2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
22 #include <linux/hash.h>
23 #include "extent_map.h"
24 #include "extent_io.h"
25 #include "ordered-data.h"
26 #include "delayed-inode.h"
29 * ordered_data_close is set by truncate when a file that used
30 * to have good data has been truncated to zero. When it is set
31 * the btrfs file release call will add this inode to the
32 * ordered operations list so that we make sure to flush out any
33 * new data the application may have written before commit.
35 #define BTRFS_INODE_ORDERED_DATA_CLOSE 0
36 #define BTRFS_INODE_ORPHAN_META_RESERVED 1
37 #define BTRFS_INODE_DUMMY 2
38 #define BTRFS_INODE_IN_DEFRAG 3
39 #define BTRFS_INODE_HAS_ORPHAN_ITEM 4
40 #define BTRFS_INODE_HAS_ASYNC_EXTENT 5
41 #define BTRFS_INODE_NEEDS_FULL_SYNC 6
42 #define BTRFS_INODE_COPY_EVERYTHING 7
43 #define BTRFS_INODE_IN_DELALLOC_LIST 8
44 #define BTRFS_INODE_READDIO_NEED_LOCK 9
45 #define BTRFS_INODE_HAS_PROPS 10
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,
64 /* the extent_tree has caches of all the extent mappings to disk */
65 struct extent_map_tree extent_tree
;
67 /* the io_tree does range state (DIRTY, LOCKED etc) */
68 struct extent_io_tree io_tree
;
70 /* special utility tree used to record which mirrors have already been
71 * tried when checksums fail for a given block
73 struct extent_io_tree io_failure_tree
;
75 /* held while logging the inode in tree-log.c */
76 struct mutex log_mutex
;
78 /* held while doing delalloc reservations */
79 struct mutex delalloc_mutex
;
81 /* used to order data wrt metadata */
82 struct btrfs_ordered_inode_tree ordered_tree
;
84 /* list of all the delalloc inodes in the FS. There are times we need
85 * to write all the delalloc pages to disk, and this list is used
88 struct list_head delalloc_inodes
;
90 /* node for the red-black tree that links inodes in subvolume root */
91 struct rb_node rb_node
;
93 unsigned long runtime_flags
;
95 /* Keep track of who's O_SYNC/fsyncing currently */
96 atomic_t sync_writers
;
98 /* full 64 bit generation number, struct vfs_inode doesn't have a big
99 * enough field for this.
104 * transid of the trans_handle that last modified this inode
109 * transid that last logged this inode
114 * log transid when this inode was last modified
118 /* a local copy of root's last_log_commit */
121 /* total number of bytes pending delalloc, used by stat to calc the
122 * real block usage of the file
127 * Total number of bytes pending delalloc that fall within a file
128 * range that is either a hole or beyond EOF (and no prealloc extent
129 * exists in the range). This is always <= delalloc_bytes.
131 u64 new_delalloc_bytes
;
134 * total number of bytes pending defrag, used by stat to check whether
140 * the size of the file stored in the metadata on disk. data=ordered
141 * means the in-memory i_size might be larger than the size on disk
142 * because not all the blocks are written yet.
147 * if this is a directory then index_cnt is the counter for the index
148 * number for new files that are created
152 /* Cache the directory index number to speed the dir/file remove */
155 /* the fsync log has some corner cases that mean we have to check
156 * directories to see if any unlinks have been done before
157 * the directory was logged. See tree-log.c for all the
160 u64 last_unlink_trans
;
163 * Track the transaction id of the last transaction used to create a
164 * hard link for the inode. This is used by the log tree (fsync).
169 * Number of bytes outstanding that are going to need csums. This is
170 * used in ENOSPC accounting.
174 /* flags field from the on disk inode */
178 * Counters to keep track of the number of extent item's we may use due
179 * to delalloc and such. outstanding_extents is the number of extent
180 * items we think we'll end up using, and reserved_extents is the number
181 * of extent items we've reserved metadata for.
183 unsigned outstanding_extents
;
185 struct btrfs_block_rsv block_rsv
;
188 * Cached values of inode properties
190 unsigned prop_compress
; /* per-file compression algorithm */
192 * Force compression on the file using the defrag ioctl, could be
193 * different from prop_compress and takes precedence if set
195 unsigned defrag_compress
;
197 struct btrfs_delayed_node
*delayed_node
;
199 /* File creation time. */
200 struct timespec i_otime
;
202 /* Hook into fs_info->delayed_iputs */
203 struct list_head delayed_iput
;
204 long delayed_iput_count
;
207 * To avoid races between lockless (i_mutex not held) direct IO writes
208 * and concurrent fsync requests. Direct IO writes must acquire read
209 * access on this semaphore for creating an extent map and its
210 * corresponding ordered extent. The fast fsync path must acquire write
211 * access on this semaphore before it collects ordered extents and
214 struct rw_semaphore dio_sem
;
216 struct inode vfs_inode
;
219 extern unsigned char btrfs_filetype_table
[];
221 static inline struct btrfs_inode
*BTRFS_I(const struct inode
*inode
)
223 return container_of(inode
, struct btrfs_inode
, vfs_inode
);
226 static inline unsigned long btrfs_inode_hash(u64 objectid
,
227 const struct btrfs_root
*root
)
229 u64 h
= objectid
^ (root
->objectid
* GOLDEN_RATIO_PRIME
);
231 #if BITS_PER_LONG == 32
232 h
= (h
>> 32) ^ (h
& 0xffffffff);
235 return (unsigned long)h
;
238 static inline void btrfs_insert_inode_hash(struct inode
*inode
)
240 unsigned long h
= btrfs_inode_hash(inode
->i_ino
, BTRFS_I(inode
)->root
);
242 __insert_inode_hash(inode
, h
);
245 static inline u64
btrfs_ino(const struct btrfs_inode
*inode
)
247 u64 ino
= inode
->location
.objectid
;
251 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
253 if (!ino
|| inode
->location
.type
== BTRFS_ROOT_ITEM_KEY
)
254 ino
= inode
->vfs_inode
.i_ino
;
258 static inline void btrfs_i_size_write(struct btrfs_inode
*inode
, u64 size
)
260 i_size_write(&inode
->vfs_inode
, size
);
261 inode
->disk_i_size
= size
;
264 static inline bool btrfs_is_free_space_inode(struct btrfs_inode
*inode
)
266 struct btrfs_root
*root
= inode
->root
;
268 if (root
== root
->fs_info
->tree_root
&&
269 btrfs_ino(inode
) != BTRFS_BTREE_INODE_OBJECTID
)
271 if (inode
->location
.objectid
== BTRFS_FREE_INO_OBJECTID
)
276 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode
*inode
,
279 lockdep_assert_held(&inode
->lock
);
280 inode
->outstanding_extents
+= mod
;
281 if (btrfs_is_free_space_inode(inode
))
283 trace_btrfs_inode_mod_outstanding_extents(inode
->root
, btrfs_ino(inode
),
287 static inline int btrfs_inode_in_log(struct btrfs_inode
*inode
, u64 generation
)
291 spin_lock(&inode
->lock
);
292 if (inode
->logged_trans
== generation
&&
293 inode
->last_sub_trans
<= inode
->last_log_commit
&&
294 inode
->last_sub_trans
<= inode
->root
->last_log_commit
) {
296 * After a ranged fsync we might have left some extent maps
297 * (that fall outside the fsync's range). So return false
298 * here if the list isn't empty, to make sure btrfs_log_inode()
299 * will be called and process those extent maps.
302 if (list_empty(&inode
->extent_tree
.modified_extents
))
305 spin_unlock(&inode
->lock
);
309 #define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1
311 struct btrfs_dio_private
{
319 /* number of bios pending for this dio */
320 atomic_t pending_bios
;
325 /* orig_bio is our btrfs_io_bio */
326 struct bio
*orig_bio
;
328 /* dio_bio came from fs/direct-io.c */
332 * The original bio may be split to several sub-bios, this is
333 * done during endio of sub-bios
335 blk_status_t (*subio_endio
)(struct inode
*, struct btrfs_io_bio
*,
340 * Disable DIO read nolock optimization, so new dio readers will be forced
341 * to grab i_mutex. It is used to avoid the endless truncate due to
342 * nonlocked dio read.
344 static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode
*inode
)
346 set_bit(BTRFS_INODE_READDIO_NEED_LOCK
, &inode
->runtime_flags
);
350 static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode
*inode
)
352 smp_mb__before_atomic();
353 clear_bit(BTRFS_INODE_READDIO_NEED_LOCK
, &inode
->runtime_flags
);
356 static inline void btrfs_print_data_csum_error(struct btrfs_inode
*inode
,
357 u64 logical_start
, u32 csum
, u32 csum_expected
, int mirror_num
)
359 struct btrfs_root
*root
= inode
->root
;
361 /* Output minus objectid, which is more meaningful */
362 if (root
->objectid
>= BTRFS_LAST_FREE_OBJECTID
)
363 btrfs_warn_rl(root
->fs_info
,
364 "csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d",
365 root
->objectid
, btrfs_ino(inode
),
366 logical_start
, csum
, csum_expected
, mirror_num
);
368 btrfs_warn_rl(root
->fs_info
,
369 "csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d",
370 root
->objectid
, btrfs_ino(inode
),
371 logical_start
, csum
, csum_expected
, mirror_num
);
374 bool btrfs_page_exists_in_range(struct inode
*inode
, loff_t start
, loff_t end
);