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