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/spinlock.h>
12#include <linux/mutex.h>
13#include <linux/rwsem.h>
14#include <linux/fs.h>
15#include <linux/mm.h>
16#include <linux/compiler.h>
17#include <linux/fscrypt.h>
18#include <linux/lockdep.h>
19#include <uapi/linux/btrfs_tree.h>
20#include <trace/events/btrfs.h>
21#include "block-rsv.h"
22#include "btrfs_inode.h"
23#include "extent_map.h"
24#include "extent_io.h"
25#include "extent-io-tree.h"
26#include "ordered-data.h"
27#include "delayed-inode.h"
28
29struct extent_state;
30struct posix_acl;
31struct iov_iter;
32struct writeback_control;
33struct btrfs_root;
34struct btrfs_fs_info;
35struct btrfs_trans_handle;
36
37/*
38 * Since we search a directory based on f_pos (struct dir_context::pos) we have
39 * to start at 2 since '.' and '..' have f_pos of 0 and 1 respectively, so
40 * everybody else has to start at 2 (see btrfs_real_readdir() and dir_emit_dots()).
41 */
42#define BTRFS_DIR_START_INDEX 2
43
44/*
45 * ordered_data_close is set by truncate when a file that used
46 * to have good data has been truncated to zero.  When it is set
47 * the btrfs file release call will add this inode to the
48 * ordered operations list so that we make sure to flush out any
49 * new data the application may have written before commit.
50 */
51enum {
52	BTRFS_INODE_FLUSH_ON_CLOSE,
53	BTRFS_INODE_DUMMY,
54	BTRFS_INODE_IN_DEFRAG,
55	BTRFS_INODE_HAS_ASYNC_EXTENT,
56	 /*
57	  * Always set under the VFS' inode lock, otherwise it can cause races
58	  * during fsync (we start as a fast fsync and then end up in a full
59	  * fsync racing with ordered extent completion).
60	  */
61	BTRFS_INODE_NEEDS_FULL_SYNC,
62	BTRFS_INODE_COPY_EVERYTHING,
63	BTRFS_INODE_HAS_PROPS,
64	BTRFS_INODE_SNAPSHOT_FLUSH,
65	/*
66	 * Set and used when logging an inode and it serves to signal that an
67	 * inode does not have xattrs, so subsequent fsyncs can avoid searching
68	 * for xattrs to log. This bit must be cleared whenever a xattr is added
69	 * to an inode.
70	 */
71	BTRFS_INODE_NO_XATTRS,
72	/*
73	 * Set when we are in a context where we need to start a transaction and
74	 * have dirty pages with the respective file range locked. This is to
75	 * ensure that when reserving space for the transaction, if we are low
76	 * on available space and need to flush delalloc, we will not flush
77	 * delalloc for this inode, because that could result in a deadlock (on
78	 * the file range, inode's io_tree).
79	 */
80	BTRFS_INODE_NO_DELALLOC_FLUSH,
81	/*
82	 * Set when we are working on enabling verity for a file. Computing and
83	 * writing the whole Merkle tree can take a while so we want to prevent
84	 * races where two separate tasks attempt to simultaneously start verity
85	 * on the same file.
86	 */
87	BTRFS_INODE_VERITY_IN_PROGRESS,
88	/* Set when this inode is a free space inode. */
89	BTRFS_INODE_FREE_SPACE_INODE,
90	/* Set when there are no capabilities in XATTs for the inode. */
91	BTRFS_INODE_NO_CAP_XATTR,
92};
93
94/* in memory btrfs inode */
95struct btrfs_inode {
96	/* which subvolume this inode belongs to */
97	struct btrfs_root *root;
98
99	/* key used to find this inode on disk.  This is used by the code
100	 * to read in roots of subvolumes
101	 */
102	struct btrfs_key location;
103
104	/* Cached value of inode property 'compression'. */
105	u8 prop_compress;
106
107	/*
108	 * Force compression on the file using the defrag ioctl, could be
109	 * different from prop_compress and takes precedence if set.
110	 */
111	u8 defrag_compress;
112
113	/*
114	 * Lock for counters and all fields used to determine if the inode is in
115	 * the log or not (last_trans, last_sub_trans, last_log_commit,
116	 * logged_trans), to access/update delalloc_bytes, new_delalloc_bytes,
117	 * defrag_bytes, disk_i_size, outstanding_extents, csum_bytes and to
118	 * update the VFS' inode number of bytes used.
119	 */
120	spinlock_t lock;
121
122	/* the extent_tree has caches of all the extent mappings to disk */
123	struct extent_map_tree extent_tree;
124
125	/* the io_tree does range state (DIRTY, LOCKED etc) */
126	struct extent_io_tree io_tree;
127
128	/*
129	 * Keep track of where the inode has extent items mapped in order to
130	 * make sure the i_size adjustments are accurate. Not required when the
131	 * filesystem is NO_HOLES, the status can't be set while mounted as
132	 * it's a mkfs-time feature.
133	 */
134	struct extent_io_tree *file_extent_tree;
135
136	/* held while logging the inode in tree-log.c */
137	struct mutex log_mutex;
138
139	/*
140	 * Counters to keep track of the number of extent item's we may use due
141	 * to delalloc and such.  outstanding_extents is the number of extent
142	 * items we think we'll end up using, and reserved_extents is the number
143	 * of extent items we've reserved metadata for. Protected by 'lock'.
144	 */
145	unsigned outstanding_extents;
146
147	/* used to order data wrt metadata */
148	spinlock_t ordered_tree_lock;
149	struct rb_root ordered_tree;
150	struct rb_node *ordered_tree_last;
151
152	/* list of all the delalloc inodes in the FS.  There are times we need
153	 * to write all the delalloc pages to disk, and this list is used
154	 * to walk them all.
155	 */
156	struct list_head delalloc_inodes;
157
158	/* node for the red-black tree that links inodes in subvolume root */
159	struct rb_node rb_node;
160
161	unsigned long runtime_flags;
162
163	/* full 64 bit generation number, struct vfs_inode doesn't have a big
164	 * enough field for this.
165	 */
166	u64 generation;
167
168	/*
169	 * ID of the transaction handle that last modified this inode.
170	 * Protected by 'lock'.
171	 */
172	u64 last_trans;
173
174	/*
175	 * ID of the transaction that last logged this inode.
176	 * Protected by 'lock'.
177	 */
178	u64 logged_trans;
179
180	/*
181	 * Log transaction ID when this inode was last modified.
182	 * Protected by 'lock'.
183	 */
184	int last_sub_trans;
185
186	/* A local copy of root's last_log_commit. Protected by 'lock'. */
187	int last_log_commit;
188
189	union {
190		/*
191		 * Total number of bytes pending delalloc, used by stat to
192		 * calculate the real block usage of the file. This is used
193		 * only for files. Protected by 'lock'.
194		 */
195		u64 delalloc_bytes;
196		/*
197		 * The lowest possible index of the next dir index key which
198		 * points to an inode that needs to be logged.
199		 * This is used only for directories.
200		 * Use the helpers btrfs_get_first_dir_index_to_log() and
201		 * btrfs_set_first_dir_index_to_log() to access this field.
202		 */
203		u64 first_dir_index_to_log;
204	};
205
206	union {
207		/*
208		 * Total number of bytes pending delalloc that fall within a file
209		 * range that is either a hole or beyond EOF (and no prealloc extent
210		 * exists in the range). This is always <= delalloc_bytes and this
211		 * is used only for files. Protected by 'lock'.
212		 */
213		u64 new_delalloc_bytes;
214		/*
215		 * The offset of the last dir index key that was logged.
216		 * This is used only for directories.
217		 */
218		u64 last_dir_index_offset;
219	};
220
221	/*
222	 * Total number of bytes pending defrag, used by stat to check whether
223	 * it needs COW. Protected by 'lock'.
224	 */
225	u64 defrag_bytes;
226
227	/*
228	 * The size of the file stored in the metadata on disk.  data=ordered
229	 * means the in-memory i_size might be larger than the size on disk
230	 * because not all the blocks are written yet. Protected by 'lock'.
231	 */
232	u64 disk_i_size;
233
234	/*
235	 * If this is a directory then index_cnt is the counter for the index
236	 * number for new files that are created. For an empty directory, this
237	 * must be initialized to BTRFS_DIR_START_INDEX.
238	 */
239	u64 index_cnt;
240
241	/* Cache the directory index number to speed the dir/file remove */
242	u64 dir_index;
243
244	/* the fsync log has some corner cases that mean we have to check
245	 * directories to see if any unlinks have been done before
246	 * the directory was logged.  See tree-log.c for all the
247	 * details
248	 */
249	u64 last_unlink_trans;
250
251	/*
252	 * The id/generation of the last transaction where this inode was
253	 * either the source or the destination of a clone/dedupe operation.
254	 * Used when logging an inode to know if there are shared extents that
255	 * need special care when logging checksum items, to avoid duplicate
256	 * checksum items in a log (which can lead to a corruption where we end
257	 * up with missing checksum ranges after log replay).
258	 * Protected by the vfs inode lock.
259	 */
260	u64 last_reflink_trans;
261
262	/*
263	 * Number of bytes outstanding that are going to need csums.  This is
264	 * used in ENOSPC accounting. Protected by 'lock'.
265	 */
266	u64 csum_bytes;
267
268	/* Backwards incompatible flags, lower half of inode_item::flags  */
269	u32 flags;
270	/* Read-only compatibility flags, upper half of inode_item::flags */
271	u32 ro_flags;
272
273	struct btrfs_block_rsv block_rsv;
274
275	struct btrfs_delayed_node *delayed_node;
276
277	/* File creation time. */
278	u64 i_otime_sec;
279	u32 i_otime_nsec;
280
281	/* Hook into fs_info->delayed_iputs */
282	struct list_head delayed_iput;
283
284	struct rw_semaphore i_mmap_lock;
285	struct inode vfs_inode;
286};
287
288static inline u64 btrfs_get_first_dir_index_to_log(const struct btrfs_inode *inode)
289{
290	return READ_ONCE(inode->first_dir_index_to_log);
291}
292
293static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode *inode,
294						    u64 index)
295{
296	WRITE_ONCE(inode->first_dir_index_to_log, index);
297}
298
299static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
300{
301	return container_of(inode, struct btrfs_inode, vfs_inode);
302}
303
304static inline unsigned long btrfs_inode_hash(u64 objectid,
305					     const struct btrfs_root *root)
306{
307	u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
308
309#if BITS_PER_LONG == 32
310	h = (h >> 32) ^ (h & 0xffffffff);
311#endif
312
313	return (unsigned long)h;
314}
315
316#if BITS_PER_LONG == 32
317
318/*
319 * On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so
320 * we use the inode's location objectid which is a u64 to avoid truncation.
321 */
322static inline u64 btrfs_ino(const struct btrfs_inode *inode)
323{
324	u64 ino = inode->location.objectid;
325
326	/* type == BTRFS_ROOT_ITEM_KEY: subvol dir */
327	if (inode->location.type == BTRFS_ROOT_ITEM_KEY)
328		ino = inode->vfs_inode.i_ino;
329	return ino;
330}
331
332#else
333
334static inline u64 btrfs_ino(const struct btrfs_inode *inode)
335{
336	return inode->vfs_inode.i_ino;
337}
338
339#endif
340
341static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
342{
343	i_size_write(&inode->vfs_inode, size);
344	inode->disk_i_size = size;
345}
346
347static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
348{
349	return test_bit(BTRFS_INODE_FREE_SPACE_INODE, &inode->runtime_flags);
350}
351
352static inline bool is_data_inode(struct inode *inode)
353{
354	return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
355}
356
357static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
358						 int mod)
359{
360	lockdep_assert_held(&inode->lock);
361	inode->outstanding_extents += mod;
362	if (btrfs_is_free_space_inode(inode))
363		return;
364	trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
365						  mod, inode->outstanding_extents);
366}
367
368/*
369 * Called every time after doing a buffered, direct IO or memory mapped write.
370 *
371 * This is to ensure that if we write to a file that was previously fsynced in
372 * the current transaction, then try to fsync it again in the same transaction,
373 * we will know that there were changes in the file and that it needs to be
374 * logged.
375 */
376static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)
377{
378	spin_lock(&inode->lock);
379	inode->last_sub_trans = inode->root->log_transid;
380	spin_unlock(&inode->lock);
381}
382
383/*
384 * Should be called while holding the inode's VFS lock in exclusive mode or in a
385 * context where no one else can access the inode concurrently (during inode
386 * creation or when loading an inode from disk).
387 */
388static inline void btrfs_set_inode_full_sync(struct btrfs_inode *inode)
389{
390	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
391	/*
392	 * The inode may have been part of a reflink operation in the last
393	 * transaction that modified it, and then a fsync has reset the
394	 * last_reflink_trans to avoid subsequent fsyncs in the same
395	 * transaction to do unnecessary work. So update last_reflink_trans
396	 * to the last_trans value (we have to be pessimistic and assume a
397	 * reflink happened).
398	 *
399	 * The ->last_trans is protected by the inode's spinlock and we can
400	 * have a concurrent ordered extent completion update it. Also set
401	 * last_reflink_trans to ->last_trans only if the former is less than
402	 * the later, because we can be called in a context where
403	 * last_reflink_trans was set to the current transaction generation
404	 * while ->last_trans was not yet updated in the current transaction,
405	 * and therefore has a lower value.
406	 */
407	spin_lock(&inode->lock);
408	if (inode->last_reflink_trans < inode->last_trans)
409		inode->last_reflink_trans = inode->last_trans;
410	spin_unlock(&inode->lock);
411}
412
413static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
414{
415	bool ret = false;
416
417	spin_lock(&inode->lock);
418	if (inode->logged_trans == generation &&
419	    inode->last_sub_trans <= inode->last_log_commit &&
420	    inode->last_sub_trans <= btrfs_get_root_last_log_commit(inode->root))
421		ret = true;
422	spin_unlock(&inode->lock);
423	return ret;
424}
425
426/*
427 * Check if the inode has flags compatible with compression
428 */
429static inline bool btrfs_inode_can_compress(const struct btrfs_inode *inode)
430{
431	if (inode->flags & BTRFS_INODE_NODATACOW ||
432	    inode->flags & BTRFS_INODE_NODATASUM)
433		return false;
434	return true;
435}
436
437/* Array of bytes with variable length, hexadecimal format 0x1234 */
438#define CSUM_FMT				"0x%*phN"
439#define CSUM_FMT_VALUE(size, bytes)		size, bytes
440
441int btrfs_check_sector_csum(struct btrfs_fs_info *fs_info, struct page *page,
442			    u32 pgoff, u8 *csum, const u8 * const csum_expected);
443bool btrfs_data_csum_ok(struct btrfs_bio *bbio, struct btrfs_device *dev,
444			u32 bio_offset, struct bio_vec *bv);
445noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
446			      u64 *orig_start, u64 *orig_block_len,
447			      u64 *ram_bytes, bool nowait, bool strict);
448
449void btrfs_del_delalloc_inode(struct btrfs_inode *inode);
450struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);
451int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index);
452int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
453		       struct btrfs_inode *dir, struct btrfs_inode *inode,
454		       const struct fscrypt_str *name);
455int btrfs_add_link(struct btrfs_trans_handle *trans,
456		   struct btrfs_inode *parent_inode, struct btrfs_inode *inode,
457		   const struct fscrypt_str *name, int add_backref, u64 index);
458int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry);
459int btrfs_truncate_block(struct btrfs_inode *inode, loff_t from, loff_t len,
460			 int front);
461
462int btrfs_start_delalloc_snapshot(struct btrfs_root *root, bool in_reclaim_context);
463int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, long nr,
464			       bool in_reclaim_context);
465int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
466			      unsigned int extra_bits,
467			      struct extent_state **cached_state);
468
469struct btrfs_new_inode_args {
470	/* Input */
471	struct inode *dir;
472	struct dentry *dentry;
473	struct inode *inode;
474	bool orphan;
475	bool subvol;
476
477	/* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */
478	struct posix_acl *default_acl;
479	struct posix_acl *acl;
480	struct fscrypt_name fname;
481};
482
483int btrfs_new_inode_prepare(struct btrfs_new_inode_args *args,
484			    unsigned int *trans_num_items);
485int btrfs_create_new_inode(struct btrfs_trans_handle *trans,
486			   struct btrfs_new_inode_args *args);
487void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args *args);
488struct inode *btrfs_new_subvol_inode(struct mnt_idmap *idmap,
489				     struct inode *dir);
490 void btrfs_set_delalloc_extent(struct btrfs_inode *inode, struct extent_state *state,
491			        u32 bits);
492void btrfs_clear_delalloc_extent(struct btrfs_inode *inode,
493				 struct extent_state *state, u32 bits);
494void btrfs_merge_delalloc_extent(struct btrfs_inode *inode, struct extent_state *new,
495				 struct extent_state *other);
496void btrfs_split_delalloc_extent(struct btrfs_inode *inode,
497				 struct extent_state *orig, u64 split);
498void btrfs_set_range_writeback(struct btrfs_inode *inode, u64 start, u64 end);
499vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf);
500void btrfs_evict_inode(struct inode *inode);
501struct inode *btrfs_alloc_inode(struct super_block *sb);
502void btrfs_destroy_inode(struct inode *inode);
503void btrfs_free_inode(struct inode *inode);
504int btrfs_drop_inode(struct inode *inode);
505int __init btrfs_init_cachep(void);
506void __cold btrfs_destroy_cachep(void);
507struct inode *btrfs_iget_path(struct super_block *s, u64 ino,
508			      struct btrfs_root *root, struct btrfs_path *path);
509struct inode *btrfs_iget(struct super_block *s, u64 ino, struct btrfs_root *root);
510struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,
511				    struct page *page, u64 start, u64 len);
512int btrfs_update_inode(struct btrfs_trans_handle *trans,
513		       struct btrfs_inode *inode);
514int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
515				struct btrfs_inode *inode);
516int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct btrfs_inode *inode);
517int btrfs_orphan_cleanup(struct btrfs_root *root);
518int btrfs_cont_expand(struct btrfs_inode *inode, loff_t oldsize, loff_t size);
519void btrfs_add_delayed_iput(struct btrfs_inode *inode);
520void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info);
521int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info);
522int btrfs_prealloc_file_range(struct inode *inode, int mode,
523			      u64 start, u64 num_bytes, u64 min_size,
524			      loff_t actual_len, u64 *alloc_hint);
525int btrfs_prealloc_file_range_trans(struct inode *inode,
526				    struct btrfs_trans_handle *trans, int mode,
527				    u64 start, u64 num_bytes, u64 min_size,
528				    loff_t actual_len, u64 *alloc_hint);
529int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct page *locked_page,
530			     u64 start, u64 end, struct writeback_control *wbc);
531int btrfs_writepage_cow_fixup(struct page *page);
532int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info *fs_info,
533					     int compress_type);
534int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode *inode,
535					  u64 file_offset, u64 disk_bytenr,
536					  u64 disk_io_size,
537					  struct page **pages);
538ssize_t btrfs_encoded_read(struct kiocb *iocb, struct iov_iter *iter,
539			   struct btrfs_ioctl_encoded_io_args *encoded);
540ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,
541			       const struct btrfs_ioctl_encoded_io_args *encoded);
542
543ssize_t btrfs_dio_read(struct kiocb *iocb, struct iov_iter *iter,
544		       size_t done_before);
545struct iomap_dio *btrfs_dio_write(struct kiocb *iocb, struct iov_iter *iter,
546				  size_t done_before);
547
548extern const struct dentry_operations btrfs_dentry_operations;
549
550/* Inode locking type flags, by default the exclusive lock is taken. */
551enum btrfs_ilock_type {
552	ENUM_BIT(BTRFS_ILOCK_SHARED),
553	ENUM_BIT(BTRFS_ILOCK_TRY),
554	ENUM_BIT(BTRFS_ILOCK_MMAP),
555};
556
557int btrfs_inode_lock(struct btrfs_inode *inode, unsigned int ilock_flags);
558void btrfs_inode_unlock(struct btrfs_inode *inode, unsigned int ilock_flags);
559void btrfs_update_inode_bytes(struct btrfs_inode *inode, const u64 add_bytes,
560			      const u64 del_bytes);
561void btrfs_assert_inode_range_clean(struct btrfs_inode *inode, u64 start, u64 end);
562
563#endif
564