xref: /linux/fs/btrfs/fs.h (revision 41f7adb676f6c4aef439d78b15c7e1119216bc2b)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 
3 #ifndef BTRFS_FS_H
4 #define BTRFS_FS_H
5 
6 #include <linux/blkdev.h>
7 #include <linux/sizes.h>
8 #include <linux/time64.h>
9 #include <linux/compiler.h>
10 #include <linux/math.h>
11 #include <linux/atomic.h>
12 #include <linux/percpu_counter.h>
13 #include <linux/completion.h>
14 #include <linux/lockdep.h>
15 #include <linux/spinlock.h>
16 #include <linux/mutex.h>
17 #include <linux/rwlock_types.h>
18 #include <linux/rwsem.h>
19 #include <linux/semaphore.h>
20 #include <linux/list.h>
21 #include <linux/radix-tree.h>
22 #include <linux/workqueue.h>
23 #include <linux/wait.h>
24 #include <linux/wait_bit.h>
25 #include <linux/sched.h>
26 #include <linux/rbtree.h>
27 #include <uapi/linux/btrfs.h>
28 #include <uapi/linux/btrfs_tree.h>
29 #include "extent-io-tree.h"
30 #include "async-thread.h"
31 #include "block-rsv.h"
32 #include "fs.h"
33 
34 struct inode;
35 struct super_block;
36 struct kobject;
37 struct reloc_control;
38 struct crypto_shash;
39 struct ulist;
40 struct btrfs_device;
41 struct btrfs_block_group;
42 struct btrfs_root;
43 struct btrfs_fs_devices;
44 struct btrfs_transaction;
45 struct btrfs_delayed_root;
46 struct btrfs_balance_control;
47 struct btrfs_subpage_info;
48 struct btrfs_stripe_hash_table;
49 struct btrfs_space_info;
50 
51 #define BTRFS_MAX_EXTENT_SIZE SZ_128M
52 
53 #define BTRFS_OLDEST_GENERATION	0ULL
54 
55 #define BTRFS_EMPTY_DIR_SIZE 0
56 
57 #define BTRFS_DIRTY_METADATA_THRESH		SZ_32M
58 
59 #define BTRFS_SUPER_INFO_OFFSET			SZ_64K
60 #define BTRFS_SUPER_INFO_SIZE			4096
61 static_assert(sizeof(struct btrfs_super_block) == BTRFS_SUPER_INFO_SIZE);
62 
63 /*
64  * Number of metadata items necessary for an unlink operation:
65  *
66  * 1 for the possible orphan item
67  * 1 for the dir item
68  * 1 for the dir index
69  * 1 for the inode ref
70  * 1 for the inode
71  * 1 for the parent inode
72  */
73 #define BTRFS_UNLINK_METADATA_UNITS		6
74 
75 /*
76  * The reserved space at the beginning of each device.  It covers the primary
77  * super block and leaves space for potential use by other tools like
78  * bootloaders or to lower potential damage of accidental overwrite.
79  */
80 #define BTRFS_DEVICE_RANGE_RESERVED			(SZ_1M)
81 /*
82  * Runtime (in-memory) states of filesystem
83  */
84 enum {
85 	/*
86 	 * Filesystem is being remounted, allow to skip some operations, like
87 	 * defrag
88 	 */
89 	BTRFS_FS_STATE_REMOUNTING,
90 	/* Filesystem in RO mode */
91 	BTRFS_FS_STATE_RO,
92 	/* Track if a transaction abort has been reported on this filesystem */
93 	BTRFS_FS_STATE_TRANS_ABORTED,
94 	/*
95 	 * Bio operations should be blocked on this filesystem because a source
96 	 * or target device is being destroyed as part of a device replace
97 	 */
98 	BTRFS_FS_STATE_DEV_REPLACING,
99 	/* The btrfs_fs_info created for self-tests */
100 	BTRFS_FS_STATE_DUMMY_FS_INFO,
101 
102 	BTRFS_FS_STATE_NO_CSUMS,
103 
104 	/* Indicates there was an error cleaning up a log tree. */
105 	BTRFS_FS_STATE_LOG_CLEANUP_ERROR,
106 
107 	BTRFS_FS_STATE_COUNT
108 };
109 
110 enum {
111 	BTRFS_FS_CLOSING_START,
112 	BTRFS_FS_CLOSING_DONE,
113 	BTRFS_FS_LOG_RECOVERING,
114 	BTRFS_FS_OPEN,
115 	BTRFS_FS_QUOTA_ENABLED,
116 	BTRFS_FS_UPDATE_UUID_TREE_GEN,
117 	BTRFS_FS_CREATING_FREE_SPACE_TREE,
118 	BTRFS_FS_BTREE_ERR,
119 	BTRFS_FS_LOG1_ERR,
120 	BTRFS_FS_LOG2_ERR,
121 	BTRFS_FS_QUOTA_OVERRIDE,
122 	/* Used to record internally whether fs has been frozen */
123 	BTRFS_FS_FROZEN,
124 	/*
125 	 * Indicate that balance has been set up from the ioctl and is in the
126 	 * main phase. The fs_info::balance_ctl is initialized.
127 	 */
128 	BTRFS_FS_BALANCE_RUNNING,
129 
130 	/*
131 	 * Indicate that relocation of a chunk has started, it's set per chunk
132 	 * and is toggled between chunks.
133 	 */
134 	BTRFS_FS_RELOC_RUNNING,
135 
136 	/* Indicate that the cleaner thread is awake and doing something. */
137 	BTRFS_FS_CLEANER_RUNNING,
138 
139 	/*
140 	 * The checksumming has an optimized version and is considered fast,
141 	 * so we don't need to offload checksums to workqueues.
142 	 */
143 	BTRFS_FS_CSUM_IMPL_FAST,
144 
145 	/* Indicate that the discard workqueue can service discards. */
146 	BTRFS_FS_DISCARD_RUNNING,
147 
148 	/* Indicate that we need to cleanup space cache v1 */
149 	BTRFS_FS_CLEANUP_SPACE_CACHE_V1,
150 
151 	/* Indicate that we can't trust the free space tree for caching yet */
152 	BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED,
153 
154 	/* Indicate whether there are any tree modification log users */
155 	BTRFS_FS_TREE_MOD_LOG_USERS,
156 
157 	/* Indicate that we want the transaction kthread to commit right now. */
158 	BTRFS_FS_COMMIT_TRANS,
159 
160 	/* Indicate we have half completed snapshot deletions pending. */
161 	BTRFS_FS_UNFINISHED_DROPS,
162 
163 	/* Indicate we have to finish a zone to do next allocation. */
164 	BTRFS_FS_NEED_ZONE_FINISH,
165 
166 	/* Indicate that we want to commit the transaction. */
167 	BTRFS_FS_NEED_TRANS_COMMIT,
168 
169 	/* This is set when active zone tracking is needed. */
170 	BTRFS_FS_ACTIVE_ZONE_TRACKING,
171 
172 	/*
173 	 * Indicate if we have some features changed, this is mostly for
174 	 * cleaner thread to update the sysfs interface.
175 	 */
176 	BTRFS_FS_FEATURE_CHANGED,
177 
178 	/*
179 	 * Indicate that we have found a tree block which is only aligned to
180 	 * sectorsize, but not to nodesize.  This should be rare nowadays.
181 	 */
182 	BTRFS_FS_UNALIGNED_TREE_BLOCK,
183 
184 #if BITS_PER_LONG == 32
185 	/* Indicate if we have error/warn message printed on 32bit systems */
186 	BTRFS_FS_32BIT_ERROR,
187 	BTRFS_FS_32BIT_WARN,
188 #endif
189 };
190 
191 /*
192  * Flags for mount options.
193  *
194  * Note: don't forget to add new options to btrfs_show_options()
195  */
196 enum {
197 	BTRFS_MOUNT_NODATASUM			= (1UL << 0),
198 	BTRFS_MOUNT_NODATACOW			= (1UL << 1),
199 	BTRFS_MOUNT_NOBARRIER			= (1UL << 2),
200 	BTRFS_MOUNT_SSD				= (1UL << 3),
201 	BTRFS_MOUNT_DEGRADED			= (1UL << 4),
202 	BTRFS_MOUNT_COMPRESS			= (1UL << 5),
203 	BTRFS_MOUNT_NOTREELOG   		= (1UL << 6),
204 	BTRFS_MOUNT_FLUSHONCOMMIT		= (1UL << 7),
205 	BTRFS_MOUNT_SSD_SPREAD			= (1UL << 8),
206 	BTRFS_MOUNT_NOSSD			= (1UL << 9),
207 	BTRFS_MOUNT_DISCARD_SYNC		= (1UL << 10),
208 	BTRFS_MOUNT_FORCE_COMPRESS      	= (1UL << 11),
209 	BTRFS_MOUNT_SPACE_CACHE			= (1UL << 12),
210 	BTRFS_MOUNT_CLEAR_CACHE			= (1UL << 13),
211 	BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED	= (1UL << 14),
212 	BTRFS_MOUNT_ENOSPC_DEBUG		= (1UL << 15),
213 	BTRFS_MOUNT_AUTO_DEFRAG			= (1UL << 16),
214 	BTRFS_MOUNT_USEBACKUPROOT		= (1UL << 17),
215 	BTRFS_MOUNT_SKIP_BALANCE		= (1UL << 18),
216 	BTRFS_MOUNT_PANIC_ON_FATAL_ERROR	= (1UL << 19),
217 	BTRFS_MOUNT_RESCAN_UUID_TREE		= (1UL << 20),
218 	BTRFS_MOUNT_FRAGMENT_DATA		= (1UL << 21),
219 	BTRFS_MOUNT_FRAGMENT_METADATA		= (1UL << 22),
220 	BTRFS_MOUNT_FREE_SPACE_TREE		= (1UL << 23),
221 	BTRFS_MOUNT_NOLOGREPLAY			= (1UL << 24),
222 	BTRFS_MOUNT_REF_VERIFY			= (1UL << 25),
223 	BTRFS_MOUNT_DISCARD_ASYNC		= (1UL << 26),
224 	BTRFS_MOUNT_IGNOREBADROOTS		= (1UL << 27),
225 	BTRFS_MOUNT_IGNOREDATACSUMS		= (1UL << 28),
226 	BTRFS_MOUNT_NODISCARD			= (1UL << 29),
227 	BTRFS_MOUNT_NOSPACECACHE		= (1UL << 30),
228 };
229 
230 /*
231  * Compat flags that we support.  If any incompat flags are set other than the
232  * ones specified below then we will fail to mount
233  */
234 #define BTRFS_FEATURE_COMPAT_SUPP		0ULL
235 #define BTRFS_FEATURE_COMPAT_SAFE_SET		0ULL
236 #define BTRFS_FEATURE_COMPAT_SAFE_CLEAR		0ULL
237 
238 #define BTRFS_FEATURE_COMPAT_RO_SUPP			\
239 	(BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE |	\
240 	 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID | \
241 	 BTRFS_FEATURE_COMPAT_RO_VERITY |		\
242 	 BTRFS_FEATURE_COMPAT_RO_BLOCK_GROUP_TREE)
243 
244 #define BTRFS_FEATURE_COMPAT_RO_SAFE_SET	0ULL
245 #define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR	0ULL
246 
247 #define BTRFS_FEATURE_INCOMPAT_SUPP_STABLE		\
248 	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
249 	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
250 	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
251 	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
252 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
253 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD |		\
254 	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
255 	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF |		\
256 	 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA |	\
257 	 BTRFS_FEATURE_INCOMPAT_NO_HOLES	|	\
258 	 BTRFS_FEATURE_INCOMPAT_METADATA_UUID	|	\
259 	 BTRFS_FEATURE_INCOMPAT_RAID1C34	|	\
260 	 BTRFS_FEATURE_INCOMPAT_ZONED		|	\
261 	 BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA)
262 
263 #ifdef CONFIG_BTRFS_DEBUG
264 	/*
265 	 * Features under developmen like Extent tree v2 support is enabled
266 	 * only under CONFIG_BTRFS_DEBUG.
267 	 */
268 #define BTRFS_FEATURE_INCOMPAT_SUPP		\
269 	(BTRFS_FEATURE_INCOMPAT_SUPP_STABLE |	\
270 	 BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE | \
271 	 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2)
272 
273 #else
274 
275 #define BTRFS_FEATURE_INCOMPAT_SUPP		\
276 	(BTRFS_FEATURE_INCOMPAT_SUPP_STABLE)
277 
278 #endif
279 
280 #define BTRFS_FEATURE_INCOMPAT_SAFE_SET			\
281 	(BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
282 #define BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR		0ULL
283 
284 #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
285 #define BTRFS_DEFAULT_MAX_INLINE	(2048)
286 
287 struct btrfs_dev_replace {
288 	/* See #define above */
289 	u64 replace_state;
290 	/* Seconds since 1-Jan-1970 */
291 	time64_t time_started;
292 	/* Seconds since 1-Jan-1970 */
293 	time64_t time_stopped;
294 	atomic64_t num_write_errors;
295 	atomic64_t num_uncorrectable_read_errors;
296 
297 	u64 cursor_left;
298 	u64 committed_cursor_left;
299 	u64 cursor_left_last_write_of_item;
300 	u64 cursor_right;
301 
302 	/* See #define above */
303 	u64 cont_reading_from_srcdev_mode;
304 
305 	int is_valid;
306 	int item_needs_writeback;
307 	struct btrfs_device *srcdev;
308 	struct btrfs_device *tgtdev;
309 
310 	struct mutex lock_finishing_cancel_unmount;
311 	struct rw_semaphore rwsem;
312 
313 	struct btrfs_scrub_progress scrub_progress;
314 
315 	struct percpu_counter bio_counter;
316 	wait_queue_head_t replace_wait;
317 };
318 
319 /*
320  * Free clusters are used to claim free space in relatively large chunks,
321  * allowing us to do less seeky writes. They are used for all metadata
322  * allocations. In ssd_spread mode they are also used for data allocations.
323  */
324 struct btrfs_free_cluster {
325 	spinlock_t lock;
326 	spinlock_t refill_lock;
327 	struct rb_root root;
328 
329 	/* Largest extent in this cluster */
330 	u64 max_size;
331 
332 	/* First extent starting offset */
333 	u64 window_start;
334 
335 	/* We did a full search and couldn't create a cluster */
336 	bool fragmented;
337 
338 	struct btrfs_block_group *block_group;
339 	/*
340 	 * When a cluster is allocated from a block group, we put the cluster
341 	 * onto a list in the block group so that it can be freed before the
342 	 * block group is freed.
343 	 */
344 	struct list_head block_group_list;
345 };
346 
347 /* Discard control. */
348 /*
349  * Async discard uses multiple lists to differentiate the discard filter
350  * parameters.  Index 0 is for completely free block groups where we need to
351  * ensure the entire block group is trimmed without being lossy.  Indices
352  * afterwards represent monotonically decreasing discard filter sizes to
353  * prioritize what should be discarded next.
354  */
355 #define BTRFS_NR_DISCARD_LISTS		3
356 #define BTRFS_DISCARD_INDEX_UNUSED	0
357 #define BTRFS_DISCARD_INDEX_START	1
358 
359 struct btrfs_discard_ctl {
360 	struct workqueue_struct *discard_workers;
361 	struct delayed_work work;
362 	spinlock_t lock;
363 	struct btrfs_block_group *block_group;
364 	struct list_head discard_list[BTRFS_NR_DISCARD_LISTS];
365 	u64 prev_discard;
366 	u64 prev_discard_time;
367 	atomic_t discardable_extents;
368 	atomic64_t discardable_bytes;
369 	u64 max_discard_size;
370 	u64 delay_ms;
371 	u32 iops_limit;
372 	u32 kbps_limit;
373 	u64 discard_extent_bytes;
374 	u64 discard_bitmap_bytes;
375 	atomic64_t discard_bytes_saved;
376 };
377 
378 /*
379  * Exclusive operations (device replace, resize, device add/remove, balance)
380  */
381 enum btrfs_exclusive_operation {
382 	BTRFS_EXCLOP_NONE,
383 	BTRFS_EXCLOP_BALANCE_PAUSED,
384 	BTRFS_EXCLOP_BALANCE,
385 	BTRFS_EXCLOP_DEV_ADD,
386 	BTRFS_EXCLOP_DEV_REMOVE,
387 	BTRFS_EXCLOP_DEV_REPLACE,
388 	BTRFS_EXCLOP_RESIZE,
389 	BTRFS_EXCLOP_SWAP_ACTIVATE,
390 };
391 
392 /* Store data about transaction commits, exported via sysfs. */
393 struct btrfs_commit_stats {
394 	/* Total number of commits */
395 	u64 commit_count;
396 	/* The maximum commit duration so far in ns */
397 	u64 max_commit_dur;
398 	/* The last commit duration in ns */
399 	u64 last_commit_dur;
400 	/* The total commit duration in ns */
401 	u64 total_commit_dur;
402 };
403 
404 struct btrfs_fs_info {
405 	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
406 	unsigned long flags;
407 	struct btrfs_root *tree_root;
408 	struct btrfs_root *chunk_root;
409 	struct btrfs_root *dev_root;
410 	struct btrfs_root *fs_root;
411 	struct btrfs_root *quota_root;
412 	struct btrfs_root *uuid_root;
413 	struct btrfs_root *data_reloc_root;
414 	struct btrfs_root *block_group_root;
415 	struct btrfs_root *stripe_root;
416 
417 	/* The log root tree is a directory of all the other log roots */
418 	struct btrfs_root *log_root_tree;
419 
420 	/* The tree that holds the global roots (csum, extent, etc) */
421 	rwlock_t global_root_lock;
422 	struct rb_root global_root_tree;
423 
424 	spinlock_t fs_roots_radix_lock;
425 	struct radix_tree_root fs_roots_radix;
426 
427 	/* Block group cache stuff */
428 	rwlock_t block_group_cache_lock;
429 	struct rb_root_cached block_group_cache_tree;
430 
431 	/* Keep track of unallocated space */
432 	atomic64_t free_chunk_space;
433 
434 	/* Track ranges which are used by log trees blocks/logged data extents */
435 	struct extent_io_tree excluded_extents;
436 
437 	/* logical->physical extent mapping */
438 	struct rb_root_cached mapping_tree;
439 	rwlock_t mapping_tree_lock;
440 
441 	/*
442 	 * Block reservation for extent, checksum, root tree and delayed dir
443 	 * index item.
444 	 */
445 	struct btrfs_block_rsv global_block_rsv;
446 	/* Block reservation for metadata operations */
447 	struct btrfs_block_rsv trans_block_rsv;
448 	/* Block reservation for chunk tree */
449 	struct btrfs_block_rsv chunk_block_rsv;
450 	/* Block reservation for delayed operations */
451 	struct btrfs_block_rsv delayed_block_rsv;
452 	/* Block reservation for delayed refs */
453 	struct btrfs_block_rsv delayed_refs_rsv;
454 
455 	struct btrfs_block_rsv empty_block_rsv;
456 
457 	/*
458 	 * Updated while holding the lock 'trans_lock'. Due to the life cycle of
459 	 * a transaction, it can be directly read while holding a transaction
460 	 * handle, everywhere else must be read with btrfs_get_fs_generation().
461 	 * Should always be updated using btrfs_set_fs_generation().
462 	 */
463 	u64 generation;
464 	/*
465 	 * Always use btrfs_get_last_trans_committed() and
466 	 * btrfs_set_last_trans_committed() to read and update this field.
467 	 */
468 	u64 last_trans_committed;
469 	/*
470 	 * Generation of the last transaction used for block group relocation
471 	 * since the filesystem was last mounted (or 0 if none happened yet).
472 	 * Must be written and read while holding btrfs_fs_info::commit_root_sem.
473 	 */
474 	u64 last_reloc_trans;
475 
476 	/*
477 	 * This is updated to the current trans every time a full commit is
478 	 * required instead of the faster short fsync log commits
479 	 */
480 	u64 last_trans_log_full_commit;
481 	unsigned long mount_opt;
482 
483 	unsigned long compress_type:4;
484 	unsigned int compress_level;
485 	u32 commit_interval;
486 	/*
487 	 * It is a suggestive number, the read side is safe even it gets a
488 	 * wrong number because we will write out the data into a regular
489 	 * extent. The write side(mount/remount) is under ->s_umount lock,
490 	 * so it is also safe.
491 	 */
492 	u64 max_inline;
493 
494 	struct btrfs_transaction *running_transaction;
495 	wait_queue_head_t transaction_throttle;
496 	wait_queue_head_t transaction_wait;
497 	wait_queue_head_t transaction_blocked_wait;
498 	wait_queue_head_t async_submit_wait;
499 
500 	/*
501 	 * Used to protect the incompat_flags, compat_flags, compat_ro_flags
502 	 * when they are updated.
503 	 *
504 	 * Because we do not clear the flags for ever, so we needn't use
505 	 * the lock on the read side.
506 	 *
507 	 * We also needn't use the lock when we mount the fs, because
508 	 * there is no other task which will update the flag.
509 	 */
510 	spinlock_t super_lock;
511 	struct btrfs_super_block *super_copy;
512 	struct btrfs_super_block *super_for_commit;
513 	struct super_block *sb;
514 	struct inode *btree_inode;
515 	struct mutex tree_log_mutex;
516 	struct mutex transaction_kthread_mutex;
517 	struct mutex cleaner_mutex;
518 	struct mutex chunk_mutex;
519 
520 	/*
521 	 * This is taken to make sure we don't set block groups ro after the
522 	 * free space cache has been allocated on them.
523 	 */
524 	struct mutex ro_block_group_mutex;
525 
526 	/*
527 	 * This is used during read/modify/write to make sure no two ios are
528 	 * trying to mod the same stripe at the same time.
529 	 */
530 	struct btrfs_stripe_hash_table *stripe_hash_table;
531 
532 	/*
533 	 * This protects the ordered operations list only while we are
534 	 * processing all of the entries on it.  This way we make sure the
535 	 * commit code doesn't find the list temporarily empty because another
536 	 * function happens to be doing non-waiting preflush before jumping
537 	 * into the main commit.
538 	 */
539 	struct mutex ordered_operations_mutex;
540 
541 	struct rw_semaphore commit_root_sem;
542 
543 	struct rw_semaphore cleanup_work_sem;
544 
545 	struct rw_semaphore subvol_sem;
546 
547 	spinlock_t trans_lock;
548 	/*
549 	 * The reloc mutex goes with the trans lock, it is taken during commit
550 	 * to protect us from the relocation code.
551 	 */
552 	struct mutex reloc_mutex;
553 
554 	struct list_head trans_list;
555 	struct list_head dead_roots;
556 	struct list_head caching_block_groups;
557 
558 	spinlock_t delayed_iput_lock;
559 	struct list_head delayed_iputs;
560 	atomic_t nr_delayed_iputs;
561 	wait_queue_head_t delayed_iputs_wait;
562 
563 	atomic64_t tree_mod_seq;
564 
565 	/* This protects tree_mod_log and tree_mod_seq_list */
566 	rwlock_t tree_mod_log_lock;
567 	struct rb_root tree_mod_log;
568 	struct list_head tree_mod_seq_list;
569 
570 	atomic_t async_delalloc_pages;
571 
572 	/* This is used to protect the following list -- ordered_roots. */
573 	spinlock_t ordered_root_lock;
574 
575 	/*
576 	 * All fs/file tree roots in which there are data=ordered extents
577 	 * pending writeback are added into this list.
578 	 *
579 	 * These can span multiple transactions and basically include every
580 	 * dirty data page that isn't from nodatacow.
581 	 */
582 	struct list_head ordered_roots;
583 
584 	struct mutex delalloc_root_mutex;
585 	spinlock_t delalloc_root_lock;
586 	/* All fs/file tree roots that have delalloc inodes. */
587 	struct list_head delalloc_roots;
588 
589 	/*
590 	 * There is a pool of worker threads for checksumming during writes and
591 	 * a pool for checksumming after reads.  This is because readers can
592 	 * run with FS locks held, and the writers may be waiting for those
593 	 * locks.  We don't want ordering in the pending list to cause
594 	 * deadlocks, and so the two are serviced separately.
595 	 *
596 	 * A third pool does submit_bio to avoid deadlocking with the other two.
597 	 */
598 	struct btrfs_workqueue *workers;
599 	struct btrfs_workqueue *delalloc_workers;
600 	struct btrfs_workqueue *flush_workers;
601 	struct workqueue_struct *endio_workers;
602 	struct workqueue_struct *endio_meta_workers;
603 	struct workqueue_struct *rmw_workers;
604 	struct workqueue_struct *compressed_write_workers;
605 	struct btrfs_workqueue *endio_write_workers;
606 	struct btrfs_workqueue *endio_freespace_worker;
607 	struct btrfs_workqueue *caching_workers;
608 
609 	/*
610 	 * Fixup workers take dirty pages that didn't properly go through the
611 	 * cow mechanism and make them safe to write.  It happens for the
612 	 * sys_munmap function call path.
613 	 */
614 	struct btrfs_workqueue *fixup_workers;
615 	struct btrfs_workqueue *delayed_workers;
616 
617 	struct task_struct *transaction_kthread;
618 	struct task_struct *cleaner_kthread;
619 	u32 thread_pool_size;
620 
621 	struct kobject *space_info_kobj;
622 	struct kobject *qgroups_kobj;
623 	struct kobject *discard_kobj;
624 
625 	/* Used to keep from writing metadata until there is a nice batch */
626 	struct percpu_counter dirty_metadata_bytes;
627 	struct percpu_counter delalloc_bytes;
628 	struct percpu_counter ordered_bytes;
629 	s32 dirty_metadata_batch;
630 	s32 delalloc_batch;
631 
632 	struct percpu_counter evictable_extent_maps;
633 	u64 extent_map_shrinker_last_root;
634 	u64 extent_map_shrinker_last_ino;
635 
636 	/* Protected by 'trans_lock'. */
637 	struct list_head dirty_cowonly_roots;
638 
639 	struct btrfs_fs_devices *fs_devices;
640 
641 	/*
642 	 * The space_info list is effectively read only after initial setup.
643 	 * It is populated at mount time and cleaned up after all block groups
644 	 * are removed.  RCU is used to protect it.
645 	 */
646 	struct list_head space_info;
647 
648 	struct btrfs_space_info *data_sinfo;
649 
650 	struct reloc_control *reloc_ctl;
651 
652 	/* data_alloc_cluster is only used in ssd_spread mode */
653 	struct btrfs_free_cluster data_alloc_cluster;
654 
655 	/* All metadata allocations go through this cluster. */
656 	struct btrfs_free_cluster meta_alloc_cluster;
657 
658 	/* Auto defrag inodes go here. */
659 	spinlock_t defrag_inodes_lock;
660 	struct rb_root defrag_inodes;
661 	atomic_t defrag_running;
662 
663 	/* Used to protect avail_{data, metadata, system}_alloc_bits */
664 	seqlock_t profiles_lock;
665 	/*
666 	 * These three are in extended format (availability of single chunks is
667 	 * denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other types are denoted
668 	 * by corresponding BTRFS_BLOCK_GROUP_* bits)
669 	 */
670 	u64 avail_data_alloc_bits;
671 	u64 avail_metadata_alloc_bits;
672 	u64 avail_system_alloc_bits;
673 
674 	/* Balance state */
675 	spinlock_t balance_lock;
676 	struct mutex balance_mutex;
677 	atomic_t balance_pause_req;
678 	atomic_t balance_cancel_req;
679 	struct btrfs_balance_control *balance_ctl;
680 	wait_queue_head_t balance_wait_q;
681 
682 	/* Cancellation requests for chunk relocation */
683 	atomic_t reloc_cancel_req;
684 
685 	u32 data_chunk_allocations;
686 	u32 metadata_ratio;
687 
688 	void *bdev_holder;
689 
690 	/* Private scrub information */
691 	struct mutex scrub_lock;
692 	atomic_t scrubs_running;
693 	atomic_t scrub_pause_req;
694 	atomic_t scrubs_paused;
695 	atomic_t scrub_cancel_req;
696 	wait_queue_head_t scrub_pause_wait;
697 	/*
698 	 * The worker pointers are NULL iff the refcount is 0, ie. scrub is not
699 	 * running.
700 	 */
701 	refcount_t scrub_workers_refcnt;
702 	struct workqueue_struct *scrub_workers;
703 	struct btrfs_subpage_info *subpage_info;
704 
705 	struct btrfs_discard_ctl discard_ctl;
706 
707 	/* Is qgroup tracking in a consistent state? */
708 	u64 qgroup_flags;
709 
710 	/* Holds configuration and tracking. Protected by qgroup_lock. */
711 	struct rb_root qgroup_tree;
712 	spinlock_t qgroup_lock;
713 
714 	/*
715 	 * Used to avoid frequently calling ulist_alloc()/ulist_free()
716 	 * when doing qgroup accounting, it must be protected by qgroup_lock.
717 	 */
718 	struct ulist *qgroup_ulist;
719 
720 	/*
721 	 * Protect user change for quota operations. If a transaction is needed,
722 	 * it must be started before locking this lock.
723 	 */
724 	struct mutex qgroup_ioctl_lock;
725 
726 	/* List of dirty qgroups to be written at next commit. */
727 	struct list_head dirty_qgroups;
728 
729 	/* Used by qgroup for an efficient tree traversal. */
730 	u64 qgroup_seq;
731 
732 	/* Qgroup rescan items. */
733 	/* Protects the progress item */
734 	struct mutex qgroup_rescan_lock;
735 	struct btrfs_key qgroup_rescan_progress;
736 	struct btrfs_workqueue *qgroup_rescan_workers;
737 	struct completion qgroup_rescan_completion;
738 	struct btrfs_work qgroup_rescan_work;
739 	/* Protected by qgroup_rescan_lock */
740 	bool qgroup_rescan_running;
741 	u8 qgroup_drop_subtree_thres;
742 	u64 qgroup_enable_gen;
743 
744 	/*
745 	 * If this is not 0, then it indicates a serious filesystem error has
746 	 * happened and it contains that error (negative errno value).
747 	 */
748 	int fs_error;
749 
750 	/* Filesystem state */
751 	unsigned long fs_state;
752 
753 	struct btrfs_delayed_root *delayed_root;
754 
755 	/* Extent buffer radix tree */
756 	spinlock_t buffer_lock;
757 	/* Entries are eb->start / sectorsize */
758 	struct radix_tree_root buffer_radix;
759 
760 	/* Next backup root to be overwritten */
761 	int backup_root_index;
762 
763 	/* Device replace state */
764 	struct btrfs_dev_replace dev_replace;
765 
766 	struct semaphore uuid_tree_rescan_sem;
767 
768 	/* Used to reclaim the metadata space in the background. */
769 	struct work_struct async_reclaim_work;
770 	struct work_struct async_data_reclaim_work;
771 	struct work_struct preempt_reclaim_work;
772 
773 	/* Reclaim partially filled block groups in the background */
774 	struct work_struct reclaim_bgs_work;
775 	/* Protected by unused_bgs_lock. */
776 	struct list_head reclaim_bgs;
777 	int bg_reclaim_threshold;
778 
779 	/* Protects the lists unused_bgs and reclaim_bgs. */
780 	spinlock_t unused_bgs_lock;
781 	/* Protected by unused_bgs_lock. */
782 	struct list_head unused_bgs;
783 	struct mutex unused_bg_unpin_mutex;
784 	/* Protect block groups that are going to be deleted */
785 	struct mutex reclaim_bgs_lock;
786 
787 	/* Cached block sizes */
788 	u32 nodesize;
789 	u32 sectorsize;
790 	/* ilog2 of sectorsize, use to avoid 64bit division */
791 	u32 sectorsize_bits;
792 	u32 csum_size;
793 	u32 csums_per_leaf;
794 	u32 stripesize;
795 
796 	/*
797 	 * Maximum size of an extent. BTRFS_MAX_EXTENT_SIZE on regular
798 	 * filesystem, on zoned it depends on the device constraints.
799 	 */
800 	u64 max_extent_size;
801 
802 	/* Block groups and devices containing active swapfiles. */
803 	spinlock_t swapfile_pins_lock;
804 	struct rb_root swapfile_pins;
805 
806 	struct crypto_shash *csum_shash;
807 
808 	/* Type of exclusive operation running, protected by super_lock */
809 	enum btrfs_exclusive_operation exclusive_operation;
810 
811 	/*
812 	 * Zone size > 0 when in ZONED mode, otherwise it's used for a check
813 	 * if the mode is enabled
814 	 */
815 	u64 zone_size;
816 
817 	/* Constraints for ZONE_APPEND commands: */
818 	struct queue_limits limits;
819 	u64 max_zone_append_size;
820 
821 	struct mutex zoned_meta_io_lock;
822 	spinlock_t treelog_bg_lock;
823 	u64 treelog_bg;
824 
825 	/*
826 	 * Start of the dedicated data relocation block group, protected by
827 	 * relocation_bg_lock.
828 	 */
829 	spinlock_t relocation_bg_lock;
830 	u64 data_reloc_bg;
831 	struct mutex zoned_data_reloc_io_lock;
832 
833 	struct btrfs_block_group *active_meta_bg;
834 	struct btrfs_block_group *active_system_bg;
835 
836 	u64 nr_global_roots;
837 
838 	spinlock_t zone_active_bgs_lock;
839 	struct list_head zone_active_bgs;
840 
841 	/* Updates are not protected by any lock */
842 	struct btrfs_commit_stats commit_stats;
843 
844 	/*
845 	 * Last generation where we dropped a non-relocation root.
846 	 * Use btrfs_set_last_root_drop_gen() and btrfs_get_last_root_drop_gen()
847 	 * to change it and to read it, respectively.
848 	 */
849 	u64 last_root_drop_gen;
850 
851 	/*
852 	 * Annotations for transaction events (structures are empty when
853 	 * compiled without lockdep).
854 	 */
855 	struct lockdep_map btrfs_trans_num_writers_map;
856 	struct lockdep_map btrfs_trans_num_extwriters_map;
857 	struct lockdep_map btrfs_state_change_map[4];
858 	struct lockdep_map btrfs_trans_pending_ordered_map;
859 	struct lockdep_map btrfs_ordered_extent_map;
860 
861 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
862 	spinlock_t ref_verify_lock;
863 	struct rb_root block_tree;
864 #endif
865 
866 #ifdef CONFIG_BTRFS_DEBUG
867 	struct kobject *debug_kobj;
868 	struct list_head allocated_roots;
869 
870 	spinlock_t eb_leak_lock;
871 	struct list_head allocated_ebs;
872 #endif
873 };
874 
875 #define page_to_inode(_page)	(BTRFS_I(_Generic((_page),			\
876 					  struct page *: (_page))->mapping->host))
877 #define folio_to_inode(_folio)	(BTRFS_I(_Generic((_folio),			\
878 					  struct folio *: (_folio))->mapping->host))
879 
880 #define page_to_fs_info(_page)	 (page_to_inode(_page)->root->fs_info)
881 #define folio_to_fs_info(_folio) (folio_to_inode(_folio)->root->fs_info)
882 
883 #define inode_to_fs_info(_inode) (BTRFS_I(_Generic((_inode),			\
884 					   struct inode *: (_inode)))->root->fs_info)
885 
886 static inline u64 btrfs_get_fs_generation(const struct btrfs_fs_info *fs_info)
887 {
888 	return READ_ONCE(fs_info->generation);
889 }
890 
891 static inline void btrfs_set_fs_generation(struct btrfs_fs_info *fs_info, u64 gen)
892 {
893 	WRITE_ONCE(fs_info->generation, gen);
894 }
895 
896 static inline u64 btrfs_get_last_trans_committed(const struct btrfs_fs_info *fs_info)
897 {
898 	return READ_ONCE(fs_info->last_trans_committed);
899 }
900 
901 static inline void btrfs_set_last_trans_committed(struct btrfs_fs_info *fs_info, u64 gen)
902 {
903 	WRITE_ONCE(fs_info->last_trans_committed, gen);
904 }
905 
906 static inline void btrfs_set_last_root_drop_gen(struct btrfs_fs_info *fs_info,
907 						u64 gen)
908 {
909 	WRITE_ONCE(fs_info->last_root_drop_gen, gen);
910 }
911 
912 static inline u64 btrfs_get_last_root_drop_gen(const struct btrfs_fs_info *fs_info)
913 {
914 	return READ_ONCE(fs_info->last_root_drop_gen);
915 }
916 
917 /*
918  * Take the number of bytes to be checksummed and figure out how many leaves
919  * it would require to store the csums for that many bytes.
920  */
921 static inline u64 btrfs_csum_bytes_to_leaves(
922 			const struct btrfs_fs_info *fs_info, u64 csum_bytes)
923 {
924 	const u64 num_csums = csum_bytes >> fs_info->sectorsize_bits;
925 
926 	return DIV_ROUND_UP_ULL(num_csums, fs_info->csums_per_leaf);
927 }
928 
929 /*
930  * Use this if we would be adding new items, as we could split nodes as we cow
931  * down the tree.
932  */
933 static inline u64 btrfs_calc_insert_metadata_size(const struct btrfs_fs_info *fs_info,
934 						  unsigned num_items)
935 {
936 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * 2 * num_items;
937 }
938 
939 /*
940  * Doing a truncate or a modification won't result in new nodes or leaves, just
941  * what we need for COW.
942  */
943 static inline u64 btrfs_calc_metadata_size(const struct btrfs_fs_info *fs_info,
944 						 unsigned num_items)
945 {
946 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * num_items;
947 }
948 
949 #define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r->fs_info) >> 4) - \
950 					sizeof(struct btrfs_item))
951 
952 static inline bool btrfs_is_zoned(const struct btrfs_fs_info *fs_info)
953 {
954 	return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && fs_info->zone_size > 0;
955 }
956 
957 /*
958  * Count how many fs_info->max_extent_size cover the @size
959  */
960 static inline u32 count_max_extents(struct btrfs_fs_info *fs_info, u64 size)
961 {
962 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
963 	if (!fs_info)
964 		return div_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, BTRFS_MAX_EXTENT_SIZE);
965 #endif
966 
967 	return div_u64(size + fs_info->max_extent_size - 1, fs_info->max_extent_size);
968 }
969 
970 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
971 			enum btrfs_exclusive_operation type);
972 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
973 				 enum btrfs_exclusive_operation type);
974 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info);
975 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info);
976 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
977 			  enum btrfs_exclusive_operation op);
978 
979 int btrfs_check_ioctl_vol_args_path(const struct btrfs_ioctl_vol_args *vol_args);
980 
981 /* Compatibility and incompatibility defines */
982 void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
983 			     const char *name);
984 void __btrfs_clear_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
985 			       const char *name);
986 void __btrfs_set_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
987 			      const char *name);
988 void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
989 				const char *name);
990 
991 #define __btrfs_fs_incompat(fs_info, flags)				\
992 	(!!(btrfs_super_incompat_flags((fs_info)->super_copy) & (flags)))
993 
994 #define __btrfs_fs_compat_ro(fs_info, flags)				\
995 	(!!(btrfs_super_compat_ro_flags((fs_info)->super_copy) & (flags)))
996 
997 #define btrfs_set_fs_incompat(__fs_info, opt)				\
998 	__btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
999 
1000 #define btrfs_clear_fs_incompat(__fs_info, opt)				\
1001 	__btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
1002 
1003 #define btrfs_fs_incompat(fs_info, opt)					\
1004 	__btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
1005 
1006 #define btrfs_set_fs_compat_ro(__fs_info, opt)				\
1007 	__btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
1008 
1009 #define btrfs_clear_fs_compat_ro(__fs_info, opt)			\
1010 	__btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
1011 
1012 #define btrfs_fs_compat_ro(fs_info, opt)				\
1013 	__btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
1014 
1015 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
1016 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
1017 #define btrfs_raw_test_opt(o, opt)	((o) & BTRFS_MOUNT_##opt)
1018 #define btrfs_test_opt(fs_info, opt)	((fs_info)->mount_opt & \
1019 					 BTRFS_MOUNT_##opt)
1020 
1021 static inline int btrfs_fs_closing(struct btrfs_fs_info *fs_info)
1022 {
1023 	/* Do it this way so we only ever do one test_bit in the normal case. */
1024 	if (test_bit(BTRFS_FS_CLOSING_START, &fs_info->flags)) {
1025 		if (test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags))
1026 			return 2;
1027 		return 1;
1028 	}
1029 	return 0;
1030 }
1031 
1032 /*
1033  * If we remount the fs to be R/O or umount the fs, the cleaner needn't do
1034  * anything except sleeping. This function is used to check the status of
1035  * the fs.
1036  * We check for BTRFS_FS_STATE_RO to avoid races with a concurrent remount,
1037  * since setting and checking for SB_RDONLY in the superblock's flags is not
1038  * atomic.
1039  */
1040 static inline int btrfs_need_cleaner_sleep(struct btrfs_fs_info *fs_info)
1041 {
1042 	return test_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state) ||
1043 		btrfs_fs_closing(fs_info);
1044 }
1045 
1046 static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info *fs_info)
1047 {
1048 	clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags);
1049 }
1050 
1051 #define BTRFS_FS_ERROR(fs_info)	(READ_ONCE((fs_info)->fs_error))
1052 
1053 #define BTRFS_FS_LOG_CLEANUP_ERROR(fs_info)				\
1054 	(unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR,		\
1055 			   &(fs_info)->fs_state)))
1056 
1057 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1058 
1059 #define EXPORT_FOR_TESTS
1060 
1061 static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
1062 {
1063 	return test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
1064 }
1065 
1066 void btrfs_test_destroy_inode(struct inode *inode);
1067 
1068 #else
1069 
1070 #define EXPORT_FOR_TESTS static
1071 
1072 static inline int btrfs_is_testing(struct btrfs_fs_info *fs_info)
1073 {
1074 	return 0;
1075 }
1076 #endif
1077 
1078 #endif
1079