Loading...
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
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.
7 *
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.
12 *
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.
17 */
18
19#ifndef __BTRFS_I__
20#define __BTRFS_I__
21
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"
27
28/*
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.
34 */
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
47
48/* in memory btrfs inode */
49struct btrfs_inode {
50 /* which subvolume this inode belongs to */
51 struct btrfs_root *root;
52
53 /* key used to find this inode on disk. This is used by the code
54 * to read in roots of subvolumes
55 */
56 struct btrfs_key location;
57
58 /*
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,
61 * logged_trans).
62 */
63 spinlock_t lock;
64
65 /* the extent_tree has caches of all the extent mappings to disk */
66 struct extent_map_tree extent_tree;
67
68 /* the io_tree does range state (DIRTY, LOCKED etc) */
69 struct extent_io_tree io_tree;
70
71 /* special utility tree used to record which mirrors have already been
72 * tried when checksums fail for a given block
73 */
74 struct extent_io_tree io_failure_tree;
75
76 /* held while logging the inode in tree-log.c */
77 struct mutex log_mutex;
78
79 /* held while doing delalloc reservations */
80 struct mutex delalloc_mutex;
81
82 /* used to order data wrt metadata */
83 struct btrfs_ordered_inode_tree ordered_tree;
84
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
87 * to walk them all.
88 */
89 struct list_head delalloc_inodes;
90
91 /* node for the red-black tree that links inodes in subvolume root */
92 struct rb_node rb_node;
93
94 unsigned long runtime_flags;
95
96 /* Keep track of who's O_SYNC/fsyncing currently */
97 atomic_t sync_writers;
98
99 /* full 64 bit generation number, struct vfs_inode doesn't have a big
100 * enough field for this.
101 */
102 u64 generation;
103
104 /*
105 * transid of the trans_handle that last modified this inode
106 */
107 u64 last_trans;
108
109 /*
110 * transid that last logged this inode
111 */
112 u64 logged_trans;
113
114 /*
115 * log transid when this inode was last modified
116 */
117 int last_sub_trans;
118
119 /* a local copy of root's last_log_commit */
120 int last_log_commit;
121
122 /* total number of bytes pending delalloc, used by stat to calc the
123 * real block usage of the file
124 */
125 u64 delalloc_bytes;
126
127 /*
128 * total number of bytes pending defrag, used by stat to check whether
129 * it needs COW.
130 */
131 u64 defrag_bytes;
132
133 /*
134 * the size of the file stored in the metadata on disk. data=ordered
135 * means the in-memory i_size might be larger than the size on disk
136 * because not all the blocks are written yet.
137 */
138 u64 disk_i_size;
139
140 /*
141 * if this is a directory then index_cnt is the counter for the index
142 * number for new files that are created
143 */
144 u64 index_cnt;
145
146 /* Cache the directory index number to speed the dir/file remove */
147 u64 dir_index;
148
149 /* the fsync log has some corner cases that mean we have to check
150 * directories to see if any unlinks have been done before
151 * the directory was logged. See tree-log.c for all the
152 * details
153 */
154 u64 last_unlink_trans;
155
156 /*
157 * Number of bytes outstanding that are going to need csums. This is
158 * used in ENOSPC accounting.
159 */
160 u64 csum_bytes;
161
162 /* flags field from the on disk inode */
163 u32 flags;
164
165 /*
166 * Counters to keep track of the number of extent item's we may use due
167 * to delalloc and such. outstanding_extents is the number of extent
168 * items we think we'll end up using, and reserved_extents is the number
169 * of extent items we've reserved metadata for.
170 */
171 unsigned outstanding_extents;
172 unsigned reserved_extents;
173
174 /*
175 * always compress this one file
176 */
177 unsigned force_compress;
178
179 struct btrfs_delayed_node *delayed_node;
180
181 /* File creation time. */
182 struct timespec i_otime;
183
184 /* Hook into fs_info->delayed_iputs */
185 struct list_head delayed_iput;
186 long delayed_iput_count;
187
188 /*
189 * To avoid races between lockless (i_mutex not held) direct IO writes
190 * and concurrent fsync requests. Direct IO writes must acquire read
191 * access on this semaphore for creating an extent map and its
192 * corresponding ordered extent. The fast fsync path must acquire write
193 * access on this semaphore before it collects ordered extents and
194 * extent maps.
195 */
196 struct rw_semaphore dio_sem;
197
198 struct inode vfs_inode;
199};
200
201extern unsigned char btrfs_filetype_table[];
202
203static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
204{
205 return container_of(inode, struct btrfs_inode, vfs_inode);
206}
207
208static inline unsigned long btrfs_inode_hash(u64 objectid,
209 const struct btrfs_root *root)
210{
211 u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
212
213#if BITS_PER_LONG == 32
214 h = (h >> 32) ^ (h & 0xffffffff);
215#endif
216
217 return (unsigned long)h;
218}
219
220static inline void btrfs_insert_inode_hash(struct inode *inode)
221{
222 unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
223
224 __insert_inode_hash(inode, h);
225}
226
227static inline u64 btrfs_ino(struct inode *inode)
228{
229 u64 ino = BTRFS_I(inode)->location.objectid;
230
231 /*
232 * !ino: btree_inode
233 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
234 */
235 if (!ino || BTRFS_I(inode)->location.type == BTRFS_ROOT_ITEM_KEY)
236 ino = inode->i_ino;
237 return ino;
238}
239
240static inline void btrfs_i_size_write(struct inode *inode, u64 size)
241{
242 i_size_write(inode, size);
243 BTRFS_I(inode)->disk_i_size = size;
244}
245
246static inline bool btrfs_is_free_space_inode(struct inode *inode)
247{
248 struct btrfs_root *root = BTRFS_I(inode)->root;
249
250 if (root == root->fs_info->tree_root &&
251 btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
252 return true;
253 if (BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID)
254 return true;
255 return false;
256}
257
258static inline int btrfs_inode_in_log(struct inode *inode, u64 generation)
259{
260 int ret = 0;
261
262 spin_lock(&BTRFS_I(inode)->lock);
263 if (BTRFS_I(inode)->logged_trans == generation &&
264 BTRFS_I(inode)->last_sub_trans <=
265 BTRFS_I(inode)->last_log_commit &&
266 BTRFS_I(inode)->last_sub_trans <=
267 BTRFS_I(inode)->root->last_log_commit) {
268 /*
269 * After a ranged fsync we might have left some extent maps
270 * (that fall outside the fsync's range). So return false
271 * here if the list isn't empty, to make sure btrfs_log_inode()
272 * will be called and process those extent maps.
273 */
274 smp_mb();
275 if (list_empty(&BTRFS_I(inode)->extent_tree.modified_extents))
276 ret = 1;
277 }
278 spin_unlock(&BTRFS_I(inode)->lock);
279 return ret;
280}
281
282#define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1
283
284struct btrfs_dio_private {
285 struct inode *inode;
286 unsigned long flags;
287 u64 logical_offset;
288 u64 disk_bytenr;
289 u64 bytes;
290 void *private;
291
292 /* number of bios pending for this dio */
293 atomic_t pending_bios;
294
295 /* IO errors */
296 int errors;
297
298 /* orig_bio is our btrfs_io_bio */
299 struct bio *orig_bio;
300
301 /* dio_bio came from fs/direct-io.c */
302 struct bio *dio_bio;
303
304 /*
305 * The original bio may be split to several sub-bios, this is
306 * done during endio of sub-bios
307 */
308 int (*subio_endio)(struct inode *, struct btrfs_io_bio *, int);
309};
310
311/*
312 * Disable DIO read nolock optimization, so new dio readers will be forced
313 * to grab i_mutex. It is used to avoid the endless truncate due to
314 * nonlocked dio read.
315 */
316static inline void btrfs_inode_block_unlocked_dio(struct inode *inode)
317{
318 set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &BTRFS_I(inode)->runtime_flags);
319 smp_mb();
320}
321
322static inline void btrfs_inode_resume_unlocked_dio(struct inode *inode)
323{
324 smp_mb__before_atomic();
325 clear_bit(BTRFS_INODE_READDIO_NEED_LOCK,
326 &BTRFS_I(inode)->runtime_flags);
327}
328
329bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);
330
331#endif
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
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.
7 *
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.
12 *
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.
17 */
18
19#ifndef __BTRFS_I__
20#define __BTRFS_I__
21
22#include "extent_map.h"
23#include "extent_io.h"
24#include "ordered-data.h"
25#include "delayed-inode.h"
26
27/* in memory btrfs inode */
28struct btrfs_inode {
29 /* which subvolume this inode belongs to */
30 struct btrfs_root *root;
31
32 /* key used to find this inode on disk. This is used by the code
33 * to read in roots of subvolumes
34 */
35 struct btrfs_key location;
36
37 /* Lock for counters */
38 spinlock_t lock;
39
40 /* the extent_tree has caches of all the extent mappings to disk */
41 struct extent_map_tree extent_tree;
42
43 /* the io_tree does range state (DIRTY, LOCKED etc) */
44 struct extent_io_tree io_tree;
45
46 /* special utility tree used to record which mirrors have already been
47 * tried when checksums fail for a given block
48 */
49 struct extent_io_tree io_failure_tree;
50
51 /* held while logging the inode in tree-log.c */
52 struct mutex log_mutex;
53
54 /* used to order data wrt metadata */
55 struct btrfs_ordered_inode_tree ordered_tree;
56
57 /* for keeping track of orphaned inodes */
58 struct list_head i_orphan;
59
60 /* list of all the delalloc inodes in the FS. There are times we need
61 * to write all the delalloc pages to disk, and this list is used
62 * to walk them all.
63 */
64 struct list_head delalloc_inodes;
65
66 /*
67 * list for tracking inodes that must be sent to disk before a
68 * rename or truncate commit
69 */
70 struct list_head ordered_operations;
71
72 /* node for the red-black tree that links inodes in subvolume root */
73 struct rb_node rb_node;
74
75 /* the space_info for where this inode's data allocations are done */
76 struct btrfs_space_info *space_info;
77
78 /* full 64 bit generation number, struct vfs_inode doesn't have a big
79 * enough field for this.
80 */
81 u64 generation;
82
83 /* sequence number for NFS changes */
84 u64 sequence;
85
86 /*
87 * transid of the trans_handle that last modified this inode
88 */
89 u64 last_trans;
90
91 /*
92 * log transid when this inode was last modified
93 */
94 u64 last_sub_trans;
95
96 /*
97 * transid that last logged this inode
98 */
99 u64 logged_trans;
100
101 /* total number of bytes pending delalloc, used by stat to calc the
102 * real block usage of the file
103 */
104 u64 delalloc_bytes;
105
106 /* total number of bytes that may be used for this inode for
107 * delalloc
108 */
109 u64 reserved_bytes;
110
111 /*
112 * the size of the file stored in the metadata on disk. data=ordered
113 * means the in-memory i_size might be larger than the size on disk
114 * because not all the blocks are written yet.
115 */
116 u64 disk_i_size;
117
118 /* flags field from the on disk inode */
119 u32 flags;
120
121 /*
122 * if this is a directory then index_cnt is the counter for the index
123 * number for new files that are created
124 */
125 u64 index_cnt;
126
127 /* the fsync log has some corner cases that mean we have to check
128 * directories to see if any unlinks have been done before
129 * the directory was logged. See tree-log.c for all the
130 * details
131 */
132 u64 last_unlink_trans;
133
134 /*
135 * Counters to keep track of the number of extent item's we may use due
136 * to delalloc and such. outstanding_extents is the number of extent
137 * items we think we'll end up using, and reserved_extents is the number
138 * of extent items we've reserved metadata for.
139 */
140 unsigned outstanding_extents;
141 unsigned reserved_extents;
142
143 /*
144 * ordered_data_close is set by truncate when a file that used
145 * to have good data has been truncated to zero. When it is set
146 * the btrfs file release call will add this inode to the
147 * ordered operations list so that we make sure to flush out any
148 * new data the application may have written before commit.
149 *
150 * yes, its silly to have a single bitflag, but we might grow more
151 * of these.
152 */
153 unsigned ordered_data_close:1;
154 unsigned orphan_meta_reserved:1;
155 unsigned dummy_inode:1;
156 unsigned in_defrag:1;
157
158 /*
159 * always compress this one file
160 */
161 unsigned force_compress:4;
162
163 struct btrfs_delayed_node *delayed_node;
164
165 struct inode vfs_inode;
166};
167
168extern unsigned char btrfs_filetype_table[];
169
170static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
171{
172 return container_of(inode, struct btrfs_inode, vfs_inode);
173}
174
175static inline u64 btrfs_ino(struct inode *inode)
176{
177 u64 ino = BTRFS_I(inode)->location.objectid;
178
179 /*
180 * !ino: btree_inode
181 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
182 */
183 if (!ino || BTRFS_I(inode)->location.type == BTRFS_ROOT_ITEM_KEY)
184 ino = inode->i_ino;
185 return ino;
186}
187
188static inline void btrfs_i_size_write(struct inode *inode, u64 size)
189{
190 i_size_write(inode, size);
191 BTRFS_I(inode)->disk_i_size = size;
192}
193
194static inline bool btrfs_is_free_space_inode(struct btrfs_root *root,
195 struct inode *inode)
196{
197 if (root == root->fs_info->tree_root ||
198 BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID)
199 return true;
200 return false;
201}
202
203#endif