overlayfs: Implement splice-read
[linux-block.git] / fs / btrfs / fs.h
CommitLineData
c7f13d42
JB
1/* SPDX-License-Identifier: GPL-2.0 */
2
3#ifndef BTRFS_FS_H
4#define BTRFS_FS_H
5
243cf8d1 6#include <linux/blkdev.h>
3683fbbc
JB
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
a56159d4
JB
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
25static_assert(sizeof(struct btrfs_super_block) == BTRFS_SUPER_INFO_SIZE);
26
5630e2bc
FM
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
a56159d4
JB
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)
ec8eb376
JB
45/*
46 * Runtime (in-memory) states of filesystem
47 */
48enum {
49 /* Global indicator of serious filesystem errors */
50 BTRFS_FS_STATE_ERROR,
51 /*
52 * Filesystem is being remounted, allow to skip some operations, like
53 * defrag
54 */
55 BTRFS_FS_STATE_REMOUNTING,
56 /* Filesystem in RO mode */
57 BTRFS_FS_STATE_RO,
58 /* Track if a transaction abort has been reported on this filesystem */
59 BTRFS_FS_STATE_TRANS_ABORTED,
60 /*
61 * Bio operations should be blocked on this filesystem because a source
62 * or target device is being destroyed as part of a device replace
63 */
64 BTRFS_FS_STATE_DEV_REPLACING,
65 /* The btrfs_fs_info created for self-tests */
66 BTRFS_FS_STATE_DUMMY_FS_INFO,
67
68 BTRFS_FS_STATE_NO_CSUMS,
69
70 /* Indicates there was an error cleaning up a log tree. */
71 BTRFS_FS_STATE_LOG_CLEANUP_ERROR,
72
73 BTRFS_FS_STATE_COUNT
74};
75
7966a6b5
JB
76enum {
77 BTRFS_FS_CLOSING_START,
78 BTRFS_FS_CLOSING_DONE,
79 BTRFS_FS_LOG_RECOVERING,
80 BTRFS_FS_OPEN,
81 BTRFS_FS_QUOTA_ENABLED,
82 BTRFS_FS_UPDATE_UUID_TREE_GEN,
83 BTRFS_FS_CREATING_FREE_SPACE_TREE,
84 BTRFS_FS_BTREE_ERR,
85 BTRFS_FS_LOG1_ERR,
86 BTRFS_FS_LOG2_ERR,
87 BTRFS_FS_QUOTA_OVERRIDE,
88 /* Used to record internally whether fs has been frozen */
89 BTRFS_FS_FROZEN,
90 /*
91 * Indicate that balance has been set up from the ioctl and is in the
92 * main phase. The fs_info::balance_ctl is initialized.
93 */
94 BTRFS_FS_BALANCE_RUNNING,
95
96 /*
97 * Indicate that relocation of a chunk has started, it's set per chunk
98 * and is toggled between chunks.
99 */
100 BTRFS_FS_RELOC_RUNNING,
101
102 /* Indicate that the cleaner thread is awake and doing something. */
103 BTRFS_FS_CLEANER_RUNNING,
104
105 /*
106 * The checksumming has an optimized version and is considered fast,
107 * so we don't need to offload checksums to workqueues.
108 */
109 BTRFS_FS_CSUM_IMPL_FAST,
110
111 /* Indicate that the discard workqueue can service discards. */
112 BTRFS_FS_DISCARD_RUNNING,
113
114 /* Indicate that we need to cleanup space cache v1 */
115 BTRFS_FS_CLEANUP_SPACE_CACHE_V1,
116
117 /* Indicate that we can't trust the free space tree for caching yet */
118 BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED,
119
120 /* Indicate whether there are any tree modification log users */
121 BTRFS_FS_TREE_MOD_LOG_USERS,
122
123 /* Indicate that we want the transaction kthread to commit right now. */
124 BTRFS_FS_COMMIT_TRANS,
125
126 /* Indicate we have half completed snapshot deletions pending. */
127 BTRFS_FS_UNFINISHED_DROPS,
128
129 /* Indicate we have to finish a zone to do next allocation. */
130 BTRFS_FS_NEED_ZONE_FINISH,
131
c52cc7b7
JB
132 /* Indicate that we want to commit the transaction. */
133 BTRFS_FS_NEED_TRANS_COMMIT,
134
bf1f1fec
JB
135 /* This is set when active zone tracking is needed. */
136 BTRFS_FS_ACTIVE_ZONE_TRACKING,
85e79ec7 137
b7625f46
QW
138 /*
139 * Indicate if we have some features changed, this is mostly for
140 * cleaner thread to update the sysfs interface.
141 */
142 BTRFS_FS_FEATURE_CHANGED,
143
7966a6b5
JB
144#if BITS_PER_LONG == 32
145 /* Indicate if we have error/warn message printed on 32bit systems */
146 BTRFS_FS_32BIT_ERROR,
147 BTRFS_FS_32BIT_WARN,
148#endif
149};
150
fc97a410
JB
151/*
152 * Flags for mount options.
153 *
154 * Note: don't forget to add new options to btrfs_show_options()
155 */
156enum {
157 BTRFS_MOUNT_NODATASUM = (1UL << 0),
158 BTRFS_MOUNT_NODATACOW = (1UL << 1),
159 BTRFS_MOUNT_NOBARRIER = (1UL << 2),
160 BTRFS_MOUNT_SSD = (1UL << 3),
161 BTRFS_MOUNT_DEGRADED = (1UL << 4),
162 BTRFS_MOUNT_COMPRESS = (1UL << 5),
163 BTRFS_MOUNT_NOTREELOG = (1UL << 6),
164 BTRFS_MOUNT_FLUSHONCOMMIT = (1UL << 7),
165 BTRFS_MOUNT_SSD_SPREAD = (1UL << 8),
166 BTRFS_MOUNT_NOSSD = (1UL << 9),
167 BTRFS_MOUNT_DISCARD_SYNC = (1UL << 10),
168 BTRFS_MOUNT_FORCE_COMPRESS = (1UL << 11),
169 BTRFS_MOUNT_SPACE_CACHE = (1UL << 12),
170 BTRFS_MOUNT_CLEAR_CACHE = (1UL << 13),
171 BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED = (1UL << 14),
172 BTRFS_MOUNT_ENOSPC_DEBUG = (1UL << 15),
173 BTRFS_MOUNT_AUTO_DEFRAG = (1UL << 16),
174 BTRFS_MOUNT_USEBACKUPROOT = (1UL << 17),
175 BTRFS_MOUNT_SKIP_BALANCE = (1UL << 18),
176 BTRFS_MOUNT_CHECK_INTEGRITY = (1UL << 19),
177 BTRFS_MOUNT_CHECK_INTEGRITY_DATA = (1UL << 20),
178 BTRFS_MOUNT_PANIC_ON_FATAL_ERROR = (1UL << 21),
179 BTRFS_MOUNT_RESCAN_UUID_TREE = (1UL << 22),
180 BTRFS_MOUNT_FRAGMENT_DATA = (1UL << 23),
181 BTRFS_MOUNT_FRAGMENT_METADATA = (1UL << 24),
182 BTRFS_MOUNT_FREE_SPACE_TREE = (1UL << 25),
183 BTRFS_MOUNT_NOLOGREPLAY = (1UL << 26),
184 BTRFS_MOUNT_REF_VERIFY = (1UL << 27),
185 BTRFS_MOUNT_DISCARD_ASYNC = (1UL << 28),
186 BTRFS_MOUNT_IGNOREBADROOTS = (1UL << 29),
187 BTRFS_MOUNT_IGNOREDATACSUMS = (1UL << 30),
188 BTRFS_MOUNT_NODISCARD = (1UL << 31),
189};
190
d83eb482
JB
191/*
192 * Compat flags that we support. If any incompat flags are set other than the
193 * ones specified below then we will fail to mount
194 */
195#define BTRFS_FEATURE_COMPAT_SUPP 0ULL
196#define BTRFS_FEATURE_COMPAT_SAFE_SET 0ULL
197#define BTRFS_FEATURE_COMPAT_SAFE_CLEAR 0ULL
198
199#define BTRFS_FEATURE_COMPAT_RO_SUPP \
200 (BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE | \
201 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID | \
202 BTRFS_FEATURE_COMPAT_RO_VERITY | \
203 BTRFS_FEATURE_COMPAT_RO_BLOCK_GROUP_TREE)
204
205#define BTRFS_FEATURE_COMPAT_RO_SAFE_SET 0ULL
206#define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR 0ULL
207
0f202b25 208#define BTRFS_FEATURE_INCOMPAT_SUPP_STABLE \
d83eb482
JB
209 (BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF | \
210 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL | \
211 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS | \
212 BTRFS_FEATURE_INCOMPAT_BIG_METADATA | \
213 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO | \
214 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD | \
215 BTRFS_FEATURE_INCOMPAT_RAID56 | \
216 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF | \
217 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA | \
218 BTRFS_FEATURE_INCOMPAT_NO_HOLES | \
219 BTRFS_FEATURE_INCOMPAT_METADATA_UUID | \
220 BTRFS_FEATURE_INCOMPAT_RAID1C34 | \
0f202b25
AJ
221 BTRFS_FEATURE_INCOMPAT_ZONED)
222
223#ifdef CONFIG_BTRFS_DEBUG
224 /*
225 * Features under developmen like Extent tree v2 support is enabled
226 * only under CONFIG_BTRFS_DEBUG.
227 */
228#define BTRFS_FEATURE_INCOMPAT_SUPP \
229 (BTRFS_FEATURE_INCOMPAT_SUPP_STABLE | \
d83eb482 230 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2)
0f202b25 231
d83eb482 232#else
0f202b25
AJ
233
234#define BTRFS_FEATURE_INCOMPAT_SUPP \
235 (BTRFS_FEATURE_INCOMPAT_SUPP_STABLE)
236
d83eb482
JB
237#endif
238
239#define BTRFS_FEATURE_INCOMPAT_SAFE_SET \
240 (BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
241#define BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR 0ULL
242
fc97a410
JB
243#define BTRFS_DEFAULT_COMMIT_INTERVAL (30)
244#define BTRFS_DEFAULT_MAX_INLINE (2048)
245
a56159d4
JB
246struct btrfs_dev_replace {
247 /* See #define above */
248 u64 replace_state;
249 /* Seconds since 1-Jan-1970 */
250 time64_t time_started;
251 /* Seconds since 1-Jan-1970 */
252 time64_t time_stopped;
253 atomic64_t num_write_errors;
254 atomic64_t num_uncorrectable_read_errors;
255
256 u64 cursor_left;
257 u64 committed_cursor_left;
258 u64 cursor_left_last_write_of_item;
259 u64 cursor_right;
260
261 /* See #define above */
262 u64 cont_reading_from_srcdev_mode;
263
264 int is_valid;
265 int item_needs_writeback;
266 struct btrfs_device *srcdev;
267 struct btrfs_device *tgtdev;
268
269 struct mutex lock_finishing_cancel_unmount;
270 struct rw_semaphore rwsem;
271
272 struct btrfs_scrub_progress scrub_progress;
273
274 struct percpu_counter bio_counter;
275 wait_queue_head_t replace_wait;
276};
277
278/*
279 * Free clusters are used to claim free space in relatively large chunks,
280 * allowing us to do less seeky writes. They are used for all metadata
281 * allocations. In ssd_spread mode they are also used for data allocations.
282 */
283struct btrfs_free_cluster {
284 spinlock_t lock;
285 spinlock_t refill_lock;
286 struct rb_root root;
287
288 /* Largest extent in this cluster */
289 u64 max_size;
290
291 /* First extent starting offset */
292 u64 window_start;
293
294 /* We did a full search and couldn't create a cluster */
295 bool fragmented;
296
297 struct btrfs_block_group *block_group;
298 /*
299 * When a cluster is allocated from a block group, we put the cluster
300 * onto a list in the block group so that it can be freed before the
301 * block group is freed.
302 */
303 struct list_head block_group_list;
304};
305
306/* Discard control. */
307/*
308 * Async discard uses multiple lists to differentiate the discard filter
309 * parameters. Index 0 is for completely free block groups where we need to
310 * ensure the entire block group is trimmed without being lossy. Indices
311 * afterwards represent monotonically decreasing discard filter sizes to
312 * prioritize what should be discarded next.
313 */
314#define BTRFS_NR_DISCARD_LISTS 3
315#define BTRFS_DISCARD_INDEX_UNUSED 0
316#define BTRFS_DISCARD_INDEX_START 1
317
318struct btrfs_discard_ctl {
319 struct workqueue_struct *discard_workers;
320 struct delayed_work work;
321 spinlock_t lock;
322 struct btrfs_block_group *block_group;
323 struct list_head discard_list[BTRFS_NR_DISCARD_LISTS];
324 u64 prev_discard;
325 u64 prev_discard_time;
326 atomic_t discardable_extents;
327 atomic64_t discardable_bytes;
328 u64 max_discard_size;
329 u64 delay_ms;
330 u32 iops_limit;
331 u32 kbps_limit;
332 u64 discard_extent_bytes;
333 u64 discard_bitmap_bytes;
334 atomic64_t discard_bytes_saved;
335};
336
337/*
338 * Exclusive operations (device replace, resize, device add/remove, balance)
339 */
340enum btrfs_exclusive_operation {
341 BTRFS_EXCLOP_NONE,
342 BTRFS_EXCLOP_BALANCE_PAUSED,
343 BTRFS_EXCLOP_BALANCE,
344 BTRFS_EXCLOP_DEV_ADD,
345 BTRFS_EXCLOP_DEV_REMOVE,
346 BTRFS_EXCLOP_DEV_REPLACE,
347 BTRFS_EXCLOP_RESIZE,
348 BTRFS_EXCLOP_SWAP_ACTIVATE,
349};
350
351/* Store data about transaction commits, exported via sysfs. */
352struct btrfs_commit_stats {
353 /* Total number of commits */
354 u64 commit_count;
355 /* The maximum commit duration so far in ns */
356 u64 max_commit_dur;
357 /* The last commit duration in ns */
358 u64 last_commit_dur;
359 /* The total commit duration in ns */
360 u64 total_commit_dur;
361};
362
363struct btrfs_fs_info {
364 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
365 unsigned long flags;
366 struct btrfs_root *tree_root;
367 struct btrfs_root *chunk_root;
368 struct btrfs_root *dev_root;
369 struct btrfs_root *fs_root;
370 struct btrfs_root *quota_root;
371 struct btrfs_root *uuid_root;
372 struct btrfs_root *data_reloc_root;
373 struct btrfs_root *block_group_root;
374
375 /* The log root tree is a directory of all the other log roots */
376 struct btrfs_root *log_root_tree;
377
378 /* The tree that holds the global roots (csum, extent, etc) */
379 rwlock_t global_root_lock;
380 struct rb_root global_root_tree;
381
382 spinlock_t fs_roots_radix_lock;
383 struct radix_tree_root fs_roots_radix;
384
385 /* Block group cache stuff */
386 rwlock_t block_group_cache_lock;
387 struct rb_root_cached block_group_cache_tree;
388
389 /* Keep track of unallocated space */
390 atomic64_t free_chunk_space;
391
392 /* Track ranges which are used by log trees blocks/logged data extents */
393 struct extent_io_tree excluded_extents;
394
395 /* logical->physical extent mapping */
396 struct extent_map_tree mapping_tree;
397
398 /*
399 * Block reservation for extent, checksum, root tree and delayed dir
400 * index item.
401 */
402 struct btrfs_block_rsv global_block_rsv;
403 /* Block reservation for metadata operations */
404 struct btrfs_block_rsv trans_block_rsv;
405 /* Block reservation for chunk tree */
406 struct btrfs_block_rsv chunk_block_rsv;
407 /* Block reservation for delayed operations */
408 struct btrfs_block_rsv delayed_block_rsv;
409 /* Block reservation for delayed refs */
410 struct btrfs_block_rsv delayed_refs_rsv;
411
412 struct btrfs_block_rsv empty_block_rsv;
413
414 u64 generation;
415 u64 last_trans_committed;
416 /*
417 * Generation of the last transaction used for block group relocation
418 * since the filesystem was last mounted (or 0 if none happened yet).
419 * Must be written and read while holding btrfs_fs_info::commit_root_sem.
420 */
421 u64 last_reloc_trans;
a56159d4
JB
422
423 /*
424 * This is updated to the current trans every time a full commit is
425 * required instead of the faster short fsync log commits
426 */
427 u64 last_trans_log_full_commit;
428 unsigned long mount_opt;
429
430 unsigned long compress_type:4;
431 unsigned int compress_level;
432 u32 commit_interval;
433 /*
434 * It is a suggestive number, the read side is safe even it gets a
435 * wrong number because we will write out the data into a regular
436 * extent. The write side(mount/remount) is under ->s_umount lock,
437 * so it is also safe.
438 */
439 u64 max_inline;
440
441 struct btrfs_transaction *running_transaction;
442 wait_queue_head_t transaction_throttle;
443 wait_queue_head_t transaction_wait;
444 wait_queue_head_t transaction_blocked_wait;
445 wait_queue_head_t async_submit_wait;
446
447 /*
448 * Used to protect the incompat_flags, compat_flags, compat_ro_flags
449 * when they are updated.
450 *
451 * Because we do not clear the flags for ever, so we needn't use
452 * the lock on the read side.
453 *
454 * We also needn't use the lock when we mount the fs, because
455 * there is no other task which will update the flag.
456 */
457 spinlock_t super_lock;
458 struct btrfs_super_block *super_copy;
459 struct btrfs_super_block *super_for_commit;
460 struct super_block *sb;
461 struct inode *btree_inode;
462 struct mutex tree_log_mutex;
463 struct mutex transaction_kthread_mutex;
464 struct mutex cleaner_mutex;
465 struct mutex chunk_mutex;
466
467 /*
468 * This is taken to make sure we don't set block groups ro after the
469 * free space cache has been allocated on them.
470 */
471 struct mutex ro_block_group_mutex;
472
473 /*
474 * This is used during read/modify/write to make sure no two ios are
475 * trying to mod the same stripe at the same time.
476 */
477 struct btrfs_stripe_hash_table *stripe_hash_table;
478
479 /*
480 * This protects the ordered operations list only while we are
481 * processing all of the entries on it. This way we make sure the
482 * commit code doesn't find the list temporarily empty because another
483 * function happens to be doing non-waiting preflush before jumping
484 * into the main commit.
485 */
486 struct mutex ordered_operations_mutex;
487
488 struct rw_semaphore commit_root_sem;
489
490 struct rw_semaphore cleanup_work_sem;
491
492 struct rw_semaphore subvol_sem;
493
494 spinlock_t trans_lock;
495 /*
496 * The reloc mutex goes with the trans lock, it is taken during commit
497 * to protect us from the relocation code.
498 */
499 struct mutex reloc_mutex;
500
501 struct list_head trans_list;
502 struct list_head dead_roots;
503 struct list_head caching_block_groups;
504
505 spinlock_t delayed_iput_lock;
506 struct list_head delayed_iputs;
507 atomic_t nr_delayed_iputs;
508 wait_queue_head_t delayed_iputs_wait;
509
510 atomic64_t tree_mod_seq;
511
512 /* This protects tree_mod_log and tree_mod_seq_list */
513 rwlock_t tree_mod_log_lock;
514 struct rb_root tree_mod_log;
515 struct list_head tree_mod_seq_list;
516
517 atomic_t async_delalloc_pages;
518
519 /* This is used to protect the following list -- ordered_roots. */
520 spinlock_t ordered_root_lock;
521
522 /*
523 * All fs/file tree roots in which there are data=ordered extents
524 * pending writeback are added into this list.
525 *
526 * These can span multiple transactions and basically include every
527 * dirty data page that isn't from nodatacow.
528 */
529 struct list_head ordered_roots;
530
531 struct mutex delalloc_root_mutex;
532 spinlock_t delalloc_root_lock;
533 /* All fs/file tree roots that have delalloc inodes. */
534 struct list_head delalloc_roots;
535
536 /*
537 * There is a pool of worker threads for checksumming during writes and
538 * a pool for checksumming after reads. This is because readers can
539 * run with FS locks held, and the writers may be waiting for those
540 * locks. We don't want ordering in the pending list to cause
541 * deadlocks, and so the two are serviced separately.
542 *
543 * A third pool does submit_bio to avoid deadlocking with the other two.
544 */
545 struct btrfs_workqueue *workers;
546 struct btrfs_workqueue *hipri_workers;
547 struct btrfs_workqueue *delalloc_workers;
548 struct btrfs_workqueue *flush_workers;
549 struct workqueue_struct *endio_workers;
550 struct workqueue_struct *endio_meta_workers;
a56159d4
JB
551 struct workqueue_struct *rmw_workers;
552 struct workqueue_struct *compressed_write_workers;
553 struct btrfs_workqueue *endio_write_workers;
554 struct btrfs_workqueue *endio_freespace_worker;
555 struct btrfs_workqueue *caching_workers;
556
557 /*
558 * Fixup workers take dirty pages that didn't properly go through the
559 * cow mechanism and make them safe to write. It happens for the
560 * sys_munmap function call path.
561 */
562 struct btrfs_workqueue *fixup_workers;
563 struct btrfs_workqueue *delayed_workers;
564
565 struct task_struct *transaction_kthread;
566 struct task_struct *cleaner_kthread;
567 u32 thread_pool_size;
568
569 struct kobject *space_info_kobj;
570 struct kobject *qgroups_kobj;
571 struct kobject *discard_kobj;
572
573 /* Used to keep from writing metadata until there is a nice batch */
574 struct percpu_counter dirty_metadata_bytes;
575 struct percpu_counter delalloc_bytes;
576 struct percpu_counter ordered_bytes;
577 s32 dirty_metadata_batch;
578 s32 delalloc_batch;
579
580 struct list_head dirty_cowonly_roots;
581
582 struct btrfs_fs_devices *fs_devices;
583
584 /*
585 * The space_info list is effectively read only after initial setup.
586 * It is populated at mount time and cleaned up after all block groups
587 * are removed. RCU is used to protect it.
588 */
589 struct list_head space_info;
590
591 struct btrfs_space_info *data_sinfo;
592
593 struct reloc_control *reloc_ctl;
594
595 /* data_alloc_cluster is only used in ssd_spread mode */
596 struct btrfs_free_cluster data_alloc_cluster;
597
598 /* All metadata allocations go through this cluster. */
599 struct btrfs_free_cluster meta_alloc_cluster;
600
601 /* Auto defrag inodes go here. */
602 spinlock_t defrag_inodes_lock;
603 struct rb_root defrag_inodes;
604 atomic_t defrag_running;
605
606 /* Used to protect avail_{data, metadata, system}_alloc_bits */
607 seqlock_t profiles_lock;
608 /*
609 * These three are in extended format (availability of single chunks is
610 * denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other types are denoted
611 * by corresponding BTRFS_BLOCK_GROUP_* bits)
612 */
613 u64 avail_data_alloc_bits;
614 u64 avail_metadata_alloc_bits;
615 u64 avail_system_alloc_bits;
616
617 /* Balance state */
618 spinlock_t balance_lock;
619 struct mutex balance_mutex;
620 atomic_t balance_pause_req;
621 atomic_t balance_cancel_req;
622 struct btrfs_balance_control *balance_ctl;
623 wait_queue_head_t balance_wait_q;
624
625 /* Cancellation requests for chunk relocation */
626 atomic_t reloc_cancel_req;
627
628 u32 data_chunk_allocations;
629 u32 metadata_ratio;
630
631 void *bdev_holder;
632
633 /* Private scrub information */
634 struct mutex scrub_lock;
635 atomic_t scrubs_running;
636 atomic_t scrub_pause_req;
637 atomic_t scrubs_paused;
638 atomic_t scrub_cancel_req;
639 wait_queue_head_t scrub_pause_wait;
640 /*
641 * The worker pointers are NULL iff the refcount is 0, ie. scrub is not
642 * running.
643 */
644 refcount_t scrub_workers_refcnt;
645 struct workqueue_struct *scrub_workers;
646 struct workqueue_struct *scrub_wr_completion_workers;
a56159d4
JB
647 struct btrfs_subpage_info *subpage_info;
648
649 struct btrfs_discard_ctl discard_ctl;
650
651#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
652 u32 check_integrity_print_mask;
653#endif
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 /* Filesystem state */
691 unsigned long fs_state;
692
693 struct btrfs_delayed_root *delayed_root;
694
695 /* Extent buffer radix tree */
696 spinlock_t buffer_lock;
697 /* Entries are eb->start / sectorsize */
698 struct radix_tree_root buffer_radix;
699
700 /* Next backup root to be overwritten */
701 int backup_root_index;
702
703 /* Device replace state */
704 struct btrfs_dev_replace dev_replace;
705
706 struct semaphore uuid_tree_rescan_sem;
707
708 /* Used to reclaim the metadata space in the background. */
709 struct work_struct async_reclaim_work;
710 struct work_struct async_data_reclaim_work;
711 struct work_struct preempt_reclaim_work;
712
713 /* Reclaim partially filled block groups in the background */
714 struct work_struct reclaim_bgs_work;
715 struct list_head reclaim_bgs;
716 int bg_reclaim_threshold;
717
718 spinlock_t unused_bgs_lock;
719 struct list_head unused_bgs;
720 struct mutex unused_bg_unpin_mutex;
721 /* Protect block groups that are going to be deleted */
722 struct mutex reclaim_bgs_lock;
723
724 /* Cached block sizes */
725 u32 nodesize;
726 u32 sectorsize;
727 /* ilog2 of sectorsize, use to avoid 64bit division */
728 u32 sectorsize_bits;
729 u32 csum_size;
730 u32 csums_per_leaf;
731 u32 stripesize;
732
733 /*
734 * Maximum size of an extent. BTRFS_MAX_EXTENT_SIZE on regular
735 * filesystem, on zoned it depends on the device constraints.
736 */
737 u64 max_extent_size;
738
739 /* Block groups and devices containing active swapfiles. */
740 spinlock_t swapfile_pins_lock;
741 struct rb_root swapfile_pins;
742
743 struct crypto_shash *csum_shash;
744
745 /* Type of exclusive operation running, protected by super_lock */
746 enum btrfs_exclusive_operation exclusive_operation;
747
748 /*
749 * Zone size > 0 when in ZONED mode, otherwise it's used for a check
750 * if the mode is enabled
751 */
752 u64 zone_size;
753
243cf8d1
CH
754 /* Constraints for ZONE_APPEND commands: */
755 struct queue_limits limits;
a56159d4 756 u64 max_zone_append_size;
243cf8d1 757
a56159d4
JB
758 struct mutex zoned_meta_io_lock;
759 spinlock_t treelog_bg_lock;
760 u64 treelog_bg;
761
762 /*
763 * Start of the dedicated data relocation block group, protected by
764 * relocation_bg_lock.
765 */
766 spinlock_t relocation_bg_lock;
767 u64 data_reloc_bg;
768 struct mutex zoned_data_reloc_io_lock;
769
770 u64 nr_global_roots;
771
772 spinlock_t zone_active_bgs_lock;
773 struct list_head zone_active_bgs;
774
775 /* Updates are not protected by any lock */
776 struct btrfs_commit_stats commit_stats;
777
778 /*
779 * Last generation where we dropped a non-relocation root.
780 * Use btrfs_set_last_root_drop_gen() and btrfs_get_last_root_drop_gen()
781 * to change it and to read it, respectively.
782 */
783 u64 last_root_drop_gen;
784
785 /*
786 * Annotations for transaction events (structures are empty when
787 * compiled without lockdep).
788 */
789 struct lockdep_map btrfs_trans_num_writers_map;
790 struct lockdep_map btrfs_trans_num_extwriters_map;
791 struct lockdep_map btrfs_state_change_map[4];
792 struct lockdep_map btrfs_trans_pending_ordered_map;
793 struct lockdep_map btrfs_ordered_extent_map;
794
795#ifdef CONFIG_BTRFS_FS_REF_VERIFY
796 spinlock_t ref_verify_lock;
797 struct rb_root block_tree;
798#endif
799
800#ifdef CONFIG_BTRFS_DEBUG
801 struct kobject *debug_kobj;
802 struct list_head allocated_roots;
803
804 spinlock_t eb_leak_lock;
805 struct list_head allocated_ebs;
806#endif
807};
808
809static inline void btrfs_set_last_root_drop_gen(struct btrfs_fs_info *fs_info,
810 u64 gen)
811{
812 WRITE_ONCE(fs_info->last_root_drop_gen, gen);
813}
814
815static inline u64 btrfs_get_last_root_drop_gen(const struct btrfs_fs_info *fs_info)
816{
817 return READ_ONCE(fs_info->last_root_drop_gen);
818}
819
a56159d4
JB
820/*
821 * Take the number of bytes to be checksummed and figure out how many leaves
822 * it would require to store the csums for that many bytes.
823 */
824static inline u64 btrfs_csum_bytes_to_leaves(
825 const struct btrfs_fs_info *fs_info, u64 csum_bytes)
826{
827 const u64 num_csums = csum_bytes >> fs_info->sectorsize_bits;
828
829 return DIV_ROUND_UP_ULL(num_csums, fs_info->csums_per_leaf);
830}
831
832/*
833 * Use this if we would be adding new items, as we could split nodes as we cow
834 * down the tree.
835 */
d1085c9c 836static inline u64 btrfs_calc_insert_metadata_size(const struct btrfs_fs_info *fs_info,
a56159d4
JB
837 unsigned num_items)
838{
839 return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * 2 * num_items;
840}
841
842/*
843 * Doing a truncate or a modification won't result in new nodes or leaves, just
844 * what we need for COW.
845 */
d1085c9c 846static inline u64 btrfs_calc_metadata_size(const struct btrfs_fs_info *fs_info,
a56159d4
JB
847 unsigned num_items)
848{
849 return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * num_items;
850}
851
852#define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r->fs_info) >> 4) - \
853 sizeof(struct btrfs_item))
854
855static inline bool btrfs_is_zoned(const struct btrfs_fs_info *fs_info)
856{
857 return fs_info->zone_size > 0;
858}
859
860/*
861 * Count how many fs_info->max_extent_size cover the @size
862 */
863static inline u32 count_max_extents(struct btrfs_fs_info *fs_info, u64 size)
864{
865#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
866 if (!fs_info)
867 return div_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, BTRFS_MAX_EXTENT_SIZE);
868#endif
869
870 return div_u64(size + fs_info->max_extent_size - 1, fs_info->max_extent_size);
871}
872
873bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
874 enum btrfs_exclusive_operation type);
875bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
876 enum btrfs_exclusive_operation type);
877void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info);
878void btrfs_exclop_finish(struct btrfs_fs_info *fs_info);
879void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
880 enum btrfs_exclusive_operation op);
881
c7f13d42
JB
882/* Compatibility and incompatibility defines */
883void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
884 const char *name);
885void __btrfs_clear_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
886 const char *name);
887void __btrfs_set_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
888 const char *name);
889void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
890 const char *name);
891
0d3a9cf8
JB
892#define __btrfs_fs_incompat(fs_info, flags) \
893 (!!(btrfs_super_incompat_flags((fs_info)->super_copy) & (flags)))
894
895#define __btrfs_fs_compat_ro(fs_info, flags) \
896 (!!(btrfs_super_compat_ro_flags((fs_info)->super_copy) & (flags)))
897
c7f13d42
JB
898#define btrfs_set_fs_incompat(__fs_info, opt) \
899 __btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
900
901#define btrfs_clear_fs_incompat(__fs_info, opt) \
902 __btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
903
904#define btrfs_fs_incompat(fs_info, opt) \
905 __btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
906
907#define btrfs_set_fs_compat_ro(__fs_info, opt) \
908 __btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
909
910#define btrfs_clear_fs_compat_ro(__fs_info, opt) \
911 __btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
912
913#define btrfs_fs_compat_ro(fs_info, opt) \
914 __btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
915
fc97a410
JB
916#define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt)
917#define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt)
918#define btrfs_raw_test_opt(o, opt) ((o) & BTRFS_MOUNT_##opt)
919#define btrfs_test_opt(fs_info, opt) ((fs_info)->mount_opt & \
920 BTRFS_MOUNT_##opt)
921
922#define btrfs_set_and_info(fs_info, opt, fmt, args...) \
923do { \
924 if (!btrfs_test_opt(fs_info, opt)) \
925 btrfs_info(fs_info, fmt, ##args); \
926 btrfs_set_opt(fs_info->mount_opt, opt); \
927} while (0)
928
929#define btrfs_clear_and_info(fs_info, opt, fmt, args...) \
930do { \
931 if (btrfs_test_opt(fs_info, opt)) \
932 btrfs_info(fs_info, fmt, ##args); \
933 btrfs_clear_opt(fs_info->mount_opt, opt); \
934} while (0)
935
c7f13d42
JB
936static inline int btrfs_fs_closing(struct btrfs_fs_info *fs_info)
937{
938 /* Do it this way so we only ever do one test_bit in the normal case. */
939 if (test_bit(BTRFS_FS_CLOSING_START, &fs_info->flags)) {
940 if (test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags))
941 return 2;
942 return 1;
943 }
944 return 0;
945}
946
947/*
948 * If we remount the fs to be R/O or umount the fs, the cleaner needn't do
949 * anything except sleeping. This function is used to check the status of
950 * the fs.
951 * We check for BTRFS_FS_STATE_RO to avoid races with a concurrent remount,
952 * since setting and checking for SB_RDONLY in the superblock's flags is not
953 * atomic.
954 */
955static inline int btrfs_need_cleaner_sleep(struct btrfs_fs_info *fs_info)
956{
957 return test_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state) ||
958 btrfs_fs_closing(fs_info);
959}
960
7966a6b5
JB
961static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info *fs_info)
962{
963 clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags);
964}
965
ec8eb376
JB
966#define BTRFS_FS_ERROR(fs_info) (unlikely(test_bit(BTRFS_FS_STATE_ERROR, \
967 &(fs_info)->fs_state)))
968#define BTRFS_FS_LOG_CLEANUP_ERROR(fs_info) \
969 (unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR, \
970 &(fs_info)->fs_state)))
971
972#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
6a6b4daf
JB
973
974#define EXPORT_FOR_TESTS
975
ec8eb376
JB
976static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
977{
978 return test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
979}
6a6b4daf
JB
980
981void btrfs_test_destroy_inode(struct inode *inode);
982
ec8eb376 983#else
6a6b4daf
JB
984
985#define EXPORT_FOR_TESTS static
986
ec8eb376
JB
987static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
988{
989 return 0;
990}
991#endif
992
c7f13d42 993#endif