xref: /linux/fs/btrfs/fs.h (revision 0eb4aaa230d725fa9b1cd758c0f17abca5597af6)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 
3 #ifndef BTRFS_FS_H
4 #define BTRFS_FS_H
5 
6 #include <linux/blkdev.h>
7 #include <linux/sizes.h>
8 #include <linux/time64.h>
9 #include <linux/compiler.h>
10 #include <linux/math.h>
11 #include <linux/atomic.h>
12 #include <linux/percpu_counter.h>
13 #include <linux/completion.h>
14 #include <linux/lockdep.h>
15 #include <linux/spinlock.h>
16 #include <linux/mutex.h>
17 #include <linux/rwsem.h>
18 #include <linux/semaphore.h>
19 #include <linux/list.h>
20 #include <linux/pagemap.h>
21 #include <linux/radix-tree.h>
22 #include <linux/workqueue.h>
23 #include <linux/wait.h>
24 #include <linux/wait_bit.h>
25 #include <linux/sched.h>
26 #include <linux/rbtree.h>
27 #include <uapi/linux/btrfs.h>
28 #include <uapi/linux/btrfs_tree.h>
29 #include "extent-io-tree.h"
30 #include "async-thread.h"
31 #include "block-rsv.h"
32 
33 struct inode;
34 struct super_block;
35 struct kobject;
36 struct reloc_control;
37 struct crypto_shash;
38 struct ulist;
39 struct btrfs_device;
40 struct btrfs_block_group;
41 struct btrfs_root;
42 struct btrfs_fs_devices;
43 struct btrfs_transaction;
44 struct btrfs_delayed_root;
45 struct btrfs_balance_control;
46 struct btrfs_subpage_info;
47 struct btrfs_stripe_hash_table;
48 struct btrfs_space_info;
49 
50 #define BTRFS_MAX_EXTENT_SIZE SZ_128M
51 
52 #define BTRFS_OLDEST_GENERATION	0ULL
53 
54 #define BTRFS_EMPTY_DIR_SIZE 0
55 
56 #define BTRFS_DIRTY_METADATA_THRESH		SZ_32M
57 
58 #define BTRFS_SUPER_INFO_OFFSET			SZ_64K
59 #define BTRFS_SUPER_INFO_SIZE			4096
60 static_assert(sizeof(struct btrfs_super_block) == BTRFS_SUPER_INFO_SIZE);
61 
62 /*
63  * Number of metadata items necessary for an unlink operation:
64  *
65  * 1 for the possible orphan item
66  * 1 for the dir item
67  * 1 for the dir index
68  * 1 for the inode ref
69  * 1 for the inode
70  * 1 for the parent inode
71  */
72 #define BTRFS_UNLINK_METADATA_UNITS		6
73 
74 /*
75  * The reserved space at the beginning of each device.  It covers the primary
76  * super block and leaves space for potential use by other tools like
77  * bootloaders or to lower potential damage of accidental overwrite.
78  */
79 #define BTRFS_DEVICE_RANGE_RESERVED			(SZ_1M)
80 /*
81  * Runtime (in-memory) states of filesystem
82  */
83 enum {
84 	/*
85 	 * Filesystem is being remounted, allow to skip some operations, like
86 	 * defrag
87 	 */
88 	BTRFS_FS_STATE_REMOUNTING,
89 	/* Filesystem in RO mode */
90 	BTRFS_FS_STATE_RO,
91 	/* Track if a transaction abort has been reported on this filesystem */
92 	BTRFS_FS_STATE_TRANS_ABORTED,
93 	/*
94 	 * Bio operations should be blocked on this filesystem because a source
95 	 * or target device is being destroyed as part of a device replace
96 	 */
97 	BTRFS_FS_STATE_DEV_REPLACING,
98 	/* The btrfs_fs_info created for self-tests */
99 	BTRFS_FS_STATE_DUMMY_FS_INFO,
100 
101 	/* Checksum errors are ignored. */
102 	BTRFS_FS_STATE_NO_DATA_CSUMS,
103 	BTRFS_FS_STATE_SKIP_META_CSUMS,
104 
105 	/* Indicates there was an error cleaning up a log tree. */
106 	BTRFS_FS_STATE_LOG_CLEANUP_ERROR,
107 
108 	BTRFS_FS_STATE_COUNT
109 };
110 
111 enum {
112 	BTRFS_FS_CLOSING_START,
113 	BTRFS_FS_CLOSING_DONE,
114 	BTRFS_FS_LOG_RECOVERING,
115 	BTRFS_FS_OPEN,
116 	BTRFS_FS_QUOTA_ENABLED,
117 	BTRFS_FS_UPDATE_UUID_TREE_GEN,
118 	BTRFS_FS_CREATING_FREE_SPACE_TREE,
119 	BTRFS_FS_BTREE_ERR,
120 	BTRFS_FS_LOG1_ERR,
121 	BTRFS_FS_LOG2_ERR,
122 	BTRFS_FS_QUOTA_OVERRIDE,
123 	/* Used to record internally whether fs has been frozen */
124 	BTRFS_FS_FROZEN,
125 	/*
126 	 * Indicate that balance has been set up from the ioctl and is in the
127 	 * main phase. The fs_info::balance_ctl is initialized.
128 	 */
129 	BTRFS_FS_BALANCE_RUNNING,
130 
131 	/*
132 	 * Indicate that relocation of a chunk has started, it's set per chunk
133 	 * and is toggled between chunks.
134 	 */
135 	BTRFS_FS_RELOC_RUNNING,
136 
137 	/* Indicate that the cleaner thread is awake and doing something. */
138 	BTRFS_FS_CLEANER_RUNNING,
139 
140 	/*
141 	 * The checksumming has an optimized version and is considered fast,
142 	 * so we don't need to offload checksums to workqueues.
143 	 */
144 	BTRFS_FS_CSUM_IMPL_FAST,
145 
146 	/* Indicate that the discard workqueue can service discards. */
147 	BTRFS_FS_DISCARD_RUNNING,
148 
149 	/* Indicate that we need to cleanup space cache v1 */
150 	BTRFS_FS_CLEANUP_SPACE_CACHE_V1,
151 
152 	/* Indicate that we can't trust the free space tree for caching yet */
153 	BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED,
154 
155 	/* Indicate whether there are any tree modification log users */
156 	BTRFS_FS_TREE_MOD_LOG_USERS,
157 
158 	/* Indicate that we want the transaction kthread to commit right now. */
159 	BTRFS_FS_COMMIT_TRANS,
160 
161 	/* Indicate we have half completed snapshot deletions pending. */
162 	BTRFS_FS_UNFINISHED_DROPS,
163 
164 	/* Indicate we have to finish a zone to do next allocation. */
165 	BTRFS_FS_NEED_ZONE_FINISH,
166 
167 	/* Indicate that we want to commit the transaction. */
168 	BTRFS_FS_NEED_TRANS_COMMIT,
169 
170 	/* This is set when active zone tracking is needed. */
171 	BTRFS_FS_ACTIVE_ZONE_TRACKING,
172 
173 	/*
174 	 * Indicate if we have some features changed, this is mostly for
175 	 * cleaner thread to update the sysfs interface.
176 	 */
177 	BTRFS_FS_FEATURE_CHANGED,
178 
179 	/*
180 	 * Indicate that we have found a tree block which is only aligned to
181 	 * sectorsize, but not to nodesize.  This should be rare nowadays.
182 	 */
183 	BTRFS_FS_UNALIGNED_TREE_BLOCK,
184 
185 #if BITS_PER_LONG == 32
186 	/* Indicate if we have error/warn message printed on 32bit systems */
187 	BTRFS_FS_32BIT_ERROR,
188 	BTRFS_FS_32BIT_WARN,
189 #endif
190 };
191 
192 /*
193  * Flags for mount options.
194  *
195  * Note: don't forget to add new options to btrfs_show_options()
196  */
197 enum {
198 	BTRFS_MOUNT_NODATASUM			= (1ULL << 0),
199 	BTRFS_MOUNT_NODATACOW			= (1ULL << 1),
200 	BTRFS_MOUNT_NOBARRIER			= (1ULL << 2),
201 	BTRFS_MOUNT_SSD				= (1ULL << 3),
202 	BTRFS_MOUNT_DEGRADED			= (1ULL << 4),
203 	BTRFS_MOUNT_COMPRESS			= (1ULL << 5),
204 	BTRFS_MOUNT_NOTREELOG			= (1ULL << 6),
205 	BTRFS_MOUNT_FLUSHONCOMMIT		= (1ULL << 7),
206 	BTRFS_MOUNT_SSD_SPREAD			= (1ULL << 8),
207 	BTRFS_MOUNT_NOSSD			= (1ULL << 9),
208 	BTRFS_MOUNT_DISCARD_SYNC		= (1ULL << 10),
209 	BTRFS_MOUNT_FORCE_COMPRESS		= (1ULL << 11),
210 	BTRFS_MOUNT_SPACE_CACHE			= (1ULL << 12),
211 	BTRFS_MOUNT_CLEAR_CACHE			= (1ULL << 13),
212 	BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED	= (1ULL << 14),
213 	BTRFS_MOUNT_ENOSPC_DEBUG		= (1ULL << 15),
214 	BTRFS_MOUNT_AUTO_DEFRAG			= (1ULL << 16),
215 	BTRFS_MOUNT_USEBACKUPROOT		= (1ULL << 17),
216 	BTRFS_MOUNT_SKIP_BALANCE		= (1ULL << 18),
217 	BTRFS_MOUNT_PANIC_ON_FATAL_ERROR	= (1ULL << 19),
218 	BTRFS_MOUNT_RESCAN_UUID_TREE		= (1ULL << 20),
219 	BTRFS_MOUNT_FRAGMENT_DATA		= (1ULL << 21),
220 	BTRFS_MOUNT_FRAGMENT_METADATA		= (1ULL << 22),
221 	BTRFS_MOUNT_FREE_SPACE_TREE		= (1ULL << 23),
222 	BTRFS_MOUNT_NOLOGREPLAY			= (1ULL << 24),
223 	BTRFS_MOUNT_REF_VERIFY			= (1ULL << 25),
224 	BTRFS_MOUNT_DISCARD_ASYNC		= (1ULL << 26),
225 	BTRFS_MOUNT_IGNOREBADROOTS		= (1ULL << 27),
226 	BTRFS_MOUNT_IGNOREDATACSUMS		= (1ULL << 28),
227 	BTRFS_MOUNT_NODISCARD			= (1ULL << 29),
228 	BTRFS_MOUNT_NOSPACECACHE		= (1ULL << 30),
229 	BTRFS_MOUNT_IGNOREMETACSUMS		= (1ULL << 31),
230 	BTRFS_MOUNT_IGNORESUPERFLAGS		= (1ULL << 32),
231 };
232 
233 /*
234  * Compat flags that we support.  If any incompat flags are set other than the
235  * ones specified below then we will fail to mount
236  */
237 #define BTRFS_FEATURE_COMPAT_SUPP		0ULL
238 #define BTRFS_FEATURE_COMPAT_SAFE_SET		0ULL
239 #define BTRFS_FEATURE_COMPAT_SAFE_CLEAR		0ULL
240 
241 #define BTRFS_FEATURE_COMPAT_RO_SUPP			\
242 	(BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE |	\
243 	 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID | \
244 	 BTRFS_FEATURE_COMPAT_RO_VERITY |		\
245 	 BTRFS_FEATURE_COMPAT_RO_BLOCK_GROUP_TREE)
246 
247 #define BTRFS_FEATURE_COMPAT_RO_SAFE_SET	0ULL
248 #define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR	0ULL
249 
250 #define BTRFS_FEATURE_INCOMPAT_SUPP_STABLE		\
251 	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
252 	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
253 	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
254 	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
255 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
256 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD |		\
257 	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
258 	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF |		\
259 	 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA |	\
260 	 BTRFS_FEATURE_INCOMPAT_NO_HOLES	|	\
261 	 BTRFS_FEATURE_INCOMPAT_METADATA_UUID	|	\
262 	 BTRFS_FEATURE_INCOMPAT_RAID1C34	|	\
263 	 BTRFS_FEATURE_INCOMPAT_ZONED		|	\
264 	 BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA)
265 
266 #ifdef CONFIG_BTRFS_EXPERIMENTAL
267 	/*
268 	 * Features under developmen like Extent tree v2 support is enabled
269 	 * only under CONFIG_BTRFS_EXPERIMENTAL
270 	 */
271 #define BTRFS_FEATURE_INCOMPAT_SUPP		\
272 	(BTRFS_FEATURE_INCOMPAT_SUPP_STABLE |	\
273 	 BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE | \
274 	 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2)
275 
276 #else
277 
278 #define BTRFS_FEATURE_INCOMPAT_SUPP		\
279 	(BTRFS_FEATURE_INCOMPAT_SUPP_STABLE)
280 
281 #endif
282 
283 #define BTRFS_FEATURE_INCOMPAT_SAFE_SET			\
284 	(BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
285 #define BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR		0ULL
286 
287 #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
288 #define BTRFS_DEFAULT_MAX_INLINE	(2048)
289 
290 struct btrfs_dev_replace {
291 	/* See #define above */
292 	u64 replace_state;
293 	/* Seconds since 1-Jan-1970 */
294 	time64_t time_started;
295 	/* Seconds since 1-Jan-1970 */
296 	time64_t time_stopped;
297 	atomic64_t num_write_errors;
298 	atomic64_t num_uncorrectable_read_errors;
299 
300 	u64 cursor_left;
301 	u64 committed_cursor_left;
302 	u64 cursor_left_last_write_of_item;
303 	u64 cursor_right;
304 
305 	/* See #define above */
306 	u64 cont_reading_from_srcdev_mode;
307 
308 	int is_valid;
309 	int item_needs_writeback;
310 	struct btrfs_device *srcdev;
311 	struct btrfs_device *tgtdev;
312 
313 	struct mutex lock_finishing_cancel_unmount;
314 	struct rw_semaphore rwsem;
315 
316 	struct btrfs_scrub_progress scrub_progress;
317 
318 	struct percpu_counter bio_counter;
319 	wait_queue_head_t replace_wait;
320 
321 	struct task_struct *replace_task;
322 };
323 
324 /*
325  * Free clusters are used to claim free space in relatively large chunks,
326  * allowing us to do less seeky writes. They are used for all metadata
327  * allocations. In ssd_spread mode they are also used for data allocations.
328  */
329 struct btrfs_free_cluster {
330 	spinlock_t lock;
331 	spinlock_t refill_lock;
332 	struct rb_root root;
333 
334 	/* Largest extent in this cluster */
335 	u64 max_size;
336 
337 	/* First extent starting offset */
338 	u64 window_start;
339 
340 	/* We did a full search and couldn't create a cluster */
341 	bool fragmented;
342 
343 	struct btrfs_block_group *block_group;
344 	/*
345 	 * When a cluster is allocated from a block group, we put the cluster
346 	 * onto a list in the block group so that it can be freed before the
347 	 * block group is freed.
348 	 */
349 	struct list_head block_group_list;
350 };
351 
352 /* Discard control. */
353 /*
354  * Async discard uses multiple lists to differentiate the discard filter
355  * parameters.  Index 0 is for completely free block groups where we need to
356  * ensure the entire block group is trimmed without being lossy.  Indices
357  * afterwards represent monotonically decreasing discard filter sizes to
358  * prioritize what should be discarded next.
359  */
360 #define BTRFS_NR_DISCARD_LISTS		3
361 #define BTRFS_DISCARD_INDEX_UNUSED	0
362 #define BTRFS_DISCARD_INDEX_START	1
363 
364 struct btrfs_discard_ctl {
365 	struct workqueue_struct *discard_workers;
366 	struct delayed_work work;
367 	spinlock_t lock;
368 	struct btrfs_block_group *block_group;
369 	struct list_head discard_list[BTRFS_NR_DISCARD_LISTS];
370 	u64 prev_discard;
371 	u64 prev_discard_time;
372 	atomic_t discardable_extents;
373 	atomic64_t discardable_bytes;
374 	u64 max_discard_size;
375 	u64 delay_ms;
376 	u32 iops_limit;
377 	u32 kbps_limit;
378 	u64 discard_extent_bytes;
379 	u64 discard_bitmap_bytes;
380 	atomic64_t discard_bytes_saved;
381 };
382 
383 /*
384  * Exclusive operations (device replace, resize, device add/remove, balance)
385  */
386 enum btrfs_exclusive_operation {
387 	BTRFS_EXCLOP_NONE,
388 	BTRFS_EXCLOP_BALANCE_PAUSED,
389 	BTRFS_EXCLOP_BALANCE,
390 	BTRFS_EXCLOP_DEV_ADD,
391 	BTRFS_EXCLOP_DEV_REMOVE,
392 	BTRFS_EXCLOP_DEV_REPLACE,
393 	BTRFS_EXCLOP_RESIZE,
394 	BTRFS_EXCLOP_SWAP_ACTIVATE,
395 };
396 
397 /* Store data about transaction commits, exported via sysfs. */
398 struct btrfs_commit_stats {
399 	/* Total number of commits */
400 	u64 commit_count;
401 	/* The maximum commit duration so far in ns */
402 	u64 max_commit_dur;
403 	/* The last commit duration in ns */
404 	u64 last_commit_dur;
405 	/* The total commit duration in ns */
406 	u64 total_commit_dur;
407 };
408 
409 struct btrfs_fs_info {
410 	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
411 	unsigned long flags;
412 	struct btrfs_root *tree_root;
413 	struct btrfs_root *chunk_root;
414 	struct btrfs_root *dev_root;
415 	struct btrfs_root *fs_root;
416 	struct btrfs_root *quota_root;
417 	struct btrfs_root *uuid_root;
418 	struct btrfs_root *data_reloc_root;
419 	struct btrfs_root *block_group_root;
420 	struct btrfs_root *stripe_root;
421 
422 	/* The log root tree is a directory of all the other log roots */
423 	struct btrfs_root *log_root_tree;
424 
425 	/* The tree that holds the global roots (csum, extent, etc) */
426 	rwlock_t global_root_lock;
427 	struct rb_root global_root_tree;
428 
429 	spinlock_t fs_roots_radix_lock;
430 	struct radix_tree_root fs_roots_radix;
431 
432 	/* Block group cache stuff */
433 	rwlock_t block_group_cache_lock;
434 	struct rb_root_cached block_group_cache_tree;
435 
436 	/* Keep track of unallocated space */
437 	atomic64_t free_chunk_space;
438 
439 	/* Track ranges which are used by log trees blocks/logged data extents */
440 	struct extent_io_tree excluded_extents;
441 
442 	/* logical->physical extent mapping */
443 	struct rb_root_cached mapping_tree;
444 	rwlock_t mapping_tree_lock;
445 
446 	/*
447 	 * Block reservation for extent, checksum, root tree and delayed dir
448 	 * index item.
449 	 */
450 	struct btrfs_block_rsv global_block_rsv;
451 	/* Block reservation for metadata operations */
452 	struct btrfs_block_rsv trans_block_rsv;
453 	/* Block reservation for chunk tree */
454 	struct btrfs_block_rsv chunk_block_rsv;
455 	/* Block reservation for delayed operations */
456 	struct btrfs_block_rsv delayed_block_rsv;
457 	/* Block reservation for delayed refs */
458 	struct btrfs_block_rsv delayed_refs_rsv;
459 
460 	struct btrfs_block_rsv empty_block_rsv;
461 
462 	/*
463 	 * Updated while holding the lock 'trans_lock'. Due to the life cycle of
464 	 * a transaction, it can be directly read while holding a transaction
465 	 * handle, everywhere else must be read with btrfs_get_fs_generation().
466 	 * Should always be updated using btrfs_set_fs_generation().
467 	 */
468 	u64 generation;
469 	/*
470 	 * Always use btrfs_get_last_trans_committed() and
471 	 * btrfs_set_last_trans_committed() to read and update this field.
472 	 */
473 	u64 last_trans_committed;
474 	/*
475 	 * Generation of the last transaction used for block group relocation
476 	 * since the filesystem was last mounted (or 0 if none happened yet).
477 	 * Must be written and read while holding btrfs_fs_info::commit_root_sem.
478 	 */
479 	u64 last_reloc_trans;
480 
481 	/*
482 	 * This is updated to the current trans every time a full commit is
483 	 * required instead of the faster short fsync log commits
484 	 */
485 	u64 last_trans_log_full_commit;
486 	unsigned long long mount_opt;
487 
488 	unsigned long compress_type:4;
489 	unsigned int compress_level;
490 	u32 commit_interval;
491 	/*
492 	 * It is a suggestive number, the read side is safe even it gets a
493 	 * wrong number because we will write out the data into a regular
494 	 * extent. The write side(mount/remount) is under ->s_umount lock,
495 	 * so it is also safe.
496 	 */
497 	u64 max_inline;
498 
499 	struct btrfs_transaction *running_transaction;
500 	wait_queue_head_t transaction_throttle;
501 	wait_queue_head_t transaction_wait;
502 	wait_queue_head_t transaction_blocked_wait;
503 	wait_queue_head_t async_submit_wait;
504 
505 	/*
506 	 * Used to protect the incompat_flags, compat_flags, compat_ro_flags
507 	 * when they are updated.
508 	 *
509 	 * Because we do not clear the flags for ever, so we needn't use
510 	 * the lock on the read side.
511 	 *
512 	 * We also needn't use the lock when we mount the fs, because
513 	 * there is no other task which will update the flag.
514 	 */
515 	spinlock_t super_lock;
516 	struct btrfs_super_block *super_copy;
517 	struct btrfs_super_block *super_for_commit;
518 	struct super_block *sb;
519 	struct inode *btree_inode;
520 	struct mutex tree_log_mutex;
521 	struct mutex transaction_kthread_mutex;
522 	struct mutex cleaner_mutex;
523 	struct mutex chunk_mutex;
524 
525 	/*
526 	 * This is taken to make sure we don't set block groups ro after the
527 	 * free space cache has been allocated on them.
528 	 */
529 	struct mutex ro_block_group_mutex;
530 
531 	/*
532 	 * This is used during read/modify/write to make sure no two ios are
533 	 * trying to mod the same stripe at the same time.
534 	 */
535 	struct btrfs_stripe_hash_table *stripe_hash_table;
536 
537 	/*
538 	 * This protects the ordered operations list only while we are
539 	 * processing all of the entries on it.  This way we make sure the
540 	 * commit code doesn't find the list temporarily empty because another
541 	 * function happens to be doing non-waiting preflush before jumping
542 	 * into the main commit.
543 	 */
544 	struct mutex ordered_operations_mutex;
545 
546 	struct rw_semaphore commit_root_sem;
547 
548 	struct rw_semaphore cleanup_work_sem;
549 
550 	struct rw_semaphore subvol_sem;
551 
552 	spinlock_t trans_lock;
553 	/*
554 	 * The reloc mutex goes with the trans lock, it is taken during commit
555 	 * to protect us from the relocation code.
556 	 */
557 	struct mutex reloc_mutex;
558 
559 	struct list_head trans_list;
560 	struct list_head dead_roots;
561 	struct list_head caching_block_groups;
562 
563 	spinlock_t delayed_iput_lock;
564 	struct list_head delayed_iputs;
565 	atomic_t nr_delayed_iputs;
566 	wait_queue_head_t delayed_iputs_wait;
567 
568 	atomic64_t tree_mod_seq;
569 
570 	/* This protects tree_mod_log and tree_mod_seq_list */
571 	rwlock_t tree_mod_log_lock;
572 	struct rb_root tree_mod_log;
573 	struct list_head tree_mod_seq_list;
574 
575 	atomic_t async_delalloc_pages;
576 
577 	/* This is used to protect the following list -- ordered_roots. */
578 	spinlock_t ordered_root_lock;
579 
580 	/*
581 	 * All fs/file tree roots in which there are data=ordered extents
582 	 * pending writeback are added into this list.
583 	 *
584 	 * These can span multiple transactions and basically include every
585 	 * dirty data page that isn't from nodatacow.
586 	 */
587 	struct list_head ordered_roots;
588 
589 	struct mutex delalloc_root_mutex;
590 	spinlock_t delalloc_root_lock;
591 	/* All fs/file tree roots that have delalloc inodes. */
592 	struct list_head delalloc_roots;
593 
594 	/*
595 	 * There is a pool of worker threads for checksumming during writes and
596 	 * a pool for checksumming after reads.  This is because readers can
597 	 * run with FS locks held, and the writers may be waiting for those
598 	 * locks.  We don't want ordering in the pending list to cause
599 	 * deadlocks, and so the two are serviced separately.
600 	 *
601 	 * A third pool does submit_bio to avoid deadlocking with the other two.
602 	 */
603 	struct btrfs_workqueue *workers;
604 	struct btrfs_workqueue *delalloc_workers;
605 	struct btrfs_workqueue *flush_workers;
606 	struct workqueue_struct *endio_workers;
607 	struct workqueue_struct *endio_meta_workers;
608 	struct workqueue_struct *rmw_workers;
609 	struct workqueue_struct *compressed_write_workers;
610 	struct btrfs_workqueue *endio_write_workers;
611 	struct btrfs_workqueue *endio_freespace_worker;
612 	struct btrfs_workqueue *caching_workers;
613 
614 	/*
615 	 * Fixup workers take dirty pages that didn't properly go through the
616 	 * cow mechanism and make them safe to write.  It happens for the
617 	 * sys_munmap function call path.
618 	 */
619 	struct btrfs_workqueue *fixup_workers;
620 	struct btrfs_workqueue *delayed_workers;
621 
622 	struct task_struct *transaction_kthread;
623 	struct task_struct *cleaner_kthread;
624 	u32 thread_pool_size;
625 
626 	struct kobject *space_info_kobj;
627 	struct kobject *qgroups_kobj;
628 	struct kobject *discard_kobj;
629 
630 	/* Track the number of blocks (sectors) read by the filesystem. */
631 	struct percpu_counter stats_read_blocks;
632 
633 	/* Used to keep from writing metadata until there is a nice batch */
634 	struct percpu_counter dirty_metadata_bytes;
635 	struct percpu_counter delalloc_bytes;
636 	struct percpu_counter ordered_bytes;
637 	s32 dirty_metadata_batch;
638 	s32 delalloc_batch;
639 
640 	struct percpu_counter evictable_extent_maps;
641 	u64 em_shrinker_last_root;
642 	u64 em_shrinker_last_ino;
643 	atomic64_t em_shrinker_nr_to_scan;
644 	struct work_struct em_shrinker_work;
645 
646 	/* Protected by 'trans_lock'. */
647 	struct list_head dirty_cowonly_roots;
648 
649 	struct btrfs_fs_devices *fs_devices;
650 
651 	/*
652 	 * The space_info list is effectively read only after initial setup.
653 	 * It is populated at mount time and cleaned up after all block groups
654 	 * are removed.  RCU is used to protect it.
655 	 */
656 	struct list_head space_info;
657 
658 	struct btrfs_space_info *data_sinfo;
659 
660 	struct reloc_control *reloc_ctl;
661 
662 	/* data_alloc_cluster is only used in ssd_spread mode */
663 	struct btrfs_free_cluster data_alloc_cluster;
664 
665 	/* All metadata allocations go through this cluster. */
666 	struct btrfs_free_cluster meta_alloc_cluster;
667 
668 	/* Auto defrag inodes go here. */
669 	spinlock_t defrag_inodes_lock;
670 	struct rb_root defrag_inodes;
671 	atomic_t defrag_running;
672 
673 	/* Used to protect avail_{data, metadata, system}_alloc_bits */
674 	seqlock_t profiles_lock;
675 	/*
676 	 * These three are in extended format (availability of single chunks is
677 	 * denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other types are denoted
678 	 * by corresponding BTRFS_BLOCK_GROUP_* bits)
679 	 */
680 	u64 avail_data_alloc_bits;
681 	u64 avail_metadata_alloc_bits;
682 	u64 avail_system_alloc_bits;
683 
684 	/* Balance state */
685 	spinlock_t balance_lock;
686 	struct mutex balance_mutex;
687 	atomic_t balance_pause_req;
688 	atomic_t balance_cancel_req;
689 	struct btrfs_balance_control *balance_ctl;
690 	wait_queue_head_t balance_wait_q;
691 
692 	/* Cancellation requests for chunk relocation */
693 	atomic_t reloc_cancel_req;
694 
695 	u32 data_chunk_allocations;
696 	u32 metadata_ratio;
697 
698 	void *bdev_holder;
699 
700 	/* Private scrub information */
701 	struct mutex scrub_lock;
702 	atomic_t scrubs_running;
703 	atomic_t scrub_pause_req;
704 	atomic_t scrubs_paused;
705 	atomic_t scrub_cancel_req;
706 	wait_queue_head_t scrub_pause_wait;
707 	/*
708 	 * The worker pointers are NULL iff the refcount is 0, ie. scrub is not
709 	 * running.
710 	 */
711 	refcount_t scrub_workers_refcnt;
712 	u32 sectors_per_page;
713 	struct workqueue_struct *scrub_workers;
714 
715 	struct btrfs_discard_ctl discard_ctl;
716 
717 	/* Is qgroup tracking in a consistent state? */
718 	u64 qgroup_flags;
719 
720 	/* Holds configuration and tracking. Protected by qgroup_lock. */
721 	struct rb_root qgroup_tree;
722 	spinlock_t qgroup_lock;
723 
724 	/*
725 	 * Used to avoid frequently calling ulist_alloc()/ulist_free()
726 	 * when doing qgroup accounting, it must be protected by qgroup_lock.
727 	 */
728 	struct ulist *qgroup_ulist;
729 
730 	/*
731 	 * Protect user change for quota operations. If a transaction is needed,
732 	 * it must be started before locking this lock.
733 	 */
734 	struct mutex qgroup_ioctl_lock;
735 
736 	/* List of dirty qgroups to be written at next commit. */
737 	struct list_head dirty_qgroups;
738 
739 	/* Used by qgroup for an efficient tree traversal. */
740 	u64 qgroup_seq;
741 
742 	/* Qgroup rescan items. */
743 	/* Protects the progress item */
744 	struct mutex qgroup_rescan_lock;
745 	struct btrfs_key qgroup_rescan_progress;
746 	struct btrfs_workqueue *qgroup_rescan_workers;
747 	struct completion qgroup_rescan_completion;
748 	struct btrfs_work qgroup_rescan_work;
749 	/* Protected by qgroup_rescan_lock */
750 	bool qgroup_rescan_running;
751 	u8 qgroup_drop_subtree_thres;
752 	u64 qgroup_enable_gen;
753 
754 	/*
755 	 * If this is not 0, then it indicates a serious filesystem error has
756 	 * happened and it contains that error (negative errno value).
757 	 */
758 	int fs_error;
759 
760 	/* Filesystem state */
761 	unsigned long fs_state;
762 
763 	struct btrfs_delayed_root *delayed_root;
764 
765 	/* Extent buffer radix tree */
766 	spinlock_t buffer_lock;
767 	/* Entries are eb->start / sectorsize */
768 	struct radix_tree_root buffer_radix;
769 
770 	/* Next backup root to be overwritten */
771 	int backup_root_index;
772 
773 	/* Device replace state */
774 	struct btrfs_dev_replace dev_replace;
775 
776 	struct semaphore uuid_tree_rescan_sem;
777 
778 	/* Used to reclaim the metadata space in the background. */
779 	struct work_struct async_reclaim_work;
780 	struct work_struct async_data_reclaim_work;
781 	struct work_struct preempt_reclaim_work;
782 
783 	/* Reclaim partially filled block groups in the background */
784 	struct work_struct reclaim_bgs_work;
785 	/* Protected by unused_bgs_lock. */
786 	struct list_head reclaim_bgs;
787 	int bg_reclaim_threshold;
788 
789 	/* Protects the lists unused_bgs and reclaim_bgs. */
790 	spinlock_t unused_bgs_lock;
791 	/* Protected by unused_bgs_lock. */
792 	struct list_head unused_bgs;
793 	struct mutex unused_bg_unpin_mutex;
794 	/* Protect block groups that are going to be deleted */
795 	struct mutex reclaim_bgs_lock;
796 
797 	/* Cached block sizes */
798 	u32 nodesize;
799 	u32 sectorsize;
800 	/* ilog2 of sectorsize, use to avoid 64bit division */
801 	u32 sectorsize_bits;
802 	u32 csum_size;
803 	u32 csums_per_leaf;
804 	u32 stripesize;
805 
806 	/*
807 	 * Maximum size of an extent. BTRFS_MAX_EXTENT_SIZE on regular
808 	 * filesystem, on zoned it depends on the device constraints.
809 	 */
810 	u64 max_extent_size;
811 
812 	/* Block groups and devices containing active swapfiles. */
813 	spinlock_t swapfile_pins_lock;
814 	struct rb_root swapfile_pins;
815 
816 	struct crypto_shash *csum_shash;
817 
818 	/* Type of exclusive operation running, protected by super_lock */
819 	enum btrfs_exclusive_operation exclusive_operation;
820 
821 	/*
822 	 * Zone size > 0 when in ZONED mode, otherwise it's used for a check
823 	 * if the mode is enabled
824 	 */
825 	u64 zone_size;
826 
827 	/* Constraints for ZONE_APPEND commands: */
828 	struct queue_limits limits;
829 	u64 max_zone_append_size;
830 
831 	struct mutex zoned_meta_io_lock;
832 	spinlock_t treelog_bg_lock;
833 	u64 treelog_bg;
834 
835 	/*
836 	 * Start of the dedicated data relocation block group, protected by
837 	 * relocation_bg_lock.
838 	 */
839 	spinlock_t relocation_bg_lock;
840 	u64 data_reloc_bg;
841 	struct mutex zoned_data_reloc_io_lock;
842 
843 	struct btrfs_block_group *active_meta_bg;
844 	struct btrfs_block_group *active_system_bg;
845 
846 	u64 nr_global_roots;
847 
848 	spinlock_t zone_active_bgs_lock;
849 	struct list_head zone_active_bgs;
850 
851 	/* Updates are not protected by any lock */
852 	struct btrfs_commit_stats commit_stats;
853 
854 	/*
855 	 * Last generation where we dropped a non-relocation root.
856 	 * Use btrfs_set_last_root_drop_gen() and btrfs_get_last_root_drop_gen()
857 	 * to change it and to read it, respectively.
858 	 */
859 	u64 last_root_drop_gen;
860 
861 	/*
862 	 * Annotations for transaction events (structures are empty when
863 	 * compiled without lockdep).
864 	 */
865 	struct lockdep_map btrfs_trans_num_writers_map;
866 	struct lockdep_map btrfs_trans_num_extwriters_map;
867 	struct lockdep_map btrfs_state_change_map[4];
868 	struct lockdep_map btrfs_trans_pending_ordered_map;
869 	struct lockdep_map btrfs_ordered_extent_map;
870 
871 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
872 	spinlock_t ref_verify_lock;
873 	struct rb_root block_tree;
874 #endif
875 
876 #ifdef CONFIG_BTRFS_DEBUG
877 	struct kobject *debug_kobj;
878 	struct list_head allocated_roots;
879 
880 	spinlock_t eb_leak_lock;
881 	struct list_head allocated_ebs;
882 #endif
883 };
884 
885 #define folio_to_inode(_folio)	(BTRFS_I(_Generic((_folio),			\
886 					  struct folio *: (_folio))->mapping->host))
887 
888 #define folio_to_fs_info(_folio) (folio_to_inode(_folio)->root->fs_info)
889 
890 #define inode_to_fs_info(_inode) (BTRFS_I(_Generic((_inode),			\
891 					   struct inode *: (_inode)))->root->fs_info)
892 
btrfs_alloc_write_mask(struct address_space * mapping)893 static inline gfp_t btrfs_alloc_write_mask(struct address_space *mapping)
894 {
895 	return mapping_gfp_constraint(mapping, ~__GFP_FS);
896 }
897 
btrfs_get_fs_generation(const struct btrfs_fs_info * fs_info)898 static inline u64 btrfs_get_fs_generation(const struct btrfs_fs_info *fs_info)
899 {
900 	return READ_ONCE(fs_info->generation);
901 }
902 
btrfs_set_fs_generation(struct btrfs_fs_info * fs_info,u64 gen)903 static inline void btrfs_set_fs_generation(struct btrfs_fs_info *fs_info, u64 gen)
904 {
905 	WRITE_ONCE(fs_info->generation, gen);
906 }
907 
btrfs_get_last_trans_committed(const struct btrfs_fs_info * fs_info)908 static inline u64 btrfs_get_last_trans_committed(const struct btrfs_fs_info *fs_info)
909 {
910 	return READ_ONCE(fs_info->last_trans_committed);
911 }
912 
btrfs_set_last_trans_committed(struct btrfs_fs_info * fs_info,u64 gen)913 static inline void btrfs_set_last_trans_committed(struct btrfs_fs_info *fs_info, u64 gen)
914 {
915 	WRITE_ONCE(fs_info->last_trans_committed, gen);
916 }
917 
btrfs_set_last_root_drop_gen(struct btrfs_fs_info * fs_info,u64 gen)918 static inline void btrfs_set_last_root_drop_gen(struct btrfs_fs_info *fs_info,
919 						u64 gen)
920 {
921 	WRITE_ONCE(fs_info->last_root_drop_gen, gen);
922 }
923 
btrfs_get_last_root_drop_gen(const struct btrfs_fs_info * fs_info)924 static inline u64 btrfs_get_last_root_drop_gen(const struct btrfs_fs_info *fs_info)
925 {
926 	return READ_ONCE(fs_info->last_root_drop_gen);
927 }
928 
929 /*
930  * Take the number of bytes to be checksummed and figure out how many leaves
931  * it would require to store the csums for that many bytes.
932  */
btrfs_csum_bytes_to_leaves(const struct btrfs_fs_info * fs_info,u64 csum_bytes)933 static inline u64 btrfs_csum_bytes_to_leaves(
934 			const struct btrfs_fs_info *fs_info, u64 csum_bytes)
935 {
936 	const u64 num_csums = csum_bytes >> fs_info->sectorsize_bits;
937 
938 	return DIV_ROUND_UP_ULL(num_csums, fs_info->csums_per_leaf);
939 }
940 
941 /*
942  * Use this if we would be adding new items, as we could split nodes as we cow
943  * down the tree.
944  */
btrfs_calc_insert_metadata_size(const struct btrfs_fs_info * fs_info,unsigned num_items)945 static inline u64 btrfs_calc_insert_metadata_size(const struct btrfs_fs_info *fs_info,
946 						  unsigned num_items)
947 {
948 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * 2 * num_items;
949 }
950 
951 /*
952  * Doing a truncate or a modification won't result in new nodes or leaves, just
953  * what we need for COW.
954  */
btrfs_calc_metadata_size(const struct btrfs_fs_info * fs_info,unsigned num_items)955 static inline u64 btrfs_calc_metadata_size(const struct btrfs_fs_info *fs_info,
956 						 unsigned num_items)
957 {
958 	return (u64)fs_info->nodesize * BTRFS_MAX_LEVEL * num_items;
959 }
960 
961 #define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r->fs_info) >> 4) - \
962 					sizeof(struct btrfs_item))
963 
964 #define BTRFS_BYTES_TO_BLKS(fs_info, bytes) ((bytes) >> (fs_info)->sectorsize_bits)
965 
btrfs_is_zoned(const struct btrfs_fs_info * fs_info)966 static inline bool btrfs_is_zoned(const struct btrfs_fs_info *fs_info)
967 {
968 	return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && fs_info->zone_size > 0;
969 }
970 
971 /*
972  * Count how many fs_info->max_extent_size cover the @size
973  */
count_max_extents(const struct btrfs_fs_info * fs_info,u64 size)974 static inline u32 count_max_extents(const struct btrfs_fs_info *fs_info, u64 size)
975 {
976 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
977 	if (!fs_info)
978 		return div_u64(size + BTRFS_MAX_EXTENT_SIZE - 1, BTRFS_MAX_EXTENT_SIZE);
979 #endif
980 
981 	return div_u64(size + fs_info->max_extent_size - 1, fs_info->max_extent_size);
982 }
983 
984 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
985 			enum btrfs_exclusive_operation type);
986 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
987 				 enum btrfs_exclusive_operation type);
988 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info);
989 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info);
990 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
991 			  enum btrfs_exclusive_operation op);
992 
993 int btrfs_check_ioctl_vol_args_path(const struct btrfs_ioctl_vol_args *vol_args);
994 
995 u16 btrfs_csum_type_size(u16 type);
996 int btrfs_super_csum_size(const struct btrfs_super_block *s);
997 const char *btrfs_super_csum_name(u16 csum_type);
998 const char *btrfs_super_csum_driver(u16 csum_type);
999 size_t __attribute_const__ btrfs_get_num_csums(void);
1000 
btrfs_is_empty_uuid(const u8 * uuid)1001 static inline bool btrfs_is_empty_uuid(const u8 *uuid)
1002 {
1003 	return uuid_is_null((const uuid_t *)uuid);
1004 }
1005 
1006 /* Compatibility and incompatibility defines */
1007 void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
1008 			     const char *name);
1009 void __btrfs_clear_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag,
1010 			       const char *name);
1011 void __btrfs_set_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
1012 			      const char *name);
1013 void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag,
1014 				const char *name);
1015 
1016 #define __btrfs_fs_incompat(fs_info, flags)				\
1017 	(!!(btrfs_super_incompat_flags((fs_info)->super_copy) & (flags)))
1018 
1019 #define __btrfs_fs_compat_ro(fs_info, flags)				\
1020 	(!!(btrfs_super_compat_ro_flags((fs_info)->super_copy) & (flags)))
1021 
1022 #define btrfs_set_fs_incompat(__fs_info, opt)				\
1023 	__btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
1024 
1025 #define btrfs_clear_fs_incompat(__fs_info, opt)				\
1026 	__btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt, #opt)
1027 
1028 #define btrfs_fs_incompat(fs_info, opt)					\
1029 	__btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
1030 
1031 #define btrfs_set_fs_compat_ro(__fs_info, opt)				\
1032 	__btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
1033 
1034 #define btrfs_clear_fs_compat_ro(__fs_info, opt)			\
1035 	__btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt, #opt)
1036 
1037 #define btrfs_fs_compat_ro(fs_info, opt)				\
1038 	__btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
1039 
1040 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
1041 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
1042 #define btrfs_raw_test_opt(o, opt)	((o) & BTRFS_MOUNT_##opt)
1043 #define btrfs_test_opt(fs_info, opt)	((fs_info)->mount_opt & \
1044 					 BTRFS_MOUNT_##opt)
1045 
btrfs_fs_closing(const struct btrfs_fs_info * fs_info)1046 static inline int btrfs_fs_closing(const struct btrfs_fs_info *fs_info)
1047 {
1048 	/* Do it this way so we only ever do one test_bit in the normal case. */
1049 	if (test_bit(BTRFS_FS_CLOSING_START, &fs_info->flags)) {
1050 		if (test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags))
1051 			return 2;
1052 		return 1;
1053 	}
1054 	return 0;
1055 }
1056 
1057 /*
1058  * If we remount the fs to be R/O or umount the fs, the cleaner needn't do
1059  * anything except sleeping. This function is used to check the status of
1060  * the fs.
1061  * We check for BTRFS_FS_STATE_RO to avoid races with a concurrent remount,
1062  * since setting and checking for SB_RDONLY in the superblock's flags is not
1063  * atomic.
1064  */
btrfs_need_cleaner_sleep(const struct btrfs_fs_info * fs_info)1065 static inline int btrfs_need_cleaner_sleep(const struct btrfs_fs_info *fs_info)
1066 {
1067 	return test_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state) ||
1068 		btrfs_fs_closing(fs_info);
1069 }
1070 
btrfs_wake_unfinished_drop(struct btrfs_fs_info * fs_info)1071 static inline void btrfs_wake_unfinished_drop(struct btrfs_fs_info *fs_info)
1072 {
1073 	clear_and_wake_up_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags);
1074 }
1075 
1076 #define BTRFS_FS_ERROR(fs_info)	(READ_ONCE((fs_info)->fs_error))
1077 
1078 #define BTRFS_FS_LOG_CLEANUP_ERROR(fs_info)				\
1079 	(unlikely(test_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR,		\
1080 			   &(fs_info)->fs_state)))
1081 
1082 /*
1083  * We use folio flag owner_2 to indicate there is an ordered extent with
1084  * unfinished IO.
1085  */
1086 #define folio_test_ordered(folio)	folio_test_owner_2(folio)
1087 #define folio_set_ordered(folio)	folio_set_owner_2(folio)
1088 #define folio_clear_ordered(folio)	folio_clear_owner_2(folio)
1089 
1090 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1091 
1092 #define EXPORT_FOR_TESTS
1093 
btrfs_is_testing(const struct btrfs_fs_info * fs_info)1094 static inline int btrfs_is_testing(const struct btrfs_fs_info *fs_info)
1095 {
1096 	return test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
1097 }
1098 
1099 void btrfs_test_destroy_inode(struct inode *inode);
1100 
1101 #else
1102 
1103 #define EXPORT_FOR_TESTS static
1104 
btrfs_is_testing(const struct btrfs_fs_info * fs_info)1105 static inline int btrfs_is_testing(const struct btrfs_fs_info *fs_info)
1106 {
1107 	return 0;
1108 }
1109 #endif
1110 
1111 #endif
1112