1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
6 #ifndef __XFS_FORMAT_H__
7 #define __XFS_FORMAT_H__
10 * XFS On Disk Format Definitions
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.
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.
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
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.
56 #define XFS_XATTR_SIZE_MAX (1 << 16)
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.
63 #define XFS_SB_VERSION_OKBITS \
64 ((XFS_SB_VERSION_NUMBITS | XFS_SB_VERSION_ALLFBITS) & \
65 ~XFS_SB_VERSION_SHAREDBIT)
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.
72 * These defines represent bits in sb_features2.
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 */
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)
89 /* Maximum size of the xfs filesystem label, no terminating NULL */
90 #define XFSLABEL_MAX 12
93 * Superblock - in core version. Must match the ondisk version below.
94 * Must be padded to 64 bit alignment.
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 */
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().
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 */
136 * End contiguous fields.
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 */
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.
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
161 uint32_t sb_bad_features2;
163 /* version 5 superblock fields start here */
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;
171 uint32_t sb_crc; /* superblock crc */
172 xfs_extlen_t sb_spino_align; /* sparse inode chunk alignment */
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 */
178 /* must be padded to 64 bit alignment */
181 #define XFS_SB_CRC_OFF offsetof(struct xfs_sb, sb_crc)
184 * Superblock - on disk version. Must match the in core version above.
185 * Must be padded to 64 bit alignment.
187 typedef 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 */
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().
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 */
227 * End contiguous fields.
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 */
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.
248 __be32 sb_bad_features2;
250 /* version 5 superblock fields start here */
253 __be32 sb_features_compat;
254 __be32 sb_features_ro_compat;
255 __be32 sb_features_incompat;
256 __be32 sb_features_log_incompat;
258 __le32 sb_crc; /* superblock crc */
259 __be32 sb_spino_align; /* sparse inode chunk alignment */
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 */
265 /* must be padded to 64 bit alignment */
270 * Misc. Flags - warning - these will be cleared by xfs_repair unless
271 * a feature bit is set when the flag is used.
273 #define XFS_SBF_NOFLAGS 0x00 /* no flags set */
274 #define XFS_SBF_READONLY 0x01 /* only read-only mounts allowed */
277 * define max. shared version we can interoperate with
279 #define XFS_SB_MAX_SHARED_VN 0
281 #define XFS_SB_VERSION_NUM(sbp) ((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
284 * The first XFS version we support is a v4 superblock with V2 directories.
286 static inline bool xfs_sb_good_v4_features(struct xfs_sb *sbp)
288 if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
290 if (!(sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT))
293 /* check for unknown features in the fs */
294 if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
295 ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
296 (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
302 static inline bool xfs_sb_good_version(struct xfs_sb *sbp)
304 if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5)
306 if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_4)
307 return xfs_sb_good_v4_features(sbp);
311 static inline bool xfs_sb_version_hasrealtime(struct xfs_sb *sbp)
313 return sbp->sb_rblocks > 0;
317 * Detect a mismatched features2 field. Older kernels read/wrote
318 * this into the wrong slot, so to be safe we keep them in sync.
320 static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
322 return sbp->sb_bad_features2 != sbp->sb_features2;
325 static inline bool xfs_sb_version_hasattr(struct xfs_sb *sbp)
327 return (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT);
330 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
332 sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
335 static inline bool xfs_sb_version_hasquota(struct xfs_sb *sbp)
337 return (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT);
340 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
342 sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
345 static inline bool xfs_sb_version_hasalign(struct xfs_sb *sbp)
347 return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
348 (sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT));
351 static inline bool xfs_sb_version_hasdalign(struct xfs_sb *sbp)
353 return (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT);
356 static inline bool xfs_sb_version_haslogv2(struct xfs_sb *sbp)
358 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
359 (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT);
362 static inline bool xfs_sb_version_hassector(struct xfs_sb *sbp)
364 return (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT);
367 static inline bool xfs_sb_version_hasasciici(struct xfs_sb *sbp)
369 return (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT);
372 static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp)
374 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
375 (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
379 * sb_features2 bit version macros.
381 static inline bool xfs_sb_version_haslazysbcount(struct xfs_sb *sbp)
383 return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
384 (xfs_sb_version_hasmorebits(sbp) &&
385 (sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT));
388 static inline bool xfs_sb_version_hasattr2(struct xfs_sb *sbp)
390 return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
391 (xfs_sb_version_hasmorebits(sbp) &&
392 (sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT));
395 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
397 sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
398 sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
401 static inline void xfs_sb_version_removeattr2(struct xfs_sb *sbp)
403 sbp->sb_features2 &= ~XFS_SB_VERSION2_ATTR2BIT;
404 if (!sbp->sb_features2)
405 sbp->sb_versionnum &= ~XFS_SB_VERSION_MOREBITSBIT;
408 static inline bool xfs_sb_version_hasprojid32bit(struct xfs_sb *sbp)
410 return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
411 (xfs_sb_version_hasmorebits(sbp) &&
412 (sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT));
415 static inline void xfs_sb_version_addprojid32bit(struct xfs_sb *sbp)
417 sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
418 sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
422 * Extended v5 superblock feature masks. These are to be used for new v5
423 * superblock features only.
425 * Compat features are new features that old kernels will not notice or affect
426 * and so can mount read-write without issues.
428 * RO-Compat (read only) are features that old kernels can read but will break
429 * if they write. Hence only read-only mounts of such filesystems are allowed on
430 * kernels that don't support the feature bit.
432 * InCompat features are features which old kernels will not understand and so
435 * Log-InCompat features are for changes to log formats or new transactions that
436 * can't be replayed on older kernels. The fields are set when the filesystem is
437 * mounted, and a clean unmount clears the fields.
439 #define XFS_SB_FEAT_COMPAT_ALL 0
440 #define XFS_SB_FEAT_COMPAT_UNKNOWN ~XFS_SB_FEAT_COMPAT_ALL
442 xfs_sb_has_compat_feature(
446 return (sbp->sb_features_compat & feature) != 0;
449 #define XFS_SB_FEAT_RO_COMPAT_FINOBT (1 << 0) /* free inode btree */
450 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT (1 << 1) /* reverse map btree */
451 #define XFS_SB_FEAT_RO_COMPAT_REFLINK (1 << 2) /* reflinked files */
452 #define XFS_SB_FEAT_RO_COMPAT_INOBTCNT (1 << 3) /* inobt block counts */
453 #define XFS_SB_FEAT_RO_COMPAT_ALL \
454 (XFS_SB_FEAT_RO_COMPAT_FINOBT | \
455 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
456 XFS_SB_FEAT_RO_COMPAT_REFLINK| \
457 XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
458 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN ~XFS_SB_FEAT_RO_COMPAT_ALL
460 xfs_sb_has_ro_compat_feature(
464 return (sbp->sb_features_ro_compat & feature) != 0;
467 #define XFS_SB_FEAT_INCOMPAT_FTYPE (1 << 0) /* filetype in dirent */
468 #define XFS_SB_FEAT_INCOMPAT_SPINODES (1 << 1) /* sparse inode chunks */
469 #define XFS_SB_FEAT_INCOMPAT_META_UUID (1 << 2) /* metadata UUID */
470 #define XFS_SB_FEAT_INCOMPAT_BIGTIME (1 << 3) /* large timestamps */
471 #define XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR (1 << 4) /* needs xfs_repair */
472 #define XFS_SB_FEAT_INCOMPAT_ALL \
473 (XFS_SB_FEAT_INCOMPAT_FTYPE| \
474 XFS_SB_FEAT_INCOMPAT_SPINODES| \
475 XFS_SB_FEAT_INCOMPAT_META_UUID| \
476 XFS_SB_FEAT_INCOMPAT_BIGTIME| \
477 XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR)
479 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN ~XFS_SB_FEAT_INCOMPAT_ALL
481 xfs_sb_has_incompat_feature(
485 return (sbp->sb_features_incompat & feature) != 0;
488 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL 0
489 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN ~XFS_SB_FEAT_INCOMPAT_LOG_ALL
491 xfs_sb_has_incompat_log_feature(
495 return (sbp->sb_features_log_incompat & feature) != 0;
499 * V5 superblock specific feature checks
501 static inline bool xfs_sb_version_hascrc(struct xfs_sb *sbp)
503 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
507 * v5 file systems support V3 inodes only, earlier file systems support
510 static inline bool xfs_sb_version_has_v3inode(struct xfs_sb *sbp)
512 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
515 static inline bool xfs_dinode_good_version(struct xfs_sb *sbp,
518 if (xfs_sb_version_has_v3inode(sbp))
520 return version == 1 || version == 2;
523 static inline bool xfs_sb_version_has_pquotino(struct xfs_sb *sbp)
525 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
528 static inline int xfs_sb_version_hasftype(struct xfs_sb *sbp)
530 return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
531 xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_FTYPE)) ||
532 (xfs_sb_version_hasmorebits(sbp) &&
533 (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE));
536 static inline bool xfs_sb_version_hasfinobt(xfs_sb_t *sbp)
538 return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
539 (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT);
542 static inline bool xfs_sb_version_hassparseinodes(struct xfs_sb *sbp)
544 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
545 xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_SPINODES);
549 * XFS_SB_FEAT_INCOMPAT_META_UUID indicates that the metadata UUID
550 * is stored separately from the user-visible UUID; this allows the
551 * user-visible UUID to be changed on V5 filesystems which have a
552 * filesystem UUID stamped into every piece of metadata.
554 static inline bool xfs_sb_version_hasmetauuid(struct xfs_sb *sbp)
556 return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
557 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID);
560 static inline bool xfs_sb_version_hasrmapbt(struct xfs_sb *sbp)
562 return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
563 (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT);
566 static inline bool xfs_sb_version_hasreflink(struct xfs_sb *sbp)
568 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
569 (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK);
572 static inline bool xfs_sb_version_hasbigtime(struct xfs_sb *sbp)
574 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
575 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_BIGTIME);
579 * Inode btree block counter. We record the number of inobt and finobt blocks
580 * in the AGI header so that we can skip the finobt walk at mount time when
581 * setting up per-AG reservations.
583 static inline bool xfs_sb_version_hasinobtcounts(struct xfs_sb *sbp)
585 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
586 (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_INOBTCNT);
589 static inline bool xfs_sb_version_needsrepair(struct xfs_sb *sbp)
591 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
592 (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR);
596 * end of superblock version macros
600 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
602 return (ino == sbp->sb_uquotino ||
603 ino == sbp->sb_gquotino ||
604 ino == sbp->sb_pquotino);
607 #define XFS_SB_DADDR ((xfs_daddr_t)0) /* daddr in filesystem/ag */
608 #define XFS_SB_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
610 #define XFS_HDR_BLOCK(mp,d) ((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
611 #define XFS_DADDR_TO_FSB(mp,d) XFS_AGB_TO_FSB(mp, \
612 xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
613 #define XFS_FSB_TO_DADDR(mp,fsbno) XFS_AGB_TO_DADDR(mp, \
614 XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
617 * File system sector to basic block conversions.
619 #define XFS_FSS_TO_BB(mp,sec) ((sec) << (mp)->m_sectbb_log)
622 * File system block to basic block conversions.
624 #define XFS_FSB_TO_BB(mp,fsbno) ((fsbno) << (mp)->m_blkbb_log)
625 #define XFS_BB_TO_FSB(mp,bb) \
626 (((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
627 #define XFS_BB_TO_FSBT(mp,bb) ((bb) >> (mp)->m_blkbb_log)
630 * File system block to byte conversions.
632 #define XFS_FSB_TO_B(mp,fsbno) ((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
633 #define XFS_B_TO_FSB(mp,b) \
634 ((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
635 #define XFS_B_TO_FSBT(mp,b) (((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
638 * Allocation group header
640 * This is divided into three structures, placed in sequential 512-byte
641 * buffers after a copy of the superblock (also in a 512-byte buffer).
643 #define XFS_AGF_MAGIC 0x58414746 /* 'XAGF' */
644 #define XFS_AGI_MAGIC 0x58414749 /* 'XAGI' */
645 #define XFS_AGFL_MAGIC 0x5841464c /* 'XAFL' */
646 #define XFS_AGF_VERSION 1
647 #define XFS_AGI_VERSION 1
649 #define XFS_AGF_GOOD_VERSION(v) ((v) == XFS_AGF_VERSION)
650 #define XFS_AGI_GOOD_VERSION(v) ((v) == XFS_AGI_VERSION)
653 * Btree number 0 is bno, 1 is cnt, 2 is rmap. This value gives the size of the
656 #define XFS_BTNUM_AGF ((int)XFS_BTNUM_RMAPi + 1)
659 * The second word of agf_levels in the first a.g. overlaps the EFS
660 * superblock's magic number. Since the magic numbers valid for EFS
661 * are > 64k, our value cannot be confused for an EFS superblock's.
664 typedef struct xfs_agf {
666 * Common allocation group header information
668 __be32 agf_magicnum; /* magic number == XFS_AGF_MAGIC */
669 __be32 agf_versionnum; /* header version == XFS_AGF_VERSION */
670 __be32 agf_seqno; /* sequence # starting from 0 */
671 __be32 agf_length; /* size in blocks of a.g. */
673 * Freespace and rmap information
675 __be32 agf_roots[XFS_BTNUM_AGF]; /* root blocks */
676 __be32 agf_levels[XFS_BTNUM_AGF]; /* btree levels */
678 __be32 agf_flfirst; /* first freelist block's index */
679 __be32 agf_fllast; /* last freelist block's index */
680 __be32 agf_flcount; /* count of blocks in freelist */
681 __be32 agf_freeblks; /* total free blocks */
683 __be32 agf_longest; /* longest free space */
684 __be32 agf_btreeblks; /* # of blocks held in AGF btrees */
685 uuid_t agf_uuid; /* uuid of filesystem */
687 __be32 agf_rmap_blocks; /* rmapbt blocks used */
688 __be32 agf_refcount_blocks; /* refcountbt blocks used */
690 __be32 agf_refcount_root; /* refcount tree root block */
691 __be32 agf_refcount_level; /* refcount btree levels */
694 * reserve some contiguous space for future logged fields before we add
695 * the unlogged fields. This makes the range logging via flags and
696 * structure offsets much simpler.
698 __be64 agf_spare64[14];
700 /* unlogged fields, written during buffer writeback. */
701 __be64 agf_lsn; /* last write sequence */
702 __be32 agf_crc; /* crc of agf sector */
705 /* structure must be padded to 64 bit alignment */
708 #define XFS_AGF_CRC_OFF offsetof(struct xfs_agf, agf_crc)
710 #define XFS_AGF_MAGICNUM 0x00000001
711 #define XFS_AGF_VERSIONNUM 0x00000002
712 #define XFS_AGF_SEQNO 0x00000004
713 #define XFS_AGF_LENGTH 0x00000008
714 #define XFS_AGF_ROOTS 0x00000010
715 #define XFS_AGF_LEVELS 0x00000020
716 #define XFS_AGF_FLFIRST 0x00000040
717 #define XFS_AGF_FLLAST 0x00000080
718 #define XFS_AGF_FLCOUNT 0x00000100
719 #define XFS_AGF_FREEBLKS 0x00000200
720 #define XFS_AGF_LONGEST 0x00000400
721 #define XFS_AGF_BTREEBLKS 0x00000800
722 #define XFS_AGF_UUID 0x00001000
723 #define XFS_AGF_RMAP_BLOCKS 0x00002000
724 #define XFS_AGF_REFCOUNT_BLOCKS 0x00004000
725 #define XFS_AGF_REFCOUNT_ROOT 0x00008000
726 #define XFS_AGF_REFCOUNT_LEVEL 0x00010000
727 #define XFS_AGF_SPARE64 0x00020000
728 #define XFS_AGF_NUM_BITS 18
729 #define XFS_AGF_ALL_BITS ((1 << XFS_AGF_NUM_BITS) - 1)
731 #define XFS_AGF_FLAGS \
732 { XFS_AGF_MAGICNUM, "MAGICNUM" }, \
733 { XFS_AGF_VERSIONNUM, "VERSIONNUM" }, \
734 { XFS_AGF_SEQNO, "SEQNO" }, \
735 { XFS_AGF_LENGTH, "LENGTH" }, \
736 { XFS_AGF_ROOTS, "ROOTS" }, \
737 { XFS_AGF_LEVELS, "LEVELS" }, \
738 { XFS_AGF_FLFIRST, "FLFIRST" }, \
739 { XFS_AGF_FLLAST, "FLLAST" }, \
740 { XFS_AGF_FLCOUNT, "FLCOUNT" }, \
741 { XFS_AGF_FREEBLKS, "FREEBLKS" }, \
742 { XFS_AGF_LONGEST, "LONGEST" }, \
743 { XFS_AGF_BTREEBLKS, "BTREEBLKS" }, \
744 { XFS_AGF_UUID, "UUID" }, \
745 { XFS_AGF_RMAP_BLOCKS, "RMAP_BLOCKS" }, \
746 { XFS_AGF_REFCOUNT_BLOCKS, "REFCOUNT_BLOCKS" }, \
747 { XFS_AGF_REFCOUNT_ROOT, "REFCOUNT_ROOT" }, \
748 { XFS_AGF_REFCOUNT_LEVEL, "REFCOUNT_LEVEL" }, \
749 { XFS_AGF_SPARE64, "SPARE64" }
751 /* disk block (xfs_daddr_t) in the AG */
752 #define XFS_AGF_DADDR(mp) ((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
753 #define XFS_AGF_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
756 * Size of the unlinked inode hash table in the agi.
758 #define XFS_AGI_UNLINKED_BUCKETS 64
760 typedef struct xfs_agi {
762 * Common allocation group header information
764 __be32 agi_magicnum; /* magic number == XFS_AGI_MAGIC */
765 __be32 agi_versionnum; /* header version == XFS_AGI_VERSION */
766 __be32 agi_seqno; /* sequence # starting from 0 */
767 __be32 agi_length; /* size in blocks of a.g. */
770 * Inodes are mapped by interpreting the inode number, so no
771 * mapping data is needed here.
773 __be32 agi_count; /* count of allocated inodes */
774 __be32 agi_root; /* root of inode btree */
775 __be32 agi_level; /* levels in inode btree */
776 __be32 agi_freecount; /* number of free inodes */
778 __be32 agi_newino; /* new inode just allocated */
779 __be32 agi_dirino; /* last directory inode chunk */
781 * Hash table of inodes which have been unlinked but are
782 * still being referenced.
784 __be32 agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
786 * This marks the end of logging region 1 and start of logging region 2.
788 uuid_t agi_uuid; /* uuid of filesystem */
789 __be32 agi_crc; /* crc of agi sector */
791 __be64 agi_lsn; /* last write sequence */
793 __be32 agi_free_root; /* root of the free inode btree */
794 __be32 agi_free_level;/* levels in free inode btree */
796 __be32 agi_iblocks; /* inobt blocks used */
797 __be32 agi_fblocks; /* finobt blocks used */
799 /* structure must be padded to 64 bit alignment */
802 #define XFS_AGI_CRC_OFF offsetof(struct xfs_agi, agi_crc)
804 #define XFS_AGI_MAGICNUM (1 << 0)
805 #define XFS_AGI_VERSIONNUM (1 << 1)
806 #define XFS_AGI_SEQNO (1 << 2)
807 #define XFS_AGI_LENGTH (1 << 3)
808 #define XFS_AGI_COUNT (1 << 4)
809 #define XFS_AGI_ROOT (1 << 5)
810 #define XFS_AGI_LEVEL (1 << 6)
811 #define XFS_AGI_FREECOUNT (1 << 7)
812 #define XFS_AGI_NEWINO (1 << 8)
813 #define XFS_AGI_DIRINO (1 << 9)
814 #define XFS_AGI_UNLINKED (1 << 10)
815 #define XFS_AGI_NUM_BITS_R1 11 /* end of the 1st agi logging region */
816 #define XFS_AGI_ALL_BITS_R1 ((1 << XFS_AGI_NUM_BITS_R1) - 1)
817 #define XFS_AGI_FREE_ROOT (1 << 11)
818 #define XFS_AGI_FREE_LEVEL (1 << 12)
819 #define XFS_AGI_IBLOCKS (1 << 13) /* both inobt/finobt block counters */
820 #define XFS_AGI_NUM_BITS_R2 14
822 /* disk block (xfs_daddr_t) in the AG */
823 #define XFS_AGI_DADDR(mp) ((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
824 #define XFS_AGI_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
827 * The third a.g. block contains the a.g. freelist, an array
828 * of block pointers to blocks owned by the allocation btree code.
830 #define XFS_AGFL_DADDR(mp) ((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
831 #define XFS_AGFL_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
832 #define XFS_BUF_TO_AGFL(bp) ((struct xfs_agfl *)((bp)->b_addr))
835 __be32 agfl_magicnum;
840 } __attribute__((packed));
842 #define XFS_AGFL_CRC_OFF offsetof(struct xfs_agfl, agfl_crc)
844 #define XFS_AGB_TO_FSB(mp,agno,agbno) \
845 (((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
846 #define XFS_FSB_TO_AGNO(mp,fsbno) \
847 ((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
848 #define XFS_FSB_TO_AGBNO(mp,fsbno) \
849 ((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
850 #define XFS_AGB_TO_DADDR(mp,agno,agbno) \
851 ((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
852 (xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
853 #define XFS_AG_DADDR(mp,agno,d) (XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
856 * For checking for bad ranges of xfs_daddr_t's, covering multiple
857 * allocation groups or a single xfs_daddr_t that's a superblock copy.
859 #define XFS_AG_CHECK_DADDR(mp,d,len) \
861 ASSERT((d) == XFS_SB_DADDR || \
862 xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
863 ASSERT(xfs_daddr_to_agno(mp, d) == \
864 xfs_daddr_to_agno(mp, (d) + (len) - 1)))
870 * Traditional ondisk inode timestamps consist of signed 32-bit counters for
871 * seconds and nanoseconds; time zero is the Unix epoch, Jan 1 00:00:00 UTC
872 * 1970, which means that the timestamp epoch is the same as the Unix epoch.
873 * Therefore, the ondisk min and max defined here can be used directly to
874 * constrain the incore timestamps on a Unix system. Note that we actually
875 * encode a __be64 value on disk.
877 * When the bigtime feature is enabled, ondisk inode timestamps become an
878 * unsigned 64-bit nanoseconds counter. This means that the bigtime inode
879 * timestamp epoch is the start of the classic timestamp range, which is
880 * Dec 31 20:45:52 UTC 1901. Because the epochs are not the same, callers
881 * /must/ use the bigtime conversion functions when encoding and decoding raw
884 typedef __be64 xfs_timestamp_t;
886 /* Legacy timestamp encoding format. */
887 struct xfs_legacy_timestamp {
888 __be32 t_sec; /* timestamp seconds */
889 __be32 t_nsec; /* timestamp nanoseconds */
893 * Smallest possible ondisk seconds value with traditional timestamps. This
894 * corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901.
896 #define XFS_LEGACY_TIME_MIN ((int64_t)S32_MIN)
899 * Largest possible ondisk seconds value with traditional timestamps. This
900 * corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038.
902 #define XFS_LEGACY_TIME_MAX ((int64_t)S32_MAX)
905 * Smallest possible ondisk seconds value with bigtime timestamps. This
906 * corresponds (after conversion to a Unix timestamp) with the traditional
907 * minimum timestamp of Dec 13 20:45:52 UTC 1901.
909 #define XFS_BIGTIME_TIME_MIN ((int64_t)0)
912 * Largest supported ondisk seconds value with bigtime timestamps. This
913 * corresponds (after conversion to a Unix timestamp) with an incore timestamp
914 * of Jul 2 20:20:24 UTC 2486.
916 * We round down the ondisk limit so that the bigtime quota and inode max
917 * timestamps will be the same.
919 #define XFS_BIGTIME_TIME_MAX ((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL))
922 * Bigtime epoch is set exactly to the minimum time value that a traditional
923 * 32-bit timestamp can represent when using the Unix epoch as a reference.
924 * Hence the Unix epoch is at a fixed offset into the supported bigtime
927 * The bigtime epoch also matches the minimum value an on-disk 32-bit XFS
928 * timestamp can represent so we will not lose any fidelity in converting
929 * to/from unix and bigtime timestamps.
931 * The following conversion factor converts a seconds counter from the Unix
932 * epoch to the bigtime epoch.
934 #define XFS_BIGTIME_EPOCH_OFFSET (-(int64_t)S32_MIN)
936 /* Convert a timestamp from the Unix epoch to the bigtime epoch. */
937 static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds)
939 return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET;
942 /* Convert a timestamp from the bigtime epoch to the Unix epoch. */
943 static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds)
945 return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET;
949 * On-disk inode structure.
951 * This is just the header or "dinode core", the inode is expanded to fill a
952 * variable size the leftover area split into a data and an attribute fork.
953 * The format of the data and attribute fork depends on the format of the
954 * inode as indicated by di_format and di_aformat. To access the data and
955 * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
958 * There is a very similar struct xfs_log_dinode which matches the layout of
959 * this structure, but is kept in native format instead of big endian.
961 * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
962 * padding field for v3 inodes.
964 #define XFS_DINODE_MAGIC 0x494e /* 'IN' */
965 typedef struct xfs_dinode {
966 __be16 di_magic; /* inode magic # = XFS_DINODE_MAGIC */
967 __be16 di_mode; /* mode and type of file */
968 __u8 di_version; /* inode version */
969 __u8 di_format; /* format of di_c data */
970 __be16 di_onlink; /* old number of links to file */
971 __be32 di_uid; /* owner's user id */
972 __be32 di_gid; /* owner's group id */
973 __be32 di_nlink; /* number of links to file */
974 __be16 di_projid_lo; /* lower part of owner's project id */
975 __be16 di_projid_hi; /* higher part owner's project id */
976 __u8 di_pad[6]; /* unused, zeroed space */
977 __be16 di_flushiter; /* incremented on flush */
978 xfs_timestamp_t di_atime; /* time last accessed */
979 xfs_timestamp_t di_mtime; /* time last modified */
980 xfs_timestamp_t di_ctime; /* time created/inode modified */
981 __be64 di_size; /* number of bytes in file */
982 __be64 di_nblocks; /* # of direct & btree blocks used */
983 __be32 di_extsize; /* basic/minimum extent size for file */
984 __be32 di_nextents; /* number of extents in data fork */
985 __be16 di_anextents; /* number of extents in attribute fork*/
986 __u8 di_forkoff; /* attr fork offs, <<3 for 64b align */
987 __s8 di_aformat; /* format of attr fork's data */
988 __be32 di_dmevmask; /* DMIG event mask */
989 __be16 di_dmstate; /* DMIG state info */
990 __be16 di_flags; /* random flags, XFS_DIFLAG_... */
991 __be32 di_gen; /* generation number */
993 /* di_next_unlinked is the only non-core field in the old dinode */
994 __be32 di_next_unlinked;/* agi unlinked list ptr */
996 /* start of the extended dinode, writable fields */
997 __le32 di_crc; /* CRC of the inode */
998 __be64 di_changecount; /* number of attribute changes */
999 __be64 di_lsn; /* flush sequence */
1000 __be64 di_flags2; /* more random flags */
1001 __be32 di_cowextsize; /* basic cow extent size for file */
1002 __u8 di_pad2[12]; /* more padding for future expansion */
1004 /* fields only written to during inode creation */
1005 xfs_timestamp_t di_crtime; /* time created */
1006 __be64 di_ino; /* inode number */
1007 uuid_t di_uuid; /* UUID of the filesystem */
1009 /* structure must be padded to 64 bit alignment */
1012 #define XFS_DINODE_CRC_OFF offsetof(struct xfs_dinode, di_crc)
1014 #define DI_MAX_FLUSH 0xffff
1017 * Size of the core inode on disk. Version 1 and 2 inodes have
1018 * the same size, but version 3 has grown a few additional fields.
1020 static inline uint xfs_dinode_size(int version)
1023 return sizeof(struct xfs_dinode);
1024 return offsetof(struct xfs_dinode, di_crc);
1028 * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
1029 * Since the pathconf interface is signed, we use 2^31 - 1 instead.
1031 #define XFS_MAXLINK ((1U << 31) - 1U)
1034 * Values for di_format
1036 * This enum is used in string mapping in xfs_trace.h; please keep the
1037 * TRACE_DEFINE_ENUMs for it up to date.
1039 enum xfs_dinode_fmt {
1040 XFS_DINODE_FMT_DEV, /* xfs_dev_t */
1041 XFS_DINODE_FMT_LOCAL, /* bulk data */
1042 XFS_DINODE_FMT_EXTENTS, /* struct xfs_bmbt_rec */
1043 XFS_DINODE_FMT_BTREE, /* struct xfs_bmdr_block */
1044 XFS_DINODE_FMT_UUID /* added long ago, but never used */
1047 #define XFS_INODE_FORMAT_STR \
1048 { XFS_DINODE_FMT_DEV, "dev" }, \
1049 { XFS_DINODE_FMT_LOCAL, "local" }, \
1050 { XFS_DINODE_FMT_EXTENTS, "extent" }, \
1051 { XFS_DINODE_FMT_BTREE, "btree" }, \
1052 { XFS_DINODE_FMT_UUID, "uuid" }
1055 * Inode minimum and maximum sizes.
1057 #define XFS_DINODE_MIN_LOG 8
1058 #define XFS_DINODE_MAX_LOG 11
1059 #define XFS_DINODE_MIN_SIZE (1 << XFS_DINODE_MIN_LOG)
1060 #define XFS_DINODE_MAX_SIZE (1 << XFS_DINODE_MAX_LOG)
1063 * Inode size for given fs.
1065 #define XFS_DINODE_SIZE(sbp) \
1066 (xfs_sb_version_has_v3inode(sbp) ? \
1067 sizeof(struct xfs_dinode) : \
1068 offsetof(struct xfs_dinode, di_crc))
1069 #define XFS_LITINO(mp) \
1070 ((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(&(mp)->m_sb))
1073 * Inode data & attribute fork sizes, per inode.
1075 #define XFS_DFORK_BOFF(dip) ((int)((dip)->di_forkoff << 3))
1077 #define XFS_DFORK_DSIZE(dip,mp) \
1078 ((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp))
1079 #define XFS_DFORK_ASIZE(dip,mp) \
1080 ((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0)
1081 #define XFS_DFORK_SIZE(dip,mp,w) \
1082 ((w) == XFS_DATA_FORK ? \
1083 XFS_DFORK_DSIZE(dip, mp) : \
1084 XFS_DFORK_ASIZE(dip, mp))
1086 #define XFS_DFORK_MAXEXT(dip, mp, w) \
1087 (XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
1090 * Return pointers to the data or attribute forks.
1092 #define XFS_DFORK_DPTR(dip) \
1093 ((char *)dip + xfs_dinode_size(dip->di_version))
1094 #define XFS_DFORK_APTR(dip) \
1095 (XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
1096 #define XFS_DFORK_PTR(dip,w) \
1097 ((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
1099 #define XFS_DFORK_FORMAT(dip,w) \
1100 ((w) == XFS_DATA_FORK ? \
1101 (dip)->di_format : \
1103 #define XFS_DFORK_NEXTENTS(dip,w) \
1104 ((w) == XFS_DATA_FORK ? \
1105 be32_to_cpu((dip)->di_nextents) : \
1106 be16_to_cpu((dip)->di_anextents))
1109 * For block and character special files the 32bit dev_t is stored at the
1110 * beginning of the data fork.
1112 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1114 return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1117 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1119 *(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1123 * Values for di_flags
1125 #define XFS_DIFLAG_REALTIME_BIT 0 /* file's blocks come from rt area */
1126 #define XFS_DIFLAG_PREALLOC_BIT 1 /* file space has been preallocated */
1127 #define XFS_DIFLAG_NEWRTBM_BIT 2 /* for rtbitmap inode, new format */
1128 #define XFS_DIFLAG_IMMUTABLE_BIT 3 /* inode is immutable */
1129 #define XFS_DIFLAG_APPEND_BIT 4 /* inode is append-only */
1130 #define XFS_DIFLAG_SYNC_BIT 5 /* inode is written synchronously */
1131 #define XFS_DIFLAG_NOATIME_BIT 6 /* do not update atime */
1132 #define XFS_DIFLAG_NODUMP_BIT 7 /* do not dump */
1133 #define XFS_DIFLAG_RTINHERIT_BIT 8 /* create with realtime bit set */
1134 #define XFS_DIFLAG_PROJINHERIT_BIT 9 /* create with parents projid */
1135 #define XFS_DIFLAG_NOSYMLINKS_BIT 10 /* disallow symlink creation */
1136 #define XFS_DIFLAG_EXTSIZE_BIT 11 /* inode extent size allocator hint */
1137 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12 /* inherit inode extent size */
1138 #define XFS_DIFLAG_NODEFRAG_BIT 13 /* do not reorganize/defragment */
1139 #define XFS_DIFLAG_FILESTREAM_BIT 14 /* use filestream allocator */
1140 /* Do not use bit 15, di_flags is legacy and unchanging now */
1142 #define XFS_DIFLAG_REALTIME (1 << XFS_DIFLAG_REALTIME_BIT)
1143 #define XFS_DIFLAG_PREALLOC (1 << XFS_DIFLAG_PREALLOC_BIT)
1144 #define XFS_DIFLAG_NEWRTBM (1 << XFS_DIFLAG_NEWRTBM_BIT)
1145 #define XFS_DIFLAG_IMMUTABLE (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1146 #define XFS_DIFLAG_APPEND (1 << XFS_DIFLAG_APPEND_BIT)
1147 #define XFS_DIFLAG_SYNC (1 << XFS_DIFLAG_SYNC_BIT)
1148 #define XFS_DIFLAG_NOATIME (1 << XFS_DIFLAG_NOATIME_BIT)
1149 #define XFS_DIFLAG_NODUMP (1 << XFS_DIFLAG_NODUMP_BIT)
1150 #define XFS_DIFLAG_RTINHERIT (1 << XFS_DIFLAG_RTINHERIT_BIT)
1151 #define XFS_DIFLAG_PROJINHERIT (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1152 #define XFS_DIFLAG_NOSYMLINKS (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1153 #define XFS_DIFLAG_EXTSIZE (1 << XFS_DIFLAG_EXTSIZE_BIT)
1154 #define XFS_DIFLAG_EXTSZINHERIT (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1155 #define XFS_DIFLAG_NODEFRAG (1 << XFS_DIFLAG_NODEFRAG_BIT)
1156 #define XFS_DIFLAG_FILESTREAM (1 << XFS_DIFLAG_FILESTREAM_BIT)
1158 #define XFS_DIFLAG_ANY \
1159 (XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1160 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1161 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1162 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1163 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1166 * Values for di_flags2 These start by being exposed to userspace in the upper
1167 * 16 bits of the XFS_XFLAG_s range.
1169 #define XFS_DIFLAG2_DAX_BIT 0 /* use DAX for this inode */
1170 #define XFS_DIFLAG2_REFLINK_BIT 1 /* file's blocks may be shared */
1171 #define XFS_DIFLAG2_COWEXTSIZE_BIT 2 /* copy on write extent size hint */
1172 #define XFS_DIFLAG2_BIGTIME_BIT 3 /* big timestamps */
1174 #define XFS_DIFLAG2_DAX (1 << XFS_DIFLAG2_DAX_BIT)
1175 #define XFS_DIFLAG2_REFLINK (1 << XFS_DIFLAG2_REFLINK_BIT)
1176 #define XFS_DIFLAG2_COWEXTSIZE (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1177 #define XFS_DIFLAG2_BIGTIME (1 << XFS_DIFLAG2_BIGTIME_BIT)
1179 #define XFS_DIFLAG2_ANY \
1180 (XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \
1181 XFS_DIFLAG2_BIGTIME)
1183 static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip)
1185 return dip->di_version >= 3 &&
1186 (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME));
1190 * Inode number format:
1191 * low inopblog bits - offset in block
1192 * next agblklog bits - block number in ag
1193 * next agno_log bits - ag number
1194 * high agno_log-agblklog-inopblog bits - 0
1196 #define XFS_INO_MASK(k) (uint32_t)((1ULL << (k)) - 1)
1197 #define XFS_INO_OFFSET_BITS(mp) (mp)->m_sb.sb_inopblog
1198 #define XFS_INO_AGBNO_BITS(mp) (mp)->m_sb.sb_agblklog
1199 #define XFS_INO_AGINO_BITS(mp) ((mp)->m_ino_geo.agino_log)
1200 #define XFS_INO_AGNO_BITS(mp) (mp)->m_agno_log
1201 #define XFS_INO_BITS(mp) \
1202 XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1203 #define XFS_INO_TO_AGNO(mp,i) \
1204 ((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1205 #define XFS_INO_TO_AGINO(mp,i) \
1206 ((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1207 #define XFS_INO_TO_AGBNO(mp,i) \
1208 (((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1209 XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1210 #define XFS_INO_TO_OFFSET(mp,i) \
1211 ((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1212 #define XFS_INO_TO_FSB(mp,i) \
1213 XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1214 #define XFS_AGINO_TO_INO(mp,a,i) \
1215 (((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1216 #define XFS_AGINO_TO_AGBNO(mp,i) ((i) >> XFS_INO_OFFSET_BITS(mp))
1217 #define XFS_AGINO_TO_OFFSET(mp,i) \
1218 ((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1219 #define XFS_OFFBNO_TO_AGINO(mp,b,o) \
1220 ((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1221 #define XFS_FSB_TO_INO(mp, b) ((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1222 #define XFS_AGB_TO_AGINO(mp, b) ((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1224 #define XFS_MAXINUMBER ((xfs_ino_t)((1ULL << 56) - 1ULL))
1225 #define XFS_MAXINUMBER_32 ((xfs_ino_t)((1ULL << 32) - 1ULL))
1228 * RealTime Device format definitions
1231 /* Min and max rt extent sizes, specified in bytes */
1232 #define XFS_MAX_RTEXTSIZE (1024 * 1024 * 1024) /* 1GB */
1233 #define XFS_DFL_RTEXTSIZE (64 * 1024) /* 64kB */
1234 #define XFS_MIN_RTEXTSIZE (4 * 1024) /* 4kB */
1236 #define XFS_BLOCKSIZE(mp) ((mp)->m_sb.sb_blocksize)
1237 #define XFS_BLOCKMASK(mp) ((mp)->m_blockmask)
1238 #define XFS_BLOCKWSIZE(mp) ((mp)->m_blockwsize)
1239 #define XFS_BLOCKWMASK(mp) ((mp)->m_blockwmask)
1242 * RT Summary and bit manipulation macros.
1244 #define XFS_SUMOFFS(mp,ls,bb) ((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb)))
1245 #define XFS_SUMOFFSTOBLOCK(mp,s) \
1246 (((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog)
1247 #define XFS_SUMPTR(mp,bp,so) \
1248 ((xfs_suminfo_t *)((bp)->b_addr + \
1249 (((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp))))
1251 #define XFS_BITTOBLOCK(mp,bi) ((bi) >> (mp)->m_blkbit_log)
1252 #define XFS_BLOCKTOBIT(mp,bb) ((bb) << (mp)->m_blkbit_log)
1253 #define XFS_BITTOWORD(mp,bi) \
1254 ((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp)))
1256 #define XFS_RTMIN(a,b) ((a) < (b) ? (a) : (b))
1257 #define XFS_RTMAX(a,b) ((a) > (b) ? (a) : (b))
1259 #define XFS_RTLOBIT(w) xfs_lowbit32(w)
1260 #define XFS_RTHIBIT(w) xfs_highbit32(w)
1262 #define XFS_RTBLOCKLOG(b) xfs_highbit64(b)
1265 * Dquot and dquot block format definitions
1267 #define XFS_DQUOT_MAGIC 0x4451 /* 'DQ' */
1268 #define XFS_DQUOT_VERSION (uint8_t)0x01 /* latest version number */
1270 #define XFS_DQTYPE_USER 0x01 /* user dquot record */
1271 #define XFS_DQTYPE_PROJ 0x02 /* project dquot record */
1272 #define XFS_DQTYPE_GROUP 0x04 /* group dquot record */
1273 #define XFS_DQTYPE_BIGTIME 0x80 /* large expiry timestamps */
1275 /* bitmask to determine if this is a user/group/project dquot */
1276 #define XFS_DQTYPE_REC_MASK (XFS_DQTYPE_USER | \
1280 #define XFS_DQTYPE_ANY (XFS_DQTYPE_REC_MASK | \
1287 * Traditional quota grace period expiration timers are an unsigned 32-bit
1288 * seconds counter; time zero is the Unix epoch, Jan 1 00:00:01 UTC 1970.
1289 * Note that an expiration value of zero means that the quota limit has not
1290 * been reached, and therefore no expiration has been set. Therefore, the
1291 * ondisk min and max defined here can be used directly to constrain the incore
1292 * quota expiration timestamps on a Unix system.
1294 * When bigtime is enabled, we trade two bits of precision to expand the
1295 * expiration timeout range to match that of big inode timestamps. The min and
1296 * max recorded here are the on-disk limits, not a Unix timestamp.
1298 * The grace period for each quota type is stored in the root dquot (id = 0)
1299 * and is applied to a non-root dquot when it exceeds the soft or hard limits.
1300 * The length of quota grace periods are unsigned 32-bit quantities measured in
1301 * units of seconds. A value of zero means to use the default period.
1305 * Smallest possible ondisk quota expiration value with traditional timestamps.
1306 * This corresponds exactly with the incore expiration Jan 1 00:00:01 UTC 1970.
1308 #define XFS_DQ_LEGACY_EXPIRY_MIN ((int64_t)1)
1311 * Largest possible ondisk quota expiration value with traditional timestamps.
1312 * This corresponds exactly with the incore expiration Feb 7 06:28:15 UTC 2106.
1314 #define XFS_DQ_LEGACY_EXPIRY_MAX ((int64_t)U32_MAX)
1317 * Smallest possible ondisk quota expiration value with bigtime timestamps.
1318 * This corresponds (after conversion to a Unix timestamp) with the incore
1319 * expiration of Jan 1 00:00:04 UTC 1970.
1321 #define XFS_DQ_BIGTIME_EXPIRY_MIN (XFS_DQ_LEGACY_EXPIRY_MIN)
1324 * Largest supported ondisk quota expiration value with bigtime timestamps.
1325 * This corresponds (after conversion to a Unix timestamp) with an incore
1326 * expiration of Jul 2 20:20:24 UTC 2486.
1328 * The ondisk field supports values up to -1U, which corresponds to an incore
1329 * expiration in 2514. This is beyond the maximum the bigtime inode timestamp,
1330 * so we cap the maximum bigtime quota expiration to the max inode timestamp.
1332 #define XFS_DQ_BIGTIME_EXPIRY_MAX ((int64_t)4074815106U)
1335 * The following conversion factors assist in converting a quota expiration
1336 * timestamp between the incore and ondisk formats.
1338 #define XFS_DQ_BIGTIME_SHIFT (2)
1339 #define XFS_DQ_BIGTIME_SLACK ((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1)
1341 /* Convert an incore quota expiration timestamp to an ondisk bigtime value. */
1342 static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds)
1345 * Round the expiration timestamp up to the nearest bigtime timestamp
1346 * that we can store, to give users the most time to fix problems.
1348 return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >>
1349 XFS_DQ_BIGTIME_SHIFT;
1352 /* Convert an ondisk bigtime quota expiration value to an incore timestamp. */
1353 static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)
1355 return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT;
1359 * Default quota grace periods, ranging from zero (use the compiled defaults)
1360 * to ~136 years. These are applied to a non-root dquot that has exceeded
1363 #define XFS_DQ_GRACE_MIN ((int64_t)0)
1364 #define XFS_DQ_GRACE_MAX ((int64_t)U32_MAX)
1367 * This is the main portion of the on-disk representation of quota information
1368 * for a user. We pad this with some more expansion room to construct the on
1371 struct xfs_disk_dquot {
1372 __be16 d_magic; /* dquot magic = XFS_DQUOT_MAGIC */
1373 __u8 d_version; /* dquot version */
1374 __u8 d_type; /* XFS_DQTYPE_USER/PROJ/GROUP */
1375 __be32 d_id; /* user,project,group id */
1376 __be64 d_blk_hardlimit;/* absolute limit on disk blks */
1377 __be64 d_blk_softlimit;/* preferred limit on disk blks */
1378 __be64 d_ino_hardlimit;/* maximum # allocated inodes */
1379 __be64 d_ino_softlimit;/* preferred inode limit */
1380 __be64 d_bcount; /* disk blocks owned by the user */
1381 __be64 d_icount; /* inodes owned by the user */
1382 __be32 d_itimer; /* zero if within inode limits if not,
1383 this is when we refuse service */
1384 __be32 d_btimer; /* similar to above; for disk blocks */
1385 __be16 d_iwarns; /* warnings issued wrt num inodes */
1386 __be16 d_bwarns; /* warnings issued wrt disk blocks */
1387 __be32 d_pad0; /* 64 bit align */
1388 __be64 d_rtb_hardlimit;/* absolute limit on realtime blks */
1389 __be64 d_rtb_softlimit;/* preferred limit on RT disk blks */
1390 __be64 d_rtbcount; /* realtime blocks owned */
1391 __be32 d_rtbtimer; /* similar to above; for RT disk blocks */
1392 __be16 d_rtbwarns; /* warnings issued wrt RT disk blocks */
1397 * This is what goes on disk. This is separated from the xfs_disk_dquot because
1398 * carrying the unnecessary padding would be a waste of memory.
1400 typedef struct xfs_dqblk {
1401 struct xfs_disk_dquot dd_diskdq; /* portion living incore as well */
1402 char dd_fill[4];/* filling for posterity */
1405 * These two are only present on filesystems with the CRC bits set.
1407 __be32 dd_crc; /* checksum */
1408 __be64 dd_lsn; /* last modification in log */
1409 uuid_t dd_uuid; /* location information */
1412 #define XFS_DQUOT_CRC_OFF offsetof(struct xfs_dqblk, dd_crc)
1415 * This defines the unit of allocation of dquots.
1417 * Currently, it is just one file system block, and a 4K blk contains 30
1418 * (136 * 30 = 4080) dquots. It's probably not worth trying to make
1419 * this more dynamic.
1421 * However, if this number is changed, we have to make sure that we don't
1422 * implicitly assume that we do allocations in chunks of a single filesystem
1423 * block in the dquot/xqm code.
1425 * This is part of the ondisk format because the structure size is not a power
1426 * of two, which leaves slack at the end of the disk block.
1428 #define XFS_DQUOT_CLUSTER_SIZE_FSB (xfs_filblks_t)1
1431 * Remote symlink format and access functions.
1433 #define XFS_SYMLINK_MAGIC 0x58534c4d /* XSLM */
1435 struct xfs_dsymlink_hdr {
1446 #define XFS_SYMLINK_CRC_OFF offsetof(struct xfs_dsymlink_hdr, sl_crc)
1448 #define XFS_SYMLINK_MAXLEN 1024
1450 * The maximum pathlen is 1024 bytes. Since the minimum file system
1451 * blocksize is 512 bytes, we can get a max of 3 extents back from
1452 * bmapi when crc headers are taken into account.
1454 #define XFS_SYMLINK_MAPS 3
1456 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize) \
1457 ((bufsize) - (xfs_sb_version_hascrc(&(mp)->m_sb) ? \
1458 sizeof(struct xfs_dsymlink_hdr) : 0))
1462 * Allocation Btree format definitions
1464 * There are two on-disk btrees, one sorted by blockno and one sorted
1465 * by blockcount and blockno. All blocks look the same to make the code
1466 * simpler; if we have time later, we'll make the optimizations.
1468 #define XFS_ABTB_MAGIC 0x41425442 /* 'ABTB' for bno tree */
1469 #define XFS_ABTB_CRC_MAGIC 0x41423342 /* 'AB3B' */
1470 #define XFS_ABTC_MAGIC 0x41425443 /* 'ABTC' for cnt tree */
1471 #define XFS_ABTC_CRC_MAGIC 0x41423343 /* 'AB3C' */
1474 * Data record/key structure
1476 typedef struct xfs_alloc_rec {
1477 __be32 ar_startblock; /* starting block number */
1478 __be32 ar_blockcount; /* count of free blocks */
1479 } xfs_alloc_rec_t, xfs_alloc_key_t;
1481 typedef struct xfs_alloc_rec_incore {
1482 xfs_agblock_t ar_startblock; /* starting block number */
1483 xfs_extlen_t ar_blockcount; /* count of free blocks */
1484 } xfs_alloc_rec_incore_t;
1486 /* btree pointer type */
1487 typedef __be32 xfs_alloc_ptr_t;
1490 * Block numbers in the AG:
1491 * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1493 #define XFS_BNO_BLOCK(mp) ((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1494 #define XFS_CNT_BLOCK(mp) ((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1498 * Inode Allocation Btree format definitions
1500 * There is a btree for the inode map per allocation group.
1502 #define XFS_IBT_MAGIC 0x49414254 /* 'IABT' */
1503 #define XFS_IBT_CRC_MAGIC 0x49414233 /* 'IAB3' */
1504 #define XFS_FIBT_MAGIC 0x46494254 /* 'FIBT' */
1505 #define XFS_FIBT_CRC_MAGIC 0x46494233 /* 'FIB3' */
1507 typedef uint64_t xfs_inofree_t;
1508 #define XFS_INODES_PER_CHUNK (NBBY * sizeof(xfs_inofree_t))
1509 #define XFS_INODES_PER_CHUNK_LOG (XFS_NBBYLOG + 3)
1510 #define XFS_INOBT_ALL_FREE ((xfs_inofree_t)-1)
1511 #define XFS_INOBT_MASK(i) ((xfs_inofree_t)1 << (i))
1513 #define XFS_INOBT_HOLEMASK_FULL 0 /* holemask for full chunk */
1514 #define XFS_INOBT_HOLEMASK_BITS (NBBY * sizeof(uint16_t))
1515 #define XFS_INODES_PER_HOLEMASK_BIT \
1516 (XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1518 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1520 return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1524 * The on-disk inode record structure has two formats. The original "full"
1525 * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1526 * and replaces the 3 high-order freecount bytes wth the holemask and inode
1529 * The holemask of the sparse record format allows an inode chunk to have holes
1530 * that refer to blocks not owned by the inode record. This facilitates inode
1531 * allocation in the event of severe free space fragmentation.
1533 typedef struct xfs_inobt_rec {
1534 __be32 ir_startino; /* starting inode number */
1537 __be32 ir_freecount; /* count of free inodes */
1540 __be16 ir_holemask;/* hole mask for sparse chunks */
1541 __u8 ir_count; /* total inode count */
1542 __u8 ir_freecount; /* count of free inodes */
1545 __be64 ir_free; /* free inode mask */
1548 typedef struct xfs_inobt_rec_incore {
1549 xfs_agino_t ir_startino; /* starting inode number */
1550 uint16_t ir_holemask; /* hole mask for sparse chunks */
1551 uint8_t ir_count; /* total inode count */
1552 uint8_t ir_freecount; /* count of free inodes (set bits) */
1553 xfs_inofree_t ir_free; /* free inode mask */
1554 } xfs_inobt_rec_incore_t;
1556 static inline bool xfs_inobt_issparse(uint16_t holemask)
1558 /* non-zero holemask represents a sparse rec. */
1565 typedef struct xfs_inobt_key {
1566 __be32 ir_startino; /* starting inode number */
1569 /* btree pointer type */
1570 typedef __be32 xfs_inobt_ptr_t;
1573 * block numbers in the AG.
1575 #define XFS_IBT_BLOCK(mp) ((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1576 #define XFS_FIBT_BLOCK(mp) ((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1579 * Reverse mapping btree format definitions
1581 * There is a btree for the reverse map per allocation group
1583 #define XFS_RMAP_CRC_MAGIC 0x524d4233 /* 'RMB3' */
1586 * Ownership info for an extent. This is used to create reverse-mapping
1589 #define XFS_OWNER_INFO_ATTR_FORK (1 << 0)
1590 #define XFS_OWNER_INFO_BMBT_BLOCK (1 << 1)
1591 struct xfs_owner_info {
1593 xfs_fileoff_t oi_offset;
1594 unsigned int oi_flags;
1598 * Special owner types.
1600 * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1601 * to tell us we have a special owner value. We use these for static metadata
1602 * allocated at mkfs/growfs time, as well as for freespace management metadata.
1604 #define XFS_RMAP_OWN_NULL (-1ULL) /* No owner, for growfs */
1605 #define XFS_RMAP_OWN_UNKNOWN (-2ULL) /* Unknown owner, for EFI recovery */
1606 #define XFS_RMAP_OWN_FS (-3ULL) /* static fs metadata */
1607 #define XFS_RMAP_OWN_LOG (-4ULL) /* static fs metadata */
1608 #define XFS_RMAP_OWN_AG (-5ULL) /* AG freespace btree blocks */
1609 #define XFS_RMAP_OWN_INOBT (-6ULL) /* Inode btree blocks */
1610 #define XFS_RMAP_OWN_INODES (-7ULL) /* Inode chunk */
1611 #define XFS_RMAP_OWN_REFC (-8ULL) /* refcount tree */
1612 #define XFS_RMAP_OWN_COW (-9ULL) /* cow allocations */
1613 #define XFS_RMAP_OWN_MIN (-10ULL) /* guard */
1615 #define XFS_RMAP_NON_INODE_OWNER(owner) (!!((owner) & (1ULL << 63)))
1618 * Data record structure
1620 struct xfs_rmap_rec {
1621 __be32 rm_startblock; /* extent start block */
1622 __be32 rm_blockcount; /* extent length */
1623 __be64 rm_owner; /* extent owner */
1624 __be64 rm_offset; /* offset within the owner */
1629 * rm_offset:63 is the attribute fork flag
1630 * rm_offset:62 is the bmbt block flag
1631 * rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1632 * rm_offset:54-60 aren't used and should be zero
1633 * rm_offset:0-53 is the block offset within the inode
1635 #define XFS_RMAP_OFF_ATTR_FORK ((uint64_t)1ULL << 63)
1636 #define XFS_RMAP_OFF_BMBT_BLOCK ((uint64_t)1ULL << 62)
1637 #define XFS_RMAP_OFF_UNWRITTEN ((uint64_t)1ULL << 61)
1639 #define XFS_RMAP_LEN_MAX ((uint32_t)~0U)
1640 #define XFS_RMAP_OFF_FLAGS (XFS_RMAP_OFF_ATTR_FORK | \
1641 XFS_RMAP_OFF_BMBT_BLOCK | \
1642 XFS_RMAP_OFF_UNWRITTEN)
1643 #define XFS_RMAP_OFF_MASK ((uint64_t)0x3FFFFFFFFFFFFFULL)
1645 #define XFS_RMAP_OFF(off) ((off) & XFS_RMAP_OFF_MASK)
1647 #define XFS_RMAP_IS_BMBT_BLOCK(off) (!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1648 #define XFS_RMAP_IS_ATTR_FORK(off) (!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1649 #define XFS_RMAP_IS_UNWRITTEN(len) (!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1651 #define RMAPBT_STARTBLOCK_BITLEN 32
1652 #define RMAPBT_BLOCKCOUNT_BITLEN 32
1653 #define RMAPBT_OWNER_BITLEN 64
1654 #define RMAPBT_ATTRFLAG_BITLEN 1
1655 #define RMAPBT_BMBTFLAG_BITLEN 1
1656 #define RMAPBT_EXNTFLAG_BITLEN 1
1657 #define RMAPBT_UNUSED_OFFSET_BITLEN 7
1658 #define RMAPBT_OFFSET_BITLEN 54
1660 #define XFS_RMAP_ATTR_FORK (1 << 0)
1661 #define XFS_RMAP_BMBT_BLOCK (1 << 1)
1662 #define XFS_RMAP_UNWRITTEN (1 << 2)
1663 #define XFS_RMAP_KEY_FLAGS (XFS_RMAP_ATTR_FORK | \
1664 XFS_RMAP_BMBT_BLOCK)
1665 #define XFS_RMAP_REC_FLAGS (XFS_RMAP_UNWRITTEN)
1666 struct xfs_rmap_irec {
1667 xfs_agblock_t rm_startblock; /* extent start block */
1668 xfs_extlen_t rm_blockcount; /* extent length */
1669 uint64_t rm_owner; /* extent owner */
1670 uint64_t rm_offset; /* offset within the owner */
1671 unsigned int rm_flags; /* state flags */
1677 * We don't use the length for lookups
1679 struct xfs_rmap_key {
1680 __be32 rm_startblock; /* extent start block */
1681 __be64 rm_owner; /* extent owner */
1682 __be64 rm_offset; /* offset within the owner */
1683 } __attribute__((packed));
1685 /* btree pointer type */
1686 typedef __be32 xfs_rmap_ptr_t;
1688 #define XFS_RMAP_BLOCK(mp) \
1689 (xfs_sb_version_hasfinobt(&((mp)->m_sb)) ? \
1690 XFS_FIBT_BLOCK(mp) + 1 : \
1691 XFS_IBT_BLOCK(mp) + 1)
1694 * Reference Count Btree format definitions
1697 #define XFS_REFC_CRC_MAGIC 0x52334643 /* 'R3FC' */
1699 unsigned int xfs_refc_block(struct xfs_mount *mp);
1702 * Data record/key structure
1704 * Each record associates a range of physical blocks (starting at
1705 * rc_startblock and ending rc_blockcount blocks later) with a reference
1706 * count (rc_refcount). Extents that are being used to stage a copy on
1707 * write (CoW) operation are recorded in the refcount btree with a
1708 * refcount of 1. All other records must have a refcount > 1 and must
1709 * track an extent mapped only by file data forks.
1711 * Extents with a single owner (attributes, metadata, non-shared file
1712 * data) are not tracked here. Free space is also not tracked here.
1713 * This is consistent with pre-reflink XFS.
1717 * Extents that are being used to stage a copy on write are stored
1718 * in the refcount btree with a refcount of 1 and the upper bit set
1719 * on the startblock. This speeds up mount time deletion of stale
1720 * staging extents because they're all at the right side of the tree.
1722 #define XFS_REFC_COW_START ((xfs_agblock_t)(1U << 31))
1723 #define REFCNTBT_COWFLAG_BITLEN 1
1724 #define REFCNTBT_AGBLOCK_BITLEN 31
1726 struct xfs_refcount_rec {
1727 __be32 rc_startblock; /* starting block number */
1728 __be32 rc_blockcount; /* count of blocks */
1729 __be32 rc_refcount; /* number of inodes linked here */
1732 struct xfs_refcount_key {
1733 __be32 rc_startblock; /* starting block number */
1736 struct xfs_refcount_irec {
1737 xfs_agblock_t rc_startblock; /* starting block number */
1738 xfs_extlen_t rc_blockcount; /* count of free blocks */
1739 xfs_nlink_t rc_refcount; /* number of inodes linked here */
1742 #define MAXREFCOUNT ((xfs_nlink_t)~0U)
1743 #define MAXREFCEXTLEN ((xfs_extlen_t)~0U)
1745 /* btree pointer type */
1746 typedef __be32 xfs_refcount_ptr_t;
1750 * BMAP Btree format definitions
1752 * This includes both the root block definition that sits inside an inode fork
1753 * and the record/pointer formats for the leaf/node in the blocks.
1755 #define XFS_BMAP_MAGIC 0x424d4150 /* 'BMAP' */
1756 #define XFS_BMAP_CRC_MAGIC 0x424d4133 /* 'BMA3' */
1759 * Bmap root header, on-disk form only.
1761 typedef struct xfs_bmdr_block {
1762 __be16 bb_level; /* 0 is a leaf */
1763 __be16 bb_numrecs; /* current # of data records */
1767 * Bmap btree record and extent descriptor.
1768 * l0:63 is an extent flag (value 1 indicates non-normal).
1769 * l0:9-62 are startoff.
1770 * l0:0-8 and l1:21-63 are startblock.
1771 * l1:0-20 are blockcount.
1773 #define BMBT_EXNTFLAG_BITLEN 1
1774 #define BMBT_STARTOFF_BITLEN 54
1775 #define BMBT_STARTBLOCK_BITLEN 52
1776 #define BMBT_BLOCKCOUNT_BITLEN 21
1778 #define BMBT_STARTOFF_MASK ((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1779 #define BMBT_BLOCKCOUNT_MASK ((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1)
1782 * bmbt records have a file offset (block) field that is 54 bits wide, so this
1783 * is the largest xfs_fileoff_t that we ever expect to see.
1785 #define XFS_MAX_FILEOFF (BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK)
1787 typedef struct xfs_bmbt_rec {
1791 typedef uint64_t xfs_bmbt_rec_base_t; /* use this for casts */
1792 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1795 * Values and macros for delayed-allocation startblock fields.
1797 #define STARTBLOCKVALBITS 17
1798 #define STARTBLOCKMASKBITS (15 + 20)
1799 #define STARTBLOCKMASK \
1800 (((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1802 static inline int isnullstartblock(xfs_fsblock_t x)
1804 return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1807 static inline xfs_fsblock_t nullstartblock(int k)
1809 ASSERT(k < (1 << STARTBLOCKVALBITS));
1810 return STARTBLOCKMASK | (k);
1813 static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1815 return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1819 * Key structure for non-leaf levels of the tree.
1821 typedef struct xfs_bmbt_key {
1822 __be64 br_startoff; /* starting file offset */
1823 } xfs_bmbt_key_t, xfs_bmdr_key_t;
1825 /* btree pointer type */
1826 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1830 * Generic Btree block format definitions
1832 * This is a combination of the actual format used on disk for short and long
1833 * format btrees. The first three fields are shared by both format, but the
1834 * pointers are different and should be used with care.
1836 * To get the size of the actual short or long form headers please use the size
1837 * macros below. Never use sizeof(xfs_btree_block).
1839 * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1840 * with the crc feature bit, and all accesses to them must be conditional on
1843 /* short form block header */
1844 struct xfs_btree_block_shdr {
1855 /* long form block header */
1856 struct xfs_btree_block_lhdr {
1865 __be32 bb_pad; /* padding for alignment */
1868 struct xfs_btree_block {
1869 __be32 bb_magic; /* magic number for block type */
1870 __be16 bb_level; /* 0 is a leaf */
1871 __be16 bb_numrecs; /* current # of data records */
1873 struct xfs_btree_block_shdr s;
1874 struct xfs_btree_block_lhdr l;
1878 /* size of a short form block */
1879 #define XFS_BTREE_SBLOCK_LEN \
1880 (offsetof(struct xfs_btree_block, bb_u) + \
1881 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1882 /* size of a long form block */
1883 #define XFS_BTREE_LBLOCK_LEN \
1884 (offsetof(struct xfs_btree_block, bb_u) + \
1885 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1887 /* sizes of CRC enabled btree blocks */
1888 #define XFS_BTREE_SBLOCK_CRC_LEN \
1889 (offsetof(struct xfs_btree_block, bb_u) + \
1890 sizeof(struct xfs_btree_block_shdr))
1891 #define XFS_BTREE_LBLOCK_CRC_LEN \
1892 (offsetof(struct xfs_btree_block, bb_u) + \
1893 sizeof(struct xfs_btree_block_lhdr))
1895 #define XFS_BTREE_SBLOCK_CRC_OFF \
1896 offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1897 #define XFS_BTREE_LBLOCK_CRC_OFF \
1898 offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1901 * On-disk XFS access control list structure.
1903 struct xfs_acl_entry {
1907 __be16 ae_pad; /* fill the implicit hole in the structure */
1912 struct xfs_acl_entry acl_entry[];
1916 * The number of ACL entries allowed is defined by the on-disk format.
1917 * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1918 * limited only by the maximum size of the xattr that stores the information.
1920 #define XFS_ACL_MAX_ENTRIES(mp) \
1921 (xfs_sb_version_hascrc(&mp->m_sb) \
1922 ? (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1923 sizeof(struct xfs_acl_entry) \
1926 #define XFS_ACL_SIZE(cnt) \
1927 (sizeof(struct xfs_acl) + \
1928 sizeof(struct xfs_acl_entry) * cnt)
1930 #define XFS_ACL_MAX_SIZE(mp) \
1931 XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1934 /* On-disk XFS extended attribute names */
1935 #define SGI_ACL_FILE "SGI_ACL_FILE"
1936 #define SGI_ACL_DEFAULT "SGI_ACL_DEFAULT"
1937 #define SGI_ACL_FILE_SIZE (sizeof(SGI_ACL_FILE)-1)
1938 #define SGI_ACL_DEFAULT_SIZE (sizeof(SGI_ACL_DEFAULT)-1)
1940 #endif /* __XFS_FORMAT_H__ */