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 */
187struct 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};
267
268/*
269 * Misc. Flags - warning - these will be cleared by xfs_repair unless
270 * a feature bit is set when the flag is used.
271 */
272#define XFS_SBF_NOFLAGS		0x00	/* no flags set */
273#define XFS_SBF_READONLY	0x01	/* only read-only mounts allowed */
274
275/*
276 * define max. shared version we can interoperate with
277 */
278#define XFS_SB_MAX_SHARED_VN	0
279
280#define	XFS_SB_VERSION_NUM(sbp)	((sbp)->sb_versionnum & XFS_SB_VERSION_NUMBITS)
281
282static inline bool xfs_sb_is_v5(struct xfs_sb *sbp)
283{
284	return XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5;
285}
286
287/*
288 * Detect a mismatched features2 field.  Older kernels read/wrote
289 * this into the wrong slot, so to be safe we keep them in sync.
290 */
291static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
292{
293	return sbp->sb_bad_features2 != sbp->sb_features2;
294}
295
296static inline bool xfs_sb_version_hasmorebits(struct xfs_sb *sbp)
297{
298	return xfs_sb_is_v5(sbp) ||
299	       (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT);
300}
301
302static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
303{
304	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
305}
306
307static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
308{
309	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
310}
311
312static inline void xfs_sb_version_addattr2(struct xfs_sb *sbp)
313{
314	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
315	sbp->sb_features2 |= XFS_SB_VERSION2_ATTR2BIT;
316}
317
318static inline void xfs_sb_version_addprojid32(struct xfs_sb *sbp)
319{
320	sbp->sb_versionnum |= XFS_SB_VERSION_MOREBITSBIT;
321	sbp->sb_features2 |= XFS_SB_VERSION2_PROJID32BIT;
322}
323
324/*
325 * Extended v5 superblock feature masks. These are to be used for new v5
326 * superblock features only.
327 *
328 * Compat features are new features that old kernels will not notice or affect
329 * and so can mount read-write without issues.
330 *
331 * RO-Compat (read only) are features that old kernels can read but will break
332 * if they write. Hence only read-only mounts of such filesystems are allowed on
333 * kernels that don't support the feature bit.
334 *
335 * InCompat features are features which old kernels will not understand and so
336 * must not mount.
337 *
338 * Log-InCompat features are for changes to log formats or new transactions that
339 * can't be replayed on older kernels. The fields are set when the filesystem is
340 * mounted, and a clean unmount clears the fields.
341 */
342#define XFS_SB_FEAT_COMPAT_ALL 0
343#define XFS_SB_FEAT_COMPAT_UNKNOWN	~XFS_SB_FEAT_COMPAT_ALL
344static inline bool
345xfs_sb_has_compat_feature(
346	struct xfs_sb	*sbp,
347	uint32_t	feature)
348{
349	return (sbp->sb_features_compat & feature) != 0;
350}
351
352#define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
353#define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
354#define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
355#define XFS_SB_FEAT_RO_COMPAT_INOBTCNT (1 << 3)		/* inobt block counts */
356#define XFS_SB_FEAT_RO_COMPAT_ALL \
357		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
358		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
359		 XFS_SB_FEAT_RO_COMPAT_REFLINK| \
360		 XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
361#define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
362static inline bool
363xfs_sb_has_ro_compat_feature(
364	struct xfs_sb	*sbp,
365	uint32_t	feature)
366{
367	return (sbp->sb_features_ro_compat & feature) != 0;
368}
369
370#define XFS_SB_FEAT_INCOMPAT_FTYPE	(1 << 0)	/* filetype in dirent */
371#define XFS_SB_FEAT_INCOMPAT_SPINODES	(1 << 1)	/* sparse inode chunks */
372#define XFS_SB_FEAT_INCOMPAT_META_UUID	(1 << 2)	/* metadata UUID */
373#define XFS_SB_FEAT_INCOMPAT_BIGTIME	(1 << 3)	/* large timestamps */
374#define XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR (1 << 4)	/* needs xfs_repair */
375#define XFS_SB_FEAT_INCOMPAT_NREXT64	(1 << 5)	/* large extent counters */
376#define XFS_SB_FEAT_INCOMPAT_ALL \
377		(XFS_SB_FEAT_INCOMPAT_FTYPE|	\
378		 XFS_SB_FEAT_INCOMPAT_SPINODES|	\
379		 XFS_SB_FEAT_INCOMPAT_META_UUID| \
380		 XFS_SB_FEAT_INCOMPAT_BIGTIME| \
381		 XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR| \
382		 XFS_SB_FEAT_INCOMPAT_NREXT64)
383
384#define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
385static inline bool
386xfs_sb_has_incompat_feature(
387	struct xfs_sb	*sbp,
388	uint32_t	feature)
389{
390	return (sbp->sb_features_incompat & feature) != 0;
391}
392
393#define XFS_SB_FEAT_INCOMPAT_LOG_XATTRS   (1 << 0)	/* Delayed Attributes */
394#define XFS_SB_FEAT_INCOMPAT_LOG_ALL \
395	(XFS_SB_FEAT_INCOMPAT_LOG_XATTRS)
396#define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
397static inline bool
398xfs_sb_has_incompat_log_feature(
399	struct xfs_sb	*sbp,
400	uint32_t	feature)
401{
402	return (sbp->sb_features_log_incompat & feature) != 0;
403}
404
405static inline void
406xfs_sb_remove_incompat_log_features(
407	struct xfs_sb	*sbp)
408{
409	sbp->sb_features_log_incompat &= ~XFS_SB_FEAT_INCOMPAT_LOG_ALL;
410}
411
412static inline void
413xfs_sb_add_incompat_log_features(
414	struct xfs_sb	*sbp,
415	unsigned int	features)
416{
417	sbp->sb_features_log_incompat |= features;
418}
419
420static inline bool xfs_sb_version_haslogxattrs(struct xfs_sb *sbp)
421{
422	return xfs_sb_is_v5(sbp) && (sbp->sb_features_log_incompat &
423		 XFS_SB_FEAT_INCOMPAT_LOG_XATTRS);
424}
425
426static inline bool
427xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
428{
429	return (ino == sbp->sb_uquotino ||
430		ino == sbp->sb_gquotino ||
431		ino == sbp->sb_pquotino);
432}
433
434#define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
435#define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
436
437#define	XFS_HDR_BLOCK(mp,d)	((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
438#define	XFS_DADDR_TO_FSB(mp,d)	XFS_AGB_TO_FSB(mp, \
439			xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
440#define	XFS_FSB_TO_DADDR(mp,fsbno)	XFS_AGB_TO_DADDR(mp, \
441			XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
442
443/*
444 * File system sector to basic block conversions.
445 */
446#define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
447
448/*
449 * File system block to basic block conversions.
450 */
451#define	XFS_FSB_TO_BB(mp,fsbno)	((fsbno) << (mp)->m_blkbb_log)
452#define	XFS_BB_TO_FSB(mp,bb)	\
453	(((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
454#define	XFS_BB_TO_FSBT(mp,bb)	((bb) >> (mp)->m_blkbb_log)
455
456/*
457 * File system block to byte conversions.
458 */
459#define XFS_FSB_TO_B(mp,fsbno)	((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
460#define XFS_B_TO_FSB(mp,b)	\
461	((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
462#define XFS_B_TO_FSBT(mp,b)	(((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
463
464/*
465 * Allocation group header
466 *
467 * This is divided into three structures, placed in sequential 512-byte
468 * buffers after a copy of the superblock (also in a 512-byte buffer).
469 */
470#define	XFS_AGF_MAGIC	0x58414746	/* 'XAGF' */
471#define	XFS_AGI_MAGIC	0x58414749	/* 'XAGI' */
472#define	XFS_AGFL_MAGIC	0x5841464c	/* 'XAFL' */
473#define	XFS_AGF_VERSION	1
474#define	XFS_AGI_VERSION	1
475
476#define	XFS_AGF_GOOD_VERSION(v)	((v) == XFS_AGF_VERSION)
477#define	XFS_AGI_GOOD_VERSION(v)	((v) == XFS_AGI_VERSION)
478
479/*
480 * Btree number 0 is bno, 1 is cnt, 2 is rmap. This value gives the size of the
481 * arrays below.
482 */
483#define	XFS_BTNUM_AGF	((int)XFS_BTNUM_RMAPi + 1)
484
485/*
486 * The second word of agf_levels in the first a.g. overlaps the EFS
487 * superblock's magic number.  Since the magic numbers valid for EFS
488 * are > 64k, our value cannot be confused for an EFS superblock's.
489 */
490
491typedef struct xfs_agf {
492	/*
493	 * Common allocation group header information
494	 */
495	__be32		agf_magicnum;	/* magic number == XFS_AGF_MAGIC */
496	__be32		agf_versionnum;	/* header version == XFS_AGF_VERSION */
497	__be32		agf_seqno;	/* sequence # starting from 0 */
498	__be32		agf_length;	/* size in blocks of a.g. */
499	/*
500	 * Freespace and rmap information
501	 */
502	__be32		agf_roots[XFS_BTNUM_AGF];	/* root blocks */
503	__be32		agf_levels[XFS_BTNUM_AGF];	/* btree levels */
504
505	__be32		agf_flfirst;	/* first freelist block's index */
506	__be32		agf_fllast;	/* last freelist block's index */
507	__be32		agf_flcount;	/* count of blocks in freelist */
508	__be32		agf_freeblks;	/* total free blocks */
509
510	__be32		agf_longest;	/* longest free space */
511	__be32		agf_btreeblks;	/* # of blocks held in AGF btrees */
512	uuid_t		agf_uuid;	/* uuid of filesystem */
513
514	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
515	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
516
517	__be32		agf_refcount_root;	/* refcount tree root block */
518	__be32		agf_refcount_level;	/* refcount btree levels */
519
520	/*
521	 * reserve some contiguous space for future logged fields before we add
522	 * the unlogged fields. This makes the range logging via flags and
523	 * structure offsets much simpler.
524	 */
525	__be64		agf_spare64[14];
526
527	/* unlogged fields, written during buffer writeback. */
528	__be64		agf_lsn;	/* last write sequence */
529	__be32		agf_crc;	/* crc of agf sector */
530	__be32		agf_spare2;
531
532	/* structure must be padded to 64 bit alignment */
533} xfs_agf_t;
534
535#define XFS_AGF_CRC_OFF		offsetof(struct xfs_agf, agf_crc)
536
537#define	XFS_AGF_MAGICNUM	(1u << 0)
538#define	XFS_AGF_VERSIONNUM	(1u << 1)
539#define	XFS_AGF_SEQNO		(1u << 2)
540#define	XFS_AGF_LENGTH		(1u << 3)
541#define	XFS_AGF_ROOTS		(1u << 4)
542#define	XFS_AGF_LEVELS		(1u << 5)
543#define	XFS_AGF_FLFIRST		(1u << 6)
544#define	XFS_AGF_FLLAST		(1u << 7)
545#define	XFS_AGF_FLCOUNT		(1u << 8)
546#define	XFS_AGF_FREEBLKS	(1u << 9)
547#define	XFS_AGF_LONGEST		(1u << 10)
548#define	XFS_AGF_BTREEBLKS	(1u << 11)
549#define	XFS_AGF_UUID		(1u << 12)
550#define	XFS_AGF_RMAP_BLOCKS	(1u << 13)
551#define	XFS_AGF_REFCOUNT_BLOCKS	(1u << 14)
552#define	XFS_AGF_REFCOUNT_ROOT	(1u << 15)
553#define	XFS_AGF_REFCOUNT_LEVEL	(1u << 16)
554#define	XFS_AGF_SPARE64		(1u << 17)
555#define	XFS_AGF_NUM_BITS	18
556#define	XFS_AGF_ALL_BITS	((1u << XFS_AGF_NUM_BITS) - 1)
557
558#define XFS_AGF_FLAGS \
559	{ XFS_AGF_MAGICNUM,	"MAGICNUM" }, \
560	{ XFS_AGF_VERSIONNUM,	"VERSIONNUM" }, \
561	{ XFS_AGF_SEQNO,	"SEQNO" }, \
562	{ XFS_AGF_LENGTH,	"LENGTH" }, \
563	{ XFS_AGF_ROOTS,	"ROOTS" }, \
564	{ XFS_AGF_LEVELS,	"LEVELS" }, \
565	{ XFS_AGF_FLFIRST,	"FLFIRST" }, \
566	{ XFS_AGF_FLLAST,	"FLLAST" }, \
567	{ XFS_AGF_FLCOUNT,	"FLCOUNT" }, \
568	{ XFS_AGF_FREEBLKS,	"FREEBLKS" }, \
569	{ XFS_AGF_LONGEST,	"LONGEST" }, \
570	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
571	{ XFS_AGF_UUID,		"UUID" }, \
572	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
573	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
574	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
575	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
576	{ XFS_AGF_SPARE64,	"SPARE64" }
577
578/* disk block (xfs_daddr_t) in the AG */
579#define XFS_AGF_DADDR(mp)	((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
580#define	XFS_AGF_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
581
582/*
583 * Size of the unlinked inode hash table in the agi.
584 */
585#define	XFS_AGI_UNLINKED_BUCKETS	64
586
587typedef struct xfs_agi {
588	/*
589	 * Common allocation group header information
590	 */
591	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
592	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
593	__be32		agi_seqno;	/* sequence # starting from 0 */
594	__be32		agi_length;	/* size in blocks of a.g. */
595	/*
596	 * Inode information
597	 * Inodes are mapped by interpreting the inode number, so no
598	 * mapping data is needed here.
599	 */
600	__be32		agi_count;	/* count of allocated inodes */
601	__be32		agi_root;	/* root of inode btree */
602	__be32		agi_level;	/* levels in inode btree */
603	__be32		agi_freecount;	/* number of free inodes */
604
605	__be32		agi_newino;	/* new inode just allocated */
606	__be32		agi_dirino;	/* last directory inode chunk */
607	/*
608	 * Hash table of inodes which have been unlinked but are
609	 * still being referenced.
610	 */
611	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
612	/*
613	 * This marks the end of logging region 1 and start of logging region 2.
614	 */
615	uuid_t		agi_uuid;	/* uuid of filesystem */
616	__be32		agi_crc;	/* crc of agi sector */
617	__be32		agi_pad32;
618	__be64		agi_lsn;	/* last write sequence */
619
620	__be32		agi_free_root; /* root of the free inode btree */
621	__be32		agi_free_level;/* levels in free inode btree */
622
623	__be32		agi_iblocks;	/* inobt blocks used */
624	__be32		agi_fblocks;	/* finobt blocks used */
625
626	/* structure must be padded to 64 bit alignment */
627} xfs_agi_t;
628
629#define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
630
631#define	XFS_AGI_MAGICNUM	(1u << 0)
632#define	XFS_AGI_VERSIONNUM	(1u << 1)
633#define	XFS_AGI_SEQNO		(1u << 2)
634#define	XFS_AGI_LENGTH		(1u << 3)
635#define	XFS_AGI_COUNT		(1u << 4)
636#define	XFS_AGI_ROOT		(1u << 5)
637#define	XFS_AGI_LEVEL		(1u << 6)
638#define	XFS_AGI_FREECOUNT	(1u << 7)
639#define	XFS_AGI_NEWINO		(1u << 8)
640#define	XFS_AGI_DIRINO		(1u << 9)
641#define	XFS_AGI_UNLINKED	(1u << 10)
642#define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
643#define	XFS_AGI_ALL_BITS_R1	((1u << XFS_AGI_NUM_BITS_R1) - 1)
644#define	XFS_AGI_FREE_ROOT	(1u << 11)
645#define	XFS_AGI_FREE_LEVEL	(1u << 12)
646#define	XFS_AGI_IBLOCKS		(1u << 13) /* both inobt/finobt block counters */
647#define	XFS_AGI_NUM_BITS_R2	14
648
649/* disk block (xfs_daddr_t) in the AG */
650#define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
651#define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
652
653/*
654 * The third a.g. block contains the a.g. freelist, an array
655 * of block pointers to blocks owned by the allocation btree code.
656 */
657#define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
658#define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
659#define	XFS_BUF_TO_AGFL(bp)	((struct xfs_agfl *)((bp)->b_addr))
660
661struct xfs_agfl {
662	__be32		agfl_magicnum;
663	__be32		agfl_seqno;
664	uuid_t		agfl_uuid;
665	__be64		agfl_lsn;
666	__be32		agfl_crc;
667} __attribute__((packed));
668
669#define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
670
671#define XFS_AGB_TO_FSB(mp,agno,agbno)	\
672	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
673#define	XFS_FSB_TO_AGNO(mp,fsbno)	\
674	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
675#define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
676	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
677#define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
678	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
679		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
680#define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
681
682/*
683 * For checking for bad ranges of xfs_daddr_t's, covering multiple
684 * allocation groups or a single xfs_daddr_t that's a superblock copy.
685 */
686#define	XFS_AG_CHECK_DADDR(mp,d,len)	\
687	((len) == 1 ? \
688	    ASSERT((d) == XFS_SB_DADDR || \
689		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
690	    ASSERT(xfs_daddr_to_agno(mp, d) == \
691		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
692
693/*
694 * XFS Timestamps
695 * ==============
696 *
697 * Traditional ondisk inode timestamps consist of signed 32-bit counters for
698 * seconds and nanoseconds; time zero is the Unix epoch, Jan  1 00:00:00 UTC
699 * 1970, which means that the timestamp epoch is the same as the Unix epoch.
700 * Therefore, the ondisk min and max defined here can be used directly to
701 * constrain the incore timestamps on a Unix system.  Note that we actually
702 * encode a __be64 value on disk.
703 *
704 * When the bigtime feature is enabled, ondisk inode timestamps become an
705 * unsigned 64-bit nanoseconds counter.  This means that the bigtime inode
706 * timestamp epoch is the start of the classic timestamp range, which is
707 * Dec 31 20:45:52 UTC 1901.  Because the epochs are not the same, callers
708 * /must/ use the bigtime conversion functions when encoding and decoding raw
709 * timestamps.
710 */
711typedef __be64 xfs_timestamp_t;
712
713/* Legacy timestamp encoding format. */
714struct xfs_legacy_timestamp {
715	__be32		t_sec;		/* timestamp seconds */
716	__be32		t_nsec;		/* timestamp nanoseconds */
717};
718
719/*
720 * Smallest possible ondisk seconds value with traditional timestamps.  This
721 * corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901.
722 */
723#define XFS_LEGACY_TIME_MIN	((int64_t)S32_MIN)
724
725/*
726 * Largest possible ondisk seconds value with traditional timestamps.  This
727 * corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038.
728 */
729#define XFS_LEGACY_TIME_MAX	((int64_t)S32_MAX)
730
731/*
732 * Smallest possible ondisk seconds value with bigtime timestamps.  This
733 * corresponds (after conversion to a Unix timestamp) with the traditional
734 * minimum timestamp of Dec 13 20:45:52 UTC 1901.
735 */
736#define XFS_BIGTIME_TIME_MIN	((int64_t)0)
737
738/*
739 * Largest supported ondisk seconds value with bigtime timestamps.  This
740 * corresponds (after conversion to a Unix timestamp) with an incore timestamp
741 * of Jul  2 20:20:24 UTC 2486.
742 *
743 * We round down the ondisk limit so that the bigtime quota and inode max
744 * timestamps will be the same.
745 */
746#define XFS_BIGTIME_TIME_MAX	((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL))
747
748/*
749 * Bigtime epoch is set exactly to the minimum time value that a traditional
750 * 32-bit timestamp can represent when using the Unix epoch as a reference.
751 * Hence the Unix epoch is at a fixed offset into the supported bigtime
752 * timestamp range.
753 *
754 * The bigtime epoch also matches the minimum value an on-disk 32-bit XFS
755 * timestamp can represent so we will not lose any fidelity in converting
756 * to/from unix and bigtime timestamps.
757 *
758 * The following conversion factor converts a seconds counter from the Unix
759 * epoch to the bigtime epoch.
760 */
761#define XFS_BIGTIME_EPOCH_OFFSET	(-(int64_t)S32_MIN)
762
763/* Convert a timestamp from the Unix epoch to the bigtime epoch. */
764static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds)
765{
766	return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET;
767}
768
769/* Convert a timestamp from the bigtime epoch to the Unix epoch. */
770static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds)
771{
772	return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET;
773}
774
775/*
776 * On-disk inode structure.
777 *
778 * This is just the header or "dinode core", the inode is expanded to fill a
779 * variable size the leftover area split into a data and an attribute fork.
780 * The format of the data and attribute fork depends on the format of the
781 * inode as indicated by di_format and di_aformat.  To access the data and
782 * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
783 * below.
784 *
785 * There is a very similar struct xfs_log_dinode which matches the layout of
786 * this structure, but is kept in native format instead of big endian.
787 *
788 * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
789 * padding field for v3 inodes.
790 */
791#define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
792struct xfs_dinode {
793	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
794	__be16		di_mode;	/* mode and type of file */
795	__u8		di_version;	/* inode version */
796	__u8		di_format;	/* format of di_c data */
797	__be16		di_onlink;	/* old number of links to file */
798	__be32		di_uid;		/* owner's user id */
799	__be32		di_gid;		/* owner's group id */
800	__be32		di_nlink;	/* number of links to file */
801	__be16		di_projid_lo;	/* lower part of owner's project id */
802	__be16		di_projid_hi;	/* higher part owner's project id */
803	union {
804		/* Number of data fork extents if NREXT64 is set */
805		__be64	di_big_nextents;
806
807		/* Padding for V3 inodes without NREXT64 set. */
808		__be64	di_v3_pad;
809
810		/* Padding and inode flush counter for V2 inodes. */
811		struct {
812			__u8	di_v2_pad[6];
813			__be16	di_flushiter;
814		};
815	};
816	xfs_timestamp_t	di_atime;	/* time last accessed */
817	xfs_timestamp_t	di_mtime;	/* time last modified */
818	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
819	__be64		di_size;	/* number of bytes in file */
820	__be64		di_nblocks;	/* # of direct & btree blocks used */
821	__be32		di_extsize;	/* basic/minimum extent size for file */
822	union {
823		/*
824		 * For V2 inodes and V3 inodes without NREXT64 set, this
825		 * is the number of data and attr fork extents.
826		 */
827		struct {
828			__be32	di_nextents;
829			__be16	di_anextents;
830		} __packed;
831
832		/* Number of attr fork extents if NREXT64 is set. */
833		struct {
834			__be32	di_big_anextents;
835			__be16	di_nrext64_pad;
836		} __packed;
837	} __packed;
838	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
839	__s8		di_aformat;	/* format of attr fork's data */
840	__be32		di_dmevmask;	/* DMIG event mask */
841	__be16		di_dmstate;	/* DMIG state info */
842	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
843	__be32		di_gen;		/* generation number */
844
845	/* di_next_unlinked is the only non-core field in the old dinode */
846	__be32		di_next_unlinked;/* agi unlinked list ptr */
847
848	/* start of the extended dinode, writable fields */
849	__le32		di_crc;		/* CRC of the inode */
850	__be64		di_changecount;	/* number of attribute changes */
851	__be64		di_lsn;		/* flush sequence */
852	__be64		di_flags2;	/* more random flags */
853	__be32		di_cowextsize;	/* basic cow extent size for file */
854	__u8		di_pad2[12];	/* more padding for future expansion */
855
856	/* fields only written to during inode creation */
857	xfs_timestamp_t	di_crtime;	/* time created */
858	__be64		di_ino;		/* inode number */
859	uuid_t		di_uuid;	/* UUID of the filesystem */
860
861	/* structure must be padded to 64 bit alignment */
862};
863
864#define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
865
866#define DI_MAX_FLUSH 0xffff
867
868/*
869 * Size of the core inode on disk.  Version 1 and 2 inodes have
870 * the same size, but version 3 has grown a few additional fields.
871 */
872static inline uint xfs_dinode_size(int version)
873{
874	if (version == 3)
875		return sizeof(struct xfs_dinode);
876	return offsetof(struct xfs_dinode, di_crc);
877}
878
879/*
880 * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
881 * Since the pathconf interface is signed, we use 2^31 - 1 instead.
882 */
883#define	XFS_MAXLINK		((1U << 31) - 1U)
884
885/*
886 * Values for di_format
887 *
888 * This enum is used in string mapping in xfs_trace.h; please keep the
889 * TRACE_DEFINE_ENUMs for it up to date.
890 */
891enum xfs_dinode_fmt {
892	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
893	XFS_DINODE_FMT_LOCAL,		/* bulk data */
894	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
895	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
896	XFS_DINODE_FMT_UUID		/* added long ago, but never used */
897};
898
899#define XFS_INODE_FORMAT_STR \
900	{ XFS_DINODE_FMT_DEV,		"dev" }, \
901	{ XFS_DINODE_FMT_LOCAL,		"local" }, \
902	{ XFS_DINODE_FMT_EXTENTS,	"extent" }, \
903	{ XFS_DINODE_FMT_BTREE,		"btree" }, \
904	{ XFS_DINODE_FMT_UUID,		"uuid" }
905
906/*
907 * Max values for extnum and aextnum.
908 *
909 * The original on-disk extent counts were held in signed fields, resulting in
910 * maximum extent counts of 2^31 and 2^15 for the data and attr forks
911 * respectively. Similarly the maximum extent length is limited to 2^21 blocks
912 * by the 21-bit wide blockcount field of a BMBT extent record.
913 *
914 * The newly introduced data fork extent counter can hold a 64-bit value,
915 * however the maximum number of extents in a file is also limited to 2^54
916 * extents by the 54-bit wide startoff field of a BMBT extent record.
917 *
918 * It is further limited by the maximum supported file size of 2^63
919 * *bytes*. This leads to a maximum extent count for maximally sized filesystem
920 * blocks (64kB) of:
921 *
922 * 2^63 bytes / 2^16 bytes per block = 2^47 blocks
923 *
924 * Rounding up 47 to the nearest multiple of bits-per-byte results in 48. Hence
925 * 2^48 was chosen as the maximum data fork extent count.
926 *
927 * The maximum file size that can be represented by the data fork extent counter
928 * in the worst case occurs when all extents are 1 block in length and each
929 * block is 1KB in size.
930 *
931 * With XFS_MAX_EXTCNT_DATA_FORK_SMALL representing maximum extent count and
932 * with 1KB sized blocks, a file can reach upto,
933 * 1KB * (2^31) = 2TB
934 *
935 * This is much larger than the theoretical maximum size of a directory
936 * i.e. XFS_DIR2_SPACE_SIZE * XFS_DIR2_MAX_SPACES = ~96GB.
937 *
938 * Hence, a directory inode can never overflow its data fork extent counter.
939 */
940#define XFS_MAX_EXTCNT_DATA_FORK_LARGE	((xfs_extnum_t)((1ULL << 48) - 1))
941#define XFS_MAX_EXTCNT_ATTR_FORK_LARGE	((xfs_extnum_t)((1ULL << 32) - 1))
942#define XFS_MAX_EXTCNT_DATA_FORK_SMALL	((xfs_extnum_t)((1ULL << 31) - 1))
943#define XFS_MAX_EXTCNT_ATTR_FORK_SMALL	((xfs_extnum_t)((1ULL << 15) - 1))
944
945/*
946 * When we upgrade an inode to the large extent counts, the maximum value by
947 * which the extent count can increase is bound by the change in size of the
948 * on-disk field. No upgrade operation should ever be adding more than a few
949 * tens of extents, so if we get a really large value it is a sign of a code bug
950 * or corruption.
951 */
952#define XFS_MAX_EXTCNT_UPGRADE_NR	\
953	min(XFS_MAX_EXTCNT_ATTR_FORK_LARGE - XFS_MAX_EXTCNT_ATTR_FORK_SMALL,	\
954	    XFS_MAX_EXTCNT_DATA_FORK_LARGE - XFS_MAX_EXTCNT_DATA_FORK_SMALL)
955
956/*
957 * Inode minimum and maximum sizes.
958 */
959#define	XFS_DINODE_MIN_LOG	8
960#define	XFS_DINODE_MAX_LOG	11
961#define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
962#define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
963
964/*
965 * Inode size for given fs.
966 */
967#define XFS_DINODE_SIZE(mp) \
968	(xfs_has_v3inodes(mp) ? \
969		sizeof(struct xfs_dinode) : \
970		offsetof(struct xfs_dinode, di_crc))
971#define XFS_LITINO(mp) \
972	((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(mp))
973
974/*
975 * Inode data & attribute fork sizes, per inode.
976 */
977#define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
978
979#define XFS_DFORK_DSIZE(dip,mp) \
980	((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp))
981#define XFS_DFORK_ASIZE(dip,mp) \
982	((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0)
983#define XFS_DFORK_SIZE(dip,mp,w) \
984	((w) == XFS_DATA_FORK ? \
985		XFS_DFORK_DSIZE(dip, mp) : \
986		XFS_DFORK_ASIZE(dip, mp))
987
988#define XFS_DFORK_MAXEXT(dip, mp, w) \
989	(XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
990
991/*
992 * Return pointers to the data or attribute forks.
993 */
994#define XFS_DFORK_DPTR(dip) \
995	((char *)dip + xfs_dinode_size(dip->di_version))
996#define XFS_DFORK_APTR(dip)	\
997	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
998#define XFS_DFORK_PTR(dip,w)	\
999	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
1000
1001#define XFS_DFORK_FORMAT(dip,w) \
1002	((w) == XFS_DATA_FORK ? \
1003		(dip)->di_format : \
1004		(dip)->di_aformat)
1005
1006/*
1007 * For block and character special files the 32bit dev_t is stored at the
1008 * beginning of the data fork.
1009 */
1010static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1011{
1012	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1013}
1014
1015static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1016{
1017	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1018}
1019
1020/*
1021 * Values for di_flags
1022 */
1023#define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
1024#define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
1025#define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
1026#define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
1027#define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
1028#define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
1029#define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
1030#define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
1031#define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
1032#define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
1033#define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
1034#define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
1035#define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
1036#define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
1037#define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
1038/* Do not use bit 15, di_flags is legacy and unchanging now */
1039
1040#define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
1041#define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
1042#define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
1043#define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1044#define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
1045#define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
1046#define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
1047#define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
1048#define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
1049#define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1050#define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1051#define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
1052#define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1053#define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
1054#define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
1055
1056#define XFS_DIFLAG_ANY \
1057	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1058	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1059	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1060	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1061	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1062
1063/*
1064 * Values for di_flags2 These start by being exposed to userspace in the upper
1065 * 16 bits of the XFS_XFLAG_s range.
1066 */
1067#define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
1068#define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
1069#define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
1070#define XFS_DIFLAG2_BIGTIME_BIT	3	/* big timestamps */
1071#define XFS_DIFLAG2_NREXT64_BIT 4	/* large extent counters */
1072
1073#define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
1074#define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
1075#define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1076#define XFS_DIFLAG2_BIGTIME	(1 << XFS_DIFLAG2_BIGTIME_BIT)
1077#define XFS_DIFLAG2_NREXT64	(1 << XFS_DIFLAG2_NREXT64_BIT)
1078
1079#define XFS_DIFLAG2_ANY \
1080	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \
1081	 XFS_DIFLAG2_BIGTIME | XFS_DIFLAG2_NREXT64)
1082
1083static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip)
1084{
1085	return dip->di_version >= 3 &&
1086	       (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME));
1087}
1088
1089static inline bool xfs_dinode_has_large_extent_counts(
1090	const struct xfs_dinode *dip)
1091{
1092	return dip->di_version >= 3 &&
1093	       (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_NREXT64));
1094}
1095
1096/*
1097 * Inode number format:
1098 * low inopblog bits - offset in block
1099 * next agblklog bits - block number in ag
1100 * next agno_log bits - ag number
1101 * high agno_log-agblklog-inopblog bits - 0
1102 */
1103#define	XFS_INO_MASK(k)			(uint32_t)((1ULL << (k)) - 1)
1104#define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1105#define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1106#define	XFS_INO_AGINO_BITS(mp)		((mp)->m_ino_geo.agino_log)
1107#define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1108#define	XFS_INO_BITS(mp)		\
1109	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1110#define	XFS_INO_TO_AGNO(mp,i)		\
1111	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1112#define	XFS_INO_TO_AGINO(mp,i)		\
1113	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1114#define	XFS_INO_TO_AGBNO(mp,i)		\
1115	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1116		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1117#define	XFS_INO_TO_OFFSET(mp,i)		\
1118	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1119#define	XFS_INO_TO_FSB(mp,i)		\
1120	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1121#define	XFS_AGINO_TO_INO(mp,a,i)	\
1122	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1123#define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1124#define	XFS_AGINO_TO_OFFSET(mp,i)	\
1125	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1126#define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1127	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1128#define	XFS_FSB_TO_INO(mp, b)	((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1129#define	XFS_AGB_TO_AGINO(mp, b)	((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1130
1131#define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1132#define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1133
1134/*
1135 * RealTime Device format definitions
1136 */
1137
1138/* Min and max rt extent sizes, specified in bytes */
1139#define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1140#define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1141#define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1142
1143#define	XFS_BLOCKSIZE(mp)	((mp)->m_sb.sb_blocksize)
1144#define	XFS_BLOCKMASK(mp)	((mp)->m_blockmask)
1145#define	XFS_BLOCKWSIZE(mp)	((mp)->m_blockwsize)
1146#define	XFS_BLOCKWMASK(mp)	((mp)->m_blockwmask)
1147
1148/*
1149 * RT Summary and bit manipulation macros.
1150 */
1151#define	XFS_SUMOFFS(mp,ls,bb)	((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb)))
1152#define	XFS_SUMOFFSTOBLOCK(mp,s)	\
1153	(((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog)
1154#define	XFS_SUMPTR(mp,bp,so)	\
1155	((xfs_suminfo_t *)((bp)->b_addr + \
1156		(((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp))))
1157
1158#define	XFS_BITTOBLOCK(mp,bi)	((bi) >> (mp)->m_blkbit_log)
1159#define	XFS_BLOCKTOBIT(mp,bb)	((bb) << (mp)->m_blkbit_log)
1160#define	XFS_BITTOWORD(mp,bi)	\
1161	((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp)))
1162
1163#define	XFS_RTMIN(a,b)	((a) < (b) ? (a) : (b))
1164#define	XFS_RTMAX(a,b)	((a) > (b) ? (a) : (b))
1165
1166#define	XFS_RTLOBIT(w)	xfs_lowbit32(w)
1167#define	XFS_RTHIBIT(w)	xfs_highbit32(w)
1168
1169#define	XFS_RTBLOCKLOG(b)	xfs_highbit64(b)
1170
1171/*
1172 * Dquot and dquot block format definitions
1173 */
1174#define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1175#define XFS_DQUOT_VERSION	(uint8_t)0x01	/* latest version number */
1176
1177#define XFS_DQTYPE_USER		(1u << 0)	/* user dquot record */
1178#define XFS_DQTYPE_PROJ		(1u << 1)	/* project dquot record */
1179#define XFS_DQTYPE_GROUP	(1u << 2)	/* group dquot record */
1180#define XFS_DQTYPE_BIGTIME	(1u << 7)	/* large expiry timestamps */
1181
1182/* bitmask to determine if this is a user/group/project dquot */
1183#define XFS_DQTYPE_REC_MASK	(XFS_DQTYPE_USER | \
1184				 XFS_DQTYPE_PROJ | \
1185				 XFS_DQTYPE_GROUP)
1186
1187#define XFS_DQTYPE_ANY		(XFS_DQTYPE_REC_MASK | \
1188				 XFS_DQTYPE_BIGTIME)
1189
1190/*
1191 * XFS Quota Timers
1192 * ================
1193 *
1194 * Traditional quota grace period expiration timers are an unsigned 32-bit
1195 * seconds counter; time zero is the Unix epoch, Jan  1 00:00:01 UTC 1970.
1196 * Note that an expiration value of zero means that the quota limit has not
1197 * been reached, and therefore no expiration has been set.  Therefore, the
1198 * ondisk min and max defined here can be used directly to constrain the incore
1199 * quota expiration timestamps on a Unix system.
1200 *
1201 * When bigtime is enabled, we trade two bits of precision to expand the
1202 * expiration timeout range to match that of big inode timestamps.  The min and
1203 * max recorded here are the on-disk limits, not a Unix timestamp.
1204 *
1205 * The grace period for each quota type is stored in the root dquot (id = 0)
1206 * and is applied to a non-root dquot when it exceeds the soft or hard limits.
1207 * The length of quota grace periods are unsigned 32-bit quantities measured in
1208 * units of seconds.  A value of zero means to use the default period.
1209 */
1210
1211/*
1212 * Smallest possible ondisk quota expiration value with traditional timestamps.
1213 * This corresponds exactly with the incore expiration Jan  1 00:00:01 UTC 1970.
1214 */
1215#define XFS_DQ_LEGACY_EXPIRY_MIN	((int64_t)1)
1216
1217/*
1218 * Largest possible ondisk quota expiration value with traditional timestamps.
1219 * This corresponds exactly with the incore expiration Feb  7 06:28:15 UTC 2106.
1220 */
1221#define XFS_DQ_LEGACY_EXPIRY_MAX	((int64_t)U32_MAX)
1222
1223/*
1224 * Smallest possible ondisk quota expiration value with bigtime timestamps.
1225 * This corresponds (after conversion to a Unix timestamp) with the incore
1226 * expiration of Jan  1 00:00:04 UTC 1970.
1227 */
1228#define XFS_DQ_BIGTIME_EXPIRY_MIN	(XFS_DQ_LEGACY_EXPIRY_MIN)
1229
1230/*
1231 * Largest supported ondisk quota expiration value with bigtime timestamps.
1232 * This corresponds (after conversion to a Unix timestamp) with an incore
1233 * expiration of Jul  2 20:20:24 UTC 2486.
1234 *
1235 * The ondisk field supports values up to -1U, which corresponds to an incore
1236 * expiration in 2514.  This is beyond the maximum the bigtime inode timestamp,
1237 * so we cap the maximum bigtime quota expiration to the max inode timestamp.
1238 */
1239#define XFS_DQ_BIGTIME_EXPIRY_MAX	((int64_t)4074815106U)
1240
1241/*
1242 * The following conversion factors assist in converting a quota expiration
1243 * timestamp between the incore and ondisk formats.
1244 */
1245#define XFS_DQ_BIGTIME_SHIFT	(2)
1246#define XFS_DQ_BIGTIME_SLACK	((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1)
1247
1248/* Convert an incore quota expiration timestamp to an ondisk bigtime value. */
1249static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds)
1250{
1251	/*
1252	 * Round the expiration timestamp up to the nearest bigtime timestamp
1253	 * that we can store, to give users the most time to fix problems.
1254	 */
1255	return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >>
1256			XFS_DQ_BIGTIME_SHIFT;
1257}
1258
1259/* Convert an ondisk bigtime quota expiration value to an incore timestamp. */
1260static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)
1261{
1262	return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT;
1263}
1264
1265/*
1266 * Default quota grace periods, ranging from zero (use the compiled defaults)
1267 * to ~136 years.  These are applied to a non-root dquot that has exceeded
1268 * either limit.
1269 */
1270#define XFS_DQ_GRACE_MIN		((int64_t)0)
1271#define XFS_DQ_GRACE_MAX		((int64_t)U32_MAX)
1272
1273/*
1274 * This is the main portion of the on-disk representation of quota information
1275 * for a user.  We pad this with some more expansion room to construct the on
1276 * disk structure.
1277 */
1278struct xfs_disk_dquot {
1279	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1280	__u8		d_version;	/* dquot version */
1281	__u8		d_type;		/* XFS_DQTYPE_USER/PROJ/GROUP */
1282	__be32		d_id;		/* user,project,group id */
1283	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1284	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1285	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1286	__be64		d_ino_softlimit;/* preferred inode limit */
1287	__be64		d_bcount;	/* disk blocks owned by the user */
1288	__be64		d_icount;	/* inodes owned by the user */
1289	__be32		d_itimer;	/* zero if within inode limits if not,
1290					   this is when we refuse service */
1291	__be32		d_btimer;	/* similar to above; for disk blocks */
1292	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1293	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1294	__be32		d_pad0;		/* 64 bit align */
1295	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1296	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1297	__be64		d_rtbcount;	/* realtime blocks owned */
1298	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1299	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1300	__be16		d_pad;
1301};
1302
1303/*
1304 * This is what goes on disk. This is separated from the xfs_disk_dquot because
1305 * carrying the unnecessary padding would be a waste of memory.
1306 */
1307struct xfs_dqblk {
1308	struct xfs_disk_dquot	dd_diskdq; /* portion living incore as well */
1309	char			dd_fill[4];/* filling for posterity */
1310
1311	/*
1312	 * These two are only present on filesystems with the CRC bits set.
1313	 */
1314	__be32		  dd_crc;	/* checksum */
1315	__be64		  dd_lsn;	/* last modification in log */
1316	uuid_t		  dd_uuid;	/* location information */
1317};
1318
1319#define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1320
1321/*
1322 * This defines the unit of allocation of dquots.
1323 *
1324 * Currently, it is just one file system block, and a 4K blk contains 30
1325 * (136 * 30 = 4080) dquots. It's probably not worth trying to make
1326 * this more dynamic.
1327 *
1328 * However, if this number is changed, we have to make sure that we don't
1329 * implicitly assume that we do allocations in chunks of a single filesystem
1330 * block in the dquot/xqm code.
1331 *
1332 * This is part of the ondisk format because the structure size is not a power
1333 * of two, which leaves slack at the end of the disk block.
1334 */
1335#define XFS_DQUOT_CLUSTER_SIZE_FSB	(xfs_filblks_t)1
1336
1337/*
1338 * Remote symlink format and access functions.
1339 */
1340#define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1341
1342struct xfs_dsymlink_hdr {
1343	__be32	sl_magic;
1344	__be32	sl_offset;
1345	__be32	sl_bytes;
1346	__be32	sl_crc;
1347	uuid_t	sl_uuid;
1348	__be64	sl_owner;
1349	__be64	sl_blkno;
1350	__be64	sl_lsn;
1351};
1352
1353#define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1354
1355#define XFS_SYMLINK_MAXLEN	1024
1356/*
1357 * The maximum pathlen is 1024 bytes. Since the minimum file system
1358 * blocksize is 512 bytes, we can get a max of 3 extents back from
1359 * bmapi when crc headers are taken into account.
1360 */
1361#define XFS_SYMLINK_MAPS 3
1362
1363#define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1364	((bufsize) - (xfs_has_crc((mp)) ? \
1365			sizeof(struct xfs_dsymlink_hdr) : 0))
1366
1367
1368/*
1369 * Allocation Btree format definitions
1370 *
1371 * There are two on-disk btrees, one sorted by blockno and one sorted
1372 * by blockcount and blockno.  All blocks look the same to make the code
1373 * simpler; if we have time later, we'll make the optimizations.
1374 */
1375#define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1376#define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1377#define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1378#define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1379
1380/*
1381 * Data record/key structure
1382 */
1383typedef struct xfs_alloc_rec {
1384	__be32		ar_startblock;	/* starting block number */
1385	__be32		ar_blockcount;	/* count of free blocks */
1386} xfs_alloc_rec_t, xfs_alloc_key_t;
1387
1388typedef struct xfs_alloc_rec_incore {
1389	xfs_agblock_t	ar_startblock;	/* starting block number */
1390	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1391} xfs_alloc_rec_incore_t;
1392
1393/* btree pointer type */
1394typedef __be32 xfs_alloc_ptr_t;
1395
1396/*
1397 * Block numbers in the AG:
1398 * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1399 */
1400#define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1401#define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1402
1403
1404/*
1405 * Inode Allocation Btree format definitions
1406 *
1407 * There is a btree for the inode map per allocation group.
1408 */
1409#define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1410#define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1411#define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1412#define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1413
1414typedef uint64_t	xfs_inofree_t;
1415#define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1416#define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1417#define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1418#define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1419
1420#define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1421#define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(uint16_t))
1422#define XFS_INODES_PER_HOLEMASK_BIT	\
1423	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1424
1425static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1426{
1427	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1428}
1429
1430/*
1431 * The on-disk inode record structure has two formats. The original "full"
1432 * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1433 * and replaces the 3 high-order freecount bytes wth the holemask and inode
1434 * count.
1435 *
1436 * The holemask of the sparse record format allows an inode chunk to have holes
1437 * that refer to blocks not owned by the inode record. This facilitates inode
1438 * allocation in the event of severe free space fragmentation.
1439 */
1440typedef struct xfs_inobt_rec {
1441	__be32		ir_startino;	/* starting inode number */
1442	union {
1443		struct {
1444			__be32	ir_freecount;	/* count of free inodes */
1445		} f;
1446		struct {
1447			__be16	ir_holemask;/* hole mask for sparse chunks */
1448			__u8	ir_count;	/* total inode count */
1449			__u8	ir_freecount;	/* count of free inodes */
1450		} sp;
1451	} ir_u;
1452	__be64		ir_free;	/* free inode mask */
1453} xfs_inobt_rec_t;
1454
1455typedef struct xfs_inobt_rec_incore {
1456	xfs_agino_t	ir_startino;	/* starting inode number */
1457	uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1458	uint8_t		ir_count;	/* total inode count */
1459	uint8_t		ir_freecount;	/* count of free inodes (set bits) */
1460	xfs_inofree_t	ir_free;	/* free inode mask */
1461} xfs_inobt_rec_incore_t;
1462
1463static inline bool xfs_inobt_issparse(uint16_t holemask)
1464{
1465	/* non-zero holemask represents a sparse rec. */
1466	return holemask;
1467}
1468
1469/*
1470 * Key structure
1471 */
1472typedef struct xfs_inobt_key {
1473	__be32		ir_startino;	/* starting inode number */
1474} xfs_inobt_key_t;
1475
1476/* btree pointer type */
1477typedef __be32 xfs_inobt_ptr_t;
1478
1479/*
1480 * block numbers in the AG.
1481 */
1482#define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1483#define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1484
1485/*
1486 * Reverse mapping btree format definitions
1487 *
1488 * There is a btree for the reverse map per allocation group
1489 */
1490#define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1491
1492/*
1493 * Ownership info for an extent.  This is used to create reverse-mapping
1494 * entries.
1495 */
1496#define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1497#define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1498struct xfs_owner_info {
1499	uint64_t		oi_owner;
1500	xfs_fileoff_t		oi_offset;
1501	unsigned int		oi_flags;
1502};
1503
1504/*
1505 * Special owner types.
1506 *
1507 * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1508 * to tell us we have a special owner value. We use these for static metadata
1509 * allocated at mkfs/growfs time, as well as for freespace management metadata.
1510 */
1511#define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1512#define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1513#define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1514#define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1515#define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1516#define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1517#define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1518#define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1519#define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1520#define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1521
1522#define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1523
1524/*
1525 * Data record structure
1526 */
1527struct xfs_rmap_rec {
1528	__be32		rm_startblock;	/* extent start block */
1529	__be32		rm_blockcount;	/* extent length */
1530	__be64		rm_owner;	/* extent owner */
1531	__be64		rm_offset;	/* offset within the owner */
1532};
1533
1534/*
1535 * rmap btree record
1536 *  rm_offset:63 is the attribute fork flag
1537 *  rm_offset:62 is the bmbt block flag
1538 *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1539 *  rm_offset:54-60 aren't used and should be zero
1540 *  rm_offset:0-53 is the block offset within the inode
1541 */
1542#define XFS_RMAP_OFF_ATTR_FORK	((uint64_t)1ULL << 63)
1543#define XFS_RMAP_OFF_BMBT_BLOCK	((uint64_t)1ULL << 62)
1544#define XFS_RMAP_OFF_UNWRITTEN	((uint64_t)1ULL << 61)
1545
1546#define XFS_RMAP_LEN_MAX	((uint32_t)~0U)
1547#define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1548				 XFS_RMAP_OFF_BMBT_BLOCK | \
1549				 XFS_RMAP_OFF_UNWRITTEN)
1550#define XFS_RMAP_OFF_MASK	((uint64_t)0x3FFFFFFFFFFFFFULL)
1551
1552#define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1553
1554#define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1555#define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1556#define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1557
1558#define RMAPBT_STARTBLOCK_BITLEN	32
1559#define RMAPBT_BLOCKCOUNT_BITLEN	32
1560#define RMAPBT_OWNER_BITLEN		64
1561#define RMAPBT_ATTRFLAG_BITLEN		1
1562#define RMAPBT_BMBTFLAG_BITLEN		1
1563#define RMAPBT_EXNTFLAG_BITLEN		1
1564#define RMAPBT_UNUSED_OFFSET_BITLEN	7
1565#define RMAPBT_OFFSET_BITLEN		54
1566
1567#define XFS_RMAP_ATTR_FORK		(1 << 0)
1568#define XFS_RMAP_BMBT_BLOCK		(1 << 1)
1569#define XFS_RMAP_UNWRITTEN		(1 << 2)
1570#define XFS_RMAP_KEY_FLAGS		(XFS_RMAP_ATTR_FORK | \
1571					 XFS_RMAP_BMBT_BLOCK)
1572#define XFS_RMAP_REC_FLAGS		(XFS_RMAP_UNWRITTEN)
1573struct xfs_rmap_irec {
1574	xfs_agblock_t	rm_startblock;	/* extent start block */
1575	xfs_extlen_t	rm_blockcount;	/* extent length */
1576	uint64_t	rm_owner;	/* extent owner */
1577	uint64_t	rm_offset;	/* offset within the owner */
1578	unsigned int	rm_flags;	/* state flags */
1579};
1580
1581/*
1582 * Key structure
1583 *
1584 * We don't use the length for lookups
1585 */
1586struct xfs_rmap_key {
1587	__be32		rm_startblock;	/* extent start block */
1588	__be64		rm_owner;	/* extent owner */
1589	__be64		rm_offset;	/* offset within the owner */
1590} __attribute__((packed));
1591
1592/* btree pointer type */
1593typedef __be32 xfs_rmap_ptr_t;
1594
1595#define	XFS_RMAP_BLOCK(mp) \
1596	(xfs_has_finobt(((mp))) ? \
1597	 XFS_FIBT_BLOCK(mp) + 1 : \
1598	 XFS_IBT_BLOCK(mp) + 1)
1599
1600/*
1601 * Reference Count Btree format definitions
1602 *
1603 */
1604#define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1605
1606unsigned int xfs_refc_block(struct xfs_mount *mp);
1607
1608/*
1609 * Data record/key structure
1610 *
1611 * Each record associates a range of physical blocks (starting at
1612 * rc_startblock and ending rc_blockcount blocks later) with a reference
1613 * count (rc_refcount).  Extents that are being used to stage a copy on
1614 * write (CoW) operation are recorded in the refcount btree with a
1615 * refcount of 1.  All other records must have a refcount > 1 and must
1616 * track an extent mapped only by file data forks.
1617 *
1618 * Extents with a single owner (attributes, metadata, non-shared file
1619 * data) are not tracked here.  Free space is also not tracked here.
1620 * This is consistent with pre-reflink XFS.
1621 */
1622
1623/*
1624 * Extents that are being used to stage a copy on write are stored
1625 * in the refcount btree with a refcount of 1 and the upper bit set
1626 * on the startblock.  This speeds up mount time deletion of stale
1627 * staging extents because they're all at the right side of the tree.
1628 */
1629#define XFS_REFC_COW_START		((xfs_agblock_t)(1U << 31))
1630#define REFCNTBT_COWFLAG_BITLEN		1
1631#define REFCNTBT_AGBLOCK_BITLEN		31
1632
1633struct xfs_refcount_rec {
1634	__be32		rc_startblock;	/* starting block number */
1635	__be32		rc_blockcount;	/* count of blocks */
1636	__be32		rc_refcount;	/* number of inodes linked here */
1637};
1638
1639struct xfs_refcount_key {
1640	__be32		rc_startblock;	/* starting block number */
1641};
1642
1643struct xfs_refcount_irec {
1644	xfs_agblock_t	rc_startblock;	/* starting block number */
1645	xfs_extlen_t	rc_blockcount;	/* count of free blocks */
1646	xfs_nlink_t	rc_refcount;	/* number of inodes linked here */
1647};
1648
1649#define MAXREFCOUNT	((xfs_nlink_t)~0U)
1650#define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
1651
1652/* btree pointer type */
1653typedef __be32 xfs_refcount_ptr_t;
1654
1655
1656/*
1657 * BMAP Btree format definitions
1658 *
1659 * This includes both the root block definition that sits inside an inode fork
1660 * and the record/pointer formats for the leaf/node in the blocks.
1661 */
1662#define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1663#define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1664
1665/*
1666 * Bmap root header, on-disk form only.
1667 */
1668typedef struct xfs_bmdr_block {
1669	__be16		bb_level;	/* 0 is a leaf */
1670	__be16		bb_numrecs;	/* current # of data records */
1671} xfs_bmdr_block_t;
1672
1673/*
1674 * Bmap btree record and extent descriptor.
1675 *  l0:63 is an extent flag (value 1 indicates non-normal).
1676 *  l0:9-62 are startoff.
1677 *  l0:0-8 and l1:21-63 are startblock.
1678 *  l1:0-20 are blockcount.
1679 */
1680#define BMBT_EXNTFLAG_BITLEN	1
1681#define BMBT_STARTOFF_BITLEN	54
1682#define BMBT_STARTBLOCK_BITLEN	52
1683#define BMBT_BLOCKCOUNT_BITLEN	21
1684
1685#define BMBT_STARTOFF_MASK	((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1686#define BMBT_BLOCKCOUNT_MASK	((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1)
1687
1688#define XFS_MAX_BMBT_EXTLEN	((xfs_extlen_t)(BMBT_BLOCKCOUNT_MASK))
1689
1690/*
1691 * bmbt records have a file offset (block) field that is 54 bits wide, so this
1692 * is the largest xfs_fileoff_t that we ever expect to see.
1693 */
1694#define XFS_MAX_FILEOFF		(BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK)
1695
1696typedef struct xfs_bmbt_rec {
1697	__be64			l0, l1;
1698} xfs_bmbt_rec_t;
1699
1700typedef uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1701typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1702
1703/*
1704 * Values and macros for delayed-allocation startblock fields.
1705 */
1706#define STARTBLOCKVALBITS	17
1707#define STARTBLOCKMASKBITS	(15 + 20)
1708#define STARTBLOCKMASK		\
1709	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1710
1711static inline int isnullstartblock(xfs_fsblock_t x)
1712{
1713	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1714}
1715
1716static inline xfs_fsblock_t nullstartblock(int k)
1717{
1718	ASSERT(k < (1 << STARTBLOCKVALBITS));
1719	return STARTBLOCKMASK | (k);
1720}
1721
1722static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1723{
1724	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1725}
1726
1727/*
1728 * Key structure for non-leaf levels of the tree.
1729 */
1730typedef struct xfs_bmbt_key {
1731	__be64		br_startoff;	/* starting file offset */
1732} xfs_bmbt_key_t, xfs_bmdr_key_t;
1733
1734/* btree pointer type */
1735typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1736
1737
1738/*
1739 * Generic Btree block format definitions
1740 *
1741 * This is a combination of the actual format used on disk for short and long
1742 * format btrees.  The first three fields are shared by both format, but the
1743 * pointers are different and should be used with care.
1744 *
1745 * To get the size of the actual short or long form headers please use the size
1746 * macros below.  Never use sizeof(xfs_btree_block).
1747 *
1748 * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1749 * with the crc feature bit, and all accesses to them must be conditional on
1750 * that flag.
1751 */
1752/* short form block header */
1753struct xfs_btree_block_shdr {
1754	__be32		bb_leftsib;
1755	__be32		bb_rightsib;
1756
1757	__be64		bb_blkno;
1758	__be64		bb_lsn;
1759	uuid_t		bb_uuid;
1760	__be32		bb_owner;
1761	__le32		bb_crc;
1762};
1763
1764/* long form block header */
1765struct xfs_btree_block_lhdr {
1766	__be64		bb_leftsib;
1767	__be64		bb_rightsib;
1768
1769	__be64		bb_blkno;
1770	__be64		bb_lsn;
1771	uuid_t		bb_uuid;
1772	__be64		bb_owner;
1773	__le32		bb_crc;
1774	__be32		bb_pad; /* padding for alignment */
1775};
1776
1777struct xfs_btree_block {
1778	__be32		bb_magic;	/* magic number for block type */
1779	__be16		bb_level;	/* 0 is a leaf */
1780	__be16		bb_numrecs;	/* current # of data records */
1781	union {
1782		struct xfs_btree_block_shdr s;
1783		struct xfs_btree_block_lhdr l;
1784	} bb_u;				/* rest */
1785};
1786
1787/* size of a short form block */
1788#define XFS_BTREE_SBLOCK_LEN \
1789	(offsetof(struct xfs_btree_block, bb_u) + \
1790	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1791/* size of a long form block */
1792#define XFS_BTREE_LBLOCK_LEN \
1793	(offsetof(struct xfs_btree_block, bb_u) + \
1794	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1795
1796/* sizes of CRC enabled btree blocks */
1797#define XFS_BTREE_SBLOCK_CRC_LEN \
1798	(offsetof(struct xfs_btree_block, bb_u) + \
1799	 sizeof(struct xfs_btree_block_shdr))
1800#define XFS_BTREE_LBLOCK_CRC_LEN \
1801	(offsetof(struct xfs_btree_block, bb_u) + \
1802	 sizeof(struct xfs_btree_block_lhdr))
1803
1804#define XFS_BTREE_SBLOCK_CRC_OFF \
1805	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1806#define XFS_BTREE_LBLOCK_CRC_OFF \
1807	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1808
1809/*
1810 * On-disk XFS access control list structure.
1811 */
1812struct xfs_acl_entry {
1813	__be32	ae_tag;
1814	__be32	ae_id;
1815	__be16	ae_perm;
1816	__be16	ae_pad;		/* fill the implicit hole in the structure */
1817};
1818
1819struct xfs_acl {
1820	__be32			acl_cnt;
1821	struct xfs_acl_entry	acl_entry[];
1822};
1823
1824/*
1825 * The number of ACL entries allowed is defined by the on-disk format.
1826 * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1827 * limited only by the maximum size of the xattr that stores the information.
1828 */
1829#define XFS_ACL_MAX_ENTRIES(mp)	\
1830	(xfs_has_crc(mp) \
1831		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1832						sizeof(struct xfs_acl_entry) \
1833		: 25)
1834
1835#define XFS_ACL_SIZE(cnt) \
1836	(sizeof(struct xfs_acl) + \
1837		sizeof(struct xfs_acl_entry) * cnt)
1838
1839#define XFS_ACL_MAX_SIZE(mp) \
1840	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1841
1842
1843/* On-disk XFS extended attribute names */
1844#define SGI_ACL_FILE		"SGI_ACL_FILE"
1845#define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
1846#define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
1847#define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
1848
1849#endif /* __XFS_FORMAT_H__ */
1850