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_DELALLOC_META_RESERVED 4
40 #define BTRFS_INODE_HAS_ORPHAN_ITEM 5
41 #define BTRFS_INODE_HAS_ASYNC_EXTENT 6
42 #define BTRFS_INODE_NEEDS_FULL_SYNC 7
43 #define BTRFS_INODE_COPY_EVERYTHING 8
44 #define BTRFS_INODE_IN_DELALLOC_LIST 9
45 #define BTRFS_INODE_READDIO_NEED_LOCK 10
46 #define BTRFS_INODE_HAS_PROPS 11
48 /* in memory btrfs inode */
50 /* which subvolume this inode belongs to */
51 struct btrfs_root *root;
53 /* key used to find this inode on disk. This is used by the code
54 * to read in roots of subvolumes
56 struct btrfs_key location;
59 * Lock for counters and all fields used to determine if the inode is in
60 * the log or not (last_trans, last_sub_trans, last_log_commit,
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;
76 /* held while logging the inode in tree-log.c */
77 struct mutex log_mutex;
79 /* held while doing delalloc reservations */
80 struct mutex delalloc_mutex;
82 /* used to order data wrt metadata */
83 struct btrfs_ordered_inode_tree ordered_tree;
85 /* list of all the delalloc inodes in the FS. There are times we need
86 * to write all the delalloc pages to disk, and this list is used
89 struct list_head delalloc_inodes;
91 /* node for the red-black tree that links inodes in subvolume root */
92 struct rb_node rb_node;
94 unsigned long runtime_flags;
96 /* Keep track of who's O_SYNC/fsyncing currently */
97 atomic_t sync_writers;
99 /* full 64 bit generation number, struct vfs_inode doesn't have a big
100 * enough field for this.
105 * transid of the trans_handle that last modified this inode
110 * transid that last logged this inode
115 * log transid when this inode was last modified
119 /* a local copy of root's last_log_commit */
122 /* total number of bytes pending delalloc, used by stat to calc the
123 * real block usage of the file
128 * Total number of bytes pending delalloc that fall within a file
129 * range that is either a hole or beyond EOF (and no prealloc extent
130 * exists in the range). This is always <= delalloc_bytes.
132 u64 new_delalloc_bytes;
135 * total number of bytes pending defrag, used by stat to check whether
141 * the size of the file stored in the metadata on disk. data=ordered
142 * means the in-memory i_size might be larger than the size on disk
143 * because not all the blocks are written yet.
148 * if this is a directory then index_cnt is the counter for the index
149 * number for new files that are created
153 /* Cache the directory index number to speed the dir/file remove */
156 /* the fsync log has some corner cases that mean we have to check
157 * directories to see if any unlinks have been done before
158 * the directory was logged. See tree-log.c for all the
161 u64 last_unlink_trans;
164 * Number of bytes outstanding that are going to need csums. This is
165 * used in ENOSPC accounting.
169 /* flags field from the on disk inode */
173 * Counters to keep track of the number of extent item's we may use due
174 * to delalloc and such. outstanding_extents is the number of extent
175 * items we think we'll end up using, and reserved_extents is the number
176 * of extent items we've reserved metadata for.
178 unsigned outstanding_extents;
179 unsigned reserved_extents;
182 * Cached values of inode properties
184 unsigned prop_compress; /* per-file compression algorithm */
186 * Force compression on the file using the defrag ioctl, could be
187 * different from prop_compress and takes precedence if set
189 unsigned defrag_compress;
191 struct btrfs_delayed_node *delayed_node;
193 /* File creation time. */
194 struct timespec i_otime;
196 /* Hook into fs_info->delayed_iputs */
197 struct list_head delayed_iput;
198 long delayed_iput_count;
201 * To avoid races between lockless (i_mutex not held) direct IO writes
202 * and concurrent fsync requests. Direct IO writes must acquire read
203 * access on this semaphore for creating an extent map and its
204 * corresponding ordered extent. The fast fsync path must acquire write
205 * access on this semaphore before it collects ordered extents and
208 struct rw_semaphore dio_sem;
210 struct inode vfs_inode;
213 extern unsigned char btrfs_filetype_table[];
215 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
217 return container_of(inode, struct btrfs_inode, vfs_inode);
220 static inline unsigned long btrfs_inode_hash(u64 objectid,
221 const struct btrfs_root *root)
223 u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
225 #if BITS_PER_LONG == 32
226 h = (h >> 32) ^ (h & 0xffffffff);
229 return (unsigned long)h;
232 static inline void btrfs_insert_inode_hash(struct inode *inode)
234 unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
236 __insert_inode_hash(inode, h);
239 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
241 u64 ino = inode->location.objectid;
245 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
247 if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
248 ino = inode->vfs_inode.i_ino;
252 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
254 i_size_write(&inode->vfs_inode, size);
255 inode->disk_i_size = size;
258 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
260 struct btrfs_root *root = inode->root;
262 if (root == root->fs_info->tree_root &&
263 btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
265 if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID)
270 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
274 spin_lock(&inode->lock);
275 if (inode->logged_trans == generation &&
276 inode->last_sub_trans <= inode->last_log_commit &&
277 inode->last_sub_trans <= inode->root->last_log_commit) {
279 * After a ranged fsync we might have left some extent maps
280 * (that fall outside the fsync's range). So return false
281 * here if the list isn't empty, to make sure btrfs_log_inode()
282 * will be called and process those extent maps.
285 if (list_empty(&inode->extent_tree.modified_extents))
288 spin_unlock(&inode->lock);
292 #define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1
294 struct btrfs_dio_private {
302 /* number of bios pending for this dio */
303 atomic_t pending_bios;
308 /* orig_bio is our btrfs_io_bio */
309 struct bio *orig_bio;
311 /* dio_bio came from fs/direct-io.c */
315 * The original bio may be split to several sub-bios, this is
316 * done during endio of sub-bios
318 blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *,
323 * Disable DIO read nolock optimization, so new dio readers will be forced
324 * to grab i_mutex. It is used to avoid the endless truncate due to
325 * nonlocked dio read.
327 static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode)
329 set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
333 static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode)
335 smp_mb__before_atomic();
336 clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
339 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
340 u64 logical_start, u32 csum, u32 csum_expected, int mirror_num)
342 struct btrfs_root *root = inode->root;
344 /* Output minus objectid, which is more meaningful */
345 if (root->objectid >= BTRFS_LAST_FREE_OBJECTID)
346 btrfs_warn_rl(root->fs_info,
347 "csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d",
348 root->objectid, btrfs_ino(inode),
349 logical_start, csum, csum_expected, mirror_num);
351 btrfs_warn_rl(root->fs_info,
352 "csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d",
353 root->objectid, btrfs_ino(inode),
354 logical_start, csum, csum_expected, mirror_num);
357 bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);