xref: /linux/fs/btrfs/btrfs_inode.h (revision 3932b9ca55b0be314a36d3e84faff3e823c081f5)
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 */
49 struct 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 	/* Lock for counters */
59 	spinlock_t lock;
60 
61 	/* the extent_tree has caches of all the extent mappings to disk */
62 	struct extent_map_tree extent_tree;
63 
64 	/* the io_tree does range state (DIRTY, LOCKED etc) */
65 	struct extent_io_tree io_tree;
66 
67 	/* special utility tree used to record which mirrors have already been
68 	 * tried when checksums fail for a given block
69 	 */
70 	struct extent_io_tree io_failure_tree;
71 
72 	/* held while logging the inode in tree-log.c */
73 	struct mutex log_mutex;
74 
75 	/* held while doing delalloc reservations */
76 	struct mutex delalloc_mutex;
77 
78 	/* used to order data wrt metadata */
79 	struct btrfs_ordered_inode_tree ordered_tree;
80 
81 	/* list of all the delalloc inodes in the FS.  There are times we need
82 	 * to write all the delalloc pages to disk, and this list is used
83 	 * to walk them all.
84 	 */
85 	struct list_head delalloc_inodes;
86 
87 	/* node for the red-black tree that links inodes in subvolume root */
88 	struct rb_node rb_node;
89 
90 	unsigned long runtime_flags;
91 
92 	/* Keep track of who's O_SYNC/fsyncing currently */
93 	atomic_t sync_writers;
94 
95 	/* full 64 bit generation number, struct vfs_inode doesn't have a big
96 	 * enough field for this.
97 	 */
98 	u64 generation;
99 
100 	/*
101 	 * transid of the trans_handle that last modified this inode
102 	 */
103 	u64 last_trans;
104 
105 	/*
106 	 * transid that last logged this inode
107 	 */
108 	u64 logged_trans;
109 
110 	/*
111 	 * log transid when this inode was last modified
112 	 */
113 	int last_sub_trans;
114 
115 	/* a local copy of root's last_log_commit */
116 	int last_log_commit;
117 
118 	/* total number of bytes pending delalloc, used by stat to calc the
119 	 * real block usage of the file
120 	 */
121 	u64 delalloc_bytes;
122 
123 	/*
124 	 * the size of the file stored in the metadata on disk.  data=ordered
125 	 * means the in-memory i_size might be larger than the size on disk
126 	 * because not all the blocks are written yet.
127 	 */
128 	u64 disk_i_size;
129 
130 	/*
131 	 * if this is a directory then index_cnt is the counter for the index
132 	 * number for new files that are created
133 	 */
134 	u64 index_cnt;
135 
136 	/* Cache the directory index number to speed the dir/file remove */
137 	u64 dir_index;
138 
139 	/* the fsync log has some corner cases that mean we have to check
140 	 * directories to see if any unlinks have been done before
141 	 * the directory was logged.  See tree-log.c for all the
142 	 * details
143 	 */
144 	u64 last_unlink_trans;
145 
146 	/*
147 	 * Number of bytes outstanding that are going to need csums.  This is
148 	 * used in ENOSPC accounting.
149 	 */
150 	u64 csum_bytes;
151 
152 	/* flags field from the on disk inode */
153 	u32 flags;
154 
155 	/*
156 	 * Counters to keep track of the number of extent item's we may use due
157 	 * to delalloc and such.  outstanding_extents is the number of extent
158 	 * items we think we'll end up using, and reserved_extents is the number
159 	 * of extent items we've reserved metadata for.
160 	 */
161 	unsigned outstanding_extents;
162 	unsigned reserved_extents;
163 
164 	/*
165 	 * always compress this one file
166 	 */
167 	unsigned force_compress;
168 
169 	struct btrfs_delayed_node *delayed_node;
170 
171 	struct inode vfs_inode;
172 };
173 
174 extern unsigned char btrfs_filetype_table[];
175 
176 static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
177 {
178 	return container_of(inode, struct btrfs_inode, vfs_inode);
179 }
180 
181 static inline unsigned long btrfs_inode_hash(u64 objectid,
182 					     const struct btrfs_root *root)
183 {
184 	u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
185 
186 #if BITS_PER_LONG == 32
187 	h = (h >> 32) ^ (h & 0xffffffff);
188 #endif
189 
190 	return (unsigned long)h;
191 }
192 
193 static inline void btrfs_insert_inode_hash(struct inode *inode)
194 {
195 	unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
196 
197 	__insert_inode_hash(inode, h);
198 }
199 
200 static inline u64 btrfs_ino(struct inode *inode)
201 {
202 	u64 ino = BTRFS_I(inode)->location.objectid;
203 
204 	/*
205 	 * !ino: btree_inode
206 	 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
207 	 */
208 	if (!ino || BTRFS_I(inode)->location.type == BTRFS_ROOT_ITEM_KEY)
209 		ino = inode->i_ino;
210 	return ino;
211 }
212 
213 static inline void btrfs_i_size_write(struct inode *inode, u64 size)
214 {
215 	i_size_write(inode, size);
216 	BTRFS_I(inode)->disk_i_size = size;
217 }
218 
219 static inline bool btrfs_is_free_space_inode(struct inode *inode)
220 {
221 	struct btrfs_root *root = BTRFS_I(inode)->root;
222 
223 	if (root == root->fs_info->tree_root &&
224 	    btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
225 		return true;
226 	if (BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID)
227 		return true;
228 	return false;
229 }
230 
231 static inline int btrfs_inode_in_log(struct inode *inode, u64 generation)
232 {
233 	if (BTRFS_I(inode)->logged_trans == generation &&
234 	    BTRFS_I(inode)->last_sub_trans <=
235 	    BTRFS_I(inode)->last_log_commit &&
236 	    BTRFS_I(inode)->last_sub_trans <=
237 	    BTRFS_I(inode)->root->last_log_commit) {
238 		/*
239 		 * After a ranged fsync we might have left some extent maps
240 		 * (that fall outside the fsync's range). So return false
241 		 * here if the list isn't empty, to make sure btrfs_log_inode()
242 		 * will be called and process those extent maps.
243 		 */
244 		smp_mb();
245 		if (list_empty(&BTRFS_I(inode)->extent_tree.modified_extents))
246 			return 1;
247 	}
248 	return 0;
249 }
250 
251 struct btrfs_dio_private {
252 	struct inode *inode;
253 	u64 logical_offset;
254 	u64 disk_bytenr;
255 	u64 bytes;
256 	void *private;
257 
258 	/* number of bios pending for this dio */
259 	atomic_t pending_bios;
260 
261 	/* IO errors */
262 	int errors;
263 
264 	/* orig_bio is our btrfs_io_bio */
265 	struct bio *orig_bio;
266 
267 	/* dio_bio came from fs/direct-io.c */
268 	struct bio *dio_bio;
269 	u8 csum[0];
270 };
271 
272 /*
273  * Disable DIO read nolock optimization, so new dio readers will be forced
274  * to grab i_mutex. It is used to avoid the endless truncate due to
275  * nonlocked dio read.
276  */
277 static inline void btrfs_inode_block_unlocked_dio(struct inode *inode)
278 {
279 	set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &BTRFS_I(inode)->runtime_flags);
280 	smp_mb();
281 }
282 
283 static inline void btrfs_inode_resume_unlocked_dio(struct inode *inode)
284 {
285 	smp_mb__before_atomic();
286 	clear_bit(BTRFS_INODE_READDIO_NEED_LOCK,
287 		  &BTRFS_I(inode)->runtime_flags);
288 }
289 
290 bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);
291 
292 #endif
293