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