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_rtbxlen_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_EXCHRANGE (1 << 6) /* exchangerange supported */ 377 #define XFS_SB_FEAT_INCOMPAT_ALL \ 378 (XFS_SB_FEAT_INCOMPAT_FTYPE | \ 379 XFS_SB_FEAT_INCOMPAT_SPINODES | \ 380 XFS_SB_FEAT_INCOMPAT_META_UUID | \ 381 XFS_SB_FEAT_INCOMPAT_BIGTIME | \ 382 XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR | \ 383 XFS_SB_FEAT_INCOMPAT_NREXT64 | \ 384 XFS_SB_FEAT_INCOMPAT_EXCHRANGE) 385 386 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN ~XFS_SB_FEAT_INCOMPAT_ALL 387 static inline bool 388 xfs_sb_has_incompat_feature( 389 struct xfs_sb *sbp, 390 uint32_t feature) 391 { 392 return (sbp->sb_features_incompat & feature) != 0; 393 } 394 395 #define XFS_SB_FEAT_INCOMPAT_LOG_XATTRS (1 << 0) /* Delayed Attributes */ 396 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL \ 397 (XFS_SB_FEAT_INCOMPAT_LOG_XATTRS) 398 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN ~XFS_SB_FEAT_INCOMPAT_LOG_ALL 399 static inline bool 400 xfs_sb_has_incompat_log_feature( 401 struct xfs_sb *sbp, 402 uint32_t feature) 403 { 404 return (sbp->sb_features_log_incompat & feature) != 0; 405 } 406 407 static inline void 408 xfs_sb_remove_incompat_log_features( 409 struct xfs_sb *sbp) 410 { 411 sbp->sb_features_log_incompat &= ~XFS_SB_FEAT_INCOMPAT_LOG_ALL; 412 } 413 414 static inline void 415 xfs_sb_add_incompat_log_features( 416 struct xfs_sb *sbp, 417 unsigned int features) 418 { 419 sbp->sb_features_log_incompat |= features; 420 } 421 422 static inline bool xfs_sb_version_haslogxattrs(struct xfs_sb *sbp) 423 { 424 return xfs_sb_is_v5(sbp) && (sbp->sb_features_log_incompat & 425 XFS_SB_FEAT_INCOMPAT_LOG_XATTRS); 426 } 427 428 static inline bool 429 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino) 430 { 431 return (ino == sbp->sb_uquotino || 432 ino == sbp->sb_gquotino || 433 ino == sbp->sb_pquotino); 434 } 435 436 #define XFS_SB_DADDR ((xfs_daddr_t)0) /* daddr in filesystem/ag */ 437 #define XFS_SB_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_SB_DADDR) 438 439 #define XFS_HDR_BLOCK(mp,d) ((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d)) 440 #define XFS_DADDR_TO_FSB(mp,d) XFS_AGB_TO_FSB(mp, \ 441 xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d)) 442 #define XFS_FSB_TO_DADDR(mp,fsbno) XFS_AGB_TO_DADDR(mp, \ 443 XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno)) 444 445 /* 446 * File system sector to basic block conversions. 447 */ 448 #define XFS_FSS_TO_BB(mp,sec) ((sec) << (mp)->m_sectbb_log) 449 450 /* 451 * File system block to basic block conversions. 452 */ 453 #define XFS_FSB_TO_BB(mp,fsbno) ((fsbno) << (mp)->m_blkbb_log) 454 #define XFS_BB_TO_FSB(mp,bb) \ 455 (((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log) 456 #define XFS_BB_TO_FSBT(mp,bb) ((bb) >> (mp)->m_blkbb_log) 457 458 /* 459 * File system block to byte conversions. 460 */ 461 #define XFS_FSB_TO_B(mp,fsbno) ((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog) 462 #define XFS_B_TO_FSB(mp,b) \ 463 ((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog) 464 #define XFS_B_TO_FSBT(mp,b) (((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog) 465 466 /* 467 * Allocation group header 468 * 469 * This is divided into three structures, placed in sequential 512-byte 470 * buffers after a copy of the superblock (also in a 512-byte buffer). 471 */ 472 #define XFS_AGF_MAGIC 0x58414746 /* 'XAGF' */ 473 #define XFS_AGI_MAGIC 0x58414749 /* 'XAGI' */ 474 #define XFS_AGFL_MAGIC 0x5841464c /* 'XAFL' */ 475 #define XFS_AGF_VERSION 1 476 #define XFS_AGI_VERSION 1 477 478 #define XFS_AGF_GOOD_VERSION(v) ((v) == XFS_AGF_VERSION) 479 #define XFS_AGI_GOOD_VERSION(v) ((v) == XFS_AGI_VERSION) 480 481 /* 482 * agf_cnt_level in the first AGF overlaps the EFS superblock's magic number. 483 * Since the magic numbers valid for EFS are > 64k, our value cannot be confused 484 * for an EFS superblock. 485 */ 486 487 typedef struct xfs_agf { 488 /* 489 * Common allocation group header information 490 */ 491 __be32 agf_magicnum; /* magic number == XFS_AGF_MAGIC */ 492 __be32 agf_versionnum; /* header version == XFS_AGF_VERSION */ 493 __be32 agf_seqno; /* sequence # starting from 0 */ 494 __be32 agf_length; /* size in blocks of a.g. */ 495 /* 496 * Freespace and rmap information 497 */ 498 __be32 agf_bno_root; /* bnobt root block */ 499 __be32 agf_cnt_root; /* cntbt root block */ 500 __be32 agf_rmap_root; /* rmapbt root block */ 501 502 __be32 agf_bno_level; /* bnobt btree levels */ 503 __be32 agf_cnt_level; /* cntbt btree levels */ 504 __be32 agf_rmap_level; /* rmapbt btree levels */ 505 506 __be32 agf_flfirst; /* first freelist block's index */ 507 __be32 agf_fllast; /* last freelist block's index */ 508 __be32 agf_flcount; /* count of blocks in freelist */ 509 __be32 agf_freeblks; /* total free blocks */ 510 511 __be32 agf_longest; /* longest free space */ 512 __be32 agf_btreeblks; /* # of blocks held in AGF btrees */ 513 uuid_t agf_uuid; /* uuid of filesystem */ 514 515 __be32 agf_rmap_blocks; /* rmapbt blocks used */ 516 __be32 agf_refcount_blocks; /* refcountbt blocks used */ 517 518 __be32 agf_refcount_root; /* refcount tree root block */ 519 __be32 agf_refcount_level; /* refcount btree levels */ 520 521 /* 522 * reserve some contiguous space for future logged fields before we add 523 * the unlogged fields. This makes the range logging via flags and 524 * structure offsets much simpler. 525 */ 526 __be64 agf_spare64[14]; 527 528 /* unlogged fields, written during buffer writeback. */ 529 __be64 agf_lsn; /* last write sequence */ 530 __be32 agf_crc; /* crc of agf sector */ 531 __be32 agf_spare2; 532 533 /* structure must be padded to 64 bit alignment */ 534 } xfs_agf_t; 535 536 #define XFS_AGF_CRC_OFF offsetof(struct xfs_agf, agf_crc) 537 538 #define XFS_AGF_MAGICNUM (1u << 0) 539 #define XFS_AGF_VERSIONNUM (1u << 1) 540 #define XFS_AGF_SEQNO (1u << 2) 541 #define XFS_AGF_LENGTH (1u << 3) 542 #define XFS_AGF_ROOTS (1u << 4) 543 #define XFS_AGF_LEVELS (1u << 5) 544 #define XFS_AGF_FLFIRST (1u << 6) 545 #define XFS_AGF_FLLAST (1u << 7) 546 #define XFS_AGF_FLCOUNT (1u << 8) 547 #define XFS_AGF_FREEBLKS (1u << 9) 548 #define XFS_AGF_LONGEST (1u << 10) 549 #define XFS_AGF_BTREEBLKS (1u << 11) 550 #define XFS_AGF_UUID (1u << 12) 551 #define XFS_AGF_RMAP_BLOCKS (1u << 13) 552 #define XFS_AGF_REFCOUNT_BLOCKS (1u << 14) 553 #define XFS_AGF_REFCOUNT_ROOT (1u << 15) 554 #define XFS_AGF_REFCOUNT_LEVEL (1u << 16) 555 #define XFS_AGF_SPARE64 (1u << 17) 556 #define XFS_AGF_NUM_BITS 18 557 #define XFS_AGF_ALL_BITS ((1u << XFS_AGF_NUM_BITS) - 1) 558 559 #define XFS_AGF_FLAGS \ 560 { XFS_AGF_MAGICNUM, "MAGICNUM" }, \ 561 { XFS_AGF_VERSIONNUM, "VERSIONNUM" }, \ 562 { XFS_AGF_SEQNO, "SEQNO" }, \ 563 { XFS_AGF_LENGTH, "LENGTH" }, \ 564 { XFS_AGF_ROOTS, "ROOTS" }, \ 565 { XFS_AGF_LEVELS, "LEVELS" }, \ 566 { XFS_AGF_FLFIRST, "FLFIRST" }, \ 567 { XFS_AGF_FLLAST, "FLLAST" }, \ 568 { XFS_AGF_FLCOUNT, "FLCOUNT" }, \ 569 { XFS_AGF_FREEBLKS, "FREEBLKS" }, \ 570 { XFS_AGF_LONGEST, "LONGEST" }, \ 571 { XFS_AGF_BTREEBLKS, "BTREEBLKS" }, \ 572 { XFS_AGF_UUID, "UUID" }, \ 573 { XFS_AGF_RMAP_BLOCKS, "RMAP_BLOCKS" }, \ 574 { XFS_AGF_REFCOUNT_BLOCKS, "REFCOUNT_BLOCKS" }, \ 575 { XFS_AGF_REFCOUNT_ROOT, "REFCOUNT_ROOT" }, \ 576 { XFS_AGF_REFCOUNT_LEVEL, "REFCOUNT_LEVEL" }, \ 577 { XFS_AGF_SPARE64, "SPARE64" } 578 579 /* disk block (xfs_daddr_t) in the AG */ 580 #define XFS_AGF_DADDR(mp) ((xfs_daddr_t)(1 << (mp)->m_sectbb_log)) 581 #define XFS_AGF_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp)) 582 583 /* 584 * Size of the unlinked inode hash table in the agi. 585 */ 586 #define XFS_AGI_UNLINKED_BUCKETS 64 587 588 typedef struct xfs_agi { 589 /* 590 * Common allocation group header information 591 */ 592 __be32 agi_magicnum; /* magic number == XFS_AGI_MAGIC */ 593 __be32 agi_versionnum; /* header version == XFS_AGI_VERSION */ 594 __be32 agi_seqno; /* sequence # starting from 0 */ 595 __be32 agi_length; /* size in blocks of a.g. */ 596 /* 597 * Inode information 598 * Inodes are mapped by interpreting the inode number, so no 599 * mapping data is needed here. 600 */ 601 __be32 agi_count; /* count of allocated inodes */ 602 __be32 agi_root; /* root of inode btree */ 603 __be32 agi_level; /* levels in inode btree */ 604 __be32 agi_freecount; /* number of free inodes */ 605 606 __be32 agi_newino; /* new inode just allocated */ 607 __be32 agi_dirino; /* last directory inode chunk */ 608 /* 609 * Hash table of inodes which have been unlinked but are 610 * still being referenced. 611 */ 612 __be32 agi_unlinked[XFS_AGI_UNLINKED_BUCKETS]; 613 /* 614 * This marks the end of logging region 1 and start of logging region 2. 615 */ 616 uuid_t agi_uuid; /* uuid of filesystem */ 617 __be32 agi_crc; /* crc of agi sector */ 618 __be32 agi_pad32; 619 __be64 agi_lsn; /* last write sequence */ 620 621 __be32 agi_free_root; /* root of the free inode btree */ 622 __be32 agi_free_level;/* levels in free inode btree */ 623 624 __be32 agi_iblocks; /* inobt blocks used */ 625 __be32 agi_fblocks; /* finobt blocks used */ 626 627 /* structure must be padded to 64 bit alignment */ 628 } xfs_agi_t; 629 630 #define XFS_AGI_CRC_OFF offsetof(struct xfs_agi, agi_crc) 631 632 #define XFS_AGI_MAGICNUM (1u << 0) 633 #define XFS_AGI_VERSIONNUM (1u << 1) 634 #define XFS_AGI_SEQNO (1u << 2) 635 #define XFS_AGI_LENGTH (1u << 3) 636 #define XFS_AGI_COUNT (1u << 4) 637 #define XFS_AGI_ROOT (1u << 5) 638 #define XFS_AGI_LEVEL (1u << 6) 639 #define XFS_AGI_FREECOUNT (1u << 7) 640 #define XFS_AGI_NEWINO (1u << 8) 641 #define XFS_AGI_DIRINO (1u << 9) 642 #define XFS_AGI_UNLINKED (1u << 10) 643 #define XFS_AGI_NUM_BITS_R1 11 /* end of the 1st agi logging region */ 644 #define XFS_AGI_ALL_BITS_R1 ((1u << XFS_AGI_NUM_BITS_R1) - 1) 645 #define XFS_AGI_FREE_ROOT (1u << 11) 646 #define XFS_AGI_FREE_LEVEL (1u << 12) 647 #define XFS_AGI_IBLOCKS (1u << 13) /* both inobt/finobt block counters */ 648 #define XFS_AGI_NUM_BITS_R2 14 649 650 /* disk block (xfs_daddr_t) in the AG */ 651 #define XFS_AGI_DADDR(mp) ((xfs_daddr_t)(2 << (mp)->m_sectbb_log)) 652 #define XFS_AGI_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp)) 653 654 /* 655 * The third a.g. block contains the a.g. freelist, an array 656 * of block pointers to blocks owned by the allocation btree code. 657 */ 658 #define XFS_AGFL_DADDR(mp) ((xfs_daddr_t)(3 << (mp)->m_sectbb_log)) 659 #define XFS_AGFL_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp)) 660 #define XFS_BUF_TO_AGFL(bp) ((struct xfs_agfl *)((bp)->b_addr)) 661 662 struct xfs_agfl { 663 __be32 agfl_magicnum; 664 __be32 agfl_seqno; 665 uuid_t agfl_uuid; 666 __be64 agfl_lsn; 667 __be32 agfl_crc; 668 } __attribute__((packed)); 669 670 #define XFS_AGFL_CRC_OFF offsetof(struct xfs_agfl, agfl_crc) 671 672 #define XFS_AGB_TO_FSB(mp,agno,agbno) \ 673 (((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno)) 674 #define XFS_FSB_TO_AGNO(mp,fsbno) \ 675 ((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog)) 676 #define XFS_FSB_TO_AGBNO(mp,fsbno) \ 677 ((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog))) 678 #define XFS_AGB_TO_DADDR(mp,agno,agbno) \ 679 ((xfs_daddr_t)XFS_FSB_TO_BB(mp, \ 680 (xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno))) 681 #define XFS_AG_DADDR(mp,agno,d) (XFS_AGB_TO_DADDR(mp, agno, 0) + (d)) 682 683 /* 684 * For checking for bad ranges of xfs_daddr_t's, covering multiple 685 * allocation groups or a single xfs_daddr_t that's a superblock copy. 686 */ 687 #define XFS_AG_CHECK_DADDR(mp,d,len) \ 688 ((len) == 1 ? \ 689 ASSERT((d) == XFS_SB_DADDR || \ 690 xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \ 691 ASSERT(xfs_daddr_to_agno(mp, d) == \ 692 xfs_daddr_to_agno(mp, (d) + (len) - 1))) 693 694 /* 695 * Realtime bitmap information is accessed by the word, which is currently 696 * stored in host-endian format. 697 */ 698 union xfs_rtword_raw { 699 __u32 old; 700 }; 701 702 /* 703 * Realtime summary counts are accessed by the word, which is currently 704 * stored in host-endian format. 705 */ 706 union xfs_suminfo_raw { 707 __u32 old; 708 }; 709 710 /* 711 * XFS Timestamps 712 * ============== 713 * 714 * Traditional ondisk inode timestamps consist of signed 32-bit counters for 715 * seconds and nanoseconds; time zero is the Unix epoch, Jan 1 00:00:00 UTC 716 * 1970, which means that the timestamp epoch is the same as the Unix epoch. 717 * Therefore, the ondisk min and max defined here can be used directly to 718 * constrain the incore timestamps on a Unix system. Note that we actually 719 * encode a __be64 value on disk. 720 * 721 * When the bigtime feature is enabled, ondisk inode timestamps become an 722 * unsigned 64-bit nanoseconds counter. This means that the bigtime inode 723 * timestamp epoch is the start of the classic timestamp range, which is 724 * Dec 13 20:45:52 UTC 1901. Because the epochs are not the same, callers 725 * /must/ use the bigtime conversion functions when encoding and decoding raw 726 * timestamps. 727 */ 728 typedef __be64 xfs_timestamp_t; 729 730 /* Legacy timestamp encoding format. */ 731 struct xfs_legacy_timestamp { 732 __be32 t_sec; /* timestamp seconds */ 733 __be32 t_nsec; /* timestamp nanoseconds */ 734 }; 735 736 /* 737 * Smallest possible ondisk seconds value with traditional timestamps. This 738 * corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901. 739 */ 740 #define XFS_LEGACY_TIME_MIN ((int64_t)S32_MIN) 741 742 /* 743 * Largest possible ondisk seconds value with traditional timestamps. This 744 * corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038. 745 */ 746 #define XFS_LEGACY_TIME_MAX ((int64_t)S32_MAX) 747 748 /* 749 * Smallest possible ondisk seconds value with bigtime timestamps. This 750 * corresponds (after conversion to a Unix timestamp) with the traditional 751 * minimum timestamp of Dec 13 20:45:52 UTC 1901. 752 */ 753 #define XFS_BIGTIME_TIME_MIN ((int64_t)0) 754 755 /* 756 * Largest supported ondisk seconds value with bigtime timestamps. This 757 * corresponds (after conversion to a Unix timestamp) with an incore timestamp 758 * of Jul 2 20:20:24 UTC 2486. 759 * 760 * We round down the ondisk limit so that the bigtime quota and inode max 761 * timestamps will be the same. 762 */ 763 #define XFS_BIGTIME_TIME_MAX ((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL)) 764 765 /* 766 * Bigtime epoch is set exactly to the minimum time value that a traditional 767 * 32-bit timestamp can represent when using the Unix epoch as a reference. 768 * Hence the Unix epoch is at a fixed offset into the supported bigtime 769 * timestamp range. 770 * 771 * The bigtime epoch also matches the minimum value an on-disk 32-bit XFS 772 * timestamp can represent so we will not lose any fidelity in converting 773 * to/from unix and bigtime timestamps. 774 * 775 * The following conversion factor converts a seconds counter from the Unix 776 * epoch to the bigtime epoch. 777 */ 778 #define XFS_BIGTIME_EPOCH_OFFSET (-(int64_t)S32_MIN) 779 780 /* Convert a timestamp from the Unix epoch to the bigtime epoch. */ 781 static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds) 782 { 783 return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET; 784 } 785 786 /* Convert a timestamp from the bigtime epoch to the Unix epoch. */ 787 static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds) 788 { 789 return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET; 790 } 791 792 /* 793 * On-disk inode structure. 794 * 795 * This is just the header or "dinode core", the inode is expanded to fill a 796 * variable size the leftover area split into a data and an attribute fork. 797 * The format of the data and attribute fork depends on the format of the 798 * inode as indicated by di_format and di_aformat. To access the data and 799 * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros 800 * below. 801 * 802 * There is a very similar struct xfs_log_dinode which matches the layout of 803 * this structure, but is kept in native format instead of big endian. 804 * 805 * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed 806 * padding field for v3 inodes. 807 */ 808 #define XFS_DINODE_MAGIC 0x494e /* 'IN' */ 809 struct xfs_dinode { 810 __be16 di_magic; /* inode magic # = XFS_DINODE_MAGIC */ 811 __be16 di_mode; /* mode and type of file */ 812 __u8 di_version; /* inode version */ 813 __u8 di_format; /* format of di_c data */ 814 __be16 di_onlink; /* old number of links to file */ 815 __be32 di_uid; /* owner's user id */ 816 __be32 di_gid; /* owner's group id */ 817 __be32 di_nlink; /* number of links to file */ 818 __be16 di_projid_lo; /* lower part of owner's project id */ 819 __be16 di_projid_hi; /* higher part owner's project id */ 820 union { 821 /* Number of data fork extents if NREXT64 is set */ 822 __be64 di_big_nextents; 823 824 /* Padding for V3 inodes without NREXT64 set. */ 825 __be64 di_v3_pad; 826 827 /* Padding and inode flush counter for V2 inodes. */ 828 struct { 829 __u8 di_v2_pad[6]; 830 __be16 di_flushiter; 831 }; 832 }; 833 xfs_timestamp_t di_atime; /* time last accessed */ 834 xfs_timestamp_t di_mtime; /* time last modified */ 835 xfs_timestamp_t di_ctime; /* time created/inode modified */ 836 __be64 di_size; /* number of bytes in file */ 837 __be64 di_nblocks; /* # of direct & btree blocks used */ 838 __be32 di_extsize; /* basic/minimum extent size for file */ 839 union { 840 /* 841 * For V2 inodes and V3 inodes without NREXT64 set, this 842 * is the number of data and attr fork extents. 843 */ 844 struct { 845 __be32 di_nextents; 846 __be16 di_anextents; 847 } __packed; 848 849 /* Number of attr fork extents if NREXT64 is set. */ 850 struct { 851 __be32 di_big_anextents; 852 __be16 di_nrext64_pad; 853 } __packed; 854 } __packed; 855 __u8 di_forkoff; /* attr fork offs, <<3 for 64b align */ 856 __s8 di_aformat; /* format of attr fork's data */ 857 __be32 di_dmevmask; /* DMIG event mask */ 858 __be16 di_dmstate; /* DMIG state info */ 859 __be16 di_flags; /* random flags, XFS_DIFLAG_... */ 860 __be32 di_gen; /* generation number */ 861 862 /* di_next_unlinked is the only non-core field in the old dinode */ 863 __be32 di_next_unlinked;/* agi unlinked list ptr */ 864 865 /* start of the extended dinode, writable fields */ 866 __le32 di_crc; /* CRC of the inode */ 867 __be64 di_changecount; /* number of attribute changes */ 868 __be64 di_lsn; /* flush sequence */ 869 __be64 di_flags2; /* more random flags */ 870 __be32 di_cowextsize; /* basic cow extent size for file */ 871 __u8 di_pad2[12]; /* more padding for future expansion */ 872 873 /* fields only written to during inode creation */ 874 xfs_timestamp_t di_crtime; /* time created */ 875 __be64 di_ino; /* inode number */ 876 uuid_t di_uuid; /* UUID of the filesystem */ 877 878 /* structure must be padded to 64 bit alignment */ 879 }; 880 881 #define XFS_DINODE_CRC_OFF offsetof(struct xfs_dinode, di_crc) 882 883 #define DI_MAX_FLUSH 0xffff 884 885 /* 886 * Size of the core inode on disk. Version 1 and 2 inodes have 887 * the same size, but version 3 has grown a few additional fields. 888 */ 889 static inline uint xfs_dinode_size(int version) 890 { 891 if (version == 3) 892 return sizeof(struct xfs_dinode); 893 return offsetof(struct xfs_dinode, di_crc); 894 } 895 896 /* 897 * The 32 bit link count in the inode theoretically maxes out at UINT_MAX. 898 * Since the pathconf interface is signed, we use 2^31 - 1 instead. 899 */ 900 #define XFS_MAXLINK ((1U << 31) - 1U) 901 902 /* 903 * Any file that hits the maximum ondisk link count should be pinned to avoid 904 * a use-after-free situation. 905 */ 906 #define XFS_NLINK_PINNED (~0U) 907 908 /* 909 * Values for di_format 910 * 911 * This enum is used in string mapping in xfs_trace.h; please keep the 912 * TRACE_DEFINE_ENUMs for it up to date. 913 */ 914 enum xfs_dinode_fmt { 915 XFS_DINODE_FMT_DEV, /* xfs_dev_t */ 916 XFS_DINODE_FMT_LOCAL, /* bulk data */ 917 XFS_DINODE_FMT_EXTENTS, /* struct xfs_bmbt_rec */ 918 XFS_DINODE_FMT_BTREE, /* struct xfs_bmdr_block */ 919 XFS_DINODE_FMT_UUID /* added long ago, but never used */ 920 }; 921 922 #define XFS_INODE_FORMAT_STR \ 923 { XFS_DINODE_FMT_DEV, "dev" }, \ 924 { XFS_DINODE_FMT_LOCAL, "local" }, \ 925 { XFS_DINODE_FMT_EXTENTS, "extent" }, \ 926 { XFS_DINODE_FMT_BTREE, "btree" }, \ 927 { XFS_DINODE_FMT_UUID, "uuid" } 928 929 /* 930 * Max values for extnum and aextnum. 931 * 932 * The original on-disk extent counts were held in signed fields, resulting in 933 * maximum extent counts of 2^31 and 2^15 for the data and attr forks 934 * respectively. Similarly the maximum extent length is limited to 2^21 blocks 935 * by the 21-bit wide blockcount field of a BMBT extent record. 936 * 937 * The newly introduced data fork extent counter can hold a 64-bit value, 938 * however the maximum number of extents in a file is also limited to 2^54 939 * extents by the 54-bit wide startoff field of a BMBT extent record. 940 * 941 * It is further limited by the maximum supported file size of 2^63 942 * *bytes*. This leads to a maximum extent count for maximally sized filesystem 943 * blocks (64kB) of: 944 * 945 * 2^63 bytes / 2^16 bytes per block = 2^47 blocks 946 * 947 * Rounding up 47 to the nearest multiple of bits-per-byte results in 48. Hence 948 * 2^48 was chosen as the maximum data fork extent count. 949 * 950 * The maximum file size that can be represented by the data fork extent counter 951 * in the worst case occurs when all extents are 1 block in length and each 952 * block is 1KB in size. 953 * 954 * With XFS_MAX_EXTCNT_DATA_FORK_SMALL representing maximum extent count and 955 * with 1KB sized blocks, a file can reach upto, 956 * 1KB * (2^31) = 2TB 957 * 958 * This is much larger than the theoretical maximum size of a directory 959 * i.e. XFS_DIR2_SPACE_SIZE * XFS_DIR2_MAX_SPACES = ~96GB. 960 * 961 * Hence, a directory inode can never overflow its data fork extent counter. 962 */ 963 #define XFS_MAX_EXTCNT_DATA_FORK_LARGE ((xfs_extnum_t)((1ULL << 48) - 1)) 964 #define XFS_MAX_EXTCNT_ATTR_FORK_LARGE ((xfs_extnum_t)((1ULL << 32) - 1)) 965 #define XFS_MAX_EXTCNT_DATA_FORK_SMALL ((xfs_extnum_t)((1ULL << 31) - 1)) 966 #define XFS_MAX_EXTCNT_ATTR_FORK_SMALL ((xfs_extnum_t)((1ULL << 15) - 1)) 967 968 /* 969 * When we upgrade an inode to the large extent counts, the maximum value by 970 * which the extent count can increase is bound by the change in size of the 971 * on-disk field. No upgrade operation should ever be adding more than a few 972 * tens of extents, so if we get a really large value it is a sign of a code bug 973 * or corruption. 974 */ 975 #define XFS_MAX_EXTCNT_UPGRADE_NR \ 976 min(XFS_MAX_EXTCNT_ATTR_FORK_LARGE - XFS_MAX_EXTCNT_ATTR_FORK_SMALL, \ 977 XFS_MAX_EXTCNT_DATA_FORK_LARGE - XFS_MAX_EXTCNT_DATA_FORK_SMALL) 978 979 /* 980 * Inode minimum and maximum sizes. 981 */ 982 #define XFS_DINODE_MIN_LOG 8 983 #define XFS_DINODE_MAX_LOG 11 984 #define XFS_DINODE_MIN_SIZE (1 << XFS_DINODE_MIN_LOG) 985 #define XFS_DINODE_MAX_SIZE (1 << XFS_DINODE_MAX_LOG) 986 987 /* 988 * Inode size for given fs. 989 */ 990 #define XFS_DINODE_SIZE(mp) \ 991 (xfs_has_v3inodes(mp) ? \ 992 sizeof(struct xfs_dinode) : \ 993 offsetof(struct xfs_dinode, di_crc)) 994 #define XFS_LITINO(mp) \ 995 ((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(mp)) 996 997 /* 998 * Inode data & attribute fork sizes, per inode. 999 */ 1000 #define XFS_DFORK_BOFF(dip) ((int)((dip)->di_forkoff << 3)) 1001 1002 #define XFS_DFORK_DSIZE(dip,mp) \ 1003 ((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp)) 1004 #define XFS_DFORK_ASIZE(dip,mp) \ 1005 ((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0) 1006 #define XFS_DFORK_SIZE(dip,mp,w) \ 1007 ((w) == XFS_DATA_FORK ? \ 1008 XFS_DFORK_DSIZE(dip, mp) : \ 1009 XFS_DFORK_ASIZE(dip, mp)) 1010 1011 #define XFS_DFORK_MAXEXT(dip, mp, w) \ 1012 (XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec)) 1013 1014 /* 1015 * Return pointers to the data or attribute forks. 1016 */ 1017 #define XFS_DFORK_DPTR(dip) \ 1018 ((void *)dip + xfs_dinode_size(dip->di_version)) 1019 #define XFS_DFORK_APTR(dip) \ 1020 (XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip)) 1021 #define XFS_DFORK_PTR(dip,w) \ 1022 ((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip)) 1023 1024 #define XFS_DFORK_FORMAT(dip,w) \ 1025 ((w) == XFS_DATA_FORK ? \ 1026 (dip)->di_format : \ 1027 (dip)->di_aformat) 1028 1029 /* 1030 * For block and character special files the 32bit dev_t is stored at the 1031 * beginning of the data fork. 1032 */ 1033 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip) 1034 { 1035 return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip)); 1036 } 1037 1038 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev) 1039 { 1040 *(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev); 1041 } 1042 1043 /* 1044 * Values for di_flags 1045 */ 1046 #define XFS_DIFLAG_REALTIME_BIT 0 /* file's blocks come from rt area */ 1047 #define XFS_DIFLAG_PREALLOC_BIT 1 /* file space has been preallocated */ 1048 #define XFS_DIFLAG_NEWRTBM_BIT 2 /* for rtbitmap inode, new format */ 1049 #define XFS_DIFLAG_IMMUTABLE_BIT 3 /* inode is immutable */ 1050 #define XFS_DIFLAG_APPEND_BIT 4 /* inode is append-only */ 1051 #define XFS_DIFLAG_SYNC_BIT 5 /* inode is written synchronously */ 1052 #define XFS_DIFLAG_NOATIME_BIT 6 /* do not update atime */ 1053 #define XFS_DIFLAG_NODUMP_BIT 7 /* do not dump */ 1054 #define XFS_DIFLAG_RTINHERIT_BIT 8 /* create with realtime bit set */ 1055 #define XFS_DIFLAG_PROJINHERIT_BIT 9 /* create with parents projid */ 1056 #define XFS_DIFLAG_NOSYMLINKS_BIT 10 /* disallow symlink creation */ 1057 #define XFS_DIFLAG_EXTSIZE_BIT 11 /* inode extent size allocator hint */ 1058 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12 /* inherit inode extent size */ 1059 #define XFS_DIFLAG_NODEFRAG_BIT 13 /* do not reorganize/defragment */ 1060 #define XFS_DIFLAG_FILESTREAM_BIT 14 /* use filestream allocator */ 1061 /* Do not use bit 15, di_flags is legacy and unchanging now */ 1062 1063 #define XFS_DIFLAG_REALTIME (1 << XFS_DIFLAG_REALTIME_BIT) 1064 #define XFS_DIFLAG_PREALLOC (1 << XFS_DIFLAG_PREALLOC_BIT) 1065 #define XFS_DIFLAG_NEWRTBM (1 << XFS_DIFLAG_NEWRTBM_BIT) 1066 #define XFS_DIFLAG_IMMUTABLE (1 << XFS_DIFLAG_IMMUTABLE_BIT) 1067 #define XFS_DIFLAG_APPEND (1 << XFS_DIFLAG_APPEND_BIT) 1068 #define XFS_DIFLAG_SYNC (1 << XFS_DIFLAG_SYNC_BIT) 1069 #define XFS_DIFLAG_NOATIME (1 << XFS_DIFLAG_NOATIME_BIT) 1070 #define XFS_DIFLAG_NODUMP (1 << XFS_DIFLAG_NODUMP_BIT) 1071 #define XFS_DIFLAG_RTINHERIT (1 << XFS_DIFLAG_RTINHERIT_BIT) 1072 #define XFS_DIFLAG_PROJINHERIT (1 << XFS_DIFLAG_PROJINHERIT_BIT) 1073 #define XFS_DIFLAG_NOSYMLINKS (1 << XFS_DIFLAG_NOSYMLINKS_BIT) 1074 #define XFS_DIFLAG_EXTSIZE (1 << XFS_DIFLAG_EXTSIZE_BIT) 1075 #define XFS_DIFLAG_EXTSZINHERIT (1 << XFS_DIFLAG_EXTSZINHERIT_BIT) 1076 #define XFS_DIFLAG_NODEFRAG (1 << XFS_DIFLAG_NODEFRAG_BIT) 1077 #define XFS_DIFLAG_FILESTREAM (1 << XFS_DIFLAG_FILESTREAM_BIT) 1078 1079 #define XFS_DIFLAG_ANY \ 1080 (XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \ 1081 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \ 1082 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \ 1083 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \ 1084 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM) 1085 1086 /* 1087 * Values for di_flags2 These start by being exposed to userspace in the upper 1088 * 16 bits of the XFS_XFLAG_s range. 1089 */ 1090 #define XFS_DIFLAG2_DAX_BIT 0 /* use DAX for this inode */ 1091 #define XFS_DIFLAG2_REFLINK_BIT 1 /* file's blocks may be shared */ 1092 #define XFS_DIFLAG2_COWEXTSIZE_BIT 2 /* copy on write extent size hint */ 1093 #define XFS_DIFLAG2_BIGTIME_BIT 3 /* big timestamps */ 1094 #define XFS_DIFLAG2_NREXT64_BIT 4 /* large extent counters */ 1095 1096 #define XFS_DIFLAG2_DAX (1 << XFS_DIFLAG2_DAX_BIT) 1097 #define XFS_DIFLAG2_REFLINK (1 << XFS_DIFLAG2_REFLINK_BIT) 1098 #define XFS_DIFLAG2_COWEXTSIZE (1 << XFS_DIFLAG2_COWEXTSIZE_BIT) 1099 #define XFS_DIFLAG2_BIGTIME (1 << XFS_DIFLAG2_BIGTIME_BIT) 1100 #define XFS_DIFLAG2_NREXT64 (1 << XFS_DIFLAG2_NREXT64_BIT) 1101 1102 #define XFS_DIFLAG2_ANY \ 1103 (XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \ 1104 XFS_DIFLAG2_BIGTIME | XFS_DIFLAG2_NREXT64) 1105 1106 static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip) 1107 { 1108 return dip->di_version >= 3 && 1109 (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME)); 1110 } 1111 1112 static inline bool xfs_dinode_has_large_extent_counts( 1113 const struct xfs_dinode *dip) 1114 { 1115 return dip->di_version >= 3 && 1116 (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_NREXT64)); 1117 } 1118 1119 /* 1120 * Inode number format: 1121 * low inopblog bits - offset in block 1122 * next agblklog bits - block number in ag 1123 * next agno_log bits - ag number 1124 * high agno_log-agblklog-inopblog bits - 0 1125 */ 1126 #define XFS_INO_MASK(k) (uint32_t)((1ULL << (k)) - 1) 1127 #define XFS_INO_OFFSET_BITS(mp) (mp)->m_sb.sb_inopblog 1128 #define XFS_INO_AGBNO_BITS(mp) (mp)->m_sb.sb_agblklog 1129 #define XFS_INO_AGINO_BITS(mp) ((mp)->m_ino_geo.agino_log) 1130 #define XFS_INO_AGNO_BITS(mp) (mp)->m_agno_log 1131 #define XFS_INO_BITS(mp) \ 1132 XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp) 1133 #define XFS_INO_TO_AGNO(mp,i) \ 1134 ((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp))) 1135 #define XFS_INO_TO_AGINO(mp,i) \ 1136 ((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp))) 1137 #define XFS_INO_TO_AGBNO(mp,i) \ 1138 (((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \ 1139 XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp))) 1140 #define XFS_INO_TO_OFFSET(mp,i) \ 1141 ((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp))) 1142 #define XFS_INO_TO_FSB(mp,i) \ 1143 XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i)) 1144 #define XFS_AGINO_TO_INO(mp,a,i) \ 1145 (((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i)) 1146 #define XFS_AGINO_TO_AGBNO(mp,i) ((i) >> XFS_INO_OFFSET_BITS(mp)) 1147 #define XFS_AGINO_TO_OFFSET(mp,i) \ 1148 ((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp))) 1149 #define XFS_OFFBNO_TO_AGINO(mp,b,o) \ 1150 ((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o))) 1151 #define XFS_FSB_TO_INO(mp, b) ((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp))) 1152 #define XFS_AGB_TO_AGINO(mp, b) ((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp))) 1153 1154 #define XFS_MAXINUMBER ((xfs_ino_t)((1ULL << 56) - 1ULL)) 1155 #define XFS_MAXINUMBER_32 ((xfs_ino_t)((1ULL << 32) - 1ULL)) 1156 1157 /* 1158 * RealTime Device format definitions 1159 */ 1160 1161 /* Min and max rt extent sizes, specified in bytes */ 1162 #define XFS_MAX_RTEXTSIZE (1024 * 1024 * 1024) /* 1GB */ 1163 #define XFS_DFL_RTEXTSIZE (64 * 1024) /* 64kB */ 1164 #define XFS_MIN_RTEXTSIZE (4 * 1024) /* 4kB */ 1165 1166 /* 1167 * Dquot and dquot block format definitions 1168 */ 1169 #define XFS_DQUOT_MAGIC 0x4451 /* 'DQ' */ 1170 #define XFS_DQUOT_VERSION (uint8_t)0x01 /* latest version number */ 1171 1172 #define XFS_DQTYPE_USER (1u << 0) /* user dquot record */ 1173 #define XFS_DQTYPE_PROJ (1u << 1) /* project dquot record */ 1174 #define XFS_DQTYPE_GROUP (1u << 2) /* group dquot record */ 1175 #define XFS_DQTYPE_BIGTIME (1u << 7) /* large expiry timestamps */ 1176 1177 /* bitmask to determine if this is a user/group/project dquot */ 1178 #define XFS_DQTYPE_REC_MASK (XFS_DQTYPE_USER | \ 1179 XFS_DQTYPE_PROJ | \ 1180 XFS_DQTYPE_GROUP) 1181 1182 #define XFS_DQTYPE_ANY (XFS_DQTYPE_REC_MASK | \ 1183 XFS_DQTYPE_BIGTIME) 1184 1185 /* 1186 * XFS Quota Timers 1187 * ================ 1188 * 1189 * Traditional quota grace period expiration timers are an unsigned 32-bit 1190 * seconds counter; time zero is the Unix epoch, Jan 1 00:00:01 UTC 1970. 1191 * Note that an expiration value of zero means that the quota limit has not 1192 * been reached, and therefore no expiration has been set. Therefore, the 1193 * ondisk min and max defined here can be used directly to constrain the incore 1194 * quota expiration timestamps on a Unix system. 1195 * 1196 * When bigtime is enabled, we trade two bits of precision to expand the 1197 * expiration timeout range to match that of big inode timestamps. The min and 1198 * max recorded here are the on-disk limits, not a Unix timestamp. 1199 * 1200 * The grace period for each quota type is stored in the root dquot (id = 0) 1201 * and is applied to a non-root dquot when it exceeds the soft or hard limits. 1202 * The length of quota grace periods are unsigned 32-bit quantities measured in 1203 * units of seconds. A value of zero means to use the default period. 1204 */ 1205 1206 /* 1207 * Smallest possible ondisk quota expiration value with traditional timestamps. 1208 * This corresponds exactly with the incore expiration Jan 1 00:00:01 UTC 1970. 1209 */ 1210 #define XFS_DQ_LEGACY_EXPIRY_MIN ((int64_t)1) 1211 1212 /* 1213 * Largest possible ondisk quota expiration value with traditional timestamps. 1214 * This corresponds exactly with the incore expiration Feb 7 06:28:15 UTC 2106. 1215 */ 1216 #define XFS_DQ_LEGACY_EXPIRY_MAX ((int64_t)U32_MAX) 1217 1218 /* 1219 * Smallest possible ondisk quota expiration value with bigtime timestamps. 1220 * This corresponds (after conversion to a Unix timestamp) with the incore 1221 * expiration of Jan 1 00:00:04 UTC 1970. 1222 */ 1223 #define XFS_DQ_BIGTIME_EXPIRY_MIN (XFS_DQ_LEGACY_EXPIRY_MIN) 1224 1225 /* 1226 * Largest supported ondisk quota expiration value with bigtime timestamps. 1227 * This corresponds (after conversion to a Unix timestamp) with an incore 1228 * expiration of Jul 2 20:20:24 UTC 2486. 1229 * 1230 * The ondisk field supports values up to -1U, which corresponds to an incore 1231 * expiration in 2514. This is beyond the maximum the bigtime inode timestamp, 1232 * so we cap the maximum bigtime quota expiration to the max inode timestamp. 1233 */ 1234 #define XFS_DQ_BIGTIME_EXPIRY_MAX ((int64_t)4074815106U) 1235 1236 /* 1237 * The following conversion factors assist in converting a quota expiration 1238 * timestamp between the incore and ondisk formats. 1239 */ 1240 #define XFS_DQ_BIGTIME_SHIFT (2) 1241 #define XFS_DQ_BIGTIME_SLACK ((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1) 1242 1243 /* Convert an incore quota expiration timestamp to an ondisk bigtime value. */ 1244 static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds) 1245 { 1246 /* 1247 * Round the expiration timestamp up to the nearest bigtime timestamp 1248 * that we can store, to give users the most time to fix problems. 1249 */ 1250 return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >> 1251 XFS_DQ_BIGTIME_SHIFT; 1252 } 1253 1254 /* Convert an ondisk bigtime quota expiration value to an incore timestamp. */ 1255 static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds) 1256 { 1257 return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT; 1258 } 1259 1260 /* 1261 * Default quota grace periods, ranging from zero (use the compiled defaults) 1262 * to ~136 years. These are applied to a non-root dquot that has exceeded 1263 * either limit. 1264 */ 1265 #define XFS_DQ_GRACE_MIN ((int64_t)0) 1266 #define XFS_DQ_GRACE_MAX ((int64_t)U32_MAX) 1267 1268 /* Maximum id value for a quota record */ 1269 #define XFS_DQ_ID_MAX (U32_MAX) 1270 1271 /* 1272 * This is the main portion of the on-disk representation of quota information 1273 * for a user. We pad this with some more expansion room to construct the on 1274 * disk structure. 1275 */ 1276 struct xfs_disk_dquot { 1277 __be16 d_magic; /* dquot magic = XFS_DQUOT_MAGIC */ 1278 __u8 d_version; /* dquot version */ 1279 __u8 d_type; /* XFS_DQTYPE_USER/PROJ/GROUP */ 1280 __be32 d_id; /* user,project,group id */ 1281 __be64 d_blk_hardlimit;/* absolute limit on disk blks */ 1282 __be64 d_blk_softlimit;/* preferred limit on disk blks */ 1283 __be64 d_ino_hardlimit;/* maximum # allocated inodes */ 1284 __be64 d_ino_softlimit;/* preferred inode limit */ 1285 __be64 d_bcount; /* disk blocks owned by the user */ 1286 __be64 d_icount; /* inodes owned by the user */ 1287 __be32 d_itimer; /* zero if within inode limits if not, 1288 this is when we refuse service */ 1289 __be32 d_btimer; /* similar to above; for disk blocks */ 1290 __be16 d_iwarns; /* warnings issued wrt num inodes */ 1291 __be16 d_bwarns; /* warnings issued wrt disk blocks */ 1292 __be32 d_pad0; /* 64 bit align */ 1293 __be64 d_rtb_hardlimit;/* absolute limit on realtime blks */ 1294 __be64 d_rtb_softlimit;/* preferred limit on RT disk blks */ 1295 __be64 d_rtbcount; /* realtime blocks owned */ 1296 __be32 d_rtbtimer; /* similar to above; for RT disk blocks */ 1297 __be16 d_rtbwarns; /* warnings issued wrt RT disk blocks */ 1298 __be16 d_pad; 1299 }; 1300 1301 /* 1302 * This is what goes on disk. This is separated from the xfs_disk_dquot because 1303 * carrying the unnecessary padding would be a waste of memory. 1304 */ 1305 struct xfs_dqblk { 1306 struct xfs_disk_dquot dd_diskdq; /* portion living incore as well */ 1307 char dd_fill[4];/* filling for posterity */ 1308 1309 /* 1310 * These two are only present on filesystems with the CRC bits set. 1311 */ 1312 __be32 dd_crc; /* checksum */ 1313 __be64 dd_lsn; /* last modification in log */ 1314 uuid_t dd_uuid; /* location information */ 1315 }; 1316 1317 #define XFS_DQUOT_CRC_OFF offsetof(struct xfs_dqblk, dd_crc) 1318 1319 /* 1320 * This defines the unit of allocation of dquots. 1321 * 1322 * Currently, it is just one file system block, and a 4K blk contains 30 1323 * (136 * 30 = 4080) dquots. It's probably not worth trying to make 1324 * this more dynamic. 1325 * 1326 * However, if this number is changed, we have to make sure that we don't 1327 * implicitly assume that we do allocations in chunks of a single filesystem 1328 * block in the dquot/xqm code. 1329 * 1330 * This is part of the ondisk format because the structure size is not a power 1331 * of two, which leaves slack at the end of the disk block. 1332 */ 1333 #define XFS_DQUOT_CLUSTER_SIZE_FSB (xfs_filblks_t)1 1334 1335 /* 1336 * Remote symlink format and access functions. 1337 */ 1338 #define XFS_SYMLINK_MAGIC 0x58534c4d /* XSLM */ 1339 1340 struct xfs_dsymlink_hdr { 1341 __be32 sl_magic; 1342 __be32 sl_offset; 1343 __be32 sl_bytes; 1344 __be32 sl_crc; 1345 uuid_t sl_uuid; 1346 __be64 sl_owner; 1347 __be64 sl_blkno; 1348 __be64 sl_lsn; 1349 }; 1350 1351 #define XFS_SYMLINK_CRC_OFF offsetof(struct xfs_dsymlink_hdr, sl_crc) 1352 1353 #define XFS_SYMLINK_MAXLEN 1024 1354 /* 1355 * The maximum pathlen is 1024 bytes. Since the minimum file system 1356 * blocksize is 512 bytes, we can get a max of 3 extents back from 1357 * bmapi when crc headers are taken into account. 1358 */ 1359 #define XFS_SYMLINK_MAPS 3 1360 1361 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize) \ 1362 ((bufsize) - (xfs_has_crc((mp)) ? \ 1363 sizeof(struct xfs_dsymlink_hdr) : 0)) 1364 1365 1366 /* 1367 * Allocation Btree format definitions 1368 * 1369 * There are two on-disk btrees, one sorted by blockno and one sorted 1370 * by blockcount and blockno. All blocks look the same to make the code 1371 * simpler; if we have time later, we'll make the optimizations. 1372 */ 1373 #define XFS_ABTB_MAGIC 0x41425442 /* 'ABTB' for bno tree */ 1374 #define XFS_ABTB_CRC_MAGIC 0x41423342 /* 'AB3B' */ 1375 #define XFS_ABTC_MAGIC 0x41425443 /* 'ABTC' for cnt tree */ 1376 #define XFS_ABTC_CRC_MAGIC 0x41423343 /* 'AB3C' */ 1377 1378 /* 1379 * Data record/key structure 1380 */ 1381 typedef struct xfs_alloc_rec { 1382 __be32 ar_startblock; /* starting block number */ 1383 __be32 ar_blockcount; /* count of free blocks */ 1384 } xfs_alloc_rec_t, xfs_alloc_key_t; 1385 1386 typedef struct xfs_alloc_rec_incore { 1387 xfs_agblock_t ar_startblock; /* starting block number */ 1388 xfs_extlen_t ar_blockcount; /* count of free blocks */ 1389 } xfs_alloc_rec_incore_t; 1390 1391 /* btree pointer type */ 1392 typedef __be32 xfs_alloc_ptr_t; 1393 1394 /* 1395 * Block numbers in the AG: 1396 * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3. 1397 */ 1398 #define XFS_BNO_BLOCK(mp) ((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1)) 1399 #define XFS_CNT_BLOCK(mp) ((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1)) 1400 1401 1402 /* 1403 * Inode Allocation Btree format definitions 1404 * 1405 * There is a btree for the inode map per allocation group. 1406 */ 1407 #define XFS_IBT_MAGIC 0x49414254 /* 'IABT' */ 1408 #define XFS_IBT_CRC_MAGIC 0x49414233 /* 'IAB3' */ 1409 #define XFS_FIBT_MAGIC 0x46494254 /* 'FIBT' */ 1410 #define XFS_FIBT_CRC_MAGIC 0x46494233 /* 'FIB3' */ 1411 1412 typedef uint64_t xfs_inofree_t; 1413 #define XFS_INODES_PER_CHUNK (NBBY * sizeof(xfs_inofree_t)) 1414 #define XFS_INODES_PER_CHUNK_LOG (XFS_NBBYLOG + 3) 1415 #define XFS_INOBT_ALL_FREE ((xfs_inofree_t)-1) 1416 #define XFS_INOBT_MASK(i) ((xfs_inofree_t)1 << (i)) 1417 1418 #define XFS_INOBT_HOLEMASK_FULL 0 /* holemask for full chunk */ 1419 #define XFS_INOBT_HOLEMASK_BITS (NBBY * sizeof(uint16_t)) 1420 #define XFS_INODES_PER_HOLEMASK_BIT \ 1421 (XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t))) 1422 1423 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n) 1424 { 1425 return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i; 1426 } 1427 1428 /* 1429 * The on-disk inode record structure has two formats. The original "full" 1430 * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount 1431 * and replaces the 3 high-order freecount bytes wth the holemask and inode 1432 * count. 1433 * 1434 * The holemask of the sparse record format allows an inode chunk to have holes 1435 * that refer to blocks not owned by the inode record. This facilitates inode 1436 * allocation in the event of severe free space fragmentation. 1437 */ 1438 typedef struct xfs_inobt_rec { 1439 __be32 ir_startino; /* starting inode number */ 1440 union { 1441 struct { 1442 __be32 ir_freecount; /* count of free inodes */ 1443 } f; 1444 struct { 1445 __be16 ir_holemask;/* hole mask for sparse chunks */ 1446 __u8 ir_count; /* total inode count */ 1447 __u8 ir_freecount; /* count of free inodes */ 1448 } sp; 1449 } ir_u; 1450 __be64 ir_free; /* free inode mask */ 1451 } xfs_inobt_rec_t; 1452 1453 typedef struct xfs_inobt_rec_incore { 1454 xfs_agino_t ir_startino; /* starting inode number */ 1455 uint16_t ir_holemask; /* hole mask for sparse chunks */ 1456 uint8_t ir_count; /* total inode count */ 1457 uint8_t ir_freecount; /* count of free inodes (set bits) */ 1458 xfs_inofree_t ir_free; /* free inode mask */ 1459 } xfs_inobt_rec_incore_t; 1460 1461 static inline bool xfs_inobt_issparse(uint16_t holemask) 1462 { 1463 /* non-zero holemask represents a sparse rec. */ 1464 return holemask; 1465 } 1466 1467 /* 1468 * Key structure 1469 */ 1470 typedef struct xfs_inobt_key { 1471 __be32 ir_startino; /* starting inode number */ 1472 } xfs_inobt_key_t; 1473 1474 /* btree pointer type */ 1475 typedef __be32 xfs_inobt_ptr_t; 1476 1477 /* 1478 * block numbers in the AG. 1479 */ 1480 #define XFS_IBT_BLOCK(mp) ((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1)) 1481 #define XFS_FIBT_BLOCK(mp) ((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1)) 1482 1483 /* 1484 * Reverse mapping btree format definitions 1485 * 1486 * There is a btree for the reverse map per allocation group 1487 */ 1488 #define XFS_RMAP_CRC_MAGIC 0x524d4233 /* 'RMB3' */ 1489 1490 /* 1491 * Ownership info for an extent. This is used to create reverse-mapping 1492 * entries. 1493 */ 1494 #define XFS_OWNER_INFO_ATTR_FORK (1 << 0) 1495 #define XFS_OWNER_INFO_BMBT_BLOCK (1 << 1) 1496 struct xfs_owner_info { 1497 uint64_t oi_owner; 1498 xfs_fileoff_t oi_offset; 1499 unsigned int oi_flags; 1500 }; 1501 1502 /* 1503 * Special owner types. 1504 * 1505 * Seeing as we only support up to 8EB, we have the upper bit of the owner field 1506 * to tell us we have a special owner value. We use these for static metadata 1507 * allocated at mkfs/growfs time, as well as for freespace management metadata. 1508 */ 1509 #define XFS_RMAP_OWN_NULL (-1ULL) /* No owner, for growfs */ 1510 #define XFS_RMAP_OWN_UNKNOWN (-2ULL) /* Unknown owner, for EFI recovery */ 1511 #define XFS_RMAP_OWN_FS (-3ULL) /* static fs metadata */ 1512 #define XFS_RMAP_OWN_LOG (-4ULL) /* static fs metadata */ 1513 #define XFS_RMAP_OWN_AG (-5ULL) /* AG freespace btree blocks */ 1514 #define XFS_RMAP_OWN_INOBT (-6ULL) /* Inode btree blocks */ 1515 #define XFS_RMAP_OWN_INODES (-7ULL) /* Inode chunk */ 1516 #define XFS_RMAP_OWN_REFC (-8ULL) /* refcount tree */ 1517 #define XFS_RMAP_OWN_COW (-9ULL) /* cow allocations */ 1518 #define XFS_RMAP_OWN_MIN (-10ULL) /* guard */ 1519 1520 #define XFS_RMAP_NON_INODE_OWNER(owner) (!!((owner) & (1ULL << 63))) 1521 1522 /* 1523 * Data record structure 1524 */ 1525 struct xfs_rmap_rec { 1526 __be32 rm_startblock; /* extent start block */ 1527 __be32 rm_blockcount; /* extent length */ 1528 __be64 rm_owner; /* extent owner */ 1529 __be64 rm_offset; /* offset within the owner */ 1530 }; 1531 1532 /* 1533 * rmap btree record 1534 * rm_offset:63 is the attribute fork flag 1535 * rm_offset:62 is the bmbt block flag 1536 * rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt) 1537 * rm_offset:54-60 aren't used and should be zero 1538 * rm_offset:0-53 is the block offset within the inode 1539 */ 1540 #define XFS_RMAP_OFF_ATTR_FORK ((uint64_t)1ULL << 63) 1541 #define XFS_RMAP_OFF_BMBT_BLOCK ((uint64_t)1ULL << 62) 1542 #define XFS_RMAP_OFF_UNWRITTEN ((uint64_t)1ULL << 61) 1543 1544 #define XFS_RMAP_LEN_MAX ((uint32_t)~0U) 1545 #define XFS_RMAP_OFF_FLAGS (XFS_RMAP_OFF_ATTR_FORK | \ 1546 XFS_RMAP_OFF_BMBT_BLOCK | \ 1547 XFS_RMAP_OFF_UNWRITTEN) 1548 #define XFS_RMAP_OFF_MASK ((uint64_t)0x3FFFFFFFFFFFFFULL) 1549 1550 #define XFS_RMAP_OFF(off) ((off) & XFS_RMAP_OFF_MASK) 1551 1552 #define XFS_RMAP_IS_BMBT_BLOCK(off) (!!((off) & XFS_RMAP_OFF_BMBT_BLOCK)) 1553 #define XFS_RMAP_IS_ATTR_FORK(off) (!!((off) & XFS_RMAP_OFF_ATTR_FORK)) 1554 #define XFS_RMAP_IS_UNWRITTEN(len) (!!((off) & XFS_RMAP_OFF_UNWRITTEN)) 1555 1556 #define RMAPBT_STARTBLOCK_BITLEN 32 1557 #define RMAPBT_BLOCKCOUNT_BITLEN 32 1558 #define RMAPBT_OWNER_BITLEN 64 1559 #define RMAPBT_ATTRFLAG_BITLEN 1 1560 #define RMAPBT_BMBTFLAG_BITLEN 1 1561 #define RMAPBT_EXNTFLAG_BITLEN 1 1562 #define RMAPBT_UNUSED_OFFSET_BITLEN 7 1563 #define RMAPBT_OFFSET_BITLEN 54 1564 1565 /* 1566 * Key structure 1567 * 1568 * We don't use the length for lookups 1569 */ 1570 struct xfs_rmap_key { 1571 __be32 rm_startblock; /* extent start block */ 1572 __be64 rm_owner; /* extent owner */ 1573 __be64 rm_offset; /* offset within the owner */ 1574 } __attribute__((packed)); 1575 1576 /* btree pointer type */ 1577 typedef __be32 xfs_rmap_ptr_t; 1578 1579 #define XFS_RMAP_BLOCK(mp) \ 1580 (xfs_has_finobt(((mp))) ? \ 1581 XFS_FIBT_BLOCK(mp) + 1 : \ 1582 XFS_IBT_BLOCK(mp) + 1) 1583 1584 /* 1585 * Reference Count Btree format definitions 1586 * 1587 */ 1588 #define XFS_REFC_CRC_MAGIC 0x52334643 /* 'R3FC' */ 1589 1590 unsigned int xfs_refc_block(struct xfs_mount *mp); 1591 1592 /* 1593 * Data record/key structure 1594 * 1595 * Each record associates a range of physical blocks (starting at 1596 * rc_startblock and ending rc_blockcount blocks later) with a reference 1597 * count (rc_refcount). Extents that are being used to stage a copy on 1598 * write (CoW) operation are recorded in the refcount btree with a 1599 * refcount of 1. All other records must have a refcount > 1 and must 1600 * track an extent mapped only by file data forks. 1601 * 1602 * Extents with a single owner (attributes, metadata, non-shared file 1603 * data) are not tracked here. Free space is also not tracked here. 1604 * This is consistent with pre-reflink XFS. 1605 */ 1606 1607 /* 1608 * Extents that are being used to stage a copy on write are stored 1609 * in the refcount btree with a refcount of 1 and the upper bit set 1610 * on the startblock. This speeds up mount time deletion of stale 1611 * staging extents because they're all at the right side of the tree. 1612 */ 1613 #define XFS_REFC_COWFLAG (1U << 31) 1614 #define REFCNTBT_COWFLAG_BITLEN 1 1615 #define REFCNTBT_AGBLOCK_BITLEN 31 1616 1617 struct xfs_refcount_rec { 1618 __be32 rc_startblock; /* starting block number */ 1619 __be32 rc_blockcount; /* count of blocks */ 1620 __be32 rc_refcount; /* number of inodes linked here */ 1621 }; 1622 1623 struct xfs_refcount_key { 1624 __be32 rc_startblock; /* starting block number */ 1625 }; 1626 1627 #define MAXREFCOUNT ((xfs_nlink_t)~0U) 1628 #define MAXREFCEXTLEN ((xfs_extlen_t)~0U) 1629 1630 /* btree pointer type */ 1631 typedef __be32 xfs_refcount_ptr_t; 1632 1633 1634 /* 1635 * BMAP Btree format definitions 1636 * 1637 * This includes both the root block definition that sits inside an inode fork 1638 * and the record/pointer formats for the leaf/node in the blocks. 1639 */ 1640 #define XFS_BMAP_MAGIC 0x424d4150 /* 'BMAP' */ 1641 #define XFS_BMAP_CRC_MAGIC 0x424d4133 /* 'BMA3' */ 1642 1643 /* 1644 * Bmap root header, on-disk form only. 1645 */ 1646 typedef struct xfs_bmdr_block { 1647 __be16 bb_level; /* 0 is a leaf */ 1648 __be16 bb_numrecs; /* current # of data records */ 1649 } xfs_bmdr_block_t; 1650 1651 /* 1652 * Bmap btree record and extent descriptor. 1653 * l0:63 is an extent flag (value 1 indicates non-normal). 1654 * l0:9-62 are startoff. 1655 * l0:0-8 and l1:21-63 are startblock. 1656 * l1:0-20 are blockcount. 1657 */ 1658 #define BMBT_EXNTFLAG_BITLEN 1 1659 #define BMBT_STARTOFF_BITLEN 54 1660 #define BMBT_STARTBLOCK_BITLEN 52 1661 #define BMBT_BLOCKCOUNT_BITLEN 21 1662 1663 #define BMBT_STARTOFF_MASK ((1ULL << BMBT_STARTOFF_BITLEN) - 1) 1664 #define BMBT_BLOCKCOUNT_MASK ((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1) 1665 1666 #define XFS_MAX_BMBT_EXTLEN ((xfs_extlen_t)(BMBT_BLOCKCOUNT_MASK)) 1667 1668 /* 1669 * bmbt records have a file offset (block) field that is 54 bits wide, so this 1670 * is the largest xfs_fileoff_t that we ever expect to see. 1671 */ 1672 #define XFS_MAX_FILEOFF (BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK) 1673 1674 typedef struct xfs_bmbt_rec { 1675 __be64 l0, l1; 1676 } xfs_bmbt_rec_t; 1677 1678 typedef uint64_t xfs_bmbt_rec_base_t; /* use this for casts */ 1679 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t; 1680 1681 /* 1682 * Values and macros for delayed-allocation startblock fields. 1683 */ 1684 #define STARTBLOCKVALBITS 17 1685 #define STARTBLOCKMASKBITS (15 + 20) 1686 #define STARTBLOCKMASK \ 1687 (((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS) 1688 1689 static inline int isnullstartblock(xfs_fsblock_t x) 1690 { 1691 return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK; 1692 } 1693 1694 static inline xfs_fsblock_t nullstartblock(int k) 1695 { 1696 ASSERT(k < (1 << STARTBLOCKVALBITS)); 1697 return STARTBLOCKMASK | (k); 1698 } 1699 1700 static inline xfs_filblks_t startblockval(xfs_fsblock_t x) 1701 { 1702 return (xfs_filblks_t)((x) & ~STARTBLOCKMASK); 1703 } 1704 1705 /* 1706 * Key structure for non-leaf levels of the tree. 1707 */ 1708 typedef struct xfs_bmbt_key { 1709 __be64 br_startoff; /* starting file offset */ 1710 } xfs_bmbt_key_t, xfs_bmdr_key_t; 1711 1712 /* btree pointer type */ 1713 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t; 1714 1715 1716 /* 1717 * Generic Btree block format definitions 1718 * 1719 * This is a combination of the actual format used on disk for short and long 1720 * format btrees. The first three fields are shared by both format, but the 1721 * pointers are different and should be used with care. 1722 * 1723 * To get the size of the actual short or long form headers please use the size 1724 * macros below. Never use sizeof(xfs_btree_block). 1725 * 1726 * The blkno, crc, lsn, owner and uuid fields are only available in filesystems 1727 * with the crc feature bit, and all accesses to them must be conditional on 1728 * that flag. 1729 */ 1730 /* short form block header */ 1731 struct xfs_btree_block_shdr { 1732 __be32 bb_leftsib; 1733 __be32 bb_rightsib; 1734 1735 __be64 bb_blkno; 1736 __be64 bb_lsn; 1737 uuid_t bb_uuid; 1738 __be32 bb_owner; 1739 __le32 bb_crc; 1740 }; 1741 1742 /* long form block header */ 1743 struct xfs_btree_block_lhdr { 1744 __be64 bb_leftsib; 1745 __be64 bb_rightsib; 1746 1747 __be64 bb_blkno; 1748 __be64 bb_lsn; 1749 uuid_t bb_uuid; 1750 __be64 bb_owner; 1751 __le32 bb_crc; 1752 __be32 bb_pad; /* padding for alignment */ 1753 }; 1754 1755 struct xfs_btree_block { 1756 __be32 bb_magic; /* magic number for block type */ 1757 __be16 bb_level; /* 0 is a leaf */ 1758 __be16 bb_numrecs; /* current # of data records */ 1759 union { 1760 struct xfs_btree_block_shdr s; 1761 struct xfs_btree_block_lhdr l; 1762 } bb_u; /* rest */ 1763 }; 1764 1765 /* size of a short form block */ 1766 #define XFS_BTREE_SBLOCK_LEN \ 1767 (offsetof(struct xfs_btree_block, bb_u) + \ 1768 offsetof(struct xfs_btree_block_shdr, bb_blkno)) 1769 /* size of a long form block */ 1770 #define XFS_BTREE_LBLOCK_LEN \ 1771 (offsetof(struct xfs_btree_block, bb_u) + \ 1772 offsetof(struct xfs_btree_block_lhdr, bb_blkno)) 1773 1774 /* sizes of CRC enabled btree blocks */ 1775 #define XFS_BTREE_SBLOCK_CRC_LEN \ 1776 (offsetof(struct xfs_btree_block, bb_u) + \ 1777 sizeof(struct xfs_btree_block_shdr)) 1778 #define XFS_BTREE_LBLOCK_CRC_LEN \ 1779 (offsetof(struct xfs_btree_block, bb_u) + \ 1780 sizeof(struct xfs_btree_block_lhdr)) 1781 1782 #define XFS_BTREE_SBLOCK_CRC_OFF \ 1783 offsetof(struct xfs_btree_block, bb_u.s.bb_crc) 1784 #define XFS_BTREE_LBLOCK_CRC_OFF \ 1785 offsetof(struct xfs_btree_block, bb_u.l.bb_crc) 1786 1787 /* 1788 * On-disk XFS access control list structure. 1789 */ 1790 struct xfs_acl_entry { 1791 __be32 ae_tag; 1792 __be32 ae_id; 1793 __be16 ae_perm; 1794 __be16 ae_pad; /* fill the implicit hole in the structure */ 1795 }; 1796 1797 struct xfs_acl { 1798 __be32 acl_cnt; 1799 struct xfs_acl_entry acl_entry[]; 1800 }; 1801 1802 /* 1803 * The number of ACL entries allowed is defined by the on-disk format. 1804 * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is 1805 * limited only by the maximum size of the xattr that stores the information. 1806 */ 1807 #define XFS_ACL_MAX_ENTRIES(mp) \ 1808 (xfs_has_crc(mp) \ 1809 ? (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \ 1810 sizeof(struct xfs_acl_entry) \ 1811 : 25) 1812 1813 #define XFS_ACL_SIZE(cnt) \ 1814 (sizeof(struct xfs_acl) + \ 1815 sizeof(struct xfs_acl_entry) * cnt) 1816 1817 #define XFS_ACL_MAX_SIZE(mp) \ 1818 XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp))) 1819 1820 1821 /* On-disk XFS extended attribute names */ 1822 #define SGI_ACL_FILE "SGI_ACL_FILE" 1823 #define SGI_ACL_DEFAULT "SGI_ACL_DEFAULT" 1824 #define SGI_ACL_FILE_SIZE (sizeof(SGI_ACL_FILE)-1) 1825 #define SGI_ACL_DEFAULT_SIZE (sizeof(SGI_ACL_DEFAULT)-1) 1826 1827 #endif /* __XFS_FORMAT_H__ */ 1828