xref: /openbmc/linux/fs/xfs/libxfs/xfs_format.h (revision 0cabf991)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4  * All Rights Reserved.
5  */
6 #ifndef __XFS_FORMAT_H__
7 #define __XFS_FORMAT_H__
8 
9 /*
10  * XFS On Disk Format Definitions
11  *
12  * This header file defines all the on-disk format definitions for
13  * general XFS objects. Directory and attribute related objects are defined in
14  * xfs_da_format.h, which log and log item formats are defined in
15  * xfs_log_format.h. Everything else goes here.
16  */
17 
18 struct xfs_mount;
19 struct xfs_trans;
20 struct xfs_inode;
21 struct xfs_buf;
22 struct xfs_ifork;
23 
24 /*
25  * Super block
26  * Fits into a sector-sized buffer at address 0 of each allocation group.
27  * Only the first of these is ever updated except during growfs.
28  */
29 #define	XFS_SB_MAGIC		0x58465342	/* 'XFSB' */
30 #define	XFS_SB_VERSION_1	1		/* 5.3, 6.0.1, 6.1 */
31 #define	XFS_SB_VERSION_2	2		/* 6.2 - attributes */
32 #define	XFS_SB_VERSION_3	3		/* 6.2 - new inode version */
33 #define	XFS_SB_VERSION_4	4		/* 6.2+ - bitmask version */
34 #define	XFS_SB_VERSION_5	5		/* CRC enabled filesystem */
35 #define	XFS_SB_VERSION_NUMBITS		0x000f
36 #define	XFS_SB_VERSION_ALLFBITS		0xfff0
37 #define	XFS_SB_VERSION_ATTRBIT		0x0010
38 #define	XFS_SB_VERSION_NLINKBIT		0x0020
39 #define	XFS_SB_VERSION_QUOTABIT		0x0040
40 #define	XFS_SB_VERSION_ALIGNBIT		0x0080
41 #define	XFS_SB_VERSION_DALIGNBIT	0x0100
42 #define	XFS_SB_VERSION_SHAREDBIT	0x0200
43 #define XFS_SB_VERSION_LOGV2BIT		0x0400
44 #define XFS_SB_VERSION_SECTORBIT	0x0800
45 #define	XFS_SB_VERSION_EXTFLGBIT	0x1000
46 #define	XFS_SB_VERSION_DIRV2BIT		0x2000
47 #define	XFS_SB_VERSION_BORGBIT		0x4000	/* ASCII only case-insens. */
48 #define	XFS_SB_VERSION_MOREBITSBIT	0x8000
49 
50 /*
51  * The size of a single extended attribute on disk is limited by
52  * the size of index values within the attribute entries themselves.
53  * These are be16 fields, so we can only support attribute data
54  * sizes up to 2^16 bytes in length.
55  */
56 #define XFS_XATTR_SIZE_MAX (1 << 16)
57 
58 /*
59  * Supported feature bit list is just all bits in the versionnum field because
60  * we've used them all up and understand them all. Except, of course, for the
61  * shared superblock bit, which nobody knows what it does and so is unsupported.
62  */
63 #define	XFS_SB_VERSION_OKBITS		\
64 	((XFS_SB_VERSION_NUMBITS | XFS_SB_VERSION_ALLFBITS) & \
65 		~XFS_SB_VERSION_SHAREDBIT)
66 
67 /*
68  * There are two words to hold XFS "feature" bits: the original
69  * word, sb_versionnum, and sb_features2.  Whenever a bit is set in
70  * sb_features2, the feature bit XFS_SB_VERSION_MOREBITSBIT must be set.
71  *
72  * These defines represent bits in sb_features2.
73  */
74 #define XFS_SB_VERSION2_RESERVED1BIT	0x00000001
75 #define XFS_SB_VERSION2_LAZYSBCOUNTBIT	0x00000002	/* Superblk counters */
76 #define XFS_SB_VERSION2_RESERVED4BIT	0x00000004
77 #define XFS_SB_VERSION2_ATTR2BIT	0x00000008	/* Inline attr rework */
78 #define XFS_SB_VERSION2_PARENTBIT	0x00000010	/* parent pointers */
79 #define XFS_SB_VERSION2_PROJID32BIT	0x00000080	/* 32 bit project id */
80 #define XFS_SB_VERSION2_CRCBIT		0x00000100	/* metadata CRCs */
81 #define XFS_SB_VERSION2_FTYPE		0x00000200	/* inode type in dir */
82 
83 #define	XFS_SB_VERSION2_OKBITS		\
84 	(XFS_SB_VERSION2_LAZYSBCOUNTBIT	| \
85 	 XFS_SB_VERSION2_ATTR2BIT	| \
86 	 XFS_SB_VERSION2_PROJID32BIT	| \
87 	 XFS_SB_VERSION2_FTYPE)
88 
89 /* Maximum size of the xfs filesystem label, no terminating NULL */
90 #define XFSLABEL_MAX			12
91 
92 /*
93  * Superblock - in core version.  Must match the ondisk version below.
94  * Must be padded to 64 bit alignment.
95  */
96 typedef struct xfs_sb {
97 	uint32_t	sb_magicnum;	/* magic number == XFS_SB_MAGIC */
98 	uint32_t	sb_blocksize;	/* logical block size, bytes */
99 	xfs_rfsblock_t	sb_dblocks;	/* number of data blocks */
100 	xfs_rfsblock_t	sb_rblocks;	/* number of realtime blocks */
101 	xfs_rtblock_t	sb_rextents;	/* number of realtime extents */
102 	uuid_t		sb_uuid;	/* user-visible file system unique id */
103 	xfs_fsblock_t	sb_logstart;	/* starting block of log if internal */
104 	xfs_ino_t	sb_rootino;	/* root inode number */
105 	xfs_ino_t	sb_rbmino;	/* bitmap inode for realtime extents */
106 	xfs_ino_t	sb_rsumino;	/* summary inode for rt bitmap */
107 	xfs_agblock_t	sb_rextsize;	/* realtime extent size, blocks */
108 	xfs_agblock_t	sb_agblocks;	/* size of an allocation group */
109 	xfs_agnumber_t	sb_agcount;	/* number of allocation groups */
110 	xfs_extlen_t	sb_rbmblocks;	/* number of rt bitmap blocks */
111 	xfs_extlen_t	sb_logblocks;	/* number of log blocks */
112 	uint16_t	sb_versionnum;	/* header version == XFS_SB_VERSION */
113 	uint16_t	sb_sectsize;	/* volume sector size, bytes */
114 	uint16_t	sb_inodesize;	/* inode size, bytes */
115 	uint16_t	sb_inopblock;	/* inodes per block */
116 	char		sb_fname[XFSLABEL_MAX]; /* file system name */
117 	uint8_t		sb_blocklog;	/* log2 of sb_blocksize */
118 	uint8_t		sb_sectlog;	/* log2 of sb_sectsize */
119 	uint8_t		sb_inodelog;	/* log2 of sb_inodesize */
120 	uint8_t		sb_inopblog;	/* log2 of sb_inopblock */
121 	uint8_t		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
122 	uint8_t		sb_rextslog;	/* log2 of sb_rextents */
123 	uint8_t		sb_inprogress;	/* mkfs is in progress, don't mount */
124 	uint8_t		sb_imax_pct;	/* max % of fs for inode space */
125 					/* statistics */
126 	/*
127 	 * These fields must remain contiguous.  If you really
128 	 * want to change their layout, make sure you fix the
129 	 * code in xfs_trans_apply_sb_deltas().
130 	 */
131 	uint64_t	sb_icount;	/* allocated inodes */
132 	uint64_t	sb_ifree;	/* free inodes */
133 	uint64_t	sb_fdblocks;	/* free data blocks */
134 	uint64_t	sb_frextents;	/* free realtime extents */
135 	/*
136 	 * End contiguous fields.
137 	 */
138 	xfs_ino_t	sb_uquotino;	/* user quota inode */
139 	xfs_ino_t	sb_gquotino;	/* group quota inode */
140 	uint16_t	sb_qflags;	/* quota flags */
141 	uint8_t		sb_flags;	/* misc. flags */
142 	uint8_t		sb_shared_vn;	/* shared version number */
143 	xfs_extlen_t	sb_inoalignmt;	/* inode chunk alignment, fsblocks */
144 	uint32_t	sb_unit;	/* stripe or raid unit */
145 	uint32_t	sb_width;	/* stripe or raid width */
146 	uint8_t		sb_dirblklog;	/* log2 of dir block size (fsbs) */
147 	uint8_t		sb_logsectlog;	/* log2 of the log sector size */
148 	uint16_t	sb_logsectsize;	/* sector size for the log, bytes */
149 	uint32_t	sb_logsunit;	/* stripe unit size for the log */
150 	uint32_t	sb_features2;	/* additional feature bits */
151 
152 	/*
153 	 * bad features2 field as a result of failing to pad the sb structure to
154 	 * 64 bits. Some machines will be using this field for features2 bits.
155 	 * Easiest just to mark it bad and not use it for anything else.
156 	 *
157 	 * This is not kept up to date in memory; it is always overwritten by
158 	 * the value in sb_features2 when formatting the incore superblock to
159 	 * the disk buffer.
160 	 */
161 	uint32_t	sb_bad_features2;
162 
163 	/* version 5 superblock fields start here */
164 
165 	/* feature masks */
166 	uint32_t	sb_features_compat;
167 	uint32_t	sb_features_ro_compat;
168 	uint32_t	sb_features_incompat;
169 	uint32_t	sb_features_log_incompat;
170 
171 	uint32_t	sb_crc;		/* superblock crc */
172 	xfs_extlen_t	sb_spino_align;	/* sparse inode chunk alignment */
173 
174 	xfs_ino_t	sb_pquotino;	/* project quota inode */
175 	xfs_lsn_t	sb_lsn;		/* last write sequence */
176 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
177 
178 	/* must be padded to 64 bit alignment */
179 } xfs_sb_t;
180 
181 #define XFS_SB_CRC_OFF		offsetof(struct xfs_sb, sb_crc)
182 
183 /*
184  * Superblock - on disk version.  Must match the in core version above.
185  * Must be padded to 64 bit alignment.
186  */
187 typedef struct xfs_dsb {
188 	__be32		sb_magicnum;	/* magic number == XFS_SB_MAGIC */
189 	__be32		sb_blocksize;	/* logical block size, bytes */
190 	__be64		sb_dblocks;	/* number of data blocks */
191 	__be64		sb_rblocks;	/* number of realtime blocks */
192 	__be64		sb_rextents;	/* number of realtime extents */
193 	uuid_t		sb_uuid;	/* user-visible file system unique id */
194 	__be64		sb_logstart;	/* starting block of log if internal */
195 	__be64		sb_rootino;	/* root inode number */
196 	__be64		sb_rbmino;	/* bitmap inode for realtime extents */
197 	__be64		sb_rsumino;	/* summary inode for rt bitmap */
198 	__be32		sb_rextsize;	/* realtime extent size, blocks */
199 	__be32		sb_agblocks;	/* size of an allocation group */
200 	__be32		sb_agcount;	/* number of allocation groups */
201 	__be32		sb_rbmblocks;	/* number of rt bitmap blocks */
202 	__be32		sb_logblocks;	/* number of log blocks */
203 	__be16		sb_versionnum;	/* header version == XFS_SB_VERSION */
204 	__be16		sb_sectsize;	/* volume sector size, bytes */
205 	__be16		sb_inodesize;	/* inode size, bytes */
206 	__be16		sb_inopblock;	/* inodes per block */
207 	char		sb_fname[XFSLABEL_MAX]; /* file system name */
208 	__u8		sb_blocklog;	/* log2 of sb_blocksize */
209 	__u8		sb_sectlog;	/* log2 of sb_sectsize */
210 	__u8		sb_inodelog;	/* log2 of sb_inodesize */
211 	__u8		sb_inopblog;	/* log2 of sb_inopblock */
212 	__u8		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
213 	__u8		sb_rextslog;	/* log2 of sb_rextents */
214 	__u8		sb_inprogress;	/* mkfs is in progress, don't mount */
215 	__u8		sb_imax_pct;	/* max % of fs for inode space */
216 					/* statistics */
217 	/*
218 	 * These fields must remain contiguous.  If you really
219 	 * want to change their layout, make sure you fix the
220 	 * code in xfs_trans_apply_sb_deltas().
221 	 */
222 	__be64		sb_icount;	/* allocated inodes */
223 	__be64		sb_ifree;	/* free inodes */
224 	__be64		sb_fdblocks;	/* free data blocks */
225 	__be64		sb_frextents;	/* free realtime extents */
226 	/*
227 	 * End contiguous fields.
228 	 */
229 	__be64		sb_uquotino;	/* user quota inode */
230 	__be64		sb_gquotino;	/* group quota inode */
231 	__be16		sb_qflags;	/* quota flags */
232 	__u8		sb_flags;	/* misc. flags */
233 	__u8		sb_shared_vn;	/* shared version number */
234 	__be32		sb_inoalignmt;	/* inode chunk alignment, fsblocks */
235 	__be32		sb_unit;	/* stripe or raid unit */
236 	__be32		sb_width;	/* stripe or raid width */
237 	__u8		sb_dirblklog;	/* log2 of dir block size (fsbs) */
238 	__u8		sb_logsectlog;	/* log2 of the log sector size */
239 	__be16		sb_logsectsize;	/* sector size for the log, bytes */
240 	__be32		sb_logsunit;	/* stripe unit size for the log */
241 	__be32		sb_features2;	/* additional feature bits */
242 	/*
243 	 * bad features2 field as a result of failing to pad the sb
244 	 * structure to 64 bits. Some machines will be using this field
245 	 * for features2 bits. Easiest just to mark it bad and not use
246 	 * it for anything else.
247 	 */
248 	__be32		sb_bad_features2;
249 
250 	/* version 5 superblock fields start here */
251 
252 	/* feature masks */
253 	__be32		sb_features_compat;
254 	__be32		sb_features_ro_compat;
255 	__be32		sb_features_incompat;
256 	__be32		sb_features_log_incompat;
257 
258 	__le32		sb_crc;		/* superblock crc */
259 	__be32		sb_spino_align;	/* sparse inode chunk alignment */
260 
261 	__be64		sb_pquotino;	/* project quota inode */
262 	__be64		sb_lsn;		/* last write sequence */
263 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
264 
265 	/* must be padded to 64 bit alignment */
266 } xfs_dsb_t;
267 
268 
269 /*
270  * Misc. Flags - warning - these will be cleared by xfs_repair unless
271  * a feature bit is set when the flag is used.
272  */
273 #define XFS_SBF_NOFLAGS		0x00	/* no flags set */
274 #define XFS_SBF_READONLY	0x01	/* only read-only mounts allowed */
275 
276 /*
277  * define max. shared version we can interoperate with
278  */
279 #define XFS_SB_MAX_SHARED_VN	0
280 
281 #define	XFS_SB_VERSION_NUM(sbp)	((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
282 
283 /*
284  * The first XFS version we support is a v4 superblock with V2 directories.
285  */
286 static inline bool xfs_sb_good_v4_features(struct xfs_sb *sbp)
287 {
288 	if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
289 		return false;
290 	if (!(sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT))
291 		return false;
292 
293 	/* check for unknown features in the fs */
294 	if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
295 	    ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
296 	     (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
297 		return false;
298 
299 	return true;
300 }
301 
302 static inline bool xfs_sb_good_version(struct xfs_sb *sbp)
303 {
304 	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5)
305 		return true;
306 	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_4)
307 		return xfs_sb_good_v4_features(sbp);
308 	return false;
309 }
310 
311 static inline bool xfs_sb_version_hasrealtime(struct xfs_sb *sbp)
312 {
313 	return sbp->sb_rblocks > 0;
314 }
315 
316 /*
317  * Detect a mismatched features2 field.  Older kernels read/wrote
318  * this into the wrong slot, so to be safe we keep them in sync.
319  */
320 static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
321 {
322 	return sbp->sb_bad_features2 != sbp->sb_features2;
323 }
324 
325 static inline bool xfs_sb_version_hasattr(struct xfs_sb *sbp)
326 {
327 	return (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT);
328 }
329 
330 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
331 {
332 	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
333 }
334 
335 static inline bool xfs_sb_version_hasquota(struct xfs_sb *sbp)
336 {
337 	return (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT);
338 }
339 
340 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
341 {
342 	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
343 }
344 
345 static inline bool xfs_sb_version_hasalign(struct xfs_sb *sbp)
346 {
347 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
348 		(sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT));
349 }
350 
351 static inline bool xfs_sb_version_hasdalign(struct xfs_sb *sbp)
352 {
353 	return (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT);
354 }
355 
356 static inline bool xfs_sb_version_haslogv2(struct xfs_sb *sbp)
357 {
358 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
359 	       (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT);
360 }
361 
362 static inline bool xfs_sb_version_hassector(struct xfs_sb *sbp)
363 {
364 	return (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT);
365 }
366 
367 static inline bool xfs_sb_version_hasasciici(struct xfs_sb *sbp)
368 {
369 	return (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT);
370 }
371 
372 static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp)
373 {
374 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
375 	       (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
376 }
377 
378 /*
379  * sb_features2 bit version macros.
380  */
381 static inline bool xfs_sb_version_haslazysbcount(struct xfs_sb *sbp)
382 {
383 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
384 	       (xfs_sb_version_hasmorebits(sbp) &&
385 		(sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT));
386 }
387 
388 static inline bool xfs_sb_version_hasattr2(struct xfs_sb *sbp)
389 {
390 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
391 	       (xfs_sb_version_hasmorebits(sbp) &&
392 		(sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT));
393 }
394 
395 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
396 {
397 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
398 	sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
399 }
400 
401 static inline void xfs_sb_version_removeattr2(struct xfs_sb *sbp)
402 {
403 	sbp->sb_features2 &= ~XFS_SB_VERSION2_ATTR2BIT;
404 	if (!sbp->sb_features2)
405 		sbp->sb_versionnum &= ~XFS_SB_VERSION_MOREBITSBIT;
406 }
407 
408 static inline bool xfs_sb_version_hasprojid32bit(struct xfs_sb *sbp)
409 {
410 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
411 	       (xfs_sb_version_hasmorebits(sbp) &&
412 		(sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT));
413 }
414 
415 static inline void xfs_sb_version_addprojid32bit(struct xfs_sb *sbp)
416 {
417 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
418 	sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
419 }
420 
421 /*
422  * Extended v5 superblock feature masks. These are to be used for new v5
423  * superblock features only.
424  *
425  * Compat features are new features that old kernels will not notice or affect
426  * and so can mount read-write without issues.
427  *
428  * RO-Compat (read only) are features that old kernels can read but will break
429  * if they write. Hence only read-only mounts of such filesystems are allowed on
430  * kernels that don't support the feature bit.
431  *
432  * InCompat features are features which old kernels will not understand and so
433  * must not mount.
434  *
435  * Log-InCompat features are for changes to log formats or new transactions that
436  * can't be replayed on older kernels. The fields are set when the filesystem is
437  * mounted, and a clean unmount clears the fields.
438  */
439 #define XFS_SB_FEAT_COMPAT_ALL 0
440 #define XFS_SB_FEAT_COMPAT_UNKNOWN	~XFS_SB_FEAT_COMPAT_ALL
441 static inline bool
442 xfs_sb_has_compat_feature(
443 	struct xfs_sb	*sbp,
444 	uint32_t	feature)
445 {
446 	return (sbp->sb_features_compat & feature) != 0;
447 }
448 
449 #define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
450 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
451 #define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
452 #define XFS_SB_FEAT_RO_COMPAT_ALL \
453 		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
454 		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
455 		 XFS_SB_FEAT_RO_COMPAT_REFLINK)
456 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
457 static inline bool
458 xfs_sb_has_ro_compat_feature(
459 	struct xfs_sb	*sbp,
460 	uint32_t	feature)
461 {
462 	return (sbp->sb_features_ro_compat & feature) != 0;
463 }
464 
465 #define XFS_SB_FEAT_INCOMPAT_FTYPE	(1 << 0)	/* filetype in dirent */
466 #define XFS_SB_FEAT_INCOMPAT_SPINODES	(1 << 1)	/* sparse inode chunks */
467 #define XFS_SB_FEAT_INCOMPAT_META_UUID	(1 << 2)	/* metadata UUID */
468 #define XFS_SB_FEAT_INCOMPAT_ALL \
469 		(XFS_SB_FEAT_INCOMPAT_FTYPE|	\
470 		 XFS_SB_FEAT_INCOMPAT_SPINODES|	\
471 		 XFS_SB_FEAT_INCOMPAT_META_UUID)
472 
473 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
474 static inline bool
475 xfs_sb_has_incompat_feature(
476 	struct xfs_sb	*sbp,
477 	uint32_t	feature)
478 {
479 	return (sbp->sb_features_incompat & feature) != 0;
480 }
481 
482 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL 0
483 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
484 static inline bool
485 xfs_sb_has_incompat_log_feature(
486 	struct xfs_sb	*sbp,
487 	uint32_t	feature)
488 {
489 	return (sbp->sb_features_log_incompat & feature) != 0;
490 }
491 
492 /*
493  * V5 superblock specific feature checks
494  */
495 static inline bool xfs_sb_version_hascrc(struct xfs_sb *sbp)
496 {
497 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
498 }
499 
500 /*
501  * v5 file systems support V3 inodes only, earlier file systems support
502  * v2 and v1 inodes.
503  */
504 static inline bool xfs_sb_version_has_v3inode(struct xfs_sb *sbp)
505 {
506 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
507 }
508 
509 static inline bool xfs_dinode_good_version(struct xfs_sb *sbp,
510 		uint8_t version)
511 {
512 	if (xfs_sb_version_has_v3inode(sbp))
513 		return version == 3;
514 	return version == 1 || version == 2;
515 }
516 
517 static inline bool xfs_sb_version_has_pquotino(struct xfs_sb *sbp)
518 {
519 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
520 }
521 
522 static inline int xfs_sb_version_hasftype(struct xfs_sb *sbp)
523 {
524 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
525 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_FTYPE)) ||
526 	       (xfs_sb_version_hasmorebits(sbp) &&
527 		 (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE));
528 }
529 
530 static inline bool xfs_sb_version_hasfinobt(xfs_sb_t *sbp)
531 {
532 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
533 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT);
534 }
535 
536 static inline bool xfs_sb_version_hassparseinodes(struct xfs_sb *sbp)
537 {
538 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
539 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_SPINODES);
540 }
541 
542 /*
543  * XFS_SB_FEAT_INCOMPAT_META_UUID indicates that the metadata UUID
544  * is stored separately from the user-visible UUID; this allows the
545  * user-visible UUID to be changed on V5 filesystems which have a
546  * filesystem UUID stamped into every piece of metadata.
547  */
548 static inline bool xfs_sb_version_hasmetauuid(struct xfs_sb *sbp)
549 {
550 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
551 		(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID);
552 }
553 
554 static inline bool xfs_sb_version_hasrmapbt(struct xfs_sb *sbp)
555 {
556 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
557 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT);
558 }
559 
560 static inline bool xfs_sb_version_hasreflink(struct xfs_sb *sbp)
561 {
562 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
563 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK);
564 }
565 
566 /*
567  * end of superblock version macros
568  */
569 
570 static inline bool
571 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
572 {
573 	return (ino == sbp->sb_uquotino ||
574 		ino == sbp->sb_gquotino ||
575 		ino == sbp->sb_pquotino);
576 }
577 
578 #define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
579 #define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
580 
581 #define	XFS_HDR_BLOCK(mp,d)	((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
582 #define	XFS_DADDR_TO_FSB(mp,d)	XFS_AGB_TO_FSB(mp, \
583 			xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
584 #define	XFS_FSB_TO_DADDR(mp,fsbno)	XFS_AGB_TO_DADDR(mp, \
585 			XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
586 
587 /*
588  * File system sector to basic block conversions.
589  */
590 #define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
591 
592 /*
593  * File system block to basic block conversions.
594  */
595 #define	XFS_FSB_TO_BB(mp,fsbno)	((fsbno) << (mp)->m_blkbb_log)
596 #define	XFS_BB_TO_FSB(mp,bb)	\
597 	(((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
598 #define	XFS_BB_TO_FSBT(mp,bb)	((bb) >> (mp)->m_blkbb_log)
599 
600 /*
601  * File system block to byte conversions.
602  */
603 #define XFS_FSB_TO_B(mp,fsbno)	((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
604 #define XFS_B_TO_FSB(mp,b)	\
605 	((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
606 #define XFS_B_TO_FSBT(mp,b)	(((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
607 #define XFS_B_FSB_OFFSET(mp,b)	((b) & (mp)->m_blockmask)
608 
609 /*
610  * Allocation group header
611  *
612  * This is divided into three structures, placed in sequential 512-byte
613  * buffers after a copy of the superblock (also in a 512-byte buffer).
614  */
615 #define	XFS_AGF_MAGIC	0x58414746	/* 'XAGF' */
616 #define	XFS_AGI_MAGIC	0x58414749	/* 'XAGI' */
617 #define	XFS_AGFL_MAGIC	0x5841464c	/* 'XAFL' */
618 #define	XFS_AGF_VERSION	1
619 #define	XFS_AGI_VERSION	1
620 
621 #define	XFS_AGF_GOOD_VERSION(v)	((v) == XFS_AGF_VERSION)
622 #define	XFS_AGI_GOOD_VERSION(v)	((v) == XFS_AGI_VERSION)
623 
624 /*
625  * Btree number 0 is bno, 1 is cnt, 2 is rmap. This value gives the size of the
626  * arrays below.
627  */
628 #define	XFS_BTNUM_AGF	((int)XFS_BTNUM_RMAPi + 1)
629 
630 /*
631  * The second word of agf_levels in the first a.g. overlaps the EFS
632  * superblock's magic number.  Since the magic numbers valid for EFS
633  * are > 64k, our value cannot be confused for an EFS superblock's.
634  */
635 
636 typedef struct xfs_agf {
637 	/*
638 	 * Common allocation group header information
639 	 */
640 	__be32		agf_magicnum;	/* magic number == XFS_AGF_MAGIC */
641 	__be32		agf_versionnum;	/* header version == XFS_AGF_VERSION */
642 	__be32		agf_seqno;	/* sequence # starting from 0 */
643 	__be32		agf_length;	/* size in blocks of a.g. */
644 	/*
645 	 * Freespace and rmap information
646 	 */
647 	__be32		agf_roots[XFS_BTNUM_AGF];	/* root blocks */
648 	__be32		agf_levels[XFS_BTNUM_AGF];	/* btree levels */
649 
650 	__be32		agf_flfirst;	/* first freelist block's index */
651 	__be32		agf_fllast;	/* last freelist block's index */
652 	__be32		agf_flcount;	/* count of blocks in freelist */
653 	__be32		agf_freeblks;	/* total free blocks */
654 
655 	__be32		agf_longest;	/* longest free space */
656 	__be32		agf_btreeblks;	/* # of blocks held in AGF btrees */
657 	uuid_t		agf_uuid;	/* uuid of filesystem */
658 
659 	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
660 	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
661 
662 	__be32		agf_refcount_root;	/* refcount tree root block */
663 	__be32		agf_refcount_level;	/* refcount btree levels */
664 
665 	/*
666 	 * reserve some contiguous space for future logged fields before we add
667 	 * the unlogged fields. This makes the range logging via flags and
668 	 * structure offsets much simpler.
669 	 */
670 	__be64		agf_spare64[14];
671 
672 	/* unlogged fields, written during buffer writeback. */
673 	__be64		agf_lsn;	/* last write sequence */
674 	__be32		agf_crc;	/* crc of agf sector */
675 	__be32		agf_spare2;
676 
677 	/* structure must be padded to 64 bit alignment */
678 } xfs_agf_t;
679 
680 #define XFS_AGF_CRC_OFF		offsetof(struct xfs_agf, agf_crc)
681 
682 #define	XFS_AGF_MAGICNUM	0x00000001
683 #define	XFS_AGF_VERSIONNUM	0x00000002
684 #define	XFS_AGF_SEQNO		0x00000004
685 #define	XFS_AGF_LENGTH		0x00000008
686 #define	XFS_AGF_ROOTS		0x00000010
687 #define	XFS_AGF_LEVELS		0x00000020
688 #define	XFS_AGF_FLFIRST		0x00000040
689 #define	XFS_AGF_FLLAST		0x00000080
690 #define	XFS_AGF_FLCOUNT		0x00000100
691 #define	XFS_AGF_FREEBLKS	0x00000200
692 #define	XFS_AGF_LONGEST		0x00000400
693 #define	XFS_AGF_BTREEBLKS	0x00000800
694 #define	XFS_AGF_UUID		0x00001000
695 #define	XFS_AGF_RMAP_BLOCKS	0x00002000
696 #define	XFS_AGF_REFCOUNT_BLOCKS	0x00004000
697 #define	XFS_AGF_REFCOUNT_ROOT	0x00008000
698 #define	XFS_AGF_REFCOUNT_LEVEL	0x00010000
699 #define	XFS_AGF_SPARE64		0x00020000
700 #define	XFS_AGF_NUM_BITS	18
701 #define	XFS_AGF_ALL_BITS	((1 << XFS_AGF_NUM_BITS) - 1)
702 
703 #define XFS_AGF_FLAGS \
704 	{ XFS_AGF_MAGICNUM,	"MAGICNUM" }, \
705 	{ XFS_AGF_VERSIONNUM,	"VERSIONNUM" }, \
706 	{ XFS_AGF_SEQNO,	"SEQNO" }, \
707 	{ XFS_AGF_LENGTH,	"LENGTH" }, \
708 	{ XFS_AGF_ROOTS,	"ROOTS" }, \
709 	{ XFS_AGF_LEVELS,	"LEVELS" }, \
710 	{ XFS_AGF_FLFIRST,	"FLFIRST" }, \
711 	{ XFS_AGF_FLLAST,	"FLLAST" }, \
712 	{ XFS_AGF_FLCOUNT,	"FLCOUNT" }, \
713 	{ XFS_AGF_FREEBLKS,	"FREEBLKS" }, \
714 	{ XFS_AGF_LONGEST,	"LONGEST" }, \
715 	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
716 	{ XFS_AGF_UUID,		"UUID" }, \
717 	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
718 	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
719 	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
720 	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
721 	{ XFS_AGF_SPARE64,	"SPARE64" }
722 
723 /* disk block (xfs_daddr_t) in the AG */
724 #define XFS_AGF_DADDR(mp)	((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
725 #define	XFS_AGF_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
726 
727 /*
728  * Size of the unlinked inode hash table in the agi.
729  */
730 #define	XFS_AGI_UNLINKED_BUCKETS	64
731 
732 typedef struct xfs_agi {
733 	/*
734 	 * Common allocation group header information
735 	 */
736 	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
737 	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
738 	__be32		agi_seqno;	/* sequence # starting from 0 */
739 	__be32		agi_length;	/* size in blocks of a.g. */
740 	/*
741 	 * Inode information
742 	 * Inodes are mapped by interpreting the inode number, so no
743 	 * mapping data is needed here.
744 	 */
745 	__be32		agi_count;	/* count of allocated inodes */
746 	__be32		agi_root;	/* root of inode btree */
747 	__be32		agi_level;	/* levels in inode btree */
748 	__be32		agi_freecount;	/* number of free inodes */
749 
750 	__be32		agi_newino;	/* new inode just allocated */
751 	__be32		agi_dirino;	/* last directory inode chunk */
752 	/*
753 	 * Hash table of inodes which have been unlinked but are
754 	 * still being referenced.
755 	 */
756 	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
757 	/*
758 	 * This marks the end of logging region 1 and start of logging region 2.
759 	 */
760 	uuid_t		agi_uuid;	/* uuid of filesystem */
761 	__be32		agi_crc;	/* crc of agi sector */
762 	__be32		agi_pad32;
763 	__be64		agi_lsn;	/* last write sequence */
764 
765 	__be32		agi_free_root; /* root of the free inode btree */
766 	__be32		agi_free_level;/* levels in free inode btree */
767 
768 	/* structure must be padded to 64 bit alignment */
769 } xfs_agi_t;
770 
771 #define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
772 
773 #define	XFS_AGI_MAGICNUM	(1 << 0)
774 #define	XFS_AGI_VERSIONNUM	(1 << 1)
775 #define	XFS_AGI_SEQNO		(1 << 2)
776 #define	XFS_AGI_LENGTH		(1 << 3)
777 #define	XFS_AGI_COUNT		(1 << 4)
778 #define	XFS_AGI_ROOT		(1 << 5)
779 #define	XFS_AGI_LEVEL		(1 << 6)
780 #define	XFS_AGI_FREECOUNT	(1 << 7)
781 #define	XFS_AGI_NEWINO		(1 << 8)
782 #define	XFS_AGI_DIRINO		(1 << 9)
783 #define	XFS_AGI_UNLINKED	(1 << 10)
784 #define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
785 #define	XFS_AGI_ALL_BITS_R1	((1 << XFS_AGI_NUM_BITS_R1) - 1)
786 #define	XFS_AGI_FREE_ROOT	(1 << 11)
787 #define	XFS_AGI_FREE_LEVEL	(1 << 12)
788 #define	XFS_AGI_NUM_BITS_R2	13
789 
790 /* disk block (xfs_daddr_t) in the AG */
791 #define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
792 #define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
793 
794 /*
795  * The third a.g. block contains the a.g. freelist, an array
796  * of block pointers to blocks owned by the allocation btree code.
797  */
798 #define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
799 #define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
800 #define	XFS_BUF_TO_AGFL(bp)	((struct xfs_agfl *)((bp)->b_addr))
801 
802 struct xfs_agfl {
803 	__be32		agfl_magicnum;
804 	__be32		agfl_seqno;
805 	uuid_t		agfl_uuid;
806 	__be64		agfl_lsn;
807 	__be32		agfl_crc;
808 } __attribute__((packed));
809 
810 #define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
811 
812 #define XFS_AGB_TO_FSB(mp,agno,agbno)	\
813 	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
814 #define	XFS_FSB_TO_AGNO(mp,fsbno)	\
815 	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
816 #define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
817 	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
818 #define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
819 	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
820 		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
821 #define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
822 
823 /*
824  * For checking for bad ranges of xfs_daddr_t's, covering multiple
825  * allocation groups or a single xfs_daddr_t that's a superblock copy.
826  */
827 #define	XFS_AG_CHECK_DADDR(mp,d,len)	\
828 	((len) == 1 ? \
829 	    ASSERT((d) == XFS_SB_DADDR || \
830 		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
831 	    ASSERT(xfs_daddr_to_agno(mp, d) == \
832 		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
833 
834 typedef struct xfs_timestamp {
835 	__be32		t_sec;		/* timestamp seconds */
836 	__be32		t_nsec;		/* timestamp nanoseconds */
837 } xfs_timestamp_t;
838 
839 /*
840  * On-disk inode structure.
841  *
842  * This is just the header or "dinode core", the inode is expanded to fill a
843  * variable size the leftover area split into a data and an attribute fork.
844  * The format of the data and attribute fork depends on the format of the
845  * inode as indicated by di_format and di_aformat.  To access the data and
846  * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
847  * below.
848  *
849  * There is a very similar struct icdinode in xfs_inode which matches the
850  * layout of the first 96 bytes of this structure, but is kept in native
851  * format instead of big endian.
852  *
853  * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
854  * padding field for v3 inodes.
855  */
856 #define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
857 typedef struct xfs_dinode {
858 	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
859 	__be16		di_mode;	/* mode and type of file */
860 	__u8		di_version;	/* inode version */
861 	__u8		di_format;	/* format of di_c data */
862 	__be16		di_onlink;	/* old number of links to file */
863 	__be32		di_uid;		/* owner's user id */
864 	__be32		di_gid;		/* owner's group id */
865 	__be32		di_nlink;	/* number of links to file */
866 	__be16		di_projid_lo;	/* lower part of owner's project id */
867 	__be16		di_projid_hi;	/* higher part owner's project id */
868 	__u8		di_pad[6];	/* unused, zeroed space */
869 	__be16		di_flushiter;	/* incremented on flush */
870 	xfs_timestamp_t	di_atime;	/* time last accessed */
871 	xfs_timestamp_t	di_mtime;	/* time last modified */
872 	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
873 	__be64		di_size;	/* number of bytes in file */
874 	__be64		di_nblocks;	/* # of direct & btree blocks used */
875 	__be32		di_extsize;	/* basic/minimum extent size for file */
876 	__be32		di_nextents;	/* number of extents in data fork */
877 	__be16		di_anextents;	/* number of extents in attribute fork*/
878 	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
879 	__s8		di_aformat;	/* format of attr fork's data */
880 	__be32		di_dmevmask;	/* DMIG event mask */
881 	__be16		di_dmstate;	/* DMIG state info */
882 	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
883 	__be32		di_gen;		/* generation number */
884 
885 	/* di_next_unlinked is the only non-core field in the old dinode */
886 	__be32		di_next_unlinked;/* agi unlinked list ptr */
887 
888 	/* start of the extended dinode, writable fields */
889 	__le32		di_crc;		/* CRC of the inode */
890 	__be64		di_changecount;	/* number of attribute changes */
891 	__be64		di_lsn;		/* flush sequence */
892 	__be64		di_flags2;	/* more random flags */
893 	__be32		di_cowextsize;	/* basic cow extent size for file */
894 	__u8		di_pad2[12];	/* more padding for future expansion */
895 
896 	/* fields only written to during inode creation */
897 	xfs_timestamp_t	di_crtime;	/* time created */
898 	__be64		di_ino;		/* inode number */
899 	uuid_t		di_uuid;	/* UUID of the filesystem */
900 
901 	/* structure must be padded to 64 bit alignment */
902 } xfs_dinode_t;
903 
904 #define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
905 
906 #define DI_MAX_FLUSH 0xffff
907 
908 /*
909  * Size of the core inode on disk.  Version 1 and 2 inodes have
910  * the same size, but version 3 has grown a few additional fields.
911  */
912 static inline uint xfs_dinode_size(int version)
913 {
914 	if (version == 3)
915 		return sizeof(struct xfs_dinode);
916 	return offsetof(struct xfs_dinode, di_crc);
917 }
918 
919 /*
920  * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
921  * Since the pathconf interface is signed, we use 2^31 - 1 instead.
922  */
923 #define	XFS_MAXLINK		((1U << 31) - 1U)
924 
925 /*
926  * Values for di_format
927  *
928  * This enum is used in string mapping in xfs_trace.h; please keep the
929  * TRACE_DEFINE_ENUMs for it up to date.
930  */
931 enum xfs_dinode_fmt {
932 	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
933 	XFS_DINODE_FMT_LOCAL,		/* bulk data */
934 	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
935 	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
936 	XFS_DINODE_FMT_UUID		/* added long ago, but never used */
937 };
938 
939 #define XFS_INODE_FORMAT_STR \
940 	{ XFS_DINODE_FMT_DEV,		"dev" }, \
941 	{ XFS_DINODE_FMT_LOCAL,		"local" }, \
942 	{ XFS_DINODE_FMT_EXTENTS,	"extent" }, \
943 	{ XFS_DINODE_FMT_BTREE,		"btree" }, \
944 	{ XFS_DINODE_FMT_UUID,		"uuid" }
945 
946 /*
947  * Inode minimum and maximum sizes.
948  */
949 #define	XFS_DINODE_MIN_LOG	8
950 #define	XFS_DINODE_MAX_LOG	11
951 #define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
952 #define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
953 
954 /*
955  * Inode size for given fs.
956  */
957 #define XFS_DINODE_SIZE(sbp) \
958 	(xfs_sb_version_has_v3inode(sbp) ? \
959 		sizeof(struct xfs_dinode) : \
960 		offsetof(struct xfs_dinode, di_crc))
961 #define XFS_LITINO(mp) \
962 	((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(&(mp)->m_sb))
963 
964 /*
965  * Inode data & attribute fork sizes, per inode.
966  */
967 #define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
968 
969 #define XFS_DFORK_DSIZE(dip,mp) \
970 	((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp))
971 #define XFS_DFORK_ASIZE(dip,mp) \
972 	((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0)
973 #define XFS_DFORK_SIZE(dip,mp,w) \
974 	((w) == XFS_DATA_FORK ? \
975 		XFS_DFORK_DSIZE(dip, mp) : \
976 		XFS_DFORK_ASIZE(dip, mp))
977 
978 #define XFS_DFORK_MAXEXT(dip, mp, w) \
979 	(XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
980 
981 /*
982  * Return pointers to the data or attribute forks.
983  */
984 #define XFS_DFORK_DPTR(dip) \
985 	((char *)dip + xfs_dinode_size(dip->di_version))
986 #define XFS_DFORK_APTR(dip)	\
987 	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
988 #define XFS_DFORK_PTR(dip,w)	\
989 	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
990 
991 #define XFS_DFORK_FORMAT(dip,w) \
992 	((w) == XFS_DATA_FORK ? \
993 		(dip)->di_format : \
994 		(dip)->di_aformat)
995 #define XFS_DFORK_NEXTENTS(dip,w) \
996 	((w) == XFS_DATA_FORK ? \
997 		be32_to_cpu((dip)->di_nextents) : \
998 		be16_to_cpu((dip)->di_anextents))
999 
1000 /*
1001  * For block and character special files the 32bit dev_t is stored at the
1002  * beginning of the data fork.
1003  */
1004 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1005 {
1006 	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1007 }
1008 
1009 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1010 {
1011 	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1012 }
1013 
1014 /*
1015  * Values for di_flags
1016  */
1017 #define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
1018 #define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
1019 #define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
1020 #define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
1021 #define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
1022 #define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
1023 #define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
1024 #define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
1025 #define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
1026 #define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
1027 #define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
1028 #define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
1029 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
1030 #define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
1031 #define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
1032 /* Do not use bit 15, di_flags is legacy and unchanging now */
1033 
1034 #define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
1035 #define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
1036 #define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
1037 #define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1038 #define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
1039 #define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
1040 #define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
1041 #define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
1042 #define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
1043 #define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1044 #define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1045 #define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
1046 #define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1047 #define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
1048 #define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
1049 
1050 #define XFS_DIFLAG_ANY \
1051 	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1052 	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1053 	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1054 	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1055 	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1056 
1057 /*
1058  * Values for di_flags2 These start by being exposed to userspace in the upper
1059  * 16 bits of the XFS_XFLAG_s range.
1060  */
1061 #define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
1062 #define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
1063 #define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
1064 #define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
1065 #define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
1066 #define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1067 
1068 #define XFS_DIFLAG2_ANY \
1069 	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)
1070 
1071 /*
1072  * Inode number format:
1073  * low inopblog bits - offset in block
1074  * next agblklog bits - block number in ag
1075  * next agno_log bits - ag number
1076  * high agno_log-agblklog-inopblog bits - 0
1077  */
1078 #define	XFS_INO_MASK(k)			(uint32_t)((1ULL << (k)) - 1)
1079 #define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1080 #define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1081 #define	XFS_INO_AGINO_BITS(mp)		((mp)->m_ino_geo.agino_log)
1082 #define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1083 #define	XFS_INO_BITS(mp)		\
1084 	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1085 #define	XFS_INO_TO_AGNO(mp,i)		\
1086 	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1087 #define	XFS_INO_TO_AGINO(mp,i)		\
1088 	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1089 #define	XFS_INO_TO_AGBNO(mp,i)		\
1090 	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1091 		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1092 #define	XFS_INO_TO_OFFSET(mp,i)		\
1093 	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1094 #define	XFS_INO_TO_FSB(mp,i)		\
1095 	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1096 #define	XFS_AGINO_TO_INO(mp,a,i)	\
1097 	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1098 #define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1099 #define	XFS_AGINO_TO_OFFSET(mp,i)	\
1100 	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1101 #define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1102 	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1103 #define	XFS_FSB_TO_INO(mp, b)	((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1104 #define	XFS_AGB_TO_AGINO(mp, b)	((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1105 
1106 #define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1107 #define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1108 
1109 /*
1110  * RealTime Device format definitions
1111  */
1112 
1113 /* Min and max rt extent sizes, specified in bytes */
1114 #define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1115 #define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1116 #define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1117 
1118 #define	XFS_BLOCKSIZE(mp)	((mp)->m_sb.sb_blocksize)
1119 #define	XFS_BLOCKMASK(mp)	((mp)->m_blockmask)
1120 #define	XFS_BLOCKWSIZE(mp)	((mp)->m_blockwsize)
1121 #define	XFS_BLOCKWMASK(mp)	((mp)->m_blockwmask)
1122 
1123 /*
1124  * RT Summary and bit manipulation macros.
1125  */
1126 #define	XFS_SUMOFFS(mp,ls,bb)	((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb)))
1127 #define	XFS_SUMOFFSTOBLOCK(mp,s)	\
1128 	(((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog)
1129 #define	XFS_SUMPTR(mp,bp,so)	\
1130 	((xfs_suminfo_t *)((bp)->b_addr + \
1131 		(((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp))))
1132 
1133 #define	XFS_BITTOBLOCK(mp,bi)	((bi) >> (mp)->m_blkbit_log)
1134 #define	XFS_BLOCKTOBIT(mp,bb)	((bb) << (mp)->m_blkbit_log)
1135 #define	XFS_BITTOWORD(mp,bi)	\
1136 	((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp)))
1137 
1138 #define	XFS_RTMIN(a,b)	((a) < (b) ? (a) : (b))
1139 #define	XFS_RTMAX(a,b)	((a) > (b) ? (a) : (b))
1140 
1141 #define	XFS_RTLOBIT(w)	xfs_lowbit32(w)
1142 #define	XFS_RTHIBIT(w)	xfs_highbit32(w)
1143 
1144 #define	XFS_RTBLOCKLOG(b)	xfs_highbit64(b)
1145 
1146 /*
1147  * Dquot and dquot block format definitions
1148  */
1149 #define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1150 #define XFS_DQUOT_VERSION	(uint8_t)0x01	/* latest version number */
1151 
1152 #define XFS_DQTYPE_USER		0x01		/* user dquot record */
1153 #define XFS_DQTYPE_PROJ		0x02		/* project dquot record */
1154 #define XFS_DQTYPE_GROUP	0x04		/* group dquot record */
1155 
1156 /* bitmask to determine if this is a user/group/project dquot */
1157 #define XFS_DQTYPE_REC_MASK	(XFS_DQTYPE_USER | \
1158 				 XFS_DQTYPE_PROJ | \
1159 				 XFS_DQTYPE_GROUP)
1160 
1161 #define XFS_DQTYPE_ANY		(XFS_DQTYPE_REC_MASK)
1162 
1163 /*
1164  * This is the main portion of the on-disk representation of quota information
1165  * for a user.  We pad this with some more expansion room to construct the on
1166  * disk structure.
1167  */
1168 struct xfs_disk_dquot {
1169 	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1170 	__u8		d_version;	/* dquot version */
1171 	__u8		d_type;		/* XFS_DQTYPE_USER/PROJ/GROUP */
1172 	__be32		d_id;		/* user,project,group id */
1173 	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1174 	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1175 	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1176 	__be64		d_ino_softlimit;/* preferred inode limit */
1177 	__be64		d_bcount;	/* disk blocks owned by the user */
1178 	__be64		d_icount;	/* inodes owned by the user */
1179 	__be32		d_itimer;	/* zero if within inode limits if not,
1180 					   this is when we refuse service */
1181 	__be32		d_btimer;	/* similar to above; for disk blocks */
1182 	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1183 	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1184 	__be32		d_pad0;		/* 64 bit align */
1185 	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1186 	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1187 	__be64		d_rtbcount;	/* realtime blocks owned */
1188 	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1189 	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1190 	__be16		d_pad;
1191 };
1192 
1193 /*
1194  * This is what goes on disk. This is separated from the xfs_disk_dquot because
1195  * carrying the unnecessary padding would be a waste of memory.
1196  */
1197 typedef struct xfs_dqblk {
1198 	struct xfs_disk_dquot	dd_diskdq; /* portion living incore as well */
1199 	char			dd_fill[4];/* filling for posterity */
1200 
1201 	/*
1202 	 * These two are only present on filesystems with the CRC bits set.
1203 	 */
1204 	__be32		  dd_crc;	/* checksum */
1205 	__be64		  dd_lsn;	/* last modification in log */
1206 	uuid_t		  dd_uuid;	/* location information */
1207 } xfs_dqblk_t;
1208 
1209 #define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1210 
1211 /*
1212  * This defines the unit of allocation of dquots.
1213  *
1214  * Currently, it is just one file system block, and a 4K blk contains 30
1215  * (136 * 30 = 4080) dquots. It's probably not worth trying to make
1216  * this more dynamic.
1217  *
1218  * However, if this number is changed, we have to make sure that we don't
1219  * implicitly assume that we do allocations in chunks of a single filesystem
1220  * block in the dquot/xqm code.
1221  *
1222  * This is part of the ondisk format because the structure size is not a power
1223  * of two, which leaves slack at the end of the disk block.
1224  */
1225 #define XFS_DQUOT_CLUSTER_SIZE_FSB	(xfs_filblks_t)1
1226 
1227 /*
1228  * Remote symlink format and access functions.
1229  */
1230 #define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1231 
1232 struct xfs_dsymlink_hdr {
1233 	__be32	sl_magic;
1234 	__be32	sl_offset;
1235 	__be32	sl_bytes;
1236 	__be32	sl_crc;
1237 	uuid_t	sl_uuid;
1238 	__be64	sl_owner;
1239 	__be64	sl_blkno;
1240 	__be64	sl_lsn;
1241 };
1242 
1243 #define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1244 
1245 #define XFS_SYMLINK_MAXLEN	1024
1246 /*
1247  * The maximum pathlen is 1024 bytes. Since the minimum file system
1248  * blocksize is 512 bytes, we can get a max of 3 extents back from
1249  * bmapi when crc headers are taken into account.
1250  */
1251 #define XFS_SYMLINK_MAPS 3
1252 
1253 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1254 	((bufsize) - (xfs_sb_version_hascrc(&(mp)->m_sb) ? \
1255 			sizeof(struct xfs_dsymlink_hdr) : 0))
1256 
1257 
1258 /*
1259  * Allocation Btree format definitions
1260  *
1261  * There are two on-disk btrees, one sorted by blockno and one sorted
1262  * by blockcount and blockno.  All blocks look the same to make the code
1263  * simpler; if we have time later, we'll make the optimizations.
1264  */
1265 #define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1266 #define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1267 #define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1268 #define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1269 
1270 /*
1271  * Data record/key structure
1272  */
1273 typedef struct xfs_alloc_rec {
1274 	__be32		ar_startblock;	/* starting block number */
1275 	__be32		ar_blockcount;	/* count of free blocks */
1276 } xfs_alloc_rec_t, xfs_alloc_key_t;
1277 
1278 typedef struct xfs_alloc_rec_incore {
1279 	xfs_agblock_t	ar_startblock;	/* starting block number */
1280 	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1281 } xfs_alloc_rec_incore_t;
1282 
1283 /* btree pointer type */
1284 typedef __be32 xfs_alloc_ptr_t;
1285 
1286 /*
1287  * Block numbers in the AG:
1288  * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1289  */
1290 #define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1291 #define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1292 
1293 
1294 /*
1295  * Inode Allocation Btree format definitions
1296  *
1297  * There is a btree for the inode map per allocation group.
1298  */
1299 #define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1300 #define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1301 #define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1302 #define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1303 
1304 typedef uint64_t	xfs_inofree_t;
1305 #define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1306 #define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1307 #define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1308 #define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1309 
1310 #define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1311 #define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(uint16_t))
1312 #define XFS_INODES_PER_HOLEMASK_BIT	\
1313 	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1314 
1315 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1316 {
1317 	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1318 }
1319 
1320 /*
1321  * The on-disk inode record structure has two formats. The original "full"
1322  * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1323  * and replaces the 3 high-order freecount bytes wth the holemask and inode
1324  * count.
1325  *
1326  * The holemask of the sparse record format allows an inode chunk to have holes
1327  * that refer to blocks not owned by the inode record. This facilitates inode
1328  * allocation in the event of severe free space fragmentation.
1329  */
1330 typedef struct xfs_inobt_rec {
1331 	__be32		ir_startino;	/* starting inode number */
1332 	union {
1333 		struct {
1334 			__be32	ir_freecount;	/* count of free inodes */
1335 		} f;
1336 		struct {
1337 			__be16	ir_holemask;/* hole mask for sparse chunks */
1338 			__u8	ir_count;	/* total inode count */
1339 			__u8	ir_freecount;	/* count of free inodes */
1340 		} sp;
1341 	} ir_u;
1342 	__be64		ir_free;	/* free inode mask */
1343 } xfs_inobt_rec_t;
1344 
1345 typedef struct xfs_inobt_rec_incore {
1346 	xfs_agino_t	ir_startino;	/* starting inode number */
1347 	uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1348 	uint8_t		ir_count;	/* total inode count */
1349 	uint8_t		ir_freecount;	/* count of free inodes (set bits) */
1350 	xfs_inofree_t	ir_free;	/* free inode mask */
1351 } xfs_inobt_rec_incore_t;
1352 
1353 static inline bool xfs_inobt_issparse(uint16_t holemask)
1354 {
1355 	/* non-zero holemask represents a sparse rec. */
1356 	return holemask;
1357 }
1358 
1359 /*
1360  * Key structure
1361  */
1362 typedef struct xfs_inobt_key {
1363 	__be32		ir_startino;	/* starting inode number */
1364 } xfs_inobt_key_t;
1365 
1366 /* btree pointer type */
1367 typedef __be32 xfs_inobt_ptr_t;
1368 
1369 /*
1370  * block numbers in the AG.
1371  */
1372 #define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1373 #define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1374 
1375 /*
1376  * Reverse mapping btree format definitions
1377  *
1378  * There is a btree for the reverse map per allocation group
1379  */
1380 #define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1381 
1382 /*
1383  * Ownership info for an extent.  This is used to create reverse-mapping
1384  * entries.
1385  */
1386 #define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1387 #define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1388 struct xfs_owner_info {
1389 	uint64_t		oi_owner;
1390 	xfs_fileoff_t		oi_offset;
1391 	unsigned int		oi_flags;
1392 };
1393 
1394 /*
1395  * Special owner types.
1396  *
1397  * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1398  * to tell us we have a special owner value. We use these for static metadata
1399  * allocated at mkfs/growfs time, as well as for freespace management metadata.
1400  */
1401 #define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1402 #define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1403 #define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1404 #define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1405 #define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1406 #define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1407 #define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1408 #define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1409 #define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1410 #define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1411 
1412 #define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1413 
1414 /*
1415  * Data record structure
1416  */
1417 struct xfs_rmap_rec {
1418 	__be32		rm_startblock;	/* extent start block */
1419 	__be32		rm_blockcount;	/* extent length */
1420 	__be64		rm_owner;	/* extent owner */
1421 	__be64		rm_offset;	/* offset within the owner */
1422 };
1423 
1424 /*
1425  * rmap btree record
1426  *  rm_offset:63 is the attribute fork flag
1427  *  rm_offset:62 is the bmbt block flag
1428  *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1429  *  rm_offset:54-60 aren't used and should be zero
1430  *  rm_offset:0-53 is the block offset within the inode
1431  */
1432 #define XFS_RMAP_OFF_ATTR_FORK	((uint64_t)1ULL << 63)
1433 #define XFS_RMAP_OFF_BMBT_BLOCK	((uint64_t)1ULL << 62)
1434 #define XFS_RMAP_OFF_UNWRITTEN	((uint64_t)1ULL << 61)
1435 
1436 #define XFS_RMAP_LEN_MAX	((uint32_t)~0U)
1437 #define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1438 				 XFS_RMAP_OFF_BMBT_BLOCK | \
1439 				 XFS_RMAP_OFF_UNWRITTEN)
1440 #define XFS_RMAP_OFF_MASK	((uint64_t)0x3FFFFFFFFFFFFFULL)
1441 
1442 #define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1443 
1444 #define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1445 #define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1446 #define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1447 
1448 #define RMAPBT_STARTBLOCK_BITLEN	32
1449 #define RMAPBT_BLOCKCOUNT_BITLEN	32
1450 #define RMAPBT_OWNER_BITLEN		64
1451 #define RMAPBT_ATTRFLAG_BITLEN		1
1452 #define RMAPBT_BMBTFLAG_BITLEN		1
1453 #define RMAPBT_EXNTFLAG_BITLEN		1
1454 #define RMAPBT_UNUSED_OFFSET_BITLEN	7
1455 #define RMAPBT_OFFSET_BITLEN		54
1456 
1457 #define XFS_RMAP_ATTR_FORK		(1 << 0)
1458 #define XFS_RMAP_BMBT_BLOCK		(1 << 1)
1459 #define XFS_RMAP_UNWRITTEN		(1 << 2)
1460 #define XFS_RMAP_KEY_FLAGS		(XFS_RMAP_ATTR_FORK | \
1461 					 XFS_RMAP_BMBT_BLOCK)
1462 #define XFS_RMAP_REC_FLAGS		(XFS_RMAP_UNWRITTEN)
1463 struct xfs_rmap_irec {
1464 	xfs_agblock_t	rm_startblock;	/* extent start block */
1465 	xfs_extlen_t	rm_blockcount;	/* extent length */
1466 	uint64_t	rm_owner;	/* extent owner */
1467 	uint64_t	rm_offset;	/* offset within the owner */
1468 	unsigned int	rm_flags;	/* state flags */
1469 };
1470 
1471 /*
1472  * Key structure
1473  *
1474  * We don't use the length for lookups
1475  */
1476 struct xfs_rmap_key {
1477 	__be32		rm_startblock;	/* extent start block */
1478 	__be64		rm_owner;	/* extent owner */
1479 	__be64		rm_offset;	/* offset within the owner */
1480 } __attribute__((packed));
1481 
1482 /* btree pointer type */
1483 typedef __be32 xfs_rmap_ptr_t;
1484 
1485 #define	XFS_RMAP_BLOCK(mp) \
1486 	(xfs_sb_version_hasfinobt(&((mp)->m_sb)) ? \
1487 	 XFS_FIBT_BLOCK(mp) + 1 : \
1488 	 XFS_IBT_BLOCK(mp) + 1)
1489 
1490 /*
1491  * Reference Count Btree format definitions
1492  *
1493  */
1494 #define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1495 
1496 unsigned int xfs_refc_block(struct xfs_mount *mp);
1497 
1498 /*
1499  * Data record/key structure
1500  *
1501  * Each record associates a range of physical blocks (starting at
1502  * rc_startblock and ending rc_blockcount blocks later) with a reference
1503  * count (rc_refcount).  Extents that are being used to stage a copy on
1504  * write (CoW) operation are recorded in the refcount btree with a
1505  * refcount of 1.  All other records must have a refcount > 1 and must
1506  * track an extent mapped only by file data forks.
1507  *
1508  * Extents with a single owner (attributes, metadata, non-shared file
1509  * data) are not tracked here.  Free space is also not tracked here.
1510  * This is consistent with pre-reflink XFS.
1511  */
1512 
1513 /*
1514  * Extents that are being used to stage a copy on write are stored
1515  * in the refcount btree with a refcount of 1 and the upper bit set
1516  * on the startblock.  This speeds up mount time deletion of stale
1517  * staging extents because they're all at the right side of the tree.
1518  */
1519 #define XFS_REFC_COW_START		((xfs_agblock_t)(1U << 31))
1520 #define REFCNTBT_COWFLAG_BITLEN		1
1521 #define REFCNTBT_AGBLOCK_BITLEN		31
1522 
1523 struct xfs_refcount_rec {
1524 	__be32		rc_startblock;	/* starting block number */
1525 	__be32		rc_blockcount;	/* count of blocks */
1526 	__be32		rc_refcount;	/* number of inodes linked here */
1527 };
1528 
1529 struct xfs_refcount_key {
1530 	__be32		rc_startblock;	/* starting block number */
1531 };
1532 
1533 struct xfs_refcount_irec {
1534 	xfs_agblock_t	rc_startblock;	/* starting block number */
1535 	xfs_extlen_t	rc_blockcount;	/* count of free blocks */
1536 	xfs_nlink_t	rc_refcount;	/* number of inodes linked here */
1537 };
1538 
1539 #define MAXREFCOUNT	((xfs_nlink_t)~0U)
1540 #define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
1541 
1542 /* btree pointer type */
1543 typedef __be32 xfs_refcount_ptr_t;
1544 
1545 
1546 /*
1547  * BMAP Btree format definitions
1548  *
1549  * This includes both the root block definition that sits inside an inode fork
1550  * and the record/pointer formats for the leaf/node in the blocks.
1551  */
1552 #define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1553 #define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1554 
1555 /*
1556  * Bmap root header, on-disk form only.
1557  */
1558 typedef struct xfs_bmdr_block {
1559 	__be16		bb_level;	/* 0 is a leaf */
1560 	__be16		bb_numrecs;	/* current # of data records */
1561 } xfs_bmdr_block_t;
1562 
1563 /*
1564  * Bmap btree record and extent descriptor.
1565  *  l0:63 is an extent flag (value 1 indicates non-normal).
1566  *  l0:9-62 are startoff.
1567  *  l0:0-8 and l1:21-63 are startblock.
1568  *  l1:0-20 are blockcount.
1569  */
1570 #define BMBT_EXNTFLAG_BITLEN	1
1571 #define BMBT_STARTOFF_BITLEN	54
1572 #define BMBT_STARTBLOCK_BITLEN	52
1573 #define BMBT_BLOCKCOUNT_BITLEN	21
1574 
1575 #define BMBT_STARTOFF_MASK	((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1576 #define BMBT_BLOCKCOUNT_MASK	((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1)
1577 
1578 /*
1579  * bmbt records have a file offset (block) field that is 54 bits wide, so this
1580  * is the largest xfs_fileoff_t that we ever expect to see.
1581  */
1582 #define XFS_MAX_FILEOFF		(BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK)
1583 
1584 typedef struct xfs_bmbt_rec {
1585 	__be64			l0, l1;
1586 } xfs_bmbt_rec_t;
1587 
1588 typedef uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1589 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1590 
1591 /*
1592  * Values and macros for delayed-allocation startblock fields.
1593  */
1594 #define STARTBLOCKVALBITS	17
1595 #define STARTBLOCKMASKBITS	(15 + 20)
1596 #define STARTBLOCKMASK		\
1597 	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1598 
1599 static inline int isnullstartblock(xfs_fsblock_t x)
1600 {
1601 	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1602 }
1603 
1604 static inline xfs_fsblock_t nullstartblock(int k)
1605 {
1606 	ASSERT(k < (1 << STARTBLOCKVALBITS));
1607 	return STARTBLOCKMASK | (k);
1608 }
1609 
1610 static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1611 {
1612 	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1613 }
1614 
1615 /*
1616  * Key structure for non-leaf levels of the tree.
1617  */
1618 typedef struct xfs_bmbt_key {
1619 	__be64		br_startoff;	/* starting file offset */
1620 } xfs_bmbt_key_t, xfs_bmdr_key_t;
1621 
1622 /* btree pointer type */
1623 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1624 
1625 
1626 /*
1627  * Generic Btree block format definitions
1628  *
1629  * This is a combination of the actual format used on disk for short and long
1630  * format btrees.  The first three fields are shared by both format, but the
1631  * pointers are different and should be used with care.
1632  *
1633  * To get the size of the actual short or long form headers please use the size
1634  * macros below.  Never use sizeof(xfs_btree_block).
1635  *
1636  * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1637  * with the crc feature bit, and all accesses to them must be conditional on
1638  * that flag.
1639  */
1640 /* short form block header */
1641 struct xfs_btree_block_shdr {
1642 	__be32		bb_leftsib;
1643 	__be32		bb_rightsib;
1644 
1645 	__be64		bb_blkno;
1646 	__be64		bb_lsn;
1647 	uuid_t		bb_uuid;
1648 	__be32		bb_owner;
1649 	__le32		bb_crc;
1650 };
1651 
1652 /* long form block header */
1653 struct xfs_btree_block_lhdr {
1654 	__be64		bb_leftsib;
1655 	__be64		bb_rightsib;
1656 
1657 	__be64		bb_blkno;
1658 	__be64		bb_lsn;
1659 	uuid_t		bb_uuid;
1660 	__be64		bb_owner;
1661 	__le32		bb_crc;
1662 	__be32		bb_pad; /* padding for alignment */
1663 };
1664 
1665 struct xfs_btree_block {
1666 	__be32		bb_magic;	/* magic number for block type */
1667 	__be16		bb_level;	/* 0 is a leaf */
1668 	__be16		bb_numrecs;	/* current # of data records */
1669 	union {
1670 		struct xfs_btree_block_shdr s;
1671 		struct xfs_btree_block_lhdr l;
1672 	} bb_u;				/* rest */
1673 };
1674 
1675 /* size of a short form block */
1676 #define XFS_BTREE_SBLOCK_LEN \
1677 	(offsetof(struct xfs_btree_block, bb_u) + \
1678 	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1679 /* size of a long form block */
1680 #define XFS_BTREE_LBLOCK_LEN \
1681 	(offsetof(struct xfs_btree_block, bb_u) + \
1682 	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1683 
1684 /* sizes of CRC enabled btree blocks */
1685 #define XFS_BTREE_SBLOCK_CRC_LEN \
1686 	(offsetof(struct xfs_btree_block, bb_u) + \
1687 	 sizeof(struct xfs_btree_block_shdr))
1688 #define XFS_BTREE_LBLOCK_CRC_LEN \
1689 	(offsetof(struct xfs_btree_block, bb_u) + \
1690 	 sizeof(struct xfs_btree_block_lhdr))
1691 
1692 #define XFS_BTREE_SBLOCK_CRC_OFF \
1693 	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1694 #define XFS_BTREE_LBLOCK_CRC_OFF \
1695 	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1696 
1697 /*
1698  * On-disk XFS access control list structure.
1699  */
1700 struct xfs_acl_entry {
1701 	__be32	ae_tag;
1702 	__be32	ae_id;
1703 	__be16	ae_perm;
1704 	__be16	ae_pad;		/* fill the implicit hole in the structure */
1705 };
1706 
1707 struct xfs_acl {
1708 	__be32			acl_cnt;
1709 	struct xfs_acl_entry	acl_entry[];
1710 };
1711 
1712 /*
1713  * The number of ACL entries allowed is defined by the on-disk format.
1714  * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1715  * limited only by the maximum size of the xattr that stores the information.
1716  */
1717 #define XFS_ACL_MAX_ENTRIES(mp)	\
1718 	(xfs_sb_version_hascrc(&mp->m_sb) \
1719 		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1720 						sizeof(struct xfs_acl_entry) \
1721 		: 25)
1722 
1723 #define XFS_ACL_SIZE(cnt) \
1724 	(sizeof(struct xfs_acl) + \
1725 		sizeof(struct xfs_acl_entry) * cnt)
1726 
1727 #define XFS_ACL_MAX_SIZE(mp) \
1728 	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1729 
1730 
1731 /* On-disk XFS extended attribute names */
1732 #define SGI_ACL_FILE		"SGI_ACL_FILE"
1733 #define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
1734 #define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
1735 #define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
1736 
1737 #endif /* __XFS_FORMAT_H__ */
1738