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 be padded to 64 bit alignment.
94 */
95 typedef struct xfs_sb {
96 uint32_t sb_magicnum; /* magic number == XFS_SB_MAGIC */
97 uint32_t sb_blocksize; /* logical block size, bytes */
98 xfs_rfsblock_t sb_dblocks; /* number of data blocks */
99 xfs_rfsblock_t sb_rblocks; /* number of realtime blocks */
100 xfs_rtbxlen_t sb_rextents; /* number of realtime extents */
101 uuid_t sb_uuid; /* user-visible file system unique id */
102 xfs_fsblock_t sb_logstart; /* starting block of log if internal */
103 xfs_ino_t sb_rootino; /* root inode number */
104 xfs_ino_t sb_rbmino; /* bitmap inode for realtime extents */
105 xfs_ino_t sb_rsumino; /* summary inode for rt bitmap */
106 xfs_agblock_t sb_rextsize; /* realtime extent size, blocks */
107 xfs_agblock_t sb_agblocks; /* size of an allocation group */
108 xfs_agnumber_t sb_agcount; /* number of allocation groups */
109 xfs_extlen_t sb_rbmblocks; /* number of rt bitmap blocks */
110 xfs_extlen_t sb_logblocks; /* number of log blocks */
111 uint16_t sb_versionnum; /* header version == XFS_SB_VERSION */
112 uint16_t sb_sectsize; /* volume sector size, bytes */
113 uint16_t sb_inodesize; /* inode size, bytes */
114 uint16_t sb_inopblock; /* inodes per block */
115 char sb_fname[XFSLABEL_MAX] __nonstring; /* file system name */
116 uint8_t sb_blocklog; /* log2 of sb_blocksize */
117 uint8_t sb_sectlog; /* log2 of sb_sectsize */
118 uint8_t sb_inodelog; /* log2 of sb_inodesize */
119 uint8_t sb_inopblog; /* log2 of sb_inopblock */
120 uint8_t sb_agblklog; /* log2 of sb_agblocks (rounded up) */
121 uint8_t sb_rextslog; /* log2 of sb_rextents */
122 uint8_t sb_inprogress; /* mkfs is in progress, don't mount */
123 uint8_t sb_imax_pct; /* max % of fs for inode space */
124 /* statistics */
125 /*
126 * These fields must remain contiguous. If you really
127 * want to change their layout, make sure you fix the
128 * code in xfs_trans_apply_sb_deltas().
129 */
130 uint64_t sb_icount; /* allocated inodes */
131 uint64_t sb_ifree; /* free inodes */
132 uint64_t sb_fdblocks; /* free data blocks */
133 uint64_t sb_frextents; /* free realtime extents */
134 /*
135 * End contiguous fields.
136 */
137 xfs_ino_t sb_uquotino; /* user quota inode */
138 xfs_ino_t sb_gquotino; /* group quota inode */
139 uint16_t sb_qflags; /* quota flags */
140 uint8_t sb_flags; /* misc. flags */
141 uint8_t sb_shared_vn; /* shared version number */
142 xfs_extlen_t sb_inoalignmt; /* inode chunk alignment, fsblocks */
143 uint32_t sb_unit; /* stripe or raid unit */
144 uint32_t sb_width; /* stripe or raid width */
145 uint8_t sb_dirblklog; /* log2 of dir block size (fsbs) */
146 uint8_t sb_logsectlog; /* log2 of the log sector size */
147 uint16_t sb_logsectsize; /* sector size for the log, bytes */
148 uint32_t sb_logsunit; /* stripe unit size for the log */
149 uint32_t sb_features2; /* additional feature bits */
150
151 /*
152 * bad features2 field as a result of failing to pad the sb structure to
153 * 64 bits. Some machines will be using this field for features2 bits.
154 * Easiest just to mark it bad and not use it for anything else.
155 *
156 * This is not kept up to date in memory; it is always overwritten by
157 * the value in sb_features2 when formatting the incore superblock to
158 * the disk buffer.
159 */
160 uint32_t sb_bad_features2;
161
162 /* version 5 superblock fields start here */
163
164 /* feature masks */
165 uint32_t sb_features_compat;
166 uint32_t sb_features_ro_compat;
167 uint32_t sb_features_incompat;
168 uint32_t sb_features_log_incompat;
169
170 uint32_t sb_crc; /* superblock crc */
171 xfs_extlen_t sb_spino_align; /* sparse inode chunk alignment */
172
173 xfs_ino_t sb_pquotino; /* project quota inode */
174 xfs_lsn_t sb_lsn; /* last write sequence */
175 uuid_t sb_meta_uuid; /* metadata file system unique id */
176
177 xfs_ino_t sb_metadirino; /* metadata directory tree root */
178
179 xfs_rgnumber_t sb_rgcount; /* number of realtime groups */
180 xfs_rtxlen_t sb_rgextents; /* size of a realtime group in rtx */
181 uint8_t sb_rgblklog; /* rt group number shift */
182 uint8_t sb_pad[7]; /* zeroes */
183 xfs_rfsblock_t sb_rtstart; /* start of internal RT section (FSB) */
184 xfs_filblks_t sb_rtreserved; /* reserved (zoned) RT blocks */
185
186 /* must be padded to 64 bit alignment */
187 } xfs_sb_t;
188
189 /*
190 * Superblock - on disk version.
191 * Must be padded to 64 bit alignment.
192 */
193 struct xfs_dsb {
194 __be32 sb_magicnum; /* magic number == XFS_SB_MAGIC */
195 __be32 sb_blocksize; /* logical block size, bytes */
196 __be64 sb_dblocks; /* number of data blocks */
197 __be64 sb_rblocks; /* number of realtime blocks */
198 __be64 sb_rextents; /* number of realtime extents */
199 uuid_t sb_uuid; /* user-visible file system unique id */
200 __be64 sb_logstart; /* starting block of log if internal */
201 __be64 sb_rootino; /* root inode number */
202 __be64 sb_rbmino; /* bitmap inode for realtime extents */
203 __be64 sb_rsumino; /* summary inode for rt bitmap */
204 __be32 sb_rextsize; /* realtime extent size, blocks */
205 __be32 sb_agblocks; /* size of an allocation group */
206 __be32 sb_agcount; /* number of allocation groups */
207 __be32 sb_rbmblocks; /* number of rt bitmap blocks */
208 __be32 sb_logblocks; /* number of log blocks */
209 __be16 sb_versionnum; /* header version == XFS_SB_VERSION */
210 __be16 sb_sectsize; /* volume sector size, bytes */
211 __be16 sb_inodesize; /* inode size, bytes */
212 __be16 sb_inopblock; /* inodes per block */
213 char sb_fname[XFSLABEL_MAX]; /* file system name */
214 __u8 sb_blocklog; /* log2 of sb_blocksize */
215 __u8 sb_sectlog; /* log2 of sb_sectsize */
216 __u8 sb_inodelog; /* log2 of sb_inodesize */
217 __u8 sb_inopblog; /* log2 of sb_inopblock */
218 __u8 sb_agblklog; /* log2 of sb_agblocks (rounded up) */
219 __u8 sb_rextslog; /* log2 of sb_rextents */
220 __u8 sb_inprogress; /* mkfs is in progress, don't mount */
221 __u8 sb_imax_pct; /* max % of fs for inode space */
222 /* statistics */
223 /*
224 * These fields must remain contiguous. If you really
225 * want to change their layout, make sure you fix the
226 * code in xfs_trans_apply_sb_deltas().
227 */
228 __be64 sb_icount; /* allocated inodes */
229 __be64 sb_ifree; /* free inodes */
230 __be64 sb_fdblocks; /* free data blocks */
231 __be64 sb_frextents; /* free realtime extents */
232 /*
233 * End contiguous fields.
234 */
235 __be64 sb_uquotino; /* user quota inode */
236 __be64 sb_gquotino; /* group quota inode */
237 __be16 sb_qflags; /* quota flags */
238 __u8 sb_flags; /* misc. flags */
239 __u8 sb_shared_vn; /* shared version number */
240 __be32 sb_inoalignmt; /* inode chunk alignment, fsblocks */
241 __be32 sb_unit; /* stripe or raid unit */
242 __be32 sb_width; /* stripe or raid width */
243 __u8 sb_dirblklog; /* log2 of dir block size (fsbs) */
244 __u8 sb_logsectlog; /* log2 of the log sector size */
245 __be16 sb_logsectsize; /* sector size for the log, bytes */
246 __be32 sb_logsunit; /* stripe unit size for the log */
247 __be32 sb_features2; /* additional feature bits */
248 /*
249 * bad features2 field as a result of failing to pad the sb
250 * structure to 64 bits. Some machines will be using this field
251 * for features2 bits. Easiest just to mark it bad and not use
252 * it for anything else.
253 */
254 __be32 sb_bad_features2;
255
256 /* version 5 superblock fields start here */
257
258 /* feature masks */
259 __be32 sb_features_compat;
260 __be32 sb_features_ro_compat;
261 __be32 sb_features_incompat;
262 __be32 sb_features_log_incompat;
263
264 __le32 sb_crc; /* superblock crc */
265 __be32 sb_spino_align; /* sparse inode chunk alignment */
266
267 __be64 sb_pquotino; /* project quota inode */
268 __be64 sb_lsn; /* last write sequence */
269 uuid_t sb_meta_uuid; /* metadata file system unique id */
270
271 __be64 sb_metadirino; /* metadata directory tree root */
272 __be32 sb_rgcount; /* # of realtime groups */
273 __be32 sb_rgextents; /* size of rtgroup in rtx */
274 __u8 sb_rgblklog; /* rt group number shift */
275 __u8 sb_pad[7]; /* zeroes */
276 __be64 sb_rtstart; /* start of internal RT section (FSB) */
277 __be64 sb_rtreserved; /* reserved (zoned) RT blocks */
278
279 /*
280 * The size of this structure must be padded to 64 bit alignment.
281 *
282 * NOTE: Don't forget to update secondary_sb_whack in xfs_repair when
283 * adding new fields here.
284 */
285 };
286
287 #define XFS_SB_CRC_OFF offsetof(struct xfs_dsb, sb_crc)
288
289 /*
290 * Misc. Flags - warning - these will be cleared by xfs_repair unless
291 * a feature bit is set when the flag is used.
292 */
293 #define XFS_SBF_NOFLAGS 0x00 /* no flags set */
294 #define XFS_SBF_READONLY 0x01 /* only read-only mounts allowed */
295
296 /*
297 * define max. shared version we can interoperate with
298 */
299 #define XFS_SB_MAX_SHARED_VN 0
300
301 #define XFS_SB_VERSION_NUM(sbp) ((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
302
xfs_sb_is_v5(const struct xfs_sb * sbp)303 static inline bool xfs_sb_is_v5(const struct xfs_sb *sbp)
304 {
305 return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
306 }
307
308 /*
309 * Detect a mismatched features2 field. Older kernels read/wrote
310 * this into the wrong slot, so to be safe we keep them in sync.
311 */
xfs_sb_has_mismatched_features2(const struct xfs_sb * sbp)312 static inline bool xfs_sb_has_mismatched_features2(const struct xfs_sb *sbp)
313 {
314 return sbp->sb_bad_features2 != sbp->sb_features2;
315 }
316
xfs_sb_version_hasmorebits(const struct xfs_sb * sbp)317 static inline bool xfs_sb_version_hasmorebits(const struct xfs_sb *sbp)
318 {
319 return xfs_sb_is_v5(sbp) ||
320 (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
321 }
322
xfs_sb_version_addattr(struct xfs_sb * sbp)323 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
324 {
325 sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
326 }
327
xfs_sb_version_addquota(struct xfs_sb * sbp)328 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
329 {
330 sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
331 }
332
xfs_sb_version_addattr2(struct xfs_sb * sbp)333 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
334 {
335 sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
336 sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
337 }
338
xfs_sb_version_addprojid32(struct xfs_sb * sbp)339 static inline void xfs_sb_version_addprojid32(struct xfs_sb *sbp)
340 {
341 sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
342 sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
343 }
344
345 /*
346 * Extended v5 superblock feature masks. These are to be used for new v5
347 * superblock features only.
348 *
349 * Compat features are new features that old kernels will not notice or affect
350 * and so can mount read-write without issues.
351 *
352 * RO-Compat (read only) are features that old kernels can read but will break
353 * if they write. Hence only read-only mounts of such filesystems are allowed on
354 * kernels that don't support the feature bit.
355 *
356 * InCompat features are features which old kernels will not understand and so
357 * must not mount.
358 *
359 * Log-InCompat features are for changes to log formats or new transactions that
360 * can't be replayed on older kernels. The fields are set when the filesystem is
361 * mounted, and a clean unmount clears the fields.
362 */
363 #define XFS_SB_FEAT_COMPAT_ALL 0
364 #define XFS_SB_FEAT_COMPAT_UNKNOWN ~XFS_SB_FEAT_COMPAT_ALL
365 static inline bool
xfs_sb_has_compat_feature(const struct xfs_sb * sbp,uint32_t feature)366 xfs_sb_has_compat_feature(
367 const struct xfs_sb *sbp,
368 uint32_t feature)
369 {
370 return (sbp->sb_features_compat & feature) != 0;
371 }
372
373 #define XFS_SB_FEAT_RO_COMPAT_FINOBT (1 << 0) /* free inode btree */
374 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT (1 << 1) /* reverse map btree */
375 #define XFS_SB_FEAT_RO_COMPAT_REFLINK (1 << 2) /* reflinked files */
376 #define XFS_SB_FEAT_RO_COMPAT_INOBTCNT (1 << 3) /* inobt block counts */
377 #define XFS_SB_FEAT_RO_COMPAT_ALL \
378 (XFS_SB_FEAT_RO_COMPAT_FINOBT | \
379 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
380 XFS_SB_FEAT_RO_COMPAT_REFLINK| \
381 XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
382 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN ~XFS_SB_FEAT_RO_COMPAT_ALL
383 static inline bool
xfs_sb_has_ro_compat_feature(const struct xfs_sb * sbp,uint32_t feature)384 xfs_sb_has_ro_compat_feature(
385 const struct xfs_sb *sbp,
386 uint32_t feature)
387 {
388 return (sbp->sb_features_ro_compat & feature) != 0;
389 }
390
391 #define XFS_SB_FEAT_INCOMPAT_FTYPE (1 << 0) /* filetype in dirent */
392 #define XFS_SB_FEAT_INCOMPAT_SPINODES (1 << 1) /* sparse inode chunks */
393 #define XFS_SB_FEAT_INCOMPAT_META_UUID (1 << 2) /* metadata UUID */
394 #define XFS_SB_FEAT_INCOMPAT_BIGTIME (1 << 3) /* large timestamps */
395 #define XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR (1 << 4) /* needs xfs_repair */
396 #define XFS_SB_FEAT_INCOMPAT_NREXT64 (1 << 5) /* large extent counters */
397 #define XFS_SB_FEAT_INCOMPAT_EXCHRANGE (1 << 6) /* exchangerange supported */
398 #define XFS_SB_FEAT_INCOMPAT_PARENT (1 << 7) /* parent pointers */
399 #define XFS_SB_FEAT_INCOMPAT_METADIR (1 << 8) /* metadata dir tree */
400 #define XFS_SB_FEAT_INCOMPAT_ZONED (1 << 9) /* zoned RT allocator */
401 #define XFS_SB_FEAT_INCOMPAT_ZONE_GAPS (1 << 10) /* RTGs have LBA gaps */
402
403 #define XFS_SB_FEAT_INCOMPAT_ALL \
404 (XFS_SB_FEAT_INCOMPAT_FTYPE | \
405 XFS_SB_FEAT_INCOMPAT_SPINODES | \
406 XFS_SB_FEAT_INCOMPAT_META_UUID | \
407 XFS_SB_FEAT_INCOMPAT_BIGTIME | \
408 XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR | \
409 XFS_SB_FEAT_INCOMPAT_NREXT64 | \
410 XFS_SB_FEAT_INCOMPAT_EXCHRANGE | \
411 XFS_SB_FEAT_INCOMPAT_PARENT | \
412 XFS_SB_FEAT_INCOMPAT_METADIR | \
413 XFS_SB_FEAT_INCOMPAT_ZONED | \
414 XFS_SB_FEAT_INCOMPAT_ZONE_GAPS)
415
416 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN ~XFS_SB_FEAT_INCOMPAT_ALL
417 static inline bool
xfs_sb_has_incompat_feature(const struct xfs_sb * sbp,uint32_t feature)418 xfs_sb_has_incompat_feature(
419 const struct xfs_sb *sbp,
420 uint32_t feature)
421 {
422 return (sbp->sb_features_incompat & feature) != 0;
423 }
424
425 #define XFS_SB_FEAT_INCOMPAT_LOG_XATTRS (1 << 0) /* Delayed Attributes */
426 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL \
427 (XFS_SB_FEAT_INCOMPAT_LOG_XATTRS)
428 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN ~XFS_SB_FEAT_INCOMPAT_LOG_ALL
429 static inline bool
xfs_sb_has_incompat_log_feature(const struct xfs_sb * sbp,uint32_t feature)430 xfs_sb_has_incompat_log_feature(
431 const struct xfs_sb *sbp,
432 uint32_t feature)
433 {
434 return (sbp->sb_features_log_incompat & feature) != 0;
435 }
436
437 static inline void
xfs_sb_remove_incompat_log_features(struct xfs_sb * sbp)438 xfs_sb_remove_incompat_log_features(
439 struct xfs_sb *sbp)
440 {
441 sbp->sb_features_log_incompat &= ~XFS_SB_FEAT_INCOMPAT_LOG_ALL;
442 }
443
444 static inline void
xfs_sb_add_incompat_log_features(struct xfs_sb * sbp,unsigned int features)445 xfs_sb_add_incompat_log_features(
446 struct xfs_sb *sbp,
447 unsigned int features)
448 {
449 sbp->sb_features_log_incompat |= features;
450 }
451
xfs_sb_version_haslogxattrs(const struct xfs_sb * sbp)452 static inline bool xfs_sb_version_haslogxattrs(const struct xfs_sb *sbp)
453 {
454 return xfs_sb_is_v5(sbp) && (sbp->sb_features_log_incompat &
455 XFS_SB_FEAT_INCOMPAT_LOG_XATTRS);
456 }
457
458 static inline bool
xfs_is_quota_inode(struct xfs_sb * sbp,xfs_ino_t ino)459 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
460 {
461 return (ino == sbp->sb_uquotino ||
462 ino == sbp->sb_gquotino ||
463 ino == sbp->sb_pquotino);
464 }
465
466 #define XFS_SB_DADDR ((xfs_daddr_t)0) /* daddr in filesystem/ag */
467 #define XFS_SB_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
468
469 #define XFS_HDR_BLOCK(mp,d) ((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
470 #define XFS_DADDR_TO_FSB(mp,d) XFS_AGB_TO_FSB(mp, \
471 xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
472 #define XFS_FSB_TO_DADDR(mp,fsbno) XFS_AGB_TO_DADDR(mp, \
473 XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
474
475 /*
476 * File system sector to basic block conversions.
477 */
478 #define XFS_FSS_TO_BB(mp,sec) ((sec) << (mp)->m_sectbb_log)
479
480 /*
481 * File system block to basic block conversions.
482 */
483 #define XFS_FSB_TO_BB(mp,fsbno) ((fsbno) << (mp)->m_blkbb_log)
484 #define XFS_BB_TO_FSB(mp,bb) \
485 (((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
486 #define XFS_BB_TO_FSBT(mp,bb) ((bb) >> (mp)->m_blkbb_log)
487
488 /*
489 * File system block to byte conversions.
490 */
491 #define XFS_FSB_TO_B(mp,fsbno) ((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
492 #define XFS_B_TO_FSB(mp,b) \
493 ((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
494 #define XFS_B_TO_FSBT(mp,b) (((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
495
496 /*
497 * Allocation group header
498 *
499 * This is divided into three structures, placed in sequential 512-byte
500 * buffers after a copy of the superblock (also in a 512-byte buffer).
501 */
502 #define XFS_AGF_MAGIC 0x58414746 /* 'XAGF' */
503 #define XFS_AGI_MAGIC 0x58414749 /* 'XAGI' */
504 #define XFS_AGFL_MAGIC 0x5841464c /* 'XAFL' */
505 #define XFS_AGF_VERSION 1
506 #define XFS_AGI_VERSION 1
507
508 #define XFS_AGF_GOOD_VERSION(v) ((v) == XFS_AGF_VERSION)
509 #define XFS_AGI_GOOD_VERSION(v) ((v) == XFS_AGI_VERSION)
510
511 /*
512 * agf_cnt_level in the first AGF overlaps the EFS superblock's magic number.
513 * Since the magic numbers valid for EFS are > 64k, our value cannot be confused
514 * for an EFS superblock.
515 */
516
517 typedef struct xfs_agf {
518 /*
519 * Common allocation group header information
520 */
521 __be32 agf_magicnum; /* magic number == XFS_AGF_MAGIC */
522 __be32 agf_versionnum; /* header version == XFS_AGF_VERSION */
523 __be32 agf_seqno; /* sequence # starting from 0 */
524 __be32 agf_length; /* size in blocks of a.g. */
525 /*
526 * Freespace and rmap information
527 */
528 __be32 agf_bno_root; /* bnobt root block */
529 __be32 agf_cnt_root; /* cntbt root block */
530 __be32 agf_rmap_root; /* rmapbt root block */
531
532 __be32 agf_bno_level; /* bnobt btree levels */
533 __be32 agf_cnt_level; /* cntbt btree levels */
534 __be32 agf_rmap_level; /* rmapbt btree levels */
535
536 __be32 agf_flfirst; /* first freelist block's index */
537 __be32 agf_fllast; /* last freelist block's index */
538 __be32 agf_flcount; /* count of blocks in freelist */
539 __be32 agf_freeblks; /* total free blocks */
540
541 __be32 agf_longest; /* longest free space */
542 __be32 agf_btreeblks; /* # of blocks held in AGF btrees */
543 uuid_t agf_uuid; /* uuid of filesystem */
544
545 __be32 agf_rmap_blocks; /* rmapbt blocks used */
546 __be32 agf_refcount_blocks; /* refcountbt blocks used */
547
548 __be32 agf_refcount_root; /* refcount tree root block */
549 __be32 agf_refcount_level; /* refcount btree levels */
550
551 /*
552 * reserve some contiguous space for future logged fields before we add
553 * the unlogged fields. This makes the range logging via flags and
554 * structure offsets much simpler.
555 */
556 __be64 agf_spare64[14];
557
558 /* unlogged fields, written during buffer writeback. */
559 __be64 agf_lsn; /* last write sequence */
560 __be32 agf_crc; /* crc of agf sector */
561 __be32 agf_spare2;
562
563 /* structure must be padded to 64 bit alignment */
564 } xfs_agf_t;
565
566 #define XFS_AGF_CRC_OFF offsetof(struct xfs_agf, agf_crc)
567
568 #define XFS_AGF_MAGICNUM (1u << 0)
569 #define XFS_AGF_VERSIONNUM (1u << 1)
570 #define XFS_AGF_SEQNO (1u << 2)
571 #define XFS_AGF_LENGTH (1u << 3)
572 #define XFS_AGF_ROOTS (1u << 4)
573 #define XFS_AGF_LEVELS (1u << 5)
574 #define XFS_AGF_FLFIRST (1u << 6)
575 #define XFS_AGF_FLLAST (1u << 7)
576 #define XFS_AGF_FLCOUNT (1u << 8)
577 #define XFS_AGF_FREEBLKS (1u << 9)
578 #define XFS_AGF_LONGEST (1u << 10)
579 #define XFS_AGF_BTREEBLKS (1u << 11)
580 #define XFS_AGF_UUID (1u << 12)
581 #define XFS_AGF_RMAP_BLOCKS (1u << 13)
582 #define XFS_AGF_REFCOUNT_BLOCKS (1u << 14)
583 #define XFS_AGF_REFCOUNT_ROOT (1u << 15)
584 #define XFS_AGF_REFCOUNT_LEVEL (1u << 16)
585 #define XFS_AGF_SPARE64 (1u << 17)
586 #define XFS_AGF_NUM_BITS 18
587 #define XFS_AGF_ALL_BITS ((1u << XFS_AGF_NUM_BITS) - 1)
588
589 #define XFS_AGF_FLAGS \
590 { XFS_AGF_MAGICNUM, "MAGICNUM" }, \
591 { XFS_AGF_VERSIONNUM, "VERSIONNUM" }, \
592 { XFS_AGF_SEQNO, "SEQNO" }, \
593 { XFS_AGF_LENGTH, "LENGTH" }, \
594 { XFS_AGF_ROOTS, "ROOTS" }, \
595 { XFS_AGF_LEVELS, "LEVELS" }, \
596 { XFS_AGF_FLFIRST, "FLFIRST" }, \
597 { XFS_AGF_FLLAST, "FLLAST" }, \
598 { XFS_AGF_FLCOUNT, "FLCOUNT" }, \
599 { XFS_AGF_FREEBLKS, "FREEBLKS" }, \
600 { XFS_AGF_LONGEST, "LONGEST" }, \
601 { XFS_AGF_BTREEBLKS, "BTREEBLKS" }, \
602 { XFS_AGF_UUID, "UUID" }, \
603 { XFS_AGF_RMAP_BLOCKS, "RMAP_BLOCKS" }, \
604 { XFS_AGF_REFCOUNT_BLOCKS, "REFCOUNT_BLOCKS" }, \
605 { XFS_AGF_REFCOUNT_ROOT, "REFCOUNT_ROOT" }, \
606 { XFS_AGF_REFCOUNT_LEVEL, "REFCOUNT_LEVEL" }, \
607 { XFS_AGF_SPARE64, "SPARE64" }
608
609 /* disk block (xfs_daddr_t) in the AG */
610 #define XFS_AGF_DADDR(mp) ((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
611 #define XFS_AGF_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
612
613 /*
614 * Size of the unlinked inode hash table in the agi.
615 */
616 #define XFS_AGI_UNLINKED_BUCKETS 64
617
618 typedef struct xfs_agi {
619 /*
620 * Common allocation group header information
621 */
622 __be32 agi_magicnum; /* magic number == XFS_AGI_MAGIC */
623 __be32 agi_versionnum; /* header version == XFS_AGI_VERSION */
624 __be32 agi_seqno; /* sequence # starting from 0 */
625 __be32 agi_length; /* size in blocks of a.g. */
626 /*
627 * Inode information
628 * Inodes are mapped by interpreting the inode number, so no
629 * mapping data is needed here.
630 */
631 __be32 agi_count; /* count of allocated inodes */
632 __be32 agi_root; /* root of inode btree */
633 __be32 agi_level; /* levels in inode btree */
634 __be32 agi_freecount; /* number of free inodes */
635
636 __be32 agi_newino; /* new inode just allocated */
637 __be32 agi_dirino; /* last directory inode chunk */
638 /*
639 * Hash table of inodes which have been unlinked but are
640 * still being referenced.
641 */
642 __be32 agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
643 /*
644 * This marks the end of logging region 1 and start of logging region 2.
645 */
646 uuid_t agi_uuid; /* uuid of filesystem */
647 __be32 agi_crc; /* crc of agi sector */
648 __be32 agi_pad32;
649 __be64 agi_lsn; /* last write sequence */
650
651 __be32 agi_free_root; /* root of the free inode btree */
652 __be32 agi_free_level;/* levels in free inode btree */
653
654 __be32 agi_iblocks; /* inobt blocks used */
655 __be32 agi_fblocks; /* finobt blocks used */
656
657 /* structure must be padded to 64 bit alignment */
658 } xfs_agi_t;
659
660 #define XFS_AGI_CRC_OFF offsetof(struct xfs_agi, agi_crc)
661
662 #define XFS_AGI_MAGICNUM (1u << 0)
663 #define XFS_AGI_VERSIONNUM (1u << 1)
664 #define XFS_AGI_SEQNO (1u << 2)
665 #define XFS_AGI_LENGTH (1u << 3)
666 #define XFS_AGI_COUNT (1u << 4)
667 #define XFS_AGI_ROOT (1u << 5)
668 #define XFS_AGI_LEVEL (1u << 6)
669 #define XFS_AGI_FREECOUNT (1u << 7)
670 #define XFS_AGI_NEWINO (1u << 8)
671 #define XFS_AGI_DIRINO (1u << 9)
672 #define XFS_AGI_UNLINKED (1u << 10)
673 #define XFS_AGI_NUM_BITS_R1 11 /* end of the 1st agi logging region */
674 #define XFS_AGI_ALL_BITS_R1 ((1u << XFS_AGI_NUM_BITS_R1) - 1)
675 #define XFS_AGI_FREE_ROOT (1u << 11)
676 #define XFS_AGI_FREE_LEVEL (1u << 12)
677 #define XFS_AGI_IBLOCKS (1u << 13) /* both inobt/finobt block counters */
678 #define XFS_AGI_NUM_BITS_R2 14
679
680 /* disk block (xfs_daddr_t) in the AG */
681 #define XFS_AGI_DADDR(mp) ((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
682 #define XFS_AGI_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
683
684 /*
685 * The third a.g. block contains the a.g. freelist, an array
686 * of block pointers to blocks owned by the allocation btree code.
687 */
688 #define XFS_AGFL_DADDR(mp) ((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
689 #define XFS_AGFL_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
690 #define XFS_BUF_TO_AGFL(bp) ((struct xfs_agfl *)((bp)->b_addr))
691
692 struct xfs_agfl {
693 __be32 agfl_magicnum;
694 __be32 agfl_seqno;
695 uuid_t agfl_uuid;
696 __be64 agfl_lsn;
697 __be32 agfl_crc;
698 } __attribute__((packed));
699
700 #define XFS_AGFL_CRC_OFF offsetof(struct xfs_agfl, agfl_crc)
701
702 #define XFS_AGB_TO_FSB(mp,agno,agbno) \
703 (((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
704 #define XFS_FSB_TO_AGNO(mp,fsbno) \
705 ((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
706 #define XFS_FSB_TO_AGBNO(mp,fsbno) \
707 ((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
708 #define XFS_AGB_TO_DADDR(mp,agno,agbno) \
709 ((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
710 (xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
711 #define XFS_AG_DADDR(mp,agno,d) (XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
712
713 /*
714 * For checking for bad ranges of xfs_daddr_t's, covering multiple
715 * allocation groups or a single xfs_daddr_t that's a superblock copy.
716 */
717 #define XFS_AG_CHECK_DADDR(mp,d,len) \
718 ((len) == 1 ? \
719 ASSERT((d) == XFS_SB_DADDR || \
720 xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
721 ASSERT(xfs_daddr_to_agno(mp, d) == \
722 xfs_daddr_to_agno(mp, (d) + (len) - 1)))
723
724 /*
725 * Realtime bitmap information is accessed by the word, which is currently
726 * stored in host-endian format. Starting with the realtime groups feature,
727 * the words are stored in be32 ondisk.
728 */
729 union xfs_rtword_raw {
730 __u32 old;
731 __be32 rtg;
732 };
733
734 /*
735 * Realtime summary counts are accessed by the word, which is currently
736 * stored in host-endian format. Starting with the realtime groups feature,
737 * the words are stored in be32 ondisk.
738 */
739 union xfs_suminfo_raw {
740 __u32 old;
741 __be32 rtg;
742 };
743
744 /*
745 * Realtime allocation groups break the rt section into multiple pieces that
746 * could be locked independently. Realtime block group numbers are 32-bit
747 * quantities. Block numbers within a group are also 32-bit quantities, but
748 * the upper bit must never be set. rtgroup 0 might have a superblock in it,
749 * so the minimum size of an rtgroup is 2 rtx.
750 */
751 #define XFS_MAX_RGBLOCKS ((xfs_rgblock_t)(1U << 31) - 1)
752 #define XFS_MIN_RGEXTENTS ((xfs_rtxlen_t)2)
753 #define XFS_MAX_RGNUMBER ((xfs_rgnumber_t)(-1U))
754
755 #define XFS_RTSB_MAGIC 0x46726F67 /* 'Frog' */
756
757 /*
758 * Realtime superblock - on disk version. Must be padded to 64 bit alignment.
759 * The first block of the realtime volume contains this superblock.
760 */
761 struct xfs_rtsb {
762 __be32 rsb_magicnum; /* magic number == XFS_RTSB_MAGIC */
763 __le32 rsb_crc; /* superblock crc */
764
765 __be32 rsb_pad; /* zero */
766 unsigned char rsb_fname[XFSLABEL_MAX]; /* file system name */
767
768 uuid_t rsb_uuid; /* user-visible file system unique id */
769 uuid_t rsb_meta_uuid; /* metadata file system unique id */
770
771 /* must be padded to 64 bit alignment */
772 };
773
774 #define XFS_RTSB_CRC_OFF offsetof(struct xfs_rtsb, rsb_crc)
775 #define XFS_RTSB_DADDR ((xfs_daddr_t)0) /* daddr in rt section */
776
777 /*
778 * XFS Timestamps
779 * ==============
780 *
781 * Traditional ondisk inode timestamps consist of signed 32-bit counters for
782 * seconds and nanoseconds; time zero is the Unix epoch, Jan 1 00:00:00 UTC
783 * 1970, which means that the timestamp epoch is the same as the Unix epoch.
784 * Therefore, the ondisk min and max defined here can be used directly to
785 * constrain the incore timestamps on a Unix system. Note that we actually
786 * encode a __be64 value on disk.
787 *
788 * When the bigtime feature is enabled, ondisk inode timestamps become an
789 * unsigned 64-bit nanoseconds counter. This means that the bigtime inode
790 * timestamp epoch is the start of the classic timestamp range, which is
791 * Dec 13 20:45:52 UTC 1901. Because the epochs are not the same, callers
792 * /must/ use the bigtime conversion functions when encoding and decoding raw
793 * timestamps.
794 */
795 typedef __be64 xfs_timestamp_t;
796
797 /* Legacy timestamp encoding format. */
798 struct xfs_legacy_timestamp {
799 __be32 t_sec; /* timestamp seconds */
800 __be32 t_nsec; /* timestamp nanoseconds */
801 };
802
803 /*
804 * Smallest possible ondisk seconds value with traditional timestamps. This
805 * corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901.
806 */
807 #define XFS_LEGACY_TIME_MIN ((int64_t)S32_MIN)
808
809 /*
810 * Largest possible ondisk seconds value with traditional timestamps. This
811 * corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038.
812 */
813 #define XFS_LEGACY_TIME_MAX ((int64_t)S32_MAX)
814
815 /*
816 * Smallest possible ondisk seconds value with bigtime timestamps. This
817 * corresponds (after conversion to a Unix timestamp) with the traditional
818 * minimum timestamp of Dec 13 20:45:52 UTC 1901.
819 */
820 #define XFS_BIGTIME_TIME_MIN ((int64_t)0)
821
822 /*
823 * Largest supported ondisk seconds value with bigtime timestamps. This
824 * corresponds (after conversion to a Unix timestamp) with an incore timestamp
825 * of Jul 2 20:20:24 UTC 2486.
826 *
827 * We round down the ondisk limit so that the bigtime quota and inode max
828 * timestamps will be the same.
829 */
830 #define XFS_BIGTIME_TIME_MAX ((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL))
831
832 /*
833 * Bigtime epoch is set exactly to the minimum time value that a traditional
834 * 32-bit timestamp can represent when using the Unix epoch as a reference.
835 * Hence the Unix epoch is at a fixed offset into the supported bigtime
836 * timestamp range.
837 *
838 * The bigtime epoch also matches the minimum value an on-disk 32-bit XFS
839 * timestamp can represent so we will not lose any fidelity in converting
840 * to/from unix and bigtime timestamps.
841 *
842 * The following conversion factor converts a seconds counter from the Unix
843 * epoch to the bigtime epoch.
844 */
845 #define XFS_BIGTIME_EPOCH_OFFSET (-(int64_t)S32_MIN)
846
847 /* Convert a timestamp from the Unix epoch to the bigtime epoch. */
xfs_unix_to_bigtime(time64_t unix_seconds)848 static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds)
849 {
850 return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET;
851 }
852
853 /* Convert a timestamp from the bigtime epoch to the Unix epoch. */
xfs_bigtime_to_unix(uint64_t ondisk_seconds)854 static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds)
855 {
856 return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET;
857 }
858
859 enum xfs_metafile_type {
860 XFS_METAFILE_UNKNOWN, /* unknown */
861 XFS_METAFILE_DIR, /* metadir directory */
862 XFS_METAFILE_USRQUOTA, /* user quota */
863 XFS_METAFILE_GRPQUOTA, /* group quota */
864 XFS_METAFILE_PRJQUOTA, /* project quota */
865 XFS_METAFILE_RTBITMAP, /* rt bitmap */
866 XFS_METAFILE_RTSUMMARY, /* rt summary */
867 XFS_METAFILE_RTRMAP, /* rt rmap */
868 XFS_METAFILE_RTREFCOUNT, /* rt refcount */
869
870 XFS_METAFILE_MAX
871 } __packed;
872
873 #define XFS_METAFILE_TYPE_STR \
874 { XFS_METAFILE_UNKNOWN, "unknown" }, \
875 { XFS_METAFILE_DIR, "dir" }, \
876 { XFS_METAFILE_USRQUOTA, "usrquota" }, \
877 { XFS_METAFILE_GRPQUOTA, "grpquota" }, \
878 { XFS_METAFILE_PRJQUOTA, "prjquota" }, \
879 { XFS_METAFILE_RTBITMAP, "rtbitmap" }, \
880 { XFS_METAFILE_RTSUMMARY, "rtsummary" }, \
881 { XFS_METAFILE_RTRMAP, "rtrmap" }, \
882 { XFS_METAFILE_RTREFCOUNT, "rtrefcount" }
883
884 /*
885 * On-disk inode structure.
886 *
887 * This is just the header or "dinode core", the inode is expanded to fill a
888 * variable size the leftover area split into a data and an attribute fork.
889 * The format of the data and attribute fork depends on the format of the
890 * inode as indicated by di_format and di_aformat. To access the data and
891 * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
892 * below.
893 *
894 * There is a very similar struct xfs_log_dinode which matches the layout of
895 * this structure, but is kept in native format instead of big endian.
896 *
897 * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
898 * padding field for v3 inodes.
899 */
900 #define XFS_DINODE_MAGIC 0x494e /* 'IN' */
901 struct xfs_dinode {
902 __be16 di_magic; /* inode magic # = XFS_DINODE_MAGIC */
903 __be16 di_mode; /* mode and type of file */
904 __u8 di_version; /* inode version */
905 __u8 di_format; /* format of di_c data */
906 __be16 di_metatype; /* XFS_METAFILE_*; was di_onlink */
907 __be32 di_uid; /* owner's user id */
908 __be32 di_gid; /* owner's group id */
909 __be32 di_nlink; /* number of links to file */
910 __be16 di_projid_lo; /* lower part of owner's project id */
911 __be16 di_projid_hi; /* higher part owner's project id */
912 union {
913 /* Number of data fork extents if NREXT64 is set */
914 __be64 di_big_nextents;
915
916 /* Padding for V3 inodes without NREXT64 set. */
917 __be64 di_v3_pad;
918
919 /* Padding and inode flush counter for V2 inodes. */
920 struct {
921 __u8 di_v2_pad[6];
922 __be16 di_flushiter;
923 };
924 };
925 xfs_timestamp_t di_atime; /* time last accessed */
926 xfs_timestamp_t di_mtime; /* time last modified */
927 xfs_timestamp_t di_ctime; /* time created/inode modified */
928 __be64 di_size; /* number of bytes in file */
929 __be64 di_nblocks; /* # of direct & btree blocks used */
930 __be32 di_extsize; /* basic/minimum extent size for file */
931 union {
932 /*
933 * For V2 inodes and V3 inodes without NREXT64 set, this
934 * is the number of data and attr fork extents.
935 */
936 struct {
937 __be32 di_nextents;
938 __be16 di_anextents;
939 } __packed;
940
941 /* Number of attr fork extents if NREXT64 is set. */
942 struct {
943 __be32 di_big_anextents;
944 __be16 di_nrext64_pad;
945 } __packed;
946 } __packed;
947 __u8 di_forkoff; /* attr fork offs, <<3 for 64b align */
948 __s8 di_aformat; /* format of attr fork's data */
949 __be32 di_dmevmask; /* DMIG event mask */
950 __be16 di_dmstate; /* DMIG state info */
951 __be16 di_flags; /* random flags, XFS_DIFLAG_... */
952 __be32 di_gen; /* generation number */
953
954 /* di_next_unlinked is the only non-core field in the old dinode */
955 __be32 di_next_unlinked;/* agi unlinked list ptr */
956
957 /* start of the extended dinode, writable fields */
958 __le32 di_crc; /* CRC of the inode */
959 __be64 di_changecount; /* number of attribute changes */
960 __be64 di_lsn; /* flush sequence */
961 __be64 di_flags2; /* more random flags */
962 union {
963 /* basic cow extent size for (regular) file */
964 __be32 di_cowextsize;
965 /* used blocks in RTG for (zoned) rtrmap inode */
966 __be32 di_used_blocks;
967 };
968 __u8 di_pad2[12]; /* more padding for future expansion */
969
970 /* fields only written to during inode creation */
971 xfs_timestamp_t di_crtime; /* time created */
972 __be64 di_ino; /* inode number */
973 uuid_t di_uuid; /* UUID of the filesystem */
974
975 /* structure must be padded to 64 bit alignment */
976 };
977
978 #define XFS_DINODE_CRC_OFF offsetof(struct xfs_dinode, di_crc)
979
980 #define DI_MAX_FLUSH 0xffff
981
982 /*
983 * Size of the core inode on disk. Version 1 and 2 inodes have
984 * the same size, but version 3 has grown a few additional fields.
985 */
xfs_dinode_size(int version)986 static inline uint xfs_dinode_size(int version)
987 {
988 if (version == 3)
989 return sizeof(struct xfs_dinode);
990 return offsetof(struct xfs_dinode, di_crc);
991 }
992
993 /*
994 * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
995 * Since the pathconf interface is signed, we use 2^31 - 1 instead.
996 */
997 #define XFS_MAXLINK ((1U << 31) - 1U)
998
999 /*
1000 * Any file that hits the maximum ondisk link count should be pinned to avoid
1001 * a use-after-free situation.
1002 */
1003 #define XFS_NLINK_PINNED (~0U)
1004
1005 /*
1006 * Values for di_format
1007 *
1008 * This enum is used in string mapping in xfs_trace.h; please keep the
1009 * TRACE_DEFINE_ENUMs for it up to date.
1010 */
1011 enum xfs_dinode_fmt {
1012 XFS_DINODE_FMT_DEV, /* xfs_dev_t */
1013 XFS_DINODE_FMT_LOCAL, /* bulk data */
1014 XFS_DINODE_FMT_EXTENTS, /* struct xfs_bmbt_rec */
1015 XFS_DINODE_FMT_BTREE, /* struct xfs_bmdr_block */
1016 XFS_DINODE_FMT_UUID, /* added long ago, but never used */
1017 XFS_DINODE_FMT_META_BTREE, /* metadata btree */
1018 };
1019
1020 #define XFS_INODE_FORMAT_STR \
1021 { XFS_DINODE_FMT_DEV, "dev" }, \
1022 { XFS_DINODE_FMT_LOCAL, "local" }, \
1023 { XFS_DINODE_FMT_EXTENTS, "extent" }, \
1024 { XFS_DINODE_FMT_BTREE, "btree" }, \
1025 { XFS_DINODE_FMT_UUID, "uuid" }, \
1026 { XFS_DINODE_FMT_META_BTREE, "meta_btree" }
1027
1028 /*
1029 * Max values for extnum and aextnum.
1030 *
1031 * The original on-disk extent counts were held in signed fields, resulting in
1032 * maximum extent counts of 2^31 and 2^15 for the data and attr forks
1033 * respectively. Similarly the maximum extent length is limited to 2^21 blocks
1034 * by the 21-bit wide blockcount field of a BMBT extent record.
1035 *
1036 * The newly introduced data fork extent counter can hold a 64-bit value,
1037 * however the maximum number of extents in a file is also limited to 2^54
1038 * extents by the 54-bit wide startoff field of a BMBT extent record.
1039 *
1040 * It is further limited by the maximum supported file size of 2^63
1041 * *bytes*. This leads to a maximum extent count for maximally sized filesystem
1042 * blocks (64kB) of:
1043 *
1044 * 2^63 bytes / 2^16 bytes per block = 2^47 blocks
1045 *
1046 * Rounding up 47 to the nearest multiple of bits-per-byte results in 48. Hence
1047 * 2^48 was chosen as the maximum data fork extent count.
1048 *
1049 * The maximum file size that can be represented by the data fork extent counter
1050 * in the worst case occurs when all extents are 1 block in length and each
1051 * block is 1KB in size.
1052 *
1053 * With XFS_MAX_EXTCNT_DATA_FORK_SMALL representing maximum extent count and
1054 * with 1KB sized blocks, a file can reach up to,
1055 * 1KB * (2^31) = 2TB
1056 *
1057 * This is much larger than the theoretical maximum size of a directory
1058 * i.e. XFS_DIR2_SPACE_SIZE * XFS_DIR2_MAX_SPACES = ~96GB.
1059 *
1060 * Hence, a directory inode can never overflow its data fork extent counter.
1061 */
1062 #define XFS_MAX_EXTCNT_DATA_FORK_LARGE ((xfs_extnum_t)((1ULL << 48) - 1))
1063 #define XFS_MAX_EXTCNT_ATTR_FORK_LARGE ((xfs_extnum_t)((1ULL << 32) - 1))
1064 #define XFS_MAX_EXTCNT_DATA_FORK_SMALL ((xfs_extnum_t)((1ULL << 31) - 1))
1065 #define XFS_MAX_EXTCNT_ATTR_FORK_SMALL ((xfs_extnum_t)((1ULL << 15) - 1))
1066
1067 /*
1068 * When we upgrade an inode to the large extent counts, the maximum value by
1069 * which the extent count can increase is bound by the change in size of the
1070 * on-disk field. No upgrade operation should ever be adding more than a few
1071 * tens of extents, so if we get a really large value it is a sign of a code bug
1072 * or corruption.
1073 */
1074 #define XFS_MAX_EXTCNT_UPGRADE_NR \
1075 min(XFS_MAX_EXTCNT_ATTR_FORK_LARGE - XFS_MAX_EXTCNT_ATTR_FORK_SMALL, \
1076 XFS_MAX_EXTCNT_DATA_FORK_LARGE - XFS_MAX_EXTCNT_DATA_FORK_SMALL)
1077
1078 /*
1079 * Inode minimum and maximum sizes.
1080 */
1081 #define XFS_DINODE_MIN_LOG 8
1082 #define XFS_DINODE_MAX_LOG 11
1083 #define XFS_DINODE_MIN_SIZE (1 << XFS_DINODE_MIN_LOG)
1084 #define XFS_DINODE_MAX_SIZE (1 << XFS_DINODE_MAX_LOG)
1085
1086 /*
1087 * Inode size for given fs.
1088 */
1089 #define XFS_DINODE_SIZE(mp) \
1090 (xfs_has_v3inodes(mp) ? \
1091 sizeof(struct xfs_dinode) : \
1092 offsetof(struct xfs_dinode, di_crc))
1093 #define XFS_LITINO(mp) \
1094 ((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(mp))
1095
1096 /*
1097 * Inode data & attribute fork sizes, per inode.
1098 */
1099 #define XFS_DFORK_BOFF(dip) ((int)((dip)->di_forkoff << 3))
1100
1101 #define XFS_DFORK_DSIZE(dip,mp) \
1102 ((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp))
1103 #define XFS_DFORK_ASIZE(dip,mp) \
1104 ((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0)
1105 #define XFS_DFORK_SIZE(dip,mp,w) \
1106 ((w) == XFS_DATA_FORK ? \
1107 XFS_DFORK_DSIZE(dip, mp) : \
1108 XFS_DFORK_ASIZE(dip, mp))
1109
1110 #define XFS_DFORK_MAXEXT(dip, mp, w) \
1111 (XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
1112
1113 /*
1114 * Return pointers to the data or attribute forks.
1115 */
1116 #define XFS_DFORK_DPTR(dip) \
1117 ((void *)dip + xfs_dinode_size(dip->di_version))
1118 #define XFS_DFORK_APTR(dip) \
1119 (XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
1120 #define XFS_DFORK_PTR(dip,w) \
1121 ((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
1122
1123 #define XFS_DFORK_FORMAT(dip,w) \
1124 ((w) == XFS_DATA_FORK ? \
1125 (dip)->di_format : \
1126 (dip)->di_aformat)
1127
1128 /*
1129 * For block and character special files the 32bit dev_t is stored at the
1130 * beginning of the data fork.
1131 */
xfs_dinode_get_rdev(struct xfs_dinode * dip)1132 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1133 {
1134 return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1135 }
1136
xfs_dinode_put_rdev(struct xfs_dinode * dip,xfs_dev_t rdev)1137 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1138 {
1139 *(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1140 }
1141
1142 /*
1143 * Values for di_flags
1144 */
1145 #define XFS_DIFLAG_REALTIME_BIT 0 /* file's blocks come from rt area */
1146 #define XFS_DIFLAG_PREALLOC_BIT 1 /* file space has been preallocated */
1147 #define XFS_DIFLAG_NEWRTBM_BIT 2 /* for rtbitmap inode, new format */
1148 #define XFS_DIFLAG_IMMUTABLE_BIT 3 /* inode is immutable */
1149 #define XFS_DIFLAG_APPEND_BIT 4 /* inode is append-only */
1150 #define XFS_DIFLAG_SYNC_BIT 5 /* inode is written synchronously */
1151 #define XFS_DIFLAG_NOATIME_BIT 6 /* do not update atime */
1152 #define XFS_DIFLAG_NODUMP_BIT 7 /* do not dump */
1153 #define XFS_DIFLAG_RTINHERIT_BIT 8 /* create with realtime bit set */
1154 #define XFS_DIFLAG_PROJINHERIT_BIT 9 /* create with parents projid */
1155 #define XFS_DIFLAG_NOSYMLINKS_BIT 10 /* disallow symlink creation */
1156 #define XFS_DIFLAG_EXTSIZE_BIT 11 /* inode extent size allocator hint */
1157 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12 /* inherit inode extent size */
1158 #define XFS_DIFLAG_NODEFRAG_BIT 13 /* do not reorganize/defragment */
1159 #define XFS_DIFLAG_FILESTREAM_BIT 14 /* use filestream allocator */
1160 /* Do not use bit 15, di_flags is legacy and unchanging now */
1161
1162 #define XFS_DIFLAG_REALTIME (1 << XFS_DIFLAG_REALTIME_BIT)
1163 #define XFS_DIFLAG_PREALLOC (1 << XFS_DIFLAG_PREALLOC_BIT)
1164 #define XFS_DIFLAG_NEWRTBM (1 << XFS_DIFLAG_NEWRTBM_BIT)
1165 #define XFS_DIFLAG_IMMUTABLE (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1166 #define XFS_DIFLAG_APPEND (1 << XFS_DIFLAG_APPEND_BIT)
1167 #define XFS_DIFLAG_SYNC (1 << XFS_DIFLAG_SYNC_BIT)
1168 #define XFS_DIFLAG_NOATIME (1 << XFS_DIFLAG_NOATIME_BIT)
1169 #define XFS_DIFLAG_NODUMP (1 << XFS_DIFLAG_NODUMP_BIT)
1170 #define XFS_DIFLAG_RTINHERIT (1 << XFS_DIFLAG_RTINHERIT_BIT)
1171 #define XFS_DIFLAG_PROJINHERIT (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1172 #define XFS_DIFLAG_NOSYMLINKS (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1173 #define XFS_DIFLAG_EXTSIZE (1 << XFS_DIFLAG_EXTSIZE_BIT)
1174 #define XFS_DIFLAG_EXTSZINHERIT (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1175 #define XFS_DIFLAG_NODEFRAG (1 << XFS_DIFLAG_NODEFRAG_BIT)
1176 #define XFS_DIFLAG_FILESTREAM (1 << XFS_DIFLAG_FILESTREAM_BIT)
1177
1178 #define XFS_DIFLAG_ANY \
1179 (XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1180 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1181 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1182 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1183 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1184
1185 /*
1186 * Values for di_flags2 These start by being exposed to userspace in the upper
1187 * 16 bits of the XFS_XFLAG_s range.
1188 */
1189 /* use DAX for this inode */
1190 #define XFS_DIFLAG2_DAX_BIT 0
1191
1192 /* file's blocks may be shared */
1193 #define XFS_DIFLAG2_REFLINK_BIT 1
1194
1195 /* copy on write extent size hint */
1196 #define XFS_DIFLAG2_COWEXTSIZE_BIT 2
1197
1198 /* big timestamps */
1199 #define XFS_DIFLAG2_BIGTIME_BIT 3
1200
1201 /* large extent counters */
1202 #define XFS_DIFLAG2_NREXT64_BIT 4
1203
1204 /*
1205 * The inode contains filesystem metadata and can be found through the metadata
1206 * directory tree. Metadata inodes must satisfy the following constraints:
1207 *
1208 * - V5 filesystem (and ftype) are enabled;
1209 * - The only valid modes are regular files and directories;
1210 * - The access bits must be zero;
1211 * - DMAPI event and state masks are zero;
1212 * - The user and group IDs must be zero;
1213 * - The project ID can be used as a u32 annotation;
1214 * - The immutable, sync, noatime, nodump, nodefrag flags must be set.
1215 * - The dax flag must not be set.
1216 * - Directories must have nosymlinks set.
1217 *
1218 * These requirements are chosen defensively to minimize the ability of
1219 * userspace to read or modify the contents, should a metadata file ever
1220 * escape to userspace.
1221 *
1222 * There are further constraints on the directory tree itself:
1223 *
1224 * - Metadata inodes must never be resolvable through the root directory;
1225 * - They must never be accessed by userspace;
1226 * - Metadata directory entries must have correct ftype.
1227 *
1228 * Superblock-rooted metadata files must have the METADATA iflag set even
1229 * though they do not have a parent directory.
1230 */
1231 #define XFS_DIFLAG2_METADATA_BIT 5
1232
1233 #define XFS_DIFLAG2_DAX (1ULL << XFS_DIFLAG2_DAX_BIT)
1234 #define XFS_DIFLAG2_REFLINK (1ULL << XFS_DIFLAG2_REFLINK_BIT)
1235 #define XFS_DIFLAG2_COWEXTSIZE (1ULL << XFS_DIFLAG2_COWEXTSIZE_BIT)
1236 #define XFS_DIFLAG2_BIGTIME (1ULL << XFS_DIFLAG2_BIGTIME_BIT)
1237 #define XFS_DIFLAG2_NREXT64 (1ULL << XFS_DIFLAG2_NREXT64_BIT)
1238 #define XFS_DIFLAG2_METADATA (1ULL << XFS_DIFLAG2_METADATA_BIT)
1239
1240 #define XFS_DIFLAG2_ANY \
1241 (XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \
1242 XFS_DIFLAG2_BIGTIME | XFS_DIFLAG2_NREXT64 | XFS_DIFLAG2_METADATA)
1243
xfs_dinode_has_bigtime(const struct xfs_dinode * dip)1244 static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip)
1245 {
1246 return dip->di_version >= 3 &&
1247 (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME));
1248 }
1249
xfs_dinode_has_large_extent_counts(const struct xfs_dinode * dip)1250 static inline bool xfs_dinode_has_large_extent_counts(
1251 const struct xfs_dinode *dip)
1252 {
1253 return dip->di_version >= 3 &&
1254 (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_NREXT64));
1255 }
1256
xfs_dinode_is_metadir(const struct xfs_dinode * dip)1257 static inline bool xfs_dinode_is_metadir(const struct xfs_dinode *dip)
1258 {
1259 return dip->di_version >= 3 &&
1260 (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_METADATA));
1261 }
1262
1263 /*
1264 * Inode number format:
1265 * low inopblog bits - offset in block
1266 * next agblklog bits - block number in ag
1267 * next agno_log bits - ag number
1268 * high agno_log-agblklog-inopblog bits - 0
1269 */
1270 #define XFS_INO_MASK(k) (uint32_t)((1ULL << (k)) - 1)
1271 #define XFS_INO_OFFSET_BITS(mp) (mp)->m_sb.sb_inopblog
1272 #define XFS_INO_AGBNO_BITS(mp) (mp)->m_sb.sb_agblklog
1273 #define XFS_INO_AGINO_BITS(mp) ((mp)->m_ino_geo.agino_log)
1274 #define XFS_INO_AGNO_BITS(mp) (mp)->m_agno_log
1275 #define XFS_INO_BITS(mp) \
1276 XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1277 #define XFS_INO_TO_AGNO(mp,i) \
1278 ((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1279 #define XFS_INODE_TO_AGNO(ip) \
1280 XFS_INO_TO_AGNO((ip)->i_mount, I_INO(ip))
1281 #define XFS_INO_TO_AGINO(mp,i) \
1282 ((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1283 #define XFS_INODE_TO_AGINO(ip) \
1284 XFS_INO_TO_AGINO((ip)->i_mount, I_INO(ip))
1285 #define XFS_INO_TO_AGBNO(mp,i) \
1286 (((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1287 XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1288 #define XFS_INO_TO_OFFSET(mp,i) \
1289 ((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1290 #define XFS_INO_TO_FSB(mp,i) \
1291 XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1292 #define XFS_INODE_TO_FSB(ip) \
1293 XFS_INO_TO_FSB((ip)->i_mount, I_INO(ip))
1294 #define XFS_AGINO_TO_INO(mp,a,i) \
1295 (((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1296 #define XFS_AGINO_TO_AGBNO(mp,i) ((i) >> XFS_INO_OFFSET_BITS(mp))
1297 #define XFS_AGINO_TO_OFFSET(mp,i) \
1298 ((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1299 #define XFS_OFFBNO_TO_AGINO(mp,b,o) \
1300 ((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1301 #define XFS_FSB_TO_INO(mp, b) ((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1302 #define XFS_AGB_TO_AGINO(mp, b) ((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1303
1304 #define XFS_MAXINUMBER ((xfs_ino_t)((1ULL << 56) - 1ULL))
1305 #define XFS_MAXINUMBER_32 ((xfs_ino_t)((1ULL << 32) - 1ULL))
1306
1307 /*
1308 * RealTime Device format definitions
1309 */
1310
1311 /* Min and max rt extent sizes, specified in bytes */
1312 #define XFS_MAX_RTEXTSIZE (1024 * 1024 * 1024) /* 1GB */
1313 #define XFS_DFL_RTEXTSIZE (64 * 1024) /* 64kB */
1314 #define XFS_MIN_RTEXTSIZE (4 * 1024) /* 4kB */
1315
1316 /*
1317 * RT bit manipulation macros.
1318 */
1319 #define XFS_RTBITMAP_MAGIC 0x424D505A /* BMPZ */
1320 #define XFS_RTSUMMARY_MAGIC 0x53554D59 /* SUMY */
1321
1322 struct xfs_rtbuf_blkinfo {
1323 __be32 rt_magic; /* validity check on block */
1324 __be32 rt_crc; /* CRC of block */
1325 __be64 rt_owner; /* inode that owns the block */
1326 __be64 rt_blkno; /* first block of the buffer */
1327 __be64 rt_lsn; /* sequence number of last write */
1328 uuid_t rt_uuid; /* filesystem we belong to */
1329 };
1330
1331 #define XFS_RTBUF_CRC_OFF \
1332 offsetof(struct xfs_rtbuf_blkinfo, rt_crc)
1333
1334 /*
1335 * Dquot and dquot block format definitions
1336 */
1337 #define XFS_DQUOT_MAGIC 0x4451 /* 'DQ' */
1338 #define XFS_DQUOT_VERSION (uint8_t)0x01 /* latest version number */
1339
1340 #define XFS_DQTYPE_USER (1u << 0) /* user dquot record */
1341 #define XFS_DQTYPE_PROJ (1u << 1) /* project dquot record */
1342 #define XFS_DQTYPE_GROUP (1u << 2) /* group dquot record */
1343 #define XFS_DQTYPE_BIGTIME (1u << 7) /* large expiry timestamps */
1344
1345 /* bitmask to determine if this is a user/group/project dquot */
1346 #define XFS_DQTYPE_REC_MASK (XFS_DQTYPE_USER | \
1347 XFS_DQTYPE_PROJ | \
1348 XFS_DQTYPE_GROUP)
1349
1350 #define XFS_DQTYPE_ANY (XFS_DQTYPE_REC_MASK | \
1351 XFS_DQTYPE_BIGTIME)
1352
1353 /*
1354 * XFS Quota Timers
1355 * ================
1356 *
1357 * Traditional quota grace period expiration timers are an unsigned 32-bit
1358 * seconds counter; time zero is the Unix epoch, Jan 1 00:00:01 UTC 1970.
1359 * Note that an expiration value of zero means that the quota limit has not
1360 * been reached, and therefore no expiration has been set. Therefore, the
1361 * ondisk min and max defined here can be used directly to constrain the incore
1362 * quota expiration timestamps on a Unix system.
1363 *
1364 * When bigtime is enabled, we trade two bits of precision to expand the
1365 * expiration timeout range to match that of big inode timestamps. The min and
1366 * max recorded here are the on-disk limits, not a Unix timestamp.
1367 *
1368 * The grace period for each quota type is stored in the root dquot (id = 0)
1369 * and is applied to a non-root dquot when it exceeds the soft or hard limits.
1370 * The length of quota grace periods are unsigned 32-bit quantities measured in
1371 * units of seconds. A value of zero means to use the default period.
1372 */
1373
1374 /*
1375 * Smallest possible ondisk quota expiration value with traditional timestamps.
1376 * This corresponds exactly with the incore expiration Jan 1 00:00:01 UTC 1970.
1377 */
1378 #define XFS_DQ_LEGACY_EXPIRY_MIN ((int64_t)1)
1379
1380 /*
1381 * Largest possible ondisk quota expiration value with traditional timestamps.
1382 * This corresponds exactly with the incore expiration Feb 7 06:28:15 UTC 2106.
1383 */
1384 #define XFS_DQ_LEGACY_EXPIRY_MAX ((int64_t)U32_MAX)
1385
1386 /*
1387 * Smallest possible ondisk quota expiration value with bigtime timestamps.
1388 * This corresponds (after conversion to a Unix timestamp) with the incore
1389 * expiration of Jan 1 00:00:04 UTC 1970.
1390 */
1391 #define XFS_DQ_BIGTIME_EXPIRY_MIN (XFS_DQ_LEGACY_EXPIRY_MIN)
1392
1393 /*
1394 * Largest supported ondisk quota expiration value with bigtime timestamps.
1395 * This corresponds (after conversion to a Unix timestamp) with an incore
1396 * expiration of Jul 2 20:20:24 UTC 2486.
1397 *
1398 * The ondisk field supports values up to -1U, which corresponds to an incore
1399 * expiration in 2514. This is beyond the maximum the bigtime inode timestamp,
1400 * so we cap the maximum bigtime quota expiration to the max inode timestamp.
1401 */
1402 #define XFS_DQ_BIGTIME_EXPIRY_MAX ((int64_t)4074815106U)
1403
1404 /*
1405 * The following conversion factors assist in converting a quota expiration
1406 * timestamp between the incore and ondisk formats.
1407 */
1408 #define XFS_DQ_BIGTIME_SHIFT (2)
1409 #define XFS_DQ_BIGTIME_SLACK ((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1)
1410
1411 /* Convert an incore quota expiration timestamp to an ondisk bigtime value. */
xfs_dq_unix_to_bigtime(time64_t unix_seconds)1412 static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds)
1413 {
1414 /*
1415 * Round the expiration timestamp up to the nearest bigtime timestamp
1416 * that we can store, to give users the most time to fix problems.
1417 */
1418 return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >>
1419 XFS_DQ_BIGTIME_SHIFT;
1420 }
1421
1422 /* Convert an ondisk bigtime quota expiration value to an incore timestamp. */
xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)1423 static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)
1424 {
1425 return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT;
1426 }
1427
1428 /*
1429 * Default quota grace periods, ranging from zero (use the compiled defaults)
1430 * to ~136 years. These are applied to a non-root dquot that has exceeded
1431 * either limit.
1432 */
1433 #define XFS_DQ_GRACE_MIN ((int64_t)0)
1434 #define XFS_DQ_GRACE_MAX ((int64_t)U32_MAX)
1435
1436 /* Maximum id value for a quota record */
1437 #define XFS_DQ_ID_MAX (U32_MAX)
1438
1439 /*
1440 * This is the main portion of the on-disk representation of quota information
1441 * for a user. We pad this with some more expansion room to construct the on
1442 * disk structure.
1443 */
1444 struct xfs_disk_dquot {
1445 __be16 d_magic; /* dquot magic = XFS_DQUOT_MAGIC */
1446 __u8 d_version; /* dquot version */
1447 __u8 d_type; /* XFS_DQTYPE_USER/PROJ/GROUP */
1448 __be32 d_id; /* user,project,group id */
1449 __be64 d_blk_hardlimit;/* absolute limit on disk blks */
1450 __be64 d_blk_softlimit;/* preferred limit on disk blks */
1451 __be64 d_ino_hardlimit;/* maximum # allocated inodes */
1452 __be64 d_ino_softlimit;/* preferred inode limit */
1453 __be64 d_bcount; /* disk blocks owned by the user */
1454 __be64 d_icount; /* inodes owned by the user */
1455 __be32 d_itimer; /* zero if within inode limits if not,
1456 this is when we refuse service */
1457 __be32 d_btimer; /* similar to above; for disk blocks */
1458 __be16 d_iwarns; /* warnings issued wrt num inodes */
1459 __be16 d_bwarns; /* warnings issued wrt disk blocks */
1460 __be32 d_pad0; /* 64 bit align */
1461 __be64 d_rtb_hardlimit;/* absolute limit on realtime blks */
1462 __be64 d_rtb_softlimit;/* preferred limit on RT disk blks */
1463 __be64 d_rtbcount; /* realtime blocks owned */
1464 __be32 d_rtbtimer; /* similar to above; for RT disk blocks */
1465 __be16 d_rtbwarns; /* warnings issued wrt RT disk blocks */
1466 __be16 d_pad;
1467 };
1468
1469 /*
1470 * This is what goes on disk. This is separated from the xfs_disk_dquot because
1471 * carrying the unnecessary padding would be a waste of memory.
1472 */
1473 struct xfs_dqblk {
1474 struct xfs_disk_dquot dd_diskdq; /* portion living incore as well */
1475 char dd_fill[4];/* filling for posterity */
1476
1477 /*
1478 * These two are only present on filesystems with the CRC bits set.
1479 */
1480 __be32 dd_crc; /* checksum */
1481 __be64 dd_lsn; /* last modification in log */
1482 uuid_t dd_uuid; /* location information */
1483 };
1484
1485 #define XFS_DQUOT_CRC_OFF offsetof(struct xfs_dqblk, dd_crc)
1486
1487 /*
1488 * This defines the unit of allocation of dquots.
1489 *
1490 * Currently, it is just one file system block, and a 4K blk contains 30
1491 * (136 * 30 = 4080) dquots. It's probably not worth trying to make
1492 * this more dynamic.
1493 *
1494 * However, if this number is changed, we have to make sure that we don't
1495 * implicitly assume that we do allocations in chunks of a single filesystem
1496 * block in the dquot/xqm code.
1497 *
1498 * This is part of the ondisk format because the structure size is not a power
1499 * of two, which leaves slack at the end of the disk block.
1500 */
1501 #define XFS_DQUOT_CLUSTER_SIZE_FSB (xfs_filblks_t)1
1502
1503 /*
1504 * Remote symlink format and access functions.
1505 */
1506 #define XFS_SYMLINK_MAGIC 0x58534c4d /* XSLM */
1507
1508 struct xfs_dsymlink_hdr {
1509 __be32 sl_magic;
1510 __be32 sl_offset;
1511 __be32 sl_bytes;
1512 __be32 sl_crc;
1513 uuid_t sl_uuid;
1514 __be64 sl_owner;
1515 __be64 sl_blkno;
1516 __be64 sl_lsn;
1517 };
1518
1519 #define XFS_SYMLINK_CRC_OFF offsetof(struct xfs_dsymlink_hdr, sl_crc)
1520
1521 #define XFS_SYMLINK_MAXLEN 1024
1522 /*
1523 * The maximum pathlen is 1024 bytes. Since the minimum file system
1524 * blocksize is 512 bytes, we can get a max of 3 extents back from
1525 * bmapi when crc headers are taken into account.
1526 */
1527 #define XFS_SYMLINK_MAPS 3
1528
1529 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize) \
1530 ((bufsize) - (xfs_has_crc((mp)) ? \
1531 sizeof(struct xfs_dsymlink_hdr) : 0))
1532
1533
1534 /*
1535 * Allocation Btree format definitions
1536 *
1537 * There are two on-disk btrees, one sorted by blockno and one sorted
1538 * by blockcount and blockno. All blocks look the same to make the code
1539 * simpler; if we have time later, we'll make the optimizations.
1540 */
1541 #define XFS_ABTB_MAGIC 0x41425442 /* 'ABTB' for bno tree */
1542 #define XFS_ABTB_CRC_MAGIC 0x41423342 /* 'AB3B' */
1543 #define XFS_ABTC_MAGIC 0x41425443 /* 'ABTC' for cnt tree */
1544 #define XFS_ABTC_CRC_MAGIC 0x41423343 /* 'AB3C' */
1545
1546 /*
1547 * Data record/key structure
1548 */
1549 typedef struct xfs_alloc_rec {
1550 __be32 ar_startblock; /* starting block number */
1551 __be32 ar_blockcount; /* count of free blocks */
1552 } xfs_alloc_rec_t, xfs_alloc_key_t;
1553
1554 typedef struct xfs_alloc_rec_incore {
1555 xfs_agblock_t ar_startblock; /* starting block number */
1556 xfs_extlen_t ar_blockcount; /* count of free blocks */
1557 } xfs_alloc_rec_incore_t;
1558
1559 /* btree pointer type */
1560 typedef __be32 xfs_alloc_ptr_t;
1561
1562 /*
1563 * Block numbers in the AG:
1564 * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1565 */
1566 #define XFS_BNO_BLOCK(mp) ((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1567 #define XFS_CNT_BLOCK(mp) ((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1568
1569
1570 /*
1571 * Inode Allocation Btree format definitions
1572 *
1573 * There is a btree for the inode map per allocation group.
1574 */
1575 #define XFS_IBT_MAGIC 0x49414254 /* 'IABT' */
1576 #define XFS_IBT_CRC_MAGIC 0x49414233 /* 'IAB3' */
1577 #define XFS_FIBT_MAGIC 0x46494254 /* 'FIBT' */
1578 #define XFS_FIBT_CRC_MAGIC 0x46494233 /* 'FIB3' */
1579
1580 typedef uint64_t xfs_inofree_t;
1581 #define XFS_INODES_PER_CHUNK (NBBY * sizeof(xfs_inofree_t))
1582 #define XFS_INODES_PER_CHUNK_LOG (XFS_NBBYLOG + 3)
1583 #define XFS_INOBT_ALL_FREE ((xfs_inofree_t)-1)
1584 #define XFS_INOBT_MASK(i) ((xfs_inofree_t)1 << (i))
1585
1586 #define XFS_INOBT_HOLEMASK_FULL 0 /* holemask for full chunk */
1587 #define XFS_INOBT_HOLEMASK_BITS (NBBY * sizeof(uint16_t))
1588 #define XFS_INODES_PER_HOLEMASK_BIT \
1589 (XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1590
xfs_inobt_maskn(int i,int n)1591 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1592 {
1593 return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1594 }
1595
1596 /*
1597 * The on-disk inode record structure has two formats. The original "full"
1598 * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1599 * and replaces the 3 high-order freecount bytes wth the holemask and inode
1600 * count.
1601 *
1602 * The holemask of the sparse record format allows an inode chunk to have holes
1603 * that refer to blocks not owned by the inode record. This facilitates inode
1604 * allocation in the event of severe free space fragmentation.
1605 */
1606 typedef struct xfs_inobt_rec {
1607 __be32 ir_startino; /* starting inode number */
1608 union {
1609 struct {
1610 __be32 ir_freecount; /* count of free inodes */
1611 } f;
1612 struct {
1613 __be16 ir_holemask;/* hole mask for sparse chunks */
1614 __u8 ir_count; /* total inode count */
1615 __u8 ir_freecount; /* count of free inodes */
1616 } sp;
1617 } ir_u;
1618 __be64 ir_free; /* free inode mask */
1619 } xfs_inobt_rec_t;
1620
1621 typedef struct xfs_inobt_rec_incore {
1622 xfs_agino_t ir_startino; /* starting inode number */
1623 uint16_t ir_holemask; /* hole mask for sparse chunks */
1624 uint8_t ir_count; /* total inode count */
1625 uint8_t ir_freecount; /* count of free inodes (set bits) */
1626 xfs_inofree_t ir_free; /* free inode mask */
1627 } xfs_inobt_rec_incore_t;
1628
xfs_inobt_issparse(uint16_t holemask)1629 static inline bool xfs_inobt_issparse(uint16_t holemask)
1630 {
1631 /* non-zero holemask represents a sparse rec. */
1632 return holemask;
1633 }
1634
1635 /*
1636 * Key structure
1637 */
1638 typedef struct xfs_inobt_key {
1639 __be32 ir_startino; /* starting inode number */
1640 } xfs_inobt_key_t;
1641
1642 /* btree pointer type */
1643 typedef __be32 xfs_inobt_ptr_t;
1644
1645 /*
1646 * block numbers in the AG.
1647 */
1648 #define XFS_IBT_BLOCK(mp) ((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1649 #define XFS_FIBT_BLOCK(mp) ((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1650
1651 /*
1652 * Reverse mapping btree format definitions
1653 *
1654 * There is a btree for the reverse map per allocation group
1655 */
1656 #define XFS_RMAP_CRC_MAGIC 0x524d4233 /* 'RMB3' */
1657
1658 /*
1659 * Ownership info for an extent. This is used to create reverse-mapping
1660 * entries.
1661 */
1662 #define XFS_OWNER_INFO_ATTR_FORK (1 << 0)
1663 #define XFS_OWNER_INFO_BMBT_BLOCK (1 << 1)
1664 struct xfs_owner_info {
1665 uint64_t oi_owner;
1666 xfs_fileoff_t oi_offset;
1667 unsigned int oi_flags;
1668 };
1669
1670 /*
1671 * Special owner types.
1672 *
1673 * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1674 * to tell us we have a special owner value. We use these for static metadata
1675 * allocated at mkfs/growfs time, as well as for freespace management metadata.
1676 */
1677 #define XFS_RMAP_OWN_NULL (-1ULL) /* No owner, for growfs */
1678 #define XFS_RMAP_OWN_UNKNOWN (-2ULL) /* Unknown owner, for EFI recovery */
1679 #define XFS_RMAP_OWN_FS (-3ULL) /* static fs metadata */
1680 #define XFS_RMAP_OWN_LOG (-4ULL) /* static fs metadata */
1681 #define XFS_RMAP_OWN_AG (-5ULL) /* AG freespace btree blocks */
1682 #define XFS_RMAP_OWN_INOBT (-6ULL) /* Inode btree blocks */
1683 #define XFS_RMAP_OWN_INODES (-7ULL) /* Inode chunk */
1684 #define XFS_RMAP_OWN_REFC (-8ULL) /* refcount tree */
1685 #define XFS_RMAP_OWN_COW (-9ULL) /* cow allocations */
1686 #define XFS_RMAP_OWN_MIN (-10ULL) /* guard */
1687
1688 #define XFS_RMAP_NON_INODE_OWNER(owner) (!!((owner) & (1ULL << 63)))
1689
1690 /*
1691 * Data record structure
1692 */
1693 struct xfs_rmap_rec {
1694 __be32 rm_startblock; /* extent start block */
1695 __be32 rm_blockcount; /* extent length */
1696 __be64 rm_owner; /* extent owner */
1697 __be64 rm_offset; /* offset within the owner */
1698 };
1699
1700 /*
1701 * rmap btree record
1702 * rm_offset:63 is the attribute fork flag
1703 * rm_offset:62 is the bmbt block flag
1704 * rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1705 * rm_offset:54-60 aren't used and should be zero
1706 * rm_offset:0-53 is the block offset within the inode
1707 */
1708 #define XFS_RMAP_OFF_ATTR_FORK ((uint64_t)1ULL << 63)
1709 #define XFS_RMAP_OFF_BMBT_BLOCK ((uint64_t)1ULL << 62)
1710 #define XFS_RMAP_OFF_UNWRITTEN ((uint64_t)1ULL << 61)
1711
1712 #define XFS_RMAP_LEN_MAX ((uint32_t)~0U)
1713 #define XFS_RMAP_OFF_FLAGS (XFS_RMAP_OFF_ATTR_FORK | \
1714 XFS_RMAP_OFF_BMBT_BLOCK | \
1715 XFS_RMAP_OFF_UNWRITTEN)
1716 #define XFS_RMAP_OFF_MASK ((uint64_t)0x3FFFFFFFFFFFFFULL)
1717
1718 #define XFS_RMAP_OFF(off) ((off) & XFS_RMAP_OFF_MASK)
1719
1720 #define XFS_RMAP_IS_BMBT_BLOCK(off) (!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1721 #define XFS_RMAP_IS_ATTR_FORK(off) (!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1722 #define XFS_RMAP_IS_UNWRITTEN(len) (!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1723
1724 #define RMAPBT_STARTBLOCK_BITLEN 32
1725 #define RMAPBT_BLOCKCOUNT_BITLEN 32
1726 #define RMAPBT_OWNER_BITLEN 64
1727 #define RMAPBT_ATTRFLAG_BITLEN 1
1728 #define RMAPBT_BMBTFLAG_BITLEN 1
1729 #define RMAPBT_EXNTFLAG_BITLEN 1
1730 #define RMAPBT_UNUSED_OFFSET_BITLEN 7
1731 #define RMAPBT_OFFSET_BITLEN 54
1732
1733 /*
1734 * Key structure
1735 *
1736 * We don't use the length for lookups
1737 */
1738 struct xfs_rmap_key {
1739 __be32 rm_startblock; /* extent start block */
1740 __be64 rm_owner; /* extent owner */
1741 __be64 rm_offset; /* offset within the owner */
1742 } __attribute__((packed));
1743
1744 /* btree pointer type */
1745 typedef __be32 xfs_rmap_ptr_t;
1746
1747 #define XFS_RMAP_BLOCK(mp) \
1748 (xfs_has_finobt(((mp))) ? \
1749 XFS_FIBT_BLOCK(mp) + 1 : \
1750 XFS_IBT_BLOCK(mp) + 1)
1751
1752 /*
1753 * Realtime Reverse mapping btree format definitions
1754 *
1755 * This is a btree for reverse mapping records for realtime volumes
1756 */
1757 #define XFS_RTRMAP_CRC_MAGIC 0x4d415052 /* 'MAPR' */
1758
1759 /*
1760 * rtrmap root header, on-disk form only.
1761 */
1762 struct xfs_rtrmap_root {
1763 __be16 bb_level; /* 0 is a leaf */
1764 __be16 bb_numrecs; /* current # of data records */
1765 };
1766
1767 /* inode-based btree pointer type */
1768 typedef __be64 xfs_rtrmap_ptr_t;
1769
1770 /*
1771 * Reference Count Btree format definitions
1772 *
1773 */
1774 #define XFS_REFC_CRC_MAGIC 0x52334643 /* 'R3FC' */
1775
1776 unsigned int xfs_refc_block(struct xfs_mount *mp);
1777
1778 /*
1779 * Data record/key structure
1780 *
1781 * Each record associates a range of physical blocks (starting at
1782 * rc_startblock and ending rc_blockcount blocks later) with a reference
1783 * count (rc_refcount). Extents that are being used to stage a copy on
1784 * write (CoW) operation are recorded in the refcount btree with a
1785 * refcount of 1. All other records must have a refcount > 1 and must
1786 * track an extent mapped only by file data forks.
1787 *
1788 * Extents with a single owner (attributes, metadata, non-shared file
1789 * data) are not tracked here. Free space is also not tracked here.
1790 * This is consistent with pre-reflink XFS.
1791 */
1792
1793 /*
1794 * Extents that are being used to stage a copy on write are stored
1795 * in the refcount btree with a refcount of 1 and the upper bit set
1796 * on the startblock. This speeds up mount time deletion of stale
1797 * staging extents because they're all at the right side of the tree.
1798 */
1799 #define XFS_REFC_COWFLAG (1U << 31)
1800 #define REFCNTBT_COWFLAG_BITLEN 1
1801 #define REFCNTBT_AGBLOCK_BITLEN 31
1802
1803 struct xfs_refcount_rec {
1804 __be32 rc_startblock; /* starting block number */
1805 __be32 rc_blockcount; /* count of blocks */
1806 __be32 rc_refcount; /* number of inodes linked here */
1807 };
1808
1809 struct xfs_refcount_key {
1810 __be32 rc_startblock; /* starting block number */
1811 };
1812
1813 #define XFS_REFC_REFCOUNT_MAX ((xfs_nlink_t)~0U)
1814 #define XFS_REFC_LEN_MAX ((xfs_extlen_t)~0U)
1815
1816 /* btree pointer type */
1817 typedef __be32 xfs_refcount_ptr_t;
1818
1819 /*
1820 * Realtime Reference Count btree format definitions
1821 *
1822 * This is a btree for reference count records for realtime volumes
1823 */
1824 #define XFS_RTREFC_CRC_MAGIC 0x52434e54 /* 'RCNT' */
1825
1826 /*
1827 * rt refcount root header, on-disk form only.
1828 */
1829 struct xfs_rtrefcount_root {
1830 __be16 bb_level; /* 0 is a leaf */
1831 __be16 bb_numrecs; /* current # of data records */
1832 };
1833
1834 /* inode-rooted btree pointer type */
1835 typedef __be64 xfs_rtrefcount_ptr_t;
1836
1837 /*
1838 * BMAP Btree format definitions
1839 *
1840 * This includes both the root block definition that sits inside an inode fork
1841 * and the record/pointer formats for the leaf/node in the blocks.
1842 */
1843 #define XFS_BMAP_MAGIC 0x424d4150 /* 'BMAP' */
1844 #define XFS_BMAP_CRC_MAGIC 0x424d4133 /* 'BMA3' */
1845
1846 /*
1847 * Bmap root header, on-disk form only.
1848 */
1849 typedef struct xfs_bmdr_block {
1850 __be16 bb_level; /* 0 is a leaf */
1851 __be16 bb_numrecs; /* current # of data records */
1852 } xfs_bmdr_block_t;
1853
1854 /*
1855 * Bmap btree record and extent descriptor.
1856 * l0:63 is an extent flag (value 1 indicates non-normal).
1857 * l0:9-62 are startoff.
1858 * l0:0-8 and l1:21-63 are startblock.
1859 * l1:0-20 are blockcount.
1860 */
1861 #define BMBT_EXNTFLAG_BITLEN 1
1862 #define BMBT_STARTOFF_BITLEN 54
1863 #define BMBT_STARTBLOCK_BITLEN 52
1864 #define BMBT_BLOCKCOUNT_BITLEN 21
1865
1866 #define BMBT_STARTOFF_MASK ((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1867 #define BMBT_BLOCKCOUNT_MASK ((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1)
1868
1869 #define XFS_MAX_BMBT_EXTLEN ((xfs_extlen_t)(BMBT_BLOCKCOUNT_MASK))
1870
1871 /*
1872 * bmbt records have a file offset (block) field that is 54 bits wide, so this
1873 * is the largest xfs_fileoff_t that we ever expect to see.
1874 */
1875 #define XFS_MAX_FILEOFF (BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK)
1876
1877 typedef struct xfs_bmbt_rec {
1878 __be64 l0, l1;
1879 } xfs_bmbt_rec_t;
1880
1881 typedef uint64_t xfs_bmbt_rec_base_t; /* use this for casts */
1882 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1883
1884 /*
1885 * Values and macros for delayed-allocation startblock fields.
1886 */
1887 #define STARTBLOCKVALBITS 17
1888 #define STARTBLOCKMASKBITS (15 + 20)
1889 #define STARTBLOCKMASK \
1890 (((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1891
isnullstartblock(xfs_fsblock_t x)1892 static inline int isnullstartblock(xfs_fsblock_t x)
1893 {
1894 return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1895 }
1896
nullstartblock(int k)1897 static inline xfs_fsblock_t nullstartblock(int k)
1898 {
1899 ASSERT(k < (1 << STARTBLOCKVALBITS));
1900 return STARTBLOCKMASK | (k);
1901 }
1902
startblockval(xfs_fsblock_t x)1903 static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1904 {
1905 return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1906 }
1907
1908 /*
1909 * Key structure for non-leaf levels of the tree.
1910 */
1911 typedef struct xfs_bmbt_key {
1912 __be64 br_startoff; /* starting file offset */
1913 } xfs_bmbt_key_t, xfs_bmdr_key_t;
1914
1915 /* btree pointer type */
1916 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1917
1918
1919 /*
1920 * Generic Btree block format definitions
1921 *
1922 * This is a combination of the actual format used on disk for short and long
1923 * format btrees. The first three fields are shared by both format, but the
1924 * pointers are different and should be used with care.
1925 *
1926 * To get the size of the actual short or long form headers please use the size
1927 * macros below. Never use sizeof(xfs_btree_block).
1928 *
1929 * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1930 * with the crc feature bit, and all accesses to them must be conditional on
1931 * that flag.
1932 */
1933 /* short form block header */
1934 struct xfs_btree_block_shdr {
1935 __be32 bb_leftsib;
1936 __be32 bb_rightsib;
1937
1938 __be64 bb_blkno;
1939 __be64 bb_lsn;
1940 uuid_t bb_uuid;
1941 __be32 bb_owner;
1942 __le32 bb_crc;
1943 };
1944
1945 /* long form block header */
1946 struct xfs_btree_block_lhdr {
1947 __be64 bb_leftsib;
1948 __be64 bb_rightsib;
1949
1950 __be64 bb_blkno;
1951 __be64 bb_lsn;
1952 uuid_t bb_uuid;
1953 __be64 bb_owner;
1954 __le32 bb_crc;
1955 __be32 bb_pad; /* padding for alignment */
1956 };
1957
1958 struct xfs_btree_block {
1959 __be32 bb_magic; /* magic number for block type */
1960 __be16 bb_level; /* 0 is a leaf */
1961 __be16 bb_numrecs; /* current # of data records */
1962 union {
1963 struct xfs_btree_block_shdr s;
1964 struct xfs_btree_block_lhdr l;
1965 } bb_u; /* rest */
1966 };
1967
1968 /* size of a short form block */
1969 #define XFS_BTREE_SBLOCK_LEN \
1970 (offsetof(struct xfs_btree_block, bb_u) + \
1971 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1972 /* size of a long form block */
1973 #define XFS_BTREE_LBLOCK_LEN \
1974 (offsetof(struct xfs_btree_block, bb_u) + \
1975 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1976
1977 /* sizes of CRC enabled btree blocks */
1978 #define XFS_BTREE_SBLOCK_CRC_LEN \
1979 (offsetof(struct xfs_btree_block, bb_u) + \
1980 sizeof(struct xfs_btree_block_shdr))
1981 #define XFS_BTREE_LBLOCK_CRC_LEN \
1982 (offsetof(struct xfs_btree_block, bb_u) + \
1983 sizeof(struct xfs_btree_block_lhdr))
1984
1985 #define XFS_BTREE_SBLOCK_CRC_OFF \
1986 offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1987 #define XFS_BTREE_LBLOCK_CRC_OFF \
1988 offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1989
1990 /*
1991 * On-disk XFS access control list structure.
1992 */
1993 struct xfs_acl_entry {
1994 __be32 ae_tag;
1995 __be32 ae_id;
1996 __be16 ae_perm;
1997 __be16 ae_pad; /* fill the implicit hole in the structure */
1998 };
1999
2000 struct xfs_acl {
2001 __be32 acl_cnt;
2002 struct xfs_acl_entry acl_entry[];
2003 };
2004
2005 /*
2006 * The number of ACL entries allowed is defined by the on-disk format.
2007 * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
2008 * limited only by the maximum size of the xattr that stores the information.
2009 */
2010 #define XFS_ACL_MAX_ENTRIES(mp) \
2011 (xfs_has_crc(mp) \
2012 ? (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
2013 sizeof(struct xfs_acl_entry) \
2014 : 25)
2015
2016 #define XFS_ACL_SIZE(cnt) \
2017 (sizeof(struct xfs_acl) + \
2018 sizeof(struct xfs_acl_entry) * cnt)
2019
2020 #define XFS_ACL_MAX_SIZE(mp) \
2021 XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
2022
2023
2024 /* On-disk XFS extended attribute names */
2025 #define SGI_ACL_FILE "SGI_ACL_FILE"
2026 #define SGI_ACL_DEFAULT "SGI_ACL_DEFAULT"
2027 #define SGI_ACL_FILE_SIZE (sizeof(SGI_ACL_FILE)-1)
2028 #define SGI_ACL_DEFAULT_SIZE (sizeof(SGI_ACL_DEFAULT)-1)
2029
2030 #endif /* __XFS_FORMAT_H__ */
2031