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 "extent_map.h"
12 #include "extent_io.h"
13 #include "ordered-data.h"
14 #include "delayed-inode.h"
17 * Since we search a directory based on f_pos (struct dir_context::pos) we have
18 * to start at 2 since '.' and '..' have f_pos of 0 and 1 respectively, so
19 * everybody else has to start at 2 (see btrfs_real_readdir() and dir_emit_dots()).
21 #define BTRFS_DIR_START_INDEX 2
24 * ordered_data_close is set by truncate when a file that used
25 * to have good data has been truncated to zero. When it is set
26 * the btrfs file release call will add this inode to the
27 * ordered operations list so that we make sure to flush out any
28 * new data the application may have written before commit.
31 BTRFS_INODE_FLUSH_ON_CLOSE,
33 BTRFS_INODE_IN_DEFRAG,
34 BTRFS_INODE_HAS_ASYNC_EXTENT,
36 * Always set under the VFS' inode lock, otherwise it can cause races
37 * during fsync (we start as a fast fsync and then end up in a full
38 * fsync racing with ordered extent completion).
40 BTRFS_INODE_NEEDS_FULL_SYNC,
41 BTRFS_INODE_COPY_EVERYTHING,
42 BTRFS_INODE_IN_DELALLOC_LIST,
43 BTRFS_INODE_HAS_PROPS,
44 BTRFS_INODE_SNAPSHOT_FLUSH,
46 * Set and used when logging an inode and it serves to signal that an
47 * inode does not have xattrs, so subsequent fsyncs can avoid searching
48 * for xattrs to log. This bit must be cleared whenever a xattr is added
51 BTRFS_INODE_NO_XATTRS,
53 * Set when we are in a context where we need to start a transaction and
54 * have dirty pages with the respective file range locked. This is to
55 * ensure that when reserving space for the transaction, if we are low
56 * on available space and need to flush delalloc, we will not flush
57 * delalloc for this inode, because that could result in a deadlock (on
58 * the file range, inode's io_tree).
60 BTRFS_INODE_NO_DELALLOC_FLUSH,
62 * Set when we are working on enabling verity for a file. Computing and
63 * writing the whole Merkle tree can take a while so we want to prevent
64 * races where two separate tasks attempt to simultaneously start verity
67 BTRFS_INODE_VERITY_IN_PROGRESS,
68 /* Set when this inode is a free space inode. */
69 BTRFS_INODE_FREE_SPACE_INODE,
72 /* in memory btrfs inode */
74 /* which subvolume this inode belongs to */
75 struct btrfs_root *root;
77 /* key used to find this inode on disk. This is used by the code
78 * to read in roots of subvolumes
80 struct btrfs_key location;
83 * Lock for counters and all fields used to determine if the inode is in
84 * the log or not (last_trans, last_sub_trans, last_log_commit,
85 * logged_trans), to access/update new_delalloc_bytes and to update the
86 * VFS' inode number of bytes used.
90 /* the extent_tree has caches of all the extent mappings to disk */
91 struct extent_map_tree extent_tree;
93 /* the io_tree does range state (DIRTY, LOCKED etc) */
94 struct extent_io_tree io_tree;
96 /* special utility tree used to record which mirrors have already been
97 * tried when checksums fail for a given block
99 struct rb_root io_failure_tree;
100 spinlock_t io_failure_lock;
103 * Keep track of where the inode has extent items mapped in order to
104 * make sure the i_size adjustments are accurate
106 struct extent_io_tree file_extent_tree;
108 /* held while logging the inode in tree-log.c */
109 struct mutex log_mutex;
111 /* used to order data wrt metadata */
112 struct btrfs_ordered_inode_tree ordered_tree;
114 /* list of all the delalloc inodes in the FS. There are times we need
115 * to write all the delalloc pages to disk, and this list is used
118 struct list_head delalloc_inodes;
120 /* node for the red-black tree that links inodes in subvolume root */
121 struct rb_node rb_node;
123 unsigned long runtime_flags;
125 /* Keep track of who's O_SYNC/fsyncing currently */
126 atomic_t sync_writers;
128 /* full 64 bit generation number, struct vfs_inode doesn't have a big
129 * enough field for this.
134 * transid of the trans_handle that last modified this inode
139 * transid that last logged this inode
144 * log transid when this inode was last modified
148 /* a local copy of root's last_log_commit */
152 * Total number of bytes pending delalloc, used by stat to calculate the
153 * real block usage of the file. This is used only for files.
159 * Total number of bytes pending delalloc that fall within a file
160 * range that is either a hole or beyond EOF (and no prealloc extent
161 * exists in the range). This is always <= delalloc_bytes and this
162 * is used only for files.
164 u64 new_delalloc_bytes;
166 * The offset of the last dir index key that was logged.
167 * This is used only for directories.
169 u64 last_dir_index_offset;
173 * total number of bytes pending defrag, used by stat to check whether
179 * the size of the file stored in the metadata on disk. data=ordered
180 * means the in-memory i_size might be larger than the size on disk
181 * because not all the blocks are written yet.
186 * If this is a directory then index_cnt is the counter for the index
187 * number for new files that are created. For an empty directory, this
188 * must be initialized to BTRFS_DIR_START_INDEX.
192 /* Cache the directory index number to speed the dir/file remove */
195 /* the fsync log has some corner cases that mean we have to check
196 * directories to see if any unlinks have been done before
197 * the directory was logged. See tree-log.c for all the
200 u64 last_unlink_trans;
203 * The id/generation of the last transaction where this inode was
204 * either the source or the destination of a clone/dedupe operation.
205 * Used when logging an inode to know if there are shared extents that
206 * need special care when logging checksum items, to avoid duplicate
207 * checksum items in a log (which can lead to a corruption where we end
208 * up with missing checksum ranges after log replay).
209 * Protected by the vfs inode lock.
211 u64 last_reflink_trans;
214 * Number of bytes outstanding that are going to need csums. This is
215 * used in ENOSPC accounting.
219 /* Backwards incompatible flags, lower half of inode_item::flags */
221 /* Read-only compatibility flags, upper half of inode_item::flags */
225 * Counters to keep track of the number of extent item's we may use due
226 * to delalloc and such. outstanding_extents is the number of extent
227 * items we think we'll end up using, and reserved_extents is the number
228 * of extent items we've reserved metadata for.
230 unsigned outstanding_extents;
232 struct btrfs_block_rsv block_rsv;
235 * Cached values of inode properties
237 unsigned prop_compress; /* per-file compression algorithm */
239 * Force compression on the file using the defrag ioctl, could be
240 * different from prop_compress and takes precedence if set
242 unsigned defrag_compress;
244 struct btrfs_delayed_node *delayed_node;
246 /* File creation time. */
247 struct timespec64 i_otime;
249 /* Hook into fs_info->delayed_iputs */
250 struct list_head delayed_iput;
252 struct rw_semaphore i_mmap_lock;
253 struct inode vfs_inode;
256 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
258 return container_of(inode, struct btrfs_inode, vfs_inode);
261 static inline unsigned long btrfs_inode_hash(u64 objectid,
262 const struct btrfs_root *root)
264 u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
266 #if BITS_PER_LONG == 32
267 h = (h >> 32) ^ (h & 0xffffffff);
270 return (unsigned long)h;
273 #if BITS_PER_LONG == 32
276 * On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so
277 * we use the inode's location objectid which is a u64 to avoid truncation.
279 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
281 u64 ino = inode->location.objectid;
283 /* type == BTRFS_ROOT_ITEM_KEY: subvol dir */
284 if (inode->location.type == BTRFS_ROOT_ITEM_KEY)
285 ino = inode->vfs_inode.i_ino;
291 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
293 return inode->vfs_inode.i_ino;
298 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
300 i_size_write(&inode->vfs_inode, size);
301 inode->disk_i_size = size;
304 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
306 return test_bit(BTRFS_INODE_FREE_SPACE_INODE, &inode->runtime_flags);
309 static inline bool is_data_inode(struct inode *inode)
311 return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
314 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
317 lockdep_assert_held(&inode->lock);
318 inode->outstanding_extents += mod;
319 if (btrfs_is_free_space_inode(inode))
321 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
326 * Called every time after doing a buffered, direct IO or memory mapped write.
328 * This is to ensure that if we write to a file that was previously fsynced in
329 * the current transaction, then try to fsync it again in the same transaction,
330 * we will know that there were changes in the file and that it needs to be
333 static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)
335 spin_lock(&inode->lock);
336 inode->last_sub_trans = inode->root->log_transid;
337 spin_unlock(&inode->lock);
341 * Should be called while holding the inode's VFS lock in exclusive mode or in a
342 * context where no one else can access the inode concurrently (during inode
343 * creation or when loading an inode from disk).
345 static inline void btrfs_set_inode_full_sync(struct btrfs_inode *inode)
347 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
349 * The inode may have been part of a reflink operation in the last
350 * transaction that modified it, and then a fsync has reset the
351 * last_reflink_trans to avoid subsequent fsyncs in the same
352 * transaction to do unnecessary work. So update last_reflink_trans
353 * to the last_trans value (we have to be pessimistic and assume a
356 * The ->last_trans is protected by the inode's spinlock and we can
357 * have a concurrent ordered extent completion update it. Also set
358 * last_reflink_trans to ->last_trans only if the former is less than
359 * the later, because we can be called in a context where
360 * last_reflink_trans was set to the current transaction generation
361 * while ->last_trans was not yet updated in the current transaction,
362 * and therefore has a lower value.
364 spin_lock(&inode->lock);
365 if (inode->last_reflink_trans < inode->last_trans)
366 inode->last_reflink_trans = inode->last_trans;
367 spin_unlock(&inode->lock);
370 static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
374 spin_lock(&inode->lock);
375 if (inode->logged_trans == generation &&
376 inode->last_sub_trans <= inode->last_log_commit &&
377 inode->last_sub_trans <= inode->root->last_log_commit)
379 spin_unlock(&inode->lock);
384 * Check if the inode has flags compatible with compression
386 static inline bool btrfs_inode_can_compress(const struct btrfs_inode *inode)
388 if (inode->flags & BTRFS_INODE_NODATACOW ||
389 inode->flags & BTRFS_INODE_NODATASUM)
395 * btrfs_inode_item stores flags in a u64, btrfs_inode stores them in two
396 * separate u32s. These two functions convert between the two representations.
398 static inline u64 btrfs_inode_combine_flags(u32 flags, u32 ro_flags)
400 return (flags | ((u64)ro_flags << 32));
403 static inline void btrfs_inode_split_flags(u64 inode_item_flags,
404 u32 *flags, u32 *ro_flags)
406 *flags = (u32)inode_item_flags;
407 *ro_flags = (u32)(inode_item_flags >> 32);
410 /* Array of bytes with variable length, hexadecimal format 0x1234 */
411 #define CSUM_FMT "0x%*phN"
412 #define CSUM_FMT_VALUE(size, bytes) size, bytes
414 void btrfs_submit_data_write_bio(struct btrfs_inode *inode, struct bio *bio, int mirror_num);
415 void btrfs_submit_data_read_bio(struct btrfs_inode *inode, struct bio *bio,
416 int mirror_num, enum btrfs_compression_type compress_type);
417 void btrfs_submit_dio_repair_bio(struct btrfs_inode *inode, struct bio *bio, int mirror_num);
418 blk_status_t btrfs_submit_bio_start(struct btrfs_inode *inode, struct bio *bio);
419 blk_status_t btrfs_submit_bio_start_direct_io(struct btrfs_inode *inode,
421 u64 dio_file_offset);
422 int btrfs_check_sector_csum(struct btrfs_fs_info *fs_info, struct page *page,
423 u32 pgoff, u8 *csum, const u8 * const csum_expected);
424 int btrfs_check_data_csum(struct inode *inode, struct btrfs_bio *bbio,
425 u32 bio_offset, struct page *page, u32 pgoff);
426 unsigned int btrfs_verify_data_csum(struct btrfs_bio *bbio,
427 u32 bio_offset, struct page *page,
429 int btrfs_check_data_csum(struct inode *inode, struct btrfs_bio *bbio,
430 u32 bio_offset, struct page *page, u32 pgoff);
431 noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
432 u64 *orig_start, u64 *orig_block_len,
433 u64 *ram_bytes, bool nowait, bool strict);
435 void __btrfs_del_delalloc_inode(struct btrfs_root *root, struct btrfs_inode *inode);
436 struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);
437 int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index);
438 int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
439 struct btrfs_inode *dir, struct btrfs_inode *inode,
440 const struct fscrypt_str *name);
441 int btrfs_add_link(struct btrfs_trans_handle *trans,
442 struct btrfs_inode *parent_inode, struct btrfs_inode *inode,
443 const struct fscrypt_str *name, int add_backref, u64 index);
444 int btrfs_delete_subvolume(struct inode *dir, struct dentry *dentry);
445 int btrfs_truncate_block(struct btrfs_inode *inode, loff_t from, loff_t len,
448 int btrfs_start_delalloc_snapshot(struct btrfs_root *root, bool in_reclaim_context);
449 int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, long nr,
450 bool in_reclaim_context);
451 int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
452 unsigned int extra_bits,
453 struct extent_state **cached_state);
455 struct btrfs_new_inode_args {
458 struct dentry *dentry;
463 /* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */
464 struct posix_acl *default_acl;
465 struct posix_acl *acl;
466 struct fscrypt_name fname;
469 int btrfs_new_inode_prepare(struct btrfs_new_inode_args *args,
470 unsigned int *trans_num_items);
471 int btrfs_create_new_inode(struct btrfs_trans_handle *trans,
472 struct btrfs_new_inode_args *args);
473 void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args *args);
474 struct inode *btrfs_new_subvol_inode(struct user_namespace *mnt_userns,
476 void btrfs_set_delalloc_extent(struct inode *inode, struct extent_state *state,
478 void btrfs_clear_delalloc_extent(struct inode *inode,
479 struct extent_state *state, u32 bits);
480 void btrfs_merge_delalloc_extent(struct inode *inode, struct extent_state *new,
481 struct extent_state *other);
482 void btrfs_split_delalloc_extent(struct inode *inode,
483 struct extent_state *orig, u64 split);
484 void btrfs_set_range_writeback(struct btrfs_inode *inode, u64 start, u64 end);
485 vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf);
486 void btrfs_evict_inode(struct inode *inode);
487 struct inode *btrfs_alloc_inode(struct super_block *sb);
488 void btrfs_destroy_inode(struct inode *inode);
489 void btrfs_free_inode(struct inode *inode);
490 int btrfs_drop_inode(struct inode *inode);
491 int __init btrfs_init_cachep(void);
492 void __cold btrfs_destroy_cachep(void);
493 struct inode *btrfs_iget_path(struct super_block *s, u64 ino,
494 struct btrfs_root *root, struct btrfs_path *path);
495 struct inode *btrfs_iget(struct super_block *s, u64 ino, struct btrfs_root *root);
496 struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,
497 struct page *page, size_t pg_offset,
499 int btrfs_update_inode(struct btrfs_trans_handle *trans,
500 struct btrfs_root *root, struct btrfs_inode *inode);
501 int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
502 struct btrfs_root *root, struct btrfs_inode *inode);
503 int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct btrfs_inode *inode);
504 int btrfs_orphan_cleanup(struct btrfs_root *root);
505 int btrfs_cont_expand(struct btrfs_inode *inode, loff_t oldsize, loff_t size);
506 void btrfs_add_delayed_iput(struct inode *inode);
507 void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info);
508 int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info);
509 int btrfs_prealloc_file_range(struct inode *inode, int mode,
510 u64 start, u64 num_bytes, u64 min_size,
511 loff_t actual_len, u64 *alloc_hint);
512 int btrfs_prealloc_file_range_trans(struct inode *inode,
513 struct btrfs_trans_handle *trans, int mode,
514 u64 start, u64 num_bytes, u64 min_size,
515 loff_t actual_len, u64 *alloc_hint);
516 int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct page *locked_page,
517 u64 start, u64 end, int *page_started,
518 unsigned long *nr_written, struct writeback_control *wbc);
519 int btrfs_writepage_cow_fixup(struct page *page);
520 void btrfs_writepage_endio_finish_ordered(struct btrfs_inode *inode,
521 struct page *page, u64 start,
522 u64 end, bool uptodate);
523 int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info *fs_info,
525 int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode *inode,
526 u64 file_offset, u64 disk_bytenr,
528 struct page **pages);
529 ssize_t btrfs_encoded_read(struct kiocb *iocb, struct iov_iter *iter,
530 struct btrfs_ioctl_encoded_io_args *encoded);
531 ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,
532 const struct btrfs_ioctl_encoded_io_args *encoded);
534 ssize_t btrfs_dio_read(struct kiocb *iocb, struct iov_iter *iter,
536 struct iomap_dio *btrfs_dio_write(struct kiocb *iocb, struct iov_iter *iter,
539 extern const struct dentry_operations btrfs_dentry_operations;
541 /* Inode locking type flags, by default the exclusive lock is taken. */
542 enum btrfs_ilock_type {
543 ENUM_BIT(BTRFS_ILOCK_SHARED),
544 ENUM_BIT(BTRFS_ILOCK_TRY),
545 ENUM_BIT(BTRFS_ILOCK_MMAP),
548 int btrfs_inode_lock(struct btrfs_inode *inode, unsigned int ilock_flags);
549 void btrfs_inode_unlock(struct btrfs_inode *inode, unsigned int ilock_flags);
550 void btrfs_update_inode_bytes(struct btrfs_inode *inode, const u64 add_bytes,
551 const u64 del_bytes);
552 void btrfs_assert_inode_range_clean(struct btrfs_inode *inode, u64 start, u64 end);