Commit | Line | Data |
---|---|---|
9888c340 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
6cbd5570 CM |
2 | /* |
3 | * Copyright (C) 2007 Oracle. All rights reserved. | |
6cbd5570 CM |
4 | */ |
5 | ||
9888c340 DS |
6 | #ifndef BTRFS_INODE_H |
7 | #define BTRFS_INODE_H | |
2c90e5d6 | 8 | |
778ba82b | 9 | #include <linux/hash.h> |
e3b318d1 | 10 | #include <linux/refcount.h> |
a52d9a80 | 11 | #include "extent_map.h" |
d1310b2e | 12 | #include "extent_io.h" |
e6dcd2dc | 13 | #include "ordered-data.h" |
16cdcec7 | 14 | #include "delayed-inode.h" |
a52d9a80 | 15 | |
528ee697 FM |
16 | /* |
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()). | |
20 | */ | |
21 | #define BTRFS_DIR_START_INDEX 2 | |
22 | ||
72ac3c0d JB |
23 | /* |
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. | |
29 | */ | |
7efc3e34 | 30 | enum { |
1fd4033d | 31 | BTRFS_INODE_FLUSH_ON_CLOSE, |
7efc3e34 OS |
32 | BTRFS_INODE_DUMMY, |
33 | BTRFS_INODE_IN_DEFRAG, | |
34 | BTRFS_INODE_HAS_ASYNC_EXTENT, | |
48778179 FM |
35 | /* |
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). | |
39 | */ | |
7efc3e34 OS |
40 | BTRFS_INODE_NEEDS_FULL_SYNC, |
41 | BTRFS_INODE_COPY_EVERYTHING, | |
42 | BTRFS_INODE_IN_DELALLOC_LIST, | |
7efc3e34 | 43 | BTRFS_INODE_HAS_PROPS, |
3cd24c69 | 44 | BTRFS_INODE_SNAPSHOT_FLUSH, |
f2f121ab FM |
45 | /* |
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 | |
49 | * to an inode. | |
50 | */ | |
51 | BTRFS_INODE_NO_XATTRS, | |
3d45f221 FM |
52 | /* |
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). | |
59 | */ | |
60 | BTRFS_INODE_NO_DELALLOC_FLUSH, | |
14605409 BB |
61 | /* |
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 | |
65 | * on the same file. | |
66 | */ | |
67 | BTRFS_INODE_VERITY_IN_PROGRESS, | |
7efc3e34 | 68 | }; |
72ac3c0d | 69 | |
f1ace244 | 70 | /* in memory btrfs inode */ |
2c90e5d6 | 71 | struct btrfs_inode { |
d352ac68 | 72 | /* which subvolume this inode belongs to */ |
d6e4a428 | 73 | struct btrfs_root *root; |
d352ac68 | 74 | |
d352ac68 CM |
75 | /* key used to find this inode on disk. This is used by the code |
76 | * to read in roots of subvolumes | |
77 | */ | |
d6e4a428 | 78 | struct btrfs_key location; |
d352ac68 | 79 | |
2f2ff0ee FM |
80 | /* |
81 | * Lock for counters and all fields used to determine if the inode is in | |
82 | * the log or not (last_trans, last_sub_trans, last_log_commit, | |
2766ff61 FM |
83 | * logged_trans), to access/update new_delalloc_bytes and to update the |
84 | * VFS' inode number of bytes used. | |
2f2ff0ee | 85 | */ |
9e0baf60 JB |
86 | spinlock_t lock; |
87 | ||
d352ac68 | 88 | /* the extent_tree has caches of all the extent mappings to disk */ |
a52d9a80 | 89 | struct extent_map_tree extent_tree; |
d352ac68 CM |
90 | |
91 | /* the io_tree does range state (DIRTY, LOCKED etc) */ | |
d1310b2e | 92 | struct extent_io_tree io_tree; |
d352ac68 CM |
93 | |
94 | /* special utility tree used to record which mirrors have already been | |
95 | * tried when checksums fail for a given block | |
96 | */ | |
7e38326f | 97 | struct extent_io_tree io_failure_tree; |
d352ac68 | 98 | |
41a2ee75 JB |
99 | /* |
100 | * Keep track of where the inode has extent items mapped in order to | |
101 | * make sure the i_size adjustments are accurate | |
102 | */ | |
103 | struct extent_io_tree file_extent_tree; | |
104 | ||
d352ac68 | 105 | /* held while logging the inode in tree-log.c */ |
e02119d5 | 106 | struct mutex log_mutex; |
d352ac68 CM |
107 | |
108 | /* used to order data wrt metadata */ | |
e6dcd2dc | 109 | struct btrfs_ordered_inode_tree ordered_tree; |
15ee9bc7 | 110 | |
d352ac68 CM |
111 | /* list of all the delalloc inodes in the FS. There are times we need |
112 | * to write all the delalloc pages to disk, and this list is used | |
113 | * to walk them all. | |
114 | */ | |
ea8c2819 CM |
115 | struct list_head delalloc_inodes; |
116 | ||
5d4f98a2 YZ |
117 | /* node for the red-black tree that links inodes in subvolume root */ |
118 | struct rb_node rb_node; | |
119 | ||
72ac3c0d JB |
120 | unsigned long runtime_flags; |
121 | ||
9c931c5a | 122 | /* Keep track of who's O_SYNC/fsyncing currently */ |
b812ce28 JB |
123 | atomic_t sync_writers; |
124 | ||
d352ac68 CM |
125 | /* full 64 bit generation number, struct vfs_inode doesn't have a big |
126 | * enough field for this. | |
127 | */ | |
e02119d5 CM |
128 | u64 generation; |
129 | ||
15ee9bc7 JB |
130 | /* |
131 | * transid of the trans_handle that last modified this inode | |
132 | */ | |
133 | u64 last_trans; | |
257c62e1 CM |
134 | |
135 | /* | |
bb14a59b | 136 | * transid that last logged this inode |
257c62e1 | 137 | */ |
bb14a59b | 138 | u64 logged_trans; |
257c62e1 | 139 | |
e02119d5 | 140 | /* |
bb14a59b | 141 | * log transid when this inode was last modified |
e02119d5 | 142 | */ |
bb14a59b MX |
143 | int last_sub_trans; |
144 | ||
145 | /* a local copy of root's last_log_commit */ | |
146 | int last_log_commit; | |
49eb7e46 | 147 | |
339d0354 FM |
148 | /* |
149 | * Total number of bytes pending delalloc, used by stat to calculate the | |
150 | * real block usage of the file. This is used only for files. | |
151 | */ | |
152 | u64 delalloc_bytes; | |
dc287224 FM |
153 | |
154 | union { | |
155 | /* | |
156 | * Total number of bytes pending delalloc that fall within a file | |
157 | * range that is either a hole or beyond EOF (and no prealloc extent | |
158 | * exists in the range). This is always <= delalloc_bytes and this | |
159 | * is used only for files. | |
160 | */ | |
161 | u64 new_delalloc_bytes; | |
162 | /* | |
163 | * The offset of the last dir index key that was logged. | |
164 | * This is used only for directories. | |
165 | */ | |
166 | u64 last_dir_index_offset; | |
167 | }; | |
a7e3b975 | 168 | |
47059d93 WS |
169 | /* |
170 | * total number of bytes pending defrag, used by stat to check whether | |
171 | * it needs COW. | |
172 | */ | |
173 | u64 defrag_bytes; | |
174 | ||
d352ac68 CM |
175 | /* |
176 | * the size of the file stored in the metadata on disk. data=ordered | |
177 | * means the in-memory i_size might be larger than the size on disk | |
178 | * because not all the blocks are written yet. | |
179 | */ | |
dbe674a9 | 180 | u64 disk_i_size; |
d352ac68 | 181 | |
aec7477b | 182 | /* |
528ee697 FM |
183 | * If this is a directory then index_cnt is the counter for the index |
184 | * number for new files that are created. For an empty directory, this | |
185 | * must be initialized to BTRFS_DIR_START_INDEX. | |
aec7477b JB |
186 | */ |
187 | u64 index_cnt; | |
d352ac68 | 188 | |
67de1176 MX |
189 | /* Cache the directory index number to speed the dir/file remove */ |
190 | u64 dir_index; | |
191 | ||
12fcfd22 CM |
192 | /* the fsync log has some corner cases that mean we have to check |
193 | * directories to see if any unlinks have been done before | |
194 | * the directory was logged. See tree-log.c for all the | |
195 | * details | |
196 | */ | |
197 | u64 last_unlink_trans; | |
198 | ||
3ebac17c FM |
199 | /* |
200 | * The id/generation of the last transaction where this inode was | |
201 | * either the source or the destination of a clone/dedupe operation. | |
202 | * Used when logging an inode to know if there are shared extents that | |
203 | * need special care when logging checksum items, to avoid duplicate | |
204 | * checksum items in a log (which can lead to a corruption where we end | |
205 | * up with missing checksum ranges after log replay). | |
206 | * Protected by the vfs inode lock. | |
207 | */ | |
208 | u64 last_reflink_trans; | |
209 | ||
7709cde3 JB |
210 | /* |
211 | * Number of bytes outstanding that are going to need csums. This is | |
212 | * used in ENOSPC accounting. | |
213 | */ | |
214 | u64 csum_bytes; | |
215 | ||
77eea05e | 216 | /* Backwards incompatible flags, lower half of inode_item::flags */ |
f1bdcc0a | 217 | u32 flags; |
77eea05e BB |
218 | /* Read-only compatibility flags, upper half of inode_item::flags */ |
219 | u32 ro_flags; | |
f1bdcc0a | 220 | |
9ed74f2d | 221 | /* |
32c00aff JB |
222 | * Counters to keep track of the number of extent item's we may use due |
223 | * to delalloc and such. outstanding_extents is the number of extent | |
224 | * items we think we'll end up using, and reserved_extents is the number | |
225 | * of extent items we've reserved metadata for. | |
9ed74f2d | 226 | */ |
9e0baf60 | 227 | unsigned outstanding_extents; |
69fe2d75 JB |
228 | |
229 | struct btrfs_block_rsv block_rsv; | |
9ed74f2d | 230 | |
1e701a32 | 231 | /* |
b52aa8c9 | 232 | * Cached values of inode properties |
1e701a32 | 233 | */ |
b52aa8c9 | 234 | unsigned prop_compress; /* per-file compression algorithm */ |
eec63c65 DS |
235 | /* |
236 | * Force compression on the file using the defrag ioctl, could be | |
237 | * different from prop_compress and takes precedence if set | |
238 | */ | |
239 | unsigned defrag_compress; | |
1e701a32 | 240 | |
16cdcec7 MX |
241 | struct btrfs_delayed_node *delayed_node; |
242 | ||
9cc97d64 | 243 | /* File creation time. */ |
d3c6be6f | 244 | struct timespec64 i_otime; |
9cc97d64 | 245 | |
8089fe62 DS |
246 | /* Hook into fs_info->delayed_iputs */ |
247 | struct list_head delayed_iput; | |
8089fe62 | 248 | |
8318ba79 | 249 | struct rw_semaphore i_mmap_lock; |
d352ac68 | 250 | struct inode vfs_inode; |
2c90e5d6 | 251 | }; |
dbe674a9 | 252 | |
6fee248d NB |
253 | static inline u32 btrfs_inode_sectorsize(const struct btrfs_inode *inode) |
254 | { | |
255 | return inode->root->fs_info->sectorsize; | |
256 | } | |
257 | ||
9a35b637 | 258 | static inline struct btrfs_inode *BTRFS_I(const struct inode *inode) |
2c90e5d6 CM |
259 | { |
260 | return container_of(inode, struct btrfs_inode, vfs_inode); | |
261 | } | |
262 | ||
778ba82b FDBM |
263 | static inline unsigned long btrfs_inode_hash(u64 objectid, |
264 | const struct btrfs_root *root) | |
265 | { | |
4fd786e6 | 266 | u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME); |
778ba82b FDBM |
267 | |
268 | #if BITS_PER_LONG == 32 | |
269 | h = (h >> 32) ^ (h & 0xffffffff); | |
270 | #endif | |
271 | ||
272 | return (unsigned long)h; | |
273 | } | |
274 | ||
275 | static inline void btrfs_insert_inode_hash(struct inode *inode) | |
276 | { | |
277 | unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root); | |
278 | ||
279 | __insert_inode_hash(inode, h); | |
280 | } | |
281 | ||
9a35b637 | 282 | static inline u64 btrfs_ino(const struct btrfs_inode *inode) |
33345d01 | 283 | { |
4a0cc7ca | 284 | u64 ino = inode->location.objectid; |
33345d01 | 285 | |
14c7cca7 LB |
286 | /* |
287 | * !ino: btree_inode | |
288 | * type == BTRFS_ROOT_ITEM_KEY: subvol dir | |
289 | */ | |
4a0cc7ca NB |
290 | if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY) |
291 | ino = inode->vfs_inode.i_ino; | |
33345d01 LZ |
292 | return ino; |
293 | } | |
294 | ||
6ef06d27 | 295 | static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size) |
dbe674a9 | 296 | { |
6ef06d27 NB |
297 | i_size_write(&inode->vfs_inode, size); |
298 | inode->disk_i_size = size; | |
dbe674a9 CM |
299 | } |
300 | ||
70ddc553 | 301 | static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode) |
2cf8572d | 302 | { |
70ddc553 | 303 | struct btrfs_root *root = inode->root; |
83eea1f1 | 304 | |
51a8cf9d | 305 | if (root == root->fs_info->tree_root && |
70ddc553 | 306 | btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID) |
51a8cf9d | 307 | return true; |
70ddc553 | 308 | if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID) |
2cf8572d CM |
309 | return true; |
310 | return false; | |
311 | } | |
312 | ||
06f2548f NB |
313 | static inline bool is_data_inode(struct inode *inode) |
314 | { | |
315 | return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID; | |
316 | } | |
317 | ||
8b62f87b JB |
318 | static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode, |
319 | int mod) | |
320 | { | |
321 | lockdep_assert_held(&inode->lock); | |
322 | inode->outstanding_extents += mod; | |
323 | if (btrfs_is_free_space_inode(inode)) | |
324 | return; | |
dd48d407 JB |
325 | trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode), |
326 | mod); | |
8b62f87b JB |
327 | } |
328 | ||
bc0939fc FM |
329 | /* |
330 | * Called every time after doing a buffered, direct IO or memory mapped write. | |
331 | * | |
332 | * This is to ensure that if we write to a file that was previously fsynced in | |
333 | * the current transaction, then try to fsync it again in the same transaction, | |
334 | * we will know that there were changes in the file and that it needs to be | |
335 | * logged. | |
336 | */ | |
337 | static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode) | |
338 | { | |
339 | spin_lock(&inode->lock); | |
340 | inode->last_sub_trans = inode->root->log_transid; | |
341 | spin_unlock(&inode->lock); | |
342 | } | |
343 | ||
23e3337f FM |
344 | /* |
345 | * Should be called while holding the inode's VFS lock in exclusive mode or in a | |
346 | * context where no one else can access the inode concurrently (during inode | |
347 | * creation or when loading an inode from disk). | |
348 | */ | |
349 | static inline void btrfs_set_inode_full_sync(struct btrfs_inode *inode) | |
350 | { | |
351 | set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); | |
352 | /* | |
353 | * The inode may have been part of a reflink operation in the last | |
354 | * transaction that modified it, and then a fsync has reset the | |
355 | * last_reflink_trans to avoid subsequent fsyncs in the same | |
356 | * transaction to do unnecessary work. So update last_reflink_trans | |
357 | * to the last_trans value (we have to be pessimistic and assume a | |
358 | * reflink happened). | |
359 | * | |
360 | * The ->last_trans is protected by the inode's spinlock and we can | |
361 | * have a concurrent ordered extent completion update it. Also set | |
362 | * last_reflink_trans to ->last_trans only if the former is less than | |
363 | * the later, because we can be called in a context where | |
364 | * last_reflink_trans was set to the current transaction generation | |
365 | * while ->last_trans was not yet updated in the current transaction, | |
366 | * and therefore has a lower value. | |
367 | */ | |
368 | spin_lock(&inode->lock); | |
369 | if (inode->last_reflink_trans < inode->last_trans) | |
370 | inode->last_reflink_trans = inode->last_trans; | |
371 | spin_unlock(&inode->lock); | |
372 | } | |
373 | ||
209ecbb8 | 374 | static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation) |
22ee6985 | 375 | { |
209ecbb8 | 376 | bool ret = false; |
2f2ff0ee | 377 | |
0f8939b8 NB |
378 | spin_lock(&inode->lock); |
379 | if (inode->logged_trans == generation && | |
380 | inode->last_sub_trans <= inode->last_log_commit && | |
209ecbb8 FM |
381 | inode->last_sub_trans <= inode->root->last_log_commit) |
382 | ret = true; | |
0f8939b8 | 383 | spin_unlock(&inode->lock); |
2f2ff0ee | 384 | return ret; |
22ee6985 JB |
385 | } |
386 | ||
facc8a22 MX |
387 | struct btrfs_dio_private { |
388 | struct inode *inode; | |
47926ab5 QW |
389 | |
390 | /* | |
391 | * Since DIO can use anonymous page, we cannot use page_offset() to | |
392 | * grab the file offset, thus need a dedicated member for file offset. | |
393 | */ | |
394 | u64 file_offset; | |
facc8a22 | 395 | u64 disk_bytenr; |
523929f1 QW |
396 | /* Used for bio::bi_size */ |
397 | u32 bytes; | |
facc8a22 | 398 | |
e3b318d1 OS |
399 | /* |
400 | * References to this structure. There is one reference per in-flight | |
401 | * bio plus one while we're still setting up. | |
402 | */ | |
403 | refcount_t refs; | |
facc8a22 | 404 | |
facc8a22 MX |
405 | /* dio_bio came from fs/direct-io.c */ |
406 | struct bio *dio_bio; | |
c1dc0896 | 407 | |
85879573 OS |
408 | /* Array of checksums */ |
409 | u8 csums[]; | |
facc8a22 MX |
410 | }; |
411 | ||
77eea05e BB |
412 | /* |
413 | * btrfs_inode_item stores flags in a u64, btrfs_inode stores them in two | |
414 | * separate u32s. These two functions convert between the two representations. | |
415 | */ | |
416 | static inline u64 btrfs_inode_combine_flags(u32 flags, u32 ro_flags) | |
417 | { | |
418 | return (flags | ((u64)ro_flags << 32)); | |
419 | } | |
420 | ||
421 | static inline void btrfs_inode_split_flags(u64 inode_item_flags, | |
422 | u32 *flags, u32 *ro_flags) | |
423 | { | |
424 | *flags = (u32)inode_item_flags; | |
425 | *ro_flags = (u32)(inode_item_flags >> 32); | |
426 | } | |
427 | ||
7ebc7e5f JT |
428 | /* Array of bytes with variable length, hexadecimal format 0x1234 */ |
429 | #define CSUM_FMT "0x%*phN" | |
430 | #define CSUM_FMT_VALUE(size, bytes) size, bytes | |
431 | ||
0970a22e | 432 | static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode, |
ea41d6b2 | 433 | u64 logical_start, u8 *csum, u8 *csum_expected, int mirror_num) |
6f6b643e | 434 | { |
0970a22e | 435 | struct btrfs_root *root = inode->root; |
223486c2 | 436 | const u32 csum_size = root->fs_info->csum_size; |
6f6b643e QW |
437 | |
438 | /* Output minus objectid, which is more meaningful */ | |
4fd786e6 | 439 | if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID) |
6f6b643e | 440 | btrfs_warn_rl(root->fs_info, |
7ebc7e5f | 441 | "csum failed root %lld ino %lld off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d", |
4fd786e6 | 442 | root->root_key.objectid, btrfs_ino(inode), |
7ebc7e5f | 443 | logical_start, |
ea41d6b2 JT |
444 | CSUM_FMT_VALUE(csum_size, csum), |
445 | CSUM_FMT_VALUE(csum_size, csum_expected), | |
7ebc7e5f | 446 | mirror_num); |
6f6b643e QW |
447 | else |
448 | btrfs_warn_rl(root->fs_info, | |
7ebc7e5f | 449 | "csum failed root %llu ino %llu off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d", |
4fd786e6 | 450 | root->root_key.objectid, btrfs_ino(inode), |
7ebc7e5f | 451 | logical_start, |
ea41d6b2 JT |
452 | CSUM_FMT_VALUE(csum_size, csum), |
453 | CSUM_FMT_VALUE(csum_size, csum_expected), | |
7ebc7e5f | 454 | mirror_num); |
6f6b643e QW |
455 | } |
456 | ||
2c90e5d6 | 457 | #endif |