xref: /openbmc/linux/fs/btrfs/fs.h (revision 1edd0337)
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  * The reserved space at the beginning of each device.  It covers the primary
29  * super block and leaves space for potential use by other tools like
30  * bootloaders or to lower potential damage of accidental overwrite.
31  */
32 #define BTRFS_DEVICE_RANGE_RESERVED			(SZ_1M)
33 /*
34  * Runtime (in-memory) states of filesystem
35  */
36 enum {
37 	/* Global indicator of serious filesystem errors */
38 	BTRFS_FS_STATE_ERROR,
39 	/*
40 	 * Filesystem is being remounted, allow to skip some operations, like
41 	 * defrag
42 	 */
43 	BTRFS_FS_STATE_REMOUNTING,
44 	/* Filesystem in RO mode */
45 	BTRFS_FS_STATE_RO,
46 	/* Track if a transaction abort has been reported on this filesystem */
47 	BTRFS_FS_STATE_TRANS_ABORTED,
48 	/*
49 	 * Bio operations should be blocked on this filesystem because a source
50 	 * or target device is being destroyed as part of a device replace
51 	 */
52 	BTRFS_FS_STATE_DEV_REPLACING,
53 	/* The btrfs_fs_info created for self-tests */
54 	BTRFS_FS_STATE_DUMMY_FS_INFO,
55 
56 	BTRFS_FS_STATE_NO_CSUMS,
57 
58 	/* Indicates there was an error cleaning up a log tree. */
59 	BTRFS_FS_STATE_LOG_CLEANUP_ERROR,
60 
61 	BTRFS_FS_STATE_COUNT
62 };
63 
64 enum {
65 	BTRFS_FS_CLOSING_START,
66 	BTRFS_FS_CLOSING_DONE,
67 	BTRFS_FS_LOG_RECOVERING,
68 	BTRFS_FS_OPEN,
69 	BTRFS_FS_QUOTA_ENABLED,
70 	BTRFS_FS_UPDATE_UUID_TREE_GEN,
71 	BTRFS_FS_CREATING_FREE_SPACE_TREE,
72 	BTRFS_FS_BTREE_ERR,
73 	BTRFS_FS_LOG1_ERR,
74 	BTRFS_FS_LOG2_ERR,
75 	BTRFS_FS_QUOTA_OVERRIDE,
76 	/* Used to record internally whether fs has been frozen */
77 	BTRFS_FS_FROZEN,
78 	/*
79 	 * Indicate that balance has been set up from the ioctl and is in the
80 	 * main phase. The fs_info::balance_ctl is initialized.
81 	 */
82 	BTRFS_FS_BALANCE_RUNNING,
83 
84 	/*
85 	 * Indicate that relocation of a chunk has started, it's set per chunk
86 	 * and is toggled between chunks.
87 	 */
88 	BTRFS_FS_RELOC_RUNNING,
89 
90 	/* Indicate that the cleaner thread is awake and doing something. */
91 	BTRFS_FS_CLEANER_RUNNING,
92 
93 	/*
94 	 * The checksumming has an optimized version and is considered fast,
95 	 * so we don't need to offload checksums to workqueues.
96 	 */
97 	BTRFS_FS_CSUM_IMPL_FAST,
98 
99 	/* Indicate that the discard workqueue can service discards. */
100 	BTRFS_FS_DISCARD_RUNNING,
101 
102 	/* Indicate that we need to cleanup space cache v1 */
103 	BTRFS_FS_CLEANUP_SPACE_CACHE_V1,
104 
105 	/* Indicate that we can't trust the free space tree for caching yet */
106 	BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED,
107 
108 	/* Indicate whether there are any tree modification log users */
109 	BTRFS_FS_TREE_MOD_LOG_USERS,
110 
111 	/* Indicate that we want the transaction kthread to commit right now. */
112 	BTRFS_FS_COMMIT_TRANS,
113 
114 	/* Indicate we have half completed snapshot deletions pending. */
115 	BTRFS_FS_UNFINISHED_DROPS,
116 
117 	/* Indicate we have to finish a zone to do next allocation. */
118 	BTRFS_FS_NEED_ZONE_FINISH,
119 
120 	/* Indicate that we want to commit the transaction. */
121 	BTRFS_FS_NEED_TRANS_COMMIT,
122 
123 	/* This is set when active zone tracking is needed. */
124 	BTRFS_FS_ACTIVE_ZONE_TRACKING,
125 
126 	/*
127 	 * Indicate if we have some features changed, this is mostly for
128 	 * cleaner thread to update the sysfs interface.
129 	 */
130 	BTRFS_FS_FEATURE_CHANGED,
131 
132 #if BITS_PER_LONG == 32
133 	/* Indicate if we have error/warn message printed on 32bit systems */
134 	BTRFS_FS_32BIT_ERROR,
135 	BTRFS_FS_32BIT_WARN,
136 #endif
137 };
138 
139 /*
140  * Flags for mount options.
141  *
142  * Note: don't forget to add new options to btrfs_show_options()
143  */
144 enum {
145 	BTRFS_MOUNT_NODATASUM			= (1UL << 0),
146 	BTRFS_MOUNT_NODATACOW			= (1UL << 1),
147 	BTRFS_MOUNT_NOBARRIER			= (1UL << 2),
148 	BTRFS_MOUNT_SSD				= (1UL << 3),
149 	BTRFS_MOUNT_DEGRADED			= (1UL << 4),
150 	BTRFS_MOUNT_COMPRESS			= (1UL << 5),
151 	BTRFS_MOUNT_NOTREELOG   		= (1UL << 6),
152 	BTRFS_MOUNT_FLUSHONCOMMIT		= (1UL << 7),
153 	BTRFS_MOUNT_SSD_SPREAD			= (1UL << 8),
154 	BTRFS_MOUNT_NOSSD			= (1UL << 9),
155 	BTRFS_MOUNT_DISCARD_SYNC		= (1UL << 10),
156 	BTRFS_MOUNT_FORCE_COMPRESS      	= (1UL << 11),
157 	BTRFS_MOUNT_SPACE_CACHE			= (1UL << 12),
158 	BTRFS_MOUNT_CLEAR_CACHE			= (1UL << 13),
159 	BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED	= (1UL << 14),
160 	BTRFS_MOUNT_ENOSPC_DEBUG		= (1UL << 15),
161 	BTRFS_MOUNT_AUTO_DEFRAG			= (1UL << 16),
162 	BTRFS_MOUNT_USEBACKUPROOT		= (1UL << 17),
163 	BTRFS_MOUNT_SKIP_BALANCE		= (1UL << 18),
164 	BTRFS_MOUNT_CHECK_INTEGRITY		= (1UL << 19),
165 	BTRFS_MOUNT_CHECK_INTEGRITY_DATA	= (1UL << 20),
166 	BTRFS_MOUNT_PANIC_ON_FATAL_ERROR	= (1UL << 21),
167 	BTRFS_MOUNT_RESCAN_UUID_TREE		= (1UL << 22),
168 	BTRFS_MOUNT_FRAGMENT_DATA		= (1UL << 23),
169 	BTRFS_MOUNT_FRAGMENT_METADATA		= (1UL << 24),
170 	BTRFS_MOUNT_FREE_SPACE_TREE		= (1UL << 25),
171 	BTRFS_MOUNT_NOLOGREPLAY			= (1UL << 26),
172 	BTRFS_MOUNT_REF_VERIFY			= (1UL << 27),
173 	BTRFS_MOUNT_DISCARD_ASYNC		= (1UL << 28),
174 	BTRFS_MOUNT_IGNOREBADROOTS		= (1UL << 29),
175 	BTRFS_MOUNT_IGNOREDATACSUMS		= (1UL << 30),
176 	BTRFS_MOUNT_NODISCARD			= (1UL << 31),
177 };
178 
179 /*
180  * Compat flags that we support.  If any incompat flags are set other than the
181  * ones specified below then we will fail to mount
182  */
183 #define BTRFS_FEATURE_COMPAT_SUPP		0ULL
184 #define BTRFS_FEATURE_COMPAT_SAFE_SET		0ULL
185 #define BTRFS_FEATURE_COMPAT_SAFE_CLEAR		0ULL
186 
187 #define BTRFS_FEATURE_COMPAT_RO_SUPP			\
188 	(BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE |	\
189 	 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID | \
190 	 BTRFS_FEATURE_COMPAT_RO_VERITY |		\
191 	 BTRFS_FEATURE_COMPAT_RO_BLOCK_GROUP_TREE)
192 
193 #define BTRFS_FEATURE_COMPAT_RO_SAFE_SET	0ULL
194 #define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR	0ULL
195 
196 #ifdef CONFIG_BTRFS_DEBUG
197 /*
198  * Extent tree v2 supported only with CONFIG_BTRFS_DEBUG
199  */
200 #define BTRFS_FEATURE_INCOMPAT_SUPP			\
201 	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
202 	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
203 	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
204 	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
205 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
206 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD |		\
207 	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
208 	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF |		\
209 	 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA |	\
210 	 BTRFS_FEATURE_INCOMPAT_NO_HOLES	|	\
211 	 BTRFS_FEATURE_INCOMPAT_METADATA_UUID	|	\
212 	 BTRFS_FEATURE_INCOMPAT_RAID1C34	|	\
213 	 BTRFS_FEATURE_INCOMPAT_ZONED		|	\
214 	 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2)
215 #else
216 #define BTRFS_FEATURE_INCOMPAT_SUPP			\
217 	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
218 	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
219 	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
220 	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
221 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
222 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD |		\
223 	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
224 	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF |		\
225 	 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA |	\
226 	 BTRFS_FEATURE_INCOMPAT_NO_HOLES	|	\
227 	 BTRFS_FEATURE_INCOMPAT_METADATA_UUID	|	\
228 	 BTRFS_FEATURE_INCOMPAT_RAID1C34	|	\
229 	 BTRFS_FEATURE_INCOMPAT_ZONED)
230 #endif
231 
232 #define BTRFS_FEATURE_INCOMPAT_SAFE_SET			\
233 	(BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
234 #define BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR		0ULL
235 
236 #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
237 #define BTRFS_DEFAULT_MAX_INLINE	(2048)
238 
239 struct btrfs_dev_replace {
240 	/* See #define above */
241 	u64 replace_state;
242 	/* Seconds since 1-Jan-1970 */
243 	time64_t time_started;
244 	/* Seconds since 1-Jan-1970 */
245 	time64_t time_stopped;
246 	atomic64_t num_write_errors;
247 	atomic64_t num_uncorrectable_read_errors;
248 
249 	u64 cursor_left;
250 	u64 committed_cursor_left;
251 	u64 cursor_left_last_write_of_item;
252 	u64 cursor_right;
253 
254 	/* See #define above */
255 	u64 cont_reading_from_srcdev_mode;
256 
257 	int is_valid;
258 	int item_needs_writeback;
259 	struct btrfs_device *srcdev;
260 	struct btrfs_device *tgtdev;
261 
262 	struct mutex lock_finishing_cancel_unmount;
263 	struct rw_semaphore rwsem;
264 
265 	struct btrfs_scrub_progress scrub_progress;
266 
267 	struct percpu_counter bio_counter;
268 	wait_queue_head_t replace_wait;
269 };
270 
271 /*
272  * Free clusters are used to claim free space in relatively large chunks,
273  * allowing us to do less seeky writes. They are used for all metadata
274  * allocations. In ssd_spread mode they are also used for data allocations.
275  */
276 struct btrfs_free_cluster {
277 	spinlock_t lock;
278 	spinlock_t refill_lock;
279 	struct rb_root root;
280 
281 	/* Largest extent in this cluster */
282 	u64 max_size;
283 
284 	/* First extent starting offset */
285 	u64 window_start;
286 
287 	/* We did a full search and couldn't create a cluster */
288 	bool fragmented;
289 
290 	struct btrfs_block_group *block_group;
291 	/*
292 	 * When a cluster is allocated from a block group, we put the cluster
293 	 * onto a list in the block group so that it can be freed before the
294 	 * block group is freed.
295 	 */
296 	struct list_head block_group_list;
297 };
298 
299 /* Discard control. */
300 /*
301  * Async discard uses multiple lists to differentiate the discard filter
302  * parameters.  Index 0 is for completely free block groups where we need to
303  * ensure the entire block group is trimmed without being lossy.  Indices
304  * afterwards represent monotonically decreasing discard filter sizes to
305  * prioritize what should be discarded next.
306  */
307 #define BTRFS_NR_DISCARD_LISTS		3
308 #define BTRFS_DISCARD_INDEX_UNUSED	0
309 #define BTRFS_DISCARD_INDEX_START	1
310 
311 struct btrfs_discard_ctl {
312 	struct workqueue_struct *discard_workers;
313 	struct delayed_work work;
314 	spinlock_t lock;
315 	struct btrfs_block_group *block_group;
316 	struct list_head discard_list[BTRFS_NR_DISCARD_LISTS];
317 	u64 prev_discard;
318 	u64 prev_discard_time;
319 	atomic_t discardable_extents;
320 	atomic64_t discardable_bytes;
321 	u64 max_discard_size;
322 	u64 delay_ms;
323 	u32 iops_limit;
324 	u32 kbps_limit;
325 	u64 discard_extent_bytes;
326 	u64 discard_bitmap_bytes;
327 	atomic64_t discard_bytes_saved;
328 };
329 
330 /*
331  * Exclusive operations (device replace, resize, device add/remove, balance)
332  */
333 enum btrfs_exclusive_operation {
334 	BTRFS_EXCLOP_NONE,
335 	BTRFS_EXCLOP_BALANCE_PAUSED,
336 	BTRFS_EXCLOP_BALANCE,
337 	BTRFS_EXCLOP_DEV_ADD,
338 	BTRFS_EXCLOP_DEV_REMOVE,
339 	BTRFS_EXCLOP_DEV_REPLACE,
340 	BTRFS_EXCLOP_RESIZE,
341 	BTRFS_EXCLOP_SWAP_ACTIVATE,
342 };
343 
344 /* Store data about transaction commits, exported via sysfs. */
345 struct btrfs_commit_stats {
346 	/* Total number of commits */
347 	u64 commit_count;
348 	/* The maximum commit duration so far in ns */
349 	u64 max_commit_dur;
350 	/* The last commit duration in ns */
351 	u64 last_commit_dur;
352 	/* The total commit duration in ns */
353 	u64 total_commit_dur;
354 };
355 
356 struct btrfs_fs_info {
357 	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
358 	unsigned long flags;
359 	struct btrfs_root *tree_root;
360 	struct btrfs_root *chunk_root;
361 	struct btrfs_root *dev_root;
362 	struct btrfs_root *fs_root;
363 	struct btrfs_root *quota_root;
364 	struct btrfs_root *uuid_root;
365 	struct btrfs_root *data_reloc_root;
366 	struct btrfs_root *block_group_root;
367 
368 	/* The log root tree is a directory of all the other log roots */
369 	struct btrfs_root *log_root_tree;
370 
371 	/* The tree that holds the global roots (csum, extent, etc) */
372 	rwlock_t global_root_lock;
373 	struct rb_root global_root_tree;
374 
375 	spinlock_t fs_roots_radix_lock;
376 	struct radix_tree_root fs_roots_radix;
377 
378 	/* Block group cache stuff */
379 	rwlock_t block_group_cache_lock;
380 	struct rb_root_cached block_group_cache_tree;
381 
382 	/* Keep track of unallocated space */
383 	atomic64_t free_chunk_space;
384 
385 	/* Track ranges which are used by log trees blocks/logged data extents */
386 	struct extent_io_tree excluded_extents;
387 
388 	/* logical->physical extent mapping */
389 	struct extent_map_tree mapping_tree;
390 
391 	/*
392 	 * Block reservation for extent, checksum, root tree and delayed dir
393 	 * index item.
394 	 */
395 	struct btrfs_block_rsv global_block_rsv;
396 	/* Block reservation for metadata operations */
397 	struct btrfs_block_rsv trans_block_rsv;
398 	/* Block reservation for chunk tree */
399 	struct btrfs_block_rsv chunk_block_rsv;
400 	/* Block reservation for delayed operations */
401 	struct btrfs_block_rsv delayed_block_rsv;
402 	/* Block reservation for delayed refs */
403 	struct btrfs_block_rsv delayed_refs_rsv;
404 
405 	struct btrfs_block_rsv empty_block_rsv;
406 
407 	u64 generation;
408 	u64 last_trans_committed;
409 	/*
410 	 * Generation of the last transaction used for block group relocation
411 	 * since the filesystem was last mounted (or 0 if none happened yet).
412 	 * Must be written and read while holding btrfs_fs_info::commit_root_sem.
413 	 */
414 	u64 last_reloc_trans;
415 	u64 avg_delayed_ref_runtime;
416 
417 	/*
418 	 * This is updated to the current trans every time a full commit is
419 	 * required instead of the faster short fsync log commits
420 	 */
421 	u64 last_trans_log_full_commit;
422 	unsigned long mount_opt;
423 
424 	unsigned long compress_type:4;
425 	unsigned int compress_level;
426 	u32 commit_interval;
427 	/*
428 	 * It is a suggestive number, the read side is safe even it gets a
429 	 * wrong number because we will write out the data into a regular
430 	 * extent. The write side(mount/remount) is under ->s_umount lock,
431 	 * so it is also safe.
432 	 */
433 	u64 max_inline;
434 
435 	struct btrfs_transaction *running_transaction;
436 	wait_queue_head_t transaction_throttle;
437 	wait_queue_head_t transaction_wait;
438 	wait_queue_head_t transaction_blocked_wait;
439 	wait_queue_head_t async_submit_wait;
440 
441 	/*
442 	 * Used to protect the incompat_flags, compat_flags, compat_ro_flags
443 	 * when they are updated.
444 	 *
445 	 * Because we do not clear the flags for ever, so we needn't use
446 	 * the lock on the read side.
447 	 *
448 	 * We also needn't use the lock when we mount the fs, because
449 	 * there is no other task which will update the flag.
450 	 */
451 	spinlock_t super_lock;
452 	struct btrfs_super_block *super_copy;
453 	struct btrfs_super_block *super_for_commit;
454 	struct super_block *sb;
455 	struct inode *btree_inode;
456 	struct mutex tree_log_mutex;
457 	struct mutex transaction_kthread_mutex;
458 	struct mutex cleaner_mutex;
459 	struct mutex chunk_mutex;
460 
461 	/*
462 	 * This is taken to make sure we don't set block groups ro after the
463 	 * free space cache has been allocated on them.
464 	 */
465 	struct mutex ro_block_group_mutex;
466 
467 	/*
468 	 * This is used during read/modify/write to make sure no two ios are
469 	 * trying to mod the same stripe at the same time.
470 	 */
471 	struct btrfs_stripe_hash_table *stripe_hash_table;
472 
473 	/*
474 	 * This protects the ordered operations list only while we are
475 	 * processing all of the entries on it.  This way we make sure the
476 	 * commit code doesn't find the list temporarily empty because another
477 	 * function happens to be doing non-waiting preflush before jumping
478 	 * into the main commit.
479 	 */
480 	struct mutex ordered_operations_mutex;
481 
482 	struct rw_semaphore commit_root_sem;
483 
484 	struct rw_semaphore cleanup_work_sem;
485 
486 	struct rw_semaphore subvol_sem;
487 
488 	spinlock_t trans_lock;
489 	/*
490 	 * The reloc mutex goes with the trans lock, it is taken during commit
491 	 * to protect us from the relocation code.
492 	 */
493 	struct mutex reloc_mutex;
494 
495 	struct list_head trans_list;
496 	struct list_head dead_roots;
497 	struct list_head caching_block_groups;
498 
499 	spinlock_t delayed_iput_lock;
500 	struct list_head delayed_iputs;
501 	atomic_t nr_delayed_iputs;
502 	wait_queue_head_t delayed_iputs_wait;
503 
504 	atomic64_t tree_mod_seq;
505 
506 	/* This protects tree_mod_log and tree_mod_seq_list */
507 	rwlock_t tree_mod_log_lock;
508 	struct rb_root tree_mod_log;
509 	struct list_head tree_mod_seq_list;
510 
511 	atomic_t async_delalloc_pages;
512 
513 	/* This is used to protect the following list -- ordered_roots. */
514 	spinlock_t ordered_root_lock;
515 
516 	/*
517 	 * All fs/file tree roots in which there are data=ordered extents
518 	 * pending writeback are added into this list.
519 	 *
520 	 * These can span multiple transactions and basically include every
521 	 * dirty data page that isn't from nodatacow.
522 	 */
523 	struct list_head ordered_roots;
524 
525 	struct mutex delalloc_root_mutex;
526 	spinlock_t delalloc_root_lock;
527 	/* All fs/file tree roots that have delalloc inodes. */
528 	struct list_head delalloc_roots;
529 
530 	/*
531 	 * There is a pool of worker threads for checksumming during writes and
532 	 * a pool for checksumming after reads.  This is because readers can
533 	 * run with FS locks held, and the writers may be waiting for those
534 	 * locks.  We don't want ordering in the pending list to cause
535 	 * deadlocks, and so the two are serviced separately.
536 	 *
537 	 * A third pool does submit_bio to avoid deadlocking with the other two.
538 	 */
539 	struct btrfs_workqueue *workers;
540 	struct btrfs_workqueue *hipri_workers;
541 	struct btrfs_workqueue *delalloc_workers;
542 	struct btrfs_workqueue *flush_workers;
543 	struct workqueue_struct *endio_workers;
544 	struct workqueue_struct *endio_meta_workers;
545 	struct workqueue_struct *rmw_workers;
546 	struct workqueue_struct *compressed_write_workers;
547 	struct btrfs_workqueue *endio_write_workers;
548 	struct btrfs_workqueue *endio_freespace_worker;
549 	struct btrfs_workqueue *caching_workers;
550 
551 	/*
552 	 * Fixup workers take dirty pages that didn't properly go through the
553 	 * cow mechanism and make them safe to write.  It happens for the
554 	 * sys_munmap function call path.
555 	 */
556 	struct btrfs_workqueue *fixup_workers;
557 	struct btrfs_workqueue *delayed_workers;
558 
559 	struct task_struct *transaction_kthread;
560 	struct task_struct *cleaner_kthread;
561 	u32 thread_pool_size;
562 
563 	struct kobject *space_info_kobj;
564 	struct kobject *qgroups_kobj;
565 	struct kobject *discard_kobj;
566 
567 	/* Used to keep from writing metadata until there is a nice batch */
568 	struct percpu_counter dirty_metadata_bytes;
569 	struct percpu_counter delalloc_bytes;
570 	struct percpu_counter ordered_bytes;
571 	s32 dirty_metadata_batch;
572 	s32 delalloc_batch;
573 
574 	struct list_head dirty_cowonly_roots;
575 
576 	struct btrfs_fs_devices *fs_devices;
577 
578 	/*
579 	 * The space_info list is effectively read only after initial setup.
580 	 * It is populated at mount time and cleaned up after all block groups
581 	 * are removed.  RCU is used to protect it.
582 	 */
583 	struct list_head space_info;
584 
585 	struct btrfs_space_info *data_sinfo;
586 
587 	struct reloc_control *reloc_ctl;
588 
589 	/* data_alloc_cluster is only used in ssd_spread mode */
590 	struct btrfs_free_cluster data_alloc_cluster;
591 
592 	/* All metadata allocations go through this cluster. */
593 	struct btrfs_free_cluster meta_alloc_cluster;
594 
595 	/* Auto defrag inodes go here. */
596 	spinlock_t defrag_inodes_lock;
597 	struct rb_root defrag_inodes;
598 	atomic_t defrag_running;
599 
600 	/* Used to protect avail_{data, metadata, system}_alloc_bits */
601 	seqlock_t profiles_lock;
602 	/*
603 	 * These three are in extended format (availability of single chunks is
604 	 * denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other types are denoted
605 	 * by corresponding BTRFS_BLOCK_GROUP_* bits)
606 	 */
607 	u64 avail_data_alloc_bits;
608 	u64 avail_metadata_alloc_bits;
609 	u64 avail_system_alloc_bits;
610 
611 	/* Balance state */
612 	spinlock_t balance_lock;
613 	struct mutex balance_mutex;
614 	atomic_t balance_pause_req;
615 	atomic_t balance_cancel_req;
616 	struct btrfs_balance_control *balance_ctl;
617 	wait_queue_head_t balance_wait_q;
618 
619 	/* Cancellation requests for chunk relocation */
620 	atomic_t reloc_cancel_req;
621 
622 	u32 data_chunk_allocations;
623 	u32 metadata_ratio;
624 
625 	void *bdev_holder;
626 
627 	/* Private scrub information */
628 	struct mutex scrub_lock;
629 	atomic_t scrubs_running;
630 	atomic_t scrub_pause_req;
631 	atomic_t scrubs_paused;
632 	atomic_t scrub_cancel_req;
633 	wait_queue_head_t scrub_pause_wait;
634 	/*
635 	 * The worker pointers are NULL iff the refcount is 0, ie. scrub is not
636 	 * running.
637 	 */
638 	refcount_t scrub_workers_refcnt;
639 	struct workqueue_struct *scrub_workers;
640 	struct workqueue_struct *scrub_wr_completion_workers;
641 	struct workqueue_struct *scrub_parity_workers;
642 	struct btrfs_subpage_info *subpage_info;
643 
644 	struct btrfs_discard_ctl discard_ctl;
645 
646 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
647 	u32 check_integrity_print_mask;
648 #endif
649 	/* Is qgroup tracking in a consistent state? */
650 	u64 qgroup_flags;
651 
652 	/* Holds configuration and tracking. Protected by qgroup_lock. */
653 	struct rb_root qgroup_tree;
654 	spinlock_t qgroup_lock;
655 
656 	/*
657 	 * Used to avoid frequently calling ulist_alloc()/ulist_free()
658 	 * when doing qgroup accounting, it must be protected by qgroup_lock.
659 	 */
660 	struct ulist *qgroup_ulist;
661 
662 	/*
663 	 * Protect user change for quota operations. If a transaction is needed,
664 	 * it must be started before locking this lock.
665 	 */
666 	struct mutex qgroup_ioctl_lock;
667 
668 	/* List of dirty qgroups to be written at next commit. */
669 	struct list_head dirty_qgroups;
670 
671 	/* Used by qgroup for an efficient tree traversal. */
672 	u64 qgroup_seq;
673 
674 	/* Qgroup rescan items. */
675 	/* Protects the progress item */
676 	struct mutex qgroup_rescan_lock;
677 	struct btrfs_key qgroup_rescan_progress;
678 	struct btrfs_workqueue *qgroup_rescan_workers;
679 	struct completion qgroup_rescan_completion;
680 	struct btrfs_work qgroup_rescan_work;
681 	/* Protected by qgroup_rescan_lock */
682 	bool qgroup_rescan_running;
683 	u8 qgroup_drop_subtree_thres;
684 
685 	/* Filesystem state */
686 	unsigned long fs_state;
687 
688 	struct btrfs_delayed_root *delayed_root;
689 
690 	/* Extent buffer radix tree */
691 	spinlock_t buffer_lock;
692 	/* Entries are eb->start / sectorsize */
693 	struct radix_tree_root buffer_radix;
694 
695 	/* Next backup root to be overwritten */
696 	int backup_root_index;
697 
698 	/* Device replace state */
699 	struct btrfs_dev_replace dev_replace;
700 
701 	struct semaphore uuid_tree_rescan_sem;
702 
703 	/* Used to reclaim the metadata space in the background. */
704 	struct work_struct async_reclaim_work;
705 	struct work_struct async_data_reclaim_work;
706 	struct work_struct preempt_reclaim_work;
707 
708 	/* Reclaim partially filled block groups in the background */
709 	struct work_struct reclaim_bgs_work;
710 	struct list_head reclaim_bgs;
711 	int bg_reclaim_threshold;
712 
713 	spinlock_t unused_bgs_lock;
714 	struct list_head unused_bgs;
715 	struct mutex unused_bg_unpin_mutex;
716 	/* Protect block groups that are going to be deleted */
717 	struct mutex reclaim_bgs_lock;
718 
719 	/* Cached block sizes */
720 	u32 nodesize;
721 	u32 sectorsize;
722 	/* ilog2 of sectorsize, use to avoid 64bit division */
723 	u32 sectorsize_bits;
724 	u32 csum_size;
725 	u32 csums_per_leaf;
726 	u32 stripesize;
727 
728 	/*
729 	 * Maximum size of an extent. BTRFS_MAX_EXTENT_SIZE on regular
730 	 * filesystem, on zoned it depends on the device constraints.
731 	 */
732 	u64 max_extent_size;
733 
734 	/* Block groups and devices containing active swapfiles. */
735 	spinlock_t swapfile_pins_lock;
736 	struct rb_root swapfile_pins;
737 
738 	struct crypto_shash *csum_shash;
739 
740 	/* Type of exclusive operation running, protected by super_lock */
741 	enum btrfs_exclusive_operation exclusive_operation;
742 
743 	/*
744 	 * Zone size > 0 when in ZONED mode, otherwise it's used for a check
745 	 * if the mode is enabled
746 	 */
747 	u64 zone_size;
748 
749 	/* Constraints for ZONE_APPEND commands: */
750 	struct queue_limits limits;
751 	u64 max_zone_append_size;
752 
753 	struct mutex zoned_meta_io_lock;
754 	spinlock_t treelog_bg_lock;
755 	u64 treelog_bg;
756 
757 	/*
758 	 * Start of the dedicated data relocation block group, protected by
759 	 * relocation_bg_lock.
760 	 */
761 	spinlock_t relocation_bg_lock;
762 	u64 data_reloc_bg;
763 	struct mutex zoned_data_reloc_io_lock;
764 
765 	u64 nr_global_roots;
766 
767 	spinlock_t zone_active_bgs_lock;
768 	struct list_head zone_active_bgs;
769 
770 	/* Updates are not protected by any lock */
771 	struct btrfs_commit_stats commit_stats;
772 
773 	/*
774 	 * Last generation where we dropped a non-relocation root.
775 	 * Use btrfs_set_last_root_drop_gen() and btrfs_get_last_root_drop_gen()
776 	 * to change it and to read it, respectively.
777 	 */
778 	u64 last_root_drop_gen;
779 
780 	/*
781 	 * Annotations for transaction events (structures are empty when
782 	 * compiled without lockdep).
783 	 */
784 	struct lockdep_map btrfs_trans_num_writers_map;
785 	struct lockdep_map btrfs_trans_num_extwriters_map;
786 	struct lockdep_map btrfs_state_change_map[4];
787 	struct lockdep_map btrfs_trans_pending_ordered_map;
788 	struct lockdep_map btrfs_ordered_extent_map;
789 
790 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
791 	spinlock_t ref_verify_lock;
792 	struct rb_root block_tree;
793 #endif
794 
795 #ifdef CONFIG_BTRFS_DEBUG
796 	struct kobject *debug_kobj;
797 	struct list_head allocated_roots;
798 
799 	spinlock_t eb_leak_lock;
800 	struct list_head allocated_ebs;
801 #endif
802 };
803 
804 static inline void btrfs_set_last_root_drop_gen(struct btrfs_fs_info *fs_info,
805 						u64 gen)
806 {
807 	WRITE_ONCE(fs_info->last_root_drop_gen, gen);
808 }
809 
810 static inline u64 btrfs_get_last_root_drop_gen(const struct btrfs_fs_info *fs_info)
811 {
812 	return READ_ONCE(fs_info->last_root_drop_gen);
813 }
814 
815 /*
816  * Take the number of bytes to be checksummed and figure out how many leaves
817  * it would require to store the csums for that many bytes.
818  */
819 static inline u64 btrfs_csum_bytes_to_leaves(
820 			const struct btrfs_fs_info *fs_info, u64 csum_bytes)
821 {
822 	const u64 num_csums = csum_bytes >> fs_info->sectorsize_bits;
823 
824 	return DIV_ROUND_UP_ULL(num_csums, fs_info->csums_per_leaf);
825 }
826 
827 /*
828  * Use this if we would be adding new items, as we could split nodes as we cow
829  * down the tree.
830  */
831 static inline u64 btrfs_calc_insert_metadata_size(struct btrfs_fs_info *fs_info,
832 						  unsigned num_items)
833 {
834 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * 2 * num_items;
835 }
836 
837 /*
838  * Doing a truncate or a modification won't result in new nodes or leaves, just
839  * what we need for COW.
840  */
841 static inline u64 btrfs_calc_metadata_size(struct btrfs_fs_info *fs_info,
842 						 unsigned num_items)
843 {
844 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * num_items;
845 }
846 
847 #define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r->fs_info) >> 4) - \
848 					sizeof(struct btrfs_item))
849 
850 static inline bool btrfs_is_zoned(const struct btrfs_fs_info *fs_info)
851 {
852 	return fs_info->zone_size > 0;
853 }
854 
855 /*
856  * Count how many fs_info->max_extent_size cover the @size
857  */
858 static inline u32 count_max_extents(struct btrfs_fs_info *fs_info, u64 size)
859 {
860 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
861 	if (!fs_info)
862 		return div_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, BTRFS_MAX_EXTENT_SIZE);
863 #endif
864 
865 	return div_u64(size + fs_info->max_extent_size - 1, fs_info->max_extent_size);
866 }
867 
868 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
869 			enum btrfs_exclusive_operation type);
870 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
871 				 enum btrfs_exclusive_operation type);
872 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info);
873 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info);
874 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
875 			  enum btrfs_exclusive_operation op);
876 
877 /* Compatibility and incompatibility defines */
878 void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
879 			     const char *name);
880 void __btrfs_clear_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
881 			       const char *name);
882 void __btrfs_set_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
883 			      const char *name);
884 void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
885 				const char *name);
886 
887 #define __btrfs_fs_incompat(fs_info, flags)				\
888 	(!!(btrfs_super_incompat_flags((fs_info)->super_copy) & (flags)))
889 
890 #define __btrfs_fs_compat_ro(fs_info, flags)				\
891 	(!!(btrfs_super_compat_ro_flags((fs_info)->super_copy) & (flags)))
892 
893 #define btrfs_set_fs_incompat(__fs_info, opt)				\
894 	__btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
895 
896 #define btrfs_clear_fs_incompat(__fs_info, opt)				\
897 	__btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
898 
899 #define btrfs_fs_incompat(fs_info, opt)					\
900 	__btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
901 
902 #define btrfs_set_fs_compat_ro(__fs_info, opt)				\
903 	__btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
904 
905 #define btrfs_clear_fs_compat_ro(__fs_info, opt)			\
906 	__btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
907 
908 #define btrfs_fs_compat_ro(fs_info, opt)				\
909 	__btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
910 
911 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
912 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
913 #define btrfs_raw_test_opt(o, opt)	((o) & BTRFS_MOUNT_##opt)
914 #define btrfs_test_opt(fs_info, opt)	((fs_info)->mount_opt & \
915 					 BTRFS_MOUNT_##opt)
916 
917 #define btrfs_set_and_info(fs_info, opt, fmt, args...)			\
918 do {									\
919 	if (!btrfs_test_opt(fs_info, opt))				\
920 		btrfs_info(fs_info, fmt, ##args);			\
921 	btrfs_set_opt(fs_info->mount_opt, opt);				\
922 } while (0)
923 
924 #define btrfs_clear_and_info(fs_info, opt, fmt, args...)		\
925 do {									\
926 	if (btrfs_test_opt(fs_info, opt))				\
927 		btrfs_info(fs_info, fmt, ##args);			\
928 	btrfs_clear_opt(fs_info->mount_opt, opt);			\
929 } while (0)
930 
931 static inline int btrfs_fs_closing(struct btrfs_fs_info *fs_info)
932 {
933 	/* Do it this way so we only ever do one test_bit in the normal case. */
934 	if (test_bit(BTRFS_FS_CLOSING_START, &fs_info->flags)) {
935 		if (test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags))
936 			return 2;
937 		return 1;
938 	}
939 	return 0;
940 }
941 
942 /*
943  * If we remount the fs to be R/O or umount the fs, the cleaner needn't do
944  * anything except sleeping. This function is used to check the status of
945  * the fs.
946  * We check for BTRFS_FS_STATE_RO to avoid races with a concurrent remount,
947  * since setting and checking for SB_RDONLY in the superblock's flags is not
948  * atomic.
949  */
950 static inline int btrfs_need_cleaner_sleep(struct btrfs_fs_info *fs_info)
951 {
952 	return test_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state) ||
953 		btrfs_fs_closing(fs_info);
954 }
955 
956 static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info *fs_info)
957 {
958 	clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags);
959 }
960 
961 #define BTRFS_FS_ERROR(fs_info)	(unlikely(test_bit(BTRFS_FS_STATE_ERROR, \
962 						   &(fs_info)->fs_state)))
963 #define BTRFS_FS_LOG_CLEANUP_ERROR(fs_info)				\
964 	(unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR,		\
965 			   &(fs_info)->fs_state)))
966 
967 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
968 
969 #define EXPORT_FOR_TESTS
970 
971 static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
972 {
973 	return test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
974 }
975 
976 void btrfs_test_destroy_inode(struct inode *inode);
977 
978 #else
979 
980 #define EXPORT_FOR_TESTS static
981 
982 static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
983 {
984 	return 0;
985 }
986 #endif
987 
988 #endif
989