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_ALL \
376		(XFS_SB_FEAT_INCOMPAT_FTYPE|	\
377		 XFS_SB_FEAT_INCOMPAT_SPINODES|	\
378		 XFS_SB_FEAT_INCOMPAT_META_UUID| \
379		 XFS_SB_FEAT_INCOMPAT_BIGTIME| \
380		 XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR)
381
382#define XFS_SB_FEAT_INCOMPAT_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_ALL
383static inline bool
384xfs_sb_has_incompat_feature(
385	struct xfs_sb	*sbp,
386	uint32_t	feature)
387{
388	return (sbp->sb_features_incompat & feature) != 0;
389}
390
391#define XFS_SB_FEAT_INCOMPAT_LOG_ALL 0
392#define XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN	~XFS_SB_FEAT_INCOMPAT_LOG_ALL
393static inline bool
394xfs_sb_has_incompat_log_feature(
395	struct xfs_sb	*sbp,
396	uint32_t	feature)
397{
398	return (sbp->sb_features_log_incompat & feature) != 0;
399}
400
401static inline void
402xfs_sb_remove_incompat_log_features(
403	struct xfs_sb	*sbp)
404{
405	sbp->sb_features_log_incompat &= ~XFS_SB_FEAT_INCOMPAT_LOG_ALL;
406}
407
408static inline void
409xfs_sb_add_incompat_log_features(
410	struct xfs_sb	*sbp,
411	unsigned int	features)
412{
413	sbp->sb_features_log_incompat |= features;
414}
415
416
417static inline bool
418xfs_is_quota_inode(struct xfs_sb *sbp, xfs_ino_t ino)
419{
420	return (ino == sbp->sb_uquotino ||
421		ino == sbp->sb_gquotino ||
422		ino == sbp->sb_pquotino);
423}
424
425#define XFS_SB_DADDR		((xfs_daddr_t)0) /* daddr in filesystem/ag */
426#define	XFS_SB_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_SB_DADDR)
427
428#define	XFS_HDR_BLOCK(mp,d)	((xfs_agblock_t)XFS_BB_TO_FSBT(mp,d))
429#define	XFS_DADDR_TO_FSB(mp,d)	XFS_AGB_TO_FSB(mp, \
430			xfs_daddr_to_agno(mp,d), xfs_daddr_to_agbno(mp,d))
431#define	XFS_FSB_TO_DADDR(mp,fsbno)	XFS_AGB_TO_DADDR(mp, \
432			XFS_FSB_TO_AGNO(mp,fsbno), XFS_FSB_TO_AGBNO(mp,fsbno))
433
434/*
435 * File system sector to basic block conversions.
436 */
437#define XFS_FSS_TO_BB(mp,sec)	((sec) << (mp)->m_sectbb_log)
438
439/*
440 * File system block to basic block conversions.
441 */
442#define	XFS_FSB_TO_BB(mp,fsbno)	((fsbno) << (mp)->m_blkbb_log)
443#define	XFS_BB_TO_FSB(mp,bb)	\
444	(((bb) + (XFS_FSB_TO_BB(mp,1) - 1)) >> (mp)->m_blkbb_log)
445#define	XFS_BB_TO_FSBT(mp,bb)	((bb) >> (mp)->m_blkbb_log)
446
447/*
448 * File system block to byte conversions.
449 */
450#define XFS_FSB_TO_B(mp,fsbno)	((xfs_fsize_t)(fsbno) << (mp)->m_sb.sb_blocklog)
451#define XFS_B_TO_FSB(mp,b)	\
452	((((uint64_t)(b)) + (mp)->m_blockmask) >> (mp)->m_sb.sb_blocklog)
453#define XFS_B_TO_FSBT(mp,b)	(((uint64_t)(b)) >> (mp)->m_sb.sb_blocklog)
454
455/*
456 * Allocation group header
457 *
458 * This is divided into three structures, placed in sequential 512-byte
459 * buffers after a copy of the superblock (also in a 512-byte buffer).
460 */
461#define	XFS_AGF_MAGIC	0x58414746	/* 'XAGF' */
462#define	XFS_AGI_MAGIC	0x58414749	/* 'XAGI' */
463#define	XFS_AGFL_MAGIC	0x5841464c	/* 'XAFL' */
464#define	XFS_AGF_VERSION	1
465#define	XFS_AGI_VERSION	1
466
467#define	XFS_AGF_GOOD_VERSION(v)	((v) == XFS_AGF_VERSION)
468#define	XFS_AGI_GOOD_VERSION(v)	((v) == XFS_AGI_VERSION)
469
470/*
471 * Btree number 0 is bno, 1 is cnt, 2 is rmap. This value gives the size of the
472 * arrays below.
473 */
474#define	XFS_BTNUM_AGF	((int)XFS_BTNUM_RMAPi + 1)
475
476/*
477 * The second word of agf_levels in the first a.g. overlaps the EFS
478 * superblock's magic number.  Since the magic numbers valid for EFS
479 * are > 64k, our value cannot be confused for an EFS superblock's.
480 */
481
482typedef struct xfs_agf {
483	/*
484	 * Common allocation group header information
485	 */
486	__be32		agf_magicnum;	/* magic number == XFS_AGF_MAGIC */
487	__be32		agf_versionnum;	/* header version == XFS_AGF_VERSION */
488	__be32		agf_seqno;	/* sequence # starting from 0 */
489	__be32		agf_length;	/* size in blocks of a.g. */
490	/*
491	 * Freespace and rmap information
492	 */
493	__be32		agf_roots[XFS_BTNUM_AGF];	/* root blocks */
494	__be32		agf_levels[XFS_BTNUM_AGF];	/* btree levels */
495
496	__be32		agf_flfirst;	/* first freelist block's index */
497	__be32		agf_fllast;	/* last freelist block's index */
498	__be32		agf_flcount;	/* count of blocks in freelist */
499	__be32		agf_freeblks;	/* total free blocks */
500
501	__be32		agf_longest;	/* longest free space */
502	__be32		agf_btreeblks;	/* # of blocks held in AGF btrees */
503	uuid_t		agf_uuid;	/* uuid of filesystem */
504
505	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
506	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
507
508	__be32		agf_refcount_root;	/* refcount tree root block */
509	__be32		agf_refcount_level;	/* refcount btree levels */
510
511	/*
512	 * reserve some contiguous space for future logged fields before we add
513	 * the unlogged fields. This makes the range logging via flags and
514	 * structure offsets much simpler.
515	 */
516	__be64		agf_spare64[14];
517
518	/* unlogged fields, written during buffer writeback. */
519	__be64		agf_lsn;	/* last write sequence */
520	__be32		agf_crc;	/* crc of agf sector */
521	__be32		agf_spare2;
522
523	/* structure must be padded to 64 bit alignment */
524} xfs_agf_t;
525
526#define XFS_AGF_CRC_OFF		offsetof(struct xfs_agf, agf_crc)
527
528#define	XFS_AGF_MAGICNUM	0x00000001
529#define	XFS_AGF_VERSIONNUM	0x00000002
530#define	XFS_AGF_SEQNO		0x00000004
531#define	XFS_AGF_LENGTH		0x00000008
532#define	XFS_AGF_ROOTS		0x00000010
533#define	XFS_AGF_LEVELS		0x00000020
534#define	XFS_AGF_FLFIRST		0x00000040
535#define	XFS_AGF_FLLAST		0x00000080
536#define	XFS_AGF_FLCOUNT		0x00000100
537#define	XFS_AGF_FREEBLKS	0x00000200
538#define	XFS_AGF_LONGEST		0x00000400
539#define	XFS_AGF_BTREEBLKS	0x00000800
540#define	XFS_AGF_UUID		0x00001000
541#define	XFS_AGF_RMAP_BLOCKS	0x00002000
542#define	XFS_AGF_REFCOUNT_BLOCKS	0x00004000
543#define	XFS_AGF_REFCOUNT_ROOT	0x00008000
544#define	XFS_AGF_REFCOUNT_LEVEL	0x00010000
545#define	XFS_AGF_SPARE64		0x00020000
546#define	XFS_AGF_NUM_BITS	18
547#define	XFS_AGF_ALL_BITS	((1 << XFS_AGF_NUM_BITS) - 1)
548
549#define XFS_AGF_FLAGS \
550	{ XFS_AGF_MAGICNUM,	"MAGICNUM" }, \
551	{ XFS_AGF_VERSIONNUM,	"VERSIONNUM" }, \
552	{ XFS_AGF_SEQNO,	"SEQNO" }, \
553	{ XFS_AGF_LENGTH,	"LENGTH" }, \
554	{ XFS_AGF_ROOTS,	"ROOTS" }, \
555	{ XFS_AGF_LEVELS,	"LEVELS" }, \
556	{ XFS_AGF_FLFIRST,	"FLFIRST" }, \
557	{ XFS_AGF_FLLAST,	"FLLAST" }, \
558	{ XFS_AGF_FLCOUNT,	"FLCOUNT" }, \
559	{ XFS_AGF_FREEBLKS,	"FREEBLKS" }, \
560	{ XFS_AGF_LONGEST,	"LONGEST" }, \
561	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
562	{ XFS_AGF_UUID,		"UUID" }, \
563	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
564	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
565	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
566	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
567	{ XFS_AGF_SPARE64,	"SPARE64" }
568
569/* disk block (xfs_daddr_t) in the AG */
570#define XFS_AGF_DADDR(mp)	((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
571#define	XFS_AGF_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
572
573/*
574 * Size of the unlinked inode hash table in the agi.
575 */
576#define	XFS_AGI_UNLINKED_BUCKETS	64
577
578typedef struct xfs_agi {
579	/*
580	 * Common allocation group header information
581	 */
582	__be32		agi_magicnum;	/* magic number == XFS_AGI_MAGIC */
583	__be32		agi_versionnum;	/* header version == XFS_AGI_VERSION */
584	__be32		agi_seqno;	/* sequence # starting from 0 */
585	__be32		agi_length;	/* size in blocks of a.g. */
586	/*
587	 * Inode information
588	 * Inodes are mapped by interpreting the inode number, so no
589	 * mapping data is needed here.
590	 */
591	__be32		agi_count;	/* count of allocated inodes */
592	__be32		agi_root;	/* root of inode btree */
593	__be32		agi_level;	/* levels in inode btree */
594	__be32		agi_freecount;	/* number of free inodes */
595
596	__be32		agi_newino;	/* new inode just allocated */
597	__be32		agi_dirino;	/* last directory inode chunk */
598	/*
599	 * Hash table of inodes which have been unlinked but are
600	 * still being referenced.
601	 */
602	__be32		agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
603	/*
604	 * This marks the end of logging region 1 and start of logging region 2.
605	 */
606	uuid_t		agi_uuid;	/* uuid of filesystem */
607	__be32		agi_crc;	/* crc of agi sector */
608	__be32		agi_pad32;
609	__be64		agi_lsn;	/* last write sequence */
610
611	__be32		agi_free_root; /* root of the free inode btree */
612	__be32		agi_free_level;/* levels in free inode btree */
613
614	__be32		agi_iblocks;	/* inobt blocks used */
615	__be32		agi_fblocks;	/* finobt blocks used */
616
617	/* structure must be padded to 64 bit alignment */
618} xfs_agi_t;
619
620#define XFS_AGI_CRC_OFF		offsetof(struct xfs_agi, agi_crc)
621
622#define	XFS_AGI_MAGICNUM	(1 << 0)
623#define	XFS_AGI_VERSIONNUM	(1 << 1)
624#define	XFS_AGI_SEQNO		(1 << 2)
625#define	XFS_AGI_LENGTH		(1 << 3)
626#define	XFS_AGI_COUNT		(1 << 4)
627#define	XFS_AGI_ROOT		(1 << 5)
628#define	XFS_AGI_LEVEL		(1 << 6)
629#define	XFS_AGI_FREECOUNT	(1 << 7)
630#define	XFS_AGI_NEWINO		(1 << 8)
631#define	XFS_AGI_DIRINO		(1 << 9)
632#define	XFS_AGI_UNLINKED	(1 << 10)
633#define	XFS_AGI_NUM_BITS_R1	11	/* end of the 1st agi logging region */
634#define	XFS_AGI_ALL_BITS_R1	((1 << XFS_AGI_NUM_BITS_R1) - 1)
635#define	XFS_AGI_FREE_ROOT	(1 << 11)
636#define	XFS_AGI_FREE_LEVEL	(1 << 12)
637#define	XFS_AGI_IBLOCKS		(1 << 13) /* both inobt/finobt block counters */
638#define	XFS_AGI_NUM_BITS_R2	14
639
640/* disk block (xfs_daddr_t) in the AG */
641#define XFS_AGI_DADDR(mp)	((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
642#define	XFS_AGI_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
643
644/*
645 * The third a.g. block contains the a.g. freelist, an array
646 * of block pointers to blocks owned by the allocation btree code.
647 */
648#define XFS_AGFL_DADDR(mp)	((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
649#define	XFS_AGFL_BLOCK(mp)	XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
650#define	XFS_BUF_TO_AGFL(bp)	((struct xfs_agfl *)((bp)->b_addr))
651
652struct xfs_agfl {
653	__be32		agfl_magicnum;
654	__be32		agfl_seqno;
655	uuid_t		agfl_uuid;
656	__be64		agfl_lsn;
657	__be32		agfl_crc;
658} __attribute__((packed));
659
660#define XFS_AGFL_CRC_OFF	offsetof(struct xfs_agfl, agfl_crc)
661
662#define XFS_AGB_TO_FSB(mp,agno,agbno)	\
663	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
664#define	XFS_FSB_TO_AGNO(mp,fsbno)	\
665	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
666#define	XFS_FSB_TO_AGBNO(mp,fsbno)	\
667	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
668#define	XFS_AGB_TO_DADDR(mp,agno,agbno)	\
669	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
670		(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
671#define	XFS_AG_DADDR(mp,agno,d)		(XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
672
673/*
674 * For checking for bad ranges of xfs_daddr_t's, covering multiple
675 * allocation groups or a single xfs_daddr_t that's a superblock copy.
676 */
677#define	XFS_AG_CHECK_DADDR(mp,d,len)	\
678	((len) == 1 ? \
679	    ASSERT((d) == XFS_SB_DADDR || \
680		   xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
681	    ASSERT(xfs_daddr_to_agno(mp, d) == \
682		   xfs_daddr_to_agno(mp, (d) + (len) - 1)))
683
684/*
685 * XFS Timestamps
686 * ==============
687 *
688 * Traditional ondisk inode timestamps consist of signed 32-bit counters for
689 * seconds and nanoseconds; time zero is the Unix epoch, Jan  1 00:00:00 UTC
690 * 1970, which means that the timestamp epoch is the same as the Unix epoch.
691 * Therefore, the ondisk min and max defined here can be used directly to
692 * constrain the incore timestamps on a Unix system.  Note that we actually
693 * encode a __be64 value on disk.
694 *
695 * When the bigtime feature is enabled, ondisk inode timestamps become an
696 * unsigned 64-bit nanoseconds counter.  This means that the bigtime inode
697 * timestamp epoch is the start of the classic timestamp range, which is
698 * Dec 31 20:45:52 UTC 1901.  Because the epochs are not the same, callers
699 * /must/ use the bigtime conversion functions when encoding and decoding raw
700 * timestamps.
701 */
702typedef __be64 xfs_timestamp_t;
703
704/* Legacy timestamp encoding format. */
705struct xfs_legacy_timestamp {
706	__be32		t_sec;		/* timestamp seconds */
707	__be32		t_nsec;		/* timestamp nanoseconds */
708};
709
710/*
711 * Smallest possible ondisk seconds value with traditional timestamps.  This
712 * corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901.
713 */
714#define XFS_LEGACY_TIME_MIN	((int64_t)S32_MIN)
715
716/*
717 * Largest possible ondisk seconds value with traditional timestamps.  This
718 * corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038.
719 */
720#define XFS_LEGACY_TIME_MAX	((int64_t)S32_MAX)
721
722/*
723 * Smallest possible ondisk seconds value with bigtime timestamps.  This
724 * corresponds (after conversion to a Unix timestamp) with the traditional
725 * minimum timestamp of Dec 13 20:45:52 UTC 1901.
726 */
727#define XFS_BIGTIME_TIME_MIN	((int64_t)0)
728
729/*
730 * Largest supported ondisk seconds value with bigtime timestamps.  This
731 * corresponds (after conversion to a Unix timestamp) with an incore timestamp
732 * of Jul  2 20:20:24 UTC 2486.
733 *
734 * We round down the ondisk limit so that the bigtime quota and inode max
735 * timestamps will be the same.
736 */
737#define XFS_BIGTIME_TIME_MAX	((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL))
738
739/*
740 * Bigtime epoch is set exactly to the minimum time value that a traditional
741 * 32-bit timestamp can represent when using the Unix epoch as a reference.
742 * Hence the Unix epoch is at a fixed offset into the supported bigtime
743 * timestamp range.
744 *
745 * The bigtime epoch also matches the minimum value an on-disk 32-bit XFS
746 * timestamp can represent so we will not lose any fidelity in converting
747 * to/from unix and bigtime timestamps.
748 *
749 * The following conversion factor converts a seconds counter from the Unix
750 * epoch to the bigtime epoch.
751 */
752#define XFS_BIGTIME_EPOCH_OFFSET	(-(int64_t)S32_MIN)
753
754/* Convert a timestamp from the Unix epoch to the bigtime epoch. */
755static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds)
756{
757	return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET;
758}
759
760/* Convert a timestamp from the bigtime epoch to the Unix epoch. */
761static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds)
762{
763	return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET;
764}
765
766/*
767 * On-disk inode structure.
768 *
769 * This is just the header or "dinode core", the inode is expanded to fill a
770 * variable size the leftover area split into a data and an attribute fork.
771 * The format of the data and attribute fork depends on the format of the
772 * inode as indicated by di_format and di_aformat.  To access the data and
773 * attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
774 * below.
775 *
776 * There is a very similar struct xfs_log_dinode which matches the layout of
777 * this structure, but is kept in native format instead of big endian.
778 *
779 * Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
780 * padding field for v3 inodes.
781 */
782#define	XFS_DINODE_MAGIC		0x494e	/* 'IN' */
783struct xfs_dinode {
784	__be16		di_magic;	/* inode magic # = XFS_DINODE_MAGIC */
785	__be16		di_mode;	/* mode and type of file */
786	__u8		di_version;	/* inode version */
787	__u8		di_format;	/* format of di_c data */
788	__be16		di_onlink;	/* old number of links to file */
789	__be32		di_uid;		/* owner's user id */
790	__be32		di_gid;		/* owner's group id */
791	__be32		di_nlink;	/* number of links to file */
792	__be16		di_projid_lo;	/* lower part of owner's project id */
793	__be16		di_projid_hi;	/* higher part owner's project id */
794	__u8		di_pad[6];	/* unused, zeroed space */
795	__be16		di_flushiter;	/* incremented on flush */
796	xfs_timestamp_t	di_atime;	/* time last accessed */
797	xfs_timestamp_t	di_mtime;	/* time last modified */
798	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
799	__be64		di_size;	/* number of bytes in file */
800	__be64		di_nblocks;	/* # of direct & btree blocks used */
801	__be32		di_extsize;	/* basic/minimum extent size for file */
802	__be32		di_nextents;	/* number of extents in data fork */
803	__be16		di_anextents;	/* number of extents in attribute fork*/
804	__u8		di_forkoff;	/* attr fork offs, <<3 for 64b align */
805	__s8		di_aformat;	/* format of attr fork's data */
806	__be32		di_dmevmask;	/* DMIG event mask */
807	__be16		di_dmstate;	/* DMIG state info */
808	__be16		di_flags;	/* random flags, XFS_DIFLAG_... */
809	__be32		di_gen;		/* generation number */
810
811	/* di_next_unlinked is the only non-core field in the old dinode */
812	__be32		di_next_unlinked;/* agi unlinked list ptr */
813
814	/* start of the extended dinode, writable fields */
815	__le32		di_crc;		/* CRC of the inode */
816	__be64		di_changecount;	/* number of attribute changes */
817	__be64		di_lsn;		/* flush sequence */
818	__be64		di_flags2;	/* more random flags */
819	__be32		di_cowextsize;	/* basic cow extent size for file */
820	__u8		di_pad2[12];	/* more padding for future expansion */
821
822	/* fields only written to during inode creation */
823	xfs_timestamp_t	di_crtime;	/* time created */
824	__be64		di_ino;		/* inode number */
825	uuid_t		di_uuid;	/* UUID of the filesystem */
826
827	/* structure must be padded to 64 bit alignment */
828};
829
830#define XFS_DINODE_CRC_OFF	offsetof(struct xfs_dinode, di_crc)
831
832#define DI_MAX_FLUSH 0xffff
833
834/*
835 * Size of the core inode on disk.  Version 1 and 2 inodes have
836 * the same size, but version 3 has grown a few additional fields.
837 */
838static inline uint xfs_dinode_size(int version)
839{
840	if (version == 3)
841		return sizeof(struct xfs_dinode);
842	return offsetof(struct xfs_dinode, di_crc);
843}
844
845/*
846 * The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
847 * Since the pathconf interface is signed, we use 2^31 - 1 instead.
848 */
849#define	XFS_MAXLINK		((1U << 31) - 1U)
850
851/*
852 * Values for di_format
853 *
854 * This enum is used in string mapping in xfs_trace.h; please keep the
855 * TRACE_DEFINE_ENUMs for it up to date.
856 */
857enum xfs_dinode_fmt {
858	XFS_DINODE_FMT_DEV,		/* xfs_dev_t */
859	XFS_DINODE_FMT_LOCAL,		/* bulk data */
860	XFS_DINODE_FMT_EXTENTS,		/* struct xfs_bmbt_rec */
861	XFS_DINODE_FMT_BTREE,		/* struct xfs_bmdr_block */
862	XFS_DINODE_FMT_UUID		/* added long ago, but never used */
863};
864
865#define XFS_INODE_FORMAT_STR \
866	{ XFS_DINODE_FMT_DEV,		"dev" }, \
867	{ XFS_DINODE_FMT_LOCAL,		"local" }, \
868	{ XFS_DINODE_FMT_EXTENTS,	"extent" }, \
869	{ XFS_DINODE_FMT_BTREE,		"btree" }, \
870	{ XFS_DINODE_FMT_UUID,		"uuid" }
871
872/*
873 * Inode minimum and maximum sizes.
874 */
875#define	XFS_DINODE_MIN_LOG	8
876#define	XFS_DINODE_MAX_LOG	11
877#define	XFS_DINODE_MIN_SIZE	(1 << XFS_DINODE_MIN_LOG)
878#define	XFS_DINODE_MAX_SIZE	(1 << XFS_DINODE_MAX_LOG)
879
880/*
881 * Inode size for given fs.
882 */
883#define XFS_DINODE_SIZE(mp) \
884	(xfs_has_v3inodes(mp) ? \
885		sizeof(struct xfs_dinode) : \
886		offsetof(struct xfs_dinode, di_crc))
887#define XFS_LITINO(mp) \
888	((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(mp))
889
890/*
891 * Inode data & attribute fork sizes, per inode.
892 */
893#define XFS_DFORK_BOFF(dip)		((int)((dip)->di_forkoff << 3))
894
895#define XFS_DFORK_DSIZE(dip,mp) \
896	((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp))
897#define XFS_DFORK_ASIZE(dip,mp) \
898	((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0)
899#define XFS_DFORK_SIZE(dip,mp,w) \
900	((w) == XFS_DATA_FORK ? \
901		XFS_DFORK_DSIZE(dip, mp) : \
902		XFS_DFORK_ASIZE(dip, mp))
903
904#define XFS_DFORK_MAXEXT(dip, mp, w) \
905	(XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
906
907/*
908 * Return pointers to the data or attribute forks.
909 */
910#define XFS_DFORK_DPTR(dip) \
911	((char *)dip + xfs_dinode_size(dip->di_version))
912#define XFS_DFORK_APTR(dip)	\
913	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
914#define XFS_DFORK_PTR(dip,w)	\
915	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
916
917#define XFS_DFORK_FORMAT(dip,w) \
918	((w) == XFS_DATA_FORK ? \
919		(dip)->di_format : \
920		(dip)->di_aformat)
921#define XFS_DFORK_NEXTENTS(dip,w) \
922	((w) == XFS_DATA_FORK ? \
923		be32_to_cpu((dip)->di_nextents) : \
924		be16_to_cpu((dip)->di_anextents))
925
926/*
927 * For block and character special files the 32bit dev_t is stored at the
928 * beginning of the data fork.
929 */
930static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
931{
932	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
933}
934
935static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
936{
937	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
938}
939
940/*
941 * Values for di_flags
942 */
943#define XFS_DIFLAG_REALTIME_BIT  0	/* file's blocks come from rt area */
944#define XFS_DIFLAG_PREALLOC_BIT  1	/* file space has been preallocated */
945#define XFS_DIFLAG_NEWRTBM_BIT   2	/* for rtbitmap inode, new format */
946#define XFS_DIFLAG_IMMUTABLE_BIT 3	/* inode is immutable */
947#define XFS_DIFLAG_APPEND_BIT    4	/* inode is append-only */
948#define XFS_DIFLAG_SYNC_BIT      5	/* inode is written synchronously */
949#define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
950#define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
951#define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
952#define XFS_DIFLAG_PROJINHERIT_BIT   9	/* create with parents projid */
953#define XFS_DIFLAG_NOSYMLINKS_BIT   10	/* disallow symlink creation */
954#define XFS_DIFLAG_EXTSIZE_BIT      11	/* inode extent size allocator hint */
955#define XFS_DIFLAG_EXTSZINHERIT_BIT 12	/* inherit inode extent size */
956#define XFS_DIFLAG_NODEFRAG_BIT     13	/* do not reorganize/defragment */
957#define XFS_DIFLAG_FILESTREAM_BIT   14  /* use filestream allocator */
958/* Do not use bit 15, di_flags is legacy and unchanging now */
959
960#define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
961#define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
962#define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
963#define XFS_DIFLAG_IMMUTABLE     (1 << XFS_DIFLAG_IMMUTABLE_BIT)
964#define XFS_DIFLAG_APPEND        (1 << XFS_DIFLAG_APPEND_BIT)
965#define XFS_DIFLAG_SYNC          (1 << XFS_DIFLAG_SYNC_BIT)
966#define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
967#define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
968#define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
969#define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
970#define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
971#define XFS_DIFLAG_EXTSIZE       (1 << XFS_DIFLAG_EXTSIZE_BIT)
972#define XFS_DIFLAG_EXTSZINHERIT  (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
973#define XFS_DIFLAG_NODEFRAG      (1 << XFS_DIFLAG_NODEFRAG_BIT)
974#define XFS_DIFLAG_FILESTREAM    (1 << XFS_DIFLAG_FILESTREAM_BIT)
975
976#define XFS_DIFLAG_ANY \
977	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
978	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
979	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
980	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
981	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
982
983/*
984 * Values for di_flags2 These start by being exposed to userspace in the upper
985 * 16 bits of the XFS_XFLAG_s range.
986 */
987#define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
988#define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
989#define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
990#define XFS_DIFLAG2_BIGTIME_BIT	3	/* big timestamps */
991
992#define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
993#define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
994#define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
995#define XFS_DIFLAG2_BIGTIME	(1 << XFS_DIFLAG2_BIGTIME_BIT)
996
997#define XFS_DIFLAG2_ANY \
998	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \
999	 XFS_DIFLAG2_BIGTIME)
1000
1001static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip)
1002{
1003	return dip->di_version >= 3 &&
1004	       (dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME));
1005}
1006
1007/*
1008 * Inode number format:
1009 * low inopblog bits - offset in block
1010 * next agblklog bits - block number in ag
1011 * next agno_log bits - ag number
1012 * high agno_log-agblklog-inopblog bits - 0
1013 */
1014#define	XFS_INO_MASK(k)			(uint32_t)((1ULL << (k)) - 1)
1015#define	XFS_INO_OFFSET_BITS(mp)		(mp)->m_sb.sb_inopblog
1016#define	XFS_INO_AGBNO_BITS(mp)		(mp)->m_sb.sb_agblklog
1017#define	XFS_INO_AGINO_BITS(mp)		((mp)->m_ino_geo.agino_log)
1018#define	XFS_INO_AGNO_BITS(mp)		(mp)->m_agno_log
1019#define	XFS_INO_BITS(mp)		\
1020	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1021#define	XFS_INO_TO_AGNO(mp,i)		\
1022	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1023#define	XFS_INO_TO_AGINO(mp,i)		\
1024	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1025#define	XFS_INO_TO_AGBNO(mp,i)		\
1026	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1027		XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1028#define	XFS_INO_TO_OFFSET(mp,i)		\
1029	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1030#define	XFS_INO_TO_FSB(mp,i)		\
1031	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1032#define	XFS_AGINO_TO_INO(mp,a,i)	\
1033	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1034#define	XFS_AGINO_TO_AGBNO(mp,i)	((i) >> XFS_INO_OFFSET_BITS(mp))
1035#define	XFS_AGINO_TO_OFFSET(mp,i)	\
1036	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1037#define	XFS_OFFBNO_TO_AGINO(mp,b,o)	\
1038	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1039#define	XFS_FSB_TO_INO(mp, b)	((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1040#define	XFS_AGB_TO_AGINO(mp, b)	((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1041
1042#define	XFS_MAXINUMBER		((xfs_ino_t)((1ULL << 56) - 1ULL))
1043#define	XFS_MAXINUMBER_32	((xfs_ino_t)((1ULL << 32) - 1ULL))
1044
1045/*
1046 * RealTime Device format definitions
1047 */
1048
1049/* Min and max rt extent sizes, specified in bytes */
1050#define	XFS_MAX_RTEXTSIZE	(1024 * 1024 * 1024)	/* 1GB */
1051#define	XFS_DFL_RTEXTSIZE	(64 * 1024)	        /* 64kB */
1052#define	XFS_MIN_RTEXTSIZE	(4 * 1024)		/* 4kB */
1053
1054#define	XFS_BLOCKSIZE(mp)	((mp)->m_sb.sb_blocksize)
1055#define	XFS_BLOCKMASK(mp)	((mp)->m_blockmask)
1056#define	XFS_BLOCKWSIZE(mp)	((mp)->m_blockwsize)
1057#define	XFS_BLOCKWMASK(mp)	((mp)->m_blockwmask)
1058
1059/*
1060 * RT Summary and bit manipulation macros.
1061 */
1062#define	XFS_SUMOFFS(mp,ls,bb)	((int)((ls) * (mp)->m_sb.sb_rbmblocks + (bb)))
1063#define	XFS_SUMOFFSTOBLOCK(mp,s)	\
1064	(((s) * (uint)sizeof(xfs_suminfo_t)) >> (mp)->m_sb.sb_blocklog)
1065#define	XFS_SUMPTR(mp,bp,so)	\
1066	((xfs_suminfo_t *)((bp)->b_addr + \
1067		(((so) * (uint)sizeof(xfs_suminfo_t)) & XFS_BLOCKMASK(mp))))
1068
1069#define	XFS_BITTOBLOCK(mp,bi)	((bi) >> (mp)->m_blkbit_log)
1070#define	XFS_BLOCKTOBIT(mp,bb)	((bb) << (mp)->m_blkbit_log)
1071#define	XFS_BITTOWORD(mp,bi)	\
1072	((int)(((bi) >> XFS_NBWORDLOG) & XFS_BLOCKWMASK(mp)))
1073
1074#define	XFS_RTMIN(a,b)	((a) < (b) ? (a) : (b))
1075#define	XFS_RTMAX(a,b)	((a) > (b) ? (a) : (b))
1076
1077#define	XFS_RTLOBIT(w)	xfs_lowbit32(w)
1078#define	XFS_RTHIBIT(w)	xfs_highbit32(w)
1079
1080#define	XFS_RTBLOCKLOG(b)	xfs_highbit64(b)
1081
1082/*
1083 * Dquot and dquot block format definitions
1084 */
1085#define XFS_DQUOT_MAGIC		0x4451		/* 'DQ' */
1086#define XFS_DQUOT_VERSION	(uint8_t)0x01	/* latest version number */
1087
1088#define XFS_DQTYPE_USER		0x01		/* user dquot record */
1089#define XFS_DQTYPE_PROJ		0x02		/* project dquot record */
1090#define XFS_DQTYPE_GROUP	0x04		/* group dquot record */
1091#define XFS_DQTYPE_BIGTIME	0x80		/* large expiry timestamps */
1092
1093/* bitmask to determine if this is a user/group/project dquot */
1094#define XFS_DQTYPE_REC_MASK	(XFS_DQTYPE_USER | \
1095				 XFS_DQTYPE_PROJ | \
1096				 XFS_DQTYPE_GROUP)
1097
1098#define XFS_DQTYPE_ANY		(XFS_DQTYPE_REC_MASK | \
1099				 XFS_DQTYPE_BIGTIME)
1100
1101/*
1102 * XFS Quota Timers
1103 * ================
1104 *
1105 * Traditional quota grace period expiration timers are an unsigned 32-bit
1106 * seconds counter; time zero is the Unix epoch, Jan  1 00:00:01 UTC 1970.
1107 * Note that an expiration value of zero means that the quota limit has not
1108 * been reached, and therefore no expiration has been set.  Therefore, the
1109 * ondisk min and max defined here can be used directly to constrain the incore
1110 * quota expiration timestamps on a Unix system.
1111 *
1112 * When bigtime is enabled, we trade two bits of precision to expand the
1113 * expiration timeout range to match that of big inode timestamps.  The min and
1114 * max recorded here are the on-disk limits, not a Unix timestamp.
1115 *
1116 * The grace period for each quota type is stored in the root dquot (id = 0)
1117 * and is applied to a non-root dquot when it exceeds the soft or hard limits.
1118 * The length of quota grace periods are unsigned 32-bit quantities measured in
1119 * units of seconds.  A value of zero means to use the default period.
1120 */
1121
1122/*
1123 * Smallest possible ondisk quota expiration value with traditional timestamps.
1124 * This corresponds exactly with the incore expiration Jan  1 00:00:01 UTC 1970.
1125 */
1126#define XFS_DQ_LEGACY_EXPIRY_MIN	((int64_t)1)
1127
1128/*
1129 * Largest possible ondisk quota expiration value with traditional timestamps.
1130 * This corresponds exactly with the incore expiration Feb  7 06:28:15 UTC 2106.
1131 */
1132#define XFS_DQ_LEGACY_EXPIRY_MAX	((int64_t)U32_MAX)
1133
1134/*
1135 * Smallest possible ondisk quota expiration value with bigtime timestamps.
1136 * This corresponds (after conversion to a Unix timestamp) with the incore
1137 * expiration of Jan  1 00:00:04 UTC 1970.
1138 */
1139#define XFS_DQ_BIGTIME_EXPIRY_MIN	(XFS_DQ_LEGACY_EXPIRY_MIN)
1140
1141/*
1142 * Largest supported ondisk quota expiration value with bigtime timestamps.
1143 * This corresponds (after conversion to a Unix timestamp) with an incore
1144 * expiration of Jul  2 20:20:24 UTC 2486.
1145 *
1146 * The ondisk field supports values up to -1U, which corresponds to an incore
1147 * expiration in 2514.  This is beyond the maximum the bigtime inode timestamp,
1148 * so we cap the maximum bigtime quota expiration to the max inode timestamp.
1149 */
1150#define XFS_DQ_BIGTIME_EXPIRY_MAX	((int64_t)4074815106U)
1151
1152/*
1153 * The following conversion factors assist in converting a quota expiration
1154 * timestamp between the incore and ondisk formats.
1155 */
1156#define XFS_DQ_BIGTIME_SHIFT	(2)
1157#define XFS_DQ_BIGTIME_SLACK	((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1)
1158
1159/* Convert an incore quota expiration timestamp to an ondisk bigtime value. */
1160static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds)
1161{
1162	/*
1163	 * Round the expiration timestamp up to the nearest bigtime timestamp
1164	 * that we can store, to give users the most time to fix problems.
1165	 */
1166	return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >>
1167			XFS_DQ_BIGTIME_SHIFT;
1168}
1169
1170/* Convert an ondisk bigtime quota expiration value to an incore timestamp. */
1171static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)
1172{
1173	return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT;
1174}
1175
1176/*
1177 * Default quota grace periods, ranging from zero (use the compiled defaults)
1178 * to ~136 years.  These are applied to a non-root dquot that has exceeded
1179 * either limit.
1180 */
1181#define XFS_DQ_GRACE_MIN		((int64_t)0)
1182#define XFS_DQ_GRACE_MAX		((int64_t)U32_MAX)
1183
1184/*
1185 * This is the main portion of the on-disk representation of quota information
1186 * for a user.  We pad this with some more expansion room to construct the on
1187 * disk structure.
1188 */
1189struct xfs_disk_dquot {
1190	__be16		d_magic;	/* dquot magic = XFS_DQUOT_MAGIC */
1191	__u8		d_version;	/* dquot version */
1192	__u8		d_type;		/* XFS_DQTYPE_USER/PROJ/GROUP */
1193	__be32		d_id;		/* user,project,group id */
1194	__be64		d_blk_hardlimit;/* absolute limit on disk blks */
1195	__be64		d_blk_softlimit;/* preferred limit on disk blks */
1196	__be64		d_ino_hardlimit;/* maximum # allocated inodes */
1197	__be64		d_ino_softlimit;/* preferred inode limit */
1198	__be64		d_bcount;	/* disk blocks owned by the user */
1199	__be64		d_icount;	/* inodes owned by the user */
1200	__be32		d_itimer;	/* zero if within inode limits if not,
1201					   this is when we refuse service */
1202	__be32		d_btimer;	/* similar to above; for disk blocks */
1203	__be16		d_iwarns;	/* warnings issued wrt num inodes */
1204	__be16		d_bwarns;	/* warnings issued wrt disk blocks */
1205	__be32		d_pad0;		/* 64 bit align */
1206	__be64		d_rtb_hardlimit;/* absolute limit on realtime blks */
1207	__be64		d_rtb_softlimit;/* preferred limit on RT disk blks */
1208	__be64		d_rtbcount;	/* realtime blocks owned */
1209	__be32		d_rtbtimer;	/* similar to above; for RT disk blocks */
1210	__be16		d_rtbwarns;	/* warnings issued wrt RT disk blocks */
1211	__be16		d_pad;
1212};
1213
1214/*
1215 * This is what goes on disk. This is separated from the xfs_disk_dquot because
1216 * carrying the unnecessary padding would be a waste of memory.
1217 */
1218struct xfs_dqblk {
1219	struct xfs_disk_dquot	dd_diskdq; /* portion living incore as well */
1220	char			dd_fill[4];/* filling for posterity */
1221
1222	/*
1223	 * These two are only present on filesystems with the CRC bits set.
1224	 */
1225	__be32		  dd_crc;	/* checksum */
1226	__be64		  dd_lsn;	/* last modification in log */
1227	uuid_t		  dd_uuid;	/* location information */
1228};
1229
1230#define XFS_DQUOT_CRC_OFF	offsetof(struct xfs_dqblk, dd_crc)
1231
1232/*
1233 * This defines the unit of allocation of dquots.
1234 *
1235 * Currently, it is just one file system block, and a 4K blk contains 30
1236 * (136 * 30 = 4080) dquots. It's probably not worth trying to make
1237 * this more dynamic.
1238 *
1239 * However, if this number is changed, we have to make sure that we don't
1240 * implicitly assume that we do allocations in chunks of a single filesystem
1241 * block in the dquot/xqm code.
1242 *
1243 * This is part of the ondisk format because the structure size is not a power
1244 * of two, which leaves slack at the end of the disk block.
1245 */
1246#define XFS_DQUOT_CLUSTER_SIZE_FSB	(xfs_filblks_t)1
1247
1248/*
1249 * Remote symlink format and access functions.
1250 */
1251#define XFS_SYMLINK_MAGIC	0x58534c4d	/* XSLM */
1252
1253struct xfs_dsymlink_hdr {
1254	__be32	sl_magic;
1255	__be32	sl_offset;
1256	__be32	sl_bytes;
1257	__be32	sl_crc;
1258	uuid_t	sl_uuid;
1259	__be64	sl_owner;
1260	__be64	sl_blkno;
1261	__be64	sl_lsn;
1262};
1263
1264#define XFS_SYMLINK_CRC_OFF	offsetof(struct xfs_dsymlink_hdr, sl_crc)
1265
1266#define XFS_SYMLINK_MAXLEN	1024
1267/*
1268 * The maximum pathlen is 1024 bytes. Since the minimum file system
1269 * blocksize is 512 bytes, we can get a max of 3 extents back from
1270 * bmapi when crc headers are taken into account.
1271 */
1272#define XFS_SYMLINK_MAPS 3
1273
1274#define XFS_SYMLINK_BUF_SPACE(mp, bufsize)	\
1275	((bufsize) - (xfs_has_crc((mp)) ? \
1276			sizeof(struct xfs_dsymlink_hdr) : 0))
1277
1278
1279/*
1280 * Allocation Btree format definitions
1281 *
1282 * There are two on-disk btrees, one sorted by blockno and one sorted
1283 * by blockcount and blockno.  All blocks look the same to make the code
1284 * simpler; if we have time later, we'll make the optimizations.
1285 */
1286#define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
1287#define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
1288#define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
1289#define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
1290
1291/*
1292 * Data record/key structure
1293 */
1294typedef struct xfs_alloc_rec {
1295	__be32		ar_startblock;	/* starting block number */
1296	__be32		ar_blockcount;	/* count of free blocks */
1297} xfs_alloc_rec_t, xfs_alloc_key_t;
1298
1299typedef struct xfs_alloc_rec_incore {
1300	xfs_agblock_t	ar_startblock;	/* starting block number */
1301	xfs_extlen_t	ar_blockcount;	/* count of free blocks */
1302} xfs_alloc_rec_incore_t;
1303
1304/* btree pointer type */
1305typedef __be32 xfs_alloc_ptr_t;
1306
1307/*
1308 * Block numbers in the AG:
1309 * SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1310 */
1311#define	XFS_BNO_BLOCK(mp)	((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1312#define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1313
1314
1315/*
1316 * Inode Allocation Btree format definitions
1317 *
1318 * There is a btree for the inode map per allocation group.
1319 */
1320#define	XFS_IBT_MAGIC		0x49414254	/* 'IABT' */
1321#define	XFS_IBT_CRC_MAGIC	0x49414233	/* 'IAB3' */
1322#define	XFS_FIBT_MAGIC		0x46494254	/* 'FIBT' */
1323#define	XFS_FIBT_CRC_MAGIC	0x46494233	/* 'FIB3' */
1324
1325typedef uint64_t	xfs_inofree_t;
1326#define	XFS_INODES_PER_CHUNK		(NBBY * sizeof(xfs_inofree_t))
1327#define	XFS_INODES_PER_CHUNK_LOG	(XFS_NBBYLOG + 3)
1328#define	XFS_INOBT_ALL_FREE		((xfs_inofree_t)-1)
1329#define	XFS_INOBT_MASK(i)		((xfs_inofree_t)1 << (i))
1330
1331#define XFS_INOBT_HOLEMASK_FULL		0	/* holemask for full chunk */
1332#define XFS_INOBT_HOLEMASK_BITS		(NBBY * sizeof(uint16_t))
1333#define XFS_INODES_PER_HOLEMASK_BIT	\
1334	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1335
1336static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1337{
1338	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1339}
1340
1341/*
1342 * The on-disk inode record structure has two formats. The original "full"
1343 * format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1344 * and replaces the 3 high-order freecount bytes wth the holemask and inode
1345 * count.
1346 *
1347 * The holemask of the sparse record format allows an inode chunk to have holes
1348 * that refer to blocks not owned by the inode record. This facilitates inode
1349 * allocation in the event of severe free space fragmentation.
1350 */
1351typedef struct xfs_inobt_rec {
1352	__be32		ir_startino;	/* starting inode number */
1353	union {
1354		struct {
1355			__be32	ir_freecount;	/* count of free inodes */
1356		} f;
1357		struct {
1358			__be16	ir_holemask;/* hole mask for sparse chunks */
1359			__u8	ir_count;	/* total inode count */
1360			__u8	ir_freecount;	/* count of free inodes */
1361		} sp;
1362	} ir_u;
1363	__be64		ir_free;	/* free inode mask */
1364} xfs_inobt_rec_t;
1365
1366typedef struct xfs_inobt_rec_incore {
1367	xfs_agino_t	ir_startino;	/* starting inode number */
1368	uint16_t	ir_holemask;	/* hole mask for sparse chunks */
1369	uint8_t		ir_count;	/* total inode count */
1370	uint8_t		ir_freecount;	/* count of free inodes (set bits) */
1371	xfs_inofree_t	ir_free;	/* free inode mask */
1372} xfs_inobt_rec_incore_t;
1373
1374static inline bool xfs_inobt_issparse(uint16_t holemask)
1375{
1376	/* non-zero holemask represents a sparse rec. */
1377	return holemask;
1378}
1379
1380/*
1381 * Key structure
1382 */
1383typedef struct xfs_inobt_key {
1384	__be32		ir_startino;	/* starting inode number */
1385} xfs_inobt_key_t;
1386
1387/* btree pointer type */
1388typedef __be32 xfs_inobt_ptr_t;
1389
1390/*
1391 * block numbers in the AG.
1392 */
1393#define	XFS_IBT_BLOCK(mp)		((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1394#define	XFS_FIBT_BLOCK(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1395
1396/*
1397 * Reverse mapping btree format definitions
1398 *
1399 * There is a btree for the reverse map per allocation group
1400 */
1401#define	XFS_RMAP_CRC_MAGIC	0x524d4233	/* 'RMB3' */
1402
1403/*
1404 * Ownership info for an extent.  This is used to create reverse-mapping
1405 * entries.
1406 */
1407#define XFS_OWNER_INFO_ATTR_FORK	(1 << 0)
1408#define XFS_OWNER_INFO_BMBT_BLOCK	(1 << 1)
1409struct xfs_owner_info {
1410	uint64_t		oi_owner;
1411	xfs_fileoff_t		oi_offset;
1412	unsigned int		oi_flags;
1413};
1414
1415/*
1416 * Special owner types.
1417 *
1418 * Seeing as we only support up to 8EB, we have the upper bit of the owner field
1419 * to tell us we have a special owner value. We use these for static metadata
1420 * allocated at mkfs/growfs time, as well as for freespace management metadata.
1421 */
1422#define XFS_RMAP_OWN_NULL	(-1ULL)	/* No owner, for growfs */
1423#define XFS_RMAP_OWN_UNKNOWN	(-2ULL)	/* Unknown owner, for EFI recovery */
1424#define XFS_RMAP_OWN_FS		(-3ULL)	/* static fs metadata */
1425#define XFS_RMAP_OWN_LOG	(-4ULL)	/* static fs metadata */
1426#define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
1427#define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
1428#define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
1429#define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
1430#define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
1431#define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
1432
1433#define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
1434
1435/*
1436 * Data record structure
1437 */
1438struct xfs_rmap_rec {
1439	__be32		rm_startblock;	/* extent start block */
1440	__be32		rm_blockcount;	/* extent length */
1441	__be64		rm_owner;	/* extent owner */
1442	__be64		rm_offset;	/* offset within the owner */
1443};
1444
1445/*
1446 * rmap btree record
1447 *  rm_offset:63 is the attribute fork flag
1448 *  rm_offset:62 is the bmbt block flag
1449 *  rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1450 *  rm_offset:54-60 aren't used and should be zero
1451 *  rm_offset:0-53 is the block offset within the inode
1452 */
1453#define XFS_RMAP_OFF_ATTR_FORK	((uint64_t)1ULL << 63)
1454#define XFS_RMAP_OFF_BMBT_BLOCK	((uint64_t)1ULL << 62)
1455#define XFS_RMAP_OFF_UNWRITTEN	((uint64_t)1ULL << 61)
1456
1457#define XFS_RMAP_LEN_MAX	((uint32_t)~0U)
1458#define XFS_RMAP_OFF_FLAGS	(XFS_RMAP_OFF_ATTR_FORK | \
1459				 XFS_RMAP_OFF_BMBT_BLOCK | \
1460				 XFS_RMAP_OFF_UNWRITTEN)
1461#define XFS_RMAP_OFF_MASK	((uint64_t)0x3FFFFFFFFFFFFFULL)
1462
1463#define XFS_RMAP_OFF(off)		((off) & XFS_RMAP_OFF_MASK)
1464
1465#define XFS_RMAP_IS_BMBT_BLOCK(off)	(!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1466#define XFS_RMAP_IS_ATTR_FORK(off)	(!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1467#define XFS_RMAP_IS_UNWRITTEN(len)	(!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1468
1469#define RMAPBT_STARTBLOCK_BITLEN	32
1470#define RMAPBT_BLOCKCOUNT_BITLEN	32
1471#define RMAPBT_OWNER_BITLEN		64
1472#define RMAPBT_ATTRFLAG_BITLEN		1
1473#define RMAPBT_BMBTFLAG_BITLEN		1
1474#define RMAPBT_EXNTFLAG_BITLEN		1
1475#define RMAPBT_UNUSED_OFFSET_BITLEN	7
1476#define RMAPBT_OFFSET_BITLEN		54
1477
1478#define XFS_RMAP_ATTR_FORK		(1 << 0)
1479#define XFS_RMAP_BMBT_BLOCK		(1 << 1)
1480#define XFS_RMAP_UNWRITTEN		(1 << 2)
1481#define XFS_RMAP_KEY_FLAGS		(XFS_RMAP_ATTR_FORK | \
1482					 XFS_RMAP_BMBT_BLOCK)
1483#define XFS_RMAP_REC_FLAGS		(XFS_RMAP_UNWRITTEN)
1484struct xfs_rmap_irec {
1485	xfs_agblock_t	rm_startblock;	/* extent start block */
1486	xfs_extlen_t	rm_blockcount;	/* extent length */
1487	uint64_t	rm_owner;	/* extent owner */
1488	uint64_t	rm_offset;	/* offset within the owner */
1489	unsigned int	rm_flags;	/* state flags */
1490};
1491
1492/*
1493 * Key structure
1494 *
1495 * We don't use the length for lookups
1496 */
1497struct xfs_rmap_key {
1498	__be32		rm_startblock;	/* extent start block */
1499	__be64		rm_owner;	/* extent owner */
1500	__be64		rm_offset;	/* offset within the owner */
1501} __attribute__((packed));
1502
1503/* btree pointer type */
1504typedef __be32 xfs_rmap_ptr_t;
1505
1506#define	XFS_RMAP_BLOCK(mp) \
1507	(xfs_has_finobt(((mp))) ? \
1508	 XFS_FIBT_BLOCK(mp) + 1 : \
1509	 XFS_IBT_BLOCK(mp) + 1)
1510
1511/*
1512 * Reference Count Btree format definitions
1513 *
1514 */
1515#define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
1516
1517unsigned int xfs_refc_block(struct xfs_mount *mp);
1518
1519/*
1520 * Data record/key structure
1521 *
1522 * Each record associates a range of physical blocks (starting at
1523 * rc_startblock and ending rc_blockcount blocks later) with a reference
1524 * count (rc_refcount).  Extents that are being used to stage a copy on
1525 * write (CoW) operation are recorded in the refcount btree with a
1526 * refcount of 1.  All other records must have a refcount > 1 and must
1527 * track an extent mapped only by file data forks.
1528 *
1529 * Extents with a single owner (attributes, metadata, non-shared file
1530 * data) are not tracked here.  Free space is also not tracked here.
1531 * This is consistent with pre-reflink XFS.
1532 */
1533
1534/*
1535 * Extents that are being used to stage a copy on write are stored
1536 * in the refcount btree with a refcount of 1 and the upper bit set
1537 * on the startblock.  This speeds up mount time deletion of stale
1538 * staging extents because they're all at the right side of the tree.
1539 */
1540#define XFS_REFC_COW_START		((xfs_agblock_t)(1U << 31))
1541#define REFCNTBT_COWFLAG_BITLEN		1
1542#define REFCNTBT_AGBLOCK_BITLEN		31
1543
1544struct xfs_refcount_rec {
1545	__be32		rc_startblock;	/* starting block number */
1546	__be32		rc_blockcount;	/* count of blocks */
1547	__be32		rc_refcount;	/* number of inodes linked here */
1548};
1549
1550struct xfs_refcount_key {
1551	__be32		rc_startblock;	/* starting block number */
1552};
1553
1554struct xfs_refcount_irec {
1555	xfs_agblock_t	rc_startblock;	/* starting block number */
1556	xfs_extlen_t	rc_blockcount;	/* count of free blocks */
1557	xfs_nlink_t	rc_refcount;	/* number of inodes linked here */
1558};
1559
1560#define MAXREFCOUNT	((xfs_nlink_t)~0U)
1561#define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
1562
1563/* btree pointer type */
1564typedef __be32 xfs_refcount_ptr_t;
1565
1566
1567/*
1568 * BMAP Btree format definitions
1569 *
1570 * This includes both the root block definition that sits inside an inode fork
1571 * and the record/pointer formats for the leaf/node in the blocks.
1572 */
1573#define XFS_BMAP_MAGIC		0x424d4150	/* 'BMAP' */
1574#define XFS_BMAP_CRC_MAGIC	0x424d4133	/* 'BMA3' */
1575
1576/*
1577 * Bmap root header, on-disk form only.
1578 */
1579typedef struct xfs_bmdr_block {
1580	__be16		bb_level;	/* 0 is a leaf */
1581	__be16		bb_numrecs;	/* current # of data records */
1582} xfs_bmdr_block_t;
1583
1584/*
1585 * Bmap btree record and extent descriptor.
1586 *  l0:63 is an extent flag (value 1 indicates non-normal).
1587 *  l0:9-62 are startoff.
1588 *  l0:0-8 and l1:21-63 are startblock.
1589 *  l1:0-20 are blockcount.
1590 */
1591#define BMBT_EXNTFLAG_BITLEN	1
1592#define BMBT_STARTOFF_BITLEN	54
1593#define BMBT_STARTBLOCK_BITLEN	52
1594#define BMBT_BLOCKCOUNT_BITLEN	21
1595
1596#define BMBT_STARTOFF_MASK	((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1597#define BMBT_BLOCKCOUNT_MASK	((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1)
1598
1599/*
1600 * bmbt records have a file offset (block) field that is 54 bits wide, so this
1601 * is the largest xfs_fileoff_t that we ever expect to see.
1602 */
1603#define XFS_MAX_FILEOFF		(BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK)
1604
1605typedef struct xfs_bmbt_rec {
1606	__be64			l0, l1;
1607} xfs_bmbt_rec_t;
1608
1609typedef uint64_t	xfs_bmbt_rec_base_t;	/* use this for casts */
1610typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1611
1612/*
1613 * Values and macros for delayed-allocation startblock fields.
1614 */
1615#define STARTBLOCKVALBITS	17
1616#define STARTBLOCKMASKBITS	(15 + 20)
1617#define STARTBLOCKMASK		\
1618	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1619
1620static inline int isnullstartblock(xfs_fsblock_t x)
1621{
1622	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1623}
1624
1625static inline xfs_fsblock_t nullstartblock(int k)
1626{
1627	ASSERT(k < (1 << STARTBLOCKVALBITS));
1628	return STARTBLOCKMASK | (k);
1629}
1630
1631static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1632{
1633	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1634}
1635
1636/*
1637 * Key structure for non-leaf levels of the tree.
1638 */
1639typedef struct xfs_bmbt_key {
1640	__be64		br_startoff;	/* starting file offset */
1641} xfs_bmbt_key_t, xfs_bmdr_key_t;
1642
1643/* btree pointer type */
1644typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1645
1646
1647/*
1648 * Generic Btree block format definitions
1649 *
1650 * This is a combination of the actual format used on disk for short and long
1651 * format btrees.  The first three fields are shared by both format, but the
1652 * pointers are different and should be used with care.
1653 *
1654 * To get the size of the actual short or long form headers please use the size
1655 * macros below.  Never use sizeof(xfs_btree_block).
1656 *
1657 * The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1658 * with the crc feature bit, and all accesses to them must be conditional on
1659 * that flag.
1660 */
1661/* short form block header */
1662struct xfs_btree_block_shdr {
1663	__be32		bb_leftsib;
1664	__be32		bb_rightsib;
1665
1666	__be64		bb_blkno;
1667	__be64		bb_lsn;
1668	uuid_t		bb_uuid;
1669	__be32		bb_owner;
1670	__le32		bb_crc;
1671};
1672
1673/* long form block header */
1674struct xfs_btree_block_lhdr {
1675	__be64		bb_leftsib;
1676	__be64		bb_rightsib;
1677
1678	__be64		bb_blkno;
1679	__be64		bb_lsn;
1680	uuid_t		bb_uuid;
1681	__be64		bb_owner;
1682	__le32		bb_crc;
1683	__be32		bb_pad; /* padding for alignment */
1684};
1685
1686struct xfs_btree_block {
1687	__be32		bb_magic;	/* magic number for block type */
1688	__be16		bb_level;	/* 0 is a leaf */
1689	__be16		bb_numrecs;	/* current # of data records */
1690	union {
1691		struct xfs_btree_block_shdr s;
1692		struct xfs_btree_block_lhdr l;
1693	} bb_u;				/* rest */
1694};
1695
1696/* size of a short form block */
1697#define XFS_BTREE_SBLOCK_LEN \
1698	(offsetof(struct xfs_btree_block, bb_u) + \
1699	 offsetof(struct xfs_btree_block_shdr, bb_blkno))
1700/* size of a long form block */
1701#define XFS_BTREE_LBLOCK_LEN \
1702	(offsetof(struct xfs_btree_block, bb_u) + \
1703	 offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1704
1705/* sizes of CRC enabled btree blocks */
1706#define XFS_BTREE_SBLOCK_CRC_LEN \
1707	(offsetof(struct xfs_btree_block, bb_u) + \
1708	 sizeof(struct xfs_btree_block_shdr))
1709#define XFS_BTREE_LBLOCK_CRC_LEN \
1710	(offsetof(struct xfs_btree_block, bb_u) + \
1711	 sizeof(struct xfs_btree_block_lhdr))
1712
1713#define XFS_BTREE_SBLOCK_CRC_OFF \
1714	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1715#define XFS_BTREE_LBLOCK_CRC_OFF \
1716	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1717
1718/*
1719 * On-disk XFS access control list structure.
1720 */
1721struct xfs_acl_entry {
1722	__be32	ae_tag;
1723	__be32	ae_id;
1724	__be16	ae_perm;
1725	__be16	ae_pad;		/* fill the implicit hole in the structure */
1726};
1727
1728struct xfs_acl {
1729	__be32			acl_cnt;
1730	struct xfs_acl_entry	acl_entry[];
1731};
1732
1733/*
1734 * The number of ACL entries allowed is defined by the on-disk format.
1735 * For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1736 * limited only by the maximum size of the xattr that stores the information.
1737 */
1738#define XFS_ACL_MAX_ENTRIES(mp)	\
1739	(xfs_has_crc(mp) \
1740		?  (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1741						sizeof(struct xfs_acl_entry) \
1742		: 25)
1743
1744#define XFS_ACL_SIZE(cnt) \
1745	(sizeof(struct xfs_acl) + \
1746		sizeof(struct xfs_acl_entry) * cnt)
1747
1748#define XFS_ACL_MAX_SIZE(mp) \
1749	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1750
1751
1752/* On-disk XFS extended attribute names */
1753#define SGI_ACL_FILE		"SGI_ACL_FILE"
1754#define SGI_ACL_DEFAULT		"SGI_ACL_DEFAULT"
1755#define SGI_ACL_FILE_SIZE	(sizeof(SGI_ACL_FILE)-1)
1756#define SGI_ACL_DEFAULT_SIZE	(sizeof(SGI_ACL_DEFAULT)-1)
1757
1758#endif /* __XFS_FORMAT_H__ */
1759