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/fscrypt.h>
12 #include <trace/events/btrfs.h>
13 #include "extent_map.h"
14 #include "extent_io.h"
15 #include "ordered-data.h"
16 #include "delayed-inode.h"
19 * Since we search a directory based on f_pos (struct dir_context::pos) we have
20 * to start at 2 since '.' and '..' have f_pos of 0 and 1 respectively, so
21 * everybody else has to start at 2 (see btrfs_real_readdir() and dir_emit_dots()).
23 #define BTRFS_DIR_START_INDEX 2
26 * ordered_data_close is set by truncate when a file that used
27 * to have good data has been truncated to zero. When it is set
28 * the btrfs file release call will add this inode to the
29 * ordered operations list so that we make sure to flush out any
30 * new data the application may have written before commit.
33 BTRFS_INODE_FLUSH_ON_CLOSE
,
35 BTRFS_INODE_IN_DEFRAG
,
36 BTRFS_INODE_HAS_ASYNC_EXTENT
,
38 * Always set under the VFS' inode lock, otherwise it can cause races
39 * during fsync (we start as a fast fsync and then end up in a full
40 * fsync racing with ordered extent completion).
42 BTRFS_INODE_NEEDS_FULL_SYNC
,
43 BTRFS_INODE_COPY_EVERYTHING
,
44 BTRFS_INODE_IN_DELALLOC_LIST
,
45 BTRFS_INODE_HAS_PROPS
,
46 BTRFS_INODE_SNAPSHOT_FLUSH
,
48 * Set and used when logging an inode and it serves to signal that an
49 * inode does not have xattrs, so subsequent fsyncs can avoid searching
50 * for xattrs to log. This bit must be cleared whenever a xattr is added
53 BTRFS_INODE_NO_XATTRS
,
55 * Set when we are in a context where we need to start a transaction and
56 * have dirty pages with the respective file range locked. This is to
57 * ensure that when reserving space for the transaction, if we are low
58 * on available space and need to flush delalloc, we will not flush
59 * delalloc for this inode, because that could result in a deadlock (on
60 * the file range, inode's io_tree).
62 BTRFS_INODE_NO_DELALLOC_FLUSH
,
64 * Set when we are working on enabling verity for a file. Computing and
65 * writing the whole Merkle tree can take a while so we want to prevent
66 * races where two separate tasks attempt to simultaneously start verity
69 BTRFS_INODE_VERITY_IN_PROGRESS
,
70 /* Set when this inode is a free space inode. */
71 BTRFS_INODE_FREE_SPACE_INODE
,
74 /* in memory btrfs inode */
76 /* which subvolume this inode belongs to */
77 struct btrfs_root
*root
;
79 /* key used to find this inode on disk. This is used by the code
80 * to read in roots of subvolumes
82 struct btrfs_key location
;
84 /* Cached value of inode property 'compression'. */
88 * Force compression on the file using the defrag ioctl, could be
89 * different from prop_compress and takes precedence if set.
94 * Lock for counters and all fields used to determine if the inode is in
95 * the log or not (last_trans, last_sub_trans, last_log_commit,
96 * logged_trans), to access/update new_delalloc_bytes and to update the
97 * VFS' inode number of bytes used.
101 /* the extent_tree has caches of all the extent mappings to disk */
102 struct extent_map_tree extent_tree
;
104 /* the io_tree does range state (DIRTY, LOCKED etc) */
105 struct extent_io_tree io_tree
;
108 * Keep track of where the inode has extent items mapped in order to
109 * make sure the i_size adjustments are accurate
111 struct extent_io_tree file_extent_tree
;
113 /* held while logging the inode in tree-log.c */
114 struct mutex log_mutex
;
116 /* used to order data wrt metadata */
117 struct btrfs_ordered_inode_tree ordered_tree
;
119 /* list of all the delalloc inodes in the FS. There are times we need
120 * to write all the delalloc pages to disk, and this list is used
123 struct list_head delalloc_inodes
;
125 /* node for the red-black tree that links inodes in subvolume root */
126 struct rb_node rb_node
;
128 unsigned long runtime_flags
;
130 /* full 64 bit generation number, struct vfs_inode doesn't have a big
131 * enough field for this.
136 * transid of the trans_handle that last modified this inode
141 * transid that last logged this inode
146 * log transid when this inode was last modified
150 /* a local copy of root's last_log_commit */
155 * Total number of bytes pending delalloc, used by stat to
156 * calculate the real block usage of the file. This is used
161 * The lowest possible index of the next dir index key which
162 * points to an inode that needs to be logged.
163 * This is used only for directories.
164 * Use the helpers btrfs_get_first_dir_index_to_log() and
165 * btrfs_set_first_dir_index_to_log() to access this field.
167 u64 first_dir_index_to_log
;
172 * Total number of bytes pending delalloc that fall within a file
173 * range that is either a hole or beyond EOF (and no prealloc extent
174 * exists in the range). This is always <= delalloc_bytes and this
175 * is used only for files.
177 u64 new_delalloc_bytes
;
179 * The offset of the last dir index key that was logged.
180 * This is used only for directories.
182 u64 last_dir_index_offset
;
186 * total number of bytes pending defrag, used by stat to check whether
192 * the size of the file stored in the metadata on disk. data=ordered
193 * means the in-memory i_size might be larger than the size on disk
194 * because not all the blocks are written yet.
199 * If this is a directory then index_cnt is the counter for the index
200 * number for new files that are created. For an empty directory, this
201 * must be initialized to BTRFS_DIR_START_INDEX.
205 /* Cache the directory index number to speed the dir/file remove */
208 /* the fsync log has some corner cases that mean we have to check
209 * directories to see if any unlinks have been done before
210 * the directory was logged. See tree-log.c for all the
213 u64 last_unlink_trans
;
216 * The id/generation of the last transaction where this inode was
217 * either the source or the destination of a clone/dedupe operation.
218 * Used when logging an inode to know if there are shared extents that
219 * need special care when logging checksum items, to avoid duplicate
220 * checksum items in a log (which can lead to a corruption where we end
221 * up with missing checksum ranges after log replay).
222 * Protected by the vfs inode lock.
224 u64 last_reflink_trans
;
227 * Number of bytes outstanding that are going to need csums. This is
228 * used in ENOSPC accounting.
232 /* Backwards incompatible flags, lower half of inode_item::flags */
234 /* Read-only compatibility flags, upper half of inode_item::flags */
238 * Counters to keep track of the number of extent item's we may use due
239 * to delalloc and such. outstanding_extents is the number of extent
240 * items we think we'll end up using, and reserved_extents is the number
241 * of extent items we've reserved metadata for.
243 unsigned outstanding_extents
;
245 struct btrfs_block_rsv block_rsv
;
247 struct btrfs_delayed_node
*delayed_node
;
249 /* File creation time. */
250 struct timespec64 i_otime
;
252 /* Hook into fs_info->delayed_iputs */
253 struct list_head delayed_iput
;
255 struct rw_semaphore i_mmap_lock
;
256 struct inode vfs_inode
;
259 static inline u64
btrfs_get_first_dir_index_to_log(const struct btrfs_inode
*inode
)
261 return READ_ONCE(inode
->first_dir_index_to_log
);
264 static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode
*inode
,
267 WRITE_ONCE(inode
->first_dir_index_to_log
, index
);
270 static inline struct btrfs_inode
*BTRFS_I(const struct inode
*inode
)
272 return container_of(inode
, struct btrfs_inode
, vfs_inode
);
275 static inline unsigned long btrfs_inode_hash(u64 objectid
,
276 const struct btrfs_root
*root
)
278 u64 h
= objectid
^ (root
->root_key
.objectid
* GOLDEN_RATIO_PRIME
);
280 #if BITS_PER_LONG == 32
281 h
= (h
>> 32) ^ (h
& 0xffffffff);
284 return (unsigned long)h
;
287 #if BITS_PER_LONG == 32
290 * On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so
291 * we use the inode's location objectid which is a u64 to avoid truncation.
293 static inline u64
btrfs_ino(const struct btrfs_inode
*inode
)
295 u64 ino
= inode
->location
.objectid
;
297 /* type == BTRFS_ROOT_ITEM_KEY: subvol dir */
298 if (inode
->location
.type
== BTRFS_ROOT_ITEM_KEY
)
299 ino
= inode
->vfs_inode
.i_ino
;
305 static inline u64
btrfs_ino(const struct btrfs_inode
*inode
)
307 return inode
->vfs_inode
.i_ino
;
312 static inline void btrfs_i_size_write(struct btrfs_inode
*inode
, u64 size
)
314 i_size_write(&inode
->vfs_inode
, size
);
315 inode
->disk_i_size
= size
;
318 static inline bool btrfs_is_free_space_inode(struct btrfs_inode
*inode
)
320 return test_bit(BTRFS_INODE_FREE_SPACE_INODE
, &inode
->runtime_flags
);
323 static inline bool is_data_inode(struct inode
*inode
)
325 return btrfs_ino(BTRFS_I(inode
)) != BTRFS_BTREE_INODE_OBJECTID
;
328 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode
*inode
,
331 lockdep_assert_held(&inode
->lock
);
332 inode
->outstanding_extents
+= mod
;
333 if (btrfs_is_free_space_inode(inode
))
335 trace_btrfs_inode_mod_outstanding_extents(inode
->root
, btrfs_ino(inode
),
336 mod
, inode
->outstanding_extents
);
340 * Called every time after doing a buffered, direct IO or memory mapped write.
342 * This is to ensure that if we write to a file that was previously fsynced in
343 * the current transaction, then try to fsync it again in the same transaction,
344 * we will know that there were changes in the file and that it needs to be
347 static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode
*inode
)
349 spin_lock(&inode
->lock
);
350 inode
->last_sub_trans
= inode
->root
->log_transid
;
351 spin_unlock(&inode
->lock
);
355 * Should be called while holding the inode's VFS lock in exclusive mode or in a
356 * context where no one else can access the inode concurrently (during inode
357 * creation or when loading an inode from disk).
359 static inline void btrfs_set_inode_full_sync(struct btrfs_inode
*inode
)
361 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
, &inode
->runtime_flags
);
363 * The inode may have been part of a reflink operation in the last
364 * transaction that modified it, and then a fsync has reset the
365 * last_reflink_trans to avoid subsequent fsyncs in the same
366 * transaction to do unnecessary work. So update last_reflink_trans
367 * to the last_trans value (we have to be pessimistic and assume a
370 * The ->last_trans is protected by the inode's spinlock and we can
371 * have a concurrent ordered extent completion update it. Also set
372 * last_reflink_trans to ->last_trans only if the former is less than
373 * the later, because we can be called in a context where
374 * last_reflink_trans was set to the current transaction generation
375 * while ->last_trans was not yet updated in the current transaction,
376 * and therefore has a lower value.
378 spin_lock(&inode
->lock
);
379 if (inode
->last_reflink_trans
< inode
->last_trans
)
380 inode
->last_reflink_trans
= inode
->last_trans
;
381 spin_unlock(&inode
->lock
);
384 static inline bool btrfs_inode_in_log(struct btrfs_inode
*inode
, u64 generation
)
388 spin_lock(&inode
->lock
);
389 if (inode
->logged_trans
== generation
&&
390 inode
->last_sub_trans
<= inode
->last_log_commit
&&
391 inode
->last_sub_trans
<= inode
->root
->last_log_commit
)
393 spin_unlock(&inode
->lock
);
398 * Check if the inode has flags compatible with compression
400 static inline bool btrfs_inode_can_compress(const struct btrfs_inode
*inode
)
402 if (inode
->flags
& BTRFS_INODE_NODATACOW
||
403 inode
->flags
& BTRFS_INODE_NODATASUM
)
408 /* Array of bytes with variable length, hexadecimal format 0x1234 */
409 #define CSUM_FMT "0x%*phN"
410 #define CSUM_FMT_VALUE(size, bytes) size, bytes
412 int btrfs_check_sector_csum(struct btrfs_fs_info
*fs_info
, struct page
*page
,
413 u32 pgoff
, u8
*csum
, const u8
* const csum_expected
);
414 bool btrfs_data_csum_ok(struct btrfs_bio
*bbio
, struct btrfs_device
*dev
,
415 u32 bio_offset
, struct bio_vec
*bv
);
416 noinline
int can_nocow_extent(struct inode
*inode
, u64 offset
, u64
*len
,
417 u64
*orig_start
, u64
*orig_block_len
,
418 u64
*ram_bytes
, bool nowait
, bool strict
);
420 void __btrfs_del_delalloc_inode(struct btrfs_root
*root
, struct btrfs_inode
*inode
);
421 struct inode
*btrfs_lookup_dentry(struct inode
*dir
, struct dentry
*dentry
);
422 int btrfs_set_inode_index(struct btrfs_inode
*dir
, u64
*index
);
423 int btrfs_unlink_inode(struct btrfs_trans_handle
*trans
,
424 struct btrfs_inode
*dir
, struct btrfs_inode
*inode
,
425 const struct fscrypt_str
*name
);
426 int btrfs_add_link(struct btrfs_trans_handle
*trans
,
427 struct btrfs_inode
*parent_inode
, struct btrfs_inode
*inode
,
428 const struct fscrypt_str
*name
, int add_backref
, u64 index
);
429 int btrfs_delete_subvolume(struct btrfs_inode
*dir
, struct dentry
*dentry
);
430 int btrfs_truncate_block(struct btrfs_inode
*inode
, loff_t from
, loff_t len
,
433 int btrfs_start_delalloc_snapshot(struct btrfs_root
*root
, bool in_reclaim_context
);
434 int btrfs_start_delalloc_roots(struct btrfs_fs_info
*fs_info
, long nr
,
435 bool in_reclaim_context
);
436 int btrfs_set_extent_delalloc(struct btrfs_inode
*inode
, u64 start
, u64 end
,
437 unsigned int extra_bits
,
438 struct extent_state
**cached_state
);
440 struct btrfs_new_inode_args
{
443 struct dentry
*dentry
;
448 /* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */
449 struct posix_acl
*default_acl
;
450 struct posix_acl
*acl
;
451 struct fscrypt_name fname
;
454 int btrfs_new_inode_prepare(struct btrfs_new_inode_args
*args
,
455 unsigned int *trans_num_items
);
456 int btrfs_create_new_inode(struct btrfs_trans_handle
*trans
,
457 struct btrfs_new_inode_args
*args
);
458 void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args
*args
);
459 struct inode
*btrfs_new_subvol_inode(struct mnt_idmap
*idmap
,
461 void btrfs_set_delalloc_extent(struct btrfs_inode
*inode
, struct extent_state
*state
,
463 void btrfs_clear_delalloc_extent(struct btrfs_inode
*inode
,
464 struct extent_state
*state
, u32 bits
);
465 void btrfs_merge_delalloc_extent(struct btrfs_inode
*inode
, struct extent_state
*new,
466 struct extent_state
*other
);
467 void btrfs_split_delalloc_extent(struct btrfs_inode
*inode
,
468 struct extent_state
*orig
, u64 split
);
469 void btrfs_set_range_writeback(struct btrfs_inode
*inode
, u64 start
, u64 end
);
470 vm_fault_t
btrfs_page_mkwrite(struct vm_fault
*vmf
);
471 void btrfs_evict_inode(struct inode
*inode
);
472 struct inode
*btrfs_alloc_inode(struct super_block
*sb
);
473 void btrfs_destroy_inode(struct inode
*inode
);
474 void btrfs_free_inode(struct inode
*inode
);
475 int btrfs_drop_inode(struct inode
*inode
);
476 int __init
btrfs_init_cachep(void);
477 void __cold
btrfs_destroy_cachep(void);
478 struct inode
*btrfs_iget_path(struct super_block
*s
, u64 ino
,
479 struct btrfs_root
*root
, struct btrfs_path
*path
);
480 struct inode
*btrfs_iget(struct super_block
*s
, u64 ino
, struct btrfs_root
*root
);
481 struct extent_map
*btrfs_get_extent(struct btrfs_inode
*inode
,
482 struct page
*page
, size_t pg_offset
,
484 int btrfs_update_inode(struct btrfs_trans_handle
*trans
,
485 struct btrfs_root
*root
, struct btrfs_inode
*inode
);
486 int btrfs_update_inode_fallback(struct btrfs_trans_handle
*trans
,
487 struct btrfs_root
*root
, struct btrfs_inode
*inode
);
488 int btrfs_orphan_add(struct btrfs_trans_handle
*trans
, struct btrfs_inode
*inode
);
489 int btrfs_orphan_cleanup(struct btrfs_root
*root
);
490 int btrfs_cont_expand(struct btrfs_inode
*inode
, loff_t oldsize
, loff_t size
);
491 void btrfs_add_delayed_iput(struct btrfs_inode
*inode
);
492 void btrfs_run_delayed_iputs(struct btrfs_fs_info
*fs_info
);
493 int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info
*fs_info
);
494 int btrfs_prealloc_file_range(struct inode
*inode
, int mode
,
495 u64 start
, u64 num_bytes
, u64 min_size
,
496 loff_t actual_len
, u64
*alloc_hint
);
497 int btrfs_prealloc_file_range_trans(struct inode
*inode
,
498 struct btrfs_trans_handle
*trans
, int mode
,
499 u64 start
, u64 num_bytes
, u64 min_size
,
500 loff_t actual_len
, u64
*alloc_hint
);
501 int btrfs_run_delalloc_range(struct btrfs_inode
*inode
, struct page
*locked_page
,
502 u64 start
, u64 end
, struct writeback_control
*wbc
);
503 int btrfs_writepage_cow_fixup(struct page
*page
);
504 int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info
*fs_info
,
506 int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode
*inode
,
507 u64 file_offset
, u64 disk_bytenr
,
509 struct page
**pages
);
510 ssize_t
btrfs_encoded_read(struct kiocb
*iocb
, struct iov_iter
*iter
,
511 struct btrfs_ioctl_encoded_io_args
*encoded
);
512 ssize_t
btrfs_do_encoded_write(struct kiocb
*iocb
, struct iov_iter
*from
,
513 const struct btrfs_ioctl_encoded_io_args
*encoded
);
515 ssize_t
btrfs_dio_read(struct kiocb
*iocb
, struct iov_iter
*iter
,
517 struct iomap_dio
*btrfs_dio_write(struct kiocb
*iocb
, struct iov_iter
*iter
,
520 extern const struct dentry_operations btrfs_dentry_operations
;
522 /* Inode locking type flags, by default the exclusive lock is taken. */
523 enum btrfs_ilock_type
{
524 ENUM_BIT(BTRFS_ILOCK_SHARED
),
525 ENUM_BIT(BTRFS_ILOCK_TRY
),
526 ENUM_BIT(BTRFS_ILOCK_MMAP
),
529 int btrfs_inode_lock(struct btrfs_inode
*inode
, unsigned int ilock_flags
);
530 void btrfs_inode_unlock(struct btrfs_inode
*inode
, unsigned int ilock_flags
);
531 void btrfs_update_inode_bytes(struct btrfs_inode
*inode
, const u64 add_bytes
,
532 const u64 del_bytes
);
533 void btrfs_assert_inode_range_clean(struct btrfs_inode
*inode
, u64 start
, u64 end
);