1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (C) 2007 Oracle. All rights reserved.
9 #include <linux/hash.h>
10 #include "extent_map.h"
11 #include "extent_io.h"
12 #include "ordered-data.h"
13 #include "delayed-inode.h"
16 * ordered_data_close is set by truncate when a file that used
17 * to have good data has been truncated to zero. When it is set
18 * the btrfs file release call will add this inode to the
19 * ordered operations list so that we make sure to flush out any
20 * new data the application may have written before commit.
23 BTRFS_INODE_ORDERED_DATA_CLOSE = 0,
25 BTRFS_INODE_IN_DEFRAG,
26 BTRFS_INODE_HAS_ASYNC_EXTENT,
27 BTRFS_INODE_NEEDS_FULL_SYNC,
28 BTRFS_INODE_COPY_EVERYTHING,
29 BTRFS_INODE_IN_DELALLOC_LIST,
30 BTRFS_INODE_READDIO_NEED_LOCK,
31 BTRFS_INODE_HAS_PROPS,
34 /* in memory btrfs inode */
36 /* which subvolume this inode belongs to */
37 struct btrfs_root *root;
39 /* key used to find this inode on disk. This is used by the code
40 * to read in roots of subvolumes
42 struct btrfs_key location;
45 * Lock for counters and all fields used to determine if the inode is in
46 * the log or not (last_trans, last_sub_trans, last_log_commit,
51 /* the extent_tree has caches of all the extent mappings to disk */
52 struct extent_map_tree extent_tree;
54 /* the io_tree does range state (DIRTY, LOCKED etc) */
55 struct extent_io_tree io_tree;
57 /* special utility tree used to record which mirrors have already been
58 * tried when checksums fail for a given block
60 struct extent_io_tree io_failure_tree;
62 /* held while logging the inode in tree-log.c */
63 struct mutex log_mutex;
65 /* held while doing delalloc reservations */
66 struct mutex delalloc_mutex;
68 /* used to order data wrt metadata */
69 struct btrfs_ordered_inode_tree ordered_tree;
71 /* list of all the delalloc inodes in the FS. There are times we need
72 * to write all the delalloc pages to disk, and this list is used
75 struct list_head delalloc_inodes;
77 /* node for the red-black tree that links inodes in subvolume root */
78 struct rb_node rb_node;
80 unsigned long runtime_flags;
82 /* Keep track of who's O_SYNC/fsyncing currently */
83 atomic_t sync_writers;
85 /* full 64 bit generation number, struct vfs_inode doesn't have a big
86 * enough field for this.
91 * transid of the trans_handle that last modified this inode
96 * transid that last logged this inode
101 * log transid when this inode was last modified
105 /* a local copy of root's last_log_commit */
108 /* total number of bytes pending delalloc, used by stat to calc the
109 * real block usage of the file
114 * Total number of bytes pending delalloc that fall within a file
115 * range that is either a hole or beyond EOF (and no prealloc extent
116 * exists in the range). This is always <= delalloc_bytes.
118 u64 new_delalloc_bytes;
121 * total number of bytes pending defrag, used by stat to check whether
127 * the size of the file stored in the metadata on disk. data=ordered
128 * means the in-memory i_size might be larger than the size on disk
129 * because not all the blocks are written yet.
134 * if this is a directory then index_cnt is the counter for the index
135 * number for new files that are created
139 /* Cache the directory index number to speed the dir/file remove */
142 /* the fsync log has some corner cases that mean we have to check
143 * directories to see if any unlinks have been done before
144 * the directory was logged. See tree-log.c for all the
147 u64 last_unlink_trans;
150 * Track the transaction id of the last transaction used to create a
151 * hard link for the inode. This is used by the log tree (fsync).
156 * Number of bytes outstanding that are going to need csums. This is
157 * used in ENOSPC accounting.
161 /* flags field from the on disk inode */
165 * Counters to keep track of the number of extent item's we may use due
166 * to delalloc and such. outstanding_extents is the number of extent
167 * items we think we'll end up using, and reserved_extents is the number
168 * of extent items we've reserved metadata for.
170 unsigned outstanding_extents;
172 struct btrfs_block_rsv block_rsv;
175 * Cached values of inode properties
177 unsigned prop_compress; /* per-file compression algorithm */
179 * Force compression on the file using the defrag ioctl, could be
180 * different from prop_compress and takes precedence if set
182 unsigned defrag_compress;
184 struct btrfs_delayed_node *delayed_node;
186 /* File creation time. */
187 struct timespec64 i_otime;
189 /* Hook into fs_info->delayed_iputs */
190 struct list_head delayed_iput;
193 * To avoid races between lockless (i_mutex not held) direct IO writes
194 * and concurrent fsync requests. Direct IO writes must acquire read
195 * access on this semaphore for creating an extent map and its
196 * corresponding ordered extent. The fast fsync path must acquire write
197 * access on this semaphore before it collects ordered extents and
200 struct rw_semaphore dio_sem;
202 struct inode vfs_inode;
205 extern unsigned char btrfs_filetype_table[];
207 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
209 return container_of(inode, struct btrfs_inode, vfs_inode);
212 static inline unsigned long btrfs_inode_hash(u64 objectid,
213 const struct btrfs_root *root)
215 u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
217 #if BITS_PER_LONG == 32
218 h = (h >> 32) ^ (h & 0xffffffff);
221 return (unsigned long)h;
224 static inline void btrfs_insert_inode_hash(struct inode *inode)
226 unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
228 __insert_inode_hash(inode, h);
231 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
233 u64 ino = inode->location.objectid;
237 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
239 if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
240 ino = inode->vfs_inode.i_ino;
244 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
246 i_size_write(&inode->vfs_inode, size);
247 inode->disk_i_size = size;
250 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
252 struct btrfs_root *root = inode->root;
254 if (root == root->fs_info->tree_root &&
255 btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
257 if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID)
262 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
265 lockdep_assert_held(&inode->lock);
266 inode->outstanding_extents += mod;
267 if (btrfs_is_free_space_inode(inode))
269 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
273 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
277 spin_lock(&inode->lock);
278 if (inode->logged_trans == generation &&
279 inode->last_sub_trans <= inode->last_log_commit &&
280 inode->last_sub_trans <= inode->root->last_log_commit) {
282 * After a ranged fsync we might have left some extent maps
283 * (that fall outside the fsync's range). So return false
284 * here if the list isn't empty, to make sure btrfs_log_inode()
285 * will be called and process those extent maps.
288 if (list_empty(&inode->extent_tree.modified_extents))
291 spin_unlock(&inode->lock);
295 #define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1
297 struct btrfs_dio_private {
305 /* number of bios pending for this dio */
306 atomic_t pending_bios;
311 /* orig_bio is our btrfs_io_bio */
312 struct bio *orig_bio;
314 /* dio_bio came from fs/direct-io.c */
318 * The original bio may be split to several sub-bios, this is
319 * done during endio of sub-bios
321 blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *,
326 * Disable DIO read nolock optimization, so new dio readers will be forced
327 * to grab i_mutex. It is used to avoid the endless truncate due to
328 * nonlocked dio read.
330 static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode)
332 set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
336 static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode)
338 smp_mb__before_atomic();
339 clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
342 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
343 u64 logical_start, u32 csum, u32 csum_expected, int mirror_num)
345 struct btrfs_root *root = inode->root;
347 /* Output minus objectid, which is more meaningful */
348 if (root->objectid >= BTRFS_LAST_FREE_OBJECTID)
349 btrfs_warn_rl(root->fs_info,
350 "csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d",
351 root->objectid, btrfs_ino(inode),
352 logical_start, csum, csum_expected, mirror_num);
354 btrfs_warn_rl(root->fs_info,
355 "csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d",
356 root->objectid, btrfs_ino(inode),
357 logical_start, csum, csum_expected, mirror_num);