xref: /linux/fs/xfs/libxfs/xfs_format.h (revision 4f77c3462308c62ffe7129cc18b9ac937f44b5a5)
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]; /* 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 	/* must be padded to 64 bit alignment */
178 } xfs_sb_t;
179 
180 /*
181  * Superblock - on disk version.
182  * Must be padded to 64 bit alignment.
183  */
184 struct xfs_dsb {
185 	__be32		sb_magicnum;	/* magic number == XFS_SB_MAGIC */
186 	__be32		sb_blocksize;	/* logical block size, bytes */
187 	__be64		sb_dblocks;	/* number of data blocks */
188 	__be64		sb_rblocks;	/* number of realtime blocks */
189 	__be64		sb_rextents;	/* number of realtime extents */
190 	uuid_t		sb_uuid;	/* user-visible file system unique id */
191 	__be64		sb_logstart;	/* starting block of log if internal */
192 	__be64		sb_rootino;	/* root inode number */
193 	__be64		sb_rbmino;	/* bitmap inode for realtime extents */
194 	__be64		sb_rsumino;	/* summary inode for rt bitmap */
195 	__be32		sb_rextsize;	/* realtime extent size, blocks */
196 	__be32		sb_agblocks;	/* size of an allocation group */
197 	__be32		sb_agcount;	/* number of allocation groups */
198 	__be32		sb_rbmblocks;	/* number of rt bitmap blocks */
199 	__be32		sb_logblocks;	/* number of log blocks */
200 	__be16		sb_versionnum;	/* header version == XFS_SB_VERSION */
201 	__be16		sb_sectsize;	/* volume sector size, bytes */
202 	__be16		sb_inodesize;	/* inode size, bytes */
203 	__be16		sb_inopblock;	/* inodes per block */
204 	char		sb_fname[XFSLABEL_MAX]; /* file system name */
205 	__u8		sb_blocklog;	/* log2 of sb_blocksize */
206 	__u8		sb_sectlog;	/* log2 of sb_sectsize */
207 	__u8		sb_inodelog;	/* log2 of sb_inodesize */
208 	__u8		sb_inopblog;	/* log2 of sb_inopblock */
209 	__u8		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
210 	__u8		sb_rextslog;	/* log2 of sb_rextents */
211 	__u8		sb_inprogress;	/* mkfs is in progress, don't mount */
212 	__u8		sb_imax_pct;	/* max % of fs for inode space */
213 					/* statistics */
214 	/*
215 	 * These fields must remain contiguous.  If you really
216 	 * want to change their layout, make sure you fix the
217 	 * code in xfs_trans_apply_sb_deltas().
218 	 */
219 	__be64		sb_icount;	/* allocated inodes */
220 	__be64		sb_ifree;	/* free inodes */
221 	__be64		sb_fdblocks;	/* free data blocks */
222 	__be64		sb_frextents;	/* free realtime extents */
223 	/*
224 	 * End contiguous fields.
225 	 */
226 	__be64		sb_uquotino;	/* user quota inode */
227 	__be64		sb_gquotino;	/* group quota inode */
228 	__be16		sb_qflags;	/* quota flags */
229 	__u8		sb_flags;	/* misc. flags */
230 	__u8		sb_shared_vn;	/* shared version number */
231 	__be32		sb_inoalignmt;	/* inode chunk alignment, fsblocks */
232 	__be32		sb_unit;	/* stripe or raid unit */
233 	__be32		sb_width;	/* stripe or raid width */
234 	__u8		sb_dirblklog;	/* log2 of dir block size (fsbs) */
235 	__u8		sb_logsectlog;	/* log2 of the log sector size */
236 	__be16		sb_logsectsize;	/* sector size for the log, bytes */
237 	__be32		sb_logsunit;	/* stripe unit size for the log */
238 	__be32		sb_features2;	/* additional feature bits */
239 	/*
240 	 * bad features2 field as a result of failing to pad the sb
241 	 * structure to 64 bits. Some machines will be using this field
242 	 * for features2 bits. Easiest just to mark it bad and not use
243 	 * it for anything else.
244 	 */
245 	__be32		sb_bad_features2;
246 
247 	/* version 5 superblock fields start here */
248 
249 	/* feature masks */
250 	__be32		sb_features_compat;
251 	__be32		sb_features_ro_compat;
252 	__be32		sb_features_incompat;
253 	__be32		sb_features_log_incompat;
254 
255 	__le32		sb_crc;		/* superblock crc */
256 	__be32		sb_spino_align;	/* sparse inode chunk alignment */
257 
258 	__be64		sb_pquotino;	/* project quota inode */
259 	__be64		sb_lsn;		/* last write sequence */
260 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
261 
262 	/* must be padded to 64 bit alignment */
263 };
264 
265 #define XFS_SB_CRC_OFF		offsetof(struct xfs_dsb, sb_crc)
266 
267 /*
268  * Misc. Flags - warning - these will be cleared by xfs_repair unless
269  * a feature bit is set when the flag is used.
270  */
271 #define XFS_SBF_NOFLAGS		0x00	/* no flags set */
272 #define XFS_SBF_READONLY	0x01	/* only read-only mounts allowed */
273 
274 /*
275  * define max. shared version we can interoperate with
276  */
277 #define XFS_SB_MAX_SHARED_VN	0
278 
279 #define	XFS_SB_VERSION_NUM(sbp)	((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
280 
281 static inline bool xfs_sb_is_v5(struct xfs_sb *sbp)
282 {
283 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
284 }
285 
286 /*
287  * Detect a mismatched features2 field.  Older kernels read/wrote
288  * this into the wrong slot, so to be safe we keep them in sync.
289  */
290 static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
291 {
292 	return sbp->sb_bad_features2 != sbp->sb_features2;
293 }
294 
295 static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp)
296 {
297 	return xfs_sb_is_v5(sbp) ||
298 	       (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
299 }
300 
301 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
302 {
303 	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
304 }
305 
306 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
307 {
308 	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
309 }
310 
311 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
312 {
313 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
314 	sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
315 }
316 
317 static inline void xfs_sb_version_addprojid32(struct xfs_sb *sbp)
318 {
319 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
320 	sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
321 }
322 
323 /*
324  * Extended v5 superblock feature masks. These are to be used for new v5
325  * superblock features only.
326  *
327  * Compat features are new features that old kernels will not notice or affect
328  * and so can mount read-write without issues.
329  *
330  * RO-Compat (read only) are features that old kernels can read but will break
331  * if they write. Hence only read-only mounts of such filesystems are allowed on
332  * kernels that don't support the feature bit.
333  *
334  * InCompat features are features which old kernels will not understand and so
335  * must not mount.
336  *
337  * Log-InCompat features are for changes to log formats or new transactions that
338  * can't be replayed on older kernels. The fields are set when the filesystem is
339  * mounted, and a clean unmount clears the fields.
340  */
341 #define XFS_SB_FEAT_COMPAT_ALL 0
342 #define XFS_SB_FEAT_COMPAT_UNKNOWN	~XFS_SB_FEAT_COMPAT_ALL
343 static inline bool
344 xfs_sb_has_compat_feature(
345 	struct xfs_sb	*sbp,
346 	uint32_t	feature)
347 {
348 	return (sbp->sb_features_compat & feature) != 0;
349 }
350 
351 #define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
352 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
353 #define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
354 #define XFS_SB_FEAT_RO_COMPAT_INOBTCNT (1 << 3)		/* inobt block counts */
355 #define XFS_SB_FEAT_RO_COMPAT_ALL \
356 		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
357 		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
358 		 XFS_SB_FEAT_RO_COMPAT_REFLINK| \
359 		 XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
360 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
361 static inline bool
362 xfs_sb_has_ro_compat_feature(
363 	struct xfs_sb	*sbp,
364 	uint32_t	feature)
365 {
366 	return (sbp->sb_features_ro_compat & feature) != 0;
367 }
368 
369 #define XFS_SB_FEAT_INCOMPAT_FTYPE	(1 << 0)  /* filetype in dirent */
370 #define XFS_SB_FEAT_INCOMPAT_SPINODES	(1 << 1)  /* sparse inode chunks */
371 #define XFS_SB_FEAT_INCOMPAT_META_UUID	(1 << 2)  /* metadata UUID */
372 #define XFS_SB_FEAT_INCOMPAT_BIGTIME	(1 << 3)  /* large timestamps */
373 #define XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR (1 << 4) /* needs xfs_repair */
374 #define XFS_SB_FEAT_INCOMPAT_NREXT64	(1 << 5)  /* large extent counters */
375 #define XFS_SB_FEAT_INCOMPAT_EXCHRANGE	(1 << 6)  /* exchangerange supported */
376 #define XFS_SB_FEAT_INCOMPAT_PARENT	(1 << 7)  /* parent pointers */
377 #define XFS_SB_FEAT_INCOMPAT_ALL \
378 		(XFS_SB_FEAT_INCOMPAT_FTYPE | \
379 		 XFS_SB_FEAT_INCOMPAT_SPINODES | \
380 		 XFS_SB_FEAT_INCOMPAT_META_UUID | \
381 		 XFS_SB_FEAT_INCOMPAT_BIGTIME | \
382 		 XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR | \
383 		 XFS_SB_FEAT_INCOMPAT_NREXT64 | \
384 		 XFS_SB_FEAT_INCOMPAT_EXCHRANGE | \
385 		 XFS_SB_FEAT_INCOMPAT_PARENT)
386 
387 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
388 static inline bool
389 xfs_sb_has_incompat_feature(
390 	struct xfs_sb	*sbp,
391 	uint32_t	feature)
392 {
393 	return (sbp->sb_features_incompat & feature) != 0;
394 }
395 
396 #define XFS_SB_FEAT_INCOMPAT_LOG_XATTRS   (1 << 0)	/* Delayed Attributes */
397 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL \
398 	(XFS_SB_FEAT_INCOMPAT_LOG_XATTRS)
399 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
400 static inline bool
401 xfs_sb_has_incompat_log_feature(
402 	struct xfs_sb	*sbp,
403 	uint32_t	feature)
404 {
405 	return (sbp->sb_features_log_incompat & feature) != 0;
406 }
407 
408 static inline void
409 xfs_sb_remove_incompat_log_features(
410 	struct xfs_sb	*sbp)
411 {
412 	sbp->sb_features_log_incompat &= ~XFS_SB_FEAT_INCOMPAT_LOG_ALL;
413 }
414 
415 static inline void
416 xfs_sb_add_incompat_log_features(
417 	struct xfs_sb	*sbp,
418 	unsigned int	features)
419 {
420 	sbp->sb_features_log_incompat |= features;
421 }
422 
423 static inline bool xfs_sb_version_haslogxattrs(struct xfs_sb *sbp)
424 {
425 	return xfs_sb_is_v5(sbp) && (sbp->sb_features_log_incompat &
426 		 XFS_SB_FEAT_INCOMPAT_LOG_XATTRS);
427 }
428 
429 static inline bool
430 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
431 {
432 	return (ino == sbp->sb_uquotino ||
433 		ino == sbp->sb_gquotino ||
434 		ino == sbp->sb_pquotino);
435 }
436 
437 #define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
438 #define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
439 
440 #define	XFS_HDR_BLOCK(mp,d)	((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
441 #define	XFS_DADDR_TO_FSB(mp,d)	XFS_AGB_TO_FSB(mp, \
442 			xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
443 #define	XFS_FSB_TO_DADDR(mp,fsbno)	XFS_AGB_TO_DADDR(mp, \
444 			XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
445 
446 /*
447  * File system sector to basic block conversions.
448  */
449 #define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
450 
451 /*
452  * File system block to basic block conversions.
453  */
454 #define	XFS_FSB_TO_BB(mp,fsbno)	((fsbno) << (mp)->m_blkbb_log)
455 #define	XFS_BB_TO_FSB(mp,bb)	\
456 	(((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
457 #define	XFS_BB_TO_FSBT(mp,bb)	((bb) >> (mp)->m_blkbb_log)
458 
459 /*
460  * File system block to byte conversions.
461  */
462 #define XFS_FSB_TO_B(mp,fsbno)	((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
463 #define XFS_B_TO_FSB(mp,b)	\
464 	((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
465 #define XFS_B_TO_FSBT(mp,b)	(((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
466 
467 /*
468  * Allocation group header
469  *
470  * This is divided into three structures, placed in sequential 512-byte
471  * buffers after a copy of the superblock (also in a 512-byte buffer).
472  */
473 #define	XFS_AGF_MAGIC	0x58414746	/* 'XAGF' */
474 #define	XFS_AGI_MAGIC	0x58414749	/* 'XAGI' */
475 #define	XFS_AGFL_MAGIC	0x5841464c	/* 'XAFL' */
476 #define	XFS_AGF_VERSION	1
477 #define	XFS_AGI_VERSION	1
478 
479 #define	XFS_AGF_GOOD_VERSION(v)	((v) == XFS_AGF_VERSION)
480 #define	XFS_AGI_GOOD_VERSION(v)	((v) == XFS_AGI_VERSION)
481 
482 /*
483  * agf_cnt_level in the first AGF overlaps the EFS superblock's magic number.
484  * Since the magic numbers valid for EFS are > 64k, our value cannot be confused
485  * for an EFS superblock.
486  */
487 
488 typedef struct xfs_agf {
489 	/*
490 	 * Common allocation group header information
491 	 */
492 	__be32		agf_magicnum;	/* magic number == XFS_AGF_MAGIC */
493 	__be32		agf_versionnum;	/* header version == XFS_AGF_VERSION */
494 	__be32		agf_seqno;	/* sequence # starting from 0 */
495 	__be32		agf_length;	/* size in blocks of a.g. */
496 	/*
497 	 * Freespace and rmap information
498 	 */
499 	__be32		agf_bno_root;	/* bnobt root block */
500 	__be32		agf_cnt_root;	/* cntbt root block */
501 	__be32		agf_rmap_root;	/* rmapbt root block */
502 
503 	__be32		agf_bno_level;	/* bnobt btree levels */
504 	__be32		agf_cnt_level;	/* cntbt btree levels */
505 	__be32		agf_rmap_level;	/* rmapbt btree levels */
506 
507 	__be32		agf_flfirst;	/* first freelist block's index */
508 	__be32		agf_fllast;	/* last freelist block's index */
509 	__be32		agf_flcount;	/* count of blocks in freelist */
510 	__be32		agf_freeblks;	/* total free blocks */
511 
512 	__be32		agf_longest;	/* longest free space */
513 	__be32		agf_btreeblks;	/* # of blocks held in AGF btrees */
514 	uuid_t		agf_uuid;	/* uuid of filesystem */
515 
516 	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
517 	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
518 
519 	__be32		agf_refcount_root;	/* refcount tree root block */
520 	__be32		agf_refcount_level;	/* refcount btree levels */
521 
522 	/*
523 	 * reserve some contiguous space for future logged fields before we add
524 	 * the unlogged fields. This makes the range logging via flags and
525 	 * structure offsets much simpler.
526 	 */
527 	__be64		agf_spare64[14];
528 
529 	/* unlogged fields, written during buffer writeback. */
530 	__be64		agf_lsn;	/* last write sequence */
531 	__be32		agf_crc;	/* crc of agf sector */
532 	__be32		agf_spare2;
533 
534 	/* structure must be padded to 64 bit alignment */
535 } xfs_agf_t;
536 
537 #define XFS_AGF_CRC_OFF		offsetof(struct xfs_agf, agf_crc)
538 
539 #define	XFS_AGF_MAGICNUM	(1u << 0)
540 #define	XFS_AGF_VERSIONNUM	(1u << 1)
541 #define	XFS_AGF_SEQNO		(1u << 2)
542 #define	XFS_AGF_LENGTH		(1u << 3)
543 #define	XFS_AGF_ROOTS		(1u << 4)
544 #define	XFS_AGF_LEVELS		(1u << 5)
545 #define	XFS_AGF_FLFIRST		(1u << 6)
546 #define	XFS_AGF_FLLAST		(1u << 7)
547 #define	XFS_AGF_FLCOUNT		(1u << 8)
548 #define	XFS_AGF_FREEBLKS	(1u << 9)
549 #define	XFS_AGF_LONGEST		(1u << 10)
550 #define	XFS_AGF_BTREEBLKS	(1u << 11)
551 #define	XFS_AGF_UUID		(1u << 12)
552 #define	XFS_AGF_RMAP_BLOCKS	(1u << 13)
553 #define	XFS_AGF_REFCOUNT_BLOCKS	(1u << 14)
554 #define	XFS_AGF_REFCOUNT_ROOT	(1u << 15)
555 #define	XFS_AGF_REFCOUNT_LEVEL	(1u << 16)
556 #define	XFS_AGF_SPARE64		(1u << 17)
557 #define	XFS_AGF_NUM_BITS	18
558 #define	XFS_AGF_ALL_BITS	((1u << XFS_AGF_NUM_BITS) - 1)
559 
560 #define XFS_AGF_FLAGS \
561 	{ XFS_AGF_MAGICNUM,	"MAGICNUM" }, \
562 	{ XFS_AGF_VERSIONNUM,	"VERSIONNUM" }, \
563 	{ XFS_AGF_SEQNO,	"SEQNO" }, \
564 	{ XFS_AGF_LENGTH,	"LENGTH" }, \
565 	{ XFS_AGF_ROOTS,	"ROOTS" }, \
566 	{ XFS_AGF_LEVELS,	"LEVELS" }, \
567 	{ XFS_AGF_FLFIRST,	"FLFIRST" }, \
568 	{ XFS_AGF_FLLAST,	"FLLAST" }, \
569 	{ XFS_AGF_FLCOUNT,	"FLCOUNT" }, \
570 	{ XFS_AGF_FREEBLKS,	"FREEBLKS" }, \
571 	{ XFS_AGF_LONGEST,	"LONGEST" }, \
572 	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
573 	{ XFS_AGF_UUID,		"UUID" }, \
574 	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
575 	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
576 	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
577 	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
578 	{ XFS_AGF_SPARE64,	"SPARE64" }
579 
580 /* disk block (xfs_daddr_t) in the AG */
581 #define XFS_AGF_DADDR(mp)	((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
582 #define	XFS_AGF_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
583 
584 /*
585  * Size of the unlinked inode hash table in the agi.
586  */
587 #define	XFS_AGI_UNLINKED_BUCKETS	64
588 
589 typedef struct xfs_agi {
590 	/*
591 	 * Common allocation group header information
592 	 */
593 	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
594 	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
595 	__be32		agi_seqno;	/* sequence # starting from 0 */
596 	__be32		agi_length;	/* size in blocks of a.g. */
597 	/*
598 	 * Inode information
599 	 * Inodes are mapped by interpreting the inode number, so no
600 	 * mapping data is needed here.
601 	 */
602 	__be32		agi_count;	/* count of allocated inodes */
603 	__be32		agi_root;	/* root of inode btree */
604 	__be32		agi_level;	/* levels in inode btree */
605 	__be32		agi_freecount;	/* number of free inodes */
606 
607 	__be32		agi_newino;	/* new inode just allocated */
608 	__be32		agi_dirino;	/* last directory inode chunk */
609 	/*
610 	 * Hash table of inodes which have been unlinked but are
611 	 * still being referenced.
612 	 */
613 	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
614 	/*
615 	 * This marks the end of logging region 1 and start of logging region 2.
616 	 */
617 	uuid_t		agi_uuid;	/* uuid of filesystem */
618 	__be32		agi_crc;	/* crc of agi sector */
619 	__be32		agi_pad32;
620 	__be64		agi_lsn;	/* last write sequence */
621 
622 	__be32		agi_free_root; /* root of the free inode btree */
623 	__be32		agi_free_level;/* levels in free inode btree */
624 
625 	__be32		agi_iblocks;	/* inobt blocks used */
626 	__be32		agi_fblocks;	/* finobt blocks used */
627 
628 	/* structure must be padded to 64 bit alignment */
629 } xfs_agi_t;
630 
631 #define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
632 
633 #define	XFS_AGI_MAGICNUM	(1u << 0)
634 #define	XFS_AGI_VERSIONNUM	(1u << 1)
635 #define	XFS_AGI_SEQNO		(1u << 2)
636 #define	XFS_AGI_LENGTH		(1u << 3)
637 #define	XFS_AGI_COUNT		(1u << 4)
638 #define	XFS_AGI_ROOT		(1u << 5)
639 #define	XFS_AGI_LEVEL		(1u << 6)
640 #define	XFS_AGI_FREECOUNT	(1u << 7)
641 #define	XFS_AGI_NEWINO		(1u << 8)
642 #define	XFS_AGI_DIRINO		(1u << 9)
643 #define	XFS_AGI_UNLINKED	(1u << 10)
644 #define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
645 #define	XFS_AGI_ALL_BITS_R1	((1u << XFS_AGI_NUM_BITS_R1) - 1)
646 #define	XFS_AGI_FREE_ROOT	(1u << 11)
647 #define	XFS_AGI_FREE_LEVEL	(1u << 12)
648 #define	XFS_AGI_IBLOCKS		(1u << 13) /* both inobt/finobt block counters */
649 #define	XFS_AGI_NUM_BITS_R2	14
650 
651 /* disk block (xfs_daddr_t) in the AG */
652 #define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
653 #define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
654 
655 /*
656  * The third a.g. block contains the a.g. freelist, an array
657  * of block pointers to blocks owned by the allocation btree code.
658  */
659 #define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
660 #define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
661 #define	XFS_BUF_TO_AGFL(bp)	((struct xfs_agfl *)((bp)->b_addr))
662 
663 struct xfs_agfl {
664 	__be32		agfl_magicnum;
665 	__be32		agfl_seqno;
666 	uuid_t		agfl_uuid;
667 	__be64		agfl_lsn;
668 	__be32		agfl_crc;
669 } __attribute__((packed));
670 
671 #define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
672 
673 #define XFS_AGB_TO_FSB(mp,agno,agbno)	\
674 	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
675 #define	XFS_FSB_TO_AGNO(mp,fsbno)	\
676 	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
677 #define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
678 	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
679 #define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
680 	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
681 		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
682 #define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
683 
684 /*
685  * For checking for bad ranges of xfs_daddr_t's, covering multiple
686  * allocation groups or a single xfs_daddr_t that's a superblock copy.
687  */
688 #define	XFS_AG_CHECK_DADDR(mp,d,len)	\
689 	((len) == 1 ? \
690 	    ASSERT((d) == XFS_SB_DADDR || \
691 		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
692 	    ASSERT(xfs_daddr_to_agno(mp, d) == \
693 		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
694 
695 /*
696  * Realtime bitmap information is accessed by the word, which is currently
697  * stored in host-endian format.
698  */
699 union xfs_rtword_raw {
700 	__u32		old;
701 };
702 
703 /*
704  * Realtime summary counts are accessed by the word, which is currently
705  * stored in host-endian format.
706  */
707 union xfs_suminfo_raw {
708 	__u32		old;
709 };
710 
711 /*
712  * XFS Timestamps
713  * ==============
714  *
715  * Traditional ondisk inode timestamps consist of signed 32-bit counters for
716  * seconds and nanoseconds; time zero is the Unix epoch, Jan  1 00:00:00 UTC
717  * 1970, which means that the timestamp epoch is the same as the Unix epoch.
718  * Therefore, the ondisk min and max defined here can be used directly to
719  * constrain the incore timestamps on a Unix system.  Note that we actually
720  * encode a __be64 value on disk.
721  *
722  * When the bigtime feature is enabled, ondisk inode timestamps become an
723  * unsigned 64-bit nanoseconds counter.  This means that the bigtime inode
724  * timestamp epoch is the start of the classic timestamp range, which is
725  * Dec 13 20:45:52 UTC 1901.  Because the epochs are not the same, callers
726  * /must/ use the bigtime conversion functions when encoding and decoding raw
727  * timestamps.
728  */
729 typedef __be64 xfs_timestamp_t;
730 
731 /* Legacy timestamp encoding format. */
732 struct xfs_legacy_timestamp {
733 	__be32		t_sec;		/* timestamp seconds */
734 	__be32		t_nsec;		/* timestamp nanoseconds */
735 };
736 
737 /*
738  * Smallest possible ondisk seconds value with traditional timestamps.  This
739  * corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901.
740  */
741 #define XFS_LEGACY_TIME_MIN	((int64_t)S32_MIN)
742 
743 /*
744  * Largest possible ondisk seconds value with traditional timestamps.  This
745  * corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038.
746  */
747 #define XFS_LEGACY_TIME_MAX	((int64_t)S32_MAX)
748 
749 /*
750  * Smallest possible ondisk seconds value with bigtime timestamps.  This
751  * corresponds (after conversion to a Unix timestamp) with the traditional
752  * minimum timestamp of Dec 13 20:45:52 UTC 1901.
753  */
754 #define XFS_BIGTIME_TIME_MIN	((int64_t)0)
755 
756 /*
757  * Largest supported ondisk seconds value with bigtime timestamps.  This
758  * corresponds (after conversion to a Unix timestamp) with an incore timestamp
759  * of Jul  2 20:20:24 UTC 2486.
760  *
761  * We round down the ondisk limit so that the bigtime quota and inode max
762  * timestamps will be the same.
763  */
764 #define XFS_BIGTIME_TIME_MAX	((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL))
765 
766 /*
767  * Bigtime epoch is set exactly to the minimum time value that a traditional
768  * 32-bit timestamp can represent when using the Unix epoch as a reference.
769  * Hence the Unix epoch is at a fixed offset into the supported bigtime
770  * timestamp range.
771  *
772  * The bigtime epoch also matches the minimum value an on-disk 32-bit XFS
773  * timestamp can represent so we will not lose any fidelity in converting
774  * to/from unix and bigtime timestamps.
775  *
776  * The following conversion factor converts a seconds counter from the Unix
777  * epoch to the bigtime epoch.
778  */
779 #define XFS_BIGTIME_EPOCH_OFFSET	(-(int64_t)S32_MIN)
780 
781 /* Convert a timestamp from the Unix epoch to the bigtime epoch. */
782 static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds)
783 {
784 	return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET;
785 }
786 
787 /* Convert a timestamp from the bigtime epoch to the Unix epoch. */
788 static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds)
789 {
790 	return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET;
791 }
792 
793 /*
794  * On-disk inode structure.
795  *
796  * This is just the header or "dinode core", the inode is expanded to fill a
797  * variable size the leftover area split into a data and an attribute fork.
798  * The format of the data and attribute fork depends on the format of the
799  * inode as indicated by di_format and di_aformat.  To access the data and
800  * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
801  * below.
802  *
803  * There is a very similar struct xfs_log_dinode which matches the layout of
804  * this structure, but is kept in native format instead of big endian.
805  *
806  * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
807  * padding field for v3 inodes.
808  */
809 #define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
810 struct xfs_dinode {
811 	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
812 	__be16		di_mode;	/* mode and type of file */
813 	__u8		di_version;	/* inode version */
814 	__u8		di_format;	/* format of di_c data */
815 	__be16		di_onlink;	/* old number of links to file */
816 	__be32		di_uid;		/* owner's user id */
817 	__be32		di_gid;		/* owner's group id */
818 	__be32		di_nlink;	/* number of links to file */
819 	__be16		di_projid_lo;	/* lower part of owner's project id */
820 	__be16		di_projid_hi;	/* higher part owner's project id */
821 	union {
822 		/* Number of data fork extents if NREXT64 is set */
823 		__be64	di_big_nextents;
824 
825 		/* Padding for V3 inodes without NREXT64 set. */
826 		__be64	di_v3_pad;
827 
828 		/* Padding and inode flush counter for V2 inodes. */
829 		struct {
830 			__u8	di_v2_pad[6];
831 			__be16	di_flushiter;
832 		};
833 	};
834 	xfs_timestamp_t	di_atime;	/* time last accessed */
835 	xfs_timestamp_t	di_mtime;	/* time last modified */
836 	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
837 	__be64		di_size;	/* number of bytes in file */
838 	__be64		di_nblocks;	/* # of direct & btree blocks used */
839 	__be32		di_extsize;	/* basic/minimum extent size for file */
840 	union {
841 		/*
842 		 * For V2 inodes and V3 inodes without NREXT64 set, this
843 		 * is the number of data and attr fork extents.
844 		 */
845 		struct {
846 			__be32	di_nextents;
847 			__be16	di_anextents;
848 		} __packed;
849 
850 		/* Number of attr fork extents if NREXT64 is set. */
851 		struct {
852 			__be32	di_big_anextents;
853 			__be16	di_nrext64_pad;
854 		} __packed;
855 	} __packed;
856 	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
857 	__s8		di_aformat;	/* format of attr fork's data */
858 	__be32		di_dmevmask;	/* DMIG event mask */
859 	__be16		di_dmstate;	/* DMIG state info */
860 	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
861 	__be32		di_gen;		/* generation number */
862 
863 	/* di_next_unlinked is the only non-core field in the old dinode */
864 	__be32		di_next_unlinked;/* agi unlinked list ptr */
865 
866 	/* start of the extended dinode, writable fields */
867 	__le32		di_crc;		/* CRC of the inode */
868 	__be64		di_changecount;	/* number of attribute changes */
869 	__be64		di_lsn;		/* flush sequence */
870 	__be64		di_flags2;	/* more random flags */
871 	__be32		di_cowextsize;	/* basic cow extent size for file */
872 	__u8		di_pad2[12];	/* more padding for future expansion */
873 
874 	/* fields only written to during inode creation */
875 	xfs_timestamp_t	di_crtime;	/* time created */
876 	__be64		di_ino;		/* inode number */
877 	uuid_t		di_uuid;	/* UUID of the filesystem */
878 
879 	/* structure must be padded to 64 bit alignment */
880 };
881 
882 #define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
883 
884 #define DI_MAX_FLUSH 0xffff
885 
886 /*
887  * Size of the core inode on disk.  Version 1 and 2 inodes have
888  * the same size, but version 3 has grown a few additional fields.
889  */
890 static inline uint xfs_dinode_size(int version)
891 {
892 	if (version == 3)
893 		return sizeof(struct xfs_dinode);
894 	return offsetof(struct xfs_dinode, di_crc);
895 }
896 
897 /*
898  * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
899  * Since the pathconf interface is signed, we use 2^31 - 1 instead.
900  */
901 #define	XFS_MAXLINK		((1U << 31) - 1U)
902 
903 /*
904  * Any file that hits the maximum ondisk link count should be pinned to avoid
905  * a use-after-free situation.
906  */
907 #define	XFS_NLINK_PINNED	(~0U)
908 
909 /*
910  * Values for di_format
911  *
912  * This enum is used in string mapping in xfs_trace.h; please keep the
913  * TRACE_DEFINE_ENUMs for it up to date.
914  */
915 enum xfs_dinode_fmt {
916 	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
917 	XFS_DINODE_FMT_LOCAL,		/* bulk data */
918 	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
919 	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
920 	XFS_DINODE_FMT_UUID		/* added long ago, but never used */
921 };
922 
923 #define XFS_INODE_FORMAT_STR \
924 	{ XFS_DINODE_FMT_DEV,		"dev" }, \
925 	{ XFS_DINODE_FMT_LOCAL,		"local" }, \
926 	{ XFS_DINODE_FMT_EXTENTS,	"extent" }, \
927 	{ XFS_DINODE_FMT_BTREE,		"btree" }, \
928 	{ XFS_DINODE_FMT_UUID,		"uuid" }
929 
930 /*
931  * Max values for extnum and aextnum.
932  *
933  * The original on-disk extent counts were held in signed fields, resulting in
934  * maximum extent counts of 2^31 and 2^15 for the data and attr forks
935  * respectively. Similarly the maximum extent length is limited to 2^21 blocks
936  * by the 21-bit wide blockcount field of a BMBT extent record.
937  *
938  * The newly introduced data fork extent counter can hold a 64-bit value,
939  * however the maximum number of extents in a file is also limited to 2^54
940  * extents by the 54-bit wide startoff field of a BMBT extent record.
941  *
942  * It is further limited by the maximum supported file size of 2^63
943  * *bytes*. This leads to a maximum extent count for maximally sized filesystem
944  * blocks (64kB) of:
945  *
946  * 2^63 bytes / 2^16 bytes per block = 2^47 blocks
947  *
948  * Rounding up 47 to the nearest multiple of bits-per-byte results in 48. Hence
949  * 2^48 was chosen as the maximum data fork extent count.
950  *
951  * The maximum file size that can be represented by the data fork extent counter
952  * in the worst case occurs when all extents are 1 block in length and each
953  * block is 1KB in size.
954  *
955  * With XFS_MAX_EXTCNT_DATA_FORK_SMALL representing maximum extent count and
956  * with 1KB sized blocks, a file can reach upto,
957  * 1KB * (2^31) = 2TB
958  *
959  * This is much larger than the theoretical maximum size of a directory
960  * i.e. XFS_DIR2_SPACE_SIZE * XFS_DIR2_MAX_SPACES = ~96GB.
961  *
962  * Hence, a directory inode can never overflow its data fork extent counter.
963  */
964 #define XFS_MAX_EXTCNT_DATA_FORK_LARGE	((xfs_extnum_t)((1ULL << 48) - 1))
965 #define XFS_MAX_EXTCNT_ATTR_FORK_LARGE	((xfs_extnum_t)((1ULL << 32) - 1))
966 #define XFS_MAX_EXTCNT_DATA_FORK_SMALL	((xfs_extnum_t)((1ULL << 31) - 1))
967 #define XFS_MAX_EXTCNT_ATTR_FORK_SMALL	((xfs_extnum_t)((1ULL << 15) - 1))
968 
969 /*
970  * When we upgrade an inode to the large extent counts, the maximum value by
971  * which the extent count can increase is bound by the change in size of the
972  * on-disk field. No upgrade operation should ever be adding more than a few
973  * tens of extents, so if we get a really large value it is a sign of a code bug
974  * or corruption.
975  */
976 #define XFS_MAX_EXTCNT_UPGRADE_NR	\
977 	min(XFS_MAX_EXTCNT_ATTR_FORK_LARGE - XFS_MAX_EXTCNT_ATTR_FORK_SMALL,	\
978 	    XFS_MAX_EXTCNT_DATA_FORK_LARGE - XFS_MAX_EXTCNT_DATA_FORK_SMALL)
979 
980 /*
981  * Inode minimum and maximum sizes.
982  */
983 #define	XFS_DINODE_MIN_LOG	8
984 #define	XFS_DINODE_MAX_LOG	11
985 #define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
986 #define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
987 
988 /*
989  * Inode size for given fs.
990  */
991 #define XFS_DINODE_SIZE(mp) \
992 	(xfs_has_v3inodes(mp) ? \
993 		sizeof(struct xfs_dinode) : \
994 		offsetof(struct xfs_dinode, di_crc))
995 #define XFS_LITINO(mp) \
996 	((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(mp))
997 
998 /*
999  * Inode data & attribute fork sizes, per inode.
1000  */
1001 #define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
1002 
1003 #define XFS_DFORK_DSIZE(dip,mp) \
1004 	((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp))
1005 #define XFS_DFORK_ASIZE(dip,mp) \
1006 	((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0)
1007 #define XFS_DFORK_SIZE(dip,mp,w) \
1008 	((w) == XFS_DATA_FORK ? \
1009 		XFS_DFORK_DSIZE(dip, mp) : \
1010 		XFS_DFORK_ASIZE(dip, mp))
1011 
1012 #define XFS_DFORK_MAXEXT(dip, mp, w) \
1013 	(XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
1014 
1015 /*
1016  * Return pointers to the data or attribute forks.
1017  */
1018 #define XFS_DFORK_DPTR(dip) \
1019 	((void *)dip + xfs_dinode_size(dip->di_version))
1020 #define XFS_DFORK_APTR(dip)	\
1021 	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
1022 #define XFS_DFORK_PTR(dip,w)	\
1023 	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
1024 
1025 #define XFS_DFORK_FORMAT(dip,w) \
1026 	((w) == XFS_DATA_FORK ? \
1027 		(dip)->di_format : \
1028 		(dip)->di_aformat)
1029 
1030 /*
1031  * For block and character special files the 32bit dev_t is stored at the
1032  * beginning of the data fork.
1033  */
1034 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1035 {
1036 	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1037 }
1038 
1039 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1040 {
1041 	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1042 }
1043 
1044 /*
1045  * Values for di_flags
1046  */
1047 #define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
1048 #define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
1049 #define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
1050 #define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
1051 #define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
1052 #define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
1053 #define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
1054 #define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
1055 #define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
1056 #define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
1057 #define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
1058 #define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
1059 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
1060 #define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
1061 #define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
1062 /* Do not use bit 15, di_flags is legacy and unchanging now */
1063 
1064 #define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
1065 #define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
1066 #define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
1067 #define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1068 #define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
1069 #define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
1070 #define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
1071 #define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
1072 #define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
1073 #define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1074 #define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1075 #define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
1076 #define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1077 #define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
1078 #define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
1079 
1080 #define XFS_DIFLAG_ANY \
1081 	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1082 	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1083 	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1084 	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1085 	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1086 
1087 /*
1088  * Values for di_flags2 These start by being exposed to userspace in the upper
1089  * 16 bits of the XFS_XFLAG_s range.
1090  */
1091 #define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
1092 #define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
1093 #define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
1094 #define XFS_DIFLAG2_BIGTIME_BIT	3	/* big timestamps */
1095 #define XFS_DIFLAG2_NREXT64_BIT 4	/* large extent counters */
1096 
1097 #define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
1098 #define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
1099 #define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1100 #define XFS_DIFLAG2_BIGTIME	(1 << XFS_DIFLAG2_BIGTIME_BIT)
1101 #define XFS_DIFLAG2_NREXT64	(1 << XFS_DIFLAG2_NREXT64_BIT)
1102 
1103 #define XFS_DIFLAG2_ANY \
1104 	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \
1105 	 XFS_DIFLAG2_BIGTIME | XFS_DIFLAG2_NREXT64)
1106 
1107 static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip)
1108 {
1109 	return dip->di_version >= 3 &&
1110 	       (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME));
1111 }
1112 
1113 static inline bool xfs_dinode_has_large_extent_counts(
1114 	const struct xfs_dinode *dip)
1115 {
1116 	return dip->di_version >= 3 &&
1117 	       (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_NREXT64));
1118 }
1119 
1120 /*
1121  * Inode number format:
1122  * low inopblog bits - offset in block
1123  * next agblklog bits - block number in ag
1124  * next agno_log bits - ag number
1125  * high agno_log-agblklog-inopblog bits - 0
1126  */
1127 #define	XFS_INO_MASK(k)			(uint32_t)((1ULL << (k)) - 1)
1128 #define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1129 #define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1130 #define	XFS_INO_AGINO_BITS(mp)		((mp)->m_ino_geo.agino_log)
1131 #define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1132 #define	XFS_INO_BITS(mp)		\
1133 	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1134 #define	XFS_INO_TO_AGNO(mp,i)		\
1135 	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1136 #define	XFS_INO_TO_AGINO(mp,i)		\
1137 	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1138 #define	XFS_INO_TO_AGBNO(mp,i)		\
1139 	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1140 		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1141 #define	XFS_INO_TO_OFFSET(mp,i)		\
1142 	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1143 #define	XFS_INO_TO_FSB(mp,i)		\
1144 	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1145 #define	XFS_AGINO_TO_INO(mp,a,i)	\
1146 	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1147 #define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1148 #define	XFS_AGINO_TO_OFFSET(mp,i)	\
1149 	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1150 #define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1151 	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1152 #define	XFS_FSB_TO_INO(mp, b)	((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1153 #define	XFS_AGB_TO_AGINO(mp, b)	((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1154 
1155 #define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1156 #define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1157 
1158 /*
1159  * RealTime Device format definitions
1160  */
1161 
1162 /* Min and max rt extent sizes, specified in bytes */
1163 #define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1164 #define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1165 #define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1166 
1167 /*
1168  * Dquot and dquot block format definitions
1169  */
1170 #define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1171 #define XFS_DQUOT_VERSION	(uint8_t)0x01	/* latest version number */
1172 
1173 #define XFS_DQTYPE_USER		(1u << 0)	/* user dquot record */
1174 #define XFS_DQTYPE_PROJ		(1u << 1)	/* project dquot record */
1175 #define XFS_DQTYPE_GROUP	(1u << 2)	/* group dquot record */
1176 #define XFS_DQTYPE_BIGTIME	(1u << 7)	/* large expiry timestamps */
1177 
1178 /* bitmask to determine if this is a user/group/project dquot */
1179 #define XFS_DQTYPE_REC_MASK	(XFS_DQTYPE_USER | \
1180 				 XFS_DQTYPE_PROJ | \
1181 				 XFS_DQTYPE_GROUP)
1182 
1183 #define XFS_DQTYPE_ANY		(XFS_DQTYPE_REC_MASK | \
1184 				 XFS_DQTYPE_BIGTIME)
1185 
1186 /*
1187  * XFS Quota Timers
1188  * ================
1189  *
1190  * Traditional quota grace period expiration timers are an unsigned 32-bit
1191  * seconds counter; time zero is the Unix epoch, Jan  1 00:00:01 UTC 1970.
1192  * Note that an expiration value of zero means that the quota limit has not
1193  * been reached, and therefore no expiration has been set.  Therefore, the
1194  * ondisk min and max defined here can be used directly to constrain the incore
1195  * quota expiration timestamps on a Unix system.
1196  *
1197  * When bigtime is enabled, we trade two bits of precision to expand the
1198  * expiration timeout range to match that of big inode timestamps.  The min and
1199  * max recorded here are the on-disk limits, not a Unix timestamp.
1200  *
1201  * The grace period for each quota type is stored in the root dquot (id = 0)
1202  * and is applied to a non-root dquot when it exceeds the soft or hard limits.
1203  * The length of quota grace periods are unsigned 32-bit quantities measured in
1204  * units of seconds.  A value of zero means to use the default period.
1205  */
1206 
1207 /*
1208  * Smallest possible ondisk quota expiration value with traditional timestamps.
1209  * This corresponds exactly with the incore expiration Jan  1 00:00:01 UTC 1970.
1210  */
1211 #define XFS_DQ_LEGACY_EXPIRY_MIN	((int64_t)1)
1212 
1213 /*
1214  * Largest possible ondisk quota expiration value with traditional timestamps.
1215  * This corresponds exactly with the incore expiration Feb  7 06:28:15 UTC 2106.
1216  */
1217 #define XFS_DQ_LEGACY_EXPIRY_MAX	((int64_t)U32_MAX)
1218 
1219 /*
1220  * Smallest possible ondisk quota expiration value with bigtime timestamps.
1221  * This corresponds (after conversion to a Unix timestamp) with the incore
1222  * expiration of Jan  1 00:00:04 UTC 1970.
1223  */
1224 #define XFS_DQ_BIGTIME_EXPIRY_MIN	(XFS_DQ_LEGACY_EXPIRY_MIN)
1225 
1226 /*
1227  * Largest supported ondisk quota expiration value with bigtime timestamps.
1228  * This corresponds (after conversion to a Unix timestamp) with an incore
1229  * expiration of Jul  2 20:20:24 UTC 2486.
1230  *
1231  * The ondisk field supports values up to -1U, which corresponds to an incore
1232  * expiration in 2514.  This is beyond the maximum the bigtime inode timestamp,
1233  * so we cap the maximum bigtime quota expiration to the max inode timestamp.
1234  */
1235 #define XFS_DQ_BIGTIME_EXPIRY_MAX	((int64_t)4074815106U)
1236 
1237 /*
1238  * The following conversion factors assist in converting a quota expiration
1239  * timestamp between the incore and ondisk formats.
1240  */
1241 #define XFS_DQ_BIGTIME_SHIFT	(2)
1242 #define XFS_DQ_BIGTIME_SLACK	((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1)
1243 
1244 /* Convert an incore quota expiration timestamp to an ondisk bigtime value. */
1245 static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds)
1246 {
1247 	/*
1248 	 * Round the expiration timestamp up to the nearest bigtime timestamp
1249 	 * that we can store, to give users the most time to fix problems.
1250 	 */
1251 	return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >>
1252 			XFS_DQ_BIGTIME_SHIFT;
1253 }
1254 
1255 /* Convert an ondisk bigtime quota expiration value to an incore timestamp. */
1256 static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)
1257 {
1258 	return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT;
1259 }
1260 
1261 /*
1262  * Default quota grace periods, ranging from zero (use the compiled defaults)
1263  * to ~136 years.  These are applied to a non-root dquot that has exceeded
1264  * either limit.
1265  */
1266 #define XFS_DQ_GRACE_MIN		((int64_t)0)
1267 #define XFS_DQ_GRACE_MAX		((int64_t)U32_MAX)
1268 
1269 /* Maximum id value for a quota record */
1270 #define XFS_DQ_ID_MAX			(U32_MAX)
1271 
1272 /*
1273  * This is the main portion of the on-disk representation of quota information
1274  * for a user.  We pad this with some more expansion room to construct the on
1275  * disk structure.
1276  */
1277 struct xfs_disk_dquot {
1278 	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1279 	__u8		d_version;	/* dquot version */
1280 	__u8		d_type;		/* XFS_DQTYPE_USER/PROJ/GROUP */
1281 	__be32		d_id;		/* user,project,group id */
1282 	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1283 	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1284 	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1285 	__be64		d_ino_softlimit;/* preferred inode limit */
1286 	__be64		d_bcount;	/* disk blocks owned by the user */
1287 	__be64		d_icount;	/* inodes owned by the user */
1288 	__be32		d_itimer;	/* zero if within inode limits if not,
1289 					   this is when we refuse service */
1290 	__be32		d_btimer;	/* similar to above; for disk blocks */
1291 	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1292 	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1293 	__be32		d_pad0;		/* 64 bit align */
1294 	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1295 	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1296 	__be64		d_rtbcount;	/* realtime blocks owned */
1297 	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1298 	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1299 	__be16		d_pad;
1300 };
1301 
1302 /*
1303  * This is what goes on disk. This is separated from the xfs_disk_dquot because
1304  * carrying the unnecessary padding would be a waste of memory.
1305  */
1306 struct xfs_dqblk {
1307 	struct xfs_disk_dquot	dd_diskdq; /* portion living incore as well */
1308 	char			dd_fill[4];/* filling for posterity */
1309 
1310 	/*
1311 	 * These two are only present on filesystems with the CRC bits set.
1312 	 */
1313 	__be32		  dd_crc;	/* checksum */
1314 	__be64		  dd_lsn;	/* last modification in log */
1315 	uuid_t		  dd_uuid;	/* location information */
1316 };
1317 
1318 #define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1319 
1320 /*
1321  * This defines the unit of allocation of dquots.
1322  *
1323  * Currently, it is just one file system block, and a 4K blk contains 30
1324  * (136 * 30 = 4080) dquots. It's probably not worth trying to make
1325  * this more dynamic.
1326  *
1327  * However, if this number is changed, we have to make sure that we don't
1328  * implicitly assume that we do allocations in chunks of a single filesystem
1329  * block in the dquot/xqm code.
1330  *
1331  * This is part of the ondisk format because the structure size is not a power
1332  * of two, which leaves slack at the end of the disk block.
1333  */
1334 #define XFS_DQUOT_CLUSTER_SIZE_FSB	(xfs_filblks_t)1
1335 
1336 /*
1337  * Remote symlink format and access functions.
1338  */
1339 #define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1340 
1341 struct xfs_dsymlink_hdr {
1342 	__be32	sl_magic;
1343 	__be32	sl_offset;
1344 	__be32	sl_bytes;
1345 	__be32	sl_crc;
1346 	uuid_t	sl_uuid;
1347 	__be64	sl_owner;
1348 	__be64	sl_blkno;
1349 	__be64	sl_lsn;
1350 };
1351 
1352 #define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1353 
1354 #define XFS_SYMLINK_MAXLEN	1024
1355 /*
1356  * The maximum pathlen is 1024 bytes. Since the minimum file system
1357  * blocksize is 512 bytes, we can get a max of 3 extents back from
1358  * bmapi when crc headers are taken into account.
1359  */
1360 #define XFS_SYMLINK_MAPS 3
1361 
1362 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1363 	((bufsize) - (xfs_has_crc((mp)) ? \
1364 			sizeof(struct xfs_dsymlink_hdr) : 0))
1365 
1366 
1367 /*
1368  * Allocation Btree format definitions
1369  *
1370  * There are two on-disk btrees, one sorted by blockno and one sorted
1371  * by blockcount and blockno.  All blocks look the same to make the code
1372  * simpler; if we have time later, we'll make the optimizations.
1373  */
1374 #define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1375 #define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1376 #define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1377 #define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1378 
1379 /*
1380  * Data record/key structure
1381  */
1382 typedef struct xfs_alloc_rec {
1383 	__be32		ar_startblock;	/* starting block number */
1384 	__be32		ar_blockcount;	/* count of free blocks */
1385 } xfs_alloc_rec_t, xfs_alloc_key_t;
1386 
1387 typedef struct xfs_alloc_rec_incore {
1388 	xfs_agblock_t	ar_startblock;	/* starting block number */
1389 	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1390 } xfs_alloc_rec_incore_t;
1391 
1392 /* btree pointer type */
1393 typedef __be32 xfs_alloc_ptr_t;
1394 
1395 /*
1396  * Block numbers in the AG:
1397  * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1398  */
1399 #define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1400 #define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1401 
1402 
1403 /*
1404  * Inode Allocation Btree format definitions
1405  *
1406  * There is a btree for the inode map per allocation group.
1407  */
1408 #define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1409 #define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1410 #define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1411 #define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1412 
1413 typedef uint64_t	xfs_inofree_t;
1414 #define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1415 #define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1416 #define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1417 #define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1418 
1419 #define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1420 #define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(uint16_t))
1421 #define XFS_INODES_PER_HOLEMASK_BIT	\
1422 	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1423 
1424 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1425 {
1426 	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1427 }
1428 
1429 /*
1430  * The on-disk inode record structure has two formats. The original "full"
1431  * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1432  * and replaces the 3 high-order freecount bytes wth the holemask and inode
1433  * count.
1434  *
1435  * The holemask of the sparse record format allows an inode chunk to have holes
1436  * that refer to blocks not owned by the inode record. This facilitates inode
1437  * allocation in the event of severe free space fragmentation.
1438  */
1439 typedef struct xfs_inobt_rec {
1440 	__be32		ir_startino;	/* starting inode number */
1441 	union {
1442 		struct {
1443 			__be32	ir_freecount;	/* count of free inodes */
1444 		} f;
1445 		struct {
1446 			__be16	ir_holemask;/* hole mask for sparse chunks */
1447 			__u8	ir_count;	/* total inode count */
1448 			__u8	ir_freecount;	/* count of free inodes */
1449 		} sp;
1450 	} ir_u;
1451 	__be64		ir_free;	/* free inode mask */
1452 } xfs_inobt_rec_t;
1453 
1454 typedef struct xfs_inobt_rec_incore {
1455 	xfs_agino_t	ir_startino;	/* starting inode number */
1456 	uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1457 	uint8_t		ir_count;	/* total inode count */
1458 	uint8_t		ir_freecount;	/* count of free inodes (set bits) */
1459 	xfs_inofree_t	ir_free;	/* free inode mask */
1460 } xfs_inobt_rec_incore_t;
1461 
1462 static inline bool xfs_inobt_issparse(uint16_t holemask)
1463 {
1464 	/* non-zero holemask represents a sparse rec. */
1465 	return holemask;
1466 }
1467 
1468 /*
1469  * Key structure
1470  */
1471 typedef struct xfs_inobt_key {
1472 	__be32		ir_startino;	/* starting inode number */
1473 } xfs_inobt_key_t;
1474 
1475 /* btree pointer type */
1476 typedef __be32 xfs_inobt_ptr_t;
1477 
1478 /*
1479  * block numbers in the AG.
1480  */
1481 #define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1482 #define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1483 
1484 /*
1485  * Reverse mapping btree format definitions
1486  *
1487  * There is a btree for the reverse map per allocation group
1488  */
1489 #define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1490 
1491 /*
1492  * Ownership info for an extent.  This is used to create reverse-mapping
1493  * entries.
1494  */
1495 #define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1496 #define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1497 struct xfs_owner_info {
1498 	uint64_t		oi_owner;
1499 	xfs_fileoff_t		oi_offset;
1500 	unsigned int		oi_flags;
1501 };
1502 
1503 /*
1504  * Special owner types.
1505  *
1506  * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1507  * to tell us we have a special owner value. We use these for static metadata
1508  * allocated at mkfs/growfs time, as well as for freespace management metadata.
1509  */
1510 #define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1511 #define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1512 #define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1513 #define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1514 #define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1515 #define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1516 #define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1517 #define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1518 #define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1519 #define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1520 
1521 #define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1522 
1523 /*
1524  * Data record structure
1525  */
1526 struct xfs_rmap_rec {
1527 	__be32		rm_startblock;	/* extent start block */
1528 	__be32		rm_blockcount;	/* extent length */
1529 	__be64		rm_owner;	/* extent owner */
1530 	__be64		rm_offset;	/* offset within the owner */
1531 };
1532 
1533 /*
1534  * rmap btree record
1535  *  rm_offset:63 is the attribute fork flag
1536  *  rm_offset:62 is the bmbt block flag
1537  *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1538  *  rm_offset:54-60 aren't used and should be zero
1539  *  rm_offset:0-53 is the block offset within the inode
1540  */
1541 #define XFS_RMAP_OFF_ATTR_FORK	((uint64_t)1ULL << 63)
1542 #define XFS_RMAP_OFF_BMBT_BLOCK	((uint64_t)1ULL << 62)
1543 #define XFS_RMAP_OFF_UNWRITTEN	((uint64_t)1ULL << 61)
1544 
1545 #define XFS_RMAP_LEN_MAX	((uint32_t)~0U)
1546 #define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1547 				 XFS_RMAP_OFF_BMBT_BLOCK | \
1548 				 XFS_RMAP_OFF_UNWRITTEN)
1549 #define XFS_RMAP_OFF_MASK	((uint64_t)0x3FFFFFFFFFFFFFULL)
1550 
1551 #define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1552 
1553 #define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1554 #define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1555 #define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1556 
1557 #define RMAPBT_STARTBLOCK_BITLEN	32
1558 #define RMAPBT_BLOCKCOUNT_BITLEN	32
1559 #define RMAPBT_OWNER_BITLEN		64
1560 #define RMAPBT_ATTRFLAG_BITLEN		1
1561 #define RMAPBT_BMBTFLAG_BITLEN		1
1562 #define RMAPBT_EXNTFLAG_BITLEN		1
1563 #define RMAPBT_UNUSED_OFFSET_BITLEN	7
1564 #define RMAPBT_OFFSET_BITLEN		54
1565 
1566 /*
1567  * Key structure
1568  *
1569  * We don't use the length for lookups
1570  */
1571 struct xfs_rmap_key {
1572 	__be32		rm_startblock;	/* extent start block */
1573 	__be64		rm_owner;	/* extent owner */
1574 	__be64		rm_offset;	/* offset within the owner */
1575 } __attribute__((packed));
1576 
1577 /* btree pointer type */
1578 typedef __be32 xfs_rmap_ptr_t;
1579 
1580 #define	XFS_RMAP_BLOCK(mp) \
1581 	(xfs_has_finobt(((mp))) ? \
1582 	 XFS_FIBT_BLOCK(mp) + 1 : \
1583 	 XFS_IBT_BLOCK(mp) + 1)
1584 
1585 /*
1586  * Reference Count Btree format definitions
1587  *
1588  */
1589 #define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1590 
1591 unsigned int xfs_refc_block(struct xfs_mount *mp);
1592 
1593 /*
1594  * Data record/key structure
1595  *
1596  * Each record associates a range of physical blocks (starting at
1597  * rc_startblock and ending rc_blockcount blocks later) with a reference
1598  * count (rc_refcount).  Extents that are being used to stage a copy on
1599  * write (CoW) operation are recorded in the refcount btree with a
1600  * refcount of 1.  All other records must have a refcount > 1 and must
1601  * track an extent mapped only by file data forks.
1602  *
1603  * Extents with a single owner (attributes, metadata, non-shared file
1604  * data) are not tracked here.  Free space is also not tracked here.
1605  * This is consistent with pre-reflink XFS.
1606  */
1607 
1608 /*
1609  * Extents that are being used to stage a copy on write are stored
1610  * in the refcount btree with a refcount of 1 and the upper bit set
1611  * on the startblock.  This speeds up mount time deletion of stale
1612  * staging extents because they're all at the right side of the tree.
1613  */
1614 #define XFS_REFC_COWFLAG		(1U << 31)
1615 #define REFCNTBT_COWFLAG_BITLEN		1
1616 #define REFCNTBT_AGBLOCK_BITLEN		31
1617 
1618 struct xfs_refcount_rec {
1619 	__be32		rc_startblock;	/* starting block number */
1620 	__be32		rc_blockcount;	/* count of blocks */
1621 	__be32		rc_refcount;	/* number of inodes linked here */
1622 };
1623 
1624 struct xfs_refcount_key {
1625 	__be32		rc_startblock;	/* starting block number */
1626 };
1627 
1628 #define MAXREFCOUNT	((xfs_nlink_t)~0U)
1629 #define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
1630 
1631 /* btree pointer type */
1632 typedef __be32 xfs_refcount_ptr_t;
1633 
1634 
1635 /*
1636  * BMAP Btree format definitions
1637  *
1638  * This includes both the root block definition that sits inside an inode fork
1639  * and the record/pointer formats for the leaf/node in the blocks.
1640  */
1641 #define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1642 #define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1643 
1644 /*
1645  * Bmap root header, on-disk form only.
1646  */
1647 typedef struct xfs_bmdr_block {
1648 	__be16		bb_level;	/* 0 is a leaf */
1649 	__be16		bb_numrecs;	/* current # of data records */
1650 } xfs_bmdr_block_t;
1651 
1652 /*
1653  * Bmap btree record and extent descriptor.
1654  *  l0:63 is an extent flag (value 1 indicates non-normal).
1655  *  l0:9-62 are startoff.
1656  *  l0:0-8 and l1:21-63 are startblock.
1657  *  l1:0-20 are blockcount.
1658  */
1659 #define BMBT_EXNTFLAG_BITLEN	1
1660 #define BMBT_STARTOFF_BITLEN	54
1661 #define BMBT_STARTBLOCK_BITLEN	52
1662 #define BMBT_BLOCKCOUNT_BITLEN	21
1663 
1664 #define BMBT_STARTOFF_MASK	((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1665 #define BMBT_BLOCKCOUNT_MASK	((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1)
1666 
1667 #define XFS_MAX_BMBT_EXTLEN	((xfs_extlen_t)(BMBT_BLOCKCOUNT_MASK))
1668 
1669 /*
1670  * bmbt records have a file offset (block) field that is 54 bits wide, so this
1671  * is the largest xfs_fileoff_t that we ever expect to see.
1672  */
1673 #define XFS_MAX_FILEOFF		(BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK)
1674 
1675 typedef struct xfs_bmbt_rec {
1676 	__be64			l0, l1;
1677 } xfs_bmbt_rec_t;
1678 
1679 typedef uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1680 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1681 
1682 /*
1683  * Values and macros for delayed-allocation startblock fields.
1684  */
1685 #define STARTBLOCKVALBITS	17
1686 #define STARTBLOCKMASKBITS	(15 + 20)
1687 #define STARTBLOCKMASK		\
1688 	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1689 
1690 static inline int isnullstartblock(xfs_fsblock_t x)
1691 {
1692 	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1693 }
1694 
1695 static inline xfs_fsblock_t nullstartblock(int k)
1696 {
1697 	ASSERT(k < (1 << STARTBLOCKVALBITS));
1698 	return STARTBLOCKMASK | (k);
1699 }
1700 
1701 static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1702 {
1703 	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1704 }
1705 
1706 /*
1707  * Key structure for non-leaf levels of the tree.
1708  */
1709 typedef struct xfs_bmbt_key {
1710 	__be64		br_startoff;	/* starting file offset */
1711 } xfs_bmbt_key_t, xfs_bmdr_key_t;
1712 
1713 /* btree pointer type */
1714 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1715 
1716 
1717 /*
1718  * Generic Btree block format definitions
1719  *
1720  * This is a combination of the actual format used on disk for short and long
1721  * format btrees.  The first three fields are shared by both format, but the
1722  * pointers are different and should be used with care.
1723  *
1724  * To get the size of the actual short or long form headers please use the size
1725  * macros below.  Never use sizeof(xfs_btree_block).
1726  *
1727  * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1728  * with the crc feature bit, and all accesses to them must be conditional on
1729  * that flag.
1730  */
1731 /* short form block header */
1732 struct xfs_btree_block_shdr {
1733 	__be32		bb_leftsib;
1734 	__be32		bb_rightsib;
1735 
1736 	__be64		bb_blkno;
1737 	__be64		bb_lsn;
1738 	uuid_t		bb_uuid;
1739 	__be32		bb_owner;
1740 	__le32		bb_crc;
1741 };
1742 
1743 /* long form block header */
1744 struct xfs_btree_block_lhdr {
1745 	__be64		bb_leftsib;
1746 	__be64		bb_rightsib;
1747 
1748 	__be64		bb_blkno;
1749 	__be64		bb_lsn;
1750 	uuid_t		bb_uuid;
1751 	__be64		bb_owner;
1752 	__le32		bb_crc;
1753 	__be32		bb_pad; /* padding for alignment */
1754 };
1755 
1756 struct xfs_btree_block {
1757 	__be32		bb_magic;	/* magic number for block type */
1758 	__be16		bb_level;	/* 0 is a leaf */
1759 	__be16		bb_numrecs;	/* current # of data records */
1760 	union {
1761 		struct xfs_btree_block_shdr s;
1762 		struct xfs_btree_block_lhdr l;
1763 	} bb_u;				/* rest */
1764 };
1765 
1766 /* size of a short form block */
1767 #define XFS_BTREE_SBLOCK_LEN \
1768 	(offsetof(struct xfs_btree_block, bb_u) + \
1769 	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1770 /* size of a long form block */
1771 #define XFS_BTREE_LBLOCK_LEN \
1772 	(offsetof(struct xfs_btree_block, bb_u) + \
1773 	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1774 
1775 /* sizes of CRC enabled btree blocks */
1776 #define XFS_BTREE_SBLOCK_CRC_LEN \
1777 	(offsetof(struct xfs_btree_block, bb_u) + \
1778 	 sizeof(struct xfs_btree_block_shdr))
1779 #define XFS_BTREE_LBLOCK_CRC_LEN \
1780 	(offsetof(struct xfs_btree_block, bb_u) + \
1781 	 sizeof(struct xfs_btree_block_lhdr))
1782 
1783 #define XFS_BTREE_SBLOCK_CRC_OFF \
1784 	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1785 #define XFS_BTREE_LBLOCK_CRC_OFF \
1786 	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1787 
1788 /*
1789  * On-disk XFS access control list structure.
1790  */
1791 struct xfs_acl_entry {
1792 	__be32	ae_tag;
1793 	__be32	ae_id;
1794 	__be16	ae_perm;
1795 	__be16	ae_pad;		/* fill the implicit hole in the structure */
1796 };
1797 
1798 struct xfs_acl {
1799 	__be32			acl_cnt;
1800 	struct xfs_acl_entry	acl_entry[];
1801 };
1802 
1803 /*
1804  * The number of ACL entries allowed is defined by the on-disk format.
1805  * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1806  * limited only by the maximum size of the xattr that stores the information.
1807  */
1808 #define XFS_ACL_MAX_ENTRIES(mp)	\
1809 	(xfs_has_crc(mp) \
1810 		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1811 						sizeof(struct xfs_acl_entry) \
1812 		: 25)
1813 
1814 #define XFS_ACL_SIZE(cnt) \
1815 	(sizeof(struct xfs_acl) + \
1816 		sizeof(struct xfs_acl_entry) * cnt)
1817 
1818 #define XFS_ACL_MAX_SIZE(mp) \
1819 	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1820 
1821 
1822 /* On-disk XFS extended attribute names */
1823 #define SGI_ACL_FILE		"SGI_ACL_FILE"
1824 #define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
1825 #define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
1826 #define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
1827 
1828 #endif /* __XFS_FORMAT_H__ */
1829