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#ifdef __KERNEL__ 8#include <linux/stddef.h> 9#else 10#include <stddef.h> 11#endif 12 13/* ASCII for _BHRfS_M, no terminating nul */ 14#define BTRFS_MAGIC 0x4D5F53665248425FULL 15 16#define BTRFS_MAX_LEVEL 8 17 18/* 19 * We can actually store much bigger names, but lets not confuse the rest of 20 * linux. 21 */ 22#define BTRFS_NAME_LEN 255 23 24/* 25 * Theoretical limit is larger, but we keep this down to a sane value. That 26 * should limit greatly the possibility of collisions on inode ref items. 27 */ 28#define BTRFS_LINK_MAX 65535U 29 30/* 31 * This header contains the structure definitions and constants used 32 * by file system objects that can be retrieved using 33 * the BTRFS_IOC_SEARCH_TREE ioctl. That means basically anything that 34 * is needed to describe a leaf node's key or item contents. 35 */ 36 37/* holds pointers to all of the tree roots */ 38#define BTRFS_ROOT_TREE_OBJECTID 1ULL 39 40/* stores information about which extents are in use, and reference counts */ 41#define BTRFS_EXTENT_TREE_OBJECTID 2ULL 42 43/* 44 * chunk tree stores translations from logical -> physical block numbering 45 * the super block points to the chunk tree 46 */ 47#define BTRFS_CHUNK_TREE_OBJECTID 3ULL 48 49/* 50 * stores information about which areas of a given device are in use. 51 * one per device. The tree of tree roots points to the device tree 52 */ 53#define BTRFS_DEV_TREE_OBJECTID 4ULL 54 55/* one per subvolume, storing files and directories */ 56#define BTRFS_FS_TREE_OBJECTID 5ULL 57 58/* directory objectid inside the root tree */ 59#define BTRFS_ROOT_TREE_DIR_OBJECTID 6ULL 60 61/* holds checksums of all the data extents */ 62#define BTRFS_CSUM_TREE_OBJECTID 7ULL 63 64/* holds quota configuration and tracking */ 65#define BTRFS_QUOTA_TREE_OBJECTID 8ULL 66 67/* for storing items that use the BTRFS_UUID_KEY* types */ 68#define BTRFS_UUID_TREE_OBJECTID 9ULL 69 70/* tracks free space in block groups. */ 71#define BTRFS_FREE_SPACE_TREE_OBJECTID 10ULL 72 73/* Holds the block group items for extent tree v2. */ 74#define BTRFS_BLOCK_GROUP_TREE_OBJECTID 11ULL 75 76/* Tracks RAID stripes in block groups. */ 77#define BTRFS_RAID_STRIPE_TREE_OBJECTID 12ULL 78 79/* device stats in the device tree */ 80#define BTRFS_DEV_STATS_OBJECTID 0ULL 81 82/* for storing balance parameters in the root tree */ 83#define BTRFS_BALANCE_OBJECTID -4ULL 84 85/* orphan objectid for tracking unlinked/truncated files */ 86#define BTRFS_ORPHAN_OBJECTID -5ULL 87 88/* does write ahead logging to speed up fsyncs */ 89#define BTRFS_TREE_LOG_OBJECTID -6ULL 90#define BTRFS_TREE_LOG_FIXUP_OBJECTID -7ULL 91 92/* for space balancing */ 93#define BTRFS_TREE_RELOC_OBJECTID -8ULL 94#define BTRFS_DATA_RELOC_TREE_OBJECTID -9ULL 95 96/* 97 * extent checksums all have this objectid 98 * this allows them to share the logging tree 99 * for fsyncs 100 */ 101#define BTRFS_EXTENT_CSUM_OBJECTID -10ULL 102 103/* For storing free space cache */ 104#define BTRFS_FREE_SPACE_OBJECTID -11ULL 105 106/* 107 * The inode number assigned to the special inode for storing 108 * free ino cache 109 */ 110#define BTRFS_FREE_INO_OBJECTID -12ULL 111 112/* dummy objectid represents multiple objectids */ 113#define BTRFS_MULTIPLE_OBJECTIDS -255ULL 114 115/* 116 * All files have objectids in this range. 117 */ 118#define BTRFS_FIRST_FREE_OBJECTID 256ULL 119#define BTRFS_LAST_FREE_OBJECTID -256ULL 120#define BTRFS_FIRST_CHUNK_TREE_OBJECTID 256ULL 121 122 123/* 124 * the device items go into the chunk tree. The key is in the form 125 * [ 1 BTRFS_DEV_ITEM_KEY device_id ] 126 */ 127#define BTRFS_DEV_ITEMS_OBJECTID 1ULL 128 129#define BTRFS_BTREE_INODE_OBJECTID 1 130 131#define BTRFS_EMPTY_SUBVOL_DIR_OBJECTID 2 132 133#define BTRFS_DEV_REPLACE_DEVID 0ULL 134 135/* 136 * inode items have the data typically returned from stat and store other 137 * info about object characteristics. There is one for every file and dir in 138 * the FS 139 */ 140#define BTRFS_INODE_ITEM_KEY 1 141#define BTRFS_INODE_REF_KEY 12 142#define BTRFS_INODE_EXTREF_KEY 13 143#define BTRFS_XATTR_ITEM_KEY 24 144 145/* 146 * fs verity items are stored under two different key types on disk. 147 * The descriptor items: 148 * [ inode objectid, BTRFS_VERITY_DESC_ITEM_KEY, offset ] 149 * 150 * At offset 0, we store a btrfs_verity_descriptor_item which tracks the size 151 * of the descriptor item and some extra data for encryption. 152 * Starting at offset 1, these hold the generic fs verity descriptor. The 153 * latter are opaque to btrfs, we just read and write them as a blob for the 154 * higher level verity code. The most common descriptor size is 256 bytes. 155 * 156 * The merkle tree items: 157 * [ inode objectid, BTRFS_VERITY_MERKLE_ITEM_KEY, offset ] 158 * 159 * These also start at offset 0, and correspond to the merkle tree bytes. When 160 * fsverity asks for page 0 of the merkle tree, we pull up one page starting at 161 * offset 0 for this key type. These are also opaque to btrfs, we're blindly 162 * storing whatever fsverity sends down. 163 */ 164#define BTRFS_VERITY_DESC_ITEM_KEY 36 165#define BTRFS_VERITY_MERKLE_ITEM_KEY 37 166 167#define BTRFS_ORPHAN_ITEM_KEY 48 168/* reserve 2-15 close to the inode for later flexibility */ 169 170/* 171 * dir items are the name -> inode pointers in a directory. There is one 172 * for every name in a directory. BTRFS_DIR_LOG_ITEM_KEY is no longer used 173 * but it's still defined here for documentation purposes and to help avoid 174 * having its numerical value reused in the future. 175 */ 176#define BTRFS_DIR_LOG_ITEM_KEY 60 177#define BTRFS_DIR_LOG_INDEX_KEY 72 178#define BTRFS_DIR_ITEM_KEY 84 179#define BTRFS_DIR_INDEX_KEY 96 180/* 181 * extent data is for file data 182 */ 183#define BTRFS_EXTENT_DATA_KEY 108 184 185/* 186 * extent csums are stored in a separate tree and hold csums for 187 * an entire extent on disk. 188 */ 189#define BTRFS_EXTENT_CSUM_KEY 128 190 191/* 192 * root items point to tree roots. They are typically in the root 193 * tree used by the super block to find all the other trees 194 */ 195#define BTRFS_ROOT_ITEM_KEY 132 196 197/* 198 * root backrefs tie subvols and snapshots to the directory entries that 199 * reference them 200 */ 201#define BTRFS_ROOT_BACKREF_KEY 144 202 203/* 204 * root refs make a fast index for listing all of the snapshots and 205 * subvolumes referenced by a given root. They point directly to the 206 * directory item in the root that references the subvol 207 */ 208#define BTRFS_ROOT_REF_KEY 156 209 210/* 211 * extent items are in the extent map tree. These record which blocks 212 * are used, and how many references there are to each block 213 */ 214#define BTRFS_EXTENT_ITEM_KEY 168 215 216/* 217 * The same as the BTRFS_EXTENT_ITEM_KEY, except it's metadata we already know 218 * the length, so we save the level in key->offset instead of the length. 219 */ 220#define BTRFS_METADATA_ITEM_KEY 169 221 222/* 223 * Special inline ref key which stores the id of the subvolume which originally 224 * created the extent. This subvolume owns the extent permanently from the 225 * perspective of simple quotas. Needed to know which subvolume to free quota 226 * usage from when the extent is deleted. 227 * 228 * Stored as an inline ref rather to avoid wasting space on a separate item on 229 * top of the existing extent item. However, unlike the other inline refs, 230 * there is one one owner ref per extent rather than one per extent. 231 * 232 * Because of this, it goes at the front of the list of inline refs, and thus 233 * must have a lower type value than any other inline ref type (to satisfy the 234 * disk format rule that inline refs have non-decreasing type). 235 */ 236#define BTRFS_EXTENT_OWNER_REF_KEY 172 237 238#define BTRFS_TREE_BLOCK_REF_KEY 176 239 240#define BTRFS_EXTENT_DATA_REF_KEY 178 241 242/* 243 * Obsolete key. Defintion removed in 6.6, value may be reused in the future. 244 * 245 * #define BTRFS_EXTENT_REF_V0_KEY 180 246 */ 247 248#define BTRFS_SHARED_BLOCK_REF_KEY 182 249 250#define BTRFS_SHARED_DATA_REF_KEY 184 251 252/* 253 * block groups give us hints into the extent allocation trees. Which 254 * blocks are free etc etc 255 */ 256#define BTRFS_BLOCK_GROUP_ITEM_KEY 192 257 258/* 259 * Every block group is represented in the free space tree by a free space info 260 * item, which stores some accounting information. It is keyed on 261 * (block_group_start, FREE_SPACE_INFO, block_group_length). 262 */ 263#define BTRFS_FREE_SPACE_INFO_KEY 198 264 265/* 266 * A free space extent tracks an extent of space that is free in a block group. 267 * It is keyed on (start, FREE_SPACE_EXTENT, length). 268 */ 269#define BTRFS_FREE_SPACE_EXTENT_KEY 199 270 271/* 272 * When a block group becomes very fragmented, we convert it to use bitmaps 273 * instead of extents. A free space bitmap is keyed on 274 * (start, FREE_SPACE_BITMAP, length); the corresponding item is a bitmap with 275 * (length / sectorsize) bits. 276 */ 277#define BTRFS_FREE_SPACE_BITMAP_KEY 200 278 279#define BTRFS_DEV_EXTENT_KEY 204 280#define BTRFS_DEV_ITEM_KEY 216 281#define BTRFS_CHUNK_ITEM_KEY 228 282 283#define BTRFS_RAID_STRIPE_KEY 230 284 285/* 286 * Records the overall state of the qgroups. 287 * There's only one instance of this key present, 288 * (0, BTRFS_QGROUP_STATUS_KEY, 0) 289 */ 290#define BTRFS_QGROUP_STATUS_KEY 240 291/* 292 * Records the currently used space of the qgroup. 293 * One key per qgroup, (0, BTRFS_QGROUP_INFO_KEY, qgroupid). 294 */ 295#define BTRFS_QGROUP_INFO_KEY 242 296/* 297 * Contains the user configured limits for the qgroup. 298 * One key per qgroup, (0, BTRFS_QGROUP_LIMIT_KEY, qgroupid). 299 */ 300#define BTRFS_QGROUP_LIMIT_KEY 244 301/* 302 * Records the child-parent relationship of qgroups. For 303 * each relation, 2 keys are present: 304 * (childid, BTRFS_QGROUP_RELATION_KEY, parentid) 305 * (parentid, BTRFS_QGROUP_RELATION_KEY, childid) 306 */ 307#define BTRFS_QGROUP_RELATION_KEY 246 308 309/* 310 * Obsolete name, see BTRFS_TEMPORARY_ITEM_KEY. 311 */ 312#define BTRFS_BALANCE_ITEM_KEY 248 313 314/* 315 * The key type for tree items that are stored persistently, but do not need to 316 * exist for extended period of time. The items can exist in any tree. 317 * 318 * [subtype, BTRFS_TEMPORARY_ITEM_KEY, data] 319 * 320 * Existing items: 321 * 322 * - balance status item 323 * (BTRFS_BALANCE_OBJECTID, BTRFS_TEMPORARY_ITEM_KEY, 0) 324 */ 325#define BTRFS_TEMPORARY_ITEM_KEY 248 326 327/* 328 * Obsolete name, see BTRFS_PERSISTENT_ITEM_KEY 329 */ 330#define BTRFS_DEV_STATS_KEY 249 331 332/* 333 * The key type for tree items that are stored persistently and usually exist 334 * for a long period, eg. filesystem lifetime. The item kinds can be status 335 * information, stats or preference values. The item can exist in any tree. 336 * 337 * [subtype, BTRFS_PERSISTENT_ITEM_KEY, data] 338 * 339 * Existing items: 340 * 341 * - device statistics, store IO stats in the device tree, one key for all 342 * stats 343 * (BTRFS_DEV_STATS_OBJECTID, BTRFS_DEV_STATS_KEY, 0) 344 */ 345#define BTRFS_PERSISTENT_ITEM_KEY 249 346 347/* 348 * Persistently stores the device replace state in the device tree. 349 * The key is built like this: (0, BTRFS_DEV_REPLACE_KEY, 0). 350 */ 351#define BTRFS_DEV_REPLACE_KEY 250 352 353/* 354 * Stores items that allow to quickly map UUIDs to something else. 355 * These items are part of the filesystem UUID tree. 356 * The key is built like this: 357 * (UUID_upper_64_bits, BTRFS_UUID_KEY*, UUID_lower_64_bits). 358 */ 359#if BTRFS_UUID_SIZE != 16 360#error "UUID items require BTRFS_UUID_SIZE == 16!" 361#endif 362#define BTRFS_UUID_KEY_SUBVOL 251 /* for UUIDs assigned to subvols */ 363#define BTRFS_UUID_KEY_RECEIVED_SUBVOL 252 /* for UUIDs assigned to 364 * received subvols */ 365 366/* 367 * string items are for debugging. They just store a short string of 368 * data in the FS 369 */ 370#define BTRFS_STRING_ITEM_KEY 253 371 372/* Maximum metadata block size (nodesize) */ 373#define BTRFS_MAX_METADATA_BLOCKSIZE 65536 374 375/* 32 bytes in various csum fields */ 376#define BTRFS_CSUM_SIZE 32 377 378/* csum types */ 379enum btrfs_csum_type { 380 BTRFS_CSUM_TYPE_CRC32 = 0, 381 BTRFS_CSUM_TYPE_XXHASH = 1, 382 BTRFS_CSUM_TYPE_SHA256 = 2, 383 BTRFS_CSUM_TYPE_BLAKE2 = 3, 384}; 385 386/* 387 * flags definitions for directory entry item type 388 * 389 * Used by: 390 * struct btrfs_dir_item.type 391 * 392 * Values 0..7 must match common file type values in fs_types.h. 393 */ 394#define BTRFS_FT_UNKNOWN 0 395#define BTRFS_FT_REG_FILE 1 396#define BTRFS_FT_DIR 2 397#define BTRFS_FT_CHRDEV 3 398#define BTRFS_FT_BLKDEV 4 399#define BTRFS_FT_FIFO 5 400#define BTRFS_FT_SOCK 6 401#define BTRFS_FT_SYMLINK 7 402#define BTRFS_FT_XATTR 8 403#define BTRFS_FT_MAX 9 404/* Directory contains encrypted data */ 405#define BTRFS_FT_ENCRYPTED 0x80 406 407static inline __u8 btrfs_dir_flags_to_ftype(__u8 flags) 408{ 409 return flags & ~BTRFS_FT_ENCRYPTED; 410} 411 412/* 413 * Inode flags 414 */ 415#define BTRFS_INODE_NODATASUM (1U << 0) 416#define BTRFS_INODE_NODATACOW (1U << 1) 417#define BTRFS_INODE_READONLY (1U << 2) 418#define BTRFS_INODE_NOCOMPRESS (1U << 3) 419#define BTRFS_INODE_PREALLOC (1U << 4) 420#define BTRFS_INODE_SYNC (1U << 5) 421#define BTRFS_INODE_IMMUTABLE (1U << 6) 422#define BTRFS_INODE_APPEND (1U << 7) 423#define BTRFS_INODE_NODUMP (1U << 8) 424#define BTRFS_INODE_NOATIME (1U << 9) 425#define BTRFS_INODE_DIRSYNC (1U << 10) 426#define BTRFS_INODE_COMPRESS (1U << 11) 427 428#define BTRFS_INODE_ROOT_ITEM_INIT (1U << 31) 429 430#define BTRFS_INODE_FLAG_MASK \ 431 (BTRFS_INODE_NODATASUM | \ 432 BTRFS_INODE_NODATACOW | \ 433 BTRFS_INODE_READONLY | \ 434 BTRFS_INODE_NOCOMPRESS | \ 435 BTRFS_INODE_PREALLOC | \ 436 BTRFS_INODE_SYNC | \ 437 BTRFS_INODE_IMMUTABLE | \ 438 BTRFS_INODE_APPEND | \ 439 BTRFS_INODE_NODUMP | \ 440 BTRFS_INODE_NOATIME | \ 441 BTRFS_INODE_DIRSYNC | \ 442 BTRFS_INODE_COMPRESS | \ 443 BTRFS_INODE_ROOT_ITEM_INIT) 444 445#define BTRFS_INODE_RO_VERITY (1U << 0) 446 447#define BTRFS_INODE_RO_FLAG_MASK (BTRFS_INODE_RO_VERITY) 448 449/* 450 * The key defines the order in the tree, and so it also defines (optimal) 451 * block layout. 452 * 453 * objectid corresponds to the inode number. 454 * 455 * type tells us things about the object, and is a kind of stream selector. 456 * so for a given inode, keys with type of 1 might refer to the inode data, 457 * type of 2 may point to file data in the btree and type == 3 may point to 458 * extents. 459 * 460 * offset is the starting byte offset for this key in the stream. 461 * 462 * btrfs_disk_key is in disk byte order. struct btrfs_key is always 463 * in cpu native order. Otherwise they are identical and their sizes 464 * should be the same (ie both packed) 465 */ 466struct btrfs_disk_key { 467 __le64 objectid; 468 __u8 type; 469 __le64 offset; 470} __attribute__ ((__packed__)); 471 472struct btrfs_key { 473 __u64 objectid; 474 __u8 type; 475 __u64 offset; 476} __attribute__ ((__packed__)); 477 478/* 479 * Every tree block (leaf or node) starts with this header. 480 */ 481struct btrfs_header { 482 /* These first four must match the super block */ 483 __u8 csum[BTRFS_CSUM_SIZE]; 484 /* FS specific uuid */ 485 __u8 fsid[BTRFS_FSID_SIZE]; 486 /* Which block this node is supposed to live in */ 487 __le64 bytenr; 488 __le64 flags; 489 490 /* Allowed to be different from the super from here on down */ 491 __u8 chunk_tree_uuid[BTRFS_UUID_SIZE]; 492 __le64 generation; 493 __le64 owner; 494 __le32 nritems; 495 __u8 level; 496} __attribute__ ((__packed__)); 497 498/* 499 * This is a very generous portion of the super block, giving us room to 500 * translate 14 chunks with 3 stripes each. 501 */ 502#define BTRFS_SYSTEM_CHUNK_ARRAY_SIZE 2048 503 504/* 505 * Just in case we somehow lose the roots and are not able to mount, we store 506 * an array of the roots from previous transactions in the super. 507 */ 508#define BTRFS_NUM_BACKUP_ROOTS 4 509struct btrfs_root_backup { 510 __le64 tree_root; 511 __le64 tree_root_gen; 512 513 __le64 chunk_root; 514 __le64 chunk_root_gen; 515 516 __le64 extent_root; 517 __le64 extent_root_gen; 518 519 __le64 fs_root; 520 __le64 fs_root_gen; 521 522 __le64 dev_root; 523 __le64 dev_root_gen; 524 525 __le64 csum_root; 526 __le64 csum_root_gen; 527 528 __le64 total_bytes; 529 __le64 bytes_used; 530 __le64 num_devices; 531 /* future */ 532 __le64 unused_64[4]; 533 534 __u8 tree_root_level; 535 __u8 chunk_root_level; 536 __u8 extent_root_level; 537 __u8 fs_root_level; 538 __u8 dev_root_level; 539 __u8 csum_root_level; 540 /* future and to align */ 541 __u8 unused_8[10]; 542} __attribute__ ((__packed__)); 543 544/* 545 * A leaf is full of items. offset and size tell us where to find the item in 546 * the leaf (relative to the start of the data area) 547 */ 548struct btrfs_item { 549 struct btrfs_disk_key key; 550 __le32 offset; 551 __le32 size; 552} __attribute__ ((__packed__)); 553 554/* 555 * Leaves have an item area and a data area: 556 * [item0, item1....itemN] [free space] [dataN...data1, data0] 557 * 558 * The data is separate from the items to get the keys closer together during 559 * searches. 560 */ 561struct btrfs_leaf { 562 struct btrfs_header header; 563 struct btrfs_item items[]; 564} __attribute__ ((__packed__)); 565 566/* 567 * All non-leaf blocks are nodes, they hold only keys and pointers to other 568 * blocks. 569 */ 570struct btrfs_key_ptr { 571 struct btrfs_disk_key key; 572 __le64 blockptr; 573 __le64 generation; 574} __attribute__ ((__packed__)); 575 576struct btrfs_node { 577 struct btrfs_header header; 578 struct btrfs_key_ptr ptrs[]; 579} __attribute__ ((__packed__)); 580 581struct btrfs_dev_item { 582 /* the internal btrfs device id */ 583 __le64 devid; 584 585 /* size of the device */ 586 __le64 total_bytes; 587 588 /* bytes used */ 589 __le64 bytes_used; 590 591 /* optimal io alignment for this device */ 592 __le32 io_align; 593 594 /* optimal io width for this device */ 595 __le32 io_width; 596 597 /* minimal io size for this device */ 598 __le32 sector_size; 599 600 /* type and info about this device */ 601 __le64 type; 602 603 /* expected generation for this device */ 604 __le64 generation; 605 606 /* 607 * starting byte of this partition on the device, 608 * to allow for stripe alignment in the future 609 */ 610 __le64 start_offset; 611 612 /* grouping information for allocation decisions */ 613 __le32 dev_group; 614 615 /* seek speed 0-100 where 100 is fastest */ 616 __u8 seek_speed; 617 618 /* bandwidth 0-100 where 100 is fastest */ 619 __u8 bandwidth; 620 621 /* btrfs generated uuid for this device */ 622 __u8 uuid[BTRFS_UUID_SIZE]; 623 624 /* uuid of FS who owns this device */ 625 __u8 fsid[BTRFS_UUID_SIZE]; 626} __attribute__ ((__packed__)); 627 628struct btrfs_stripe { 629 __le64 devid; 630 __le64 offset; 631 __u8 dev_uuid[BTRFS_UUID_SIZE]; 632} __attribute__ ((__packed__)); 633 634struct btrfs_chunk { 635 /* size of this chunk in bytes */ 636 __le64 length; 637 638 /* objectid of the root referencing this chunk */ 639 __le64 owner; 640 641 __le64 stripe_len; 642 __le64 type; 643 644 /* optimal io alignment for this chunk */ 645 __le32 io_align; 646 647 /* optimal io width for this chunk */ 648 __le32 io_width; 649 650 /* minimal io size for this chunk */ 651 __le32 sector_size; 652 653 /* 2^16 stripes is quite a lot, a second limit is the size of a single 654 * item in the btree 655 */ 656 __le16 num_stripes; 657 658 /* sub stripes only matter for raid10 */ 659 __le16 sub_stripes; 660 struct btrfs_stripe stripe; 661 /* additional stripes go here */ 662} __attribute__ ((__packed__)); 663 664/* 665 * The super block basically lists the main trees of the FS. 666 */ 667struct btrfs_super_block { 668 /* The first 4 fields must match struct btrfs_header */ 669 __u8 csum[BTRFS_CSUM_SIZE]; 670 /* FS specific UUID, visible to user */ 671 __u8 fsid[BTRFS_FSID_SIZE]; 672 /* This block number */ 673 __le64 bytenr; 674 __le64 flags; 675 676 /* Allowed to be different from the btrfs_header from here own down */ 677 __le64 magic; 678 __le64 generation; 679 __le64 root; 680 __le64 chunk_root; 681 __le64 log_root; 682 683 /* 684 * This member has never been utilized since the very beginning, thus 685 * it's always 0 regardless of kernel version. We always use 686 * generation + 1 to read log tree root. So here we mark it deprecated. 687 */ 688 __le64 __unused_log_root_transid; 689 __le64 total_bytes; 690 __le64 bytes_used; 691 __le64 root_dir_objectid; 692 __le64 num_devices; 693 __le32 sectorsize; 694 __le32 nodesize; 695 __le32 __unused_leafsize; 696 __le32 stripesize; 697 __le32 sys_chunk_array_size; 698 __le64 chunk_root_generation; 699 __le64 compat_flags; 700 __le64 compat_ro_flags; 701 __le64 incompat_flags; 702 __le16 csum_type; 703 __u8 root_level; 704 __u8 chunk_root_level; 705 __u8 log_root_level; 706 struct btrfs_dev_item dev_item; 707 708 char label[BTRFS_LABEL_SIZE]; 709 710 __le64 cache_generation; 711 __le64 uuid_tree_generation; 712 713 /* The UUID written into btree blocks */ 714 __u8 metadata_uuid[BTRFS_FSID_SIZE]; 715 716 __u64 nr_global_roots; 717 718 /* Future expansion */ 719 __le64 reserved[27]; 720 __u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE]; 721 struct btrfs_root_backup super_roots[BTRFS_NUM_BACKUP_ROOTS]; 722 723 /* Padded to 4096 bytes */ 724 __u8 padding[565]; 725} __attribute__ ((__packed__)); 726 727#define BTRFS_FREE_SPACE_EXTENT 1 728#define BTRFS_FREE_SPACE_BITMAP 2 729 730struct btrfs_free_space_entry { 731 __le64 offset; 732 __le64 bytes; 733 __u8 type; 734} __attribute__ ((__packed__)); 735 736struct btrfs_free_space_header { 737 struct btrfs_disk_key location; 738 __le64 generation; 739 __le64 num_entries; 740 __le64 num_bitmaps; 741} __attribute__ ((__packed__)); 742 743struct btrfs_raid_stride { 744 /* The id of device this raid extent lives on. */ 745 __le64 devid; 746 /* The physical location on disk. */ 747 __le64 physical; 748} __attribute__ ((__packed__)); 749 750/* The stripe_extent::encoding, 1:1 mapping of enum btrfs_raid_types. */ 751#define BTRFS_STRIPE_RAID0 1 752#define BTRFS_STRIPE_RAID1 2 753#define BTRFS_STRIPE_DUP 3 754#define BTRFS_STRIPE_RAID10 4 755#define BTRFS_STRIPE_RAID5 5 756#define BTRFS_STRIPE_RAID6 6 757#define BTRFS_STRIPE_RAID1C3 7 758#define BTRFS_STRIPE_RAID1C4 8 759 760struct btrfs_stripe_extent { 761 __u8 encoding; 762 __u8 reserved[7]; 763 /* An array of raid strides this stripe is composed of. */ 764 struct btrfs_raid_stride strides[]; 765} __attribute__ ((__packed__)); 766 767#define BTRFS_HEADER_FLAG_WRITTEN (1ULL << 0) 768#define BTRFS_HEADER_FLAG_RELOC (1ULL << 1) 769 770/* Super block flags */ 771/* Errors detected */ 772#define BTRFS_SUPER_FLAG_ERROR (1ULL << 2) 773 774#define BTRFS_SUPER_FLAG_SEEDING (1ULL << 32) 775#define BTRFS_SUPER_FLAG_METADUMP (1ULL << 33) 776#define BTRFS_SUPER_FLAG_METADUMP_V2 (1ULL << 34) 777#define BTRFS_SUPER_FLAG_CHANGING_FSID (1ULL << 35) 778#define BTRFS_SUPER_FLAG_CHANGING_FSID_V2 (1ULL << 36) 779 780 781/* 782 * items in the extent btree are used to record the objectid of the 783 * owner of the block and the number of references 784 */ 785 786struct btrfs_extent_item { 787 __le64 refs; 788 __le64 generation; 789 __le64 flags; 790} __attribute__ ((__packed__)); 791 792struct btrfs_extent_item_v0 { 793 __le32 refs; 794} __attribute__ ((__packed__)); 795 796 797#define BTRFS_EXTENT_FLAG_DATA (1ULL << 0) 798#define BTRFS_EXTENT_FLAG_TREE_BLOCK (1ULL << 1) 799 800/* following flags only apply to tree blocks */ 801 802/* use full backrefs for extent pointers in the block */ 803#define BTRFS_BLOCK_FLAG_FULL_BACKREF (1ULL << 8) 804 805#define BTRFS_BACKREF_REV_MAX 256 806#define BTRFS_BACKREF_REV_SHIFT 56 807#define BTRFS_BACKREF_REV_MASK (((u64)BTRFS_BACKREF_REV_MAX - 1) << \ 808 BTRFS_BACKREF_REV_SHIFT) 809 810#define BTRFS_OLD_BACKREF_REV 0 811#define BTRFS_MIXED_BACKREF_REV 1 812 813/* 814 * this flag is only used internally by scrub and may be changed at any time 815 * it is only declared here to avoid collisions 816 */ 817#define BTRFS_EXTENT_FLAG_SUPER (1ULL << 48) 818 819struct btrfs_tree_block_info { 820 struct btrfs_disk_key key; 821 __u8 level; 822} __attribute__ ((__packed__)); 823 824struct btrfs_extent_data_ref { 825 __le64 root; 826 __le64 objectid; 827 __le64 offset; 828 __le32 count; 829} __attribute__ ((__packed__)); 830 831struct btrfs_shared_data_ref { 832 __le32 count; 833} __attribute__ ((__packed__)); 834 835struct btrfs_extent_owner_ref { 836 __le64 root_id; 837} __attribute__ ((__packed__)); 838 839struct btrfs_extent_inline_ref { 840 __u8 type; 841 __le64 offset; 842} __attribute__ ((__packed__)); 843 844/* dev extents record free space on individual devices. The owner 845 * field points back to the chunk allocation mapping tree that allocated 846 * the extent. The chunk tree uuid field is a way to double check the owner 847 */ 848struct btrfs_dev_extent { 849 __le64 chunk_tree; 850 __le64 chunk_objectid; 851 __le64 chunk_offset; 852 __le64 length; 853 __u8 chunk_tree_uuid[BTRFS_UUID_SIZE]; 854} __attribute__ ((__packed__)); 855 856struct btrfs_inode_ref { 857 __le64 index; 858 __le16 name_len; 859 /* name goes here */ 860} __attribute__ ((__packed__)); 861 862struct btrfs_inode_extref { 863 __le64 parent_objectid; 864 __le64 index; 865 __le16 name_len; 866 __u8 name[]; 867 /* name goes here */ 868} __attribute__ ((__packed__)); 869 870struct btrfs_timespec { 871 __le64 sec; 872 __le32 nsec; 873} __attribute__ ((__packed__)); 874 875struct btrfs_inode_item { 876 /* nfs style generation number */ 877 __le64 generation; 878 /* transid that last touched this inode */ 879 __le64 transid; 880 __le64 size; 881 __le64 nbytes; 882 __le64 block_group; 883 __le32 nlink; 884 __le32 uid; 885 __le32 gid; 886 __le32 mode; 887 __le64 rdev; 888 __le64 flags; 889 890 /* modification sequence number for NFS */ 891 __le64 sequence; 892 893 /* 894 * a little future expansion, for more than this we can 895 * just grow the inode item and version it 896 */ 897 __le64 reserved[4]; 898 struct btrfs_timespec atime; 899 struct btrfs_timespec ctime; 900 struct btrfs_timespec mtime; 901 struct btrfs_timespec otime; 902} __attribute__ ((__packed__)); 903 904struct btrfs_dir_log_item { 905 __le64 end; 906} __attribute__ ((__packed__)); 907 908struct btrfs_dir_item { 909 struct btrfs_disk_key location; 910 __le64 transid; 911 __le16 data_len; 912 __le16 name_len; 913 __u8 type; 914} __attribute__ ((__packed__)); 915 916#define BTRFS_ROOT_SUBVOL_RDONLY (1ULL << 0) 917 918/* 919 * Internal in-memory flag that a subvolume has been marked for deletion but 920 * still visible as a directory 921 */ 922#define BTRFS_ROOT_SUBVOL_DEAD (1ULL << 48) 923 924struct btrfs_root_item { 925 struct btrfs_inode_item inode; 926 __le64 generation; 927 __le64 root_dirid; 928 __le64 bytenr; 929 __le64 byte_limit; 930 __le64 bytes_used; 931 __le64 last_snapshot; 932 __le64 flags; 933 __le32 refs; 934 struct btrfs_disk_key drop_progress; 935 __u8 drop_level; 936 __u8 level; 937 938 /* 939 * The following fields appear after subvol_uuids+subvol_times 940 * were introduced. 941 */ 942 943 /* 944 * This generation number is used to test if the new fields are valid 945 * and up to date while reading the root item. Every time the root item 946 * is written out, the "generation" field is copied into this field. If 947 * anyone ever mounted the fs with an older kernel, we will have 948 * mismatching generation values here and thus must invalidate the 949 * new fields. See btrfs_update_root and btrfs_find_last_root for 950 * details. 951 * the offset of generation_v2 is also used as the start for the memset 952 * when invalidating the fields. 953 */ 954 __le64 generation_v2; 955 __u8 uuid[BTRFS_UUID_SIZE]; 956 __u8 parent_uuid[BTRFS_UUID_SIZE]; 957 __u8 received_uuid[BTRFS_UUID_SIZE]; 958 __le64 ctransid; /* updated when an inode changes */ 959 __le64 otransid; /* trans when created */ 960 __le64 stransid; /* trans when sent. non-zero for received subvol */ 961 __le64 rtransid; /* trans when received. non-zero for received subvol */ 962 struct btrfs_timespec ctime; 963 struct btrfs_timespec otime; 964 struct btrfs_timespec stime; 965 struct btrfs_timespec rtime; 966 __le64 reserved[8]; /* for future */ 967} __attribute__ ((__packed__)); 968 969/* 970 * Btrfs root item used to be smaller than current size. The old format ends 971 * at where member generation_v2 is. 972 */ 973static inline __u32 btrfs_legacy_root_item_size(void) 974{ 975 return offsetof(struct btrfs_root_item, generation_v2); 976} 977 978/* 979 * this is used for both forward and backward root refs 980 */ 981struct btrfs_root_ref { 982 __le64 dirid; 983 __le64 sequence; 984 __le16 name_len; 985} __attribute__ ((__packed__)); 986 987struct btrfs_disk_balance_args { 988 /* 989 * profiles to operate on, single is denoted by 990 * BTRFS_AVAIL_ALLOC_BIT_SINGLE 991 */ 992 __le64 profiles; 993 994 /* 995 * usage filter 996 * BTRFS_BALANCE_ARGS_USAGE with a single value means '0..N' 997 * BTRFS_BALANCE_ARGS_USAGE_RANGE - range syntax, min..max 998 */ 999 union { 1000 __le64 usage; 1001 struct { 1002 __le32 usage_min; 1003 __le32 usage_max; 1004 }; 1005 }; 1006 1007 /* devid filter */ 1008 __le64 devid; 1009 1010 /* devid subset filter [pstart..pend) */ 1011 __le64 pstart; 1012 __le64 pend; 1013 1014 /* btrfs virtual address space subset filter [vstart..vend) */ 1015 __le64 vstart; 1016 __le64 vend; 1017 1018 /* 1019 * profile to convert to, single is denoted by 1020 * BTRFS_AVAIL_ALLOC_BIT_SINGLE 1021 */ 1022 __le64 target; 1023 1024 /* BTRFS_BALANCE_ARGS_* */ 1025 __le64 flags; 1026 1027 /* 1028 * BTRFS_BALANCE_ARGS_LIMIT with value 'limit' 1029 * BTRFS_BALANCE_ARGS_LIMIT_RANGE - the extend version can use minimum 1030 * and maximum 1031 */ 1032 union { 1033 __le64 limit; 1034 struct { 1035 __le32 limit_min; 1036 __le32 limit_max; 1037 }; 1038 }; 1039 1040 /* 1041 * Process chunks that cross stripes_min..stripes_max devices, 1042 * BTRFS_BALANCE_ARGS_STRIPES_RANGE 1043 */ 1044 __le32 stripes_min; 1045 __le32 stripes_max; 1046 1047 __le64 unused[6]; 1048} __attribute__ ((__packed__)); 1049 1050/* 1051 * store balance parameters to disk so that balance can be properly 1052 * resumed after crash or unmount 1053 */ 1054struct btrfs_balance_item { 1055 /* BTRFS_BALANCE_* */ 1056 __le64 flags; 1057 1058 struct btrfs_disk_balance_args data; 1059 struct btrfs_disk_balance_args meta; 1060 struct btrfs_disk_balance_args sys; 1061 1062 __le64 unused[4]; 1063} __attribute__ ((__packed__)); 1064 1065enum { 1066 BTRFS_FILE_EXTENT_INLINE = 0, 1067 BTRFS_FILE_EXTENT_REG = 1, 1068 BTRFS_FILE_EXTENT_PREALLOC = 2, 1069 BTRFS_NR_FILE_EXTENT_TYPES = 3, 1070}; 1071 1072struct btrfs_file_extent_item { 1073 /* 1074 * transaction id that created this extent 1075 */ 1076 __le64 generation; 1077 /* 1078 * max number of bytes to hold this extent in ram 1079 * when we split a compressed extent we can't know how big 1080 * each of the resulting pieces will be. So, this is 1081 * an upper limit on the size of the extent in ram instead of 1082 * an exact limit. 1083 */ 1084 __le64 ram_bytes; 1085 1086 /* 1087 * 32 bits for the various ways we might encode the data, 1088 * including compression and encryption. If any of these 1089 * are set to something a given disk format doesn't understand 1090 * it is treated like an incompat flag for reading and writing, 1091 * but not for stat. 1092 */ 1093 __u8 compression; 1094 __u8 encryption; 1095 __le16 other_encoding; /* spare for later use */ 1096 1097 /* are we inline data or a real extent? */ 1098 __u8 type; 1099 1100 /* 1101 * disk space consumed by the extent, checksum blocks are included 1102 * in these numbers 1103 * 1104 * At this offset in the structure, the inline extent data start. 1105 */ 1106 __le64 disk_bytenr; 1107 __le64 disk_num_bytes; 1108 /* 1109 * the logical offset in file blocks (no csums) 1110 * this extent record is for. This allows a file extent to point 1111 * into the middle of an existing extent on disk, sharing it 1112 * between two snapshots (useful if some bytes in the middle of the 1113 * extent have changed 1114 */ 1115 __le64 offset; 1116 /* 1117 * the logical number of file blocks (no csums included). This 1118 * always reflects the size uncompressed and without encoding. 1119 */ 1120 __le64 num_bytes; 1121 1122} __attribute__ ((__packed__)); 1123 1124struct btrfs_csum_item { 1125 __u8 csum; 1126} __attribute__ ((__packed__)); 1127 1128struct btrfs_dev_stats_item { 1129 /* 1130 * grow this item struct at the end for future enhancements and keep 1131 * the existing values unchanged 1132 */ 1133 __le64 values[BTRFS_DEV_STAT_VALUES_MAX]; 1134} __attribute__ ((__packed__)); 1135 1136#define BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_ALWAYS 0 1137#define BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_AVOID 1 1138 1139struct btrfs_dev_replace_item { 1140 /* 1141 * grow this item struct at the end for future enhancements and keep 1142 * the existing values unchanged 1143 */ 1144 __le64 src_devid; 1145 __le64 cursor_left; 1146 __le64 cursor_right; 1147 __le64 cont_reading_from_srcdev_mode; 1148 1149 __le64 replace_state; 1150 __le64 time_started; 1151 __le64 time_stopped; 1152 __le64 num_write_errors; 1153 __le64 num_uncorrectable_read_errors; 1154} __attribute__ ((__packed__)); 1155 1156/* different types of block groups (and chunks) */ 1157#define BTRFS_BLOCK_GROUP_DATA (1ULL << 0) 1158#define BTRFS_BLOCK_GROUP_SYSTEM (1ULL << 1) 1159#define BTRFS_BLOCK_GROUP_METADATA (1ULL << 2) 1160#define BTRFS_BLOCK_GROUP_RAID0 (1ULL << 3) 1161#define BTRFS_BLOCK_GROUP_RAID1 (1ULL << 4) 1162#define BTRFS_BLOCK_GROUP_DUP (1ULL << 5) 1163#define BTRFS_BLOCK_GROUP_RAID10 (1ULL << 6) 1164#define BTRFS_BLOCK_GROUP_RAID5 (1ULL << 7) 1165#define BTRFS_BLOCK_GROUP_RAID6 (1ULL << 8) 1166#define BTRFS_BLOCK_GROUP_RAID1C3 (1ULL << 9) 1167#define BTRFS_BLOCK_GROUP_RAID1C4 (1ULL << 10) 1168#define BTRFS_BLOCK_GROUP_RESERVED (BTRFS_AVAIL_ALLOC_BIT_SINGLE | \ 1169 BTRFS_SPACE_INFO_GLOBAL_RSV) 1170 1171#define BTRFS_BLOCK_GROUP_TYPE_MASK (BTRFS_BLOCK_GROUP_DATA | \ 1172 BTRFS_BLOCK_GROUP_SYSTEM | \ 1173 BTRFS_BLOCK_GROUP_METADATA) 1174 1175#define BTRFS_BLOCK_GROUP_PROFILE_MASK (BTRFS_BLOCK_GROUP_RAID0 | \ 1176 BTRFS_BLOCK_GROUP_RAID1 | \ 1177 BTRFS_BLOCK_GROUP_RAID1C3 | \ 1178 BTRFS_BLOCK_GROUP_RAID1C4 | \ 1179 BTRFS_BLOCK_GROUP_RAID5 | \ 1180 BTRFS_BLOCK_GROUP_RAID6 | \ 1181 BTRFS_BLOCK_GROUP_DUP | \ 1182 BTRFS_BLOCK_GROUP_RAID10) 1183#define BTRFS_BLOCK_GROUP_RAID56_MASK (BTRFS_BLOCK_GROUP_RAID5 | \ 1184 BTRFS_BLOCK_GROUP_RAID6) 1185 1186#define BTRFS_BLOCK_GROUP_RAID1_MASK (BTRFS_BLOCK_GROUP_RAID1 | \ 1187 BTRFS_BLOCK_GROUP_RAID1C3 | \ 1188 BTRFS_BLOCK_GROUP_RAID1C4) 1189 1190/* 1191 * We need a bit for restriper to be able to tell when chunks of type 1192 * SINGLE are available. This "extended" profile format is used in 1193 * fs_info->avail_*_alloc_bits (in-memory) and balance item fields 1194 * (on-disk). The corresponding on-disk bit in chunk.type is reserved 1195 * to avoid remappings between two formats in future. 1196 */ 1197#define BTRFS_AVAIL_ALLOC_BIT_SINGLE (1ULL << 48) 1198 1199/* 1200 * A fake block group type that is used to communicate global block reserve 1201 * size to userspace via the SPACE_INFO ioctl. 1202 */ 1203#define BTRFS_SPACE_INFO_GLOBAL_RSV (1ULL << 49) 1204 1205#define BTRFS_EXTENDED_PROFILE_MASK (BTRFS_BLOCK_GROUP_PROFILE_MASK | \ 1206 BTRFS_AVAIL_ALLOC_BIT_SINGLE) 1207 1208static inline __u64 chunk_to_extended(__u64 flags) 1209{ 1210 if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0) 1211 flags |= BTRFS_AVAIL_ALLOC_BIT_SINGLE; 1212 1213 return flags; 1214} 1215static inline __u64 extended_to_chunk(__u64 flags) 1216{ 1217 return flags & ~BTRFS_AVAIL_ALLOC_BIT_SINGLE; 1218} 1219 1220struct btrfs_block_group_item { 1221 __le64 used; 1222 __le64 chunk_objectid; 1223 __le64 flags; 1224} __attribute__ ((__packed__)); 1225 1226struct btrfs_free_space_info { 1227 __le32 extent_count; 1228 __le32 flags; 1229} __attribute__ ((__packed__)); 1230 1231#define BTRFS_FREE_SPACE_USING_BITMAPS (1ULL << 0) 1232 1233#define BTRFS_QGROUP_LEVEL_SHIFT 48 1234static inline __u16 btrfs_qgroup_level(__u64 qgroupid) 1235{ 1236 return (__u16)(qgroupid >> BTRFS_QGROUP_LEVEL_SHIFT); 1237} 1238 1239/* 1240 * is subvolume quota turned on? 1241 */ 1242#define BTRFS_QGROUP_STATUS_FLAG_ON (1ULL << 0) 1243/* 1244 * RESCAN is set during the initialization phase 1245 */ 1246#define BTRFS_QGROUP_STATUS_FLAG_RESCAN (1ULL << 1) 1247/* 1248 * Some qgroup entries are known to be out of date, 1249 * either because the configuration has changed in a way that 1250 * makes a rescan necessary, or because the fs has been mounted 1251 * with a non-qgroup-aware version. 1252 * Turning qouta off and on again makes it inconsistent, too. 1253 */ 1254#define BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT (1ULL << 2) 1255 1256/* 1257 * Whether or not this filesystem is using simple quotas. Not exactly the 1258 * incompat bit, because we support using simple quotas, disabling it, then 1259 * going back to full qgroup quotas. 1260 */ 1261#define BTRFS_QGROUP_STATUS_FLAG_SIMPLE_MODE (1ULL << 3) 1262 1263#define BTRFS_QGROUP_STATUS_FLAGS_MASK (BTRFS_QGROUP_STATUS_FLAG_ON | \ 1264 BTRFS_QGROUP_STATUS_FLAG_RESCAN | \ 1265 BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT | \ 1266 BTRFS_QGROUP_STATUS_FLAG_SIMPLE_MODE) 1267 1268#define BTRFS_QGROUP_STATUS_VERSION 1 1269 1270struct btrfs_qgroup_status_item { 1271 __le64 version; 1272 /* 1273 * the generation is updated during every commit. As older 1274 * versions of btrfs are not aware of qgroups, it will be 1275 * possible to detect inconsistencies by checking the 1276 * generation on mount time 1277 */ 1278 __le64 generation; 1279 1280 /* flag definitions see above */ 1281 __le64 flags; 1282 1283 /* 1284 * only used during scanning to record the progress 1285 * of the scan. It contains a logical address 1286 */ 1287 __le64 rescan; 1288 1289 /* 1290 * The generation when quotas were last enabled. Used by simple quotas to 1291 * avoid decrementing when freeing an extent that was written before 1292 * enable. 1293 * 1294 * Set only if flags contain BTRFS_QGROUP_STATUS_FLAG_SIMPLE_MODE. 1295 */ 1296 __le64 enable_gen; 1297} __attribute__ ((__packed__)); 1298 1299struct btrfs_qgroup_info_item { 1300 __le64 generation; 1301 __le64 rfer; 1302 __le64 rfer_cmpr; 1303 __le64 excl; 1304 __le64 excl_cmpr; 1305} __attribute__ ((__packed__)); 1306 1307struct btrfs_qgroup_limit_item { 1308 /* 1309 * only updated when any of the other values change 1310 */ 1311 __le64 flags; 1312 __le64 max_rfer; 1313 __le64 max_excl; 1314 __le64 rsv_rfer; 1315 __le64 rsv_excl; 1316} __attribute__ ((__packed__)); 1317 1318struct btrfs_verity_descriptor_item { 1319 /* Size of the verity descriptor in bytes */ 1320 __le64 size; 1321 /* 1322 * When we implement support for fscrypt, we will need to encrypt the 1323 * Merkle tree for encrypted verity files. These 128 bits are for the 1324 * eventual storage of an fscrypt initialization vector. 1325 */ 1326 __le64 reserved[2]; 1327 __u8 encryption; 1328} __attribute__ ((__packed__)); 1329 1330#endif /* _BTRFS_CTREE_H_ */ 1331