• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 	/*
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 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.
131 	 */
132 	u64 new_delalloc_bytes;
133 
134 	/*
135 	 * total number of bytes pending defrag, used by stat to check whether
136 	 * it needs COW.
137 	 */
138 	u64 defrag_bytes;
139 
140 	/*
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.
144 	 */
145 	u64 disk_i_size;
146 
147 	/*
148 	 * if this is a directory then index_cnt is the counter for the index
149 	 * number for new files that are created
150 	 */
151 	u64 index_cnt;
152 
153 	/* Cache the directory index number to speed the dir/file remove */
154 	u64 dir_index;
155 
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
159 	 * details
160 	 */
161 	u64 last_unlink_trans;
162 
163 	/*
164 	 * Track the transaction id of the last transaction used to create a
165 	 * hard link for the inode. This is used by the log tree (fsync).
166 	 */
167 	u64 last_link_trans;
168 
169 	/*
170 	 * Number of bytes outstanding that are going to need csums.  This is
171 	 * used in ENOSPC accounting.
172 	 */
173 	u64 csum_bytes;
174 
175 	/* flags field from the on disk inode */
176 	u32 flags;
177 
178 	/*
179 	 * Counters to keep track of the number of extent item's we may use due
180 	 * to delalloc and such.  outstanding_extents is the number of extent
181 	 * items we think we'll end up using, and reserved_extents is the number
182 	 * of extent items we've reserved metadata for.
183 	 */
184 	unsigned outstanding_extents;
185 	unsigned reserved_extents;
186 
187 	/*
188 	 * Cached values of inode properties
189 	 */
190 	unsigned prop_compress;		/* per-file compression algorithm */
191 	/*
192 	 * Force compression on the file using the defrag ioctl, could be
193 	 * different from prop_compress and takes precedence if set
194 	 */
195 	unsigned defrag_compress;
196 
197 	struct btrfs_delayed_node *delayed_node;
198 
199 	/* File creation time. */
200 	struct timespec i_otime;
201 
202 	/* Hook into fs_info->delayed_iputs */
203 	struct list_head delayed_iput;
204 	long delayed_iput_count;
205 
206 	/*
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
212 	 * extent maps.
213 	 */
214 	struct rw_semaphore dio_sem;
215 
216 	struct inode vfs_inode;
217 };
218 
219 extern unsigned char btrfs_filetype_table[];
220 
BTRFS_I(const struct inode * inode)221 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
222 {
223 	return container_of(inode, struct btrfs_inode, vfs_inode);
224 }
225 
btrfs_inode_hash(u64 objectid,const struct btrfs_root * root)226 static inline unsigned long btrfs_inode_hash(u64 objectid,
227 					     const struct btrfs_root *root)
228 {
229 	u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
230 
231 #if BITS_PER_LONG == 32
232 	h = (h >> 32) ^ (h & 0xffffffff);
233 #endif
234 
235 	return (unsigned long)h;
236 }
237 
btrfs_insert_inode_hash(struct inode * inode)238 static inline void btrfs_insert_inode_hash(struct inode *inode)
239 {
240 	unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
241 
242 	__insert_inode_hash(inode, h);
243 }
244 
btrfs_ino(const struct btrfs_inode * inode)245 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
246 {
247 	u64 ino = inode->location.objectid;
248 
249 	/*
250 	 * !ino: btree_inode
251 	 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
252 	 */
253 	if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
254 		ino = inode->vfs_inode.i_ino;
255 	return ino;
256 }
257 
btrfs_i_size_write(struct btrfs_inode * inode,u64 size)258 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
259 {
260 	i_size_write(&inode->vfs_inode, size);
261 	inode->disk_i_size = size;
262 }
263 
btrfs_is_free_space_inode(struct btrfs_inode * inode)264 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
265 {
266 	struct btrfs_root *root = inode->root;
267 
268 	if (root == root->fs_info->tree_root &&
269 	    btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
270 		return true;
271 	if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID)
272 		return true;
273 	return false;
274 }
275 
btrfs_inode_in_log(struct btrfs_inode * inode,u64 generation)276 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
277 {
278 	int ret = 0;
279 
280 	spin_lock(&inode->lock);
281 	if (inode->logged_trans == generation &&
282 	    inode->last_sub_trans <= inode->last_log_commit &&
283 	    inode->last_sub_trans <= inode->root->last_log_commit) {
284 		/*
285 		 * After a ranged fsync we might have left some extent maps
286 		 * (that fall outside the fsync's range). So return false
287 		 * here if the list isn't empty, to make sure btrfs_log_inode()
288 		 * will be called and process those extent maps.
289 		 */
290 		smp_mb();
291 		if (list_empty(&inode->extent_tree.modified_extents))
292 			ret = 1;
293 	}
294 	spin_unlock(&inode->lock);
295 	return ret;
296 }
297 
298 #define BTRFS_DIO_ORIG_BIO_SUBMITTED	0x1
299 
300 struct btrfs_dio_private {
301 	struct inode *inode;
302 	unsigned long flags;
303 	u64 logical_offset;
304 	u64 disk_bytenr;
305 	u64 bytes;
306 	void *private;
307 
308 	/* number of bios pending for this dio */
309 	atomic_t pending_bios;
310 
311 	/* IO errors */
312 	int errors;
313 
314 	/* orig_bio is our btrfs_io_bio */
315 	struct bio *orig_bio;
316 
317 	/* dio_bio came from fs/direct-io.c */
318 	struct bio *dio_bio;
319 
320 	/*
321 	 * The original bio may be split to several sub-bios, this is
322 	 * done during endio of sub-bios
323 	 */
324 	blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *,
325 			blk_status_t);
326 };
327 
328 /*
329  * Disable DIO read nolock optimization, so new dio readers will be forced
330  * to grab i_mutex. It is used to avoid the endless truncate due to
331  * nonlocked dio read.
332  */
btrfs_inode_block_unlocked_dio(struct btrfs_inode * inode)333 static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode)
334 {
335 	set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
336 	smp_mb();
337 }
338 
btrfs_inode_resume_unlocked_dio(struct btrfs_inode * inode)339 static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode)
340 {
341 	smp_mb__before_atomic();
342 	clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
343 }
344 
btrfs_print_data_csum_error(struct btrfs_inode * inode,u64 logical_start,u32 csum,u32 csum_expected,int mirror_num)345 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
346 		u64 logical_start, u32 csum, u32 csum_expected, int mirror_num)
347 {
348 	struct btrfs_root *root = inode->root;
349 
350 	/* Output minus objectid, which is more meaningful */
351 	if (root->objectid >= BTRFS_LAST_FREE_OBJECTID)
352 		btrfs_warn_rl(root->fs_info,
353 	"csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d",
354 			root->objectid, btrfs_ino(inode),
355 			logical_start, csum, csum_expected, mirror_num);
356 	else
357 		btrfs_warn_rl(root->fs_info,
358 	"csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d",
359 			root->objectid, btrfs_ino(inode),
360 			logical_start, csum, csum_expected, mirror_num);
361 }
362 
363 bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);
364 
365 #endif
366