1	/* SPDX-License-Identifier: GPL-2.0 */
2	/*
3	* Copyright (c) 2000-2005 Silicon Graphics, Inc.
4	* All Rights Reserved.
5	*/
6	#ifndef __XFS_FORMAT_H__
7	#define __XFS_FORMAT_H__
8
9	/*
10	* XFS On Disk Format Definitions
11	*
12	* This header file defines all the on-disk format definitions for
13	* general XFS objects. Directory and attribute related objects are defined in
14	* xfs_da_format.h, which log and log item formats are defined in
15	* xfs_log_format.h. Everything else goes here.
16	*/
17
18	struct xfs_mount;
19	struct xfs_trans;
20	struct xfs_inode;
21	struct xfs_buf;
22	struct xfs_ifork;
23
24	/*
25	* Super block
26	* Fits into a sector-sized buffer at address 0 of each allocation group.
27	* Only the first of these is ever updated except during growfs.
28	*/
29	#define XFS_SB_MAGIC 0x58465342 /* 'XFSB' */
30	#define XFS_SB_VERSION_1 1 /* 5.3, 6.0.1, 6.1 */
31	#define XFS_SB_VERSION_2 2 /* 6.2 - attributes */
32	#define XFS_SB_VERSION_3 3 /* 6.2 - new inode version */
33	#define XFS_SB_VERSION_4 4 /* 6.2+ - bitmask version */
34	#define XFS_SB_VERSION_5 5 /* CRC enabled filesystem */
35	#define XFS_SB_VERSION_NUMBITS 0x000f
36	#define XFS_SB_VERSION_ALLFBITS 0xfff0
37	#define XFS_SB_VERSION_ATTRBIT 0x0010
38	#define XFS_SB_VERSION_NLINKBIT 0x0020
39	#define XFS_SB_VERSION_QUOTABIT 0x0040
40	#define XFS_SB_VERSION_ALIGNBIT 0x0080
41	#define XFS_SB_VERSION_DALIGNBIT 0x0100
42	#define XFS_SB_VERSION_SHAREDBIT 0x0200
43	#define XFS_SB_VERSION_LOGV2BIT 0x0400
44	#define XFS_SB_VERSION_SECTORBIT 0x0800
45	#define XFS_SB_VERSION_EXTFLGBIT 0x1000
46	#define XFS_SB_VERSION_DIRV2BIT 0x2000
47	#define XFS_SB_VERSION_BORGBIT 0x4000 /* ASCII only case-insens. */
48	#define XFS_SB_VERSION_MOREBITSBIT 0x8000
49
50	/*
51	* The size of a single extended attribute on disk is limited by
52	* the size of index values within the attribute entries themselves.
53	* These are be16 fields, so we can only support attribute data
54	* sizes up to 2^16 bytes in length.
55	*/
56	#define XFS_XATTR_SIZE_MAX (1 << 16)
57
58	/*
59	* Supported feature bit list is just all bits in the versionnum field because
60	* we've used them all up and understand them all. Except, of course, for the
61	* shared superblock bit, which nobody knows what it does and so is unsupported.
62	*/
63	#define XFS_SB_VERSION_OKBITS \
64	((XFS_SB_VERSION_NUMBITS | XFS_SB_VERSION_ALLFBITS) & \
65	~XFS_SB_VERSION_SHAREDBIT)
66
67	/*
68	* There are two words to hold XFS "feature" bits: the original
69	* word, sb_versionnum, and sb_features2. Whenever a bit is set in
70	* sb_features2, the feature bit XFS_SB_VERSION_MOREBITSBIT must be set.
71	*
72	* These defines represent bits in sb_features2.
73	*/
74	#define XFS_SB_VERSION2_RESERVED1BIT 0x00000001
75	#define XFS_SB_VERSION2_LAZYSBCOUNTBIT 0x00000002 /* Superblk counters */
76	#define XFS_SB_VERSION2_RESERVED4BIT 0x00000004
77	#define XFS_SB_VERSION2_ATTR2BIT 0x00000008 /* Inline attr rework */
78	#define XFS_SB_VERSION2_PARENTBIT 0x00000010 /* parent pointers */
79	#define XFS_SB_VERSION2_PROJID32BIT 0x00000080 /* 32 bit project id */
80	#define XFS_SB_VERSION2_CRCBIT 0x00000100 /* metadata CRCs */
81	#define XFS_SB_VERSION2_FTYPE 0x00000200 /* inode type in dir */
82
83	#define XFS_SB_VERSION2_OKBITS \
84	(XFS_SB_VERSION2_LAZYSBCOUNTBIT | \
85	XFS_SB_VERSION2_ATTR2BIT | \
86	XFS_SB_VERSION2_PROJID32BIT | \
87	XFS_SB_VERSION2_FTYPE)
88
89	/* Maximum size of the xfs filesystem label, no terminating NULL */
90	#define XFSLABEL_MAX 12
91
92	/*
93	* Superblock - in core version. Must match the ondisk version below.
94	* Must be padded to 64 bit alignment.
95	*/
96	typedef 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_rtbxlen_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	*/
187	struct xfs_dsb {
188	__be32 sb_magicnum; /* magic number == XFS_SB_MAGIC */
189	__be32 sb_blocksize; /* logical block size, bytes */
190	__be64 sb_dblocks; /* number of data blocks */
191	__be64 sb_rblocks; /* number of realtime blocks */
192	__be64 sb_rextents; /* number of realtime extents */
193	uuid_t sb_uuid; /* user-visible file system unique id */
194	__be64 sb_logstart; /* starting block of log if internal */
195	__be64 sb_rootino; /* root inode number */
196	__be64 sb_rbmino; /* bitmap inode for realtime extents */
197	__be64 sb_rsumino; /* summary inode for rt bitmap */
198	__be32 sb_rextsize; /* realtime extent size, blocks */
199	__be32 sb_agblocks; /* size of an allocation group */
200	__be32 sb_agcount; /* number of allocation groups */
201	__be32 sb_rbmblocks; /* number of rt bitmap blocks */
202	__be32 sb_logblocks; /* number of log blocks */
203	__be16 sb_versionnum; /* header version == XFS_SB_VERSION */
204	__be16 sb_sectsize; /* volume sector size, bytes */
205	__be16 sb_inodesize; /* inode size, bytes */
206	__be16 sb_inopblock; /* inodes per block */
207	char sb_fname[XFSLABEL_MAX]; /* file system name */
208	__u8 sb_blocklog; /* log2 of sb_blocksize */
209	__u8 sb_sectlog; /* log2 of sb_sectsize */
210	__u8 sb_inodelog; /* log2 of sb_inodesize */
211	__u8 sb_inopblog; /* log2 of sb_inopblock */
212	__u8 sb_agblklog; /* log2 of sb_agblocks (rounded up) */
213	__u8 sb_rextslog; /* log2 of sb_rextents */
214	__u8 sb_inprogress; /* mkfs is in progress, don't mount */
215	__u8 sb_imax_pct; /* max % of fs for inode space */
216	/* statistics */
217	/*
218	* These fields must remain contiguous. If you really
219	* want to change their layout, make sure you fix the
220	* code in xfs_trans_apply_sb_deltas().
221	*/
222	__be64 sb_icount; /* allocated inodes */
223	__be64 sb_ifree; /* free inodes */
224	__be64 sb_fdblocks; /* free data blocks */
225	__be64 sb_frextents; /* free realtime extents */
226	/*
227	* End contiguous fields.
228	*/
229	__be64 sb_uquotino; /* user quota inode */
230	__be64 sb_gquotino; /* group quota inode */
231	__be16 sb_qflags; /* quota flags */
232	__u8 sb_flags; /* misc. flags */
233	__u8 sb_shared_vn; /* shared version number */
234	__be32 sb_inoalignmt; /* inode chunk alignment, fsblocks */
235	__be32 sb_unit; /* stripe or raid unit */
236	__be32 sb_width; /* stripe or raid width */
237	__u8 sb_dirblklog; /* log2 of dir block size (fsbs) */
238	__u8 sb_logsectlog; /* log2 of the log sector size */
239	__be16 sb_logsectsize; /* sector size for the log, bytes */
240	__be32 sb_logsunit; /* stripe unit size for the log */
241	__be32 sb_features2; /* additional feature bits */
242	/*
243	* bad features2 field as a result of failing to pad the sb
244	* structure to 64 bits. Some machines will be using this field
245	* for features2 bits. Easiest just to mark it bad and not use
246	* it for anything else.
247	*/
248	__be32 sb_bad_features2;
249
250	/* version 5 superblock fields start here */
251
252	/* feature masks */
253	__be32 sb_features_compat;
254	__be32 sb_features_ro_compat;
255	__be32 sb_features_incompat;
256	__be32 sb_features_log_incompat;
257
258	__le32 sb_crc; /* superblock crc */
259	__be32 sb_spino_align; /* sparse inode chunk alignment */
260
261	__be64 sb_pquotino; /* project quota inode */
262	__be64 sb_lsn; /* last write sequence */
263	uuid_t sb_meta_uuid; /* metadata file system unique id */
264
265	/* must be padded to 64 bit alignment */
266	};
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
282	static 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	*/
291	static inline bool xfs_sb_has_mismatched_features2(struct xfs_sb *sbp)
292	{
293	return sbp->sb_bad_features2 != sbp->sb_features2;
294	}
295
296	static 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
302	static inline void xfs_sb_version_addattr(struct xfs_sb *sbp)
303	{
304	sbp->sb_versionnum |= XFS_SB_VERSION_ATTRBIT;
305	}
306
307	static inline void xfs_sb_version_addquota(struct xfs_sb *sbp)
308	{
309	sbp->sb_versionnum |= XFS_SB_VERSION_QUOTABIT;
310	}
311
312	static 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
318	static 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
344	static inline bool
345	xfs_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
362	static inline bool
363	xfs_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
385	static inline bool
386	xfs_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
397	static inline bool
398	xfs_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
405	static inline void
406	xfs_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
412	static inline void
413	xfs_sb_add_incompat_log_features(
414	struct xfs_sb *sbp,
415	unsigned int features)
416	{
417	sbp->sb_features_log_incompat |= features;
418	}
419
420	static 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
426	static inline bool
427	xfs_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	* agf_cnt_level in the first AGF overlaps the EFS superblock's magic number.
481	* Since the magic numbers valid for EFS are > 64k, our value cannot be confused
482	* for an EFS superblock.
483	*/
484
485	typedef struct xfs_agf {
486	/*
487	* Common allocation group header information
488	*/
489	__be32 agf_magicnum; /* magic number == XFS_AGF_MAGIC */
490	__be32 agf_versionnum; /* header version == XFS_AGF_VERSION */
491	__be32 agf_seqno; /* sequence # starting from 0 */
492	__be32 agf_length; /* size in blocks of a.g. */
493	/*
494	* Freespace and rmap information
495	*/
496	__be32 agf_bno_root; /* bnobt root block */
497	__be32 agf_cnt_root; /* cntbt root block */
498	__be32 agf_rmap_root; /* rmapbt root block */
499
500	__be32 agf_bno_level; /* bnobt btree levels */
501	__be32 agf_cnt_level; /* cntbt btree levels */
502	__be32 agf_rmap_level; /* rmapbt btree levels */
503
504	__be32 agf_flfirst; /* first freelist block's index */
505	__be32 agf_fllast; /* last freelist block's index */
506	__be32 agf_flcount; /* count of blocks in freelist */
507	__be32 agf_freeblks; /* total free blocks */
508
509	__be32 agf_longest; /* longest free space */
510	__be32 agf_btreeblks; /* # of blocks held in AGF btrees */
511	uuid_t agf_uuid; /* uuid of filesystem */
512
513	__be32 agf_rmap_blocks; /* rmapbt blocks used */
514	__be32 agf_refcount_blocks; /* refcountbt blocks used */
515
516	__be32 agf_refcount_root; /* refcount tree root block */
517	__be32 agf_refcount_level; /* refcount btree levels */
518
519	/*
520	* reserve some contiguous space for future logged fields before we add
521	* the unlogged fields. This makes the range logging via flags and
522	* structure offsets much simpler.
523	*/
524	__be64 agf_spare64[14];
525
526	/* unlogged fields, written during buffer writeback. */
527	__be64 agf_lsn; /* last write sequence */
528	__be32 agf_crc; /* crc of agf sector */
529	__be32 agf_spare2;
530
531	/* structure must be padded to 64 bit alignment */
532	} xfs_agf_t;
533
534	#define XFS_AGF_CRC_OFF offsetof(struct xfs_agf, agf_crc)
535
536	#define XFS_AGF_MAGICNUM (1u << 0)
537	#define XFS_AGF_VERSIONNUM (1u << 1)
538	#define XFS_AGF_SEQNO (1u << 2)
539	#define XFS_AGF_LENGTH (1u << 3)
540	#define XFS_AGF_ROOTS (1u << 4)
541	#define XFS_AGF_LEVELS (1u << 5)
542	#define XFS_AGF_FLFIRST (1u << 6)
543	#define XFS_AGF_FLLAST (1u << 7)
544	#define XFS_AGF_FLCOUNT (1u << 8)
545	#define XFS_AGF_FREEBLKS (1u << 9)
546	#define XFS_AGF_LONGEST (1u << 10)
547	#define XFS_AGF_BTREEBLKS (1u << 11)
548	#define XFS_AGF_UUID (1u << 12)
549	#define XFS_AGF_RMAP_BLOCKS (1u << 13)
550	#define XFS_AGF_REFCOUNT_BLOCKS (1u << 14)
551	#define XFS_AGF_REFCOUNT_ROOT (1u << 15)
552	#define XFS_AGF_REFCOUNT_LEVEL (1u << 16)
553	#define XFS_AGF_SPARE64 (1u << 17)
554	#define XFS_AGF_NUM_BITS 18
555	#define XFS_AGF_ALL_BITS ((1u << XFS_AGF_NUM_BITS) - 1)
556
557	#define XFS_AGF_FLAGS \
558	{ XFS_AGF_MAGICNUM, "MAGICNUM" }, \
559	{ XFS_AGF_VERSIONNUM, "VERSIONNUM" }, \
560	{ XFS_AGF_SEQNO, "SEQNO" }, \
561	{ XFS_AGF_LENGTH, "LENGTH" }, \
562	{ XFS_AGF_ROOTS, "ROOTS" }, \
563	{ XFS_AGF_LEVELS, "LEVELS" }, \
564	{ XFS_AGF_FLFIRST, "FLFIRST" }, \
565	{ XFS_AGF_FLLAST, "FLLAST" }, \
566	{ XFS_AGF_FLCOUNT, "FLCOUNT" }, \
567	{ XFS_AGF_FREEBLKS, "FREEBLKS" }, \
568	{ XFS_AGF_LONGEST, "LONGEST" }, \
569	{ XFS_AGF_BTREEBLKS, "BTREEBLKS" }, \
570	{ XFS_AGF_UUID, "UUID" }, \
571	{ XFS_AGF_RMAP_BLOCKS, "RMAP_BLOCKS" }, \
572	{ XFS_AGF_REFCOUNT_BLOCKS, "REFCOUNT_BLOCKS" }, \
573	{ XFS_AGF_REFCOUNT_ROOT, "REFCOUNT_ROOT" }, \
574	{ XFS_AGF_REFCOUNT_LEVEL, "REFCOUNT_LEVEL" }, \
575	{ XFS_AGF_SPARE64, "SPARE64" }
576
577	/* disk block (xfs_daddr_t) in the AG */
578	#define XFS_AGF_DADDR(mp) ((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
579	#define XFS_AGF_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
580
581	/*
582	* Size of the unlinked inode hash table in the agi.
583	*/
584	#define XFS_AGI_UNLINKED_BUCKETS 64
585
586	typedef struct xfs_agi {
587	/*
588	* Common allocation group header information
589	*/
590	__be32 agi_magicnum; /* magic number == XFS_AGI_MAGIC */
591	__be32 agi_versionnum; /* header version == XFS_AGI_VERSION */
592	__be32 agi_seqno; /* sequence # starting from 0 */
593	__be32 agi_length; /* size in blocks of a.g. */
594	/*
595	* Inode information
596	* Inodes are mapped by interpreting the inode number, so no
597	* mapping data is needed here.
598	*/
599	__be32 agi_count; /* count of allocated inodes */
600	__be32 agi_root; /* root of inode btree */
601	__be32 agi_level; /* levels in inode btree */
602	__be32 agi_freecount; /* number of free inodes */
603
604	__be32 agi_newino; /* new inode just allocated */
605	__be32 agi_dirino; /* last directory inode chunk */
606	/*
607	* Hash table of inodes which have been unlinked but are
608	* still being referenced.
609	*/
610	__be32 agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
611	/*
612	* This marks the end of logging region 1 and start of logging region 2.
613	*/
614	uuid_t agi_uuid; /* uuid of filesystem */
615	__be32 agi_crc; /* crc of agi sector */
616	__be32 agi_pad32;
617	__be64 agi_lsn; /* last write sequence */
618
619	__be32 agi_free_root; /* root of the free inode btree */
620	__be32 agi_free_level;/* levels in free inode btree */
621
622	__be32 agi_iblocks; /* inobt blocks used */
623	__be32 agi_fblocks; /* finobt blocks used */
624
625	/* structure must be padded to 64 bit alignment */
626	} xfs_agi_t;
627
628	#define XFS_AGI_CRC_OFF offsetof(struct xfs_agi, agi_crc)
629
630	#define XFS_AGI_MAGICNUM (1u << 0)
631	#define XFS_AGI_VERSIONNUM (1u << 1)
632	#define XFS_AGI_SEQNO (1u << 2)
633	#define XFS_AGI_LENGTH (1u << 3)
634	#define XFS_AGI_COUNT (1u << 4)
635	#define XFS_AGI_ROOT (1u << 5)
636	#define XFS_AGI_LEVEL (1u << 6)
637	#define XFS_AGI_FREECOUNT (1u << 7)
638	#define XFS_AGI_NEWINO (1u << 8)
639	#define XFS_AGI_DIRINO (1u << 9)
640	#define XFS_AGI_UNLINKED (1u << 10)
641	#define XFS_AGI_NUM_BITS_R1 11 /* end of the 1st agi logging region */
642	#define XFS_AGI_ALL_BITS_R1 ((1u << XFS_AGI_NUM_BITS_R1) - 1)
643	#define XFS_AGI_FREE_ROOT (1u << 11)
644	#define XFS_AGI_FREE_LEVEL (1u << 12)
645	#define XFS_AGI_IBLOCKS (1u << 13) /* both inobt/finobt block counters */
646	#define XFS_AGI_NUM_BITS_R2 14
647
648	/* disk block (xfs_daddr_t) in the AG */
649	#define XFS_AGI_DADDR(mp) ((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
650	#define XFS_AGI_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
651
652	/*
653	* The third a.g. block contains the a.g. freelist, an array
654	* of block pointers to blocks owned by the allocation btree code.
655	*/
656	#define XFS_AGFL_DADDR(mp) ((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
657	#define XFS_AGFL_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
658	#define XFS_BUF_TO_AGFL(bp) ((struct xfs_agfl *)((bp)->b_addr))
659
660	struct xfs_agfl {
661	__be32 agfl_magicnum;
662	__be32 agfl_seqno;
663	uuid_t agfl_uuid;
664	__be64 agfl_lsn;
665	__be32 agfl_crc;
666	} __attribute__((packed));
667
668	#define XFS_AGFL_CRC_OFF offsetof(struct xfs_agfl, agfl_crc)
669
670	#define XFS_AGB_TO_FSB(mp,agno,agbno) \
671	(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
672	#define XFS_FSB_TO_AGNO(mp,fsbno) \
673	((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
674	#define XFS_FSB_TO_AGBNO(mp,fsbno) \
675	((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
676	#define XFS_AGB_TO_DADDR(mp,agno,agbno) \
677	((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
678	(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
679	#define XFS_AG_DADDR(mp,agno,d) (XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
680
681	/*
682	* For checking for bad ranges of xfs_daddr_t's, covering multiple
683	* allocation groups or a single xfs_daddr_t that's a superblock copy.
684	*/
685	#define XFS_AG_CHECK_DADDR(mp,d,len) \
686	((len) == 1 ? \
687	ASSERT((d) == XFS_SB_DADDR || \
688	xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
689	ASSERT(xfs_daddr_to_agno(mp, d) == \
690	xfs_daddr_to_agno(mp, (d) + (len) - 1)))
691
692	/*
693	* Realtime bitmap information is accessed by the word, which is currently
694	* stored in host-endian format.
695	*/
696	union xfs_rtword_raw {
697	__u32 old;
698	};
699
700	/*
701	* Realtime summary counts are accessed by the word, which is currently
702	* stored in host-endian format.
703	*/
704	union xfs_suminfo_raw {
705	__u32 old;
706	};
707
708	/*
709	* XFS Timestamps
710	* ==============
711	*
712	* Traditional ondisk inode timestamps consist of signed 32-bit counters for
713	* seconds and nanoseconds; time zero is the Unix epoch, Jan 1 00:00:00 UTC
714	* 1970, which means that the timestamp epoch is the same as the Unix epoch.
715	* Therefore, the ondisk min and max defined here can be used directly to
716	* constrain the incore timestamps on a Unix system. Note that we actually
717	* encode a __be64 value on disk.
718	*
719	* When the bigtime feature is enabled, ondisk inode timestamps become an
720	* unsigned 64-bit nanoseconds counter. This means that the bigtime inode
721	* timestamp epoch is the start of the classic timestamp range, which is
722	* Dec 13 20:45:52 UTC 1901. Because the epochs are not the same, callers
723	* /must/ use the bigtime conversion functions when encoding and decoding raw
724	* timestamps.
725	*/
726	typedef __be64 xfs_timestamp_t;
727
728	/* Legacy timestamp encoding format. */
729	struct xfs_legacy_timestamp {
730	__be32 t_sec; /* timestamp seconds */
731	__be32 t_nsec; /* timestamp nanoseconds */
732	};
733
734	/*
735	* Smallest possible ondisk seconds value with traditional timestamps. This
736	* corresponds exactly with the incore timestamp Dec 13 20:45:52 UTC 1901.
737	*/
738	#define XFS_LEGACY_TIME_MIN ((int64_t)S32_MIN)
739
740	/*
741	* Largest possible ondisk seconds value with traditional timestamps. This
742	* corresponds exactly with the incore timestamp Jan 19 03:14:07 UTC 2038.
743	*/
744	#define XFS_LEGACY_TIME_MAX ((int64_t)S32_MAX)
745
746	/*
747	* Smallest possible ondisk seconds value with bigtime timestamps. This
748	* corresponds (after conversion to a Unix timestamp) with the traditional
749	* minimum timestamp of Dec 13 20:45:52 UTC 1901.
750	*/
751	#define XFS_BIGTIME_TIME_MIN ((int64_t)0)
752
753	/*
754	* Largest supported ondisk seconds value with bigtime timestamps. This
755	* corresponds (after conversion to a Unix timestamp) with an incore timestamp
756	* of Jul 2 20:20:24 UTC 2486.
757	*
758	* We round down the ondisk limit so that the bigtime quota and inode max
759	* timestamps will be the same.
760	*/
761	#define XFS_BIGTIME_TIME_MAX ((int64_t)((-1ULL / NSEC_PER_SEC) & ~0x3ULL))
762
763	/*
764	* Bigtime epoch is set exactly to the minimum time value that a traditional
765	* 32-bit timestamp can represent when using the Unix epoch as a reference.
766	* Hence the Unix epoch is at a fixed offset into the supported bigtime
767	* timestamp range.
768	*
769	* The bigtime epoch also matches the minimum value an on-disk 32-bit XFS
770	* timestamp can represent so we will not lose any fidelity in converting
771	* to/from unix and bigtime timestamps.
772	*
773	* The following conversion factor converts a seconds counter from the Unix
774	* epoch to the bigtime epoch.
775	*/
776	#define XFS_BIGTIME_EPOCH_OFFSET (-(int64_t)S32_MIN)
777
778	/* Convert a timestamp from the Unix epoch to the bigtime epoch. */
779	static inline uint64_t xfs_unix_to_bigtime(time64_t unix_seconds)
780	{
781	return (uint64_t)unix_seconds + XFS_BIGTIME_EPOCH_OFFSET;
782	}
783
784	/* Convert a timestamp from the bigtime epoch to the Unix epoch. */
785	static inline time64_t xfs_bigtime_to_unix(uint64_t ondisk_seconds)
786	{
787	return (time64_t)ondisk_seconds - XFS_BIGTIME_EPOCH_OFFSET;
788	}
789
790	/*
791	* On-disk inode structure.
792	*
793	* This is just the header or "dinode core", the inode is expanded to fill a
794	* variable size the leftover area split into a data and an attribute fork.
795	* The format of the data and attribute fork depends on the format of the
796	* inode as indicated by di_format and di_aformat. To access the data and
797	* attribute use the XFS_DFORK_DPTR, XFS_DFORK_APTR, and XFS_DFORK_PTR macros
798	* below.
799	*
800	* There is a very similar struct xfs_log_dinode which matches the layout of
801	* this structure, but is kept in native format instead of big endian.
802	*
803	* Note: di_flushiter is only used by v1/2 inodes - it's effectively a zeroed
804	* padding field for v3 inodes.
805	*/
806	#define XFS_DINODE_MAGIC 0x494e /* 'IN' */
807	struct xfs_dinode {
808	__be16 di_magic; /* inode magic # = XFS_DINODE_MAGIC */
809	__be16 di_mode; /* mode and type of file */
810	__u8 di_version; /* inode version */
811	__u8 di_format; /* format of di_c data */
812	__be16 di_onlink; /* old number of links to file */
813	__be32 di_uid; /* owner's user id */
814	__be32 di_gid; /* owner's group id */
815	__be32 di_nlink; /* number of links to file */
816	__be16 di_projid_lo; /* lower part of owner's project id */
817	__be16 di_projid_hi; /* higher part owner's project id */
818	union {
819	/* Number of data fork extents if NREXT64 is set */
820	__be64 di_big_nextents;
821
822	/* Padding for V3 inodes without NREXT64 set. */
823	__be64 di_v3_pad;
824
825	/* Padding and inode flush counter for V2 inodes. */
826	struct {
827	__u8 di_v2_pad[6];
828	__be16 di_flushiter;
829	};
830	};
831	xfs_timestamp_t di_atime; /* time last accessed */
832	xfs_timestamp_t di_mtime; /* time last modified */
833	xfs_timestamp_t di_ctime; /* time created/inode modified */
834	__be64 di_size; /* number of bytes in file */
835	__be64 di_nblocks; /* # of direct & btree blocks used */
836	__be32 di_extsize; /* basic/minimum extent size for file */
837	union {
838	/*
839	* For V2 inodes and V3 inodes without NREXT64 set, this
840	* is the number of data and attr fork extents.
841	*/
842	struct {
843	__be32 di_nextents;
844	__be16 di_anextents;
845	} __packed;
846
847	/* Number of attr fork extents if NREXT64 is set. */
848	struct {
849	__be32 di_big_anextents;
850	__be16 di_nrext64_pad;
851	} __packed;
852	} __packed;
853	__u8 di_forkoff; /* attr fork offs, <<3 for 64b align */
854	__s8 di_aformat; /* format of attr fork's data */
855	__be32 di_dmevmask; /* DMIG event mask */
856	__be16 di_dmstate; /* DMIG state info */
857	__be16 di_flags; /* random flags, XFS_DIFLAG_... */
858	__be32 di_gen; /* generation number */
859
860	/* di_next_unlinked is the only non-core field in the old dinode */
861	__be32 di_next_unlinked;/* agi unlinked list ptr */
862
863	/* start of the extended dinode, writable fields */
864	__le32 di_crc; /* CRC of the inode */
865	__be64 di_changecount; /* number of attribute changes */
866	__be64 di_lsn; /* flush sequence */
867	__be64 di_flags2; /* more random flags */
868	__be32 di_cowextsize; /* basic cow extent size for file */
869	__u8 di_pad2[12]; /* more padding for future expansion */
870
871	/* fields only written to during inode creation */
872	xfs_timestamp_t di_crtime; /* time created */
873	__be64 di_ino; /* inode number */
874	uuid_t di_uuid; /* UUID of the filesystem */
875
876	/* structure must be padded to 64 bit alignment */
877	};
878
879	#define XFS_DINODE_CRC_OFF offsetof(struct xfs_dinode, di_crc)
880
881	#define DI_MAX_FLUSH 0xffff
882
883	/*
884	* Size of the core inode on disk. Version 1 and 2 inodes have
885	* the same size, but version 3 has grown a few additional fields.
886	*/
887	static inline uint xfs_dinode_size(int version)
888	{
889	if (version == 3)
890	return sizeof(struct xfs_dinode);
891	return offsetof(struct xfs_dinode, di_crc);
892	}
893
894	/*
895	* The 32 bit link count in the inode theoretically maxes out at UINT_MAX.
896	* Since the pathconf interface is signed, we use 2^31 - 1 instead.
897	*/
898	#define XFS_MAXLINK ((1U << 31) - 1U)
899
900	/*
901	* Values for di_format
902	*
903	* This enum is used in string mapping in xfs_trace.h; please keep the
904	* TRACE_DEFINE_ENUMs for it up to date.
905	*/
906	enum xfs_dinode_fmt {
907	XFS_DINODE_FMT_DEV, /* xfs_dev_t */
908	XFS_DINODE_FMT_LOCAL, /* bulk data */
909	XFS_DINODE_FMT_EXTENTS, /* struct xfs_bmbt_rec */
910	XFS_DINODE_FMT_BTREE, /* struct xfs_bmdr_block */
911	XFS_DINODE_FMT_UUID /* added long ago, but never used */
912	};
913
914	#define XFS_INODE_FORMAT_STR \
915	{ XFS_DINODE_FMT_DEV, "dev" }, \
916	{ XFS_DINODE_FMT_LOCAL, "local" }, \
917	{ XFS_DINODE_FMT_EXTENTS, "extent" }, \
918	{ XFS_DINODE_FMT_BTREE, "btree" }, \
919	{ XFS_DINODE_FMT_UUID, "uuid" }
920
921	/*
922	* Max values for extnum and aextnum.
923	*
924	* The original on-disk extent counts were held in signed fields, resulting in
925	* maximum extent counts of 2^31 and 2^15 for the data and attr forks
926	* respectively. Similarly the maximum extent length is limited to 2^21 blocks
927	* by the 21-bit wide blockcount field of a BMBT extent record.
928	*
929	* The newly introduced data fork extent counter can hold a 64-bit value,
930	* however the maximum number of extents in a file is also limited to 2^54
931	* extents by the 54-bit wide startoff field of a BMBT extent record.
932	*
933	* It is further limited by the maximum supported file size of 2^63
934	* *bytes*. This leads to a maximum extent count for maximally sized filesystem
935	* blocks (64kB) of:
936	*
937	* 2^63 bytes / 2^16 bytes per block = 2^47 blocks
938	*
939	* Rounding up 47 to the nearest multiple of bits-per-byte results in 48. Hence
940	* 2^48 was chosen as the maximum data fork extent count.
941	*
942	* The maximum file size that can be represented by the data fork extent counter
943	* in the worst case occurs when all extents are 1 block in length and each
944	* block is 1KB in size.
945	*
946	* With XFS_MAX_EXTCNT_DATA_FORK_SMALL representing maximum extent count and
947	* with 1KB sized blocks, a file can reach upto,
948	* 1KB * (2^31) = 2TB
949	*
950	* This is much larger than the theoretical maximum size of a directory
951	* i.e. XFS_DIR2_SPACE_SIZE * XFS_DIR2_MAX_SPACES = ~96GB.
952	*
953	* Hence, a directory inode can never overflow its data fork extent counter.
954	*/
955	#define XFS_MAX_EXTCNT_DATA_FORK_LARGE ((xfs_extnum_t)((1ULL << 48) - 1))
956	#define XFS_MAX_EXTCNT_ATTR_FORK_LARGE ((xfs_extnum_t)((1ULL << 32) - 1))
957	#define XFS_MAX_EXTCNT_DATA_FORK_SMALL ((xfs_extnum_t)((1ULL << 31) - 1))
958	#define XFS_MAX_EXTCNT_ATTR_FORK_SMALL ((xfs_extnum_t)((1ULL << 15) - 1))
959
960	/*
961	* When we upgrade an inode to the large extent counts, the maximum value by
962	* which the extent count can increase is bound by the change in size of the
963	* on-disk field. No upgrade operation should ever be adding more than a few
964	* tens of extents, so if we get a really large value it is a sign of a code bug
965	* or corruption.
966	*/
967	#define XFS_MAX_EXTCNT_UPGRADE_NR \
968	min(XFS_MAX_EXTCNT_ATTR_FORK_LARGE - XFS_MAX_EXTCNT_ATTR_FORK_SMALL, \
969	XFS_MAX_EXTCNT_DATA_FORK_LARGE - XFS_MAX_EXTCNT_DATA_FORK_SMALL)
970
971	/*
972	* Inode minimum and maximum sizes.
973	*/
974	#define XFS_DINODE_MIN_LOG 8
975	#define XFS_DINODE_MAX_LOG 11
976	#define XFS_DINODE_MIN_SIZE (1 << XFS_DINODE_MIN_LOG)
977	#define XFS_DINODE_MAX_SIZE (1 << XFS_DINODE_MAX_LOG)
978
979	/*
980	* Inode size for given fs.
981	*/
982	#define XFS_DINODE_SIZE(mp) \
983	(xfs_has_v3inodes(mp) ? \
984	sizeof(struct xfs_dinode) : \
985	offsetof(struct xfs_dinode, di_crc))
986	#define XFS_LITINO(mp) \
987	((mp)->m_sb.sb_inodesize - XFS_DINODE_SIZE(mp))
988
989	/*
990	* Inode data & attribute fork sizes, per inode.
991	*/
992	#define XFS_DFORK_BOFF(dip) ((int)((dip)->di_forkoff << 3))
993
994	#define XFS_DFORK_DSIZE(dip,mp) \
995	((dip)->di_forkoff ? XFS_DFORK_BOFF(dip) : XFS_LITINO(mp))
996	#define XFS_DFORK_ASIZE(dip,mp) \
997	((dip)->di_forkoff ? XFS_LITINO(mp) - XFS_DFORK_BOFF(dip) : 0)
998	#define XFS_DFORK_SIZE(dip,mp,w) \
999	((w) == XFS_DATA_FORK ? \
1000	XFS_DFORK_DSIZE(dip, mp) : \
1001	XFS_DFORK_ASIZE(dip, mp))
1002
1003	#define XFS_DFORK_MAXEXT(dip, mp, w) \
1004	(XFS_DFORK_SIZE(dip, mp, w) / sizeof(struct xfs_bmbt_rec))
1005
1006	/*
1007	* Return pointers to the data or attribute forks.
1008	*/
1009	#define XFS_DFORK_DPTR(dip) \
1010	((void *)dip + xfs_dinode_size(dip->di_version))
1011	#define XFS_DFORK_APTR(dip) \
1012	(XFS_DFORK_DPTR(dip) + XFS_DFORK_BOFF(dip))
1013	#define XFS_DFORK_PTR(dip,w) \
1014	((w) == XFS_DATA_FORK ? XFS_DFORK_DPTR(dip) : XFS_DFORK_APTR(dip))
1015
1016	#define XFS_DFORK_FORMAT(dip,w) \
1017	((w) == XFS_DATA_FORK ? \
1018	(dip)->di_format : \
1019	(dip)->di_aformat)
1020
1021	/*
1022	* For block and character special files the 32bit dev_t is stored at the
1023	* beginning of the data fork.
1024	*/
1025	static inline xfs_dev_t xfs_dinode_get_rdev(struct xfs_dinode *dip)
1026	{
1027	return be32_to_cpu(*(__be32 *)XFS_DFORK_DPTR(dip));
1028	}
1029
1030	static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
1031	{
1032	*(__be32 *)XFS_DFORK_DPTR(dip) = cpu_to_be32(rdev);
1033	}
1034
1035	/*
1036	* Values for di_flags
1037	*/
1038	#define XFS_DIFLAG_REALTIME_BIT 0 /* file's blocks come from rt area */
1039	#define XFS_DIFLAG_PREALLOC_BIT 1 /* file space has been preallocated */
1040	#define XFS_DIFLAG_NEWRTBM_BIT 2 /* for rtbitmap inode, new format */
1041	#define XFS_DIFLAG_IMMUTABLE_BIT 3 /* inode is immutable */
1042	#define XFS_DIFLAG_APPEND_BIT 4 /* inode is append-only */
1043	#define XFS_DIFLAG_SYNC_BIT 5 /* inode is written synchronously */
1044	#define XFS_DIFLAG_NOATIME_BIT 6 /* do not update atime */
1045	#define XFS_DIFLAG_NODUMP_BIT 7 /* do not dump */
1046	#define XFS_DIFLAG_RTINHERIT_BIT 8 /* create with realtime bit set */
1047	#define XFS_DIFLAG_PROJINHERIT_BIT 9 /* create with parents projid */
1048	#define XFS_DIFLAG_NOSYMLINKS_BIT 10 /* disallow symlink creation */
1049	#define XFS_DIFLAG_EXTSIZE_BIT 11 /* inode extent size allocator hint */
1050	#define XFS_DIFLAG_EXTSZINHERIT_BIT 12 /* inherit inode extent size */
1051	#define XFS_DIFLAG_NODEFRAG_BIT 13 /* do not reorganize/defragment */
1052	#define XFS_DIFLAG_FILESTREAM_BIT 14 /* use filestream allocator */
1053	/* Do not use bit 15, di_flags is legacy and unchanging now */
1054
1055	#define XFS_DIFLAG_REALTIME (1 << XFS_DIFLAG_REALTIME_BIT)
1056	#define XFS_DIFLAG_PREALLOC (1 << XFS_DIFLAG_PREALLOC_BIT)
1057	#define XFS_DIFLAG_NEWRTBM (1 << XFS_DIFLAG_NEWRTBM_BIT)
1058	#define XFS_DIFLAG_IMMUTABLE (1 << XFS_DIFLAG_IMMUTABLE_BIT)
1059	#define XFS_DIFLAG_APPEND (1 << XFS_DIFLAG_APPEND_BIT)
1060	#define XFS_DIFLAG_SYNC (1 << XFS_DIFLAG_SYNC_BIT)
1061	#define XFS_DIFLAG_NOATIME (1 << XFS_DIFLAG_NOATIME_BIT)
1062	#define XFS_DIFLAG_NODUMP (1 << XFS_DIFLAG_NODUMP_BIT)
1063	#define XFS_DIFLAG_RTINHERIT (1 << XFS_DIFLAG_RTINHERIT_BIT)
1064	#define XFS_DIFLAG_PROJINHERIT (1 << XFS_DIFLAG_PROJINHERIT_BIT)
1065	#define XFS_DIFLAG_NOSYMLINKS (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
1066	#define XFS_DIFLAG_EXTSIZE (1 << XFS_DIFLAG_EXTSIZE_BIT)
1067	#define XFS_DIFLAG_EXTSZINHERIT (1 << XFS_DIFLAG_EXTSZINHERIT_BIT)
1068	#define XFS_DIFLAG_NODEFRAG (1 << XFS_DIFLAG_NODEFRAG_BIT)
1069	#define XFS_DIFLAG_FILESTREAM (1 << XFS_DIFLAG_FILESTREAM_BIT)
1070
1071	#define XFS_DIFLAG_ANY \
1072	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
1073	XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
1074	XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
1075	XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS | XFS_DIFLAG_EXTSIZE | \
1076	XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
1077
1078	/*
1079	* Values for di_flags2 These start by being exposed to userspace in the upper
1080	* 16 bits of the XFS_XFLAG_s range.
1081	*/
1082	#define XFS_DIFLAG2_DAX_BIT 0 /* use DAX for this inode */
1083	#define XFS_DIFLAG2_REFLINK_BIT 1 /* file's blocks may be shared */
1084	#define XFS_DIFLAG2_COWEXTSIZE_BIT 2 /* copy on write extent size hint */
1085	#define XFS_DIFLAG2_BIGTIME_BIT 3 /* big timestamps */
1086	#define XFS_DIFLAG2_NREXT64_BIT 4 /* large extent counters */
1087
1088	#define XFS_DIFLAG2_DAX (1 << XFS_DIFLAG2_DAX_BIT)
1089	#define XFS_DIFLAG2_REFLINK (1 << XFS_DIFLAG2_REFLINK_BIT)
1090	#define XFS_DIFLAG2_COWEXTSIZE (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
1091	#define XFS_DIFLAG2_BIGTIME (1 << XFS_DIFLAG2_BIGTIME_BIT)
1092	#define XFS_DIFLAG2_NREXT64 (1 << XFS_DIFLAG2_NREXT64_BIT)
1093
1094	#define XFS_DIFLAG2_ANY \
1095	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE | \
1096	XFS_DIFLAG2_BIGTIME | XFS_DIFLAG2_NREXT64)
1097
1098	static inline bool xfs_dinode_has_bigtime(const struct xfs_dinode *dip)
1099	{
1100	return dip->di_version >= 3 &&
1101	(dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_BIGTIME));
1102	}
1103
1104	static inline bool xfs_dinode_has_large_extent_counts(
1105	const struct xfs_dinode *dip)
1106	{
1107	return dip->di_version >= 3 &&
1108	(dip->di_flags2 & cpu_to_be64(XFS_DIFLAG2_NREXT64));
1109	}
1110
1111	/*
1112	* Inode number format:
1113	* low inopblog bits - offset in block
1114	* next agblklog bits - block number in ag
1115	* next agno_log bits - ag number
1116	* high agno_log-agblklog-inopblog bits - 0
1117	*/
1118	#define XFS_INO_MASK(k) (uint32_t)((1ULL << (k)) - 1)
1119	#define XFS_INO_OFFSET_BITS(mp) (mp)->m_sb.sb_inopblog
1120	#define XFS_INO_AGBNO_BITS(mp) (mp)->m_sb.sb_agblklog
1121	#define XFS_INO_AGINO_BITS(mp) ((mp)->m_ino_geo.agino_log)
1122	#define XFS_INO_AGNO_BITS(mp) (mp)->m_agno_log
1123	#define XFS_INO_BITS(mp) \
1124	XFS_INO_AGNO_BITS(mp) + XFS_INO_AGINO_BITS(mp)
1125	#define XFS_INO_TO_AGNO(mp,i) \
1126	((xfs_agnumber_t)((i) >> XFS_INO_AGINO_BITS(mp)))
1127	#define XFS_INO_TO_AGINO(mp,i) \
1128	((xfs_agino_t)(i) & XFS_INO_MASK(XFS_INO_AGINO_BITS(mp)))
1129	#define XFS_INO_TO_AGBNO(mp,i) \
1130	(((xfs_agblock_t)(i) >> XFS_INO_OFFSET_BITS(mp)) & \
1131	XFS_INO_MASK(XFS_INO_AGBNO_BITS(mp)))
1132	#define XFS_INO_TO_OFFSET(mp,i) \
1133	((int)(i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1134	#define XFS_INO_TO_FSB(mp,i) \
1135	XFS_AGB_TO_FSB(mp, XFS_INO_TO_AGNO(mp,i), XFS_INO_TO_AGBNO(mp,i))
1136	#define XFS_AGINO_TO_INO(mp,a,i) \
1137	(((xfs_ino_t)(a) << XFS_INO_AGINO_BITS(mp)) | (i))
1138	#define XFS_AGINO_TO_AGBNO(mp,i) ((i) >> XFS_INO_OFFSET_BITS(mp))
1139	#define XFS_AGINO_TO_OFFSET(mp,i) \
1140	((i) & XFS_INO_MASK(XFS_INO_OFFSET_BITS(mp)))
1141	#define XFS_OFFBNO_TO_AGINO(mp,b,o) \
1142	((xfs_agino_t)(((b) << XFS_INO_OFFSET_BITS(mp)) | (o)))
1143	#define XFS_FSB_TO_INO(mp, b) ((xfs_ino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1144	#define XFS_AGB_TO_AGINO(mp, b) ((xfs_agino_t)((b) << XFS_INO_OFFSET_BITS(mp)))
1145
1146	#define XFS_MAXINUMBER ((xfs_ino_t)((1ULL << 56) - 1ULL))
1147	#define XFS_MAXINUMBER_32 ((xfs_ino_t)((1ULL << 32) - 1ULL))
1148
1149	/*
1150	* RealTime Device format definitions
1151	*/
1152
1153	/* Min and max rt extent sizes, specified in bytes */
1154	#define XFS_MAX_RTEXTSIZE (1024 * 1024 * 1024) /* 1GB */
1155	#define XFS_DFL_RTEXTSIZE (64 * 1024) /* 64kB */
1156	#define XFS_MIN_RTEXTSIZE (4 * 1024) /* 4kB */
1157
1158	/*
1159	* Dquot and dquot block format definitions
1160	*/
1161	#define XFS_DQUOT_MAGIC 0x4451 /* 'DQ' */
1162	#define XFS_DQUOT_VERSION (uint8_t)0x01 /* latest version number */
1163
1164	#define XFS_DQTYPE_USER (1u << 0) /* user dquot record */
1165	#define XFS_DQTYPE_PROJ (1u << 1) /* project dquot record */
1166	#define XFS_DQTYPE_GROUP (1u << 2) /* group dquot record */
1167	#define XFS_DQTYPE_BIGTIME (1u << 7) /* large expiry timestamps */
1168
1169	/* bitmask to determine if this is a user/group/project dquot */
1170	#define XFS_DQTYPE_REC_MASK (XFS_DQTYPE_USER | \
1171	XFS_DQTYPE_PROJ | \
1172	XFS_DQTYPE_GROUP)
1173
1174	#define XFS_DQTYPE_ANY (XFS_DQTYPE_REC_MASK | \
1175	XFS_DQTYPE_BIGTIME)
1176
1177	/*
1178	* XFS Quota Timers
1179	* ================
1180	*
1181	* Traditional quota grace period expiration timers are an unsigned 32-bit
1182	* seconds counter; time zero is the Unix epoch, Jan 1 00:00:01 UTC 1970.
1183	* Note that an expiration value of zero means that the quota limit has not
1184	* been reached, and therefore no expiration has been set. Therefore, the
1185	* ondisk min and max defined here can be used directly to constrain the incore
1186	* quota expiration timestamps on a Unix system.
1187	*
1188	* When bigtime is enabled, we trade two bits of precision to expand the
1189	* expiration timeout range to match that of big inode timestamps. The min and
1190	* max recorded here are the on-disk limits, not a Unix timestamp.
1191	*
1192	* The grace period for each quota type is stored in the root dquot (id = 0)
1193	* and is applied to a non-root dquot when it exceeds the soft or hard limits.
1194	* The length of quota grace periods are unsigned 32-bit quantities measured in
1195	* units of seconds. A value of zero means to use the default period.
1196	*/
1197
1198	/*
1199	* Smallest possible ondisk quota expiration value with traditional timestamps.
1200	* This corresponds exactly with the incore expiration Jan 1 00:00:01 UTC 1970.
1201	*/
1202	#define XFS_DQ_LEGACY_EXPIRY_MIN ((int64_t)1)
1203
1204	/*
1205	* Largest possible ondisk quota expiration value with traditional timestamps.
1206	* This corresponds exactly with the incore expiration Feb 7 06:28:15 UTC 2106.
1207	*/
1208	#define XFS_DQ_LEGACY_EXPIRY_MAX ((int64_t)U32_MAX)
1209
1210	/*
1211	* Smallest possible ondisk quota expiration value with bigtime timestamps.
1212	* This corresponds (after conversion to a Unix timestamp) with the incore
1213	* expiration of Jan 1 00:00:04 UTC 1970.
1214	*/
1215	#define XFS_DQ_BIGTIME_EXPIRY_MIN (XFS_DQ_LEGACY_EXPIRY_MIN)
1216
1217	/*
1218	* Largest supported ondisk quota expiration value with bigtime timestamps.
1219	* This corresponds (after conversion to a Unix timestamp) with an incore
1220	* expiration of Jul 2 20:20:24 UTC 2486.
1221	*
1222	* The ondisk field supports values up to -1U, which corresponds to an incore
1223	* expiration in 2514. This is beyond the maximum the bigtime inode timestamp,
1224	* so we cap the maximum bigtime quota expiration to the max inode timestamp.
1225	*/
1226	#define XFS_DQ_BIGTIME_EXPIRY_MAX ((int64_t)4074815106U)
1227
1228	/*
1229	* The following conversion factors assist in converting a quota expiration
1230	* timestamp between the incore and ondisk formats.
1231	*/
1232	#define XFS_DQ_BIGTIME_SHIFT (2)
1233	#define XFS_DQ_BIGTIME_SLACK ((int64_t)(1ULL << XFS_DQ_BIGTIME_SHIFT) - 1)
1234
1235	/* Convert an incore quota expiration timestamp to an ondisk bigtime value. */
1236	static inline uint32_t xfs_dq_unix_to_bigtime(time64_t unix_seconds)
1237	{
1238	/*
1239	* Round the expiration timestamp up to the nearest bigtime timestamp
1240	* that we can store, to give users the most time to fix problems.
1241	*/
1242	return ((uint64_t)unix_seconds + XFS_DQ_BIGTIME_SLACK) >>
1243	XFS_DQ_BIGTIME_SHIFT;
1244	}
1245
1246	/* Convert an ondisk bigtime quota expiration value to an incore timestamp. */
1247	static inline time64_t xfs_dq_bigtime_to_unix(uint32_t ondisk_seconds)
1248	{
1249	return (time64_t)ondisk_seconds << XFS_DQ_BIGTIME_SHIFT;
1250	}
1251
1252	/*
1253	* Default quota grace periods, ranging from zero (use the compiled defaults)
1254	* to ~136 years. These are applied to a non-root dquot that has exceeded
1255	* either limit.
1256	*/
1257	#define XFS_DQ_GRACE_MIN ((int64_t)0)
1258	#define XFS_DQ_GRACE_MAX ((int64_t)U32_MAX)
1259
1260	/* Maximum id value for a quota record */
1261	#define XFS_DQ_ID_MAX (U32_MAX)
1262
1263	/*
1264	* This is the main portion of the on-disk representation of quota information
1265	* for a user. We pad this with some more expansion room to construct the on
1266	* disk structure.
1267	*/
1268	struct xfs_disk_dquot {
1269	__be16 d_magic; /* dquot magic = XFS_DQUOT_MAGIC */
1270	__u8 d_version; /* dquot version */
1271	__u8 d_type; /* XFS_DQTYPE_USER/PROJ/GROUP */
1272	__be32 d_id; /* user,project,group id */
1273	__be64 d_blk_hardlimit;/* absolute limit on disk blks */
1274	__be64 d_blk_softlimit;/* preferred limit on disk blks */
1275	__be64 d_ino_hardlimit;/* maximum # allocated inodes */
1276	__be64 d_ino_softlimit;/* preferred inode limit */
1277	__be64 d_bcount; /* disk blocks owned by the user */
1278	__be64 d_icount; /* inodes owned by the user */
1279	__be32 d_itimer; /* zero if within inode limits if not,
1280	this is when we refuse service */
1281	__be32 d_btimer; /* similar to above; for disk blocks */
1282	__be16 d_iwarns; /* warnings issued wrt num inodes */
1283	__be16 d_bwarns; /* warnings issued wrt disk blocks */
1284	__be32 d_pad0; /* 64 bit align */
1285	__be64 d_rtb_hardlimit;/* absolute limit on realtime blks */
1286	__be64 d_rtb_softlimit;/* preferred limit on RT disk blks */
1287	__be64 d_rtbcount; /* realtime blocks owned */
1288	__be32 d_rtbtimer; /* similar to above; for RT disk blocks */
1289	__be16 d_rtbwarns; /* warnings issued wrt RT disk blocks */
1290	__be16 d_pad;
1291	};
1292
1293	/*
1294	* This is what goes on disk. This is separated from the xfs_disk_dquot because
1295	* carrying the unnecessary padding would be a waste of memory.
1296	*/
1297	struct xfs_dqblk {
1298	struct xfs_disk_dquot dd_diskdq; /* portion living incore as well */
1299	char dd_fill[4];/* filling for posterity */
1300
1301	/*
1302	* These two are only present on filesystems with the CRC bits set.
1303	*/
1304	__be32 dd_crc; /* checksum */
1305	__be64 dd_lsn; /* last modification in log */
1306	uuid_t dd_uuid; /* location information */
1307	};
1308
1309	#define XFS_DQUOT_CRC_OFF offsetof(struct xfs_dqblk, dd_crc)
1310
1311	/*
1312	* This defines the unit of allocation of dquots.
1313	*
1314	* Currently, it is just one file system block, and a 4K blk contains 30
1315	* (136 * 30 = 4080) dquots. It's probably not worth trying to make
1316	* this more dynamic.
1317	*
1318	* However, if this number is changed, we have to make sure that we don't
1319	* implicitly assume that we do allocations in chunks of a single filesystem
1320	* block in the dquot/xqm code.
1321	*
1322	* This is part of the ondisk format because the structure size is not a power
1323	* of two, which leaves slack at the end of the disk block.
1324	*/
1325	#define XFS_DQUOT_CLUSTER_SIZE_FSB (xfs_filblks_t)1
1326
1327	/*
1328	* Remote symlink format and access functions.
1329	*/
1330	#define XFS_SYMLINK_MAGIC 0x58534c4d /* XSLM */
1331
1332	struct xfs_dsymlink_hdr {
1333	__be32 sl_magic;
1334	__be32 sl_offset;
1335	__be32 sl_bytes;
1336	__be32 sl_crc;
1337	uuid_t sl_uuid;
1338	__be64 sl_owner;
1339	__be64 sl_blkno;
1340	__be64 sl_lsn;
1341	};
1342
1343	#define XFS_SYMLINK_CRC_OFF offsetof(struct xfs_dsymlink_hdr, sl_crc)
1344
1345	#define XFS_SYMLINK_MAXLEN 1024
1346	/*
1347	* The maximum pathlen is 1024 bytes. Since the minimum file system
1348	* blocksize is 512 bytes, we can get a max of 3 extents back from
1349	* bmapi when crc headers are taken into account.
1350	*/
1351	#define XFS_SYMLINK_MAPS 3
1352
1353	#define XFS_SYMLINK_BUF_SPACE(mp, bufsize) \
1354	((bufsize) - (xfs_has_crc((mp)) ? \
1355	sizeof(struct xfs_dsymlink_hdr) : 0))
1356
1357
1358	/*
1359	* Allocation Btree format definitions
1360	*
1361	* There are two on-disk btrees, one sorted by blockno and one sorted
1362	* by blockcount and blockno. All blocks look the same to make the code
1363	* simpler; if we have time later, we'll make the optimizations.
1364	*/
1365	#define XFS_ABTB_MAGIC 0x41425442 /* 'ABTB' for bno tree */
1366	#define XFS_ABTB_CRC_MAGIC 0x41423342 /* 'AB3B' */
1367	#define XFS_ABTC_MAGIC 0x41425443 /* 'ABTC' for cnt tree */
1368	#define XFS_ABTC_CRC_MAGIC 0x41423343 /* 'AB3C' */
1369
1370	/*
1371	* Data record/key structure
1372	*/
1373	typedef struct xfs_alloc_rec {
1374	__be32 ar_startblock; /* starting block number */
1375	__be32 ar_blockcount; /* count of free blocks */
1376	} xfs_alloc_rec_t, xfs_alloc_key_t;
1377
1378	typedef struct xfs_alloc_rec_incore {
1379	xfs_agblock_t ar_startblock; /* starting block number */
1380	xfs_extlen_t ar_blockcount; /* count of free blocks */
1381	} xfs_alloc_rec_incore_t;
1382
1383	/* btree pointer type */
1384	typedef __be32 xfs_alloc_ptr_t;
1385
1386	/*
1387	* Block numbers in the AG:
1388	* SB is sector 0, AGF is sector 1, AGI is sector 2, AGFL is sector 3.
1389	*/
1390	#define XFS_BNO_BLOCK(mp) ((xfs_agblock_t)(XFS_AGFL_BLOCK(mp) + 1))
1391	#define XFS_CNT_BLOCK(mp) ((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
1392
1393
1394	/*
1395	* Inode Allocation Btree format definitions
1396	*
1397	* There is a btree for the inode map per allocation group.
1398	*/
1399	#define XFS_IBT_MAGIC 0x49414254 /* 'IABT' */
1400	#define XFS_IBT_CRC_MAGIC 0x49414233 /* 'IAB3' */
1401	#define XFS_FIBT_MAGIC 0x46494254 /* 'FIBT' */
1402	#define XFS_FIBT_CRC_MAGIC 0x46494233 /* 'FIB3' */
1403
1404	typedef uint64_t xfs_inofree_t;
1405	#define XFS_INODES_PER_CHUNK (NBBY * sizeof(xfs_inofree_t))
1406	#define XFS_INODES_PER_CHUNK_LOG (XFS_NBBYLOG + 3)
1407	#define XFS_INOBT_ALL_FREE ((xfs_inofree_t)-1)
1408	#define XFS_INOBT_MASK(i) ((xfs_inofree_t)1 << (i))
1409
1410	#define XFS_INOBT_HOLEMASK_FULL 0 /* holemask for full chunk */
1411	#define XFS_INOBT_HOLEMASK_BITS (NBBY * sizeof(uint16_t))
1412	#define XFS_INODES_PER_HOLEMASK_BIT \
1413	(XFS_INODES_PER_CHUNK / (NBBY * sizeof(uint16_t)))
1414
1415	static inline xfs_inofree_t xfs_inobt_maskn(int i, int n)
1416	{
1417	return ((n >= XFS_INODES_PER_CHUNK ? 0 : XFS_INOBT_MASK(n)) - 1) << i;
1418	}
1419
1420	/*
1421	* The on-disk inode record structure has two formats. The original "full"
1422	* format uses a 4-byte freecount. The "sparse" format uses a 1-byte freecount
1423	* and replaces the 3 high-order freecount bytes wth the holemask and inode
1424	* count.
1425	*
1426	* The holemask of the sparse record format allows an inode chunk to have holes
1427	* that refer to blocks not owned by the inode record. This facilitates inode
1428	* allocation in the event of severe free space fragmentation.
1429	*/
1430	typedef struct xfs_inobt_rec {
1431	__be32 ir_startino; /* starting inode number */
1432	union {
1433	struct {
1434	__be32 ir_freecount; /* count of free inodes */
1435	} f;
1436	struct {
1437	__be16 ir_holemask;/* hole mask for sparse chunks */
1438	__u8 ir_count; /* total inode count */
1439	__u8 ir_freecount; /* count of free inodes */
1440	} sp;
1441	} ir_u;
1442	__be64 ir_free; /* free inode mask */
1443	} xfs_inobt_rec_t;
1444
1445	typedef struct xfs_inobt_rec_incore {
1446	xfs_agino_t ir_startino; /* starting inode number */
1447	uint16_t ir_holemask; /* hole mask for sparse chunks */
1448	uint8_t ir_count; /* total inode count */
1449	uint8_t ir_freecount; /* count of free inodes (set bits) */
1450	xfs_inofree_t ir_free; /* free inode mask */
1451	} xfs_inobt_rec_incore_t;
1452
1453	static inline bool xfs_inobt_issparse(uint16_t holemask)
1454	{
1455	/* non-zero holemask represents a sparse rec. */
1456	return holemask;
1457	}
1458
1459	/*
1460	* Key structure
1461	*/
1462	typedef struct xfs_inobt_key {
1463	__be32 ir_startino; /* starting inode number */
1464	} xfs_inobt_key_t;
1465
1466	/* btree pointer type */
1467	typedef __be32 xfs_inobt_ptr_t;
1468
1469	/*
1470	* block numbers in the AG.
1471	*/
1472	#define XFS_IBT_BLOCK(mp) ((xfs_agblock_t)(XFS_CNT_BLOCK(mp) + 1))
1473	#define XFS_FIBT_BLOCK(mp) ((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
1474
1475	/*
1476	* Reverse mapping btree format definitions
1477	*
1478	* There is a btree for the reverse map per allocation group
1479	*/
1480	#define XFS_RMAP_CRC_MAGIC 0x524d4233 /* 'RMB3' */
1481
1482	/*
1483	* Ownership info for an extent. This is used to create reverse-mapping
1484	* entries.
1485	*/
1486	#define XFS_OWNER_INFO_ATTR_FORK (1 << 0)
1487	#define XFS_OWNER_INFO_BMBT_BLOCK (1 << 1)
1488	struct xfs_owner_info {
1489	uint64_t oi_owner;
1490	xfs_fileoff_t oi_offset;
1491	unsigned int oi_flags;
1492	};
1493
1494	/*
1495	* Special owner types.
1496	*
1497	* Seeing as we only support up to 8EB, we have the upper bit of the owner field
1498	* to tell us we have a special owner value. We use these for static metadata
1499	* allocated at mkfs/growfs time, as well as for freespace management metadata.
1500	*/
1501	#define XFS_RMAP_OWN_NULL (-1ULL) /* No owner, for growfs */
1502	#define XFS_RMAP_OWN_UNKNOWN (-2ULL) /* Unknown owner, for EFI recovery */
1503	#define XFS_RMAP_OWN_FS (-3ULL) /* static fs metadata */
1504	#define XFS_RMAP_OWN_LOG (-4ULL) /* static fs metadata */
1505	#define XFS_RMAP_OWN_AG (-5ULL) /* AG freespace btree blocks */
1506	#define XFS_RMAP_OWN_INOBT (-6ULL) /* Inode btree blocks */
1507	#define XFS_RMAP_OWN_INODES (-7ULL) /* Inode chunk */
1508	#define XFS_RMAP_OWN_REFC (-8ULL) /* refcount tree */
1509	#define XFS_RMAP_OWN_COW (-9ULL) /* cow allocations */
1510	#define XFS_RMAP_OWN_MIN (-10ULL) /* guard */
1511
1512	#define XFS_RMAP_NON_INODE_OWNER(owner) (!!((owner) & (1ULL << 63)))
1513
1514	/*
1515	* Data record structure
1516	*/
1517	struct xfs_rmap_rec {
1518	__be32 rm_startblock; /* extent start block */
1519	__be32 rm_blockcount; /* extent length */
1520	__be64 rm_owner; /* extent owner */
1521	__be64 rm_offset; /* offset within the owner */
1522	};
1523
1524	/*
1525	* rmap btree record
1526	* rm_offset:63 is the attribute fork flag
1527	* rm_offset:62 is the bmbt block flag
1528	* rm_offset:61 is the unwritten extent flag (same as l0:63 in bmbt)
1529	* rm_offset:54-60 aren't used and should be zero
1530	* rm_offset:0-53 is the block offset within the inode
1531	*/
1532	#define XFS_RMAP_OFF_ATTR_FORK ((uint64_t)1ULL << 63)
1533	#define XFS_RMAP_OFF_BMBT_BLOCK ((uint64_t)1ULL << 62)
1534	#define XFS_RMAP_OFF_UNWRITTEN ((uint64_t)1ULL << 61)
1535
1536	#define XFS_RMAP_LEN_MAX ((uint32_t)~0U)
1537	#define XFS_RMAP_OFF_FLAGS (XFS_RMAP_OFF_ATTR_FORK | \
1538	XFS_RMAP_OFF_BMBT_BLOCK | \
1539	XFS_RMAP_OFF_UNWRITTEN)
1540	#define XFS_RMAP_OFF_MASK ((uint64_t)0x3FFFFFFFFFFFFFULL)
1541
1542	#define XFS_RMAP_OFF(off) ((off) & XFS_RMAP_OFF_MASK)
1543
1544	#define XFS_RMAP_IS_BMBT_BLOCK(off) (!!((off) & XFS_RMAP_OFF_BMBT_BLOCK))
1545	#define XFS_RMAP_IS_ATTR_FORK(off) (!!((off) & XFS_RMAP_OFF_ATTR_FORK))
1546	#define XFS_RMAP_IS_UNWRITTEN(len) (!!((off) & XFS_RMAP_OFF_UNWRITTEN))
1547
1548	#define RMAPBT_STARTBLOCK_BITLEN 32
1549	#define RMAPBT_BLOCKCOUNT_BITLEN 32
1550	#define RMAPBT_OWNER_BITLEN 64
1551	#define RMAPBT_ATTRFLAG_BITLEN 1
1552	#define RMAPBT_BMBTFLAG_BITLEN 1
1553	#define RMAPBT_EXNTFLAG_BITLEN 1
1554	#define RMAPBT_UNUSED_OFFSET_BITLEN 7
1555	#define RMAPBT_OFFSET_BITLEN 54
1556
1557	/*
1558	* Key structure
1559	*
1560	* We don't use the length for lookups
1561	*/
1562	struct xfs_rmap_key {
1563	__be32 rm_startblock; /* extent start block */
1564	__be64 rm_owner; /* extent owner */
1565	__be64 rm_offset; /* offset within the owner */
1566	} __attribute__((packed));
1567
1568	/* btree pointer type */
1569	typedef __be32 xfs_rmap_ptr_t;
1570
1571	#define XFS_RMAP_BLOCK(mp) \
1572	(xfs_has_finobt(((mp))) ? \
1573	XFS_FIBT_BLOCK(mp) + 1 : \
1574	XFS_IBT_BLOCK(mp) + 1)
1575
1576	/*
1577	* Reference Count Btree format definitions
1578	*
1579	*/
1580	#define XFS_REFC_CRC_MAGIC 0x52334643 /* 'R3FC' */
1581
1582	unsigned int xfs_refc_block(struct xfs_mount *mp);
1583
1584	/*
1585	* Data record/key structure
1586	*
1587	* Each record associates a range of physical blocks (starting at
1588	* rc_startblock and ending rc_blockcount blocks later) with a reference
1589	* count (rc_refcount). Extents that are being used to stage a copy on
1590	* write (CoW) operation are recorded in the refcount btree with a
1591	* refcount of 1. All other records must have a refcount > 1 and must
1592	* track an extent mapped only by file data forks.
1593	*
1594	* Extents with a single owner (attributes, metadata, non-shared file
1595	* data) are not tracked here. Free space is also not tracked here.
1596	* This is consistent with pre-reflink XFS.
1597	*/
1598
1599	/*
1600	* Extents that are being used to stage a copy on write are stored
1601	* in the refcount btree with a refcount of 1 and the upper bit set
1602	* on the startblock. This speeds up mount time deletion of stale
1603	* staging extents because they're all at the right side of the tree.
1604	*/
1605	#define XFS_REFC_COWFLAG (1U << 31)
1606	#define REFCNTBT_COWFLAG_BITLEN 1
1607	#define REFCNTBT_AGBLOCK_BITLEN 31
1608
1609	struct xfs_refcount_rec {
1610	__be32 rc_startblock; /* starting block number */
1611	__be32 rc_blockcount; /* count of blocks */
1612	__be32 rc_refcount; /* number of inodes linked here */
1613	};
1614
1615	struct xfs_refcount_key {
1616	__be32 rc_startblock; /* starting block number */
1617	};
1618
1619	#define MAXREFCOUNT ((xfs_nlink_t)~0U)
1620	#define MAXREFCEXTLEN ((xfs_extlen_t)~0U)
1621
1622	/* btree pointer type */
1623	typedef __be32 xfs_refcount_ptr_t;
1624
1625
1626	/*
1627	* BMAP Btree format definitions
1628	*
1629	* This includes both the root block definition that sits inside an inode fork
1630	* and the record/pointer formats for the leaf/node in the blocks.
1631	*/
1632	#define XFS_BMAP_MAGIC 0x424d4150 /* 'BMAP' */
1633	#define XFS_BMAP_CRC_MAGIC 0x424d4133 /* 'BMA3' */
1634
1635	/*
1636	* Bmap root header, on-disk form only.
1637	*/
1638	typedef struct xfs_bmdr_block {
1639	__be16 bb_level; /* 0 is a leaf */
1640	__be16 bb_numrecs; /* current # of data records */
1641	} xfs_bmdr_block_t;
1642
1643	/*
1644	* Bmap btree record and extent descriptor.
1645	* l0:63 is an extent flag (value 1 indicates non-normal).
1646	* l0:9-62 are startoff.
1647	* l0:0-8 and l1:21-63 are startblock.
1648	* l1:0-20 are blockcount.
1649	*/
1650	#define BMBT_EXNTFLAG_BITLEN 1
1651	#define BMBT_STARTOFF_BITLEN 54
1652	#define BMBT_STARTBLOCK_BITLEN 52
1653	#define BMBT_BLOCKCOUNT_BITLEN 21
1654
1655	#define BMBT_STARTOFF_MASK ((1ULL << BMBT_STARTOFF_BITLEN) - 1)
1656	#define BMBT_BLOCKCOUNT_MASK ((1ULL << BMBT_BLOCKCOUNT_BITLEN) - 1)
1657
1658	#define XFS_MAX_BMBT_EXTLEN ((xfs_extlen_t)(BMBT_BLOCKCOUNT_MASK))
1659
1660	/*
1661	* bmbt records have a file offset (block) field that is 54 bits wide, so this
1662	* is the largest xfs_fileoff_t that we ever expect to see.
1663	*/
1664	#define XFS_MAX_FILEOFF (BMBT_STARTOFF_MASK + BMBT_BLOCKCOUNT_MASK)
1665
1666	typedef struct xfs_bmbt_rec {
1667	__be64 l0, l1;
1668	} xfs_bmbt_rec_t;
1669
1670	typedef uint64_t xfs_bmbt_rec_base_t; /* use this for casts */
1671	typedef xfs_bmbt_rec_t xfs_bmdr_rec_t;
1672
1673	/*
1674	* Values and macros for delayed-allocation startblock fields.
1675	*/
1676	#define STARTBLOCKVALBITS 17
1677	#define STARTBLOCKMASKBITS (15 + 20)
1678	#define STARTBLOCKMASK \
1679	(((((xfs_fsblock_t)1) << STARTBLOCKMASKBITS) - 1) << STARTBLOCKVALBITS)
1680
1681	static inline int isnullstartblock(xfs_fsblock_t x)
1682	{
1683	return ((x) & STARTBLOCKMASK) == STARTBLOCKMASK;
1684	}
1685
1686	static inline xfs_fsblock_t nullstartblock(int k)
1687	{
1688	ASSERT(k < (1 << STARTBLOCKVALBITS));
1689	return STARTBLOCKMASK | (k);
1690	}
1691
1692	static inline xfs_filblks_t startblockval(xfs_fsblock_t x)
1693	{
1694	return (xfs_filblks_t)((x) & ~STARTBLOCKMASK);
1695	}
1696
1697	/*
1698	* Key structure for non-leaf levels of the tree.
1699	*/
1700	typedef struct xfs_bmbt_key {
1701	__be64 br_startoff; /* starting file offset */
1702	} xfs_bmbt_key_t, xfs_bmdr_key_t;
1703
1704	/* btree pointer type */
1705	typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
1706
1707
1708	/*
1709	* Generic Btree block format definitions
1710	*
1711	* This is a combination of the actual format used on disk for short and long
1712	* format btrees. The first three fields are shared by both format, but the
1713	* pointers are different and should be used with care.
1714	*
1715	* To get the size of the actual short or long form headers please use the size
1716	* macros below. Never use sizeof(xfs_btree_block).
1717	*
1718	* The blkno, crc, lsn, owner and uuid fields are only available in filesystems
1719	* with the crc feature bit, and all accesses to them must be conditional on
1720	* that flag.
1721	*/
1722	/* short form block header */
1723	struct xfs_btree_block_shdr {
1724	__be32 bb_leftsib;
1725	__be32 bb_rightsib;
1726
1727	__be64 bb_blkno;
1728	__be64 bb_lsn;
1729	uuid_t bb_uuid;
1730	__be32 bb_owner;
1731	__le32 bb_crc;
1732	};
1733
1734	/* long form block header */
1735	struct xfs_btree_block_lhdr {
1736	__be64 bb_leftsib;
1737	__be64 bb_rightsib;
1738
1739	__be64 bb_blkno;
1740	__be64 bb_lsn;
1741	uuid_t bb_uuid;
1742	__be64 bb_owner;
1743	__le32 bb_crc;
1744	__be32 bb_pad; /* padding for alignment */
1745	};
1746
1747	struct xfs_btree_block {
1748	__be32 bb_magic; /* magic number for block type */
1749	__be16 bb_level; /* 0 is a leaf */
1750	__be16 bb_numrecs; /* current # of data records */
1751	union {
1752	struct xfs_btree_block_shdr s;
1753	struct xfs_btree_block_lhdr l;
1754	} bb_u; /* rest */
1755	};
1756
1757	/* size of a short form block */
1758	#define XFS_BTREE_SBLOCK_LEN \
1759	(offsetof(struct xfs_btree_block, bb_u) + \
1760	offsetof(struct xfs_btree_block_shdr, bb_blkno))
1761	/* size of a long form block */
1762	#define XFS_BTREE_LBLOCK_LEN \
1763	(offsetof(struct xfs_btree_block, bb_u) + \
1764	offsetof(struct xfs_btree_block_lhdr, bb_blkno))
1765
1766	/* sizes of CRC enabled btree blocks */
1767	#define XFS_BTREE_SBLOCK_CRC_LEN \
1768	(offsetof(struct xfs_btree_block, bb_u) + \
1769	sizeof(struct xfs_btree_block_shdr))
1770	#define XFS_BTREE_LBLOCK_CRC_LEN \
1771	(offsetof(struct xfs_btree_block, bb_u) + \
1772	sizeof(struct xfs_btree_block_lhdr))
1773
1774	#define XFS_BTREE_SBLOCK_CRC_OFF \
1775	offsetof(struct xfs_btree_block, bb_u.s.bb_crc)
1776	#define XFS_BTREE_LBLOCK_CRC_OFF \
1777	offsetof(struct xfs_btree_block, bb_u.l.bb_crc)
1778
1779	/*
1780	* On-disk XFS access control list structure.
1781	*/
1782	struct xfs_acl_entry {
1783	__be32 ae_tag;
1784	__be32 ae_id;
1785	__be16 ae_perm;
1786	__be16 ae_pad; /* fill the implicit hole in the structure */
1787	};
1788
1789	struct xfs_acl {
1790	__be32 acl_cnt;
1791	struct xfs_acl_entry acl_entry[];
1792	};
1793
1794	/*
1795	* The number of ACL entries allowed is defined by the on-disk format.
1796	* For v4 superblocks, that is limited to 25 entries. For v5 superblocks, it is
1797	* limited only by the maximum size of the xattr that stores the information.
1798	*/
1799	#define XFS_ACL_MAX_ENTRIES(mp) \
1800	(xfs_has_crc(mp) \
1801	? (XFS_XATTR_SIZE_MAX - sizeof(struct xfs_acl)) / \
1802	sizeof(struct xfs_acl_entry) \
1803	: 25)
1804
1805	#define XFS_ACL_SIZE(cnt) \
1806	(sizeof(struct xfs_acl) + \
1807	sizeof(struct xfs_acl_entry) * cnt)
1808
1809	#define XFS_ACL_MAX_SIZE(mp) \
1810	XFS_ACL_SIZE(XFS_ACL_MAX_ENTRIES((mp)))
1811
1812
1813	/* On-disk XFS extended attribute names */
1814	#define SGI_ACL_FILE "SGI_ACL_FILE"
1815	#define SGI_ACL_DEFAULT "SGI_ACL_DEFAULT"
1816	#define SGI_ACL_FILE_SIZE (sizeof(SGI_ACL_FILE)-1)
1817	#define SGI_ACL_DEFAULT_SIZE (sizeof(SGI_ACL_DEFAULT)-1)
1818
1819	#endif /* __XFS_FORMAT_H__ */
1820