blob: ca15be569d69c2656c3c7e737b4773c4a6b9d37e [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
Josef Bacik69fe2d72017-10-19 14:15:57 -040039#define BTRFS_INODE_HAS_ORPHAN_ITEM 4
40#define BTRFS_INODE_HAS_ASYNC_EXTENT 5
41#define BTRFS_INODE_NEEDS_FULL_SYNC 6
42#define BTRFS_INODE_COPY_EVERYTHING 7
43#define BTRFS_INODE_IN_DELALLOC_LIST 8
44#define BTRFS_INODE_READDIO_NEED_LOCK 9
45#define BTRFS_INODE_HAS_PROPS 10
Josef Bacik72ac3c02012-05-23 14:13:11 -040046
Aneeshf1ace242007-06-13 16:18:26 -040047/* in memory btrfs inode */
Chris Mason2c90e5d2007-04-02 10:50:19 -040048struct btrfs_inode {
Chris Masond352ac62008-09-29 15:18:18 -040049 /* which subvolume this inode belongs to */
Chris Masond6e4a422007-04-06 15:37:36 -040050 struct btrfs_root *root;
Chris Masond352ac62008-09-29 15:18:18 -040051
Chris Masond352ac62008-09-29 15:18:18 -040052 /* key used to find this inode on disk. This is used by the code
53 * to read in roots of subvolumes
54 */
Chris Masond6e4a422007-04-06 15:37:36 -040055 struct btrfs_key location;
Chris Masond352ac62008-09-29 15:18:18 -040056
Filipe Manana2f2ff0e2015-03-20 17:19:46 +000057 /*
58 * Lock for counters and all fields used to determine if the inode is in
59 * the log or not (last_trans, last_sub_trans, last_log_commit,
60 * logged_trans).
61 */
Josef Bacik9e0baf62011-07-15 15:16:44 +000062 spinlock_t lock;
63
Chris Masond352ac62008-09-29 15:18:18 -040064 /* the extent_tree has caches of all the extent mappings to disk */
Chris Masona52d9a82007-08-27 16:49:44 -040065 struct extent_map_tree extent_tree;
Chris Masond352ac62008-09-29 15:18:18 -040066
67 /* the io_tree does range state (DIRTY, LOCKED etc) */
Chris Masond1310b22008-01-24 16:13:08 -050068 struct extent_io_tree io_tree;
Chris Masond352ac62008-09-29 15:18:18 -040069
70 /* special utility tree used to record which mirrors have already been
71 * tried when checksums fail for a given block
72 */
Chris Mason7e383262008-04-09 16:28:12 -040073 struct extent_io_tree io_failure_tree;
Chris Masond352ac62008-09-29 15:18:18 -040074
Chris Masond352ac62008-09-29 15:18:18 -040075 /* held while logging the inode in tree-log.c */
Chris Masone02119d2008-09-05 16:13:11 -040076 struct mutex log_mutex;
Chris Masond352ac62008-09-29 15:18:18 -040077
Josef Bacikf2486792012-01-13 12:09:22 -050078 /* held while doing delalloc reservations */
79 struct mutex delalloc_mutex;
80
Chris Masond352ac62008-09-29 15:18:18 -040081 /* used to order data wrt metadata */
Chris Masone6dcd2d2008-07-17 12:53:50 -040082 struct btrfs_ordered_inode_tree ordered_tree;
Josef Bacik15ee9bc2007-08-10 16:22:09 -040083
Chris Masond352ac62008-09-29 15:18:18 -040084 /* list of all the delalloc inodes in the FS. There are times we need
85 * to write all the delalloc pages to disk, and this list is used
86 * to walk them all.
87 */
Chris Masonea8c2812008-08-04 23:17:27 -040088 struct list_head delalloc_inodes;
89
Yan Zheng5d4f98a2009-06-10 10:45:14 -040090 /* node for the red-black tree that links inodes in subvolume root */
91 struct rb_node rb_node;
92
Josef Bacik72ac3c02012-05-23 14:13:11 -040093 unsigned long runtime_flags;
94
Nathaniel Yazdani9c931c52013-04-15 00:44:02 +000095 /* Keep track of who's O_SYNC/fsyncing currently */
Josef Bacikb812ce22012-11-16 13:56:32 -050096 atomic_t sync_writers;
97
Chris Masond352ac62008-09-29 15:18:18 -040098 /* full 64 bit generation number, struct vfs_inode doesn't have a big
99 * enough field for this.
100 */
Chris Masone02119d2008-09-05 16:13:11 -0400101 u64 generation;
102
Josef Bacik15ee9bc2007-08-10 16:22:09 -0400103 /*
104 * transid of the trans_handle that last modified this inode
105 */
106 u64 last_trans;
Chris Mason257c62e2009-10-13 13:21:08 -0400107
108 /*
Chris Masone02119d2008-09-05 16:13:11 -0400109 * transid that last logged this inode
110 */
111 u64 logged_trans;
Chris Mason49eb7e42008-09-11 15:53:12 -0400112
Miao Xiebb14a592014-02-20 18:08:56 +0800113 /*
114 * log transid when this inode was last modified
115 */
116 int last_sub_trans;
117
118 /* a local copy of root's last_log_commit */
119 int last_log_commit;
120
Chris Masond352ac62008-09-29 15:18:18 -0400121 /* total number of bytes pending delalloc, used by stat to calc the
122 * real block usage of the file
123 */
Chris Mason90692182008-02-08 13:49:28 -0500124 u64 delalloc_bytes;
Chris Masond352ac62008-09-29 15:18:18 -0400125
126 /*
Filipe Mananaa7e3b972017-04-03 10:45:46 +0100127 * Total number of bytes pending delalloc that fall within a file
128 * range that is either a hole or beyond EOF (and no prealloc extent
129 * exists in the range). This is always <= delalloc_bytes.
130 */
131 u64 new_delalloc_bytes;
132
133 /*
Wang Shilong47059d92014-07-03 18:22:07 +0800134 * total number of bytes pending defrag, used by stat to check whether
135 * it needs COW.
136 */
137 u64 defrag_bytes;
138
139 /*
Chris Masond352ac62008-09-29 15:18:18 -0400140 * the size of the file stored in the metadata on disk. data=ordered
141 * means the in-memory i_size might be larger than the size on disk
142 * because not all the blocks are written yet.
143 */
Chris Masondbe674a2008-07-17 12:54:05 -0400144 u64 disk_i_size;
Chris Masond352ac62008-09-29 15:18:18 -0400145
Josef Bacikaec74772008-07-24 12:12:38 -0400146 /*
147 * if this is a directory then index_cnt is the counter for the index
148 * number for new files that are created
149 */
150 u64 index_cnt;
Chris Masond352ac62008-09-29 15:18:18 -0400151
Miao Xie67de1172013-12-26 13:07:06 +0800152 /* Cache the directory index number to speed the dir/file remove */
153 u64 dir_index;
154
Chris Mason12fcfd22009-03-24 10:24:20 -0400155 /* the fsync log has some corner cases that mean we have to check
156 * directories to see if any unlinks have been done before
157 * the directory was logged. See tree-log.c for all the
158 * details
159 */
160 u64 last_unlink_trans;
161
Josef Bacik7709cde2011-08-04 10:25:02 -0400162 /*
163 * Number of bytes outstanding that are going to need csums. This is
164 * used in ENOSPC accounting.
165 */
166 u64 csum_bytes;
167
Josef Bacikf1bdcc02011-07-14 14:28:08 -0400168 /* flags field from the on disk inode */
169 u32 flags;
170
Chris Mason5a3f23d2009-03-31 13:27:11 -0400171 /*
Josef Bacik32c00af2009-10-08 13:34:05 -0400172 * Counters to keep track of the number of extent item's we may use due
173 * to delalloc and such. outstanding_extents is the number of extent
174 * items we think we'll end up using, and reserved_extents is the number
175 * of extent items we've reserved metadata for.
Josef Bacik9ed74f22009-09-11 16:12:44 -0400176 */
Josef Bacik9e0baf62011-07-15 15:16:44 +0000177 unsigned outstanding_extents;
Josef Bacik69fe2d72017-10-19 14:15:57 -0400178
179 struct btrfs_block_rsv block_rsv;
Josef Bacik9ed74f22009-09-11 16:12:44 -0400180
181 /*
David Sterbab52aa8c2017-07-17 19:17:20 +0200182 * Cached values of inode properties
Chris Mason1e701a32010-03-11 09:42:04 -0500183 */
David Sterbab52aa8c2017-07-17 19:17:20 +0200184 unsigned prop_compress; /* per-file compression algorithm */
David Sterbaeec63c62017-07-17 19:41:31 +0200185 /*
186 * Force compression on the file using the defrag ioctl, could be
187 * different from prop_compress and takes precedence if set
188 */
189 unsigned defrag_compress;
Chris Mason1e701a32010-03-11 09:42:04 -0500190
Miao Xie16cdcec2011-04-22 18:12:22 +0800191 struct btrfs_delayed_node *delayed_node;
192
chandan r9cc97d62012-07-04 12:48:07 +0530193 /* File creation time. */
194 struct timespec i_otime;
195
David Sterba8089fe62015-11-19 14:15:51 +0100196 /* Hook into fs_info->delayed_iputs */
197 struct list_head delayed_iput;
David Sterba8089fe62015-11-19 14:15:51 +0100198
Filipe Manana5f9a8a52016-05-12 13:53:36 +0100199 /*
200 * To avoid races between lockless (i_mutex not held) direct IO writes
201 * and concurrent fsync requests. Direct IO writes must acquire read
202 * access on this semaphore for creating an extent map and its
203 * corresponding ordered extent. The fast fsync path must acquire write
204 * access on this semaphore before it collects ordered extents and
205 * extent maps.
206 */
207 struct rw_semaphore dio_sem;
208
Chris Masond352ac62008-09-29 15:18:18 -0400209 struct inode vfs_inode;
Chris Mason2c90e5d2007-04-02 10:50:19 -0400210};
Chris Masondbe674a2008-07-17 12:54:05 -0400211
Miao Xie16cdcec2011-04-22 18:12:22 +0800212extern unsigned char btrfs_filetype_table[];
213
Jeff Mahoney9a35b632017-06-28 21:56:54 -0600214static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
Chris Mason2c90e5d2007-04-02 10:50:19 -0400215{
216 return container_of(inode, struct btrfs_inode, vfs_inode);
217}
218
Filipe David Borba Manana778ba822013-10-06 22:22:33 +0100219static inline unsigned long btrfs_inode_hash(u64 objectid,
220 const struct btrfs_root *root)
221{
222 u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
223
224#if BITS_PER_LONG == 32
225 h = (h >> 32) ^ (h & 0xffffffff);
226#endif
227
228 return (unsigned long)h;
229}
230
231static inline void btrfs_insert_inode_hash(struct inode *inode)
232{
233 unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
234
235 __insert_inode_hash(inode, h);
236}
237
Jeff Mahoney9a35b632017-06-28 21:56:54 -0600238static inline u64 btrfs_ino(const struct btrfs_inode *inode)
Li Zefan33345d012011-04-20 10:31:50 +0800239{
Nikolay Borisov4a0cc7c2017-01-10 20:35:31 +0200240 u64 ino = inode->location.objectid;
Li Zefan33345d012011-04-20 10:31:50 +0800241
Liu Bo14c7cca2011-09-11 10:52:24 -0400242 /*
243 * !ino: btree_inode
244 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
245 */
Nikolay Borisov4a0cc7c2017-01-10 20:35:31 +0200246 if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
247 ino = inode->vfs_inode.i_ino;
Li Zefan33345d012011-04-20 10:31:50 +0800248 return ino;
249}
250
Nikolay Borisov6ef06d22017-02-20 13:50:34 +0200251static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
Chris Masondbe674a2008-07-17 12:54:05 -0400252{
Nikolay Borisov6ef06d22017-02-20 13:50:34 +0200253 i_size_write(&inode->vfs_inode, size);
254 inode->disk_i_size = size;
Chris Masondbe674a2008-07-17 12:54:05 -0400255}
256
Nikolay Borisov70ddc552017-02-20 13:50:35 +0200257static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
Chris Mason2cf85722011-07-26 15:35:09 -0400258{
Nikolay Borisov70ddc552017-02-20 13:50:35 +0200259 struct btrfs_root *root = inode->root;
Liu Bo83eea1f2012-07-10 05:28:39 -0600260
Liu Bo51a8cf92012-07-10 05:28:38 -0600261 if (root == root->fs_info->tree_root &&
Nikolay Borisov70ddc552017-02-20 13:50:35 +0200262 btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
Liu Bo51a8cf92012-07-10 05:28:38 -0600263 return true;
Nikolay Borisov70ddc552017-02-20 13:50:35 +0200264 if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID)
Chris Mason2cf85722011-07-26 15:35:09 -0400265 return true;
266 return false;
267}
268
Josef Bacik8b62f872017-10-19 14:15:55 -0400269static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
270 int mod)
271{
272 lockdep_assert_held(&inode->lock);
273 inode->outstanding_extents += mod;
274 if (btrfs_is_free_space_inode(inode))
275 return;
Josef Bacikdd48d402017-10-19 14:15:56 -0400276 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
277 mod);
Josef Bacik8b62f872017-10-19 14:15:55 -0400278}
279
Nikolay Borisov0f8939b2017-01-18 00:31:30 +0200280static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
Josef Bacik22ee6985d2012-05-29 16:57:49 -0400281{
Filipe Manana2f2ff0e2015-03-20 17:19:46 +0000282 int ret = 0;
283
Nikolay Borisov0f8939b2017-01-18 00:31:30 +0200284 spin_lock(&inode->lock);
285 if (inode->logged_trans == generation &&
286 inode->last_sub_trans <= inode->last_log_commit &&
287 inode->last_sub_trans <= inode->root->last_log_commit) {
Filipe Manana125c4cf92014-09-11 21:22:14 +0100288 /*
289 * After a ranged fsync we might have left some extent maps
290 * (that fall outside the fsync's range). So return false
291 * here if the list isn't empty, to make sure btrfs_log_inode()
292 * will be called and process those extent maps.
293 */
294 smp_mb();
Nikolay Borisov0f8939b2017-01-18 00:31:30 +0200295 if (list_empty(&inode->extent_tree.modified_extents))
Filipe Manana2f2ff0e2015-03-20 17:19:46 +0000296 ret = 1;
Filipe Manana125c4cf92014-09-11 21:22:14 +0100297 }
Nikolay Borisov0f8939b2017-01-18 00:31:30 +0200298 spin_unlock(&inode->lock);
Filipe Manana2f2ff0e2015-03-20 17:19:46 +0000299 return ret;
Josef Bacik22ee6985d2012-05-29 16:57:49 -0400300}
301
Miao Xiec1dc0892014-09-12 18:43:56 +0800302#define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1
303
Miao Xiefacc8a222013-07-25 19:22:34 +0800304struct btrfs_dio_private {
305 struct inode *inode;
Miao Xiec1dc0892014-09-12 18:43:56 +0800306 unsigned long flags;
Miao Xiefacc8a222013-07-25 19:22:34 +0800307 u64 logical_offset;
308 u64 disk_bytenr;
309 u64 bytes;
310 void *private;
311
312 /* number of bios pending for this dio */
313 atomic_t pending_bios;
314
315 /* IO errors */
316 int errors;
317
318 /* orig_bio is our btrfs_io_bio */
319 struct bio *orig_bio;
320
321 /* dio_bio came from fs/direct-io.c */
322 struct bio *dio_bio;
Miao Xiec1dc0892014-09-12 18:43:56 +0800323
324 /*
Nicholas D Steeves01327612016-05-19 21:18:45 -0400325 * The original bio may be split to several sub-bios, this is
Miao Xiec1dc0892014-09-12 18:43:56 +0800326 * done during endio of sub-bios
327 */
Christoph Hellwig4e4cbee2017-06-03 09:38:06 +0200328 blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *,
329 blk_status_t);
Miao Xiefacc8a222013-07-25 19:22:34 +0800330};
331
Miao Xie2e60a512013-02-08 07:01:08 +0000332/*
333 * Disable DIO read nolock optimization, so new dio readers will be forced
334 * to grab i_mutex. It is used to avoid the endless truncate due to
335 * nonlocked dio read.
336 */
Nikolay Borisovabcefb12017-02-20 13:51:10 +0200337static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode)
Miao Xie2e60a512013-02-08 07:01:08 +0000338{
Nikolay Borisovabcefb12017-02-20 13:51:10 +0200339 set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
Miao Xie2e60a512013-02-08 07:01:08 +0000340 smp_mb();
341}
342
Nikolay Borisov0b581702017-02-20 13:51:11 +0200343static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode)
Miao Xie2e60a512013-02-08 07:01:08 +0000344{
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100345 smp_mb__before_atomic();
Nikolay Borisov0b581702017-02-20 13:51:11 +0200346 clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
Miao Xie2e60a512013-02-08 07:01:08 +0000347}
348
Nikolay Borisov0970a222017-02-20 13:50:53 +0200349static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
Qu Wenruo6f6b6432017-02-09 10:45:06 +0800350 u64 logical_start, u32 csum, u32 csum_expected, int mirror_num)
351{
Nikolay Borisov0970a222017-02-20 13:50:53 +0200352 struct btrfs_root *root = inode->root;
Qu Wenruo6f6b6432017-02-09 10:45:06 +0800353
354 /* Output minus objectid, which is more meaningful */
355 if (root->objectid >= BTRFS_LAST_FREE_OBJECTID)
356 btrfs_warn_rl(root->fs_info,
357 "csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d",
Nikolay Borisov0970a222017-02-20 13:50:53 +0200358 root->objectid, btrfs_ino(inode),
Qu Wenruo6f6b6432017-02-09 10:45:06 +0800359 logical_start, csum, csum_expected, mirror_num);
360 else
361 btrfs_warn_rl(root->fs_info,
362 "csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d",
Nikolay Borisov0970a222017-02-20 13:50:53 +0200363 root->objectid, btrfs_ino(inode),
Qu Wenruo6f6b6432017-02-09 10:45:06 +0800364 logical_start, csum, csum_expected, mirror_num);
365}
366
Chris Mason2c90e5d2007-04-02 10:50:19 -0400367#endif