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,
72 /* Set when there are no capabilities in XATTs for the inode. */
73 BTRFS_INODE_NO_CAP_XATTR,
76 /* in memory btrfs inode */
78 /* which subvolume this inode belongs to */
79 struct btrfs_root *root;
81 /* key used to find this inode on disk. This is used by the code
82 * to read in roots of subvolumes
84 struct btrfs_key location;
86 /* Cached value of inode property 'compression'. */
90 * Force compression on the file using the defrag ioctl, could be
91 * different from prop_compress and takes precedence if set.
96 * Lock for counters and all fields used to determine if the inode is in
97 * the log or not (last_trans, last_sub_trans, last_log_commit,
98 * logged_trans), to access/update delalloc_bytes, new_delalloc_bytes,
99 * defrag_bytes, disk_i_size, outstanding_extents, csum_bytes and to
100 * update the VFS' inode number of bytes used.
104 /* the extent_tree has caches of all the extent mappings to disk */
105 struct extent_map_tree extent_tree;
107 /* the io_tree does range state (DIRTY, LOCKED etc) */
108 struct extent_io_tree io_tree;
111 * Keep track of where the inode has extent items mapped in order to
112 * make sure the i_size adjustments are accurate. Not required when the
113 * filesystem is NO_HOLES, the status can't be set while mounted as
114 * it's a mkfs-time feature.
116 struct extent_io_tree *file_extent_tree;
118 /* held while logging the inode in tree-log.c */
119 struct mutex log_mutex;
122 * Counters to keep track of the number of extent item's we may use due
123 * to delalloc and such. outstanding_extents is the number of extent
124 * items we think we'll end up using, and reserved_extents is the number
125 * of extent items we've reserved metadata for. Protected by 'lock'.
127 unsigned outstanding_extents;
129 /* used to order data wrt metadata */
130 spinlock_t ordered_tree_lock;
131 struct rb_root ordered_tree;
132 struct rb_node *ordered_tree_last;
134 /* list of all the delalloc inodes in the FS. There are times we need
135 * to write all the delalloc pages to disk, and this list is used
138 struct list_head delalloc_inodes;
140 /* node for the red-black tree that links inodes in subvolume root */
141 struct rb_node rb_node;
143 unsigned long runtime_flags;
145 /* full 64 bit generation number, struct vfs_inode doesn't have a big
146 * enough field for this.
151 * ID of the transaction handle that last modified this inode.
152 * Protected by 'lock'.
157 * ID of the transaction that last logged this inode.
158 * Protected by 'lock'.
163 * Log transaction ID when this inode was last modified.
164 * Protected by 'lock'.
168 /* A local copy of root's last_log_commit. Protected by 'lock'. */
173 * Total number of bytes pending delalloc, used by stat to
174 * calculate the real block usage of the file. This is used
175 * only for files. Protected by 'lock'.
179 * The lowest possible index of the next dir index key which
180 * points to an inode that needs to be logged.
181 * This is used only for directories.
182 * Use the helpers btrfs_get_first_dir_index_to_log() and
183 * btrfs_set_first_dir_index_to_log() to access this field.
185 u64 first_dir_index_to_log;
190 * Total number of bytes pending delalloc that fall within a file
191 * range that is either a hole or beyond EOF (and no prealloc extent
192 * exists in the range). This is always <= delalloc_bytes and this
193 * is used only for files. Protected by 'lock'.
195 u64 new_delalloc_bytes;
197 * The offset of the last dir index key that was logged.
198 * This is used only for directories.
200 u64 last_dir_index_offset;
204 * Total number of bytes pending defrag, used by stat to check whether
205 * it needs COW. Protected by 'lock'.
210 * The size of the file stored in the metadata on disk. data=ordered
211 * means the in-memory i_size might be larger than the size on disk
212 * because not all the blocks are written yet. Protected by 'lock'.
217 * If this is a directory then index_cnt is the counter for the index
218 * number for new files that are created. For an empty directory, this
219 * must be initialized to BTRFS_DIR_START_INDEX.
223 /* Cache the directory index number to speed the dir/file remove */
226 /* the fsync log has some corner cases that mean we have to check
227 * directories to see if any unlinks have been done before
228 * the directory was logged. See tree-log.c for all the
231 u64 last_unlink_trans;
234 * The id/generation of the last transaction where this inode was
235 * either the source or the destination of a clone/dedupe operation.
236 * Used when logging an inode to know if there are shared extents that
237 * need special care when logging checksum items, to avoid duplicate
238 * checksum items in a log (which can lead to a corruption where we end
239 * up with missing checksum ranges after log replay).
240 * Protected by the vfs inode lock.
242 u64 last_reflink_trans;
245 * Number of bytes outstanding that are going to need csums. This is
246 * used in ENOSPC accounting. Protected by 'lock'.
250 /* Backwards incompatible flags, lower half of inode_item::flags */
252 /* Read-only compatibility flags, upper half of inode_item::flags */
255 struct btrfs_block_rsv block_rsv;
257 struct btrfs_delayed_node *delayed_node;
259 /* File creation time. */
263 /* Hook into fs_info->delayed_iputs */
264 struct list_head delayed_iput;
266 struct rw_semaphore i_mmap_lock;
267 struct inode vfs_inode;
270 static inline u64 btrfs_get_first_dir_index_to_log(const struct btrfs_inode *inode)
272 return READ_ONCE(inode->first_dir_index_to_log);
275 static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode *inode,
278 WRITE_ONCE(inode->first_dir_index_to_log, index);
281 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
283 return container_of(inode, struct btrfs_inode, vfs_inode);
286 static inline unsigned long btrfs_inode_hash(u64 objectid,
287 const struct btrfs_root *root)
289 u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
291 #if BITS_PER_LONG == 32
292 h = (h >> 32) ^ (h & 0xffffffff);
295 return (unsigned long)h;
298 #if BITS_PER_LONG == 32
301 * On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so
302 * we use the inode's location objectid which is a u64 to avoid truncation.
304 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
306 u64 ino = inode->location.objectid;
308 /* type == BTRFS_ROOT_ITEM_KEY: subvol dir */
309 if (inode->location.type == BTRFS_ROOT_ITEM_KEY)
310 ino = inode->vfs_inode.i_ino;
316 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
318 return inode->vfs_inode.i_ino;
323 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
325 i_size_write(&inode->vfs_inode, size);
326 inode->disk_i_size = size;
329 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
331 return test_bit(BTRFS_INODE_FREE_SPACE_INODE, &inode->runtime_flags);
334 static inline bool is_data_inode(struct inode *inode)
336 return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
339 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
342 lockdep_assert_held(&inode->lock);
343 inode->outstanding_extents += mod;
344 if (btrfs_is_free_space_inode(inode))
346 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
347 mod, inode->outstanding_extents);
351 * Called every time after doing a buffered, direct IO or memory mapped write.
353 * This is to ensure that if we write to a file that was previously fsynced in
354 * the current transaction, then try to fsync it again in the same transaction,
355 * we will know that there were changes in the file and that it needs to be
358 static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)
360 spin_lock(&inode->lock);
361 inode->last_sub_trans = inode->root->log_transid;
362 spin_unlock(&inode->lock);
366 * Should be called while holding the inode's VFS lock in exclusive mode or in a
367 * context where no one else can access the inode concurrently (during inode
368 * creation or when loading an inode from disk).
370 static inline void btrfs_set_inode_full_sync(struct btrfs_inode *inode)
372 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
374 * The inode may have been part of a reflink operation in the last
375 * transaction that modified it, and then a fsync has reset the
376 * last_reflink_trans to avoid subsequent fsyncs in the same
377 * transaction to do unnecessary work. So update last_reflink_trans
378 * to the last_trans value (we have to be pessimistic and assume a
381 * The ->last_trans is protected by the inode's spinlock and we can
382 * have a concurrent ordered extent completion update it. Also set
383 * last_reflink_trans to ->last_trans only if the former is less than
384 * the later, because we can be called in a context where
385 * last_reflink_trans was set to the current transaction generation
386 * while ->last_trans was not yet updated in the current transaction,
387 * and therefore has a lower value.
389 spin_lock(&inode->lock);
390 if (inode->last_reflink_trans < inode->last_trans)
391 inode->last_reflink_trans = inode->last_trans;
392 spin_unlock(&inode->lock);
395 static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
399 spin_lock(&inode->lock);
400 if (inode->logged_trans == generation &&
401 inode->last_sub_trans <= inode->last_log_commit &&
402 inode->last_sub_trans <= btrfs_get_root_last_log_commit(inode->root))
404 spin_unlock(&inode->lock);
409 * Check if the inode has flags compatible with compression
411 static inline bool btrfs_inode_can_compress(const struct btrfs_inode *inode)
413 if (inode->flags & BTRFS_INODE_NODATACOW ||
414 inode->flags & BTRFS_INODE_NODATASUM)
419 /* Array of bytes with variable length, hexadecimal format 0x1234 */
420 #define CSUM_FMT "0x%*phN"
421 #define CSUM_FMT_VALUE(size, bytes) size, bytes
423 int btrfs_check_sector_csum(struct btrfs_fs_info *fs_info, struct page *page,
424 u32 pgoff, u8 *csum, const u8 * const csum_expected);
425 bool btrfs_data_csum_ok(struct btrfs_bio *bbio, struct btrfs_device *dev,
426 u32 bio_offset, struct bio_vec *bv);
427 noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
428 u64 *orig_start, u64 *orig_block_len,
429 u64 *ram_bytes, bool nowait, bool strict);
431 void __btrfs_del_delalloc_inode(struct btrfs_root *root, struct btrfs_inode *inode);
432 struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);
433 int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index);
434 int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
435 struct btrfs_inode *dir, struct btrfs_inode *inode,
436 const struct fscrypt_str *name);
437 int btrfs_add_link(struct btrfs_trans_handle *trans,
438 struct btrfs_inode *parent_inode, struct btrfs_inode *inode,
439 const struct fscrypt_str *name, int add_backref, u64 index);
440 int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry);
441 int btrfs_truncate_block(struct btrfs_inode *inode, loff_t from, loff_t len,
444 int btrfs_start_delalloc_snapshot(struct btrfs_root *root, bool in_reclaim_context);
445 int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, long nr,
446 bool in_reclaim_context);
447 int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
448 unsigned int extra_bits,
449 struct extent_state **cached_state);
451 struct btrfs_new_inode_args {
454 struct dentry *dentry;
459 /* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */
460 struct posix_acl *default_acl;
461 struct posix_acl *acl;
462 struct fscrypt_name fname;
465 int btrfs_new_inode_prepare(struct btrfs_new_inode_args *args,
466 unsigned int *trans_num_items);
467 int btrfs_create_new_inode(struct btrfs_trans_handle *trans,
468 struct btrfs_new_inode_args *args);
469 void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args *args);
470 struct inode *btrfs_new_subvol_inode(struct mnt_idmap *idmap,
472 void btrfs_set_delalloc_extent(struct btrfs_inode *inode, struct extent_state *state,
474 void btrfs_clear_delalloc_extent(struct btrfs_inode *inode,
475 struct extent_state *state, u32 bits);
476 void btrfs_merge_delalloc_extent(struct btrfs_inode *inode, struct extent_state *new,
477 struct extent_state *other);
478 void btrfs_split_delalloc_extent(struct btrfs_inode *inode,
479 struct extent_state *orig, u64 split);
480 void btrfs_set_range_writeback(struct btrfs_inode *inode, u64 start, u64 end);
481 vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf);
482 void btrfs_evict_inode(struct inode *inode);
483 struct inode *btrfs_alloc_inode(struct super_block *sb);
484 void btrfs_destroy_inode(struct inode *inode);
485 void btrfs_free_inode(struct inode *inode);
486 int btrfs_drop_inode(struct inode *inode);
487 int __init btrfs_init_cachep(void);
488 void __cold btrfs_destroy_cachep(void);
489 struct inode *btrfs_iget_path(struct super_block *s, u64 ino,
490 struct btrfs_root *root, struct btrfs_path *path);
491 struct inode *btrfs_iget(struct super_block *s, u64 ino, struct btrfs_root *root);
492 struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,
493 struct page *page, size_t pg_offset,
495 int btrfs_update_inode(struct btrfs_trans_handle *trans,
496 struct btrfs_inode *inode);
497 int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
498 struct btrfs_inode *inode);
499 int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct btrfs_inode *inode);
500 int btrfs_orphan_cleanup(struct btrfs_root *root);
501 int btrfs_cont_expand(struct btrfs_inode *inode, loff_t oldsize, loff_t size);
502 void btrfs_add_delayed_iput(struct btrfs_inode *inode);
503 void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info);
504 int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info);
505 int btrfs_prealloc_file_range(struct inode *inode, int mode,
506 u64 start, u64 num_bytes, u64 min_size,
507 loff_t actual_len, u64 *alloc_hint);
508 int btrfs_prealloc_file_range_trans(struct inode *inode,
509 struct btrfs_trans_handle *trans, int mode,
510 u64 start, u64 num_bytes, u64 min_size,
511 loff_t actual_len, u64 *alloc_hint);
512 int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct page *locked_page,
513 u64 start, u64 end, struct writeback_control *wbc);
514 int btrfs_writepage_cow_fixup(struct page *page);
515 int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info *fs_info,
517 int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode *inode,
518 u64 file_offset, u64 disk_bytenr,
520 struct page **pages);
521 ssize_t btrfs_encoded_read(struct kiocb *iocb, struct iov_iter *iter,
522 struct btrfs_ioctl_encoded_io_args *encoded);
523 ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,
524 const struct btrfs_ioctl_encoded_io_args *encoded);
526 ssize_t btrfs_dio_read(struct kiocb *iocb, struct iov_iter *iter,
528 struct iomap_dio *btrfs_dio_write(struct kiocb *iocb, struct iov_iter *iter,
531 extern const struct dentry_operations btrfs_dentry_operations;
533 /* Inode locking type flags, by default the exclusive lock is taken. */
534 enum btrfs_ilock_type {
535 ENUM_BIT(BTRFS_ILOCK_SHARED),
536 ENUM_BIT(BTRFS_ILOCK_TRY),
537 ENUM_BIT(BTRFS_ILOCK_MMAP),
540 int btrfs_inode_lock(struct btrfs_inode *inode, unsigned int ilock_flags);
541 void btrfs_inode_unlock(struct btrfs_inode *inode, unsigned int ilock_flags);
542 void btrfs_update_inode_bytes(struct btrfs_inode *inode, const u64 add_bytes,
543 const u64 del_bytes);
544 void btrfs_assert_inode_range_clean(struct btrfs_inode *inode, u64 start, u64 end);