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 <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 * The following 3 bits are meant only for the btree inode.
 49 * When any of them is set, it means an error happened while writing an
 50 * extent buffer belonging to:
 51 * 1) a non-log btree
 52 * 2) a log btree and first log sub-transaction
 53 * 3) a log btree and second log sub-transaction
 54 */
 55#define BTRFS_INODE_BTREE_ERR		        12
 56#define BTRFS_INODE_BTREE_LOG1_ERR		13
 57#define BTRFS_INODE_BTREE_LOG2_ERR		14
 58
 59/* in memory btrfs inode */
 60struct btrfs_inode {
 61	/* which subvolume this inode belongs to */
 62	struct btrfs_root *root;
 63
 64	/* key used to find this inode on disk.  This is used by the code
 65	 * to read in roots of subvolumes
 66	 */
 67	struct btrfs_key location;
 68
 69	/*
 70	 * Lock for counters and all fields used to determine if the inode is in
 71	 * the log or not (last_trans, last_sub_trans, last_log_commit,
 72	 * logged_trans).
 73	 */
 74	spinlock_t lock;
 75
 76	/* the extent_tree has caches of all the extent mappings to disk */
 77	struct extent_map_tree extent_tree;
 78
 79	/* the io_tree does range state (DIRTY, LOCKED etc) */
 80	struct extent_io_tree io_tree;
 81
 82	/* special utility tree used to record which mirrors have already been
 83	 * tried when checksums fail for a given block
 84	 */
 85	struct extent_io_tree io_failure_tree;
 86
 87	/* held while logging the inode in tree-log.c */
 88	struct mutex log_mutex;
 89
 90	/* held while doing delalloc reservations */
 91	struct mutex delalloc_mutex;
 92
 93	/* used to order data wrt metadata */
 94	struct btrfs_ordered_inode_tree ordered_tree;
 95
 96	/* list of all the delalloc inodes in the FS.  There are times we need
 97	 * to write all the delalloc pages to disk, and this list is used
 98	 * to walk them all.
 99	 */
100	struct list_head delalloc_inodes;
101
102	/* node for the red-black tree that links inodes in subvolume root */
103	struct rb_node rb_node;
104
105	unsigned long runtime_flags;
106
107	/* Keep track of who's O_SYNC/fsyncing currently */
108	atomic_t sync_writers;
109
110	/* full 64 bit generation number, struct vfs_inode doesn't have a big
111	 * enough field for this.
112	 */
113	u64 generation;
114
115	/*
116	 * transid of the trans_handle that last modified this inode
117	 */
118	u64 last_trans;
119
120	/*
121	 * transid that last logged this inode
122	 */
123	u64 logged_trans;
124
125	/*
126	 * log transid when this inode was last modified
127	 */
128	int last_sub_trans;
129
130	/* a local copy of root's last_log_commit */
131	int last_log_commit;
132
133	/* total number of bytes pending delalloc, used by stat to calc the
134	 * real block usage of the file
135	 */
136	u64 delalloc_bytes;
137
138	/*
139	 * total number of bytes pending defrag, used by stat to check whether
140	 * it needs COW.
141	 */
142	u64 defrag_bytes;
143
144	/*
145	 * the size of the file stored in the metadata on disk.  data=ordered
146	 * means the in-memory i_size might be larger than the size on disk
147	 * because not all the blocks are written yet.
148	 */
149	u64 disk_i_size;
150
151	/*
152	 * if this is a directory then index_cnt is the counter for the index
153	 * number for new files that are created
154	 */
155	u64 index_cnt;
156
157	/* Cache the directory index number to speed the dir/file remove */
158	u64 dir_index;
159
160	/* the fsync log has some corner cases that mean we have to check
161	 * directories to see if any unlinks have been done before
162	 * the directory was logged.  See tree-log.c for all the
163	 * details
164	 */
165	u64 last_unlink_trans;
166
167	/*
168	 * Number of bytes outstanding that are going to need csums.  This is
169	 * used in ENOSPC accounting.
170	 */
171	u64 csum_bytes;
172
173	/* flags field from the on disk inode */
174	u32 flags;
175
176	/*
177	 * Counters to keep track of the number of extent item's we may use due
178	 * to delalloc and such.  outstanding_extents is the number of extent
179	 * items we think we'll end up using, and reserved_extents is the number
180	 * of extent items we've reserved metadata for.
181	 */
182	unsigned outstanding_extents;
183	unsigned reserved_extents;
184
185	/*
186	 * always compress this one file
187	 */
188	unsigned force_compress;
189
190	struct btrfs_delayed_node *delayed_node;
191
192	/* File creation time. */
193	struct timespec i_otime;
194
195	/* Hook into fs_info->delayed_iputs */
196	struct list_head delayed_iput;
197	long delayed_iput_count;
198
 
 
 
 
 
 
 
 
 
 
199	struct inode vfs_inode;
200};
201
202extern unsigned char btrfs_filetype_table[];
203
204static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
205{
206	return container_of(inode, struct btrfs_inode, vfs_inode);
207}
208
209static inline unsigned long btrfs_inode_hash(u64 objectid,
210					     const struct btrfs_root *root)
211{
212	u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
213
214#if BITS_PER_LONG == 32
215	h = (h >> 32) ^ (h & 0xffffffff);
216#endif
217
218	return (unsigned long)h;
219}
220
221static inline void btrfs_insert_inode_hash(struct inode *inode)
222{
223	unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
224
225	__insert_inode_hash(inode, h);
226}
227
228static inline u64 btrfs_ino(struct inode *inode)
229{
230	u64 ino = BTRFS_I(inode)->location.objectid;
231
232	/*
233	 * !ino: btree_inode
234	 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
235	 */
236	if (!ino || BTRFS_I(inode)->location.type == BTRFS_ROOT_ITEM_KEY)
237		ino = inode->i_ino;
238	return ino;
239}
240
241static inline void btrfs_i_size_write(struct inode *inode, u64 size)
242{
243	i_size_write(inode, size);
244	BTRFS_I(inode)->disk_i_size = size;
245}
246
247static inline bool btrfs_is_free_space_inode(struct inode *inode)
248{
249	struct btrfs_root *root = BTRFS_I(inode)->root;
250
251	if (root == root->fs_info->tree_root &&
252	    btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
253		return true;
254	if (BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID)
255		return true;
256	return false;
257}
258
259static inline int btrfs_inode_in_log(struct inode *inode, u64 generation)
260{
261	int ret = 0;
262
263	spin_lock(&BTRFS_I(inode)->lock);
264	if (BTRFS_I(inode)->logged_trans == generation &&
265	    BTRFS_I(inode)->last_sub_trans <=
266	    BTRFS_I(inode)->last_log_commit &&
267	    BTRFS_I(inode)->last_sub_trans <=
268	    BTRFS_I(inode)->root->last_log_commit) {
269		/*
270		 * After a ranged fsync we might have left some extent maps
271		 * (that fall outside the fsync's range). So return false
272		 * here if the list isn't empty, to make sure btrfs_log_inode()
273		 * will be called and process those extent maps.
274		 */
275		smp_mb();
276		if (list_empty(&BTRFS_I(inode)->extent_tree.modified_extents))
277			ret = 1;
278	}
279	spin_unlock(&BTRFS_I(inode)->lock);
280	return ret;
281}
282
283#define BTRFS_DIO_ORIG_BIO_SUBMITTED	0x1
284
285struct btrfs_dio_private {
286	struct inode *inode;
287	unsigned long flags;
288	u64 logical_offset;
289	u64 disk_bytenr;
290	u64 bytes;
291	void *private;
292
293	/* number of bios pending for this dio */
294	atomic_t pending_bios;
295
296	/* IO errors */
297	int errors;
298
299	/* orig_bio is our btrfs_io_bio */
300	struct bio *orig_bio;
301
302	/* dio_bio came from fs/direct-io.c */
303	struct bio *dio_bio;
304
305	/*
306	 * The original bio may be splited to several sub-bios, this is
307	 * done during endio of sub-bios
308	 */
309	int (*subio_endio)(struct inode *, struct btrfs_io_bio *, int);
310};
311
312/*
313 * Disable DIO read nolock optimization, so new dio readers will be forced
314 * to grab i_mutex. It is used to avoid the endless truncate due to
315 * nonlocked dio read.
316 */
317static inline void btrfs_inode_block_unlocked_dio(struct inode *inode)
318{
319	set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &BTRFS_I(inode)->runtime_flags);
320	smp_mb();
321}
322
323static inline void btrfs_inode_resume_unlocked_dio(struct inode *inode)
324{
325	smp_mb__before_atomic();
326	clear_bit(BTRFS_INODE_READDIO_NEED_LOCK,
327		  &BTRFS_I(inode)->runtime_flags);
328}
329
330bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);
331
332#endif