xref: /openbmc/linux/fs/xfs/libxfs/xfs_format.h (revision 6396bb221514d2876fd6dc0aa2a1f240d99b37bb)
1 /*
2  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU General Public License as
7  * published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope that it would be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12  * GNU General Public License for more details.
13  *
14  * You should have received a copy of the GNU General Public License
15  * along with this program; if not, write the Free Software Foundation,
16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17  */
18 #ifndef __XFS_FORMAT_H__
19 #define __XFS_FORMAT_H__
20 
21 /*
22  * XFS On Disk Format Definitions
23  *
24  * This header file defines all the on-disk format definitions for
25  * general XFS objects. Directory and attribute related objects are defined in
26  * xfs_da_format.h, which log and log item formats are defined in
27  * xfs_log_format.h. Everything else goes here.
28  */
29 
30 struct xfs_mount;
31 struct xfs_trans;
32 struct xfs_inode;
33 struct xfs_buf;
34 struct xfs_ifork;
35 
36 /*
37  * Super block
38  * Fits into a sector-sized buffer at address 0 of each allocation group.
39  * Only the first of these is ever updated except during growfs.
40  */
41 #define	XFS_SB_MAGIC		0x58465342	/* 'XFSB' */
42 #define	XFS_SB_VERSION_1	1		/* 5.3, 6.0.1, 6.1 */
43 #define	XFS_SB_VERSION_2	2		/* 6.2 - attributes */
44 #define	XFS_SB_VERSION_3	3		/* 6.2 - new inode version */
45 #define	XFS_SB_VERSION_4	4		/* 6.2+ - bitmask version */
46 #define	XFS_SB_VERSION_5	5		/* CRC enabled filesystem */
47 #define	XFS_SB_VERSION_NUMBITS		0x000f
48 #define	XFS_SB_VERSION_ALLFBITS		0xfff0
49 #define	XFS_SB_VERSION_ATTRBIT		0x0010
50 #define	XFS_SB_VERSION_NLINKBIT		0x0020
51 #define	XFS_SB_VERSION_QUOTABIT		0x0040
52 #define	XFS_SB_VERSION_ALIGNBIT		0x0080
53 #define	XFS_SB_VERSION_DALIGNBIT	0x0100
54 #define	XFS_SB_VERSION_SHAREDBIT	0x0200
55 #define XFS_SB_VERSION_LOGV2BIT		0x0400
56 #define XFS_SB_VERSION_SECTORBIT	0x0800
57 #define	XFS_SB_VERSION_EXTFLGBIT	0x1000
58 #define	XFS_SB_VERSION_DIRV2BIT		0x2000
59 #define	XFS_SB_VERSION_BORGBIT		0x4000	/* ASCII only case-insens. */
60 #define	XFS_SB_VERSION_MOREBITSBIT	0x8000
61 
62 /*
63  * The size of a single extended attribute on disk is limited by
64  * the size of index values within the attribute entries themselves.
65  * These are be16 fields, so we can only support attribute data
66  * sizes up to 2^16 bytes in length.
67  */
68 #define XFS_XATTR_SIZE_MAX (1 << 16)
69 
70 /*
71  * Supported feature bit list is just all bits in the versionnum field because
72  * we've used them all up and understand them all. Except, of course, for the
73  * shared superblock bit, which nobody knows what it does and so is unsupported.
74  */
75 #define	XFS_SB_VERSION_OKBITS		\
76 	((XFS_SB_VERSION_NUMBITS | XFS_SB_VERSION_ALLFBITS) & \
77 		~XFS_SB_VERSION_SHAREDBIT)
78 
79 /*
80  * There are two words to hold XFS "feature" bits: the original
81  * word, sb_versionnum, and sb_features2.  Whenever a bit is set in
82  * sb_features2, the feature bit XFS_SB_VERSION_MOREBITSBIT must be set.
83  *
84  * These defines represent bits in sb_features2.
85  */
86 #define XFS_SB_VERSION2_RESERVED1BIT	0x00000001
87 #define XFS_SB_VERSION2_LAZYSBCOUNTBIT	0x00000002	/* Superblk counters */
88 #define XFS_SB_VERSION2_RESERVED4BIT	0x00000004
89 #define XFS_SB_VERSION2_ATTR2BIT	0x00000008	/* Inline attr rework */
90 #define XFS_SB_VERSION2_PARENTBIT	0x00000010	/* parent pointers */
91 #define XFS_SB_VERSION2_PROJID32BIT	0x00000080	/* 32 bit project id */
92 #define XFS_SB_VERSION2_CRCBIT		0x00000100	/* metadata CRCs */
93 #define XFS_SB_VERSION2_FTYPE		0x00000200	/* inode type in dir */
94 
95 #define	XFS_SB_VERSION2_OKBITS		\
96 	(XFS_SB_VERSION2_LAZYSBCOUNTBIT	| \
97 	 XFS_SB_VERSION2_ATTR2BIT	| \
98 	 XFS_SB_VERSION2_PROJID32BIT	| \
99 	 XFS_SB_VERSION2_FTYPE)
100 
101 /* Maximum size of the xfs filesystem label, no terminating NULL */
102 #define XFSLABEL_MAX			12
103 
104 /*
105  * Superblock - in core version.  Must match the ondisk version below.
106  * Must be padded to 64 bit alignment.
107  */
108 typedef struct xfs_sb {
109 	uint32_t	sb_magicnum;	/* magic number == XFS_SB_MAGIC */
110 	uint32_t	sb_blocksize;	/* logical block size, bytes */
111 	xfs_rfsblock_t	sb_dblocks;	/* number of data blocks */
112 	xfs_rfsblock_t	sb_rblocks;	/* number of realtime blocks */
113 	xfs_rtblock_t	sb_rextents;	/* number of realtime extents */
114 	uuid_t		sb_uuid;	/* user-visible file system unique id */
115 	xfs_fsblock_t	sb_logstart;	/* starting block of log if internal */
116 	xfs_ino_t	sb_rootino;	/* root inode number */
117 	xfs_ino_t	sb_rbmino;	/* bitmap inode for realtime extents */
118 	xfs_ino_t	sb_rsumino;	/* summary inode for rt bitmap */
119 	xfs_agblock_t	sb_rextsize;	/* realtime extent size, blocks */
120 	xfs_agblock_t	sb_agblocks;	/* size of an allocation group */
121 	xfs_agnumber_t	sb_agcount;	/* number of allocation groups */
122 	xfs_extlen_t	sb_rbmblocks;	/* number of rt bitmap blocks */
123 	xfs_extlen_t	sb_logblocks;	/* number of log blocks */
124 	uint16_t	sb_versionnum;	/* header version == XFS_SB_VERSION */
125 	uint16_t	sb_sectsize;	/* volume sector size, bytes */
126 	uint16_t	sb_inodesize;	/* inode size, bytes */
127 	uint16_t	sb_inopblock;	/* inodes per block */
128 	char		sb_fname[XFSLABEL_MAX]; /* file system name */
129 	uint8_t		sb_blocklog;	/* log2 of sb_blocksize */
130 	uint8_t		sb_sectlog;	/* log2 of sb_sectsize */
131 	uint8_t		sb_inodelog;	/* log2 of sb_inodesize */
132 	uint8_t		sb_inopblog;	/* log2 of sb_inopblock */
133 	uint8_t		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
134 	uint8_t		sb_rextslog;	/* log2 of sb_rextents */
135 	uint8_t		sb_inprogress;	/* mkfs is in progress, don't mount */
136 	uint8_t		sb_imax_pct;	/* max % of fs for inode space */
137 					/* statistics */
138 	/*
139 	 * These fields must remain contiguous.  If you really
140 	 * want to change their layout, make sure you fix the
141 	 * code in xfs_trans_apply_sb_deltas().
142 	 */
143 	uint64_t	sb_icount;	/* allocated inodes */
144 	uint64_t	sb_ifree;	/* free inodes */
145 	uint64_t	sb_fdblocks;	/* free data blocks */
146 	uint64_t	sb_frextents;	/* free realtime extents */
147 	/*
148 	 * End contiguous fields.
149 	 */
150 	xfs_ino_t	sb_uquotino;	/* user quota inode */
151 	xfs_ino_t	sb_gquotino;	/* group quota inode */
152 	uint16_t	sb_qflags;	/* quota flags */
153 	uint8_t		sb_flags;	/* misc. flags */
154 	uint8_t		sb_shared_vn;	/* shared version number */
155 	xfs_extlen_t	sb_inoalignmt;	/* inode chunk alignment, fsblocks */
156 	uint32_t	sb_unit;	/* stripe or raid unit */
157 	uint32_t	sb_width;	/* stripe or raid width */
158 	uint8_t		sb_dirblklog;	/* log2 of dir block size (fsbs) */
159 	uint8_t		sb_logsectlog;	/* log2 of the log sector size */
160 	uint16_t	sb_logsectsize;	/* sector size for the log, bytes */
161 	uint32_t	sb_logsunit;	/* stripe unit size for the log */
162 	uint32_t	sb_features2;	/* additional feature bits */
163 
164 	/*
165 	 * bad features2 field as a result of failing to pad the sb structure to
166 	 * 64 bits. Some machines will be using this field for features2 bits.
167 	 * Easiest just to mark it bad and not use it for anything else.
168 	 *
169 	 * This is not kept up to date in memory; it is always overwritten by
170 	 * the value in sb_features2 when formatting the incore superblock to
171 	 * the disk buffer.
172 	 */
173 	uint32_t	sb_bad_features2;
174 
175 	/* version 5 superblock fields start here */
176 
177 	/* feature masks */
178 	uint32_t	sb_features_compat;
179 	uint32_t	sb_features_ro_compat;
180 	uint32_t	sb_features_incompat;
181 	uint32_t	sb_features_log_incompat;
182 
183 	uint32_t	sb_crc;		/* superblock crc */
184 	xfs_extlen_t	sb_spino_align;	/* sparse inode chunk alignment */
185 
186 	xfs_ino_t	sb_pquotino;	/* project quota inode */
187 	xfs_lsn_t	sb_lsn;		/* last write sequence */
188 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
189 
190 	/* must be padded to 64 bit alignment */
191 } xfs_sb_t;
192 
193 #define XFS_SB_CRC_OFF		offsetof(struct xfs_sb, sb_crc)
194 
195 /*
196  * Superblock - on disk version.  Must match the in core version above.
197  * Must be padded to 64 bit alignment.
198  */
199 typedef struct xfs_dsb {
200 	__be32		sb_magicnum;	/* magic number == XFS_SB_MAGIC */
201 	__be32		sb_blocksize;	/* logical block size, bytes */
202 	__be64		sb_dblocks;	/* number of data blocks */
203 	__be64		sb_rblocks;	/* number of realtime blocks */
204 	__be64		sb_rextents;	/* number of realtime extents */
205 	uuid_t		sb_uuid;	/* user-visible file system unique id */
206 	__be64		sb_logstart;	/* starting block of log if internal */
207 	__be64		sb_rootino;	/* root inode number */
208 	__be64		sb_rbmino;	/* bitmap inode for realtime extents */
209 	__be64		sb_rsumino;	/* summary inode for rt bitmap */
210 	__be32		sb_rextsize;	/* realtime extent size, blocks */
211 	__be32		sb_agblocks;	/* size of an allocation group */
212 	__be32		sb_agcount;	/* number of allocation groups */
213 	__be32		sb_rbmblocks;	/* number of rt bitmap blocks */
214 	__be32		sb_logblocks;	/* number of log blocks */
215 	__be16		sb_versionnum;	/* header version == XFS_SB_VERSION */
216 	__be16		sb_sectsize;	/* volume sector size, bytes */
217 	__be16		sb_inodesize;	/* inode size, bytes */
218 	__be16		sb_inopblock;	/* inodes per block */
219 	char		sb_fname[XFSLABEL_MAX]; /* file system name */
220 	__u8		sb_blocklog;	/* log2 of sb_blocksize */
221 	__u8		sb_sectlog;	/* log2 of sb_sectsize */
222 	__u8		sb_inodelog;	/* log2 of sb_inodesize */
223 	__u8		sb_inopblog;	/* log2 of sb_inopblock */
224 	__u8		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
225 	__u8		sb_rextslog;	/* log2 of sb_rextents */
226 	__u8		sb_inprogress;	/* mkfs is in progress, don't mount */
227 	__u8		sb_imax_pct;	/* max % of fs for inode space */
228 					/* statistics */
229 	/*
230 	 * These fields must remain contiguous.  If you really
231 	 * want to change their layout, make sure you fix the
232 	 * code in xfs_trans_apply_sb_deltas().
233 	 */
234 	__be64		sb_icount;	/* allocated inodes */
235 	__be64		sb_ifree;	/* free inodes */
236 	__be64		sb_fdblocks;	/* free data blocks */
237 	__be64		sb_frextents;	/* free realtime extents */
238 	/*
239 	 * End contiguous fields.
240 	 */
241 	__be64		sb_uquotino;	/* user quota inode */
242 	__be64		sb_gquotino;	/* group quota inode */
243 	__be16		sb_qflags;	/* quota flags */
244 	__u8		sb_flags;	/* misc. flags */
245 	__u8		sb_shared_vn;	/* shared version number */
246 	__be32		sb_inoalignmt;	/* inode chunk alignment, fsblocks */
247 	__be32		sb_unit;	/* stripe or raid unit */
248 	__be32		sb_width;	/* stripe or raid width */
249 	__u8		sb_dirblklog;	/* log2 of dir block size (fsbs) */
250 	__u8		sb_logsectlog;	/* log2 of the log sector size */
251 	__be16		sb_logsectsize;	/* sector size for the log, bytes */
252 	__be32		sb_logsunit;	/* stripe unit size for the log */
253 	__be32		sb_features2;	/* additional feature bits */
254 	/*
255 	 * bad features2 field as a result of failing to pad the sb
256 	 * structure to 64 bits. Some machines will be using this field
257 	 * for features2 bits. Easiest just to mark it bad and not use
258 	 * it for anything else.
259 	 */
260 	__be32		sb_bad_features2;
261 
262 	/* version 5 superblock fields start here */
263 
264 	/* feature masks */
265 	__be32		sb_features_compat;
266 	__be32		sb_features_ro_compat;
267 	__be32		sb_features_incompat;
268 	__be32		sb_features_log_incompat;
269 
270 	__le32		sb_crc;		/* superblock crc */
271 	__be32		sb_spino_align;	/* sparse inode chunk alignment */
272 
273 	__be64		sb_pquotino;	/* project quota inode */
274 	__be64		sb_lsn;		/* last write sequence */
275 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
276 
277 	/* must be padded to 64 bit alignment */
278 } xfs_dsb_t;
279 
280 
281 /*
282  * Misc. Flags - warning - these will be cleared by xfs_repair unless
283  * a feature bit is set when the flag is used.
284  */
285 #define XFS_SBF_NOFLAGS		0x00	/* no flags set */
286 #define XFS_SBF_READONLY	0x01	/* only read-only mounts allowed */
287 
288 /*
289  * define max. shared version we can interoperate with
290  */
291 #define XFS_SB_MAX_SHARED_VN	0
292 
293 #define	XFS_SB_VERSION_NUM(sbp)	((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
294 
295 /*
296  * The first XFS version we support is a v4 superblock with V2 directories.
297  */
298 static inline bool xfs_sb_good_v4_features(struct xfs_sb *sbp)
299 {
300 	if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
301 		return false;
302 
303 	/* check for unknown features in the fs */
304 	if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
305 	    ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
306 	     (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
307 		return false;
308 
309 	return true;
310 }
311 
312 static inline bool xfs_sb_good_version(struct xfs_sb *sbp)
313 {
314 	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5)
315 		return true;
316 	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_4)
317 		return xfs_sb_good_v4_features(sbp);
318 	return false;
319 }
320 
321 static inline bool xfs_sb_version_hasrealtime(struct xfs_sb *sbp)
322 {
323 	return sbp->sb_rblocks > 0;
324 }
325 
326 /*
327  * Detect a mismatched features2 field.  Older kernels read/wrote
328  * this into the wrong slot, so to be safe we keep them in sync.
329  */
330 static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
331 {
332 	return sbp->sb_bad_features2 != sbp->sb_features2;
333 }
334 
335 static inline bool xfs_sb_version_hasattr(struct xfs_sb *sbp)
336 {
337 	return (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT);
338 }
339 
340 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
341 {
342 	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
343 }
344 
345 static inline bool xfs_sb_version_hasquota(struct xfs_sb *sbp)
346 {
347 	return (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT);
348 }
349 
350 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
351 {
352 	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
353 }
354 
355 static inline bool xfs_sb_version_hasalign(struct xfs_sb *sbp)
356 {
357 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
358 		(sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT));
359 }
360 
361 static inline bool xfs_sb_version_hasdalign(struct xfs_sb *sbp)
362 {
363 	return (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT);
364 }
365 
366 static inline bool xfs_sb_version_haslogv2(struct xfs_sb *sbp)
367 {
368 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
369 	       (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT);
370 }
371 
372 static inline bool xfs_sb_version_hasextflgbit(struct xfs_sb *sbp)
373 {
374 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
375 	       (sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT);
376 }
377 
378 static inline bool xfs_sb_version_hassector(struct xfs_sb *sbp)
379 {
380 	return (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT);
381 }
382 
383 static inline bool xfs_sb_version_hasasciici(struct xfs_sb *sbp)
384 {
385 	return (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT);
386 }
387 
388 static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp)
389 {
390 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
391 	       (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
392 }
393 
394 /*
395  * sb_features2 bit version macros.
396  */
397 static inline bool xfs_sb_version_haslazysbcount(struct xfs_sb *sbp)
398 {
399 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
400 	       (xfs_sb_version_hasmorebits(sbp) &&
401 		(sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT));
402 }
403 
404 static inline bool xfs_sb_version_hasattr2(struct xfs_sb *sbp)
405 {
406 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
407 	       (xfs_sb_version_hasmorebits(sbp) &&
408 		(sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT));
409 }
410 
411 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
412 {
413 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
414 	sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
415 }
416 
417 static inline void xfs_sb_version_removeattr2(struct xfs_sb *sbp)
418 {
419 	sbp->sb_features2 &= ~XFS_SB_VERSION2_ATTR2BIT;
420 	if (!sbp->sb_features2)
421 		sbp->sb_versionnum &= ~XFS_SB_VERSION_MOREBITSBIT;
422 }
423 
424 static inline bool xfs_sb_version_hasprojid32bit(struct xfs_sb *sbp)
425 {
426 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
427 	       (xfs_sb_version_hasmorebits(sbp) &&
428 		(sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT));
429 }
430 
431 static inline void xfs_sb_version_addprojid32bit(struct xfs_sb *sbp)
432 {
433 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
434 	sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
435 }
436 
437 /*
438  * Extended v5 superblock feature masks. These are to be used for new v5
439  * superblock features only.
440  *
441  * Compat features are new features that old kernels will not notice or affect
442  * and so can mount read-write without issues.
443  *
444  * RO-Compat (read only) are features that old kernels can read but will break
445  * if they write. Hence only read-only mounts of such filesystems are allowed on
446  * kernels that don't support the feature bit.
447  *
448  * InCompat features are features which old kernels will not understand and so
449  * must not mount.
450  *
451  * Log-InCompat features are for changes to log formats or new transactions that
452  * can't be replayed on older kernels. The fields are set when the filesystem is
453  * mounted, and a clean unmount clears the fields.
454  */
455 #define XFS_SB_FEAT_COMPAT_ALL 0
456 #define XFS_SB_FEAT_COMPAT_UNKNOWN	~XFS_SB_FEAT_COMPAT_ALL
457 static inline bool
458 xfs_sb_has_compat_feature(
459 	struct xfs_sb	*sbp,
460 	uint32_t	feature)
461 {
462 	return (sbp->sb_features_compat & feature) != 0;
463 }
464 
465 #define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
466 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
467 #define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
468 #define XFS_SB_FEAT_RO_COMPAT_ALL \
469 		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
470 		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
471 		 XFS_SB_FEAT_RO_COMPAT_REFLINK)
472 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
473 static inline bool
474 xfs_sb_has_ro_compat_feature(
475 	struct xfs_sb	*sbp,
476 	uint32_t	feature)
477 {
478 	return (sbp->sb_features_ro_compat & feature) != 0;
479 }
480 
481 #define XFS_SB_FEAT_INCOMPAT_FTYPE	(1 << 0)	/* filetype in dirent */
482 #define XFS_SB_FEAT_INCOMPAT_SPINODES	(1 << 1)	/* sparse inode chunks */
483 #define XFS_SB_FEAT_INCOMPAT_META_UUID	(1 << 2)	/* metadata UUID */
484 #define XFS_SB_FEAT_INCOMPAT_ALL \
485 		(XFS_SB_FEAT_INCOMPAT_FTYPE|	\
486 		 XFS_SB_FEAT_INCOMPAT_SPINODES|	\
487 		 XFS_SB_FEAT_INCOMPAT_META_UUID)
488 
489 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
490 static inline bool
491 xfs_sb_has_incompat_feature(
492 	struct xfs_sb	*sbp,
493 	uint32_t	feature)
494 {
495 	return (sbp->sb_features_incompat & feature) != 0;
496 }
497 
498 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL 0
499 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
500 static inline bool
501 xfs_sb_has_incompat_log_feature(
502 	struct xfs_sb	*sbp,
503 	uint32_t	feature)
504 {
505 	return (sbp->sb_features_log_incompat & feature) != 0;
506 }
507 
508 /*
509  * V5 superblock specific feature checks
510  */
511 static inline bool xfs_sb_version_hascrc(struct xfs_sb *sbp)
512 {
513 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
514 }
515 
516 static inline bool xfs_sb_version_has_pquotino(struct xfs_sb *sbp)
517 {
518 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
519 }
520 
521 static inline int xfs_sb_version_hasftype(struct xfs_sb *sbp)
522 {
523 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
524 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_FTYPE)) ||
525 	       (xfs_sb_version_hasmorebits(sbp) &&
526 		 (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE));
527 }
528 
529 static inline bool xfs_sb_version_hasfinobt(xfs_sb_t *sbp)
530 {
531 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
532 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT);
533 }
534 
535 static inline bool xfs_sb_version_hassparseinodes(struct xfs_sb *sbp)
536 {
537 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
538 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_SPINODES);
539 }
540 
541 /*
542  * XFS_SB_FEAT_INCOMPAT_META_UUID indicates that the metadata UUID
543  * is stored separately from the user-visible UUID; this allows the
544  * user-visible UUID to be changed on V5 filesystems which have a
545  * filesystem UUID stamped into every piece of metadata.
546  */
547 static inline bool xfs_sb_version_hasmetauuid(struct xfs_sb *sbp)
548 {
549 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
550 		(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID);
551 }
552 
553 static inline bool xfs_sb_version_hasrmapbt(struct xfs_sb *sbp)
554 {
555 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
556 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT);
557 }
558 
559 static inline bool xfs_sb_version_hasreflink(struct xfs_sb *sbp)
560 {
561 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
562 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK);
563 }
564 
565 /*
566  * end of superblock version macros
567  */
568 
569 static inline bool
570 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
571 {
572 	return (ino == sbp->sb_uquotino ||
573 		ino == sbp->sb_gquotino ||
574 		ino == sbp->sb_pquotino);
575 }
576 
577 #define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
578 #define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
579 #define XFS_BUF_TO_SBP(bp)	((xfs_dsb_t *)((bp)->b_addr))
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 #define	XFS_BUF_TO_AGF(bp)	((xfs_agf_t *)((bp)->b_addr))
727 
728 /*
729  * Size of the unlinked inode hash table in the agi.
730  */
731 #define	XFS_AGI_UNLINKED_BUCKETS	64
732 
733 typedef struct xfs_agi {
734 	/*
735 	 * Common allocation group header information
736 	 */
737 	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
738 	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
739 	__be32		agi_seqno;	/* sequence # starting from 0 */
740 	__be32		agi_length;	/* size in blocks of a.g. */
741 	/*
742 	 * Inode information
743 	 * Inodes are mapped by interpreting the inode number, so no
744 	 * mapping data is needed here.
745 	 */
746 	__be32		agi_count;	/* count of allocated inodes */
747 	__be32		agi_root;	/* root of inode btree */
748 	__be32		agi_level;	/* levels in inode btree */
749 	__be32		agi_freecount;	/* number of free inodes */
750 
751 	__be32		agi_newino;	/* new inode just allocated */
752 	__be32		agi_dirino;	/* last directory inode chunk */
753 	/*
754 	 * Hash table of inodes which have been unlinked but are
755 	 * still being referenced.
756 	 */
757 	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
758 	/*
759 	 * This marks the end of logging region 1 and start of logging region 2.
760 	 */
761 	uuid_t		agi_uuid;	/* uuid of filesystem */
762 	__be32		agi_crc;	/* crc of agi sector */
763 	__be32		agi_pad32;
764 	__be64		agi_lsn;	/* last write sequence */
765 
766 	__be32		agi_free_root; /* root of the free inode btree */
767 	__be32		agi_free_level;/* levels in free inode btree */
768 
769 	/* structure must be padded to 64 bit alignment */
770 } xfs_agi_t;
771 
772 #define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
773 
774 #define	XFS_AGI_MAGICNUM	(1 << 0)
775 #define	XFS_AGI_VERSIONNUM	(1 << 1)
776 #define	XFS_AGI_SEQNO		(1 << 2)
777 #define	XFS_AGI_LENGTH		(1 << 3)
778 #define	XFS_AGI_COUNT		(1 << 4)
779 #define	XFS_AGI_ROOT		(1 << 5)
780 #define	XFS_AGI_LEVEL		(1 << 6)
781 #define	XFS_AGI_FREECOUNT	(1 << 7)
782 #define	XFS_AGI_NEWINO		(1 << 8)
783 #define	XFS_AGI_DIRINO		(1 << 9)
784 #define	XFS_AGI_UNLINKED	(1 << 10)
785 #define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
786 #define	XFS_AGI_ALL_BITS_R1	((1 << XFS_AGI_NUM_BITS_R1) - 1)
787 #define	XFS_AGI_FREE_ROOT	(1 << 11)
788 #define	XFS_AGI_FREE_LEVEL	(1 << 12)
789 #define	XFS_AGI_NUM_BITS_R2	13
790 
791 /* disk block (xfs_daddr_t) in the AG */
792 #define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
793 #define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
794 #define	XFS_BUF_TO_AGI(bp)	((xfs_agi_t *)((bp)->b_addr))
795 
796 /*
797  * The third a.g. block contains the a.g. freelist, an array
798  * of block pointers to blocks owned by the allocation btree code.
799  */
800 #define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
801 #define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
802 #define	XFS_BUF_TO_AGFL(bp)	((xfs_agfl_t *)((bp)->b_addr))
803 
804 #define XFS_BUF_TO_AGFL_BNO(mp, bp) \
805 	(xfs_sb_version_hascrc(&((mp)->m_sb)) ? \
806 		&(XFS_BUF_TO_AGFL(bp)->agfl_bno[0]) : \
807 		(__be32 *)(bp)->b_addr)
808 
809 typedef struct xfs_agfl {
810 	__be32		agfl_magicnum;
811 	__be32		agfl_seqno;
812 	uuid_t		agfl_uuid;
813 	__be64		agfl_lsn;
814 	__be32		agfl_crc;
815 	__be32		agfl_bno[];	/* actually xfs_agfl_size(mp) */
816 } __attribute__((packed)) xfs_agfl_t;
817 
818 #define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
819 
820 #define XFS_AGB_TO_FSB(mp,agno,agbno)	\
821 	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
822 #define	XFS_FSB_TO_AGNO(mp,fsbno)	\
823 	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
824 #define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
825 	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
826 #define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
827 	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
828 		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
829 #define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
830 
831 /*
832  * For checking for bad ranges of xfs_daddr_t's, covering multiple
833  * allocation groups or a single xfs_daddr_t that's a superblock copy.
834  */
835 #define	XFS_AG_CHECK_DADDR(mp,d,len)	\
836 	((len) == 1 ? \
837 	    ASSERT((d) == XFS_SB_DADDR || \
838 		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
839 	    ASSERT(xfs_daddr_to_agno(mp, d) == \
840 		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
841 
842 typedef struct xfs_timestamp {
843 	__be32		t_sec;		/* timestamp seconds */
844 	__be32		t_nsec;		/* timestamp nanoseconds */
845 } xfs_timestamp_t;
846 
847 /*
848  * On-disk inode structure.
849  *
850  * This is just the header or "dinode core", the inode is expanded to fill a
851  * variable size the leftover area split into a data and an attribute fork.
852  * The format of the data and attribute fork depends on the format of the
853  * inode as indicated by di_format and di_aformat.  To access the data and
854  * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
855  * below.
856  *
857  * There is a very similar struct icdinode in xfs_inode which matches the
858  * layout of the first 96 bytes of this structure, but is kept in native
859  * format instead of big endian.
860  *
861  * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
862  * padding field for v3 inodes.
863  */
864 #define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
865 typedef struct xfs_dinode {
866 	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
867 	__be16		di_mode;	/* mode and type of file */
868 	__u8		di_version;	/* inode version */
869 	__u8		di_format;	/* format of di_c data */
870 	__be16		di_onlink;	/* old number of links to file */
871 	__be32		di_uid;		/* owner's user id */
872 	__be32		di_gid;		/* owner's group id */
873 	__be32		di_nlink;	/* number of links to file */
874 	__be16		di_projid_lo;	/* lower part of owner's project id */
875 	__be16		di_projid_hi;	/* higher part owner's project id */
876 	__u8		di_pad[6];	/* unused, zeroed space */
877 	__be16		di_flushiter;	/* incremented on flush */
878 	xfs_timestamp_t	di_atime;	/* time last accessed */
879 	xfs_timestamp_t	di_mtime;	/* time last modified */
880 	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
881 	__be64		di_size;	/* number of bytes in file */
882 	__be64		di_nblocks;	/* # of direct & btree blocks used */
883 	__be32		di_extsize;	/* basic/minimum extent size for file */
884 	__be32		di_nextents;	/* number of extents in data fork */
885 	__be16		di_anextents;	/* number of extents in attribute fork*/
886 	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
887 	__s8		di_aformat;	/* format of attr fork's data */
888 	__be32		di_dmevmask;	/* DMIG event mask */
889 	__be16		di_dmstate;	/* DMIG state info */
890 	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
891 	__be32		di_gen;		/* generation number */
892 
893 	/* di_next_unlinked is the only non-core field in the old dinode */
894 	__be32		di_next_unlinked;/* agi unlinked list ptr */
895 
896 	/* start of the extended dinode, writable fields */
897 	__le32		di_crc;		/* CRC of the inode */
898 	__be64		di_changecount;	/* number of attribute changes */
899 	__be64		di_lsn;		/* flush sequence */
900 	__be64		di_flags2;	/* more random flags */
901 	__be32		di_cowextsize;	/* basic cow extent size for file */
902 	__u8		di_pad2[12];	/* more padding for future expansion */
903 
904 	/* fields only written to during inode creation */
905 	xfs_timestamp_t	di_crtime;	/* time created */
906 	__be64		di_ino;		/* inode number */
907 	uuid_t		di_uuid;	/* UUID of the filesystem */
908 
909 	/* structure must be padded to 64 bit alignment */
910 } xfs_dinode_t;
911 
912 #define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
913 
914 #define DI_MAX_FLUSH 0xffff
915 
916 /*
917  * Size of the core inode on disk.  Version 1 and 2 inodes have
918  * the same size, but version 3 has grown a few additional fields.
919  */
920 static inline uint xfs_dinode_size(int version)
921 {
922 	if (version == 3)
923 		return sizeof(struct xfs_dinode);
924 	return offsetof(struct xfs_dinode, di_crc);
925 }
926 
927 /*
928  * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
929  * Since the pathconf interface is signed, we use 2^31 - 1 instead.
930  */
931 #define	XFS_MAXLINK		((1U << 31) - 1U)
932 
933 /*
934  * Values for di_format
935  */
936 typedef enum xfs_dinode_fmt {
937 	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
938 	XFS_DINODE_FMT_LOCAL,		/* bulk data */
939 	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
940 	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
941 	XFS_DINODE_FMT_UUID		/* added long ago, but never used */
942 } xfs_dinode_fmt_t;
943 
944 /*
945  * Inode minimum and maximum sizes.
946  */
947 #define	XFS_DINODE_MIN_LOG	8
948 #define	XFS_DINODE_MAX_LOG	11
949 #define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
950 #define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
951 
952 /*
953  * Inode size for given fs.
954  */
955 #define XFS_LITINO(mp, version) \
956 	((int)(((mp)->m_sb.sb_inodesize) - xfs_dinode_size(version)))
957 
958 /*
959  * Inode data & attribute fork sizes, per inode.
960  */
961 #define XFS_DFORK_Q(dip)		((dip)->di_forkoff != 0)
962 #define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
963 
964 #define XFS_DFORK_DSIZE(dip,mp) \
965 	(XFS_DFORK_Q(dip) ? \
966 		XFS_DFORK_BOFF(dip) : \
967 		XFS_LITINO(mp, (dip)->di_version))
968 #define XFS_DFORK_ASIZE(dip,mp) \
969 	(XFS_DFORK_Q(dip) ? \
970 		XFS_LITINO(mp, (dip)->di_version) - XFS_DFORK_BOFF(dip) : \
971 		0)
972 #define XFS_DFORK_SIZE(dip,mp,w) \
973 	((w) == XFS_DATA_FORK ? \
974 		XFS_DFORK_DSIZE(dip, mp) : \
975 		XFS_DFORK_ASIZE(dip, mp))
976 
977 /*
978  * Return pointers to the data or attribute forks.
979  */
980 #define XFS_DFORK_DPTR(dip) \
981 	((char *)dip + xfs_dinode_size(dip->di_version))
982 #define XFS_DFORK_APTR(dip)	\
983 	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
984 #define XFS_DFORK_PTR(dip,w)	\
985 	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
986 
987 #define XFS_DFORK_FORMAT(dip,w) \
988 	((w) == XFS_DATA_FORK ? \
989 		(dip)->di_format : \
990 		(dip)->di_aformat)
991 #define XFS_DFORK_NEXTENTS(dip,w) \
992 	((w) == XFS_DATA_FORK ? \
993 		be32_to_cpu((dip)->di_nextents) : \
994 		be16_to_cpu((dip)->di_anextents))
995 
996 /*
997  * For block and character special files the 32bit dev_t is stored at the
998  * beginning of the data fork.
999  */
1000 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1001 {
1002 	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1003 }
1004 
1005 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1006 {
1007 	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1008 }
1009 
1010 /*
1011  * Values for di_flags
1012  */
1013 #define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
1014 #define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
1015 #define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
1016 #define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
1017 #define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
1018 #define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
1019 #define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
1020 #define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
1021 #define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
1022 #define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
1023 #define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
1024 #define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
1025 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
1026 #define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
1027 #define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
1028 #define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
1029 #define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
1030 #define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
1031 #define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1032 #define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
1033 #define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
1034 #define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
1035 #define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
1036 #define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
1037 #define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1038 #define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1039 #define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
1040 #define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1041 #define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
1042 #define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
1043 
1044 #define XFS_DIFLAG_ANY \
1045 	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1046 	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1047 	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1048 	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1049 	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1050 
1051 /*
1052  * Values for di_flags2 These start by being exposed to userspace in the upper
1053  * 16 bits of the XFS_XFLAG_s range.
1054  */
1055 #define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
1056 #define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
1057 #define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
1058 #define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
1059 #define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
1060 #define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1061 
1062 #define XFS_DIFLAG2_ANY \
1063 	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)
1064 
1065 /*
1066  * Inode number format:
1067  * low inopblog bits - offset in block
1068  * next agblklog bits - block number in ag
1069  * next agno_log bits - ag number
1070  * high agno_log-agblklog-inopblog bits - 0
1071  */
1072 #define	XFS_INO_MASK(k)			(uint32_t)((1ULL << (k)) - 1)
1073 #define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1074 #define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1075 #define	XFS_INO_AGINO_BITS(mp)		(mp)->m_agino_log
1076 #define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1077 #define	XFS_INO_BITS(mp)		\
1078 	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1079 #define	XFS_INO_TO_AGNO(mp,i)		\
1080 	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1081 #define	XFS_INO_TO_AGINO(mp,i)		\
1082 	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1083 #define	XFS_INO_TO_AGBNO(mp,i)		\
1084 	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1085 		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1086 #define	XFS_INO_TO_OFFSET(mp,i)		\
1087 	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1088 #define	XFS_INO_TO_FSB(mp,i)		\
1089 	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1090 #define	XFS_AGINO_TO_INO(mp,a,i)	\
1091 	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1092 #define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1093 #define	XFS_AGINO_TO_OFFSET(mp,i)	\
1094 	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1095 #define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1096 	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1097 
1098 #define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1099 #define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1100 
1101 /*
1102  * RealTime Device format definitions
1103  */
1104 
1105 /* Min and max rt extent sizes, specified in bytes */
1106 #define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1107 #define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1108 #define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1109 
1110 #define	XFS_BLOCKSIZE(mp)	((mp)->m_sb.sb_blocksize)
1111 #define	XFS_BLOCKMASK(mp)	((mp)->m_blockmask)
1112 #define	XFS_BLOCKWSIZE(mp)	((mp)->m_blockwsize)
1113 #define	XFS_BLOCKWMASK(mp)	((mp)->m_blockwmask)
1114 
1115 /*
1116  * RT Summary and bit manipulation macros.
1117  */
1118 #define	XFS_SUMOFFS(mp,ls,bb)	((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb)))
1119 #define	XFS_SUMOFFSTOBLOCK(mp,s)	\
1120 	(((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog)
1121 #define	XFS_SUMPTR(mp,bp,so)	\
1122 	((xfs_suminfo_t *)((bp)->b_addr + \
1123 		(((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp))))
1124 
1125 #define	XFS_BITTOBLOCK(mp,bi)	((bi) >> (mp)->m_blkbit_log)
1126 #define	XFS_BLOCKTOBIT(mp,bb)	((bb) << (mp)->m_blkbit_log)
1127 #define	XFS_BITTOWORD(mp,bi)	\
1128 	((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp)))
1129 
1130 #define	XFS_RTMIN(a,b)	((a) < (b) ? (a) : (b))
1131 #define	XFS_RTMAX(a,b)	((a) > (b) ? (a) : (b))
1132 
1133 #define	XFS_RTLOBIT(w)	xfs_lowbit32(w)
1134 #define	XFS_RTHIBIT(w)	xfs_highbit32(w)
1135 
1136 #define	XFS_RTBLOCKLOG(b)	xfs_highbit64(b)
1137 
1138 /*
1139  * Dquot and dquot block format definitions
1140  */
1141 #define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1142 #define XFS_DQUOT_VERSION	(uint8_t)0x01	/* latest version number */
1143 
1144 /*
1145  * This is the main portion of the on-disk representation of quota
1146  * information for a user. This is the q_core of the xfs_dquot_t that
1147  * is kept in kernel memory. We pad this with some more expansion room
1148  * to construct the on disk structure.
1149  */
1150 typedef struct	xfs_disk_dquot {
1151 	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1152 	__u8		d_version;	/* dquot version */
1153 	__u8		d_flags;	/* XFS_DQ_USER/PROJ/GROUP */
1154 	__be32		d_id;		/* user,project,group id */
1155 	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1156 	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1157 	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1158 	__be64		d_ino_softlimit;/* preferred inode limit */
1159 	__be64		d_bcount;	/* disk blocks owned by the user */
1160 	__be64		d_icount;	/* inodes owned by the user */
1161 	__be32		d_itimer;	/* zero if within inode limits if not,
1162 					   this is when we refuse service */
1163 	__be32		d_btimer;	/* similar to above; for disk blocks */
1164 	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1165 	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1166 	__be32		d_pad0;		/* 64 bit align */
1167 	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1168 	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1169 	__be64		d_rtbcount;	/* realtime blocks owned */
1170 	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1171 	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1172 	__be16		d_pad;
1173 } xfs_disk_dquot_t;
1174 
1175 /*
1176  * This is what goes on disk. This is separated from the xfs_disk_dquot because
1177  * carrying the unnecessary padding would be a waste of memory.
1178  */
1179 typedef struct xfs_dqblk {
1180 	xfs_disk_dquot_t  dd_diskdq;	/* portion that lives incore as well */
1181 	char		  dd_fill[4];	/* filling for posterity */
1182 
1183 	/*
1184 	 * These two are only present on filesystems with the CRC bits set.
1185 	 */
1186 	__be32		  dd_crc;	/* checksum */
1187 	__be64		  dd_lsn;	/* last modification in log */
1188 	uuid_t		  dd_uuid;	/* location information */
1189 } xfs_dqblk_t;
1190 
1191 #define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1192 
1193 /*
1194  * Remote symlink format and access functions.
1195  */
1196 #define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1197 
1198 struct xfs_dsymlink_hdr {
1199 	__be32	sl_magic;
1200 	__be32	sl_offset;
1201 	__be32	sl_bytes;
1202 	__be32	sl_crc;
1203 	uuid_t	sl_uuid;
1204 	__be64	sl_owner;
1205 	__be64	sl_blkno;
1206 	__be64	sl_lsn;
1207 };
1208 
1209 #define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1210 
1211 #define XFS_SYMLINK_MAXLEN	1024
1212 /*
1213  * The maximum pathlen is 1024 bytes. Since the minimum file system
1214  * blocksize is 512 bytes, we can get a max of 3 extents back from
1215  * bmapi when crc headers are taken into account.
1216  */
1217 #define XFS_SYMLINK_MAPS 3
1218 
1219 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1220 	((bufsize) - (xfs_sb_version_hascrc(&(mp)->m_sb) ? \
1221 			sizeof(struct xfs_dsymlink_hdr) : 0))
1222 
1223 
1224 /*
1225  * Allocation Btree format definitions
1226  *
1227  * There are two on-disk btrees, one sorted by blockno and one sorted
1228  * by blockcount and blockno.  All blocks look the same to make the code
1229  * simpler; if we have time later, we'll make the optimizations.
1230  */
1231 #define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1232 #define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1233 #define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1234 #define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1235 
1236 /*
1237  * Data record/key structure
1238  */
1239 typedef struct xfs_alloc_rec {
1240 	__be32		ar_startblock;	/* starting block number */
1241 	__be32		ar_blockcount;	/* count of free blocks */
1242 } xfs_alloc_rec_t, xfs_alloc_key_t;
1243 
1244 typedef struct xfs_alloc_rec_incore {
1245 	xfs_agblock_t	ar_startblock;	/* starting block number */
1246 	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1247 } xfs_alloc_rec_incore_t;
1248 
1249 /* btree pointer type */
1250 typedef __be32 xfs_alloc_ptr_t;
1251 
1252 /*
1253  * Block numbers in the AG:
1254  * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1255  */
1256 #define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1257 #define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1258 
1259 
1260 /*
1261  * Inode Allocation Btree format definitions
1262  *
1263  * There is a btree for the inode map per allocation group.
1264  */
1265 #define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1266 #define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1267 #define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1268 #define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1269 
1270 typedef uint64_t	xfs_inofree_t;
1271 #define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1272 #define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1273 #define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1274 #define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1275 
1276 #define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1277 #define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(uint16_t))
1278 #define XFS_INODES_PER_HOLEMASK_BIT	\
1279 	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1280 
1281 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1282 {
1283 	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1284 }
1285 
1286 /*
1287  * The on-disk inode record structure has two formats. The original "full"
1288  * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1289  * and replaces the 3 high-order freecount bytes wth the holemask and inode
1290  * count.
1291  *
1292  * The holemask of the sparse record format allows an inode chunk to have holes
1293  * that refer to blocks not owned by the inode record. This facilitates inode
1294  * allocation in the event of severe free space fragmentation.
1295  */
1296 typedef struct xfs_inobt_rec {
1297 	__be32		ir_startino;	/* starting inode number */
1298 	union {
1299 		struct {
1300 			__be32	ir_freecount;	/* count of free inodes */
1301 		} f;
1302 		struct {
1303 			__be16	ir_holemask;/* hole mask for sparse chunks */
1304 			__u8	ir_count;	/* total inode count */
1305 			__u8	ir_freecount;	/* count of free inodes */
1306 		} sp;
1307 	} ir_u;
1308 	__be64		ir_free;	/* free inode mask */
1309 } xfs_inobt_rec_t;
1310 
1311 typedef struct xfs_inobt_rec_incore {
1312 	xfs_agino_t	ir_startino;	/* starting inode number */
1313 	uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1314 	uint8_t		ir_count;	/* total inode count */
1315 	uint8_t		ir_freecount;	/* count of free inodes (set bits) */
1316 	xfs_inofree_t	ir_free;	/* free inode mask */
1317 } xfs_inobt_rec_incore_t;
1318 
1319 static inline bool xfs_inobt_issparse(uint16_t holemask)
1320 {
1321 	/* non-zero holemask represents a sparse rec. */
1322 	return holemask;
1323 }
1324 
1325 /*
1326  * Key structure
1327  */
1328 typedef struct xfs_inobt_key {
1329 	__be32		ir_startino;	/* starting inode number */
1330 } xfs_inobt_key_t;
1331 
1332 /* btree pointer type */
1333 typedef __be32 xfs_inobt_ptr_t;
1334 
1335 /*
1336  * block numbers in the AG.
1337  */
1338 #define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1339 #define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1340 
1341 /*
1342  * Reverse mapping btree format definitions
1343  *
1344  * There is a btree for the reverse map per allocation group
1345  */
1346 #define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1347 
1348 /*
1349  * Ownership info for an extent.  This is used to create reverse-mapping
1350  * entries.
1351  */
1352 #define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1353 #define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1354 struct xfs_owner_info {
1355 	uint64_t		oi_owner;
1356 	xfs_fileoff_t		oi_offset;
1357 	unsigned int		oi_flags;
1358 };
1359 
1360 /*
1361  * Special owner types.
1362  *
1363  * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1364  * to tell us we have a special owner value. We use these for static metadata
1365  * allocated at mkfs/growfs time, as well as for freespace management metadata.
1366  */
1367 #define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1368 #define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1369 #define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1370 #define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1371 #define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1372 #define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1373 #define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1374 #define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1375 #define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1376 #define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1377 
1378 #define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1379 
1380 /*
1381  * Data record structure
1382  */
1383 struct xfs_rmap_rec {
1384 	__be32		rm_startblock;	/* extent start block */
1385 	__be32		rm_blockcount;	/* extent length */
1386 	__be64		rm_owner;	/* extent owner */
1387 	__be64		rm_offset;	/* offset within the owner */
1388 };
1389 
1390 /*
1391  * rmap btree record
1392  *  rm_offset:63 is the attribute fork flag
1393  *  rm_offset:62 is the bmbt block flag
1394  *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1395  *  rm_offset:54-60 aren't used and should be zero
1396  *  rm_offset:0-53 is the block offset within the inode
1397  */
1398 #define XFS_RMAP_OFF_ATTR_FORK	((uint64_t)1ULL << 63)
1399 #define XFS_RMAP_OFF_BMBT_BLOCK	((uint64_t)1ULL << 62)
1400 #define XFS_RMAP_OFF_UNWRITTEN	((uint64_t)1ULL << 61)
1401 
1402 #define XFS_RMAP_LEN_MAX	((uint32_t)~0U)
1403 #define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1404 				 XFS_RMAP_OFF_BMBT_BLOCK | \
1405 				 XFS_RMAP_OFF_UNWRITTEN)
1406 #define XFS_RMAP_OFF_MASK	((uint64_t)0x3FFFFFFFFFFFFFULL)
1407 
1408 #define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1409 
1410 #define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1411 #define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1412 #define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1413 
1414 #define RMAPBT_STARTBLOCK_BITLEN	32
1415 #define RMAPBT_BLOCKCOUNT_BITLEN	32
1416 #define RMAPBT_OWNER_BITLEN		64
1417 #define RMAPBT_ATTRFLAG_BITLEN		1
1418 #define RMAPBT_BMBTFLAG_BITLEN		1
1419 #define RMAPBT_EXNTFLAG_BITLEN		1
1420 #define RMAPBT_UNUSED_OFFSET_BITLEN	7
1421 #define RMAPBT_OFFSET_BITLEN		54
1422 
1423 #define XFS_RMAP_ATTR_FORK		(1 << 0)
1424 #define XFS_RMAP_BMBT_BLOCK		(1 << 1)
1425 #define XFS_RMAP_UNWRITTEN		(1 << 2)
1426 #define XFS_RMAP_KEY_FLAGS		(XFS_RMAP_ATTR_FORK | \
1427 					 XFS_RMAP_BMBT_BLOCK)
1428 #define XFS_RMAP_REC_FLAGS		(XFS_RMAP_UNWRITTEN)
1429 struct xfs_rmap_irec {
1430 	xfs_agblock_t	rm_startblock;	/* extent start block */
1431 	xfs_extlen_t	rm_blockcount;	/* extent length */
1432 	uint64_t	rm_owner;	/* extent owner */
1433 	uint64_t	rm_offset;	/* offset within the owner */
1434 	unsigned int	rm_flags;	/* state flags */
1435 };
1436 
1437 /*
1438  * Key structure
1439  *
1440  * We don't use the length for lookups
1441  */
1442 struct xfs_rmap_key {
1443 	__be32		rm_startblock;	/* extent start block */
1444 	__be64		rm_owner;	/* extent owner */
1445 	__be64		rm_offset;	/* offset within the owner */
1446 } __attribute__((packed));
1447 
1448 /* btree pointer type */
1449 typedef __be32 xfs_rmap_ptr_t;
1450 
1451 #define	XFS_RMAP_BLOCK(mp) \
1452 	(xfs_sb_version_hasfinobt(&((mp)->m_sb)) ? \
1453 	 XFS_FIBT_BLOCK(mp) + 1 : \
1454 	 XFS_IBT_BLOCK(mp) + 1)
1455 
1456 /*
1457  * Reference Count Btree format definitions
1458  *
1459  */
1460 #define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1461 
1462 unsigned int xfs_refc_block(struct xfs_mount *mp);
1463 
1464 /*
1465  * Data record/key structure
1466  *
1467  * Each record associates a range of physical blocks (starting at
1468  * rc_startblock and ending rc_blockcount blocks later) with a reference
1469  * count (rc_refcount).  Extents that are being used to stage a copy on
1470  * write (CoW) operation are recorded in the refcount btree with a
1471  * refcount of 1.  All other records must have a refcount > 1 and must
1472  * track an extent mapped only by file data forks.
1473  *
1474  * Extents with a single owner (attributes, metadata, non-shared file
1475  * data) are not tracked here.  Free space is also not tracked here.
1476  * This is consistent with pre-reflink XFS.
1477  */
1478 
1479 /*
1480  * Extents that are being used to stage a copy on write are stored
1481  * in the refcount btree with a refcount of 1 and the upper bit set
1482  * on the startblock.  This speeds up mount time deletion of stale
1483  * staging extents because they're all at the right side of the tree.
1484  */
1485 #define XFS_REFC_COW_START		((xfs_agblock_t)(1U << 31))
1486 #define REFCNTBT_COWFLAG_BITLEN		1
1487 #define REFCNTBT_AGBLOCK_BITLEN		31
1488 
1489 struct xfs_refcount_rec {
1490 	__be32		rc_startblock;	/* starting block number */
1491 	__be32		rc_blockcount;	/* count of blocks */
1492 	__be32		rc_refcount;	/* number of inodes linked here */
1493 };
1494 
1495 struct xfs_refcount_key {
1496 	__be32		rc_startblock;	/* starting block number */
1497 };
1498 
1499 struct xfs_refcount_irec {
1500 	xfs_agblock_t	rc_startblock;	/* starting block number */
1501 	xfs_extlen_t	rc_blockcount;	/* count of free blocks */
1502 	xfs_nlink_t	rc_refcount;	/* number of inodes linked here */
1503 };
1504 
1505 #define MAXREFCOUNT	((xfs_nlink_t)~0U)
1506 #define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
1507 
1508 /* btree pointer type */
1509 typedef __be32 xfs_refcount_ptr_t;
1510 
1511 
1512 /*
1513  * BMAP Btree format definitions
1514  *
1515  * This includes both the root block definition that sits inside an inode fork
1516  * and the record/pointer formats for the leaf/node in the blocks.
1517  */
1518 #define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1519 #define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1520 
1521 /*
1522  * Bmap root header, on-disk form only.
1523  */
1524 typedef struct xfs_bmdr_block {
1525 	__be16		bb_level;	/* 0 is a leaf */
1526 	__be16		bb_numrecs;	/* current # of data records */
1527 } xfs_bmdr_block_t;
1528 
1529 /*
1530  * Bmap btree record and extent descriptor.
1531  *  l0:63 is an extent flag (value 1 indicates non-normal).
1532  *  l0:9-62 are startoff.
1533  *  l0:0-8 and l1:21-63 are startblock.
1534  *  l1:0-20 are blockcount.
1535  */
1536 #define BMBT_EXNTFLAG_BITLEN	1
1537 #define BMBT_STARTOFF_BITLEN	54
1538 #define BMBT_STARTBLOCK_BITLEN	52
1539 #define BMBT_BLOCKCOUNT_BITLEN	21
1540 
1541 typedef struct xfs_bmbt_rec {
1542 	__be64			l0, l1;
1543 } xfs_bmbt_rec_t;
1544 
1545 typedef uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1546 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1547 
1548 /*
1549  * Values and macros for delayed-allocation startblock fields.
1550  */
1551 #define STARTBLOCKVALBITS	17
1552 #define STARTBLOCKMASKBITS	(15 + 20)
1553 #define STARTBLOCKMASK		\
1554 	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1555 
1556 static inline int isnullstartblock(xfs_fsblock_t x)
1557 {
1558 	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1559 }
1560 
1561 static inline xfs_fsblock_t nullstartblock(int k)
1562 {
1563 	ASSERT(k < (1 << STARTBLOCKVALBITS));
1564 	return STARTBLOCKMASK | (k);
1565 }
1566 
1567 static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1568 {
1569 	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1570 }
1571 
1572 /*
1573  * Key structure for non-leaf levels of the tree.
1574  */
1575 typedef struct xfs_bmbt_key {
1576 	__be64		br_startoff;	/* starting file offset */
1577 } xfs_bmbt_key_t, xfs_bmdr_key_t;
1578 
1579 /* btree pointer type */
1580 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1581 
1582 
1583 /*
1584  * Generic Btree block format definitions
1585  *
1586  * This is a combination of the actual format used on disk for short and long
1587  * format btrees.  The first three fields are shared by both format, but the
1588  * pointers are different and should be used with care.
1589  *
1590  * To get the size of the actual short or long form headers please use the size
1591  * macros below.  Never use sizeof(xfs_btree_block).
1592  *
1593  * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1594  * with the crc feature bit, and all accesses to them must be conditional on
1595  * that flag.
1596  */
1597 /* short form block header */
1598 struct xfs_btree_block_shdr {
1599 	__be32		bb_leftsib;
1600 	__be32		bb_rightsib;
1601 
1602 	__be64		bb_blkno;
1603 	__be64		bb_lsn;
1604 	uuid_t		bb_uuid;
1605 	__be32		bb_owner;
1606 	__le32		bb_crc;
1607 };
1608 
1609 /* long form block header */
1610 struct xfs_btree_block_lhdr {
1611 	__be64		bb_leftsib;
1612 	__be64		bb_rightsib;
1613 
1614 	__be64		bb_blkno;
1615 	__be64		bb_lsn;
1616 	uuid_t		bb_uuid;
1617 	__be64		bb_owner;
1618 	__le32		bb_crc;
1619 	__be32		bb_pad; /* padding for alignment */
1620 };
1621 
1622 struct xfs_btree_block {
1623 	__be32		bb_magic;	/* magic number for block type */
1624 	__be16		bb_level;	/* 0 is a leaf */
1625 	__be16		bb_numrecs;	/* current # of data records */
1626 	union {
1627 		struct xfs_btree_block_shdr s;
1628 		struct xfs_btree_block_lhdr l;
1629 	} bb_u;				/* rest */
1630 };
1631 
1632 /* size of a short form block */
1633 #define XFS_BTREE_SBLOCK_LEN \
1634 	(offsetof(struct xfs_btree_block, bb_u) + \
1635 	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1636 /* size of a long form block */
1637 #define XFS_BTREE_LBLOCK_LEN \
1638 	(offsetof(struct xfs_btree_block, bb_u) + \
1639 	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1640 
1641 /* sizes of CRC enabled btree blocks */
1642 #define XFS_BTREE_SBLOCK_CRC_LEN \
1643 	(offsetof(struct xfs_btree_block, bb_u) + \
1644 	 sizeof(struct xfs_btree_block_shdr))
1645 #define XFS_BTREE_LBLOCK_CRC_LEN \
1646 	(offsetof(struct xfs_btree_block, bb_u) + \
1647 	 sizeof(struct xfs_btree_block_lhdr))
1648 
1649 #define XFS_BTREE_SBLOCK_CRC_OFF \
1650 	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1651 #define XFS_BTREE_LBLOCK_CRC_OFF \
1652 	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1653 
1654 /*
1655  * On-disk XFS access control list structure.
1656  */
1657 struct xfs_acl_entry {
1658 	__be32	ae_tag;
1659 	__be32	ae_id;
1660 	__be16	ae_perm;
1661 	__be16	ae_pad;		/* fill the implicit hole in the structure */
1662 };
1663 
1664 struct xfs_acl {
1665 	__be32			acl_cnt;
1666 	struct xfs_acl_entry	acl_entry[0];
1667 };
1668 
1669 /*
1670  * The number of ACL entries allowed is defined by the on-disk format.
1671  * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1672  * limited only by the maximum size of the xattr that stores the information.
1673  */
1674 #define XFS_ACL_MAX_ENTRIES(mp)	\
1675 	(xfs_sb_version_hascrc(&mp->m_sb) \
1676 		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1677 						sizeof(struct xfs_acl_entry) \
1678 		: 25)
1679 
1680 #define XFS_ACL_SIZE(cnt) \
1681 	(sizeof(struct xfs_acl) + \
1682 		sizeof(struct xfs_acl_entry) * cnt)
1683 
1684 #define XFS_ACL_MAX_SIZE(mp) \
1685 	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1686 
1687 
1688 /* On-disk XFS extended attribute names */
1689 #define SGI_ACL_FILE		"SGI_ACL_FILE"
1690 #define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
1691 #define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
1692 #define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
1693 
1694 #endif /* __XFS_FORMAT_H__ */
1695