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_ORDERED_DATA_CLOSE,
26 BTRFS_INODE_IN_DEFRAG,
27 BTRFS_INODE_HAS_ASYNC_EXTENT,
28 BTRFS_INODE_NEEDS_FULL_SYNC,
29 BTRFS_INODE_COPY_EVERYTHING,
30 BTRFS_INODE_IN_DELALLOC_LIST,
31 BTRFS_INODE_HAS_PROPS,
32 BTRFS_INODE_SNAPSHOT_FLUSH,
35 /* in memory btrfs inode */
37 /* which subvolume this inode belongs to */
38 struct btrfs_root *root;
40 /* key used to find this inode on disk. This is used by the code
41 * to read in roots of subvolumes
43 struct btrfs_key location;
46 * Lock for counters and all fields used to determine if the inode is in
47 * the log or not (last_trans, last_sub_trans, last_log_commit,
52 /* the extent_tree has caches of all the extent mappings to disk */
53 struct extent_map_tree extent_tree;
55 /* the io_tree does range state (DIRTY, LOCKED etc) */
56 struct extent_io_tree io_tree;
58 /* special utility tree used to record which mirrors have already been
59 * tried when checksums fail for a given block
61 struct extent_io_tree io_failure_tree;
64 * Keep track of where the inode has extent items mapped in order to
65 * make sure the i_size adjustments are accurate
67 struct extent_io_tree file_extent_tree;
69 /* held while logging the inode in tree-log.c */
70 struct mutex log_mutex;
72 /* used to order data wrt metadata */
73 struct btrfs_ordered_inode_tree ordered_tree;
75 /* list of all the delalloc inodes in the FS. There are times we need
76 * to write all the delalloc pages to disk, and this list is used
79 struct list_head delalloc_inodes;
81 /* node for the red-black tree that links inodes in subvolume root */
82 struct rb_node rb_node;
84 unsigned long runtime_flags;
86 /* Keep track of who's O_SYNC/fsyncing currently */
87 atomic_t sync_writers;
89 /* full 64 bit generation number, struct vfs_inode doesn't have a big
90 * enough field for this.
95 * transid of the trans_handle that last modified this inode
100 * transid that last logged this inode
105 * log transid when this inode was last modified
109 /* a local copy of root's last_log_commit */
112 /* total number of bytes pending delalloc, used by stat to calc the
113 * real block usage of the file
118 * Total number of bytes pending delalloc that fall within a file
119 * range that is either a hole or beyond EOF (and no prealloc extent
120 * exists in the range). This is always <= delalloc_bytes.
122 u64 new_delalloc_bytes;
125 * total number of bytes pending defrag, used by stat to check whether
131 * the size of the file stored in the metadata on disk. data=ordered
132 * means the in-memory i_size might be larger than the size on disk
133 * because not all the blocks are written yet.
138 * if this is a directory then index_cnt is the counter for the index
139 * number for new files that are created
143 /* Cache the directory index number to speed the dir/file remove */
146 /* the fsync log has some corner cases that mean we have to check
147 * directories to see if any unlinks have been done before
148 * the directory was logged. See tree-log.c for all the
151 u64 last_unlink_trans;
154 * Number of bytes outstanding that are going to need csums. This is
155 * used in ENOSPC accounting.
159 /* flags field from the on disk inode */
163 * Counters to keep track of the number of extent item's we may use due
164 * to delalloc and such. outstanding_extents is the number of extent
165 * items we think we'll end up using, and reserved_extents is the number
166 * of extent items we've reserved metadata for.
168 unsigned outstanding_extents;
170 struct btrfs_block_rsv block_rsv;
173 * Cached values of inode properties
175 unsigned prop_compress; /* per-file compression algorithm */
177 * Force compression on the file using the defrag ioctl, could be
178 * different from prop_compress and takes precedence if set
180 unsigned defrag_compress;
182 struct btrfs_delayed_node *delayed_node;
184 /* File creation time. */
185 struct timespec64 i_otime;
187 /* Hook into fs_info->delayed_iputs */
188 struct list_head delayed_iput;
191 * To avoid races between lockless (i_mutex not held) direct IO writes
192 * and concurrent fsync requests. Direct IO writes must acquire read
193 * access on this semaphore for creating an extent map and its
194 * corresponding ordered extent. The fast fsync path must acquire write
195 * access on this semaphore before it collects ordered extents and
198 struct rw_semaphore dio_sem;
200 struct inode vfs_inode;
203 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
205 return container_of(inode, struct btrfs_inode, vfs_inode);
208 static inline unsigned long btrfs_inode_hash(u64 objectid,
209 const struct btrfs_root *root)
211 u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
213 #if BITS_PER_LONG == 32
214 h = (h >> 32) ^ (h & 0xffffffff);
217 return (unsigned long)h;
220 static inline void btrfs_insert_inode_hash(struct inode *inode)
222 unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
224 __insert_inode_hash(inode, h);
227 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
229 u64 ino = inode->location.objectid;
233 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
235 if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
236 ino = inode->vfs_inode.i_ino;
240 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
242 i_size_write(&inode->vfs_inode, size);
243 inode->disk_i_size = size;
246 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
248 struct btrfs_root *root = inode->root;
250 if (root == root->fs_info->tree_root &&
251 btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
253 if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID)
258 static inline bool is_data_inode(struct inode *inode)
260 return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
263 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
266 lockdep_assert_held(&inode->lock);
267 inode->outstanding_extents += mod;
268 if (btrfs_is_free_space_inode(inode))
270 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
274 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
278 spin_lock(&inode->lock);
279 if (inode->logged_trans == generation &&
280 inode->last_sub_trans <= inode->last_log_commit &&
281 inode->last_sub_trans <= inode->root->last_log_commit) {
283 * After a ranged fsync we might have left some extent maps
284 * (that fall outside the fsync's range). So return false
285 * here if the list isn't empty, to make sure btrfs_log_inode()
286 * will be called and process those extent maps.
289 if (list_empty(&inode->extent_tree.modified_extents))
292 spin_unlock(&inode->lock);
296 struct btrfs_dio_private {
303 * References to this structure. There is one reference per in-flight
304 * bio plus one while we're still setting up.
308 /* dio_bio came from fs/direct-io.c */
311 /* Array of checksums */
315 /* Array of bytes with variable length, hexadecimal format 0x1234 */
316 #define CSUM_FMT "0x%*phN"
317 #define CSUM_FMT_VALUE(size, bytes) size, bytes
319 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
320 u64 logical_start, u8 *csum, u8 *csum_expected, int mirror_num)
322 struct btrfs_root *root = inode->root;
323 struct btrfs_super_block *sb = root->fs_info->super_copy;
324 const u16 csum_size = btrfs_super_csum_size(sb);
326 /* Output minus objectid, which is more meaningful */
327 if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID)
328 btrfs_warn_rl(root->fs_info,
329 "csum failed root %lld ino %lld off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d",
330 root->root_key.objectid, btrfs_ino(inode),
332 CSUM_FMT_VALUE(csum_size, csum),
333 CSUM_FMT_VALUE(csum_size, csum_expected),
336 btrfs_warn_rl(root->fs_info,
337 "csum failed root %llu ino %llu off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d",
338 root->root_key.objectid, btrfs_ino(inode),
340 CSUM_FMT_VALUE(csum_size, csum),
341 CSUM_FMT_VALUE(csum_size, csum_expected),