blob: 61205e3bbefac237afadf36f961e32ce5e8e1661 [file] [log] [blame]
Chris Mason6cbd5572007-06-12 09:07:21 -04001/*
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
Chris Mason2c90e5d2007-04-02 10:50:19 -040019#ifndef __BTRFS_I__
20#define __BTRFS_I__
21
Filipe David Borba Manana778ba822013-10-06 22:22:33 +010022#include <linux/hash.h>
Chris Masona52d9a82007-08-27 16:49:44 -040023#include "extent_map.h"
Chris Masond1310b22008-01-24 16:13:08 -050024#include "extent_io.h"
Chris Masone6dcd2d2008-07-17 12:53:50 -040025#include "ordered-data.h"
Miao Xie16cdcec2011-04-22 18:12:22 +080026#include "delayed-inode.h"
Chris Masona52d9a82007-08-27 16:49:44 -040027
Josef Bacik72ac3c02012-05-23 14:13:11 -040028/*
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
Josef Bacik8a35d952012-05-23 14:26:42 -040040#define BTRFS_INODE_HAS_ORPHAN_ITEM 5
Josef Bacik7ddf5a42012-06-08 15:26:47 -040041#define BTRFS_INODE_HAS_ASYNC_EXTENT 6
Josef Bacik5dc562c2012-08-17 13:14:17 -040042#define BTRFS_INODE_NEEDS_FULL_SYNC 7
Josef Bacike9976152012-10-11 15:53:56 -040043#define BTRFS_INODE_COPY_EVERYTHING 8
Miao Xiedf0af1a2013-01-29 10:11:59 +000044#define BTRFS_INODE_IN_DELALLOC_LIST 9
Miao Xie2e60a512013-02-08 07:01:08 +000045#define BTRFS_INODE_READDIO_NEED_LOCK 10
Filipe David Borba Manana63541922014-01-07 11:47:46 +000046#define BTRFS_INODE_HAS_PROPS 11
Filipe Manana656f30d2014-09-26 12:25:56 +010047/*
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
Josef Bacik72ac3c02012-05-23 14:13:11 -040058
Aneeshf1ace242007-06-13 16:18:26 -040059/* in memory btrfs inode */
Chris Mason2c90e5d2007-04-02 10:50:19 -040060struct btrfs_inode {
Chris Masond352ac62008-09-29 15:18:18 -040061 /* which subvolume this inode belongs to */
Chris Masond6e4a422007-04-06 15:37:36 -040062 struct btrfs_root *root;
Chris Masond352ac62008-09-29 15:18:18 -040063
Chris Masond352ac62008-09-29 15:18:18 -040064 /* key used to find this inode on disk. This is used by the code
65 * to read in roots of subvolumes
66 */
Chris Masond6e4a422007-04-06 15:37:36 -040067 struct btrfs_key location;
Chris Masond352ac62008-09-29 15:18:18 -040068
Filipe Manana2f2ff0e2015-03-20 17:19:46 +000069 /*
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 */
Josef Bacik9e0baf62011-07-15 15:16:44 +000074 spinlock_t lock;
75
Chris Masond352ac62008-09-29 15:18:18 -040076 /* the extent_tree has caches of all the extent mappings to disk */
Chris Masona52d9a82007-08-27 16:49:44 -040077 struct extent_map_tree extent_tree;
Chris Masond352ac62008-09-29 15:18:18 -040078
79 /* the io_tree does range state (DIRTY, LOCKED etc) */
Chris Masond1310b22008-01-24 16:13:08 -050080 struct extent_io_tree io_tree;
Chris Masond352ac62008-09-29 15:18:18 -040081
82 /* special utility tree used to record which mirrors have already been
83 * tried when checksums fail for a given block
84 */
Chris Mason7e383262008-04-09 16:28:12 -040085 struct extent_io_tree io_failure_tree;
Chris Masond352ac62008-09-29 15:18:18 -040086
Chris Masond352ac62008-09-29 15:18:18 -040087 /* held while logging the inode in tree-log.c */
Chris Masone02119d2008-09-05 16:13:11 -040088 struct mutex log_mutex;
Chris Masond352ac62008-09-29 15:18:18 -040089
Josef Bacikf2486792012-01-13 12:09:22 -050090 /* held while doing delalloc reservations */
91 struct mutex delalloc_mutex;
92
Chris Masond352ac62008-09-29 15:18:18 -040093 /* used to order data wrt metadata */
Chris Masone6dcd2d2008-07-17 12:53:50 -040094 struct btrfs_ordered_inode_tree ordered_tree;
Josef Bacik15ee9bc2007-08-10 16:22:09 -040095
Chris Masond352ac62008-09-29 15:18:18 -040096 /* 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 */
Chris Masonea8c2812008-08-04 23:17:27 -0400100 struct list_head delalloc_inodes;
101
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400102 /* node for the red-black tree that links inodes in subvolume root */
103 struct rb_node rb_node;
104
Josef Bacik72ac3c02012-05-23 14:13:11 -0400105 unsigned long runtime_flags;
106
Nathaniel Yazdani9c931c52013-04-15 00:44:02 +0000107 /* Keep track of who's O_SYNC/fsyncing currently */
Josef Bacikb812ce22012-11-16 13:56:32 -0500108 atomic_t sync_writers;
109
Chris Masond352ac62008-09-29 15:18:18 -0400110 /* full 64 bit generation number, struct vfs_inode doesn't have a big
111 * enough field for this.
112 */
Chris Masone02119d2008-09-05 16:13:11 -0400113 u64 generation;
114
Josef Bacik15ee9bc2007-08-10 16:22:09 -0400115 /*
116 * transid of the trans_handle that last modified this inode
117 */
118 u64 last_trans;
Chris Mason257c62e2009-10-13 13:21:08 -0400119
120 /*
Chris Masone02119d2008-09-05 16:13:11 -0400121 * transid that last logged this inode
122 */
123 u64 logged_trans;
Chris Mason49eb7e42008-09-11 15:53:12 -0400124
Miao Xiebb14a592014-02-20 18:08:56 +0800125 /*
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
Chris Masond352ac62008-09-29 15:18:18 -0400133 /* total number of bytes pending delalloc, used by stat to calc the
134 * real block usage of the file
135 */
Chris Mason90692182008-02-08 13:49:28 -0500136 u64 delalloc_bytes;
Chris Masond352ac62008-09-29 15:18:18 -0400137
138 /*
Wang Shilong47059d92014-07-03 18:22:07 +0800139 * total number of bytes pending defrag, used by stat to check whether
140 * it needs COW.
141 */
142 u64 defrag_bytes;
143
144 /*
Chris Masond352ac62008-09-29 15:18:18 -0400145 * 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 */
Chris Masondbe674a2008-07-17 12:54:05 -0400149 u64 disk_i_size;
Chris Masond352ac62008-09-29 15:18:18 -0400150
Josef Bacikaec74772008-07-24 12:12:38 -0400151 /*
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;
Chris Masond352ac62008-09-29 15:18:18 -0400156
Miao Xie67de1172013-12-26 13:07:06 +0800157 /* Cache the directory index number to speed the dir/file remove */
158 u64 dir_index;
159
Chris Mason12fcfd22009-03-24 10:24:20 -0400160 /* 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
Josef Bacik7709cde2011-08-04 10:25:02 -0400167 /*
168 * Number of bytes outstanding that are going to need csums. This is
169 * used in ENOSPC accounting.
170 */
171 u64 csum_bytes;
172
Josef Bacikf1bdcc02011-07-14 14:28:08 -0400173 /* flags field from the on disk inode */
174 u32 flags;
175
Chris Mason5a3f23d2009-03-31 13:27:11 -0400176 /*
Josef Bacik32c00af2009-10-08 13:34:05 -0400177 * 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.
Josef Bacik9ed74f22009-09-11 16:12:44 -0400181 */
Josef Bacik9e0baf62011-07-15 15:16:44 +0000182 unsigned outstanding_extents;
183 unsigned reserved_extents;
Josef Bacik9ed74f22009-09-11 16:12:44 -0400184
185 /*
Chris Mason1e701a32010-03-11 09:42:04 -0500186 * always compress this one file
187 */
Josef Bacik72ac3c02012-05-23 14:13:11 -0400188 unsigned force_compress;
Chris Mason1e701a32010-03-11 09:42:04 -0500189
Miao Xie16cdcec2011-04-22 18:12:22 +0800190 struct btrfs_delayed_node *delayed_node;
191
chandan r9cc97d62012-07-04 12:48:07 +0530192 /* File creation time. */
193 struct timespec i_otime;
194
David Sterba8089fe62015-11-19 14:15:51 +0100195 /* Hook into fs_info->delayed_iputs */
196 struct list_head delayed_iput;
197 long delayed_iput_count;
198
Chris Masond352ac62008-09-29 15:18:18 -0400199 struct inode vfs_inode;
Chris Mason2c90e5d2007-04-02 10:50:19 -0400200};
Chris Masondbe674a2008-07-17 12:54:05 -0400201
Miao Xie16cdcec2011-04-22 18:12:22 +0800202extern unsigned char btrfs_filetype_table[];
203
Chris Mason2c90e5d2007-04-02 10:50:19 -0400204static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
205{
206 return container_of(inode, struct btrfs_inode, vfs_inode);
207}
208
Filipe David Borba Manana778ba822013-10-06 22:22:33 +0100209static 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
Li Zefan33345d012011-04-20 10:31:50 +0800228static inline u64 btrfs_ino(struct inode *inode)
229{
230 u64 ino = BTRFS_I(inode)->location.objectid;
231
Liu Bo14c7cca2011-09-11 10:52:24 -0400232 /*
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)
Li Zefan33345d012011-04-20 10:31:50 +0800237 ino = inode->i_ino;
238 return ino;
239}
240
Chris Masondbe674a2008-07-17 12:54:05 -0400241static inline void btrfs_i_size_write(struct inode *inode, u64 size)
242{
Yan, Zhengc2167752009-11-12 09:34:21 +0000243 i_size_write(inode, size);
Chris Masondbe674a2008-07-17 12:54:05 -0400244 BTRFS_I(inode)->disk_i_size = size;
245}
246
Liu Bo83eea1f2012-07-10 05:28:39 -0600247static inline bool btrfs_is_free_space_inode(struct inode *inode)
Chris Mason2cf85722011-07-26 15:35:09 -0400248{
Liu Bo83eea1f2012-07-10 05:28:39 -0600249 struct btrfs_root *root = BTRFS_I(inode)->root;
250
Liu Bo51a8cf92012-07-10 05:28:38 -0600251 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)
Chris Mason2cf85722011-07-26 15:35:09 -0400255 return true;
256 return false;
257}
258
Josef Bacik22ee6982012-05-29 16:57:49 -0400259static inline int btrfs_inode_in_log(struct inode *inode, u64 generation)
260{
Filipe Manana2f2ff0e2015-03-20 17:19:46 +0000261 int ret = 0;
262
263 spin_lock(&BTRFS_I(inode)->lock);
Josef Bacik22ee6982012-05-29 16:57:49 -0400264 if (BTRFS_I(inode)->logged_trans == generation &&
Josef Bacika5874ce2013-09-11 09:55:42 -0400265 BTRFS_I(inode)->last_sub_trans <=
266 BTRFS_I(inode)->last_log_commit &&
267 BTRFS_I(inode)->last_sub_trans <=
Filipe Manana125c4cf92014-09-11 21:22:14 +0100268 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))
Filipe Manana2f2ff0e2015-03-20 17:19:46 +0000277 ret = 1;
Filipe Manana125c4cf92014-09-11 21:22:14 +0100278 }
Filipe Manana2f2ff0e2015-03-20 17:19:46 +0000279 spin_unlock(&BTRFS_I(inode)->lock);
280 return ret;
Josef Bacik22ee6982012-05-29 16:57:49 -0400281}
282
Miao Xiec1dc0892014-09-12 18:43:56 +0800283#define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1
284
Miao Xiefacc8a222013-07-25 19:22:34 +0800285struct btrfs_dio_private {
286 struct inode *inode;
Miao Xiec1dc0892014-09-12 18:43:56 +0800287 unsigned long flags;
Miao Xiefacc8a222013-07-25 19:22:34 +0800288 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;
Miao Xiec1dc0892014-09-12 18:43:56 +0800304
305 /*
306 * The original bio may be splited to several sub-bios, this is
307 * done during endio of sub-bios
308 */
Miao Xie8b110e32014-09-12 18:44:03 +0800309 int (*subio_endio)(struct inode *, struct btrfs_io_bio *, int);
Miao Xiefacc8a222013-07-25 19:22:34 +0800310};
311
Miao Xie2e60a512013-02-08 07:01:08 +0000312/*
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{
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100325 smp_mb__before_atomic();
Miao Xie2e60a512013-02-08 07:01:08 +0000326 clear_bit(BTRFS_INODE_READDIO_NEED_LOCK,
327 &BTRFS_I(inode)->runtime_flags);
328}
329
Alex Gartrellfc4adbf2014-05-20 13:07:56 -0700330bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);
331
Chris Mason2c90e5d2007-04-02 10:50:19 -0400332#endif