xref: /linux/fs/xfs/libxfs/xfs_format.h (revision b477ff98d903618a1ab8247861f2ea6e70c0f0f8)
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 be padded to 64 bit alignment.
94  */
95 typedef struct xfs_sb {
96 	uint32_t	sb_magicnum;	/* magic number == XFS_SB_MAGIC */
97 	uint32_t	sb_blocksize;	/* logical block size, bytes */
98 	xfs_rfsblock_t	sb_dblocks;	/* number of data blocks */
99 	xfs_rfsblock_t	sb_rblocks;	/* number of realtime blocks */
100 	xfs_rtbxlen_t	sb_rextents;	/* number of realtime extents */
101 	uuid_t		sb_uuid;	/* user-visible file system unique id */
102 	xfs_fsblock_t	sb_logstart;	/* starting block of log if internal */
103 	xfs_ino_t	sb_rootino;	/* root inode number */
104 	xfs_ino_t	sb_rbmino;	/* bitmap inode for realtime extents */
105 	xfs_ino_t	sb_rsumino;	/* summary inode for rt bitmap */
106 	xfs_agblock_t	sb_rextsize;	/* realtime extent size, blocks */
107 	xfs_agblock_t	sb_agblocks;	/* size of an allocation group */
108 	xfs_agnumber_t	sb_agcount;	/* number of allocation groups */
109 	xfs_extlen_t	sb_rbmblocks;	/* number of rt bitmap blocks */
110 	xfs_extlen_t	sb_logblocks;	/* number of log blocks */
111 	uint16_t	sb_versionnum;	/* header version == XFS_SB_VERSION */
112 	uint16_t	sb_sectsize;	/* volume sector size, bytes */
113 	uint16_t	sb_inodesize;	/* inode size, bytes */
114 	uint16_t	sb_inopblock;	/* inodes per block */
115 	char		sb_fname[XFSLABEL_MAX]; /* file system name */
116 	uint8_t		sb_blocklog;	/* log2 of sb_blocksize */
117 	uint8_t		sb_sectlog;	/* log2 of sb_sectsize */
118 	uint8_t		sb_inodelog;	/* log2 of sb_inodesize */
119 	uint8_t		sb_inopblog;	/* log2 of sb_inopblock */
120 	uint8_t		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
121 	uint8_t		sb_rextslog;	/* log2 of sb_rextents */
122 	uint8_t		sb_inprogress;	/* mkfs is in progress, don't mount */
123 	uint8_t		sb_imax_pct;	/* max % of fs for inode space */
124 					/* statistics */
125 	/*
126 	 * These fields must remain contiguous.  If you really
127 	 * want to change their layout, make sure you fix the
128 	 * code in xfs_trans_apply_sb_deltas().
129 	 */
130 	uint64_t	sb_icount;	/* allocated inodes */
131 	uint64_t	sb_ifree;	/* free inodes */
132 	uint64_t	sb_fdblocks;	/* free data blocks */
133 	uint64_t	sb_frextents;	/* free realtime extents */
134 	/*
135 	 * End contiguous fields.
136 	 */
137 	xfs_ino_t	sb_uquotino;	/* user quota inode */
138 	xfs_ino_t	sb_gquotino;	/* group quota inode */
139 	uint16_t	sb_qflags;	/* quota flags */
140 	uint8_t		sb_flags;	/* misc. flags */
141 	uint8_t		sb_shared_vn;	/* shared version number */
142 	xfs_extlen_t	sb_inoalignmt;	/* inode chunk alignment, fsblocks */
143 	uint32_t	sb_unit;	/* stripe or raid unit */
144 	uint32_t	sb_width;	/* stripe or raid width */
145 	uint8_t		sb_dirblklog;	/* log2 of dir block size (fsbs) */
146 	uint8_t		sb_logsectlog;	/* log2 of the log sector size */
147 	uint16_t	sb_logsectsize;	/* sector size for the log, bytes */
148 	uint32_t	sb_logsunit;	/* stripe unit size for the log */
149 	uint32_t	sb_features2;	/* additional feature bits */
150 
151 	/*
152 	 * bad features2 field as a result of failing to pad the sb structure to
153 	 * 64 bits. Some machines will be using this field for features2 bits.
154 	 * Easiest just to mark it bad and not use it for anything else.
155 	 *
156 	 * This is not kept up to date in memory; it is always overwritten by
157 	 * the value in sb_features2 when formatting the incore superblock to
158 	 * the disk buffer.
159 	 */
160 	uint32_t	sb_bad_features2;
161 
162 	/* version 5 superblock fields start here */
163 
164 	/* feature masks */
165 	uint32_t	sb_features_compat;
166 	uint32_t	sb_features_ro_compat;
167 	uint32_t	sb_features_incompat;
168 	uint32_t	sb_features_log_incompat;
169 
170 	uint32_t	sb_crc;		/* superblock crc */
171 	xfs_extlen_t	sb_spino_align;	/* sparse inode chunk alignment */
172 
173 	xfs_ino_t	sb_pquotino;	/* project quota inode */
174 	xfs_lsn_t	sb_lsn;		/* last write sequence */
175 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
176 
177 	xfs_ino_t	sb_metadirino;	/* metadata directory tree root */
178 
179 	xfs_rgnumber_t	sb_rgcount;	/* number of realtime groups */
180 	xfs_rtxlen_t	sb_rgextents;	/* size of a realtime group in rtx */
181 
182 	uint8_t		sb_rgblklog;    /* rt group number shift */
183 	uint8_t		sb_pad[7];	/* zeroes */
184 
185 	/* must be padded to 64 bit alignment */
186 } xfs_sb_t;
187 
188 /*
189  * Superblock - on disk version.
190  * Must be padded to 64 bit alignment.
191  */
192 struct xfs_dsb {
193 	__be32		sb_magicnum;	/* magic number == XFS_SB_MAGIC */
194 	__be32		sb_blocksize;	/* logical block size, bytes */
195 	__be64		sb_dblocks;	/* number of data blocks */
196 	__be64		sb_rblocks;	/* number of realtime blocks */
197 	__be64		sb_rextents;	/* number of realtime extents */
198 	uuid_t		sb_uuid;	/* user-visible file system unique id */
199 	__be64		sb_logstart;	/* starting block of log if internal */
200 	__be64		sb_rootino;	/* root inode number */
201 	__be64		sb_rbmino;	/* bitmap inode for realtime extents */
202 	__be64		sb_rsumino;	/* summary inode for rt bitmap */
203 	__be32		sb_rextsize;	/* realtime extent size, blocks */
204 	__be32		sb_agblocks;	/* size of an allocation group */
205 	__be32		sb_agcount;	/* number of allocation groups */
206 	__be32		sb_rbmblocks;	/* number of rt bitmap blocks */
207 	__be32		sb_logblocks;	/* number of log blocks */
208 	__be16		sb_versionnum;	/* header version == XFS_SB_VERSION */
209 	__be16		sb_sectsize;	/* volume sector size, bytes */
210 	__be16		sb_inodesize;	/* inode size, bytes */
211 	__be16		sb_inopblock;	/* inodes per block */
212 	char		sb_fname[XFSLABEL_MAX]; /* file system name */
213 	__u8		sb_blocklog;	/* log2 of sb_blocksize */
214 	__u8		sb_sectlog;	/* log2 of sb_sectsize */
215 	__u8		sb_inodelog;	/* log2 of sb_inodesize */
216 	__u8		sb_inopblog;	/* log2 of sb_inopblock */
217 	__u8		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
218 	__u8		sb_rextslog;	/* log2 of sb_rextents */
219 	__u8		sb_inprogress;	/* mkfs is in progress, don't mount */
220 	__u8		sb_imax_pct;	/* max % of fs for inode space */
221 					/* statistics */
222 	/*
223 	 * These fields must remain contiguous.  If you really
224 	 * want to change their layout, make sure you fix the
225 	 * code in xfs_trans_apply_sb_deltas().
226 	 */
227 	__be64		sb_icount;	/* allocated inodes */
228 	__be64		sb_ifree;	/* free inodes */
229 	__be64		sb_fdblocks;	/* free data blocks */
230 	__be64		sb_frextents;	/* free realtime extents */
231 	/*
232 	 * End contiguous fields.
233 	 */
234 	__be64		sb_uquotino;	/* user quota inode */
235 	__be64		sb_gquotino;	/* group quota inode */
236 	__be16		sb_qflags;	/* quota flags */
237 	__u8		sb_flags;	/* misc. flags */
238 	__u8		sb_shared_vn;	/* shared version number */
239 	__be32		sb_inoalignmt;	/* inode chunk alignment, fsblocks */
240 	__be32		sb_unit;	/* stripe or raid unit */
241 	__be32		sb_width;	/* stripe or raid width */
242 	__u8		sb_dirblklog;	/* log2 of dir block size (fsbs) */
243 	__u8		sb_logsectlog;	/* log2 of the log sector size */
244 	__be16		sb_logsectsize;	/* sector size for the log, bytes */
245 	__be32		sb_logsunit;	/* stripe unit size for the log */
246 	__be32		sb_features2;	/* additional feature bits */
247 	/*
248 	 * bad features2 field as a result of failing to pad the sb
249 	 * structure to 64 bits. Some machines will be using this field
250 	 * for features2 bits. Easiest just to mark it bad and not use
251 	 * it for anything else.
252 	 */
253 	__be32		sb_bad_features2;
254 
255 	/* version 5 superblock fields start here */
256 
257 	/* feature masks */
258 	__be32		sb_features_compat;
259 	__be32		sb_features_ro_compat;
260 	__be32		sb_features_incompat;
261 	__be32		sb_features_log_incompat;
262 
263 	__le32		sb_crc;		/* superblock crc */
264 	__be32		sb_spino_align;	/* sparse inode chunk alignment */
265 
266 	__be64		sb_pquotino;	/* project quota inode */
267 	__be64		sb_lsn;		/* last write sequence */
268 	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
269 
270 	__be64		sb_metadirino;	/* metadata directory tree root */
271 	__be32		sb_rgcount;	/* # of realtime groups */
272 	__be32		sb_rgextents;	/* size of rtgroup in rtx */
273 
274 	__u8		sb_rgblklog;    /* rt group number shift */
275 	__u8		sb_pad[7];	/* zeroes */
276 
277 	/*
278 	 * The size of this structure must be padded to 64 bit alignment.
279 	 *
280 	 * NOTE: Don't forget to update secondary_sb_whack in xfs_repair when
281 	 * adding new fields here.
282 	 */
283 };
284 
285 #define XFS_SB_CRC_OFF		offsetof(struct xfs_dsb, sb_crc)
286 
287 /*
288  * Misc. Flags - warning - these will be cleared by xfs_repair unless
289  * a feature bit is set when the flag is used.
290  */
291 #define XFS_SBF_NOFLAGS		0x00	/* no flags set */
292 #define XFS_SBF_READONLY	0x01	/* only read-only mounts allowed */
293 
294 /*
295  * define max. shared version we can interoperate with
296  */
297 #define XFS_SB_MAX_SHARED_VN	0
298 
299 #define	XFS_SB_VERSION_NUM(sbp)	((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
300 
xfs_sb_is_v5(const struct xfs_sb * sbp)301 static inline bool xfs_sb_is_v5(const struct xfs_sb *sbp)
302 {
303 	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
304 }
305 
306 /*
307  * Detect a mismatched features2 field.  Older kernels read/wrote
308  * this into the wrong slot, so to be safe we keep them in sync.
309  */
xfs_sb_has_mismatched_features2(const struct xfs_sb * sbp)310 static inline bool xfs_sb_has_mismatched_features2(const struct xfs_sb *sbp)
311 {
312 	return sbp->sb_bad_features2 != sbp->sb_features2;
313 }
314 
xfs_sb_version_hasmorebits(const struct xfs_sb * sbp)315 static inline bool xfs_sb_version_hasmorebits(const struct xfs_sb *sbp)
316 {
317 	return xfs_sb_is_v5(sbp) ||
318 	       (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
319 }
320 
xfs_sb_version_addattr(struct xfs_sb * sbp)321 static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
322 {
323 	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
324 }
325 
xfs_sb_version_addquota(struct xfs_sb * sbp)326 static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
327 {
328 	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
329 }
330 
xfs_sb_version_addattr2(struct xfs_sb * sbp)331 static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
332 {
333 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
334 	sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
335 }
336 
xfs_sb_version_addprojid32(struct xfs_sb * sbp)337 static inline void xfs_sb_version_addprojid32(struct xfs_sb *sbp)
338 {
339 	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
340 	sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
341 }
342 
343 /*
344  * Extended v5 superblock feature masks. These are to be used for new v5
345  * superblock features only.
346  *
347  * Compat features are new features that old kernels will not notice or affect
348  * and so can mount read-write without issues.
349  *
350  * RO-Compat (read only) are features that old kernels can read but will break
351  * if they write. Hence only read-only mounts of such filesystems are allowed on
352  * kernels that don't support the feature bit.
353  *
354  * InCompat features are features which old kernels will not understand and so
355  * must not mount.
356  *
357  * Log-InCompat features are for changes to log formats or new transactions that
358  * can't be replayed on older kernels. The fields are set when the filesystem is
359  * mounted, and a clean unmount clears the fields.
360  */
361 #define XFS_SB_FEAT_COMPAT_ALL 0
362 #define XFS_SB_FEAT_COMPAT_UNKNOWN	~XFS_SB_FEAT_COMPAT_ALL
363 static inline bool
xfs_sb_has_compat_feature(const struct xfs_sb * sbp,uint32_t feature)364 xfs_sb_has_compat_feature(
365 	const struct xfs_sb	*sbp,
366 	uint32_t		feature)
367 {
368 	return (sbp->sb_features_compat & feature) != 0;
369 }
370 
371 #define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
372 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
373 #define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
374 #define XFS_SB_FEAT_RO_COMPAT_INOBTCNT (1 << 3)		/* inobt block counts */
375 #define XFS_SB_FEAT_RO_COMPAT_ALL \
376 		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
377 		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
378 		 XFS_SB_FEAT_RO_COMPAT_REFLINK| \
379 		 XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
380 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
381 static inline bool
xfs_sb_has_ro_compat_feature(const struct xfs_sb * sbp,uint32_t feature)382 xfs_sb_has_ro_compat_feature(
383 	const struct xfs_sb	*sbp,
384 	uint32_t		feature)
385 {
386 	return (sbp->sb_features_ro_compat & feature) != 0;
387 }
388 
389 #define XFS_SB_FEAT_INCOMPAT_FTYPE	(1 << 0)  /* filetype in dirent */
390 #define XFS_SB_FEAT_INCOMPAT_SPINODES	(1 << 1)  /* sparse inode chunks */
391 #define XFS_SB_FEAT_INCOMPAT_META_UUID	(1 << 2)  /* metadata UUID */
392 #define XFS_SB_FEAT_INCOMPAT_BIGTIME	(1 << 3)  /* large timestamps */
393 #define XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR (1 << 4) /* needs xfs_repair */
394 #define XFS_SB_FEAT_INCOMPAT_NREXT64	(1 << 5)  /* large extent counters */
395 #define XFS_SB_FEAT_INCOMPAT_EXCHRANGE	(1 << 6)  /* exchangerange supported */
396 #define XFS_SB_FEAT_INCOMPAT_PARENT	(1 << 7)  /* parent pointers */
397 #define XFS_SB_FEAT_INCOMPAT_METADIR	(1 << 8)  /* metadata dir tree */
398 #define XFS_SB_FEAT_INCOMPAT_ALL \
399 		(XFS_SB_FEAT_INCOMPAT_FTYPE | \
400 		 XFS_SB_FEAT_INCOMPAT_SPINODES | \
401 		 XFS_SB_FEAT_INCOMPAT_META_UUID | \
402 		 XFS_SB_FEAT_INCOMPAT_BIGTIME | \
403 		 XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR | \
404 		 XFS_SB_FEAT_INCOMPAT_NREXT64 | \
405 		 XFS_SB_FEAT_INCOMPAT_EXCHRANGE | \
406 		 XFS_SB_FEAT_INCOMPAT_PARENT | \
407 		 XFS_SB_FEAT_INCOMPAT_METADIR)
408 
409 #define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
410 static inline bool
xfs_sb_has_incompat_feature(const struct xfs_sb * sbp,uint32_t feature)411 xfs_sb_has_incompat_feature(
412 	const struct xfs_sb	*sbp,
413 	uint32_t		feature)
414 {
415 	return (sbp->sb_features_incompat & feature) != 0;
416 }
417 
418 #define XFS_SB_FEAT_INCOMPAT_LOG_XATTRS   (1 << 0)	/* Delayed Attributes */
419 #define XFS_SB_FEAT_INCOMPAT_LOG_ALL \
420 	(XFS_SB_FEAT_INCOMPAT_LOG_XATTRS)
421 #define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
422 static inline bool
xfs_sb_has_incompat_log_feature(const struct xfs_sb * sbp,uint32_t feature)423 xfs_sb_has_incompat_log_feature(
424 	const struct xfs_sb	*sbp,
425 	uint32_t		feature)
426 {
427 	return (sbp->sb_features_log_incompat & feature) != 0;
428 }
429 
430 static inline void
xfs_sb_remove_incompat_log_features(struct xfs_sb * sbp)431 xfs_sb_remove_incompat_log_features(
432 	struct xfs_sb	*sbp)
433 {
434 	sbp->sb_features_log_incompat &= ~XFS_SB_FEAT_INCOMPAT_LOG_ALL;
435 }
436 
437 static inline void
xfs_sb_add_incompat_log_features(struct xfs_sb * sbp,unsigned int features)438 xfs_sb_add_incompat_log_features(
439 	struct xfs_sb	*sbp,
440 	unsigned int	features)
441 {
442 	sbp->sb_features_log_incompat |= features;
443 }
444 
xfs_sb_version_haslogxattrs(const struct xfs_sb * sbp)445 static inline bool xfs_sb_version_haslogxattrs(const struct xfs_sb *sbp)
446 {
447 	return xfs_sb_is_v5(sbp) && (sbp->sb_features_log_incompat &
448 		 XFS_SB_FEAT_INCOMPAT_LOG_XATTRS);
449 }
450 
451 static inline bool
xfs_is_quota_inode(struct xfs_sb * sbp,xfs_ino_t ino)452 xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
453 {
454 	return (ino == sbp->sb_uquotino ||
455 		ino == sbp->sb_gquotino ||
456 		ino == sbp->sb_pquotino);
457 }
458 
459 #define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
460 #define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
461 
462 #define	XFS_HDR_BLOCK(mp,d)	((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
463 #define	XFS_DADDR_TO_FSB(mp,d)	XFS_AGB_TO_FSB(mp, \
464 			xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
465 #define	XFS_FSB_TO_DADDR(mp,fsbno)	XFS_AGB_TO_DADDR(mp, \
466 			XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
467 
468 /*
469  * File system sector to basic block conversions.
470  */
471 #define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
472 
473 /*
474  * File system block to basic block conversions.
475  */
476 #define	XFS_FSB_TO_BB(mp,fsbno)	((fsbno) << (mp)->m_blkbb_log)
477 #define	XFS_BB_TO_FSB(mp,bb)	\
478 	(((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
479 #define	XFS_BB_TO_FSBT(mp,bb)	((bb) >> (mp)->m_blkbb_log)
480 
481 /*
482  * File system block to byte conversions.
483  */
484 #define XFS_FSB_TO_B(mp,fsbno)	((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
485 #define XFS_B_TO_FSB(mp,b)	\
486 	((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
487 #define XFS_B_TO_FSBT(mp,b)	(((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
488 
489 /*
490  * Allocation group header
491  *
492  * This is divided into three structures, placed in sequential 512-byte
493  * buffers after a copy of the superblock (also in a 512-byte buffer).
494  */
495 #define	XFS_AGF_MAGIC	0x58414746	/* 'XAGF' */
496 #define	XFS_AGI_MAGIC	0x58414749	/* 'XAGI' */
497 #define	XFS_AGFL_MAGIC	0x5841464c	/* 'XAFL' */
498 #define	XFS_AGF_VERSION	1
499 #define	XFS_AGI_VERSION	1
500 
501 #define	XFS_AGF_GOOD_VERSION(v)	((v) == XFS_AGF_VERSION)
502 #define	XFS_AGI_GOOD_VERSION(v)	((v) == XFS_AGI_VERSION)
503 
504 /*
505  * agf_cnt_level in the first AGF overlaps the EFS superblock's magic number.
506  * Since the magic numbers valid for EFS are > 64k, our value cannot be confused
507  * for an EFS superblock.
508  */
509 
510 typedef struct xfs_agf {
511 	/*
512 	 * Common allocation group header information
513 	 */
514 	__be32		agf_magicnum;	/* magic number == XFS_AGF_MAGIC */
515 	__be32		agf_versionnum;	/* header version == XFS_AGF_VERSION */
516 	__be32		agf_seqno;	/* sequence # starting from 0 */
517 	__be32		agf_length;	/* size in blocks of a.g. */
518 	/*
519 	 * Freespace and rmap information
520 	 */
521 	__be32		agf_bno_root;	/* bnobt root block */
522 	__be32		agf_cnt_root;	/* cntbt root block */
523 	__be32		agf_rmap_root;	/* rmapbt root block */
524 
525 	__be32		agf_bno_level;	/* bnobt btree levels */
526 	__be32		agf_cnt_level;	/* cntbt btree levels */
527 	__be32		agf_rmap_level;	/* rmapbt btree levels */
528 
529 	__be32		agf_flfirst;	/* first freelist block's index */
530 	__be32		agf_fllast;	/* last freelist block's index */
531 	__be32		agf_flcount;	/* count of blocks in freelist */
532 	__be32		agf_freeblks;	/* total free blocks */
533 
534 	__be32		agf_longest;	/* longest free space */
535 	__be32		agf_btreeblks;	/* # of blocks held in AGF btrees */
536 	uuid_t		agf_uuid;	/* uuid of filesystem */
537 
538 	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
539 	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
540 
541 	__be32		agf_refcount_root;	/* refcount tree root block */
542 	__be32		agf_refcount_level;	/* refcount btree levels */
543 
544 	/*
545 	 * reserve some contiguous space for future logged fields before we add
546 	 * the unlogged fields. This makes the range logging via flags and
547 	 * structure offsets much simpler.
548 	 */
549 	__be64		agf_spare64[14];
550 
551 	/* unlogged fields, written during buffer writeback. */
552 	__be64		agf_lsn;	/* last write sequence */
553 	__be32		agf_crc;	/* crc of agf sector */
554 	__be32		agf_spare2;
555 
556 	/* structure must be padded to 64 bit alignment */
557 } xfs_agf_t;
558 
559 #define XFS_AGF_CRC_OFF		offsetof(struct xfs_agf, agf_crc)
560 
561 #define	XFS_AGF_MAGICNUM	(1u << 0)
562 #define	XFS_AGF_VERSIONNUM	(1u << 1)
563 #define	XFS_AGF_SEQNO		(1u << 2)
564 #define	XFS_AGF_LENGTH		(1u << 3)
565 #define	XFS_AGF_ROOTS		(1u << 4)
566 #define	XFS_AGF_LEVELS		(1u << 5)
567 #define	XFS_AGF_FLFIRST		(1u << 6)
568 #define	XFS_AGF_FLLAST		(1u << 7)
569 #define	XFS_AGF_FLCOUNT		(1u << 8)
570 #define	XFS_AGF_FREEBLKS	(1u << 9)
571 #define	XFS_AGF_LONGEST		(1u << 10)
572 #define	XFS_AGF_BTREEBLKS	(1u << 11)
573 #define	XFS_AGF_UUID		(1u << 12)
574 #define	XFS_AGF_RMAP_BLOCKS	(1u << 13)
575 #define	XFS_AGF_REFCOUNT_BLOCKS	(1u << 14)
576 #define	XFS_AGF_REFCOUNT_ROOT	(1u << 15)
577 #define	XFS_AGF_REFCOUNT_LEVEL	(1u << 16)
578 #define	XFS_AGF_SPARE64		(1u << 17)
579 #define	XFS_AGF_NUM_BITS	18
580 #define	XFS_AGF_ALL_BITS	((1u << XFS_AGF_NUM_BITS) - 1)
581 
582 #define XFS_AGF_FLAGS \
583 	{ XFS_AGF_MAGICNUM,	"MAGICNUM" }, \
584 	{ XFS_AGF_VERSIONNUM,	"VERSIONNUM" }, \
585 	{ XFS_AGF_SEQNO,	"SEQNO" }, \
586 	{ XFS_AGF_LENGTH,	"LENGTH" }, \
587 	{ XFS_AGF_ROOTS,	"ROOTS" }, \
588 	{ XFS_AGF_LEVELS,	"LEVELS" }, \
589 	{ XFS_AGF_FLFIRST,	"FLFIRST" }, \
590 	{ XFS_AGF_FLLAST,	"FLLAST" }, \
591 	{ XFS_AGF_FLCOUNT,	"FLCOUNT" }, \
592 	{ XFS_AGF_FREEBLKS,	"FREEBLKS" }, \
593 	{ XFS_AGF_LONGEST,	"LONGEST" }, \
594 	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
595 	{ XFS_AGF_UUID,		"UUID" }, \
596 	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
597 	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
598 	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
599 	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
600 	{ XFS_AGF_SPARE64,	"SPARE64" }
601 
602 /* disk block (xfs_daddr_t) in the AG */
603 #define XFS_AGF_DADDR(mp)	((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
604 #define	XFS_AGF_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
605 
606 /*
607  * Size of the unlinked inode hash table in the agi.
608  */
609 #define	XFS_AGI_UNLINKED_BUCKETS	64
610 
611 typedef struct xfs_agi {
612 	/*
613 	 * Common allocation group header information
614 	 */
615 	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
616 	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
617 	__be32		agi_seqno;	/* sequence # starting from 0 */
618 	__be32		agi_length;	/* size in blocks of a.g. */
619 	/*
620 	 * Inode information
621 	 * Inodes are mapped by interpreting the inode number, so no
622 	 * mapping data is needed here.
623 	 */
624 	__be32		agi_count;	/* count of allocated inodes */
625 	__be32		agi_root;	/* root of inode btree */
626 	__be32		agi_level;	/* levels in inode btree */
627 	__be32		agi_freecount;	/* number of free inodes */
628 
629 	__be32		agi_newino;	/* new inode just allocated */
630 	__be32		agi_dirino;	/* last directory inode chunk */
631 	/*
632 	 * Hash table of inodes which have been unlinked but are
633 	 * still being referenced.
634 	 */
635 	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
636 	/*
637 	 * This marks the end of logging region 1 and start of logging region 2.
638 	 */
639 	uuid_t		agi_uuid;	/* uuid of filesystem */
640 	__be32		agi_crc;	/* crc of agi sector */
641 	__be32		agi_pad32;
642 	__be64		agi_lsn;	/* last write sequence */
643 
644 	__be32		agi_free_root; /* root of the free inode btree */
645 	__be32		agi_free_level;/* levels in free inode btree */
646 
647 	__be32		agi_iblocks;	/* inobt blocks used */
648 	__be32		agi_fblocks;	/* finobt blocks used */
649 
650 	/* structure must be padded to 64 bit alignment */
651 } xfs_agi_t;
652 
653 #define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
654 
655 #define	XFS_AGI_MAGICNUM	(1u << 0)
656 #define	XFS_AGI_VERSIONNUM	(1u << 1)
657 #define	XFS_AGI_SEQNO		(1u << 2)
658 #define	XFS_AGI_LENGTH		(1u << 3)
659 #define	XFS_AGI_COUNT		(1u << 4)
660 #define	XFS_AGI_ROOT		(1u << 5)
661 #define	XFS_AGI_LEVEL		(1u << 6)
662 #define	XFS_AGI_FREECOUNT	(1u << 7)
663 #define	XFS_AGI_NEWINO		(1u << 8)
664 #define	XFS_AGI_DIRINO		(1u << 9)
665 #define	XFS_AGI_UNLINKED	(1u << 10)
666 #define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
667 #define	XFS_AGI_ALL_BITS_R1	((1u << XFS_AGI_NUM_BITS_R1) - 1)
668 #define	XFS_AGI_FREE_ROOT	(1u << 11)
669 #define	XFS_AGI_FREE_LEVEL	(1u << 12)
670 #define	XFS_AGI_IBLOCKS		(1u << 13) /* both inobt/finobt block counters */
671 #define	XFS_AGI_NUM_BITS_R2	14
672 
673 /* disk block (xfs_daddr_t) in the AG */
674 #define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
675 #define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
676 
677 /*
678  * The third a.g. block contains the a.g. freelist, an array
679  * of block pointers to blocks owned by the allocation btree code.
680  */
681 #define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
682 #define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
683 #define	XFS_BUF_TO_AGFL(bp)	((struct xfs_agfl *)((bp)->b_addr))
684 
685 struct xfs_agfl {
686 	__be32		agfl_magicnum;
687 	__be32		agfl_seqno;
688 	uuid_t		agfl_uuid;
689 	__be64		agfl_lsn;
690 	__be32		agfl_crc;
691 } __attribute__((packed));
692 
693 #define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
694 
695 #define XFS_AGB_TO_FSB(mp,agno,agbno)	\
696 	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
697 #define	XFS_FSB_TO_AGNO(mp,fsbno)	\
698 	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
699 #define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
700 	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
701 #define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
702 	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
703 		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
704 #define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
705 
706 /*
707  * For checking for bad ranges of xfs_daddr_t's, covering multiple
708  * allocation groups or a single xfs_daddr_t that's a superblock copy.
709  */
710 #define	XFS_AG_CHECK_DADDR(mp,d,len)	\
711 	((len) == 1 ? \
712 	    ASSERT((d) == XFS_SB_DADDR || \
713 		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
714 	    ASSERT(xfs_daddr_to_agno(mp, d) == \
715 		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
716 
717 /*
718  * Realtime bitmap information is accessed by the word, which is currently
719  * stored in host-endian format.  Starting with the realtime groups feature,
720  * the words are stored in be32 ondisk.
721  */
722 union xfs_rtword_raw {
723 	__u32		old;
724 	__be32		rtg;
725 };
726 
727 /*
728  * Realtime summary counts are accessed by the word, which is currently
729  * stored in host-endian format.  Starting with the realtime groups feature,
730  * the words are stored in be32 ondisk.
731  */
732 union xfs_suminfo_raw {
733 	__u32		old;
734 	__be32		rtg;
735 };
736 
737 /*
738  * Realtime allocation groups break the rt section into multiple pieces that
739  * could be locked independently.  Realtime block group numbers are 32-bit
740  * quantities.  Block numbers within a group are also 32-bit quantities, but
741  * the upper bit must never be set.  rtgroup 0 might have a superblock in it,
742  * so the minimum size of an rtgroup is 2 rtx.
743  */
744 #define XFS_MAX_RGBLOCKS	((xfs_rgblock_t)(1U << 31) - 1)
745 #define XFS_MIN_RGEXTENTS	((xfs_rtxlen_t)2)
746 #define XFS_MAX_RGNUMBER	((xfs_rgnumber_t)(-1U))
747 
748 #define XFS_RTSB_MAGIC	0x46726F67	/* 'Frog' */
749 
750 /*
751  * Realtime superblock - on disk version.  Must be padded to 64 bit alignment.
752  * The first block of the realtime volume contains this superblock.
753  */
754 struct xfs_rtsb {
755 	__be32		rsb_magicnum;	/* magic number == XFS_RTSB_MAGIC */
756 	__le32		rsb_crc;	/* superblock crc */
757 
758 	__be32		rsb_pad;	/* zero */
759 	unsigned char	rsb_fname[XFSLABEL_MAX]; /* file system name */
760 
761 	uuid_t		rsb_uuid;	/* user-visible file system unique id */
762 	uuid_t		rsb_meta_uuid;	/* metadata file system unique id */
763 
764 	/* must be padded to 64 bit alignment */
765 };
766 
767 #define XFS_RTSB_CRC_OFF	offsetof(struct xfs_rtsb, rsb_crc)
768 #define XFS_RTSB_DADDR		((xfs_daddr_t)0) /* daddr in rt section */
769 
770 /*
771  * XFS Timestamps
772  * ==============
773  *
774  * Traditional ondisk inode timestamps consist of signed 32-bit counters for
775  * seconds and nanoseconds; time zero is the Unix epoch, Jan  1 00:00:00 UTC
776  * 1970, which means that the timestamp epoch is the same as the Unix epoch.
777  * Therefore, the ondisk min and max defined here can be used directly to
778  * constrain the incore timestamps on a Unix system.  Note that we actually
779  * encode a __be64 value on disk.
780  *
781  * When the bigtime feature is enabled, ondisk inode timestamps become an
782  * unsigned 64-bit nanoseconds counter.  This means that the bigtime inode
783  * timestamp epoch is the start of the classic timestamp range, which is
784  * Dec 13 20:45:52 UTC 1901.  Because the epochs are not the same, callers
785  * /must/ use the bigtime conversion functions when encoding and decoding raw
786  * timestamps.
787  */
788 typedef __be64 xfs_timestamp_t;
789 
790 /* Legacy timestamp encoding format. */
791 struct xfs_legacy_timestamp {
792 	__be32		t_sec;		/* timestamp seconds */
793 	__be32		t_nsec;		/* timestamp nanoseconds */
794 };
795 
796 /*
797  * Smallest possible ondisk seconds value with traditional timestamps.  This
798  * corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901.
799  */
800 #define XFS_LEGACY_TIME_MIN	((int64_t)S32_MIN)
801 
802 /*
803  * Largest possible ondisk seconds value with traditional timestamps.  This
804  * corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038.
805  */
806 #define XFS_LEGACY_TIME_MAX	((int64_t)S32_MAX)
807 
808 /*
809  * Smallest possible ondisk seconds value with bigtime timestamps.  This
810  * corresponds (after conversion to a Unix timestamp) with the traditional
811  * minimum timestamp of Dec 13 20:45:52 UTC 1901.
812  */
813 #define XFS_BIGTIME_TIME_MIN	((int64_t)0)
814 
815 /*
816  * Largest supported ondisk seconds value with bigtime timestamps.  This
817  * corresponds (after conversion to a Unix timestamp) with an incore timestamp
818  * of Jul  2 20:20:24 UTC 2486.
819  *
820  * We round down the ondisk limit so that the bigtime quota and inode max
821  * timestamps will be the same.
822  */
823 #define XFS_BIGTIME_TIME_MAX	((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL))
824 
825 /*
826  * Bigtime epoch is set exactly to the minimum time value that a traditional
827  * 32-bit timestamp can represent when using the Unix epoch as a reference.
828  * Hence the Unix epoch is at a fixed offset into the supported bigtime
829  * timestamp range.
830  *
831  * The bigtime epoch also matches the minimum value an on-disk 32-bit XFS
832  * timestamp can represent so we will not lose any fidelity in converting
833  * to/from unix and bigtime timestamps.
834  *
835  * The following conversion factor converts a seconds counter from the Unix
836  * epoch to the bigtime epoch.
837  */
838 #define XFS_BIGTIME_EPOCH_OFFSET	(-(int64_t)S32_MIN)
839 
840 /* Convert a timestamp from the Unix epoch to the bigtime epoch. */
xfs_unix_to_bigtime(time64_t unix_seconds)841 static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds)
842 {
843 	return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET;
844 }
845 
846 /* Convert a timestamp from the bigtime epoch to the Unix epoch. */
xfs_bigtime_to_unix(uint64_t ondisk_seconds)847 static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds)
848 {
849 	return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET;
850 }
851 
852 enum xfs_metafile_type {
853 	XFS_METAFILE_UNKNOWN,		/* unknown */
854 	XFS_METAFILE_DIR,		/* metadir directory */
855 	XFS_METAFILE_USRQUOTA,		/* user quota */
856 	XFS_METAFILE_GRPQUOTA,		/* group quota */
857 	XFS_METAFILE_PRJQUOTA,		/* project quota */
858 	XFS_METAFILE_RTBITMAP,		/* rt bitmap */
859 	XFS_METAFILE_RTSUMMARY,		/* rt summary */
860 	XFS_METAFILE_RTRMAP,		/* rt rmap */
861 	XFS_METAFILE_RTREFCOUNT,	/* rt refcount */
862 
863 	XFS_METAFILE_MAX
864 } __packed;
865 
866 #define XFS_METAFILE_TYPE_STR \
867 	{ XFS_METAFILE_UNKNOWN,		"unknown" }, \
868 	{ XFS_METAFILE_DIR,		"dir" }, \
869 	{ XFS_METAFILE_USRQUOTA,	"usrquota" }, \
870 	{ XFS_METAFILE_GRPQUOTA,	"grpquota" }, \
871 	{ XFS_METAFILE_PRJQUOTA,	"prjquota" }, \
872 	{ XFS_METAFILE_RTBITMAP,	"rtbitmap" }, \
873 	{ XFS_METAFILE_RTSUMMARY,	"rtsummary" }, \
874 	{ XFS_METAFILE_RTRMAP,		"rtrmap" }, \
875 	{ XFS_METAFILE_RTREFCOUNT,	"rtrefcount" }
876 
877 /*
878  * On-disk inode structure.
879  *
880  * This is just the header or "dinode core", the inode is expanded to fill a
881  * variable size the leftover area split into a data and an attribute fork.
882  * The format of the data and attribute fork depends on the format of the
883  * inode as indicated by di_format and di_aformat.  To access the data and
884  * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
885  * below.
886  *
887  * There is a very similar struct xfs_log_dinode which matches the layout of
888  * this structure, but is kept in native format instead of big endian.
889  *
890  * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
891  * padding field for v3 inodes.
892  */
893 #define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
894 struct xfs_dinode {
895 	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
896 	__be16		di_mode;	/* mode and type of file */
897 	__u8		di_version;	/* inode version */
898 	__u8		di_format;	/* format of di_c data */
899 	__be16		di_metatype;	/* XFS_METAFILE_*; was di_onlink */
900 	__be32		di_uid;		/* owner's user id */
901 	__be32		di_gid;		/* owner's group id */
902 	__be32		di_nlink;	/* number of links to file */
903 	__be16		di_projid_lo;	/* lower part of owner's project id */
904 	__be16		di_projid_hi;	/* higher part owner's project id */
905 	union {
906 		/* Number of data fork extents if NREXT64 is set */
907 		__be64	di_big_nextents;
908 
909 		/* Padding for V3 inodes without NREXT64 set. */
910 		__be64	di_v3_pad;
911 
912 		/* Padding and inode flush counter for V2 inodes. */
913 		struct {
914 			__u8	di_v2_pad[6];
915 			__be16	di_flushiter;
916 		};
917 	};
918 	xfs_timestamp_t	di_atime;	/* time last accessed */
919 	xfs_timestamp_t	di_mtime;	/* time last modified */
920 	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
921 	__be64		di_size;	/* number of bytes in file */
922 	__be64		di_nblocks;	/* # of direct & btree blocks used */
923 	__be32		di_extsize;	/* basic/minimum extent size for file */
924 	union {
925 		/*
926 		 * For V2 inodes and V3 inodes without NREXT64 set, this
927 		 * is the number of data and attr fork extents.
928 		 */
929 		struct {
930 			__be32	di_nextents;
931 			__be16	di_anextents;
932 		} __packed;
933 
934 		/* Number of attr fork extents if NREXT64 is set. */
935 		struct {
936 			__be32	di_big_anextents;
937 			__be16	di_nrext64_pad;
938 		} __packed;
939 	} __packed;
940 	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
941 	__s8		di_aformat;	/* format of attr fork's data */
942 	__be32		di_dmevmask;	/* DMIG event mask */
943 	__be16		di_dmstate;	/* DMIG state info */
944 	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
945 	__be32		di_gen;		/* generation number */
946 
947 	/* di_next_unlinked is the only non-core field in the old dinode */
948 	__be32		di_next_unlinked;/* agi unlinked list ptr */
949 
950 	/* start of the extended dinode, writable fields */
951 	__le32		di_crc;		/* CRC of the inode */
952 	__be64		di_changecount;	/* number of attribute changes */
953 	__be64		di_lsn;		/* flush sequence */
954 	__be64		di_flags2;	/* more random flags */
955 	__be32		di_cowextsize;	/* basic cow extent size for file */
956 	__u8		di_pad2[12];	/* more padding for future expansion */
957 
958 	/* fields only written to during inode creation */
959 	xfs_timestamp_t	di_crtime;	/* time created */
960 	__be64		di_ino;		/* inode number */
961 	uuid_t		di_uuid;	/* UUID of the filesystem */
962 
963 	/* structure must be padded to 64 bit alignment */
964 };
965 
966 #define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
967 
968 #define DI_MAX_FLUSH 0xffff
969 
970 /*
971  * Size of the core inode on disk.  Version 1 and 2 inodes have
972  * the same size, but version 3 has grown a few additional fields.
973  */
xfs_dinode_size(int version)974 static inline uint xfs_dinode_size(int version)
975 {
976 	if (version == 3)
977 		return sizeof(struct xfs_dinode);
978 	return offsetof(struct xfs_dinode, di_crc);
979 }
980 
981 /*
982  * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
983  * Since the pathconf interface is signed, we use 2^31 - 1 instead.
984  */
985 #define	XFS_MAXLINK		((1U << 31) - 1U)
986 
987 /*
988  * Any file that hits the maximum ondisk link count should be pinned to avoid
989  * a use-after-free situation.
990  */
991 #define	XFS_NLINK_PINNED	(~0U)
992 
993 /*
994  * Values for di_format
995  *
996  * This enum is used in string mapping in xfs_trace.h; please keep the
997  * TRACE_DEFINE_ENUMs for it up to date.
998  */
999 enum xfs_dinode_fmt {
1000 	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
1001 	XFS_DINODE_FMT_LOCAL,		/* bulk data */
1002 	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
1003 	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
1004 	XFS_DINODE_FMT_UUID,		/* added long ago, but never used */
1005 	XFS_DINODE_FMT_META_BTREE,	/* metadata btree */
1006 };
1007 
1008 #define XFS_INODE_FORMAT_STR \
1009 	{ XFS_DINODE_FMT_DEV,		"dev" }, \
1010 	{ XFS_DINODE_FMT_LOCAL,		"local" }, \
1011 	{ XFS_DINODE_FMT_EXTENTS,	"extent" }, \
1012 	{ XFS_DINODE_FMT_BTREE,		"btree" }, \
1013 	{ XFS_DINODE_FMT_UUID,		"uuid" }, \
1014 	{ XFS_DINODE_FMT_META_BTREE,	"meta_btree" }
1015 
1016 /*
1017  * Max values for extnum and aextnum.
1018  *
1019  * The original on-disk extent counts were held in signed fields, resulting in
1020  * maximum extent counts of 2^31 and 2^15 for the data and attr forks
1021  * respectively. Similarly the maximum extent length is limited to 2^21 blocks
1022  * by the 21-bit wide blockcount field of a BMBT extent record.
1023  *
1024  * The newly introduced data fork extent counter can hold a 64-bit value,
1025  * however the maximum number of extents in a file is also limited to 2^54
1026  * extents by the 54-bit wide startoff field of a BMBT extent record.
1027  *
1028  * It is further limited by the maximum supported file size of 2^63
1029  * *bytes*. This leads to a maximum extent count for maximally sized filesystem
1030  * blocks (64kB) of:
1031  *
1032  * 2^63 bytes / 2^16 bytes per block = 2^47 blocks
1033  *
1034  * Rounding up 47 to the nearest multiple of bits-per-byte results in 48. Hence
1035  * 2^48 was chosen as the maximum data fork extent count.
1036  *
1037  * The maximum file size that can be represented by the data fork extent counter
1038  * in the worst case occurs when all extents are 1 block in length and each
1039  * block is 1KB in size.
1040  *
1041  * With XFS_MAX_EXTCNT_DATA_FORK_SMALL representing maximum extent count and
1042  * with 1KB sized blocks, a file can reach upto,
1043  * 1KB * (2^31) = 2TB
1044  *
1045  * This is much larger than the theoretical maximum size of a directory
1046  * i.e. XFS_DIR2_SPACE_SIZE * XFS_DIR2_MAX_SPACES = ~96GB.
1047  *
1048  * Hence, a directory inode can never overflow its data fork extent counter.
1049  */
1050 #define XFS_MAX_EXTCNT_DATA_FORK_LARGE	((xfs_extnum_t)((1ULL << 48) - 1))
1051 #define XFS_MAX_EXTCNT_ATTR_FORK_LARGE	((xfs_extnum_t)((1ULL << 32) - 1))
1052 #define XFS_MAX_EXTCNT_DATA_FORK_SMALL	((xfs_extnum_t)((1ULL << 31) - 1))
1053 #define XFS_MAX_EXTCNT_ATTR_FORK_SMALL	((xfs_extnum_t)((1ULL << 15) - 1))
1054 
1055 /*
1056  * When we upgrade an inode to the large extent counts, the maximum value by
1057  * which the extent count can increase is bound by the change in size of the
1058  * on-disk field. No upgrade operation should ever be adding more than a few
1059  * tens of extents, so if we get a really large value it is a sign of a code bug
1060  * or corruption.
1061  */
1062 #define XFS_MAX_EXTCNT_UPGRADE_NR	\
1063 	min(XFS_MAX_EXTCNT_ATTR_FORK_LARGE - XFS_MAX_EXTCNT_ATTR_FORK_SMALL,	\
1064 	    XFS_MAX_EXTCNT_DATA_FORK_LARGE - XFS_MAX_EXTCNT_DATA_FORK_SMALL)
1065 
1066 /*
1067  * Inode minimum and maximum sizes.
1068  */
1069 #define	XFS_DINODE_MIN_LOG	8
1070 #define	XFS_DINODE_MAX_LOG	11
1071 #define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
1072 #define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
1073 
1074 /*
1075  * Inode size for given fs.
1076  */
1077 #define XFS_DINODE_SIZE(mp) \
1078 	(xfs_has_v3inodes(mp) ? \
1079 		sizeof(struct xfs_dinode) : \
1080 		offsetof(struct xfs_dinode, di_crc))
1081 #define XFS_LITINO(mp) \
1082 	((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(mp))
1083 
1084 /*
1085  * Inode data & attribute fork sizes, per inode.
1086  */
1087 #define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
1088 
1089 #define XFS_DFORK_DSIZE(dip,mp) \
1090 	((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp))
1091 #define XFS_DFORK_ASIZE(dip,mp) \
1092 	((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0)
1093 #define XFS_DFORK_SIZE(dip,mp,w) \
1094 	((w) == XFS_DATA_FORK ? \
1095 		XFS_DFORK_DSIZE(dip, mp) : \
1096 		XFS_DFORK_ASIZE(dip, mp))
1097 
1098 #define XFS_DFORK_MAXEXT(dip, mp, w) \
1099 	(XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
1100 
1101 /*
1102  * Return pointers to the data or attribute forks.
1103  */
1104 #define XFS_DFORK_DPTR(dip) \
1105 	((void *)dip + xfs_dinode_size(dip->di_version))
1106 #define XFS_DFORK_APTR(dip)	\
1107 	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
1108 #define XFS_DFORK_PTR(dip,w)	\
1109 	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
1110 
1111 #define XFS_DFORK_FORMAT(dip,w) \
1112 	((w) == XFS_DATA_FORK ? \
1113 		(dip)->di_format : \
1114 		(dip)->di_aformat)
1115 
1116 /*
1117  * For block and character special files the 32bit dev_t is stored at the
1118  * beginning of the data fork.
1119  */
xfs_dinode_get_rdev(struct xfs_dinode * dip)1120 static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1121 {
1122 	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1123 }
1124 
xfs_dinode_put_rdev(struct xfs_dinode * dip,xfs_dev_t rdev)1125 static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1126 {
1127 	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1128 }
1129 
1130 /*
1131  * Values for di_flags
1132  */
1133 #define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
1134 #define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
1135 #define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
1136 #define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
1137 #define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
1138 #define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
1139 #define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
1140 #define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
1141 #define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
1142 #define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
1143 #define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
1144 #define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
1145 #define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
1146 #define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
1147 #define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
1148 /* Do not use bit 15, di_flags is legacy and unchanging now */
1149 
1150 #define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
1151 #define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
1152 #define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
1153 #define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1154 #define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
1155 #define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
1156 #define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
1157 #define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
1158 #define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
1159 #define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1160 #define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1161 #define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
1162 #define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1163 #define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
1164 #define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
1165 
1166 #define XFS_DIFLAG_ANY \
1167 	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1168 	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1169 	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1170 	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1171 	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1172 
1173 /*
1174  * Values for di_flags2 These start by being exposed to userspace in the upper
1175  * 16 bits of the XFS_XFLAG_s range.
1176  */
1177 /* use DAX for this inode */
1178 #define XFS_DIFLAG2_DAX_BIT		0
1179 
1180 /* file's blocks may be shared */
1181 #define XFS_DIFLAG2_REFLINK_BIT		1
1182 
1183 /* copy on write extent size hint */
1184 #define XFS_DIFLAG2_COWEXTSIZE_BIT	2
1185 
1186 /* big timestamps */
1187 #define XFS_DIFLAG2_BIGTIME_BIT		3
1188 
1189 /* large extent counters */
1190 #define XFS_DIFLAG2_NREXT64_BIT		4
1191 
1192 /*
1193  * The inode contains filesystem metadata and can be found through the metadata
1194  * directory tree.  Metadata inodes must satisfy the following constraints:
1195  *
1196  * - V5 filesystem (and ftype) are enabled;
1197  * - The only valid modes are regular files and directories;
1198  * - The access bits must be zero;
1199  * - DMAPI event and state masks are zero;
1200  * - The user and group IDs must be zero;
1201  * - The project ID can be used as a u32 annotation;
1202  * - The immutable, sync, noatime, nodump, nodefrag flags must be set.
1203  * - The dax flag must not be set.
1204  * - Directories must have nosymlinks set.
1205  *
1206  * These requirements are chosen defensively to minimize the ability of
1207  * userspace to read or modify the contents, should a metadata file ever
1208  * escape to userspace.
1209  *
1210  * There are further constraints on the directory tree itself:
1211  *
1212  * - Metadata inodes must never be resolvable through the root directory;
1213  * - They must never be accessed by userspace;
1214  * - Metadata directory entries must have correct ftype.
1215  *
1216  * Superblock-rooted metadata files must have the METADATA iflag set even
1217  * though they do not have a parent directory.
1218  */
1219 #define XFS_DIFLAG2_METADATA_BIT	5
1220 
1221 #define XFS_DIFLAG2_DAX		(1ULL << XFS_DIFLAG2_DAX_BIT)
1222 #define XFS_DIFLAG2_REFLINK	(1ULL << XFS_DIFLAG2_REFLINK_BIT)
1223 #define XFS_DIFLAG2_COWEXTSIZE	(1ULL << XFS_DIFLAG2_COWEXTSIZE_BIT)
1224 #define XFS_DIFLAG2_BIGTIME	(1ULL << XFS_DIFLAG2_BIGTIME_BIT)
1225 #define XFS_DIFLAG2_NREXT64	(1ULL << XFS_DIFLAG2_NREXT64_BIT)
1226 #define XFS_DIFLAG2_METADATA	(1ULL << XFS_DIFLAG2_METADATA_BIT)
1227 
1228 #define XFS_DIFLAG2_ANY \
1229 	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \
1230 	 XFS_DIFLAG2_BIGTIME | XFS_DIFLAG2_NREXT64 | XFS_DIFLAG2_METADATA)
1231 
xfs_dinode_has_bigtime(const struct xfs_dinode * dip)1232 static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip)
1233 {
1234 	return dip->di_version >= 3 &&
1235 	       (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME));
1236 }
1237 
xfs_dinode_has_large_extent_counts(const struct xfs_dinode * dip)1238 static inline bool xfs_dinode_has_large_extent_counts(
1239 	const struct xfs_dinode *dip)
1240 {
1241 	return dip->di_version >= 3 &&
1242 	       (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_NREXT64));
1243 }
1244 
xfs_dinode_is_metadir(const struct xfs_dinode * dip)1245 static inline bool xfs_dinode_is_metadir(const struct xfs_dinode *dip)
1246 {
1247 	return dip->di_version >= 3 &&
1248 	       (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_METADATA));
1249 }
1250 
1251 /*
1252  * Inode number format:
1253  * low inopblog bits - offset in block
1254  * next agblklog bits - block number in ag
1255  * next agno_log bits - ag number
1256  * high agno_log-agblklog-inopblog bits - 0
1257  */
1258 #define	XFS_INO_MASK(k)			(uint32_t)((1ULL << (k)) - 1)
1259 #define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1260 #define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1261 #define	XFS_INO_AGINO_BITS(mp)		((mp)->m_ino_geo.agino_log)
1262 #define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1263 #define	XFS_INO_BITS(mp)		\
1264 	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1265 #define	XFS_INO_TO_AGNO(mp,i)		\
1266 	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1267 #define	XFS_INO_TO_AGINO(mp,i)		\
1268 	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1269 #define	XFS_INO_TO_AGBNO(mp,i)		\
1270 	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1271 		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1272 #define	XFS_INO_TO_OFFSET(mp,i)		\
1273 	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1274 #define	XFS_INO_TO_FSB(mp,i)		\
1275 	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1276 #define	XFS_AGINO_TO_INO(mp,a,i)	\
1277 	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1278 #define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1279 #define	XFS_AGINO_TO_OFFSET(mp,i)	\
1280 	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1281 #define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1282 	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1283 #define	XFS_FSB_TO_INO(mp, b)	((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1284 #define	XFS_AGB_TO_AGINO(mp, b)	((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1285 
1286 #define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1287 #define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1288 
1289 /*
1290  * RealTime Device format definitions
1291  */
1292 
1293 /* Min and max rt extent sizes, specified in bytes */
1294 #define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1295 #define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1296 #define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1297 
1298 /*
1299  * RT bit manipulation macros.
1300  */
1301 #define XFS_RTBITMAP_MAGIC	0x424D505A	/* BMPZ */
1302 #define XFS_RTSUMMARY_MAGIC	0x53554D59	/* SUMY */
1303 
1304 struct xfs_rtbuf_blkinfo {
1305 	__be32		rt_magic;	/* validity check on block */
1306 	__be32		rt_crc;		/* CRC of block */
1307 	__be64		rt_owner;	/* inode that owns the block */
1308 	__be64		rt_blkno;	/* first block of the buffer */
1309 	__be64		rt_lsn;		/* sequence number of last write */
1310 	uuid_t		rt_uuid;	/* filesystem we belong to */
1311 };
1312 
1313 #define XFS_RTBUF_CRC_OFF \
1314 	offsetof(struct xfs_rtbuf_blkinfo, rt_crc)
1315 
1316 /*
1317  * Dquot and dquot block format definitions
1318  */
1319 #define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1320 #define XFS_DQUOT_VERSION	(uint8_t)0x01	/* latest version number */
1321 
1322 #define XFS_DQTYPE_USER		(1u << 0)	/* user dquot record */
1323 #define XFS_DQTYPE_PROJ		(1u << 1)	/* project dquot record */
1324 #define XFS_DQTYPE_GROUP	(1u << 2)	/* group dquot record */
1325 #define XFS_DQTYPE_BIGTIME	(1u << 7)	/* large expiry timestamps */
1326 
1327 /* bitmask to determine if this is a user/group/project dquot */
1328 #define XFS_DQTYPE_REC_MASK	(XFS_DQTYPE_USER | \
1329 				 XFS_DQTYPE_PROJ | \
1330 				 XFS_DQTYPE_GROUP)
1331 
1332 #define XFS_DQTYPE_ANY		(XFS_DQTYPE_REC_MASK | \
1333 				 XFS_DQTYPE_BIGTIME)
1334 
1335 /*
1336  * XFS Quota Timers
1337  * ================
1338  *
1339  * Traditional quota grace period expiration timers are an unsigned 32-bit
1340  * seconds counter; time zero is the Unix epoch, Jan  1 00:00:01 UTC 1970.
1341  * Note that an expiration value of zero means that the quota limit has not
1342  * been reached, and therefore no expiration has been set.  Therefore, the
1343  * ondisk min and max defined here can be used directly to constrain the incore
1344  * quota expiration timestamps on a Unix system.
1345  *
1346  * When bigtime is enabled, we trade two bits of precision to expand the
1347  * expiration timeout range to match that of big inode timestamps.  The min and
1348  * max recorded here are the on-disk limits, not a Unix timestamp.
1349  *
1350  * The grace period for each quota type is stored in the root dquot (id = 0)
1351  * and is applied to a non-root dquot when it exceeds the soft or hard limits.
1352  * The length of quota grace periods are unsigned 32-bit quantities measured in
1353  * units of seconds.  A value of zero means to use the default period.
1354  */
1355 
1356 /*
1357  * Smallest possible ondisk quota expiration value with traditional timestamps.
1358  * This corresponds exactly with the incore expiration Jan  1 00:00:01 UTC 1970.
1359  */
1360 #define XFS_DQ_LEGACY_EXPIRY_MIN	((int64_t)1)
1361 
1362 /*
1363  * Largest possible ondisk quota expiration value with traditional timestamps.
1364  * This corresponds exactly with the incore expiration Feb  7 06:28:15 UTC 2106.
1365  */
1366 #define XFS_DQ_LEGACY_EXPIRY_MAX	((int64_t)U32_MAX)
1367 
1368 /*
1369  * Smallest possible ondisk quota expiration value with bigtime timestamps.
1370  * This corresponds (after conversion to a Unix timestamp) with the incore
1371  * expiration of Jan  1 00:00:04 UTC 1970.
1372  */
1373 #define XFS_DQ_BIGTIME_EXPIRY_MIN	(XFS_DQ_LEGACY_EXPIRY_MIN)
1374 
1375 /*
1376  * Largest supported ondisk quota expiration value with bigtime timestamps.
1377  * This corresponds (after conversion to a Unix timestamp) with an incore
1378  * expiration of Jul  2 20:20:24 UTC 2486.
1379  *
1380  * The ondisk field supports values up to -1U, which corresponds to an incore
1381  * expiration in 2514.  This is beyond the maximum the bigtime inode timestamp,
1382  * so we cap the maximum bigtime quota expiration to the max inode timestamp.
1383  */
1384 #define XFS_DQ_BIGTIME_EXPIRY_MAX	((int64_t)4074815106U)
1385 
1386 /*
1387  * The following conversion factors assist in converting a quota expiration
1388  * timestamp between the incore and ondisk formats.
1389  */
1390 #define XFS_DQ_BIGTIME_SHIFT	(2)
1391 #define XFS_DQ_BIGTIME_SLACK	((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1)
1392 
1393 /* Convert an incore quota expiration timestamp to an ondisk bigtime value. */
xfs_dq_unix_to_bigtime(time64_t unix_seconds)1394 static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds)
1395 {
1396 	/*
1397 	 * Round the expiration timestamp up to the nearest bigtime timestamp
1398 	 * that we can store, to give users the most time to fix problems.
1399 	 */
1400 	return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >>
1401 			XFS_DQ_BIGTIME_SHIFT;
1402 }
1403 
1404 /* Convert an ondisk bigtime quota expiration value to an incore timestamp. */
xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)1405 static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)
1406 {
1407 	return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT;
1408 }
1409 
1410 /*
1411  * Default quota grace periods, ranging from zero (use the compiled defaults)
1412  * to ~136 years.  These are applied to a non-root dquot that has exceeded
1413  * either limit.
1414  */
1415 #define XFS_DQ_GRACE_MIN		((int64_t)0)
1416 #define XFS_DQ_GRACE_MAX		((int64_t)U32_MAX)
1417 
1418 /* Maximum id value for a quota record */
1419 #define XFS_DQ_ID_MAX			(U32_MAX)
1420 
1421 /*
1422  * This is the main portion of the on-disk representation of quota information
1423  * for a user.  We pad this with some more expansion room to construct the on
1424  * disk structure.
1425  */
1426 struct xfs_disk_dquot {
1427 	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1428 	__u8		d_version;	/* dquot version */
1429 	__u8		d_type;		/* XFS_DQTYPE_USER/PROJ/GROUP */
1430 	__be32		d_id;		/* user,project,group id */
1431 	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1432 	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1433 	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1434 	__be64		d_ino_softlimit;/* preferred inode limit */
1435 	__be64		d_bcount;	/* disk blocks owned by the user */
1436 	__be64		d_icount;	/* inodes owned by the user */
1437 	__be32		d_itimer;	/* zero if within inode limits if not,
1438 					   this is when we refuse service */
1439 	__be32		d_btimer;	/* similar to above; for disk blocks */
1440 	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1441 	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1442 	__be32		d_pad0;		/* 64 bit align */
1443 	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1444 	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1445 	__be64		d_rtbcount;	/* realtime blocks owned */
1446 	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1447 	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1448 	__be16		d_pad;
1449 };
1450 
1451 /*
1452  * This is what goes on disk. This is separated from the xfs_disk_dquot because
1453  * carrying the unnecessary padding would be a waste of memory.
1454  */
1455 struct xfs_dqblk {
1456 	struct xfs_disk_dquot	dd_diskdq; /* portion living incore as well */
1457 	char			dd_fill[4];/* filling for posterity */
1458 
1459 	/*
1460 	 * These two are only present on filesystems with the CRC bits set.
1461 	 */
1462 	__be32		  dd_crc;	/* checksum */
1463 	__be64		  dd_lsn;	/* last modification in log */
1464 	uuid_t		  dd_uuid;	/* location information */
1465 };
1466 
1467 #define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1468 
1469 /*
1470  * This defines the unit of allocation of dquots.
1471  *
1472  * Currently, it is just one file system block, and a 4K blk contains 30
1473  * (136 * 30 = 4080) dquots. It's probably not worth trying to make
1474  * this more dynamic.
1475  *
1476  * However, if this number is changed, we have to make sure that we don't
1477  * implicitly assume that we do allocations in chunks of a single filesystem
1478  * block in the dquot/xqm code.
1479  *
1480  * This is part of the ondisk format because the structure size is not a power
1481  * of two, which leaves slack at the end of the disk block.
1482  */
1483 #define XFS_DQUOT_CLUSTER_SIZE_FSB	(xfs_filblks_t)1
1484 
1485 /*
1486  * Remote symlink format and access functions.
1487  */
1488 #define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1489 
1490 struct xfs_dsymlink_hdr {
1491 	__be32	sl_magic;
1492 	__be32	sl_offset;
1493 	__be32	sl_bytes;
1494 	__be32	sl_crc;
1495 	uuid_t	sl_uuid;
1496 	__be64	sl_owner;
1497 	__be64	sl_blkno;
1498 	__be64	sl_lsn;
1499 };
1500 
1501 #define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1502 
1503 #define XFS_SYMLINK_MAXLEN	1024
1504 /*
1505  * The maximum pathlen is 1024 bytes. Since the minimum file system
1506  * blocksize is 512 bytes, we can get a max of 3 extents back from
1507  * bmapi when crc headers are taken into account.
1508  */
1509 #define XFS_SYMLINK_MAPS 3
1510 
1511 #define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1512 	((bufsize) - (xfs_has_crc((mp)) ? \
1513 			sizeof(struct xfs_dsymlink_hdr) : 0))
1514 
1515 
1516 /*
1517  * Allocation Btree format definitions
1518  *
1519  * There are two on-disk btrees, one sorted by blockno and one sorted
1520  * by blockcount and blockno.  All blocks look the same to make the code
1521  * simpler; if we have time later, we'll make the optimizations.
1522  */
1523 #define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1524 #define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1525 #define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1526 #define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1527 
1528 /*
1529  * Data record/key structure
1530  */
1531 typedef struct xfs_alloc_rec {
1532 	__be32		ar_startblock;	/* starting block number */
1533 	__be32		ar_blockcount;	/* count of free blocks */
1534 } xfs_alloc_rec_t, xfs_alloc_key_t;
1535 
1536 typedef struct xfs_alloc_rec_incore {
1537 	xfs_agblock_t	ar_startblock;	/* starting block number */
1538 	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1539 } xfs_alloc_rec_incore_t;
1540 
1541 /* btree pointer type */
1542 typedef __be32 xfs_alloc_ptr_t;
1543 
1544 /*
1545  * Block numbers in the AG:
1546  * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1547  */
1548 #define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1549 #define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1550 
1551 
1552 /*
1553  * Inode Allocation Btree format definitions
1554  *
1555  * There is a btree for the inode map per allocation group.
1556  */
1557 #define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1558 #define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1559 #define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1560 #define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1561 
1562 typedef uint64_t	xfs_inofree_t;
1563 #define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1564 #define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1565 #define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1566 #define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1567 
1568 #define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1569 #define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(uint16_t))
1570 #define XFS_INODES_PER_HOLEMASK_BIT	\
1571 	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1572 
xfs_inobt_maskn(int i,int n)1573 static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1574 {
1575 	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1576 }
1577 
1578 /*
1579  * The on-disk inode record structure has two formats. The original "full"
1580  * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1581  * and replaces the 3 high-order freecount bytes wth the holemask and inode
1582  * count.
1583  *
1584  * The holemask of the sparse record format allows an inode chunk to have holes
1585  * that refer to blocks not owned by the inode record. This facilitates inode
1586  * allocation in the event of severe free space fragmentation.
1587  */
1588 typedef struct xfs_inobt_rec {
1589 	__be32		ir_startino;	/* starting inode number */
1590 	union {
1591 		struct {
1592 			__be32	ir_freecount;	/* count of free inodes */
1593 		} f;
1594 		struct {
1595 			__be16	ir_holemask;/* hole mask for sparse chunks */
1596 			__u8	ir_count;	/* total inode count */
1597 			__u8	ir_freecount;	/* count of free inodes */
1598 		} sp;
1599 	} ir_u;
1600 	__be64		ir_free;	/* free inode mask */
1601 } xfs_inobt_rec_t;
1602 
1603 typedef struct xfs_inobt_rec_incore {
1604 	xfs_agino_t	ir_startino;	/* starting inode number */
1605 	uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1606 	uint8_t		ir_count;	/* total inode count */
1607 	uint8_t		ir_freecount;	/* count of free inodes (set bits) */
1608 	xfs_inofree_t	ir_free;	/* free inode mask */
1609 } xfs_inobt_rec_incore_t;
1610 
xfs_inobt_issparse(uint16_t holemask)1611 static inline bool xfs_inobt_issparse(uint16_t holemask)
1612 {
1613 	/* non-zero holemask represents a sparse rec. */
1614 	return holemask;
1615 }
1616 
1617 /*
1618  * Key structure
1619  */
1620 typedef struct xfs_inobt_key {
1621 	__be32		ir_startino;	/* starting inode number */
1622 } xfs_inobt_key_t;
1623 
1624 /* btree pointer type */
1625 typedef __be32 xfs_inobt_ptr_t;
1626 
1627 /*
1628  * block numbers in the AG.
1629  */
1630 #define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1631 #define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1632 
1633 /*
1634  * Reverse mapping btree format definitions
1635  *
1636  * There is a btree for the reverse map per allocation group
1637  */
1638 #define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1639 
1640 /*
1641  * Ownership info for an extent.  This is used to create reverse-mapping
1642  * entries.
1643  */
1644 #define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1645 #define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1646 struct xfs_owner_info {
1647 	uint64_t		oi_owner;
1648 	xfs_fileoff_t		oi_offset;
1649 	unsigned int		oi_flags;
1650 };
1651 
1652 /*
1653  * Special owner types.
1654  *
1655  * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1656  * to tell us we have a special owner value. We use these for static metadata
1657  * allocated at mkfs/growfs time, as well as for freespace management metadata.
1658  */
1659 #define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1660 #define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1661 #define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1662 #define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1663 #define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1664 #define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1665 #define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1666 #define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1667 #define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1668 #define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1669 
1670 #define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1671 
1672 /*
1673  * Data record structure
1674  */
1675 struct xfs_rmap_rec {
1676 	__be32		rm_startblock;	/* extent start block */
1677 	__be32		rm_blockcount;	/* extent length */
1678 	__be64		rm_owner;	/* extent owner */
1679 	__be64		rm_offset;	/* offset within the owner */
1680 };
1681 
1682 /*
1683  * rmap btree record
1684  *  rm_offset:63 is the attribute fork flag
1685  *  rm_offset:62 is the bmbt block flag
1686  *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1687  *  rm_offset:54-60 aren't used and should be zero
1688  *  rm_offset:0-53 is the block offset within the inode
1689  */
1690 #define XFS_RMAP_OFF_ATTR_FORK	((uint64_t)1ULL << 63)
1691 #define XFS_RMAP_OFF_BMBT_BLOCK	((uint64_t)1ULL << 62)
1692 #define XFS_RMAP_OFF_UNWRITTEN	((uint64_t)1ULL << 61)
1693 
1694 #define XFS_RMAP_LEN_MAX	((uint32_t)~0U)
1695 #define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1696 				 XFS_RMAP_OFF_BMBT_BLOCK | \
1697 				 XFS_RMAP_OFF_UNWRITTEN)
1698 #define XFS_RMAP_OFF_MASK	((uint64_t)0x3FFFFFFFFFFFFFULL)
1699 
1700 #define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1701 
1702 #define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1703 #define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1704 #define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1705 
1706 #define RMAPBT_STARTBLOCK_BITLEN	32
1707 #define RMAPBT_BLOCKCOUNT_BITLEN	32
1708 #define RMAPBT_OWNER_BITLEN		64
1709 #define RMAPBT_ATTRFLAG_BITLEN		1
1710 #define RMAPBT_BMBTFLAG_BITLEN		1
1711 #define RMAPBT_EXNTFLAG_BITLEN		1
1712 #define RMAPBT_UNUSED_OFFSET_BITLEN	7
1713 #define RMAPBT_OFFSET_BITLEN		54
1714 
1715 /*
1716  * Key structure
1717  *
1718  * We don't use the length for lookups
1719  */
1720 struct xfs_rmap_key {
1721 	__be32		rm_startblock;	/* extent start block */
1722 	__be64		rm_owner;	/* extent owner */
1723 	__be64		rm_offset;	/* offset within the owner */
1724 } __attribute__((packed));
1725 
1726 /* btree pointer type */
1727 typedef __be32 xfs_rmap_ptr_t;
1728 
1729 #define	XFS_RMAP_BLOCK(mp) \
1730 	(xfs_has_finobt(((mp))) ? \
1731 	 XFS_FIBT_BLOCK(mp) + 1 : \
1732 	 XFS_IBT_BLOCK(mp) + 1)
1733 
1734 /*
1735  * Realtime Reverse mapping btree format definitions
1736  *
1737  * This is a btree for reverse mapping records for realtime volumes
1738  */
1739 #define	XFS_RTRMAP_CRC_MAGIC	0x4d415052	/* 'MAPR' */
1740 
1741 /*
1742  * rtrmap root header, on-disk form only.
1743  */
1744 struct xfs_rtrmap_root {
1745 	__be16		bb_level;	/* 0 is a leaf */
1746 	__be16		bb_numrecs;	/* current # of data records */
1747 };
1748 
1749 /* inode-based btree pointer type */
1750 typedef __be64 xfs_rtrmap_ptr_t;
1751 
1752 /*
1753  * Reference Count Btree format definitions
1754  *
1755  */
1756 #define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1757 
1758 unsigned int xfs_refc_block(struct xfs_mount *mp);
1759 
1760 /*
1761  * Data record/key structure
1762  *
1763  * Each record associates a range of physical blocks (starting at
1764  * rc_startblock and ending rc_blockcount blocks later) with a reference
1765  * count (rc_refcount).  Extents that are being used to stage a copy on
1766  * write (CoW) operation are recorded in the refcount btree with a
1767  * refcount of 1.  All other records must have a refcount > 1 and must
1768  * track an extent mapped only by file data forks.
1769  *
1770  * Extents with a single owner (attributes, metadata, non-shared file
1771  * data) are not tracked here.  Free space is also not tracked here.
1772  * This is consistent with pre-reflink XFS.
1773  */
1774 
1775 /*
1776  * Extents that are being used to stage a copy on write are stored
1777  * in the refcount btree with a refcount of 1 and the upper bit set
1778  * on the startblock.  This speeds up mount time deletion of stale
1779  * staging extents because they're all at the right side of the tree.
1780  */
1781 #define XFS_REFC_COWFLAG		(1U << 31)
1782 #define REFCNTBT_COWFLAG_BITLEN		1
1783 #define REFCNTBT_AGBLOCK_BITLEN		31
1784 
1785 struct xfs_refcount_rec {
1786 	__be32		rc_startblock;	/* starting block number */
1787 	__be32		rc_blockcount;	/* count of blocks */
1788 	__be32		rc_refcount;	/* number of inodes linked here */
1789 };
1790 
1791 struct xfs_refcount_key {
1792 	__be32		rc_startblock;	/* starting block number */
1793 };
1794 
1795 #define XFS_REFC_REFCOUNT_MAX	((xfs_nlink_t)~0U)
1796 #define XFS_REFC_LEN_MAX	((xfs_extlen_t)~0U)
1797 
1798 /* btree pointer type */
1799 typedef __be32 xfs_refcount_ptr_t;
1800 
1801 /*
1802  * Realtime Reference Count btree format definitions
1803  *
1804  * This is a btree for reference count records for realtime volumes
1805  */
1806 #define	XFS_RTREFC_CRC_MAGIC	0x52434e54	/* 'RCNT' */
1807 
1808 /*
1809  * rt refcount root header, on-disk form only.
1810  */
1811 struct xfs_rtrefcount_root {
1812 	__be16		bb_level;	/* 0 is a leaf */
1813 	__be16		bb_numrecs;	/* current # of data records */
1814 };
1815 
1816 /* inode-rooted btree pointer type */
1817 typedef __be64 xfs_rtrefcount_ptr_t;
1818 
1819 /*
1820  * BMAP Btree format definitions
1821  *
1822  * This includes both the root block definition that sits inside an inode fork
1823  * and the record/pointer formats for the leaf/node in the blocks.
1824  */
1825 #define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1826 #define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1827 
1828 /*
1829  * Bmap root header, on-disk form only.
1830  */
1831 typedef struct xfs_bmdr_block {
1832 	__be16		bb_level;	/* 0 is a leaf */
1833 	__be16		bb_numrecs;	/* current # of data records */
1834 } xfs_bmdr_block_t;
1835 
1836 /*
1837  * Bmap btree record and extent descriptor.
1838  *  l0:63 is an extent flag (value 1 indicates non-normal).
1839  *  l0:9-62 are startoff.
1840  *  l0:0-8 and l1:21-63 are startblock.
1841  *  l1:0-20 are blockcount.
1842  */
1843 #define BMBT_EXNTFLAG_BITLEN	1
1844 #define BMBT_STARTOFF_BITLEN	54
1845 #define BMBT_STARTBLOCK_BITLEN	52
1846 #define BMBT_BLOCKCOUNT_BITLEN	21
1847 
1848 #define BMBT_STARTOFF_MASK	((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1849 #define BMBT_BLOCKCOUNT_MASK	((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1)
1850 
1851 #define XFS_MAX_BMBT_EXTLEN	((xfs_extlen_t)(BMBT_BLOCKCOUNT_MASK))
1852 
1853 /*
1854  * bmbt records have a file offset (block) field that is 54 bits wide, so this
1855  * is the largest xfs_fileoff_t that we ever expect to see.
1856  */
1857 #define XFS_MAX_FILEOFF		(BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK)
1858 
1859 typedef struct xfs_bmbt_rec {
1860 	__be64			l0, l1;
1861 } xfs_bmbt_rec_t;
1862 
1863 typedef uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1864 typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1865 
1866 /*
1867  * Values and macros for delayed-allocation startblock fields.
1868  */
1869 #define STARTBLOCKVALBITS	17
1870 #define STARTBLOCKMASKBITS	(15 + 20)
1871 #define STARTBLOCKMASK		\
1872 	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1873 
isnullstartblock(xfs_fsblock_t x)1874 static inline int isnullstartblock(xfs_fsblock_t x)
1875 {
1876 	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1877 }
1878 
nullstartblock(int k)1879 static inline xfs_fsblock_t nullstartblock(int k)
1880 {
1881 	ASSERT(k < (1 << STARTBLOCKVALBITS));
1882 	return STARTBLOCKMASK | (k);
1883 }
1884 
startblockval(xfs_fsblock_t x)1885 static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1886 {
1887 	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1888 }
1889 
1890 /*
1891  * Key structure for non-leaf levels of the tree.
1892  */
1893 typedef struct xfs_bmbt_key {
1894 	__be64		br_startoff;	/* starting file offset */
1895 } xfs_bmbt_key_t, xfs_bmdr_key_t;
1896 
1897 /* btree pointer type */
1898 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1899 
1900 
1901 /*
1902  * Generic Btree block format definitions
1903  *
1904  * This is a combination of the actual format used on disk for short and long
1905  * format btrees.  The first three fields are shared by both format, but the
1906  * pointers are different and should be used with care.
1907  *
1908  * To get the size of the actual short or long form headers please use the size
1909  * macros below.  Never use sizeof(xfs_btree_block).
1910  *
1911  * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1912  * with the crc feature bit, and all accesses to them must be conditional on
1913  * that flag.
1914  */
1915 /* short form block header */
1916 struct xfs_btree_block_shdr {
1917 	__be32		bb_leftsib;
1918 	__be32		bb_rightsib;
1919 
1920 	__be64		bb_blkno;
1921 	__be64		bb_lsn;
1922 	uuid_t		bb_uuid;
1923 	__be32		bb_owner;
1924 	__le32		bb_crc;
1925 };
1926 
1927 /* long form block header */
1928 struct xfs_btree_block_lhdr {
1929 	__be64		bb_leftsib;
1930 	__be64		bb_rightsib;
1931 
1932 	__be64		bb_blkno;
1933 	__be64		bb_lsn;
1934 	uuid_t		bb_uuid;
1935 	__be64		bb_owner;
1936 	__le32		bb_crc;
1937 	__be32		bb_pad; /* padding for alignment */
1938 };
1939 
1940 struct xfs_btree_block {
1941 	__be32		bb_magic;	/* magic number for block type */
1942 	__be16		bb_level;	/* 0 is a leaf */
1943 	__be16		bb_numrecs;	/* current # of data records */
1944 	union {
1945 		struct xfs_btree_block_shdr s;
1946 		struct xfs_btree_block_lhdr l;
1947 	} bb_u;				/* rest */
1948 };
1949 
1950 /* size of a short form block */
1951 #define XFS_BTREE_SBLOCK_LEN \
1952 	(offsetof(struct xfs_btree_block, bb_u) + \
1953 	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1954 /* size of a long form block */
1955 #define XFS_BTREE_LBLOCK_LEN \
1956 	(offsetof(struct xfs_btree_block, bb_u) + \
1957 	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1958 
1959 /* sizes of CRC enabled btree blocks */
1960 #define XFS_BTREE_SBLOCK_CRC_LEN \
1961 	(offsetof(struct xfs_btree_block, bb_u) + \
1962 	 sizeof(struct xfs_btree_block_shdr))
1963 #define XFS_BTREE_LBLOCK_CRC_LEN \
1964 	(offsetof(struct xfs_btree_block, bb_u) + \
1965 	 sizeof(struct xfs_btree_block_lhdr))
1966 
1967 #define XFS_BTREE_SBLOCK_CRC_OFF \
1968 	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1969 #define XFS_BTREE_LBLOCK_CRC_OFF \
1970 	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1971 
1972 /*
1973  * On-disk XFS access control list structure.
1974  */
1975 struct xfs_acl_entry {
1976 	__be32	ae_tag;
1977 	__be32	ae_id;
1978 	__be16	ae_perm;
1979 	__be16	ae_pad;		/* fill the implicit hole in the structure */
1980 };
1981 
1982 struct xfs_acl {
1983 	__be32			acl_cnt;
1984 	struct xfs_acl_entry	acl_entry[];
1985 };
1986 
1987 /*
1988  * The number of ACL entries allowed is defined by the on-disk format.
1989  * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1990  * limited only by the maximum size of the xattr that stores the information.
1991  */
1992 #define XFS_ACL_MAX_ENTRIES(mp)	\
1993 	(xfs_has_crc(mp) \
1994 		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1995 						sizeof(struct xfs_acl_entry) \
1996 		: 25)
1997 
1998 #define XFS_ACL_SIZE(cnt) \
1999 	(sizeof(struct xfs_acl) + \
2000 		sizeof(struct xfs_acl_entry) * cnt)
2001 
2002 #define XFS_ACL_MAX_SIZE(mp) \
2003 	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
2004 
2005 
2006 /* On-disk XFS extended attribute names */
2007 #define SGI_ACL_FILE		"SGI_ACL_FILE"
2008 #define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
2009 #define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
2010 #define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
2011 
2012 #endif /* __XFS_FORMAT_H__ */
2013