ctree.h 71 KB
Newer Older
Chris Mason's avatar
Chris Mason committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

Chris Mason's avatar
Chris Mason committed
19
20
#ifndef __BTRFS_CTREE__
#define __BTRFS_CTREE__
21

22
#include <linux/version.h>
23
24
#include <linux/mm.h>
#include <linux/highmem.h>
Chris Mason's avatar
Chris Mason committed
25
#include <linux/fs.h>
26
#include <linux/completion.h>
Chris Mason's avatar
Chris Mason committed
27
#include <linux/backing-dev.h>
28
#include <linux/wait.h>
29
#include <asm/kmap_types.h>
30
#include "extent_io.h"
31
#include "extent_map.h"
32
#include "async-thread.h"
Chris Mason's avatar
Chris Mason committed
33

34
struct btrfs_trans_handle;
Chris Mason's avatar
Chris Mason committed
35
struct btrfs_transaction;
36
37
38
extern struct kmem_cache *btrfs_trans_handle_cachep;
extern struct kmem_cache *btrfs_transaction_cachep;
extern struct kmem_cache *btrfs_bit_radix_cachep;
Chris Mason's avatar
Chris Mason committed
39
extern struct kmem_cache *btrfs_path_cachep;
40
struct btrfs_ordered_sum;
41

42
#define BTRFS_MAGIC "_BHRfS_M"
43

Josef Bacik's avatar
Josef Bacik committed
44
45
#define BTRFS_ACL_NOT_CACHED    ((void *)-1)

46
#define BTRFS_MAX_LEVEL 8
47

48
49
50
51
52
53
54
/*
 * files bigger than this get some pre-flushing when they are added
 * to the ordered operations list.  That way we limit the total
 * work done by the commit
 */
#define BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT (8 * 1024 * 1024)

55
/* holds pointers to all of the tree roots */
56
#define BTRFS_ROOT_TREE_OBJECTID 1ULL
57
58

/* stores information about which extents are in use, and reference counts */
Chris Mason's avatar
Chris Mason committed
59
#define BTRFS_EXTENT_TREE_OBJECTID 2ULL
60
61
62
63
64

/*
 * chunk tree stores translations from logical -> physical block numbering
 * the super block points to the chunk tree
 */
65
#define BTRFS_CHUNK_TREE_OBJECTID 3ULL
66
67
68
69
70

/*
 * stores information about which areas of a given device are in use.
 * one per device.  The tree of tree roots points to the device tree
 */
71
72
73
74
75
76
77
#define BTRFS_DEV_TREE_OBJECTID 4ULL

/* one per subvolume, storing files and directories */
#define BTRFS_FS_TREE_OBJECTID 5ULL

/* directory objectid inside the root tree */
#define BTRFS_ROOT_TREE_DIR_OBJECTID 6ULL
78

79
80
81
/* holds checksums of all the data extents */
#define BTRFS_CSUM_TREE_OBJECTID 7ULL

82
83
84
/* orhpan objectid for tracking unlinked/truncated files */
#define BTRFS_ORPHAN_OBJECTID -5ULL

85
86
87
88
/* does write ahead logging to speed up fsyncs */
#define BTRFS_TREE_LOG_OBJECTID -6ULL
#define BTRFS_TREE_LOG_FIXUP_OBJECTID -7ULL

Zheng Yan's avatar
Zheng Yan committed
89
90
91
92
/* for space balancing */
#define BTRFS_TREE_RELOC_OBJECTID -8ULL
#define BTRFS_DATA_RELOC_TREE_OBJECTID -9ULL

93
94
95
96
97
98
99
/*
 * extent checksums all have this objectid
 * this allows them to share the logging tree
 * for fsyncs
 */
#define BTRFS_EXTENT_CSUM_OBJECTID -10ULL

Zheng Yan's avatar
Zheng Yan committed
100
101
102
/* dummy objectid represents multiple objectids */
#define BTRFS_MULTIPLE_OBJECTIDS -255ULL

103
/*
104
 * All files have objectids in this range.
105
 */
106
#define BTRFS_FIRST_FREE_OBJECTID 256ULL
107
#define BTRFS_LAST_FREE_OBJECTID -256ULL
108
#define BTRFS_FIRST_CHUNK_TREE_OBJECTID 256ULL
109

110
111
112
113
114
115
116

/*
 * the device items go into the chunk tree.  The key is in the form
 * [ 1 BTRFS_DEV_ITEM_KEY device_id ]
 */
#define BTRFS_DEV_ITEMS_OBJECTID 1ULL

Chris Mason's avatar
Chris Mason committed
117
118
119
120
121
122
/*
 * we can actually store much bigger names, but lets not confuse the rest
 * of linux
 */
#define BTRFS_NAME_LEN 255

Chris Mason's avatar
Chris Mason committed
123
124
/* 32 bytes in various csum fields */
#define BTRFS_CSUM_SIZE 32
125
126
127
128
129
130

/* csum types */
#define BTRFS_CSUM_TYPE_CRC32	0

static int btrfs_csum_sizes[] = { 4, 0 };

131
/* four bytes for CRC32 */
132
#define BTRFS_EMPTY_DIR_SIZE 0
Chris Mason's avatar
Chris Mason committed
133

Chris Mason's avatar
Chris Mason committed
134
135
136
137
138
139
140
141
#define BTRFS_FT_UNKNOWN	0
#define BTRFS_FT_REG_FILE	1
#define BTRFS_FT_DIR		2
#define BTRFS_FT_CHRDEV		3
#define BTRFS_FT_BLKDEV		4
#define BTRFS_FT_FIFO		5
#define BTRFS_FT_SOCK		6
#define BTRFS_FT_SYMLINK	7
Josef Bacik's avatar
Josef Bacik committed
142
143
#define BTRFS_FT_XATTR		8
#define BTRFS_FT_MAX		9
Chris Mason's avatar
Chris Mason committed
144

145
/*
Wu Fengguang's avatar
Wu Fengguang committed
146
147
148
149
150
151
152
153
154
 * The key defines the order in the tree, and so it also defines (optimal)
 * block layout.
 *
 * objectid corresponds to the inode number.
 *
 * type tells us things about the object, and is a kind of stream selector.
 * so for a given inode, keys with type of 1 might refer to the inode data,
 * type of 2 may point to file data in the btree and type == 3 may point to
 * extents.
155
156
 *
 * offset is the starting byte offset for this key in the stream.
Chris Mason's avatar
Chris Mason committed
157
158
159
160
 *
 * btrfs_disk_key is in disk byte order.  struct btrfs_key is always
 * in cpu native order.  Otherwise they are identical and their sizes
 * should be the same (ie both packed)
161
 */
Chris Mason's avatar
Chris Mason committed
162
163
struct btrfs_disk_key {
	__le64 objectid;
164
	u8 type;
165
	__le64 offset;
Chris Mason's avatar
Chris Mason committed
166
167
168
} __attribute__ ((__packed__));

struct btrfs_key {
169
	u64 objectid;
170
	u8 type;
171
	u64 offset;
172
173
} __attribute__ ((__packed__));

174
175
176
177
struct btrfs_mapping_tree {
	struct extent_map_tree map_tree;
};

178
#define BTRFS_UUID_SIZE 16
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
struct btrfs_dev_item {
	/* the internal btrfs device id */
	__le64 devid;

	/* size of the device */
	__le64 total_bytes;

	/* bytes used */
	__le64 bytes_used;

	/* optimal io alignment for this device */
	__le32 io_align;

	/* optimal io width for this device */
	__le32 io_width;

	/* minimal io size for this device */
	__le32 sector_size;

	/* type and info about this device */
	__le64 type;

Yan Zheng's avatar
Yan Zheng committed
201
202
203
	/* expected generation for this device */
	__le64 generation;

204
205
	/*
	 * starting byte of this partition on the device,
Wu Fengguang's avatar
Wu Fengguang committed
206
	 * to allow for stripe alignment in the future
207
208
209
	 */
	__le64 start_offset;

210
211
212
213
214
215
216
217
218
	/* grouping information for allocation decisions */
	__le32 dev_group;

	/* seek speed 0-100 where 100 is fastest */
	u8 seek_speed;

	/* bandwidth 0-100 where 100 is fastest */
	u8 bandwidth;

219
	/* btrfs generated uuid for this device */
220
	u8 uuid[BTRFS_UUID_SIZE];
Yan Zheng's avatar
Yan Zheng committed
221
222
223

	/* uuid of FS who owns this device */
	u8 fsid[BTRFS_UUID_SIZE];
224
225
226
227
228
} __attribute__ ((__packed__));

struct btrfs_stripe {
	__le64 devid;
	__le64 offset;
229
	u8 dev_uuid[BTRFS_UUID_SIZE];
230
231
232
} __attribute__ ((__packed__));

struct btrfs_chunk {
233
234
235
236
	/* size of this chunk in bytes */
	__le64 length;

	/* objectid of the root referencing this chunk */
237
	__le64 owner;
238

239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
	__le64 stripe_len;
	__le64 type;

	/* optimal io alignment for this chunk */
	__le32 io_align;

	/* optimal io width for this chunk */
	__le32 io_width;

	/* minimal io size for this chunk */
	__le32 sector_size;

	/* 2^16 stripes is quite a lot, a second limit is the size of a single
	 * item in the btree
	 */
	__le16 num_stripes;
Chris Mason's avatar
Chris Mason committed
255
256
257

	/* sub stripes only matter for raid10 */
	__le16 sub_stripes;
258
259
260
261
262
263
264
265
266
267
268
	struct btrfs_stripe stripe;
	/* additional stripes go here */
} __attribute__ ((__packed__));

static inline unsigned long btrfs_chunk_item_size(int num_stripes)
{
	BUG_ON(num_stripes == 0);
	return sizeof(struct btrfs_chunk) +
		sizeof(struct btrfs_stripe) * (num_stripes - 1);
}

269
#define BTRFS_FSID_SIZE 16
270
271
#define BTRFS_HEADER_FLAG_WRITTEN (1 << 0)

272
273
274
/*
 * every tree block (leaf or node) starts with this header.
 */
275
struct btrfs_header {
276
	/* these first four must match the super block */
Chris Mason's avatar
Chris Mason committed
277
	u8 csum[BTRFS_CSUM_SIZE];
278
	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
279
	__le64 bytenr; /* which block this node is supposed to live in */
280
	__le64 flags;
281
282
283

	/* allowed to be different from the super from here on down */
	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
284
	__le64 generation;
285
	__le64 owner;
286
	__le32 nritems;
287
	u8 level;
288
289
} __attribute__ ((__packed__));

290
#define BTRFS_NODEPTRS_PER_BLOCK(r) (((r)->nodesize - \
291
292
				      sizeof(struct btrfs_header)) / \
				     sizeof(struct btrfs_key_ptr))
293
#define __BTRFS_LEAF_DATA_SIZE(bs) ((bs) - sizeof(struct btrfs_header))
294
#define BTRFS_LEAF_DATA_SIZE(r) (__BTRFS_LEAF_DATA_SIZE(r->leafsize))
295
296
297
#define BTRFS_MAX_INLINE_DATA_SIZE(r) (BTRFS_LEAF_DATA_SIZE(r) - \
					sizeof(struct btrfs_item) - \
					sizeof(struct btrfs_file_extent_item))
298

Yan Zheng's avatar
Yan Zheng committed
299
#define BTRFS_SUPER_FLAG_SEEDING (1ULL << 32)
300
301
302
303
304
305

/*
 * this is a very generous portion of the super block, giving us
 * room to translate 14 chunks with 3 stripes each.
 */
#define BTRFS_SYSTEM_CHUNK_ARRAY_SIZE 2048
306
#define BTRFS_LABEL_SIZE 256
307

308
309
310
311
/*
 * the super block basically lists the main trees of the FS
 * it currently lacks any block count etc etc
 */
312
struct btrfs_super_block {
Chris Mason's avatar
Chris Mason committed
313
	u8 csum[BTRFS_CSUM_SIZE];
314
	/* the first 4 fields must match struct btrfs_header */
Yan Zheng's avatar
Yan Zheng committed
315
	u8 fsid[BTRFS_FSID_SIZE];    /* FS specific uuid */
316
	__le64 bytenr; /* this block number */
317
	__le64 flags;
318
319

	/* allowed to be different from the btrfs_header from here own down */
320
321
322
	__le64 magic;
	__le64 generation;
	__le64 root;
323
	__le64 chunk_root;
324
	__le64 log_root;
325
326
327

	/* this will help find the new super based on the log root */
	__le64 log_root_transid;
328
329
	__le64 total_bytes;
	__le64 bytes_used;
330
	__le64 root_dir_objectid;
331
	__le64 num_devices;
332
333
334
	__le32 sectorsize;
	__le32 nodesize;
	__le32 leafsize;
335
	__le32 stripesize;
336
	__le32 sys_chunk_array_size;
337
	__le64 chunk_root_generation;
338
339
340
	__le64 compat_flags;
	__le64 compat_ro_flags;
	__le64 incompat_flags;
341
	__le16 csum_type;
342
	u8 root_level;
343
	u8 chunk_root_level;
344
	u8 log_root_level;
345
	struct btrfs_dev_item dev_item;
346

347
	char label[BTRFS_LABEL_SIZE];
348
349
350

	/* future expansion */
	__le64 reserved[32];
351
	u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE];
Chris Mason's avatar
Chris Mason committed
352
353
} __attribute__ ((__packed__));

354
355
356
357
358
359
360
361
/*
 * Compat flags that we support.  If any incompat flags are set other than the
 * ones specified below then we will fail to mount
 */
#define BTRFS_FEATURE_COMPAT_SUPP	0x0
#define BTRFS_FEATURE_COMPAT_RO_SUPP	0x0
#define BTRFS_FEATURE_INCOMPAT_SUPP	0x0

362
/*
363
 * A leaf is full of items. offset and size tell us where to find
364
365
 * the item in the leaf (relative to the start of the data area)
 */
Chris Mason's avatar
Chris Mason committed
366
struct btrfs_item {
Chris Mason's avatar
Chris Mason committed
367
	struct btrfs_disk_key key;
368
	__le32 offset;
369
	__le32 size;
370
371
} __attribute__ ((__packed__));

372
373
374
375
376
377
378
/*
 * leaves have an item area and a data area:
 * [item0, item1....itemN] [free space] [dataN...data1, data0]
 *
 * The data is separate from the items to get the keys closer together
 * during searches.
 */
379
struct btrfs_leaf {
380
	struct btrfs_header header;
381
	struct btrfs_item items[];
382
383
} __attribute__ ((__packed__));

384
385
386
387
/*
 * all non-leaf blocks are nodes, they hold only keys and pointers to
 * other blocks
 */
388
389
390
struct btrfs_key_ptr {
	struct btrfs_disk_key key;
	__le64 blockptr;
391
	__le64 generation;
392
393
} __attribute__ ((__packed__));

394
struct btrfs_node {
395
	struct btrfs_header header;
396
	struct btrfs_key_ptr ptrs[];
397
398
} __attribute__ ((__packed__));

399
/*
400
401
 * btrfs_paths remember the path taken from the root down to the leaf.
 * level 0 is always the leaf, and nodes[1...BTRFS_MAX_LEVEL] will point
402
403
404
405
406
 * to any other levels that are present.
 *
 * The slots array records the index of the item or block pointer
 * used while walking the tree.
 */
407
struct btrfs_path {
408
	struct extent_buffer *nodes[BTRFS_MAX_LEVEL];
409
	int slots[BTRFS_MAX_LEVEL];
410
411
	/* if there is real range locking, this locks field will change */
	int locks[BTRFS_MAX_LEVEL];
412
	int reada;
413
	/* keep some upper locks as we walk down */
414
	int lowest_level;
415
416
417
418
419

	/*
	 * set by btrfs_split_item, tells search_slot to keep all locks
	 * and to force calls to keep space in the nodes
	 */
420
421
422
423
	unsigned int search_for_split:1;
	unsigned int keep_locks:1;
	unsigned int skip_locking:1;
	unsigned int leave_spinning:1;
424
};
Chris Mason's avatar
Chris Mason committed
425

426
427
428
429
430
431
/*
 * items in the extent btree are used to record the objectid of the
 * owner of the block and the number of references
 */
struct btrfs_extent_item {
	__le32 refs;
432
433
434
435
436
437
} __attribute__ ((__packed__));

struct btrfs_extent_ref {
	__le64 root;
	__le64 generation;
	__le64 objectid;
Zheng Yan's avatar
Zheng Yan committed
438
	__le32 num_refs;
439
440
} __attribute__ ((__packed__));

441
442
/* dev extents record free space on individual devices.  The owner
 * field points back to the chunk allocation mapping tree that allocated
443
 * the extent.  The chunk tree uuid field is a way to double check the owner
444
445
 */
struct btrfs_dev_extent {
446
447
448
	__le64 chunk_tree;
	__le64 chunk_objectid;
	__le64 chunk_offset;
449
	__le64 length;
450
	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
451
452
} __attribute__ ((__packed__));

453
struct btrfs_inode_ref {
454
	__le64 index;
455
456
457
458
	__le16 name_len;
	/* name goes here */
} __attribute__ ((__packed__));

459
struct btrfs_timespec {
Chris Mason's avatar
Chris Mason committed
460
	__le64 sec;
Chris Mason's avatar
Chris Mason committed
461
462
463
	__le32 nsec;
} __attribute__ ((__packed__));

Jan Engelhardt's avatar
Jan Engelhardt committed
464
enum btrfs_compression_type {
465
466
467
	BTRFS_COMPRESS_NONE = 0,
	BTRFS_COMPRESS_ZLIB = 1,
	BTRFS_COMPRESS_LAST = 2,
Jan Engelhardt's avatar
Jan Engelhardt committed
468
};
469

Chris Mason's avatar
Chris Mason committed
470
struct btrfs_inode_item {
471
	/* nfs style generation number */
Chris Mason's avatar
Chris Mason committed
472
	__le64 generation;
473
474
	/* transid that last touched this inode */
	__le64 transid;
Chris Mason's avatar
Chris Mason committed
475
	__le64 size;
476
	__le64 nbytes;
477
	__le64 block_group;
Chris Mason's avatar
Chris Mason committed
478
479
480
481
	__le32 nlink;
	__le32 uid;
	__le32 gid;
	__le32 mode;
482
	__le64 rdev;
483
	__le64 flags;
484

485
486
487
488
489
490
491
492
	/* modification sequence number for NFS */
	__le64 sequence;

	/*
	 * a little future expansion, for more than this we can
	 * just grow the inode item and version it
	 */
	__le64 reserved[4];
493
494
495
496
	struct btrfs_timespec atime;
	struct btrfs_timespec ctime;
	struct btrfs_timespec mtime;
	struct btrfs_timespec otime;
Chris Mason's avatar
Chris Mason committed
497
498
} __attribute__ ((__packed__));

499
500
501
502
struct btrfs_dir_log_item {
	__le64 end;
} __attribute__ ((__packed__));

503
struct btrfs_dir_item {
504
	struct btrfs_disk_key location;
505
	__le64 transid;
Josef Bacik's avatar
Josef Bacik committed
506
	__le16 data_len;
507
	__le16 name_len;
508
509
510
511
	u8 type;
} __attribute__ ((__packed__));

struct btrfs_root_item {
512
	struct btrfs_inode_item inode;
513
	__le64 generation;
514
	__le64 root_dirid;
515
516
517
	__le64 bytenr;
	__le64 byte_limit;
	__le64 bytes_used;
Yan Zheng's avatar
Yan Zheng committed
518
	__le64 last_snapshot;
519
	__le64 flags;
520
	__le32 refs;
521
522
	struct btrfs_disk_key drop_progress;
	u8 drop_level;
523
	u8 level;
524
} __attribute__ ((__packed__));
525

526
527
528
529
530
531
532
533
534
/*
 * this is used for both forward and backward root refs
 */
struct btrfs_root_ref {
	__le64 dirid;
	__le64 sequence;
	__le16 name_len;
} __attribute__ ((__packed__));

Yan Zheng's avatar
Yan Zheng committed
535
536
537
#define BTRFS_FILE_EXTENT_INLINE 0
#define BTRFS_FILE_EXTENT_REG 1
#define BTRFS_FILE_EXTENT_PREALLOC 2
538

539
struct btrfs_file_extent_item {
540
541
542
	/*
	 * transaction id that created this extent
	 */
543
	__le64 generation;
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
	/*
	 * max number of bytes to hold this extent in ram
	 * when we split a compressed extent we can't know how big
	 * each of the resulting pieces will be.  So, this is
	 * an upper limit on the size of the extent in ram instead of
	 * an exact limit.
	 */
	__le64 ram_bytes;

	/*
	 * 32 bits for the various ways we might encode the data,
	 * including compression and encryption.  If any of these
	 * are set to something a given disk format doesn't understand
	 * it is treated like an incompat flag for reading and writing,
	 * but not for stat.
	 */
	u8 compression;
	u8 encryption;
	__le16 other_encoding; /* spare for later use */

	/* are we inline data or a real extent? */
565
	u8 type;
566

567
568
569
570
	/*
	 * disk space consumed by the extent, checksum blocks are included
	 * in these numbers
	 */
571
572
	__le64 disk_bytenr;
	__le64 disk_num_bytes;
573
	/*
Chris Mason's avatar
Chris Mason committed
574
	 * the logical offset in file blocks (no csums)
575
576
577
578
579
580
581
	 * this extent record is for.  This allows a file extent to point
	 * into the middle of an existing extent on disk, sharing it
	 * between two snapshots (useful if some bytes in the middle of the
	 * extent have changed
	 */
	__le64 offset;
	/*
582
583
	 * the logical number of file blocks (no csums included).  This
	 * always reflects the size uncompressed and without encoding.
584
	 */
585
	__le64 num_bytes;
586

587
588
} __attribute__ ((__packed__));

Chris Mason's avatar
Chris Mason committed
589
struct btrfs_csum_item {
590
	u8 csum;
Chris Mason's avatar
Chris Mason committed
591
592
} __attribute__ ((__packed__));

593
594
595
596
/* different types of block groups (and chunks) */
#define BTRFS_BLOCK_GROUP_DATA     (1 << 0)
#define BTRFS_BLOCK_GROUP_SYSTEM   (1 << 1)
#define BTRFS_BLOCK_GROUP_METADATA (1 << 2)
597
#define BTRFS_BLOCK_GROUP_RAID0    (1 << 3)
598
#define BTRFS_BLOCK_GROUP_RAID1    (1 << 4)
599
#define BTRFS_BLOCK_GROUP_DUP	   (1 << 5)
Chris Mason's avatar
Chris Mason committed
600
#define BTRFS_BLOCK_GROUP_RAID10   (1 << 6)
Chris Mason's avatar
Chris Mason committed
601

Chris Mason's avatar
Chris Mason committed
602
603
struct btrfs_block_group_item {
	__le64 used;
604
605
	__le64 chunk_objectid;
	__le64 flags;
Chris Mason's avatar
Chris Mason committed
606
607
} __attribute__ ((__packed__));

608
609
struct btrfs_space_info {
	u64 flags;
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630

	u64 total_bytes;	/* total bytes in the space */
	u64 bytes_used;		/* total bytes used on disk */
	u64 bytes_pinned;	/* total bytes pinned, will be freed when the
				   transaction finishes */
	u64 bytes_reserved;	/* total bytes the allocator has reserved for
				   current allocations */
	u64 bytes_readonly;	/* total bytes that are read only */

	/* delalloc accounting */
	u64 bytes_delalloc;	/* number of bytes reserved for allocation,
				   this space is not necessarily reserved yet
				   by the allocator */
	u64 bytes_may_use;	/* number of bytes that may be used for
				   delalloc */

	int full;		/* indicates that we cannot allocate any more
				   chunks for this space */
	int force_alloc;	/* set if we need to force a chunk alloc for
				   this space */

631
	struct list_head list;
632
633
634
635

	/* for block groups in our same type */
	struct list_head block_groups;
	spinlock_t lock;
636
	struct rw_semaphore groups_sem;
637
638
};

639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
/*
 * free clusters are used to claim free space in relatively large chunks,
 * allowing us to do less seeky writes.  They are used for all metadata
 * allocations and data allocations in ssd mode.
 */
struct btrfs_free_cluster {
	spinlock_t lock;
	spinlock_t refill_lock;
	struct rb_root root;

	/* largest extent in this cluster */
	u64 max_size;

	/* first extent starting offset */
	u64 window_start;

	struct btrfs_block_group_cache *block_group;
	/*
	 * when a cluster is allocated from a block group, we put the
	 * cluster onto a list in the block group so that it can
	 * be freed before the block group is freed.
	 */
	struct list_head block_group_list;
662
663
};

Chris Mason's avatar
Chris Mason committed
664
665
666
struct btrfs_block_group_cache {
	struct btrfs_key key;
	struct btrfs_block_group_item item;
667
	spinlock_t lock;
668
	struct mutex cache_mutex;
669
	u64 pinned;
670
	u64 reserved;
671
672
	u64 flags;
	int cached;
673
	int ro;
674
675
676
677
678
	int dirty;

	struct btrfs_space_info *space_info;

	/* free space cache stuff */
679
	spinlock_t tree_lock;
680
681
682
683
684
685
686
687
	struct rb_root free_space_bytes;
	struct rb_root free_space_offset;

	/* block group cache stuff */
	struct rb_node cache_node;

	/* for block groups in the same raid type */
	struct list_head list;
688
689
690

	/* usage count */
	atomic_t count;
691
692
693
694
695

	/* List of struct btrfs_free_clusters for this block group.
	 * Today it will only have one thing on it, but that may change
	 */
	struct list_head cluster_list;
Chris Mason's avatar
Chris Mason committed
696
};
697

Zheng Yan's avatar
Zheng Yan committed
698
699
700
701
702
703
struct btrfs_leaf_ref_tree {
	struct rb_root root;
	struct list_head list;
	spinlock_t lock;
};

704
struct btrfs_device;
705
struct btrfs_fs_devices;
706
struct btrfs_fs_info {
707
	u8 fsid[BTRFS_FSID_SIZE];
708
	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
709
710
	struct btrfs_root *extent_root;
	struct btrfs_root *tree_root;
711
712
	struct btrfs_root *chunk_root;
	struct btrfs_root *dev_root;
713
	struct btrfs_root *fs_root;
714
	struct btrfs_root *csum_root;
715
716
717

	/* the log root tree is a directory of all the other log roots */
	struct btrfs_root *log_root_tree;
718
	struct radix_tree_root fs_roots_radix;
719

720
721
722
723
	/* block group cache stuff */
	spinlock_t block_group_cache_lock;
	struct rb_root block_group_cache_tree;

724
	struct extent_io_tree pinned_extents;
725

726
727
728
	/* logical->physical extent mapping */
	struct btrfs_mapping_tree mapping_tree;

729
	u64 generation;
730
	u64 last_trans_committed;
731
732
733
734
735
736

	/*
	 * this is updated to the current trans every time a full commit
	 * is required instead of the faster short fsync log commits
	 */
	u64 last_trans_log_full_commit;
737
	u64 open_ioctl_trans;
738
	unsigned long mount_opt;
739
	u64 max_extent;
740
	u64 max_inline;
741
	u64 alloc_start;
Chris Mason's avatar
Chris Mason committed
742
	struct btrfs_transaction *running_transaction;
743
	wait_queue_head_t transaction_throttle;
744
	wait_queue_head_t transaction_wait;
745
	wait_queue_head_t async_submit_wait;
746

747
	struct btrfs_super_block super_copy;
748
	struct btrfs_super_block super_for_commit;
749
	struct block_device *__bdev;
Chris Mason's avatar
Chris Mason committed
750
	struct super_block *sb;
751
	struct inode *btree_inode;
Chris Mason's avatar
Chris Mason committed
752
	struct backing_dev_info bdi;
Chris Mason's avatar
Chris Mason committed
753
	struct mutex trans_mutex;
754
	struct mutex tree_log_mutex;
755
756
	struct mutex transaction_kthread_mutex;
	struct mutex cleaner_mutex;
757
	struct mutex chunk_mutex;
758
	struct mutex drop_mutex;
759
	struct mutex volume_mutex;
Zheng Yan's avatar
Zheng Yan committed
760
	struct mutex tree_reloc_mutex;
761

762
763
764
765
766
767
768
769
770
	/*
	 * this protects the ordered operations list only while we are
	 * processing all of the entries on it.  This way we make
	 * sure the commit code doesn't find the list temporarily empty
	 * because another function happens to be doing non-waiting preflush
	 * before jumping into the main commit.
	 */
	struct mutex ordered_operations_mutex;

Chris Mason's avatar
Chris Mason committed
771
	struct list_head trans_list;
772
	struct list_head hashers;
773
	struct list_head dead_roots;
774

775
	atomic_t nr_async_submits;
776
	atomic_t async_submit_draining;
777
	atomic_t nr_async_bios;
778
	atomic_t async_delalloc_pages;
779

780
781
782
783
784
	/*
	 * this is used by the balancing code to wait for all the pending
	 * ordered extents
	 */
	spinlock_t ordered_extent_lock;
785
786
787
788
789
790

	/*
	 * all of the data=ordered extents pending writeback
	 * these can span multiple transactions and basically include
	 * every dirty data page that isn't from nodatacow
	 */
791
	struct list_head ordered_extents;
792
793
794
795
796
797

	/*
	 * all of the inodes that have delalloc bytes.  It is possible for
	 * this list to be empty even when there is still dirty data=ordered
	 * extents waiting to finish IO.
	 */
798
	struct list_head delalloc_inodes;
799

800
801
802
803
804
805
806
	/*
	 * special rename and truncate targets that must be on disk before
	 * we're allowed to commit.  This is basically the ext3 style
	 * data=ordered list.
	 */
	struct list_head ordered_operations;

807
808
809
810
811
812
	/*
	 * there is a pool of worker threads for checksumming during writes
	 * and a pool for checksumming after reads.  This is because readers
	 * can run with FS locks held, and the writers may be waiting for
	 * those locks.  We don't want ordering in the pending list to cause
	 * deadlocks, and so the two are serviced separately.
813
814
815
	 *
	 * A third pool does submit_bio to avoid deadlocking with the other
	 * two
816
817
	 */
	struct btrfs_workers workers;
818
	struct btrfs_workers delalloc_workers;
819
	struct btrfs_workers endio_workers;
820
	struct btrfs_workers endio_meta_workers;
821
	struct btrfs_workers endio_meta_write_workers;
822
	struct btrfs_workers endio_write_workers;
823
	struct btrfs_workers submit_workers;
824
825
826
827
828
829
	/*
	 * fixup workers take dirty pages that didn't properly go through
	 * the cow mechanism and make them safe to write.  It happens
	 * for the sys_munmap function call path
	 */
	struct btrfs_workers fixup_workers;
830
831
	struct task_struct *transaction_kthread;
	struct task_struct *cleaner_kthread;
832
	int thread_pool_size;
833

Zheng Yan's avatar
Zheng Yan committed
834
835
836
	/* tree relocation relocated fields */
	struct list_head dead_reloc_roots;
	struct btrfs_leaf_ref_tree reloc_ref_tree;
Zheng Yan's avatar
Zheng Yan committed
837
838
	struct btrfs_leaf_ref_tree shared_ref_tree;

839
840
	struct kobject super_kobj;
	struct completion kobj_unregister;
841
	int do_barriers;
842
	int closing;
843
	int log_root_recovering;
844
	atomic_t throttles;
845
	atomic_t throttle_gen;
846

847
	u64 total_pinned;
848
849
850
851
852

	/* protected by the delalloc lock, used to keep from writing
	 * metadata until there is a nice batch
	 */
	u64 dirty_metadata_bytes;
853
854
	struct list_head dirty_cowonly_roots;

855
	struct btrfs_fs_devices *fs_devices;
856
857
858
859
860
861

	/*
	 * the space_info list is almost entirely read only.  It only changes
	 * when we add a new raid type to the FS, and that happens
	 * very rarely.  RCU is used to protect it.
	 */
862
	struct list_head space_info;
863

864
	spinlock_t delalloc_lock;
865
	spinlock_t new_trans_lock;
866
	u64 delalloc_bytes;
867
868
869
870
871
872

	/* data_alloc_cluster is only used in ssd mode */
	struct btrfs_free_cluster data_alloc_cluster;

	/* all metadata allocations go through this cluster */
	struct btrfs_free_cluster meta_alloc_cluster;
873

Yan Zheng's avatar
Yan Zheng committed
874
875
876
	spinlock_t ref_cache_lock;
	u64 total_ref_cache_size;

877
878
879
880
881
882
	u64 avail_data_alloc_bits;
	u64 avail_metadata_alloc_bits;
	u64 avail_system_alloc_bits;
	u64 data_alloc_profile;
	u64 metadata_alloc_profile;
	u64 system_alloc_profile;
883
884

	void *bdev_holder;
885
};
886

887
888
/*
 * in ram representation of the tree.  extent_root is used for all allocations
889
 * and for the extent tree extent_root root.
890
 */
891
struct btrfs_dirty_root;
892
struct btrfs_root {
893
	struct extent_buffer *node;
894
895
896
897

	/* the node lock is held while changing the node pointer */
	spinlock_t node_lock;

898
	struct extent_buffer *commit_root;
Yan Zheng's avatar
Yan Zheng committed
899
	struct btrfs_leaf_ref_tree *ref_tree;
900
	struct btrfs_leaf_ref_tree ref_tree_struct;
901
	struct btrfs_dirty_root *dirty_root;
902
	struct btrfs_root *log_root;
Zheng Yan's avatar
Zheng Yan committed
903
	struct btrfs_root *reloc_root;
Yan Zheng's avatar
Yan Zheng committed
904

905
906
	struct btrfs_root_item root_item;
	struct btrfs_key root_key;
907
	struct btrfs_fs_info *fs_info;
908
909
	struct extent_io_tree dirty_log_pages;

910
911
	struct kobject root_kobj;
	struct completion kobj_unregister;
912
	struct mutex objectid_mutex;
Yan Zheng's avatar
Yan Zheng committed
913

914
	struct mutex log_mutex;
Yan Zheng's avatar
Yan Zheng committed
915
916
917
918
919
920
	wait_queue_head_t log_writer_wait;
	wait_queue_head_t log_commit_wait[2];
	atomic_t log_writers;
	atomic_t log_commit[2];
	unsigned long log_transid;
	unsigned long log_batch;
921

922
923
	u64 objectid;
	u64 last_trans;
924
925
926
927
928
929
930
931
932
933

	/* data allocations are done in sectorsize units */
	u32 sectorsize;

	/* node allocations are done in nodesize units */
	u32 nodesize;

	/* leaf allocations are done in leafsize units */
	u32 leafsize;

934
935
	u32 stripesize;

936
	u32 type;
Chris Mason's avatar
Chris Mason committed
937
938
	u64 highest_inode;
	u64 last_inode_alloc;
939
	int ref_cows;
940
	int track_dirty;
941
	u64 defrag_trans_start;
942
	struct btrfs_key defrag_progress;
943
	struct btrfs_key defrag_max;
944
945
	int defrag_running;
	int defrag_level;
946
	char *name;
947
	int in_sysfs;
948
949
950

	/* the dirty list is only used by non-reference counted roots */
	struct list_head dirty_list;
951

952
953
	spinlock_t list_lock;
	struct list_head dead_list;
954
	struct list_head orphan_list;
955
956
957
958
959
960

	/*
	 * right now this just gets used so that a root has its own devid
	 * for stat.  It may be used for more later
	 */
	struct super_block anon_super;
961
962
};

Chris Mason's avatar
Chris Mason committed
963
964
965
966
967
/*
 * inode items have the data typically returned from stat and store other
 * info about object characteristics.  There is one for every file and dir in
 * the FS
 */
Chris Mason's avatar
Chris Mason committed
968
#define BTRFS_INODE_ITEM_KEY		1
969
970
971
#define BTRFS_INODE_REF_KEY		12
#define BTRFS_XATTR_ITEM_KEY		24
#define BTRFS_ORPHAN_ITEM_KEY		48
Chris Mason's avatar
Chris Mason committed
972
/* reserve 2-15 close to the inode for later flexibility */
Chris Mason's avatar
Chris Mason committed
973
974
975
976
977

/*
 * dir items are the name -> inode pointers in a directory.  There is one
 * for every name in a directory.
 */
978
979
980
981
#define BTRFS_DIR_LOG_ITEM_KEY  60
#define BTRFS_DIR_LOG_INDEX_KEY 72
#define BTRFS_DIR_ITEM_KEY	84
#define BTRFS_DIR_INDEX_KEY	96
Chris Mason's avatar
Chris Mason committed
982
/*
Chris Mason's avatar
Chris Mason committed
983
 * extent data is for file data
Chris Mason's avatar
Chris Mason committed
984
 */
985
#define BTRFS_EXTENT_DATA_KEY	108
986

Chris Mason's avatar
Chris Mason committed
987
/*
988
989
 * extent csums are stored in a separate tree and hold csums for
 * an entire extent on disk.
Chris Mason's avatar
Chris Mason committed
990
 */
991
#define BTRFS_EXTENT_CSUM_KEY	128
Chris Mason's avatar
Chris Mason committed
992

Chris Mason's avatar
Chris Mason committed
993
/*
Wu Fengguang's avatar
Wu Fengguang committed
994
 * root items point to tree roots.  They are typically in the root
Chris Mason's avatar
Chris Mason committed
995
996
 * tree used by the super block to find all the other trees
 */
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
#define BTRFS_ROOT_ITEM_KEY	132

/*
 * root backrefs tie subvols and snapshots to the directory entries that
 * reference them
 */
#define BTRFS_ROOT_BACKREF_KEY	144

/*
 * root refs make a fast index for listing all of the snapshots and
 * subvolumes referenced by a given root.  They point directly to the
 * directory item in the root that references the subvol
 */
#define BTRFS_ROOT_REF_KEY	156

Chris Mason's avatar
Chris Mason committed
1012
1013
1014
1015
/*
 * extent items are in the extent map tree.  These record which blocks
 * are used, and how many references there are to each block
 */
1016
1017
#define BTRFS_EXTENT_ITEM_KEY	168
#define BTRFS_EXTENT_REF_KEY	180
Chris Mason's avatar
Chris Mason committed
1018
1019
1020
1021
1022

/*
 * block groups give us hints into the extent allocation trees.  Which
 * blocks are free etc etc
 */
1023
#define BTRFS_BLOCK_GROUP_ITEM_KEY 192
1024

1025
1026
1027
#define BTRFS_DEV_EXTENT_KEY	204
#define BTRFS_DEV_ITEM_KEY	216
#define BTRFS_CHUNK_ITEM_KEY	228
1028

Chris Mason's avatar
Chris Mason committed
1029
1030
1031
1032
/*
 * string items are for debugging.  They just store a short string of
 * data in the FS
 */
Chris Mason's avatar