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 { |
bbe339cc | 24 | BTRFS_INODE_ORDERED_DATA_CLOSE, |
7efc3e34 OS |
25 | BTRFS_INODE_DUMMY, |
26 | BTRFS_INODE_IN_DEFRAG, | |
27 | BTRFS_INODE_HAS_ASYNC_EXTENT, | |
28 | BTRFS_INODE_NEEDS_FULL_SYNC, | |
29 | BTRFS_INODE_COPY_EVERYTHING, | |
30 | BTRFS_INODE_IN_DELALLOC_LIST, | |
7efc3e34 | 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 | |
41a2ee75 JB |
63 | /* |
64 | * Keep track of where the inode has extent items mapped in order to | |
65 | * make sure the i_size adjustments are accurate | |
66 | */ | |
67 | struct extent_io_tree file_extent_tree; | |
68 | ||
d352ac68 | 69 | /* held while logging the inode in tree-log.c */ |
e02119d5 | 70 | struct mutex log_mutex; |
d352ac68 CM |
71 | |
72 | /* used to order data wrt metadata */ | |
e6dcd2dc | 73 | struct btrfs_ordered_inode_tree ordered_tree; |
15ee9bc7 | 74 | |
d352ac68 CM |
75 | /* list of all the delalloc inodes in the FS. There are times we need |
76 | * to write all the delalloc pages to disk, and this list is used | |
77 | * to walk them all. | |
78 | */ | |
ea8c2819 CM |
79 | struct list_head delalloc_inodes; |
80 | ||
5d4f98a2 YZ |
81 | /* node for the red-black tree that links inodes in subvolume root */ |
82 | struct rb_node rb_node; | |
83 | ||
72ac3c0d JB |
84 | unsigned long runtime_flags; |
85 | ||
9c931c5a | 86 | /* Keep track of who's O_SYNC/fsyncing currently */ |
b812ce28 JB |
87 | atomic_t sync_writers; |
88 | ||
d352ac68 CM |
89 | /* full 64 bit generation number, struct vfs_inode doesn't have a big |
90 | * enough field for this. | |
91 | */ | |
e02119d5 CM |
92 | u64 generation; |
93 | ||
15ee9bc7 JB |
94 | /* |
95 | * transid of the trans_handle that last modified this inode | |
96 | */ | |
97 | u64 last_trans; | |
257c62e1 CM |
98 | |
99 | /* | |
bb14a59b | 100 | * transid that last logged this inode |
257c62e1 | 101 | */ |
bb14a59b | 102 | u64 logged_trans; |
257c62e1 | 103 | |
e02119d5 | 104 | /* |
bb14a59b | 105 | * log transid when this inode was last modified |
e02119d5 | 106 | */ |
bb14a59b MX |
107 | int last_sub_trans; |
108 | ||
109 | /* a local copy of root's last_log_commit */ | |
110 | int last_log_commit; | |
49eb7e46 | 111 | |
d352ac68 CM |
112 | /* total number of bytes pending delalloc, used by stat to calc the |
113 | * real block usage of the file | |
114 | */ | |
9069218d | 115 | u64 delalloc_bytes; |
d352ac68 | 116 | |
a7e3b975 FM |
117 | /* |
118 | * Total number of bytes pending delalloc that fall within a file | |
119 | * range that is either a hole or beyond EOF (and no prealloc extent | |
120 | * exists in the range). This is always <= delalloc_bytes. | |
121 | */ | |
122 | u64 new_delalloc_bytes; | |
123 | ||
47059d93 WS |
124 | /* |
125 | * total number of bytes pending defrag, used by stat to check whether | |
126 | * it needs COW. | |
127 | */ | |
128 | u64 defrag_bytes; | |
129 | ||
d352ac68 CM |
130 | /* |
131 | * the size of the file stored in the metadata on disk. data=ordered | |
132 | * means the in-memory i_size might be larger than the size on disk | |
133 | * because not all the blocks are written yet. | |
134 | */ | |
dbe674a9 | 135 | u64 disk_i_size; |
d352ac68 | 136 | |
aec7477b JB |
137 | /* |
138 | * if this is a directory then index_cnt is the counter for the index | |
139 | * number for new files that are created | |
140 | */ | |
141 | u64 index_cnt; | |
d352ac68 | 142 | |
67de1176 MX |
143 | /* Cache the directory index number to speed the dir/file remove */ |
144 | u64 dir_index; | |
145 | ||
12fcfd22 CM |
146 | /* the fsync log has some corner cases that mean we have to check |
147 | * directories to see if any unlinks have been done before | |
148 | * the directory was logged. See tree-log.c for all the | |
149 | * details | |
150 | */ | |
151 | u64 last_unlink_trans; | |
152 | ||
7709cde3 JB |
153 | /* |
154 | * Number of bytes outstanding that are going to need csums. This is | |
155 | * used in ENOSPC accounting. | |
156 | */ | |
157 | u64 csum_bytes; | |
158 | ||
f1bdcc0a JB |
159 | /* flags field from the on disk inode */ |
160 | u32 flags; | |
161 | ||
9ed74f2d | 162 | /* |
32c00aff JB |
163 | * Counters to keep track of the number of extent item's we may use due |
164 | * to delalloc and such. outstanding_extents is the number of extent | |
165 | * items we think we'll end up using, and reserved_extents is the number | |
166 | * of extent items we've reserved metadata for. | |
9ed74f2d | 167 | */ |
9e0baf60 | 168 | unsigned outstanding_extents; |
69fe2d75 JB |
169 | |
170 | struct btrfs_block_rsv block_rsv; | |
9ed74f2d | 171 | |
1e701a32 | 172 | /* |
b52aa8c9 | 173 | * Cached values of inode properties |
1e701a32 | 174 | */ |
b52aa8c9 | 175 | unsigned prop_compress; /* per-file compression algorithm */ |
eec63c65 DS |
176 | /* |
177 | * Force compression on the file using the defrag ioctl, could be | |
178 | * different from prop_compress and takes precedence if set | |
179 | */ | |
180 | unsigned defrag_compress; | |
1e701a32 | 181 | |
16cdcec7 MX |
182 | struct btrfs_delayed_node *delayed_node; |
183 | ||
9cc97d64 | 184 | /* File creation time. */ |
d3c6be6f | 185 | struct timespec64 i_otime; |
9cc97d64 | 186 | |
8089fe62 DS |
187 | /* Hook into fs_info->delayed_iputs */ |
188 | struct list_head delayed_iput; | |
8089fe62 | 189 | |
5f9a8a51 FM |
190 | /* |
191 | * To avoid races between lockless (i_mutex not held) direct IO writes | |
192 | * and concurrent fsync requests. Direct IO writes must acquire read | |
193 | * access on this semaphore for creating an extent map and its | |
194 | * corresponding ordered extent. The fast fsync path must acquire write | |
195 | * access on this semaphore before it collects ordered extents and | |
196 | * extent maps. | |
197 | */ | |
198 | struct rw_semaphore dio_sem; | |
199 | ||
d352ac68 | 200 | struct inode vfs_inode; |
2c90e5d6 | 201 | }; |
dbe674a9 | 202 | |
9a35b637 | 203 | static inline struct btrfs_inode *BTRFS_I(const struct inode *inode) |
2c90e5d6 CM |
204 | { |
205 | return container_of(inode, struct btrfs_inode, vfs_inode); | |
206 | } | |
207 | ||
778ba82b FDBM |
208 | static inline unsigned long btrfs_inode_hash(u64 objectid, |
209 | const struct btrfs_root *root) | |
210 | { | |
4fd786e6 | 211 | u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME); |
778ba82b FDBM |
212 | |
213 | #if BITS_PER_LONG == 32 | |
214 | h = (h >> 32) ^ (h & 0xffffffff); | |
215 | #endif | |
216 | ||
217 | return (unsigned long)h; | |
218 | } | |
219 | ||
220 | static inline void btrfs_insert_inode_hash(struct inode *inode) | |
221 | { | |
222 | unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root); | |
223 | ||
224 | __insert_inode_hash(inode, h); | |
225 | } | |
226 | ||
9a35b637 | 227 | static inline u64 btrfs_ino(const struct btrfs_inode *inode) |
33345d01 | 228 | { |
4a0cc7ca | 229 | u64 ino = inode->location.objectid; |
33345d01 | 230 | |
14c7cca7 LB |
231 | /* |
232 | * !ino: btree_inode | |
233 | * type == BTRFS_ROOT_ITEM_KEY: subvol dir | |
234 | */ | |
4a0cc7ca NB |
235 | if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY) |
236 | ino = inode->vfs_inode.i_ino; | |
33345d01 LZ |
237 | return ino; |
238 | } | |
239 | ||
6ef06d27 | 240 | static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size) |
dbe674a9 | 241 | { |
6ef06d27 NB |
242 | i_size_write(&inode->vfs_inode, size); |
243 | inode->disk_i_size = size; | |
dbe674a9 CM |
244 | } |
245 | ||
70ddc553 | 246 | static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode) |
2cf8572d | 247 | { |
70ddc553 | 248 | struct btrfs_root *root = inode->root; |
83eea1f1 | 249 | |
51a8cf9d | 250 | if (root == root->fs_info->tree_root && |
70ddc553 | 251 | btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID) |
51a8cf9d | 252 | return true; |
70ddc553 | 253 | if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID) |
2cf8572d CM |
254 | return true; |
255 | return false; | |
256 | } | |
257 | ||
06f2548f NB |
258 | static inline bool is_data_inode(struct inode *inode) |
259 | { | |
260 | return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID; | |
261 | } | |
262 | ||
8b62f87b JB |
263 | static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode, |
264 | int mod) | |
265 | { | |
266 | lockdep_assert_held(&inode->lock); | |
267 | inode->outstanding_extents += mod; | |
268 | if (btrfs_is_free_space_inode(inode)) | |
269 | return; | |
dd48d407 JB |
270 | trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode), |
271 | mod); | |
8b62f87b JB |
272 | } |
273 | ||
0f8939b8 | 274 | static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation) |
22ee6985 | 275 | { |
2f2ff0ee FM |
276 | int ret = 0; |
277 | ||
0f8939b8 NB |
278 | spin_lock(&inode->lock); |
279 | if (inode->logged_trans == generation && | |
280 | inode->last_sub_trans <= inode->last_log_commit && | |
281 | inode->last_sub_trans <= inode->root->last_log_commit) { | |
125c4cf9 FM |
282 | /* |
283 | * After a ranged fsync we might have left some extent maps | |
284 | * (that fall outside the fsync's range). So return false | |
285 | * here if the list isn't empty, to make sure btrfs_log_inode() | |
286 | * will be called and process those extent maps. | |
287 | */ | |
288 | smp_mb(); | |
0f8939b8 | 289 | if (list_empty(&inode->extent_tree.modified_extents)) |
2f2ff0ee | 290 | ret = 1; |
125c4cf9 | 291 | } |
0f8939b8 | 292 | spin_unlock(&inode->lock); |
2f2ff0ee | 293 | return ret; |
22ee6985 JB |
294 | } |
295 | ||
facc8a22 MX |
296 | struct btrfs_dio_private { |
297 | struct inode *inode; | |
298 | u64 logical_offset; | |
299 | u64 disk_bytenr; | |
300 | u64 bytes; | |
facc8a22 | 301 | |
e3b318d1 OS |
302 | /* |
303 | * References to this structure. There is one reference per in-flight | |
304 | * bio plus one while we're still setting up. | |
305 | */ | |
306 | refcount_t refs; | |
facc8a22 | 307 | |
facc8a22 MX |
308 | /* dio_bio came from fs/direct-io.c */ |
309 | struct bio *dio_bio; | |
c1dc0896 | 310 | |
85879573 OS |
311 | /* Array of checksums */ |
312 | u8 csums[]; | |
facc8a22 MX |
313 | }; |
314 | ||
7ebc7e5f JT |
315 | /* Array of bytes with variable length, hexadecimal format 0x1234 */ |
316 | #define CSUM_FMT "0x%*phN" | |
317 | #define CSUM_FMT_VALUE(size, bytes) size, bytes | |
318 | ||
0970a22e | 319 | static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode, |
ea41d6b2 | 320 | u64 logical_start, u8 *csum, u8 *csum_expected, int mirror_num) |
6f6b643e | 321 | { |
0970a22e | 322 | struct btrfs_root *root = inode->root; |
7ebc7e5f JT |
323 | struct btrfs_super_block *sb = root->fs_info->super_copy; |
324 | const u16 csum_size = btrfs_super_csum_size(sb); | |
6f6b643e QW |
325 | |
326 | /* Output minus objectid, which is more meaningful */ | |
4fd786e6 | 327 | if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID) |
6f6b643e | 328 | btrfs_warn_rl(root->fs_info, |
7ebc7e5f | 329 | "csum failed root %lld ino %lld off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d", |
4fd786e6 | 330 | root->root_key.objectid, btrfs_ino(inode), |
7ebc7e5f | 331 | logical_start, |
ea41d6b2 JT |
332 | CSUM_FMT_VALUE(csum_size, csum), |
333 | CSUM_FMT_VALUE(csum_size, csum_expected), | |
7ebc7e5f | 334 | mirror_num); |
6f6b643e QW |
335 | else |
336 | btrfs_warn_rl(root->fs_info, | |
7ebc7e5f | 337 | "csum failed root %llu ino %llu off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d", |
4fd786e6 | 338 | root->root_key.objectid, btrfs_ino(inode), |
7ebc7e5f | 339 | logical_start, |
ea41d6b2 JT |
340 | CSUM_FMT_VALUE(csum_size, csum), |
341 | CSUM_FMT_VALUE(csum_size, csum_expected), | |
7ebc7e5f | 342 | mirror_num); |
6f6b643e QW |
343 | } |
344 | ||
2c90e5d6 | 345 | #endif |