1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * Copyright (c) 2000-2005 Silicon Graphics, Inc. 4 * All Rights Reserved. 5 */ 6 #ifndef __XFS_FORMAT_H__ 7 #define __XFS_FORMAT_H__ 8 9 /* 10 * XFS On Disk Format Definitions 11 * 12 * This header file defines all the on-disk format definitions for 13 * general XFS objects. Directory and attribute related objects are defined in 14 * xfs_da_format.h, which log and log item formats are defined in 15 * xfs_log_format.h. Everything else goes here. 16 */ 17 18 struct xfs_mount; 19 struct xfs_trans; 20 struct xfs_inode; 21 struct xfs_buf; 22 struct xfs_ifork; 23 24 /* 25 * Super block 26 * Fits into a sector-sized buffer at address 0 of each allocation group. 27 * Only the first of these is ever updated except during growfs. 28 */ 29 #define XFS_SB_MAGIC 0x58465342 /* 'XFSB' */ 30 #define XFS_SB_VERSION_1 1 /* 5.3, 6.0.1, 6.1 */ 31 #define XFS_SB_VERSION_2 2 /* 6.2 - attributes */ 32 #define XFS_SB_VERSION_3 3 /* 6.2 - new inode version */ 33 #define XFS_SB_VERSION_4 4 /* 6.2+ - bitmask version */ 34 #define XFS_SB_VERSION_5 5 /* CRC enabled filesystem */ 35 #define XFS_SB_VERSION_NUMBITS 0x000f 36 #define XFS_SB_VERSION_ALLFBITS 0xfff0 37 #define XFS_SB_VERSION_ATTRBIT 0x0010 38 #define XFS_SB_VERSION_NLINKBIT 0x0020 39 #define XFS_SB_VERSION_QUOTABIT 0x0040 40 #define XFS_SB_VERSION_ALIGNBIT 0x0080 41 #define XFS_SB_VERSION_DALIGNBIT 0x0100 42 #define XFS_SB_VERSION_SHAREDBIT 0x0200 43 #define XFS_SB_VERSION_LOGV2BIT 0x0400 44 #define XFS_SB_VERSION_SECTORBIT 0x0800 45 #define XFS_SB_VERSION_EXTFLGBIT 0x1000 46 #define XFS_SB_VERSION_DIRV2BIT 0x2000 47 #define XFS_SB_VERSION_BORGBIT 0x4000 /* ASCII only case-insens. */ 48 #define XFS_SB_VERSION_MOREBITSBIT 0x8000 49 50 /* 51 * The size of a single extended attribute on disk is limited by 52 * the size of index values within the attribute entries themselves. 53 * These are be16 fields, so we can only support attribute data 54 * sizes up to 2^16 bytes in length. 55 */ 56 #define XFS_XATTR_SIZE_MAX (1 << 16) 57 58 /* 59 * Supported feature bit list is just all bits in the versionnum field because 60 * we've used them all up and understand them all. Except, of course, for the 61 * shared superblock bit, which nobody knows what it does and so is unsupported. 62 */ 63 #define XFS_SB_VERSION_OKBITS \ 64 ((XFS_SB_VERSION_NUMBITS | XFS_SB_VERSION_ALLFBITS) & \ 65 ~XFS_SB_VERSION_SHAREDBIT) 66 67 /* 68 * There are two words to hold XFS "feature" bits: the original 69 * word, sb_versionnum, and sb_features2. Whenever a bit is set in 70 * sb_features2, the feature bit XFS_SB_VERSION_MOREBITSBIT must be set. 71 * 72 * These defines represent bits in sb_features2. 73 */ 74 #define XFS_SB_VERSION2_RESERVED1BIT 0x00000001 75 #define XFS_SB_VERSION2_LAZYSBCOUNTBIT 0x00000002 /* Superblk counters */ 76 #define XFS_SB_VERSION2_RESERVED4BIT 0x00000004 77 #define XFS_SB_VERSION2_ATTR2BIT 0x00000008 /* Inline attr rework */ 78 #define XFS_SB_VERSION2_PARENTBIT 0x00000010 /* parent pointers */ 79 #define XFS_SB_VERSION2_PROJID32BIT 0x00000080 /* 32 bit project id */ 80 #define XFS_SB_VERSION2_CRCBIT 0x00000100 /* metadata CRCs */ 81 #define XFS_SB_VERSION2_FTYPE 0x00000200 /* inode type in dir */ 82 83 #define XFS_SB_VERSION2_OKBITS \ 84 (XFS_SB_VERSION2_LAZYSBCOUNTBIT | \ 85 XFS_SB_VERSION2_ATTR2BIT | \ 86 XFS_SB_VERSION2_PROJID32BIT | \ 87 XFS_SB_VERSION2_FTYPE) 88 89 /* Maximum size of the xfs filesystem label, no terminating NULL */ 90 #define XFSLABEL_MAX 12 91 92 /* 93 * Superblock - in core version. Must match the ondisk version below. 94 * Must be padded to 64 bit alignment. 95 */ 96 typedef struct xfs_sb { 97 uint32_t sb_magicnum; /* magic number == XFS_SB_MAGIC */ 98 uint32_t sb_blocksize; /* logical block size, bytes */ 99 xfs_rfsblock_t sb_dblocks; /* number of data blocks */ 100 xfs_rfsblock_t sb_rblocks; /* number of realtime blocks */ 101 xfs_rtblock_t sb_rextents; /* number of realtime extents */ 102 uuid_t sb_uuid; /* user-visible file system unique id */ 103 xfs_fsblock_t sb_logstart; /* starting block of log if internal */ 104 xfs_ino_t sb_rootino; /* root inode number */ 105 xfs_ino_t sb_rbmino; /* bitmap inode for realtime extents */ 106 xfs_ino_t sb_rsumino; /* summary inode for rt bitmap */ 107 xfs_agblock_t sb_rextsize; /* realtime extent size, blocks */ 108 xfs_agblock_t sb_agblocks; /* size of an allocation group */ 109 xfs_agnumber_t sb_agcount; /* number of allocation groups */ 110 xfs_extlen_t sb_rbmblocks; /* number of rt bitmap blocks */ 111 xfs_extlen_t sb_logblocks; /* number of log blocks */ 112 uint16_t sb_versionnum; /* header version == XFS_SB_VERSION */ 113 uint16_t sb_sectsize; /* volume sector size, bytes */ 114 uint16_t sb_inodesize; /* inode size, bytes */ 115 uint16_t sb_inopblock; /* inodes per block */ 116 char sb_fname[XFSLABEL_MAX]; /* file system name */ 117 uint8_t sb_blocklog; /* log2 of sb_blocksize */ 118 uint8_t sb_sectlog; /* log2 of sb_sectsize */ 119 uint8_t sb_inodelog; /* log2 of sb_inodesize */ 120 uint8_t sb_inopblog; /* log2 of sb_inopblock */ 121 uint8_t sb_agblklog; /* log2 of sb_agblocks (rounded up) */ 122 uint8_t sb_rextslog; /* log2 of sb_rextents */ 123 uint8_t sb_inprogress; /* mkfs is in progress, don't mount */ 124 uint8_t sb_imax_pct; /* max % of fs for inode space */ 125 /* statistics */ 126 /* 127 * These fields must remain contiguous. If you really 128 * want to change their layout, make sure you fix the 129 * code in xfs_trans_apply_sb_deltas(). 130 */ 131 uint64_t sb_icount; /* allocated inodes */ 132 uint64_t sb_ifree; /* free inodes */ 133 uint64_t sb_fdblocks; /* free data blocks */ 134 uint64_t sb_frextents; /* free realtime extents */ 135 /* 136 * End contiguous fields. 137 */ 138 xfs_ino_t sb_uquotino; /* user quota inode */ 139 xfs_ino_t sb_gquotino; /* group quota inode */ 140 uint16_t sb_qflags; /* quota flags */ 141 uint8_t sb_flags; /* misc. flags */ 142 uint8_t sb_shared_vn; /* shared version number */ 143 xfs_extlen_t sb_inoalignmt; /* inode chunk alignment, fsblocks */ 144 uint32_t sb_unit; /* stripe or raid unit */ 145 uint32_t sb_width; /* stripe or raid width */ 146 uint8_t sb_dirblklog; /* log2 of dir block size (fsbs) */ 147 uint8_t sb_logsectlog; /* log2 of the log sector size */ 148 uint16_t sb_logsectsize; /* sector size for the log, bytes */ 149 uint32_t sb_logsunit; /* stripe unit size for the log */ 150 uint32_t sb_features2; /* additional feature bits */ 151 152 /* 153 * bad features2 field as a result of failing to pad the sb structure to 154 * 64 bits. Some machines will be using this field for features2 bits. 155 * Easiest just to mark it bad and not use it for anything else. 156 * 157 * This is not kept up to date in memory; it is always overwritten by 158 * the value in sb_features2 when formatting the incore superblock to 159 * the disk buffer. 160 */ 161 uint32_t sb_bad_features2; 162 163 /* version 5 superblock fields start here */ 164 165 /* feature masks */ 166 uint32_t sb_features_compat; 167 uint32_t sb_features_ro_compat; 168 uint32_t sb_features_incompat; 169 uint32_t sb_features_log_incompat; 170 171 uint32_t sb_crc; /* superblock crc */ 172 xfs_extlen_t sb_spino_align; /* sparse inode chunk alignment */ 173 174 xfs_ino_t sb_pquotino; /* project quota inode */ 175 xfs_lsn_t sb_lsn; /* last write sequence */ 176 uuid_t sb_meta_uuid; /* metadata file system unique id */ 177 178 /* must be padded to 64 bit alignment */ 179 } xfs_sb_t; 180 181 #define XFS_SB_CRC_OFF offsetof(struct xfs_sb, sb_crc) 182 183 /* 184 * Superblock - on disk version. Must match the in core version above. 185 * Must be padded to 64 bit alignment. 186 */ 187 struct xfs_dsb { 188 __be32 sb_magicnum; /* magic number == XFS_SB_MAGIC */ 189 __be32 sb_blocksize; /* logical block size, bytes */ 190 __be64 sb_dblocks; /* number of data blocks */ 191 __be64 sb_rblocks; /* number of realtime blocks */ 192 __be64 sb_rextents; /* number of realtime extents */ 193 uuid_t sb_uuid; /* user-visible file system unique id */ 194 __be64 sb_logstart; /* starting block of log if internal */ 195 __be64 sb_rootino; /* root inode number */ 196 __be64 sb_rbmino; /* bitmap inode for realtime extents */ 197 __be64 sb_rsumino; /* summary inode for rt bitmap */ 198 __be32 sb_rextsize; /* realtime extent size, blocks */ 199 __be32 sb_agblocks; /* size of an allocation group */ 200 __be32 sb_agcount; /* number of allocation groups */ 201 __be32 sb_rbmblocks; /* number of rt bitmap blocks */ 202 __be32 sb_logblocks; /* number of log blocks */ 203 __be16 sb_versionnum; /* header version == XFS_SB_VERSION */ 204 __be16 sb_sectsize; /* volume sector size, bytes */ 205 __be16 sb_inodesize; /* inode size, bytes */ 206 __be16 sb_inopblock; /* inodes per block */ 207 char sb_fname[XFSLABEL_MAX]; /* file system name */ 208 __u8 sb_blocklog; /* log2 of sb_blocksize */ 209 __u8 sb_sectlog; /* log2 of sb_sectsize */ 210 __u8 sb_inodelog; /* log2 of sb_inodesize */ 211 __u8 sb_inopblog; /* log2 of sb_inopblock */ 212 __u8 sb_agblklog; /* log2 of sb_agblocks (rounded up) */ 213 __u8 sb_rextslog; /* log2 of sb_rextents */ 214 __u8 sb_inprogress; /* mkfs is in progress, don't mount */ 215 __u8 sb_imax_pct; /* max % of fs for inode space */ 216 /* statistics */ 217 /* 218 * These fields must remain contiguous. If you really 219 * want to change their layout, make sure you fix the 220 * code in xfs_trans_apply_sb_deltas(). 221 */ 222 __be64 sb_icount; /* allocated inodes */ 223 __be64 sb_ifree; /* free inodes */ 224 __be64 sb_fdblocks; /* free data blocks */ 225 __be64 sb_frextents; /* free realtime extents */ 226 /* 227 * End contiguous fields. 228 */ 229 __be64 sb_uquotino; /* user quota inode */ 230 __be64 sb_gquotino; /* group quota inode */ 231 __be16 sb_qflags; /* quota flags */ 232 __u8 sb_flags; /* misc. flags */ 233 __u8 sb_shared_vn; /* shared version number */ 234 __be32 sb_inoalignmt; /* inode chunk alignment, fsblocks */ 235 __be32 sb_unit; /* stripe or raid unit */ 236 __be32 sb_width; /* stripe or raid width */ 237 __u8 sb_dirblklog; /* log2 of dir block size (fsbs) */ 238 __u8 sb_logsectlog; /* log2 of the log sector size */ 239 __be16 sb_logsectsize; /* sector size for the log, bytes */ 240 __be32 sb_logsunit; /* stripe unit size for the log */ 241 __be32 sb_features2; /* additional feature bits */ 242 /* 243 * bad features2 field as a result of failing to pad the sb 244 * structure to 64 bits. Some machines will be using this field 245 * for features2 bits. Easiest just to mark it bad and not use 246 * it for anything else. 247 */ 248 __be32 sb_bad_features2; 249 250 /* version 5 superblock fields start here */ 251 252 /* feature masks */ 253 __be32 sb_features_compat; 254 __be32 sb_features_ro_compat; 255 __be32 sb_features_incompat; 256 __be32 sb_features_log_incompat; 257 258 __le32 sb_crc; /* superblock crc */ 259 __be32 sb_spino_align; /* sparse inode chunk alignment */ 260 261 __be64 sb_pquotino; /* project quota inode */ 262 __be64 sb_lsn; /* last write sequence */ 263 uuid_t sb_meta_uuid; /* metadata file system unique id */ 264 265 /* must be padded to 64 bit alignment */ 266 }; 267 268 /* 269 * Misc. Flags - warning - these will be cleared by xfs_repair unless 270 * a feature bit is set when the flag is used. 271 */ 272 #define XFS_SBF_NOFLAGS 0x00 /* no flags set */ 273 #define XFS_SBF_READONLY 0x01 /* only read-only mounts allowed */ 274 275 /* 276 * define max. shared version we can interoperate with 277 */ 278 #define XFS_SB_MAX_SHARED_VN 0 279 280 #define XFS_SB_VERSION_NUM(sbp) ((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS) 281 282 static inline bool xfs_sb_is_v5(struct xfs_sb *sbp) 283 { 284 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5; 285 } 286 287 /* 288 * Detect a mismatched features2 field. Older kernels read/wrote 289 * this into the wrong slot, so to be safe we keep them in sync. 290 */ 291 static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp) 292 { 293 return sbp->sb_bad_features2 != sbp->sb_features2; 294 } 295 296 static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp) 297 { 298 return xfs_sb_is_v5(sbp) || 299 (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT); 300 } 301 302 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp) 303 { 304 sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT; 305 } 306 307 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp) 308 { 309 sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT; 310 } 311 312 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp) 313 { 314 sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT; 315 sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT; 316 } 317 318 static inline void xfs_sb_version_addprojid32(struct xfs_sb *sbp) 319 { 320 sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT; 321 sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT; 322 } 323 324 /* 325 * Extended v5 superblock feature masks. These are to be used for new v5 326 * superblock features only. 327 * 328 * Compat features are new features that old kernels will not notice or affect 329 * and so can mount read-write without issues. 330 * 331 * RO-Compat (read only) are features that old kernels can read but will break 332 * if they write. Hence only read-only mounts of such filesystems are allowed on 333 * kernels that don't support the feature bit. 334 * 335 * InCompat features are features which old kernels will not understand and so 336 * must not mount. 337 * 338 * Log-InCompat features are for changes to log formats or new transactions that 339 * can't be replayed on older kernels. The fields are set when the filesystem is 340 * mounted, and a clean unmount clears the fields. 341 */ 342 #define XFS_SB_FEAT_COMPAT_ALL 0 343 #define XFS_SB_FEAT_COMPAT_UNKNOWN ~XFS_SB_FEAT_COMPAT_ALL 344 static inline bool 345 xfs_sb_has_compat_feature( 346 struct xfs_sb *sbp, 347 uint32_t feature) 348 { 349 return (sbp->sb_features_compat & feature) != 0; 350 } 351 352 #define XFS_SB_FEAT_RO_COMPAT_FINOBT (1 << 0) /* free inode btree */ 353 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT (1 << 1) /* reverse map btree */ 354 #define XFS_SB_FEAT_RO_COMPAT_REFLINK (1 << 2) /* reflinked files */ 355 #define XFS_SB_FEAT_RO_COMPAT_INOBTCNT (1 << 3) /* inobt block counts */ 356 #define XFS_SB_FEAT_RO_COMPAT_ALL \ 357 (XFS_SB_FEAT_RO_COMPAT_FINOBT | \ 358 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \ 359 XFS_SB_FEAT_RO_COMPAT_REFLINK| \ 360 XFS_SB_FEAT_RO_COMPAT_INOBTCNT) 361 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN ~XFS_SB_FEAT_RO_COMPAT_ALL 362 static inline bool 363 xfs_sb_has_ro_compat_feature( 364 struct xfs_sb *sbp, 365 uint32_t feature) 366 { 367 return (sbp->sb_features_ro_compat & feature) != 0; 368 } 369 370 #define XFS_SB_FEAT_INCOMPAT_FTYPE (1 << 0) /* filetype in dirent */ 371 #define XFS_SB_FEAT_INCOMPAT_SPINODES (1 << 1) /* sparse inode chunks */ 372 #define XFS_SB_FEAT_INCOMPAT_META_UUID (1 << 2) /* metadata UUID */ 373 #define XFS_SB_FEAT_INCOMPAT_BIGTIME (1 << 3) /* large timestamps */ 374 #define XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR (1 << 4) /* needs xfs_repair */ 375 #define XFS_SB_FEAT_INCOMPAT_NREXT64 (1 << 5) /* large extent counters */ 376 #define XFS_SB_FEAT_INCOMPAT_ALL \ 377 (XFS_SB_FEAT_INCOMPAT_FTYPE| \ 378 XFS_SB_FEAT_INCOMPAT_SPINODES| \ 379 XFS_SB_FEAT_INCOMPAT_META_UUID| \ 380 XFS_SB_FEAT_INCOMPAT_BIGTIME| \ 381 XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR) 382 383 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN ~XFS_SB_FEAT_INCOMPAT_ALL 384 static inline bool 385 xfs_sb_has_incompat_feature( 386 struct xfs_sb *sbp, 387 uint32_t feature) 388 { 389 return (sbp->sb_features_incompat & feature) != 0; 390 } 391 392 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL 0 393 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN ~XFS_SB_FEAT_INCOMPAT_LOG_ALL 394 static inline bool 395 xfs_sb_has_incompat_log_feature( 396 struct xfs_sb *sbp, 397 uint32_t feature) 398 { 399 return (sbp->sb_features_log_incompat & feature) != 0; 400 } 401 402 static inline void 403 xfs_sb_remove_incompat_log_features( 404 struct xfs_sb *sbp) 405 { 406 sbp->sb_features_log_incompat &= ~XFS_SB_FEAT_INCOMPAT_LOG_ALL; 407 } 408 409 static inline void 410 xfs_sb_add_incompat_log_features( 411 struct xfs_sb *sbp, 412 unsigned int features) 413 { 414 sbp->sb_features_log_incompat |= features; 415 } 416 417 418 static inline bool 419 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino) 420 { 421 return (ino == sbp->sb_uquotino || 422 ino == sbp->sb_gquotino || 423 ino == sbp->sb_pquotino); 424 } 425 426 #define XFS_SB_DADDR ((xfs_daddr_t)0) /* daddr in filesystem/ag */ 427 #define XFS_SB_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_SB_DADDR) 428 429 #define XFS_HDR_BLOCK(mp,d) ((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d)) 430 #define XFS_DADDR_TO_FSB(mp,d) XFS_AGB_TO_FSB(mp, \ 431 xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d)) 432 #define XFS_FSB_TO_DADDR(mp,fsbno) XFS_AGB_TO_DADDR(mp, \ 433 XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno)) 434 435 /* 436 * File system sector to basic block conversions. 437 */ 438 #define XFS_FSS_TO_BB(mp,sec) ((sec) << (mp)->m_sectbb_log) 439 440 /* 441 * File system block to basic block conversions. 442 */ 443 #define XFS_FSB_TO_BB(mp,fsbno) ((fsbno) << (mp)->m_blkbb_log) 444 #define XFS_BB_TO_FSB(mp,bb) \ 445 (((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log) 446 #define XFS_BB_TO_FSBT(mp,bb) ((bb) >> (mp)->m_blkbb_log) 447 448 /* 449 * File system block to byte conversions. 450 */ 451 #define XFS_FSB_TO_B(mp,fsbno) ((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog) 452 #define XFS_B_TO_FSB(mp,b) \ 453 ((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog) 454 #define XFS_B_TO_FSBT(mp,b) (((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog) 455 456 /* 457 * Allocation group header 458 * 459 * This is divided into three structures, placed in sequential 512-byte 460 * buffers after a copy of the superblock (also in a 512-byte buffer). 461 */ 462 #define XFS_AGF_MAGIC 0x58414746 /* 'XAGF' */ 463 #define XFS_AGI_MAGIC 0x58414749 /* 'XAGI' */ 464 #define XFS_AGFL_MAGIC 0x5841464c /* 'XAFL' */ 465 #define XFS_AGF_VERSION 1 466 #define XFS_AGI_VERSION 1 467 468 #define XFS_AGF_GOOD_VERSION(v) ((v) == XFS_AGF_VERSION) 469 #define XFS_AGI_GOOD_VERSION(v) ((v) == XFS_AGI_VERSION) 470 471 /* 472 * Btree number 0 is bno, 1 is cnt, 2 is rmap. This value gives the size of the 473 * arrays below. 474 */ 475 #define XFS_BTNUM_AGF ((int)XFS_BTNUM_RMAPi + 1) 476 477 /* 478 * The second word of agf_levels in the first a.g. overlaps the EFS 479 * superblock's magic number. Since the magic numbers valid for EFS 480 * are > 64k, our value cannot be confused for an EFS superblock's. 481 */ 482 483 typedef struct xfs_agf { 484 /* 485 * Common allocation group header information 486 */ 487 __be32 agf_magicnum; /* magic number == XFS_AGF_MAGIC */ 488 __be32 agf_versionnum; /* header version == XFS_AGF_VERSION */ 489 __be32 agf_seqno; /* sequence # starting from 0 */ 490 __be32 agf_length; /* size in blocks of a.g. */ 491 /* 492 * Freespace and rmap information 493 */ 494 __be32 agf_roots[XFS_BTNUM_AGF]; /* root blocks */ 495 __be32 agf_levels[XFS_BTNUM_AGF]; /* btree levels */ 496 497 __be32 agf_flfirst; /* first freelist block's index */ 498 __be32 agf_fllast; /* last freelist block's index */ 499 __be32 agf_flcount; /* count of blocks in freelist */ 500 __be32 agf_freeblks; /* total free blocks */ 501 502 __be32 agf_longest; /* longest free space */ 503 __be32 agf_btreeblks; /* # of blocks held in AGF btrees */ 504 uuid_t agf_uuid; /* uuid of filesystem */ 505 506 __be32 agf_rmap_blocks; /* rmapbt blocks used */ 507 __be32 agf_refcount_blocks; /* refcountbt blocks used */ 508 509 __be32 agf_refcount_root; /* refcount tree root block */ 510 __be32 agf_refcount_level; /* refcount btree levels */ 511 512 /* 513 * reserve some contiguous space for future logged fields before we add 514 * the unlogged fields. This makes the range logging via flags and 515 * structure offsets much simpler. 516 */ 517 __be64 agf_spare64[14]; 518 519 /* unlogged fields, written during buffer writeback. */ 520 __be64 agf_lsn; /* last write sequence */ 521 __be32 agf_crc; /* crc of agf sector */ 522 __be32 agf_spare2; 523 524 /* structure must be padded to 64 bit alignment */ 525 } xfs_agf_t; 526 527 #define XFS_AGF_CRC_OFF offsetof(struct xfs_agf, agf_crc) 528 529 #define XFS_AGF_MAGICNUM 0x00000001 530 #define XFS_AGF_VERSIONNUM 0x00000002 531 #define XFS_AGF_SEQNO 0x00000004 532 #define XFS_AGF_LENGTH 0x00000008 533 #define XFS_AGF_ROOTS 0x00000010 534 #define XFS_AGF_LEVELS 0x00000020 535 #define XFS_AGF_FLFIRST 0x00000040 536 #define XFS_AGF_FLLAST 0x00000080 537 #define XFS_AGF_FLCOUNT 0x00000100 538 #define XFS_AGF_FREEBLKS 0x00000200 539 #define XFS_AGF_LONGEST 0x00000400 540 #define XFS_AGF_BTREEBLKS 0x00000800 541 #define XFS_AGF_UUID 0x00001000 542 #define XFS_AGF_RMAP_BLOCKS 0x00002000 543 #define XFS_AGF_REFCOUNT_BLOCKS 0x00004000 544 #define XFS_AGF_REFCOUNT_ROOT 0x00008000 545 #define XFS_AGF_REFCOUNT_LEVEL 0x00010000 546 #define XFS_AGF_SPARE64 0x00020000 547 #define XFS_AGF_NUM_BITS 18 548 #define XFS_AGF_ALL_BITS ((1 << XFS_AGF_NUM_BITS) - 1) 549 550 #define XFS_AGF_FLAGS \ 551 { XFS_AGF_MAGICNUM, "MAGICNUM" }, \ 552 { XFS_AGF_VERSIONNUM, "VERSIONNUM" }, \ 553 { XFS_AGF_SEQNO, "SEQNO" }, \ 554 { XFS_AGF_LENGTH, "LENGTH" }, \ 555 { XFS_AGF_ROOTS, "ROOTS" }, \ 556 { XFS_AGF_LEVELS, "LEVELS" }, \ 557 { XFS_AGF_FLFIRST, "FLFIRST" }, \ 558 { XFS_AGF_FLLAST, "FLLAST" }, \ 559 { XFS_AGF_FLCOUNT, "FLCOUNT" }, \ 560 { XFS_AGF_FREEBLKS, "FREEBLKS" }, \ 561 { XFS_AGF_LONGEST, "LONGEST" }, \ 562 { XFS_AGF_BTREEBLKS, "BTREEBLKS" }, \ 563 { XFS_AGF_UUID, "UUID" }, \ 564 { XFS_AGF_RMAP_BLOCKS, "RMAP_BLOCKS" }, \ 565 { XFS_AGF_REFCOUNT_BLOCKS, "REFCOUNT_BLOCKS" }, \ 566 { XFS_AGF_REFCOUNT_ROOT, "REFCOUNT_ROOT" }, \ 567 { XFS_AGF_REFCOUNT_LEVEL, "REFCOUNT_LEVEL" }, \ 568 { XFS_AGF_SPARE64, "SPARE64" } 569 570 /* disk block (xfs_daddr_t) in the AG */ 571 #define XFS_AGF_DADDR(mp) ((xfs_daddr_t)(1 << (mp)->m_sectbb_log)) 572 #define XFS_AGF_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp)) 573 574 /* 575 * Size of the unlinked inode hash table in the agi. 576 */ 577 #define XFS_AGI_UNLINKED_BUCKETS 64 578 579 typedef struct xfs_agi { 580 /* 581 * Common allocation group header information 582 */ 583 __be32 agi_magicnum; /* magic number == XFS_AGI_MAGIC */ 584 __be32 agi_versionnum; /* header version == XFS_AGI_VERSION */ 585 __be32 agi_seqno; /* sequence # starting from 0 */ 586 __be32 agi_length; /* size in blocks of a.g. */ 587 /* 588 * Inode information 589 * Inodes are mapped by interpreting the inode number, so no 590 * mapping data is needed here. 591 */ 592 __be32 agi_count; /* count of allocated inodes */ 593 __be32 agi_root; /* root of inode btree */ 594 __be32 agi_level; /* levels in inode btree */ 595 __be32 agi_freecount; /* number of free inodes */ 596 597 __be32 agi_newino; /* new inode just allocated */ 598 __be32 agi_dirino; /* last directory inode chunk */ 599 /* 600 * Hash table of inodes which have been unlinked but are 601 * still being referenced. 602 */ 603 __be32 agi_unlinked[XFS_AGI_UNLINKED_BUCKETS]; 604 /* 605 * This marks the end of logging region 1 and start of logging region 2. 606 */ 607 uuid_t agi_uuid; /* uuid of filesystem */ 608 __be32 agi_crc; /* crc of agi sector */ 609 __be32 agi_pad32; 610 __be64 agi_lsn; /* last write sequence */ 611 612 __be32 agi_free_root; /* root of the free inode btree */ 613 __be32 agi_free_level;/* levels in free inode btree */ 614 615 __be32 agi_iblocks; /* inobt blocks used */ 616 __be32 agi_fblocks; /* finobt blocks used */ 617 618 /* structure must be padded to 64 bit alignment */ 619 } xfs_agi_t; 620 621 #define XFS_AGI_CRC_OFF offsetof(struct xfs_agi, agi_crc) 622 623 #define XFS_AGI_MAGICNUM (1 << 0) 624 #define XFS_AGI_VERSIONNUM (1 << 1) 625 #define XFS_AGI_SEQNO (1 << 2) 626 #define XFS_AGI_LENGTH (1 << 3) 627 #define XFS_AGI_COUNT (1 << 4) 628 #define XFS_AGI_ROOT (1 << 5) 629 #define XFS_AGI_LEVEL (1 << 6) 630 #define XFS_AGI_FREECOUNT (1 << 7) 631 #define XFS_AGI_NEWINO (1 << 8) 632 #define XFS_AGI_DIRINO (1 << 9) 633 #define XFS_AGI_UNLINKED (1 << 10) 634 #define XFS_AGI_NUM_BITS_R1 11 /* end of the 1st agi logging region */ 635 #define XFS_AGI_ALL_BITS_R1 ((1 << XFS_AGI_NUM_BITS_R1) - 1) 636 #define XFS_AGI_FREE_ROOT (1 << 11) 637 #define XFS_AGI_FREE_LEVEL (1 << 12) 638 #define XFS_AGI_IBLOCKS (1 << 13) /* both inobt/finobt block counters */ 639 #define XFS_AGI_NUM_BITS_R2 14 640 641 /* disk block (xfs_daddr_t) in the AG */ 642 #define XFS_AGI_DADDR(mp) ((xfs_daddr_t)(2 << (mp)->m_sectbb_log)) 643 #define XFS_AGI_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp)) 644 645 /* 646 * The third a.g. block contains the a.g. freelist, an array 647 * of block pointers to blocks owned by the allocation btree code. 648 */ 649 #define XFS_AGFL_DADDR(mp) ((xfs_daddr_t)(3 << (mp)->m_sectbb_log)) 650 #define XFS_AGFL_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp)) 651 #define XFS_BUF_TO_AGFL(bp) ((struct xfs_agfl *)((bp)->b_addr)) 652 653 struct xfs_agfl { 654 __be32 agfl_magicnum; 655 __be32 agfl_seqno; 656 uuid_t agfl_uuid; 657 __be64 agfl_lsn; 658 __be32 agfl_crc; 659 } __attribute__((packed)); 660 661 #define XFS_AGFL_CRC_OFF offsetof(struct xfs_agfl, agfl_crc) 662 663 #define XFS_AGB_TO_FSB(mp,agno,agbno) \ 664 (((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno)) 665 #define XFS_FSB_TO_AGNO(mp,fsbno) \ 666 ((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog)) 667 #define XFS_FSB_TO_AGBNO(mp,fsbno) \ 668 ((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog))) 669 #define XFS_AGB_TO_DADDR(mp,agno,agbno) \ 670 ((xfs_daddr_t)XFS_FSB_TO_BB(mp, \ 671 (xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno))) 672 #define XFS_AG_DADDR(mp,agno,d) (XFS_AGB_TO_DADDR(mp, agno, 0) + (d)) 673 674 /* 675 * For checking for bad ranges of xfs_daddr_t's, covering multiple 676 * allocation groups or a single xfs_daddr_t that's a superblock copy. 677 */ 678 #define XFS_AG_CHECK_DADDR(mp,d,len) \ 679 ((len) == 1 ? \ 680 ASSERT((d) == XFS_SB_DADDR || \ 681 xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \ 682 ASSERT(xfs_daddr_to_agno(mp, d) == \ 683 xfs_daddr_to_agno(mp, (d) + (len) - 1))) 684 685 /* 686 * XFS Timestamps 687 * ============== 688 * 689 * Traditional ondisk inode timestamps consist of signed 32-bit counters for 690 * seconds and nanoseconds; time zero is the Unix epoch, Jan 1 00:00:00 UTC 691 * 1970, which means that the timestamp epoch is the same as the Unix epoch. 692 * Therefore, the ondisk min and max defined here can be used directly to 693 * constrain the incore timestamps on a Unix system. Note that we actually 694 * encode a __be64 value on disk. 695 * 696 * When the bigtime feature is enabled, ondisk inode timestamps become an 697 * unsigned 64-bit nanoseconds counter. This means that the bigtime inode 698 * timestamp epoch is the start of the classic timestamp range, which is 699 * Dec 31 20:45:52 UTC 1901. Because the epochs are not the same, callers 700 * /must/ use the bigtime conversion functions when encoding and decoding raw 701 * timestamps. 702 */ 703 typedef __be64 xfs_timestamp_t; 704 705 /* Legacy timestamp encoding format. */ 706 struct xfs_legacy_timestamp { 707 __be32 t_sec; /* timestamp seconds */ 708 __be32 t_nsec; /* timestamp nanoseconds */ 709 }; 710 711 /* 712 * Smallest possible ondisk seconds value with traditional timestamps. This 713 * corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901. 714 */ 715 #define XFS_LEGACY_TIME_MIN ((int64_t)S32_MIN) 716 717 /* 718 * Largest possible ondisk seconds value with traditional timestamps. This 719 * corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038. 720 */ 721 #define XFS_LEGACY_TIME_MAX ((int64_t)S32_MAX) 722 723 /* 724 * Smallest possible ondisk seconds value with bigtime timestamps. This 725 * corresponds (after conversion to a Unix timestamp) with the traditional 726 * minimum timestamp of Dec 13 20:45:52 UTC 1901. 727 */ 728 #define XFS_BIGTIME_TIME_MIN ((int64_t)0) 729 730 /* 731 * Largest supported ondisk seconds value with bigtime timestamps. This 732 * corresponds (after conversion to a Unix timestamp) with an incore timestamp 733 * of Jul 2 20:20:24 UTC 2486. 734 * 735 * We round down the ondisk limit so that the bigtime quota and inode max 736 * timestamps will be the same. 737 */ 738 #define XFS_BIGTIME_TIME_MAX ((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL)) 739 740 /* 741 * Bigtime epoch is set exactly to the minimum time value that a traditional 742 * 32-bit timestamp can represent when using the Unix epoch as a reference. 743 * Hence the Unix epoch is at a fixed offset into the supported bigtime 744 * timestamp range. 745 * 746 * The bigtime epoch also matches the minimum value an on-disk 32-bit XFS 747 * timestamp can represent so we will not lose any fidelity in converting 748 * to/from unix and bigtime timestamps. 749 * 750 * The following conversion factor converts a seconds counter from the Unix 751 * epoch to the bigtime epoch. 752 */ 753 #define XFS_BIGTIME_EPOCH_OFFSET (-(int64_t)S32_MIN) 754 755 /* Convert a timestamp from the Unix epoch to the bigtime epoch. */ 756 static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds) 757 { 758 return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET; 759 } 760 761 /* Convert a timestamp from the bigtime epoch to the Unix epoch. */ 762 static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds) 763 { 764 return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET; 765 } 766 767 /* 768 * On-disk inode structure. 769 * 770 * This is just the header or "dinode core", the inode is expanded to fill a 771 * variable size the leftover area split into a data and an attribute fork. 772 * The format of the data and attribute fork depends on the format of the 773 * inode as indicated by di_format and di_aformat. To access the data and 774 * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros 775 * below. 776 * 777 * There is a very similar struct xfs_log_dinode which matches the layout of 778 * this structure, but is kept in native format instead of big endian. 779 * 780 * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed 781 * padding field for v3 inodes. 782 */ 783 #define XFS_DINODE_MAGIC 0x494e /* 'IN' */ 784 struct xfs_dinode { 785 __be16 di_magic; /* inode magic # = XFS_DINODE_MAGIC */ 786 __be16 di_mode; /* mode and type of file */ 787 __u8 di_version; /* inode version */ 788 __u8 di_format; /* format of di_c data */ 789 __be16 di_onlink; /* old number of links to file */ 790 __be32 di_uid; /* owner's user id */ 791 __be32 di_gid; /* owner's group id */ 792 __be32 di_nlink; /* number of links to file */ 793 __be16 di_projid_lo; /* lower part of owner's project id */ 794 __be16 di_projid_hi; /* higher part owner's project id */ 795 union { 796 /* Number of data fork extents if NREXT64 is set */ 797 __be64 di_big_nextents; 798 799 /* Padding for V3 inodes without NREXT64 set. */ 800 __be64 di_v3_pad; 801 802 /* Padding and inode flush counter for V2 inodes. */ 803 struct { 804 __u8 di_v2_pad[6]; 805 __be16 di_flushiter; 806 }; 807 }; 808 xfs_timestamp_t di_atime; /* time last accessed */ 809 xfs_timestamp_t di_mtime; /* time last modified */ 810 xfs_timestamp_t di_ctime; /* time created/inode modified */ 811 __be64 di_size; /* number of bytes in file */ 812 __be64 di_nblocks; /* # of direct & btree blocks used */ 813 __be32 di_extsize; /* basic/minimum extent size for file */ 814 union { 815 /* 816 * For V2 inodes and V3 inodes without NREXT64 set, this 817 * is the number of data and attr fork extents. 818 */ 819 struct { 820 __be32 di_nextents; 821 __be16 di_anextents; 822 } __packed; 823 824 /* Number of attr fork extents if NREXT64 is set. */ 825 struct { 826 __be32 di_big_anextents; 827 __be16 di_nrext64_pad; 828 } __packed; 829 } __packed; 830 __u8 di_forkoff; /* attr fork offs, <<3 for 64b align */ 831 __s8 di_aformat; /* format of attr fork's data */ 832 __be32 di_dmevmask; /* DMIG event mask */ 833 __be16 di_dmstate; /* DMIG state info */ 834 __be16 di_flags; /* random flags, XFS_DIFLAG_... */ 835 __be32 di_gen; /* generation number */ 836 837 /* di_next_unlinked is the only non-core field in the old dinode */ 838 __be32 di_next_unlinked;/* agi unlinked list ptr */ 839 840 /* start of the extended dinode, writable fields */ 841 __le32 di_crc; /* CRC of the inode */ 842 __be64 di_changecount; /* number of attribute changes */ 843 __be64 di_lsn; /* flush sequence */ 844 __be64 di_flags2; /* more random flags */ 845 __be32 di_cowextsize; /* basic cow extent size for file */ 846 __u8 di_pad2[12]; /* more padding for future expansion */ 847 848 /* fields only written to during inode creation */ 849 xfs_timestamp_t di_crtime; /* time created */ 850 __be64 di_ino; /* inode number */ 851 uuid_t di_uuid; /* UUID of the filesystem */ 852 853 /* structure must be padded to 64 bit alignment */ 854 }; 855 856 #define XFS_DINODE_CRC_OFF offsetof(struct xfs_dinode, di_crc) 857 858 #define DI_MAX_FLUSH 0xffff 859 860 /* 861 * Size of the core inode on disk. Version 1 and 2 inodes have 862 * the same size, but version 3 has grown a few additional fields. 863 */ 864 static inline uint xfs_dinode_size(int version) 865 { 866 if (version == 3) 867 return sizeof(struct xfs_dinode); 868 return offsetof(struct xfs_dinode, di_crc); 869 } 870 871 /* 872 * The 32 bit link count in the inode theoretically maxes out at UINT_MAX. 873 * Since the pathconf interface is signed, we use 2^31 - 1 instead. 874 */ 875 #define XFS_MAXLINK ((1U << 31) - 1U) 876 877 /* 878 * Values for di_format 879 * 880 * This enum is used in string mapping in xfs_trace.h; please keep the 881 * TRACE_DEFINE_ENUMs for it up to date. 882 */ 883 enum xfs_dinode_fmt { 884 XFS_DINODE_FMT_DEV, /* xfs_dev_t */ 885 XFS_DINODE_FMT_LOCAL, /* bulk data */ 886 XFS_DINODE_FMT_EXTENTS, /* struct xfs_bmbt_rec */ 887 XFS_DINODE_FMT_BTREE, /* struct xfs_bmdr_block */ 888 XFS_DINODE_FMT_UUID /* added long ago, but never used */ 889 }; 890 891 #define XFS_INODE_FORMAT_STR \ 892 { XFS_DINODE_FMT_DEV, "dev" }, \ 893 { XFS_DINODE_FMT_LOCAL, "local" }, \ 894 { XFS_DINODE_FMT_EXTENTS, "extent" }, \ 895 { XFS_DINODE_FMT_BTREE, "btree" }, \ 896 { XFS_DINODE_FMT_UUID, "uuid" } 897 898 /* 899 * Max values for extnum and aextnum. 900 * 901 * The original on-disk extent counts were held in signed fields, resulting in 902 * maximum extent counts of 2^31 and 2^15 for the data and attr forks 903 * respectively. Similarly the maximum extent length is limited to 2^21 blocks 904 * by the 21-bit wide blockcount field of a BMBT extent record. 905 * 906 * The newly introduced data fork extent counter can hold a 64-bit value, 907 * however the maximum number of extents in a file is also limited to 2^54 908 * extents by the 54-bit wide startoff field of a BMBT extent record. 909 * 910 * It is further limited by the maximum supported file size of 2^63 911 * *bytes*. This leads to a maximum extent count for maximally sized filesystem 912 * blocks (64kB) of: 913 * 914 * 2^63 bytes / 2^16 bytes per block = 2^47 blocks 915 * 916 * Rounding up 47 to the nearest multiple of bits-per-byte results in 48. Hence 917 * 2^48 was chosen as the maximum data fork extent count. 918 * 919 * The maximum file size that can be represented by the data fork extent counter 920 * in the worst case occurs when all extents are 1 block in length and each 921 * block is 1KB in size. 922 * 923 * With XFS_MAX_EXTCNT_DATA_FORK_SMALL representing maximum extent count and 924 * with 1KB sized blocks, a file can reach upto, 925 * 1KB * (2^31) = 2TB 926 * 927 * This is much larger than the theoretical maximum size of a directory 928 * i.e. XFS_DIR2_SPACE_SIZE * XFS_DIR2_MAX_SPACES = ~96GB. 929 * 930 * Hence, a directory inode can never overflow its data fork extent counter. 931 */ 932 #define XFS_MAX_EXTCNT_DATA_FORK_LARGE ((xfs_extnum_t)((1ULL << 48) - 1)) 933 #define XFS_MAX_EXTCNT_ATTR_FORK_LARGE ((xfs_extnum_t)((1ULL << 32) - 1)) 934 #define XFS_MAX_EXTCNT_DATA_FORK_SMALL ((xfs_extnum_t)((1ULL << 31) - 1)) 935 #define XFS_MAX_EXTCNT_ATTR_FORK_SMALL ((xfs_extnum_t)((1ULL << 15) - 1)) 936 937 /* 938 * Inode minimum and maximum sizes. 939 */ 940 #define XFS_DINODE_MIN_LOG 8 941 #define XFS_DINODE_MAX_LOG 11 942 #define XFS_DINODE_MIN_SIZE (1 << XFS_DINODE_MIN_LOG) 943 #define XFS_DINODE_MAX_SIZE (1 << XFS_DINODE_MAX_LOG) 944 945 /* 946 * Inode size for given fs. 947 */ 948 #define XFS_DINODE_SIZE(mp) \ 949 (xfs_has_v3inodes(mp) ? \ 950 sizeof(struct xfs_dinode) : \ 951 offsetof(struct xfs_dinode, di_crc)) 952 #define XFS_LITINO(mp) \ 953 ((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(mp)) 954 955 /* 956 * Inode data & attribute fork sizes, per inode. 957 */ 958 #define XFS_DFORK_BOFF(dip) ((int)((dip)->di_forkoff << 3)) 959 960 #define XFS_DFORK_DSIZE(dip,mp) \ 961 ((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp)) 962 #define XFS_DFORK_ASIZE(dip,mp) \ 963 ((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0) 964 #define XFS_DFORK_SIZE(dip,mp,w) \ 965 ((w) == XFS_DATA_FORK ? \ 966 XFS_DFORK_DSIZE(dip, mp) : \ 967 XFS_DFORK_ASIZE(dip, mp)) 968 969 #define XFS_DFORK_MAXEXT(dip, mp, w) \ 970 (XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec)) 971 972 /* 973 * Return pointers to the data or attribute forks. 974 */ 975 #define XFS_DFORK_DPTR(dip) \ 976 ((char *)dip + xfs_dinode_size(dip->di_version)) 977 #define XFS_DFORK_APTR(dip) \ 978 (XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip)) 979 #define XFS_DFORK_PTR(dip,w) \ 980 ((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip)) 981 982 #define XFS_DFORK_FORMAT(dip,w) \ 983 ((w) == XFS_DATA_FORK ? \ 984 (dip)->di_format : \ 985 (dip)->di_aformat) 986 987 /* 988 * For block and character special files the 32bit dev_t is stored at the 989 * beginning of the data fork. 990 */ 991 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip) 992 { 993 return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip)); 994 } 995 996 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev) 997 { 998 *(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev); 999 } 1000 1001 /* 1002 * Values for di_flags 1003 */ 1004 #define XFS_DIFLAG_REALTIME_BIT 0 /* file's blocks come from rt area */ 1005 #define XFS_DIFLAG_PREALLOC_BIT 1 /* file space has been preallocated */ 1006 #define XFS_DIFLAG_NEWRTBM_BIT 2 /* for rtbitmap inode, new format */ 1007 #define XFS_DIFLAG_IMMUTABLE_BIT 3 /* inode is immutable */ 1008 #define XFS_DIFLAG_APPEND_BIT 4 /* inode is append-only */ 1009 #define XFS_DIFLAG_SYNC_BIT 5 /* inode is written synchronously */ 1010 #define XFS_DIFLAG_NOATIME_BIT 6 /* do not update atime */ 1011 #define XFS_DIFLAG_NODUMP_BIT 7 /* do not dump */ 1012 #define XFS_DIFLAG_RTINHERIT_BIT 8 /* create with realtime bit set */ 1013 #define XFS_DIFLAG_PROJINHERIT_BIT 9 /* create with parents projid */ 1014 #define XFS_DIFLAG_NOSYMLINKS_BIT 10 /* disallow symlink creation */ 1015 #define XFS_DIFLAG_EXTSIZE_BIT 11 /* inode extent size allocator hint */ 1016 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12 /* inherit inode extent size */ 1017 #define XFS_DIFLAG_NODEFRAG_BIT 13 /* do not reorganize/defragment */ 1018 #define XFS_DIFLAG_FILESTREAM_BIT 14 /* use filestream allocator */ 1019 /* Do not use bit 15, di_flags is legacy and unchanging now */ 1020 1021 #define XFS_DIFLAG_REALTIME (1 << XFS_DIFLAG_REALTIME_BIT) 1022 #define XFS_DIFLAG_PREALLOC (1 << XFS_DIFLAG_PREALLOC_BIT) 1023 #define XFS_DIFLAG_NEWRTBM (1 << XFS_DIFLAG_NEWRTBM_BIT) 1024 #define XFS_DIFLAG_IMMUTABLE (1 << XFS_DIFLAG_IMMUTABLE_BIT) 1025 #define XFS_DIFLAG_APPEND (1 << XFS_DIFLAG_APPEND_BIT) 1026 #define XFS_DIFLAG_SYNC (1 << XFS_DIFLAG_SYNC_BIT) 1027 #define XFS_DIFLAG_NOATIME (1 << XFS_DIFLAG_NOATIME_BIT) 1028 #define XFS_DIFLAG_NODUMP (1 << XFS_DIFLAG_NODUMP_BIT) 1029 #define XFS_DIFLAG_RTINHERIT (1 << XFS_DIFLAG_RTINHERIT_BIT) 1030 #define XFS_DIFLAG_PROJINHERIT (1 << XFS_DIFLAG_PROJINHERIT_BIT) 1031 #define XFS_DIFLAG_NOSYMLINKS (1 << XFS_DIFLAG_NOSYMLINKS_BIT) 1032 #define XFS_DIFLAG_EXTSIZE (1 << XFS_DIFLAG_EXTSIZE_BIT) 1033 #define XFS_DIFLAG_EXTSZINHERIT (1 << XFS_DIFLAG_EXTSZINHERIT_BIT) 1034 #define XFS_DIFLAG_NODEFRAG (1 << XFS_DIFLAG_NODEFRAG_BIT) 1035 #define XFS_DIFLAG_FILESTREAM (1 << XFS_DIFLAG_FILESTREAM_BIT) 1036 1037 #define XFS_DIFLAG_ANY \ 1038 (XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \ 1039 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \ 1040 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \ 1041 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \ 1042 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM) 1043 1044 /* 1045 * Values for di_flags2 These start by being exposed to userspace in the upper 1046 * 16 bits of the XFS_XFLAG_s range. 1047 */ 1048 #define XFS_DIFLAG2_DAX_BIT 0 /* use DAX for this inode */ 1049 #define XFS_DIFLAG2_REFLINK_BIT 1 /* file's blocks may be shared */ 1050 #define XFS_DIFLAG2_COWEXTSIZE_BIT 2 /* copy on write extent size hint */ 1051 #define XFS_DIFLAG2_BIGTIME_BIT 3 /* big timestamps */ 1052 #define XFS_DIFLAG2_NREXT64_BIT 4 /* large extent counters */ 1053 1054 #define XFS_DIFLAG2_DAX (1 << XFS_DIFLAG2_DAX_BIT) 1055 #define XFS_DIFLAG2_REFLINK (1 << XFS_DIFLAG2_REFLINK_BIT) 1056 #define XFS_DIFLAG2_COWEXTSIZE (1 << XFS_DIFLAG2_COWEXTSIZE_BIT) 1057 #define XFS_DIFLAG2_BIGTIME (1 << XFS_DIFLAG2_BIGTIME_BIT) 1058 #define XFS_DIFLAG2_NREXT64 (1 << XFS_DIFLAG2_NREXT64_BIT) 1059 1060 #define XFS_DIFLAG2_ANY \ 1061 (XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \ 1062 XFS_DIFLAG2_BIGTIME | XFS_DIFLAG2_NREXT64) 1063 1064 static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip) 1065 { 1066 return dip->di_version >= 3 && 1067 (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME)); 1068 } 1069 1070 static inline bool xfs_dinode_has_large_extent_counts( 1071 const struct xfs_dinode *dip) 1072 { 1073 return dip->di_version >= 3 && 1074 (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_NREXT64)); 1075 } 1076 1077 /* 1078 * Inode number format: 1079 * low inopblog bits - offset in block 1080 * next agblklog bits - block number in ag 1081 * next agno_log bits - ag number 1082 * high agno_log-agblklog-inopblog bits - 0 1083 */ 1084 #define XFS_INO_MASK(k) (uint32_t)((1ULL << (k)) - 1) 1085 #define XFS_INO_OFFSET_BITS(mp) (mp)->m_sb.sb_inopblog 1086 #define XFS_INO_AGBNO_BITS(mp) (mp)->m_sb.sb_agblklog 1087 #define XFS_INO_AGINO_BITS(mp) ((mp)->m_ino_geo.agino_log) 1088 #define XFS_INO_AGNO_BITS(mp) (mp)->m_agno_log 1089 #define XFS_INO_BITS(mp) \ 1090 XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp) 1091 #define XFS_INO_TO_AGNO(mp,i) \ 1092 ((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp))) 1093 #define XFS_INO_TO_AGINO(mp,i) \ 1094 ((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp))) 1095 #define XFS_INO_TO_AGBNO(mp,i) \ 1096 (((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \ 1097 XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp))) 1098 #define XFS_INO_TO_OFFSET(mp,i) \ 1099 ((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp))) 1100 #define XFS_INO_TO_FSB(mp,i) \ 1101 XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i)) 1102 #define XFS_AGINO_TO_INO(mp,a,i) \ 1103 (((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i)) 1104 #define XFS_AGINO_TO_AGBNO(mp,i) ((i) >> XFS_INO_OFFSET_BITS(mp)) 1105 #define XFS_AGINO_TO_OFFSET(mp,i) \ 1106 ((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp))) 1107 #define XFS_OFFBNO_TO_AGINO(mp,b,o) \ 1108 ((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o))) 1109 #define XFS_FSB_TO_INO(mp, b) ((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp))) 1110 #define XFS_AGB_TO_AGINO(mp, b) ((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp))) 1111 1112 #define XFS_MAXINUMBER ((xfs_ino_t)((1ULL << 56) - 1ULL)) 1113 #define XFS_MAXINUMBER_32 ((xfs_ino_t)((1ULL << 32) - 1ULL)) 1114 1115 /* 1116 * RealTime Device format definitions 1117 */ 1118 1119 /* Min and max rt extent sizes, specified in bytes */ 1120 #define XFS_MAX_RTEXTSIZE (1024 * 1024 * 1024) /* 1GB */ 1121 #define XFS_DFL_RTEXTSIZE (64 * 1024) /* 64kB */ 1122 #define XFS_MIN_RTEXTSIZE (4 * 1024) /* 4kB */ 1123 1124 #define XFS_BLOCKSIZE(mp) ((mp)->m_sb.sb_blocksize) 1125 #define XFS_BLOCKMASK(mp) ((mp)->m_blockmask) 1126 #define XFS_BLOCKWSIZE(mp) ((mp)->m_blockwsize) 1127 #define XFS_BLOCKWMASK(mp) ((mp)->m_blockwmask) 1128 1129 /* 1130 * RT Summary and bit manipulation macros. 1131 */ 1132 #define XFS_SUMOFFS(mp,ls,bb) ((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb))) 1133 #define XFS_SUMOFFSTOBLOCK(mp,s) \ 1134 (((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog) 1135 #define XFS_SUMPTR(mp,bp,so) \ 1136 ((xfs_suminfo_t *)((bp)->b_addr + \ 1137 (((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp)))) 1138 1139 #define XFS_BITTOBLOCK(mp,bi) ((bi) >> (mp)->m_blkbit_log) 1140 #define XFS_BLOCKTOBIT(mp,bb) ((bb) << (mp)->m_blkbit_log) 1141 #define XFS_BITTOWORD(mp,bi) \ 1142 ((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp))) 1143 1144 #define XFS_RTMIN(a,b) ((a) < (b) ? (a) : (b)) 1145 #define XFS_RTMAX(a,b) ((a) > (b) ? (a) : (b)) 1146 1147 #define XFS_RTLOBIT(w) xfs_lowbit32(w) 1148 #define XFS_RTHIBIT(w) xfs_highbit32(w) 1149 1150 #define XFS_RTBLOCKLOG(b) xfs_highbit64(b) 1151 1152 /* 1153 * Dquot and dquot block format definitions 1154 */ 1155 #define XFS_DQUOT_MAGIC 0x4451 /* 'DQ' */ 1156 #define XFS_DQUOT_VERSION (uint8_t)0x01 /* latest version number */ 1157 1158 #define XFS_DQTYPE_USER 0x01 /* user dquot record */ 1159 #define XFS_DQTYPE_PROJ 0x02 /* project dquot record */ 1160 #define XFS_DQTYPE_GROUP 0x04 /* group dquot record */ 1161 #define XFS_DQTYPE_BIGTIME 0x80 /* large expiry timestamps */ 1162 1163 /* bitmask to determine if this is a user/group/project dquot */ 1164 #define XFS_DQTYPE_REC_MASK (XFS_DQTYPE_USER | \ 1165 XFS_DQTYPE_PROJ | \ 1166 XFS_DQTYPE_GROUP) 1167 1168 #define XFS_DQTYPE_ANY (XFS_DQTYPE_REC_MASK | \ 1169 XFS_DQTYPE_BIGTIME) 1170 1171 /* 1172 * XFS Quota Timers 1173 * ================ 1174 * 1175 * Traditional quota grace period expiration timers are an unsigned 32-bit 1176 * seconds counter; time zero is the Unix epoch, Jan 1 00:00:01 UTC 1970. 1177 * Note that an expiration value of zero means that the quota limit has not 1178 * been reached, and therefore no expiration has been set. Therefore, the 1179 * ondisk min and max defined here can be used directly to constrain the incore 1180 * quota expiration timestamps on a Unix system. 1181 * 1182 * When bigtime is enabled, we trade two bits of precision to expand the 1183 * expiration timeout range to match that of big inode timestamps. The min and 1184 * max recorded here are the on-disk limits, not a Unix timestamp. 1185 * 1186 * The grace period for each quota type is stored in the root dquot (id = 0) 1187 * and is applied to a non-root dquot when it exceeds the soft or hard limits. 1188 * The length of quota grace periods are unsigned 32-bit quantities measured in 1189 * units of seconds. A value of zero means to use the default period. 1190 */ 1191 1192 /* 1193 * Smallest possible ondisk quota expiration value with traditional timestamps. 1194 * This corresponds exactly with the incore expiration Jan 1 00:00:01 UTC 1970. 1195 */ 1196 #define XFS_DQ_LEGACY_EXPIRY_MIN ((int64_t)1) 1197 1198 /* 1199 * Largest possible ondisk quota expiration value with traditional timestamps. 1200 * This corresponds exactly with the incore expiration Feb 7 06:28:15 UTC 2106. 1201 */ 1202 #define XFS_DQ_LEGACY_EXPIRY_MAX ((int64_t)U32_MAX) 1203 1204 /* 1205 * Smallest possible ondisk quota expiration value with bigtime timestamps. 1206 * This corresponds (after conversion to a Unix timestamp) with the incore 1207 * expiration of Jan 1 00:00:04 UTC 1970. 1208 */ 1209 #define XFS_DQ_BIGTIME_EXPIRY_MIN (XFS_DQ_LEGACY_EXPIRY_MIN) 1210 1211 /* 1212 * Largest supported ondisk quota expiration value with bigtime timestamps. 1213 * This corresponds (after conversion to a Unix timestamp) with an incore 1214 * expiration of Jul 2 20:20:24 UTC 2486. 1215 * 1216 * The ondisk field supports values up to -1U, which corresponds to an incore 1217 * expiration in 2514. This is beyond the maximum the bigtime inode timestamp, 1218 * so we cap the maximum bigtime quota expiration to the max inode timestamp. 1219 */ 1220 #define XFS_DQ_BIGTIME_EXPIRY_MAX ((int64_t)4074815106U) 1221 1222 /* 1223 * The following conversion factors assist in converting a quota expiration 1224 * timestamp between the incore and ondisk formats. 1225 */ 1226 #define XFS_DQ_BIGTIME_SHIFT (2) 1227 #define XFS_DQ_BIGTIME_SLACK ((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1) 1228 1229 /* Convert an incore quota expiration timestamp to an ondisk bigtime value. */ 1230 static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds) 1231 { 1232 /* 1233 * Round the expiration timestamp up to the nearest bigtime timestamp 1234 * that we can store, to give users the most time to fix problems. 1235 */ 1236 return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >> 1237 XFS_DQ_BIGTIME_SHIFT; 1238 } 1239 1240 /* Convert an ondisk bigtime quota expiration value to an incore timestamp. */ 1241 static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds) 1242 { 1243 return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT; 1244 } 1245 1246 /* 1247 * Default quota grace periods, ranging from zero (use the compiled defaults) 1248 * to ~136 years. These are applied to a non-root dquot that has exceeded 1249 * either limit. 1250 */ 1251 #define XFS_DQ_GRACE_MIN ((int64_t)0) 1252 #define XFS_DQ_GRACE_MAX ((int64_t)U32_MAX) 1253 1254 /* 1255 * This is the main portion of the on-disk representation of quota information 1256 * for a user. We pad this with some more expansion room to construct the on 1257 * disk structure. 1258 */ 1259 struct xfs_disk_dquot { 1260 __be16 d_magic; /* dquot magic = XFS_DQUOT_MAGIC */ 1261 __u8 d_version; /* dquot version */ 1262 __u8 d_type; /* XFS_DQTYPE_USER/PROJ/GROUP */ 1263 __be32 d_id; /* user,project,group id */ 1264 __be64 d_blk_hardlimit;/* absolute limit on disk blks */ 1265 __be64 d_blk_softlimit;/* preferred limit on disk blks */ 1266 __be64 d_ino_hardlimit;/* maximum # allocated inodes */ 1267 __be64 d_ino_softlimit;/* preferred inode limit */ 1268 __be64 d_bcount; /* disk blocks owned by the user */ 1269 __be64 d_icount; /* inodes owned by the user */ 1270 __be32 d_itimer; /* zero if within inode limits if not, 1271 this is when we refuse service */ 1272 __be32 d_btimer; /* similar to above; for disk blocks */ 1273 __be16 d_iwarns; /* warnings issued wrt num inodes */ 1274 __be16 d_bwarns; /* warnings issued wrt disk blocks */ 1275 __be32 d_pad0; /* 64 bit align */ 1276 __be64 d_rtb_hardlimit;/* absolute limit on realtime blks */ 1277 __be64 d_rtb_softlimit;/* preferred limit on RT disk blks */ 1278 __be64 d_rtbcount; /* realtime blocks owned */ 1279 __be32 d_rtbtimer; /* similar to above; for RT disk blocks */ 1280 __be16 d_rtbwarns; /* warnings issued wrt RT disk blocks */ 1281 __be16 d_pad; 1282 }; 1283 1284 /* 1285 * This is what goes on disk. This is separated from the xfs_disk_dquot because 1286 * carrying the unnecessary padding would be a waste of memory. 1287 */ 1288 struct xfs_dqblk { 1289 struct xfs_disk_dquot dd_diskdq; /* portion living incore as well */ 1290 char dd_fill[4];/* filling for posterity */ 1291 1292 /* 1293 * These two are only present on filesystems with the CRC bits set. 1294 */ 1295 __be32 dd_crc; /* checksum */ 1296 __be64 dd_lsn; /* last modification in log */ 1297 uuid_t dd_uuid; /* location information */ 1298 }; 1299 1300 #define XFS_DQUOT_CRC_OFF offsetof(struct xfs_dqblk, dd_crc) 1301 1302 /* 1303 * This defines the unit of allocation of dquots. 1304 * 1305 * Currently, it is just one file system block, and a 4K blk contains 30 1306 * (136 * 30 = 4080) dquots. It's probably not worth trying to make 1307 * this more dynamic. 1308 * 1309 * However, if this number is changed, we have to make sure that we don't 1310 * implicitly assume that we do allocations in chunks of a single filesystem 1311 * block in the dquot/xqm code. 1312 * 1313 * This is part of the ondisk format because the structure size is not a power 1314 * of two, which leaves slack at the end of the disk block. 1315 */ 1316 #define XFS_DQUOT_CLUSTER_SIZE_FSB (xfs_filblks_t)1 1317 1318 /* 1319 * Remote symlink format and access functions. 1320 */ 1321 #define XFS_SYMLINK_MAGIC 0x58534c4d /* XSLM */ 1322 1323 struct xfs_dsymlink_hdr { 1324 __be32 sl_magic; 1325 __be32 sl_offset; 1326 __be32 sl_bytes; 1327 __be32 sl_crc; 1328 uuid_t sl_uuid; 1329 __be64 sl_owner; 1330 __be64 sl_blkno; 1331 __be64 sl_lsn; 1332 }; 1333 1334 #define XFS_SYMLINK_CRC_OFF offsetof(struct xfs_dsymlink_hdr, sl_crc) 1335 1336 #define XFS_SYMLINK_MAXLEN 1024 1337 /* 1338 * The maximum pathlen is 1024 bytes. Since the minimum file system 1339 * blocksize is 512 bytes, we can get a max of 3 extents back from 1340 * bmapi when crc headers are taken into account. 1341 */ 1342 #define XFS_SYMLINK_MAPS 3 1343 1344 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize) \ 1345 ((bufsize) - (xfs_has_crc((mp)) ? \ 1346 sizeof(struct xfs_dsymlink_hdr) : 0)) 1347 1348 1349 /* 1350 * Allocation Btree format definitions 1351 * 1352 * There are two on-disk btrees, one sorted by blockno and one sorted 1353 * by blockcount and blockno. All blocks look the same to make the code 1354 * simpler; if we have time later, we'll make the optimizations. 1355 */ 1356 #define XFS_ABTB_MAGIC 0x41425442 /* 'ABTB' for bno tree */ 1357 #define XFS_ABTB_CRC_MAGIC 0x41423342 /* 'AB3B' */ 1358 #define XFS_ABTC_MAGIC 0x41425443 /* 'ABTC' for cnt tree */ 1359 #define XFS_ABTC_CRC_MAGIC 0x41423343 /* 'AB3C' */ 1360 1361 /* 1362 * Data record/key structure 1363 */ 1364 typedef struct xfs_alloc_rec { 1365 __be32 ar_startblock; /* starting block number */ 1366 __be32 ar_blockcount; /* count of free blocks */ 1367 } xfs_alloc_rec_t, xfs_alloc_key_t; 1368 1369 typedef struct xfs_alloc_rec_incore { 1370 xfs_agblock_t ar_startblock; /* starting block number */ 1371 xfs_extlen_t ar_blockcount; /* count of free blocks */ 1372 } xfs_alloc_rec_incore_t; 1373 1374 /* btree pointer type */ 1375 typedef __be32 xfs_alloc_ptr_t; 1376 1377 /* 1378 * Block numbers in the AG: 1379 * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3. 1380 */ 1381 #define XFS_BNO_BLOCK(mp) ((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1)) 1382 #define XFS_CNT_BLOCK(mp) ((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1)) 1383 1384 1385 /* 1386 * Inode Allocation Btree format definitions 1387 * 1388 * There is a btree for the inode map per allocation group. 1389 */ 1390 #define XFS_IBT_MAGIC 0x49414254 /* 'IABT' */ 1391 #define XFS_IBT_CRC_MAGIC 0x49414233 /* 'IAB3' */ 1392 #define XFS_FIBT_MAGIC 0x46494254 /* 'FIBT' */ 1393 #define XFS_FIBT_CRC_MAGIC 0x46494233 /* 'FIB3' */ 1394 1395 typedef uint64_t xfs_inofree_t; 1396 #define XFS_INODES_PER_CHUNK (NBBY * sizeof(xfs_inofree_t)) 1397 #define XFS_INODES_PER_CHUNK_LOG (XFS_NBBYLOG + 3) 1398 #define XFS_INOBT_ALL_FREE ((xfs_inofree_t)-1) 1399 #define XFS_INOBT_MASK(i) ((xfs_inofree_t)1 << (i)) 1400 1401 #define XFS_INOBT_HOLEMASK_FULL 0 /* holemask for full chunk */ 1402 #define XFS_INOBT_HOLEMASK_BITS (NBBY * sizeof(uint16_t)) 1403 #define XFS_INODES_PER_HOLEMASK_BIT \ 1404 (XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t))) 1405 1406 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n) 1407 { 1408 return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i; 1409 } 1410 1411 /* 1412 * The on-disk inode record structure has two formats. The original "full" 1413 * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount 1414 * and replaces the 3 high-order freecount bytes wth the holemask and inode 1415 * count. 1416 * 1417 * The holemask of the sparse record format allows an inode chunk to have holes 1418 * that refer to blocks not owned by the inode record. This facilitates inode 1419 * allocation in the event of severe free space fragmentation. 1420 */ 1421 typedef struct xfs_inobt_rec { 1422 __be32 ir_startino; /* starting inode number */ 1423 union { 1424 struct { 1425 __be32 ir_freecount; /* count of free inodes */ 1426 } f; 1427 struct { 1428 __be16 ir_holemask;/* hole mask for sparse chunks */ 1429 __u8 ir_count; /* total inode count */ 1430 __u8 ir_freecount; /* count of free inodes */ 1431 } sp; 1432 } ir_u; 1433 __be64 ir_free; /* free inode mask */ 1434 } xfs_inobt_rec_t; 1435 1436 typedef struct xfs_inobt_rec_incore { 1437 xfs_agino_t ir_startino; /* starting inode number */ 1438 uint16_t ir_holemask; /* hole mask for sparse chunks */ 1439 uint8_t ir_count; /* total inode count */ 1440 uint8_t ir_freecount; /* count of free inodes (set bits) */ 1441 xfs_inofree_t ir_free; /* free inode mask */ 1442 } xfs_inobt_rec_incore_t; 1443 1444 static inline bool xfs_inobt_issparse(uint16_t holemask) 1445 { 1446 /* non-zero holemask represents a sparse rec. */ 1447 return holemask; 1448 } 1449 1450 /* 1451 * Key structure 1452 */ 1453 typedef struct xfs_inobt_key { 1454 __be32 ir_startino; /* starting inode number */ 1455 } xfs_inobt_key_t; 1456 1457 /* btree pointer type */ 1458 typedef __be32 xfs_inobt_ptr_t; 1459 1460 /* 1461 * block numbers in the AG. 1462 */ 1463 #define XFS_IBT_BLOCK(mp) ((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1)) 1464 #define XFS_FIBT_BLOCK(mp) ((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1)) 1465 1466 /* 1467 * Reverse mapping btree format definitions 1468 * 1469 * There is a btree for the reverse map per allocation group 1470 */ 1471 #define XFS_RMAP_CRC_MAGIC 0x524d4233 /* 'RMB3' */ 1472 1473 /* 1474 * Ownership info for an extent. This is used to create reverse-mapping 1475 * entries. 1476 */ 1477 #define XFS_OWNER_INFO_ATTR_FORK (1 << 0) 1478 #define XFS_OWNER_INFO_BMBT_BLOCK (1 << 1) 1479 struct xfs_owner_info { 1480 uint64_t oi_owner; 1481 xfs_fileoff_t oi_offset; 1482 unsigned int oi_flags; 1483 }; 1484 1485 /* 1486 * Special owner types. 1487 * 1488 * Seeing as we only support up to 8EB, we have the upper bit of the owner field 1489 * to tell us we have a special owner value. We use these for static metadata 1490 * allocated at mkfs/growfs time, as well as for freespace management metadata. 1491 */ 1492 #define XFS_RMAP_OWN_NULL (-1ULL) /* No owner, for growfs */ 1493 #define XFS_RMAP_OWN_UNKNOWN (-2ULL) /* Unknown owner, for EFI recovery */ 1494 #define XFS_RMAP_OWN_FS (-3ULL) /* static fs metadata */ 1495 #define XFS_RMAP_OWN_LOG (-4ULL) /* static fs metadata */ 1496 #define XFS_RMAP_OWN_AG (-5ULL) /* AG freespace btree blocks */ 1497 #define XFS_RMAP_OWN_INOBT (-6ULL) /* Inode btree blocks */ 1498 #define XFS_RMAP_OWN_INODES (-7ULL) /* Inode chunk */ 1499 #define XFS_RMAP_OWN_REFC (-8ULL) /* refcount tree */ 1500 #define XFS_RMAP_OWN_COW (-9ULL) /* cow allocations */ 1501 #define XFS_RMAP_OWN_MIN (-10ULL) /* guard */ 1502 1503 #define XFS_RMAP_NON_INODE_OWNER(owner) (!!((owner) & (1ULL << 63))) 1504 1505 /* 1506 * Data record structure 1507 */ 1508 struct xfs_rmap_rec { 1509 __be32 rm_startblock; /* extent start block */ 1510 __be32 rm_blockcount; /* extent length */ 1511 __be64 rm_owner; /* extent owner */ 1512 __be64 rm_offset; /* offset within the owner */ 1513 }; 1514 1515 /* 1516 * rmap btree record 1517 * rm_offset:63 is the attribute fork flag 1518 * rm_offset:62 is the bmbt block flag 1519 * rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt) 1520 * rm_offset:54-60 aren't used and should be zero 1521 * rm_offset:0-53 is the block offset within the inode 1522 */ 1523 #define XFS_RMAP_OFF_ATTR_FORK ((uint64_t)1ULL << 63) 1524 #define XFS_RMAP_OFF_BMBT_BLOCK ((uint64_t)1ULL << 62) 1525 #define XFS_RMAP_OFF_UNWRITTEN ((uint64_t)1ULL << 61) 1526 1527 #define XFS_RMAP_LEN_MAX ((uint32_t)~0U) 1528 #define XFS_RMAP_OFF_FLAGS (XFS_RMAP_OFF_ATTR_FORK | \ 1529 XFS_RMAP_OFF_BMBT_BLOCK | \ 1530 XFS_RMAP_OFF_UNWRITTEN) 1531 #define XFS_RMAP_OFF_MASK ((uint64_t)0x3FFFFFFFFFFFFFULL) 1532 1533 #define XFS_RMAP_OFF(off) ((off) & XFS_RMAP_OFF_MASK) 1534 1535 #define XFS_RMAP_IS_BMBT_BLOCK(off) (!!((off) & XFS_RMAP_OFF_BMBT_BLOCK)) 1536 #define XFS_RMAP_IS_ATTR_FORK(off) (!!((off) & XFS_RMAP_OFF_ATTR_FORK)) 1537 #define XFS_RMAP_IS_UNWRITTEN(len) (!!((off) & XFS_RMAP_OFF_UNWRITTEN)) 1538 1539 #define RMAPBT_STARTBLOCK_BITLEN 32 1540 #define RMAPBT_BLOCKCOUNT_BITLEN 32 1541 #define RMAPBT_OWNER_BITLEN 64 1542 #define RMAPBT_ATTRFLAG_BITLEN 1 1543 #define RMAPBT_BMBTFLAG_BITLEN 1 1544 #define RMAPBT_EXNTFLAG_BITLEN 1 1545 #define RMAPBT_UNUSED_OFFSET_BITLEN 7 1546 #define RMAPBT_OFFSET_BITLEN 54 1547 1548 #define XFS_RMAP_ATTR_FORK (1 << 0) 1549 #define XFS_RMAP_BMBT_BLOCK (1 << 1) 1550 #define XFS_RMAP_UNWRITTEN (1 << 2) 1551 #define XFS_RMAP_KEY_FLAGS (XFS_RMAP_ATTR_FORK | \ 1552 XFS_RMAP_BMBT_BLOCK) 1553 #define XFS_RMAP_REC_FLAGS (XFS_RMAP_UNWRITTEN) 1554 struct xfs_rmap_irec { 1555 xfs_agblock_t rm_startblock; /* extent start block */ 1556 xfs_extlen_t rm_blockcount; /* extent length */ 1557 uint64_t rm_owner; /* extent owner */ 1558 uint64_t rm_offset; /* offset within the owner */ 1559 unsigned int rm_flags; /* state flags */ 1560 }; 1561 1562 /* 1563 * Key structure 1564 * 1565 * We don't use the length for lookups 1566 */ 1567 struct xfs_rmap_key { 1568 __be32 rm_startblock; /* extent start block */ 1569 __be64 rm_owner; /* extent owner */ 1570 __be64 rm_offset; /* offset within the owner */ 1571 } __attribute__((packed)); 1572 1573 /* btree pointer type */ 1574 typedef __be32 xfs_rmap_ptr_t; 1575 1576 #define XFS_RMAP_BLOCK(mp) \ 1577 (xfs_has_finobt(((mp))) ? \ 1578 XFS_FIBT_BLOCK(mp) + 1 : \ 1579 XFS_IBT_BLOCK(mp) + 1) 1580 1581 /* 1582 * Reference Count Btree format definitions 1583 * 1584 */ 1585 #define XFS_REFC_CRC_MAGIC 0x52334643 /* 'R3FC' */ 1586 1587 unsigned int xfs_refc_block(struct xfs_mount *mp); 1588 1589 /* 1590 * Data record/key structure 1591 * 1592 * Each record associates a range of physical blocks (starting at 1593 * rc_startblock and ending rc_blockcount blocks later) with a reference 1594 * count (rc_refcount). Extents that are being used to stage a copy on 1595 * write (CoW) operation are recorded in the refcount btree with a 1596 * refcount of 1. All other records must have a refcount > 1 and must 1597 * track an extent mapped only by file data forks. 1598 * 1599 * Extents with a single owner (attributes, metadata, non-shared file 1600 * data) are not tracked here. Free space is also not tracked here. 1601 * This is consistent with pre-reflink XFS. 1602 */ 1603 1604 /* 1605 * Extents that are being used to stage a copy on write are stored 1606 * in the refcount btree with a refcount of 1 and the upper bit set 1607 * on the startblock. This speeds up mount time deletion of stale 1608 * staging extents because they're all at the right side of the tree. 1609 */ 1610 #define XFS_REFC_COW_START ((xfs_agblock_t)(1U << 31)) 1611 #define REFCNTBT_COWFLAG_BITLEN 1 1612 #define REFCNTBT_AGBLOCK_BITLEN 31 1613 1614 struct xfs_refcount_rec { 1615 __be32 rc_startblock; /* starting block number */ 1616 __be32 rc_blockcount; /* count of blocks */ 1617 __be32 rc_refcount; /* number of inodes linked here */ 1618 }; 1619 1620 struct xfs_refcount_key { 1621 __be32 rc_startblock; /* starting block number */ 1622 }; 1623 1624 struct xfs_refcount_irec { 1625 xfs_agblock_t rc_startblock; /* starting block number */ 1626 xfs_extlen_t rc_blockcount; /* count of free blocks */ 1627 xfs_nlink_t rc_refcount; /* number of inodes linked here */ 1628 }; 1629 1630 #define MAXREFCOUNT ((xfs_nlink_t)~0U) 1631 #define MAXREFCEXTLEN ((xfs_extlen_t)~0U) 1632 1633 /* btree pointer type */ 1634 typedef __be32 xfs_refcount_ptr_t; 1635 1636 1637 /* 1638 * BMAP Btree format definitions 1639 * 1640 * This includes both the root block definition that sits inside an inode fork 1641 * and the record/pointer formats for the leaf/node in the blocks. 1642 */ 1643 #define XFS_BMAP_MAGIC 0x424d4150 /* 'BMAP' */ 1644 #define XFS_BMAP_CRC_MAGIC 0x424d4133 /* 'BMA3' */ 1645 1646 /* 1647 * Bmap root header, on-disk form only. 1648 */ 1649 typedef struct xfs_bmdr_block { 1650 __be16 bb_level; /* 0 is a leaf */ 1651 __be16 bb_numrecs; /* current # of data records */ 1652 } xfs_bmdr_block_t; 1653 1654 /* 1655 * Bmap btree record and extent descriptor. 1656 * l0:63 is an extent flag (value 1 indicates non-normal). 1657 * l0:9-62 are startoff. 1658 * l0:0-8 and l1:21-63 are startblock. 1659 * l1:0-20 are blockcount. 1660 */ 1661 #define BMBT_EXNTFLAG_BITLEN 1 1662 #define BMBT_STARTOFF_BITLEN 54 1663 #define BMBT_STARTBLOCK_BITLEN 52 1664 #define BMBT_BLOCKCOUNT_BITLEN 21 1665 1666 #define BMBT_STARTOFF_MASK ((1ULL << BMBT_STARTOFF_BITLEN) - 1) 1667 #define BMBT_BLOCKCOUNT_MASK ((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1) 1668 1669 #define XFS_MAX_BMBT_EXTLEN ((xfs_extlen_t)(BMBT_BLOCKCOUNT_MASK)) 1670 1671 /* 1672 * bmbt records have a file offset (block) field that is 54 bits wide, so this 1673 * is the largest xfs_fileoff_t that we ever expect to see. 1674 */ 1675 #define XFS_MAX_FILEOFF (BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK) 1676 1677 typedef struct xfs_bmbt_rec { 1678 __be64 l0, l1; 1679 } xfs_bmbt_rec_t; 1680 1681 typedef uint64_t xfs_bmbt_rec_base_t; /* use this for casts */ 1682 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t; 1683 1684 /* 1685 * Values and macros for delayed-allocation startblock fields. 1686 */ 1687 #define STARTBLOCKVALBITS 17 1688 #define STARTBLOCKMASKBITS (15 + 20) 1689 #define STARTBLOCKMASK \ 1690 (((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS) 1691 1692 static inline int isnullstartblock(xfs_fsblock_t x) 1693 { 1694 return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK; 1695 } 1696 1697 static inline xfs_fsblock_t nullstartblock(int k) 1698 { 1699 ASSERT(k < (1 << STARTBLOCKVALBITS)); 1700 return STARTBLOCKMASK | (k); 1701 } 1702 1703 static inline xfs_filblks_t startblockval(xfs_fsblock_t x) 1704 { 1705 return (xfs_filblks_t)((x) & ~STARTBLOCKMASK); 1706 } 1707 1708 /* 1709 * Key structure for non-leaf levels of the tree. 1710 */ 1711 typedef struct xfs_bmbt_key { 1712 __be64 br_startoff; /* starting file offset */ 1713 } xfs_bmbt_key_t, xfs_bmdr_key_t; 1714 1715 /* btree pointer type */ 1716 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t; 1717 1718 1719 /* 1720 * Generic Btree block format definitions 1721 * 1722 * This is a combination of the actual format used on disk for short and long 1723 * format btrees. The first three fields are shared by both format, but the 1724 * pointers are different and should be used with care. 1725 * 1726 * To get the size of the actual short or long form headers please use the size 1727 * macros below. Never use sizeof(xfs_btree_block). 1728 * 1729 * The blkno, crc, lsn, owner and uuid fields are only available in filesystems 1730 * with the crc feature bit, and all accesses to them must be conditional on 1731 * that flag. 1732 */ 1733 /* short form block header */ 1734 struct xfs_btree_block_shdr { 1735 __be32 bb_leftsib; 1736 __be32 bb_rightsib; 1737 1738 __be64 bb_blkno; 1739 __be64 bb_lsn; 1740 uuid_t bb_uuid; 1741 __be32 bb_owner; 1742 __le32 bb_crc; 1743 }; 1744 1745 /* long form block header */ 1746 struct xfs_btree_block_lhdr { 1747 __be64 bb_leftsib; 1748 __be64 bb_rightsib; 1749 1750 __be64 bb_blkno; 1751 __be64 bb_lsn; 1752 uuid_t bb_uuid; 1753 __be64 bb_owner; 1754 __le32 bb_crc; 1755 __be32 bb_pad; /* padding for alignment */ 1756 }; 1757 1758 struct xfs_btree_block { 1759 __be32 bb_magic; /* magic number for block type */ 1760 __be16 bb_level; /* 0 is a leaf */ 1761 __be16 bb_numrecs; /* current # of data records */ 1762 union { 1763 struct xfs_btree_block_shdr s; 1764 struct xfs_btree_block_lhdr l; 1765 } bb_u; /* rest */ 1766 }; 1767 1768 /* size of a short form block */ 1769 #define XFS_BTREE_SBLOCK_LEN \ 1770 (offsetof(struct xfs_btree_block, bb_u) + \ 1771 offsetof(struct xfs_btree_block_shdr, bb_blkno)) 1772 /* size of a long form block */ 1773 #define XFS_BTREE_LBLOCK_LEN \ 1774 (offsetof(struct xfs_btree_block, bb_u) + \ 1775 offsetof(struct xfs_btree_block_lhdr, bb_blkno)) 1776 1777 /* sizes of CRC enabled btree blocks */ 1778 #define XFS_BTREE_SBLOCK_CRC_LEN \ 1779 (offsetof(struct xfs_btree_block, bb_u) + \ 1780 sizeof(struct xfs_btree_block_shdr)) 1781 #define XFS_BTREE_LBLOCK_CRC_LEN \ 1782 (offsetof(struct xfs_btree_block, bb_u) + \ 1783 sizeof(struct xfs_btree_block_lhdr)) 1784 1785 #define XFS_BTREE_SBLOCK_CRC_OFF \ 1786 offsetof(struct xfs_btree_block, bb_u.s.bb_crc) 1787 #define XFS_BTREE_LBLOCK_CRC_OFF \ 1788 offsetof(struct xfs_btree_block, bb_u.l.bb_crc) 1789 1790 /* 1791 * On-disk XFS access control list structure. 1792 */ 1793 struct xfs_acl_entry { 1794 __be32 ae_tag; 1795 __be32 ae_id; 1796 __be16 ae_perm; 1797 __be16 ae_pad; /* fill the implicit hole in the structure */ 1798 }; 1799 1800 struct xfs_acl { 1801 __be32 acl_cnt; 1802 struct xfs_acl_entry acl_entry[]; 1803 }; 1804 1805 /* 1806 * The number of ACL entries allowed is defined by the on-disk format. 1807 * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is 1808 * limited only by the maximum size of the xattr that stores the information. 1809 */ 1810 #define XFS_ACL_MAX_ENTRIES(mp) \ 1811 (xfs_has_crc(mp) \ 1812 ? (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \ 1813 sizeof(struct xfs_acl_entry) \ 1814 : 25) 1815 1816 #define XFS_ACL_SIZE(cnt) \ 1817 (sizeof(struct xfs_acl) + \ 1818 sizeof(struct xfs_acl_entry) * cnt) 1819 1820 #define XFS_ACL_MAX_SIZE(mp) \ 1821 XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp))) 1822 1823 1824 /* On-disk XFS extended attribute names */ 1825 #define SGI_ACL_FILE "SGI_ACL_FILE" 1826 #define SGI_ACL_DEFAULT "SGI_ACL_DEFAULT" 1827 #define SGI_ACL_FILE_SIZE (sizeof(SGI_ACL_FILE)-1) 1828 #define SGI_ACL_DEFAULT_SIZE (sizeof(SGI_ACL_DEFAULT)-1) 1829 1830 #endif /* __XFS_FORMAT_H__ */ 1831