1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (C) 2007 Oracle. All rights reserved.
9 #include <linux/hash.h>
10 #include <linux/refcount.h>
11 #include <linux/spinlock.h>
12 #include <linux/mutex.h>
13 #include <linux/rwsem.h>
16 #include <linux/compiler.h>
17 #include <linux/fscrypt.h>
18 #include <linux/lockdep.h>
19 #include <uapi/linux/btrfs_tree.h>
20 #include <trace/events/btrfs.h>
21 #include "block-rsv.h"
22 #include "extent_map.h"
23 #include "extent_io.h"
24 #include "extent-io-tree.h"
25 #include "ordered-data.h"
26 #include "delayed-inode.h"
31 struct writeback_control
;
34 struct btrfs_trans_handle
;
37 * Since we search a directory based on f_pos (struct dir_context::pos) we have
38 * to start at 2 since '.' and '..' have f_pos of 0 and 1 respectively, so
39 * everybody else has to start at 2 (see btrfs_real_readdir() and dir_emit_dots()).
41 #define BTRFS_DIR_START_INDEX 2
44 * ordered_data_close is set by truncate when a file that used
45 * to have good data has been truncated to zero. When it is set
46 * the btrfs file release call will add this inode to the
47 * ordered operations list so that we make sure to flush out any
48 * new data the application may have written before commit.
51 BTRFS_INODE_FLUSH_ON_CLOSE
,
53 BTRFS_INODE_IN_DEFRAG
,
54 BTRFS_INODE_HAS_ASYNC_EXTENT
,
56 * Always set under the VFS' inode lock, otherwise it can cause races
57 * during fsync (we start as a fast fsync and then end up in a full
58 * fsync racing with ordered extent completion).
60 BTRFS_INODE_NEEDS_FULL_SYNC
,
61 BTRFS_INODE_COPY_EVERYTHING
,
62 BTRFS_INODE_HAS_PROPS
,
63 BTRFS_INODE_SNAPSHOT_FLUSH
,
65 * Set and used when logging an inode and it serves to signal that an
66 * inode does not have xattrs, so subsequent fsyncs can avoid searching
67 * for xattrs to log. This bit must be cleared whenever a xattr is added
70 BTRFS_INODE_NO_XATTRS
,
72 * Set when we are in a context where we need to start a transaction and
73 * have dirty pages with the respective file range locked. This is to
74 * ensure that when reserving space for the transaction, if we are low
75 * on available space and need to flush delalloc, we will not flush
76 * delalloc for this inode, because that could result in a deadlock (on
77 * the file range, inode's io_tree).
79 BTRFS_INODE_NO_DELALLOC_FLUSH
,
81 * Set when we are working on enabling verity for a file. Computing and
82 * writing the whole Merkle tree can take a while so we want to prevent
83 * races where two separate tasks attempt to simultaneously start verity
86 BTRFS_INODE_VERITY_IN_PROGRESS
,
87 /* Set when this inode is a free space inode. */
88 BTRFS_INODE_FREE_SPACE_INODE
,
89 /* Set when there are no capabilities in XATTs for the inode. */
90 BTRFS_INODE_NO_CAP_XATTR
,
92 * Set if an error happened when doing a COW write before submitting a
93 * bio or during writeback. Used for both buffered writes and direct IO
94 * writes. This is to signal a fast fsync that it has to wait for
95 * ordered extents to complete and therefore not log extent maps that
96 * point to unwritten extents (when an ordered extent completes and it
97 * has the BTRFS_ORDERED_IOERR flag set, it drops extent maps in its
100 BTRFS_INODE_COW_WRITE_ERROR
,
102 * Indicate this is a directory that points to a subvolume for which
103 * there is no root reference item. That's a case like the following:
105 * $ btrfs subvolume create /mnt/parent
106 * $ btrfs subvolume create /mnt/parent/child
107 * $ btrfs subvolume snapshot /mnt/parent /mnt/snap
109 * If subvolume "parent" is root 256, subvolume "child" is root 257 and
110 * snapshot "snap" is root 258, then there's no root reference item (key
111 * BTRFS_ROOT_REF_KEY in the root tree) for the subvolume "child"
112 * associated to root 258 (the snapshot) - there's only for the root
113 * of the "parent" subvolume (root 256). In the chunk root we have a
114 * (256 BTRFS_ROOT_REF_KEY 257) key but we don't have a
115 * (258 BTRFS_ROOT_REF_KEY 257) key - the sames goes for backrefs, we
116 * have a (257 BTRFS_ROOT_BACKREF_KEY 256) but we don't have a
117 * (257 BTRFS_ROOT_BACKREF_KEY 258) key.
119 * So when opening the "child" dentry from the snapshot's directory,
120 * we don't find a root ref item and we create a stub inode. This is
121 * done at new_simple_dir(), called from btrfs_lookup_dentry().
123 BTRFS_INODE_ROOT_STUB
,
126 /* in memory btrfs inode */
128 /* which subvolume this inode belongs to */
129 struct btrfs_root
*root
;
131 #if BITS_PER_LONG == 32
133 * The objectid of the corresponding BTRFS_INODE_ITEM_KEY.
134 * On 64 bits platforms we can get it from vfs_inode.i_ino, which is an
135 * unsigned long and therefore 64 bits on such platforms.
140 /* Cached value of inode property 'compression'. */
144 * Force compression on the file using the defrag ioctl, could be
145 * different from prop_compress and takes precedence if set.
150 * Lock for counters and all fields used to determine if the inode is in
151 * the log or not (last_trans, last_sub_trans, last_log_commit,
152 * logged_trans), to access/update delalloc_bytes, new_delalloc_bytes,
153 * defrag_bytes, disk_i_size, outstanding_extents, csum_bytes and to
154 * update the VFS' inode number of bytes used.
155 * Also protects setting struct file::private_data.
159 /* the extent_tree has caches of all the extent mappings to disk */
160 struct extent_map_tree extent_tree
;
162 /* the io_tree does range state (DIRTY, LOCKED etc) */
163 struct extent_io_tree io_tree
;
166 * Keep track of where the inode has extent items mapped in order to
167 * make sure the i_size adjustments are accurate. Not required when the
168 * filesystem is NO_HOLES, the status can't be set while mounted as
169 * it's a mkfs-time feature.
171 struct extent_io_tree
*file_extent_tree
;
173 /* held while logging the inode in tree-log.c */
174 struct mutex log_mutex
;
177 * Counters to keep track of the number of extent item's we may use due
178 * to delalloc and such. outstanding_extents is the number of extent
179 * items we think we'll end up using, and reserved_extents is the number
180 * of extent items we've reserved metadata for. Protected by 'lock'.
182 unsigned outstanding_extents
;
184 /* used to order data wrt metadata */
185 spinlock_t ordered_tree_lock
;
186 struct rb_root ordered_tree
;
187 struct rb_node
*ordered_tree_last
;
189 /* list of all the delalloc inodes in the FS. There are times we need
190 * to write all the delalloc pages to disk, and this list is used
193 struct list_head delalloc_inodes
;
195 unsigned long runtime_flags
;
197 /* full 64 bit generation number, struct vfs_inode doesn't have a big
198 * enough field for this.
203 * ID of the transaction handle that last modified this inode.
204 * Protected by 'lock'.
209 * ID of the transaction that last logged this inode.
210 * Protected by 'lock'.
215 * Log transaction ID when this inode was last modified.
216 * Protected by 'lock'.
220 /* A local copy of root's last_log_commit. Protected by 'lock'. */
225 * Total number of bytes pending delalloc, used by stat to
226 * calculate the real block usage of the file. This is used
227 * only for files. Protected by 'lock'.
231 * The lowest possible index of the next dir index key which
232 * points to an inode that needs to be logged.
233 * This is used only for directories.
234 * Use the helpers btrfs_get_first_dir_index_to_log() and
235 * btrfs_set_first_dir_index_to_log() to access this field.
237 u64 first_dir_index_to_log
;
242 * Total number of bytes pending delalloc that fall within a file
243 * range that is either a hole or beyond EOF (and no prealloc extent
244 * exists in the range). This is always <= delalloc_bytes and this
245 * is used only for files. Protected by 'lock'.
247 u64 new_delalloc_bytes
;
249 * The offset of the last dir index key that was logged.
250 * This is used only for directories.
252 u64 last_dir_index_offset
;
257 * Total number of bytes pending defrag, used by stat to check whether
258 * it needs COW. Protected by 'lock'.
259 * Used by inodes other than the data relocation inode.
264 * Logical address of the block group being relocated.
265 * Used only by the data relocation inode.
267 u64 reloc_block_group_start
;
271 * The size of the file stored in the metadata on disk. data=ordered
272 * means the in-memory i_size might be larger than the size on disk
273 * because not all the blocks are written yet. Protected by 'lock'.
279 * If this is a directory then index_cnt is the counter for the
280 * index number for new files that are created. For an empty
281 * directory, this must be initialized to BTRFS_DIR_START_INDEX.
286 * If this is not a directory, this is the number of bytes
287 * outstanding that are going to need csums. This is used in
288 * ENOSPC accounting. Protected by 'lock'.
293 /* Cache the directory index number to speed the dir/file remove */
296 /* the fsync log has some corner cases that mean we have to check
297 * directories to see if any unlinks have been done before
298 * the directory was logged. See tree-log.c for all the
301 u64 last_unlink_trans
;
305 * The id/generation of the last transaction where this inode
306 * was either the source or the destination of a clone/dedupe
307 * operation. Used when logging an inode to know if there are
308 * shared extents that need special care when logging checksum
309 * items, to avoid duplicate checksum items in a log (which can
310 * lead to a corruption where we end up with missing checksum
311 * ranges after log replay). Protected by the VFS inode lock.
312 * Used for regular files only.
314 u64 last_reflink_trans
;
317 * In case this a root stub inode (BTRFS_INODE_ROOT_STUB flag set),
318 * the ID of that root.
323 /* Backwards incompatible flags, lower half of inode_item::flags */
325 /* Read-only compatibility flags, upper half of inode_item::flags */
328 struct btrfs_block_rsv block_rsv
;
330 struct btrfs_delayed_node
*delayed_node
;
332 /* File creation time. */
336 /* Hook into fs_info->delayed_iputs */
337 struct list_head delayed_iput
;
339 struct rw_semaphore i_mmap_lock
;
340 struct inode vfs_inode
;
343 static inline u64
btrfs_get_first_dir_index_to_log(const struct btrfs_inode
*inode
)
345 return READ_ONCE(inode
->first_dir_index_to_log
);
348 static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode
*inode
,
351 WRITE_ONCE(inode
->first_dir_index_to_log
, index
);
354 /* Type checked and const-preserving VFS inode -> btrfs inode. */
355 #define BTRFS_I(_inode) \
357 struct inode *: container_of(_inode, struct btrfs_inode, vfs_inode), \
358 const struct inode *: (const struct btrfs_inode *)container_of( \
359 _inode, const struct btrfs_inode, vfs_inode))
361 static inline unsigned long btrfs_inode_hash(u64 objectid
,
362 const struct btrfs_root
*root
)
364 u64 h
= objectid
^ (root
->root_key
.objectid
* GOLDEN_RATIO_PRIME
);
366 #if BITS_PER_LONG == 32
367 h
= (h
>> 32) ^ (h
& 0xffffffff);
370 return (unsigned long)h
;
373 #if BITS_PER_LONG == 32
376 * On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so
377 * we use the inode's location objectid which is a u64 to avoid truncation.
379 static inline u64
btrfs_ino(const struct btrfs_inode
*inode
)
381 u64 ino
= inode
->objectid
;
383 if (test_bit(BTRFS_INODE_ROOT_STUB
, &inode
->runtime_flags
))
384 ino
= inode
->vfs_inode
.i_ino
;
390 static inline u64
btrfs_ino(const struct btrfs_inode
*inode
)
392 return inode
->vfs_inode
.i_ino
;
397 static inline void btrfs_get_inode_key(const struct btrfs_inode
*inode
,
398 struct btrfs_key
*key
)
400 key
->objectid
= btrfs_ino(inode
);
401 key
->type
= BTRFS_INODE_ITEM_KEY
;
405 static inline void btrfs_set_inode_number(struct btrfs_inode
*inode
, u64 ino
)
407 #if BITS_PER_LONG == 32
408 inode
->objectid
= ino
;
410 inode
->vfs_inode
.i_ino
= ino
;
413 static inline void btrfs_i_size_write(struct btrfs_inode
*inode
, u64 size
)
415 i_size_write(&inode
->vfs_inode
, size
);
416 inode
->disk_i_size
= size
;
419 static inline bool btrfs_is_free_space_inode(const struct btrfs_inode
*inode
)
421 return test_bit(BTRFS_INODE_FREE_SPACE_INODE
, &inode
->runtime_flags
);
424 static inline bool is_data_inode(const struct btrfs_inode
*inode
)
426 return btrfs_ino(inode
) != BTRFS_BTREE_INODE_OBJECTID
;
429 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode
*inode
,
432 lockdep_assert_held(&inode
->lock
);
433 inode
->outstanding_extents
+= mod
;
434 if (btrfs_is_free_space_inode(inode
))
436 trace_btrfs_inode_mod_outstanding_extents(inode
->root
, btrfs_ino(inode
),
437 mod
, inode
->outstanding_extents
);
441 * Called every time after doing a buffered, direct IO or memory mapped write.
443 * This is to ensure that if we write to a file that was previously fsynced in
444 * the current transaction, then try to fsync it again in the same transaction,
445 * we will know that there were changes in the file and that it needs to be
448 static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode
*inode
)
450 spin_lock(&inode
->lock
);
451 inode
->last_sub_trans
= inode
->root
->log_transid
;
452 spin_unlock(&inode
->lock
);
456 * Should be called while holding the inode's VFS lock in exclusive mode, or
457 * while holding the inode's mmap lock (struct btrfs_inode::i_mmap_lock) in
458 * either shared or exclusive mode, or in a context where no one else can access
459 * the inode concurrently (during inode creation or when loading an inode from
462 static inline void btrfs_set_inode_full_sync(struct btrfs_inode
*inode
)
464 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
, &inode
->runtime_flags
);
466 * The inode may have been part of a reflink operation in the last
467 * transaction that modified it, and then a fsync has reset the
468 * last_reflink_trans to avoid subsequent fsyncs in the same
469 * transaction to do unnecessary work. So update last_reflink_trans
470 * to the last_trans value (we have to be pessimistic and assume a
473 * The ->last_trans is protected by the inode's spinlock and we can
474 * have a concurrent ordered extent completion update it. Also set
475 * last_reflink_trans to ->last_trans only if the former is less than
476 * the later, because we can be called in a context where
477 * last_reflink_trans was set to the current transaction generation
478 * while ->last_trans was not yet updated in the current transaction,
479 * and therefore has a lower value.
481 spin_lock(&inode
->lock
);
482 if (inode
->last_reflink_trans
< inode
->last_trans
)
483 inode
->last_reflink_trans
= inode
->last_trans
;
484 spin_unlock(&inode
->lock
);
487 static inline bool btrfs_inode_in_log(struct btrfs_inode
*inode
, u64 generation
)
491 spin_lock(&inode
->lock
);
492 if (inode
->logged_trans
== generation
&&
493 inode
->last_sub_trans
<= inode
->last_log_commit
&&
494 inode
->last_sub_trans
<= btrfs_get_root_last_log_commit(inode
->root
))
496 spin_unlock(&inode
->lock
);
501 * Check if the inode has flags compatible with compression
503 static inline bool btrfs_inode_can_compress(const struct btrfs_inode
*inode
)
505 if (inode
->flags
& BTRFS_INODE_NODATACOW
||
506 inode
->flags
& BTRFS_INODE_NODATASUM
)
511 static inline void btrfs_assert_inode_locked(struct btrfs_inode
*inode
)
513 /* Immediately trigger a crash if the inode is not locked. */
514 ASSERT(inode_is_locked(&inode
->vfs_inode
));
515 /* Trigger a splat in dmesg if this task is not holding the lock. */
516 lockdep_assert_held(&inode
->vfs_inode
.i_rwsem
);
519 /* Array of bytes with variable length, hexadecimal format 0x1234 */
520 #define CSUM_FMT "0x%*phN"
521 #define CSUM_FMT_VALUE(size, bytes) size, bytes
523 int btrfs_check_sector_csum(struct btrfs_fs_info
*fs_info
, struct page
*page
,
524 u32 pgoff
, u8
*csum
, const u8
* const csum_expected
);
525 bool btrfs_data_csum_ok(struct btrfs_bio
*bbio
, struct btrfs_device
*dev
,
526 u32 bio_offset
, struct bio_vec
*bv
);
527 noinline
int can_nocow_extent(struct inode
*inode
, u64 offset
, u64
*len
,
528 struct btrfs_file_extent
*file_extent
,
529 bool nowait
, bool strict
);
531 void btrfs_del_delalloc_inode(struct btrfs_inode
*inode
);
532 struct inode
*btrfs_lookup_dentry(struct inode
*dir
, struct dentry
*dentry
);
533 int btrfs_set_inode_index(struct btrfs_inode
*dir
, u64
*index
);
534 int btrfs_unlink_inode(struct btrfs_trans_handle
*trans
,
535 struct btrfs_inode
*dir
, struct btrfs_inode
*inode
,
536 const struct fscrypt_str
*name
);
537 int btrfs_add_link(struct btrfs_trans_handle
*trans
,
538 struct btrfs_inode
*parent_inode
, struct btrfs_inode
*inode
,
539 const struct fscrypt_str
*name
, int add_backref
, u64 index
);
540 int btrfs_delete_subvolume(struct btrfs_inode
*dir
, struct dentry
*dentry
);
541 int btrfs_truncate_block(struct btrfs_inode
*inode
, loff_t from
, loff_t len
,
544 int btrfs_start_delalloc_snapshot(struct btrfs_root
*root
, bool in_reclaim_context
);
545 int btrfs_start_delalloc_roots(struct btrfs_fs_info
*fs_info
, long nr
,
546 bool in_reclaim_context
);
547 int btrfs_set_extent_delalloc(struct btrfs_inode
*inode
, u64 start
, u64 end
,
548 unsigned int extra_bits
,
549 struct extent_state
**cached_state
);
551 struct btrfs_new_inode_args
{
554 struct dentry
*dentry
;
559 /* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */
560 struct posix_acl
*default_acl
;
561 struct posix_acl
*acl
;
562 struct fscrypt_name fname
;
565 int btrfs_new_inode_prepare(struct btrfs_new_inode_args
*args
,
566 unsigned int *trans_num_items
);
567 int btrfs_create_new_inode(struct btrfs_trans_handle
*trans
,
568 struct btrfs_new_inode_args
*args
);
569 void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args
*args
);
570 struct inode
*btrfs_new_subvol_inode(struct mnt_idmap
*idmap
,
572 void btrfs_set_delalloc_extent(struct btrfs_inode
*inode
, struct extent_state
*state
,
574 void btrfs_clear_delalloc_extent(struct btrfs_inode
*inode
,
575 struct extent_state
*state
, u32 bits
);
576 void btrfs_merge_delalloc_extent(struct btrfs_inode
*inode
, struct extent_state
*new,
577 struct extent_state
*other
);
578 void btrfs_split_delalloc_extent(struct btrfs_inode
*inode
,
579 struct extent_state
*orig
, u64 split
);
580 void btrfs_evict_inode(struct inode
*inode
);
581 struct inode
*btrfs_alloc_inode(struct super_block
*sb
);
582 void btrfs_destroy_inode(struct inode
*inode
);
583 void btrfs_free_inode(struct inode
*inode
);
584 int btrfs_drop_inode(struct inode
*inode
);
585 int __init
btrfs_init_cachep(void);
586 void __cold
btrfs_destroy_cachep(void);
587 struct inode
*btrfs_iget_path(u64 ino
, struct btrfs_root
*root
,
588 struct btrfs_path
*path
);
589 struct inode
*btrfs_iget(u64 ino
, struct btrfs_root
*root
);
590 struct extent_map
*btrfs_get_extent(struct btrfs_inode
*inode
,
591 struct folio
*folio
, u64 start
, u64 len
);
592 int btrfs_update_inode(struct btrfs_trans_handle
*trans
,
593 struct btrfs_inode
*inode
);
594 int btrfs_update_inode_fallback(struct btrfs_trans_handle
*trans
,
595 struct btrfs_inode
*inode
);
596 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct btrfs_inode
*inode
);
597 int btrfs_orphan_cleanup(struct btrfs_root
*root
);
598 int btrfs_cont_expand(struct btrfs_inode
*inode
, loff_t oldsize
, loff_t size
);
599 void btrfs_add_delayed_iput(struct btrfs_inode
*inode
);
600 void btrfs_run_delayed_iputs(struct btrfs_fs_info
*fs_info
);
601 int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info
*fs_info
);
602 int btrfs_prealloc_file_range(struct inode
*inode
, int mode
,
603 u64 start
, u64 num_bytes
, u64 min_size
,
604 loff_t actual_len
, u64
*alloc_hint
);
605 int btrfs_prealloc_file_range_trans(struct inode
*inode
,
606 struct btrfs_trans_handle
*trans
, int mode
,
607 u64 start
, u64 num_bytes
, u64 min_size
,
608 loff_t actual_len
, u64
*alloc_hint
);
609 int btrfs_run_delalloc_range(struct btrfs_inode
*inode
, struct folio
*locked_folio
,
610 u64 start
, u64 end
, struct writeback_control
*wbc
);
611 int btrfs_writepage_cow_fixup(struct folio
*folio
);
612 int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info
*fs_info
,
614 int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode
*inode
,
615 u64 disk_bytenr
, u64 disk_io_size
,
616 struct page
**pages
, void *uring_ctx
);
617 ssize_t
btrfs_encoded_read(struct kiocb
*iocb
, struct iov_iter
*iter
,
618 struct btrfs_ioctl_encoded_io_args
*encoded
,
619 struct extent_state
**cached_state
,
620 u64
*disk_bytenr
, u64
*disk_io_size
);
621 ssize_t
btrfs_encoded_read_regular(struct kiocb
*iocb
, struct iov_iter
*iter
,
622 u64 start
, u64 lockend
,
623 struct extent_state
**cached_state
,
624 u64 disk_bytenr
, u64 disk_io_size
,
625 size_t count
, bool compressed
, bool *unlocked
);
626 ssize_t
btrfs_do_encoded_write(struct kiocb
*iocb
, struct iov_iter
*from
,
627 const struct btrfs_ioctl_encoded_io_args
*encoded
);
629 struct btrfs_inode
*btrfs_find_first_inode(struct btrfs_root
*root
, u64 min_ino
);
631 extern const struct dentry_operations btrfs_dentry_operations
;
633 /* Inode locking type flags, by default the exclusive lock is taken. */
634 enum btrfs_ilock_type
{
635 ENUM_BIT(BTRFS_ILOCK_SHARED
),
636 ENUM_BIT(BTRFS_ILOCK_TRY
),
637 ENUM_BIT(BTRFS_ILOCK_MMAP
),
640 int btrfs_inode_lock(struct btrfs_inode
*inode
, unsigned int ilock_flags
);
641 void btrfs_inode_unlock(struct btrfs_inode
*inode
, unsigned int ilock_flags
);
642 void btrfs_update_inode_bytes(struct btrfs_inode
*inode
, const u64 add_bytes
,
643 const u64 del_bytes
);
644 void btrfs_assert_inode_range_clean(struct btrfs_inode
*inode
, u64 start
, u64 end
);
645 u64
btrfs_get_extent_allocation_hint(struct btrfs_inode
*inode
, u64 start
,
647 struct extent_map
*btrfs_create_io_em(struct btrfs_inode
*inode
, u64 start
,
648 const struct btrfs_file_extent
*file_extent
,