xref: /linux/fs/btrfs/btrfs_inode.h (revision 2f804aca48322f02a8f44cca540663845ee80fb1)
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 	/* Keep track of who's O_SYNC/fsyncing currently */
120 	atomic_t sync_writers;
121 
122 	/* full 64 bit generation number, struct vfs_inode doesn't have a big
123 	 * enough field for this.
124 	 */
125 	u64 generation;
126 
127 	/*
128 	 * transid of the trans_handle that last modified this inode
129 	 */
130 	u64 last_trans;
131 
132 	/*
133 	 * transid that last logged this inode
134 	 */
135 	u64 logged_trans;
136 
137 	/*
138 	 * log transid when this inode was last modified
139 	 */
140 	int last_sub_trans;
141 
142 	/* a local copy of root's last_log_commit */
143 	int last_log_commit;
144 
145 	union {
146 		/*
147 		 * Total number of bytes pending delalloc, used by stat to
148 		 * calculate the real block usage of the file. This is used
149 		 * only for files.
150 		 */
151 		u64 delalloc_bytes;
152 		/*
153 		 * The lowest possible index of the next dir index key which
154 		 * points to an inode that needs to be logged.
155 		 * This is used only for directories.
156 		 * Use the helpers btrfs_get_first_dir_index_to_log() and
157 		 * btrfs_set_first_dir_index_to_log() to access this field.
158 		 */
159 		u64 first_dir_index_to_log;
160 	};
161 
162 	union {
163 		/*
164 		 * Total number of bytes pending delalloc that fall within a file
165 		 * range that is either a hole or beyond EOF (and no prealloc extent
166 		 * exists in the range). This is always <= delalloc_bytes and this
167 		 * is used only for files.
168 		 */
169 		u64 new_delalloc_bytes;
170 		/*
171 		 * The offset of the last dir index key that was logged.
172 		 * This is used only for directories.
173 		 */
174 		u64 last_dir_index_offset;
175 	};
176 
177 	/*
178 	 * total number of bytes pending defrag, used by stat to check whether
179 	 * it needs COW.
180 	 */
181 	u64 defrag_bytes;
182 
183 	/*
184 	 * the size of the file stored in the metadata on disk.  data=ordered
185 	 * means the in-memory i_size might be larger than the size on disk
186 	 * because not all the blocks are written yet.
187 	 */
188 	u64 disk_i_size;
189 
190 	/*
191 	 * If this is a directory then index_cnt is the counter for the index
192 	 * number for new files that are created. For an empty directory, this
193 	 * must be initialized to BTRFS_DIR_START_INDEX.
194 	 */
195 	u64 index_cnt;
196 
197 	/* Cache the directory index number to speed the dir/file remove */
198 	u64 dir_index;
199 
200 	/* the fsync log has some corner cases that mean we have to check
201 	 * directories to see if any unlinks have been done before
202 	 * the directory was logged.  See tree-log.c for all the
203 	 * details
204 	 */
205 	u64 last_unlink_trans;
206 
207 	/*
208 	 * The id/generation of the last transaction where this inode was
209 	 * either the source or the destination of a clone/dedupe operation.
210 	 * Used when logging an inode to know if there are shared extents that
211 	 * need special care when logging checksum items, to avoid duplicate
212 	 * checksum items in a log (which can lead to a corruption where we end
213 	 * up with missing checksum ranges after log replay).
214 	 * Protected by the vfs inode lock.
215 	 */
216 	u64 last_reflink_trans;
217 
218 	/*
219 	 * Number of bytes outstanding that are going to need csums.  This is
220 	 * used in ENOSPC accounting.
221 	 */
222 	u64 csum_bytes;
223 
224 	/* Backwards incompatible flags, lower half of inode_item::flags  */
225 	u32 flags;
226 	/* Read-only compatibility flags, upper half of inode_item::flags */
227 	u32 ro_flags;
228 
229 	/*
230 	 * Counters to keep track of the number of extent item's we may use due
231 	 * to delalloc and such.  outstanding_extents is the number of extent
232 	 * items we think we'll end up using, and reserved_extents is the number
233 	 * of extent items we've reserved metadata for.
234 	 */
235 	unsigned outstanding_extents;
236 
237 	struct btrfs_block_rsv block_rsv;
238 
239 	/*
240 	 * Cached values of inode properties
241 	 */
242 	unsigned prop_compress;		/* per-file compression algorithm */
243 	/*
244 	 * Force compression on the file using the defrag ioctl, could be
245 	 * different from prop_compress and takes precedence if set
246 	 */
247 	unsigned defrag_compress;
248 
249 	struct btrfs_delayed_node *delayed_node;
250 
251 	/* File creation time. */
252 	struct timespec64 i_otime;
253 
254 	/* Hook into fs_info->delayed_iputs */
255 	struct list_head delayed_iput;
256 
257 	struct rw_semaphore i_mmap_lock;
258 	struct inode vfs_inode;
259 };
260 
261 static inline u64 btrfs_get_first_dir_index_to_log(const struct btrfs_inode *inode)
262 {
263 	return READ_ONCE(inode->first_dir_index_to_log);
264 }
265 
266 static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode *inode,
267 						    u64 index)
268 {
269 	WRITE_ONCE(inode->first_dir_index_to_log, index);
270 }
271 
272 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
273 {
274 	return container_of(inode, struct btrfs_inode, vfs_inode);
275 }
276 
277 static inline unsigned long btrfs_inode_hash(u64 objectid,
278 					     const struct btrfs_root *root)
279 {
280 	u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
281 
282 #if BITS_PER_LONG == 32
283 	h = (h >> 32) ^ (h & 0xffffffff);
284 #endif
285 
286 	return (unsigned long)h;
287 }
288 
289 #if BITS_PER_LONG == 32
290 
291 /*
292  * On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so
293  * we use the inode's location objectid which is a u64 to avoid truncation.
294  */
295 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
296 {
297 	u64 ino = inode->location.objectid;
298 
299 	/* type == BTRFS_ROOT_ITEM_KEY: subvol dir */
300 	if (inode->location.type == BTRFS_ROOT_ITEM_KEY)
301 		ino = inode->vfs_inode.i_ino;
302 	return ino;
303 }
304 
305 #else
306 
307 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
308 {
309 	return inode->vfs_inode.i_ino;
310 }
311 
312 #endif
313 
314 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
315 {
316 	i_size_write(&inode->vfs_inode, size);
317 	inode->disk_i_size = size;
318 }
319 
320 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
321 {
322 	return test_bit(BTRFS_INODE_FREE_SPACE_INODE, &inode->runtime_flags);
323 }
324 
325 static inline bool is_data_inode(struct inode *inode)
326 {
327 	return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
328 }
329 
330 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
331 						 int mod)
332 {
333 	lockdep_assert_held(&inode->lock);
334 	inode->outstanding_extents += mod;
335 	if (btrfs_is_free_space_inode(inode))
336 		return;
337 	trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
338 						  mod);
339 }
340 
341 /*
342  * Called every time after doing a buffered, direct IO or memory mapped write.
343  *
344  * This is to ensure that if we write to a file that was previously fsynced in
345  * the current transaction, then try to fsync it again in the same transaction,
346  * we will know that there were changes in the file and that it needs to be
347  * logged.
348  */
349 static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)
350 {
351 	spin_lock(&inode->lock);
352 	inode->last_sub_trans = inode->root->log_transid;
353 	spin_unlock(&inode->lock);
354 }
355 
356 /*
357  * Should be called while holding the inode's VFS lock in exclusive mode or in a
358  * context where no one else can access the inode concurrently (during inode
359  * creation or when loading an inode from disk).
360  */
361 static inline void btrfs_set_inode_full_sync(struct btrfs_inode *inode)
362 {
363 	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
364 	/*
365 	 * The inode may have been part of a reflink operation in the last
366 	 * transaction that modified it, and then a fsync has reset the
367 	 * last_reflink_trans to avoid subsequent fsyncs in the same
368 	 * transaction to do unnecessary work. So update last_reflink_trans
369 	 * to the last_trans value (we have to be pessimistic and assume a
370 	 * reflink happened).
371 	 *
372 	 * The ->last_trans is protected by the inode's spinlock and we can
373 	 * have a concurrent ordered extent completion update it. Also set
374 	 * last_reflink_trans to ->last_trans only if the former is less than
375 	 * the later, because we can be called in a context where
376 	 * last_reflink_trans was set to the current transaction generation
377 	 * while ->last_trans was not yet updated in the current transaction,
378 	 * and therefore has a lower value.
379 	 */
380 	spin_lock(&inode->lock);
381 	if (inode->last_reflink_trans < inode->last_trans)
382 		inode->last_reflink_trans = inode->last_trans;
383 	spin_unlock(&inode->lock);
384 }
385 
386 static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
387 {
388 	bool ret = false;
389 
390 	spin_lock(&inode->lock);
391 	if (inode->logged_trans == generation &&
392 	    inode->last_sub_trans <= inode->last_log_commit &&
393 	    inode->last_sub_trans <= inode->root->last_log_commit)
394 		ret = true;
395 	spin_unlock(&inode->lock);
396 	return ret;
397 }
398 
399 /*
400  * Check if the inode has flags compatible with compression
401  */
402 static inline bool btrfs_inode_can_compress(const struct btrfs_inode *inode)
403 {
404 	if (inode->flags & BTRFS_INODE_NODATACOW ||
405 	    inode->flags & BTRFS_INODE_NODATASUM)
406 		return false;
407 	return true;
408 }
409 
410 /*
411  * btrfs_inode_item stores flags in a u64, btrfs_inode stores them in two
412  * separate u32s. These two functions convert between the two representations.
413  */
414 static inline u64 btrfs_inode_combine_flags(u32 flags, u32 ro_flags)
415 {
416 	return (flags | ((u64)ro_flags << 32));
417 }
418 
419 static inline void btrfs_inode_split_flags(u64 inode_item_flags,
420 					   u32 *flags, u32 *ro_flags)
421 {
422 	*flags = (u32)inode_item_flags;
423 	*ro_flags = (u32)(inode_item_flags >> 32);
424 }
425 
426 /* Array of bytes with variable length, hexadecimal format 0x1234 */
427 #define CSUM_FMT				"0x%*phN"
428 #define CSUM_FMT_VALUE(size, bytes)		size, bytes
429 
430 int btrfs_check_sector_csum(struct btrfs_fs_info *fs_info, struct page *page,
431 			    u32 pgoff, u8 *csum, const u8 * const csum_expected);
432 int btrfs_extract_ordered_extent(struct btrfs_bio *bbio,
433 				 struct btrfs_ordered_extent *ordered);
434 bool btrfs_data_csum_ok(struct btrfs_bio *bbio, struct btrfs_device *dev,
435 			u32 bio_offset, struct bio_vec *bv);
436 noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
437 			      u64 *orig_start, u64 *orig_block_len,
438 			      u64 *ram_bytes, bool nowait, bool strict);
439 
440 void __btrfs_del_delalloc_inode(struct btrfs_root *root, struct btrfs_inode *inode);
441 struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);
442 int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index);
443 int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
444 		       struct btrfs_inode *dir, struct btrfs_inode *inode,
445 		       const struct fscrypt_str *name);
446 int btrfs_add_link(struct btrfs_trans_handle *trans,
447 		   struct btrfs_inode *parent_inode, struct btrfs_inode *inode,
448 		   const struct fscrypt_str *name, int add_backref, u64 index);
449 int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry);
450 int btrfs_truncate_block(struct btrfs_inode *inode, loff_t from, loff_t len,
451 			 int front);
452 
453 int btrfs_start_delalloc_snapshot(struct btrfs_root *root, bool in_reclaim_context);
454 int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, long nr,
455 			       bool in_reclaim_context);
456 int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
457 			      unsigned int extra_bits,
458 			      struct extent_state **cached_state);
459 
460 struct btrfs_new_inode_args {
461 	/* Input */
462 	struct inode *dir;
463 	struct dentry *dentry;
464 	struct inode *inode;
465 	bool orphan;
466 	bool subvol;
467 
468 	/* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */
469 	struct posix_acl *default_acl;
470 	struct posix_acl *acl;
471 	struct fscrypt_name fname;
472 };
473 
474 int btrfs_new_inode_prepare(struct btrfs_new_inode_args *args,
475 			    unsigned int *trans_num_items);
476 int btrfs_create_new_inode(struct btrfs_trans_handle *trans,
477 			   struct btrfs_new_inode_args *args);
478 void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args *args);
479 struct inode *btrfs_new_subvol_inode(struct mnt_idmap *idmap,
480 				     struct inode *dir);
481  void btrfs_set_delalloc_extent(struct btrfs_inode *inode, struct extent_state *state,
482 			        u32 bits);
483 void btrfs_clear_delalloc_extent(struct btrfs_inode *inode,
484 				 struct extent_state *state, u32 bits);
485 void btrfs_merge_delalloc_extent(struct btrfs_inode *inode, struct extent_state *new,
486 				 struct extent_state *other);
487 void btrfs_split_delalloc_extent(struct btrfs_inode *inode,
488 				 struct extent_state *orig, u64 split);
489 void btrfs_set_range_writeback(struct btrfs_inode *inode, u64 start, u64 end);
490 vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf);
491 void btrfs_evict_inode(struct inode *inode);
492 struct inode *btrfs_alloc_inode(struct super_block *sb);
493 void btrfs_destroy_inode(struct inode *inode);
494 void btrfs_free_inode(struct inode *inode);
495 int btrfs_drop_inode(struct inode *inode);
496 int __init btrfs_init_cachep(void);
497 void __cold btrfs_destroy_cachep(void);
498 struct inode *btrfs_iget_path(struct super_block *s, u64 ino,
499 			      struct btrfs_root *root, struct btrfs_path *path);
500 struct inode *btrfs_iget(struct super_block *s, u64 ino, struct btrfs_root *root);
501 struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,
502 				    struct page *page, size_t pg_offset,
503 				    u64 start, u64 end);
504 int btrfs_update_inode(struct btrfs_trans_handle *trans,
505 		       struct btrfs_root *root, struct btrfs_inode *inode);
506 int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
507 				struct btrfs_root *root, struct btrfs_inode *inode);
508 int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct btrfs_inode *inode);
509 int btrfs_orphan_cleanup(struct btrfs_root *root);
510 int btrfs_cont_expand(struct btrfs_inode *inode, loff_t oldsize, loff_t size);
511 void btrfs_add_delayed_iput(struct btrfs_inode *inode);
512 void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info);
513 int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info);
514 int btrfs_prealloc_file_range(struct inode *inode, int mode,
515 			      u64 start, u64 num_bytes, u64 min_size,
516 			      loff_t actual_len, u64 *alloc_hint);
517 int btrfs_prealloc_file_range_trans(struct inode *inode,
518 				    struct btrfs_trans_handle *trans, int mode,
519 				    u64 start, u64 num_bytes, u64 min_size,
520 				    loff_t actual_len, u64 *alloc_hint);
521 int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct page *locked_page,
522 			     u64 start, u64 end, int *page_started,
523 			     unsigned long *nr_written, struct writeback_control *wbc);
524 int btrfs_writepage_cow_fixup(struct page *page);
525 void btrfs_writepage_endio_finish_ordered(struct btrfs_inode *inode,
526 					  struct page *page, u64 start,
527 					  u64 end, bool uptodate);
528 int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info *fs_info,
529 					     int compress_type);
530 int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode *inode,
531 					  u64 file_offset, u64 disk_bytenr,
532 					  u64 disk_io_size,
533 					  struct page **pages);
534 ssize_t btrfs_encoded_read(struct kiocb *iocb, struct iov_iter *iter,
535 			   struct btrfs_ioctl_encoded_io_args *encoded);
536 ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,
537 			       const struct btrfs_ioctl_encoded_io_args *encoded);
538 
539 ssize_t btrfs_dio_read(struct kiocb *iocb, struct iov_iter *iter,
540 		       size_t done_before);
541 struct iomap_dio *btrfs_dio_write(struct kiocb *iocb, struct iov_iter *iter,
542 				  size_t done_before);
543 
544 extern const struct dentry_operations btrfs_dentry_operations;
545 
546 /* Inode locking type flags, by default the exclusive lock is taken. */
547 enum btrfs_ilock_type {
548 	ENUM_BIT(BTRFS_ILOCK_SHARED),
549 	ENUM_BIT(BTRFS_ILOCK_TRY),
550 	ENUM_BIT(BTRFS_ILOCK_MMAP),
551 };
552 
553 int btrfs_inode_lock(struct btrfs_inode *inode, unsigned int ilock_flags);
554 void btrfs_inode_unlock(struct btrfs_inode *inode, unsigned int ilock_flags);
555 void btrfs_update_inode_bytes(struct btrfs_inode *inode, const u64 add_bytes,
556 			      const u64 del_bytes);
557 void btrfs_assert_inode_range_clean(struct btrfs_inode *inode, u64 start, u64 end);
558 
559 #endif
560