Line data Source code
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 "extent_map.h"
12 : #include "extent_io.h"
13 : #include "ordered-data.h"
14 : #include "delayed-inode.h"
15 :
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 :
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 : */
30 : enum {
31 : BTRFS_INODE_FLUSH_ON_CLOSE,
32 : BTRFS_INODE_DUMMY,
33 : BTRFS_INODE_IN_DEFRAG,
34 : BTRFS_INODE_HAS_ASYNC_EXTENT,
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 : */
40 : BTRFS_INODE_NEEDS_FULL_SYNC,
41 : BTRFS_INODE_COPY_EVERYTHING,
42 : BTRFS_INODE_IN_DELALLOC_LIST,
43 : BTRFS_INODE_HAS_PROPS,
44 : BTRFS_INODE_SNAPSHOT_FLUSH,
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,
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,
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,
68 : /* Set when this inode is a free space inode. */
69 : BTRFS_INODE_FREE_SPACE_INODE,
70 : };
71 :
72 : /* in memory btrfs inode */
73 : struct btrfs_inode {
74 : /* which subvolume this inode belongs to */
75 : struct btrfs_root *root;
76 :
77 : /* key used to find this inode on disk. This is used by the code
78 : * to read in roots of subvolumes
79 : */
80 : struct btrfs_key location;
81 :
82 : /*
83 : * Lock for counters and all fields used to determine if the inode is in
84 : * the log or not (last_trans, last_sub_trans, last_log_commit,
85 : * logged_trans), to access/update new_delalloc_bytes and to update the
86 : * VFS' inode number of bytes used.
87 : */
88 : spinlock_t lock;
89 :
90 : /* the extent_tree has caches of all the extent mappings to disk */
91 : struct extent_map_tree extent_tree;
92 :
93 : /* the io_tree does range state (DIRTY, LOCKED etc) */
94 : struct extent_io_tree io_tree;
95 :
96 : /*
97 : * Keep track of where the inode has extent items mapped in order to
98 : * make sure the i_size adjustments are accurate
99 : */
100 : struct extent_io_tree file_extent_tree;
101 :
102 : /* held while logging the inode in tree-log.c */
103 : struct mutex log_mutex;
104 :
105 : /* used to order data wrt metadata */
106 : struct btrfs_ordered_inode_tree ordered_tree;
107 :
108 : /* list of all the delalloc inodes in the FS. There are times we need
109 : * to write all the delalloc pages to disk, and this list is used
110 : * to walk them all.
111 : */
112 : struct list_head delalloc_inodes;
113 :
114 : /* node for the red-black tree that links inodes in subvolume root */
115 : struct rb_node rb_node;
116 :
117 : unsigned long runtime_flags;
118 :
119 : /* full 64 bit generation number, struct vfs_inode doesn't have a big
120 : * enough field for this.
121 : */
122 : u64 generation;
123 :
124 : /*
125 : * transid of the trans_handle that last modified this inode
126 : */
127 : u64 last_trans;
128 :
129 : /*
130 : * transid that last logged this inode
131 : */
132 : u64 logged_trans;
133 :
134 : /*
135 : * log transid when this inode was last modified
136 : */
137 : int last_sub_trans;
138 :
139 : /* a local copy of root's last_log_commit */
140 : int last_log_commit;
141 :
142 : union {
143 : /*
144 : * Total number of bytes pending delalloc, used by stat to
145 : * calculate the real block usage of the file. This is used
146 : * only for files.
147 : */
148 : u64 delalloc_bytes;
149 : /*
150 : * The lowest possible index of the next dir index key which
151 : * points to an inode that needs to be logged.
152 : * This is used only for directories.
153 : * Use the helpers btrfs_get_first_dir_index_to_log() and
154 : * btrfs_set_first_dir_index_to_log() to access this field.
155 : */
156 : u64 first_dir_index_to_log;
157 : };
158 :
159 : union {
160 : /*
161 : * Total number of bytes pending delalloc that fall within a file
162 : * range that is either a hole or beyond EOF (and no prealloc extent
163 : * exists in the range). This is always <= delalloc_bytes and this
164 : * is used only for files.
165 : */
166 : u64 new_delalloc_bytes;
167 : /*
168 : * The offset of the last dir index key that was logged.
169 : * This is used only for directories.
170 : */
171 : u64 last_dir_index_offset;
172 : };
173 :
174 : /*
175 : * total number of bytes pending defrag, used by stat to check whether
176 : * it needs COW.
177 : */
178 : u64 defrag_bytes;
179 :
180 : /*
181 : * the size of the file stored in the metadata on disk. data=ordered
182 : * means the in-memory i_size might be larger than the size on disk
183 : * because not all the blocks are written yet.
184 : */
185 : u64 disk_i_size;
186 :
187 : /*
188 : * If this is a directory then index_cnt is the counter for the index
189 : * number for new files that are created. For an empty directory, this
190 : * must be initialized to BTRFS_DIR_START_INDEX.
191 : */
192 : u64 index_cnt;
193 :
194 : /* Cache the directory index number to speed the dir/file remove */
195 : u64 dir_index;
196 :
197 : /* the fsync log has some corner cases that mean we have to check
198 : * directories to see if any unlinks have been done before
199 : * the directory was logged. See tree-log.c for all the
200 : * details
201 : */
202 : u64 last_unlink_trans;
203 :
204 : /*
205 : * The id/generation of the last transaction where this inode was
206 : * either the source or the destination of a clone/dedupe operation.
207 : * Used when logging an inode to know if there are shared extents that
208 : * need special care when logging checksum items, to avoid duplicate
209 : * checksum items in a log (which can lead to a corruption where we end
210 : * up with missing checksum ranges after log replay).
211 : * Protected by the vfs inode lock.
212 : */
213 : u64 last_reflink_trans;
214 :
215 : /*
216 : * Number of bytes outstanding that are going to need csums. This is
217 : * used in ENOSPC accounting.
218 : */
219 : u64 csum_bytes;
220 :
221 : /* Backwards incompatible flags, lower half of inode_item::flags */
222 : u32 flags;
223 : /* Read-only compatibility flags, upper half of inode_item::flags */
224 : u32 ro_flags;
225 :
226 : /*
227 : * Counters to keep track of the number of extent item's we may use due
228 : * to delalloc and such. outstanding_extents is the number of extent
229 : * items we think we'll end up using, and reserved_extents is the number
230 : * of extent items we've reserved metadata for.
231 : */
232 : unsigned outstanding_extents;
233 :
234 : struct btrfs_block_rsv block_rsv;
235 :
236 : /*
237 : * Cached values of inode properties
238 : */
239 : unsigned prop_compress; /* per-file compression algorithm */
240 : /*
241 : * Force compression on the file using the defrag ioctl, could be
242 : * different from prop_compress and takes precedence if set
243 : */
244 : unsigned defrag_compress;
245 :
246 : struct btrfs_delayed_node *delayed_node;
247 :
248 : /* File creation time. */
249 : struct timespec64 i_otime;
250 :
251 : /* Hook into fs_info->delayed_iputs */
252 : struct list_head delayed_iput;
253 :
254 : struct rw_semaphore i_mmap_lock;
255 : struct inode vfs_inode;
256 : };
257 :
258 : static inline u64 btrfs_get_first_dir_index_to_log(const struct btrfs_inode *inode)
259 : {
260 134 : return READ_ONCE(inode->first_dir_index_to_log);
261 : }
262 :
263 : static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode *inode,
264 : u64 index)
265 : {
266 61 : WRITE_ONCE(inode->first_dir_index_to_log, index);
267 71 : }
268 :
269 : static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
270 : {
271 584252852 : return container_of(inode, struct btrfs_inode, vfs_inode);
272 : }
273 :
274 : static inline unsigned long btrfs_inode_hash(u64 objectid,
275 : const struct btrfs_root *root)
276 : {
277 4196138 : u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
278 :
279 : #if BITS_PER_LONG == 32
280 : h = (h >> 32) ^ (h & 0xffffffff);
281 : #endif
282 :
283 4196138 : return (unsigned long)h;
284 : }
285 :
286 : #if BITS_PER_LONG == 32
287 :
288 : /*
289 : * On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so
290 : * we use the inode's location objectid which is a u64 to avoid truncation.
291 : */
292 : static inline u64 btrfs_ino(const struct btrfs_inode *inode)
293 : {
294 : u64 ino = inode->location.objectid;
295 :
296 : /* type == BTRFS_ROOT_ITEM_KEY: subvol dir */
297 : if (inode->location.type == BTRFS_ROOT_ITEM_KEY)
298 : ino = inode->vfs_inode.i_ino;
299 : return ino;
300 : }
301 :
302 : #else
303 :
304 : static inline u64 btrfs_ino(const struct btrfs_inode *inode)
305 : {
306 555907170 : return inode->vfs_inode.i_ino;
307 : }
308 :
309 : #endif
310 :
311 : static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
312 : {
313 7729222 : i_size_write(&inode->vfs_inode, size);
314 5946380 : inode->disk_i_size = size;
315 1782842 : }
316 :
317 : static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
318 : {
319 207540111 : return test_bit(BTRFS_INODE_FREE_SPACE_INODE, &inode->runtime_flags);
320 : }
321 :
322 : static inline bool is_data_inode(struct inode *inode)
323 : {
324 18874950 : return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
325 : }
326 :
327 209114059 : static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
328 : int mod)
329 : {
330 209114059 : lockdep_assert_held(&inode->lock);
331 209114059 : inode->outstanding_extents += mod;
332 418228118 : if (btrfs_is_free_space_inode(inode))
333 : return;
334 208955331 : trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
335 : mod, inode->outstanding_extents);
336 : }
337 :
338 : /*
339 : * Called every time after doing a buffered, direct IO or memory mapped write.
340 : *
341 : * This is to ensure that if we write to a file that was previously fsynced in
342 : * the current transaction, then try to fsync it again in the same transaction,
343 : * we will know that there were changes in the file and that it needs to be
344 : * logged.
345 : */
346 35669405 : static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)
347 : {
348 35669405 : spin_lock(&inode->lock);
349 35731653 : inode->last_sub_trans = inode->root->log_transid;
350 35731653 : spin_unlock(&inode->lock);
351 35746453 : }
352 :
353 : /*
354 : * Should be called while holding the inode's VFS lock in exclusive mode or in a
355 : * context where no one else can access the inode concurrently (during inode
356 : * creation or when loading an inode from disk).
357 : */
358 19911324 : static inline void btrfs_set_inode_full_sync(struct btrfs_inode *inode)
359 : {
360 19911324 : set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
361 : /*
362 : * The inode may have been part of a reflink operation in the last
363 : * transaction that modified it, and then a fsync has reset the
364 : * last_reflink_trans to avoid subsequent fsyncs in the same
365 : * transaction to do unnecessary work. So update last_reflink_trans
366 : * to the last_trans value (we have to be pessimistic and assume a
367 : * reflink happened).
368 : *
369 : * The ->last_trans is protected by the inode's spinlock and we can
370 : * have a concurrent ordered extent completion update it. Also set
371 : * last_reflink_trans to ->last_trans only if the former is less than
372 : * the later, because we can be called in a context where
373 : * last_reflink_trans was set to the current transaction generation
374 : * while ->last_trans was not yet updated in the current transaction,
375 : * and therefore has a lower value.
376 : */
377 19911413 : spin_lock(&inode->lock);
378 19911328 : if (inode->last_reflink_trans < inode->last_trans)
379 1154954 : inode->last_reflink_trans = inode->last_trans;
380 19911328 : spin_unlock(&inode->lock);
381 19911317 : }
382 :
383 532425 : static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
384 : {
385 532425 : bool ret = false;
386 :
387 532425 : spin_lock(&inode->lock);
388 532431 : if (inode->logged_trans == generation &&
389 499079 : inode->last_sub_trans <= inode->last_log_commit &&
390 19191 : inode->last_sub_trans <= inode->root->last_log_commit)
391 2213 : ret = true;
392 532431 : spin_unlock(&inode->lock);
393 532425 : return ret;
394 : }
395 :
396 : /*
397 : * Check if the inode has flags compatible with compression
398 : */
399 : static inline bool btrfs_inode_can_compress(const struct btrfs_inode *inode)
400 : {
401 4544796 : if (inode->flags & BTRFS_INODE_NODATACOW ||
402 : inode->flags & BTRFS_INODE_NODATASUM)
403 : return false;
404 : return true;
405 : }
406 :
407 : /* Array of bytes with variable length, hexadecimal format 0x1234 */
408 : #define CSUM_FMT "0x%*phN"
409 : #define CSUM_FMT_VALUE(size, bytes) size, bytes
410 :
411 : int btrfs_check_sector_csum(struct btrfs_fs_info *fs_info, struct page *page,
412 : u32 pgoff, u8 *csum, const u8 * const csum_expected);
413 : bool btrfs_data_csum_ok(struct btrfs_bio *bbio, struct btrfs_device *dev,
414 : u32 bio_offset, struct bio_vec *bv);
415 : noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
416 : u64 *orig_start, u64 *orig_block_len,
417 : u64 *ram_bytes, bool nowait, bool strict);
418 :
419 : void __btrfs_del_delalloc_inode(struct btrfs_root *root, struct btrfs_inode *inode);
420 : struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);
421 : int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index);
422 : int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
423 : struct btrfs_inode *dir, struct btrfs_inode *inode,
424 : const struct fscrypt_str *name);
425 : int btrfs_add_link(struct btrfs_trans_handle *trans,
426 : struct btrfs_inode *parent_inode, struct btrfs_inode *inode,
427 : const struct fscrypt_str *name, int add_backref, u64 index);
428 : int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry);
429 : int btrfs_truncate_block(struct btrfs_inode *inode, loff_t from, loff_t len,
430 : int front);
431 :
432 : int btrfs_start_delalloc_snapshot(struct btrfs_root *root, bool in_reclaim_context);
433 : int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, long nr,
434 : bool in_reclaim_context);
435 : int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
436 : unsigned int extra_bits,
437 : struct extent_state **cached_state);
438 :
439 : struct btrfs_new_inode_args {
440 : /* Input */
441 : struct inode *dir;
442 : struct dentry *dentry;
443 : struct inode *inode;
444 : bool orphan;
445 : bool subvol;
446 :
447 : /* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */
448 : struct posix_acl *default_acl;
449 : struct posix_acl *acl;
450 : struct fscrypt_name fname;
451 : };
452 :
453 : int btrfs_new_inode_prepare(struct btrfs_new_inode_args *args,
454 : unsigned int *trans_num_items);
455 : int btrfs_create_new_inode(struct btrfs_trans_handle *trans,
456 : struct btrfs_new_inode_args *args);
457 : void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args *args);
458 : struct inode *btrfs_new_subvol_inode(struct mnt_idmap *idmap,
459 : struct inode *dir);
460 : void btrfs_set_delalloc_extent(struct btrfs_inode *inode, struct extent_state *state,
461 : u32 bits);
462 : void btrfs_clear_delalloc_extent(struct btrfs_inode *inode,
463 : struct extent_state *state, u32 bits);
464 : void btrfs_merge_delalloc_extent(struct btrfs_inode *inode, struct extent_state *new,
465 : struct extent_state *other);
466 : void btrfs_split_delalloc_extent(struct btrfs_inode *inode,
467 : struct extent_state *orig, u64 split);
468 : void btrfs_set_range_writeback(struct btrfs_inode *inode, u64 start, u64 end);
469 : vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf);
470 : void btrfs_evict_inode(struct inode *inode);
471 : struct inode *btrfs_alloc_inode(struct super_block *sb);
472 : void btrfs_destroy_inode(struct inode *inode);
473 : void btrfs_free_inode(struct inode *inode);
474 : int btrfs_drop_inode(struct inode *inode);
475 : int __init btrfs_init_cachep(void);
476 : void __cold btrfs_destroy_cachep(void);
477 : struct inode *btrfs_iget_path(struct super_block *s, u64 ino,
478 : struct btrfs_root *root, struct btrfs_path *path);
479 : struct inode *btrfs_iget(struct super_block *s, u64 ino, struct btrfs_root *root);
480 : struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,
481 : struct page *page, size_t pg_offset,
482 : u64 start, u64 end);
483 : int btrfs_update_inode(struct btrfs_trans_handle *trans,
484 : struct btrfs_root *root, struct btrfs_inode *inode);
485 : int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
486 : struct btrfs_root *root, struct btrfs_inode *inode);
487 : int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct btrfs_inode *inode);
488 : int btrfs_orphan_cleanup(struct btrfs_root *root);
489 : int btrfs_cont_expand(struct btrfs_inode *inode, loff_t oldsize, loff_t size);
490 : void btrfs_add_delayed_iput(struct btrfs_inode *inode);
491 : void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info);
492 : int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info);
493 : int btrfs_prealloc_file_range(struct inode *inode, int mode,
494 : u64 start, u64 num_bytes, u64 min_size,
495 : loff_t actual_len, u64 *alloc_hint);
496 : int btrfs_prealloc_file_range_trans(struct inode *inode,
497 : struct btrfs_trans_handle *trans, int mode,
498 : u64 start, u64 num_bytes, u64 min_size,
499 : loff_t actual_len, u64 *alloc_hint);
500 : int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct page *locked_page,
501 : u64 start, u64 end, int *page_started,
502 : unsigned long *nr_written, struct writeback_control *wbc);
503 : int btrfs_writepage_cow_fixup(struct page *page);
504 : void btrfs_writepage_endio_finish_ordered(struct btrfs_inode *inode,
505 : struct page *page, u64 start,
506 : u64 end, bool uptodate);
507 : int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info *fs_info,
508 : int compress_type);
509 : int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode *inode,
510 : u64 file_offset, u64 disk_bytenr,
511 : u64 disk_io_size,
512 : struct page **pages);
513 : ssize_t btrfs_encoded_read(struct kiocb *iocb, struct iov_iter *iter,
514 : struct btrfs_ioctl_encoded_io_args *encoded);
515 : ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,
516 : const struct btrfs_ioctl_encoded_io_args *encoded);
517 :
518 : ssize_t btrfs_dio_read(struct kiocb *iocb, struct iov_iter *iter,
519 : size_t done_before);
520 : struct iomap_dio *btrfs_dio_write(struct kiocb *iocb, struct iov_iter *iter,
521 : size_t done_before);
522 :
523 : extern const struct dentry_operations btrfs_dentry_operations;
524 :
525 : /* Inode locking type flags, by default the exclusive lock is taken. */
526 : enum btrfs_ilock_type {
527 : ENUM_BIT(BTRFS_ILOCK_SHARED),
528 : ENUM_BIT(BTRFS_ILOCK_TRY),
529 : ENUM_BIT(BTRFS_ILOCK_MMAP),
530 : };
531 :
532 : int btrfs_inode_lock(struct btrfs_inode *inode, unsigned int ilock_flags);
533 : void btrfs_inode_unlock(struct btrfs_inode *inode, unsigned int ilock_flags);
534 : void btrfs_update_inode_bytes(struct btrfs_inode *inode, const u64 add_bytes,
535 : const u64 del_bytes);
536 : void btrfs_assert_inode_range_clean(struct btrfs_inode *inode, u64 start, u64 end);
537 :
538 : #endif
|
