• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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  */
xfs_sb_good_v4_features(struct xfs_sb * sbp)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 
xfs_sb_good_version(struct xfs_sb * sbp)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 /*
319  * Detect a mismatched features2 field.  Older kernels read/wrote
320  * this into the wrong slot, so to be safe we keep them in sync.
321  */
xfs_sb_has_mismatched_features2(struct xfs_sb * sbp)322 static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
323 {
324 	return sbp->sb_bad_features2 != sbp->sb_features2;
325 }
326 
xfs_sb_version_hasattr(struct xfs_sb * sbp)327 static inline bool xfs_sb_version_hasattr(struct xfs_sb *sbp)
328 {
329 	return (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT);
330 }
331 
xfs_sb_version_addattr(struct xfs_sb * sbp)332 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
333 {
334 	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
335 }
336 
xfs_sb_version_hasquota(struct xfs_sb * sbp)337 static inline bool xfs_sb_version_hasquota(struct xfs_sb *sbp)
338 {
339 	return (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT);
340 }
341 
xfs_sb_version_addquota(struct xfs_sb * sbp)342 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
343 {
344 	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
345 }
346 
xfs_sb_version_hasalign(struct xfs_sb * sbp)347 static inline bool xfs_sb_version_hasalign(struct xfs_sb *sbp)
348 {
349 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
350 		(sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT));
351 }
352 
xfs_sb_version_hasdalign(struct xfs_sb * sbp)353 static inline bool xfs_sb_version_hasdalign(struct xfs_sb *sbp)
354 {
355 	return (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT);
356 }
357 
xfs_sb_version_haslogv2(struct xfs_sb * sbp)358 static inline bool xfs_sb_version_haslogv2(struct xfs_sb *sbp)
359 {
360 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
361 	       (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT);
362 }
363 
xfs_sb_version_hasextflgbit(struct xfs_sb * sbp)364 static inline bool xfs_sb_version_hasextflgbit(struct xfs_sb *sbp)
365 {
366 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
367 	       (sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT);
368 }
369 
xfs_sb_version_hassector(struct xfs_sb * sbp)370 static inline bool xfs_sb_version_hassector(struct xfs_sb *sbp)
371 {
372 	return (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT);
373 }
374 
xfs_sb_version_hasasciici(struct xfs_sb * sbp)375 static inline bool xfs_sb_version_hasasciici(struct xfs_sb *sbp)
376 {
377 	return (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT);
378 }
379 
xfs_sb_version_hasmorebits(struct xfs_sb * sbp)380 static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp)
381 {
382 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
383 	       (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
384 }
385 
386 /*
387  * sb_features2 bit version macros.
388  */
xfs_sb_version_haslazysbcount(struct xfs_sb * sbp)389 static inline bool xfs_sb_version_haslazysbcount(struct xfs_sb *sbp)
390 {
391 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
392 	       (xfs_sb_version_hasmorebits(sbp) &&
393 		(sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT));
394 }
395 
xfs_sb_version_hasattr2(struct xfs_sb * sbp)396 static inline bool xfs_sb_version_hasattr2(struct xfs_sb *sbp)
397 {
398 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
399 	       (xfs_sb_version_hasmorebits(sbp) &&
400 		(sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT));
401 }
402 
xfs_sb_version_addattr2(struct xfs_sb * sbp)403 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
404 {
405 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
406 	sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
407 }
408 
xfs_sb_version_removeattr2(struct xfs_sb * sbp)409 static inline void xfs_sb_version_removeattr2(struct xfs_sb *sbp)
410 {
411 	sbp->sb_features2 &= ~XFS_SB_VERSION2_ATTR2BIT;
412 	if (!sbp->sb_features2)
413 		sbp->sb_versionnum &= ~XFS_SB_VERSION_MOREBITSBIT;
414 }
415 
xfs_sb_version_hasprojid32bit(struct xfs_sb * sbp)416 static inline bool xfs_sb_version_hasprojid32bit(struct xfs_sb *sbp)
417 {
418 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
419 	       (xfs_sb_version_hasmorebits(sbp) &&
420 		(sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT));
421 }
422 
xfs_sb_version_addprojid32bit(struct xfs_sb * sbp)423 static inline void xfs_sb_version_addprojid32bit(struct xfs_sb *sbp)
424 {
425 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
426 	sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
427 }
428 
429 /*
430  * Extended v5 superblock feature masks. These are to be used for new v5
431  * superblock features only.
432  *
433  * Compat features are new features that old kernels will not notice or affect
434  * and so can mount read-write without issues.
435  *
436  * RO-Compat (read only) are features that old kernels can read but will break
437  * if they write. Hence only read-only mounts of such filesystems are allowed on
438  * kernels that don't support the feature bit.
439  *
440  * InCompat features are features which old kernels will not understand and so
441  * must not mount.
442  *
443  * Log-InCompat features are for changes to log formats or new transactions that
444  * can't be replayed on older kernels. The fields are set when the filesystem is
445  * mounted, and a clean unmount clears the fields.
446  */
447 #define XFS_SB_FEAT_COMPAT_ALL 0
448 #define XFS_SB_FEAT_COMPAT_UNKNOWN	~XFS_SB_FEAT_COMPAT_ALL
449 static inline bool
xfs_sb_has_compat_feature(struct xfs_sb * sbp,__uint32_t feature)450 xfs_sb_has_compat_feature(
451 	struct xfs_sb	*sbp,
452 	__uint32_t	feature)
453 {
454 	return (sbp->sb_features_compat & feature) != 0;
455 }
456 
457 #define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
458 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
459 #define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
460 #define XFS_SB_FEAT_RO_COMPAT_ALL \
461 		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
462 		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
463 		 XFS_SB_FEAT_RO_COMPAT_REFLINK)
464 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
465 static inline bool
xfs_sb_has_ro_compat_feature(struct xfs_sb * sbp,__uint32_t feature)466 xfs_sb_has_ro_compat_feature(
467 	struct xfs_sb	*sbp,
468 	__uint32_t	feature)
469 {
470 	return (sbp->sb_features_ro_compat & feature) != 0;
471 }
472 
473 #define XFS_SB_FEAT_INCOMPAT_FTYPE	(1 << 0)	/* filetype in dirent */
474 #define XFS_SB_FEAT_INCOMPAT_SPINODES	(1 << 1)	/* sparse inode chunks */
475 #define XFS_SB_FEAT_INCOMPAT_META_UUID	(1 << 2)	/* metadata UUID */
476 #define XFS_SB_FEAT_INCOMPAT_ALL \
477 		(XFS_SB_FEAT_INCOMPAT_FTYPE|	\
478 		 XFS_SB_FEAT_INCOMPAT_SPINODES|	\
479 		 XFS_SB_FEAT_INCOMPAT_META_UUID)
480 
481 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
482 static inline bool
xfs_sb_has_incompat_feature(struct xfs_sb * sbp,__uint32_t feature)483 xfs_sb_has_incompat_feature(
484 	struct xfs_sb	*sbp,
485 	__uint32_t	feature)
486 {
487 	return (sbp->sb_features_incompat & feature) != 0;
488 }
489 
490 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL 0
491 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
492 static inline bool
xfs_sb_has_incompat_log_feature(struct xfs_sb * sbp,__uint32_t feature)493 xfs_sb_has_incompat_log_feature(
494 	struct xfs_sb	*sbp,
495 	__uint32_t	feature)
496 {
497 	return (sbp->sb_features_log_incompat & feature) != 0;
498 }
499 
500 /*
501  * V5 superblock specific feature checks
502  */
xfs_sb_version_hascrc(struct xfs_sb * sbp)503 static inline int xfs_sb_version_hascrc(struct xfs_sb *sbp)
504 {
505 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
506 }
507 
xfs_sb_version_has_pquotino(struct xfs_sb * sbp)508 static inline int xfs_sb_version_has_pquotino(struct xfs_sb *sbp)
509 {
510 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
511 }
512 
xfs_sb_version_hasftype(struct xfs_sb * sbp)513 static inline int xfs_sb_version_hasftype(struct xfs_sb *sbp)
514 {
515 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
516 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_FTYPE)) ||
517 	       (xfs_sb_version_hasmorebits(sbp) &&
518 		 (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE));
519 }
520 
xfs_sb_version_hasfinobt(xfs_sb_t * sbp)521 static inline int xfs_sb_version_hasfinobt(xfs_sb_t *sbp)
522 {
523 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
524 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT);
525 }
526 
xfs_sb_version_hassparseinodes(struct xfs_sb * sbp)527 static inline bool xfs_sb_version_hassparseinodes(struct xfs_sb *sbp)
528 {
529 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
530 		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_SPINODES);
531 }
532 
533 /*
534  * XFS_SB_FEAT_INCOMPAT_META_UUID indicates that the metadata UUID
535  * is stored separately from the user-visible UUID; this allows the
536  * user-visible UUID to be changed on V5 filesystems which have a
537  * filesystem UUID stamped into every piece of metadata.
538  */
xfs_sb_version_hasmetauuid(struct xfs_sb * sbp)539 static inline bool xfs_sb_version_hasmetauuid(struct xfs_sb *sbp)
540 {
541 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
542 		(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID);
543 }
544 
xfs_sb_version_hasrmapbt(struct xfs_sb * sbp)545 static inline bool xfs_sb_version_hasrmapbt(struct xfs_sb *sbp)
546 {
547 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
548 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT);
549 }
550 
xfs_sb_version_hasreflink(struct xfs_sb * sbp)551 static inline bool xfs_sb_version_hasreflink(struct xfs_sb *sbp)
552 {
553 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
554 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK);
555 }
556 
557 /*
558  * end of superblock version macros
559  */
560 
561 static inline bool
xfs_is_quota_inode(struct xfs_sb * sbp,xfs_ino_t ino)562 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
563 {
564 	return (ino == sbp->sb_uquotino ||
565 		ino == sbp->sb_gquotino ||
566 		ino == sbp->sb_pquotino);
567 }
568 
569 #define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
570 #define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
571 #define XFS_BUF_TO_SBP(bp)	((xfs_dsb_t *)((bp)->b_addr))
572 
573 #define	XFS_HDR_BLOCK(mp,d)	((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
574 #define	XFS_DADDR_TO_FSB(mp,d)	XFS_AGB_TO_FSB(mp, \
575 			xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
576 #define	XFS_FSB_TO_DADDR(mp,fsbno)	XFS_AGB_TO_DADDR(mp, \
577 			XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
578 
579 /*
580  * File system sector to basic block conversions.
581  */
582 #define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
583 
584 /*
585  * File system block to basic block conversions.
586  */
587 #define	XFS_FSB_TO_BB(mp,fsbno)	((fsbno) << (mp)->m_blkbb_log)
588 #define	XFS_BB_TO_FSB(mp,bb)	\
589 	(((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
590 #define	XFS_BB_TO_FSBT(mp,bb)	((bb) >> (mp)->m_blkbb_log)
591 
592 /*
593  * File system block to byte conversions.
594  */
595 #define XFS_FSB_TO_B(mp,fsbno)	((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
596 #define XFS_B_TO_FSB(mp,b)	\
597 	((((__uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
598 #define XFS_B_TO_FSBT(mp,b)	(((__uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
599 #define XFS_B_FSB_OFFSET(mp,b)	((b) & (mp)->m_blockmask)
600 
601 /*
602  * Allocation group header
603  *
604  * This is divided into three structures, placed in sequential 512-byte
605  * buffers after a copy of the superblock (also in a 512-byte buffer).
606  */
607 #define	XFS_AGF_MAGIC	0x58414746	/* 'XAGF' */
608 #define	XFS_AGI_MAGIC	0x58414749	/* 'XAGI' */
609 #define	XFS_AGFL_MAGIC	0x5841464c	/* 'XAFL' */
610 #define	XFS_AGF_VERSION	1
611 #define	XFS_AGI_VERSION	1
612 
613 #define	XFS_AGF_GOOD_VERSION(v)	((v) == XFS_AGF_VERSION)
614 #define	XFS_AGI_GOOD_VERSION(v)	((v) == XFS_AGI_VERSION)
615 
616 /*
617  * Btree number 0 is bno, 1 is cnt, 2 is rmap. This value gives the size of the
618  * arrays below.
619  */
620 #define	XFS_BTNUM_AGF	((int)XFS_BTNUM_RMAPi + 1)
621 
622 /*
623  * The second word of agf_levels in the first a.g. overlaps the EFS
624  * superblock's magic number.  Since the magic numbers valid for EFS
625  * are > 64k, our value cannot be confused for an EFS superblock's.
626  */
627 
628 typedef struct xfs_agf {
629 	/*
630 	 * Common allocation group header information
631 	 */
632 	__be32		agf_magicnum;	/* magic number == XFS_AGF_MAGIC */
633 	__be32		agf_versionnum;	/* header version == XFS_AGF_VERSION */
634 	__be32		agf_seqno;	/* sequence # starting from 0 */
635 	__be32		agf_length;	/* size in blocks of a.g. */
636 	/*
637 	 * Freespace and rmap information
638 	 */
639 	__be32		agf_roots[XFS_BTNUM_AGF];	/* root blocks */
640 	__be32		agf_levels[XFS_BTNUM_AGF];	/* btree levels */
641 
642 	__be32		agf_flfirst;	/* first freelist block's index */
643 	__be32		agf_fllast;	/* last freelist block's index */
644 	__be32		agf_flcount;	/* count of blocks in freelist */
645 	__be32		agf_freeblks;	/* total free blocks */
646 
647 	__be32		agf_longest;	/* longest free space */
648 	__be32		agf_btreeblks;	/* # of blocks held in AGF btrees */
649 	uuid_t		agf_uuid;	/* uuid of filesystem */
650 
651 	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
652 	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
653 
654 	__be32		agf_refcount_root;	/* refcount tree root block */
655 	__be32		agf_refcount_level;	/* refcount btree levels */
656 
657 	/*
658 	 * reserve some contiguous space for future logged fields before we add
659 	 * the unlogged fields. This makes the range logging via flags and
660 	 * structure offsets much simpler.
661 	 */
662 	__be64		agf_spare64[14];
663 
664 	/* unlogged fields, written during buffer writeback. */
665 	__be64		agf_lsn;	/* last write sequence */
666 	__be32		agf_crc;	/* crc of agf sector */
667 	__be32		agf_spare2;
668 
669 	/* structure must be padded to 64 bit alignment */
670 } xfs_agf_t;
671 
672 #define XFS_AGF_CRC_OFF		offsetof(struct xfs_agf, agf_crc)
673 
674 #define	XFS_AGF_MAGICNUM	0x00000001
675 #define	XFS_AGF_VERSIONNUM	0x00000002
676 #define	XFS_AGF_SEQNO		0x00000004
677 #define	XFS_AGF_LENGTH		0x00000008
678 #define	XFS_AGF_ROOTS		0x00000010
679 #define	XFS_AGF_LEVELS		0x00000020
680 #define	XFS_AGF_FLFIRST		0x00000040
681 #define	XFS_AGF_FLLAST		0x00000080
682 #define	XFS_AGF_FLCOUNT		0x00000100
683 #define	XFS_AGF_FREEBLKS	0x00000200
684 #define	XFS_AGF_LONGEST		0x00000400
685 #define	XFS_AGF_BTREEBLKS	0x00000800
686 #define	XFS_AGF_UUID		0x00001000
687 #define	XFS_AGF_RMAP_BLOCKS	0x00002000
688 #define	XFS_AGF_REFCOUNT_BLOCKS	0x00004000
689 #define	XFS_AGF_REFCOUNT_ROOT	0x00008000
690 #define	XFS_AGF_REFCOUNT_LEVEL	0x00010000
691 #define	XFS_AGF_SPARE64		0x00020000
692 #define	XFS_AGF_NUM_BITS	18
693 #define	XFS_AGF_ALL_BITS	((1 << XFS_AGF_NUM_BITS) - 1)
694 
695 #define XFS_AGF_FLAGS \
696 	{ XFS_AGF_MAGICNUM,	"MAGICNUM" }, \
697 	{ XFS_AGF_VERSIONNUM,	"VERSIONNUM" }, \
698 	{ XFS_AGF_SEQNO,	"SEQNO" }, \
699 	{ XFS_AGF_LENGTH,	"LENGTH" }, \
700 	{ XFS_AGF_ROOTS,	"ROOTS" }, \
701 	{ XFS_AGF_LEVELS,	"LEVELS" }, \
702 	{ XFS_AGF_FLFIRST,	"FLFIRST" }, \
703 	{ XFS_AGF_FLLAST,	"FLLAST" }, \
704 	{ XFS_AGF_FLCOUNT,	"FLCOUNT" }, \
705 	{ XFS_AGF_FREEBLKS,	"FREEBLKS" }, \
706 	{ XFS_AGF_LONGEST,	"LONGEST" }, \
707 	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
708 	{ XFS_AGF_UUID,		"UUID" }, \
709 	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
710 	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
711 	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
712 	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
713 	{ XFS_AGF_SPARE64,	"SPARE64" }
714 
715 /* disk block (xfs_daddr_t) in the AG */
716 #define XFS_AGF_DADDR(mp)	((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
717 #define	XFS_AGF_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
718 #define	XFS_BUF_TO_AGF(bp)	((xfs_agf_t *)((bp)->b_addr))
719 
720 /*
721  * Size of the unlinked inode hash table in the agi.
722  */
723 #define	XFS_AGI_UNLINKED_BUCKETS	64
724 
725 typedef struct xfs_agi {
726 	/*
727 	 * Common allocation group header information
728 	 */
729 	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
730 	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
731 	__be32		agi_seqno;	/* sequence # starting from 0 */
732 	__be32		agi_length;	/* size in blocks of a.g. */
733 	/*
734 	 * Inode information
735 	 * Inodes are mapped by interpreting the inode number, so no
736 	 * mapping data is needed here.
737 	 */
738 	__be32		agi_count;	/* count of allocated inodes */
739 	__be32		agi_root;	/* root of inode btree */
740 	__be32		agi_level;	/* levels in inode btree */
741 	__be32		agi_freecount;	/* number of free inodes */
742 
743 	__be32		agi_newino;	/* new inode just allocated */
744 	__be32		agi_dirino;	/* last directory inode chunk */
745 	/*
746 	 * Hash table of inodes which have been unlinked but are
747 	 * still being referenced.
748 	 */
749 	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
750 	/*
751 	 * This marks the end of logging region 1 and start of logging region 2.
752 	 */
753 	uuid_t		agi_uuid;	/* uuid of filesystem */
754 	__be32		agi_crc;	/* crc of agi sector */
755 	__be32		agi_pad32;
756 	__be64		agi_lsn;	/* last write sequence */
757 
758 	__be32		agi_free_root; /* root of the free inode btree */
759 	__be32		agi_free_level;/* levels in free inode btree */
760 
761 	/* structure must be padded to 64 bit alignment */
762 } xfs_agi_t;
763 
764 #define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
765 
766 #define	XFS_AGI_MAGICNUM	(1 << 0)
767 #define	XFS_AGI_VERSIONNUM	(1 << 1)
768 #define	XFS_AGI_SEQNO		(1 << 2)
769 #define	XFS_AGI_LENGTH		(1 << 3)
770 #define	XFS_AGI_COUNT		(1 << 4)
771 #define	XFS_AGI_ROOT		(1 << 5)
772 #define	XFS_AGI_LEVEL		(1 << 6)
773 #define	XFS_AGI_FREECOUNT	(1 << 7)
774 #define	XFS_AGI_NEWINO		(1 << 8)
775 #define	XFS_AGI_DIRINO		(1 << 9)
776 #define	XFS_AGI_UNLINKED	(1 << 10)
777 #define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
778 #define	XFS_AGI_ALL_BITS_R1	((1 << XFS_AGI_NUM_BITS_R1) - 1)
779 #define	XFS_AGI_FREE_ROOT	(1 << 11)
780 #define	XFS_AGI_FREE_LEVEL	(1 << 12)
781 #define	XFS_AGI_NUM_BITS_R2	13
782 
783 /* disk block (xfs_daddr_t) in the AG */
784 #define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
785 #define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
786 #define	XFS_BUF_TO_AGI(bp)	((xfs_agi_t *)((bp)->b_addr))
787 
788 /*
789  * The third a.g. block contains the a.g. freelist, an array
790  * of block pointers to blocks owned by the allocation btree code.
791  */
792 #define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
793 #define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
794 #define	XFS_BUF_TO_AGFL(bp)	((xfs_agfl_t *)((bp)->b_addr))
795 
796 #define XFS_BUF_TO_AGFL_BNO(mp, bp) \
797 	(xfs_sb_version_hascrc(&((mp)->m_sb)) ? \
798 		&(XFS_BUF_TO_AGFL(bp)->agfl_bno[0]) : \
799 		(__be32 *)(bp)->b_addr)
800 
801 /*
802  * Size of the AGFL.  For CRC-enabled filesystes we steal a couple of
803  * slots in the beginning of the block for a proper header with the
804  * location information and CRC.
805  */
806 #define XFS_AGFL_SIZE(mp) \
807 	(((mp)->m_sb.sb_sectsize - \
808 	 (xfs_sb_version_hascrc(&((mp)->m_sb)) ? \
809 		sizeof(struct xfs_agfl) : 0)) / \
810 	  sizeof(xfs_agblock_t))
811 
812 typedef struct xfs_agfl {
813 	__be32		agfl_magicnum;
814 	__be32		agfl_seqno;
815 	uuid_t		agfl_uuid;
816 	__be64		agfl_lsn;
817 	__be32		agfl_crc;
818 	__be32		agfl_bno[];	/* actually XFS_AGFL_SIZE(mp) */
819 } __attribute__((packed)) xfs_agfl_t;
820 
821 #define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
822 
823 #define XFS_AGB_TO_FSB(mp,agno,agbno)	\
824 	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
825 #define	XFS_FSB_TO_AGNO(mp,fsbno)	\
826 	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
827 #define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
828 	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
829 #define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
830 	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
831 		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
832 #define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
833 
834 /*
835  * For checking for bad ranges of xfs_daddr_t's, covering multiple
836  * allocation groups or a single xfs_daddr_t that's a superblock copy.
837  */
838 #define	XFS_AG_CHECK_DADDR(mp,d,len)	\
839 	((len) == 1 ? \
840 	    ASSERT((d) == XFS_SB_DADDR || \
841 		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
842 	    ASSERT(xfs_daddr_to_agno(mp, d) == \
843 		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
844 
845 typedef struct xfs_timestamp {
846 	__be32		t_sec;		/* timestamp seconds */
847 	__be32		t_nsec;		/* timestamp nanoseconds */
848 } xfs_timestamp_t;
849 
850 /*
851  * On-disk inode structure.
852  *
853  * This is just the header or "dinode core", the inode is expanded to fill a
854  * variable size the leftover area split into a data and an attribute fork.
855  * The format of the data and attribute fork depends on the format of the
856  * inode as indicated by di_format and di_aformat.  To access the data and
857  * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
858  * below.
859  *
860  * There is a very similar struct icdinode in xfs_inode which matches the
861  * layout of the first 96 bytes of this structure, but is kept in native
862  * format instead of big endian.
863  *
864  * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
865  * padding field for v3 inodes.
866  */
867 #define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
868 typedef struct xfs_dinode {
869 	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
870 	__be16		di_mode;	/* mode and type of file */
871 	__u8		di_version;	/* inode version */
872 	__u8		di_format;	/* format of di_c data */
873 	__be16		di_onlink;	/* old number of links to file */
874 	__be32		di_uid;		/* owner's user id */
875 	__be32		di_gid;		/* owner's group id */
876 	__be32		di_nlink;	/* number of links to file */
877 	__be16		di_projid_lo;	/* lower part of owner's project id */
878 	__be16		di_projid_hi;	/* higher part owner's project id */
879 	__u8		di_pad[6];	/* unused, zeroed space */
880 	__be16		di_flushiter;	/* incremented on flush */
881 	xfs_timestamp_t	di_atime;	/* time last accessed */
882 	xfs_timestamp_t	di_mtime;	/* time last modified */
883 	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
884 	__be64		di_size;	/* number of bytes in file */
885 	__be64		di_nblocks;	/* # of direct & btree blocks used */
886 	__be32		di_extsize;	/* basic/minimum extent size for file */
887 	__be32		di_nextents;	/* number of extents in data fork */
888 	__be16		di_anextents;	/* number of extents in attribute fork*/
889 	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
890 	__s8		di_aformat;	/* format of attr fork's data */
891 	__be32		di_dmevmask;	/* DMIG event mask */
892 	__be16		di_dmstate;	/* DMIG state info */
893 	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
894 	__be32		di_gen;		/* generation number */
895 
896 	/* di_next_unlinked is the only non-core field in the old dinode */
897 	__be32		di_next_unlinked;/* agi unlinked list ptr */
898 
899 	/* start of the extended dinode, writable fields */
900 	__le32		di_crc;		/* CRC of the inode */
901 	__be64		di_changecount;	/* number of attribute changes */
902 	__be64		di_lsn;		/* flush sequence */
903 	__be64		di_flags2;	/* more random flags */
904 	__be32		di_cowextsize;	/* basic cow extent size for file */
905 	__u8		di_pad2[12];	/* more padding for future expansion */
906 
907 	/* fields only written to during inode creation */
908 	xfs_timestamp_t	di_crtime;	/* time created */
909 	__be64		di_ino;		/* inode number */
910 	uuid_t		di_uuid;	/* UUID of the filesystem */
911 
912 	/* structure must be padded to 64 bit alignment */
913 } xfs_dinode_t;
914 
915 #define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
916 
917 #define DI_MAX_FLUSH 0xffff
918 
919 /*
920  * Size of the core inode on disk.  Version 1 and 2 inodes have
921  * the same size, but version 3 has grown a few additional fields.
922  */
xfs_dinode_size(int version)923 static inline uint xfs_dinode_size(int version)
924 {
925 	if (version == 3)
926 		return sizeof(struct xfs_dinode);
927 	return offsetof(struct xfs_dinode, di_crc);
928 }
929 
930 /*
931  * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
932  * Since the pathconf interface is signed, we use 2^31 - 1 instead.
933  * The old inode format had a 16 bit link count, so its maximum is USHRT_MAX.
934  */
935 #define	XFS_MAXLINK		((1U << 31) - 1U)
936 #define	XFS_MAXLINK_1		65535U
937 
938 /*
939  * Values for di_format
940  */
941 typedef enum xfs_dinode_fmt {
942 	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
943 	XFS_DINODE_FMT_LOCAL,		/* bulk data */
944 	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
945 	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
946 	XFS_DINODE_FMT_UUID		/* uuid_t */
947 } xfs_dinode_fmt_t;
948 
949 /*
950  * Inode minimum and maximum sizes.
951  */
952 #define	XFS_DINODE_MIN_LOG	8
953 #define	XFS_DINODE_MAX_LOG	11
954 #define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
955 #define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
956 
957 /*
958  * Inode size for given fs.
959  */
960 #define XFS_LITINO(mp, version) \
961 	((int)(((mp)->m_sb.sb_inodesize) - xfs_dinode_size(version)))
962 
963 /*
964  * Inode data & attribute fork sizes, per inode.
965  */
966 #define XFS_DFORK_Q(dip)		((dip)->di_forkoff != 0)
967 #define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
968 
969 #define XFS_DFORK_DSIZE(dip,mp) \
970 	(XFS_DFORK_Q(dip) ? \
971 		XFS_DFORK_BOFF(dip) : \
972 		XFS_LITINO(mp, (dip)->di_version))
973 #define XFS_DFORK_ASIZE(dip,mp) \
974 	(XFS_DFORK_Q(dip) ? \
975 		XFS_LITINO(mp, (dip)->di_version) - XFS_DFORK_BOFF(dip) : \
976 		0)
977 #define XFS_DFORK_SIZE(dip,mp,w) \
978 	((w) == XFS_DATA_FORK ? \
979 		XFS_DFORK_DSIZE(dip, mp) : \
980 		XFS_DFORK_ASIZE(dip, mp))
981 
982 /*
983  * Return pointers to the data or attribute forks.
984  */
985 #define XFS_DFORK_DPTR(dip) \
986 	((char *)dip + xfs_dinode_size(dip->di_version))
987 #define XFS_DFORK_APTR(dip)	\
988 	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
989 #define XFS_DFORK_PTR(dip,w)	\
990 	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
991 
992 #define XFS_DFORK_FORMAT(dip,w) \
993 	((w) == XFS_DATA_FORK ? \
994 		(dip)->di_format : \
995 		(dip)->di_aformat)
996 #define XFS_DFORK_NEXTENTS(dip,w) \
997 	((w) == XFS_DATA_FORK ? \
998 		be32_to_cpu((dip)->di_nextents) : \
999 		be16_to_cpu((dip)->di_anextents))
1000 
1001 /*
1002  * For block and character special files the 32bit dev_t is stored at the
1003  * beginning of the data fork.
1004  */
xfs_dinode_get_rdev(struct xfs_dinode * dip)1005 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1006 {
1007 	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1008 }
1009 
xfs_dinode_put_rdev(struct xfs_dinode * dip,xfs_dev_t rdev)1010 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1011 {
1012 	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1013 }
1014 
1015 /*
1016  * Values for di_flags
1017  */
1018 #define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
1019 #define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
1020 #define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
1021 #define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
1022 #define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
1023 #define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
1024 #define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
1025 #define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
1026 #define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
1027 #define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
1028 #define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
1029 #define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
1030 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
1031 #define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
1032 #define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
1033 #define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
1034 #define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
1035 #define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
1036 #define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1037 #define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
1038 #define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
1039 #define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
1040 #define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
1041 #define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
1042 #define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1043 #define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1044 #define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
1045 #define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1046 #define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
1047 #define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
1048 
1049 #define XFS_DIFLAG_ANY \
1050 	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1051 	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1052 	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1053 	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1054 	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1055 
1056 /*
1057  * Values for di_flags2 These start by being exposed to userspace in the upper
1058  * 16 bits of the XFS_XFLAG_s range.
1059  */
1060 #define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
1061 #define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
1062 #define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
1063 #define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
1064 #define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
1065 #define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1066 
1067 #define XFS_DIFLAG2_ANY \
1068 	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)
1069 
1070 /*
1071  * Inode number format:
1072  * low inopblog bits - offset in block
1073  * next agblklog bits - block number in ag
1074  * next agno_log bits - ag number
1075  * high agno_log-agblklog-inopblog bits - 0
1076  */
1077 #define	XFS_INO_MASK(k)			(__uint32_t)((1ULL << (k)) - 1)
1078 #define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1079 #define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1080 #define	XFS_INO_AGINO_BITS(mp)		(mp)->m_agino_log
1081 #define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1082 #define	XFS_INO_BITS(mp)		\
1083 	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1084 #define	XFS_INO_TO_AGNO(mp,i)		\
1085 	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1086 #define	XFS_INO_TO_AGINO(mp,i)		\
1087 	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1088 #define	XFS_INO_TO_AGBNO(mp,i)		\
1089 	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1090 		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1091 #define	XFS_INO_TO_OFFSET(mp,i)		\
1092 	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1093 #define	XFS_INO_TO_FSB(mp,i)		\
1094 	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1095 #define	XFS_AGINO_TO_INO(mp,a,i)	\
1096 	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1097 #define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1098 #define	XFS_AGINO_TO_OFFSET(mp,i)	\
1099 	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1100 #define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1101 	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1102 
1103 #define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1104 #define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1105 
1106 /*
1107  * RealTime Device format definitions
1108  */
1109 
1110 /* Min and max rt extent sizes, specified in bytes */
1111 #define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1112 #define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1113 #define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1114 
1115 #define	XFS_BLOCKSIZE(mp)	((mp)->m_sb.sb_blocksize)
1116 #define	XFS_BLOCKMASK(mp)	((mp)->m_blockmask)
1117 #define	XFS_BLOCKWSIZE(mp)	((mp)->m_blockwsize)
1118 #define	XFS_BLOCKWMASK(mp)	((mp)->m_blockwmask)
1119 
1120 /*
1121  * RT Summary and bit manipulation macros.
1122  */
1123 #define	XFS_SUMOFFS(mp,ls,bb)	((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb)))
1124 #define	XFS_SUMOFFSTOBLOCK(mp,s)	\
1125 	(((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog)
1126 #define	XFS_SUMPTR(mp,bp,so)	\
1127 	((xfs_suminfo_t *)((bp)->b_addr + \
1128 		(((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp))))
1129 
1130 #define	XFS_BITTOBLOCK(mp,bi)	((bi) >> (mp)->m_blkbit_log)
1131 #define	XFS_BLOCKTOBIT(mp,bb)	((bb) << (mp)->m_blkbit_log)
1132 #define	XFS_BITTOWORD(mp,bi)	\
1133 	((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp)))
1134 
1135 #define	XFS_RTMIN(a,b)	((a) < (b) ? (a) : (b))
1136 #define	XFS_RTMAX(a,b)	((a) > (b) ? (a) : (b))
1137 
1138 #define	XFS_RTLOBIT(w)	xfs_lowbit32(w)
1139 #define	XFS_RTHIBIT(w)	xfs_highbit32(w)
1140 
1141 #define	XFS_RTBLOCKLOG(b)	xfs_highbit64(b)
1142 
1143 /*
1144  * Dquot and dquot block format definitions
1145  */
1146 #define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1147 #define XFS_DQUOT_VERSION	(u_int8_t)0x01	/* latest version number */
1148 
1149 /*
1150  * This is the main portion of the on-disk representation of quota
1151  * information for a user. This is the q_core of the xfs_dquot_t that
1152  * is kept in kernel memory. We pad this with some more expansion room
1153  * to construct the on disk structure.
1154  */
1155 typedef struct	xfs_disk_dquot {
1156 	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1157 	__u8		d_version;	/* dquot version */
1158 	__u8		d_flags;	/* XFS_DQ_USER/PROJ/GROUP */
1159 	__be32		d_id;		/* user,project,group id */
1160 	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1161 	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1162 	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1163 	__be64		d_ino_softlimit;/* preferred inode limit */
1164 	__be64		d_bcount;	/* disk blocks owned by the user */
1165 	__be64		d_icount;	/* inodes owned by the user */
1166 	__be32		d_itimer;	/* zero if within inode limits if not,
1167 					   this is when we refuse service */
1168 	__be32		d_btimer;	/* similar to above; for disk blocks */
1169 	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1170 	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1171 	__be32		d_pad0;		/* 64 bit align */
1172 	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1173 	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1174 	__be64		d_rtbcount;	/* realtime blocks owned */
1175 	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1176 	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1177 	__be16		d_pad;
1178 } xfs_disk_dquot_t;
1179 
1180 /*
1181  * This is what goes on disk. This is separated from the xfs_disk_dquot because
1182  * carrying the unnecessary padding would be a waste of memory.
1183  */
1184 typedef struct xfs_dqblk {
1185 	xfs_disk_dquot_t  dd_diskdq;	/* portion that lives incore as well */
1186 	char		  dd_fill[4];	/* filling for posterity */
1187 
1188 	/*
1189 	 * These two are only present on filesystems with the CRC bits set.
1190 	 */
1191 	__be32		  dd_crc;	/* checksum */
1192 	__be64		  dd_lsn;	/* last modification in log */
1193 	uuid_t		  dd_uuid;	/* location information */
1194 } xfs_dqblk_t;
1195 
1196 #define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1197 
1198 /*
1199  * Remote symlink format and access functions.
1200  */
1201 #define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1202 
1203 struct xfs_dsymlink_hdr {
1204 	__be32	sl_magic;
1205 	__be32	sl_offset;
1206 	__be32	sl_bytes;
1207 	__be32	sl_crc;
1208 	uuid_t	sl_uuid;
1209 	__be64	sl_owner;
1210 	__be64	sl_blkno;
1211 	__be64	sl_lsn;
1212 };
1213 
1214 #define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1215 
1216 /*
1217  * The maximum pathlen is 1024 bytes. Since the minimum file system
1218  * blocksize is 512 bytes, we can get a max of 3 extents back from
1219  * bmapi when crc headers are taken into account.
1220  */
1221 #define XFS_SYMLINK_MAPS 3
1222 
1223 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1224 	((bufsize) - (xfs_sb_version_hascrc(&(mp)->m_sb) ? \
1225 			sizeof(struct xfs_dsymlink_hdr) : 0))
1226 
1227 
1228 /*
1229  * Allocation Btree format definitions
1230  *
1231  * There are two on-disk btrees, one sorted by blockno and one sorted
1232  * by blockcount and blockno.  All blocks look the same to make the code
1233  * simpler; if we have time later, we'll make the optimizations.
1234  */
1235 #define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1236 #define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1237 #define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1238 #define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1239 
1240 /*
1241  * Data record/key structure
1242  */
1243 typedef struct xfs_alloc_rec {
1244 	__be32		ar_startblock;	/* starting block number */
1245 	__be32		ar_blockcount;	/* count of free blocks */
1246 } xfs_alloc_rec_t, xfs_alloc_key_t;
1247 
1248 typedef struct xfs_alloc_rec_incore {
1249 	xfs_agblock_t	ar_startblock;	/* starting block number */
1250 	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1251 } xfs_alloc_rec_incore_t;
1252 
1253 /* btree pointer type */
1254 typedef __be32 xfs_alloc_ptr_t;
1255 
1256 /*
1257  * Block numbers in the AG:
1258  * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1259  */
1260 #define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1261 #define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1262 
1263 
1264 /*
1265  * Inode Allocation Btree format definitions
1266  *
1267  * There is a btree for the inode map per allocation group.
1268  */
1269 #define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1270 #define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1271 #define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1272 #define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1273 
1274 typedef	__uint64_t	xfs_inofree_t;
1275 #define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1276 #define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1277 #define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1278 #define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1279 
1280 #define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1281 #define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(__uint16_t))
1282 #define XFS_INODES_PER_HOLEMASK_BIT	\
1283 	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(__uint16_t)))
1284 
xfs_inobt_maskn(int i,int n)1285 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1286 {
1287 	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1288 }
1289 
1290 /*
1291  * The on-disk inode record structure has two formats. The original "full"
1292  * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1293  * and replaces the 3 high-order freecount bytes wth the holemask and inode
1294  * count.
1295  *
1296  * The holemask of the sparse record format allows an inode chunk to have holes
1297  * that refer to blocks not owned by the inode record. This facilitates inode
1298  * allocation in the event of severe free space fragmentation.
1299  */
1300 typedef struct xfs_inobt_rec {
1301 	__be32		ir_startino;	/* starting inode number */
1302 	union {
1303 		struct {
1304 			__be32	ir_freecount;	/* count of free inodes */
1305 		} f;
1306 		struct {
1307 			__be16	ir_holemask;/* hole mask for sparse chunks */
1308 			__u8	ir_count;	/* total inode count */
1309 			__u8	ir_freecount;	/* count of free inodes */
1310 		} sp;
1311 	} ir_u;
1312 	__be64		ir_free;	/* free inode mask */
1313 } xfs_inobt_rec_t;
1314 
1315 typedef struct xfs_inobt_rec_incore {
1316 	xfs_agino_t	ir_startino;	/* starting inode number */
1317 	__uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1318 	__uint8_t	ir_count;	/* total inode count */
1319 	__uint8_t	ir_freecount;	/* count of free inodes (set bits) */
1320 	xfs_inofree_t	ir_free;	/* free inode mask */
1321 } xfs_inobt_rec_incore_t;
1322 
xfs_inobt_issparse(uint16_t holemask)1323 static inline bool xfs_inobt_issparse(uint16_t holemask)
1324 {
1325 	/* non-zero holemask represents a sparse rec. */
1326 	return holemask;
1327 }
1328 
1329 /*
1330  * Key structure
1331  */
1332 typedef struct xfs_inobt_key {
1333 	__be32		ir_startino;	/* starting inode number */
1334 } xfs_inobt_key_t;
1335 
1336 /* btree pointer type */
1337 typedef __be32 xfs_inobt_ptr_t;
1338 
1339 /*
1340  * block numbers in the AG.
1341  */
1342 #define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1343 #define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1344 
1345 /*
1346  * Reverse mapping btree format definitions
1347  *
1348  * There is a btree for the reverse map per allocation group
1349  */
1350 #define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1351 
1352 /*
1353  * Ownership info for an extent.  This is used to create reverse-mapping
1354  * entries.
1355  */
1356 #define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1357 #define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1358 struct xfs_owner_info {
1359 	uint64_t		oi_owner;
1360 	xfs_fileoff_t		oi_offset;
1361 	unsigned int		oi_flags;
1362 };
1363 
1364 /*
1365  * Special owner types.
1366  *
1367  * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1368  * to tell us we have a special owner value. We use these for static metadata
1369  * allocated at mkfs/growfs time, as well as for freespace management metadata.
1370  */
1371 #define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1372 #define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1373 #define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1374 #define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1375 #define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1376 #define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1377 #define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1378 #define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1379 #define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1380 #define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1381 
1382 #define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1383 
1384 /*
1385  * Data record structure
1386  */
1387 struct xfs_rmap_rec {
1388 	__be32		rm_startblock;	/* extent start block */
1389 	__be32		rm_blockcount;	/* extent length */
1390 	__be64		rm_owner;	/* extent owner */
1391 	__be64		rm_offset;	/* offset within the owner */
1392 };
1393 
1394 /*
1395  * rmap btree record
1396  *  rm_offset:63 is the attribute fork flag
1397  *  rm_offset:62 is the bmbt block flag
1398  *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1399  *  rm_offset:54-60 aren't used and should be zero
1400  *  rm_offset:0-53 is the block offset within the inode
1401  */
1402 #define XFS_RMAP_OFF_ATTR_FORK	((__uint64_t)1ULL << 63)
1403 #define XFS_RMAP_OFF_BMBT_BLOCK	((__uint64_t)1ULL << 62)
1404 #define XFS_RMAP_OFF_UNWRITTEN	((__uint64_t)1ULL << 61)
1405 
1406 #define XFS_RMAP_LEN_MAX	((__uint32_t)~0U)
1407 #define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1408 				 XFS_RMAP_OFF_BMBT_BLOCK | \
1409 				 XFS_RMAP_OFF_UNWRITTEN)
1410 #define XFS_RMAP_OFF_MASK	((__uint64_t)0x3FFFFFFFFFFFFFULL)
1411 
1412 #define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1413 
1414 #define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1415 #define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1416 #define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1417 
1418 #define RMAPBT_STARTBLOCK_BITLEN	32
1419 #define RMAPBT_BLOCKCOUNT_BITLEN	32
1420 #define RMAPBT_OWNER_BITLEN		64
1421 #define RMAPBT_ATTRFLAG_BITLEN		1
1422 #define RMAPBT_BMBTFLAG_BITLEN		1
1423 #define RMAPBT_EXNTFLAG_BITLEN		1
1424 #define RMAPBT_UNUSED_OFFSET_BITLEN	7
1425 #define RMAPBT_OFFSET_BITLEN		54
1426 
1427 #define XFS_RMAP_ATTR_FORK		(1 << 0)
1428 #define XFS_RMAP_BMBT_BLOCK		(1 << 1)
1429 #define XFS_RMAP_UNWRITTEN		(1 << 2)
1430 #define XFS_RMAP_KEY_FLAGS		(XFS_RMAP_ATTR_FORK | \
1431 					 XFS_RMAP_BMBT_BLOCK)
1432 #define XFS_RMAP_REC_FLAGS		(XFS_RMAP_UNWRITTEN)
1433 struct xfs_rmap_irec {
1434 	xfs_agblock_t	rm_startblock;	/* extent start block */
1435 	xfs_extlen_t	rm_blockcount;	/* extent length */
1436 	__uint64_t	rm_owner;	/* extent owner */
1437 	__uint64_t	rm_offset;	/* offset within the owner */
1438 	unsigned int	rm_flags;	/* state flags */
1439 };
1440 
1441 /*
1442  * Key structure
1443  *
1444  * We don't use the length for lookups
1445  */
1446 struct xfs_rmap_key {
1447 	__be32		rm_startblock;	/* extent start block */
1448 	__be64		rm_owner;	/* extent owner */
1449 	__be64		rm_offset;	/* offset within the owner */
1450 } __attribute__((packed));
1451 
1452 /* btree pointer type */
1453 typedef __be32 xfs_rmap_ptr_t;
1454 
1455 #define	XFS_RMAP_BLOCK(mp) \
1456 	(xfs_sb_version_hasfinobt(&((mp)->m_sb)) ? \
1457 	 XFS_FIBT_BLOCK(mp) + 1 : \
1458 	 XFS_IBT_BLOCK(mp) + 1)
1459 
1460 /*
1461  * Reference Count Btree format definitions
1462  *
1463  */
1464 #define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1465 
1466 unsigned int xfs_refc_block(struct xfs_mount *mp);
1467 
1468 /*
1469  * Data record/key structure
1470  *
1471  * Each record associates a range of physical blocks (starting at
1472  * rc_startblock and ending rc_blockcount blocks later) with a reference
1473  * count (rc_refcount).  Extents that are being used to stage a copy on
1474  * write (CoW) operation are recorded in the refcount btree with a
1475  * refcount of 1.  All other records must have a refcount > 1 and must
1476  * track an extent mapped only by file data forks.
1477  *
1478  * Extents with a single owner (attributes, metadata, non-shared file
1479  * data) are not tracked here.  Free space is also not tracked here.
1480  * This is consistent with pre-reflink XFS.
1481  */
1482 
1483 /*
1484  * Extents that are being used to stage a copy on write are stored
1485  * in the refcount btree with a refcount of 1 and the upper bit set
1486  * on the startblock.  This speeds up mount time deletion of stale
1487  * staging extents because they're all at the right side of the tree.
1488  */
1489 #define XFS_REFC_COW_START		((xfs_agblock_t)(1U << 31))
1490 #define REFCNTBT_COWFLAG_BITLEN		1
1491 #define REFCNTBT_AGBLOCK_BITLEN		31
1492 
1493 struct xfs_refcount_rec {
1494 	__be32		rc_startblock;	/* starting block number */
1495 	__be32		rc_blockcount;	/* count of blocks */
1496 	__be32		rc_refcount;	/* number of inodes linked here */
1497 };
1498 
1499 struct xfs_refcount_key {
1500 	__be32		rc_startblock;	/* starting block number */
1501 };
1502 
1503 struct xfs_refcount_irec {
1504 	xfs_agblock_t	rc_startblock;	/* starting block number */
1505 	xfs_extlen_t	rc_blockcount;	/* count of free blocks */
1506 	xfs_nlink_t	rc_refcount;	/* number of inodes linked here */
1507 };
1508 
1509 #define MAXREFCOUNT	((xfs_nlink_t)~0U)
1510 #define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
1511 
1512 /* btree pointer type */
1513 typedef __be32 xfs_refcount_ptr_t;
1514 
1515 
1516 /*
1517  * BMAP Btree format definitions
1518  *
1519  * This includes both the root block definition that sits inside an inode fork
1520  * and the record/pointer formats for the leaf/node in the blocks.
1521  */
1522 #define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1523 #define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1524 
1525 /*
1526  * Bmap root header, on-disk form only.
1527  */
1528 typedef struct xfs_bmdr_block {
1529 	__be16		bb_level;	/* 0 is a leaf */
1530 	__be16		bb_numrecs;	/* current # of data records */
1531 } xfs_bmdr_block_t;
1532 
1533 /*
1534  * Bmap btree record and extent descriptor.
1535  *  l0:63 is an extent flag (value 1 indicates non-normal).
1536  *  l0:9-62 are startoff.
1537  *  l0:0-8 and l1:21-63 are startblock.
1538  *  l1:0-20 are blockcount.
1539  */
1540 #define BMBT_EXNTFLAG_BITLEN	1
1541 #define BMBT_STARTOFF_BITLEN	54
1542 #define BMBT_STARTBLOCK_BITLEN	52
1543 #define BMBT_BLOCKCOUNT_BITLEN	21
1544 
1545 typedef struct xfs_bmbt_rec {
1546 	__be64			l0, l1;
1547 } xfs_bmbt_rec_t;
1548 
1549 typedef __uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1550 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1551 
1552 typedef struct xfs_bmbt_rec_host {
1553 	__uint64_t		l0, l1;
1554 } xfs_bmbt_rec_host_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 
isnullstartblock(xfs_fsblock_t x)1564 static inline int isnullstartblock(xfs_fsblock_t x)
1565 {
1566 	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1567 }
1568 
nullstartblock(int k)1569 static inline xfs_fsblock_t nullstartblock(int k)
1570 {
1571 	ASSERT(k < (1 << STARTBLOCKVALBITS));
1572 	return STARTBLOCKMASK | (k);
1573 }
1574 
startblockval(xfs_fsblock_t x)1575 static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1576 {
1577 	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1578 }
1579 
1580 /*
1581  * Possible extent formats.
1582  */
1583 typedef enum {
1584 	XFS_EXTFMT_NOSTATE = 0,
1585 	XFS_EXTFMT_HASSTATE
1586 } xfs_exntfmt_t;
1587 
1588 /*
1589  * Possible extent states.
1590  */
1591 typedef enum {
1592 	XFS_EXT_NORM, XFS_EXT_UNWRITTEN,
1593 	XFS_EXT_DMAPI_OFFLINE, XFS_EXT_INVALID
1594 } xfs_exntst_t;
1595 
1596 /*
1597  * Incore version of above.
1598  */
1599 typedef struct xfs_bmbt_irec
1600 {
1601 	xfs_fileoff_t	br_startoff;	/* starting file offset */
1602 	xfs_fsblock_t	br_startblock;	/* starting block number */
1603 	xfs_filblks_t	br_blockcount;	/* number of blocks */
1604 	xfs_exntst_t	br_state;	/* extent state */
1605 } xfs_bmbt_irec_t;
1606 
1607 /*
1608  * Key structure for non-leaf levels of the tree.
1609  */
1610 typedef struct xfs_bmbt_key {
1611 	__be64		br_startoff;	/* starting file offset */
1612 } xfs_bmbt_key_t, xfs_bmdr_key_t;
1613 
1614 /* btree pointer type */
1615 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1616 
1617 
1618 /*
1619  * Generic Btree block format definitions
1620  *
1621  * This is a combination of the actual format used on disk for short and long
1622  * format btrees.  The first three fields are shared by both format, but the
1623  * pointers are different and should be used with care.
1624  *
1625  * To get the size of the actual short or long form headers please use the size
1626  * macros below.  Never use sizeof(xfs_btree_block).
1627  *
1628  * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1629  * with the crc feature bit, and all accesses to them must be conditional on
1630  * that flag.
1631  */
1632 /* short form block header */
1633 struct xfs_btree_block_shdr {
1634 	__be32		bb_leftsib;
1635 	__be32		bb_rightsib;
1636 
1637 	__be64		bb_blkno;
1638 	__be64		bb_lsn;
1639 	uuid_t		bb_uuid;
1640 	__be32		bb_owner;
1641 	__le32		bb_crc;
1642 };
1643 
1644 /* long form block header */
1645 struct xfs_btree_block_lhdr {
1646 	__be64		bb_leftsib;
1647 	__be64		bb_rightsib;
1648 
1649 	__be64		bb_blkno;
1650 	__be64		bb_lsn;
1651 	uuid_t		bb_uuid;
1652 	__be64		bb_owner;
1653 	__le32		bb_crc;
1654 	__be32		bb_pad; /* padding for alignment */
1655 };
1656 
1657 struct xfs_btree_block {
1658 	__be32		bb_magic;	/* magic number for block type */
1659 	__be16		bb_level;	/* 0 is a leaf */
1660 	__be16		bb_numrecs;	/* current # of data records */
1661 	union {
1662 		struct xfs_btree_block_shdr s;
1663 		struct xfs_btree_block_lhdr l;
1664 	} bb_u;				/* rest */
1665 };
1666 
1667 /* size of a short form block */
1668 #define XFS_BTREE_SBLOCK_LEN \
1669 	(offsetof(struct xfs_btree_block, bb_u) + \
1670 	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1671 /* size of a long form block */
1672 #define XFS_BTREE_LBLOCK_LEN \
1673 	(offsetof(struct xfs_btree_block, bb_u) + \
1674 	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1675 
1676 /* sizes of CRC enabled btree blocks */
1677 #define XFS_BTREE_SBLOCK_CRC_LEN \
1678 	(offsetof(struct xfs_btree_block, bb_u) + \
1679 	 sizeof(struct xfs_btree_block_shdr))
1680 #define XFS_BTREE_LBLOCK_CRC_LEN \
1681 	(offsetof(struct xfs_btree_block, bb_u) + \
1682 	 sizeof(struct xfs_btree_block_lhdr))
1683 
1684 #define XFS_BTREE_SBLOCK_CRC_OFF \
1685 	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1686 #define XFS_BTREE_LBLOCK_CRC_OFF \
1687 	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1688 
1689 /*
1690  * On-disk XFS access control list structure.
1691  */
1692 struct xfs_acl_entry {
1693 	__be32	ae_tag;
1694 	__be32	ae_id;
1695 	__be16	ae_perm;
1696 	__be16	ae_pad;		/* fill the implicit hole in the structure */
1697 };
1698 
1699 struct xfs_acl {
1700 	__be32			acl_cnt;
1701 	struct xfs_acl_entry	acl_entry[0];
1702 };
1703 
1704 /*
1705  * The number of ACL entries allowed is defined by the on-disk format.
1706  * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1707  * limited only by the maximum size of the xattr that stores the information.
1708  */
1709 #define XFS_ACL_MAX_ENTRIES(mp)	\
1710 	(xfs_sb_version_hascrc(&mp->m_sb) \
1711 		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1712 						sizeof(struct xfs_acl_entry) \
1713 		: 25)
1714 
1715 #define XFS_ACL_SIZE(cnt) \
1716 	(sizeof(struct xfs_acl) + \
1717 		sizeof(struct xfs_acl_entry) * cnt)
1718 
1719 #define XFS_ACL_MAX_SIZE(mp) \
1720 	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1721 
1722 
1723 /* On-disk XFS extended attribute names */
1724 #define SGI_ACL_FILE		"SGI_ACL_FILE"
1725 #define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
1726 #define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
1727 #define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
1728 
1729 #endif /* __XFS_FORMAT_H__ */
1730