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_DEFAULT_MAX_INLINE (2048)
304
305 struct btrfs_dev_replace {
306 /* See #define above */
307 u64 replace_state;
308 /* Seconds since 1-Jan-1970 */
309 time64_t time_started;
310 /* Seconds since 1-Jan-1970 */
311 time64_t time_stopped;
312 atomic64_t num_write_errors;
313 atomic64_t num_uncorrectable_read_errors;
314
315 u64 cursor_left;
316 u64 committed_cursor_left;
317 u64 cursor_left_last_write_of_item;
318 u64 cursor_right;
319
320 /* See #define above */
321 u64 cont_reading_from_srcdev_mode;
322
323 int is_valid;
324 int item_needs_writeback;
325 struct btrfs_device *srcdev;
326 struct btrfs_device *tgtdev;
327
328 struct mutex lock_finishing_cancel_unmount;
329 struct rw_semaphore rwsem;
330
331 struct btrfs_scrub_progress scrub_progress;
332
333 struct percpu_counter bio_counter;
334 wait_queue_head_t replace_wait;
335
336 struct task_struct *replace_task;
337 };
338
339 /*
340 * Free clusters are used to claim free space in relatively large chunks,
341 * allowing us to do less seeky writes. They are used for all metadata
342 * allocations. In ssd_spread mode they are also used for data allocations.
343 */
344 struct btrfs_free_cluster {
345 spinlock_t lock;
346 spinlock_t refill_lock;
347 struct rb_root root;
348
349 /* Largest extent in this cluster */
350 u64 max_size;
351
352 /* First extent starting offset */
353 u64 window_start;
354
355 /* We did a full search and couldn't create a cluster */
356 bool fragmented;
357
358 struct btrfs_block_group *block_group;
359 /*
360 * When a cluster is allocated from a block group, we put the cluster
361 * onto a list in the block group so that it can be freed before the
362 * block group is freed.
363 */
364 struct list_head block_group_list;
365 };
366
367 /* Discard control. */
368 /*
369 * Async discard uses multiple lists to differentiate the discard filter
370 * parameters. Index 0 is for completely free block groups where we need to
371 * ensure the entire block group is trimmed without being lossy. Indices
372 * afterwards represent monotonically decreasing discard filter sizes to
373 * prioritize what should be discarded next.
374 */
375 #define BTRFS_NR_DISCARD_LISTS 3
376 #define BTRFS_DISCARD_INDEX_UNUSED 0
377 #define BTRFS_DISCARD_INDEX_START 1
378
379 struct btrfs_discard_ctl {
380 struct workqueue_struct *discard_workers;
381 struct delayed_work work;
382 spinlock_t lock;
383 struct btrfs_block_group *block_group;
384 struct list_head discard_list[BTRFS_NR_DISCARD_LISTS];
385 u64 prev_discard;
386 u64 prev_discard_time;
387 atomic_t discardable_extents;
388 atomic64_t discardable_bytes;
389 u64 max_discard_size;
390 u64 delay_ms;
391 u32 iops_limit;
392 u32 kbps_limit;
393 u64 discard_extent_bytes;
394 u64 discard_bitmap_bytes;
395 atomic64_t discard_bytes_saved;
396 };
397
398 /*
399 * Exclusive operations (device replace, resize, device add/remove, balance)
400 */
401 enum btrfs_exclusive_operation {
402 BTRFS_EXCLOP_NONE,
403 BTRFS_EXCLOP_BALANCE_PAUSED,
404 BTRFS_EXCLOP_BALANCE,
405 BTRFS_EXCLOP_DEV_ADD,
406 BTRFS_EXCLOP_DEV_REMOVE,
407 BTRFS_EXCLOP_DEV_REPLACE,
408 BTRFS_EXCLOP_RESIZE,
409 BTRFS_EXCLOP_SWAP_ACTIVATE,
410 };
411
412 /* Store data about transaction commits, exported via sysfs. */
413 struct btrfs_commit_stats {
414 /* Total number of commits */
415 u64 commit_count;
416 /* The maximum commit duration so far in ns */
417 u64 max_commit_dur;
418 /* The last commit duration in ns */
419 u64 last_commit_dur;
420 /* The total commit duration in ns */
421 u64 total_commit_dur;
422 };
423
424 struct btrfs_fs_info {
425 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
426 unsigned long flags;
427 struct btrfs_root *tree_root;
428 struct btrfs_root *chunk_root;
429 struct btrfs_root *dev_root;
430 struct btrfs_root *fs_root;
431 struct btrfs_root *quota_root;
432 struct btrfs_root *uuid_root;
433 struct btrfs_root *data_reloc_root;
434 struct btrfs_root *block_group_root;
435 struct btrfs_root *stripe_root;
436
437 /* The log root tree is a directory of all the other log roots */
438 struct btrfs_root *log_root_tree;
439
440 /* The tree that holds the global roots (csum, extent, etc) */
441 rwlock_t global_root_lock;
442 struct rb_root global_root_tree;
443
444 spinlock_t fs_roots_radix_lock;
445 struct radix_tree_root fs_roots_radix;
446
447 /* Block group cache stuff */
448 rwlock_t block_group_cache_lock;
449 struct rb_root_cached block_group_cache_tree;
450
451 /* Keep track of unallocated space */
452 atomic64_t free_chunk_space;
453
454 /* Track ranges which are used by log trees blocks/logged data extents */
455 struct extent_io_tree excluded_extents;
456
457 /* logical->physical extent mapping */
458 struct rb_root_cached mapping_tree;
459 rwlock_t mapping_tree_lock;
460
461 /*
462 * Block reservation for extent, checksum, root tree and delayed dir
463 * index item.
464 */
465 struct btrfs_block_rsv global_block_rsv;
466 /* Block reservation for metadata operations */
467 struct btrfs_block_rsv trans_block_rsv;
468 /* Block reservation for chunk tree */
469 struct btrfs_block_rsv chunk_block_rsv;
470 /* Block reservation for delayed operations */
471 struct btrfs_block_rsv delayed_block_rsv;
472 /* Block reservation for delayed refs */
473 struct btrfs_block_rsv delayed_refs_rsv;
474
475 struct btrfs_block_rsv empty_block_rsv;
476
477 /*
478 * Updated while holding the lock 'trans_lock'. Due to the life cycle of
479 * a transaction, it can be directly read while holding a transaction
480 * handle, everywhere else must be read with btrfs_get_fs_generation().
481 * Should always be updated using btrfs_set_fs_generation().
482 */
483 u64 generation;
484 /*
485 * Always use btrfs_get_last_trans_committed() and
486 * btrfs_set_last_trans_committed() to read and update this field.
487 */
488 u64 last_trans_committed;
489 /*
490 * Generation of the last transaction used for block group relocation
491 * since the filesystem was last mounted (or 0 if none happened yet).
492 * Must be written and read while holding btrfs_fs_info::commit_root_sem.
493 */
494 u64 last_reloc_trans;
495
496 /*
497 * This is updated to the current trans every time a full commit is
498 * required instead of the faster short fsync log commits
499 */
500 u64 last_trans_log_full_commit;
501 unsigned long long mount_opt;
502
503 int compress_type;
504 int compress_level;
505 u32 commit_interval;
506 /*
507 * It is a suggestive number, the read side is safe even it gets a
508 * wrong number because we will write out the data into a regular
509 * extent. The write side(mount/remount) is under ->s_umount lock,
510 * so it is also safe.
511 */
512 u64 max_inline;
513
514 struct btrfs_transaction *running_transaction;
515 wait_queue_head_t transaction_throttle;
516 wait_queue_head_t transaction_wait;
517 wait_queue_head_t transaction_blocked_wait;
518 wait_queue_head_t async_submit_wait;
519
520 /*
521 * Used to protect the incompat_flags, compat_flags, compat_ro_flags
522 * when they are updated.
523 *
524 * Because we do not clear the flags for ever, so we needn't use
525 * the lock on the read side.
526 *
527 * We also needn't use the lock when we mount the fs, because
528 * there is no other task which will update the flag.
529 */
530 spinlock_t super_lock;
531 struct btrfs_super_block *super_copy;
532 struct btrfs_super_block *super_for_commit;
533 struct super_block *sb;
534 struct inode *btree_inode;
535 struct mutex tree_log_mutex;
536 struct mutex transaction_kthread_mutex;
537 struct mutex cleaner_mutex;
538 struct mutex chunk_mutex;
539
540 /*
541 * This is taken to make sure we don't set block groups ro after the
542 * free space cache has been allocated on them.
543 */
544 struct mutex ro_block_group_mutex;
545
546 /*
547 * This is used during read/modify/write to make sure no two ios are
548 * trying to mod the same stripe at the same time.
549 */
550 struct btrfs_stripe_hash_table *stripe_hash_table;
551
552 /*
553 * This protects the ordered operations list only while we are
554 * processing all of the entries on it. This way we make sure the
555 * commit code doesn't find the list temporarily empty because another
556 * function happens to be doing non-waiting preflush before jumping
557 * into the main commit.
558 */
559 struct mutex ordered_operations_mutex;
560
561 struct rw_semaphore commit_root_sem;
562
563 struct rw_semaphore cleanup_work_sem;
564
565 struct rw_semaphore subvol_sem;
566
567 spinlock_t trans_lock;
568 /*
569 * The reloc mutex goes with the trans lock, it is taken during commit
570 * to protect us from the relocation code.
571 */
572 struct mutex reloc_mutex;
573
574 struct list_head trans_list;
575 struct list_head dead_roots;
576 struct list_head caching_block_groups;
577
578 spinlock_t delayed_iput_lock;
579 struct list_head delayed_iputs;
580 atomic_t nr_delayed_iputs;
581 wait_queue_head_t delayed_iputs_wait;
582
583 atomic64_t tree_mod_seq;
584
585 /* This protects tree_mod_log and tree_mod_seq_list */
586 rwlock_t tree_mod_log_lock;
587 struct rb_root tree_mod_log;
588 struct list_head tree_mod_seq_list;
589
590 atomic_t async_delalloc_pages;
591
592 /* This is used to protect the following list -- ordered_roots. */
593 spinlock_t ordered_root_lock;
594
595 /*
596 * All fs/file tree roots in which there are data=ordered extents
597 * pending writeback are added into this list.
598 *
599 * These can span multiple transactions and basically include every
600 * dirty data page that isn't from nodatacow.
601 */
602 struct list_head ordered_roots;
603
604 struct mutex delalloc_root_mutex;
605 spinlock_t delalloc_root_lock;
606 /* All fs/file tree roots that have delalloc inodes. */
607 struct list_head delalloc_roots;
608
609 /*
610 * There is a pool of worker threads for checksumming during writes and
611 * a pool for checksumming after reads. This is because readers can
612 * run with FS locks held, and the writers may be waiting for those
613 * locks. We don't want ordering in the pending list to cause
614 * deadlocks, and so the two are serviced separately.
615 *
616 * A third pool does submit_bio to avoid deadlocking with the other two.
617 */
618 struct btrfs_workqueue *workers;
619 struct btrfs_workqueue *delalloc_workers;
620 struct btrfs_workqueue *flush_workers;
621 struct workqueue_struct *endio_workers;
622 struct workqueue_struct *endio_meta_workers;
623 struct workqueue_struct *rmw_workers;
624 struct workqueue_struct *compressed_write_workers;
625 struct btrfs_workqueue *endio_write_workers;
626 struct btrfs_workqueue *endio_freespace_worker;
627 struct btrfs_workqueue *caching_workers;
628
629 /*
630 * Fixup workers take dirty pages that didn't properly go through the
631 * cow mechanism and make them safe to write. It happens for the
632 * sys_munmap function call path.
633 */
634 struct btrfs_workqueue *fixup_workers;
635 struct btrfs_workqueue *delayed_workers;
636
637 struct task_struct *transaction_kthread;
638 struct task_struct *cleaner_kthread;
639 u32 thread_pool_size;
640
641 struct kobject *space_info_kobj;
642 struct kobject *qgroups_kobj;
643 struct kobject *discard_kobj;
644
645 /* Track the number of blocks (sectors) read by the filesystem. */
646 struct percpu_counter stats_read_blocks;
647
648 /* Used to keep from writing metadata until there is a nice batch */
649 struct percpu_counter dirty_metadata_bytes;
650 struct percpu_counter delalloc_bytes;
651 struct percpu_counter ordered_bytes;
652 s32 dirty_metadata_batch;
653 s32 delalloc_batch;
654
655 struct percpu_counter evictable_extent_maps;
656 u64 em_shrinker_last_root;
657 u64 em_shrinker_last_ino;
658 atomic64_t em_shrinker_nr_to_scan;
659 struct work_struct em_shrinker_work;
660
661 /* Protected by 'trans_lock'. */
662 struct list_head dirty_cowonly_roots;
663
664 struct btrfs_fs_devices *fs_devices;
665
666 /*
667 * The space_info list is effectively read only after initial setup.
668 * It is populated at mount time and cleaned up after all block groups
669 * are removed. RCU is used to protect it.
670 */
671 struct list_head space_info;
672
673 struct btrfs_space_info *data_sinfo;
674
675 struct reloc_control *reloc_ctl;
676
677 /* data_alloc_cluster is only used in ssd_spread mode */
678 struct btrfs_free_cluster data_alloc_cluster;
679
680 /* All metadata allocations go through this cluster. */
681 struct btrfs_free_cluster meta_alloc_cluster;
682
683 /* Auto defrag inodes go here. */
684 spinlock_t defrag_inodes_lock;
685 struct rb_root defrag_inodes;
686 atomic_t defrag_running;
687
688 /* Used to protect avail_{data, metadata, system}_alloc_bits */
689 seqlock_t profiles_lock;
690 /*
691 * These three are in extended format (availability of single chunks is
692 * denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other types are denoted
693 * by corresponding BTRFS_BLOCK_GROUP_* bits)
694 */
695 u64 avail_data_alloc_bits;
696 u64 avail_metadata_alloc_bits;
697 u64 avail_system_alloc_bits;
698
699 /* Balance state */
700 spinlock_t balance_lock;
701 struct mutex balance_mutex;
702 atomic_t balance_pause_req;
703 atomic_t balance_cancel_req;
704 struct btrfs_balance_control *balance_ctl;
705 wait_queue_head_t balance_wait_q;
706
707 /* Cancellation requests for chunk relocation */
708 atomic_t reloc_cancel_req;
709
710 u32 data_chunk_allocations;
711 u32 metadata_ratio;
712
713 void *bdev_holder;
714
715 /* Private scrub information */
716 struct mutex scrub_lock;
717 atomic_t scrubs_running;
718 atomic_t scrub_pause_req;
719 atomic_t scrubs_paused;
720 atomic_t scrub_cancel_req;
721 wait_queue_head_t scrub_pause_wait;
722 /*
723 * The worker pointers are NULL iff the refcount is 0, ie. scrub is not
724 * running.
725 */
726 refcount_t scrub_workers_refcnt;
727 struct workqueue_struct *scrub_workers;
728
729 struct btrfs_discard_ctl discard_ctl;
730
731 /* Is qgroup tracking in a consistent state? */
732 u64 qgroup_flags;
733
734 /* Holds configuration and tracking. Protected by qgroup_lock. */
735 struct rb_root qgroup_tree;
736 spinlock_t qgroup_lock;
737
738 /*
739 * Used to avoid frequently calling ulist_alloc()/ulist_free()
740 * when doing qgroup accounting, it must be protected by qgroup_lock.
741 */
742 struct ulist *qgroup_ulist;
743
744 /*
745 * Protect user change for quota operations. If a transaction is needed,
746 * it must be started before locking this lock.
747 */
748 struct mutex qgroup_ioctl_lock;
749
750 /* List of dirty qgroups to be written at next commit. */
751 struct list_head dirty_qgroups;
752
753 /* Used by qgroup for an efficient tree traversal. */
754 u64 qgroup_seq;
755
756 /* Qgroup rescan items. */
757 /* Protects the progress item */
758 struct mutex qgroup_rescan_lock;
759 struct btrfs_key qgroup_rescan_progress;
760 struct btrfs_workqueue *qgroup_rescan_workers;
761 struct completion qgroup_rescan_completion;
762 struct btrfs_work qgroup_rescan_work;
763 /* Protected by qgroup_rescan_lock */
764 bool qgroup_rescan_running;
765 u8 qgroup_drop_subtree_thres;
766 u64 qgroup_enable_gen;
767
768 /*
769 * If this is not 0, then it indicates a serious filesystem error has
770 * happened and it contains that error (negative errno value).
771 */
772 int fs_error;
773
774 /* Filesystem state */
775 unsigned long fs_state;
776
777 struct btrfs_delayed_root *delayed_root;
778
779 /* Extent buffer radix tree */
780 spinlock_t buffer_lock;
781 /* Entries are eb->start / sectorsize */
782 struct radix_tree_root buffer_radix;
783
784 /* Next backup root to be overwritten */
785 int backup_root_index;
786
787 /* Device replace state */
788 struct btrfs_dev_replace dev_replace;
789
790 struct semaphore uuid_tree_rescan_sem;
791
792 /* Used to reclaim the metadata space in the background. */
793 struct work_struct async_reclaim_work;
794 struct work_struct async_data_reclaim_work;
795 struct work_struct preempt_reclaim_work;
796
797 /* Reclaim partially filled block groups in the background */
798 struct work_struct reclaim_bgs_work;
799 /* Protected by unused_bgs_lock. */
800 struct list_head reclaim_bgs;
801 int bg_reclaim_threshold;
802
803 /* Protects the lists unused_bgs and reclaim_bgs. */
804 spinlock_t unused_bgs_lock;
805 /* Protected by unused_bgs_lock. */
806 struct list_head unused_bgs;
807 struct mutex unused_bg_unpin_mutex;
808 /* Protect block groups that are going to be deleted */
809 struct mutex reclaim_bgs_lock;
810
811 /* Cached block sizes */
812 u32 nodesize;
813 u32 sectorsize;
814 /* ilog2 of sectorsize, use to avoid 64bit division */
815 u32 sectorsize_bits;
816 u32 csum_size;
817 u32 csums_per_leaf;
818 u32 stripesize;
819
820 /*
821 * Maximum size of an extent. BTRFS_MAX_EXTENT_SIZE on regular
822 * filesystem, on zoned it depends on the device constraints.
823 */
824 u64 max_extent_size;
825
826 /* Block groups and devices containing active swapfiles. */
827 spinlock_t swapfile_pins_lock;
828 struct rb_root swapfile_pins;
829
830 struct crypto_shash *csum_shash;
831
832 /* Type of exclusive operation running, protected by super_lock */
833 enum btrfs_exclusive_operation exclusive_operation;
834
835 /*
836 * Zone size > 0 when in ZONED mode, otherwise it's used for a check
837 * if the mode is enabled
838 */
839 u64 zone_size;
840
841 /* Constraints for ZONE_APPEND commands: */
842 struct queue_limits limits;
843 u64 max_zone_append_size;
844
845 struct mutex zoned_meta_io_lock;
846 spinlock_t treelog_bg_lock;
847 u64 treelog_bg;
848
849 /*
850 * Start of the dedicated data relocation block group, protected by
851 * relocation_bg_lock.
852 */
853 spinlock_t relocation_bg_lock;
854 u64 data_reloc_bg;
855 struct mutex zoned_data_reloc_io_lock;
856
857 struct btrfs_block_group *active_meta_bg;
858 struct btrfs_block_group *active_system_bg;
859
860 u64 nr_global_roots;
861
862 spinlock_t zone_active_bgs_lock;
863 struct list_head zone_active_bgs;
864
865 /* Updates are not protected by any lock */
866 struct btrfs_commit_stats commit_stats;
867
868 /*
869 * Last generation where we dropped a non-relocation root.
870 * Use btrfs_set_last_root_drop_gen() and btrfs_get_last_root_drop_gen()
871 * to change it and to read it, respectively.
872 */
873 u64 last_root_drop_gen;
874
875 /*
876 * Annotations for transaction events (structures are empty when
877 * compiled without lockdep).
878 */
879 struct lockdep_map btrfs_trans_num_writers_map;
880 struct lockdep_map btrfs_trans_num_extwriters_map;
881 struct lockdep_map btrfs_state_change_map[4];
882 struct lockdep_map btrfs_trans_pending_ordered_map;
883 struct lockdep_map btrfs_ordered_extent_map;
884
885 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
886 spinlock_t ref_verify_lock;
887 struct rb_root block_tree;
888 #endif
889
890 #ifdef CONFIG_BTRFS_DEBUG
891 struct kobject *debug_kobj;
892 struct list_head allocated_roots;
893
894 spinlock_t eb_leak_lock;
895 struct list_head allocated_ebs;
896 #endif
897 };
898
899 #define folio_to_inode(_folio) (BTRFS_I(_Generic((_folio), \
900 struct folio *: (_folio))->mapping->host))
901
902 #define folio_to_fs_info(_folio) (folio_to_inode(_folio)->root->fs_info)
903
904 #define inode_to_fs_info(_inode) (BTRFS_I(_Generic((_inode), \
905 struct inode *: (_inode)))->root->fs_info)
906
btrfs_alloc_write_mask(struct address_space * mapping)907 static inline gfp_t btrfs_alloc_write_mask(struct address_space *mapping)
908 {
909 return mapping_gfp_constraint(mapping, ~__GFP_FS);
910 }
911
btrfs_get_fs_generation(const struct btrfs_fs_info * fs_info)912 static inline u64 btrfs_get_fs_generation(const struct btrfs_fs_info *fs_info)
913 {
914 return READ_ONCE(fs_info->generation);
915 }
916
btrfs_set_fs_generation(struct btrfs_fs_info * fs_info,u64 gen)917 static inline void btrfs_set_fs_generation(struct btrfs_fs_info *fs_info, u64 gen)
918 {
919 WRITE_ONCE(fs_info->generation, gen);
920 }
921
btrfs_get_last_trans_committed(const struct btrfs_fs_info * fs_info)922 static inline u64 btrfs_get_last_trans_committed(const struct btrfs_fs_info *fs_info)
923 {
924 return READ_ONCE(fs_info->last_trans_committed);
925 }
926
btrfs_set_last_trans_committed(struct btrfs_fs_info * fs_info,u64 gen)927 static inline void btrfs_set_last_trans_committed(struct btrfs_fs_info *fs_info, u64 gen)
928 {
929 WRITE_ONCE(fs_info->last_trans_committed, gen);
930 }
931
btrfs_set_last_root_drop_gen(struct btrfs_fs_info * fs_info,u64 gen)932 static inline void btrfs_set_last_root_drop_gen(struct btrfs_fs_info *fs_info,
933 u64 gen)
934 {
935 WRITE_ONCE(fs_info->last_root_drop_gen, gen);
936 }
937
btrfs_get_last_root_drop_gen(const struct btrfs_fs_info * fs_info)938 static inline u64 btrfs_get_last_root_drop_gen(const struct btrfs_fs_info *fs_info)
939 {
940 return READ_ONCE(fs_info->last_root_drop_gen);
941 }
942
943 /*
944 * Take the number of bytes to be checksummed and figure out how many leaves
945 * it would require to store the csums for that many bytes.
946 */
btrfs_csum_bytes_to_leaves(const struct btrfs_fs_info * fs_info,u64 csum_bytes)947 static inline u64 btrfs_csum_bytes_to_leaves(
948 const struct btrfs_fs_info *fs_info, u64 csum_bytes)
949 {
950 const u64 num_csums = csum_bytes >> fs_info->sectorsize_bits;
951
952 return DIV_ROUND_UP_ULL(num_csums, fs_info->csums_per_leaf);
953 }
954
955 /*
956 * Use this if we would be adding new items, as we could split nodes as we cow
957 * down the tree.
958 */
btrfs_calc_insert_metadata_size(const struct btrfs_fs_info * fs_info,unsigned num_items)959 static inline u64 btrfs_calc_insert_metadata_size(const struct btrfs_fs_info *fs_info,
960 unsigned num_items)
961 {
962 return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * 2 * num_items;
963 }
964
965 /*
966 * Doing a truncate or a modification won't result in new nodes or leaves, just
967 * what we need for COW.
968 */
btrfs_calc_metadata_size(const struct btrfs_fs_info * fs_info,unsigned num_items)969 static inline u64 btrfs_calc_metadata_size(const struct btrfs_fs_info *fs_info,
970 unsigned num_items)
971 {
972 return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * num_items;
973 }
974
975 #define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r->fs_info) >> 4) - \
976 sizeof(struct btrfs_item))
977
978 #define BTRFS_BYTES_TO_BLKS(fs_info, bytes) ((bytes) >> (fs_info)->sectorsize_bits)
979
btrfs_is_zoned(const struct btrfs_fs_info * fs_info)980 static inline bool btrfs_is_zoned(const struct btrfs_fs_info *fs_info)
981 {
982 return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && fs_info->zone_size > 0;
983 }
984
985 /*
986 * Count how many fs_info->max_extent_size cover the @size
987 */
count_max_extents(const struct btrfs_fs_info * fs_info,u64 size)988 static inline u32 count_max_extents(const struct btrfs_fs_info *fs_info, u64 size)
989 {
990 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
991 if (!fs_info)
992 return div_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, BTRFS_MAX_EXTENT_SIZE);
993 #endif
994
995 return div_u64(size + fs_info->max_extent_size - 1, fs_info->max_extent_size);
996 }
997
btrfs_blocks_per_folio(const struct btrfs_fs_info * fs_info,const struct folio * folio)998 static inline unsigned int btrfs_blocks_per_folio(const struct btrfs_fs_info *fs_info,
999 const struct folio *folio)
1000 {
1001 return folio_size(folio) >> fs_info->sectorsize_bits;
1002 }
1003
1004 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
1005 enum btrfs_exclusive_operation type);
1006 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
1007 enum btrfs_exclusive_operation type);
1008 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info);
1009 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info);
1010 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
1011 enum btrfs_exclusive_operation op);
1012
1013 int btrfs_check_ioctl_vol_args_path(const struct btrfs_ioctl_vol_args *vol_args);
1014
1015 u16 btrfs_csum_type_size(u16 type);
1016 int btrfs_super_csum_size(const struct btrfs_super_block *s);
1017 const char *btrfs_super_csum_name(u16 csum_type);
1018 const char *btrfs_super_csum_driver(u16 csum_type);
1019 size_t __attribute_const__ btrfs_get_num_csums(void);
1020
btrfs_is_empty_uuid(const u8 * uuid)1021 static inline bool btrfs_is_empty_uuid(const u8 *uuid)
1022 {
1023 return uuid_is_null((const uuid_t *)uuid);
1024 }
1025
1026 /* Compatibility and incompatibility defines */
1027 void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
1028 const char *name);
1029 void __btrfs_clear_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
1030 const char *name);
1031 void __btrfs_set_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
1032 const char *name);
1033 void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
1034 const char *name);
1035
1036 #define __btrfs_fs_incompat(fs_info, flags) \
1037 (!!(btrfs_super_incompat_flags((fs_info)->super_copy) & (flags)))
1038
1039 #define __btrfs_fs_compat_ro(fs_info, flags) \
1040 (!!(btrfs_super_compat_ro_flags((fs_info)->super_copy) & (flags)))
1041
1042 #define btrfs_set_fs_incompat(__fs_info, opt) \
1043 __btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
1044
1045 #define btrfs_clear_fs_incompat(__fs_info, opt) \
1046 __btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
1047
1048 #define btrfs_fs_incompat(fs_info, opt) \
1049 __btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
1050
1051 #define btrfs_set_fs_compat_ro(__fs_info, opt) \
1052 __btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
1053
1054 #define btrfs_clear_fs_compat_ro(__fs_info, opt) \
1055 __btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
1056
1057 #define btrfs_fs_compat_ro(fs_info, opt) \
1058 __btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
1059
1060 #define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt)
1061 #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt)
1062 #define btrfs_raw_test_opt(o, opt) ((o) & BTRFS_MOUNT_##opt)
1063 #define btrfs_test_opt(fs_info, opt) ((fs_info)->mount_opt & \
1064 BTRFS_MOUNT_##opt)
1065
btrfs_fs_closing(const struct btrfs_fs_info * fs_info)1066 static inline int btrfs_fs_closing(const struct btrfs_fs_info *fs_info)
1067 {
1068 /* Do it this way so we only ever do one test_bit in the normal case. */
1069 if (test_bit(BTRFS_FS_CLOSING_START, &fs_info->flags)) {
1070 if (test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags))
1071 return 2;
1072 return 1;
1073 }
1074 return 0;
1075 }
1076
1077 /*
1078 * If we remount the fs to be R/O or umount the fs, the cleaner needn't do
1079 * anything except sleeping. This function is used to check the status of
1080 * the fs.
1081 * We check for BTRFS_FS_STATE_RO to avoid races with a concurrent remount,
1082 * since setting and checking for SB_RDONLY in the superblock's flags is not
1083 * atomic.
1084 */
btrfs_need_cleaner_sleep(const struct btrfs_fs_info * fs_info)1085 static inline int btrfs_need_cleaner_sleep(const struct btrfs_fs_info *fs_info)
1086 {
1087 return test_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state) ||
1088 btrfs_fs_closing(fs_info);
1089 }
1090
btrfs_wake_unfinished_drop(struct btrfs_fs_info * fs_info)1091 static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info *fs_info)
1092 {
1093 clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags);
1094 }
1095
1096 #define BTRFS_FS_ERROR(fs_info) (READ_ONCE((fs_info)->fs_error))
1097
1098 #define BTRFS_FS_LOG_CLEANUP_ERROR(fs_info) \
1099 (unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR, \
1100 &(fs_info)->fs_state)))
1101
1102 /*
1103 * We use folio flag owner_2 to indicate there is an ordered extent with
1104 * unfinished IO.
1105 */
1106 #define folio_test_ordered(folio) folio_test_owner_2(folio)
1107 #define folio_set_ordered(folio) folio_set_owner_2(folio)
1108 #define folio_clear_ordered(folio) folio_clear_owner_2(folio)
1109
1110 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1111
1112 #define EXPORT_FOR_TESTS
1113
btrfs_is_testing(const struct btrfs_fs_info * fs_info)1114 static inline int btrfs_is_testing(const struct btrfs_fs_info *fs_info)
1115 {
1116 return test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
1117 }
1118
1119 void btrfs_test_destroy_inode(struct inode *inode);
1120
1121 #else
1122
1123 #define EXPORT_FOR_TESTS static
1124
btrfs_is_testing(const struct btrfs_fs_info * fs_info)1125 static inline int btrfs_is_testing(const struct btrfs_fs_info *fs_info)
1126 {
1127 return 0;
1128 }
1129 #endif
1130
1131 #endif
1132