Holds configuration and tracking. Protected by qgroup_lock. */
655 struct rb_root qgroup_tree
;
656 spinlock_t qgroup_lock
;
659 * Used to avoid frequently calling ulist_alloc()/ulist_free()
660 * when doing qgroup accounting, it must be protected by qgroup_lock.
662 struct ulist
*qgroup_ulist
;
665 * Protect user change for quota operations. If a transaction is needed,
666 * it must be started before locking this lock.
668 struct mutex qgroup_ioctl_lock
;
670 /* List of dirty qgroups to be written at next commit. */
671 struct list_head dirty_qgroups
;
673 /* Used by qgroup for an efficient tree traversal. */
676 /* Qgroup rescan items. */
677 /* Protects the progress item */
678 struct mutex qgroup_rescan_lock
;
679 struct btrfs_key qgroup_rescan_progress
;
680 struct btrfs_workqueue
*qgroup_rescan_workers
;
681 struct completion qgroup_rescan_completion
;
682 struct btrfs_work qgroup_rescan_work
;
683 /* Protected by qgroup_rescan_lock */
684 bool qgroup_rescan_running
;
685 u8 qgroup_drop_subtree_thres
;
688 * If this is not 0, then it indicates a serious filesystem error has
689 * happened and it contains that error (negative errno value).
693 /* Filesystem state */
694 unsigned long fs_state
;
696 struct btrfs_delayed_root
*delayed_root
;
698 /* Extent buffer radix tree */
699 spinlock_t buffer_lock
;
700 /* Entries are eb->start / sectorsize */
701 struct radix_tree_root buffer_radix
;
703 /* Next backup root to be overwritten */
704 int backup_root_index
;
706 /* Device replace state */
707 struct btrfs_dev_replace dev_replace
;
709 struct semaphore uuid_tree_rescan_sem
;
711 /* Used to reclaim the metadata space in the background. */
712 struct work_struct async_reclaim_work
;
713 struct work_struct async_data_reclaim_work
;
714 struct work_struct preempt_reclaim_work
;
716 /* Reclaim partially filled block groups in the background */
717 struct work_struct reclaim_bgs_work
;
718 struct list_head reclaim_bgs
;
719 int bg_reclaim_threshold
;
721 spinlock_t unused_bgs_lock
;
722 struct list_head unused_bgs
;
723 struct mutex unused_bg_unpin_mutex
;
724 /* Protect block groups that are going to be deleted */
725 struct mutex reclaim_bgs_lock
;
727 /* Cached block sizes */
730 /* ilog2 of sectorsize, use to avoid 64bit division */
737 * Maximum size of an extent. BTRFS_MAX_EXTENT_SIZE on regular
738 * filesystem, on zoned it depends on the device constraints.
742 /* Block groups and devices containing active swapfiles. */
743 spinlock_t swapfile_pins_lock
;
744 struct rb_root swapfile_pins
;
746 struct crypto_shash
*csum_shash
;
748 /* Type of exclusive operation running, protected by super_lock */
749 enum btrfs_exclusive_operation exclusive_operation
;
752 * Zone size > 0 when in ZONED mode, otherwise it's used for a check
753 * if the mode is enabled
757 /* Constraints for ZONE_APPEND commands: */
758 struct queue_limits limits
;
759 u64 max_zone_append_size
;
761 struct mutex zoned_meta_io_lock
;
762 spinlock_t treelog_bg_lock
;
766 * Start of the dedicated data relocation block group, protected by
767 * relocation_bg_lock.
769 spinlock_t relocation_bg_lock
;
771 struct mutex zoned_data_reloc_io_lock
;
773 struct btrfs_block_group
*active_meta_bg
;
774 struct btrfs_block_group
*active_system_bg
;
778 spinlock_t zone_active_bgs_lock
;
779 struct list_head zone_active_bgs
;
781 /* Updates are not protected by any lock */
782 struct btrfs_commit_stats commit_stats
;
785 * Last generation where we dropped a non-relocation root.
786 * Use btrfs_set_last_root_drop_gen() and btrfs_get_last_root_drop_gen()
787 * to change it and to read it, respectively.
789 u64 last_root_drop_gen
;
792 * Annotations for transaction events (structures are empty when
793 * compiled without lockdep).
795 struct lockdep_map btrfs_trans_num_writers_map
;
796 struct lockdep_map btrfs_trans_num_extwriters_map
;
797 struct lockdep_map btrfs_state_change_map
[4];
798 struct lockdep_map btrfs_trans_pending_ordered_map
;
799 struct lockdep_map btrfs_ordered_extent_map
;
801 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
802 spinlock_t ref_verify_lock
;
803 struct rb_root block_tree
;
806 #ifdef CONFIG_BTRFS_DEBUG
807 struct kobject
*debug_kobj
;
808 struct list_head allocated_roots
;
810 spinlock_t eb_leak_lock
;
811 struct list_head allocated_ebs
;
815 static inline void btrfs_set_last_root_drop_gen(struct btrfs_fs_info
*fs_info
,
818 WRITE_ONCE(fs_info
->last_root_drop_gen
, gen
);
821 static inline u64
btrfs_get_last_root_drop_gen(const struct btrfs_fs_info
*fs_info
)
823 return READ_ONCE(fs_info
->last_root_drop_gen
);
827 * Take the number of bytes to be checksummed and figure out how many leaves
828 * it would require to store the csums for that many bytes.
830 static inline u64
btrfs_csum_bytes_to_leaves(
831 const struct btrfs_fs_info
*fs_info
, u64 csum_bytes
)
833 const u64 num_csums
= csum_bytes
>> fs_info
->sectorsize_bits
;
835 return DIV_ROUND_UP_ULL(num_csums
, fs_info
->csums_per_leaf
);
839 * Use this if we would be adding new items, as we could split nodes as we cow
842 static inline u64
btrfs_calc_insert_metadata_size(const struct btrfs_fs_info
*fs_info
,
845 return (u64
)fs_info
->nodesize
* BTRFS_MAX_LEVEL
* 2 * num_items
;
849 * Doing a truncate or a modification won't result in new nodes or leaves, just
850 * what we need for COW.
852 static inline u64
btrfs_calc_metadata_size(const struct btrfs_fs_info
*fs_info
,
855 return (u64
)fs_info
->nodesize
* BTRFS_MAX_LEVEL
* num_items
;
858 #define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r->fs_info) >> 4) - \
859 sizeof(struct btrfs_item))
861 static inline bool btrfs_is_zoned(const struct btrfs_fs_info
*fs_info
)
863 return IS_ENABLED(CONFIG_BLK_DEV_ZONED
) && fs_info
->zone_size
> 0;
867 * Count how many fs_info->max_extent_size cover the @size
869 static inline u32
count_max_extents(struct btrfs_fs_info
*fs_info
, u64 size
)
871 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
873 return div_u64(size
+ BTRFS_MAX_EXTENT_SIZE
- 1, BTRFS_MAX_EXTENT_SIZE
);
876 return div_u64(size
+ fs_info
->max_extent_size
- 1, fs_info
->max_extent_size
);
879 bool btrfs_exclop_start(struct btrfs_fs_info
*fs_info
,
880 enum btrfs_exclusive_operation type
);
881 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info
*fs_info
,
882 enum btrfs_exclusive_operation type
);
883 void btrfs_exclop_start_unlock(struct btrfs_fs_info
*fs_info
);
884 void btrfs_exclop_finish(struct btrfs_fs_info
*fs_info
);
885 void btrfs_exclop_balance(struct btrfs_fs_info
*fs_info
,
886 enum btrfs_exclusive_operation op
);
888 /* Compatibility and incompatibility defines */
889 void __btrfs_set_fs_incompat(struct btrfs_fs_info
*fs_info
, u64 flag
,
891 void __btrfs_clear_fs_incompat(struct btrfs_fs_info
*fs_info
, u64 flag
,
893 void __btrfs_set_fs_compat_ro(struct btrfs_fs_info
*fs_info
, u64 flag
,
895 void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info
*fs_info
, u64 flag
,
898 #define __btrfs_fs_incompat(fs_info, flags) \
899 (!!(btrfs_super_incompat_flags((fs_info)->super_copy) & (flags)))
901 #define __btrfs_fs_compat_ro(fs_info, flags) \
902 (!!(btrfs_super_compat_ro_flags((fs_info)->super_copy) & (flags)))
904 #define btrfs_set_fs_incompat(__fs_info, opt) \
905 __btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
907 #define btrfs_clear_fs_incompat(__fs_info, opt) \
908 __btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
910 #define btrfs_fs_incompat(fs_info, opt) \
911 __btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
913 #define btrfs_set_fs_compat_ro(__fs_info, opt) \
914 __btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
916 #define btrfs_clear_fs_compat_ro(__fs_info, opt) \
917 __btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
919 #define btrfs_fs_compat_ro(fs_info, opt) \
920 __btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
922 #define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt)
923 #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt)
924 #define btrfs_raw_test_opt(o, opt) ((o) & BTRFS_MOUNT_##opt)
925 #define btrfs_test_opt(fs_info, opt) ((fs_info)->mount_opt & \
928 #define btrfs_set_and_info(fs_info, opt, fmt, args...) \
930 if (!btrfs_test_opt(fs_info, opt)) \
931 btrfs_info(fs_info, fmt, ##args); \
932 btrfs_set_opt(fs_info->mount_opt, opt); \
935 #define btrfs_clear_and_info(fs_info, opt, fmt, args...) \
937 if (btrfs_test_opt(fs_info, opt)) \
938 btrfs_info(fs_info, fmt, ##args); \
939 btrfs_clear_opt(fs_info->mount_opt, opt); \
942 static inline int btrfs_fs_closing(struct btrfs_fs_info
*fs_info
)
944 /* Do it this way so we only ever do one test_bit in the normal case. */
945 if (test_bit(BTRFS_FS_CLOSING_START
, &fs_info
->flags
)) {
946 if (test_bit(BTRFS_FS_CLOSING_DONE
, &fs_info
->flags
))
954 * If we remount the fs to be R/O or umount the fs, the cleaner needn't do
955 * anything except sleeping. This function is used to check the status of
957 * We check for BTRFS_FS_STATE_RO to avoid races with a concurrent remount,
958 * since setting and checking for SB_RDONLY in the superblock's flags is not
961 static inline int btrfs_need_cleaner_sleep(struct btrfs_fs_info
*fs_info
)
963 return test_bit(BTRFS_FS_STATE_RO
, &fs_info
->fs_state
) ||
964 btrfs_fs_closing(fs_info
);
967 static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info
*fs_info
)
969 clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS
, &fs_info
->flags
);
972 #define BTRFS_FS_ERROR(fs_info) (READ_ONCE((fs_info)->fs_error))
974 #define BTRFS_FS_LOG_CLEANUP_ERROR(fs_info) \
975 (unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR, \
976 &(fs_info)->fs_state)))
978 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
980 #define EXPORT_FOR_TESTS
982 static inline int btrfs_is_testing(struct btrfs_fs_info
*fs_info
)
984 return test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO
, &fs_info
->fs_state
);
987 void btrfs_test_destroy_inode(struct inode
*inode
);
991 #define EXPORT_FOR_TESTS static
993 static inline int btrfs_is_testing(struct btrfs_fs_info
*fs_info
)