blob: 1a8fa46ff87eb6eccae9be16d5e600ff8de0ef8b [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
Josef Bacik72ac3c02012-05-23 14:13:11 -040047
Aneeshf1ace242007-06-13 16:18:26 -040048/* in memory btrfs inode */
Chris Mason2c90e5d2007-04-02 10:50:19 -040049struct btrfs_inode {
Chris Masond352ac62008-09-29 15:18:18 -040050 /* which subvolume this inode belongs to */
Chris Masond6e4a422007-04-06 15:37:36 -040051 struct btrfs_root *root;
Chris Masond352ac62008-09-29 15:18:18 -040052
Chris Masond352ac62008-09-29 15:18:18 -040053 /* key used to find this inode on disk. This is used by the code
54 * to read in roots of subvolumes
55 */
Chris Masond6e4a422007-04-06 15:37:36 -040056 struct btrfs_key location;
Chris Masond352ac62008-09-29 15:18:18 -040057
Filipe Manana2f2ff0e2015-03-20 17:19:46 +000058 /*
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 */
Josef Bacik9e0baf62011-07-15 15:16:44 +000063 spinlock_t lock;
64
Chris Masond352ac62008-09-29 15:18:18 -040065 /* the extent_tree has caches of all the extent mappings to disk */
Chris Masona52d9a82007-08-27 16:49:44 -040066 struct extent_map_tree extent_tree;
Chris Masond352ac62008-09-29 15:18:18 -040067
68 /* the io_tree does range state (DIRTY, LOCKED etc) */
Chris Masond1310b22008-01-24 16:13:08 -050069 struct extent_io_tree io_tree;
Chris Masond352ac62008-09-29 15:18:18 -040070
71 /* special utility tree used to record which mirrors have already been
72 * tried when checksums fail for a given block
73 */
Chris Mason7e383262008-04-09 16:28:12 -040074 struct extent_io_tree io_failure_tree;
Chris Masond352ac62008-09-29 15:18:18 -040075
Chris Masond352ac62008-09-29 15:18:18 -040076 /* held while logging the inode in tree-log.c */
Chris Masone02119d2008-09-05 16:13:11 -040077 struct mutex log_mutex;
Chris Masond352ac62008-09-29 15:18:18 -040078
Josef Bacikf2486792012-01-13 12:09:22 -050079 /* held while doing delalloc reservations */
80 struct mutex delalloc_mutex;
81
Chris Masond352ac62008-09-29 15:18:18 -040082 /* used to order data wrt metadata */
Chris Masone6dcd2d2008-07-17 12:53:50 -040083 struct btrfs_ordered_inode_tree ordered_tree;
Josef Bacik15ee9bc2007-08-10 16:22:09 -040084
Chris Masond352ac62008-09-29 15:18:18 -040085 /* 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 */
Chris Masonea8c2812008-08-04 23:17:27 -040089 struct list_head delalloc_inodes;
90
Yan Zheng5d4f98a2009-06-10 10:45:14 -040091 /* node for the red-black tree that links inodes in subvolume root */
92 struct rb_node rb_node;
93
Josef Bacik72ac3c02012-05-23 14:13:11 -040094 unsigned long runtime_flags;
95
Nathaniel Yazdani9c931c52013-04-15 00:44:02 +000096 /* Keep track of who's O_SYNC/fsyncing currently */
Josef Bacikb812ce22012-11-16 13:56:32 -050097 atomic_t sync_writers;
98
Chris Masond352ac62008-09-29 15:18:18 -040099 /* full 64 bit generation number, struct vfs_inode doesn't have a big
100 * enough field for this.
101 */
Chris Masone02119d2008-09-05 16:13:11 -0400102 u64 generation;
103
Josef Bacik15ee9bc2007-08-10 16:22:09 -0400104 /*
105 * transid of the trans_handle that last modified this inode
106 */
107 u64 last_trans;
Chris Mason257c62e2009-10-13 13:21:08 -0400108
109 /*
Chris Masone02119d2008-09-05 16:13:11 -0400110 * transid that last logged this inode
111 */
112 u64 logged_trans;
Chris Mason49eb7e42008-09-11 15:53:12 -0400113
Miao Xiebb14a592014-02-20 18:08:56 +0800114 /*
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
Chris Masond352ac62008-09-29 15:18:18 -0400122 /* total number of bytes pending delalloc, used by stat to calc the
123 * real block usage of the file
124 */
Chris Mason90692182008-02-08 13:49:28 -0500125 u64 delalloc_bytes;
Chris Masond352ac62008-09-29 15:18:18 -0400126
127 /*
Wang Shilong47059d92014-07-03 18:22:07 +0800128 * total number of bytes pending defrag, used by stat to check whether
129 * it needs COW.
130 */
131 u64 defrag_bytes;
132
133 /*
Chris Masond352ac62008-09-29 15:18:18 -0400134 * 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 */
Chris Masondbe674a2008-07-17 12:54:05 -0400138 u64 disk_i_size;
Chris Masond352ac62008-09-29 15:18:18 -0400139
Josef Bacikaec74772008-07-24 12:12:38 -0400140 /*
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;
Chris Masond352ac62008-09-29 15:18:18 -0400145
Miao Xie67de1172013-12-26 13:07:06 +0800146 /* Cache the directory index number to speed the dir/file remove */
147 u64 dir_index;
148
Chris Mason12fcfd22009-03-24 10:24:20 -0400149 /* 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
Josef Bacik7709cde2011-08-04 10:25:02 -0400156 /*
157 * Number of bytes outstanding that are going to need csums. This is
158 * used in ENOSPC accounting.
159 */
160 u64 csum_bytes;
161
Josef Bacikf1bdcc02011-07-14 14:28:08 -0400162 /* flags field from the on disk inode */
163 u32 flags;
164
Chris Mason5a3f23d2009-03-31 13:27:11 -0400165 /*
Josef Bacik32c00af2009-10-08 13:34:05 -0400166 * 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.
Josef Bacik9ed74f22009-09-11 16:12:44 -0400170 */
Josef Bacik9e0baf62011-07-15 15:16:44 +0000171 unsigned outstanding_extents;
172 unsigned reserved_extents;
Josef Bacik9ed74f22009-09-11 16:12:44 -0400173
174 /*
Chris Mason1e701a32010-03-11 09:42:04 -0500175 * always compress this one file
176 */
Josef Bacik72ac3c02012-05-23 14:13:11 -0400177 unsigned force_compress;
Chris Mason1e701a32010-03-11 09:42:04 -0500178
Miao Xie16cdcec2011-04-22 18:12:22 +0800179 struct btrfs_delayed_node *delayed_node;
180
chandan r9cc97d62012-07-04 12:48:07 +0530181 /* File creation time. */
182 struct timespec i_otime;
183
David Sterba8089fe62015-11-19 14:15:51 +0100184 /* Hook into fs_info->delayed_iputs */
185 struct list_head delayed_iput;
186 long delayed_iput_count;
187
Filipe Manana5f9a8a52016-05-12 13:53:36 +0100188 /*
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
Chris Masond352ac62008-09-29 15:18:18 -0400198 struct inode vfs_inode;
Chris Mason2c90e5d2007-04-02 10:50:19 -0400199};
Chris Masondbe674a2008-07-17 12:54:05 -0400200
Miao Xie16cdcec2011-04-22 18:12:22 +0800201extern unsigned char btrfs_filetype_table[];
202
Chris Mason2c90e5d2007-04-02 10:50:19 -0400203static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
204{
205 return container_of(inode, struct btrfs_inode, vfs_inode);
206}
207
Filipe David Borba Manana778ba822013-10-06 22:22:33 +0100208static 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
Li Zefan33345d012011-04-20 10:31:50 +0800227static inline u64 btrfs_ino(struct inode *inode)
228{
229 u64 ino = BTRFS_I(inode)->location.objectid;
230
Liu Bo14c7cca2011-09-11 10:52:24 -0400231 /*
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)
Li Zefan33345d012011-04-20 10:31:50 +0800236 ino = inode->i_ino;
237 return ino;
238}
239
Chris Masondbe674a2008-07-17 12:54:05 -0400240static inline void btrfs_i_size_write(struct inode *inode, u64 size)
241{
Yan, Zhengc2167752009-11-12 09:34:21 +0000242 i_size_write(inode, size);
Chris Masondbe674a2008-07-17 12:54:05 -0400243 BTRFS_I(inode)->disk_i_size = size;
244}
245
Liu Bo83eea1f2012-07-10 05:28:39 -0600246static inline bool btrfs_is_free_space_inode(struct inode *inode)
Chris Mason2cf85722011-07-26 15:35:09 -0400247{
Liu Bo83eea1f2012-07-10 05:28:39 -0600248 struct btrfs_root *root = BTRFS_I(inode)->root;
249
Liu Bo51a8cf92012-07-10 05:28:38 -0600250 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)
Chris Mason2cf85722011-07-26 15:35:09 -0400254 return true;
255 return false;
256}
257
Josef Bacik22ee6982012-05-29 16:57:49 -0400258static inline int btrfs_inode_in_log(struct inode *inode, u64 generation)
259{
Filipe Manana2f2ff0e2015-03-20 17:19:46 +0000260 int ret = 0;
261
262 spin_lock(&BTRFS_I(inode)->lock);
Josef Bacik22ee6982012-05-29 16:57:49 -0400263 if (BTRFS_I(inode)->logged_trans == generation &&
Josef Bacika5874ce2013-09-11 09:55:42 -0400264 BTRFS_I(inode)->last_sub_trans <=
265 BTRFS_I(inode)->last_log_commit &&
266 BTRFS_I(inode)->last_sub_trans <=
Filipe Manana125c4cf92014-09-11 21:22:14 +0100267 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))
Filipe Manana2f2ff0e2015-03-20 17:19:46 +0000276 ret = 1;
Filipe Manana125c4cf92014-09-11 21:22:14 +0100277 }
Filipe Manana2f2ff0e2015-03-20 17:19:46 +0000278 spin_unlock(&BTRFS_I(inode)->lock);
279 return ret;
Josef Bacik22ee6982012-05-29 16:57:49 -0400280}
281
Miao Xiec1dc0892014-09-12 18:43:56 +0800282#define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1
283
Miao Xiefacc8a222013-07-25 19:22:34 +0800284struct btrfs_dio_private {
285 struct inode *inode;
Miao Xiec1dc0892014-09-12 18:43:56 +0800286 unsigned long flags;
Miao Xiefacc8a222013-07-25 19:22:34 +0800287 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;
Miao Xiec1dc0892014-09-12 18:43:56 +0800303
304 /*
Nicholas D Steeves01327612016-05-19 21:18:45 -0400305 * The original bio may be split to several sub-bios, this is
Miao Xiec1dc0892014-09-12 18:43:56 +0800306 * done during endio of sub-bios
307 */
Miao Xie8b110e32014-09-12 18:44:03 +0800308 int (*subio_endio)(struct inode *, struct btrfs_io_bio *, int);
Miao Xiefacc8a222013-07-25 19:22:34 +0800309};
310
Miao Xie2e60a512013-02-08 07:01:08 +0000311/*
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{
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100324 smp_mb__before_atomic();
Miao Xie2e60a512013-02-08 07:01:08 +0000325 clear_bit(BTRFS_INODE_READDIO_NEED_LOCK,
326 &BTRFS_I(inode)->runtime_flags);
327}
328
Alex Gartrellfc4adbf2014-05-20 13:07:56 -0700329bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);
330
Chris Mason2c90e5d2007-04-02 10:50:19 -0400331#endif