master
1/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2#ifndef _BTRFS_CTREE_H_
3#define _BTRFS_CTREE_H_
4
5#include <linux/btrfs.h>
6#include <linux/types.h>
7#include <stddef.h>
8
9/* ASCII for _BHRfS_M, no terminating nul */
10#define BTRFS_MAGIC 0x4D5F53665248425FULL
11
12#define BTRFS_MAX_LEVEL 8
13
14/*
15 * We can actually store much bigger names, but lets not confuse the rest of
16 * linux.
17 */
18#define BTRFS_NAME_LEN 255
19
20/*
21 * Theoretical limit is larger, but we keep this down to a sane value. That
22 * should limit greatly the possibility of collisions on inode ref items.
23 */
24#define BTRFS_LINK_MAX 65535U
25
26/*
27 * This header contains the structure definitions and constants used
28 * by file system objects that can be retrieved using
29 * the BTRFS_IOC_SEARCH_TREE ioctl. That means basically anything that
30 * is needed to describe a leaf node's key or item contents.
31 */
32
33/* holds pointers to all of the tree roots */
34#define BTRFS_ROOT_TREE_OBJECTID 1ULL
35
36/* stores information about which extents are in use, and reference counts */
37#define BTRFS_EXTENT_TREE_OBJECTID 2ULL
38
39/*
40 * chunk tree stores translations from logical -> physical block numbering
41 * the super block points to the chunk tree
42 */
43#define BTRFS_CHUNK_TREE_OBJECTID 3ULL
44
45/*
46 * stores information about which areas of a given device are in use.
47 * one per device. The tree of tree roots points to the device tree
48 */
49#define BTRFS_DEV_TREE_OBJECTID 4ULL
50
51/* one per subvolume, storing files and directories */
52#define BTRFS_FS_TREE_OBJECTID 5ULL
53
54/* directory objectid inside the root tree */
55#define BTRFS_ROOT_TREE_DIR_OBJECTID 6ULL
56
57/* holds checksums of all the data extents */
58#define BTRFS_CSUM_TREE_OBJECTID 7ULL
59
60/* holds quota configuration and tracking */
61#define BTRFS_QUOTA_TREE_OBJECTID 8ULL
62
63/* for storing items that use the BTRFS_UUID_KEY* types */
64#define BTRFS_UUID_TREE_OBJECTID 9ULL
65
66/* tracks free space in block groups. */
67#define BTRFS_FREE_SPACE_TREE_OBJECTID 10ULL
68
69/* Holds the block group items for extent tree v2. */
70#define BTRFS_BLOCK_GROUP_TREE_OBJECTID 11ULL
71
72/* Tracks RAID stripes in block groups. */
73#define BTRFS_RAID_STRIPE_TREE_OBJECTID 12ULL
74
75/* device stats in the device tree */
76#define BTRFS_DEV_STATS_OBJECTID 0ULL
77
78/* for storing balance parameters in the root tree */
79#define BTRFS_BALANCE_OBJECTID -4ULL
80
81/* orphan objectid for tracking unlinked/truncated files */
82#define BTRFS_ORPHAN_OBJECTID -5ULL
83
84/* does write ahead logging to speed up fsyncs */
85#define BTRFS_TREE_LOG_OBJECTID -6ULL
86#define BTRFS_TREE_LOG_FIXUP_OBJECTID -7ULL
87
88/* for space balancing */
89#define BTRFS_TREE_RELOC_OBJECTID -8ULL
90#define BTRFS_DATA_RELOC_TREE_OBJECTID -9ULL
91
92/*
93 * extent checksums all have this objectid
94 * this allows them to share the logging tree
95 * for fsyncs
96 */
97#define BTRFS_EXTENT_CSUM_OBJECTID -10ULL
98
99/* For storing free space cache */
100#define BTRFS_FREE_SPACE_OBJECTID -11ULL
101
102/*
103 * The inode number assigned to the special inode for storing
104 * free ino cache
105 */
106#define BTRFS_FREE_INO_OBJECTID -12ULL
107
108/* dummy objectid represents multiple objectids */
109#define BTRFS_MULTIPLE_OBJECTIDS -255ULL
110
111/*
112 * All files have objectids in this range.
113 */
114#define BTRFS_FIRST_FREE_OBJECTID 256ULL
115#define BTRFS_LAST_FREE_OBJECTID -256ULL
116#define BTRFS_FIRST_CHUNK_TREE_OBJECTID 256ULL
117
118
119/*
120 * the device items go into the chunk tree. The key is in the form
121 * [ 1 BTRFS_DEV_ITEM_KEY device_id ]
122 */
123#define BTRFS_DEV_ITEMS_OBJECTID 1ULL
124
125#define BTRFS_BTREE_INODE_OBJECTID 1
126
127#define BTRFS_EMPTY_SUBVOL_DIR_OBJECTID 2
128
129#define BTRFS_DEV_REPLACE_DEVID 0ULL
130
131/*
132 * inode items have the data typically returned from stat and store other
133 * info about object characteristics. There is one for every file and dir in
134 * the FS
135 */
136#define BTRFS_INODE_ITEM_KEY 1
137#define BTRFS_INODE_REF_KEY 12
138#define BTRFS_INODE_EXTREF_KEY 13
139#define BTRFS_XATTR_ITEM_KEY 24
140
141/*
142 * fs verity items are stored under two different key types on disk.
143 * The descriptor items:
144 * [ inode objectid, BTRFS_VERITY_DESC_ITEM_KEY, offset ]
145 *
146 * At offset 0, we store a btrfs_verity_descriptor_item which tracks the size
147 * of the descriptor item and some extra data for encryption.
148 * Starting at offset 1, these hold the generic fs verity descriptor. The
149 * latter are opaque to btrfs, we just read and write them as a blob for the
150 * higher level verity code. The most common descriptor size is 256 bytes.
151 *
152 * The merkle tree items:
153 * [ inode objectid, BTRFS_VERITY_MERKLE_ITEM_KEY, offset ]
154 *
155 * These also start at offset 0, and correspond to the merkle tree bytes. When
156 * fsverity asks for page 0 of the merkle tree, we pull up one page starting at
157 * offset 0 for this key type. These are also opaque to btrfs, we're blindly
158 * storing whatever fsverity sends down.
159 */
160#define BTRFS_VERITY_DESC_ITEM_KEY 36
161#define BTRFS_VERITY_MERKLE_ITEM_KEY 37
162
163#define BTRFS_ORPHAN_ITEM_KEY 48
164/* reserve 2-15 close to the inode for later flexibility */
165
166/*
167 * dir items are the name -> inode pointers in a directory. There is one
168 * for every name in a directory. BTRFS_DIR_LOG_ITEM_KEY is no longer used
169 * but it's still defined here for documentation purposes and to help avoid
170 * having its numerical value reused in the future.
171 */
172#define BTRFS_DIR_LOG_ITEM_KEY 60
173#define BTRFS_DIR_LOG_INDEX_KEY 72
174#define BTRFS_DIR_ITEM_KEY 84
175#define BTRFS_DIR_INDEX_KEY 96
176/*
177 * extent data is for file data
178 */
179#define BTRFS_EXTENT_DATA_KEY 108
180
181/*
182 * extent csums are stored in a separate tree and hold csums for
183 * an entire extent on disk.
184 */
185#define BTRFS_EXTENT_CSUM_KEY 128
186
187/*
188 * root items point to tree roots. They are typically in the root
189 * tree used by the super block to find all the other trees
190 */
191#define BTRFS_ROOT_ITEM_KEY 132
192
193/*
194 * root backrefs tie subvols and snapshots to the directory entries that
195 * reference them
196 */
197#define BTRFS_ROOT_BACKREF_KEY 144
198
199/*
200 * root refs make a fast index for listing all of the snapshots and
201 * subvolumes referenced by a given root. They point directly to the
202 * directory item in the root that references the subvol
203 */
204#define BTRFS_ROOT_REF_KEY 156
205
206/*
207 * extent items are in the extent map tree. These record which blocks
208 * are used, and how many references there are to each block
209 */
210#define BTRFS_EXTENT_ITEM_KEY 168
211
212/*
213 * The same as the BTRFS_EXTENT_ITEM_KEY, except it's metadata we already know
214 * the length, so we save the level in key->offset instead of the length.
215 */
216#define BTRFS_METADATA_ITEM_KEY 169
217
218/*
219 * Special __inline__ ref key which stores the id of the subvolume which originally
220 * created the extent. This subvolume owns the extent permanently from the
221 * perspective of simple quotas. Needed to know which subvolume to free quota
222 * usage from when the extent is deleted.
223 *
224 * Stored as an __inline__ ref rather to avoid wasting space on a separate item on
225 * top of the existing extent item. However, unlike the other __inline__ refs,
226 * there is one one owner ref per extent rather than one per extent.
227 *
228 * Because of this, it goes at the front of the list of __inline__ refs, and thus
229 * must have a lower type value than any other __inline__ ref type (to satisfy the
230 * disk format rule that __inline__ refs have non-decreasing type).
231 */
232#define BTRFS_EXTENT_OWNER_REF_KEY 172
233
234#define BTRFS_TREE_BLOCK_REF_KEY 176
235
236#define BTRFS_EXTENT_DATA_REF_KEY 178
237
238/*
239 * Obsolete key. Defintion removed in 6.6, value may be reused in the future.
240 *
241 * #define BTRFS_EXTENT_REF_V0_KEY 180
242 */
243
244#define BTRFS_SHARED_BLOCK_REF_KEY 182
245
246#define BTRFS_SHARED_DATA_REF_KEY 184
247
248/*
249 * block groups give us hints into the extent allocation trees. Which
250 * blocks are free etc etc
251 */
252#define BTRFS_BLOCK_GROUP_ITEM_KEY 192
253
254/*
255 * Every block group is represented in the free space tree by a free space info
256 * item, which stores some accounting information. It is keyed on
257 * (block_group_start, FREE_SPACE_INFO, block_group_length).
258 */
259#define BTRFS_FREE_SPACE_INFO_KEY 198
260
261/*
262 * A free space extent tracks an extent of space that is free in a block group.
263 * It is keyed on (start, FREE_SPACE_EXTENT, length).
264 */
265#define BTRFS_FREE_SPACE_EXTENT_KEY 199
266
267/*
268 * When a block group becomes very fragmented, we convert it to use bitmaps
269 * instead of extents. A free space bitmap is keyed on
270 * (start, FREE_SPACE_BITMAP, length); the corresponding item is a bitmap with
271 * (length / sectorsize) bits.
272 */
273#define BTRFS_FREE_SPACE_BITMAP_KEY 200
274
275#define BTRFS_DEV_EXTENT_KEY 204
276#define BTRFS_DEV_ITEM_KEY 216
277#define BTRFS_CHUNK_ITEM_KEY 228
278
279#define BTRFS_RAID_STRIPE_KEY 230
280
281/*
282 * Records the overall state of the qgroups.
283 * There's only one instance of this key present,
284 * (0, BTRFS_QGROUP_STATUS_KEY, 0)
285 */
286#define BTRFS_QGROUP_STATUS_KEY 240
287/*
288 * Records the currently used space of the qgroup.
289 * One key per qgroup, (0, BTRFS_QGROUP_INFO_KEY, qgroupid).
290 */
291#define BTRFS_QGROUP_INFO_KEY 242
292/*
293 * Contains the user configured limits for the qgroup.
294 * One key per qgroup, (0, BTRFS_QGROUP_LIMIT_KEY, qgroupid).
295 */
296#define BTRFS_QGROUP_LIMIT_KEY 244
297/*
298 * Records the child-parent relationship of qgroups. For
299 * each relation, 2 keys are present:
300 * (childid, BTRFS_QGROUP_RELATION_KEY, parentid)
301 * (parentid, BTRFS_QGROUP_RELATION_KEY, childid)
302 */
303#define BTRFS_QGROUP_RELATION_KEY 246
304
305/*
306 * Obsolete name, see BTRFS_TEMPORARY_ITEM_KEY.
307 */
308#define BTRFS_BALANCE_ITEM_KEY 248
309
310/*
311 * The key type for tree items that are stored persistently, but do not need to
312 * exist for extended period of time. The items can exist in any tree.
313 *
314 * [subtype, BTRFS_TEMPORARY_ITEM_KEY, data]
315 *
316 * Existing items:
317 *
318 * - balance status item
319 * (BTRFS_BALANCE_OBJECTID, BTRFS_TEMPORARY_ITEM_KEY, 0)
320 */
321#define BTRFS_TEMPORARY_ITEM_KEY 248
322
323/*
324 * Obsolete name, see BTRFS_PERSISTENT_ITEM_KEY
325 */
326#define BTRFS_DEV_STATS_KEY 249
327
328/*
329 * The key type for tree items that are stored persistently and usually exist
330 * for a long period, eg. filesystem lifetime. The item kinds can be status
331 * information, stats or preference values. The item can exist in any tree.
332 *
333 * [subtype, BTRFS_PERSISTENT_ITEM_KEY, data]
334 *
335 * Existing items:
336 *
337 * - device statistics, store IO stats in the device tree, one key for all
338 * stats
339 * (BTRFS_DEV_STATS_OBJECTID, BTRFS_DEV_STATS_KEY, 0)
340 */
341#define BTRFS_PERSISTENT_ITEM_KEY 249
342
343/*
344 * Persistently stores the device replace state in the device tree.
345 * The key is built like this: (0, BTRFS_DEV_REPLACE_KEY, 0).
346 */
347#define BTRFS_DEV_REPLACE_KEY 250
348
349/*
350 * Stores items that allow to quickly map UUIDs to something else.
351 * These items are part of the filesystem UUID tree.
352 * The key is built like this:
353 * (UUID_upper_64_bits, BTRFS_UUID_KEY*, UUID_lower_64_bits).
354 */
355#if BTRFS_UUID_SIZE != 16
356#error "UUID items require BTRFS_UUID_SIZE == 16!"
357#endif
358#define BTRFS_UUID_KEY_SUBVOL 251 /* for UUIDs assigned to subvols */
359#define BTRFS_UUID_KEY_RECEIVED_SUBVOL 252 /* for UUIDs assigned to
360 * received subvols */
361
362/*
363 * string items are for debugging. They just store a short string of
364 * data in the FS
365 */
366#define BTRFS_STRING_ITEM_KEY 253
367
368/* Maximum metadata block size (nodesize) */
369#define BTRFS_MAX_METADATA_BLOCKSIZE 65536
370
371/* 32 bytes in various csum fields */
372#define BTRFS_CSUM_SIZE 32
373
374/* csum types */
375enum btrfs_csum_type {
376 BTRFS_CSUM_TYPE_CRC32 = 0,
377 BTRFS_CSUM_TYPE_XXHASH = 1,
378 BTRFS_CSUM_TYPE_SHA256 = 2,
379 BTRFS_CSUM_TYPE_BLAKE2 = 3,
380};
381
382/*
383 * flags definitions for directory entry item type
384 *
385 * Used by:
386 * struct btrfs_dir_item.type
387 *
388 * Values 0..7 must match common file type values in fs_types.h.
389 */
390#define BTRFS_FT_UNKNOWN 0
391#define BTRFS_FT_REG_FILE 1
392#define BTRFS_FT_DIR 2
393#define BTRFS_FT_CHRDEV 3
394#define BTRFS_FT_BLKDEV 4
395#define BTRFS_FT_FIFO 5
396#define BTRFS_FT_SOCK 6
397#define BTRFS_FT_SYMLINK 7
398#define BTRFS_FT_XATTR 8
399#define BTRFS_FT_MAX 9
400/* Directory contains encrypted data */
401#define BTRFS_FT_ENCRYPTED 0x80
402
403static __inline__ __u8 btrfs_dir_flags_to_ftype(__u8 flags)
404{
405 return flags & ~BTRFS_FT_ENCRYPTED;
406}
407
408/*
409 * Inode flags
410 */
411#define BTRFS_INODE_NODATASUM (1U << 0)
412#define BTRFS_INODE_NODATACOW (1U << 1)
413#define BTRFS_INODE_READONLY (1U << 2)
414#define BTRFS_INODE_NOCOMPRESS (1U << 3)
415#define BTRFS_INODE_PREALLOC (1U << 4)
416#define BTRFS_INODE_SYNC (1U << 5)
417#define BTRFS_INODE_IMMUTABLE (1U << 6)
418#define BTRFS_INODE_APPEND (1U << 7)
419#define BTRFS_INODE_NODUMP (1U << 8)
420#define BTRFS_INODE_NOATIME (1U << 9)
421#define BTRFS_INODE_DIRSYNC (1U << 10)
422#define BTRFS_INODE_COMPRESS (1U << 11)
423
424#define BTRFS_INODE_ROOT_ITEM_INIT (1U << 31)
425
426#define BTRFS_INODE_FLAG_MASK \
427 (BTRFS_INODE_NODATASUM | \
428 BTRFS_INODE_NODATACOW | \
429 BTRFS_INODE_READONLY | \
430 BTRFS_INODE_NOCOMPRESS | \
431 BTRFS_INODE_PREALLOC | \
432 BTRFS_INODE_SYNC | \
433 BTRFS_INODE_IMMUTABLE | \
434 BTRFS_INODE_APPEND | \
435 BTRFS_INODE_NODUMP | \
436 BTRFS_INODE_NOATIME | \
437 BTRFS_INODE_DIRSYNC | \
438 BTRFS_INODE_COMPRESS | \
439 BTRFS_INODE_ROOT_ITEM_INIT)
440
441#define BTRFS_INODE_RO_VERITY (1U << 0)
442
443#define BTRFS_INODE_RO_FLAG_MASK (BTRFS_INODE_RO_VERITY)
444
445/*
446 * The key defines the order in the tree, and so it also defines (optimal)
447 * block layout.
448 *
449 * objectid corresponds to the inode number.
450 *
451 * type tells us things about the object, and is a kind of stream selector.
452 * so for a given inode, keys with type of 1 might refer to the inode data,
453 * type of 2 may point to file data in the btree and type == 3 may point to
454 * extents.
455 *
456 * offset is the starting byte offset for this key in the stream.
457 *
458 * btrfs_disk_key is in disk byte order. struct btrfs_key is always
459 * in cpu native order. Otherwise they are identical and their sizes
460 * should be the same (ie both packed)
461 */
462struct btrfs_disk_key {
463 __le64 objectid;
464 __u8 type;
465 __le64 offset;
466} __attribute__ ((__packed__));
467
468struct btrfs_key {
469 __u64 objectid;
470 __u8 type;
471 __u64 offset;
472} __attribute__ ((__packed__));
473
474/*
475 * Every tree block (leaf or node) starts with this header.
476 */
477struct btrfs_header {
478 /* These first four must match the super block */
479 __u8 csum[BTRFS_CSUM_SIZE];
480 /* FS specific uuid */
481 __u8 fsid[BTRFS_FSID_SIZE];
482 /* Which block this node is supposed to live in */
483 __le64 bytenr;
484 __le64 flags;
485
486 /* Allowed to be different from the super from here on down */
487 __u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
488 __le64 generation;
489 __le64 owner;
490 __le32 nritems;
491 __u8 level;
492} __attribute__ ((__packed__));
493
494/*
495 * This is a very generous portion of the super block, giving us room to
496 * translate 14 chunks with 3 stripes each.
497 */
498#define BTRFS_SYSTEM_CHUNK_ARRAY_SIZE 2048
499
500/*
501 * Just in case we somehow lose the roots and are not able to mount, we store
502 * an array of the roots from previous transactions in the super.
503 */
504#define BTRFS_NUM_BACKUP_ROOTS 4
505struct btrfs_root_backup {
506 __le64 tree_root;
507 __le64 tree_root_gen;
508
509 __le64 chunk_root;
510 __le64 chunk_root_gen;
511
512 __le64 extent_root;
513 __le64 extent_root_gen;
514
515 __le64 fs_root;
516 __le64 fs_root_gen;
517
518 __le64 dev_root;
519 __le64 dev_root_gen;
520
521 __le64 csum_root;
522 __le64 csum_root_gen;
523
524 __le64 total_bytes;
525 __le64 bytes_used;
526 __le64 num_devices;
527 /* future */
528 __le64 unused_64[4];
529
530 __u8 tree_root_level;
531 __u8 chunk_root_level;
532 __u8 extent_root_level;
533 __u8 fs_root_level;
534 __u8 dev_root_level;
535 __u8 csum_root_level;
536 /* future and to align */
537 __u8 unused_8[10];
538} __attribute__ ((__packed__));
539
540/*
541 * A leaf is full of items. offset and size tell us where to find the item in
542 * the leaf (relative to the start of the data area)
543 */
544struct btrfs_item {
545 struct btrfs_disk_key key;
546 __le32 offset;
547 __le32 size;
548} __attribute__ ((__packed__));
549
550/*
551 * Leaves have an item area and a data area:
552 * [item0, item1....itemN] [free space] [dataN...data1, data0]
553 *
554 * The data is separate from the items to get the keys closer together during
555 * searches.
556 */
557struct btrfs_leaf {
558 struct btrfs_header header;
559 struct btrfs_item items[];
560} __attribute__ ((__packed__));
561
562/*
563 * All non-leaf blocks are nodes, they hold only keys and pointers to other
564 * blocks.
565 */
566struct btrfs_key_ptr {
567 struct btrfs_disk_key key;
568 __le64 blockptr;
569 __le64 generation;
570} __attribute__ ((__packed__));
571
572struct btrfs_node {
573 struct btrfs_header header;
574 struct btrfs_key_ptr ptrs[];
575} __attribute__ ((__packed__));
576
577struct btrfs_dev_item {
578 /* the internal btrfs device id */
579 __le64 devid;
580
581 /* size of the device */
582 __le64 total_bytes;
583
584 /* bytes used */
585 __le64 bytes_used;
586
587 /* optimal io alignment for this device */
588 __le32 io_align;
589
590 /* optimal io width for this device */
591 __le32 io_width;
592
593 /* minimal io size for this device */
594 __le32 sector_size;
595
596 /* type and info about this device */
597 __le64 type;
598
599 /* expected generation for this device */
600 __le64 generation;
601
602 /*
603 * starting byte of this partition on the device,
604 * to allow for stripe alignment in the future
605 */
606 __le64 start_offset;
607
608 /* grouping information for allocation decisions */
609 __le32 dev_group;
610
611 /* seek speed 0-100 where 100 is fastest */
612 __u8 seek_speed;
613
614 /* bandwidth 0-100 where 100 is fastest */
615 __u8 bandwidth;
616
617 /* btrfs generated uuid for this device */
618 __u8 uuid[BTRFS_UUID_SIZE];
619
620 /* uuid of FS who owns this device */
621 __u8 fsid[BTRFS_UUID_SIZE];
622} __attribute__ ((__packed__));
623
624struct btrfs_stripe {
625 __le64 devid;
626 __le64 offset;
627 __u8 dev_uuid[BTRFS_UUID_SIZE];
628} __attribute__ ((__packed__));
629
630struct btrfs_chunk {
631 /* size of this chunk in bytes */
632 __le64 length;
633
634 /* objectid of the root referencing this chunk */
635 __le64 owner;
636
637 __le64 stripe_len;
638 __le64 type;
639
640 /* optimal io alignment for this chunk */
641 __le32 io_align;
642
643 /* optimal io width for this chunk */
644 __le32 io_width;
645
646 /* minimal io size for this chunk */
647 __le32 sector_size;
648
649 /* 2^16 stripes is quite a lot, a second limit is the size of a single
650 * item in the btree
651 */
652 __le16 num_stripes;
653
654 /* sub stripes only matter for raid10 */
655 __le16 sub_stripes;
656 struct btrfs_stripe stripe;
657 /* additional stripes go here */
658} __attribute__ ((__packed__));
659
660/*
661 * The super block basically lists the main trees of the FS.
662 */
663struct btrfs_super_block {
664 /* The first 4 fields must match struct btrfs_header */
665 __u8 csum[BTRFS_CSUM_SIZE];
666 /* FS specific UUID, visible to user */
667 __u8 fsid[BTRFS_FSID_SIZE];
668 /* This block number */
669 __le64 bytenr;
670 __le64 flags;
671
672 /* Allowed to be different from the btrfs_header from here own down */
673 __le64 magic;
674 __le64 generation;
675 __le64 root;
676 __le64 chunk_root;
677 __le64 log_root;
678
679 /*
680 * This member has never been utilized since the very beginning, thus
681 * it's always 0 regardless of kernel version. We always use
682 * generation + 1 to read log tree root. So here we mark it deprecated.
683 */
684 __le64 __unused_log_root_transid;
685 __le64 total_bytes;
686 __le64 bytes_used;
687 __le64 root_dir_objectid;
688 __le64 num_devices;
689 __le32 sectorsize;
690 __le32 nodesize;
691 __le32 __unused_leafsize;
692 __le32 stripesize;
693 __le32 sys_chunk_array_size;
694 __le64 chunk_root_generation;
695 __le64 compat_flags;
696 __le64 compat_ro_flags;
697 __le64 incompat_flags;
698 __le16 csum_type;
699 __u8 root_level;
700 __u8 chunk_root_level;
701 __u8 log_root_level;
702 struct btrfs_dev_item dev_item;
703
704 char label[BTRFS_LABEL_SIZE];
705
706 __le64 cache_generation;
707 __le64 uuid_tree_generation;
708
709 /* The UUID written into btree blocks */
710 __u8 metadata_uuid[BTRFS_FSID_SIZE];
711
712 __u64 nr_global_roots;
713
714 /* Future expansion */
715 __le64 reserved[27];
716 __u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE];
717 struct btrfs_root_backup super_roots[BTRFS_NUM_BACKUP_ROOTS];
718
719 /* Padded to 4096 bytes */
720 __u8 padding[565];
721} __attribute__ ((__packed__));
722
723#define BTRFS_FREE_SPACE_EXTENT 1
724#define BTRFS_FREE_SPACE_BITMAP 2
725
726struct btrfs_free_space_entry {
727 __le64 offset;
728 __le64 bytes;
729 __u8 type;
730} __attribute__ ((__packed__));
731
732struct btrfs_free_space_header {
733 struct btrfs_disk_key location;
734 __le64 generation;
735 __le64 num_entries;
736 __le64 num_bitmaps;
737} __attribute__ ((__packed__));
738
739struct btrfs_raid_stride {
740 /* The id of device this raid extent lives on. */
741 __le64 devid;
742 /* The physical location on disk. */
743 __le64 physical;
744} __attribute__ ((__packed__));
745
746struct btrfs_stripe_extent {
747 /* An array of raid strides this stripe is composed of. */
748 __DECLARE_FLEX_ARRAY(struct btrfs_raid_stride, strides);
749} __attribute__ ((__packed__));
750
751#define BTRFS_HEADER_FLAG_WRITTEN (1ULL << 0)
752#define BTRFS_HEADER_FLAG_RELOC (1ULL << 1)
753
754/* Super block flags */
755/* Errors detected */
756#define BTRFS_SUPER_FLAG_ERROR (1ULL << 2)
757
758#define BTRFS_SUPER_FLAG_SEEDING (1ULL << 32)
759#define BTRFS_SUPER_FLAG_METADUMP (1ULL << 33)
760#define BTRFS_SUPER_FLAG_METADUMP_V2 (1ULL << 34)
761#define BTRFS_SUPER_FLAG_CHANGING_FSID (1ULL << 35)
762#define BTRFS_SUPER_FLAG_CHANGING_FSID_V2 (1ULL << 36)
763
764/*
765 * Those are temporaray flags utilized by btrfs-progs to do offline conversion.
766 * They are rejected by kernel.
767 * But still keep them all here to avoid conflicts.
768 */
769#define BTRFS_SUPER_FLAG_CHANGING_BG_TREE (1ULL << 38)
770#define BTRFS_SUPER_FLAG_CHANGING_DATA_CSUM (1ULL << 39)
771#define BTRFS_SUPER_FLAG_CHANGING_META_CSUM (1ULL << 40)
772
773/*
774 * items in the extent btree are used to record the objectid of the
775 * owner of the block and the number of references
776 */
777
778struct btrfs_extent_item {
779 __le64 refs;
780 __le64 generation;
781 __le64 flags;
782} __attribute__ ((__packed__));
783
784struct btrfs_extent_item_v0 {
785 __le32 refs;
786} __attribute__ ((__packed__));
787
788
789#define BTRFS_EXTENT_FLAG_DATA (1ULL << 0)
790#define BTRFS_EXTENT_FLAG_TREE_BLOCK (1ULL << 1)
791
792/* following flags only apply to tree blocks */
793
794/* use full backrefs for extent pointers in the block */
795#define BTRFS_BLOCK_FLAG_FULL_BACKREF (1ULL << 8)
796
797#define BTRFS_BACKREF_REV_MAX 256
798#define BTRFS_BACKREF_REV_SHIFT 56
799#define BTRFS_BACKREF_REV_MASK (((u64)BTRFS_BACKREF_REV_MAX - 1) << \
800 BTRFS_BACKREF_REV_SHIFT)
801
802#define BTRFS_OLD_BACKREF_REV 0
803#define BTRFS_MIXED_BACKREF_REV 1
804
805/*
806 * this flag is only used internally by scrub and may be changed at any time
807 * it is only declared here to avoid collisions
808 */
809#define BTRFS_EXTENT_FLAG_SUPER (1ULL << 48)
810
811struct btrfs_tree_block_info {
812 struct btrfs_disk_key key;
813 __u8 level;
814} __attribute__ ((__packed__));
815
816struct btrfs_extent_data_ref {
817 __le64 root;
818 __le64 objectid;
819 __le64 offset;
820 __le32 count;
821} __attribute__ ((__packed__));
822
823struct btrfs_shared_data_ref {
824 __le32 count;
825} __attribute__ ((__packed__));
826
827struct btrfs_extent_owner_ref {
828 __le64 root_id;
829} __attribute__ ((__packed__));
830
831struct btrfs_extent_inline_ref {
832 __u8 type;
833 __le64 offset;
834} __attribute__ ((__packed__));
835
836/* dev extents record free space on individual devices. The owner
837 * field points back to the chunk allocation mapping tree that allocated
838 * the extent. The chunk tree uuid field is a way to double check the owner
839 */
840struct btrfs_dev_extent {
841 __le64 chunk_tree;
842 __le64 chunk_objectid;
843 __le64 chunk_offset;
844 __le64 length;
845 __u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
846} __attribute__ ((__packed__));
847
848struct btrfs_inode_ref {
849 __le64 index;
850 __le16 name_len;
851 /* name goes here */
852} __attribute__ ((__packed__));
853
854struct btrfs_inode_extref {
855 __le64 parent_objectid;
856 __le64 index;
857 __le16 name_len;
858 __u8 name[];
859 /* name goes here */
860} __attribute__ ((__packed__));
861
862struct btrfs_timespec {
863 __le64 sec;
864 __le32 nsec;
865} __attribute__ ((__packed__));
866
867struct btrfs_inode_item {
868 /* nfs style generation number */
869 __le64 generation;
870 /* transid that last touched this inode */
871 __le64 transid;
872 __le64 size;
873 __le64 nbytes;
874 __le64 block_group;
875 __le32 nlink;
876 __le32 uid;
877 __le32 gid;
878 __le32 mode;
879 __le64 rdev;
880 __le64 flags;
881
882 /* modification sequence number for NFS */
883 __le64 sequence;
884
885 /*
886 * a little future expansion, for more than this we can
887 * just grow the inode item and version it
888 */
889 __le64 reserved[4];
890 struct btrfs_timespec atime;
891 struct btrfs_timespec ctime;
892 struct btrfs_timespec mtime;
893 struct btrfs_timespec otime;
894} __attribute__ ((__packed__));
895
896struct btrfs_dir_log_item {
897 __le64 end;
898} __attribute__ ((__packed__));
899
900struct btrfs_dir_item {
901 struct btrfs_disk_key location;
902 __le64 transid;
903 __le16 data_len;
904 __le16 name_len;
905 __u8 type;
906} __attribute__ ((__packed__));
907
908#define BTRFS_ROOT_SUBVOL_RDONLY (1ULL << 0)
909
910/*
911 * Internal in-memory flag that a subvolume has been marked for deletion but
912 * still visible as a directory
913 */
914#define BTRFS_ROOT_SUBVOL_DEAD (1ULL << 48)
915
916struct btrfs_root_item {
917 struct btrfs_inode_item inode;
918 __le64 generation;
919 __le64 root_dirid;
920 __le64 bytenr;
921 __le64 byte_limit;
922 __le64 bytes_used;
923 __le64 last_snapshot;
924 __le64 flags;
925 __le32 refs;
926 struct btrfs_disk_key drop_progress;
927 __u8 drop_level;
928 __u8 level;
929
930 /*
931 * The following fields appear after subvol_uuids+subvol_times
932 * were introduced.
933 */
934
935 /*
936 * This generation number is used to test if the new fields are valid
937 * and up to date while reading the root item. Every time the root item
938 * is written out, the "generation" field is copied into this field. If
939 * anyone ever mounted the fs with an older kernel, we will have
940 * mismatching generation values here and thus must invalidate the
941 * new fields. See btrfs_update_root and btrfs_find_last_root for
942 * details.
943 * the offset of generation_v2 is also used as the start for the memset
944 * when invalidating the fields.
945 */
946 __le64 generation_v2;
947 __u8 uuid[BTRFS_UUID_SIZE];
948 __u8 parent_uuid[BTRFS_UUID_SIZE];
949 __u8 received_uuid[BTRFS_UUID_SIZE];
950 __le64 ctransid; /* updated when an inode changes */
951 __le64 otransid; /* trans when created */
952 __le64 stransid; /* trans when sent. non-zero for received subvol */
953 __le64 rtransid; /* trans when received. non-zero for received subvol */
954 struct btrfs_timespec ctime;
955 struct btrfs_timespec otime;
956 struct btrfs_timespec stime;
957 struct btrfs_timespec rtime;
958 __le64 reserved[8]; /* for future */
959} __attribute__ ((__packed__));
960
961/*
962 * Btrfs root item used to be smaller than current size. The old format ends
963 * at where member generation_v2 is.
964 */
965static __inline__ __u32 btrfs_legacy_root_item_size(void)
966{
967 return offsetof(struct btrfs_root_item, generation_v2);
968}
969
970/*
971 * this is used for both forward and backward root refs
972 */
973struct btrfs_root_ref {
974 __le64 dirid;
975 __le64 sequence;
976 __le16 name_len;
977} __attribute__ ((__packed__));
978
979struct btrfs_disk_balance_args {
980 /*
981 * profiles to operate on, single is denoted by
982 * BTRFS_AVAIL_ALLOC_BIT_SINGLE
983 */
984 __le64 profiles;
985
986 /*
987 * usage filter
988 * BTRFS_BALANCE_ARGS_USAGE with a single value means '0..N'
989 * BTRFS_BALANCE_ARGS_USAGE_RANGE - range syntax, min..max
990 */
991 union {
992 __le64 usage;
993 struct {
994 __le32 usage_min;
995 __le32 usage_max;
996 };
997 };
998
999 /* devid filter */
1000 __le64 devid;
1001
1002 /* devid subset filter [pstart..pend) */
1003 __le64 pstart;
1004 __le64 pend;
1005
1006 /* btrfs virtual address space subset filter [vstart..vend) */
1007 __le64 vstart;
1008 __le64 vend;
1009
1010 /*
1011 * profile to convert to, single is denoted by
1012 * BTRFS_AVAIL_ALLOC_BIT_SINGLE
1013 */
1014 __le64 target;
1015
1016 /* BTRFS_BALANCE_ARGS_* */
1017 __le64 flags;
1018
1019 /*
1020 * BTRFS_BALANCE_ARGS_LIMIT with value 'limit'
1021 * BTRFS_BALANCE_ARGS_LIMIT_RANGE - the extend version can use minimum
1022 * and maximum
1023 */
1024 union {
1025 __le64 limit;
1026 struct {
1027 __le32 limit_min;
1028 __le32 limit_max;
1029 };
1030 };
1031
1032 /*
1033 * Process chunks that cross stripes_min..stripes_max devices,
1034 * BTRFS_BALANCE_ARGS_STRIPES_RANGE
1035 */
1036 __le32 stripes_min;
1037 __le32 stripes_max;
1038
1039 __le64 unused[6];
1040} __attribute__ ((__packed__));
1041
1042/*
1043 * store balance parameters to disk so that balance can be properly
1044 * resumed after crash or unmount
1045 */
1046struct btrfs_balance_item {
1047 /* BTRFS_BALANCE_* */
1048 __le64 flags;
1049
1050 struct btrfs_disk_balance_args data;
1051 struct btrfs_disk_balance_args meta;
1052 struct btrfs_disk_balance_args sys;
1053
1054 __le64 unused[4];
1055} __attribute__ ((__packed__));
1056
1057enum {
1058 BTRFS_FILE_EXTENT_INLINE = 0,
1059 BTRFS_FILE_EXTENT_REG = 1,
1060 BTRFS_FILE_EXTENT_PREALLOC = 2,
1061 BTRFS_NR_FILE_EXTENT_TYPES = 3,
1062};
1063
1064struct btrfs_file_extent_item {
1065 /*
1066 * transaction id that created this extent
1067 */
1068 __le64 generation;
1069 /*
1070 * max number of bytes to hold this extent in ram
1071 * when we split a compressed extent we can't know how big
1072 * each of the resulting pieces will be. So, this is
1073 * an upper limit on the size of the extent in ram instead of
1074 * an exact limit.
1075 */
1076 __le64 ram_bytes;
1077
1078 /*
1079 * 32 bits for the various ways we might encode the data,
1080 * including compression and encryption. If any of these
1081 * are set to something a given disk format doesn't understand
1082 * it is treated like an incompat flag for reading and writing,
1083 * but not for stat.
1084 */
1085 __u8 compression;
1086 __u8 encryption;
1087 __le16 other_encoding; /* spare for later use */
1088
1089 /* are we __inline__ data or a real extent? */
1090 __u8 type;
1091
1092 /*
1093 * disk space consumed by the extent, checksum blocks are included
1094 * in these numbers
1095 *
1096 * At this offset in the structure, the __inline__ extent data start.
1097 */
1098 __le64 disk_bytenr;
1099 __le64 disk_num_bytes;
1100 /*
1101 * the logical offset in file blocks (no csums)
1102 * this extent record is for. This allows a file extent to point
1103 * into the middle of an existing extent on disk, sharing it
1104 * between two snapshots (useful if some bytes in the middle of the
1105 * extent have changed
1106 */
1107 __le64 offset;
1108 /*
1109 * the logical number of file blocks (no csums included). This
1110 * always reflects the size uncompressed and without encoding.
1111 */
1112 __le64 num_bytes;
1113
1114} __attribute__ ((__packed__));
1115
1116struct btrfs_csum_item {
1117 __u8 csum;
1118} __attribute__ ((__packed__));
1119
1120struct btrfs_dev_stats_item {
1121 /*
1122 * grow this item struct at the end for future enhancements and keep
1123 * the existing values unchanged
1124 */
1125 __le64 values[BTRFS_DEV_STAT_VALUES_MAX];
1126} __attribute__ ((__packed__));
1127
1128#define BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_ALWAYS 0
1129#define BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_AVOID 1
1130
1131struct btrfs_dev_replace_item {
1132 /*
1133 * grow this item struct at the end for future enhancements and keep
1134 * the existing values unchanged
1135 */
1136 __le64 src_devid;
1137 __le64 cursor_left;
1138 __le64 cursor_right;
1139 __le64 cont_reading_from_srcdev_mode;
1140
1141 __le64 replace_state;
1142 __le64 time_started;
1143 __le64 time_stopped;
1144 __le64 num_write_errors;
1145 __le64 num_uncorrectable_read_errors;
1146} __attribute__ ((__packed__));
1147
1148/* different types of block groups (and chunks) */
1149#define BTRFS_BLOCK_GROUP_DATA (1ULL << 0)
1150#define BTRFS_BLOCK_GROUP_SYSTEM (1ULL << 1)
1151#define BTRFS_BLOCK_GROUP_METADATA (1ULL << 2)
1152#define BTRFS_BLOCK_GROUP_RAID0 (1ULL << 3)
1153#define BTRFS_BLOCK_GROUP_RAID1 (1ULL << 4)
1154#define BTRFS_BLOCK_GROUP_DUP (1ULL << 5)
1155#define BTRFS_BLOCK_GROUP_RAID10 (1ULL << 6)
1156#define BTRFS_BLOCK_GROUP_RAID5 (1ULL << 7)
1157#define BTRFS_BLOCK_GROUP_RAID6 (1ULL << 8)
1158#define BTRFS_BLOCK_GROUP_RAID1C3 (1ULL << 9)
1159#define BTRFS_BLOCK_GROUP_RAID1C4 (1ULL << 10)
1160#define BTRFS_BLOCK_GROUP_RESERVED (BTRFS_AVAIL_ALLOC_BIT_SINGLE | \
1161 BTRFS_SPACE_INFO_GLOBAL_RSV)
1162
1163#define BTRFS_BLOCK_GROUP_TYPE_MASK (BTRFS_BLOCK_GROUP_DATA | \
1164 BTRFS_BLOCK_GROUP_SYSTEM | \
1165 BTRFS_BLOCK_GROUP_METADATA)
1166
1167#define BTRFS_BLOCK_GROUP_PROFILE_MASK (BTRFS_BLOCK_GROUP_RAID0 | \
1168 BTRFS_BLOCK_GROUP_RAID1 | \
1169 BTRFS_BLOCK_GROUP_RAID1C3 | \
1170 BTRFS_BLOCK_GROUP_RAID1C4 | \
1171 BTRFS_BLOCK_GROUP_RAID5 | \
1172 BTRFS_BLOCK_GROUP_RAID6 | \
1173 BTRFS_BLOCK_GROUP_DUP | \
1174 BTRFS_BLOCK_GROUP_RAID10)
1175#define BTRFS_BLOCK_GROUP_RAID56_MASK (BTRFS_BLOCK_GROUP_RAID5 | \
1176 BTRFS_BLOCK_GROUP_RAID6)
1177
1178#define BTRFS_BLOCK_GROUP_RAID1_MASK (BTRFS_BLOCK_GROUP_RAID1 | \
1179 BTRFS_BLOCK_GROUP_RAID1C3 | \
1180 BTRFS_BLOCK_GROUP_RAID1C4)
1181
1182/*
1183 * We need a bit for restriper to be able to tell when chunks of type
1184 * SINGLE are available. This "extended" profile format is used in
1185 * fs_info->avail_*_alloc_bits (in-memory) and balance item fields
1186 * (on-disk). The corresponding on-disk bit in chunk.type is reserved
1187 * to avoid remappings between two formats in future.
1188 */
1189#define BTRFS_AVAIL_ALLOC_BIT_SINGLE (1ULL << 48)
1190
1191/*
1192 * A fake block group type that is used to communicate global block reserve
1193 * size to userspace via the SPACE_INFO ioctl.
1194 */
1195#define BTRFS_SPACE_INFO_GLOBAL_RSV (1ULL << 49)
1196
1197#define BTRFS_EXTENDED_PROFILE_MASK (BTRFS_BLOCK_GROUP_PROFILE_MASK | \
1198 BTRFS_AVAIL_ALLOC_BIT_SINGLE)
1199
1200static __inline__ __u64 chunk_to_extended(__u64 flags)
1201{
1202 if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0)
1203 flags |= BTRFS_AVAIL_ALLOC_BIT_SINGLE;
1204
1205 return flags;
1206}
1207static __inline__ __u64 extended_to_chunk(__u64 flags)
1208{
1209 return flags & ~BTRFS_AVAIL_ALLOC_BIT_SINGLE;
1210}
1211
1212struct btrfs_block_group_item {
1213 __le64 used;
1214 __le64 chunk_objectid;
1215 __le64 flags;
1216} __attribute__ ((__packed__));
1217
1218struct btrfs_free_space_info {
1219 __le32 extent_count;
1220 __le32 flags;
1221} __attribute__ ((__packed__));
1222
1223#define BTRFS_FREE_SPACE_USING_BITMAPS (1ULL << 0)
1224
1225#define BTRFS_QGROUP_LEVEL_SHIFT 48
1226static __inline__ __u16 btrfs_qgroup_level(__u64 qgroupid)
1227{
1228 return (__u16)(qgroupid >> BTRFS_QGROUP_LEVEL_SHIFT);
1229}
1230
1231/*
1232 * is subvolume quota turned on?
1233 */
1234#define BTRFS_QGROUP_STATUS_FLAG_ON (1ULL << 0)
1235/*
1236 * RESCAN is set during the initialization phase
1237 */
1238#define BTRFS_QGROUP_STATUS_FLAG_RESCAN (1ULL << 1)
1239/*
1240 * Some qgroup entries are known to be out of date,
1241 * either because the configuration has changed in a way that
1242 * makes a rescan necessary, or because the fs has been mounted
1243 * with a non-qgroup-aware version.
1244 * Turning qouta off and on again makes it inconsistent, too.
1245 */
1246#define BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT (1ULL << 2)
1247
1248/*
1249 * Whether or not this filesystem is using simple quotas. Not exactly the
1250 * incompat bit, because we support using simple quotas, disabling it, then
1251 * going back to full qgroup quotas.
1252 */
1253#define BTRFS_QGROUP_STATUS_FLAG_SIMPLE_MODE (1ULL << 3)
1254
1255#define BTRFS_QGROUP_STATUS_FLAGS_MASK (BTRFS_QGROUP_STATUS_FLAG_ON | \
1256 BTRFS_QGROUP_STATUS_FLAG_RESCAN | \
1257 BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT | \
1258 BTRFS_QGROUP_STATUS_FLAG_SIMPLE_MODE)
1259
1260#define BTRFS_QGROUP_STATUS_VERSION 1
1261
1262struct btrfs_qgroup_status_item {
1263 __le64 version;
1264 /*
1265 * the generation is updated during every commit. As older
1266 * versions of btrfs are not aware of qgroups, it will be
1267 * possible to detect inconsistencies by checking the
1268 * generation on mount time
1269 */
1270 __le64 generation;
1271
1272 /* flag definitions see above */
1273 __le64 flags;
1274
1275 /*
1276 * only used during scanning to record the progress
1277 * of the scan. It contains a logical address
1278 */
1279 __le64 rescan;
1280
1281 /*
1282 * The generation when quotas were last enabled. Used by simple quotas to
1283 * avoid decrementing when freeing an extent that was written before
1284 * enable.
1285 *
1286 * Set only if flags contain BTRFS_QGROUP_STATUS_FLAG_SIMPLE_MODE.
1287 */
1288 __le64 enable_gen;
1289} __attribute__ ((__packed__));
1290
1291struct btrfs_qgroup_info_item {
1292 __le64 generation;
1293 __le64 rfer;
1294 __le64 rfer_cmpr;
1295 __le64 excl;
1296 __le64 excl_cmpr;
1297} __attribute__ ((__packed__));
1298
1299struct btrfs_qgroup_limit_item {
1300 /*
1301 * only updated when any of the other values change
1302 */
1303 __le64 flags;
1304 __le64 max_rfer;
1305 __le64 max_excl;
1306 __le64 rsv_rfer;
1307 __le64 rsv_excl;
1308} __attribute__ ((__packed__));
1309
1310struct btrfs_verity_descriptor_item {
1311 /* Size of the verity descriptor in bytes */
1312 __le64 size;
1313 /*
1314 * When we implement support for fscrypt, we will need to encrypt the
1315 * Merkle tree for encrypted verity files. These 128 bits are for the
1316 * eventual storage of an fscrypt initialization vector.
1317 */
1318 __le64 reserved[2];
1319 __u8 encryption;
1320} __attribute__ ((__packed__));
1321
1322#endif /* _BTRFS_CTREE_H_ */