xref: /openbmc/linux/fs/xfs/libxfs/xfs_format.h (revision bf070bb0)
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 /*
102  * Superblock - in core version.  Must match the ondisk version below.
103  * Must be padded to 64 bit alignment.
104  */
105 typedef struct xfs_sb {
106 	uint32_t	sb_magicnum;	/* magic number == XFS_SB_MAGIC */
107 	uint32_t	sb_blocksize;	/* logical block size, bytes */
108 	xfs_rfsblock_t	sb_dblocks;	/* number of data blocks */
109 	xfs_rfsblock_t	sb_rblocks;	/* number of realtime blocks */
110 	xfs_rtblock_t	sb_rextents;	/* number of realtime extents */
111 	uuid_t		sb_uuid;	/* user-visible file system unique id */
112 	xfs_fsblock_t	sb_logstart;	/* starting block of log if internal */
113 	xfs_ino_t	sb_rootino;	/* root inode number */
114 	xfs_ino_t	sb_rbmino;	/* bitmap inode for realtime extents */
115 	xfs_ino_t	sb_rsumino;	/* summary inode for rt bitmap */
116 	xfs_agblock_t	sb_rextsize;	/* realtime extent size, blocks */
117 	xfs_agblock_t	sb_agblocks;	/* size of an allocation group */
118 	xfs_agnumber_t	sb_agcount;	/* number of allocation groups */
119 	xfs_extlen_t	sb_rbmblocks;	/* number of rt bitmap blocks */
120 	xfs_extlen_t	sb_logblocks;	/* number of log blocks */
121 	uint16_t	sb_versionnum;	/* header version == XFS_SB_VERSION */
122 	uint16_t	sb_sectsize;	/* volume sector size, bytes */
123 	uint16_t	sb_inodesize;	/* inode size, bytes */
124 	uint16_t	sb_inopblock;	/* inodes per block */
125 	char		sb_fname[12];	/* file system name */
126 	uint8_t		sb_blocklog;	/* log2 of sb_blocksize */
127 	uint8_t		sb_sectlog;	/* log2 of sb_sectsize */
128 	uint8_t		sb_inodelog;	/* log2 of sb_inodesize */
129 	uint8_t		sb_inopblog;	/* log2 of sb_inopblock */
130 	uint8_t		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
131 	uint8_t		sb_rextslog;	/* log2 of sb_rextents */
132 	uint8_t		sb_inprogress;	/* mkfs is in progress, don't mount */
133 	uint8_t		sb_imax_pct;	/* max % of fs for inode space */
134 					/* statistics */
135 	/*
136 	 * These fields must remain contiguous.  If you really
137 	 * want to change their layout, make sure you fix the
138 	 * code in xfs_trans_apply_sb_deltas().
139 	 */
140 	uint64_t	sb_icount;	/* allocated inodes */
141 	uint64_t	sb_ifree;	/* free inodes */
142 	uint64_t	sb_fdblocks;	/* free data blocks */
143 	uint64_t	sb_frextents;	/* free realtime extents */
144 	/*
145 	 * End contiguous fields.
146 	 */
147 	xfs_ino_t	sb_uquotino;	/* user quota inode */
148 	xfs_ino_t	sb_gquotino;	/* group quota inode */
149 	uint16_t	sb_qflags;	/* quota flags */
150 	uint8_t		sb_flags;	/* misc. flags */
151 	uint8_t		sb_shared_vn;	/* shared version number */
152 	xfs_extlen_t	sb_inoalignmt;	/* inode chunk alignment, fsblocks */
153 	uint32_t	sb_unit;	/* stripe or raid unit */
154 	uint32_t	sb_width;	/* stripe or raid width */
155 	uint8_t		sb_dirblklog;	/* log2 of dir block size (fsbs) */
156 	uint8_t		sb_logsectlog;	/* log2 of the log sector size */
157 	uint16_t	sb_logsectsize;	/* sector size for the log, bytes */
158 	uint32_t	sb_logsunit;	/* stripe unit size for the log */
159 	uint32_t	sb_features2;	/* additional feature bits */
160 
161 	/*
162 	 * bad features2 field as a result of failing to pad the sb structure to
163 	 * 64 bits. Some machines will be using this field for features2 bits.
164 	 * Easiest just to mark it bad and not use it for anything else.
165 	 *
166 	 * This is not kept up to date in memory; it is always overwritten by
167 	 * the value in sb_features2 when formatting the incore superblock to
168 	 * the disk buffer.
169 	 */
170 	uint32_t	sb_bad_features2;
171 
172 	/* version 5 superblock fields start here */
173 
174 	/* feature masks */
175 	uint32_t	sb_features_compat;
176 	uint32_t	sb_features_ro_compat;
177 	uint32_t	sb_features_incompat;
178 	uint32_t	sb_features_log_incompat;
179 
180 	uint32_t	sb_crc;		/* superblock crc */
181 	xfs_extlen_t	sb_spino_align;	/* sparse inode chunk alignment */
182 
183 	xfs_ino_t	sb_pquotino;	/* project quota inode */
184 	xfs_lsn_t	sb_lsn;		/* last write sequence */
185 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
186 
187 	/* must be padded to 64 bit alignment */
188 } xfs_sb_t;
189 
190 #define XFS_SB_CRC_OFF		offsetof(struct xfs_sb, sb_crc)
191 
192 /*
193  * Superblock - on disk version.  Must match the in core version above.
194  * Must be padded to 64 bit alignment.
195  */
196 typedef struct xfs_dsb {
197 	__be32		sb_magicnum;	/* magic number == XFS_SB_MAGIC */
198 	__be32		sb_blocksize;	/* logical block size, bytes */
199 	__be64		sb_dblocks;	/* number of data blocks */
200 	__be64		sb_rblocks;	/* number of realtime blocks */
201 	__be64		sb_rextents;	/* number of realtime extents */
202 	uuid_t		sb_uuid;	/* user-visible file system unique id */
203 	__be64		sb_logstart;	/* starting block of log if internal */
204 	__be64		sb_rootino;	/* root inode number */
205 	__be64		sb_rbmino;	/* bitmap inode for realtime extents */
206 	__be64		sb_rsumino;	/* summary inode for rt bitmap */
207 	__be32		sb_rextsize;	/* realtime extent size, blocks */
208 	__be32		sb_agblocks;	/* size of an allocation group */
209 	__be32		sb_agcount;	/* number of allocation groups */
210 	__be32		sb_rbmblocks;	/* number of rt bitmap blocks */
211 	__be32		sb_logblocks;	/* number of log blocks */
212 	__be16		sb_versionnum;	/* header version == XFS_SB_VERSION */
213 	__be16		sb_sectsize;	/* volume sector size, bytes */
214 	__be16		sb_inodesize;	/* inode size, bytes */
215 	__be16		sb_inopblock;	/* inodes per block */
216 	char		sb_fname[12];	/* file system name */
217 	__u8		sb_blocklog;	/* log2 of sb_blocksize */
218 	__u8		sb_sectlog;	/* log2 of sb_sectsize */
219 	__u8		sb_inodelog;	/* log2 of sb_inodesize */
220 	__u8		sb_inopblog;	/* log2 of sb_inopblock */
221 	__u8		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
222 	__u8		sb_rextslog;	/* log2 of sb_rextents */
223 	__u8		sb_inprogress;	/* mkfs is in progress, don't mount */
224 	__u8		sb_imax_pct;	/* max % of fs for inode space */
225 					/* statistics */
226 	/*
227 	 * These fields must remain contiguous.  If you really
228 	 * want to change their layout, make sure you fix the
229 	 * code in xfs_trans_apply_sb_deltas().
230 	 */
231 	__be64		sb_icount;	/* allocated inodes */
232 	__be64		sb_ifree;	/* free inodes */
233 	__be64		sb_fdblocks;	/* free data blocks */
234 	__be64		sb_frextents;	/* free realtime extents */
235 	/*
236 	 * End contiguous fields.
237 	 */
238 	__be64		sb_uquotino;	/* user quota inode */
239 	__be64		sb_gquotino;	/* group quota inode */
240 	__be16		sb_qflags;	/* quota flags */
241 	__u8		sb_flags;	/* misc. flags */
242 	__u8		sb_shared_vn;	/* shared version number */
243 	__be32		sb_inoalignmt;	/* inode chunk alignment, fsblocks */
244 	__be32		sb_unit;	/* stripe or raid unit */
245 	__be32		sb_width;	/* stripe or raid width */
246 	__u8		sb_dirblklog;	/* log2 of dir block size (fsbs) */
247 	__u8		sb_logsectlog;	/* log2 of the log sector size */
248 	__be16		sb_logsectsize;	/* sector size for the log, bytes */
249 	__be32		sb_logsunit;	/* stripe unit size for the log */
250 	__be32		sb_features2;	/* additional feature bits */
251 	/*
252 	 * bad features2 field as a result of failing to pad the sb
253 	 * structure to 64 bits. Some machines will be using this field
254 	 * for features2 bits. Easiest just to mark it bad and not use
255 	 * it for anything else.
256 	 */
257 	__be32		sb_bad_features2;
258 
259 	/* version 5 superblock fields start here */
260 
261 	/* feature masks */
262 	__be32		sb_features_compat;
263 	__be32		sb_features_ro_compat;
264 	__be32		sb_features_incompat;
265 	__be32		sb_features_log_incompat;
266 
267 	__le32		sb_crc;		/* superblock crc */
268 	__be32		sb_spino_align;	/* sparse inode chunk alignment */
269 
270 	__be64		sb_pquotino;	/* project quota inode */
271 	__be64		sb_lsn;		/* last write sequence */
272 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
273 
274 	/* must be padded to 64 bit alignment */
275 } xfs_dsb_t;
276 
277 
278 /*
279  * Misc. Flags - warning - these will be cleared by xfs_repair unless
280  * a feature bit is set when the flag is used.
281  */
282 #define XFS_SBF_NOFLAGS		0x00	/* no flags set */
283 #define XFS_SBF_READONLY	0x01	/* only read-only mounts allowed */
284 
285 /*
286  * define max. shared version we can interoperate with
287  */
288 #define XFS_SB_MAX_SHARED_VN	0
289 
290 #define	XFS_SB_VERSION_NUM(sbp)	((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
291 
292 /*
293  * The first XFS version we support is a v4 superblock with V2 directories.
294  */
295 static inline bool xfs_sb_good_v4_features(struct xfs_sb *sbp)
296 {
297 	if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
298 		return false;
299 
300 	/* check for unknown features in the fs */
301 	if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
302 	    ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
303 	     (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
304 		return false;
305 
306 	return true;
307 }
308 
309 static inline bool xfs_sb_good_version(struct xfs_sb *sbp)
310 {
311 	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5)
312 		return true;
313 	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_4)
314 		return xfs_sb_good_v4_features(sbp);
315 	return false;
316 }
317 
318 static inline bool xfs_sb_version_hasrealtime(struct xfs_sb *sbp)
319 {
320 	return sbp->sb_rblocks > 0;
321 }
322 
323 /*
324  * Detect a mismatched features2 field.  Older kernels read/wrote
325  * this into the wrong slot, so to be safe we keep them in sync.
326  */
327 static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
328 {
329 	return sbp->sb_bad_features2 != sbp->sb_features2;
330 }
331 
332 static inline bool xfs_sb_version_hasattr(struct xfs_sb *sbp)
333 {
334 	return (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT);
335 }
336 
337 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
338 {
339 	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
340 }
341 
342 static inline bool xfs_sb_version_hasquota(struct xfs_sb *sbp)
343 {
344 	return (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT);
345 }
346 
347 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
348 {
349 	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
350 }
351 
352 static inline bool xfs_sb_version_hasalign(struct xfs_sb *sbp)
353 {
354 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
355 		(sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT));
356 }
357 
358 static inline bool xfs_sb_version_hasdalign(struct xfs_sb *sbp)
359 {
360 	return (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT);
361 }
362 
363 static inline bool xfs_sb_version_haslogv2(struct xfs_sb *sbp)
364 {
365 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
366 	       (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT);
367 }
368 
369 static inline bool xfs_sb_version_hasextflgbit(struct xfs_sb *sbp)
370 {
371 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
372 	       (sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT);
373 }
374 
375 static inline bool xfs_sb_version_hassector(struct xfs_sb *sbp)
376 {
377 	return (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT);
378 }
379 
380 static inline bool xfs_sb_version_hasasciici(struct xfs_sb *sbp)
381 {
382 	return (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT);
383 }
384 
385 static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp)
386 {
387 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
388 	       (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
389 }
390 
391 /*
392  * sb_features2 bit version macros.
393  */
394 static inline bool xfs_sb_version_haslazysbcount(struct xfs_sb *sbp)
395 {
396 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
397 	       (xfs_sb_version_hasmorebits(sbp) &&
398 		(sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT));
399 }
400 
401 static inline bool xfs_sb_version_hasattr2(struct xfs_sb *sbp)
402 {
403 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
404 	       (xfs_sb_version_hasmorebits(sbp) &&
405 		(sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT));
406 }
407 
408 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
409 {
410 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
411 	sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
412 }
413 
414 static inline void xfs_sb_version_removeattr2(struct xfs_sb *sbp)
415 {
416 	sbp->sb_features2 &= ~XFS_SB_VERSION2_ATTR2BIT;
417 	if (!sbp->sb_features2)
418 		sbp->sb_versionnum &= ~XFS_SB_VERSION_MOREBITSBIT;
419 }
420 
421 static inline bool xfs_sb_version_hasprojid32bit(struct xfs_sb *sbp)
422 {
423 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
424 	       (xfs_sb_version_hasmorebits(sbp) &&
425 		(sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT));
426 }
427 
428 static inline void xfs_sb_version_addprojid32bit(struct xfs_sb *sbp)
429 {
430 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
431 	sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
432 }
433 
434 /*
435  * Extended v5 superblock feature masks. These are to be used for new v5
436  * superblock features only.
437  *
438  * Compat features are new features that old kernels will not notice or affect
439  * and so can mount read-write without issues.
440  *
441  * RO-Compat (read only) are features that old kernels can read but will break
442  * if they write. Hence only read-only mounts of such filesystems are allowed on
443  * kernels that don't support the feature bit.
444  *
445  * InCompat features are features which old kernels will not understand and so
446  * must not mount.
447  *
448  * Log-InCompat features are for changes to log formats or new transactions that
449  * can't be replayed on older kernels. The fields are set when the filesystem is
450  * mounted, and a clean unmount clears the fields.
451  */
452 #define XFS_SB_FEAT_COMPAT_ALL 0
453 #define XFS_SB_FEAT_COMPAT_UNKNOWN	~XFS_SB_FEAT_COMPAT_ALL
454 static inline bool
455 xfs_sb_has_compat_feature(
456 	struct xfs_sb	*sbp,
457 	uint32_t	feature)
458 {
459 	return (sbp->sb_features_compat & feature) != 0;
460 }
461 
462 #define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
463 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
464 #define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
465 #define XFS_SB_FEAT_RO_COMPAT_ALL \
466 		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
467 		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
468 		 XFS_SB_FEAT_RO_COMPAT_REFLINK)
469 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
470 static inline bool
471 xfs_sb_has_ro_compat_feature(
472 	struct xfs_sb	*sbp,
473 	uint32_t	feature)
474 {
475 	return (sbp->sb_features_ro_compat & feature) != 0;
476 }
477 
478 #define XFS_SB_FEAT_INCOMPAT_FTYPE	(1 << 0)	/* filetype in dirent */
479 #define XFS_SB_FEAT_INCOMPAT_SPINODES	(1 << 1)	/* sparse inode chunks */
480 #define XFS_SB_FEAT_INCOMPAT_META_UUID	(1 << 2)	/* metadata UUID */
481 #define XFS_SB_FEAT_INCOMPAT_ALL \
482 		(XFS_SB_FEAT_INCOMPAT_FTYPE|	\
483 		 XFS_SB_FEAT_INCOMPAT_SPINODES|	\
484 		 XFS_SB_FEAT_INCOMPAT_META_UUID)
485 
486 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
487 static inline bool
488 xfs_sb_has_incompat_feature(
489 	struct xfs_sb	*sbp,
490 	uint32_t	feature)
491 {
492 	return (sbp->sb_features_incompat & feature) != 0;
493 }
494 
495 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL 0
496 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
497 static inline bool
498 xfs_sb_has_incompat_log_feature(
499 	struct xfs_sb	*sbp,
500 	uint32_t	feature)
501 {
502 	return (sbp->sb_features_log_incompat & feature) != 0;
503 }
504 
505 /*
506  * V5 superblock specific feature checks
507  */
508 static inline bool xfs_sb_version_hascrc(struct xfs_sb *sbp)
509 {
510 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
511 }
512 
513 static inline bool xfs_sb_version_has_pquotino(struct xfs_sb *sbp)
514 {
515 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
516 }
517 
518 static inline int xfs_sb_version_hasftype(struct xfs_sb *sbp)
519 {
520 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
521 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_FTYPE)) ||
522 	       (xfs_sb_version_hasmorebits(sbp) &&
523 		 (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE));
524 }
525 
526 static inline bool xfs_sb_version_hasfinobt(xfs_sb_t *sbp)
527 {
528 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
529 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT);
530 }
531 
532 static inline bool xfs_sb_version_hassparseinodes(struct xfs_sb *sbp)
533 {
534 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
535 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_SPINODES);
536 }
537 
538 /*
539  * XFS_SB_FEAT_INCOMPAT_META_UUID indicates that the metadata UUID
540  * is stored separately from the user-visible UUID; this allows the
541  * user-visible UUID to be changed on V5 filesystems which have a
542  * filesystem UUID stamped into every piece of metadata.
543  */
544 static inline bool xfs_sb_version_hasmetauuid(struct xfs_sb *sbp)
545 {
546 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
547 		(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID);
548 }
549 
550 static inline bool xfs_sb_version_hasrmapbt(struct xfs_sb *sbp)
551 {
552 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
553 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT);
554 }
555 
556 static inline bool xfs_sb_version_hasreflink(struct xfs_sb *sbp)
557 {
558 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
559 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK);
560 }
561 
562 /*
563  * end of superblock version macros
564  */
565 
566 static inline bool
567 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
568 {
569 	return (ino == sbp->sb_uquotino ||
570 		ino == sbp->sb_gquotino ||
571 		ino == sbp->sb_pquotino);
572 }
573 
574 #define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
575 #define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
576 #define XFS_BUF_TO_SBP(bp)	((xfs_dsb_t *)((bp)->b_addr))
577 
578 #define	XFS_HDR_BLOCK(mp,d)	((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
579 #define	XFS_DADDR_TO_FSB(mp,d)	XFS_AGB_TO_FSB(mp, \
580 			xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
581 #define	XFS_FSB_TO_DADDR(mp,fsbno)	XFS_AGB_TO_DADDR(mp, \
582 			XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
583 
584 /*
585  * File system sector to basic block conversions.
586  */
587 #define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
588 
589 /*
590  * File system block to basic block conversions.
591  */
592 #define	XFS_FSB_TO_BB(mp,fsbno)	((fsbno) << (mp)->m_blkbb_log)
593 #define	XFS_BB_TO_FSB(mp,bb)	\
594 	(((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
595 #define	XFS_BB_TO_FSBT(mp,bb)	((bb) >> (mp)->m_blkbb_log)
596 
597 /*
598  * File system block to byte conversions.
599  */
600 #define XFS_FSB_TO_B(mp,fsbno)	((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
601 #define XFS_B_TO_FSB(mp,b)	\
602 	((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
603 #define XFS_B_TO_FSBT(mp,b)	(((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
604 #define XFS_B_FSB_OFFSET(mp,b)	((b) & (mp)->m_blockmask)
605 
606 /*
607  * Allocation group header
608  *
609  * This is divided into three structures, placed in sequential 512-byte
610  * buffers after a copy of the superblock (also in a 512-byte buffer).
611  */
612 #define	XFS_AGF_MAGIC	0x58414746	/* 'XAGF' */
613 #define	XFS_AGI_MAGIC	0x58414749	/* 'XAGI' */
614 #define	XFS_AGFL_MAGIC	0x5841464c	/* 'XAFL' */
615 #define	XFS_AGF_VERSION	1
616 #define	XFS_AGI_VERSION	1
617 
618 #define	XFS_AGF_GOOD_VERSION(v)	((v) == XFS_AGF_VERSION)
619 #define	XFS_AGI_GOOD_VERSION(v)	((v) == XFS_AGI_VERSION)
620 
621 /*
622  * Btree number 0 is bno, 1 is cnt, 2 is rmap. This value gives the size of the
623  * arrays below.
624  */
625 #define	XFS_BTNUM_AGF	((int)XFS_BTNUM_RMAPi + 1)
626 
627 /*
628  * The second word of agf_levels in the first a.g. overlaps the EFS
629  * superblock's magic number.  Since the magic numbers valid for EFS
630  * are > 64k, our value cannot be confused for an EFS superblock's.
631  */
632 
633 typedef struct xfs_agf {
634 	/*
635 	 * Common allocation group header information
636 	 */
637 	__be32		agf_magicnum;	/* magic number == XFS_AGF_MAGIC */
638 	__be32		agf_versionnum;	/* header version == XFS_AGF_VERSION */
639 	__be32		agf_seqno;	/* sequence # starting from 0 */
640 	__be32		agf_length;	/* size in blocks of a.g. */
641 	/*
642 	 * Freespace and rmap information
643 	 */
644 	__be32		agf_roots[XFS_BTNUM_AGF];	/* root blocks */
645 	__be32		agf_levels[XFS_BTNUM_AGF];	/* btree levels */
646 
647 	__be32		agf_flfirst;	/* first freelist block's index */
648 	__be32		agf_fllast;	/* last freelist block's index */
649 	__be32		agf_flcount;	/* count of blocks in freelist */
650 	__be32		agf_freeblks;	/* total free blocks */
651 
652 	__be32		agf_longest;	/* longest free space */
653 	__be32		agf_btreeblks;	/* # of blocks held in AGF btrees */
654 	uuid_t		agf_uuid;	/* uuid of filesystem */
655 
656 	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
657 	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
658 
659 	__be32		agf_refcount_root;	/* refcount tree root block */
660 	__be32		agf_refcount_level;	/* refcount btree levels */
661 
662 	/*
663 	 * reserve some contiguous space for future logged fields before we add
664 	 * the unlogged fields. This makes the range logging via flags and
665 	 * structure offsets much simpler.
666 	 */
667 	__be64		agf_spare64[14];
668 
669 	/* unlogged fields, written during buffer writeback. */
670 	__be64		agf_lsn;	/* last write sequence */
671 	__be32		agf_crc;	/* crc of agf sector */
672 	__be32		agf_spare2;
673 
674 	/* structure must be padded to 64 bit alignment */
675 } xfs_agf_t;
676 
677 #define XFS_AGF_CRC_OFF		offsetof(struct xfs_agf, agf_crc)
678 
679 #define	XFS_AGF_MAGICNUM	0x00000001
680 #define	XFS_AGF_VERSIONNUM	0x00000002
681 #define	XFS_AGF_SEQNO		0x00000004
682 #define	XFS_AGF_LENGTH		0x00000008
683 #define	XFS_AGF_ROOTS		0x00000010
684 #define	XFS_AGF_LEVELS		0x00000020
685 #define	XFS_AGF_FLFIRST		0x00000040
686 #define	XFS_AGF_FLLAST		0x00000080
687 #define	XFS_AGF_FLCOUNT		0x00000100
688 #define	XFS_AGF_FREEBLKS	0x00000200
689 #define	XFS_AGF_LONGEST		0x00000400
690 #define	XFS_AGF_BTREEBLKS	0x00000800
691 #define	XFS_AGF_UUID		0x00001000
692 #define	XFS_AGF_RMAP_BLOCKS	0x00002000
693 #define	XFS_AGF_REFCOUNT_BLOCKS	0x00004000
694 #define	XFS_AGF_REFCOUNT_ROOT	0x00008000
695 #define	XFS_AGF_REFCOUNT_LEVEL	0x00010000
696 #define	XFS_AGF_SPARE64		0x00020000
697 #define	XFS_AGF_NUM_BITS	18
698 #define	XFS_AGF_ALL_BITS	((1 << XFS_AGF_NUM_BITS) - 1)
699 
700 #define XFS_AGF_FLAGS \
701 	{ XFS_AGF_MAGICNUM,	"MAGICNUM" }, \
702 	{ XFS_AGF_VERSIONNUM,	"VERSIONNUM" }, \
703 	{ XFS_AGF_SEQNO,	"SEQNO" }, \
704 	{ XFS_AGF_LENGTH,	"LENGTH" }, \
705 	{ XFS_AGF_ROOTS,	"ROOTS" }, \
706 	{ XFS_AGF_LEVELS,	"LEVELS" }, \
707 	{ XFS_AGF_FLFIRST,	"FLFIRST" }, \
708 	{ XFS_AGF_FLLAST,	"FLLAST" }, \
709 	{ XFS_AGF_FLCOUNT,	"FLCOUNT" }, \
710 	{ XFS_AGF_FREEBLKS,	"FREEBLKS" }, \
711 	{ XFS_AGF_LONGEST,	"LONGEST" }, \
712 	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
713 	{ XFS_AGF_UUID,		"UUID" }, \
714 	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
715 	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
716 	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
717 	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
718 	{ XFS_AGF_SPARE64,	"SPARE64" }
719 
720 /* disk block (xfs_daddr_t) in the AG */
721 #define XFS_AGF_DADDR(mp)	((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
722 #define	XFS_AGF_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
723 #define	XFS_BUF_TO_AGF(bp)	((xfs_agf_t *)((bp)->b_addr))
724 
725 /*
726  * Size of the unlinked inode hash table in the agi.
727  */
728 #define	XFS_AGI_UNLINKED_BUCKETS	64
729 
730 typedef struct xfs_agi {
731 	/*
732 	 * Common allocation group header information
733 	 */
734 	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
735 	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
736 	__be32		agi_seqno;	/* sequence # starting from 0 */
737 	__be32		agi_length;	/* size in blocks of a.g. */
738 	/*
739 	 * Inode information
740 	 * Inodes are mapped by interpreting the inode number, so no
741 	 * mapping data is needed here.
742 	 */
743 	__be32		agi_count;	/* count of allocated inodes */
744 	__be32		agi_root;	/* root of inode btree */
745 	__be32		agi_level;	/* levels in inode btree */
746 	__be32		agi_freecount;	/* number of free inodes */
747 
748 	__be32		agi_newino;	/* new inode just allocated */
749 	__be32		agi_dirino;	/* last directory inode chunk */
750 	/*
751 	 * Hash table of inodes which have been unlinked but are
752 	 * still being referenced.
753 	 */
754 	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
755 	/*
756 	 * This marks the end of logging region 1 and start of logging region 2.
757 	 */
758 	uuid_t		agi_uuid;	/* uuid of filesystem */
759 	__be32		agi_crc;	/* crc of agi sector */
760 	__be32		agi_pad32;
761 	__be64		agi_lsn;	/* last write sequence */
762 
763 	__be32		agi_free_root; /* root of the free inode btree */
764 	__be32		agi_free_level;/* levels in free inode btree */
765 
766 	/* structure must be padded to 64 bit alignment */
767 } xfs_agi_t;
768 
769 #define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
770 
771 #define	XFS_AGI_MAGICNUM	(1 << 0)
772 #define	XFS_AGI_VERSIONNUM	(1 << 1)
773 #define	XFS_AGI_SEQNO		(1 << 2)
774 #define	XFS_AGI_LENGTH		(1 << 3)
775 #define	XFS_AGI_COUNT		(1 << 4)
776 #define	XFS_AGI_ROOT		(1 << 5)
777 #define	XFS_AGI_LEVEL		(1 << 6)
778 #define	XFS_AGI_FREECOUNT	(1 << 7)
779 #define	XFS_AGI_NEWINO		(1 << 8)
780 #define	XFS_AGI_DIRINO		(1 << 9)
781 #define	XFS_AGI_UNLINKED	(1 << 10)
782 #define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
783 #define	XFS_AGI_ALL_BITS_R1	((1 << XFS_AGI_NUM_BITS_R1) - 1)
784 #define	XFS_AGI_FREE_ROOT	(1 << 11)
785 #define	XFS_AGI_FREE_LEVEL	(1 << 12)
786 #define	XFS_AGI_NUM_BITS_R2	13
787 
788 /* disk block (xfs_daddr_t) in the AG */
789 #define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
790 #define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
791 #define	XFS_BUF_TO_AGI(bp)	((xfs_agi_t *)((bp)->b_addr))
792 
793 /*
794  * The third a.g. block contains the a.g. freelist, an array
795  * of block pointers to blocks owned by the allocation btree code.
796  */
797 #define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
798 #define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
799 #define	XFS_BUF_TO_AGFL(bp)	((xfs_agfl_t *)((bp)->b_addr))
800 
801 #define XFS_BUF_TO_AGFL_BNO(mp, bp) \
802 	(xfs_sb_version_hascrc(&((mp)->m_sb)) ? \
803 		&(XFS_BUF_TO_AGFL(bp)->agfl_bno[0]) : \
804 		(__be32 *)(bp)->b_addr)
805 
806 /*
807  * Size of the AGFL.  For CRC-enabled filesystes we steal a couple of
808  * slots in the beginning of the block for a proper header with the
809  * location information and CRC.
810  */
811 #define XFS_AGFL_SIZE(mp) \
812 	(((mp)->m_sb.sb_sectsize - \
813 	 (xfs_sb_version_hascrc(&((mp)->m_sb)) ? \
814 		sizeof(struct xfs_agfl) : 0)) / \
815 	  sizeof(xfs_agblock_t))
816 
817 typedef struct xfs_agfl {
818 	__be32		agfl_magicnum;
819 	__be32		agfl_seqno;
820 	uuid_t		agfl_uuid;
821 	__be64		agfl_lsn;
822 	__be32		agfl_crc;
823 	__be32		agfl_bno[];	/* actually XFS_AGFL_SIZE(mp) */
824 } __attribute__((packed)) xfs_agfl_t;
825 
826 #define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
827 
828 #define XFS_AGB_TO_FSB(mp,agno,agbno)	\
829 	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
830 #define	XFS_FSB_TO_AGNO(mp,fsbno)	\
831 	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
832 #define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
833 	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
834 #define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
835 	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
836 		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
837 #define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
838 
839 /*
840  * For checking for bad ranges of xfs_daddr_t's, covering multiple
841  * allocation groups or a single xfs_daddr_t that's a superblock copy.
842  */
843 #define	XFS_AG_CHECK_DADDR(mp,d,len)	\
844 	((len) == 1 ? \
845 	    ASSERT((d) == XFS_SB_DADDR || \
846 		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
847 	    ASSERT(xfs_daddr_to_agno(mp, d) == \
848 		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
849 
850 typedef struct xfs_timestamp {
851 	__be32		t_sec;		/* timestamp seconds */
852 	__be32		t_nsec;		/* timestamp nanoseconds */
853 } xfs_timestamp_t;
854 
855 /*
856  * On-disk inode structure.
857  *
858  * This is just the header or "dinode core", the inode is expanded to fill a
859  * variable size the leftover area split into a data and an attribute fork.
860  * The format of the data and attribute fork depends on the format of the
861  * inode as indicated by di_format and di_aformat.  To access the data and
862  * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
863  * below.
864  *
865  * There is a very similar struct icdinode in xfs_inode which matches the
866  * layout of the first 96 bytes of this structure, but is kept in native
867  * format instead of big endian.
868  *
869  * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
870  * padding field for v3 inodes.
871  */
872 #define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
873 typedef struct xfs_dinode {
874 	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
875 	__be16		di_mode;	/* mode and type of file */
876 	__u8		di_version;	/* inode version */
877 	__u8		di_format;	/* format of di_c data */
878 	__be16		di_onlink;	/* old number of links to file */
879 	__be32		di_uid;		/* owner's user id */
880 	__be32		di_gid;		/* owner's group id */
881 	__be32		di_nlink;	/* number of links to file */
882 	__be16		di_projid_lo;	/* lower part of owner's project id */
883 	__be16		di_projid_hi;	/* higher part owner's project id */
884 	__u8		di_pad[6];	/* unused, zeroed space */
885 	__be16		di_flushiter;	/* incremented on flush */
886 	xfs_timestamp_t	di_atime;	/* time last accessed */
887 	xfs_timestamp_t	di_mtime;	/* time last modified */
888 	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
889 	__be64		di_size;	/* number of bytes in file */
890 	__be64		di_nblocks;	/* # of direct & btree blocks used */
891 	__be32		di_extsize;	/* basic/minimum extent size for file */
892 	__be32		di_nextents;	/* number of extents in data fork */
893 	__be16		di_anextents;	/* number of extents in attribute fork*/
894 	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
895 	__s8		di_aformat;	/* format of attr fork's data */
896 	__be32		di_dmevmask;	/* DMIG event mask */
897 	__be16		di_dmstate;	/* DMIG state info */
898 	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
899 	__be32		di_gen;		/* generation number */
900 
901 	/* di_next_unlinked is the only non-core field in the old dinode */
902 	__be32		di_next_unlinked;/* agi unlinked list ptr */
903 
904 	/* start of the extended dinode, writable fields */
905 	__le32		di_crc;		/* CRC of the inode */
906 	__be64		di_changecount;	/* number of attribute changes */
907 	__be64		di_lsn;		/* flush sequence */
908 	__be64		di_flags2;	/* more random flags */
909 	__be32		di_cowextsize;	/* basic cow extent size for file */
910 	__u8		di_pad2[12];	/* more padding for future expansion */
911 
912 	/* fields only written to during inode creation */
913 	xfs_timestamp_t	di_crtime;	/* time created */
914 	__be64		di_ino;		/* inode number */
915 	uuid_t		di_uuid;	/* UUID of the filesystem */
916 
917 	/* structure must be padded to 64 bit alignment */
918 } xfs_dinode_t;
919 
920 #define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
921 
922 #define DI_MAX_FLUSH 0xffff
923 
924 /*
925  * Size of the core inode on disk.  Version 1 and 2 inodes have
926  * the same size, but version 3 has grown a few additional fields.
927  */
928 static inline uint xfs_dinode_size(int version)
929 {
930 	if (version == 3)
931 		return sizeof(struct xfs_dinode);
932 	return offsetof(struct xfs_dinode, di_crc);
933 }
934 
935 /*
936  * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
937  * Since the pathconf interface is signed, we use 2^31 - 1 instead.
938  */
939 #define	XFS_MAXLINK		((1U << 31) - 1U)
940 
941 /*
942  * Values for di_format
943  */
944 typedef enum xfs_dinode_fmt {
945 	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
946 	XFS_DINODE_FMT_LOCAL,		/* bulk data */
947 	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
948 	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
949 	XFS_DINODE_FMT_UUID		/* added long ago, but never used */
950 } xfs_dinode_fmt_t;
951 
952 /*
953  * Inode minimum and maximum sizes.
954  */
955 #define	XFS_DINODE_MIN_LOG	8
956 #define	XFS_DINODE_MAX_LOG	11
957 #define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
958 #define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
959 
960 /*
961  * Inode size for given fs.
962  */
963 #define XFS_LITINO(mp, version) \
964 	((int)(((mp)->m_sb.sb_inodesize) - xfs_dinode_size(version)))
965 
966 /*
967  * Inode data & attribute fork sizes, per inode.
968  */
969 #define XFS_DFORK_Q(dip)		((dip)->di_forkoff != 0)
970 #define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
971 
972 #define XFS_DFORK_DSIZE(dip,mp) \
973 	(XFS_DFORK_Q(dip) ? \
974 		XFS_DFORK_BOFF(dip) : \
975 		XFS_LITINO(mp, (dip)->di_version))
976 #define XFS_DFORK_ASIZE(dip,mp) \
977 	(XFS_DFORK_Q(dip) ? \
978 		XFS_LITINO(mp, (dip)->di_version) - XFS_DFORK_BOFF(dip) : \
979 		0)
980 #define XFS_DFORK_SIZE(dip,mp,w) \
981 	((w) == XFS_DATA_FORK ? \
982 		XFS_DFORK_DSIZE(dip, mp) : \
983 		XFS_DFORK_ASIZE(dip, mp))
984 
985 /*
986  * Return pointers to the data or attribute forks.
987  */
988 #define XFS_DFORK_DPTR(dip) \
989 	((char *)dip + xfs_dinode_size(dip->di_version))
990 #define XFS_DFORK_APTR(dip)	\
991 	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
992 #define XFS_DFORK_PTR(dip,w)	\
993 	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
994 
995 #define XFS_DFORK_FORMAT(dip,w) \
996 	((w) == XFS_DATA_FORK ? \
997 		(dip)->di_format : \
998 		(dip)->di_aformat)
999 #define XFS_DFORK_NEXTENTS(dip,w) \
1000 	((w) == XFS_DATA_FORK ? \
1001 		be32_to_cpu((dip)->di_nextents) : \
1002 		be16_to_cpu((dip)->di_anextents))
1003 
1004 /*
1005  * For block and character special files the 32bit dev_t is stored at the
1006  * beginning of the data fork.
1007  */
1008 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1009 {
1010 	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1011 }
1012 
1013 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1014 {
1015 	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1016 }
1017 
1018 /*
1019  * Values for di_flags
1020  */
1021 #define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
1022 #define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
1023 #define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
1024 #define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
1025 #define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
1026 #define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
1027 #define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
1028 #define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
1029 #define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
1030 #define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
1031 #define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
1032 #define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
1033 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
1034 #define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
1035 #define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
1036 #define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
1037 #define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
1038 #define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
1039 #define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1040 #define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
1041 #define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
1042 #define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
1043 #define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
1044 #define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
1045 #define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1046 #define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1047 #define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
1048 #define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1049 #define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
1050 #define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
1051 
1052 #define XFS_DIFLAG_ANY \
1053 	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1054 	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1055 	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1056 	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1057 	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1058 
1059 /*
1060  * Values for di_flags2 These start by being exposed to userspace in the upper
1061  * 16 bits of the XFS_XFLAG_s range.
1062  */
1063 #define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
1064 #define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
1065 #define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
1066 #define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
1067 #define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
1068 #define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1069 
1070 #define XFS_DIFLAG2_ANY \
1071 	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)
1072 
1073 /*
1074  * Inode number format:
1075  * low inopblog bits - offset in block
1076  * next agblklog bits - block number in ag
1077  * next agno_log bits - ag number
1078  * high agno_log-agblklog-inopblog bits - 0
1079  */
1080 #define	XFS_INO_MASK(k)			(uint32_t)((1ULL << (k)) - 1)
1081 #define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1082 #define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1083 #define	XFS_INO_AGINO_BITS(mp)		(mp)->m_agino_log
1084 #define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1085 #define	XFS_INO_BITS(mp)		\
1086 	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1087 #define	XFS_INO_TO_AGNO(mp,i)		\
1088 	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1089 #define	XFS_INO_TO_AGINO(mp,i)		\
1090 	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1091 #define	XFS_INO_TO_AGBNO(mp,i)		\
1092 	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1093 		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1094 #define	XFS_INO_TO_OFFSET(mp,i)		\
1095 	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1096 #define	XFS_INO_TO_FSB(mp,i)		\
1097 	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1098 #define	XFS_AGINO_TO_INO(mp,a,i)	\
1099 	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1100 #define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1101 #define	XFS_AGINO_TO_OFFSET(mp,i)	\
1102 	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1103 #define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1104 	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1105 
1106 #define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1107 #define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1108 
1109 /*
1110  * RealTime Device format definitions
1111  */
1112 
1113 /* Min and max rt extent sizes, specified in bytes */
1114 #define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1115 #define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1116 #define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1117 
1118 #define	XFS_BLOCKSIZE(mp)	((mp)->m_sb.sb_blocksize)
1119 #define	XFS_BLOCKMASK(mp)	((mp)->m_blockmask)
1120 #define	XFS_BLOCKWSIZE(mp)	((mp)->m_blockwsize)
1121 #define	XFS_BLOCKWMASK(mp)	((mp)->m_blockwmask)
1122 
1123 /*
1124  * RT Summary and bit manipulation macros.
1125  */
1126 #define	XFS_SUMOFFS(mp,ls,bb)	((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb)))
1127 #define	XFS_SUMOFFSTOBLOCK(mp,s)	\
1128 	(((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog)
1129 #define	XFS_SUMPTR(mp,bp,so)	\
1130 	((xfs_suminfo_t *)((bp)->b_addr + \
1131 		(((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp))))
1132 
1133 #define	XFS_BITTOBLOCK(mp,bi)	((bi) >> (mp)->m_blkbit_log)
1134 #define	XFS_BLOCKTOBIT(mp,bb)	((bb) << (mp)->m_blkbit_log)
1135 #define	XFS_BITTOWORD(mp,bi)	\
1136 	((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp)))
1137 
1138 #define	XFS_RTMIN(a,b)	((a) < (b) ? (a) : (b))
1139 #define	XFS_RTMAX(a,b)	((a) > (b) ? (a) : (b))
1140 
1141 #define	XFS_RTLOBIT(w)	xfs_lowbit32(w)
1142 #define	XFS_RTHIBIT(w)	xfs_highbit32(w)
1143 
1144 #define	XFS_RTBLOCKLOG(b)	xfs_highbit64(b)
1145 
1146 /*
1147  * Dquot and dquot block format definitions
1148  */
1149 #define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1150 #define XFS_DQUOT_VERSION	(uint8_t)0x01	/* latest version number */
1151 
1152 /*
1153  * This is the main portion of the on-disk representation of quota
1154  * information for a user. This is the q_core of the xfs_dquot_t that
1155  * is kept in kernel memory. We pad this with some more expansion room
1156  * to construct the on disk structure.
1157  */
1158 typedef struct	xfs_disk_dquot {
1159 	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1160 	__u8		d_version;	/* dquot version */
1161 	__u8		d_flags;	/* XFS_DQ_USER/PROJ/GROUP */
1162 	__be32		d_id;		/* user,project,group id */
1163 	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1164 	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1165 	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1166 	__be64		d_ino_softlimit;/* preferred inode limit */
1167 	__be64		d_bcount;	/* disk blocks owned by the user */
1168 	__be64		d_icount;	/* inodes owned by the user */
1169 	__be32		d_itimer;	/* zero if within inode limits if not,
1170 					   this is when we refuse service */
1171 	__be32		d_btimer;	/* similar to above; for disk blocks */
1172 	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1173 	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1174 	__be32		d_pad0;		/* 64 bit align */
1175 	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1176 	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1177 	__be64		d_rtbcount;	/* realtime blocks owned */
1178 	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1179 	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1180 	__be16		d_pad;
1181 } xfs_disk_dquot_t;
1182 
1183 /*
1184  * This is what goes on disk. This is separated from the xfs_disk_dquot because
1185  * carrying the unnecessary padding would be a waste of memory.
1186  */
1187 typedef struct xfs_dqblk {
1188 	xfs_disk_dquot_t  dd_diskdq;	/* portion that lives incore as well */
1189 	char		  dd_fill[4];	/* filling for posterity */
1190 
1191 	/*
1192 	 * These two are only present on filesystems with the CRC bits set.
1193 	 */
1194 	__be32		  dd_crc;	/* checksum */
1195 	__be64		  dd_lsn;	/* last modification in log */
1196 	uuid_t		  dd_uuid;	/* location information */
1197 } xfs_dqblk_t;
1198 
1199 #define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1200 
1201 /*
1202  * Remote symlink format and access functions.
1203  */
1204 #define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1205 
1206 struct xfs_dsymlink_hdr {
1207 	__be32	sl_magic;
1208 	__be32	sl_offset;
1209 	__be32	sl_bytes;
1210 	__be32	sl_crc;
1211 	uuid_t	sl_uuid;
1212 	__be64	sl_owner;
1213 	__be64	sl_blkno;
1214 	__be64	sl_lsn;
1215 };
1216 
1217 #define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1218 
1219 #define XFS_SYMLINK_MAXLEN	1024
1220 /*
1221  * The maximum pathlen is 1024 bytes. Since the minimum file system
1222  * blocksize is 512 bytes, we can get a max of 3 extents back from
1223  * bmapi when crc headers are taken into account.
1224  */
1225 #define XFS_SYMLINK_MAPS 3
1226 
1227 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1228 	((bufsize) - (xfs_sb_version_hascrc(&(mp)->m_sb) ? \
1229 			sizeof(struct xfs_dsymlink_hdr) : 0))
1230 
1231 
1232 /*
1233  * Allocation Btree format definitions
1234  *
1235  * There are two on-disk btrees, one sorted by blockno and one sorted
1236  * by blockcount and blockno.  All blocks look the same to make the code
1237  * simpler; if we have time later, we'll make the optimizations.
1238  */
1239 #define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1240 #define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1241 #define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1242 #define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1243 
1244 /*
1245  * Data record/key structure
1246  */
1247 typedef struct xfs_alloc_rec {
1248 	__be32		ar_startblock;	/* starting block number */
1249 	__be32		ar_blockcount;	/* count of free blocks */
1250 } xfs_alloc_rec_t, xfs_alloc_key_t;
1251 
1252 typedef struct xfs_alloc_rec_incore {
1253 	xfs_agblock_t	ar_startblock;	/* starting block number */
1254 	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1255 } xfs_alloc_rec_incore_t;
1256 
1257 /* btree pointer type */
1258 typedef __be32 xfs_alloc_ptr_t;
1259 
1260 /*
1261  * Block numbers in the AG:
1262  * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1263  */
1264 #define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1265 #define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1266 
1267 
1268 /*
1269  * Inode Allocation Btree format definitions
1270  *
1271  * There is a btree for the inode map per allocation group.
1272  */
1273 #define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1274 #define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1275 #define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1276 #define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1277 
1278 typedef uint64_t	xfs_inofree_t;
1279 #define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1280 #define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1281 #define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1282 #define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1283 
1284 #define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1285 #define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(uint16_t))
1286 #define XFS_INODES_PER_HOLEMASK_BIT	\
1287 	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1288 
1289 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1290 {
1291 	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1292 }
1293 
1294 /*
1295  * The on-disk inode record structure has two formats. The original "full"
1296  * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1297  * and replaces the 3 high-order freecount bytes wth the holemask and inode
1298  * count.
1299  *
1300  * The holemask of the sparse record format allows an inode chunk to have holes
1301  * that refer to blocks not owned by the inode record. This facilitates inode
1302  * allocation in the event of severe free space fragmentation.
1303  */
1304 typedef struct xfs_inobt_rec {
1305 	__be32		ir_startino;	/* starting inode number */
1306 	union {
1307 		struct {
1308 			__be32	ir_freecount;	/* count of free inodes */
1309 		} f;
1310 		struct {
1311 			__be16	ir_holemask;/* hole mask for sparse chunks */
1312 			__u8	ir_count;	/* total inode count */
1313 			__u8	ir_freecount;	/* count of free inodes */
1314 		} sp;
1315 	} ir_u;
1316 	__be64		ir_free;	/* free inode mask */
1317 } xfs_inobt_rec_t;
1318 
1319 typedef struct xfs_inobt_rec_incore {
1320 	xfs_agino_t	ir_startino;	/* starting inode number */
1321 	uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1322 	uint8_t		ir_count;	/* total inode count */
1323 	uint8_t		ir_freecount;	/* count of free inodes (set bits) */
1324 	xfs_inofree_t	ir_free;	/* free inode mask */
1325 } xfs_inobt_rec_incore_t;
1326 
1327 static inline bool xfs_inobt_issparse(uint16_t holemask)
1328 {
1329 	/* non-zero holemask represents a sparse rec. */
1330 	return holemask;
1331 }
1332 
1333 /*
1334  * Key structure
1335  */
1336 typedef struct xfs_inobt_key {
1337 	__be32		ir_startino;	/* starting inode number */
1338 } xfs_inobt_key_t;
1339 
1340 /* btree pointer type */
1341 typedef __be32 xfs_inobt_ptr_t;
1342 
1343 /*
1344  * block numbers in the AG.
1345  */
1346 #define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1347 #define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1348 
1349 /*
1350  * Reverse mapping btree format definitions
1351  *
1352  * There is a btree for the reverse map per allocation group
1353  */
1354 #define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1355 
1356 /*
1357  * Ownership info for an extent.  This is used to create reverse-mapping
1358  * entries.
1359  */
1360 #define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1361 #define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1362 struct xfs_owner_info {
1363 	uint64_t		oi_owner;
1364 	xfs_fileoff_t		oi_offset;
1365 	unsigned int		oi_flags;
1366 };
1367 
1368 /*
1369  * Special owner types.
1370  *
1371  * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1372  * to tell us we have a special owner value. We use these for static metadata
1373  * allocated at mkfs/growfs time, as well as for freespace management metadata.
1374  */
1375 #define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1376 #define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1377 #define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1378 #define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1379 #define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1380 #define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1381 #define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1382 #define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1383 #define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1384 #define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1385 
1386 #define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1387 
1388 /*
1389  * Data record structure
1390  */
1391 struct xfs_rmap_rec {
1392 	__be32		rm_startblock;	/* extent start block */
1393 	__be32		rm_blockcount;	/* extent length */
1394 	__be64		rm_owner;	/* extent owner */
1395 	__be64		rm_offset;	/* offset within the owner */
1396 };
1397 
1398 /*
1399  * rmap btree record
1400  *  rm_offset:63 is the attribute fork flag
1401  *  rm_offset:62 is the bmbt block flag
1402  *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1403  *  rm_offset:54-60 aren't used and should be zero
1404  *  rm_offset:0-53 is the block offset within the inode
1405  */
1406 #define XFS_RMAP_OFF_ATTR_FORK	((uint64_t)1ULL << 63)
1407 #define XFS_RMAP_OFF_BMBT_BLOCK	((uint64_t)1ULL << 62)
1408 #define XFS_RMAP_OFF_UNWRITTEN	((uint64_t)1ULL << 61)
1409 
1410 #define XFS_RMAP_LEN_MAX	((uint32_t)~0U)
1411 #define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1412 				 XFS_RMAP_OFF_BMBT_BLOCK | \
1413 				 XFS_RMAP_OFF_UNWRITTEN)
1414 #define XFS_RMAP_OFF_MASK	((uint64_t)0x3FFFFFFFFFFFFFULL)
1415 
1416 #define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1417 
1418 #define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1419 #define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1420 #define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1421 
1422 #define RMAPBT_STARTBLOCK_BITLEN	32
1423 #define RMAPBT_BLOCKCOUNT_BITLEN	32
1424 #define RMAPBT_OWNER_BITLEN		64
1425 #define RMAPBT_ATTRFLAG_BITLEN		1
1426 #define RMAPBT_BMBTFLAG_BITLEN		1
1427 #define RMAPBT_EXNTFLAG_BITLEN		1
1428 #define RMAPBT_UNUSED_OFFSET_BITLEN	7
1429 #define RMAPBT_OFFSET_BITLEN		54
1430 
1431 #define XFS_RMAP_ATTR_FORK		(1 << 0)
1432 #define XFS_RMAP_BMBT_BLOCK		(1 << 1)
1433 #define XFS_RMAP_UNWRITTEN		(1 << 2)
1434 #define XFS_RMAP_KEY_FLAGS		(XFS_RMAP_ATTR_FORK | \
1435 					 XFS_RMAP_BMBT_BLOCK)
1436 #define XFS_RMAP_REC_FLAGS		(XFS_RMAP_UNWRITTEN)
1437 struct xfs_rmap_irec {
1438 	xfs_agblock_t	rm_startblock;	/* extent start block */
1439 	xfs_extlen_t	rm_blockcount;	/* extent length */
1440 	uint64_t	rm_owner;	/* extent owner */
1441 	uint64_t	rm_offset;	/* offset within the owner */
1442 	unsigned int	rm_flags;	/* state flags */
1443 };
1444 
1445 /*
1446  * Key structure
1447  *
1448  * We don't use the length for lookups
1449  */
1450 struct xfs_rmap_key {
1451 	__be32		rm_startblock;	/* extent start block */
1452 	__be64		rm_owner;	/* extent owner */
1453 	__be64		rm_offset;	/* offset within the owner */
1454 } __attribute__((packed));
1455 
1456 /* btree pointer type */
1457 typedef __be32 xfs_rmap_ptr_t;
1458 
1459 #define	XFS_RMAP_BLOCK(mp) \
1460 	(xfs_sb_version_hasfinobt(&((mp)->m_sb)) ? \
1461 	 XFS_FIBT_BLOCK(mp) + 1 : \
1462 	 XFS_IBT_BLOCK(mp) + 1)
1463 
1464 /*
1465  * Reference Count Btree format definitions
1466  *
1467  */
1468 #define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1469 
1470 unsigned int xfs_refc_block(struct xfs_mount *mp);
1471 
1472 /*
1473  * Data record/key structure
1474  *
1475  * Each record associates a range of physical blocks (starting at
1476  * rc_startblock and ending rc_blockcount blocks later) with a reference
1477  * count (rc_refcount).  Extents that are being used to stage a copy on
1478  * write (CoW) operation are recorded in the refcount btree with a
1479  * refcount of 1.  All other records must have a refcount > 1 and must
1480  * track an extent mapped only by file data forks.
1481  *
1482  * Extents with a single owner (attributes, metadata, non-shared file
1483  * data) are not tracked here.  Free space is also not tracked here.
1484  * This is consistent with pre-reflink XFS.
1485  */
1486 
1487 /*
1488  * Extents that are being used to stage a copy on write are stored
1489  * in the refcount btree with a refcount of 1 and the upper bit set
1490  * on the startblock.  This speeds up mount time deletion of stale
1491  * staging extents because they're all at the right side of the tree.
1492  */
1493 #define XFS_REFC_COW_START		((xfs_agblock_t)(1U << 31))
1494 #define REFCNTBT_COWFLAG_BITLEN		1
1495 #define REFCNTBT_AGBLOCK_BITLEN		31
1496 
1497 struct xfs_refcount_rec {
1498 	__be32		rc_startblock;	/* starting block number */
1499 	__be32		rc_blockcount;	/* count of blocks */
1500 	__be32		rc_refcount;	/* number of inodes linked here */
1501 };
1502 
1503 struct xfs_refcount_key {
1504 	__be32		rc_startblock;	/* starting block number */
1505 };
1506 
1507 struct xfs_refcount_irec {
1508 	xfs_agblock_t	rc_startblock;	/* starting block number */
1509 	xfs_extlen_t	rc_blockcount;	/* count of free blocks */
1510 	xfs_nlink_t	rc_refcount;	/* number of inodes linked here */
1511 };
1512 
1513 #define MAXREFCOUNT	((xfs_nlink_t)~0U)
1514 #define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
1515 
1516 /* btree pointer type */
1517 typedef __be32 xfs_refcount_ptr_t;
1518 
1519 
1520 /*
1521  * BMAP Btree format definitions
1522  *
1523  * This includes both the root block definition that sits inside an inode fork
1524  * and the record/pointer formats for the leaf/node in the blocks.
1525  */
1526 #define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1527 #define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1528 
1529 /*
1530  * Bmap root header, on-disk form only.
1531  */
1532 typedef struct xfs_bmdr_block {
1533 	__be16		bb_level;	/* 0 is a leaf */
1534 	__be16		bb_numrecs;	/* current # of data records */
1535 } xfs_bmdr_block_t;
1536 
1537 /*
1538  * Bmap btree record and extent descriptor.
1539  *  l0:63 is an extent flag (value 1 indicates non-normal).
1540  *  l0:9-62 are startoff.
1541  *  l0:0-8 and l1:21-63 are startblock.
1542  *  l1:0-20 are blockcount.
1543  */
1544 #define BMBT_EXNTFLAG_BITLEN	1
1545 #define BMBT_STARTOFF_BITLEN	54
1546 #define BMBT_STARTBLOCK_BITLEN	52
1547 #define BMBT_BLOCKCOUNT_BITLEN	21
1548 
1549 typedef struct xfs_bmbt_rec {
1550 	__be64			l0, l1;
1551 } xfs_bmbt_rec_t;
1552 
1553 typedef uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1554 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1555 
1556 /*
1557  * Values and macros for delayed-allocation startblock fields.
1558  */
1559 #define STARTBLOCKVALBITS	17
1560 #define STARTBLOCKMASKBITS	(15 + 20)
1561 #define STARTBLOCKMASK		\
1562 	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1563 
1564 static inline int isnullstartblock(xfs_fsblock_t x)
1565 {
1566 	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1567 }
1568 
1569 static inline xfs_fsblock_t nullstartblock(int k)
1570 {
1571 	ASSERT(k < (1 << STARTBLOCKVALBITS));
1572 	return STARTBLOCKMASK | (k);
1573 }
1574 
1575 static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1576 {
1577 	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1578 }
1579 
1580 /*
1581  * Key structure for non-leaf levels of the tree.
1582  */
1583 typedef struct xfs_bmbt_key {
1584 	__be64		br_startoff;	/* starting file offset */
1585 } xfs_bmbt_key_t, xfs_bmdr_key_t;
1586 
1587 /* btree pointer type */
1588 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1589 
1590 
1591 /*
1592  * Generic Btree block format definitions
1593  *
1594  * This is a combination of the actual format used on disk for short and long
1595  * format btrees.  The first three fields are shared by both format, but the
1596  * pointers are different and should be used with care.
1597  *
1598  * To get the size of the actual short or long form headers please use the size
1599  * macros below.  Never use sizeof(xfs_btree_block).
1600  *
1601  * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1602  * with the crc feature bit, and all accesses to them must be conditional on
1603  * that flag.
1604  */
1605 /* short form block header */
1606 struct xfs_btree_block_shdr {
1607 	__be32		bb_leftsib;
1608 	__be32		bb_rightsib;
1609 
1610 	__be64		bb_blkno;
1611 	__be64		bb_lsn;
1612 	uuid_t		bb_uuid;
1613 	__be32		bb_owner;
1614 	__le32		bb_crc;
1615 };
1616 
1617 /* long form block header */
1618 struct xfs_btree_block_lhdr {
1619 	__be64		bb_leftsib;
1620 	__be64		bb_rightsib;
1621 
1622 	__be64		bb_blkno;
1623 	__be64		bb_lsn;
1624 	uuid_t		bb_uuid;
1625 	__be64		bb_owner;
1626 	__le32		bb_crc;
1627 	__be32		bb_pad; /* padding for alignment */
1628 };
1629 
1630 struct xfs_btree_block {
1631 	__be32		bb_magic;	/* magic number for block type */
1632 	__be16		bb_level;	/* 0 is a leaf */
1633 	__be16		bb_numrecs;	/* current # of data records */
1634 	union {
1635 		struct xfs_btree_block_shdr s;
1636 		struct xfs_btree_block_lhdr l;
1637 	} bb_u;				/* rest */
1638 };
1639 
1640 /* size of a short form block */
1641 #define XFS_BTREE_SBLOCK_LEN \
1642 	(offsetof(struct xfs_btree_block, bb_u) + \
1643 	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1644 /* size of a long form block */
1645 #define XFS_BTREE_LBLOCK_LEN \
1646 	(offsetof(struct xfs_btree_block, bb_u) + \
1647 	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1648 
1649 /* sizes of CRC enabled btree blocks */
1650 #define XFS_BTREE_SBLOCK_CRC_LEN \
1651 	(offsetof(struct xfs_btree_block, bb_u) + \
1652 	 sizeof(struct xfs_btree_block_shdr))
1653 #define XFS_BTREE_LBLOCK_CRC_LEN \
1654 	(offsetof(struct xfs_btree_block, bb_u) + \
1655 	 sizeof(struct xfs_btree_block_lhdr))
1656 
1657 #define XFS_BTREE_SBLOCK_CRC_OFF \
1658 	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1659 #define XFS_BTREE_LBLOCK_CRC_OFF \
1660 	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1661 
1662 /*
1663  * On-disk XFS access control list structure.
1664  */
1665 struct xfs_acl_entry {
1666 	__be32	ae_tag;
1667 	__be32	ae_id;
1668 	__be16	ae_perm;
1669 	__be16	ae_pad;		/* fill the implicit hole in the structure */
1670 };
1671 
1672 struct xfs_acl {
1673 	__be32			acl_cnt;
1674 	struct xfs_acl_entry	acl_entry[0];
1675 };
1676 
1677 /*
1678  * The number of ACL entries allowed is defined by the on-disk format.
1679  * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1680  * limited only by the maximum size of the xattr that stores the information.
1681  */
1682 #define XFS_ACL_MAX_ENTRIES(mp)	\
1683 	(xfs_sb_version_hascrc(&mp->m_sb) \
1684 		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1685 						sizeof(struct xfs_acl_entry) \
1686 		: 25)
1687 
1688 #define XFS_ACL_SIZE(cnt) \
1689 	(sizeof(struct xfs_acl) + \
1690 		sizeof(struct xfs_acl_entry) * cnt)
1691 
1692 #define XFS_ACL_MAX_SIZE(mp) \
1693 	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1694 
1695 
1696 /* On-disk XFS extended attribute names */
1697 #define SGI_ACL_FILE		"SGI_ACL_FILE"
1698 #define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
1699 #define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
1700 #define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
1701 
1702 #endif /* __XFS_FORMAT_H__ */
1703