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,
46 * Set when we are in a context where we need to start a transaction and
47 * have dirty pages with the respective file range locked. This is to
48 * ensure that when reserving space for the transaction, if we are low
49 * on available space and need to flush delalloc, we will not flush
50 * delalloc for this inode, because that could result in a deadlock (on
51 * the file range, inode's io_tree).
53 BTRFS_INODE_NO_DELALLOC_FLUSH,
55 * Set when we are working on enabling verity for a file. Computing and
56 * writing the whole Merkle tree can take a while so we want to prevent
57 * races where two separate tasks attempt to simultaneously start verity
60 BTRFS_INODE_VERITY_IN_PROGRESS,
63 /* in memory btrfs inode */
65 /* which subvolume this inode belongs to */
66 struct btrfs_root *root;
68 /* key used to find this inode on disk. This is used by the code
69 * to read in roots of subvolumes
71 struct btrfs_key location;
74 * Lock for counters and all fields used to determine if the inode is in
75 * the log or not (last_trans, last_sub_trans, last_log_commit,
76 * logged_trans), to access/update new_delalloc_bytes and to update the
77 * VFS' inode number of bytes used.
81 /* the extent_tree has caches of all the extent mappings to disk */
82 struct extent_map_tree extent_tree;
84 /* the io_tree does range state (DIRTY, LOCKED etc) */
85 struct extent_io_tree io_tree;
87 /* special utility tree used to record which mirrors have already been
88 * tried when checksums fail for a given block
90 struct extent_io_tree io_failure_tree;
93 * Keep track of where the inode has extent items mapped in order to
94 * make sure the i_size adjustments are accurate
96 struct extent_io_tree file_extent_tree;
98 /* held while logging the inode in tree-log.c */
99 struct mutex log_mutex;
101 /* used to order data wrt metadata */
102 struct btrfs_ordered_inode_tree ordered_tree;
104 /* list of all the delalloc inodes in the FS. There are times we need
105 * to write all the delalloc pages to disk, and this list is used
108 struct list_head delalloc_inodes;
110 /* node for the red-black tree that links inodes in subvolume root */
111 struct rb_node rb_node;
113 unsigned long runtime_flags;
115 /* Keep track of who's O_SYNC/fsyncing currently */
116 atomic_t sync_writers;
118 /* full 64 bit generation number, struct vfs_inode doesn't have a big
119 * enough field for this.
124 * transid of the trans_handle that last modified this inode
129 * transid that last logged this inode
134 * log transid when this inode was last modified
138 /* a local copy of root's last_log_commit */
143 * Total number of bytes pending delalloc, used by stat to
144 * calculate the real block usage of the file. This is used
149 * The offset of the last dir item key that was logged.
150 * This is used only for directories.
152 u64 last_dir_item_offset;
157 * Total number of bytes pending delalloc that fall within a file
158 * range that is either a hole or beyond EOF (and no prealloc extent
159 * exists in the range). This is always <= delalloc_bytes and this
160 * is used only for files.
162 u64 new_delalloc_bytes;
164 * The offset of the last dir index key that was logged.
165 * This is used only for directories.
167 u64 last_dir_index_offset;
171 * total number of bytes pending defrag, used by stat to check whether
177 * the size of the file stored in the metadata on disk. data=ordered
178 * means the in-memory i_size might be larger than the size on disk
179 * because not all the blocks are written yet.
184 * if this is a directory then index_cnt is the counter for the index
185 * number for new files that are created
189 /* Cache the directory index number to speed the dir/file remove */
192 /* the fsync log has some corner cases that mean we have to check
193 * directories to see if any unlinks have been done before
194 * the directory was logged. See tree-log.c for all the
197 u64 last_unlink_trans;
200 * The id/generation of the last transaction where this inode was
201 * either the source or the destination of a clone/dedupe operation.
202 * Used when logging an inode to know if there are shared extents that
203 * need special care when logging checksum items, to avoid duplicate
204 * checksum items in a log (which can lead to a corruption where we end
205 * up with missing checksum ranges after log replay).
206 * Protected by the vfs inode lock.
208 u64 last_reflink_trans;
211 * Number of bytes outstanding that are going to need csums. This is
212 * used in ENOSPC accounting.
216 /* Backwards incompatible flags, lower half of inode_item::flags */
218 /* Read-only compatibility flags, upper half of inode_item::flags */
222 * Counters to keep track of the number of extent item's we may use due
223 * to delalloc and such. outstanding_extents is the number of extent
224 * items we think we'll end up using, and reserved_extents is the number
225 * of extent items we've reserved metadata for.
227 unsigned outstanding_extents;
229 struct btrfs_block_rsv block_rsv;
232 * Cached values of inode properties
234 unsigned prop_compress; /* per-file compression algorithm */
236 * Force compression on the file using the defrag ioctl, could be
237 * different from prop_compress and takes precedence if set
239 unsigned defrag_compress;
241 struct btrfs_delayed_node *delayed_node;
243 /* File creation time. */
244 struct timespec64 i_otime;
246 /* Hook into fs_info->delayed_iputs */
247 struct list_head delayed_iput;
249 struct rw_semaphore i_mmap_lock;
250 struct inode vfs_inode;
253 static inline u32 btrfs_inode_sectorsize(const struct btrfs_inode *inode)
255 return inode->root->fs_info->sectorsize;
258 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
260 return container_of(inode, struct btrfs_inode, vfs_inode);
263 static inline unsigned long btrfs_inode_hash(u64 objectid,
264 const struct btrfs_root *root)
266 u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
268 #if BITS_PER_LONG == 32
269 h = (h >> 32) ^ (h & 0xffffffff);
272 return (unsigned long)h;
275 static inline void btrfs_insert_inode_hash(struct inode *inode)
277 unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
279 __insert_inode_hash(inode, h);
282 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
284 u64 ino = inode->location.objectid;
288 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
290 if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
291 ino = inode->vfs_inode.i_ino;
295 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
297 i_size_write(&inode->vfs_inode, size);
298 inode->disk_i_size = size;
301 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
303 struct btrfs_root *root = inode->root;
305 if (root == root->fs_info->tree_root &&
306 btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
308 if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID)
313 static inline bool is_data_inode(struct inode *inode)
315 return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
318 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
321 lockdep_assert_held(&inode->lock);
322 inode->outstanding_extents += mod;
323 if (btrfs_is_free_space_inode(inode))
325 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
330 * Called every time after doing a buffered, direct IO or memory mapped write.
332 * This is to ensure that if we write to a file that was previously fsynced in
333 * the current transaction, then try to fsync it again in the same transaction,
334 * we will know that there were changes in the file and that it needs to be
337 static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)
339 spin_lock(&inode->lock);
340 inode->last_sub_trans = inode->root->log_transid;
341 spin_unlock(&inode->lock);
344 static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
348 spin_lock(&inode->lock);
349 if (inode->logged_trans == generation &&
350 inode->last_sub_trans <= inode->last_log_commit &&
351 inode->last_sub_trans <= inode->root->last_log_commit)
353 spin_unlock(&inode->lock);
357 struct btrfs_dio_private {
361 * Since DIO can use anonymous page, we cannot use page_offset() to
362 * grab the file offset, thus need a dedicated member for file offset.
366 /* Used for bio::bi_size */
370 * References to this structure. There is one reference per in-flight
371 * bio plus one while we're still setting up.
375 /* dio_bio came from fs/direct-io.c */
378 /* Array of checksums */
383 * btrfs_inode_item stores flags in a u64, btrfs_inode stores them in two
384 * separate u32s. These two functions convert between the two representations.
386 static inline u64 btrfs_inode_combine_flags(u32 flags, u32 ro_flags)
388 return (flags | ((u64)ro_flags << 32));
391 static inline void btrfs_inode_split_flags(u64 inode_item_flags,
392 u32 *flags, u32 *ro_flags)
394 *flags = (u32)inode_item_flags;
395 *ro_flags = (u32)(inode_item_flags >> 32);
398 /* Array of bytes with variable length, hexadecimal format 0x1234 */
399 #define CSUM_FMT "0x%*phN"
400 #define CSUM_FMT_VALUE(size, bytes) size, bytes
402 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
403 u64 logical_start, u8 *csum, u8 *csum_expected, int mirror_num)
405 struct btrfs_root *root = inode->root;
406 const u32 csum_size = root->fs_info->csum_size;
408 /* Output minus objectid, which is more meaningful */
409 if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID)
410 btrfs_warn_rl(root->fs_info,
411 "csum failed root %lld ino %lld off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d",
412 root->root_key.objectid, btrfs_ino(inode),
414 CSUM_FMT_VALUE(csum_size, csum),
415 CSUM_FMT_VALUE(csum_size, csum_expected),
418 btrfs_warn_rl(root->fs_info,
419 "csum failed root %llu ino %llu off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d",
420 root->root_key.objectid, btrfs_ino(inode),
422 CSUM_FMT_VALUE(csum_size, csum),
423 CSUM_FMT_VALUE(csum_size, csum_expected),