xref: /openbmc/linux/fs/btrfs/fs.h (revision c0d3b831)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 
3 #ifndef BTRFS_FS_H
4 #define BTRFS_FS_H
5 
6 #include <linux/fs.h>
7 #include <linux/btrfs_tree.h>
8 #include <linux/sizes.h>
9 #include "extent-io-tree.h"
10 #include "extent_map.h"
11 #include "async-thread.h"
12 #include "block-rsv.h"
13 
14 #define BTRFS_MAX_EXTENT_SIZE SZ_128M
15 
16 #define BTRFS_OLDEST_GENERATION	0ULL
17 
18 #define BTRFS_EMPTY_DIR_SIZE 0
19 
20 #define BTRFS_DIRTY_METADATA_THRESH		SZ_32M
21 
22 #define BTRFS_SUPER_INFO_OFFSET			SZ_64K
23 #define BTRFS_SUPER_INFO_SIZE			4096
24 static_assert(sizeof(struct btrfs_super_block) == BTRFS_SUPER_INFO_SIZE);
25 
26 /*
27  * The reserved space at the beginning of each device.  It covers the primary
28  * super block and leaves space for potential use by other tools like
29  * bootloaders or to lower potential damage of accidental overwrite.
30  */
31 #define BTRFS_DEVICE_RANGE_RESERVED			(SZ_1M)
32 /*
33  * Runtime (in-memory) states of filesystem
34  */
35 enum {
36 	/* Global indicator of serious filesystem errors */
37 	BTRFS_FS_STATE_ERROR,
38 	/*
39 	 * Filesystem is being remounted, allow to skip some operations, like
40 	 * defrag
41 	 */
42 	BTRFS_FS_STATE_REMOUNTING,
43 	/* Filesystem in RO mode */
44 	BTRFS_FS_STATE_RO,
45 	/* Track if a transaction abort has been reported on this filesystem */
46 	BTRFS_FS_STATE_TRANS_ABORTED,
47 	/*
48 	 * Bio operations should be blocked on this filesystem because a source
49 	 * or target device is being destroyed as part of a device replace
50 	 */
51 	BTRFS_FS_STATE_DEV_REPLACING,
52 	/* The btrfs_fs_info created for self-tests */
53 	BTRFS_FS_STATE_DUMMY_FS_INFO,
54 
55 	BTRFS_FS_STATE_NO_CSUMS,
56 
57 	/* Indicates there was an error cleaning up a log tree. */
58 	BTRFS_FS_STATE_LOG_CLEANUP_ERROR,
59 
60 	BTRFS_FS_STATE_COUNT
61 };
62 
63 enum {
64 	BTRFS_FS_CLOSING_START,
65 	BTRFS_FS_CLOSING_DONE,
66 	BTRFS_FS_LOG_RECOVERING,
67 	BTRFS_FS_OPEN,
68 	BTRFS_FS_QUOTA_ENABLED,
69 	BTRFS_FS_UPDATE_UUID_TREE_GEN,
70 	BTRFS_FS_CREATING_FREE_SPACE_TREE,
71 	BTRFS_FS_BTREE_ERR,
72 	BTRFS_FS_LOG1_ERR,
73 	BTRFS_FS_LOG2_ERR,
74 	BTRFS_FS_QUOTA_OVERRIDE,
75 	/* Used to record internally whether fs has been frozen */
76 	BTRFS_FS_FROZEN,
77 	/*
78 	 * Indicate that balance has been set up from the ioctl and is in the
79 	 * main phase. The fs_info::balance_ctl is initialized.
80 	 */
81 	BTRFS_FS_BALANCE_RUNNING,
82 
83 	/*
84 	 * Indicate that relocation of a chunk has started, it's set per chunk
85 	 * and is toggled between chunks.
86 	 */
87 	BTRFS_FS_RELOC_RUNNING,
88 
89 	/* Indicate that the cleaner thread is awake and doing something. */
90 	BTRFS_FS_CLEANER_RUNNING,
91 
92 	/*
93 	 * The checksumming has an optimized version and is considered fast,
94 	 * so we don't need to offload checksums to workqueues.
95 	 */
96 	BTRFS_FS_CSUM_IMPL_FAST,
97 
98 	/* Indicate that the discard workqueue can service discards. */
99 	BTRFS_FS_DISCARD_RUNNING,
100 
101 	/* Indicate that we need to cleanup space cache v1 */
102 	BTRFS_FS_CLEANUP_SPACE_CACHE_V1,
103 
104 	/* Indicate that we can't trust the free space tree for caching yet */
105 	BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED,
106 
107 	/* Indicate whether there are any tree modification log users */
108 	BTRFS_FS_TREE_MOD_LOG_USERS,
109 
110 	/* Indicate that we want the transaction kthread to commit right now. */
111 	BTRFS_FS_COMMIT_TRANS,
112 
113 	/* Indicate we have half completed snapshot deletions pending. */
114 	BTRFS_FS_UNFINISHED_DROPS,
115 
116 	/* Indicate we have to finish a zone to do next allocation. */
117 	BTRFS_FS_NEED_ZONE_FINISH,
118 
119 	/* Indicate that we want to commit the transaction. */
120 	BTRFS_FS_NEED_TRANS_COMMIT,
121 
122 	/*
123 	 * Indicate metadata over-commit is disabled. This is set when active
124 	 * zone tracking is needed.
125 	 */
126 	BTRFS_FS_NO_OVERCOMMIT,
127 
128 #if BITS_PER_LONG == 32
129 	/* Indicate if we have error/warn message printed on 32bit systems */
130 	BTRFS_FS_32BIT_ERROR,
131 	BTRFS_FS_32BIT_WARN,
132 #endif
133 };
134 
135 /*
136  * Flags for mount options.
137  *
138  * Note: don't forget to add new options to btrfs_show_options()
139  */
140 enum {
141 	BTRFS_MOUNT_NODATASUM			= (1UL << 0),
142 	BTRFS_MOUNT_NODATACOW			= (1UL << 1),
143 	BTRFS_MOUNT_NOBARRIER			= (1UL << 2),
144 	BTRFS_MOUNT_SSD				= (1UL << 3),
145 	BTRFS_MOUNT_DEGRADED			= (1UL << 4),
146 	BTRFS_MOUNT_COMPRESS			= (1UL << 5),
147 	BTRFS_MOUNT_NOTREELOG   		= (1UL << 6),
148 	BTRFS_MOUNT_FLUSHONCOMMIT		= (1UL << 7),
149 	BTRFS_MOUNT_SSD_SPREAD			= (1UL << 8),
150 	BTRFS_MOUNT_NOSSD			= (1UL << 9),
151 	BTRFS_MOUNT_DISCARD_SYNC		= (1UL << 10),
152 	BTRFS_MOUNT_FORCE_COMPRESS      	= (1UL << 11),
153 	BTRFS_MOUNT_SPACE_CACHE			= (1UL << 12),
154 	BTRFS_MOUNT_CLEAR_CACHE			= (1UL << 13),
155 	BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED	= (1UL << 14),
156 	BTRFS_MOUNT_ENOSPC_DEBUG		= (1UL << 15),
157 	BTRFS_MOUNT_AUTO_DEFRAG			= (1UL << 16),
158 	BTRFS_MOUNT_USEBACKUPROOT		= (1UL << 17),
159 	BTRFS_MOUNT_SKIP_BALANCE		= (1UL << 18),
160 	BTRFS_MOUNT_CHECK_INTEGRITY		= (1UL << 19),
161 	BTRFS_MOUNT_CHECK_INTEGRITY_DATA	= (1UL << 20),
162 	BTRFS_MOUNT_PANIC_ON_FATAL_ERROR	= (1UL << 21),
163 	BTRFS_MOUNT_RESCAN_UUID_TREE		= (1UL << 22),
164 	BTRFS_MOUNT_FRAGMENT_DATA		= (1UL << 23),
165 	BTRFS_MOUNT_FRAGMENT_METADATA		= (1UL << 24),
166 	BTRFS_MOUNT_FREE_SPACE_TREE		= (1UL << 25),
167 	BTRFS_MOUNT_NOLOGREPLAY			= (1UL << 26),
168 	BTRFS_MOUNT_REF_VERIFY			= (1UL << 27),
169 	BTRFS_MOUNT_DISCARD_ASYNC		= (1UL << 28),
170 	BTRFS_MOUNT_IGNOREBADROOTS		= (1UL << 29),
171 	BTRFS_MOUNT_IGNOREDATACSUMS		= (1UL << 30),
172 	BTRFS_MOUNT_NODISCARD			= (1UL << 31),
173 };
174 
175 /*
176  * Compat flags that we support.  If any incompat flags are set other than the
177  * ones specified below then we will fail to mount
178  */
179 #define BTRFS_FEATURE_COMPAT_SUPP		0ULL
180 #define BTRFS_FEATURE_COMPAT_SAFE_SET		0ULL
181 #define BTRFS_FEATURE_COMPAT_SAFE_CLEAR		0ULL
182 
183 #define BTRFS_FEATURE_COMPAT_RO_SUPP			\
184 	(BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE |	\
185 	 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID | \
186 	 BTRFS_FEATURE_COMPAT_RO_VERITY |		\
187 	 BTRFS_FEATURE_COMPAT_RO_BLOCK_GROUP_TREE)
188 
189 #define BTRFS_FEATURE_COMPAT_RO_SAFE_SET	0ULL
190 #define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR	0ULL
191 
192 #ifdef CONFIG_BTRFS_DEBUG
193 /*
194  * Extent tree v2 supported only with CONFIG_BTRFS_DEBUG
195  */
196 #define BTRFS_FEATURE_INCOMPAT_SUPP			\
197 	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
198 	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
199 	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
200 	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
201 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
202 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD |		\
203 	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
204 	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF |		\
205 	 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA |	\
206 	 BTRFS_FEATURE_INCOMPAT_NO_HOLES	|	\
207 	 BTRFS_FEATURE_INCOMPAT_METADATA_UUID	|	\
208 	 BTRFS_FEATURE_INCOMPAT_RAID1C34	|	\
209 	 BTRFS_FEATURE_INCOMPAT_ZONED		|	\
210 	 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2)
211 #else
212 #define BTRFS_FEATURE_INCOMPAT_SUPP			\
213 	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
214 	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
215 	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
216 	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
217 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
218 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD |		\
219 	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
220 	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF |		\
221 	 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA |	\
222 	 BTRFS_FEATURE_INCOMPAT_NO_HOLES	|	\
223 	 BTRFS_FEATURE_INCOMPAT_METADATA_UUID	|	\
224 	 BTRFS_FEATURE_INCOMPAT_RAID1C34	|	\
225 	 BTRFS_FEATURE_INCOMPAT_ZONED)
226 #endif
227 
228 #define BTRFS_FEATURE_INCOMPAT_SAFE_SET			\
229 	(BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
230 #define BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR		0ULL
231 
232 #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
233 #define BTRFS_DEFAULT_MAX_INLINE	(2048)
234 
235 struct btrfs_dev_replace {
236 	/* See #define above */
237 	u64 replace_state;
238 	/* Seconds since 1-Jan-1970 */
239 	time64_t time_started;
240 	/* Seconds since 1-Jan-1970 */
241 	time64_t time_stopped;
242 	atomic64_t num_write_errors;
243 	atomic64_t num_uncorrectable_read_errors;
244 
245 	u64 cursor_left;
246 	u64 committed_cursor_left;
247 	u64 cursor_left_last_write_of_item;
248 	u64 cursor_right;
249 
250 	/* See #define above */
251 	u64 cont_reading_from_srcdev_mode;
252 
253 	int is_valid;
254 	int item_needs_writeback;
255 	struct btrfs_device *srcdev;
256 	struct btrfs_device *tgtdev;
257 
258 	struct mutex lock_finishing_cancel_unmount;
259 	struct rw_semaphore rwsem;
260 
261 	struct btrfs_scrub_progress scrub_progress;
262 
263 	struct percpu_counter bio_counter;
264 	wait_queue_head_t replace_wait;
265 };
266 
267 /*
268  * Free clusters are used to claim free space in relatively large chunks,
269  * allowing us to do less seeky writes. They are used for all metadata
270  * allocations. In ssd_spread mode they are also used for data allocations.
271  */
272 struct btrfs_free_cluster {
273 	spinlock_t lock;
274 	spinlock_t refill_lock;
275 	struct rb_root root;
276 
277 	/* Largest extent in this cluster */
278 	u64 max_size;
279 
280 	/* First extent starting offset */
281 	u64 window_start;
282 
283 	/* We did a full search and couldn't create a cluster */
284 	bool fragmented;
285 
286 	struct btrfs_block_group *block_group;
287 	/*
288 	 * When a cluster is allocated from a block group, we put the cluster
289 	 * onto a list in the block group so that it can be freed before the
290 	 * block group is freed.
291 	 */
292 	struct list_head block_group_list;
293 };
294 
295 /* Discard control. */
296 /*
297  * Async discard uses multiple lists to differentiate the discard filter
298  * parameters.  Index 0 is for completely free block groups where we need to
299  * ensure the entire block group is trimmed without being lossy.  Indices
300  * afterwards represent monotonically decreasing discard filter sizes to
301  * prioritize what should be discarded next.
302  */
303 #define BTRFS_NR_DISCARD_LISTS		3
304 #define BTRFS_DISCARD_INDEX_UNUSED	0
305 #define BTRFS_DISCARD_INDEX_START	1
306 
307 struct btrfs_discard_ctl {
308 	struct workqueue_struct *discard_workers;
309 	struct delayed_work work;
310 	spinlock_t lock;
311 	struct btrfs_block_group *block_group;
312 	struct list_head discard_list[BTRFS_NR_DISCARD_LISTS];
313 	u64 prev_discard;
314 	u64 prev_discard_time;
315 	atomic_t discardable_extents;
316 	atomic64_t discardable_bytes;
317 	u64 max_discard_size;
318 	u64 delay_ms;
319 	u32 iops_limit;
320 	u32 kbps_limit;
321 	u64 discard_extent_bytes;
322 	u64 discard_bitmap_bytes;
323 	atomic64_t discard_bytes_saved;
324 };
325 
326 /*
327  * Exclusive operations (device replace, resize, device add/remove, balance)
328  */
329 enum btrfs_exclusive_operation {
330 	BTRFS_EXCLOP_NONE,
331 	BTRFS_EXCLOP_BALANCE_PAUSED,
332 	BTRFS_EXCLOP_BALANCE,
333 	BTRFS_EXCLOP_DEV_ADD,
334 	BTRFS_EXCLOP_DEV_REMOVE,
335 	BTRFS_EXCLOP_DEV_REPLACE,
336 	BTRFS_EXCLOP_RESIZE,
337 	BTRFS_EXCLOP_SWAP_ACTIVATE,
338 };
339 
340 /* Store data about transaction commits, exported via sysfs. */
341 struct btrfs_commit_stats {
342 	/* Total number of commits */
343 	u64 commit_count;
344 	/* The maximum commit duration so far in ns */
345 	u64 max_commit_dur;
346 	/* The last commit duration in ns */
347 	u64 last_commit_dur;
348 	/* The total commit duration in ns */
349 	u64 total_commit_dur;
350 };
351 
352 struct btrfs_fs_info {
353 	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
354 	unsigned long flags;
355 	struct btrfs_root *tree_root;
356 	struct btrfs_root *chunk_root;
357 	struct btrfs_root *dev_root;
358 	struct btrfs_root *fs_root;
359 	struct btrfs_root *quota_root;
360 	struct btrfs_root *uuid_root;
361 	struct btrfs_root *data_reloc_root;
362 	struct btrfs_root *block_group_root;
363 
364 	/* The log root tree is a directory of all the other log roots */
365 	struct btrfs_root *log_root_tree;
366 
367 	/* The tree that holds the global roots (csum, extent, etc) */
368 	rwlock_t global_root_lock;
369 	struct rb_root global_root_tree;
370 
371 	spinlock_t fs_roots_radix_lock;
372 	struct radix_tree_root fs_roots_radix;
373 
374 	/* Block group cache stuff */
375 	rwlock_t block_group_cache_lock;
376 	struct rb_root_cached block_group_cache_tree;
377 
378 	/* Keep track of unallocated space */
379 	atomic64_t free_chunk_space;
380 
381 	/* Track ranges which are used by log trees blocks/logged data extents */
382 	struct extent_io_tree excluded_extents;
383 
384 	/* logical->physical extent mapping */
385 	struct extent_map_tree mapping_tree;
386 
387 	/*
388 	 * Block reservation for extent, checksum, root tree and delayed dir
389 	 * index item.
390 	 */
391 	struct btrfs_block_rsv global_block_rsv;
392 	/* Block reservation for metadata operations */
393 	struct btrfs_block_rsv trans_block_rsv;
394 	/* Block reservation for chunk tree */
395 	struct btrfs_block_rsv chunk_block_rsv;
396 	/* Block reservation for delayed operations */
397 	struct btrfs_block_rsv delayed_block_rsv;
398 	/* Block reservation for delayed refs */
399 	struct btrfs_block_rsv delayed_refs_rsv;
400 
401 	struct btrfs_block_rsv empty_block_rsv;
402 
403 	u64 generation;
404 	u64 last_trans_committed;
405 	/*
406 	 * Generation of the last transaction used for block group relocation
407 	 * since the filesystem was last mounted (or 0 if none happened yet).
408 	 * Must be written and read while holding btrfs_fs_info::commit_root_sem.
409 	 */
410 	u64 last_reloc_trans;
411 	u64 avg_delayed_ref_runtime;
412 
413 	/*
414 	 * This is updated to the current trans every time a full commit is
415 	 * required instead of the faster short fsync log commits
416 	 */
417 	u64 last_trans_log_full_commit;
418 	unsigned long mount_opt;
419 
420 	unsigned long compress_type:4;
421 	unsigned int compress_level;
422 	u32 commit_interval;
423 	/*
424 	 * It is a suggestive number, the read side is safe even it gets a
425 	 * wrong number because we will write out the data into a regular
426 	 * extent. The write side(mount/remount) is under ->s_umount lock,
427 	 * so it is also safe.
428 	 */
429 	u64 max_inline;
430 
431 	struct btrfs_transaction *running_transaction;
432 	wait_queue_head_t transaction_throttle;
433 	wait_queue_head_t transaction_wait;
434 	wait_queue_head_t transaction_blocked_wait;
435 	wait_queue_head_t async_submit_wait;
436 
437 	/*
438 	 * Used to protect the incompat_flags, compat_flags, compat_ro_flags
439 	 * when they are updated.
440 	 *
441 	 * Because we do not clear the flags for ever, so we needn't use
442 	 * the lock on the read side.
443 	 *
444 	 * We also needn't use the lock when we mount the fs, because
445 	 * there is no other task which will update the flag.
446 	 */
447 	spinlock_t super_lock;
448 	struct btrfs_super_block *super_copy;
449 	struct btrfs_super_block *super_for_commit;
450 	struct super_block *sb;
451 	struct inode *btree_inode;
452 	struct mutex tree_log_mutex;
453 	struct mutex transaction_kthread_mutex;
454 	struct mutex cleaner_mutex;
455 	struct mutex chunk_mutex;
456 
457 	/*
458 	 * This is taken to make sure we don't set block groups ro after the
459 	 * free space cache has been allocated on them.
460 	 */
461 	struct mutex ro_block_group_mutex;
462 
463 	/*
464 	 * This is used during read/modify/write to make sure no two ios are
465 	 * trying to mod the same stripe at the same time.
466 	 */
467 	struct btrfs_stripe_hash_table *stripe_hash_table;
468 
469 	/*
470 	 * This protects the ordered operations list only while we are
471 	 * processing all of the entries on it.  This way we make sure the
472 	 * commit code doesn't find the list temporarily empty because another
473 	 * function happens to be doing non-waiting preflush before jumping
474 	 * into the main commit.
475 	 */
476 	struct mutex ordered_operations_mutex;
477 
478 	struct rw_semaphore commit_root_sem;
479 
480 	struct rw_semaphore cleanup_work_sem;
481 
482 	struct rw_semaphore subvol_sem;
483 
484 	spinlock_t trans_lock;
485 	/*
486 	 * The reloc mutex goes with the trans lock, it is taken during commit
487 	 * to protect us from the relocation code.
488 	 */
489 	struct mutex reloc_mutex;
490 
491 	struct list_head trans_list;
492 	struct list_head dead_roots;
493 	struct list_head caching_block_groups;
494 
495 	spinlock_t delayed_iput_lock;
496 	struct list_head delayed_iputs;
497 	atomic_t nr_delayed_iputs;
498 	wait_queue_head_t delayed_iputs_wait;
499 
500 	atomic64_t tree_mod_seq;
501 
502 	/* This protects tree_mod_log and tree_mod_seq_list */
503 	rwlock_t tree_mod_log_lock;
504 	struct rb_root tree_mod_log;
505 	struct list_head tree_mod_seq_list;
506 
507 	atomic_t async_delalloc_pages;
508 
509 	/* This is used to protect the following list -- ordered_roots. */
510 	spinlock_t ordered_root_lock;
511 
512 	/*
513 	 * All fs/file tree roots in which there are data=ordered extents
514 	 * pending writeback are added into this list.
515 	 *
516 	 * These can span multiple transactions and basically include every
517 	 * dirty data page that isn't from nodatacow.
518 	 */
519 	struct list_head ordered_roots;
520 
521 	struct mutex delalloc_root_mutex;
522 	spinlock_t delalloc_root_lock;
523 	/* All fs/file tree roots that have delalloc inodes. */
524 	struct list_head delalloc_roots;
525 
526 	/*
527 	 * There is a pool of worker threads for checksumming during writes and
528 	 * a pool for checksumming after reads.  This is because readers can
529 	 * run with FS locks held, and the writers may be waiting for those
530 	 * locks.  We don't want ordering in the pending list to cause
531 	 * deadlocks, and so the two are serviced separately.
532 	 *
533 	 * A third pool does submit_bio to avoid deadlocking with the other two.
534 	 */
535 	struct btrfs_workqueue *workers;
536 	struct btrfs_workqueue *hipri_workers;
537 	struct btrfs_workqueue *delalloc_workers;
538 	struct btrfs_workqueue *flush_workers;
539 	struct workqueue_struct *endio_workers;
540 	struct workqueue_struct *endio_meta_workers;
541 	struct workqueue_struct *rmw_workers;
542 	struct workqueue_struct *compressed_write_workers;
543 	struct btrfs_workqueue *endio_write_workers;
544 	struct btrfs_workqueue *endio_freespace_worker;
545 	struct btrfs_workqueue *caching_workers;
546 
547 	/*
548 	 * Fixup workers take dirty pages that didn't properly go through the
549 	 * cow mechanism and make them safe to write.  It happens for the
550 	 * sys_munmap function call path.
551 	 */
552 	struct btrfs_workqueue *fixup_workers;
553 	struct btrfs_workqueue *delayed_workers;
554 
555 	struct task_struct *transaction_kthread;
556 	struct task_struct *cleaner_kthread;
557 	u32 thread_pool_size;
558 
559 	struct kobject *space_info_kobj;
560 	struct kobject *qgroups_kobj;
561 	struct kobject *discard_kobj;
562 
563 	/* Used to keep from writing metadata until there is a nice batch */
564 	struct percpu_counter dirty_metadata_bytes;
565 	struct percpu_counter delalloc_bytes;
566 	struct percpu_counter ordered_bytes;
567 	s32 dirty_metadata_batch;
568 	s32 delalloc_batch;
569 
570 	struct list_head dirty_cowonly_roots;
571 
572 	struct btrfs_fs_devices *fs_devices;
573 
574 	/*
575 	 * The space_info list is effectively read only after initial setup.
576 	 * It is populated at mount time and cleaned up after all block groups
577 	 * are removed.  RCU is used to protect it.
578 	 */
579 	struct list_head space_info;
580 
581 	struct btrfs_space_info *data_sinfo;
582 
583 	struct reloc_control *reloc_ctl;
584 
585 	/* data_alloc_cluster is only used in ssd_spread mode */
586 	struct btrfs_free_cluster data_alloc_cluster;
587 
588 	/* All metadata allocations go through this cluster. */
589 	struct btrfs_free_cluster meta_alloc_cluster;
590 
591 	/* Auto defrag inodes go here. */
592 	spinlock_t defrag_inodes_lock;
593 	struct rb_root defrag_inodes;
594 	atomic_t defrag_running;
595 
596 	/* Used to protect avail_{data, metadata, system}_alloc_bits */
597 	seqlock_t profiles_lock;
598 	/*
599 	 * These three are in extended format (availability of single chunks is
600 	 * denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other types are denoted
601 	 * by corresponding BTRFS_BLOCK_GROUP_* bits)
602 	 */
603 	u64 avail_data_alloc_bits;
604 	u64 avail_metadata_alloc_bits;
605 	u64 avail_system_alloc_bits;
606 
607 	/* Balance state */
608 	spinlock_t balance_lock;
609 	struct mutex balance_mutex;
610 	atomic_t balance_pause_req;
611 	atomic_t balance_cancel_req;
612 	struct btrfs_balance_control *balance_ctl;
613 	wait_queue_head_t balance_wait_q;
614 
615 	/* Cancellation requests for chunk relocation */
616 	atomic_t reloc_cancel_req;
617 
618 	u32 data_chunk_allocations;
619 	u32 metadata_ratio;
620 
621 	void *bdev_holder;
622 
623 	/* Private scrub information */
624 	struct mutex scrub_lock;
625 	atomic_t scrubs_running;
626 	atomic_t scrub_pause_req;
627 	atomic_t scrubs_paused;
628 	atomic_t scrub_cancel_req;
629 	wait_queue_head_t scrub_pause_wait;
630 	/*
631 	 * The worker pointers are NULL iff the refcount is 0, ie. scrub is not
632 	 * running.
633 	 */
634 	refcount_t scrub_workers_refcnt;
635 	struct workqueue_struct *scrub_workers;
636 	struct workqueue_struct *scrub_wr_completion_workers;
637 	struct workqueue_struct *scrub_parity_workers;
638 	struct btrfs_subpage_info *subpage_info;
639 
640 	struct btrfs_discard_ctl discard_ctl;
641 
642 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
643 	u32 check_integrity_print_mask;
644 #endif
645 	/* Is qgroup tracking in a consistent state? */
646 	u64 qgroup_flags;
647 
648 	/* Holds configuration and tracking. Protected by qgroup_lock. */
649 	struct rb_root qgroup_tree;
650 	spinlock_t qgroup_lock;
651 
652 	/*
653 	 * Used to avoid frequently calling ulist_alloc()/ulist_free()
654 	 * when doing qgroup accounting, it must be protected by qgroup_lock.
655 	 */
656 	struct ulist *qgroup_ulist;
657 
658 	/*
659 	 * Protect user change for quota operations. If a transaction is needed,
660 	 * it must be started before locking this lock.
661 	 */
662 	struct mutex qgroup_ioctl_lock;
663 
664 	/* List of dirty qgroups to be written at next commit. */
665 	struct list_head dirty_qgroups;
666 
667 	/* Used by qgroup for an efficient tree traversal. */
668 	u64 qgroup_seq;
669 
670 	/* Qgroup rescan items. */
671 	/* Protects the progress item */
672 	struct mutex qgroup_rescan_lock;
673 	struct btrfs_key qgroup_rescan_progress;
674 	struct btrfs_workqueue *qgroup_rescan_workers;
675 	struct completion qgroup_rescan_completion;
676 	struct btrfs_work qgroup_rescan_work;
677 	/* Protected by qgroup_rescan_lock */
678 	bool qgroup_rescan_running;
679 	u8 qgroup_drop_subtree_thres;
680 
681 	/* Filesystem state */
682 	unsigned long fs_state;
683 
684 	struct btrfs_delayed_root *delayed_root;
685 
686 	/* Extent buffer radix tree */
687 	spinlock_t buffer_lock;
688 	/* Entries are eb->start / sectorsize */
689 	struct radix_tree_root buffer_radix;
690 
691 	/* Next backup root to be overwritten */
692 	int backup_root_index;
693 
694 	/* Device replace state */
695 	struct btrfs_dev_replace dev_replace;
696 
697 	struct semaphore uuid_tree_rescan_sem;
698 
699 	/* Used to reclaim the metadata space in the background. */
700 	struct work_struct async_reclaim_work;
701 	struct work_struct async_data_reclaim_work;
702 	struct work_struct preempt_reclaim_work;
703 
704 	/* Reclaim partially filled block groups in the background */
705 	struct work_struct reclaim_bgs_work;
706 	struct list_head reclaim_bgs;
707 	int bg_reclaim_threshold;
708 
709 	spinlock_t unused_bgs_lock;
710 	struct list_head unused_bgs;
711 	struct mutex unused_bg_unpin_mutex;
712 	/* Protect block groups that are going to be deleted */
713 	struct mutex reclaim_bgs_lock;
714 
715 	/* Cached block sizes */
716 	u32 nodesize;
717 	u32 sectorsize;
718 	/* ilog2 of sectorsize, use to avoid 64bit division */
719 	u32 sectorsize_bits;
720 	u32 csum_size;
721 	u32 csums_per_leaf;
722 	u32 stripesize;
723 
724 	/*
725 	 * Maximum size of an extent. BTRFS_MAX_EXTENT_SIZE on regular
726 	 * filesystem, on zoned it depends on the device constraints.
727 	 */
728 	u64 max_extent_size;
729 
730 	/* Block groups and devices containing active swapfiles. */
731 	spinlock_t swapfile_pins_lock;
732 	struct rb_root swapfile_pins;
733 
734 	struct crypto_shash *csum_shash;
735 
736 	/* Type of exclusive operation running, protected by super_lock */
737 	enum btrfs_exclusive_operation exclusive_operation;
738 
739 	/*
740 	 * Zone size > 0 when in ZONED mode, otherwise it's used for a check
741 	 * if the mode is enabled
742 	 */
743 	u64 zone_size;
744 
745 	/* Max size to emit ZONE_APPEND write command */
746 	u64 max_zone_append_size;
747 	struct mutex zoned_meta_io_lock;
748 	spinlock_t treelog_bg_lock;
749 	u64 treelog_bg;
750 
751 	/*
752 	 * Start of the dedicated data relocation block group, protected by
753 	 * relocation_bg_lock.
754 	 */
755 	spinlock_t relocation_bg_lock;
756 	u64 data_reloc_bg;
757 	struct mutex zoned_data_reloc_io_lock;
758 
759 	u64 nr_global_roots;
760 
761 	spinlock_t zone_active_bgs_lock;
762 	struct list_head zone_active_bgs;
763 
764 	/* Updates are not protected by any lock */
765 	struct btrfs_commit_stats commit_stats;
766 
767 	/*
768 	 * Last generation where we dropped a non-relocation root.
769 	 * Use btrfs_set_last_root_drop_gen() and btrfs_get_last_root_drop_gen()
770 	 * to change it and to read it, respectively.
771 	 */
772 	u64 last_root_drop_gen;
773 
774 	/*
775 	 * Annotations for transaction events (structures are empty when
776 	 * compiled without lockdep).
777 	 */
778 	struct lockdep_map btrfs_trans_num_writers_map;
779 	struct lockdep_map btrfs_trans_num_extwriters_map;
780 	struct lockdep_map btrfs_state_change_map[4];
781 	struct lockdep_map btrfs_trans_pending_ordered_map;
782 	struct lockdep_map btrfs_ordered_extent_map;
783 
784 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
785 	spinlock_t ref_verify_lock;
786 	struct rb_root block_tree;
787 #endif
788 
789 #ifdef CONFIG_BTRFS_DEBUG
790 	struct kobject *debug_kobj;
791 	struct list_head allocated_roots;
792 
793 	spinlock_t eb_leak_lock;
794 	struct list_head allocated_ebs;
795 #endif
796 };
797 
798 static inline void btrfs_set_last_root_drop_gen(struct btrfs_fs_info *fs_info,
799 						u64 gen)
800 {
801 	WRITE_ONCE(fs_info->last_root_drop_gen, gen);
802 }
803 
804 static inline u64 btrfs_get_last_root_drop_gen(const struct btrfs_fs_info *fs_info)
805 {
806 	return READ_ONCE(fs_info->last_root_drop_gen);
807 }
808 
809 /*
810  * Take the number of bytes to be checksummed and figure out how many leaves
811  * it would require to store the csums for that many bytes.
812  */
813 static inline u64 btrfs_csum_bytes_to_leaves(
814 			const struct btrfs_fs_info *fs_info, u64 csum_bytes)
815 {
816 	const u64 num_csums = csum_bytes >> fs_info->sectorsize_bits;
817 
818 	return DIV_ROUND_UP_ULL(num_csums, fs_info->csums_per_leaf);
819 }
820 
821 /*
822  * Use this if we would be adding new items, as we could split nodes as we cow
823  * down the tree.
824  */
825 static inline u64 btrfs_calc_insert_metadata_size(struct btrfs_fs_info *fs_info,
826 						  unsigned num_items)
827 {
828 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * 2 * num_items;
829 }
830 
831 /*
832  * Doing a truncate or a modification won't result in new nodes or leaves, just
833  * what we need for COW.
834  */
835 static inline u64 btrfs_calc_metadata_size(struct btrfs_fs_info *fs_info,
836 						 unsigned num_items)
837 {
838 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * num_items;
839 }
840 
841 #define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r->fs_info) >> 4) - \
842 					sizeof(struct btrfs_item))
843 
844 static inline bool btrfs_is_zoned(const struct btrfs_fs_info *fs_info)
845 {
846 	return fs_info->zone_size > 0;
847 }
848 
849 /*
850  * Count how many fs_info->max_extent_size cover the @size
851  */
852 static inline u32 count_max_extents(struct btrfs_fs_info *fs_info, u64 size)
853 {
854 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
855 	if (!fs_info)
856 		return div_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, BTRFS_MAX_EXTENT_SIZE);
857 #endif
858 
859 	return div_u64(size + fs_info->max_extent_size - 1, fs_info->max_extent_size);
860 }
861 
862 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
863 			enum btrfs_exclusive_operation type);
864 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
865 				 enum btrfs_exclusive_operation type);
866 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info);
867 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info);
868 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
869 			  enum btrfs_exclusive_operation op);
870 
871 /* Compatibility and incompatibility defines */
872 void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
873 			     const char *name);
874 void __btrfs_clear_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
875 			       const char *name);
876 void __btrfs_set_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
877 			      const char *name);
878 void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
879 				const char *name);
880 
881 #define __btrfs_fs_incompat(fs_info, flags)				\
882 	(!!(btrfs_super_incompat_flags((fs_info)->super_copy) & (flags)))
883 
884 #define __btrfs_fs_compat_ro(fs_info, flags)				\
885 	(!!(btrfs_super_compat_ro_flags((fs_info)->super_copy) & (flags)))
886 
887 #define btrfs_set_fs_incompat(__fs_info, opt)				\
888 	__btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
889 
890 #define btrfs_clear_fs_incompat(__fs_info, opt)				\
891 	__btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
892 
893 #define btrfs_fs_incompat(fs_info, opt)					\
894 	__btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
895 
896 #define btrfs_set_fs_compat_ro(__fs_info, opt)				\
897 	__btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
898 
899 #define btrfs_clear_fs_compat_ro(__fs_info, opt)			\
900 	__btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
901 
902 #define btrfs_fs_compat_ro(fs_info, opt)				\
903 	__btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
904 
905 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
906 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
907 #define btrfs_raw_test_opt(o, opt)	((o) & BTRFS_MOUNT_##opt)
908 #define btrfs_test_opt(fs_info, opt)	((fs_info)->mount_opt & \
909 					 BTRFS_MOUNT_##opt)
910 
911 #define btrfs_set_and_info(fs_info, opt, fmt, args...)			\
912 do {									\
913 	if (!btrfs_test_opt(fs_info, opt))				\
914 		btrfs_info(fs_info, fmt, ##args);			\
915 	btrfs_set_opt(fs_info->mount_opt, opt);				\
916 } while (0)
917 
918 #define btrfs_clear_and_info(fs_info, opt, fmt, args...)		\
919 do {									\
920 	if (btrfs_test_opt(fs_info, opt))				\
921 		btrfs_info(fs_info, fmt, ##args);			\
922 	btrfs_clear_opt(fs_info->mount_opt, opt);			\
923 } while (0)
924 
925 static inline int btrfs_fs_closing(struct btrfs_fs_info *fs_info)
926 {
927 	/* Do it this way so we only ever do one test_bit in the normal case. */
928 	if (test_bit(BTRFS_FS_CLOSING_START, &fs_info->flags)) {
929 		if (test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags))
930 			return 2;
931 		return 1;
932 	}
933 	return 0;
934 }
935 
936 /*
937  * If we remount the fs to be R/O or umount the fs, the cleaner needn't do
938  * anything except sleeping. This function is used to check the status of
939  * the fs.
940  * We check for BTRFS_FS_STATE_RO to avoid races with a concurrent remount,
941  * since setting and checking for SB_RDONLY in the superblock's flags is not
942  * atomic.
943  */
944 static inline int btrfs_need_cleaner_sleep(struct btrfs_fs_info *fs_info)
945 {
946 	return test_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state) ||
947 		btrfs_fs_closing(fs_info);
948 }
949 
950 static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info *fs_info)
951 {
952 	clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags);
953 }
954 
955 #define BTRFS_FS_ERROR(fs_info)	(unlikely(test_bit(BTRFS_FS_STATE_ERROR, \
956 						   &(fs_info)->fs_state)))
957 #define BTRFS_FS_LOG_CLEANUP_ERROR(fs_info)				\
958 	(unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR,		\
959 			   &(fs_info)->fs_state)))
960 
961 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
962 
963 #define EXPORT_FOR_TESTS
964 
965 static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
966 {
967 	return test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
968 }
969 
970 void btrfs_test_destroy_inode(struct inode *inode);
971 
972 #else
973 
974 #define EXPORT_FOR_TESTS static
975 
976 static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
977 {
978 	return 0;
979 }
980 #endif
981 
982 #endif
983