1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (C) 2007 Oracle. All rights reserved.
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"
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.
24 BTRFS_INODE_FLUSH_ON_CLOSE
,
26 BTRFS_INODE_IN_DEFRAG
,
27 BTRFS_INODE_HAS_ASYNC_EXTENT
,
29 * Always set under the VFS' inode lock, otherwise it can cause races
30 * during fsync (we start as a fast fsync and then end up in a full
31 * fsync racing with ordered extent completion).
33 BTRFS_INODE_NEEDS_FULL_SYNC
,
34 BTRFS_INODE_COPY_EVERYTHING
,
35 BTRFS_INODE_IN_DELALLOC_LIST
,
36 BTRFS_INODE_HAS_PROPS
,
37 BTRFS_INODE_SNAPSHOT_FLUSH
,
39 * Set and used when logging an inode and it serves to signal that an
40 * inode does not have xattrs, so subsequent fsyncs can avoid searching
41 * for xattrs to log. This bit must be cleared whenever a xattr is added
44 BTRFS_INODE_NO_XATTRS
,
47 /* in memory btrfs inode */
49 /* which subvolume this inode belongs to */
50 struct btrfs_root
*root
;
52 /* key used to find this inode on disk. This is used by the code
53 * to read in roots of subvolumes
55 struct btrfs_key location
;
58 * Lock for counters and all fields used to determine if the inode is in
59 * the log or not (last_trans, last_sub_trans, last_log_commit,
60 * logged_trans), to access/update new_delalloc_bytes and to update the
61 * VFS' inode number of bytes used.
65 /* the extent_tree has caches of all the extent mappings to disk */
66 struct extent_map_tree extent_tree
;
68 /* the io_tree does range state (DIRTY, LOCKED etc) */
69 struct extent_io_tree io_tree
;
71 /* special utility tree used to record which mirrors have already been
72 * tried when checksums fail for a given block
74 struct extent_io_tree io_failure_tree
;
77 * Keep track of where the inode has extent items mapped in order to
78 * make sure the i_size adjustments are accurate
80 struct extent_io_tree file_extent_tree
;
82 /* held while logging the inode in tree-log.c */
83 struct mutex log_mutex
;
85 /* used to order data wrt metadata */
86 struct btrfs_ordered_inode_tree ordered_tree
;
88 /* list of all the delalloc inodes in the FS. There are times we need
89 * to write all the delalloc pages to disk, and this list is used
92 struct list_head delalloc_inodes
;
94 /* node for the red-black tree that links inodes in subvolume root */
95 struct rb_node rb_node
;
97 unsigned long runtime_flags
;
99 /* Keep track of who's O_SYNC/fsyncing currently */
100 atomic_t sync_writers
;
102 /* full 64 bit generation number, struct vfs_inode doesn't have a big
103 * enough field for this.
108 * transid of the trans_handle that last modified this inode
113 * transid that last logged this inode
118 * log transid when this inode was last modified
122 /* a local copy of root's last_log_commit */
125 /* total number of bytes pending delalloc, used by stat to calc the
126 * real block usage of the file
131 * Total number of bytes pending delalloc that fall within a file
132 * range that is either a hole or beyond EOF (and no prealloc extent
133 * exists in the range). This is always <= delalloc_bytes.
135 u64 new_delalloc_bytes
;
138 * total number of bytes pending defrag, used by stat to check whether
144 * the size of the file stored in the metadata on disk. data=ordered
145 * means the in-memory i_size might be larger than the size on disk
146 * because not all the blocks are written yet.
151 * if this is a directory then index_cnt is the counter for the index
152 * number for new files that are created
156 /* Cache the directory index number to speed the dir/file remove */
159 /* the fsync log has some corner cases that mean we have to check
160 * directories to see if any unlinks have been done before
161 * the directory was logged. See tree-log.c for all the
164 u64 last_unlink_trans
;
167 * The id/generation of the last transaction where this inode was
168 * either the source or the destination of a clone/dedupe operation.
169 * Used when logging an inode to know if there are shared extents that
170 * need special care when logging checksum items, to avoid duplicate
171 * checksum items in a log (which can lead to a corruption where we end
172 * up with missing checksum ranges after log replay).
173 * Protected by the vfs inode lock.
175 u64 last_reflink_trans
;
178 * Number of bytes outstanding that are going to need csums. This is
179 * used in ENOSPC accounting.
183 /* flags field from the on disk inode */
187 * Counters to keep track of the number of extent item's we may use due
188 * to delalloc and such. outstanding_extents is the number of extent
189 * items we think we'll end up using, and reserved_extents is the number
190 * of extent items we've reserved metadata for.
192 unsigned outstanding_extents
;
194 struct btrfs_block_rsv block_rsv
;
197 * Cached values of inode properties
199 unsigned prop_compress
; /* per-file compression algorithm */
201 * Force compression on the file using the defrag ioctl, could be
202 * different from prop_compress and takes precedence if set
204 unsigned defrag_compress
;
206 struct btrfs_delayed_node
*delayed_node
;
208 /* File creation time. */
209 struct timespec64 i_otime
;
211 /* Hook into fs_info->delayed_iputs */
212 struct list_head delayed_iput
;
214 struct inode vfs_inode
;
217 static inline u32
btrfs_inode_sectorsize(const struct btrfs_inode
*inode
)
219 return inode
->root
->fs_info
->sectorsize
;
222 static inline struct btrfs_inode
*BTRFS_I(const struct inode
*inode
)
224 return container_of(inode
, struct btrfs_inode
, vfs_inode
);
227 static inline unsigned long btrfs_inode_hash(u64 objectid
,
228 const struct btrfs_root
*root
)
230 u64 h
= objectid
^ (root
->root_key
.objectid
* GOLDEN_RATIO_PRIME
);
232 #if BITS_PER_LONG == 32
233 h
= (h
>> 32) ^ (h
& 0xffffffff);
236 return (unsigned long)h
;
239 static inline void btrfs_insert_inode_hash(struct inode
*inode
)
241 unsigned long h
= btrfs_inode_hash(inode
->i_ino
, BTRFS_I(inode
)->root
);
243 __insert_inode_hash(inode
, h
);
246 static inline u64
btrfs_ino(const struct btrfs_inode
*inode
)
248 u64 ino
= inode
->location
.objectid
;
252 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
254 if (!ino
|| inode
->location
.type
== BTRFS_ROOT_ITEM_KEY
)
255 ino
= inode
->vfs_inode
.i_ino
;
259 static inline void btrfs_i_size_write(struct btrfs_inode
*inode
, u64 size
)
261 i_size_write(&inode
->vfs_inode
, size
);
262 inode
->disk_i_size
= size
;
265 static inline bool btrfs_is_free_space_inode(struct btrfs_inode
*inode
)
267 struct btrfs_root
*root
= inode
->root
;
269 if (root
== root
->fs_info
->tree_root
&&
270 btrfs_ino(inode
) != BTRFS_BTREE_INODE_OBJECTID
)
272 if (inode
->location
.objectid
== BTRFS_FREE_INO_OBJECTID
)
277 static inline bool is_data_inode(struct inode
*inode
)
279 return btrfs_ino(BTRFS_I(inode
)) != BTRFS_BTREE_INODE_OBJECTID
;
282 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode
*inode
,
285 lockdep_assert_held(&inode
->lock
);
286 inode
->outstanding_extents
+= mod
;
287 if (btrfs_is_free_space_inode(inode
))
289 trace_btrfs_inode_mod_outstanding_extents(inode
->root
, btrfs_ino(inode
),
293 static inline int btrfs_inode_in_log(struct btrfs_inode
*inode
, u64 generation
)
297 spin_lock(&inode
->lock
);
298 if (inode
->logged_trans
== generation
&&
299 inode
->last_sub_trans
<= inode
->last_log_commit
&&
300 inode
->last_sub_trans
<= inode
->root
->last_log_commit
) {
302 * After a ranged fsync we might have left some extent maps
303 * (that fall outside the fsync's range). So return false
304 * here if the list isn't empty, to make sure btrfs_log_inode()
305 * will be called and process those extent maps.
308 if (list_empty(&inode
->extent_tree
.modified_extents
))
311 spin_unlock(&inode
->lock
);
315 struct btrfs_dio_private
{
322 * References to this structure. There is one reference per in-flight
323 * bio plus one while we're still setting up.
327 /* dio_bio came from fs/direct-io.c */
330 /* Array of checksums */
334 /* Array of bytes with variable length, hexadecimal format 0x1234 */
335 #define CSUM_FMT "0x%*phN"
336 #define CSUM_FMT_VALUE(size, bytes) size, bytes
338 static inline void btrfs_print_data_csum_error(struct btrfs_inode
*inode
,
339 u64 logical_start
, u8
*csum
, u8
*csum_expected
, int mirror_num
)
341 struct btrfs_root
*root
= inode
->root
;
342 const u32 csum_size
= root
->fs_info
->csum_size
;
344 /* Output minus objectid, which is more meaningful */
345 if (root
->root_key
.objectid
>= BTRFS_LAST_FREE_OBJECTID
)
346 btrfs_warn_rl(root
->fs_info
,
347 "csum failed root %lld ino %lld off %llu csum " CSUM_FMT
" expected csum " CSUM_FMT
" mirror %d",
348 root
->root_key
.objectid
, btrfs_ino(inode
),
350 CSUM_FMT_VALUE(csum_size
, csum
),
351 CSUM_FMT_VALUE(csum_size
, csum_expected
),
354 btrfs_warn_rl(root
->fs_info
,
355 "csum failed root %llu ino %llu off %llu csum " CSUM_FMT
" expected csum " CSUM_FMT
" mirror %d",
356 root
->root_key
.objectid
, btrfs_ino(inode
),
358 CSUM_FMT_VALUE(csum_size
, csum
),
359 CSUM_FMT_VALUE(csum_size
, csum_expected
),