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> |
a52d9a80 | 10 | #include "extent_map.h" |
d1310b2e | 11 | #include "extent_io.h" |
e6dcd2dc | 12 | #include "ordered-data.h" |
16cdcec7 | 13 | #include "delayed-inode.h" |
a52d9a80 | 14 | |
72ac3c0d JB |
15 | /* |
16 | * ordered_data_close is set by truncate when a file that used | |
17 | * to have good data has been truncated to zero. When it is set | |
18 | * the btrfs file release call will add this inode to the | |
19 | * ordered operations list so that we make sure to flush out any | |
20 | * new data the application may have written before commit. | |
21 | */ | |
7efc3e34 | 22 | enum { |
bbe339cc | 23 | BTRFS_INODE_ORDERED_DATA_CLOSE, |
7efc3e34 OS |
24 | BTRFS_INODE_DUMMY, |
25 | BTRFS_INODE_IN_DEFRAG, | |
26 | BTRFS_INODE_HAS_ASYNC_EXTENT, | |
27 | BTRFS_INODE_NEEDS_FULL_SYNC, | |
28 | BTRFS_INODE_COPY_EVERYTHING, | |
29 | BTRFS_INODE_IN_DELALLOC_LIST, | |
30 | BTRFS_INODE_READDIO_NEED_LOCK, | |
31 | BTRFS_INODE_HAS_PROPS, | |
3cd24c69 | 32 | BTRFS_INODE_SNAPSHOT_FLUSH, |
7efc3e34 | 33 | }; |
72ac3c0d | 34 | |
f1ace244 | 35 | /* in memory btrfs inode */ |
2c90e5d6 | 36 | struct btrfs_inode { |
d352ac68 | 37 | /* which subvolume this inode belongs to */ |
d6e4a428 | 38 | struct btrfs_root *root; |
d352ac68 | 39 | |
d352ac68 CM |
40 | /* key used to find this inode on disk. This is used by the code |
41 | * to read in roots of subvolumes | |
42 | */ | |
d6e4a428 | 43 | struct btrfs_key location; |
d352ac68 | 44 | |
2f2ff0ee FM |
45 | /* |
46 | * Lock for counters and all fields used to determine if the inode is in | |
47 | * the log or not (last_trans, last_sub_trans, last_log_commit, | |
48 | * logged_trans). | |
49 | */ | |
9e0baf60 JB |
50 | spinlock_t lock; |
51 | ||
d352ac68 | 52 | /* the extent_tree has caches of all the extent mappings to disk */ |
a52d9a80 | 53 | struct extent_map_tree extent_tree; |
d352ac68 CM |
54 | |
55 | /* the io_tree does range state (DIRTY, LOCKED etc) */ | |
d1310b2e | 56 | struct extent_io_tree io_tree; |
d352ac68 CM |
57 | |
58 | /* special utility tree used to record which mirrors have already been | |
59 | * tried when checksums fail for a given block | |
60 | */ | |
7e38326f | 61 | struct extent_io_tree io_failure_tree; |
d352ac68 | 62 | |
d352ac68 | 63 | /* held while logging the inode in tree-log.c */ |
e02119d5 | 64 | struct mutex log_mutex; |
d352ac68 | 65 | |
f248679e JB |
66 | /* held while doing delalloc reservations */ |
67 | struct mutex delalloc_mutex; | |
68 | ||
d352ac68 | 69 | /* used to order data wrt metadata */ |
e6dcd2dc | 70 | struct btrfs_ordered_inode_tree ordered_tree; |
15ee9bc7 | 71 | |
d352ac68 CM |
72 | /* list of all the delalloc inodes in the FS. There are times we need |
73 | * to write all the delalloc pages to disk, and this list is used | |
74 | * to walk them all. | |
75 | */ | |
ea8c2819 CM |
76 | struct list_head delalloc_inodes; |
77 | ||
5d4f98a2 YZ |
78 | /* node for the red-black tree that links inodes in subvolume root */ |
79 | struct rb_node rb_node; | |
80 | ||
72ac3c0d JB |
81 | unsigned long runtime_flags; |
82 | ||
9c931c5a | 83 | /* Keep track of who's O_SYNC/fsyncing currently */ |
b812ce28 JB |
84 | atomic_t sync_writers; |
85 | ||
d352ac68 CM |
86 | /* full 64 bit generation number, struct vfs_inode doesn't have a big |
87 | * enough field for this. | |
88 | */ | |
e02119d5 CM |
89 | u64 generation; |
90 | ||
15ee9bc7 JB |
91 | /* |
92 | * transid of the trans_handle that last modified this inode | |
93 | */ | |
94 | u64 last_trans; | |
257c62e1 CM |
95 | |
96 | /* | |
bb14a59b | 97 | * transid that last logged this inode |
257c62e1 | 98 | */ |
bb14a59b | 99 | u64 logged_trans; |
257c62e1 | 100 | |
e02119d5 | 101 | /* |
bb14a59b | 102 | * log transid when this inode was last modified |
e02119d5 | 103 | */ |
bb14a59b MX |
104 | int last_sub_trans; |
105 | ||
106 | /* a local copy of root's last_log_commit */ | |
107 | int last_log_commit; | |
49eb7e46 | 108 | |
d352ac68 CM |
109 | /* total number of bytes pending delalloc, used by stat to calc the |
110 | * real block usage of the file | |
111 | */ | |
9069218d | 112 | u64 delalloc_bytes; |
d352ac68 | 113 | |
a7e3b975 FM |
114 | /* |
115 | * Total number of bytes pending delalloc that fall within a file | |
116 | * range that is either a hole or beyond EOF (and no prealloc extent | |
117 | * exists in the range). This is always <= delalloc_bytes. | |
118 | */ | |
119 | u64 new_delalloc_bytes; | |
120 | ||
47059d93 WS |
121 | /* |
122 | * total number of bytes pending defrag, used by stat to check whether | |
123 | * it needs COW. | |
124 | */ | |
125 | u64 defrag_bytes; | |
126 | ||
d352ac68 CM |
127 | /* |
128 | * the size of the file stored in the metadata on disk. data=ordered | |
129 | * means the in-memory i_size might be larger than the size on disk | |
130 | * because not all the blocks are written yet. | |
131 | */ | |
dbe674a9 | 132 | u64 disk_i_size; |
d352ac68 | 133 | |
aec7477b JB |
134 | /* |
135 | * if this is a directory then index_cnt is the counter for the index | |
136 | * number for new files that are created | |
137 | */ | |
138 | u64 index_cnt; | |
d352ac68 | 139 | |
67de1176 MX |
140 | /* Cache the directory index number to speed the dir/file remove */ |
141 | u64 dir_index; | |
142 | ||
12fcfd22 CM |
143 | /* the fsync log has some corner cases that mean we have to check |
144 | * directories to see if any unlinks have been done before | |
145 | * the directory was logged. See tree-log.c for all the | |
146 | * details | |
147 | */ | |
148 | u64 last_unlink_trans; | |
149 | ||
41bd6067 FM |
150 | /* |
151 | * Track the transaction id of the last transaction used to create a | |
152 | * hard link for the inode. This is used by the log tree (fsync). | |
153 | */ | |
154 | u64 last_link_trans; | |
155 | ||
7709cde3 JB |
156 | /* |
157 | * Number of bytes outstanding that are going to need csums. This is | |
158 | * used in ENOSPC accounting. | |
159 | */ | |
160 | u64 csum_bytes; | |
161 | ||
f1bdcc0a JB |
162 | /* flags field from the on disk inode */ |
163 | u32 flags; | |
164 | ||
9ed74f2d | 165 | /* |
32c00aff JB |
166 | * Counters to keep track of the number of extent item's we may use due |
167 | * to delalloc and such. outstanding_extents is the number of extent | |
168 | * items we think we'll end up using, and reserved_extents is the number | |
169 | * of extent items we've reserved metadata for. | |
9ed74f2d | 170 | */ |
9e0baf60 | 171 | unsigned outstanding_extents; |
69fe2d75 JB |
172 | |
173 | struct btrfs_block_rsv block_rsv; | |
9ed74f2d | 174 | |
1e701a32 | 175 | /* |
b52aa8c9 | 176 | * Cached values of inode properties |
1e701a32 | 177 | */ |
b52aa8c9 | 178 | unsigned prop_compress; /* per-file compression algorithm */ |
eec63c65 DS |
179 | /* |
180 | * Force compression on the file using the defrag ioctl, could be | |
181 | * different from prop_compress and takes precedence if set | |
182 | */ | |
183 | unsigned defrag_compress; | |
1e701a32 | 184 | |
16cdcec7 MX |
185 | struct btrfs_delayed_node *delayed_node; |
186 | ||
9cc97d64 | 187 | /* File creation time. */ |
d3c6be6f | 188 | struct timespec64 i_otime; |
9cc97d64 | 189 | |
8089fe62 DS |
190 | /* Hook into fs_info->delayed_iputs */ |
191 | struct list_head delayed_iput; | |
8089fe62 | 192 | |
5f9a8a51 FM |
193 | /* |
194 | * To avoid races between lockless (i_mutex not held) direct IO writes | |
195 | * and concurrent fsync requests. Direct IO writes must acquire read | |
196 | * access on this semaphore for creating an extent map and its | |
197 | * corresponding ordered extent. The fast fsync path must acquire write | |
198 | * access on this semaphore before it collects ordered extents and | |
199 | * extent maps. | |
200 | */ | |
201 | struct rw_semaphore dio_sem; | |
202 | ||
d352ac68 | 203 | struct inode vfs_inode; |
2c90e5d6 | 204 | }; |
dbe674a9 | 205 | |
16cdcec7 MX |
206 | extern unsigned char btrfs_filetype_table[]; |
207 | ||
9a35b637 | 208 | static inline struct btrfs_inode *BTRFS_I(const struct inode *inode) |
2c90e5d6 CM |
209 | { |
210 | return container_of(inode, struct btrfs_inode, vfs_inode); | |
211 | } | |
212 | ||
778ba82b FDBM |
213 | static inline unsigned long btrfs_inode_hash(u64 objectid, |
214 | const struct btrfs_root *root) | |
215 | { | |
4fd786e6 | 216 | u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME); |
778ba82b FDBM |
217 | |
218 | #if BITS_PER_LONG == 32 | |
219 | h = (h >> 32) ^ (h & 0xffffffff); | |
220 | #endif | |
221 | ||
222 | return (unsigned long)h; | |
223 | } | |
224 | ||
225 | static inline void btrfs_insert_inode_hash(struct inode *inode) | |
226 | { | |
227 | unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root); | |
228 | ||
229 | __insert_inode_hash(inode, h); | |
230 | } | |
231 | ||
9a35b637 | 232 | static inline u64 btrfs_ino(const struct btrfs_inode *inode) |
33345d01 | 233 | { |
4a0cc7ca | 234 | u64 ino = inode->location.objectid; |
33345d01 | 235 | |
14c7cca7 LB |
236 | /* |
237 | * !ino: btree_inode | |
238 | * type == BTRFS_ROOT_ITEM_KEY: subvol dir | |
239 | */ | |
4a0cc7ca NB |
240 | if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY) |
241 | ino = inode->vfs_inode.i_ino; | |
33345d01 LZ |
242 | return ino; |
243 | } | |
244 | ||
6ef06d27 | 245 | static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size) |
dbe674a9 | 246 | { |
6ef06d27 NB |
247 | i_size_write(&inode->vfs_inode, size); |
248 | inode->disk_i_size = size; | |
dbe674a9 CM |
249 | } |
250 | ||
70ddc553 | 251 | static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode) |
2cf8572d | 252 | { |
70ddc553 | 253 | struct btrfs_root *root = inode->root; |
83eea1f1 | 254 | |
51a8cf9d | 255 | if (root == root->fs_info->tree_root && |
70ddc553 | 256 | btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID) |
51a8cf9d | 257 | return true; |
70ddc553 | 258 | if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID) |
2cf8572d CM |
259 | return true; |
260 | return false; | |
261 | } | |
262 | ||
06f2548f NB |
263 | static inline bool is_data_inode(struct inode *inode) |
264 | { | |
265 | return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID; | |
266 | } | |
267 | ||
8b62f87b JB |
268 | static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode, |
269 | int mod) | |
270 | { | |
271 | lockdep_assert_held(&inode->lock); | |
272 | inode->outstanding_extents += mod; | |
273 | if (btrfs_is_free_space_inode(inode)) | |
274 | return; | |
dd48d407 JB |
275 | trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode), |
276 | mod); | |
8b62f87b JB |
277 | } |
278 | ||
0f8939b8 | 279 | static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation) |
22ee6985 | 280 | { |
2f2ff0ee FM |
281 | int ret = 0; |
282 | ||
0f8939b8 NB |
283 | spin_lock(&inode->lock); |
284 | if (inode->logged_trans == generation && | |
285 | inode->last_sub_trans <= inode->last_log_commit && | |
286 | inode->last_sub_trans <= inode->root->last_log_commit) { | |
125c4cf9 FM |
287 | /* |
288 | * After a ranged fsync we might have left some extent maps | |
289 | * (that fall outside the fsync's range). So return false | |
290 | * here if the list isn't empty, to make sure btrfs_log_inode() | |
291 | * will be called and process those extent maps. | |
292 | */ | |
293 | smp_mb(); | |
0f8939b8 | 294 | if (list_empty(&inode->extent_tree.modified_extents)) |
2f2ff0ee | 295 | ret = 1; |
125c4cf9 | 296 | } |
0f8939b8 | 297 | spin_unlock(&inode->lock); |
2f2ff0ee | 298 | return ret; |
22ee6985 JB |
299 | } |
300 | ||
c1dc0896 MX |
301 | #define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1 |
302 | ||
facc8a22 MX |
303 | struct btrfs_dio_private { |
304 | struct inode *inode; | |
c1dc0896 | 305 | unsigned long flags; |
facc8a22 MX |
306 | u64 logical_offset; |
307 | u64 disk_bytenr; | |
308 | u64 bytes; | |
309 | void *private; | |
310 | ||
311 | /* number of bios pending for this dio */ | |
312 | atomic_t pending_bios; | |
313 | ||
314 | /* IO errors */ | |
315 | int errors; | |
316 | ||
317 | /* orig_bio is our btrfs_io_bio */ | |
318 | struct bio *orig_bio; | |
319 | ||
320 | /* dio_bio came from fs/direct-io.c */ | |
321 | struct bio *dio_bio; | |
c1dc0896 MX |
322 | |
323 | /* | |
01327610 | 324 | * The original bio may be split to several sub-bios, this is |
c1dc0896 MX |
325 | * done during endio of sub-bios |
326 | */ | |
4e4cbee9 CH |
327 | blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *, |
328 | blk_status_t); | |
facc8a22 MX |
329 | }; |
330 | ||
2e60a51e MX |
331 | /* |
332 | * Disable DIO read nolock optimization, so new dio readers will be forced | |
333 | * to grab i_mutex. It is used to avoid the endless truncate due to | |
334 | * nonlocked dio read. | |
335 | */ | |
abcefb1e | 336 | static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode) |
2e60a51e | 337 | { |
abcefb1e | 338 | set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags); |
2e60a51e MX |
339 | smp_mb(); |
340 | } | |
341 | ||
0b581701 | 342 | static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode) |
2e60a51e | 343 | { |
4e857c58 | 344 | smp_mb__before_atomic(); |
0b581701 | 345 | clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags); |
2e60a51e MX |
346 | } |
347 | ||
0970a22e | 348 | static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode, |
6f6b643e QW |
349 | u64 logical_start, u32 csum, u32 csum_expected, int mirror_num) |
350 | { | |
0970a22e | 351 | struct btrfs_root *root = inode->root; |
6f6b643e QW |
352 | |
353 | /* Output minus objectid, which is more meaningful */ | |
4fd786e6 | 354 | if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID) |
6f6b643e QW |
355 | btrfs_warn_rl(root->fs_info, |
356 | "csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d", | |
4fd786e6 | 357 | root->root_key.objectid, btrfs_ino(inode), |
6f6b643e QW |
358 | logical_start, csum, csum_expected, mirror_num); |
359 | else | |
360 | btrfs_warn_rl(root->fs_info, | |
361 | "csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d", | |
4fd786e6 | 362 | root->root_key.objectid, btrfs_ino(inode), |
6f6b643e QW |
363 | logical_start, csum, csum_expected, mirror_num); |
364 | } | |
365 | ||
2c90e5d6 | 366 | #endif |