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 | |
72ac3c0d JB |
16 | /* |
17 | * ordered_data_close is set by truncate when a file that used | |
18 | * to have good data has been truncated to zero. When it is set | |
19 | * the btrfs file release call will add this inode to the | |
20 | * ordered operations list so that we make sure to flush out any | |
21 | * new data the application may have written before commit. | |
22 | */ | |
7efc3e34 | 23 | enum { |
1fd4033d | 24 | BTRFS_INODE_FLUSH_ON_CLOSE, |
7efc3e34 OS |
25 | BTRFS_INODE_DUMMY, |
26 | BTRFS_INODE_IN_DEFRAG, | |
27 | BTRFS_INODE_HAS_ASYNC_EXTENT, | |
48778179 FM |
28 | /* |
29 | * Always set under the VFS' inode lock, otherwise it can cause races | |
30 | * during fsync (we start as a fast fsync and then end up in a full | |
31 | * fsync racing with ordered extent completion). | |
32 | */ | |
7efc3e34 OS |
33 | BTRFS_INODE_NEEDS_FULL_SYNC, |
34 | BTRFS_INODE_COPY_EVERYTHING, | |
35 | BTRFS_INODE_IN_DELALLOC_LIST, | |
7efc3e34 | 36 | BTRFS_INODE_HAS_PROPS, |
3cd24c69 | 37 | BTRFS_INODE_SNAPSHOT_FLUSH, |
f2f121ab FM |
38 | /* |
39 | * Set and used when logging an inode and it serves to signal that an | |
40 | * inode does not have xattrs, so subsequent fsyncs can avoid searching | |
41 | * for xattrs to log. This bit must be cleared whenever a xattr is added | |
42 | * to an inode. | |
43 | */ | |
44 | BTRFS_INODE_NO_XATTRS, | |
3d45f221 FM |
45 | /* |
46 | * Set when we are in a context where we need to start a transaction and | |
47 | * have dirty pages with the respective file range locked. This is to | |
48 | * ensure that when reserving space for the transaction, if we are low | |
49 | * on available space and need to flush delalloc, we will not flush | |
50 | * delalloc for this inode, because that could result in a deadlock (on | |
51 | * the file range, inode's io_tree). | |
52 | */ | |
53 | BTRFS_INODE_NO_DELALLOC_FLUSH, | |
7efc3e34 | 54 | }; |
72ac3c0d | 55 | |
f1ace244 | 56 | /* in memory btrfs inode */ |
2c90e5d6 | 57 | struct btrfs_inode { |
d352ac68 | 58 | /* which subvolume this inode belongs to */ |
d6e4a428 | 59 | struct btrfs_root *root; |
d352ac68 | 60 | |
d352ac68 CM |
61 | /* key used to find this inode on disk. This is used by the code |
62 | * to read in roots of subvolumes | |
63 | */ | |
d6e4a428 | 64 | struct btrfs_key location; |
d352ac68 | 65 | |
2f2ff0ee FM |
66 | /* |
67 | * Lock for counters and all fields used to determine if the inode is in | |
68 | * the log or not (last_trans, last_sub_trans, last_log_commit, | |
2766ff61 FM |
69 | * logged_trans), to access/update new_delalloc_bytes and to update the |
70 | * VFS' inode number of bytes used. | |
2f2ff0ee | 71 | */ |
9e0baf60 JB |
72 | spinlock_t lock; |
73 | ||
d352ac68 | 74 | /* the extent_tree has caches of all the extent mappings to disk */ |
a52d9a80 | 75 | struct extent_map_tree extent_tree; |
d352ac68 CM |
76 | |
77 | /* the io_tree does range state (DIRTY, LOCKED etc) */ | |
d1310b2e | 78 | struct extent_io_tree io_tree; |
d352ac68 CM |
79 | |
80 | /* special utility tree used to record which mirrors have already been | |
81 | * tried when checksums fail for a given block | |
82 | */ | |
7e38326f | 83 | struct extent_io_tree io_failure_tree; |
d352ac68 | 84 | |
41a2ee75 JB |
85 | /* |
86 | * Keep track of where the inode has extent items mapped in order to | |
87 | * make sure the i_size adjustments are accurate | |
88 | */ | |
89 | struct extent_io_tree file_extent_tree; | |
90 | ||
d352ac68 | 91 | /* held while logging the inode in tree-log.c */ |
e02119d5 | 92 | struct mutex log_mutex; |
d352ac68 CM |
93 | |
94 | /* used to order data wrt metadata */ | |
e6dcd2dc | 95 | struct btrfs_ordered_inode_tree ordered_tree; |
15ee9bc7 | 96 | |
d352ac68 CM |
97 | /* list of all the delalloc inodes in the FS. There are times we need |
98 | * to write all the delalloc pages to disk, and this list is used | |
99 | * to walk them all. | |
100 | */ | |
ea8c2819 CM |
101 | struct list_head delalloc_inodes; |
102 | ||
5d4f98a2 YZ |
103 | /* node for the red-black tree that links inodes in subvolume root */ |
104 | struct rb_node rb_node; | |
105 | ||
72ac3c0d JB |
106 | unsigned long runtime_flags; |
107 | ||
9c931c5a | 108 | /* Keep track of who's O_SYNC/fsyncing currently */ |
b812ce28 JB |
109 | atomic_t sync_writers; |
110 | ||
d352ac68 CM |
111 | /* full 64 bit generation number, struct vfs_inode doesn't have a big |
112 | * enough field for this. | |
113 | */ | |
e02119d5 CM |
114 | u64 generation; |
115 | ||
15ee9bc7 JB |
116 | /* |
117 | * transid of the trans_handle that last modified this inode | |
118 | */ | |
119 | u64 last_trans; | |
257c62e1 CM |
120 | |
121 | /* | |
bb14a59b | 122 | * transid that last logged this inode |
257c62e1 | 123 | */ |
bb14a59b | 124 | u64 logged_trans; |
257c62e1 | 125 | |
e02119d5 | 126 | /* |
bb14a59b | 127 | * log transid when this inode was last modified |
e02119d5 | 128 | */ |
bb14a59b MX |
129 | int last_sub_trans; |
130 | ||
131 | /* a local copy of root's last_log_commit */ | |
132 | int last_log_commit; | |
49eb7e46 | 133 | |
d352ac68 CM |
134 | /* total number of bytes pending delalloc, used by stat to calc the |
135 | * real block usage of the file | |
136 | */ | |
9069218d | 137 | u64 delalloc_bytes; |
d352ac68 | 138 | |
a7e3b975 FM |
139 | /* |
140 | * Total number of bytes pending delalloc that fall within a file | |
141 | * range that is either a hole or beyond EOF (and no prealloc extent | |
142 | * exists in the range). This is always <= delalloc_bytes. | |
143 | */ | |
144 | u64 new_delalloc_bytes; | |
145 | ||
47059d93 WS |
146 | /* |
147 | * total number of bytes pending defrag, used by stat to check whether | |
148 | * it needs COW. | |
149 | */ | |
150 | u64 defrag_bytes; | |
151 | ||
d352ac68 CM |
152 | /* |
153 | * the size of the file stored in the metadata on disk. data=ordered | |
154 | * means the in-memory i_size might be larger than the size on disk | |
155 | * because not all the blocks are written yet. | |
156 | */ | |
dbe674a9 | 157 | u64 disk_i_size; |
d352ac68 | 158 | |
aec7477b JB |
159 | /* |
160 | * if this is a directory then index_cnt is the counter for the index | |
161 | * number for new files that are created | |
162 | */ | |
163 | u64 index_cnt; | |
d352ac68 | 164 | |
67de1176 MX |
165 | /* Cache the directory index number to speed the dir/file remove */ |
166 | u64 dir_index; | |
167 | ||
12fcfd22 CM |
168 | /* the fsync log has some corner cases that mean we have to check |
169 | * directories to see if any unlinks have been done before | |
170 | * the directory was logged. See tree-log.c for all the | |
171 | * details | |
172 | */ | |
173 | u64 last_unlink_trans; | |
174 | ||
3ebac17c FM |
175 | /* |
176 | * The id/generation of the last transaction where this inode was | |
177 | * either the source or the destination of a clone/dedupe operation. | |
178 | * Used when logging an inode to know if there are shared extents that | |
179 | * need special care when logging checksum items, to avoid duplicate | |
180 | * checksum items in a log (which can lead to a corruption where we end | |
181 | * up with missing checksum ranges after log replay). | |
182 | * Protected by the vfs inode lock. | |
183 | */ | |
184 | u64 last_reflink_trans; | |
185 | ||
7709cde3 JB |
186 | /* |
187 | * Number of bytes outstanding that are going to need csums. This is | |
188 | * used in ENOSPC accounting. | |
189 | */ | |
190 | u64 csum_bytes; | |
191 | ||
f1bdcc0a JB |
192 | /* flags field from the on disk inode */ |
193 | u32 flags; | |
194 | ||
9ed74f2d | 195 | /* |
32c00aff JB |
196 | * Counters to keep track of the number of extent item's we may use due |
197 | * to delalloc and such. outstanding_extents is the number of extent | |
198 | * items we think we'll end up using, and reserved_extents is the number | |
199 | * of extent items we've reserved metadata for. | |
9ed74f2d | 200 | */ |
9e0baf60 | 201 | unsigned outstanding_extents; |
69fe2d75 JB |
202 | |
203 | struct btrfs_block_rsv block_rsv; | |
9ed74f2d | 204 | |
1e701a32 | 205 | /* |
b52aa8c9 | 206 | * Cached values of inode properties |
1e701a32 | 207 | */ |
b52aa8c9 | 208 | unsigned prop_compress; /* per-file compression algorithm */ |
eec63c65 DS |
209 | /* |
210 | * Force compression on the file using the defrag ioctl, could be | |
211 | * different from prop_compress and takes precedence if set | |
212 | */ | |
213 | unsigned defrag_compress; | |
1e701a32 | 214 | |
16cdcec7 MX |
215 | struct btrfs_delayed_node *delayed_node; |
216 | ||
9cc97d64 | 217 | /* File creation time. */ |
d3c6be6f | 218 | struct timespec64 i_otime; |
9cc97d64 | 219 | |
8089fe62 DS |
220 | /* Hook into fs_info->delayed_iputs */ |
221 | struct list_head delayed_iput; | |
8089fe62 | 222 | |
8318ba79 | 223 | struct rw_semaphore i_mmap_lock; |
d352ac68 | 224 | struct inode vfs_inode; |
2c90e5d6 | 225 | }; |
dbe674a9 | 226 | |
6fee248d NB |
227 | static inline u32 btrfs_inode_sectorsize(const struct btrfs_inode *inode) |
228 | { | |
229 | return inode->root->fs_info->sectorsize; | |
230 | } | |
231 | ||
9a35b637 | 232 | static inline struct btrfs_inode *BTRFS_I(const struct inode *inode) |
2c90e5d6 CM |
233 | { |
234 | return container_of(inode, struct btrfs_inode, vfs_inode); | |
235 | } | |
236 | ||
778ba82b FDBM |
237 | static inline unsigned long btrfs_inode_hash(u64 objectid, |
238 | const struct btrfs_root *root) | |
239 | { | |
4fd786e6 | 240 | u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME); |
778ba82b FDBM |
241 | |
242 | #if BITS_PER_LONG == 32 | |
243 | h = (h >> 32) ^ (h & 0xffffffff); | |
244 | #endif | |
245 | ||
246 | return (unsigned long)h; | |
247 | } | |
248 | ||
249 | static inline void btrfs_insert_inode_hash(struct inode *inode) | |
250 | { | |
251 | unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root); | |
252 | ||
253 | __insert_inode_hash(inode, h); | |
254 | } | |
255 | ||
9a35b637 | 256 | static inline u64 btrfs_ino(const struct btrfs_inode *inode) |
33345d01 | 257 | { |
4a0cc7ca | 258 | u64 ino = inode->location.objectid; |
33345d01 | 259 | |
14c7cca7 LB |
260 | /* |
261 | * !ino: btree_inode | |
262 | * type == BTRFS_ROOT_ITEM_KEY: subvol dir | |
263 | */ | |
4a0cc7ca NB |
264 | if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY) |
265 | ino = inode->vfs_inode.i_ino; | |
33345d01 LZ |
266 | return ino; |
267 | } | |
268 | ||
6ef06d27 | 269 | static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size) |
dbe674a9 | 270 | { |
6ef06d27 NB |
271 | i_size_write(&inode->vfs_inode, size); |
272 | inode->disk_i_size = size; | |
dbe674a9 CM |
273 | } |
274 | ||
70ddc553 | 275 | static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode) |
2cf8572d | 276 | { |
70ddc553 | 277 | struct btrfs_root *root = inode->root; |
83eea1f1 | 278 | |
51a8cf9d | 279 | if (root == root->fs_info->tree_root && |
70ddc553 | 280 | btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID) |
51a8cf9d | 281 | return true; |
70ddc553 | 282 | if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID) |
2cf8572d CM |
283 | return true; |
284 | return false; | |
285 | } | |
286 | ||
06f2548f NB |
287 | static inline bool is_data_inode(struct inode *inode) |
288 | { | |
289 | return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID; | |
290 | } | |
291 | ||
8b62f87b JB |
292 | static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode, |
293 | int mod) | |
294 | { | |
295 | lockdep_assert_held(&inode->lock); | |
296 | inode->outstanding_extents += mod; | |
297 | if (btrfs_is_free_space_inode(inode)) | |
298 | return; | |
dd48d407 JB |
299 | trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode), |
300 | mod); | |
8b62f87b JB |
301 | } |
302 | ||
bc0939fc FM |
303 | /* |
304 | * Called every time after doing a buffered, direct IO or memory mapped write. | |
305 | * | |
306 | * This is to ensure that if we write to a file that was previously fsynced in | |
307 | * the current transaction, then try to fsync it again in the same transaction, | |
308 | * we will know that there were changes in the file and that it needs to be | |
309 | * logged. | |
310 | */ | |
311 | static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode) | |
312 | { | |
313 | spin_lock(&inode->lock); | |
314 | inode->last_sub_trans = inode->root->log_transid; | |
315 | spin_unlock(&inode->lock); | |
316 | } | |
317 | ||
209ecbb8 | 318 | static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation) |
22ee6985 | 319 | { |
209ecbb8 | 320 | bool ret = false; |
2f2ff0ee | 321 | |
0f8939b8 NB |
322 | spin_lock(&inode->lock); |
323 | if (inode->logged_trans == generation && | |
324 | inode->last_sub_trans <= inode->last_log_commit && | |
209ecbb8 FM |
325 | inode->last_sub_trans <= inode->root->last_log_commit) |
326 | ret = true; | |
0f8939b8 | 327 | spin_unlock(&inode->lock); |
2f2ff0ee | 328 | return ret; |
22ee6985 JB |
329 | } |
330 | ||
facc8a22 MX |
331 | struct btrfs_dio_private { |
332 | struct inode *inode; | |
333 | u64 logical_offset; | |
334 | u64 disk_bytenr; | |
523929f1 QW |
335 | /* Used for bio::bi_size */ |
336 | u32 bytes; | |
facc8a22 | 337 | |
e3b318d1 OS |
338 | /* |
339 | * References to this structure. There is one reference per in-flight | |
340 | * bio plus one while we're still setting up. | |
341 | */ | |
342 | refcount_t refs; | |
facc8a22 | 343 | |
facc8a22 MX |
344 | /* dio_bio came from fs/direct-io.c */ |
345 | struct bio *dio_bio; | |
c1dc0896 | 346 | |
85879573 OS |
347 | /* Array of checksums */ |
348 | u8 csums[]; | |
facc8a22 MX |
349 | }; |
350 | ||
7ebc7e5f JT |
351 | /* Array of bytes with variable length, hexadecimal format 0x1234 */ |
352 | #define CSUM_FMT "0x%*phN" | |
353 | #define CSUM_FMT_VALUE(size, bytes) size, bytes | |
354 | ||
0970a22e | 355 | static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode, |
ea41d6b2 | 356 | u64 logical_start, u8 *csum, u8 *csum_expected, int mirror_num) |
6f6b643e | 357 | { |
0970a22e | 358 | struct btrfs_root *root = inode->root; |
223486c2 | 359 | const u32 csum_size = root->fs_info->csum_size; |
6f6b643e QW |
360 | |
361 | /* Output minus objectid, which is more meaningful */ | |
4fd786e6 | 362 | if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID) |
6f6b643e | 363 | btrfs_warn_rl(root->fs_info, |
7ebc7e5f | 364 | "csum failed root %lld ino %lld off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d", |
4fd786e6 | 365 | root->root_key.objectid, btrfs_ino(inode), |
7ebc7e5f | 366 | logical_start, |
ea41d6b2 JT |
367 | CSUM_FMT_VALUE(csum_size, csum), |
368 | CSUM_FMT_VALUE(csum_size, csum_expected), | |
7ebc7e5f | 369 | mirror_num); |
6f6b643e QW |
370 | else |
371 | btrfs_warn_rl(root->fs_info, | |
7ebc7e5f | 372 | "csum failed root %llu ino %llu off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d", |
4fd786e6 | 373 | root->root_key.objectid, btrfs_ino(inode), |
7ebc7e5f | 374 | logical_start, |
ea41d6b2 JT |
375 | CSUM_FMT_VALUE(csum_size, csum), |
376 | CSUM_FMT_VALUE(csum_size, csum_expected), | |
7ebc7e5f | 377 | mirror_num); |
6f6b643e QW |
378 | } |
379 | ||
2c90e5d6 | 380 | #endif |