xref: /openbmc/linux/fs/xfs/libxfs/xfs_format.h (revision ae213c44)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4  * All Rights Reserved.
5  */
6 #ifndef __XFS_FORMAT_H__
7 #define __XFS_FORMAT_H__
8 
9 /*
10  * XFS On Disk Format Definitions
11  *
12  * This header file defines all the on-disk format definitions for
13  * general XFS objects. Directory and attribute related objects are defined in
14  * xfs_da_format.h, which log and log item formats are defined in
15  * xfs_log_format.h. Everything else goes here.
16  */
17 
18 struct xfs_mount;
19 struct xfs_trans;
20 struct xfs_inode;
21 struct xfs_buf;
22 struct xfs_ifork;
23 
24 /*
25  * Super block
26  * Fits into a sector-sized buffer at address 0 of each allocation group.
27  * Only the first of these is ever updated except during growfs.
28  */
29 #define	XFS_SB_MAGIC		0x58465342	/* 'XFSB' */
30 #define	XFS_SB_VERSION_1	1		/* 5.3, 6.0.1, 6.1 */
31 #define	XFS_SB_VERSION_2	2		/* 6.2 - attributes */
32 #define	XFS_SB_VERSION_3	3		/* 6.2 - new inode version */
33 #define	XFS_SB_VERSION_4	4		/* 6.2+ - bitmask version */
34 #define	XFS_SB_VERSION_5	5		/* CRC enabled filesystem */
35 #define	XFS_SB_VERSION_NUMBITS		0x000f
36 #define	XFS_SB_VERSION_ALLFBITS		0xfff0
37 #define	XFS_SB_VERSION_ATTRBIT		0x0010
38 #define	XFS_SB_VERSION_NLINKBIT		0x0020
39 #define	XFS_SB_VERSION_QUOTABIT		0x0040
40 #define	XFS_SB_VERSION_ALIGNBIT		0x0080
41 #define	XFS_SB_VERSION_DALIGNBIT	0x0100
42 #define	XFS_SB_VERSION_SHAREDBIT	0x0200
43 #define XFS_SB_VERSION_LOGV2BIT		0x0400
44 #define XFS_SB_VERSION_SECTORBIT	0x0800
45 #define	XFS_SB_VERSION_EXTFLGBIT	0x1000
46 #define	XFS_SB_VERSION_DIRV2BIT		0x2000
47 #define	XFS_SB_VERSION_BORGBIT		0x4000	/* ASCII only case-insens. */
48 #define	XFS_SB_VERSION_MOREBITSBIT	0x8000
49 
50 /*
51  * The size of a single extended attribute on disk is limited by
52  * the size of index values within the attribute entries themselves.
53  * These are be16 fields, so we can only support attribute data
54  * sizes up to 2^16 bytes in length.
55  */
56 #define XFS_XATTR_SIZE_MAX (1 << 16)
57 
58 /*
59  * Supported feature bit list is just all bits in the versionnum field because
60  * we've used them all up and understand them all. Except, of course, for the
61  * shared superblock bit, which nobody knows what it does and so is unsupported.
62  */
63 #define	XFS_SB_VERSION_OKBITS		\
64 	((XFS_SB_VERSION_NUMBITS | XFS_SB_VERSION_ALLFBITS) & \
65 		~XFS_SB_VERSION_SHAREDBIT)
66 
67 /*
68  * There are two words to hold XFS "feature" bits: the original
69  * word, sb_versionnum, and sb_features2.  Whenever a bit is set in
70  * sb_features2, the feature bit XFS_SB_VERSION_MOREBITSBIT must be set.
71  *
72  * These defines represent bits in sb_features2.
73  */
74 #define XFS_SB_VERSION2_RESERVED1BIT	0x00000001
75 #define XFS_SB_VERSION2_LAZYSBCOUNTBIT	0x00000002	/* Superblk counters */
76 #define XFS_SB_VERSION2_RESERVED4BIT	0x00000004
77 #define XFS_SB_VERSION2_ATTR2BIT	0x00000008	/* Inline attr rework */
78 #define XFS_SB_VERSION2_PARENTBIT	0x00000010	/* parent pointers */
79 #define XFS_SB_VERSION2_PROJID32BIT	0x00000080	/* 32 bit project id */
80 #define XFS_SB_VERSION2_CRCBIT		0x00000100	/* metadata CRCs */
81 #define XFS_SB_VERSION2_FTYPE		0x00000200	/* inode type in dir */
82 
83 #define	XFS_SB_VERSION2_OKBITS		\
84 	(XFS_SB_VERSION2_LAZYSBCOUNTBIT	| \
85 	 XFS_SB_VERSION2_ATTR2BIT	| \
86 	 XFS_SB_VERSION2_PROJID32BIT	| \
87 	 XFS_SB_VERSION2_FTYPE)
88 
89 /* Maximum size of the xfs filesystem label, no terminating NULL */
90 #define XFSLABEL_MAX			12
91 
92 /*
93  * Superblock - in core version.  Must match the ondisk version below.
94  * Must be padded to 64 bit alignment.
95  */
96 typedef struct xfs_sb {
97 	uint32_t	sb_magicnum;	/* magic number == XFS_SB_MAGIC */
98 	uint32_t	sb_blocksize;	/* logical block size, bytes */
99 	xfs_rfsblock_t	sb_dblocks;	/* number of data blocks */
100 	xfs_rfsblock_t	sb_rblocks;	/* number of realtime blocks */
101 	xfs_rtblock_t	sb_rextents;	/* number of realtime extents */
102 	uuid_t		sb_uuid;	/* user-visible file system unique id */
103 	xfs_fsblock_t	sb_logstart;	/* starting block of log if internal */
104 	xfs_ino_t	sb_rootino;	/* root inode number */
105 	xfs_ino_t	sb_rbmino;	/* bitmap inode for realtime extents */
106 	xfs_ino_t	sb_rsumino;	/* summary inode for rt bitmap */
107 	xfs_agblock_t	sb_rextsize;	/* realtime extent size, blocks */
108 	xfs_agblock_t	sb_agblocks;	/* size of an allocation group */
109 	xfs_agnumber_t	sb_agcount;	/* number of allocation groups */
110 	xfs_extlen_t	sb_rbmblocks;	/* number of rt bitmap blocks */
111 	xfs_extlen_t	sb_logblocks;	/* number of log blocks */
112 	uint16_t	sb_versionnum;	/* header version == XFS_SB_VERSION */
113 	uint16_t	sb_sectsize;	/* volume sector size, bytes */
114 	uint16_t	sb_inodesize;	/* inode size, bytes */
115 	uint16_t	sb_inopblock;	/* inodes per block */
116 	char		sb_fname[XFSLABEL_MAX]; /* file system name */
117 	uint8_t		sb_blocklog;	/* log2 of sb_blocksize */
118 	uint8_t		sb_sectlog;	/* log2 of sb_sectsize */
119 	uint8_t		sb_inodelog;	/* log2 of sb_inodesize */
120 	uint8_t		sb_inopblog;	/* log2 of sb_inopblock */
121 	uint8_t		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
122 	uint8_t		sb_rextslog;	/* log2 of sb_rextents */
123 	uint8_t		sb_inprogress;	/* mkfs is in progress, don't mount */
124 	uint8_t		sb_imax_pct;	/* max % of fs for inode space */
125 					/* statistics */
126 	/*
127 	 * These fields must remain contiguous.  If you really
128 	 * want to change their layout, make sure you fix the
129 	 * code in xfs_trans_apply_sb_deltas().
130 	 */
131 	uint64_t	sb_icount;	/* allocated inodes */
132 	uint64_t	sb_ifree;	/* free inodes */
133 	uint64_t	sb_fdblocks;	/* free data blocks */
134 	uint64_t	sb_frextents;	/* free realtime extents */
135 	/*
136 	 * End contiguous fields.
137 	 */
138 	xfs_ino_t	sb_uquotino;	/* user quota inode */
139 	xfs_ino_t	sb_gquotino;	/* group quota inode */
140 	uint16_t	sb_qflags;	/* quota flags */
141 	uint8_t		sb_flags;	/* misc. flags */
142 	uint8_t		sb_shared_vn;	/* shared version number */
143 	xfs_extlen_t	sb_inoalignmt;	/* inode chunk alignment, fsblocks */
144 	uint32_t	sb_unit;	/* stripe or raid unit */
145 	uint32_t	sb_width;	/* stripe or raid width */
146 	uint8_t		sb_dirblklog;	/* log2 of dir block size (fsbs) */
147 	uint8_t		sb_logsectlog;	/* log2 of the log sector size */
148 	uint16_t	sb_logsectsize;	/* sector size for the log, bytes */
149 	uint32_t	sb_logsunit;	/* stripe unit size for the log */
150 	uint32_t	sb_features2;	/* additional feature bits */
151 
152 	/*
153 	 * bad features2 field as a result of failing to pad the sb structure to
154 	 * 64 bits. Some machines will be using this field for features2 bits.
155 	 * Easiest just to mark it bad and not use it for anything else.
156 	 *
157 	 * This is not kept up to date in memory; it is always overwritten by
158 	 * the value in sb_features2 when formatting the incore superblock to
159 	 * the disk buffer.
160 	 */
161 	uint32_t	sb_bad_features2;
162 
163 	/* version 5 superblock fields start here */
164 
165 	/* feature masks */
166 	uint32_t	sb_features_compat;
167 	uint32_t	sb_features_ro_compat;
168 	uint32_t	sb_features_incompat;
169 	uint32_t	sb_features_log_incompat;
170 
171 	uint32_t	sb_crc;		/* superblock crc */
172 	xfs_extlen_t	sb_spino_align;	/* sparse inode chunk alignment */
173 
174 	xfs_ino_t	sb_pquotino;	/* project quota inode */
175 	xfs_lsn_t	sb_lsn;		/* last write sequence */
176 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
177 
178 	/* must be padded to 64 bit alignment */
179 } xfs_sb_t;
180 
181 #define XFS_SB_CRC_OFF		offsetof(struct xfs_sb, sb_crc)
182 
183 /*
184  * Superblock - on disk version.  Must match the in core version above.
185  * Must be padded to 64 bit alignment.
186  */
187 typedef struct xfs_dsb {
188 	__be32		sb_magicnum;	/* magic number == XFS_SB_MAGIC */
189 	__be32		sb_blocksize;	/* logical block size, bytes */
190 	__be64		sb_dblocks;	/* number of data blocks */
191 	__be64		sb_rblocks;	/* number of realtime blocks */
192 	__be64		sb_rextents;	/* number of realtime extents */
193 	uuid_t		sb_uuid;	/* user-visible file system unique id */
194 	__be64		sb_logstart;	/* starting block of log if internal */
195 	__be64		sb_rootino;	/* root inode number */
196 	__be64		sb_rbmino;	/* bitmap inode for realtime extents */
197 	__be64		sb_rsumino;	/* summary inode for rt bitmap */
198 	__be32		sb_rextsize;	/* realtime extent size, blocks */
199 	__be32		sb_agblocks;	/* size of an allocation group */
200 	__be32		sb_agcount;	/* number of allocation groups */
201 	__be32		sb_rbmblocks;	/* number of rt bitmap blocks */
202 	__be32		sb_logblocks;	/* number of log blocks */
203 	__be16		sb_versionnum;	/* header version == XFS_SB_VERSION */
204 	__be16		sb_sectsize;	/* volume sector size, bytes */
205 	__be16		sb_inodesize;	/* inode size, bytes */
206 	__be16		sb_inopblock;	/* inodes per block */
207 	char		sb_fname[XFSLABEL_MAX]; /* file system name */
208 	__u8		sb_blocklog;	/* log2 of sb_blocksize */
209 	__u8		sb_sectlog;	/* log2 of sb_sectsize */
210 	__u8		sb_inodelog;	/* log2 of sb_inodesize */
211 	__u8		sb_inopblog;	/* log2 of sb_inopblock */
212 	__u8		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
213 	__u8		sb_rextslog;	/* log2 of sb_rextents */
214 	__u8		sb_inprogress;	/* mkfs is in progress, don't mount */
215 	__u8		sb_imax_pct;	/* max % of fs for inode space */
216 					/* statistics */
217 	/*
218 	 * These fields must remain contiguous.  If you really
219 	 * want to change their layout, make sure you fix the
220 	 * code in xfs_trans_apply_sb_deltas().
221 	 */
222 	__be64		sb_icount;	/* allocated inodes */
223 	__be64		sb_ifree;	/* free inodes */
224 	__be64		sb_fdblocks;	/* free data blocks */
225 	__be64		sb_frextents;	/* free realtime extents */
226 	/*
227 	 * End contiguous fields.
228 	 */
229 	__be64		sb_uquotino;	/* user quota inode */
230 	__be64		sb_gquotino;	/* group quota inode */
231 	__be16		sb_qflags;	/* quota flags */
232 	__u8		sb_flags;	/* misc. flags */
233 	__u8		sb_shared_vn;	/* shared version number */
234 	__be32		sb_inoalignmt;	/* inode chunk alignment, fsblocks */
235 	__be32		sb_unit;	/* stripe or raid unit */
236 	__be32		sb_width;	/* stripe or raid width */
237 	__u8		sb_dirblklog;	/* log2 of dir block size (fsbs) */
238 	__u8		sb_logsectlog;	/* log2 of the log sector size */
239 	__be16		sb_logsectsize;	/* sector size for the log, bytes */
240 	__be32		sb_logsunit;	/* stripe unit size for the log */
241 	__be32		sb_features2;	/* additional feature bits */
242 	/*
243 	 * bad features2 field as a result of failing to pad the sb
244 	 * structure to 64 bits. Some machines will be using this field
245 	 * for features2 bits. Easiest just to mark it bad and not use
246 	 * it for anything else.
247 	 */
248 	__be32		sb_bad_features2;
249 
250 	/* version 5 superblock fields start here */
251 
252 	/* feature masks */
253 	__be32		sb_features_compat;
254 	__be32		sb_features_ro_compat;
255 	__be32		sb_features_incompat;
256 	__be32		sb_features_log_incompat;
257 
258 	__le32		sb_crc;		/* superblock crc */
259 	__be32		sb_spino_align;	/* sparse inode chunk alignment */
260 
261 	__be64		sb_pquotino;	/* project quota inode */
262 	__be64		sb_lsn;		/* last write sequence */
263 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
264 
265 	/* must be padded to 64 bit alignment */
266 } xfs_dsb_t;
267 
268 
269 /*
270  * Misc. Flags - warning - these will be cleared by xfs_repair unless
271  * a feature bit is set when the flag is used.
272  */
273 #define XFS_SBF_NOFLAGS		0x00	/* no flags set */
274 #define XFS_SBF_READONLY	0x01	/* only read-only mounts allowed */
275 
276 /*
277  * define max. shared version we can interoperate with
278  */
279 #define XFS_SB_MAX_SHARED_VN	0
280 
281 #define	XFS_SB_VERSION_NUM(sbp)	((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
282 
283 /*
284  * The first XFS version we support is a v4 superblock with V2 directories.
285  */
286 static inline bool xfs_sb_good_v4_features(struct xfs_sb *sbp)
287 {
288 	if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
289 		return false;
290 	if (!(sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT))
291 		return false;
292 
293 	/* check for unknown features in the fs */
294 	if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
295 	    ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
296 	     (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
297 		return false;
298 
299 	return true;
300 }
301 
302 static inline bool xfs_sb_good_version(struct xfs_sb *sbp)
303 {
304 	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5)
305 		return true;
306 	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_4)
307 		return xfs_sb_good_v4_features(sbp);
308 	return false;
309 }
310 
311 static inline bool xfs_sb_version_hasrealtime(struct xfs_sb *sbp)
312 {
313 	return sbp->sb_rblocks > 0;
314 }
315 
316 /*
317  * Detect a mismatched features2 field.  Older kernels read/wrote
318  * this into the wrong slot, so to be safe we keep them in sync.
319  */
320 static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
321 {
322 	return sbp->sb_bad_features2 != sbp->sb_features2;
323 }
324 
325 static inline bool xfs_sb_version_hasattr(struct xfs_sb *sbp)
326 {
327 	return (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT);
328 }
329 
330 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
331 {
332 	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
333 }
334 
335 static inline bool xfs_sb_version_hasquota(struct xfs_sb *sbp)
336 {
337 	return (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT);
338 }
339 
340 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
341 {
342 	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
343 }
344 
345 static inline bool xfs_sb_version_hasalign(struct xfs_sb *sbp)
346 {
347 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
348 		(sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT));
349 }
350 
351 static inline bool xfs_sb_version_hasdalign(struct xfs_sb *sbp)
352 {
353 	return (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT);
354 }
355 
356 static inline bool xfs_sb_version_haslogv2(struct xfs_sb *sbp)
357 {
358 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
359 	       (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT);
360 }
361 
362 static inline bool xfs_sb_version_hassector(struct xfs_sb *sbp)
363 {
364 	return (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT);
365 }
366 
367 static inline bool xfs_sb_version_hasasciici(struct xfs_sb *sbp)
368 {
369 	return (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT);
370 }
371 
372 static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp)
373 {
374 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
375 	       (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
376 }
377 
378 /*
379  * sb_features2 bit version macros.
380  */
381 static inline bool xfs_sb_version_haslazysbcount(struct xfs_sb *sbp)
382 {
383 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
384 	       (xfs_sb_version_hasmorebits(sbp) &&
385 		(sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT));
386 }
387 
388 static inline bool xfs_sb_version_hasattr2(struct xfs_sb *sbp)
389 {
390 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
391 	       (xfs_sb_version_hasmorebits(sbp) &&
392 		(sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT));
393 }
394 
395 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
396 {
397 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
398 	sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
399 }
400 
401 static inline void xfs_sb_version_removeattr2(struct xfs_sb *sbp)
402 {
403 	sbp->sb_features2 &= ~XFS_SB_VERSION2_ATTR2BIT;
404 	if (!sbp->sb_features2)
405 		sbp->sb_versionnum &= ~XFS_SB_VERSION_MOREBITSBIT;
406 }
407 
408 static inline bool xfs_sb_version_hasprojid32bit(struct xfs_sb *sbp)
409 {
410 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
411 	       (xfs_sb_version_hasmorebits(sbp) &&
412 		(sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT));
413 }
414 
415 static inline void xfs_sb_version_addprojid32bit(struct xfs_sb *sbp)
416 {
417 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
418 	sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
419 }
420 
421 /*
422  * Extended v5 superblock feature masks. These are to be used for new v5
423  * superblock features only.
424  *
425  * Compat features are new features that old kernels will not notice or affect
426  * and so can mount read-write without issues.
427  *
428  * RO-Compat (read only) are features that old kernels can read but will break
429  * if they write. Hence only read-only mounts of such filesystems are allowed on
430  * kernels that don't support the feature bit.
431  *
432  * InCompat features are features which old kernels will not understand and so
433  * must not mount.
434  *
435  * Log-InCompat features are for changes to log formats or new transactions that
436  * can't be replayed on older kernels. The fields are set when the filesystem is
437  * mounted, and a clean unmount clears the fields.
438  */
439 #define XFS_SB_FEAT_COMPAT_ALL 0
440 #define XFS_SB_FEAT_COMPAT_UNKNOWN	~XFS_SB_FEAT_COMPAT_ALL
441 static inline bool
442 xfs_sb_has_compat_feature(
443 	struct xfs_sb	*sbp,
444 	uint32_t	feature)
445 {
446 	return (sbp->sb_features_compat & feature) != 0;
447 }
448 
449 #define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
450 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
451 #define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
452 #define XFS_SB_FEAT_RO_COMPAT_ALL \
453 		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
454 		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
455 		 XFS_SB_FEAT_RO_COMPAT_REFLINK)
456 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
457 static inline bool
458 xfs_sb_has_ro_compat_feature(
459 	struct xfs_sb	*sbp,
460 	uint32_t	feature)
461 {
462 	return (sbp->sb_features_ro_compat & feature) != 0;
463 }
464 
465 #define XFS_SB_FEAT_INCOMPAT_FTYPE	(1 << 0)	/* filetype in dirent */
466 #define XFS_SB_FEAT_INCOMPAT_SPINODES	(1 << 1)	/* sparse inode chunks */
467 #define XFS_SB_FEAT_INCOMPAT_META_UUID	(1 << 2)	/* metadata UUID */
468 #define XFS_SB_FEAT_INCOMPAT_ALL \
469 		(XFS_SB_FEAT_INCOMPAT_FTYPE|	\
470 		 XFS_SB_FEAT_INCOMPAT_SPINODES|	\
471 		 XFS_SB_FEAT_INCOMPAT_META_UUID)
472 
473 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
474 static inline bool
475 xfs_sb_has_incompat_feature(
476 	struct xfs_sb	*sbp,
477 	uint32_t	feature)
478 {
479 	return (sbp->sb_features_incompat & feature) != 0;
480 }
481 
482 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL 0
483 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
484 static inline bool
485 xfs_sb_has_incompat_log_feature(
486 	struct xfs_sb	*sbp,
487 	uint32_t	feature)
488 {
489 	return (sbp->sb_features_log_incompat & feature) != 0;
490 }
491 
492 /*
493  * V5 superblock specific feature checks
494  */
495 static inline bool xfs_sb_version_hascrc(struct xfs_sb *sbp)
496 {
497 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
498 }
499 
500 static inline bool xfs_sb_version_has_pquotino(struct xfs_sb *sbp)
501 {
502 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
503 }
504 
505 static inline int xfs_sb_version_hasftype(struct xfs_sb *sbp)
506 {
507 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
508 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_FTYPE)) ||
509 	       (xfs_sb_version_hasmorebits(sbp) &&
510 		 (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE));
511 }
512 
513 static inline bool xfs_sb_version_hasfinobt(xfs_sb_t *sbp)
514 {
515 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
516 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT);
517 }
518 
519 static inline bool xfs_sb_version_hassparseinodes(struct xfs_sb *sbp)
520 {
521 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
522 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_SPINODES);
523 }
524 
525 /*
526  * XFS_SB_FEAT_INCOMPAT_META_UUID indicates that the metadata UUID
527  * is stored separately from the user-visible UUID; this allows the
528  * user-visible UUID to be changed on V5 filesystems which have a
529  * filesystem UUID stamped into every piece of metadata.
530  */
531 static inline bool xfs_sb_version_hasmetauuid(struct xfs_sb *sbp)
532 {
533 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
534 		(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID);
535 }
536 
537 static inline bool xfs_sb_version_hasrmapbt(struct xfs_sb *sbp)
538 {
539 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
540 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT);
541 }
542 
543 static inline bool xfs_sb_version_hasreflink(struct xfs_sb *sbp)
544 {
545 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
546 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK);
547 }
548 
549 /*
550  * end of superblock version macros
551  */
552 
553 static inline bool
554 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
555 {
556 	return (ino == sbp->sb_uquotino ||
557 		ino == sbp->sb_gquotino ||
558 		ino == sbp->sb_pquotino);
559 }
560 
561 #define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
562 #define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
563 #define XFS_BUF_TO_SBP(bp)	((xfs_dsb_t *)((bp)->b_addr))
564 
565 #define	XFS_HDR_BLOCK(mp,d)	((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
566 #define	XFS_DADDR_TO_FSB(mp,d)	XFS_AGB_TO_FSB(mp, \
567 			xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
568 #define	XFS_FSB_TO_DADDR(mp,fsbno)	XFS_AGB_TO_DADDR(mp, \
569 			XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
570 
571 /*
572  * File system sector to basic block conversions.
573  */
574 #define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
575 
576 /*
577  * File system block to basic block conversions.
578  */
579 #define	XFS_FSB_TO_BB(mp,fsbno)	((fsbno) << (mp)->m_blkbb_log)
580 #define	XFS_BB_TO_FSB(mp,bb)	\
581 	(((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
582 #define	XFS_BB_TO_FSBT(mp,bb)	((bb) >> (mp)->m_blkbb_log)
583 
584 /*
585  * File system block to byte conversions.
586  */
587 #define XFS_FSB_TO_B(mp,fsbno)	((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
588 #define XFS_B_TO_FSB(mp,b)	\
589 	((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
590 #define XFS_B_TO_FSBT(mp,b)	(((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
591 #define XFS_B_FSB_OFFSET(mp,b)	((b) & (mp)->m_blockmask)
592 
593 /*
594  * Allocation group header
595  *
596  * This is divided into three structures, placed in sequential 512-byte
597  * buffers after a copy of the superblock (also in a 512-byte buffer).
598  */
599 #define	XFS_AGF_MAGIC	0x58414746	/* 'XAGF' */
600 #define	XFS_AGI_MAGIC	0x58414749	/* 'XAGI' */
601 #define	XFS_AGFL_MAGIC	0x5841464c	/* 'XAFL' */
602 #define	XFS_AGF_VERSION	1
603 #define	XFS_AGI_VERSION	1
604 
605 #define	XFS_AGF_GOOD_VERSION(v)	((v) == XFS_AGF_VERSION)
606 #define	XFS_AGI_GOOD_VERSION(v)	((v) == XFS_AGI_VERSION)
607 
608 /*
609  * Btree number 0 is bno, 1 is cnt, 2 is rmap. This value gives the size of the
610  * arrays below.
611  */
612 #define	XFS_BTNUM_AGF	((int)XFS_BTNUM_RMAPi + 1)
613 
614 /*
615  * The second word of agf_levels in the first a.g. overlaps the EFS
616  * superblock's magic number.  Since the magic numbers valid for EFS
617  * are > 64k, our value cannot be confused for an EFS superblock's.
618  */
619 
620 typedef struct xfs_agf {
621 	/*
622 	 * Common allocation group header information
623 	 */
624 	__be32		agf_magicnum;	/* magic number == XFS_AGF_MAGIC */
625 	__be32		agf_versionnum;	/* header version == XFS_AGF_VERSION */
626 	__be32		agf_seqno;	/* sequence # starting from 0 */
627 	__be32		agf_length;	/* size in blocks of a.g. */
628 	/*
629 	 * Freespace and rmap information
630 	 */
631 	__be32		agf_roots[XFS_BTNUM_AGF];	/* root blocks */
632 	__be32		agf_levels[XFS_BTNUM_AGF];	/* btree levels */
633 
634 	__be32		agf_flfirst;	/* first freelist block's index */
635 	__be32		agf_fllast;	/* last freelist block's index */
636 	__be32		agf_flcount;	/* count of blocks in freelist */
637 	__be32		agf_freeblks;	/* total free blocks */
638 
639 	__be32		agf_longest;	/* longest free space */
640 	__be32		agf_btreeblks;	/* # of blocks held in AGF btrees */
641 	uuid_t		agf_uuid;	/* uuid of filesystem */
642 
643 	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
644 	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
645 
646 	__be32		agf_refcount_root;	/* refcount tree root block */
647 	__be32		agf_refcount_level;	/* refcount btree levels */
648 
649 	/*
650 	 * reserve some contiguous space for future logged fields before we add
651 	 * the unlogged fields. This makes the range logging via flags and
652 	 * structure offsets much simpler.
653 	 */
654 	__be64		agf_spare64[14];
655 
656 	/* unlogged fields, written during buffer writeback. */
657 	__be64		agf_lsn;	/* last write sequence */
658 	__be32		agf_crc;	/* crc of agf sector */
659 	__be32		agf_spare2;
660 
661 	/* structure must be padded to 64 bit alignment */
662 } xfs_agf_t;
663 
664 #define XFS_AGF_CRC_OFF		offsetof(struct xfs_agf, agf_crc)
665 
666 #define	XFS_AGF_MAGICNUM	0x00000001
667 #define	XFS_AGF_VERSIONNUM	0x00000002
668 #define	XFS_AGF_SEQNO		0x00000004
669 #define	XFS_AGF_LENGTH		0x00000008
670 #define	XFS_AGF_ROOTS		0x00000010
671 #define	XFS_AGF_LEVELS		0x00000020
672 #define	XFS_AGF_FLFIRST		0x00000040
673 #define	XFS_AGF_FLLAST		0x00000080
674 #define	XFS_AGF_FLCOUNT		0x00000100
675 #define	XFS_AGF_FREEBLKS	0x00000200
676 #define	XFS_AGF_LONGEST		0x00000400
677 #define	XFS_AGF_BTREEBLKS	0x00000800
678 #define	XFS_AGF_UUID		0x00001000
679 #define	XFS_AGF_RMAP_BLOCKS	0x00002000
680 #define	XFS_AGF_REFCOUNT_BLOCKS	0x00004000
681 #define	XFS_AGF_REFCOUNT_ROOT	0x00008000
682 #define	XFS_AGF_REFCOUNT_LEVEL	0x00010000
683 #define	XFS_AGF_SPARE64		0x00020000
684 #define	XFS_AGF_NUM_BITS	18
685 #define	XFS_AGF_ALL_BITS	((1 << XFS_AGF_NUM_BITS) - 1)
686 
687 #define XFS_AGF_FLAGS \
688 	{ XFS_AGF_MAGICNUM,	"MAGICNUM" }, \
689 	{ XFS_AGF_VERSIONNUM,	"VERSIONNUM" }, \
690 	{ XFS_AGF_SEQNO,	"SEQNO" }, \
691 	{ XFS_AGF_LENGTH,	"LENGTH" }, \
692 	{ XFS_AGF_ROOTS,	"ROOTS" }, \
693 	{ XFS_AGF_LEVELS,	"LEVELS" }, \
694 	{ XFS_AGF_FLFIRST,	"FLFIRST" }, \
695 	{ XFS_AGF_FLLAST,	"FLLAST" }, \
696 	{ XFS_AGF_FLCOUNT,	"FLCOUNT" }, \
697 	{ XFS_AGF_FREEBLKS,	"FREEBLKS" }, \
698 	{ XFS_AGF_LONGEST,	"LONGEST" }, \
699 	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
700 	{ XFS_AGF_UUID,		"UUID" }, \
701 	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
702 	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
703 	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
704 	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
705 	{ XFS_AGF_SPARE64,	"SPARE64" }
706 
707 /* disk block (xfs_daddr_t) in the AG */
708 #define XFS_AGF_DADDR(mp)	((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
709 #define	XFS_AGF_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
710 #define	XFS_BUF_TO_AGF(bp)	((xfs_agf_t *)((bp)->b_addr))
711 
712 /*
713  * Size of the unlinked inode hash table in the agi.
714  */
715 #define	XFS_AGI_UNLINKED_BUCKETS	64
716 
717 typedef struct xfs_agi {
718 	/*
719 	 * Common allocation group header information
720 	 */
721 	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
722 	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
723 	__be32		agi_seqno;	/* sequence # starting from 0 */
724 	__be32		agi_length;	/* size in blocks of a.g. */
725 	/*
726 	 * Inode information
727 	 * Inodes are mapped by interpreting the inode number, so no
728 	 * mapping data is needed here.
729 	 */
730 	__be32		agi_count;	/* count of allocated inodes */
731 	__be32		agi_root;	/* root of inode btree */
732 	__be32		agi_level;	/* levels in inode btree */
733 	__be32		agi_freecount;	/* number of free inodes */
734 
735 	__be32		agi_newino;	/* new inode just allocated */
736 	__be32		agi_dirino;	/* last directory inode chunk */
737 	/*
738 	 * Hash table of inodes which have been unlinked but are
739 	 * still being referenced.
740 	 */
741 	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
742 	/*
743 	 * This marks the end of logging region 1 and start of logging region 2.
744 	 */
745 	uuid_t		agi_uuid;	/* uuid of filesystem */
746 	__be32		agi_crc;	/* crc of agi sector */
747 	__be32		agi_pad32;
748 	__be64		agi_lsn;	/* last write sequence */
749 
750 	__be32		agi_free_root; /* root of the free inode btree */
751 	__be32		agi_free_level;/* levels in free inode btree */
752 
753 	/* structure must be padded to 64 bit alignment */
754 } xfs_agi_t;
755 
756 #define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
757 
758 #define	XFS_AGI_MAGICNUM	(1 << 0)
759 #define	XFS_AGI_VERSIONNUM	(1 << 1)
760 #define	XFS_AGI_SEQNO		(1 << 2)
761 #define	XFS_AGI_LENGTH		(1 << 3)
762 #define	XFS_AGI_COUNT		(1 << 4)
763 #define	XFS_AGI_ROOT		(1 << 5)
764 #define	XFS_AGI_LEVEL		(1 << 6)
765 #define	XFS_AGI_FREECOUNT	(1 << 7)
766 #define	XFS_AGI_NEWINO		(1 << 8)
767 #define	XFS_AGI_DIRINO		(1 << 9)
768 #define	XFS_AGI_UNLINKED	(1 << 10)
769 #define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
770 #define	XFS_AGI_ALL_BITS_R1	((1 << XFS_AGI_NUM_BITS_R1) - 1)
771 #define	XFS_AGI_FREE_ROOT	(1 << 11)
772 #define	XFS_AGI_FREE_LEVEL	(1 << 12)
773 #define	XFS_AGI_NUM_BITS_R2	13
774 
775 /* disk block (xfs_daddr_t) in the AG */
776 #define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
777 #define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
778 #define	XFS_BUF_TO_AGI(bp)	((xfs_agi_t *)((bp)->b_addr))
779 
780 /*
781  * The third a.g. block contains the a.g. freelist, an array
782  * of block pointers to blocks owned by the allocation btree code.
783  */
784 #define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
785 #define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
786 #define	XFS_BUF_TO_AGFL(bp)	((xfs_agfl_t *)((bp)->b_addr))
787 
788 #define XFS_BUF_TO_AGFL_BNO(mp, bp) \
789 	(xfs_sb_version_hascrc(&((mp)->m_sb)) ? \
790 		&(XFS_BUF_TO_AGFL(bp)->agfl_bno[0]) : \
791 		(__be32 *)(bp)->b_addr)
792 
793 typedef struct xfs_agfl {
794 	__be32		agfl_magicnum;
795 	__be32		agfl_seqno;
796 	uuid_t		agfl_uuid;
797 	__be64		agfl_lsn;
798 	__be32		agfl_crc;
799 	__be32		agfl_bno[];	/* actually xfs_agfl_size(mp) */
800 } __attribute__((packed)) xfs_agfl_t;
801 
802 #define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
803 
804 #define XFS_AGB_TO_FSB(mp,agno,agbno)	\
805 	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
806 #define	XFS_FSB_TO_AGNO(mp,fsbno)	\
807 	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
808 #define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
809 	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
810 #define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
811 	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
812 		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
813 #define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
814 
815 /*
816  * For checking for bad ranges of xfs_daddr_t's, covering multiple
817  * allocation groups or a single xfs_daddr_t that's a superblock copy.
818  */
819 #define	XFS_AG_CHECK_DADDR(mp,d,len)	\
820 	((len) == 1 ? \
821 	    ASSERT((d) == XFS_SB_DADDR || \
822 		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
823 	    ASSERT(xfs_daddr_to_agno(mp, d) == \
824 		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
825 
826 typedef struct xfs_timestamp {
827 	__be32		t_sec;		/* timestamp seconds */
828 	__be32		t_nsec;		/* timestamp nanoseconds */
829 } xfs_timestamp_t;
830 
831 /*
832  * On-disk inode structure.
833  *
834  * This is just the header or "dinode core", the inode is expanded to fill a
835  * variable size the leftover area split into a data and an attribute fork.
836  * The format of the data and attribute fork depends on the format of the
837  * inode as indicated by di_format and di_aformat.  To access the data and
838  * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
839  * below.
840  *
841  * There is a very similar struct icdinode in xfs_inode which matches the
842  * layout of the first 96 bytes of this structure, but is kept in native
843  * format instead of big endian.
844  *
845  * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
846  * padding field for v3 inodes.
847  */
848 #define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
849 typedef struct xfs_dinode {
850 	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
851 	__be16		di_mode;	/* mode and type of file */
852 	__u8		di_version;	/* inode version */
853 	__u8		di_format;	/* format of di_c data */
854 	__be16		di_onlink;	/* old number of links to file */
855 	__be32		di_uid;		/* owner's user id */
856 	__be32		di_gid;		/* owner's group id */
857 	__be32		di_nlink;	/* number of links to file */
858 	__be16		di_projid_lo;	/* lower part of owner's project id */
859 	__be16		di_projid_hi;	/* higher part owner's project id */
860 	__u8		di_pad[6];	/* unused, zeroed space */
861 	__be16		di_flushiter;	/* incremented on flush */
862 	xfs_timestamp_t	di_atime;	/* time last accessed */
863 	xfs_timestamp_t	di_mtime;	/* time last modified */
864 	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
865 	__be64		di_size;	/* number of bytes in file */
866 	__be64		di_nblocks;	/* # of direct & btree blocks used */
867 	__be32		di_extsize;	/* basic/minimum extent size for file */
868 	__be32		di_nextents;	/* number of extents in data fork */
869 	__be16		di_anextents;	/* number of extents in attribute fork*/
870 	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
871 	__s8		di_aformat;	/* format of attr fork's data */
872 	__be32		di_dmevmask;	/* DMIG event mask */
873 	__be16		di_dmstate;	/* DMIG state info */
874 	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
875 	__be32		di_gen;		/* generation number */
876 
877 	/* di_next_unlinked is the only non-core field in the old dinode */
878 	__be32		di_next_unlinked;/* agi unlinked list ptr */
879 
880 	/* start of the extended dinode, writable fields */
881 	__le32		di_crc;		/* CRC of the inode */
882 	__be64		di_changecount;	/* number of attribute changes */
883 	__be64		di_lsn;		/* flush sequence */
884 	__be64		di_flags2;	/* more random flags */
885 	__be32		di_cowextsize;	/* basic cow extent size for file */
886 	__u8		di_pad2[12];	/* more padding for future expansion */
887 
888 	/* fields only written to during inode creation */
889 	xfs_timestamp_t	di_crtime;	/* time created */
890 	__be64		di_ino;		/* inode number */
891 	uuid_t		di_uuid;	/* UUID of the filesystem */
892 
893 	/* structure must be padded to 64 bit alignment */
894 } xfs_dinode_t;
895 
896 #define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
897 
898 #define DI_MAX_FLUSH 0xffff
899 
900 /*
901  * Size of the core inode on disk.  Version 1 and 2 inodes have
902  * the same size, but version 3 has grown a few additional fields.
903  */
904 static inline uint xfs_dinode_size(int version)
905 {
906 	if (version == 3)
907 		return sizeof(struct xfs_dinode);
908 	return offsetof(struct xfs_dinode, di_crc);
909 }
910 
911 /*
912  * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
913  * Since the pathconf interface is signed, we use 2^31 - 1 instead.
914  */
915 #define	XFS_MAXLINK		((1U << 31) - 1U)
916 
917 /*
918  * Values for di_format
919  *
920  * This enum is used in string mapping in xfs_trace.h; please keep the
921  * TRACE_DEFINE_ENUMs for it up to date.
922  */
923 typedef enum xfs_dinode_fmt {
924 	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
925 	XFS_DINODE_FMT_LOCAL,		/* bulk data */
926 	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
927 	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
928 	XFS_DINODE_FMT_UUID		/* added long ago, but never used */
929 } xfs_dinode_fmt_t;
930 
931 #define XFS_INODE_FORMAT_STR \
932 	{ XFS_DINODE_FMT_DEV,		"dev" }, \
933 	{ XFS_DINODE_FMT_LOCAL,		"local" }, \
934 	{ XFS_DINODE_FMT_EXTENTS,	"extent" }, \
935 	{ XFS_DINODE_FMT_BTREE,		"btree" }, \
936 	{ XFS_DINODE_FMT_UUID,		"uuid" }
937 
938 /*
939  * Inode minimum and maximum sizes.
940  */
941 #define	XFS_DINODE_MIN_LOG	8
942 #define	XFS_DINODE_MAX_LOG	11
943 #define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
944 #define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
945 
946 /*
947  * Inode size for given fs.
948  */
949 #define XFS_LITINO(mp, version) \
950 	((int)(((mp)->m_sb.sb_inodesize) - xfs_dinode_size(version)))
951 
952 /*
953  * Inode data & attribute fork sizes, per inode.
954  */
955 #define XFS_DFORK_Q(dip)		((dip)->di_forkoff != 0)
956 #define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
957 
958 #define XFS_DFORK_DSIZE(dip,mp) \
959 	(XFS_DFORK_Q(dip) ? \
960 		XFS_DFORK_BOFF(dip) : \
961 		XFS_LITINO(mp, (dip)->di_version))
962 #define XFS_DFORK_ASIZE(dip,mp) \
963 	(XFS_DFORK_Q(dip) ? \
964 		XFS_LITINO(mp, (dip)->di_version) - XFS_DFORK_BOFF(dip) : \
965 		0)
966 #define XFS_DFORK_SIZE(dip,mp,w) \
967 	((w) == XFS_DATA_FORK ? \
968 		XFS_DFORK_DSIZE(dip, mp) : \
969 		XFS_DFORK_ASIZE(dip, mp))
970 
971 #define XFS_DFORK_MAXEXT(dip, mp, w) \
972 	(XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
973 
974 /*
975  * Return pointers to the data or attribute forks.
976  */
977 #define XFS_DFORK_DPTR(dip) \
978 	((char *)dip + xfs_dinode_size(dip->di_version))
979 #define XFS_DFORK_APTR(dip)	\
980 	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
981 #define XFS_DFORK_PTR(dip,w)	\
982 	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
983 
984 #define XFS_DFORK_FORMAT(dip,w) \
985 	((w) == XFS_DATA_FORK ? \
986 		(dip)->di_format : \
987 		(dip)->di_aformat)
988 #define XFS_DFORK_NEXTENTS(dip,w) \
989 	((w) == XFS_DATA_FORK ? \
990 		be32_to_cpu((dip)->di_nextents) : \
991 		be16_to_cpu((dip)->di_anextents))
992 
993 /*
994  * For block and character special files the 32bit dev_t is stored at the
995  * beginning of the data fork.
996  */
997 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
998 {
999 	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1000 }
1001 
1002 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1003 {
1004 	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1005 }
1006 
1007 /*
1008  * Values for di_flags
1009  */
1010 #define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
1011 #define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
1012 #define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
1013 #define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
1014 #define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
1015 #define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
1016 #define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
1017 #define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
1018 #define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
1019 #define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
1020 #define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
1021 #define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
1022 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
1023 #define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
1024 #define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
1025 /* Do not use bit 15, di_flags is legacy and unchanging now */
1026 
1027 #define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
1028 #define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
1029 #define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
1030 #define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1031 #define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
1032 #define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
1033 #define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
1034 #define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
1035 #define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
1036 #define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1037 #define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1038 #define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
1039 #define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1040 #define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
1041 #define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
1042 
1043 #define XFS_DIFLAG_ANY \
1044 	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1045 	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1046 	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1047 	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1048 	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1049 
1050 /*
1051  * Values for di_flags2 These start by being exposed to userspace in the upper
1052  * 16 bits of the XFS_XFLAG_s range.
1053  */
1054 #define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
1055 #define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
1056 #define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
1057 #define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
1058 #define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
1059 #define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1060 
1061 #define XFS_DIFLAG2_ANY \
1062 	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)
1063 
1064 /*
1065  * Inode number format:
1066  * low inopblog bits - offset in block
1067  * next agblklog bits - block number in ag
1068  * next agno_log bits - ag number
1069  * high agno_log-agblklog-inopblog bits - 0
1070  */
1071 #define	XFS_INO_MASK(k)			(uint32_t)((1ULL << (k)) - 1)
1072 #define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1073 #define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1074 #define	XFS_INO_AGINO_BITS(mp)		(mp)->m_agino_log
1075 #define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1076 #define	XFS_INO_BITS(mp)		\
1077 	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1078 #define	XFS_INO_TO_AGNO(mp,i)		\
1079 	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1080 #define	XFS_INO_TO_AGINO(mp,i)		\
1081 	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1082 #define	XFS_INO_TO_AGBNO(mp,i)		\
1083 	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1084 		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1085 #define	XFS_INO_TO_OFFSET(mp,i)		\
1086 	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1087 #define	XFS_INO_TO_FSB(mp,i)		\
1088 	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1089 #define	XFS_AGINO_TO_INO(mp,a,i)	\
1090 	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1091 #define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1092 #define	XFS_AGINO_TO_OFFSET(mp,i)	\
1093 	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1094 #define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1095 	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1096 #define	XFS_FSB_TO_INO(mp, b)	((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1097 #define	XFS_AGB_TO_AGINO(mp, b)	((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1098 
1099 #define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1100 #define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1101 
1102 /*
1103  * RealTime Device format definitions
1104  */
1105 
1106 /* Min and max rt extent sizes, specified in bytes */
1107 #define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1108 #define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1109 #define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1110 
1111 #define	XFS_BLOCKSIZE(mp)	((mp)->m_sb.sb_blocksize)
1112 #define	XFS_BLOCKMASK(mp)	((mp)->m_blockmask)
1113 #define	XFS_BLOCKWSIZE(mp)	((mp)->m_blockwsize)
1114 #define	XFS_BLOCKWMASK(mp)	((mp)->m_blockwmask)
1115 
1116 /*
1117  * RT Summary and bit manipulation macros.
1118  */
1119 #define	XFS_SUMOFFS(mp,ls,bb)	((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb)))
1120 #define	XFS_SUMOFFSTOBLOCK(mp,s)	\
1121 	(((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog)
1122 #define	XFS_SUMPTR(mp,bp,so)	\
1123 	((xfs_suminfo_t *)((bp)->b_addr + \
1124 		(((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp))))
1125 
1126 #define	XFS_BITTOBLOCK(mp,bi)	((bi) >> (mp)->m_blkbit_log)
1127 #define	XFS_BLOCKTOBIT(mp,bb)	((bb) << (mp)->m_blkbit_log)
1128 #define	XFS_BITTOWORD(mp,bi)	\
1129 	((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp)))
1130 
1131 #define	XFS_RTMIN(a,b)	((a) < (b) ? (a) : (b))
1132 #define	XFS_RTMAX(a,b)	((a) > (b) ? (a) : (b))
1133 
1134 #define	XFS_RTLOBIT(w)	xfs_lowbit32(w)
1135 #define	XFS_RTHIBIT(w)	xfs_highbit32(w)
1136 
1137 #define	XFS_RTBLOCKLOG(b)	xfs_highbit64(b)
1138 
1139 /*
1140  * Dquot and dquot block format definitions
1141  */
1142 #define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1143 #define XFS_DQUOT_VERSION	(uint8_t)0x01	/* latest version number */
1144 
1145 /*
1146  * This is the main portion of the on-disk representation of quota
1147  * information for a user. This is the q_core of the xfs_dquot_t that
1148  * is kept in kernel memory. We pad this with some more expansion room
1149  * to construct the on disk structure.
1150  */
1151 typedef struct	xfs_disk_dquot {
1152 	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1153 	__u8		d_version;	/* dquot version */
1154 	__u8		d_flags;	/* XFS_DQ_USER/PROJ/GROUP */
1155 	__be32		d_id;		/* user,project,group id */
1156 	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1157 	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1158 	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1159 	__be64		d_ino_softlimit;/* preferred inode limit */
1160 	__be64		d_bcount;	/* disk blocks owned by the user */
1161 	__be64		d_icount;	/* inodes owned by the user */
1162 	__be32		d_itimer;	/* zero if within inode limits if not,
1163 					   this is when we refuse service */
1164 	__be32		d_btimer;	/* similar to above; for disk blocks */
1165 	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1166 	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1167 	__be32		d_pad0;		/* 64 bit align */
1168 	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1169 	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1170 	__be64		d_rtbcount;	/* realtime blocks owned */
1171 	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1172 	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1173 	__be16		d_pad;
1174 } xfs_disk_dquot_t;
1175 
1176 /*
1177  * This is what goes on disk. This is separated from the xfs_disk_dquot because
1178  * carrying the unnecessary padding would be a waste of memory.
1179  */
1180 typedef struct xfs_dqblk {
1181 	xfs_disk_dquot_t  dd_diskdq;	/* portion that lives incore as well */
1182 	char		  dd_fill[4];	/* filling for posterity */
1183 
1184 	/*
1185 	 * These two are only present on filesystems with the CRC bits set.
1186 	 */
1187 	__be32		  dd_crc;	/* checksum */
1188 	__be64		  dd_lsn;	/* last modification in log */
1189 	uuid_t		  dd_uuid;	/* location information */
1190 } xfs_dqblk_t;
1191 
1192 #define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1193 
1194 /*
1195  * Remote symlink format and access functions.
1196  */
1197 #define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1198 
1199 struct xfs_dsymlink_hdr {
1200 	__be32	sl_magic;
1201 	__be32	sl_offset;
1202 	__be32	sl_bytes;
1203 	__be32	sl_crc;
1204 	uuid_t	sl_uuid;
1205 	__be64	sl_owner;
1206 	__be64	sl_blkno;
1207 	__be64	sl_lsn;
1208 };
1209 
1210 #define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1211 
1212 #define XFS_SYMLINK_MAXLEN	1024
1213 /*
1214  * The maximum pathlen is 1024 bytes. Since the minimum file system
1215  * blocksize is 512 bytes, we can get a max of 3 extents back from
1216  * bmapi when crc headers are taken into account.
1217  */
1218 #define XFS_SYMLINK_MAPS 3
1219 
1220 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1221 	((bufsize) - (xfs_sb_version_hascrc(&(mp)->m_sb) ? \
1222 			sizeof(struct xfs_dsymlink_hdr) : 0))
1223 
1224 
1225 /*
1226  * Allocation Btree format definitions
1227  *
1228  * There are two on-disk btrees, one sorted by blockno and one sorted
1229  * by blockcount and blockno.  All blocks look the same to make the code
1230  * simpler; if we have time later, we'll make the optimizations.
1231  */
1232 #define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1233 #define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1234 #define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1235 #define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1236 
1237 /*
1238  * Data record/key structure
1239  */
1240 typedef struct xfs_alloc_rec {
1241 	__be32		ar_startblock;	/* starting block number */
1242 	__be32		ar_blockcount;	/* count of free blocks */
1243 } xfs_alloc_rec_t, xfs_alloc_key_t;
1244 
1245 typedef struct xfs_alloc_rec_incore {
1246 	xfs_agblock_t	ar_startblock;	/* starting block number */
1247 	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1248 } xfs_alloc_rec_incore_t;
1249 
1250 /* btree pointer type */
1251 typedef __be32 xfs_alloc_ptr_t;
1252 
1253 /*
1254  * Block numbers in the AG:
1255  * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1256  */
1257 #define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1258 #define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1259 
1260 
1261 /*
1262  * Inode Allocation Btree format definitions
1263  *
1264  * There is a btree for the inode map per allocation group.
1265  */
1266 #define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1267 #define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1268 #define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1269 #define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1270 
1271 typedef uint64_t	xfs_inofree_t;
1272 #define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1273 #define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1274 #define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1275 #define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1276 
1277 #define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1278 #define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(uint16_t))
1279 #define XFS_INODES_PER_HOLEMASK_BIT	\
1280 	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1281 
1282 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1283 {
1284 	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1285 }
1286 
1287 /*
1288  * The on-disk inode record structure has two formats. The original "full"
1289  * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1290  * and replaces the 3 high-order freecount bytes wth the holemask and inode
1291  * count.
1292  *
1293  * The holemask of the sparse record format allows an inode chunk to have holes
1294  * that refer to blocks not owned by the inode record. This facilitates inode
1295  * allocation in the event of severe free space fragmentation.
1296  */
1297 typedef struct xfs_inobt_rec {
1298 	__be32		ir_startino;	/* starting inode number */
1299 	union {
1300 		struct {
1301 			__be32	ir_freecount;	/* count of free inodes */
1302 		} f;
1303 		struct {
1304 			__be16	ir_holemask;/* hole mask for sparse chunks */
1305 			__u8	ir_count;	/* total inode count */
1306 			__u8	ir_freecount;	/* count of free inodes */
1307 		} sp;
1308 	} ir_u;
1309 	__be64		ir_free;	/* free inode mask */
1310 } xfs_inobt_rec_t;
1311 
1312 typedef struct xfs_inobt_rec_incore {
1313 	xfs_agino_t	ir_startino;	/* starting inode number */
1314 	uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1315 	uint8_t		ir_count;	/* total inode count */
1316 	uint8_t		ir_freecount;	/* count of free inodes (set bits) */
1317 	xfs_inofree_t	ir_free;	/* free inode mask */
1318 } xfs_inobt_rec_incore_t;
1319 
1320 static inline bool xfs_inobt_issparse(uint16_t holemask)
1321 {
1322 	/* non-zero holemask represents a sparse rec. */
1323 	return holemask;
1324 }
1325 
1326 /*
1327  * Key structure
1328  */
1329 typedef struct xfs_inobt_key {
1330 	__be32		ir_startino;	/* starting inode number */
1331 } xfs_inobt_key_t;
1332 
1333 /* btree pointer type */
1334 typedef __be32 xfs_inobt_ptr_t;
1335 
1336 /*
1337  * block numbers in the AG.
1338  */
1339 #define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1340 #define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1341 
1342 /*
1343  * Reverse mapping btree format definitions
1344  *
1345  * There is a btree for the reverse map per allocation group
1346  */
1347 #define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1348 
1349 /*
1350  * Ownership info for an extent.  This is used to create reverse-mapping
1351  * entries.
1352  */
1353 #define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1354 #define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1355 struct xfs_owner_info {
1356 	uint64_t		oi_owner;
1357 	xfs_fileoff_t		oi_offset;
1358 	unsigned int		oi_flags;
1359 };
1360 
1361 /*
1362  * Special owner types.
1363  *
1364  * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1365  * to tell us we have a special owner value. We use these for static metadata
1366  * allocated at mkfs/growfs time, as well as for freespace management metadata.
1367  */
1368 #define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1369 #define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1370 #define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1371 #define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1372 #define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1373 #define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1374 #define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1375 #define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1376 #define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1377 #define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1378 
1379 #define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1380 
1381 /*
1382  * Data record structure
1383  */
1384 struct xfs_rmap_rec {
1385 	__be32		rm_startblock;	/* extent start block */
1386 	__be32		rm_blockcount;	/* extent length */
1387 	__be64		rm_owner;	/* extent owner */
1388 	__be64		rm_offset;	/* offset within the owner */
1389 };
1390 
1391 /*
1392  * rmap btree record
1393  *  rm_offset:63 is the attribute fork flag
1394  *  rm_offset:62 is the bmbt block flag
1395  *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1396  *  rm_offset:54-60 aren't used and should be zero
1397  *  rm_offset:0-53 is the block offset within the inode
1398  */
1399 #define XFS_RMAP_OFF_ATTR_FORK	((uint64_t)1ULL << 63)
1400 #define XFS_RMAP_OFF_BMBT_BLOCK	((uint64_t)1ULL << 62)
1401 #define XFS_RMAP_OFF_UNWRITTEN	((uint64_t)1ULL << 61)
1402 
1403 #define XFS_RMAP_LEN_MAX	((uint32_t)~0U)
1404 #define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1405 				 XFS_RMAP_OFF_BMBT_BLOCK | \
1406 				 XFS_RMAP_OFF_UNWRITTEN)
1407 #define XFS_RMAP_OFF_MASK	((uint64_t)0x3FFFFFFFFFFFFFULL)
1408 
1409 #define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1410 
1411 #define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1412 #define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1413 #define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1414 
1415 #define RMAPBT_STARTBLOCK_BITLEN	32
1416 #define RMAPBT_BLOCKCOUNT_BITLEN	32
1417 #define RMAPBT_OWNER_BITLEN		64
1418 #define RMAPBT_ATTRFLAG_BITLEN		1
1419 #define RMAPBT_BMBTFLAG_BITLEN		1
1420 #define RMAPBT_EXNTFLAG_BITLEN		1
1421 #define RMAPBT_UNUSED_OFFSET_BITLEN	7
1422 #define RMAPBT_OFFSET_BITLEN		54
1423 
1424 #define XFS_RMAP_ATTR_FORK		(1 << 0)
1425 #define XFS_RMAP_BMBT_BLOCK		(1 << 1)
1426 #define XFS_RMAP_UNWRITTEN		(1 << 2)
1427 #define XFS_RMAP_KEY_FLAGS		(XFS_RMAP_ATTR_FORK | \
1428 					 XFS_RMAP_BMBT_BLOCK)
1429 #define XFS_RMAP_REC_FLAGS		(XFS_RMAP_UNWRITTEN)
1430 struct xfs_rmap_irec {
1431 	xfs_agblock_t	rm_startblock;	/* extent start block */
1432 	xfs_extlen_t	rm_blockcount;	/* extent length */
1433 	uint64_t	rm_owner;	/* extent owner */
1434 	uint64_t	rm_offset;	/* offset within the owner */
1435 	unsigned int	rm_flags;	/* state flags */
1436 };
1437 
1438 /*
1439  * Key structure
1440  *
1441  * We don't use the length for lookups
1442  */
1443 struct xfs_rmap_key {
1444 	__be32		rm_startblock;	/* extent start block */
1445 	__be64		rm_owner;	/* extent owner */
1446 	__be64		rm_offset;	/* offset within the owner */
1447 } __attribute__((packed));
1448 
1449 /* btree pointer type */
1450 typedef __be32 xfs_rmap_ptr_t;
1451 
1452 #define	XFS_RMAP_BLOCK(mp) \
1453 	(xfs_sb_version_hasfinobt(&((mp)->m_sb)) ? \
1454 	 XFS_FIBT_BLOCK(mp) + 1 : \
1455 	 XFS_IBT_BLOCK(mp) + 1)
1456 
1457 /*
1458  * Reference Count Btree format definitions
1459  *
1460  */
1461 #define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1462 
1463 unsigned int xfs_refc_block(struct xfs_mount *mp);
1464 
1465 /*
1466  * Data record/key structure
1467  *
1468  * Each record associates a range of physical blocks (starting at
1469  * rc_startblock and ending rc_blockcount blocks later) with a reference
1470  * count (rc_refcount).  Extents that are being used to stage a copy on
1471  * write (CoW) operation are recorded in the refcount btree with a
1472  * refcount of 1.  All other records must have a refcount > 1 and must
1473  * track an extent mapped only by file data forks.
1474  *
1475  * Extents with a single owner (attributes, metadata, non-shared file
1476  * data) are not tracked here.  Free space is also not tracked here.
1477  * This is consistent with pre-reflink XFS.
1478  */
1479 
1480 /*
1481  * Extents that are being used to stage a copy on write are stored
1482  * in the refcount btree with a refcount of 1 and the upper bit set
1483  * on the startblock.  This speeds up mount time deletion of stale
1484  * staging extents because they're all at the right side of the tree.
1485  */
1486 #define XFS_REFC_COW_START		((xfs_agblock_t)(1U << 31))
1487 #define REFCNTBT_COWFLAG_BITLEN		1
1488 #define REFCNTBT_AGBLOCK_BITLEN		31
1489 
1490 struct xfs_refcount_rec {
1491 	__be32		rc_startblock;	/* starting block number */
1492 	__be32		rc_blockcount;	/* count of blocks */
1493 	__be32		rc_refcount;	/* number of inodes linked here */
1494 };
1495 
1496 struct xfs_refcount_key {
1497 	__be32		rc_startblock;	/* starting block number */
1498 };
1499 
1500 struct xfs_refcount_irec {
1501 	xfs_agblock_t	rc_startblock;	/* starting block number */
1502 	xfs_extlen_t	rc_blockcount;	/* count of free blocks */
1503 	xfs_nlink_t	rc_refcount;	/* number of inodes linked here */
1504 };
1505 
1506 #define MAXREFCOUNT	((xfs_nlink_t)~0U)
1507 #define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
1508 
1509 /* btree pointer type */
1510 typedef __be32 xfs_refcount_ptr_t;
1511 
1512 
1513 /*
1514  * BMAP Btree format definitions
1515  *
1516  * This includes both the root block definition that sits inside an inode fork
1517  * and the record/pointer formats for the leaf/node in the blocks.
1518  */
1519 #define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1520 #define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1521 
1522 /*
1523  * Bmap root header, on-disk form only.
1524  */
1525 typedef struct xfs_bmdr_block {
1526 	__be16		bb_level;	/* 0 is a leaf */
1527 	__be16		bb_numrecs;	/* current # of data records */
1528 } xfs_bmdr_block_t;
1529 
1530 /*
1531  * Bmap btree record and extent descriptor.
1532  *  l0:63 is an extent flag (value 1 indicates non-normal).
1533  *  l0:9-62 are startoff.
1534  *  l0:0-8 and l1:21-63 are startblock.
1535  *  l1:0-20 are blockcount.
1536  */
1537 #define BMBT_EXNTFLAG_BITLEN	1
1538 #define BMBT_STARTOFF_BITLEN	54
1539 #define BMBT_STARTBLOCK_BITLEN	52
1540 #define BMBT_BLOCKCOUNT_BITLEN	21
1541 
1542 #define BMBT_STARTOFF_MASK	((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1543 
1544 typedef struct xfs_bmbt_rec {
1545 	__be64			l0, l1;
1546 } xfs_bmbt_rec_t;
1547 
1548 typedef uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1549 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1550 
1551 /*
1552  * Values and macros for delayed-allocation startblock fields.
1553  */
1554 #define STARTBLOCKVALBITS	17
1555 #define STARTBLOCKMASKBITS	(15 + 20)
1556 #define STARTBLOCKMASK		\
1557 	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1558 
1559 static inline int isnullstartblock(xfs_fsblock_t x)
1560 {
1561 	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1562 }
1563 
1564 static inline xfs_fsblock_t nullstartblock(int k)
1565 {
1566 	ASSERT(k < (1 << STARTBLOCKVALBITS));
1567 	return STARTBLOCKMASK | (k);
1568 }
1569 
1570 static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1571 {
1572 	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1573 }
1574 
1575 /*
1576  * Key structure for non-leaf levels of the tree.
1577  */
1578 typedef struct xfs_bmbt_key {
1579 	__be64		br_startoff;	/* starting file offset */
1580 } xfs_bmbt_key_t, xfs_bmdr_key_t;
1581 
1582 /* btree pointer type */
1583 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1584 
1585 
1586 /*
1587  * Generic Btree block format definitions
1588  *
1589  * This is a combination of the actual format used on disk for short and long
1590  * format btrees.  The first three fields are shared by both format, but the
1591  * pointers are different and should be used with care.
1592  *
1593  * To get the size of the actual short or long form headers please use the size
1594  * macros below.  Never use sizeof(xfs_btree_block).
1595  *
1596  * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1597  * with the crc feature bit, and all accesses to them must be conditional on
1598  * that flag.
1599  */
1600 /* short form block header */
1601 struct xfs_btree_block_shdr {
1602 	__be32		bb_leftsib;
1603 	__be32		bb_rightsib;
1604 
1605 	__be64		bb_blkno;
1606 	__be64		bb_lsn;
1607 	uuid_t		bb_uuid;
1608 	__be32		bb_owner;
1609 	__le32		bb_crc;
1610 };
1611 
1612 /* long form block header */
1613 struct xfs_btree_block_lhdr {
1614 	__be64		bb_leftsib;
1615 	__be64		bb_rightsib;
1616 
1617 	__be64		bb_blkno;
1618 	__be64		bb_lsn;
1619 	uuid_t		bb_uuid;
1620 	__be64		bb_owner;
1621 	__le32		bb_crc;
1622 	__be32		bb_pad; /* padding for alignment */
1623 };
1624 
1625 struct xfs_btree_block {
1626 	__be32		bb_magic;	/* magic number for block type */
1627 	__be16		bb_level;	/* 0 is a leaf */
1628 	__be16		bb_numrecs;	/* current # of data records */
1629 	union {
1630 		struct xfs_btree_block_shdr s;
1631 		struct xfs_btree_block_lhdr l;
1632 	} bb_u;				/* rest */
1633 };
1634 
1635 /* size of a short form block */
1636 #define XFS_BTREE_SBLOCK_LEN \
1637 	(offsetof(struct xfs_btree_block, bb_u) + \
1638 	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1639 /* size of a long form block */
1640 #define XFS_BTREE_LBLOCK_LEN \
1641 	(offsetof(struct xfs_btree_block, bb_u) + \
1642 	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1643 
1644 /* sizes of CRC enabled btree blocks */
1645 #define XFS_BTREE_SBLOCK_CRC_LEN \
1646 	(offsetof(struct xfs_btree_block, bb_u) + \
1647 	 sizeof(struct xfs_btree_block_shdr))
1648 #define XFS_BTREE_LBLOCK_CRC_LEN \
1649 	(offsetof(struct xfs_btree_block, bb_u) + \
1650 	 sizeof(struct xfs_btree_block_lhdr))
1651 
1652 #define XFS_BTREE_SBLOCK_CRC_OFF \
1653 	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1654 #define XFS_BTREE_LBLOCK_CRC_OFF \
1655 	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1656 
1657 /*
1658  * On-disk XFS access control list structure.
1659  */
1660 struct xfs_acl_entry {
1661 	__be32	ae_tag;
1662 	__be32	ae_id;
1663 	__be16	ae_perm;
1664 	__be16	ae_pad;		/* fill the implicit hole in the structure */
1665 };
1666 
1667 struct xfs_acl {
1668 	__be32			acl_cnt;
1669 	struct xfs_acl_entry	acl_entry[0];
1670 };
1671 
1672 /*
1673  * The number of ACL entries allowed is defined by the on-disk format.
1674  * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1675  * limited only by the maximum size of the xattr that stores the information.
1676  */
1677 #define XFS_ACL_MAX_ENTRIES(mp)	\
1678 	(xfs_sb_version_hascrc(&mp->m_sb) \
1679 		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1680 						sizeof(struct xfs_acl_entry) \
1681 		: 25)
1682 
1683 #define XFS_ACL_SIZE(cnt) \
1684 	(sizeof(struct xfs_acl) + \
1685 		sizeof(struct xfs_acl_entry) * cnt)
1686 
1687 #define XFS_ACL_MAX_SIZE(mp) \
1688 	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1689 
1690 
1691 /* On-disk XFS extended attribute names */
1692 #define SGI_ACL_FILE		"SGI_ACL_FILE"
1693 #define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
1694 #define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
1695 #define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
1696 
1697 #endif /* __XFS_FORMAT_H__ */
1698