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
18struct xfs_mount;
19struct xfs_trans;
20struct xfs_inode;
21struct xfs_buf;
22struct xfs_ifork;
23
24/*
25 * Super block
26 * Fits into a sector-sized buffer at address 0 of each allocation group.
27 * Only the first of these is ever updated except during growfs.
28 */
29#define	XFS_SB_MAGIC		0x58465342	/* 'XFSB' */
30#define	XFS_SB_VERSION_1	1		/* 5.3, 6.0.1, 6.1 */
31#define	XFS_SB_VERSION_2	2		/* 6.2 - attributes */
32#define	XFS_SB_VERSION_3	3		/* 6.2 - new inode version */
33#define	XFS_SB_VERSION_4	4		/* 6.2+ - bitmask version */
34#define	XFS_SB_VERSION_5	5		/* CRC enabled filesystem */
35#define	XFS_SB_VERSION_NUMBITS		0x000f
36#define	XFS_SB_VERSION_ALLFBITS		0xfff0
37#define	XFS_SB_VERSION_ATTRBIT		0x0010
38#define	XFS_SB_VERSION_NLINKBIT		0x0020
39#define	XFS_SB_VERSION_QUOTABIT		0x0040
40#define	XFS_SB_VERSION_ALIGNBIT		0x0080
41#define	XFS_SB_VERSION_DALIGNBIT	0x0100
42#define	XFS_SB_VERSION_SHAREDBIT	0x0200
43#define XFS_SB_VERSION_LOGV2BIT		0x0400
44#define XFS_SB_VERSION_SECTORBIT	0x0800
45#define	XFS_SB_VERSION_EXTFLGBIT	0x1000
46#define	XFS_SB_VERSION_DIRV2BIT		0x2000
47#define	XFS_SB_VERSION_BORGBIT		0x4000	/* ASCII only case-insens. */
48#define	XFS_SB_VERSION_MOREBITSBIT	0x8000
49
50/*
51 * The size of a single extended attribute on disk is limited by
52 * the size of index values within the attribute entries themselves.
53 * These are be16 fields, so we can only support attribute data
54 * sizes up to 2^16 bytes in length.
55 */
56#define XFS_XATTR_SIZE_MAX (1 << 16)
57
58/*
59 * Supported feature bit list is just all bits in the versionnum field because
60 * we've used them all up and understand them all. Except, of course, for the
61 * shared superblock bit, which nobody knows what it does and so is unsupported.
62 */
63#define	XFS_SB_VERSION_OKBITS		\
64	((XFS_SB_VERSION_NUMBITS | XFS_SB_VERSION_ALLFBITS) & \
65		~XFS_SB_VERSION_SHAREDBIT)
66
67/*
68 * There are two words to hold XFS "feature" bits: the original
69 * word, sb_versionnum, and sb_features2.  Whenever a bit is set in
70 * sb_features2, the feature bit XFS_SB_VERSION_MOREBITSBIT must be set.
71 *
72 * These defines represent bits in sb_features2.
73 */
74#define XFS_SB_VERSION2_RESERVED1BIT	0x00000001
75#define XFS_SB_VERSION2_LAZYSBCOUNTBIT	0x00000002	/* Superblk counters */
76#define XFS_SB_VERSION2_RESERVED4BIT	0x00000004
77#define XFS_SB_VERSION2_ATTR2BIT	0x00000008	/* Inline attr rework */
78#define XFS_SB_VERSION2_PARENTBIT	0x00000010	/* parent pointers */
79#define XFS_SB_VERSION2_PROJID32BIT	0x00000080	/* 32 bit project id */
80#define XFS_SB_VERSION2_CRCBIT		0x00000100	/* metadata CRCs */
81#define XFS_SB_VERSION2_FTYPE		0x00000200	/* inode type in dir */
82
83#define	XFS_SB_VERSION2_OKBITS		\
84	(XFS_SB_VERSION2_LAZYSBCOUNTBIT	| \
85	 XFS_SB_VERSION2_ATTR2BIT	| \
86	 XFS_SB_VERSION2_PROJID32BIT	| \
87	 XFS_SB_VERSION2_FTYPE)
88
89/* Maximum size of the xfs filesystem label, no terminating NULL */
90#define XFSLABEL_MAX			12
91
92/*
93 * Superblock - in core version.  Must match the ondisk version below.
94 * Must be padded to 64 bit alignment.
95 */
96typedef struct xfs_sb {
97	uint32_t	sb_magicnum;	/* magic number == XFS_SB_MAGIC */
98	uint32_t	sb_blocksize;	/* logical block size, bytes */
99	xfs_rfsblock_t	sb_dblocks;	/* number of data blocks */
100	xfs_rfsblock_t	sb_rblocks;	/* number of realtime blocks */
101	xfs_rtblock_t	sb_rextents;	/* number of realtime extents */
102	uuid_t		sb_uuid;	/* user-visible file system unique id */
103	xfs_fsblock_t	sb_logstart;	/* starting block of log if internal */
104	xfs_ino_t	sb_rootino;	/* root inode number */
105	xfs_ino_t	sb_rbmino;	/* bitmap inode for realtime extents */
106	xfs_ino_t	sb_rsumino;	/* summary inode for rt bitmap */
107	xfs_agblock_t	sb_rextsize;	/* realtime extent size, blocks */
108	xfs_agblock_t	sb_agblocks;	/* size of an allocation group */
109	xfs_agnumber_t	sb_agcount;	/* number of allocation groups */
110	xfs_extlen_t	sb_rbmblocks;	/* number of rt bitmap blocks */
111	xfs_extlen_t	sb_logblocks;	/* number of log blocks */
112	uint16_t	sb_versionnum;	/* header version == XFS_SB_VERSION */
113	uint16_t	sb_sectsize;	/* volume sector size, bytes */
114	uint16_t	sb_inodesize;	/* inode size, bytes */
115	uint16_t	sb_inopblock;	/* inodes per block */
116	char		sb_fname[XFSLABEL_MAX]; /* file system name */
117	uint8_t		sb_blocklog;	/* log2 of sb_blocksize */
118	uint8_t		sb_sectlog;	/* log2 of sb_sectsize */
119	uint8_t		sb_inodelog;	/* log2 of sb_inodesize */
120	uint8_t		sb_inopblog;	/* log2 of sb_inopblock */
121	uint8_t		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
122	uint8_t		sb_rextslog;	/* log2 of sb_rextents */
123	uint8_t		sb_inprogress;	/* mkfs is in progress, don't mount */
124	uint8_t		sb_imax_pct;	/* max % of fs for inode space */
125					/* statistics */
126	/*
127	 * These fields must remain contiguous.  If you really
128	 * want to change their layout, make sure you fix the
129	 * code in xfs_trans_apply_sb_deltas().
130	 */
131	uint64_t	sb_icount;	/* allocated inodes */
132	uint64_t	sb_ifree;	/* free inodes */
133	uint64_t	sb_fdblocks;	/* free data blocks */
134	uint64_t	sb_frextents;	/* free realtime extents */
135	/*
136	 * End contiguous fields.
137	 */
138	xfs_ino_t	sb_uquotino;	/* user quota inode */
139	xfs_ino_t	sb_gquotino;	/* group quota inode */
140	uint16_t	sb_qflags;	/* quota flags */
141	uint8_t		sb_flags;	/* misc. flags */
142	uint8_t		sb_shared_vn;	/* shared version number */
143	xfs_extlen_t	sb_inoalignmt;	/* inode chunk alignment, fsblocks */
144	uint32_t	sb_unit;	/* stripe or raid unit */
145	uint32_t	sb_width;	/* stripe or raid width */
146	uint8_t		sb_dirblklog;	/* log2 of dir block size (fsbs) */
147	uint8_t		sb_logsectlog;	/* log2 of the log sector size */
148	uint16_t	sb_logsectsize;	/* sector size for the log, bytes */
149	uint32_t	sb_logsunit;	/* stripe unit size for the log */
150	uint32_t	sb_features2;	/* additional feature bits */
151
152	/*
153	 * bad features2 field as a result of failing to pad the sb structure to
154	 * 64 bits. Some machines will be using this field for features2 bits.
155	 * Easiest just to mark it bad and not use it for anything else.
156	 *
157	 * This is not kept up to date in memory; it is always overwritten by
158	 * the value in sb_features2 when formatting the incore superblock to
159	 * the disk buffer.
160	 */
161	uint32_t	sb_bad_features2;
162
163	/* version 5 superblock fields start here */
164
165	/* feature masks */
166	uint32_t	sb_features_compat;
167	uint32_t	sb_features_ro_compat;
168	uint32_t	sb_features_incompat;
169	uint32_t	sb_features_log_incompat;
170
171	uint32_t	sb_crc;		/* superblock crc */
172	xfs_extlen_t	sb_spino_align;	/* sparse inode chunk alignment */
173
174	xfs_ino_t	sb_pquotino;	/* project quota inode */
175	xfs_lsn_t	sb_lsn;		/* last write sequence */
176	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
177
178	/* must be padded to 64 bit alignment */
179} xfs_sb_t;
180
181#define XFS_SB_CRC_OFF		offsetof(struct xfs_sb, sb_crc)
182
183/*
184 * Superblock - on disk version.  Must match the in core version above.
185 * Must be padded to 64 bit alignment.
186 */
187typedef struct xfs_dsb {
188	__be32		sb_magicnum;	/* magic number == XFS_SB_MAGIC */
189	__be32		sb_blocksize;	/* logical block size, bytes */
190	__be64		sb_dblocks;	/* number of data blocks */
191	__be64		sb_rblocks;	/* number of realtime blocks */
192	__be64		sb_rextents;	/* number of realtime extents */
193	uuid_t		sb_uuid;	/* user-visible file system unique id */
194	__be64		sb_logstart;	/* starting block of log if internal */
195	__be64		sb_rootino;	/* root inode number */
196	__be64		sb_rbmino;	/* bitmap inode for realtime extents */
197	__be64		sb_rsumino;	/* summary inode for rt bitmap */
198	__be32		sb_rextsize;	/* realtime extent size, blocks */
199	__be32		sb_agblocks;	/* size of an allocation group */
200	__be32		sb_agcount;	/* number of allocation groups */
201	__be32		sb_rbmblocks;	/* number of rt bitmap blocks */
202	__be32		sb_logblocks;	/* number of log blocks */
203	__be16		sb_versionnum;	/* header version == XFS_SB_VERSION */
204	__be16		sb_sectsize;	/* volume sector size, bytes */
205	__be16		sb_inodesize;	/* inode size, bytes */
206	__be16		sb_inopblock;	/* inodes per block */
207	char		sb_fname[XFSLABEL_MAX]; /* file system name */
208	__u8		sb_blocklog;	/* log2 of sb_blocksize */
209	__u8		sb_sectlog;	/* log2 of sb_sectsize */
210	__u8		sb_inodelog;	/* log2 of sb_inodesize */
211	__u8		sb_inopblog;	/* log2 of sb_inopblock */
212	__u8		sb_agblklog;	/* log2 of sb_agblocks (rounded up) */
213	__u8		sb_rextslog;	/* log2 of sb_rextents */
214	__u8		sb_inprogress;	/* mkfs is in progress, don't mount */
215	__u8		sb_imax_pct;	/* max % of fs for inode space */
216					/* statistics */
217	/*
218	 * These fields must remain contiguous.  If you really
219	 * want to change their layout, make sure you fix the
220	 * code in xfs_trans_apply_sb_deltas().
221	 */
222	__be64		sb_icount;	/* allocated inodes */
223	__be64		sb_ifree;	/* free inodes */
224	__be64		sb_fdblocks;	/* free data blocks */
225	__be64		sb_frextents;	/* free realtime extents */
226	/*
227	 * End contiguous fields.
228	 */
229	__be64		sb_uquotino;	/* user quota inode */
230	__be64		sb_gquotino;	/* group quota inode */
231	__be16		sb_qflags;	/* quota flags */
232	__u8		sb_flags;	/* misc. flags */
233	__u8		sb_shared_vn;	/* shared version number */
234	__be32		sb_inoalignmt;	/* inode chunk alignment, fsblocks */
235	__be32		sb_unit;	/* stripe or raid unit */
236	__be32		sb_width;	/* stripe or raid width */
237	__u8		sb_dirblklog;	/* log2 of dir block size (fsbs) */
238	__u8		sb_logsectlog;	/* log2 of the log sector size */
239	__be16		sb_logsectsize;	/* sector size for the log, bytes */
240	__be32		sb_logsunit;	/* stripe unit size for the log */
241	__be32		sb_features2;	/* additional feature bits */
242	/*
243	 * bad features2 field as a result of failing to pad the sb
244	 * structure to 64 bits. Some machines will be using this field
245	 * for features2 bits. Easiest just to mark it bad and not use
246	 * it for anything else.
247	 */
248	__be32		sb_bad_features2;
249
250	/* version 5 superblock fields start here */
251
252	/* feature masks */
253	__be32		sb_features_compat;
254	__be32		sb_features_ro_compat;
255	__be32		sb_features_incompat;
256	__be32		sb_features_log_incompat;
257
258	__le32		sb_crc;		/* superblock crc */
259	__be32		sb_spino_align;	/* sparse inode chunk alignment */
260
261	__be64		sb_pquotino;	/* project quota inode */
262	__be64		sb_lsn;		/* last write sequence */
263	uuid_t		sb_meta_uuid;	/* metadata file system unique id */
264
265	/* must be padded to 64 bit alignment */
266} xfs_dsb_t;
267
268
269/*
270 * Misc. Flags - warning - these will be cleared by xfs_repair unless
271 * a feature bit is set when the flag is used.
272 */
273#define XFS_SBF_NOFLAGS		0x00	/* no flags set */
274#define XFS_SBF_READONLY	0x01	/* only read-only mounts allowed */
275
276/*
277 * define max. shared version we can interoperate with
278 */
279#define XFS_SB_MAX_SHARED_VN	0
280
281#define	XFS_SB_VERSION_NUM(sbp)	((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
282
283/*
284 * The first XFS version we support is a v4 superblock with V2 directories.
285 */
286static inline bool xfs_sb_good_v4_features(struct xfs_sb *sbp)
287{
288	if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
289		return false;
290	if (!(sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT))
291		return false;
292
293	/* check for unknown features in the fs */
294	if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
295	    ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
296	     (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
297		return false;
298
299	return true;
300}
301
302static inline bool xfs_sb_good_version(struct xfs_sb *sbp)
303{
304	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5)
305		return true;
306	if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_4)
307		return xfs_sb_good_v4_features(sbp);
308	return false;
309}
310
311static inline bool xfs_sb_version_hasrealtime(struct xfs_sb *sbp)
312{
313	return sbp->sb_rblocks > 0;
314}
315
316/*
317 * Detect a mismatched features2 field.  Older kernels read/wrote
318 * this into the wrong slot, so to be safe we keep them in sync.
319 */
320static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
321{
322	return sbp->sb_bad_features2 != sbp->sb_features2;
323}
324
325static inline bool xfs_sb_version_hasattr(struct xfs_sb *sbp)
326{
327	return (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT);
328}
329
330static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
331{
332	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
333}
334
335static inline bool xfs_sb_version_hasquota(struct xfs_sb *sbp)
336{
337	return (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT);
338}
339
340static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
341{
342	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
343}
344
345static inline bool xfs_sb_version_hasalign(struct xfs_sb *sbp)
346{
347	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
348		(sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT));
349}
350
351static inline bool xfs_sb_version_hasdalign(struct xfs_sb *sbp)
352{
353	return (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT);
354}
355
356static inline bool xfs_sb_version_haslogv2(struct xfs_sb *sbp)
357{
358	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
359	       (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT);
360}
361
362static inline bool xfs_sb_version_hassector(struct xfs_sb *sbp)
363{
364	return (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT);
365}
366
367static inline bool xfs_sb_version_hasasciici(struct xfs_sb *sbp)
368{
369	return (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT);
370}
371
372static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp)
373{
374	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 ||
375	       (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
376}
377
378/*
379 * sb_features2 bit version macros.
380 */
381static inline bool xfs_sb_version_haslazysbcount(struct xfs_sb *sbp)
382{
383	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
384	       (xfs_sb_version_hasmorebits(sbp) &&
385		(sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT));
386}
387
388static inline bool xfs_sb_version_hasattr2(struct xfs_sb *sbp)
389{
390	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
391	       (xfs_sb_version_hasmorebits(sbp) &&
392		(sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT));
393}
394
395static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
396{
397	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
398	sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
399}
400
401static inline void xfs_sb_version_removeattr2(struct xfs_sb *sbp)
402{
403	sbp->sb_features2 &= ~XFS_SB_VERSION2_ATTR2BIT;
404	if (!sbp->sb_features2)
405		sbp->sb_versionnum &= ~XFS_SB_VERSION_MOREBITSBIT;
406}
407
408static inline bool xfs_sb_version_hasprojid32bit(struct xfs_sb *sbp)
409{
410	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) ||
411	       (xfs_sb_version_hasmorebits(sbp) &&
412		(sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT));
413}
414
415static inline void xfs_sb_version_addprojid32bit(struct xfs_sb *sbp)
416{
417	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
418	sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
419}
420
421/*
422 * Extended v5 superblock feature masks. These are to be used for new v5
423 * superblock features only.
424 *
425 * Compat features are new features that old kernels will not notice or affect
426 * and so can mount read-write without issues.
427 *
428 * RO-Compat (read only) are features that old kernels can read but will break
429 * if they write. Hence only read-only mounts of such filesystems are allowed on
430 * kernels that don't support the feature bit.
431 *
432 * InCompat features are features which old kernels will not understand and so
433 * must not mount.
434 *
435 * Log-InCompat features are for changes to log formats or new transactions that
436 * can't be replayed on older kernels. The fields are set when the filesystem is
437 * mounted, and a clean unmount clears the fields.
438 */
439#define XFS_SB_FEAT_COMPAT_ALL 0
440#define XFS_SB_FEAT_COMPAT_UNKNOWN	~XFS_SB_FEAT_COMPAT_ALL
441static inline bool
442xfs_sb_has_compat_feature(
443	struct xfs_sb	*sbp,
444	uint32_t	feature)
445{
446	return (sbp->sb_features_compat & feature) != 0;
447}
448
449#define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
450#define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
451#define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
452#define XFS_SB_FEAT_RO_COMPAT_INOBTCNT (1 << 3)		/* inobt block counts */
453#define XFS_SB_FEAT_RO_COMPAT_ALL \
454		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
455		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
456		 XFS_SB_FEAT_RO_COMPAT_REFLINK| \
457		 XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
458#define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
459static inline bool
460xfs_sb_has_ro_compat_feature(
461	struct xfs_sb	*sbp,
462	uint32_t	feature)
463{
464	return (sbp->sb_features_ro_compat & feature) != 0;
465}
466
467#define XFS_SB_FEAT_INCOMPAT_FTYPE	(1 << 0)	/* filetype in dirent */
468#define XFS_SB_FEAT_INCOMPAT_SPINODES	(1 << 1)	/* sparse inode chunks */
469#define XFS_SB_FEAT_INCOMPAT_META_UUID	(1 << 2)	/* metadata UUID */
470#define XFS_SB_FEAT_INCOMPAT_BIGTIME	(1 << 3)	/* large timestamps */
471#define XFS_SB_FEAT_INCOMPAT_ALL \
472		(XFS_SB_FEAT_INCOMPAT_FTYPE|	\
473		 XFS_SB_FEAT_INCOMPAT_SPINODES|	\
474		 XFS_SB_FEAT_INCOMPAT_META_UUID| \
475		 XFS_SB_FEAT_INCOMPAT_BIGTIME)
476
477#define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
478static inline bool
479xfs_sb_has_incompat_feature(
480	struct xfs_sb	*sbp,
481	uint32_t	feature)
482{
483	return (sbp->sb_features_incompat & feature) != 0;
484}
485
486#define XFS_SB_FEAT_INCOMPAT_LOG_ALL 0
487#define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
488static inline bool
489xfs_sb_has_incompat_log_feature(
490	struct xfs_sb	*sbp,
491	uint32_t	feature)
492{
493	return (sbp->sb_features_log_incompat & feature) != 0;
494}
495
496/*
497 * V5 superblock specific feature checks
498 */
499static inline bool xfs_sb_version_hascrc(struct xfs_sb *sbp)
500{
501	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
502}
503
504/*
505 * v5 file systems support V3 inodes only, earlier file systems support
506 * v2 and v1 inodes.
507 */
508static inline bool xfs_sb_version_has_v3inode(struct xfs_sb *sbp)
509{
510	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
511}
512
513static inline bool xfs_dinode_good_version(struct xfs_sb *sbp,
514		uint8_t version)
515{
516	if (xfs_sb_version_has_v3inode(sbp))
517		return version == 3;
518	return version == 1 || version == 2;
519}
520
521static inline bool xfs_sb_version_has_pquotino(struct xfs_sb *sbp)
522{
523	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
524}
525
526static inline int xfs_sb_version_hasftype(struct xfs_sb *sbp)
527{
528	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
529		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_FTYPE)) ||
530	       (xfs_sb_version_hasmorebits(sbp) &&
531		 (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE));
532}
533
534static inline bool xfs_sb_version_hasfinobt(xfs_sb_t *sbp)
535{
536	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
537		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT);
538}
539
540static inline bool xfs_sb_version_hassparseinodes(struct xfs_sb *sbp)
541{
542	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
543		xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_SPINODES);
544}
545
546/*
547 * XFS_SB_FEAT_INCOMPAT_META_UUID indicates that the metadata UUID
548 * is stored separately from the user-visible UUID; this allows the
549 * user-visible UUID to be changed on V5 filesystems which have a
550 * filesystem UUID stamped into every piece of metadata.
551 */
552static inline bool xfs_sb_version_hasmetauuid(struct xfs_sb *sbp)
553{
554	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
555		(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID);
556}
557
558static inline bool xfs_sb_version_hasrmapbt(struct xfs_sb *sbp)
559{
560	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
561		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT);
562}
563
564static inline bool xfs_sb_version_hasreflink(struct xfs_sb *sbp)
565{
566	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
567		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK);
568}
569
570static inline bool xfs_sb_version_hasbigtime(struct xfs_sb *sbp)
571{
572	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
573		(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_BIGTIME);
574}
575
576/*
577 * Inode btree block counter.  We record the number of inobt and finobt blocks
578 * in the AGI header so that we can skip the finobt walk at mount time when
579 * setting up per-AG reservations.
580 */
581static inline bool xfs_sb_version_hasinobtcounts(struct xfs_sb *sbp)
582{
583	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 &&
584		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_INOBTCNT);
585}
586
587/*
588 * end of superblock version macros
589 */
590
591static inline bool
592xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
593{
594	return (ino == sbp->sb_uquotino ||
595		ino == sbp->sb_gquotino ||
596		ino == sbp->sb_pquotino);
597}
598
599#define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
600#define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
601
602#define	XFS_HDR_BLOCK(mp,d)	((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
603#define	XFS_DADDR_TO_FSB(mp,d)	XFS_AGB_TO_FSB(mp, \
604			xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
605#define	XFS_FSB_TO_DADDR(mp,fsbno)	XFS_AGB_TO_DADDR(mp, \
606			XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
607
608/*
609 * File system sector to basic block conversions.
610 */
611#define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
612
613/*
614 * File system block to basic block conversions.
615 */
616#define	XFS_FSB_TO_BB(mp,fsbno)	((fsbno) << (mp)->m_blkbb_log)
617#define	XFS_BB_TO_FSB(mp,bb)	\
618	(((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
619#define	XFS_BB_TO_FSBT(mp,bb)	((bb) >> (mp)->m_blkbb_log)
620
621/*
622 * File system block to byte conversions.
623 */
624#define XFS_FSB_TO_B(mp,fsbno)	((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
625#define XFS_B_TO_FSB(mp,b)	\
626	((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
627#define XFS_B_TO_FSBT(mp,b)	(((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
628#define XFS_B_FSB_OFFSET(mp,b)	((b) & (mp)->m_blockmask)
629
630/*
631 * Allocation group header
632 *
633 * This is divided into three structures, placed in sequential 512-byte
634 * buffers after a copy of the superblock (also in a 512-byte buffer).
635 */
636#define	XFS_AGF_MAGIC	0x58414746	/* 'XAGF' */
637#define	XFS_AGI_MAGIC	0x58414749	/* 'XAGI' */
638#define	XFS_AGFL_MAGIC	0x5841464c	/* 'XAFL' */
639#define	XFS_AGF_VERSION	1
640#define	XFS_AGI_VERSION	1
641
642#define	XFS_AGF_GOOD_VERSION(v)	((v) == XFS_AGF_VERSION)
643#define	XFS_AGI_GOOD_VERSION(v)	((v) == XFS_AGI_VERSION)
644
645/*
646 * Btree number 0 is bno, 1 is cnt, 2 is rmap. This value gives the size of the
647 * arrays below.
648 */
649#define	XFS_BTNUM_AGF	((int)XFS_BTNUM_RMAPi + 1)
650
651/*
652 * The second word of agf_levels in the first a.g. overlaps the EFS
653 * superblock's magic number.  Since the magic numbers valid for EFS
654 * are > 64k, our value cannot be confused for an EFS superblock's.
655 */
656
657typedef struct xfs_agf {
658	/*
659	 * Common allocation group header information
660	 */
661	__be32		agf_magicnum;	/* magic number == XFS_AGF_MAGIC */
662	__be32		agf_versionnum;	/* header version == XFS_AGF_VERSION */
663	__be32		agf_seqno;	/* sequence # starting from 0 */
664	__be32		agf_length;	/* size in blocks of a.g. */
665	/*
666	 * Freespace and rmap information
667	 */
668	__be32		agf_roots[XFS_BTNUM_AGF];	/* root blocks */
669	__be32		agf_levels[XFS_BTNUM_AGF];	/* btree levels */
670
671	__be32		agf_flfirst;	/* first freelist block's index */
672	__be32		agf_fllast;	/* last freelist block's index */
673	__be32		agf_flcount;	/* count of blocks in freelist */
674	__be32		agf_freeblks;	/* total free blocks */
675
676	__be32		agf_longest;	/* longest free space */
677	__be32		agf_btreeblks;	/* # of blocks held in AGF btrees */
678	uuid_t		agf_uuid;	/* uuid of filesystem */
679
680	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
681	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
682
683	__be32		agf_refcount_root;	/* refcount tree root block */
684	__be32		agf_refcount_level;	/* refcount btree levels */
685
686	/*
687	 * reserve some contiguous space for future logged fields before we add
688	 * the unlogged fields. This makes the range logging via flags and
689	 * structure offsets much simpler.
690	 */
691	__be64		agf_spare64[14];
692
693	/* unlogged fields, written during buffer writeback. */
694	__be64		agf_lsn;	/* last write sequence */
695	__be32		agf_crc;	/* crc of agf sector */
696	__be32		agf_spare2;
697
698	/* structure must be padded to 64 bit alignment */
699} xfs_agf_t;
700
701#define XFS_AGF_CRC_OFF		offsetof(struct xfs_agf, agf_crc)
702
703#define	XFS_AGF_MAGICNUM	0x00000001
704#define	XFS_AGF_VERSIONNUM	0x00000002
705#define	XFS_AGF_SEQNO		0x00000004
706#define	XFS_AGF_LENGTH		0x00000008
707#define	XFS_AGF_ROOTS		0x00000010
708#define	XFS_AGF_LEVELS		0x00000020
709#define	XFS_AGF_FLFIRST		0x00000040
710#define	XFS_AGF_FLLAST		0x00000080
711#define	XFS_AGF_FLCOUNT		0x00000100
712#define	XFS_AGF_FREEBLKS	0x00000200
713#define	XFS_AGF_LONGEST		0x00000400
714#define	XFS_AGF_BTREEBLKS	0x00000800
715#define	XFS_AGF_UUID		0x00001000
716#define	XFS_AGF_RMAP_BLOCKS	0x00002000
717#define	XFS_AGF_REFCOUNT_BLOCKS	0x00004000
718#define	XFS_AGF_REFCOUNT_ROOT	0x00008000
719#define	XFS_AGF_REFCOUNT_LEVEL	0x00010000
720#define	XFS_AGF_SPARE64		0x00020000
721#define	XFS_AGF_NUM_BITS	18
722#define	XFS_AGF_ALL_BITS	((1 << XFS_AGF_NUM_BITS) - 1)
723
724#define XFS_AGF_FLAGS \
725	{ XFS_AGF_MAGICNUM,	"MAGICNUM" }, \
726	{ XFS_AGF_VERSIONNUM,	"VERSIONNUM" }, \
727	{ XFS_AGF_SEQNO,	"SEQNO" }, \
728	{ XFS_AGF_LENGTH,	"LENGTH" }, \
729	{ XFS_AGF_ROOTS,	"ROOTS" }, \
730	{ XFS_AGF_LEVELS,	"LEVELS" }, \
731	{ XFS_AGF_FLFIRST,	"FLFIRST" }, \
732	{ XFS_AGF_FLLAST,	"FLLAST" }, \
733	{ XFS_AGF_FLCOUNT,	"FLCOUNT" }, \
734	{ XFS_AGF_FREEBLKS,	"FREEBLKS" }, \
735	{ XFS_AGF_LONGEST,	"LONGEST" }, \
736	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
737	{ XFS_AGF_UUID,		"UUID" }, \
738	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
739	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
740	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
741	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
742	{ XFS_AGF_SPARE64,	"SPARE64" }
743
744/* disk block (xfs_daddr_t) in the AG */
745#define XFS_AGF_DADDR(mp)	((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
746#define	XFS_AGF_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
747
748/*
749 * Size of the unlinked inode hash table in the agi.
750 */
751#define	XFS_AGI_UNLINKED_BUCKETS	64
752
753typedef struct xfs_agi {
754	/*
755	 * Common allocation group header information
756	 */
757	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
758	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
759	__be32		agi_seqno;	/* sequence # starting from 0 */
760	__be32		agi_length;	/* size in blocks of a.g. */
761	/*
762	 * Inode information
763	 * Inodes are mapped by interpreting the inode number, so no
764	 * mapping data is needed here.
765	 */
766	__be32		agi_count;	/* count of allocated inodes */
767	__be32		agi_root;	/* root of inode btree */
768	__be32		agi_level;	/* levels in inode btree */
769	__be32		agi_freecount;	/* number of free inodes */
770
771	__be32		agi_newino;	/* new inode just allocated */
772	__be32		agi_dirino;	/* last directory inode chunk */
773	/*
774	 * Hash table of inodes which have been unlinked but are
775	 * still being referenced.
776	 */
777	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
778	/*
779	 * This marks the end of logging region 1 and start of logging region 2.
780	 */
781	uuid_t		agi_uuid;	/* uuid of filesystem */
782	__be32		agi_crc;	/* crc of agi sector */
783	__be32		agi_pad32;
784	__be64		agi_lsn;	/* last write sequence */
785
786	__be32		agi_free_root; /* root of the free inode btree */
787	__be32		agi_free_level;/* levels in free inode btree */
788
789	__be32		agi_iblocks;	/* inobt blocks used */
790	__be32		agi_fblocks;	/* finobt blocks used */
791
792	/* structure must be padded to 64 bit alignment */
793} xfs_agi_t;
794
795#define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
796
797#define	XFS_AGI_MAGICNUM	(1 << 0)
798#define	XFS_AGI_VERSIONNUM	(1 << 1)
799#define	XFS_AGI_SEQNO		(1 << 2)
800#define	XFS_AGI_LENGTH		(1 << 3)
801#define	XFS_AGI_COUNT		(1 << 4)
802#define	XFS_AGI_ROOT		(1 << 5)
803#define	XFS_AGI_LEVEL		(1 << 6)
804#define	XFS_AGI_FREECOUNT	(1 << 7)
805#define	XFS_AGI_NEWINO		(1 << 8)
806#define	XFS_AGI_DIRINO		(1 << 9)
807#define	XFS_AGI_UNLINKED	(1 << 10)
808#define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
809#define	XFS_AGI_ALL_BITS_R1	((1 << XFS_AGI_NUM_BITS_R1) - 1)
810#define	XFS_AGI_FREE_ROOT	(1 << 11)
811#define	XFS_AGI_FREE_LEVEL	(1 << 12)
812#define	XFS_AGI_IBLOCKS		(1 << 13) /* both inobt/finobt block counters */
813#define	XFS_AGI_NUM_BITS_R2	14
814
815/* disk block (xfs_daddr_t) in the AG */
816#define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
817#define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
818
819/*
820 * The third a.g. block contains the a.g. freelist, an array
821 * of block pointers to blocks owned by the allocation btree code.
822 */
823#define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
824#define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
825#define	XFS_BUF_TO_AGFL(bp)	((struct xfs_agfl *)((bp)->b_addr))
826
827struct xfs_agfl {
828	__be32		agfl_magicnum;
829	__be32		agfl_seqno;
830	uuid_t		agfl_uuid;
831	__be64		agfl_lsn;
832	__be32		agfl_crc;
833} __attribute__((packed));
834
835#define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
836
837#define XFS_AGB_TO_FSB(mp,agno,agbno)	\
838	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
839#define	XFS_FSB_TO_AGNO(mp,fsbno)	\
840	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
841#define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
842	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
843#define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
844	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
845		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
846#define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
847
848/*
849 * For checking for bad ranges of xfs_daddr_t's, covering multiple
850 * allocation groups or a single xfs_daddr_t that's a superblock copy.
851 */
852#define	XFS_AG_CHECK_DADDR(mp,d,len)	\
853	((len) == 1 ? \
854	    ASSERT((d) == XFS_SB_DADDR || \
855		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
856	    ASSERT(xfs_daddr_to_agno(mp, d) == \
857		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
858
859/*
860 * XFS Timestamps
861 * ==============
862 *
863 * Traditional ondisk inode timestamps consist of signed 32-bit counters for
864 * seconds and nanoseconds; time zero is the Unix epoch, Jan  1 00:00:00 UTC
865 * 1970, which means that the timestamp epoch is the same as the Unix epoch.
866 * Therefore, the ondisk min and max defined here can be used directly to
867 * constrain the incore timestamps on a Unix system.  Note that we actually
868 * encode a __be64 value on disk.
869 *
870 * When the bigtime feature is enabled, ondisk inode timestamps become an
871 * unsigned 64-bit nanoseconds counter.  This means that the bigtime inode
872 * timestamp epoch is the start of the classic timestamp range, which is
873 * Dec 31 20:45:52 UTC 1901.  Because the epochs are not the same, callers
874 * /must/ use the bigtime conversion functions when encoding and decoding raw
875 * timestamps.
876 */
877typedef __be64 xfs_timestamp_t;
878
879/* Legacy timestamp encoding format. */
880struct xfs_legacy_timestamp {
881	__be32		t_sec;		/* timestamp seconds */
882	__be32		t_nsec;		/* timestamp nanoseconds */
883};
884
885/*
886 * Smallest possible ondisk seconds value with traditional timestamps.  This
887 * corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901.
888 */
889#define XFS_LEGACY_TIME_MIN	((int64_t)S32_MIN)
890
891/*
892 * Largest possible ondisk seconds value with traditional timestamps.  This
893 * corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038.
894 */
895#define XFS_LEGACY_TIME_MAX	((int64_t)S32_MAX)
896
897/*
898 * Smallest possible ondisk seconds value with bigtime timestamps.  This
899 * corresponds (after conversion to a Unix timestamp) with the traditional
900 * minimum timestamp of Dec 13 20:45:52 UTC 1901.
901 */
902#define XFS_BIGTIME_TIME_MIN	((int64_t)0)
903
904/*
905 * Largest supported ondisk seconds value with bigtime timestamps.  This
906 * corresponds (after conversion to a Unix timestamp) with an incore timestamp
907 * of Jul  2 20:20:24 UTC 2486.
908 *
909 * We round down the ondisk limit so that the bigtime quota and inode max
910 * timestamps will be the same.
911 */
912#define XFS_BIGTIME_TIME_MAX	((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL))
913
914/*
915 * Bigtime epoch is set exactly to the minimum time value that a traditional
916 * 32-bit timestamp can represent when using the Unix epoch as a reference.
917 * Hence the Unix epoch is at a fixed offset into the supported bigtime
918 * timestamp range.
919 *
920 * The bigtime epoch also matches the minimum value an on-disk 32-bit XFS
921 * timestamp can represent so we will not lose any fidelity in converting
922 * to/from unix and bigtime timestamps.
923 *
924 * The following conversion factor converts a seconds counter from the Unix
925 * epoch to the bigtime epoch.
926 */
927#define XFS_BIGTIME_EPOCH_OFFSET	(-(int64_t)S32_MIN)
928
929/* Convert a timestamp from the Unix epoch to the bigtime epoch. */
930static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds)
931{
932	return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET;
933}
934
935/* Convert a timestamp from the bigtime epoch to the Unix epoch. */
936static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds)
937{
938	return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET;
939}
940
941/*
942 * On-disk inode structure.
943 *
944 * This is just the header or "dinode core", the inode is expanded to fill a
945 * variable size the leftover area split into a data and an attribute fork.
946 * The format of the data and attribute fork depends on the format of the
947 * inode as indicated by di_format and di_aformat.  To access the data and
948 * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
949 * below.
950 *
951 * There is a very similar struct icdinode in xfs_inode which matches the
952 * layout of the first 96 bytes of this structure, but is kept in native
953 * format instead of big endian.
954 *
955 * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
956 * padding field for v3 inodes.
957 */
958#define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
959typedef struct xfs_dinode {
960	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
961	__be16		di_mode;	/* mode and type of file */
962	__u8		di_version;	/* inode version */
963	__u8		di_format;	/* format of di_c data */
964	__be16		di_onlink;	/* old number of links to file */
965	__be32		di_uid;		/* owner's user id */
966	__be32		di_gid;		/* owner's group id */
967	__be32		di_nlink;	/* number of links to file */
968	__be16		di_projid_lo;	/* lower part of owner's project id */
969	__be16		di_projid_hi;	/* higher part owner's project id */
970	__u8		di_pad[6];	/* unused, zeroed space */
971	__be16		di_flushiter;	/* incremented on flush */
972	xfs_timestamp_t	di_atime;	/* time last accessed */
973	xfs_timestamp_t	di_mtime;	/* time last modified */
974	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
975	__be64		di_size;	/* number of bytes in file */
976	__be64		di_nblocks;	/* # of direct & btree blocks used */
977	__be32		di_extsize;	/* basic/minimum extent size for file */
978	__be32		di_nextents;	/* number of extents in data fork */
979	__be16		di_anextents;	/* number of extents in attribute fork*/
980	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
981	__s8		di_aformat;	/* format of attr fork's data */
982	__be32		di_dmevmask;	/* DMIG event mask */
983	__be16		di_dmstate;	/* DMIG state info */
984	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
985	__be32		di_gen;		/* generation number */
986
987	/* di_next_unlinked is the only non-core field in the old dinode */
988	__be32		di_next_unlinked;/* agi unlinked list ptr */
989
990	/* start of the extended dinode, writable fields */
991	__le32		di_crc;		/* CRC of the inode */
992	__be64		di_changecount;	/* number of attribute changes */
993	__be64		di_lsn;		/* flush sequence */
994	__be64		di_flags2;	/* more random flags */
995	__be32		di_cowextsize;	/* basic cow extent size for file */
996	__u8		di_pad2[12];	/* more padding for future expansion */
997
998	/* fields only written to during inode creation */
999	xfs_timestamp_t	di_crtime;	/* time created */
1000	__be64		di_ino;		/* inode number */
1001	uuid_t		di_uuid;	/* UUID of the filesystem */
1002
1003	/* structure must be padded to 64 bit alignment */
1004} xfs_dinode_t;
1005
1006#define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
1007
1008#define DI_MAX_FLUSH 0xffff
1009
1010/*
1011 * Size of the core inode on disk.  Version 1 and 2 inodes have
1012 * the same size, but version 3 has grown a few additional fields.
1013 */
1014static inline uint xfs_dinode_size(int version)
1015{
1016	if (version == 3)
1017		return sizeof(struct xfs_dinode);
1018	return offsetof(struct xfs_dinode, di_crc);
1019}
1020
1021/*
1022 * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
1023 * Since the pathconf interface is signed, we use 2^31 - 1 instead.
1024 */
1025#define	XFS_MAXLINK		((1U << 31) - 1U)
1026
1027/*
1028 * Values for di_format
1029 *
1030 * This enum is used in string mapping in xfs_trace.h; please keep the
1031 * TRACE_DEFINE_ENUMs for it up to date.
1032 */
1033enum xfs_dinode_fmt {
1034	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
1035	XFS_DINODE_FMT_LOCAL,		/* bulk data */
1036	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
1037	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
1038	XFS_DINODE_FMT_UUID		/* added long ago, but never used */
1039};
1040
1041#define XFS_INODE_FORMAT_STR \
1042	{ XFS_DINODE_FMT_DEV,		"dev" }, \
1043	{ XFS_DINODE_FMT_LOCAL,		"local" }, \
1044	{ XFS_DINODE_FMT_EXTENTS,	"extent" }, \
1045	{ XFS_DINODE_FMT_BTREE,		"btree" }, \
1046	{ XFS_DINODE_FMT_UUID,		"uuid" }
1047
1048/*
1049 * Inode minimum and maximum sizes.
1050 */
1051#define	XFS_DINODE_MIN_LOG	8
1052#define	XFS_DINODE_MAX_LOG	11
1053#define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
1054#define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
1055
1056/*
1057 * Inode size for given fs.
1058 */
1059#define XFS_DINODE_SIZE(sbp) \
1060	(xfs_sb_version_has_v3inode(sbp) ? \
1061		sizeof(struct xfs_dinode) : \
1062		offsetof(struct xfs_dinode, di_crc))
1063#define XFS_LITINO(mp) \
1064	((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(&(mp)->m_sb))
1065
1066/*
1067 * Inode data & attribute fork sizes, per inode.
1068 */
1069#define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
1070
1071#define XFS_DFORK_DSIZE(dip,mp) \
1072	((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp))
1073#define XFS_DFORK_ASIZE(dip,mp) \
1074	((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0)
1075#define XFS_DFORK_SIZE(dip,mp,w) \
1076	((w) == XFS_DATA_FORK ? \
1077		XFS_DFORK_DSIZE(dip, mp) : \
1078		XFS_DFORK_ASIZE(dip, mp))
1079
1080#define XFS_DFORK_MAXEXT(dip, mp, w) \
1081	(XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
1082
1083/*
1084 * Return pointers to the data or attribute forks.
1085 */
1086#define XFS_DFORK_DPTR(dip) \
1087	((char *)dip + xfs_dinode_size(dip->di_version))
1088#define XFS_DFORK_APTR(dip)	\
1089	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
1090#define XFS_DFORK_PTR(dip,w)	\
1091	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
1092
1093#define XFS_DFORK_FORMAT(dip,w) \
1094	((w) == XFS_DATA_FORK ? \
1095		(dip)->di_format : \
1096		(dip)->di_aformat)
1097#define XFS_DFORK_NEXTENTS(dip,w) \
1098	((w) == XFS_DATA_FORK ? \
1099		be32_to_cpu((dip)->di_nextents) : \
1100		be16_to_cpu((dip)->di_anextents))
1101
1102/*
1103 * For block and character special files the 32bit dev_t is stored at the
1104 * beginning of the data fork.
1105 */
1106static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1107{
1108	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1109}
1110
1111static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1112{
1113	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1114}
1115
1116/*
1117 * Values for di_flags
1118 */
1119#define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
1120#define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
1121#define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
1122#define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
1123#define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
1124#define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
1125#define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
1126#define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
1127#define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
1128#define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
1129#define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
1130#define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
1131#define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
1132#define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
1133#define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
1134/* Do not use bit 15, di_flags is legacy and unchanging now */
1135
1136#define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
1137#define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
1138#define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
1139#define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1140#define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
1141#define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
1142#define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
1143#define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
1144#define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
1145#define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1146#define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1147#define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
1148#define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1149#define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
1150#define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
1151
1152#define XFS_DIFLAG_ANY \
1153	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1154	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1155	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1156	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1157	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1158
1159/*
1160 * Values for di_flags2 These start by being exposed to userspace in the upper
1161 * 16 bits of the XFS_XFLAG_s range.
1162 */
1163#define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
1164#define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
1165#define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
1166#define XFS_DIFLAG2_BIGTIME_BIT	3	/* big timestamps */
1167
1168#define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
1169#define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
1170#define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1171#define XFS_DIFLAG2_BIGTIME	(1 << XFS_DIFLAG2_BIGTIME_BIT)
1172
1173#define XFS_DIFLAG2_ANY \
1174	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \
1175	 XFS_DIFLAG2_BIGTIME)
1176
1177static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip)
1178{
1179	return dip->di_version >= 3 &&
1180	       (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME));
1181}
1182
1183/*
1184 * Inode number format:
1185 * low inopblog bits - offset in block
1186 * next agblklog bits - block number in ag
1187 * next agno_log bits - ag number
1188 * high agno_log-agblklog-inopblog bits - 0
1189 */
1190#define	XFS_INO_MASK(k)			(uint32_t)((1ULL << (k)) - 1)
1191#define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1192#define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1193#define	XFS_INO_AGINO_BITS(mp)		((mp)->m_ino_geo.agino_log)
1194#define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1195#define	XFS_INO_BITS(mp)		\
1196	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1197#define	XFS_INO_TO_AGNO(mp,i)		\
1198	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1199#define	XFS_INO_TO_AGINO(mp,i)		\
1200	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1201#define	XFS_INO_TO_AGBNO(mp,i)		\
1202	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1203		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1204#define	XFS_INO_TO_OFFSET(mp,i)		\
1205	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1206#define	XFS_INO_TO_FSB(mp,i)		\
1207	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1208#define	XFS_AGINO_TO_INO(mp,a,i)	\
1209	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1210#define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1211#define	XFS_AGINO_TO_OFFSET(mp,i)	\
1212	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1213#define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1214	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1215#define	XFS_FSB_TO_INO(mp, b)	((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1216#define	XFS_AGB_TO_AGINO(mp, b)	((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1217
1218#define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1219#define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1220
1221/*
1222 * RealTime Device format definitions
1223 */
1224
1225/* Min and max rt extent sizes, specified in bytes */
1226#define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1227#define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1228#define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1229
1230#define	XFS_BLOCKSIZE(mp)	((mp)->m_sb.sb_blocksize)
1231#define	XFS_BLOCKMASK(mp)	((mp)->m_blockmask)
1232#define	XFS_BLOCKWSIZE(mp)	((mp)->m_blockwsize)
1233#define	XFS_BLOCKWMASK(mp)	((mp)->m_blockwmask)
1234
1235/*
1236 * RT Summary and bit manipulation macros.
1237 */
1238#define	XFS_SUMOFFS(mp,ls,bb)	((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb)))
1239#define	XFS_SUMOFFSTOBLOCK(mp,s)	\
1240	(((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog)
1241#define	XFS_SUMPTR(mp,bp,so)	\
1242	((xfs_suminfo_t *)((bp)->b_addr + \
1243		(((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp))))
1244
1245#define	XFS_BITTOBLOCK(mp,bi)	((bi) >> (mp)->m_blkbit_log)
1246#define	XFS_BLOCKTOBIT(mp,bb)	((bb) << (mp)->m_blkbit_log)
1247#define	XFS_BITTOWORD(mp,bi)	\
1248	((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp)))
1249
1250#define	XFS_RTMIN(a,b)	((a) < (b) ? (a) : (b))
1251#define	XFS_RTMAX(a,b)	((a) > (b) ? (a) : (b))
1252
1253#define	XFS_RTLOBIT(w)	xfs_lowbit32(w)
1254#define	XFS_RTHIBIT(w)	xfs_highbit32(w)
1255
1256#define	XFS_RTBLOCKLOG(b)	xfs_highbit64(b)
1257
1258/*
1259 * Dquot and dquot block format definitions
1260 */
1261#define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1262#define XFS_DQUOT_VERSION	(uint8_t)0x01	/* latest version number */
1263
1264#define XFS_DQTYPE_USER		0x01		/* user dquot record */
1265#define XFS_DQTYPE_PROJ		0x02		/* project dquot record */
1266#define XFS_DQTYPE_GROUP	0x04		/* group dquot record */
1267#define XFS_DQTYPE_BIGTIME	0x80		/* large expiry timestamps */
1268
1269/* bitmask to determine if this is a user/group/project dquot */
1270#define XFS_DQTYPE_REC_MASK	(XFS_DQTYPE_USER | \
1271				 XFS_DQTYPE_PROJ | \
1272				 XFS_DQTYPE_GROUP)
1273
1274#define XFS_DQTYPE_ANY		(XFS_DQTYPE_REC_MASK | \
1275				 XFS_DQTYPE_BIGTIME)
1276
1277/*
1278 * XFS Quota Timers
1279 * ================
1280 *
1281 * Traditional quota grace period expiration timers are an unsigned 32-bit
1282 * seconds counter; time zero is the Unix epoch, Jan  1 00:00:01 UTC 1970.
1283 * Note that an expiration value of zero means that the quota limit has not
1284 * been reached, and therefore no expiration has been set.  Therefore, the
1285 * ondisk min and max defined here can be used directly to constrain the incore
1286 * quota expiration timestamps on a Unix system.
1287 *
1288 * When bigtime is enabled, we trade two bits of precision to expand the
1289 * expiration timeout range to match that of big inode timestamps.  The min and
1290 * max recorded here are the on-disk limits, not a Unix timestamp.
1291 *
1292 * The grace period for each quota type is stored in the root dquot (id = 0)
1293 * and is applied to a non-root dquot when it exceeds the soft or hard limits.
1294 * The length of quota grace periods are unsigned 32-bit quantities measured in
1295 * units of seconds.  A value of zero means to use the default period.
1296 */
1297
1298/*
1299 * Smallest possible ondisk quota expiration value with traditional timestamps.
1300 * This corresponds exactly with the incore expiration Jan  1 00:00:01 UTC 1970.
1301 */
1302#define XFS_DQ_LEGACY_EXPIRY_MIN	((int64_t)1)
1303
1304/*
1305 * Largest possible ondisk quota expiration value with traditional timestamps.
1306 * This corresponds exactly with the incore expiration Feb  7 06:28:15 UTC 2106.
1307 */
1308#define XFS_DQ_LEGACY_EXPIRY_MAX	((int64_t)U32_MAX)
1309
1310/*
1311 * Smallest possible ondisk quota expiration value with bigtime timestamps.
1312 * This corresponds (after conversion to a Unix timestamp) with the incore
1313 * expiration of Jan  1 00:00:04 UTC 1970.
1314 */
1315#define XFS_DQ_BIGTIME_EXPIRY_MIN	(XFS_DQ_LEGACY_EXPIRY_MIN)
1316
1317/*
1318 * Largest supported ondisk quota expiration value with bigtime timestamps.
1319 * This corresponds (after conversion to a Unix timestamp) with an incore
1320 * expiration of Jul  2 20:20:24 UTC 2486.
1321 *
1322 * The ondisk field supports values up to -1U, which corresponds to an incore
1323 * expiration in 2514.  This is beyond the maximum the bigtime inode timestamp,
1324 * so we cap the maximum bigtime quota expiration to the max inode timestamp.
1325 */
1326#define XFS_DQ_BIGTIME_EXPIRY_MAX	((int64_t)4074815106U)
1327
1328/*
1329 * The following conversion factors assist in converting a quota expiration
1330 * timestamp between the incore and ondisk formats.
1331 */
1332#define XFS_DQ_BIGTIME_SHIFT	(2)
1333#define XFS_DQ_BIGTIME_SLACK	((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1)
1334
1335/* Convert an incore quota expiration timestamp to an ondisk bigtime value. */
1336static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds)
1337{
1338	/*
1339	 * Round the expiration timestamp up to the nearest bigtime timestamp
1340	 * that we can store, to give users the most time to fix problems.
1341	 */
1342	return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >>
1343			XFS_DQ_BIGTIME_SHIFT;
1344}
1345
1346/* Convert an ondisk bigtime quota expiration value to an incore timestamp. */
1347static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)
1348{
1349	return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT;
1350}
1351
1352/*
1353 * Default quota grace periods, ranging from zero (use the compiled defaults)
1354 * to ~136 years.  These are applied to a non-root dquot that has exceeded
1355 * either limit.
1356 */
1357#define XFS_DQ_GRACE_MIN		((int64_t)0)
1358#define XFS_DQ_GRACE_MAX		((int64_t)U32_MAX)
1359
1360/*
1361 * This is the main portion of the on-disk representation of quota information
1362 * for a user.  We pad this with some more expansion room to construct the on
1363 * disk structure.
1364 */
1365struct xfs_disk_dquot {
1366	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1367	__u8		d_version;	/* dquot version */
1368	__u8		d_type;		/* XFS_DQTYPE_USER/PROJ/GROUP */
1369	__be32		d_id;		/* user,project,group id */
1370	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1371	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1372	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1373	__be64		d_ino_softlimit;/* preferred inode limit */
1374	__be64		d_bcount;	/* disk blocks owned by the user */
1375	__be64		d_icount;	/* inodes owned by the user */
1376	__be32		d_itimer;	/* zero if within inode limits if not,
1377					   this is when we refuse service */
1378	__be32		d_btimer;	/* similar to above; for disk blocks */
1379	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1380	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1381	__be32		d_pad0;		/* 64 bit align */
1382	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1383	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1384	__be64		d_rtbcount;	/* realtime blocks owned */
1385	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1386	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1387	__be16		d_pad;
1388};
1389
1390/*
1391 * This is what goes on disk. This is separated from the xfs_disk_dquot because
1392 * carrying the unnecessary padding would be a waste of memory.
1393 */
1394typedef struct xfs_dqblk {
1395	struct xfs_disk_dquot	dd_diskdq; /* portion living incore as well */
1396	char			dd_fill[4];/* filling for posterity */
1397
1398	/*
1399	 * These two are only present on filesystems with the CRC bits set.
1400	 */
1401	__be32		  dd_crc;	/* checksum */
1402	__be64		  dd_lsn;	/* last modification in log */
1403	uuid_t		  dd_uuid;	/* location information */
1404} xfs_dqblk_t;
1405
1406#define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1407
1408/*
1409 * This defines the unit of allocation of dquots.
1410 *
1411 * Currently, it is just one file system block, and a 4K blk contains 30
1412 * (136 * 30 = 4080) dquots. It's probably not worth trying to make
1413 * this more dynamic.
1414 *
1415 * However, if this number is changed, we have to make sure that we don't
1416 * implicitly assume that we do allocations in chunks of a single filesystem
1417 * block in the dquot/xqm code.
1418 *
1419 * This is part of the ondisk format because the structure size is not a power
1420 * of two, which leaves slack at the end of the disk block.
1421 */
1422#define XFS_DQUOT_CLUSTER_SIZE_FSB	(xfs_filblks_t)1
1423
1424/*
1425 * Remote symlink format and access functions.
1426 */
1427#define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1428
1429struct xfs_dsymlink_hdr {
1430	__be32	sl_magic;
1431	__be32	sl_offset;
1432	__be32	sl_bytes;
1433	__be32	sl_crc;
1434	uuid_t	sl_uuid;
1435	__be64	sl_owner;
1436	__be64	sl_blkno;
1437	__be64	sl_lsn;
1438};
1439
1440#define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1441
1442#define XFS_SYMLINK_MAXLEN	1024
1443/*
1444 * The maximum pathlen is 1024 bytes. Since the minimum file system
1445 * blocksize is 512 bytes, we can get a max of 3 extents back from
1446 * bmapi when crc headers are taken into account.
1447 */
1448#define XFS_SYMLINK_MAPS 3
1449
1450#define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1451	((bufsize) - (xfs_sb_version_hascrc(&(mp)->m_sb) ? \
1452			sizeof(struct xfs_dsymlink_hdr) : 0))
1453
1454
1455/*
1456 * Allocation Btree format definitions
1457 *
1458 * There are two on-disk btrees, one sorted by blockno and one sorted
1459 * by blockcount and blockno.  All blocks look the same to make the code
1460 * simpler; if we have time later, we'll make the optimizations.
1461 */
1462#define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1463#define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1464#define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1465#define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1466
1467/*
1468 * Data record/key structure
1469 */
1470typedef struct xfs_alloc_rec {
1471	__be32		ar_startblock;	/* starting block number */
1472	__be32		ar_blockcount;	/* count of free blocks */
1473} xfs_alloc_rec_t, xfs_alloc_key_t;
1474
1475typedef struct xfs_alloc_rec_incore {
1476	xfs_agblock_t	ar_startblock;	/* starting block number */
1477	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1478} xfs_alloc_rec_incore_t;
1479
1480/* btree pointer type */
1481typedef __be32 xfs_alloc_ptr_t;
1482
1483/*
1484 * Block numbers in the AG:
1485 * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1486 */
1487#define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1488#define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1489
1490
1491/*
1492 * Inode Allocation Btree format definitions
1493 *
1494 * There is a btree for the inode map per allocation group.
1495 */
1496#define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1497#define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1498#define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1499#define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1500
1501typedef uint64_t	xfs_inofree_t;
1502#define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1503#define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1504#define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1505#define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1506
1507#define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1508#define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(uint16_t))
1509#define XFS_INODES_PER_HOLEMASK_BIT	\
1510	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1511
1512static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1513{
1514	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1515}
1516
1517/*
1518 * The on-disk inode record structure has two formats. The original "full"
1519 * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1520 * and replaces the 3 high-order freecount bytes wth the holemask and inode
1521 * count.
1522 *
1523 * The holemask of the sparse record format allows an inode chunk to have holes
1524 * that refer to blocks not owned by the inode record. This facilitates inode
1525 * allocation in the event of severe free space fragmentation.
1526 */
1527typedef struct xfs_inobt_rec {
1528	__be32		ir_startino;	/* starting inode number */
1529	union {
1530		struct {
1531			__be32	ir_freecount;	/* count of free inodes */
1532		} f;
1533		struct {
1534			__be16	ir_holemask;/* hole mask for sparse chunks */
1535			__u8	ir_count;	/* total inode count */
1536			__u8	ir_freecount;	/* count of free inodes */
1537		} sp;
1538	} ir_u;
1539	__be64		ir_free;	/* free inode mask */
1540} xfs_inobt_rec_t;
1541
1542typedef struct xfs_inobt_rec_incore {
1543	xfs_agino_t	ir_startino;	/* starting inode number */
1544	uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1545	uint8_t		ir_count;	/* total inode count */
1546	uint8_t		ir_freecount;	/* count of free inodes (set bits) */
1547	xfs_inofree_t	ir_free;	/* free inode mask */
1548} xfs_inobt_rec_incore_t;
1549
1550static inline bool xfs_inobt_issparse(uint16_t holemask)
1551{
1552	/* non-zero holemask represents a sparse rec. */
1553	return holemask;
1554}
1555
1556/*
1557 * Key structure
1558 */
1559typedef struct xfs_inobt_key {
1560	__be32		ir_startino;	/* starting inode number */
1561} xfs_inobt_key_t;
1562
1563/* btree pointer type */
1564typedef __be32 xfs_inobt_ptr_t;
1565
1566/*
1567 * block numbers in the AG.
1568 */
1569#define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1570#define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1571
1572/*
1573 * Reverse mapping btree format definitions
1574 *
1575 * There is a btree for the reverse map per allocation group
1576 */
1577#define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1578
1579/*
1580 * Ownership info for an extent.  This is used to create reverse-mapping
1581 * entries.
1582 */
1583#define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1584#define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1585struct xfs_owner_info {
1586	uint64_t		oi_owner;
1587	xfs_fileoff_t		oi_offset;
1588	unsigned int		oi_flags;
1589};
1590
1591/*
1592 * Special owner types.
1593 *
1594 * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1595 * to tell us we have a special owner value. We use these for static metadata
1596 * allocated at mkfs/growfs time, as well as for freespace management metadata.
1597 */
1598#define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1599#define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1600#define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1601#define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1602#define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1603#define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1604#define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1605#define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1606#define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1607#define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1608
1609#define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1610
1611/*
1612 * Data record structure
1613 */
1614struct xfs_rmap_rec {
1615	__be32		rm_startblock;	/* extent start block */
1616	__be32		rm_blockcount;	/* extent length */
1617	__be64		rm_owner;	/* extent owner */
1618	__be64		rm_offset;	/* offset within the owner */
1619};
1620
1621/*
1622 * rmap btree record
1623 *  rm_offset:63 is the attribute fork flag
1624 *  rm_offset:62 is the bmbt block flag
1625 *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1626 *  rm_offset:54-60 aren't used and should be zero
1627 *  rm_offset:0-53 is the block offset within the inode
1628 */
1629#define XFS_RMAP_OFF_ATTR_FORK	((uint64_t)1ULL << 63)
1630#define XFS_RMAP_OFF_BMBT_BLOCK	((uint64_t)1ULL << 62)
1631#define XFS_RMAP_OFF_UNWRITTEN	((uint64_t)1ULL << 61)
1632
1633#define XFS_RMAP_LEN_MAX	((uint32_t)~0U)
1634#define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1635				 XFS_RMAP_OFF_BMBT_BLOCK | \
1636				 XFS_RMAP_OFF_UNWRITTEN)
1637#define XFS_RMAP_OFF_MASK	((uint64_t)0x3FFFFFFFFFFFFFULL)
1638
1639#define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1640
1641#define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1642#define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1643#define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1644
1645#define RMAPBT_STARTBLOCK_BITLEN	32
1646#define RMAPBT_BLOCKCOUNT_BITLEN	32
1647#define RMAPBT_OWNER_BITLEN		64
1648#define RMAPBT_ATTRFLAG_BITLEN		1
1649#define RMAPBT_BMBTFLAG_BITLEN		1
1650#define RMAPBT_EXNTFLAG_BITLEN		1
1651#define RMAPBT_UNUSED_OFFSET_BITLEN	7
1652#define RMAPBT_OFFSET_BITLEN		54
1653
1654#define XFS_RMAP_ATTR_FORK		(1 << 0)
1655#define XFS_RMAP_BMBT_BLOCK		(1 << 1)
1656#define XFS_RMAP_UNWRITTEN		(1 << 2)
1657#define XFS_RMAP_KEY_FLAGS		(XFS_RMAP_ATTR_FORK | \
1658					 XFS_RMAP_BMBT_BLOCK)
1659#define XFS_RMAP_REC_FLAGS		(XFS_RMAP_UNWRITTEN)
1660struct xfs_rmap_irec {
1661	xfs_agblock_t	rm_startblock;	/* extent start block */
1662	xfs_extlen_t	rm_blockcount;	/* extent length */
1663	uint64_t	rm_owner;	/* extent owner */
1664	uint64_t	rm_offset;	/* offset within the owner */
1665	unsigned int	rm_flags;	/* state flags */
1666};
1667
1668/*
1669 * Key structure
1670 *
1671 * We don't use the length for lookups
1672 */
1673struct xfs_rmap_key {
1674	__be32		rm_startblock;	/* extent start block */
1675	__be64		rm_owner;	/* extent owner */
1676	__be64		rm_offset;	/* offset within the owner */
1677} __attribute__((packed));
1678
1679/* btree pointer type */
1680typedef __be32 xfs_rmap_ptr_t;
1681
1682#define	XFS_RMAP_BLOCK(mp) \
1683	(xfs_sb_version_hasfinobt(&((mp)->m_sb)) ? \
1684	 XFS_FIBT_BLOCK(mp) + 1 : \
1685	 XFS_IBT_BLOCK(mp) + 1)
1686
1687/*
1688 * Reference Count Btree format definitions
1689 *
1690 */
1691#define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1692
1693unsigned int xfs_refc_block(struct xfs_mount *mp);
1694
1695/*
1696 * Data record/key structure
1697 *
1698 * Each record associates a range of physical blocks (starting at
1699 * rc_startblock and ending rc_blockcount blocks later) with a reference
1700 * count (rc_refcount).  Extents that are being used to stage a copy on
1701 * write (CoW) operation are recorded in the refcount btree with a
1702 * refcount of 1.  All other records must have a refcount > 1 and must
1703 * track an extent mapped only by file data forks.
1704 *
1705 * Extents with a single owner (attributes, metadata, non-shared file
1706 * data) are not tracked here.  Free space is also not tracked here.
1707 * This is consistent with pre-reflink XFS.
1708 */
1709
1710/*
1711 * Extents that are being used to stage a copy on write are stored
1712 * in the refcount btree with a refcount of 1 and the upper bit set
1713 * on the startblock.  This speeds up mount time deletion of stale
1714 * staging extents because they're all at the right side of the tree.
1715 */
1716#define XFS_REFC_COW_START		((xfs_agblock_t)(1U << 31))
1717#define REFCNTBT_COWFLAG_BITLEN		1
1718#define REFCNTBT_AGBLOCK_BITLEN		31
1719
1720struct xfs_refcount_rec {
1721	__be32		rc_startblock;	/* starting block number */
1722	__be32		rc_blockcount;	/* count of blocks */
1723	__be32		rc_refcount;	/* number of inodes linked here */
1724};
1725
1726struct xfs_refcount_key {
1727	__be32		rc_startblock;	/* starting block number */
1728};
1729
1730struct xfs_refcount_irec {
1731	xfs_agblock_t	rc_startblock;	/* starting block number */
1732	xfs_extlen_t	rc_blockcount;	/* count of free blocks */
1733	xfs_nlink_t	rc_refcount;	/* number of inodes linked here */
1734};
1735
1736#define MAXREFCOUNT	((xfs_nlink_t)~0U)
1737#define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
1738
1739/* btree pointer type */
1740typedef __be32 xfs_refcount_ptr_t;
1741
1742
1743/*
1744 * BMAP Btree format definitions
1745 *
1746 * This includes both the root block definition that sits inside an inode fork
1747 * and the record/pointer formats for the leaf/node in the blocks.
1748 */
1749#define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1750#define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1751
1752/*
1753 * Bmap root header, on-disk form only.
1754 */
1755typedef struct xfs_bmdr_block {
1756	__be16		bb_level;	/* 0 is a leaf */
1757	__be16		bb_numrecs;	/* current # of data records */
1758} xfs_bmdr_block_t;
1759
1760/*
1761 * Bmap btree record and extent descriptor.
1762 *  l0:63 is an extent flag (value 1 indicates non-normal).
1763 *  l0:9-62 are startoff.
1764 *  l0:0-8 and l1:21-63 are startblock.
1765 *  l1:0-20 are blockcount.
1766 */
1767#define BMBT_EXNTFLAG_BITLEN	1
1768#define BMBT_STARTOFF_BITLEN	54
1769#define BMBT_STARTBLOCK_BITLEN	52
1770#define BMBT_BLOCKCOUNT_BITLEN	21
1771
1772#define BMBT_STARTOFF_MASK	((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1773#define BMBT_BLOCKCOUNT_MASK	((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1)
1774
1775/*
1776 * bmbt records have a file offset (block) field that is 54 bits wide, so this
1777 * is the largest xfs_fileoff_t that we ever expect to see.
1778 */
1779#define XFS_MAX_FILEOFF		(BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK)
1780
1781typedef struct xfs_bmbt_rec {
1782	__be64			l0, l1;
1783} xfs_bmbt_rec_t;
1784
1785typedef uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1786typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1787
1788/*
1789 * Values and macros for delayed-allocation startblock fields.
1790 */
1791#define STARTBLOCKVALBITS	17
1792#define STARTBLOCKMASKBITS	(15 + 20)
1793#define STARTBLOCKMASK		\
1794	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1795
1796static inline int isnullstartblock(xfs_fsblock_t x)
1797{
1798	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1799}
1800
1801static inline xfs_fsblock_t nullstartblock(int k)
1802{
1803	ASSERT(k < (1 << STARTBLOCKVALBITS));
1804	return STARTBLOCKMASK | (k);
1805}
1806
1807static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1808{
1809	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1810}
1811
1812/*
1813 * Key structure for non-leaf levels of the tree.
1814 */
1815typedef struct xfs_bmbt_key {
1816	__be64		br_startoff;	/* starting file offset */
1817} xfs_bmbt_key_t, xfs_bmdr_key_t;
1818
1819/* btree pointer type */
1820typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1821
1822
1823/*
1824 * Generic Btree block format definitions
1825 *
1826 * This is a combination of the actual format used on disk for short and long
1827 * format btrees.  The first three fields are shared by both format, but the
1828 * pointers are different and should be used with care.
1829 *
1830 * To get the size of the actual short or long form headers please use the size
1831 * macros below.  Never use sizeof(xfs_btree_block).
1832 *
1833 * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1834 * with the crc feature bit, and all accesses to them must be conditional on
1835 * that flag.
1836 */
1837/* short form block header */
1838struct xfs_btree_block_shdr {
1839	__be32		bb_leftsib;
1840	__be32		bb_rightsib;
1841
1842	__be64		bb_blkno;
1843	__be64		bb_lsn;
1844	uuid_t		bb_uuid;
1845	__be32		bb_owner;
1846	__le32		bb_crc;
1847};
1848
1849/* long form block header */
1850struct xfs_btree_block_lhdr {
1851	__be64		bb_leftsib;
1852	__be64		bb_rightsib;
1853
1854	__be64		bb_blkno;
1855	__be64		bb_lsn;
1856	uuid_t		bb_uuid;
1857	__be64		bb_owner;
1858	__le32		bb_crc;
1859	__be32		bb_pad; /* padding for alignment */
1860};
1861
1862struct xfs_btree_block {
1863	__be32		bb_magic;	/* magic number for block type */
1864	__be16		bb_level;	/* 0 is a leaf */
1865	__be16		bb_numrecs;	/* current # of data records */
1866	union {
1867		struct xfs_btree_block_shdr s;
1868		struct xfs_btree_block_lhdr l;
1869	} bb_u;				/* rest */
1870};
1871
1872/* size of a short form block */
1873#define XFS_BTREE_SBLOCK_LEN \
1874	(offsetof(struct xfs_btree_block, bb_u) + \
1875	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1876/* size of a long form block */
1877#define XFS_BTREE_LBLOCK_LEN \
1878	(offsetof(struct xfs_btree_block, bb_u) + \
1879	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1880
1881/* sizes of CRC enabled btree blocks */
1882#define XFS_BTREE_SBLOCK_CRC_LEN \
1883	(offsetof(struct xfs_btree_block, bb_u) + \
1884	 sizeof(struct xfs_btree_block_shdr))
1885#define XFS_BTREE_LBLOCK_CRC_LEN \
1886	(offsetof(struct xfs_btree_block, bb_u) + \
1887	 sizeof(struct xfs_btree_block_lhdr))
1888
1889#define XFS_BTREE_SBLOCK_CRC_OFF \
1890	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1891#define XFS_BTREE_LBLOCK_CRC_OFF \
1892	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1893
1894/*
1895 * On-disk XFS access control list structure.
1896 */
1897struct xfs_acl_entry {
1898	__be32	ae_tag;
1899	__be32	ae_id;
1900	__be16	ae_perm;
1901	__be16	ae_pad;		/* fill the implicit hole in the structure */
1902};
1903
1904struct xfs_acl {
1905	__be32			acl_cnt;
1906	struct xfs_acl_entry	acl_entry[];
1907};
1908
1909/*
1910 * The number of ACL entries allowed is defined by the on-disk format.
1911 * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1912 * limited only by the maximum size of the xattr that stores the information.
1913 */
1914#define XFS_ACL_MAX_ENTRIES(mp)	\
1915	(xfs_sb_version_hascrc(&mp->m_sb) \
1916		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1917						sizeof(struct xfs_acl_entry) \
1918		: 25)
1919
1920#define XFS_ACL_SIZE(cnt) \
1921	(sizeof(struct xfs_acl) + \
1922		sizeof(struct xfs_acl_entry) * cnt)
1923
1924#define XFS_ACL_MAX_SIZE(mp) \
1925	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1926
1927
1928/* On-disk XFS extended attribute names */
1929#define SGI_ACL_FILE		"SGI_ACL_FILE"
1930#define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
1931#define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
1932#define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
1933
1934#endif /* __XFS_FORMAT_H__ */
1935