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 <linux/fscrypt.h>
12#include <trace/events/btrfs.h>
13#include "extent_map.h"
14#include "extent_io.h"
15#include "ordered-data.h"
16#include "delayed-inode.h"
17
18/*
19 * Since we search a directory based on f_pos (struct dir_context::pos) we have
20 * to start at 2 since '.' and '..' have f_pos of 0 and 1 respectively, so
21 * everybody else has to start at 2 (see btrfs_real_readdir() and dir_emit_dots()).
22 */
23#define BTRFS_DIR_START_INDEX 2
24
25/*
26 * ordered_data_close is set by truncate when a file that used
27 * to have good data has been truncated to zero. When it is set
28 * the btrfs file release call will add this inode to the
29 * ordered operations list so that we make sure to flush out any
30 * new data the application may have written before commit.
31 */
32enum {
33 BTRFS_INODE_FLUSH_ON_CLOSE,
34 BTRFS_INODE_DUMMY,
35 BTRFS_INODE_IN_DEFRAG,
36 BTRFS_INODE_HAS_ASYNC_EXTENT,
37 /*
38 * Always set under the VFS' inode lock, otherwise it can cause races
39 * during fsync (we start as a fast fsync and then end up in a full
40 * fsync racing with ordered extent completion).
41 */
42 BTRFS_INODE_NEEDS_FULL_SYNC,
43 BTRFS_INODE_COPY_EVERYTHING,
44 BTRFS_INODE_IN_DELALLOC_LIST,
45 BTRFS_INODE_HAS_PROPS,
46 BTRFS_INODE_SNAPSHOT_FLUSH,
47 /*
48 * Set and used when logging an inode and it serves to signal that an
49 * inode does not have xattrs, so subsequent fsyncs can avoid searching
50 * for xattrs to log. This bit must be cleared whenever a xattr is added
51 * to an inode.
52 */
53 BTRFS_INODE_NO_XATTRS,
54 /*
55 * Set when we are in a context where we need to start a transaction and
56 * have dirty pages with the respective file range locked. This is to
57 * ensure that when reserving space for the transaction, if we are low
58 * on available space and need to flush delalloc, we will not flush
59 * delalloc for this inode, because that could result in a deadlock (on
60 * the file range, inode's io_tree).
61 */
62 BTRFS_INODE_NO_DELALLOC_FLUSH,
63 /*
64 * Set when we are working on enabling verity for a file. Computing and
65 * writing the whole Merkle tree can take a while so we want to prevent
66 * races where two separate tasks attempt to simultaneously start verity
67 * on the same file.
68 */
69 BTRFS_INODE_VERITY_IN_PROGRESS,
70 /* Set when this inode is a free space inode. */
71 BTRFS_INODE_FREE_SPACE_INODE,
72};
73
74/* in memory btrfs inode */
75struct btrfs_inode {
76 /* which subvolume this inode belongs to */
77 struct btrfs_root *root;
78
79 /* key used to find this inode on disk. This is used by the code
80 * to read in roots of subvolumes
81 */
82 struct btrfs_key location;
83
84 /* Cached value of inode property 'compression'. */
85 u8 prop_compress;
86
87 /*
88 * Force compression on the file using the defrag ioctl, could be
89 * different from prop_compress and takes precedence if set.
90 */
91 u8 defrag_compress;
92
93 /*
94 * Lock for counters and all fields used to determine if the inode is in
95 * the log or not (last_trans, last_sub_trans, last_log_commit,
96 * logged_trans), to access/update delalloc_bytes, new_delalloc_bytes,
97 * defrag_bytes, disk_i_size, outstanding_extents, csum_bytes and to
98 * update the VFS' inode number of bytes used.
99 */
100 spinlock_t lock;
101
102 /* the extent_tree has caches of all the extent mappings to disk */
103 struct extent_map_tree extent_tree;
104
105 /* the io_tree does range state (DIRTY, LOCKED etc) */
106 struct extent_io_tree io_tree;
107
108 /*
109 * Keep track of where the inode has extent items mapped in order to
110 * make sure the i_size adjustments are accurate
111 */
112 struct extent_io_tree file_extent_tree;
113
114 /* held while logging the inode in tree-log.c */
115 struct mutex log_mutex;
116
117 /*
118 * Counters to keep track of the number of extent item's we may use due
119 * to delalloc and such. outstanding_extents is the number of extent
120 * items we think we'll end up using, and reserved_extents is the number
121 * of extent items we've reserved metadata for. Protected by 'lock'.
122 */
123 unsigned outstanding_extents;
124
125 /* used to order data wrt metadata */
126 spinlock_t ordered_tree_lock;
127 struct rb_root ordered_tree;
128 struct rb_node *ordered_tree_last;
129
130 /* list of all the delalloc inodes in the FS. There are times we need
131 * to write all the delalloc pages to disk, and this list is used
132 * to walk them all.
133 */
134 struct list_head delalloc_inodes;
135
136 /* node for the red-black tree that links inodes in subvolume root */
137 struct rb_node rb_node;
138
139 unsigned long runtime_flags;
140
141 /* full 64 bit generation number, struct vfs_inode doesn't have a big
142 * enough field for this.
143 */
144 u64 generation;
145
146 /*
147 * ID of the transaction handle that last modified this inode.
148 * Protected by 'lock'.
149 */
150 u64 last_trans;
151
152 /*
153 * ID of the transaction that last logged this inode.
154 * Protected by 'lock'.
155 */
156 u64 logged_trans;
157
158 /*
159 * Log transaction ID when this inode was last modified.
160 * Protected by 'lock'.
161 */
162 int last_sub_trans;
163
164 /* A local copy of root's last_log_commit. Protected by 'lock'. */
165 int last_log_commit;
166
167 union {
168 /*
169 * Total number of bytes pending delalloc, used by stat to
170 * calculate the real block usage of the file. This is used
171 * only for files. Protected by 'lock'.
172 */
173 u64 delalloc_bytes;
174 /*
175 * The lowest possible index of the next dir index key which
176 * points to an inode that needs to be logged.
177 * This is used only for directories.
178 * Use the helpers btrfs_get_first_dir_index_to_log() and
179 * btrfs_set_first_dir_index_to_log() to access this field.
180 */
181 u64 first_dir_index_to_log;
182 };
183
184 union {
185 /*
186 * Total number of bytes pending delalloc that fall within a file
187 * range that is either a hole or beyond EOF (and no prealloc extent
188 * exists in the range). This is always <= delalloc_bytes and this
189 * is used only for files. Protected by 'lock'.
190 */
191 u64 new_delalloc_bytes;
192 /*
193 * The offset of the last dir index key that was logged.
194 * This is used only for directories.
195 */
196 u64 last_dir_index_offset;
197 };
198
199 /*
200 * Total number of bytes pending defrag, used by stat to check whether
201 * it needs COW. Protected by 'lock'.
202 */
203 u64 defrag_bytes;
204
205 /*
206 * The size of the file stored in the metadata on disk. data=ordered
207 * means the in-memory i_size might be larger than the size on disk
208 * because not all the blocks are written yet. Protected by 'lock'.
209 */
210 u64 disk_i_size;
211
212 /*
213 * If this is a directory then index_cnt is the counter for the index
214 * number for new files that are created. For an empty directory, this
215 * must be initialized to BTRFS_DIR_START_INDEX.
216 */
217 u64 index_cnt;
218
219 /* Cache the directory index number to speed the dir/file remove */
220 u64 dir_index;
221
222 /* the fsync log has some corner cases that mean we have to check
223 * directories to see if any unlinks have been done before
224 * the directory was logged. See tree-log.c for all the
225 * details
226 */
227 u64 last_unlink_trans;
228
229 /*
230 * The id/generation of the last transaction where this inode was
231 * either the source or the destination of a clone/dedupe operation.
232 * Used when logging an inode to know if there are shared extents that
233 * need special care when logging checksum items, to avoid duplicate
234 * checksum items in a log (which can lead to a corruption where we end
235 * up with missing checksum ranges after log replay).
236 * Protected by the vfs inode lock.
237 */
238 u64 last_reflink_trans;
239
240 /*
241 * Number of bytes outstanding that are going to need csums. This is
242 * used in ENOSPC accounting. Protected by 'lock'.
243 */
244 u64 csum_bytes;
245
246 /* Backwards incompatible flags, lower half of inode_item::flags */
247 u32 flags;
248 /* Read-only compatibility flags, upper half of inode_item::flags */
249 u32 ro_flags;
250
251 struct btrfs_block_rsv block_rsv;
252
253 struct btrfs_delayed_node *delayed_node;
254
255 /* File creation time. */
256 u64 i_otime_sec;
257 u32 i_otime_nsec;
258
259 /* Hook into fs_info->delayed_iputs */
260 struct list_head delayed_iput;
261
262 struct rw_semaphore i_mmap_lock;
263 struct inode vfs_inode;
264};
265
266static inline u64 btrfs_get_first_dir_index_to_log(const struct btrfs_inode *inode)
267{
268 return READ_ONCE(inode->first_dir_index_to_log);
269}
270
271static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode *inode,
272 u64 index)
273{
274 WRITE_ONCE(inode->first_dir_index_to_log, index);
275}
276
277static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
278{
279 return container_of(inode, struct btrfs_inode, vfs_inode);
280}
281
282static inline unsigned long btrfs_inode_hash(u64 objectid,
283 const struct btrfs_root *root)
284{
285 u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
286
287#if BITS_PER_LONG == 32
288 h = (h >> 32) ^ (h & 0xffffffff);
289#endif
290
291 return (unsigned long)h;
292}
293
294#if BITS_PER_LONG == 32
295
296/*
297 * On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so
298 * we use the inode's location objectid which is a u64 to avoid truncation.
299 */
300static inline u64 btrfs_ino(const struct btrfs_inode *inode)
301{
302 u64 ino = inode->location.objectid;
303
304 /* type == BTRFS_ROOT_ITEM_KEY: subvol dir */
305 if (inode->location.type == BTRFS_ROOT_ITEM_KEY)
306 ino = inode->vfs_inode.i_ino;
307 return ino;
308}
309
310#else
311
312static inline u64 btrfs_ino(const struct btrfs_inode *inode)
313{
314 return inode->vfs_inode.i_ino;
315}
316
317#endif
318
319static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
320{
321 i_size_write(inode: &inode->vfs_inode, i_size: size);
322 inode->disk_i_size = size;
323}
324
325static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
326{
327 return test_bit(BTRFS_INODE_FREE_SPACE_INODE, &inode->runtime_flags);
328}
329
330static inline bool is_data_inode(struct inode *inode)
331{
332 return btrfs_ino(inode: BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
333}
334
335static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
336 int mod)
337{
338 lockdep_assert_held(&inode->lock);
339 inode->outstanding_extents += mod;
340 if (btrfs_is_free_space_inode(inode))
341 return;
342 trace_btrfs_inode_mod_outstanding_extents(root: inode->root, ino: btrfs_ino(inode),
343 mod, outstanding: inode->outstanding_extents);
344}
345
346/*
347 * Called every time after doing a buffered, direct IO or memory mapped write.
348 *
349 * This is to ensure that if we write to a file that was previously fsynced in
350 * the current transaction, then try to fsync it again in the same transaction,
351 * we will know that there were changes in the file and that it needs to be
352 * logged.
353 */
354static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)
355{
356 spin_lock(lock: &inode->lock);
357 inode->last_sub_trans = inode->root->log_transid;
358 spin_unlock(lock: &inode->lock);
359}
360
361/*
362 * Should be called while holding the inode's VFS lock in exclusive mode or in a
363 * context where no one else can access the inode concurrently (during inode
364 * creation or when loading an inode from disk).
365 */
366static inline void btrfs_set_inode_full_sync(struct btrfs_inode *inode)
367{
368 set_bit(nr: BTRFS_INODE_NEEDS_FULL_SYNC, addr: &inode->runtime_flags);
369 /*
370 * The inode may have been part of a reflink operation in the last
371 * transaction that modified it, and then a fsync has reset the
372 * last_reflink_trans to avoid subsequent fsyncs in the same
373 * transaction to do unnecessary work. So update last_reflink_trans
374 * to the last_trans value (we have to be pessimistic and assume a
375 * reflink happened).
376 *
377 * The ->last_trans is protected by the inode's spinlock and we can
378 * have a concurrent ordered extent completion update it. Also set
379 * last_reflink_trans to ->last_trans only if the former is less than
380 * the later, because we can be called in a context where
381 * last_reflink_trans was set to the current transaction generation
382 * while ->last_trans was not yet updated in the current transaction,
383 * and therefore has a lower value.
384 */
385 spin_lock(lock: &inode->lock);
386 if (inode->last_reflink_trans < inode->last_trans)
387 inode->last_reflink_trans = inode->last_trans;
388 spin_unlock(lock: &inode->lock);
389}
390
391static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
392{
393 bool ret = false;
394
395 spin_lock(lock: &inode->lock);
396 if (inode->logged_trans == generation &&
397 inode->last_sub_trans <= inode->last_log_commit &&
398 inode->last_sub_trans <= btrfs_get_root_last_log_commit(root: inode->root))
399 ret = true;
400 spin_unlock(lock: &inode->lock);
401 return ret;
402}
403
404/*
405 * Check if the inode has flags compatible with compression
406 */
407static inline bool btrfs_inode_can_compress(const struct btrfs_inode *inode)
408{
409 if (inode->flags & BTRFS_INODE_NODATACOW ||
410 inode->flags & BTRFS_INODE_NODATASUM)
411 return false;
412 return true;
413}
414
415/* Array of bytes with variable length, hexadecimal format 0x1234 */
416#define CSUM_FMT "0x%*phN"
417#define CSUM_FMT_VALUE(size, bytes) size, bytes
418
419int btrfs_check_sector_csum(struct btrfs_fs_info *fs_info, struct page *page,
420 u32 pgoff, u8 *csum, const u8 * const csum_expected);
421bool btrfs_data_csum_ok(struct btrfs_bio *bbio, struct btrfs_device *dev,
422 u32 bio_offset, struct bio_vec *bv);
423noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
424 u64 *orig_start, u64 *orig_block_len,
425 u64 *ram_bytes, bool nowait, bool strict);
426
427void __btrfs_del_delalloc_inode(struct btrfs_root *root, struct btrfs_inode *inode);
428struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);
429int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index);
430int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
431 struct btrfs_inode *dir, struct btrfs_inode *inode,
432 const struct fscrypt_str *name);
433int btrfs_add_link(struct btrfs_trans_handle *trans,
434 struct btrfs_inode *parent_inode, struct btrfs_inode *inode,
435 const struct fscrypt_str *name, int add_backref, u64 index);
436int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry);
437int btrfs_truncate_block(struct btrfs_inode *inode, loff_t from, loff_t len,
438 int front);
439
440int btrfs_start_delalloc_snapshot(struct btrfs_root *root, bool in_reclaim_context);
441int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, long nr,
442 bool in_reclaim_context);
443int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
444 unsigned int extra_bits,
445 struct extent_state **cached_state);
446
447struct btrfs_new_inode_args {
448 /* Input */
449 struct inode *dir;
450 struct dentry *dentry;
451 struct inode *inode;
452 bool orphan;
453 bool subvol;
454
455 /* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */
456 struct posix_acl *default_acl;
457 struct posix_acl *acl;
458 struct fscrypt_name fname;
459};
460
461int btrfs_new_inode_prepare(struct btrfs_new_inode_args *args,
462 unsigned int *trans_num_items);
463int btrfs_create_new_inode(struct btrfs_trans_handle *trans,
464 struct btrfs_new_inode_args *args);
465void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args *args);
466struct inode *btrfs_new_subvol_inode(struct mnt_idmap *idmap,
467 struct inode *dir);
468 void btrfs_set_delalloc_extent(struct btrfs_inode *inode, struct extent_state *state,
469 u32 bits);
470void btrfs_clear_delalloc_extent(struct btrfs_inode *inode,
471 struct extent_state *state, u32 bits);
472void btrfs_merge_delalloc_extent(struct btrfs_inode *inode, struct extent_state *new,
473 struct extent_state *other);
474void btrfs_split_delalloc_extent(struct btrfs_inode *inode,
475 struct extent_state *orig, u64 split);
476void btrfs_set_range_writeback(struct btrfs_inode *inode, u64 start, u64 end);
477vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf);
478void btrfs_evict_inode(struct inode *inode);
479struct inode *btrfs_alloc_inode(struct super_block *sb);
480void btrfs_destroy_inode(struct inode *inode);
481void btrfs_free_inode(struct inode *inode);
482int btrfs_drop_inode(struct inode *inode);
483int __init btrfs_init_cachep(void);
484void __cold btrfs_destroy_cachep(void);
485struct inode *btrfs_iget_path(struct super_block *s, u64 ino,
486 struct btrfs_root *root, struct btrfs_path *path);
487struct inode *btrfs_iget(struct super_block *s, u64 ino, struct btrfs_root *root);
488struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,
489 struct page *page, size_t pg_offset,
490 u64 start, u64 end);
491int btrfs_update_inode(struct btrfs_trans_handle *trans,
492 struct btrfs_inode *inode);
493int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
494 struct btrfs_inode *inode);
495int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct btrfs_inode *inode);
496int btrfs_orphan_cleanup(struct btrfs_root *root);
497int btrfs_cont_expand(struct btrfs_inode *inode, loff_t oldsize, loff_t size);
498void btrfs_add_delayed_iput(struct btrfs_inode *inode);
499void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info);
500int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info);
501int btrfs_prealloc_file_range(struct inode *inode, int mode,
502 u64 start, u64 num_bytes, u64 min_size,
503 loff_t actual_len, u64 *alloc_hint);
504int btrfs_prealloc_file_range_trans(struct inode *inode,
505 struct btrfs_trans_handle *trans, int mode,
506 u64 start, u64 num_bytes, u64 min_size,
507 loff_t actual_len, u64 *alloc_hint);
508int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct page *locked_page,
509 u64 start, u64 end, struct writeback_control *wbc);
510int btrfs_writepage_cow_fixup(struct page *page);
511int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info *fs_info,
512 int compress_type);
513int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode *inode,
514 u64 file_offset, u64 disk_bytenr,
515 u64 disk_io_size,
516 struct page **pages);
517ssize_t btrfs_encoded_read(struct kiocb *iocb, struct iov_iter *iter,
518 struct btrfs_ioctl_encoded_io_args *encoded);
519ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,
520 const struct btrfs_ioctl_encoded_io_args *encoded);
521
522ssize_t btrfs_dio_read(struct kiocb *iocb, struct iov_iter *iter,
523 size_t done_before);
524struct iomap_dio *btrfs_dio_write(struct kiocb *iocb, struct iov_iter *iter,
525 size_t done_before);
526
527extern const struct dentry_operations btrfs_dentry_operations;
528
529/* Inode locking type flags, by default the exclusive lock is taken. */
530enum btrfs_ilock_type {
531 ENUM_BIT(BTRFS_ILOCK_SHARED),
532 ENUM_BIT(BTRFS_ILOCK_TRY),
533 ENUM_BIT(BTRFS_ILOCK_MMAP),
534};
535
536int btrfs_inode_lock(struct btrfs_inode *inode, unsigned int ilock_flags);
537void btrfs_inode_unlock(struct btrfs_inode *inode, unsigned int ilock_flags);
538void btrfs_update_inode_bytes(struct btrfs_inode *inode, const u64 add_bytes,
539 const u64 del_bytes);
540void btrfs_assert_inode_range_clean(struct btrfs_inode *inode, u64 start, u64 end);
541
542#endif
543

source code of linux/fs/btrfs/btrfs_inode.h