1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6 #ifndef BTRFS_INODE_H
7 #define BTRFS_INODE_H
8
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>
14 #include <linux/fs.h>
15 #include <linux/mm.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 "btrfs_inode.h"
23 #include "extent_map.h"
24 #include "extent_io.h"
25 #include "extent-io-tree.h"
26 #include "ordered-data.h"
27 #include "delayed-inode.h"
28
29 struct extent_state;
30 struct posix_acl;
31 struct iov_iter;
32 struct writeback_control;
33 struct btrfs_root;
34 struct btrfs_fs_info;
35 struct btrfs_trans_handle;
36
37 /*
38 * Since we search a directory based on f_pos (struct dir_context::pos) we have
39 * to start at 2 since '.' and '..' have f_pos of 0 and 1 respectively, so
40 * everybody else has to start at 2 (see btrfs_real_readdir() and dir_emit_dots()).
41 */
42 #define BTRFS_DIR_START_INDEX 2
43
44 /*
45 * ordered_data_close is set by truncate when a file that used
46 * to have good data has been truncated to zero. When it is set
47 * the btrfs file release call will add this inode to the
48 * ordered operations list so that we make sure to flush out any
49 * new data the application may have written before commit.
50 */
51 enum {
52 BTRFS_INODE_FLUSH_ON_CLOSE,
53 BTRFS_INODE_DUMMY,
54 BTRFS_INODE_IN_DEFRAG,
55 BTRFS_INODE_HAS_ASYNC_EXTENT,
56 /*
57 * Always set under the VFS' inode lock, otherwise it can cause races
58 * during fsync (we start as a fast fsync and then end up in a full
59 * fsync racing with ordered extent completion).
60 */
61 BTRFS_INODE_NEEDS_FULL_SYNC,
62 BTRFS_INODE_COPY_EVERYTHING,
63 BTRFS_INODE_HAS_PROPS,
64 BTRFS_INODE_SNAPSHOT_FLUSH,
65 /*
66 * Set and used when logging an inode and it serves to signal that an
67 * inode does not have xattrs, so subsequent fsyncs can avoid searching
68 * for xattrs to log. This bit must be cleared whenever a xattr is added
69 * to an inode.
70 */
71 BTRFS_INODE_NO_XATTRS,
72 /*
73 * Set when we are in a context where we need to start a transaction and
74 * have dirty pages with the respective file range locked. This is to
75 * ensure that when reserving space for the transaction, if we are low
76 * on available space and need to flush delalloc, we will not flush
77 * delalloc for this inode, because that could result in a deadlock (on
78 * the file range, inode's io_tree).
79 */
80 BTRFS_INODE_NO_DELALLOC_FLUSH,
81 /*
82 * Set when we are working on enabling verity for a file. Computing and
83 * writing the whole Merkle tree can take a while so we want to prevent
84 * races where two separate tasks attempt to simultaneously start verity
85 * on the same file.
86 */
87 BTRFS_INODE_VERITY_IN_PROGRESS,
88 /* Set when this inode is a free space inode. */
89 BTRFS_INODE_FREE_SPACE_INODE,
90 /* Set when there are no capabilities in XATTs for the inode. */
91 BTRFS_INODE_NO_CAP_XATTR,
92 /*
93 * Set if an error happened when doing a COW write before submitting a
94 * bio or during writeback. Used for both buffered writes and direct IO
95 * writes. This is to signal a fast fsync that it has to wait for
96 * ordered extents to complete and therefore not log extent maps that
97 * point to unwritten extents (when an ordered extent completes and it
98 * has the BTRFS_ORDERED_IOERR flag set, it drops extent maps in its
99 * range).
100 */
101 BTRFS_INODE_COW_WRITE_ERROR,
102 };
103
104 /* in memory btrfs inode */
105 struct btrfs_inode {
106 /* which subvolume this inode belongs to */
107 struct btrfs_root *root;
108
109 /* key used to find this inode on disk. This is used by the code
110 * to read in roots of subvolumes
111 */
112 struct btrfs_key location;
113
114 /* Cached value of inode property 'compression'. */
115 u8 prop_compress;
116
117 /*
118 * Force compression on the file using the defrag ioctl, could be
119 * different from prop_compress and takes precedence if set.
120 */
121 u8 defrag_compress;
122
123 /*
124 * Lock for counters and all fields used to determine if the inode is in
125 * the log or not (last_trans, last_sub_trans, last_log_commit,
126 * logged_trans), to access/update delalloc_bytes, new_delalloc_bytes,
127 * defrag_bytes, disk_i_size, outstanding_extents, csum_bytes and to
128 * update the VFS' inode number of bytes used.
129 */
130 spinlock_t lock;
131
132 /* the extent_tree has caches of all the extent mappings to disk */
133 struct extent_map_tree extent_tree;
134
135 /* the io_tree does range state (DIRTY, LOCKED etc) */
136 struct extent_io_tree io_tree;
137
138 /*
139 * Keep track of where the inode has extent items mapped in order to
140 * make sure the i_size adjustments are accurate. Not required when the
141 * filesystem is NO_HOLES, the status can't be set while mounted as
142 * it's a mkfs-time feature.
143 */
144 struct extent_io_tree *file_extent_tree;
145
146 /* held while logging the inode in tree-log.c */
147 struct mutex log_mutex;
148
149 /*
150 * Counters to keep track of the number of extent item's we may use due
151 * to delalloc and such. outstanding_extents is the number of extent
152 * items we think we'll end up using, and reserved_extents is the number
153 * of extent items we've reserved metadata for. Protected by 'lock'.
154 */
155 unsigned outstanding_extents;
156
157 /* used to order data wrt metadata */
158 spinlock_t ordered_tree_lock;
159 struct rb_root ordered_tree;
160 struct rb_node *ordered_tree_last;
161
162 /* list of all the delalloc inodes in the FS. There are times we need
163 * to write all the delalloc pages to disk, and this list is used
164 * to walk them all.
165 */
166 struct list_head delalloc_inodes;
167
168 /* node for the red-black tree that links inodes in subvolume root */
169 struct rb_node rb_node;
170
171 unsigned long runtime_flags;
172
173 /* full 64 bit generation number, struct vfs_inode doesn't have a big
174 * enough field for this.
175 */
176 u64 generation;
177
178 /*
179 * ID of the transaction handle that last modified this inode.
180 * Protected by 'lock'.
181 */
182 u64 last_trans;
183
184 /*
185 * ID of the transaction that last logged this inode.
186 * Protected by 'lock'.
187 */
188 u64 logged_trans;
189
190 /*
191 * Log transaction ID when this inode was last modified.
192 * Protected by 'lock'.
193 */
194 int last_sub_trans;
195
196 /* A local copy of root's last_log_commit. Protected by 'lock'. */
197 int last_log_commit;
198
199 union {
200 /*
201 * Total number of bytes pending delalloc, used by stat to
202 * calculate the real block usage of the file. This is used
203 * only for files. Protected by 'lock'.
204 */
205 u64 delalloc_bytes;
206 /*
207 * The lowest possible index of the next dir index key which
208 * points to an inode that needs to be logged.
209 * This is used only for directories.
210 * Use the helpers btrfs_get_first_dir_index_to_log() and
211 * btrfs_set_first_dir_index_to_log() to access this field.
212 */
213 u64 first_dir_index_to_log;
214 };
215
216 union {
217 /*
218 * Total number of bytes pending delalloc that fall within a file
219 * range that is either a hole or beyond EOF (and no prealloc extent
220 * exists in the range). This is always <= delalloc_bytes and this
221 * is used only for files. Protected by 'lock'.
222 */
223 u64 new_delalloc_bytes;
224 /*
225 * The offset of the last dir index key that was logged.
226 * This is used only for directories.
227 */
228 u64 last_dir_index_offset;
229 };
230
231 /*
232 * Total number of bytes pending defrag, used by stat to check whether
233 * it needs COW. Protected by 'lock'.
234 */
235 u64 defrag_bytes;
236
237 /*
238 * The size of the file stored in the metadata on disk. data=ordered
239 * means the in-memory i_size might be larger than the size on disk
240 * because not all the blocks are written yet. Protected by 'lock'.
241 */
242 u64 disk_i_size;
243
244 /*
245 * If this is a directory then index_cnt is the counter for the index
246 * number for new files that are created. For an empty directory, this
247 * must be initialized to BTRFS_DIR_START_INDEX.
248 */
249 u64 index_cnt;
250
251 /* Cache the directory index number to speed the dir/file remove */
252 u64 dir_index;
253
254 /* the fsync log has some corner cases that mean we have to check
255 * directories to see if any unlinks have been done before
256 * the directory was logged. See tree-log.c for all the
257 * details
258 */
259 u64 last_unlink_trans;
260
261 /*
262 * The id/generation of the last transaction where this inode was
263 * either the source or the destination of a clone/dedupe operation.
264 * Used when logging an inode to know if there are shared extents that
265 * need special care when logging checksum items, to avoid duplicate
266 * checksum items in a log (which can lead to a corruption where we end
267 * up with missing checksum ranges after log replay).
268 * Protected by the vfs inode lock.
269 */
270 u64 last_reflink_trans;
271
272 /*
273 * Number of bytes outstanding that are going to need csums. This is
274 * used in ENOSPC accounting. Protected by 'lock'.
275 */
276 u64 csum_bytes;
277
278 /* Backwards incompatible flags, lower half of inode_item::flags */
279 u32 flags;
280 /* Read-only compatibility flags, upper half of inode_item::flags */
281 u32 ro_flags;
282
283 struct btrfs_block_rsv block_rsv;
284
285 struct btrfs_delayed_node *delayed_node;
286
287 /* File creation time. */
288 u64 i_otime_sec;
289 u32 i_otime_nsec;
290
291 /* Hook into fs_info->delayed_iputs */
292 struct list_head delayed_iput;
293
294 struct rw_semaphore i_mmap_lock;
295 struct inode vfs_inode;
296 };
297
btrfs_get_first_dir_index_to_log(const struct btrfs_inode * inode)298 static inline u64 btrfs_get_first_dir_index_to_log(const struct btrfs_inode *inode)
299 {
300 return READ_ONCE(inode->first_dir_index_to_log);
301 }
302
btrfs_set_first_dir_index_to_log(struct btrfs_inode * inode,u64 index)303 static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode *inode,
304 u64 index)
305 {
306 WRITE_ONCE(inode->first_dir_index_to_log, index);
307 }
308
BTRFS_I(const struct inode * inode)309 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
310 {
311 return container_of(inode, struct btrfs_inode, vfs_inode);
312 }
313
btrfs_inode_hash(u64 objectid,const struct btrfs_root * root)314 static inline unsigned long btrfs_inode_hash(u64 objectid,
315 const struct btrfs_root *root)
316 {
317 u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
318
319 #if BITS_PER_LONG == 32
320 h = (h >> 32) ^ (h & 0xffffffff);
321 #endif
322
323 return (unsigned long)h;
324 }
325
326 #if BITS_PER_LONG == 32
327
328 /*
329 * On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so
330 * we use the inode's location objectid which is a u64 to avoid truncation.
331 */
btrfs_ino(const struct btrfs_inode * inode)332 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
333 {
334 u64 ino = inode->location.objectid;
335
336 /* type == BTRFS_ROOT_ITEM_KEY: subvol dir */
337 if (inode->location.type == BTRFS_ROOT_ITEM_KEY)
338 ino = inode->vfs_inode.i_ino;
339 return ino;
340 }
341
342 #else
343
btrfs_ino(const struct btrfs_inode * inode)344 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
345 {
346 return inode->vfs_inode.i_ino;
347 }
348
349 #endif
350
btrfs_i_size_write(struct btrfs_inode * inode,u64 size)351 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
352 {
353 i_size_write(&inode->vfs_inode, size);
354 inode->disk_i_size = size;
355 }
356
btrfs_is_free_space_inode(struct btrfs_inode * inode)357 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
358 {
359 return test_bit(BTRFS_INODE_FREE_SPACE_INODE, &inode->runtime_flags);
360 }
361
is_data_inode(struct inode * inode)362 static inline bool is_data_inode(struct inode *inode)
363 {
364 return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
365 }
366
btrfs_mod_outstanding_extents(struct btrfs_inode * inode,int mod)367 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
368 int mod)
369 {
370 lockdep_assert_held(&inode->lock);
371 inode->outstanding_extents += mod;
372 if (btrfs_is_free_space_inode(inode))
373 return;
374 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
375 mod, inode->outstanding_extents);
376 }
377
378 /*
379 * Called every time after doing a buffered, direct IO or memory mapped write.
380 *
381 * This is to ensure that if we write to a file that was previously fsynced in
382 * the current transaction, then try to fsync it again in the same transaction,
383 * we will know that there were changes in the file and that it needs to be
384 * logged.
385 */
btrfs_set_inode_last_sub_trans(struct btrfs_inode * inode)386 static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)
387 {
388 spin_lock(&inode->lock);
389 inode->last_sub_trans = inode->root->log_transid;
390 spin_unlock(&inode->lock);
391 }
392
393 /*
394 * Should be called while holding the inode's VFS lock in exclusive mode, or
395 * while holding the inode's mmap lock (struct btrfs_inode::i_mmap_lock) in
396 * either shared or exclusive mode, or in a context where no one else can access
397 * the inode concurrently (during inode creation or when loading an inode from
398 * disk).
399 */
btrfs_set_inode_full_sync(struct btrfs_inode * inode)400 static inline void btrfs_set_inode_full_sync(struct btrfs_inode *inode)
401 {
402 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
403 /*
404 * The inode may have been part of a reflink operation in the last
405 * transaction that modified it, and then a fsync has reset the
406 * last_reflink_trans to avoid subsequent fsyncs in the same
407 * transaction to do unnecessary work. So update last_reflink_trans
408 * to the last_trans value (we have to be pessimistic and assume a
409 * reflink happened).
410 *
411 * The ->last_trans is protected by the inode's spinlock and we can
412 * have a concurrent ordered extent completion update it. Also set
413 * last_reflink_trans to ->last_trans only if the former is less than
414 * the later, because we can be called in a context where
415 * last_reflink_trans was set to the current transaction generation
416 * while ->last_trans was not yet updated in the current transaction,
417 * and therefore has a lower value.
418 */
419 spin_lock(&inode->lock);
420 if (inode->last_reflink_trans < inode->last_trans)
421 inode->last_reflink_trans = inode->last_trans;
422 spin_unlock(&inode->lock);
423 }
424
btrfs_inode_in_log(struct btrfs_inode * inode,u64 generation)425 static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
426 {
427 bool ret = false;
428
429 spin_lock(&inode->lock);
430 if (inode->logged_trans == generation &&
431 inode->last_sub_trans <= inode->last_log_commit &&
432 inode->last_sub_trans <= btrfs_get_root_last_log_commit(inode->root))
433 ret = true;
434 spin_unlock(&inode->lock);
435 return ret;
436 }
437
438 /*
439 * Check if the inode has flags compatible with compression
440 */
btrfs_inode_can_compress(const struct btrfs_inode * inode)441 static inline bool btrfs_inode_can_compress(const struct btrfs_inode *inode)
442 {
443 if (inode->flags & BTRFS_INODE_NODATACOW ||
444 inode->flags & BTRFS_INODE_NODATASUM)
445 return false;
446 return true;
447 }
448
449 /* Array of bytes with variable length, hexadecimal format 0x1234 */
450 #define CSUM_FMT "0x%*phN"
451 #define CSUM_FMT_VALUE(size, bytes) size, bytes
452
453 int btrfs_check_sector_csum(struct btrfs_fs_info *fs_info, struct page *page,
454 u32 pgoff, u8 *csum, const u8 * const csum_expected);
455 bool btrfs_data_csum_ok(struct btrfs_bio *bbio, struct btrfs_device *dev,
456 u32 bio_offset, struct bio_vec *bv);
457 noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
458 u64 *orig_start, u64 *orig_block_len,
459 u64 *ram_bytes, bool nowait, bool strict);
460
461 void btrfs_del_delalloc_inode(struct btrfs_inode *inode);
462 struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);
463 int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index);
464 int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
465 struct btrfs_inode *dir, struct btrfs_inode *inode,
466 const struct fscrypt_str *name);
467 int btrfs_add_link(struct btrfs_trans_handle *trans,
468 struct btrfs_inode *parent_inode, struct btrfs_inode *inode,
469 const struct fscrypt_str *name, int add_backref, u64 index);
470 int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry);
471 int btrfs_truncate_block(struct btrfs_inode *inode, loff_t from, loff_t len,
472 int front);
473
474 int btrfs_start_delalloc_snapshot(struct btrfs_root *root, bool in_reclaim_context);
475 int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, long nr,
476 bool in_reclaim_context);
477 int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
478 unsigned int extra_bits,
479 struct extent_state **cached_state);
480
481 struct btrfs_new_inode_args {
482 /* Input */
483 struct inode *dir;
484 struct dentry *dentry;
485 struct inode *inode;
486 bool orphan;
487 bool subvol;
488
489 /* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */
490 struct posix_acl *default_acl;
491 struct posix_acl *acl;
492 struct fscrypt_name fname;
493 };
494
495 int btrfs_new_inode_prepare(struct btrfs_new_inode_args *args,
496 unsigned int *trans_num_items);
497 int btrfs_create_new_inode(struct btrfs_trans_handle *trans,
498 struct btrfs_new_inode_args *args);
499 void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args *args);
500 struct inode *btrfs_new_subvol_inode(struct mnt_idmap *idmap,
501 struct inode *dir);
502 void btrfs_set_delalloc_extent(struct btrfs_inode *inode, struct extent_state *state,
503 u32 bits);
504 void btrfs_clear_delalloc_extent(struct btrfs_inode *inode,
505 struct extent_state *state, u32 bits);
506 void btrfs_merge_delalloc_extent(struct btrfs_inode *inode, struct extent_state *new,
507 struct extent_state *other);
508 void btrfs_split_delalloc_extent(struct btrfs_inode *inode,
509 struct extent_state *orig, u64 split);
510 void btrfs_set_range_writeback(struct btrfs_inode *inode, u64 start, u64 end);
511 void btrfs_evict_inode(struct inode *inode);
512 struct inode *btrfs_alloc_inode(struct super_block *sb);
513 void btrfs_destroy_inode(struct inode *inode);
514 void btrfs_free_inode(struct inode *inode);
515 int btrfs_drop_inode(struct inode *inode);
516 int __init btrfs_init_cachep(void);
517 void __cold btrfs_destroy_cachep(void);
518 struct inode *btrfs_iget_path(struct super_block *s, u64 ino,
519 struct btrfs_root *root, struct btrfs_path *path);
520 struct inode *btrfs_iget(struct super_block *s, u64 ino, struct btrfs_root *root);
521 struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,
522 struct page *page, u64 start, u64 len);
523 int btrfs_update_inode(struct btrfs_trans_handle *trans,
524 struct btrfs_inode *inode);
525 int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
526 struct btrfs_inode *inode);
527 int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct btrfs_inode *inode);
528 int btrfs_orphan_cleanup(struct btrfs_root *root);
529 int btrfs_cont_expand(struct btrfs_inode *inode, loff_t oldsize, loff_t size);
530 void btrfs_add_delayed_iput(struct btrfs_inode *inode);
531 void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info);
532 int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info);
533 int btrfs_prealloc_file_range(struct inode *inode, int mode,
534 u64 start, u64 num_bytes, u64 min_size,
535 loff_t actual_len, u64 *alloc_hint);
536 int btrfs_prealloc_file_range_trans(struct inode *inode,
537 struct btrfs_trans_handle *trans, int mode,
538 u64 start, u64 num_bytes, u64 min_size,
539 loff_t actual_len, u64 *alloc_hint);
540 int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct page *locked_page,
541 u64 start, u64 end, struct writeback_control *wbc);
542 int btrfs_writepage_cow_fixup(struct page *page);
543 int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info *fs_info,
544 int compress_type);
545 int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode *inode,
546 u64 file_offset, u64 disk_bytenr,
547 u64 disk_io_size,
548 struct page **pages);
549 ssize_t btrfs_encoded_read(struct kiocb *iocb, struct iov_iter *iter,
550 struct btrfs_ioctl_encoded_io_args *encoded);
551 ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,
552 const struct btrfs_ioctl_encoded_io_args *encoded);
553
554 ssize_t btrfs_dio_read(struct kiocb *iocb, struct iov_iter *iter,
555 size_t done_before);
556 struct iomap_dio *btrfs_dio_write(struct kiocb *iocb, struct iov_iter *iter,
557 size_t done_before);
558 struct btrfs_inode *btrfs_find_first_inode(struct btrfs_root *root, u64 min_ino);
559
560 extern const struct dentry_operations btrfs_dentry_operations;
561
562 /* Inode locking type flags, by default the exclusive lock is taken. */
563 enum btrfs_ilock_type {
564 ENUM_BIT(BTRFS_ILOCK_SHARED),
565 ENUM_BIT(BTRFS_ILOCK_TRY),
566 ENUM_BIT(BTRFS_ILOCK_MMAP),
567 };
568
569 int btrfs_inode_lock(struct btrfs_inode *inode, unsigned int ilock_flags);
570 void btrfs_inode_unlock(struct btrfs_inode *inode, unsigned int ilock_flags);
571 void btrfs_update_inode_bytes(struct btrfs_inode *inode, const u64 add_bytes,
572 const u64 del_bytes);
573 void btrfs_assert_inode_range_clean(struct btrfs_inode *inode, u64 start, u64 end);
574
575 #endif
576