xref: /linux/fs/btrfs/fs.h (revision eb2f932100288dbb881eadfed02e1459c6b9504c)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 
3 #ifndef BTRFS_FS_H
4 #define BTRFS_FS_H
5 
6 #include <linux/blkdev.h>
7 #include <linux/fs.h>
8 #include <linux/btrfs_tree.h>
9 #include <linux/sizes.h>
10 #include "extent-io-tree.h"
11 #include "extent_map.h"
12 #include "async-thread.h"
13 #include "block-rsv.h"
14 
15 #define BTRFS_MAX_EXTENT_SIZE SZ_128M
16 
17 #define BTRFS_OLDEST_GENERATION	0ULL
18 
19 #define BTRFS_EMPTY_DIR_SIZE 0
20 
21 #define BTRFS_DIRTY_METADATA_THRESH		SZ_32M
22 
23 #define BTRFS_SUPER_INFO_OFFSET			SZ_64K
24 #define BTRFS_SUPER_INFO_SIZE			4096
25 static_assert(sizeof(struct btrfs_super_block) == BTRFS_SUPER_INFO_SIZE);
26 
27 /*
28  * Number of metadata items necessary for an unlink operation:
29  *
30  * 1 for the possible orphan item
31  * 1 for the dir item
32  * 1 for the dir index
33  * 1 for the inode ref
34  * 1 for the inode
35  * 1 for the parent inode
36  */
37 #define BTRFS_UNLINK_METADATA_UNITS		6
38 
39 /*
40  * The reserved space at the beginning of each device.  It covers the primary
41  * super block and leaves space for potential use by other tools like
42  * bootloaders or to lower potential damage of accidental overwrite.
43  */
44 #define BTRFS_DEVICE_RANGE_RESERVED			(SZ_1M)
45 /*
46  * Runtime (in-memory) states of filesystem
47  */
48 enum {
49 	/*
50 	 * Filesystem is being remounted, allow to skip some operations, like
51 	 * defrag
52 	 */
53 	BTRFS_FS_STATE_REMOUNTING,
54 	/* Filesystem in RO mode */
55 	BTRFS_FS_STATE_RO,
56 	/* Track if a transaction abort has been reported on this filesystem */
57 	BTRFS_FS_STATE_TRANS_ABORTED,
58 	/*
59 	 * Bio operations should be blocked on this filesystem because a source
60 	 * or target device is being destroyed as part of a device replace
61 	 */
62 	BTRFS_FS_STATE_DEV_REPLACING,
63 	/* The btrfs_fs_info created for self-tests */
64 	BTRFS_FS_STATE_DUMMY_FS_INFO,
65 
66 	BTRFS_FS_STATE_NO_CSUMS,
67 
68 	/* Indicates there was an error cleaning up a log tree. */
69 	BTRFS_FS_STATE_LOG_CLEANUP_ERROR,
70 
71 	BTRFS_FS_STATE_COUNT
72 };
73 
74 enum {
75 	BTRFS_FS_CLOSING_START,
76 	BTRFS_FS_CLOSING_DONE,
77 	BTRFS_FS_LOG_RECOVERING,
78 	BTRFS_FS_OPEN,
79 	BTRFS_FS_QUOTA_ENABLED,
80 	BTRFS_FS_UPDATE_UUID_TREE_GEN,
81 	BTRFS_FS_CREATING_FREE_SPACE_TREE,
82 	BTRFS_FS_BTREE_ERR,
83 	BTRFS_FS_LOG1_ERR,
84 	BTRFS_FS_LOG2_ERR,
85 	BTRFS_FS_QUOTA_OVERRIDE,
86 	/* Used to record internally whether fs has been frozen */
87 	BTRFS_FS_FROZEN,
88 	/*
89 	 * Indicate that balance has been set up from the ioctl and is in the
90 	 * main phase. The fs_info::balance_ctl is initialized.
91 	 */
92 	BTRFS_FS_BALANCE_RUNNING,
93 
94 	/*
95 	 * Indicate that relocation of a chunk has started, it's set per chunk
96 	 * and is toggled between chunks.
97 	 */
98 	BTRFS_FS_RELOC_RUNNING,
99 
100 	/* Indicate that the cleaner thread is awake and doing something. */
101 	BTRFS_FS_CLEANER_RUNNING,
102 
103 	/*
104 	 * The checksumming has an optimized version and is considered fast,
105 	 * so we don't need to offload checksums to workqueues.
106 	 */
107 	BTRFS_FS_CSUM_IMPL_FAST,
108 
109 	/* Indicate that the discard workqueue can service discards. */
110 	BTRFS_FS_DISCARD_RUNNING,
111 
112 	/* Indicate that we need to cleanup space cache v1 */
113 	BTRFS_FS_CLEANUP_SPACE_CACHE_V1,
114 
115 	/* Indicate that we can't trust the free space tree for caching yet */
116 	BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED,
117 
118 	/* Indicate whether there are any tree modification log users */
119 	BTRFS_FS_TREE_MOD_LOG_USERS,
120 
121 	/* Indicate that we want the transaction kthread to commit right now. */
122 	BTRFS_FS_COMMIT_TRANS,
123 
124 	/* Indicate we have half completed snapshot deletions pending. */
125 	BTRFS_FS_UNFINISHED_DROPS,
126 
127 	/* Indicate we have to finish a zone to do next allocation. */
128 	BTRFS_FS_NEED_ZONE_FINISH,
129 
130 	/* Indicate that we want to commit the transaction. */
131 	BTRFS_FS_NEED_TRANS_COMMIT,
132 
133 	/* This is set when active zone tracking is needed. */
134 	BTRFS_FS_ACTIVE_ZONE_TRACKING,
135 
136 	/*
137 	 * Indicate if we have some features changed, this is mostly for
138 	 * cleaner thread to update the sysfs interface.
139 	 */
140 	BTRFS_FS_FEATURE_CHANGED,
141 
142 	/*
143 	 * Indicate that we have found a tree block which is only aligned to
144 	 * sectorsize, but not to nodesize.  This should be rare nowadays.
145 	 */
146 	BTRFS_FS_UNALIGNED_TREE_BLOCK,
147 
148 #if BITS_PER_LONG == 32
149 	/* Indicate if we have error/warn message printed on 32bit systems */
150 	BTRFS_FS_32BIT_ERROR,
151 	BTRFS_FS_32BIT_WARN,
152 #endif
153 };
154 
155 /*
156  * Flags for mount options.
157  *
158  * Note: don't forget to add new options to btrfs_show_options()
159  */
160 enum {
161 	BTRFS_MOUNT_NODATASUM			= (1UL << 0),
162 	BTRFS_MOUNT_NODATACOW			= (1UL << 1),
163 	BTRFS_MOUNT_NOBARRIER			= (1UL << 2),
164 	BTRFS_MOUNT_SSD				= (1UL << 3),
165 	BTRFS_MOUNT_DEGRADED			= (1UL << 4),
166 	BTRFS_MOUNT_COMPRESS			= (1UL << 5),
167 	BTRFS_MOUNT_NOTREELOG   		= (1UL << 6),
168 	BTRFS_MOUNT_FLUSHONCOMMIT		= (1UL << 7),
169 	BTRFS_MOUNT_SSD_SPREAD			= (1UL << 8),
170 	BTRFS_MOUNT_NOSSD			= (1UL << 9),
171 	BTRFS_MOUNT_DISCARD_SYNC		= (1UL << 10),
172 	BTRFS_MOUNT_FORCE_COMPRESS      	= (1UL << 11),
173 	BTRFS_MOUNT_SPACE_CACHE			= (1UL << 12),
174 	BTRFS_MOUNT_CLEAR_CACHE			= (1UL << 13),
175 	BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED	= (1UL << 14),
176 	BTRFS_MOUNT_ENOSPC_DEBUG		= (1UL << 15),
177 	BTRFS_MOUNT_AUTO_DEFRAG			= (1UL << 16),
178 	BTRFS_MOUNT_USEBACKUPROOT		= (1UL << 17),
179 	BTRFS_MOUNT_SKIP_BALANCE		= (1UL << 18),
180 	BTRFS_MOUNT_PANIC_ON_FATAL_ERROR	= (1UL << 19),
181 	BTRFS_MOUNT_RESCAN_UUID_TREE		= (1UL << 20),
182 	BTRFS_MOUNT_FRAGMENT_DATA		= (1UL << 21),
183 	BTRFS_MOUNT_FRAGMENT_METADATA		= (1UL << 22),
184 	BTRFS_MOUNT_FREE_SPACE_TREE		= (1UL << 23),
185 	BTRFS_MOUNT_NOLOGREPLAY			= (1UL << 24),
186 	BTRFS_MOUNT_REF_VERIFY			= (1UL << 25),
187 	BTRFS_MOUNT_DISCARD_ASYNC		= (1UL << 26),
188 	BTRFS_MOUNT_IGNOREBADROOTS		= (1UL << 27),
189 	BTRFS_MOUNT_IGNOREDATACSUMS		= (1UL << 28),
190 	BTRFS_MOUNT_NODISCARD			= (1UL << 29),
191 	BTRFS_MOUNT_NOSPACECACHE		= (1UL << 30),
192 };
193 
194 /*
195  * Compat flags that we support.  If any incompat flags are set other than the
196  * ones specified below then we will fail to mount
197  */
198 #define BTRFS_FEATURE_COMPAT_SUPP		0ULL
199 #define BTRFS_FEATURE_COMPAT_SAFE_SET		0ULL
200 #define BTRFS_FEATURE_COMPAT_SAFE_CLEAR		0ULL
201 
202 #define BTRFS_FEATURE_COMPAT_RO_SUPP			\
203 	(BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE |	\
204 	 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID | \
205 	 BTRFS_FEATURE_COMPAT_RO_VERITY |		\
206 	 BTRFS_FEATURE_COMPAT_RO_BLOCK_GROUP_TREE)
207 
208 #define BTRFS_FEATURE_COMPAT_RO_SAFE_SET	0ULL
209 #define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR	0ULL
210 
211 #define BTRFS_FEATURE_INCOMPAT_SUPP_STABLE		\
212 	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
213 	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
214 	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
215 	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
216 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
217 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD |		\
218 	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
219 	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF |		\
220 	 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA |	\
221 	 BTRFS_FEATURE_INCOMPAT_NO_HOLES	|	\
222 	 BTRFS_FEATURE_INCOMPAT_METADATA_UUID	|	\
223 	 BTRFS_FEATURE_INCOMPAT_RAID1C34	|	\
224 	 BTRFS_FEATURE_INCOMPAT_ZONED		|	\
225 	 BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA)
226 
227 #ifdef CONFIG_BTRFS_DEBUG
228 	/*
229 	 * Features under developmen like Extent tree v2 support is enabled
230 	 * only under CONFIG_BTRFS_DEBUG.
231 	 */
232 #define BTRFS_FEATURE_INCOMPAT_SUPP		\
233 	(BTRFS_FEATURE_INCOMPAT_SUPP_STABLE |	\
234 	 BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE | \
235 	 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2)
236 
237 #else
238 
239 #define BTRFS_FEATURE_INCOMPAT_SUPP		\
240 	(BTRFS_FEATURE_INCOMPAT_SUPP_STABLE)
241 
242 #endif
243 
244 #define BTRFS_FEATURE_INCOMPAT_SAFE_SET			\
245 	(BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
246 #define BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR		0ULL
247 
248 #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
249 #define BTRFS_DEFAULT_MAX_INLINE	(2048)
250 
251 struct btrfs_dev_replace {
252 	/* See #define above */
253 	u64 replace_state;
254 	/* Seconds since 1-Jan-1970 */
255 	time64_t time_started;
256 	/* Seconds since 1-Jan-1970 */
257 	time64_t time_stopped;
258 	atomic64_t num_write_errors;
259 	atomic64_t num_uncorrectable_read_errors;
260 
261 	u64 cursor_left;
262 	u64 committed_cursor_left;
263 	u64 cursor_left_last_write_of_item;
264 	u64 cursor_right;
265 
266 	/* See #define above */
267 	u64 cont_reading_from_srcdev_mode;
268 
269 	int is_valid;
270 	int item_needs_writeback;
271 	struct btrfs_device *srcdev;
272 	struct btrfs_device *tgtdev;
273 
274 	struct mutex lock_finishing_cancel_unmount;
275 	struct rw_semaphore rwsem;
276 
277 	struct btrfs_scrub_progress scrub_progress;
278 
279 	struct percpu_counter bio_counter;
280 	wait_queue_head_t replace_wait;
281 };
282 
283 /*
284  * Free clusters are used to claim free space in relatively large chunks,
285  * allowing us to do less seeky writes. They are used for all metadata
286  * allocations. In ssd_spread mode they are also used for data allocations.
287  */
288 struct btrfs_free_cluster {
289 	spinlock_t lock;
290 	spinlock_t refill_lock;
291 	struct rb_root root;
292 
293 	/* Largest extent in this cluster */
294 	u64 max_size;
295 
296 	/* First extent starting offset */
297 	u64 window_start;
298 
299 	/* We did a full search and couldn't create a cluster */
300 	bool fragmented;
301 
302 	struct btrfs_block_group *block_group;
303 	/*
304 	 * When a cluster is allocated from a block group, we put the cluster
305 	 * onto a list in the block group so that it can be freed before the
306 	 * block group is freed.
307 	 */
308 	struct list_head block_group_list;
309 };
310 
311 /* Discard control. */
312 /*
313  * Async discard uses multiple lists to differentiate the discard filter
314  * parameters.  Index 0 is for completely free block groups where we need to
315  * ensure the entire block group is trimmed without being lossy.  Indices
316  * afterwards represent monotonically decreasing discard filter sizes to
317  * prioritize what should be discarded next.
318  */
319 #define BTRFS_NR_DISCARD_LISTS		3
320 #define BTRFS_DISCARD_INDEX_UNUSED	0
321 #define BTRFS_DISCARD_INDEX_START	1
322 
323 struct btrfs_discard_ctl {
324 	struct workqueue_struct *discard_workers;
325 	struct delayed_work work;
326 	spinlock_t lock;
327 	struct btrfs_block_group *block_group;
328 	struct list_head discard_list[BTRFS_NR_DISCARD_LISTS];
329 	u64 prev_discard;
330 	u64 prev_discard_time;
331 	atomic_t discardable_extents;
332 	atomic64_t discardable_bytes;
333 	u64 max_discard_size;
334 	u64 delay_ms;
335 	u32 iops_limit;
336 	u32 kbps_limit;
337 	u64 discard_extent_bytes;
338 	u64 discard_bitmap_bytes;
339 	atomic64_t discard_bytes_saved;
340 };
341 
342 /*
343  * Exclusive operations (device replace, resize, device add/remove, balance)
344  */
345 enum btrfs_exclusive_operation {
346 	BTRFS_EXCLOP_NONE,
347 	BTRFS_EXCLOP_BALANCE_PAUSED,
348 	BTRFS_EXCLOP_BALANCE,
349 	BTRFS_EXCLOP_DEV_ADD,
350 	BTRFS_EXCLOP_DEV_REMOVE,
351 	BTRFS_EXCLOP_DEV_REPLACE,
352 	BTRFS_EXCLOP_RESIZE,
353 	BTRFS_EXCLOP_SWAP_ACTIVATE,
354 };
355 
356 /* Store data about transaction commits, exported via sysfs. */
357 struct btrfs_commit_stats {
358 	/* Total number of commits */
359 	u64 commit_count;
360 	/* The maximum commit duration so far in ns */
361 	u64 max_commit_dur;
362 	/* The last commit duration in ns */
363 	u64 last_commit_dur;
364 	/* The total commit duration in ns */
365 	u64 total_commit_dur;
366 };
367 
368 struct btrfs_fs_info {
369 	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
370 	unsigned long flags;
371 	struct btrfs_root *tree_root;
372 	struct btrfs_root *chunk_root;
373 	struct btrfs_root *dev_root;
374 	struct btrfs_root *fs_root;
375 	struct btrfs_root *quota_root;
376 	struct btrfs_root *uuid_root;
377 	struct btrfs_root *data_reloc_root;
378 	struct btrfs_root *block_group_root;
379 	struct btrfs_root *stripe_root;
380 
381 	/* The log root tree is a directory of all the other log roots */
382 	struct btrfs_root *log_root_tree;
383 
384 	/* The tree that holds the global roots (csum, extent, etc) */
385 	rwlock_t global_root_lock;
386 	struct rb_root global_root_tree;
387 
388 	spinlock_t fs_roots_radix_lock;
389 	struct radix_tree_root fs_roots_radix;
390 
391 	/* Block group cache stuff */
392 	rwlock_t block_group_cache_lock;
393 	struct rb_root_cached block_group_cache_tree;
394 
395 	/* Keep track of unallocated space */
396 	atomic64_t free_chunk_space;
397 
398 	/* Track ranges which are used by log trees blocks/logged data extents */
399 	struct extent_io_tree excluded_extents;
400 
401 	/* logical->physical extent mapping */
402 	struct rb_root_cached mapping_tree;
403 	rwlock_t mapping_tree_lock;
404 
405 	/*
406 	 * Block reservation for extent, checksum, root tree and delayed dir
407 	 * index item.
408 	 */
409 	struct btrfs_block_rsv global_block_rsv;
410 	/* Block reservation for metadata operations */
411 	struct btrfs_block_rsv trans_block_rsv;
412 	/* Block reservation for chunk tree */
413 	struct btrfs_block_rsv chunk_block_rsv;
414 	/* Block reservation for delayed operations */
415 	struct btrfs_block_rsv delayed_block_rsv;
416 	/* Block reservation for delayed refs */
417 	struct btrfs_block_rsv delayed_refs_rsv;
418 
419 	struct btrfs_block_rsv empty_block_rsv;
420 
421 	/*
422 	 * Updated while holding the lock 'trans_lock'. Due to the life cycle of
423 	 * a transaction, it can be directly read while holding a transaction
424 	 * handle, everywhere else must be read with btrfs_get_fs_generation().
425 	 * Should always be updated using btrfs_set_fs_generation().
426 	 */
427 	u64 generation;
428 	/*
429 	 * Always use btrfs_get_last_trans_committed() and
430 	 * btrfs_set_last_trans_committed() to read and update this field.
431 	 */
432 	u64 last_trans_committed;
433 	/*
434 	 * Generation of the last transaction used for block group relocation
435 	 * since the filesystem was last mounted (or 0 if none happened yet).
436 	 * Must be written and read while holding btrfs_fs_info::commit_root_sem.
437 	 */
438 	u64 last_reloc_trans;
439 
440 	/*
441 	 * This is updated to the current trans every time a full commit is
442 	 * required instead of the faster short fsync log commits
443 	 */
444 	u64 last_trans_log_full_commit;
445 	unsigned long mount_opt;
446 
447 	unsigned long compress_type:4;
448 	unsigned int compress_level;
449 	u32 commit_interval;
450 	/*
451 	 * It is a suggestive number, the read side is safe even it gets a
452 	 * wrong number because we will write out the data into a regular
453 	 * extent. The write side(mount/remount) is under ->s_umount lock,
454 	 * so it is also safe.
455 	 */
456 	u64 max_inline;
457 
458 	struct btrfs_transaction *running_transaction;
459 	wait_queue_head_t transaction_throttle;
460 	wait_queue_head_t transaction_wait;
461 	wait_queue_head_t transaction_blocked_wait;
462 	wait_queue_head_t async_submit_wait;
463 
464 	/*
465 	 * Used to protect the incompat_flags, compat_flags, compat_ro_flags
466 	 * when they are updated.
467 	 *
468 	 * Because we do not clear the flags for ever, so we needn't use
469 	 * the lock on the read side.
470 	 *
471 	 * We also needn't use the lock when we mount the fs, because
472 	 * there is no other task which will update the flag.
473 	 */
474 	spinlock_t super_lock;
475 	struct btrfs_super_block *super_copy;
476 	struct btrfs_super_block *super_for_commit;
477 	struct super_block *sb;
478 	struct inode *btree_inode;
479 	struct mutex tree_log_mutex;
480 	struct mutex transaction_kthread_mutex;
481 	struct mutex cleaner_mutex;
482 	struct mutex chunk_mutex;
483 
484 	/*
485 	 * This is taken to make sure we don't set block groups ro after the
486 	 * free space cache has been allocated on them.
487 	 */
488 	struct mutex ro_block_group_mutex;
489 
490 	/*
491 	 * This is used during read/modify/write to make sure no two ios are
492 	 * trying to mod the same stripe at the same time.
493 	 */
494 	struct btrfs_stripe_hash_table *stripe_hash_table;
495 
496 	/*
497 	 * This protects the ordered operations list only while we are
498 	 * processing all of the entries on it.  This way we make sure the
499 	 * commit code doesn't find the list temporarily empty because another
500 	 * function happens to be doing non-waiting preflush before jumping
501 	 * into the main commit.
502 	 */
503 	struct mutex ordered_operations_mutex;
504 
505 	struct rw_semaphore commit_root_sem;
506 
507 	struct rw_semaphore cleanup_work_sem;
508 
509 	struct rw_semaphore subvol_sem;
510 
511 	spinlock_t trans_lock;
512 	/*
513 	 * The reloc mutex goes with the trans lock, it is taken during commit
514 	 * to protect us from the relocation code.
515 	 */
516 	struct mutex reloc_mutex;
517 
518 	struct list_head trans_list;
519 	struct list_head dead_roots;
520 	struct list_head caching_block_groups;
521 
522 	spinlock_t delayed_iput_lock;
523 	struct list_head delayed_iputs;
524 	atomic_t nr_delayed_iputs;
525 	wait_queue_head_t delayed_iputs_wait;
526 
527 	atomic64_t tree_mod_seq;
528 
529 	/* This protects tree_mod_log and tree_mod_seq_list */
530 	rwlock_t tree_mod_log_lock;
531 	struct rb_root tree_mod_log;
532 	struct list_head tree_mod_seq_list;
533 
534 	atomic_t async_delalloc_pages;
535 
536 	/* This is used to protect the following list -- ordered_roots. */
537 	spinlock_t ordered_root_lock;
538 
539 	/*
540 	 * All fs/file tree roots in which there are data=ordered extents
541 	 * pending writeback are added into this list.
542 	 *
543 	 * These can span multiple transactions and basically include every
544 	 * dirty data page that isn't from nodatacow.
545 	 */
546 	struct list_head ordered_roots;
547 
548 	struct mutex delalloc_root_mutex;
549 	spinlock_t delalloc_root_lock;
550 	/* All fs/file tree roots that have delalloc inodes. */
551 	struct list_head delalloc_roots;
552 
553 	/*
554 	 * There is a pool of worker threads for checksumming during writes and
555 	 * a pool for checksumming after reads.  This is because readers can
556 	 * run with FS locks held, and the writers may be waiting for those
557 	 * locks.  We don't want ordering in the pending list to cause
558 	 * deadlocks, and so the two are serviced separately.
559 	 *
560 	 * A third pool does submit_bio to avoid deadlocking with the other two.
561 	 */
562 	struct btrfs_workqueue *workers;
563 	struct btrfs_workqueue *delalloc_workers;
564 	struct btrfs_workqueue *flush_workers;
565 	struct workqueue_struct *endio_workers;
566 	struct workqueue_struct *endio_meta_workers;
567 	struct workqueue_struct *rmw_workers;
568 	struct workqueue_struct *compressed_write_workers;
569 	struct btrfs_workqueue *endio_write_workers;
570 	struct btrfs_workqueue *endio_freespace_worker;
571 	struct btrfs_workqueue *caching_workers;
572 
573 	/*
574 	 * Fixup workers take dirty pages that didn't properly go through the
575 	 * cow mechanism and make them safe to write.  It happens for the
576 	 * sys_munmap function call path.
577 	 */
578 	struct btrfs_workqueue *fixup_workers;
579 	struct btrfs_workqueue *delayed_workers;
580 
581 	struct task_struct *transaction_kthread;
582 	struct task_struct *cleaner_kthread;
583 	u32 thread_pool_size;
584 
585 	struct kobject *space_info_kobj;
586 	struct kobject *qgroups_kobj;
587 	struct kobject *discard_kobj;
588 
589 	/* Used to keep from writing metadata until there is a nice batch */
590 	struct percpu_counter dirty_metadata_bytes;
591 	struct percpu_counter delalloc_bytes;
592 	struct percpu_counter ordered_bytes;
593 	s32 dirty_metadata_batch;
594 	s32 delalloc_batch;
595 
596 	/* Protected by 'trans_lock'. */
597 	struct list_head dirty_cowonly_roots;
598 
599 	struct btrfs_fs_devices *fs_devices;
600 
601 	/*
602 	 * The space_info list is effectively read only after initial setup.
603 	 * It is populated at mount time and cleaned up after all block groups
604 	 * are removed.  RCU is used to protect it.
605 	 */
606 	struct list_head space_info;
607 
608 	struct btrfs_space_info *data_sinfo;
609 
610 	struct reloc_control *reloc_ctl;
611 
612 	/* data_alloc_cluster is only used in ssd_spread mode */
613 	struct btrfs_free_cluster data_alloc_cluster;
614 
615 	/* All metadata allocations go through this cluster. */
616 	struct btrfs_free_cluster meta_alloc_cluster;
617 
618 	/* Auto defrag inodes go here. */
619 	spinlock_t defrag_inodes_lock;
620 	struct rb_root defrag_inodes;
621 	atomic_t defrag_running;
622 
623 	/* Used to protect avail_{data, metadata, system}_alloc_bits */
624 	seqlock_t profiles_lock;
625 	/*
626 	 * These three are in extended format (availability of single chunks is
627 	 * denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other types are denoted
628 	 * by corresponding BTRFS_BLOCK_GROUP_* bits)
629 	 */
630 	u64 avail_data_alloc_bits;
631 	u64 avail_metadata_alloc_bits;
632 	u64 avail_system_alloc_bits;
633 
634 	/* Balance state */
635 	spinlock_t balance_lock;
636 	struct mutex balance_mutex;
637 	atomic_t balance_pause_req;
638 	atomic_t balance_cancel_req;
639 	struct btrfs_balance_control *balance_ctl;
640 	wait_queue_head_t balance_wait_q;
641 
642 	/* Cancellation requests for chunk relocation */
643 	atomic_t reloc_cancel_req;
644 
645 	u32 data_chunk_allocations;
646 	u32 metadata_ratio;
647 
648 	void *bdev_holder;
649 
650 	/* Private scrub information */
651 	struct mutex scrub_lock;
652 	atomic_t scrubs_running;
653 	atomic_t scrub_pause_req;
654 	atomic_t scrubs_paused;
655 	atomic_t scrub_cancel_req;
656 	wait_queue_head_t scrub_pause_wait;
657 	/*
658 	 * The worker pointers are NULL iff the refcount is 0, ie. scrub is not
659 	 * running.
660 	 */
661 	refcount_t scrub_workers_refcnt;
662 	struct workqueue_struct *scrub_workers;
663 	struct btrfs_subpage_info *subpage_info;
664 
665 	struct btrfs_discard_ctl discard_ctl;
666 
667 	/* Is qgroup tracking in a consistent state? */
668 	u64 qgroup_flags;
669 
670 	/* Holds configuration and tracking. Protected by qgroup_lock. */
671 	struct rb_root qgroup_tree;
672 	spinlock_t qgroup_lock;
673 
674 	/*
675 	 * Used to avoid frequently calling ulist_alloc()/ulist_free()
676 	 * when doing qgroup accounting, it must be protected by qgroup_lock.
677 	 */
678 	struct ulist *qgroup_ulist;
679 
680 	/*
681 	 * Protect user change for quota operations. If a transaction is needed,
682 	 * it must be started before locking this lock.
683 	 */
684 	struct mutex qgroup_ioctl_lock;
685 
686 	/* List of dirty qgroups to be written at next commit. */
687 	struct list_head dirty_qgroups;
688 
689 	/* Used by qgroup for an efficient tree traversal. */
690 	u64 qgroup_seq;
691 
692 	/* Qgroup rescan items. */
693 	/* Protects the progress item */
694 	struct mutex qgroup_rescan_lock;
695 	struct btrfs_key qgroup_rescan_progress;
696 	struct btrfs_workqueue *qgroup_rescan_workers;
697 	struct completion qgroup_rescan_completion;
698 	struct btrfs_work qgroup_rescan_work;
699 	/* Protected by qgroup_rescan_lock */
700 	bool qgroup_rescan_running;
701 	u8 qgroup_drop_subtree_thres;
702 	u64 qgroup_enable_gen;
703 
704 	/*
705 	 * If this is not 0, then it indicates a serious filesystem error has
706 	 * happened and it contains that error (negative errno value).
707 	 */
708 	int fs_error;
709 
710 	/* Filesystem state */
711 	unsigned long fs_state;
712 
713 	struct btrfs_delayed_root *delayed_root;
714 
715 	/* Extent buffer radix tree */
716 	spinlock_t buffer_lock;
717 	/* Entries are eb->start / sectorsize */
718 	struct radix_tree_root buffer_radix;
719 
720 	/* Next backup root to be overwritten */
721 	int backup_root_index;
722 
723 	/* Device replace state */
724 	struct btrfs_dev_replace dev_replace;
725 
726 	struct semaphore uuid_tree_rescan_sem;
727 
728 	/* Used to reclaim the metadata space in the background. */
729 	struct work_struct async_reclaim_work;
730 	struct work_struct async_data_reclaim_work;
731 	struct work_struct preempt_reclaim_work;
732 
733 	/* Reclaim partially filled block groups in the background */
734 	struct work_struct reclaim_bgs_work;
735 	struct list_head reclaim_bgs;
736 	int bg_reclaim_threshold;
737 
738 	spinlock_t unused_bgs_lock;
739 	struct list_head unused_bgs;
740 	struct mutex unused_bg_unpin_mutex;
741 	/* Protect block groups that are going to be deleted */
742 	struct mutex reclaim_bgs_lock;
743 
744 	/* Cached block sizes */
745 	u32 nodesize;
746 	u32 sectorsize;
747 	/* ilog2 of sectorsize, use to avoid 64bit division */
748 	u32 sectorsize_bits;
749 	u32 csum_size;
750 	u32 csums_per_leaf;
751 	u32 stripesize;
752 
753 	/*
754 	 * Maximum size of an extent. BTRFS_MAX_EXTENT_SIZE on regular
755 	 * filesystem, on zoned it depends on the device constraints.
756 	 */
757 	u64 max_extent_size;
758 
759 	/* Block groups and devices containing active swapfiles. */
760 	spinlock_t swapfile_pins_lock;
761 	struct rb_root swapfile_pins;
762 
763 	struct crypto_shash *csum_shash;
764 
765 	/* Type of exclusive operation running, protected by super_lock */
766 	enum btrfs_exclusive_operation exclusive_operation;
767 
768 	/*
769 	 * Zone size > 0 when in ZONED mode, otherwise it's used for a check
770 	 * if the mode is enabled
771 	 */
772 	u64 zone_size;
773 
774 	/* Constraints for ZONE_APPEND commands: */
775 	struct queue_limits limits;
776 	u64 max_zone_append_size;
777 
778 	struct mutex zoned_meta_io_lock;
779 	spinlock_t treelog_bg_lock;
780 	u64 treelog_bg;
781 
782 	/*
783 	 * Start of the dedicated data relocation block group, protected by
784 	 * relocation_bg_lock.
785 	 */
786 	spinlock_t relocation_bg_lock;
787 	u64 data_reloc_bg;
788 	struct mutex zoned_data_reloc_io_lock;
789 
790 	struct btrfs_block_group *active_meta_bg;
791 	struct btrfs_block_group *active_system_bg;
792 
793 	u64 nr_global_roots;
794 
795 	spinlock_t zone_active_bgs_lock;
796 	struct list_head zone_active_bgs;
797 
798 	/* Updates are not protected by any lock */
799 	struct btrfs_commit_stats commit_stats;
800 
801 	/*
802 	 * Last generation where we dropped a non-relocation root.
803 	 * Use btrfs_set_last_root_drop_gen() and btrfs_get_last_root_drop_gen()
804 	 * to change it and to read it, respectively.
805 	 */
806 	u64 last_root_drop_gen;
807 
808 	/*
809 	 * Annotations for transaction events (structures are empty when
810 	 * compiled without lockdep).
811 	 */
812 	struct lockdep_map btrfs_trans_num_writers_map;
813 	struct lockdep_map btrfs_trans_num_extwriters_map;
814 	struct lockdep_map btrfs_state_change_map[4];
815 	struct lockdep_map btrfs_trans_pending_ordered_map;
816 	struct lockdep_map btrfs_ordered_extent_map;
817 
818 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
819 	spinlock_t ref_verify_lock;
820 	struct rb_root block_tree;
821 #endif
822 
823 #ifdef CONFIG_BTRFS_DEBUG
824 	struct kobject *debug_kobj;
825 	struct list_head allocated_roots;
826 
827 	spinlock_t eb_leak_lock;
828 	struct list_head allocated_ebs;
829 #endif
830 };
831 
832 static inline u64 btrfs_get_fs_generation(const struct btrfs_fs_info *fs_info)
833 {
834 	return READ_ONCE(fs_info->generation);
835 }
836 
837 static inline void btrfs_set_fs_generation(struct btrfs_fs_info *fs_info, u64 gen)
838 {
839 	WRITE_ONCE(fs_info->generation, gen);
840 }
841 
842 static inline u64 btrfs_get_last_trans_committed(const struct btrfs_fs_info *fs_info)
843 {
844 	return READ_ONCE(fs_info->last_trans_committed);
845 }
846 
847 static inline void btrfs_set_last_trans_committed(struct btrfs_fs_info *fs_info, u64 gen)
848 {
849 	WRITE_ONCE(fs_info->last_trans_committed, gen);
850 }
851 
852 static inline void btrfs_set_last_root_drop_gen(struct btrfs_fs_info *fs_info,
853 						u64 gen)
854 {
855 	WRITE_ONCE(fs_info->last_root_drop_gen, gen);
856 }
857 
858 static inline u64 btrfs_get_last_root_drop_gen(const struct btrfs_fs_info *fs_info)
859 {
860 	return READ_ONCE(fs_info->last_root_drop_gen);
861 }
862 
863 /*
864  * Take the number of bytes to be checksummed and figure out how many leaves
865  * it would require to store the csums for that many bytes.
866  */
867 static inline u64 btrfs_csum_bytes_to_leaves(
868 			const struct btrfs_fs_info *fs_info, u64 csum_bytes)
869 {
870 	const u64 num_csums = csum_bytes >> fs_info->sectorsize_bits;
871 
872 	return DIV_ROUND_UP_ULL(num_csums, fs_info->csums_per_leaf);
873 }
874 
875 /*
876  * Use this if we would be adding new items, as we could split nodes as we cow
877  * down the tree.
878  */
879 static inline u64 btrfs_calc_insert_metadata_size(const struct btrfs_fs_info *fs_info,
880 						  unsigned num_items)
881 {
882 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * 2 * num_items;
883 }
884 
885 /*
886  * Doing a truncate or a modification won't result in new nodes or leaves, just
887  * what we need for COW.
888  */
889 static inline u64 btrfs_calc_metadata_size(const struct btrfs_fs_info *fs_info,
890 						 unsigned num_items)
891 {
892 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * num_items;
893 }
894 
895 #define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r->fs_info) >> 4) - \
896 					sizeof(struct btrfs_item))
897 
898 static inline bool btrfs_is_zoned(const struct btrfs_fs_info *fs_info)
899 {
900 	return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && fs_info->zone_size > 0;
901 }
902 
903 /*
904  * Count how many fs_info->max_extent_size cover the @size
905  */
906 static inline u32 count_max_extents(struct btrfs_fs_info *fs_info, u64 size)
907 {
908 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
909 	if (!fs_info)
910 		return div_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, BTRFS_MAX_EXTENT_SIZE);
911 #endif
912 
913 	return div_u64(size + fs_info->max_extent_size - 1, fs_info->max_extent_size);
914 }
915 
916 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
917 			enum btrfs_exclusive_operation type);
918 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
919 				 enum btrfs_exclusive_operation type);
920 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info);
921 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info);
922 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
923 			  enum btrfs_exclusive_operation op);
924 
925 /* Compatibility and incompatibility defines */
926 void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
927 			     const char *name);
928 void __btrfs_clear_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
929 			       const char *name);
930 void __btrfs_set_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
931 			      const char *name);
932 void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
933 				const char *name);
934 
935 #define __btrfs_fs_incompat(fs_info, flags)				\
936 	(!!(btrfs_super_incompat_flags((fs_info)->super_copy) & (flags)))
937 
938 #define __btrfs_fs_compat_ro(fs_info, flags)				\
939 	(!!(btrfs_super_compat_ro_flags((fs_info)->super_copy) & (flags)))
940 
941 #define btrfs_set_fs_incompat(__fs_info, opt)				\
942 	__btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
943 
944 #define btrfs_clear_fs_incompat(__fs_info, opt)				\
945 	__btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
946 
947 #define btrfs_fs_incompat(fs_info, opt)					\
948 	__btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
949 
950 #define btrfs_set_fs_compat_ro(__fs_info, opt)				\
951 	__btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
952 
953 #define btrfs_clear_fs_compat_ro(__fs_info, opt)			\
954 	__btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
955 
956 #define btrfs_fs_compat_ro(fs_info, opt)				\
957 	__btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
958 
959 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
960 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
961 #define btrfs_raw_test_opt(o, opt)	((o) & BTRFS_MOUNT_##opt)
962 #define btrfs_test_opt(fs_info, opt)	((fs_info)->mount_opt & \
963 					 BTRFS_MOUNT_##opt)
964 
965 static inline int btrfs_fs_closing(struct btrfs_fs_info *fs_info)
966 {
967 	/* Do it this way so we only ever do one test_bit in the normal case. */
968 	if (test_bit(BTRFS_FS_CLOSING_START, &fs_info->flags)) {
969 		if (test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags))
970 			return 2;
971 		return 1;
972 	}
973 	return 0;
974 }
975 
976 /*
977  * If we remount the fs to be R/O or umount the fs, the cleaner needn't do
978  * anything except sleeping. This function is used to check the status of
979  * the fs.
980  * We check for BTRFS_FS_STATE_RO to avoid races with a concurrent remount,
981  * since setting and checking for SB_RDONLY in the superblock's flags is not
982  * atomic.
983  */
984 static inline int btrfs_need_cleaner_sleep(struct btrfs_fs_info *fs_info)
985 {
986 	return test_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state) ||
987 		btrfs_fs_closing(fs_info);
988 }
989 
990 static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info *fs_info)
991 {
992 	clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags);
993 }
994 
995 #define BTRFS_FS_ERROR(fs_info)	(READ_ONCE((fs_info)->fs_error))
996 
997 #define BTRFS_FS_LOG_CLEANUP_ERROR(fs_info)				\
998 	(unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR,		\
999 			   &(fs_info)->fs_state)))
1000 
1001 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1002 
1003 #define EXPORT_FOR_TESTS
1004 
1005 static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
1006 {
1007 	return test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
1008 }
1009 
1010 void btrfs_test_destroy_inode(struct inode *inode);
1011 
1012 #else
1013 
1014 #define EXPORT_FOR_TESTS static
1015 
1016 static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
1017 {
1018 	return 0;
1019 }
1020 #endif
1021 
1022 #endif
1023