1	// SPDX-License-Identifier: GPL-2.0
2	/*
3	* Copyright (c) 2000-2005 Silicon Graphics, Inc.
4	* All Rights Reserved.
5	*/
6	#include "xfs.h"
7	#include "xfs_fs.h"
8	#include "xfs_shared.h"
9	#include "xfs_format.h"
10	#include "xfs_log_format.h"
11	#include "xfs_trans_resv.h"
12	#include "xfs_bit.h"
13	#include "xfs_sb.h"
14	#include "xfs_mount.h"
15	#include "xfs_ialloc.h"
16	#include "xfs_alloc.h"
17	#include "xfs_error.h"
18	#include "xfs_trans.h"
19	#include "xfs_buf_item.h"
20	#include "xfs_bmap_btree.h"
21	#include "xfs_alloc_btree.h"
22	#include "xfs_log.h"
23	#include "xfs_rmap_btree.h"
24	#include "xfs_refcount_btree.h"
25	#include "xfs_da_format.h"
26	#include "xfs_health.h"
27	#include "xfs_ag.h"
28	#include "xfs_rtbitmap.h"
29	#include "xfs_exchrange.h"
30	#include "xfs_rtgroup.h"
31	#include "xfs_rtrmap_btree.h"
32	#include "xfs_rtrefcount_btree.h"
33
34	/*
35	* Physical superblock buffer manipulations. Shared with libxfs in userspace.
36	*/
37
38	/*
39	* Check that all the V4 feature bits that the V5 filesystem format requires are
40	* correctly set.
41	*/
42	static bool
43	xfs_sb_validate_v5_features(
44	struct xfs_sb *sbp)
45	{
46	/* We must not have any unknown V4 feature bits set */
47	if (sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS)
48	return false;
49
50	/*
51	* The CRC bit is considered an invalid V4 flag, so we have to add it
52	* manually to the OKBITS mask.
53	*/
54	if (sbp->sb_features2 & ~(XFS_SB_VERSION2_OKBITS |
55	XFS_SB_VERSION2_CRCBIT))
56	return false;
57
58	/* Now check all the required V4 feature flags are set. */
59
60	#define V5_VERS_FLAGS (XFS_SB_VERSION_NLINKBIT | \
61	XFS_SB_VERSION_ALIGNBIT | \
62	XFS_SB_VERSION_LOGV2BIT | \
63	XFS_SB_VERSION_EXTFLGBIT | \
64	XFS_SB_VERSION_DIRV2BIT | \
65	XFS_SB_VERSION_MOREBITSBIT)
66
67	#define V5_FEAT_FLAGS (XFS_SB_VERSION2_LAZYSBCOUNTBIT | \
68	XFS_SB_VERSION2_ATTR2BIT | \
69	XFS_SB_VERSION2_PROJID32BIT | \
70	XFS_SB_VERSION2_CRCBIT)
71
72	if ((sbp->sb_versionnum & V5_VERS_FLAGS) != V5_VERS_FLAGS)
73	return false;
74	if ((sbp->sb_features2 & V5_FEAT_FLAGS) != V5_FEAT_FLAGS)
75	return false;
76	return true;
77	}
78
79	/*
80	* We current support XFS v5 formats with known features and v4 superblocks with
81	* at least V2 directories.
82	*/
83	bool
84	xfs_sb_good_version(
85	struct xfs_sb *sbp)
86	{
87	/*
88	* All v5 filesystems are supported, but we must check that all the
89	* required v4 feature flags are enabled correctly as the code checks
90	* those flags and not for v5 support.
91	*/
92	if (xfs_sb_is_v5(sbp))
93	return xfs_sb_validate_v5_features(sbp);
94
95	/* versions prior to v4 are not supported */
96	if (XFS_SB_VERSION_NUM(sbp) != XFS_SB_VERSION_4)
97	return false;
98
99	/* We must not have any unknown v4 feature bits set */
100	if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
101	((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
102	(sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
103	return false;
104
105	/* V4 filesystems need v2 directories and unwritten extents */
106	if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
107	return false;
108	if (!(sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT))
109	return false;
110
111	/* It's a supported v4 filesystem */
112	return true;
113	}
114
115	uint64_t
116	xfs_sb_version_to_features(
117	struct xfs_sb *sbp)
118	{
119	uint64_t features = 0;
120
121	/* optional V4 features */
122	if (sbp->sb_rblocks > 0)
123	features |= XFS_FEAT_REALTIME;
124	if (sbp->sb_versionnum & XFS_SB_VERSION_NLINKBIT)
125	features |= XFS_FEAT_NLINK;
126	if (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT)
127	features |= XFS_FEAT_ATTR;
128	if (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT)
129	features |= XFS_FEAT_QUOTA;
130	if (sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT)
131	features |= XFS_FEAT_ALIGN;
132	if (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT)
133	features |= XFS_FEAT_LOGV2;
134	if (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT)
135	features |= XFS_FEAT_DALIGN;
136	if (sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT)
137	features |= XFS_FEAT_EXTFLG;
138	if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT)
139	features |= XFS_FEAT_SECTOR;
140	if (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT)
141	features |= XFS_FEAT_ASCIICI;
142	if (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) {
143	if (sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT)
144	features |= XFS_FEAT_LAZYSBCOUNT;
145	if (sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT)
146	features |= XFS_FEAT_PROJID32;
147	if (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE)
148	features |= XFS_FEAT_FTYPE;
149	}
150
151	if (!xfs_sb_is_v5(sbp))
152	return features;
153
154	/* Always on V5 features */
155	features |= XFS_FEAT_ALIGN | XFS_FEAT_LOGV2 | XFS_FEAT_EXTFLG |
156	XFS_FEAT_LAZYSBCOUNT | XFS_FEAT_PROJID32 |
157	XFS_FEAT_V3INODES | XFS_FEAT_CRC | XFS_FEAT_PQUOTINO;
158
159	/* Optional V5 features */
160	if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT)
161	features |= XFS_FEAT_FINOBT;
162	if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT)
163	features |= XFS_FEAT_RMAPBT;
164	if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK)
165	features |= XFS_FEAT_REFLINK;
166	if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
167	features |= XFS_FEAT_INOBTCNT;
168	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_FTYPE)
169	features |= XFS_FEAT_FTYPE;
170	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES)
171	features |= XFS_FEAT_SPINODES;
172	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID)
173	features |= XFS_FEAT_META_UUID;
174	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_BIGTIME)
175	features |= XFS_FEAT_BIGTIME;
176	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR)
177	features |= XFS_FEAT_NEEDSREPAIR;
178	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_NREXT64)
179	features |= XFS_FEAT_NREXT64;
180	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_EXCHRANGE)
181	features |= XFS_FEAT_EXCHANGE_RANGE;
182	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_PARENT)
183	features |= XFS_FEAT_PARENT;
184	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)
185	features |= XFS_FEAT_METADIR;
186	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED)
187	features |= XFS_FEAT_ZONED;
188
189	return features;
190	}
191
192	/* Check all the superblock fields we care about when reading one in. */
193	STATIC int
194	xfs_validate_sb_read(
195	struct xfs_mount *mp,
196	struct xfs_sb *sbp)
197	{
198	if (!xfs_sb_is_v5(sbp))
199	return 0;
200
201	/*
202	* Version 5 superblock feature mask validation. Reject combinations
203	* the kernel cannot support up front before checking anything else.
204	*/
205	if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) {
206	xfs_warn(mp,
207	"Superblock has unknown compatible features (0x%x) enabled.",
208	(sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN));
209	xfs_warn(mp,
210	"Using a more recent kernel is recommended.");
211	}
212
213	if (xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
214	xfs_alert(mp,
215	"Superblock has unknown read-only compatible features (0x%x) enabled.",
216	(sbp->sb_features_ro_compat &
217	XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
218	if (!xfs_is_readonly(mp)) {
219	xfs_warn(mp,
220	"Attempted to mount read-only compatible filesystem read-write.");
221	xfs_warn(mp,
222	"Filesystem can only be safely mounted read only.");
223
224	return -EINVAL;
225	}
226	}
227	if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
228	xfs_warn(mp,
229	"Superblock has unknown incompatible features (0x%x) enabled.",
230	(sbp->sb_features_incompat &
231	XFS_SB_FEAT_INCOMPAT_UNKNOWN));
232	xfs_warn(mp,
233	"Filesystem cannot be safely mounted by this kernel.");
234	return -EINVAL;
235	}
236
237	return 0;
238	}
239
240	/* Return the number of extents covered by a single rt bitmap file */
241	static xfs_rtbxlen_t
242	xfs_extents_per_rbm(
243	struct xfs_sb *sbp)
244	{
245	if (xfs_sb_is_v5(sbp) &&
246	(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR))
247	return sbp->sb_rgextents;
248	return sbp->sb_rextents;
249	}
250
251	/*
252	* Return the payload size of a single rt bitmap block (without the metadata
253	* header if any).
254	*/
255	static inline unsigned int
256	xfs_rtbmblock_size(
257	struct xfs_sb *sbp)
258	{
259	if (xfs_sb_is_v5(sbp) &&
260	(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR))
261	return sbp->sb_blocksize - sizeof(struct xfs_rtbuf_blkinfo);
262	return sbp->sb_blocksize;
263	}
264
265	static uint64_t
266	xfs_expected_rbmblocks(
267	struct xfs_sb *sbp)
268	{
269	if (xfs_sb_is_v5(sbp) &&
270	(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED))
271	return 0;
272	return howmany_64(xfs_extents_per_rbm(sbp),
273	NBBY * xfs_rtbmblock_size(sbp));
274	}
275
276	/* Validate the realtime geometry */
277	bool
278	xfs_validate_rt_geometry(
279	struct xfs_sb *sbp)
280	{
281	if (xfs_sb_is_v5(sbp) &&
282	(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED)) {
283	if (sbp->sb_rextsize != 1)
284	return false;
285	} else {
286	if (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE ||
287	sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)
288	return false;
289	}
290
291	if (sbp->sb_rblocks == 0) {
292	if (sbp->sb_rextents != 0 || sbp->sb_rbmblocks != 0 ||
293	sbp->sb_rextslog != 0 || sbp->sb_frextents != 0)
294	return false;
295	return true;
296	}
297
298	if (sbp->sb_rextents == 0 ||
299	sbp->sb_rextents != div_u64(sbp->sb_rblocks, sbp->sb_rextsize) ||
300	sbp->sb_rextslog != xfs_compute_rextslog(sbp->sb_rextents) ||
301	sbp->sb_rbmblocks != xfs_expected_rbmblocks(sbp))
302	return false;
303
304	if (xfs_sb_is_v5(sbp) &&
305	(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED)) {
306	uint32_t mod;
307
308	/*
309	* Zoned RT devices must be aligned to the RT group size,
310	* because garbage collection assumes that all zones have the
311	* same size to avoid insane complexity if that weren't the
312	* case.
313	*/
314	div_u64_rem(sbp->sb_rextents, sbp->sb_rgextents, &mod);
315	if (mod)
316	return false;
317	}
318
319	return true;
320	}
321
322	/* Check all the superblock fields we care about when writing one out. */
323	STATIC int
324	xfs_validate_sb_write(
325	struct xfs_mount *mp,
326	struct xfs_buf *bp,
327	struct xfs_sb *sbp)
328	{
329	/*
330	* Carry out additional sb summary counter sanity checks when we write
331	* the superblock. We skip this in the read validator because there
332	* could be newer superblocks in the log and if the values are garbage
333	* even after replay we'll recalculate them at the end of log mount.
334	*
335	* mkfs has traditionally written zeroed counters to inprogress and
336	* secondary superblocks, so allow this usage to continue because
337	* we never read counters from such superblocks.
338	*/
339	if (xfs_buf_daddr(bp) == XFS_SB_DADDR && !sbp->sb_inprogress &&
340	(sbp->sb_fdblocks > sbp->sb_dblocks ||
341	!xfs_verify_icount(mp, sbp->sb_icount) ||
342	sbp->sb_ifree > sbp->sb_icount)) {
343	xfs_warn(mp, "SB summary counter sanity check failed");
344	return -EFSCORRUPTED;
345	}
346
347	if (!xfs_sb_is_v5(sbp))
348	return 0;
349
350	/*
351	* Version 5 superblock feature mask validation. Reject combinations
352	* the kernel cannot support since we checked for unsupported bits in
353	* the read verifier, which means that memory is corrupt.
354	*/
355	if (!xfs_is_readonly(mp) &&
356	xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
357	xfs_alert(mp,
358	"Corruption detected in superblock read-only compatible features (0x%x)!",
359	(sbp->sb_features_ro_compat &
360	XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
361	return -EFSCORRUPTED;
362	}
363	if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
364	xfs_warn(mp,
365	"Corruption detected in superblock incompatible features (0x%x)!",
366	(sbp->sb_features_incompat &
367	XFS_SB_FEAT_INCOMPAT_UNKNOWN));
368	return -EFSCORRUPTED;
369	}
370	if (xfs_sb_has_incompat_log_feature(sbp,
371	XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN)) {
372	xfs_warn(mp,
373	"Corruption detected in superblock incompatible log features (0x%x)!",
374	(sbp->sb_features_log_incompat &
375	XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN));
376	return -EFSCORRUPTED;
377	}
378
379	/*
380	* We can't read verify the sb LSN because the read verifier is called
381	* before the log is allocated and processed. We know the log is set up
382	* before write verifier calls, so check it here.
383	*/
384	if (!xfs_log_check_lsn(mp, sbp->sb_lsn))
385	return -EFSCORRUPTED;
386
387	return 0;
388	}
389
390	int
391	xfs_compute_rgblklog(
392	xfs_rtxlen_t rgextents,
393	xfs_rgblock_t rextsize)
394	{
395	uint64_t rgblocks = (uint64_t)rgextents * rextsize;
396
397	return xfs_highbit64(rgblocks - 1) + 1;
398	}
399
400	static int
401	xfs_validate_sb_rtgroups(
402	struct xfs_mount *mp,
403	struct xfs_sb *sbp)
404	{
405	uint64_t groups;
406	int rgblklog;
407
408	if (sbp->sb_rextsize == 0) {
409	xfs_warn(mp,
410	"Realtime extent size must not be zero.");
411	return -EINVAL;
412	}
413
414	if (sbp->sb_rgextents > XFS_MAX_RGBLOCKS / sbp->sb_rextsize) {
415	xfs_warn(mp,
416	"Realtime group size (%u) must be less than %u rt extents.",
417	sbp->sb_rgextents,
418	XFS_MAX_RGBLOCKS / sbp->sb_rextsize);
419	return -EINVAL;
420	}
421
422	if (sbp->sb_rgextents < XFS_MIN_RGEXTENTS) {
423	xfs_warn(mp,
424	"Realtime group size (%u) must be at least %u rt extents.",
425	sbp->sb_rgextents, XFS_MIN_RGEXTENTS);
426	return -EINVAL;
427	}
428
429	if (sbp->sb_rgcount > XFS_MAX_RGNUMBER) {
430	xfs_warn(mp,
431	"Realtime groups (%u) must be less than %u.",
432	sbp->sb_rgcount, XFS_MAX_RGNUMBER);
433	return -EINVAL;
434	}
435
436	groups = howmany_64(sbp->sb_rextents, sbp->sb_rgextents);
437	if (groups != sbp->sb_rgcount) {
438	xfs_warn(mp,
439	"Realtime groups (%u) do not cover the entire rt section; need (%llu) groups.",
440	sbp->sb_rgcount, groups);
441	return -EINVAL;
442	}
443
444	/* Exchange-range is required for fsr to work on realtime files */
445	if (!(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_EXCHRANGE)) {
446	xfs_warn(mp,
447	"Realtime groups feature requires exchange-range support.");
448	return -EINVAL;
449	}
450
451	rgblklog = xfs_compute_rgblklog(sbp->sb_rgextents, sbp->sb_rextsize);
452	if (sbp->sb_rgblklog != rgblklog) {
453	xfs_warn(mp,
454	"Realtime group log (%d) does not match expected value (%d).",
455	sbp->sb_rgblklog, rgblklog);
456	return -EINVAL;
457	}
458
459	return 0;
460	}
461
462	static int
463	xfs_validate_sb_zoned(
464	struct xfs_mount *mp,
465	struct xfs_sb *sbp)
466	{
467	if (sbp->sb_frextents != 0) {
468	xfs_warn(mp,
469	"sb_frextents must be zero for zoned file systems.");
470	return -EINVAL;
471	}
472
473	if (sbp->sb_rtstart && sbp->sb_rtstart < sbp->sb_dblocks) {
474	xfs_warn(mp,
475	"sb_rtstart (%lld) overlaps sb_dblocks (%lld).",
476	sbp->sb_rtstart, sbp->sb_dblocks);
477	return -EINVAL;
478	}
479
480	if (sbp->sb_rtreserved && sbp->sb_rtreserved >= sbp->sb_rblocks) {
481	xfs_warn(mp,
482	"sb_rtreserved (%lld) larger than sb_rblocks (%lld).",
483	sbp->sb_rtreserved, sbp->sb_rblocks);
484	return -EINVAL;
485	}
486
487	return 0;
488	}
489
490	/* Check the validity of the SB. */
491	STATIC int
492	xfs_validate_sb_common(
493	struct xfs_mount *mp,
494	struct xfs_buf *bp,
495	struct xfs_sb *sbp)
496	{
497	struct xfs_dsb *dsb = bp->b_addr;
498	uint32_t agcount = 0;
499	uint32_t rem;
500	bool has_dalign;
501	int error;
502
503	if (!xfs_verify_magic(bp, dsb->sb_magicnum)) {
504	xfs_warn(mp,
505	"Superblock has bad magic number 0x%x. Not an XFS filesystem?",
506	be32_to_cpu(dsb->sb_magicnum));
507	return -EWRONGFS;
508	}
509
510	if (!xfs_sb_good_version(sbp)) {
511	xfs_warn(mp,
512	"Superblock has unknown features enabled or corrupted feature masks.");
513	return -EWRONGFS;
514	}
515
516	/*
517	* Validate feature flags and state
518	*/
519	if (xfs_sb_is_v5(sbp)) {
520	if (sbp->sb_blocksize < XFS_MIN_CRC_BLOCKSIZE) {
521	xfs_notice(mp,
522	"Block size (%u bytes) too small for Version 5 superblock (minimum %d bytes)",
523	sbp->sb_blocksize, XFS_MIN_CRC_BLOCKSIZE);
524	return -EFSCORRUPTED;
525	}
526
527	/* V5 has a separate project quota inode */
528	if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
529	xfs_notice(mp,
530	"Version 5 of Super block has XFS_OQUOTA bits.");
531	return -EFSCORRUPTED;
532	}
533
534	/*
535	* Full inode chunks must be aligned to inode chunk size when
536	* sparse inodes are enabled to support the sparse chunk
537	* allocation algorithm and prevent overlapping inode records.
538	*/
539	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES) {
540	uint32_t align;
541
542	align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
543	>> sbp->sb_blocklog;
544	if (sbp->sb_inoalignmt != align) {
545	xfs_warn(mp,
546	"Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
547	sbp->sb_inoalignmt, align);
548	return -EINVAL;
549	}
550
551	if (sbp->sb_spino_align &&
552	(sbp->sb_spino_align > sbp->sb_inoalignmt ||
553	(sbp->sb_inoalignmt % sbp->sb_spino_align) != 0)) {
554	xfs_warn(mp,
555	"Sparse inode alignment (%u) is invalid, must be integer factor of (%u).",
556	sbp->sb_spino_align,
557	sbp->sb_inoalignmt);
558	return -EINVAL;
559	}
560	} else if (sbp->sb_spino_align) {
561	xfs_warn(mp,
562	"Sparse inode alignment (%u) should be zero.",
563	sbp->sb_spino_align);
564	return -EINVAL;
565	}
566
567	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR) {
568	if (memchr_inv(sbp->sb_pad, 0, sizeof(sbp->sb_pad))) {
569	xfs_warn(mp,
570	"Metadir superblock padding fields must be zero.");
571	return -EINVAL;
572	}
573
574	error = xfs_validate_sb_rtgroups(mp, sbp);
575	if (error)
576	return error;
577	}
578	if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED) {
579	error = xfs_validate_sb_zoned(mp, sbp);
580	if (error)
581	return error;
582	}
583	} else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
584	XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
585	xfs_notice(mp,
586	"Superblock earlier than Version 5 has XFS_{P|G}QUOTA_{ENFD|CHKD} bits.");
587	return -EFSCORRUPTED;
588	}
589
590	if (unlikely(
591	sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
592	xfs_warn(mp,
593	"filesystem is marked as having an external log; "
594	"specify logdev on the mount command line.");
595	return -EINVAL;
596	}
597
598	if (unlikely(
599	sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
600	xfs_warn(mp,
601	"filesystem is marked as having an internal log; "
602	"do not specify logdev on the mount command line.");
603	return -EINVAL;
604	}
605
606	/* Compute agcount for this number of dblocks and agblocks */
607	if (sbp->sb_agblocks) {
608	agcount = div_u64_rem(sbp->sb_dblocks, sbp->sb_agblocks, &rem);
609	if (rem)
610	agcount++;
611	}
612
613	/*
614	* More sanity checking. Most of these were stolen directly from
615	* xfs_repair.
616	*/
617	if (unlikely(
618	sbp->sb_agcount <= 0 ||
619	sbp->sb_sectsize < XFS_MIN_SECTORSIZE ||
620	sbp->sb_sectsize > XFS_MAX_SECTORSIZE ||
621	sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG ||
622	sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG ||
623	sbp->sb_sectsize != (1 << sbp->sb_sectlog) ||
624	sbp->sb_blocksize < XFS_MIN_BLOCKSIZE ||
625	sbp->sb_blocksize > XFS_MAX_BLOCKSIZE ||
626	sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG ||
627	sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
628	sbp->sb_blocksize != (1 << sbp->sb_blocklog) ||
629	sbp->sb_dirblklog + sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
630	sbp->sb_inodesize < XFS_DINODE_MIN_SIZE ||
631	sbp->sb_inodesize > XFS_DINODE_MAX_SIZE ||
632	sbp->sb_inodelog < XFS_DINODE_MIN_LOG ||
633	sbp->sb_inodelog > XFS_DINODE_MAX_LOG ||
634	sbp->sb_inodesize != (1 << sbp->sb_inodelog) ||
635	sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
636	XFS_FSB_TO_B(mp, sbp->sb_agblocks) < XFS_MIN_AG_BYTES ||
637	XFS_FSB_TO_B(mp, sbp->sb_agblocks) > XFS_MAX_AG_BYTES ||
638	sbp->sb_agblklog != xfs_highbit32(sbp->sb_agblocks - 1) + 1 ||
639	agcount == 0 || agcount != sbp->sb_agcount ||
640	(sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) ||
641	(sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) ||
642	(sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) ||
643	(sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) ||
644	sbp->sb_dblocks == 0 ||
645	sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) ||
646	sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) ||
647	sbp->sb_shared_vn != 0)) {
648	xfs_notice(mp, "SB sanity check failed");
649	return -EFSCORRUPTED;
650	}
651
652	/*
653	* Logs that are too large are not supported at all. Reject them
654	* outright. Logs that are too small are tolerated on v4 filesystems,
655	* but we can only check that when mounting the log. Hence we skip
656	* those checks here.
657	*/
658	if (sbp->sb_logblocks > XFS_MAX_LOG_BLOCKS) {
659	xfs_notice(mp,
660	"Log size 0x%x blocks too large, maximum size is 0x%llx blocks",
661	sbp->sb_logblocks, XFS_MAX_LOG_BLOCKS);
662	return -EFSCORRUPTED;
663	}
664
665	if (XFS_FSB_TO_B(mp, sbp->sb_logblocks) > XFS_MAX_LOG_BYTES) {
666	xfs_warn(mp,
667	"log size 0x%llx bytes too large, maximum size is 0x%llx bytes",
668	XFS_FSB_TO_B(mp, sbp->sb_logblocks),
669	XFS_MAX_LOG_BYTES);
670	return -EFSCORRUPTED;
671	}
672
673	/*
674	* Do not allow filesystems with corrupted log sector or stripe units to
675	* be mounted. We cannot safely size the iclogs or write to the log if
676	* the log stripe unit is not valid.
677	*/
678	if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT) {
679	if (sbp->sb_logsectsize != (1U << sbp->sb_logsectlog)) {
680	xfs_notice(mp,
681	"log sector size in bytes/log2 (0x%x/0x%x) must match",
682	sbp->sb_logsectsize, 1U << sbp->sb_logsectlog);
683	return -EFSCORRUPTED;
684	}
685	} else if (sbp->sb_logsectsize || sbp->sb_logsectlog) {
686	xfs_notice(mp,
687	"log sector size in bytes/log2 (0x%x/0x%x) are not zero",
688	sbp->sb_logsectsize, sbp->sb_logsectlog);
689	return -EFSCORRUPTED;
690	}
691
692	if (sbp->sb_logsunit > 1) {
693	if (sbp->sb_logsunit % sbp->sb_blocksize) {
694	xfs_notice(mp,
695	"log stripe unit 0x%x bytes must be a multiple of block size",
696	sbp->sb_logsunit);
697	return -EFSCORRUPTED;
698	}
699	if (sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE) {
700	xfs_notice(mp,
701	"log stripe unit 0x%x bytes over maximum size (0x%x bytes)",
702	sbp->sb_logsunit, XLOG_MAX_RECORD_BSIZE);
703	return -EFSCORRUPTED;
704	}
705	}
706
707	if (!xfs_validate_rt_geometry(sbp)) {
708	xfs_notice(mp,
709	"realtime %sgeometry check failed",
710	sbp->sb_rblocks ? "" : "zeroed ");
711	return -EFSCORRUPTED;
712	}
713
714	/*
715	* Either (sb_unit and !hasdalign) or (!sb_unit and hasdalign)
716	* would imply the image is corrupted.
717	*/
718	has_dalign = sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT;
719	if (!!sbp->sb_unit ^ has_dalign) {
720	xfs_notice(mp, "SB stripe alignment sanity check failed");
721	return -EFSCORRUPTED;
722	}
723
724	if (!xfs_validate_stripe_geometry(mp, XFS_FSB_TO_B(mp, sbp->sb_unit),
725	XFS_FSB_TO_B(mp, sbp->sb_width), 0,
726	xfs_buf_daddr(bp) == XFS_SB_DADDR, false))
727	return -EFSCORRUPTED;
728
729	/*
730	* Currently only very few inode sizes are supported.
731	*/
732	switch (sbp->sb_inodesize) {
733	case 256:
734	case 512:
735	case 1024:
736	case 2048:
737	break;
738	default:
739	xfs_warn(mp, "inode size of %d bytes not supported",
740	sbp->sb_inodesize);
741	return -ENOSYS;
742	}
743
744	return 0;
745	}
746
747	void
748	xfs_sb_quota_from_disk(struct xfs_sb *sbp)
749	{
750	if (xfs_sb_is_v5(sbp) &&
751	(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)) {
752	sbp->sb_uquotino = NULLFSINO;
753	sbp->sb_gquotino = NULLFSINO;
754	sbp->sb_pquotino = NULLFSINO;
755	return;
756	}
757
758	/*
759	* older mkfs doesn't initialize quota inodes to NULLFSINO. This
760	* leads to in-core values having two different values for a quota
761	* inode to be invalid: 0 and NULLFSINO. Change it to a single value
762	* NULLFSINO.
763	*
764	* Note that this change affect only the in-core values. These
765	* values are not written back to disk unless any quota information
766	* is written to the disk. Even in that case, sb_pquotino field is
767	* not written to disk unless the superblock supports pquotino.
768	*/
769	if (sbp->sb_uquotino == 0)
770	sbp->sb_uquotino = NULLFSINO;
771	if (sbp->sb_gquotino == 0)
772	sbp->sb_gquotino = NULLFSINO;
773	if (sbp->sb_pquotino == 0)
774	sbp->sb_pquotino = NULLFSINO;
775
776	/*
777	* We need to do these manipilations only if we are working
778	* with an older version of on-disk superblock.
779	*/
780	if (xfs_sb_is_v5(sbp))
781	return;
782
783	if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
784	sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
785	XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
786	if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
787	sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
788	XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
789	sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
790
791	if (sbp->sb_qflags & XFS_PQUOTA_ACCT &&
792	sbp->sb_gquotino != NULLFSINO) {
793	/*
794	* In older version of superblock, on-disk superblock only
795	* has sb_gquotino, and in-core superblock has both sb_gquotino
796	* and sb_pquotino. But, only one of them is supported at any
797	* point of time. So, if PQUOTA is set in disk superblock,
798	* copy over sb_gquotino to sb_pquotino. The NULLFSINO test
799	* above is to make sure we don't do this twice and wipe them
800	* both out!
801	*/
802	sbp->sb_pquotino = sbp->sb_gquotino;
803	sbp->sb_gquotino = NULLFSINO;
804	}
805	}
806
807	static void
808	__xfs_sb_from_disk(
809	struct xfs_sb *to,
810	struct xfs_dsb *from,
811	bool convert_xquota)
812	{
813	to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
814	to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
815	to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
816	to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
817	to->sb_rextents = be64_to_cpu(from->sb_rextents);
818	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
819	to->sb_logstart = be64_to_cpu(from->sb_logstart);
820	to->sb_rootino = be64_to_cpu(from->sb_rootino);
821	to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
822	to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
823	to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
824	to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
825	to->sb_agcount = be32_to_cpu(from->sb_agcount);
826	to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
827	to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
828	to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
829	to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
830	to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
831	to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
832	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
833	to->sb_blocklog = from->sb_blocklog;
834	to->sb_sectlog = from->sb_sectlog;
835	to->sb_inodelog = from->sb_inodelog;
836	to->sb_inopblog = from->sb_inopblog;
837	to->sb_agblklog = from->sb_agblklog;
838	to->sb_rextslog = from->sb_rextslog;
839	to->sb_inprogress = from->sb_inprogress;
840	to->sb_imax_pct = from->sb_imax_pct;
841	to->sb_icount = be64_to_cpu(from->sb_icount);
842	to->sb_ifree = be64_to_cpu(from->sb_ifree);
843	to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
844	to->sb_frextents = be64_to_cpu(from->sb_frextents);
845	to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
846	to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
847	to->sb_qflags = be16_to_cpu(from->sb_qflags);
848	to->sb_flags = from->sb_flags;
849	to->sb_shared_vn = from->sb_shared_vn;
850	to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
851	to->sb_unit = be32_to_cpu(from->sb_unit);
852	to->sb_width = be32_to_cpu(from->sb_width);
853	to->sb_dirblklog = from->sb_dirblklog;
854	to->sb_logsectlog = from->sb_logsectlog;
855	to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
856	to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
857	to->sb_features2 = be32_to_cpu(from->sb_features2);
858	to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
859	to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
860	to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
861	to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
862	to->sb_features_log_incompat =
863	be32_to_cpu(from->sb_features_log_incompat);
864	/* crc is only used on disk, not in memory; just init to 0 here. */
865	to->sb_crc = 0;
866	to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
867	to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
868	to->sb_lsn = be64_to_cpu(from->sb_lsn);
869	/*
870	* sb_meta_uuid is only on disk if it differs from sb_uuid and the
871	* feature flag is set; if not set we keep it only in memory.
872	*/
873	if (xfs_sb_is_v5(to) &&
874	(to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID))
875	uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
876	else
877	uuid_copy(&to->sb_meta_uuid, &from->sb_uuid);
878	/* Convert on-disk flags to in-memory flags? */
879	if (convert_xquota)
880	xfs_sb_quota_from_disk(sbp: to);
881
882	if (to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR) {
883	to->sb_metadirino = be64_to_cpu(from->sb_metadirino);
884	to->sb_rgblklog = from->sb_rgblklog;
885	memcpy(to->sb_pad, from->sb_pad, sizeof(to->sb_pad));
886	to->sb_rgcount = be32_to_cpu(from->sb_rgcount);
887	to->sb_rgextents = be32_to_cpu(from->sb_rgextents);
888	to->sb_rbmino = NULLFSINO;
889	to->sb_rsumino = NULLFSINO;
890	} else {
891	to->sb_metadirino = NULLFSINO;
892	to->sb_rgcount = 1;
893	to->sb_rgextents = 0;
894	}
895
896	if (to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED) {
897	to->sb_rtstart = be64_to_cpu(from->sb_rtstart);
898	to->sb_rtreserved = be64_to_cpu(from->sb_rtreserved);
899	} else {
900	to->sb_rtstart = 0;
901	to->sb_rtreserved = 0;
902	}
903	}
904
905	void
906	xfs_sb_from_disk(
907	struct xfs_sb *to,
908	struct xfs_dsb *from)
909	{
910	__xfs_sb_from_disk(to, from, true);
911	}
912
913	static void
914	xfs_sb_quota_to_disk(
915	struct xfs_dsb *to,
916	struct xfs_sb *from)
917	{
918	uint16_t qflags = from->sb_qflags;
919
920	if (xfs_sb_is_v5(from) &&
921	(from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)) {
922	to->sb_qflags = cpu_to_be16(from->sb_qflags);
923	to->sb_uquotino = cpu_to_be64(0);
924	to->sb_gquotino = cpu_to_be64(0);
925	to->sb_pquotino = cpu_to_be64(0);
926	return;
927	}
928
929	to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
930
931	/*
932	* The in-memory superblock quota state matches the v5 on-disk format so
933	* just write them out and return
934	*/
935	if (xfs_sb_is_v5(from)) {
936	to->sb_qflags = cpu_to_be16(from->sb_qflags);
937	to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
938	to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
939	return;
940	}
941
942	/*
943	* For older superblocks (v4), the in-core version of sb_qflags do not
944	* have XFS_OQUOTA_* flags, whereas the on-disk version does. So,
945	* convert incore XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
946	*/
947	qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
948	XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
949
950	if (from->sb_qflags &
951	(XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
952	qflags |= XFS_OQUOTA_ENFD;
953	if (from->sb_qflags &
954	(XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
955	qflags |= XFS_OQUOTA_CHKD;
956	to->sb_qflags = cpu_to_be16(qflags);
957
958	/*
959	* GQUOTINO and PQUOTINO cannot be used together in versions
960	* of superblock that do not have pquotino. from->sb_flags
961	* tells us which quota is active and should be copied to
962	* disk. If neither are active, we should NULL the inode.
963	*
964	* In all cases, the separate pquotino must remain 0 because it
965	* is beyond the "end" of the valid non-pquotino superblock.
966	*/
967	if (from->sb_qflags & XFS_GQUOTA_ACCT)
968	to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
969	else if (from->sb_qflags & XFS_PQUOTA_ACCT)
970	to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
971	else {
972	/*
973	* We can't rely on just the fields being logged to tell us
974	* that it is safe to write NULLFSINO - we should only do that
975	* if quotas are not actually enabled. Hence only write
976	* NULLFSINO if both in-core quota inodes are NULL.
977	*/
978	if (from->sb_gquotino == NULLFSINO &&
979	from->sb_pquotino == NULLFSINO)
980	to->sb_gquotino = cpu_to_be64(NULLFSINO);
981	}
982
983	to->sb_pquotino = 0;
984	}
985
986	void
987	xfs_sb_to_disk(
988	struct xfs_dsb *to,
989	struct xfs_sb *from)
990	{
991	xfs_sb_quota_to_disk(to, from);
992
993	to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
994	to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
995	to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
996	to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
997	to->sb_rextents = cpu_to_be64(from->sb_rextents);
998	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
999	to->sb_logstart = cpu_to_be64(from->sb_logstart);
1000	to->sb_rootino = cpu_to_be64(from->sb_rootino);
1001	to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
1002	to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
1003	to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
1004	to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
1005	to->sb_agcount = cpu_to_be32(from->sb_agcount);
1006	to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
1007	to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
1008	to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
1009	to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
1010	to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
1011	to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
1012	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
1013	to->sb_blocklog = from->sb_blocklog;
1014	to->sb_sectlog = from->sb_sectlog;
1015	to->sb_inodelog = from->sb_inodelog;
1016	to->sb_inopblog = from->sb_inopblog;
1017	to->sb_agblklog = from->sb_agblklog;
1018	to->sb_rextslog = from->sb_rextslog;
1019	to->sb_inprogress = from->sb_inprogress;
1020	to->sb_imax_pct = from->sb_imax_pct;
1021	to->sb_icount = cpu_to_be64(from->sb_icount);
1022	to->sb_ifree = cpu_to_be64(from->sb_ifree);
1023	to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
1024	to->sb_frextents = cpu_to_be64(from->sb_frextents);
1025
1026	to->sb_flags = from->sb_flags;
1027	to->sb_shared_vn = from->sb_shared_vn;
1028	to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
1029	to->sb_unit = cpu_to_be32(from->sb_unit);
1030	to->sb_width = cpu_to_be32(from->sb_width);
1031	to->sb_dirblklog = from->sb_dirblklog;
1032	to->sb_logsectlog = from->sb_logsectlog;
1033	to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
1034	to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
1035
1036	/*
1037	* We need to ensure that bad_features2 always matches features2.
1038	* Hence we enforce that here rather than having to remember to do it
1039	* everywhere else that updates features2.
1040	*/
1041	from->sb_bad_features2 = from->sb_features2;
1042	to->sb_features2 = cpu_to_be32(from->sb_features2);
1043	to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
1044
1045	if (!xfs_sb_is_v5(from))
1046	return;
1047
1048	to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
1049	to->sb_features_ro_compat =
1050	cpu_to_be32(from->sb_features_ro_compat);
1051	to->sb_features_incompat =
1052	cpu_to_be32(from->sb_features_incompat);
1053	to->sb_features_log_incompat =
1054	cpu_to_be32(from->sb_features_log_incompat);
1055	to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
1056	to->sb_lsn = cpu_to_be64(from->sb_lsn);
1057	if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID)
1058	uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
1059
1060	if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR) {
1061	to->sb_metadirino = cpu_to_be64(from->sb_metadirino);
1062	to->sb_rgblklog = from->sb_rgblklog;
1063	memset(to->sb_pad, 0, sizeof(to->sb_pad));
1064	to->sb_rgcount = cpu_to_be32(from->sb_rgcount);
1065	to->sb_rgextents = cpu_to_be32(from->sb_rgextents);
1066	to->sb_rbmino = cpu_to_be64(0);
1067	to->sb_rsumino = cpu_to_be64(0);
1068	}
1069
1070	if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_ZONED) {
1071	to->sb_rtstart = cpu_to_be64(from->sb_rtstart);
1072	to->sb_rtreserved = cpu_to_be64(from->sb_rtreserved);
1073	}
1074	}
1075
1076	/*
1077	* If the superblock has the CRC feature bit set or the CRC field is non-null,
1078	* check that the CRC is valid. We check the CRC field is non-null because a
1079	* single bit error could clear the feature bit and unused parts of the
1080	* superblock are supposed to be zero. Hence a non-null crc field indicates that
1081	* we've potentially lost a feature bit and we should check it anyway.
1082	*
1083	* However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
1084	* last field in V4 secondary superblocks. So for secondary superblocks,
1085	* we are more forgiving, and ignore CRC failures if the primary doesn't
1086	* indicate that the fs version is V5.
1087	*/
1088	static void
1089	xfs_sb_read_verify(
1090	struct xfs_buf *bp)
1091	{
1092	struct xfs_sb sb;
1093	struct xfs_mount *mp = bp->b_mount;
1094	struct xfs_dsb *dsb = bp->b_addr;
1095	int error;
1096
1097	/*
1098	* open code the version check to avoid needing to convert the entire
1099	* superblock from disk order just to check the version number
1100	*/
1101	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
1102	(((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
1103	XFS_SB_VERSION_5) ||
1104	dsb->sb_crc != 0)) {
1105
1106	if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
1107	/* Only fail bad secondaries on a known V5 filesystem */
1108	if (xfs_buf_daddr(bp) == XFS_SB_DADDR ||
1109	xfs_has_crc(mp)) {
1110	error = -EFSBADCRC;
1111	goto out_error;
1112	}
1113	}
1114	}
1115
1116	/*
1117	* Check all the superblock fields. Don't byteswap the xquota flags
1118	* because _verify_common checks the on-disk values.
1119	*/
1120	__xfs_sb_from_disk(&sb, dsb, false);
1121	error = xfs_validate_sb_common(mp, bp, &sb);
1122	if (error)
1123	goto out_error;
1124	error = xfs_validate_sb_read(mp, &sb);
1125
1126	out_error:
1127	if (error == -EFSCORRUPTED || error == -EFSBADCRC)
1128	xfs_verifier_error(bp, error, __this_address);
1129	else if (error)
1130	xfs_buf_ioerror(bp, error);
1131	}
1132
1133	/*
1134	* We may be probed for a filesystem match, so we may not want to emit
1135	* messages when the superblock buffer is not actually an XFS superblock.
1136	* If we find an XFS superblock, then run a normal, noisy mount because we are
1137	* really going to mount it and want to know about errors.
1138	*/
1139	static void
1140	xfs_sb_quiet_read_verify(
1141	struct xfs_buf *bp)
1142	{
1143	struct xfs_dsb *dsb = bp->b_addr;
1144
1145	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
1146	/* XFS filesystem, verify noisily! */
1147	xfs_sb_read_verify(bp);
1148	return;
1149	}
1150	/* quietly fail */
1151	xfs_buf_ioerror(bp, -EWRONGFS);
1152	}
1153
1154	static void
1155	xfs_sb_write_verify(
1156	struct xfs_buf *bp)
1157	{
1158	struct xfs_sb sb;
1159	struct xfs_mount *mp = bp->b_mount;
1160	struct xfs_buf_log_item *bip = bp->b_log_item;
1161	struct xfs_dsb *dsb = bp->b_addr;
1162	int error;
1163
1164	/*
1165	* Check all the superblock fields. Don't byteswap the xquota flags
1166	* because _verify_common checks the on-disk values.
1167	*/
1168	__xfs_sb_from_disk(&sb, dsb, false);
1169	error = xfs_validate_sb_common(mp, bp, &sb);
1170	if (error)
1171	goto out_error;
1172	error = xfs_validate_sb_write(mp, bp, &sb);
1173	if (error)
1174	goto out_error;
1175
1176	if (!xfs_sb_is_v5(&sb))
1177	return;
1178
1179	if (bip)
1180	dsb->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
1181
1182	xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
1183	return;
1184
1185	out_error:
1186	xfs_verifier_error(bp, error, __this_address);
1187	}
1188
1189	const struct xfs_buf_ops xfs_sb_buf_ops = {
1190	.name = "xfs_sb",
1191	.magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
1192	.verify_read = xfs_sb_read_verify,
1193	.verify_write = xfs_sb_write_verify,
1194	};
1195
1196	const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
1197	.name = "xfs_sb_quiet",
1198	.magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
1199	.verify_read = xfs_sb_quiet_read_verify,
1200	.verify_write = xfs_sb_write_verify,
1201	};
1202
1203	/* Compute cached rt geometry from the incore sb. */
1204	void
1205	xfs_sb_mount_rextsize(
1206	struct xfs_mount *mp,
1207	struct xfs_sb *sbp)
1208	{
1209	struct xfs_groups *rgs = &mp->m_groups[XG_TYPE_RTG];
1210
1211	mp->m_rtxblklog = log2_if_power2(sbp->sb_rextsize);
1212	mp->m_rtxblkmask = mask64_if_power2(sbp->sb_rextsize);
1213
1214	if (xfs_sb_is_v5(sbp) &&
1215	(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR)) {
1216	rgs->blocks = sbp->sb_rgextents * sbp->sb_rextsize;
1217	rgs->blklog = mp->m_sb.sb_rgblklog;
1218	rgs->blkmask = xfs_mask32lo(mp->m_sb.sb_rgblklog);
1219	rgs->start_fsb = mp->m_sb.sb_rtstart;
1220	if (xfs_sb_has_incompat_feature(sbp,
1221	XFS_SB_FEAT_INCOMPAT_ZONE_GAPS))
1222	rgs->has_daddr_gaps = true;
1223	} else {
1224	rgs->blocks = 0;
1225	rgs->blklog = 0;
1226	rgs->blkmask = (uint64_t)-1;
1227	}
1228	}
1229
1230	/* Update incore sb rt extent size, then recompute the cached rt geometry. */
1231	void
1232	xfs_mount_sb_set_rextsize(
1233	struct xfs_mount *mp,
1234	struct xfs_sb *sbp,
1235	xfs_agblock_t rextsize)
1236	{
1237	sbp->sb_rextsize = rextsize;
1238	if (xfs_sb_is_v5(sbp) &&
1239	(sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_METADIR))
1240	sbp->sb_rgblklog = xfs_compute_rgblklog(sbp->sb_rgextents,
1241	rextsize);
1242
1243	xfs_sb_mount_rextsize(mp, sbp);
1244	}
1245
1246	/*
1247	* xfs_mount_common
1248	*
1249	* Mount initialization code establishing various mount
1250	* fields from the superblock associated with the given
1251	* mount structure.
1252	*
1253	* Inode geometry are calculated in xfs_ialloc_setup_geometry.
1254	*/
1255	void
1256	xfs_sb_mount_common(
1257	struct xfs_mount *mp,
1258	struct xfs_sb *sbp)
1259	{
1260	struct xfs_groups *ags = &mp->m_groups[XG_TYPE_AG];
1261
1262	mp->m_agfrotor = 0;
1263	atomic_set(&mp->m_agirotor, 0);
1264	mp->m_maxagi = mp->m_sb.sb_agcount;
1265	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
1266	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
1267	mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
1268	mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
1269	mp->m_blockmask = sbp->sb_blocksize - 1;
1270	mp->m_blockwsize = xfs_rtbmblock_size(sbp) >> XFS_WORDLOG;
1271	mp->m_rtx_per_rbmblock = mp->m_blockwsize << XFS_NBWORDLOG;
1272
1273	ags->blocks = mp->m_sb.sb_agblocks;
1274	ags->blklog = mp->m_sb.sb_agblklog;
1275	ags->blkmask = xfs_mask32lo(mp->m_sb.sb_agblklog);
1276
1277	xfs_sb_mount_rextsize(mp, sbp);
1278
1279	mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, true);
1280	mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, false);
1281	mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
1282	mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
1283
1284	mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, true);
1285	mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, false);
1286	mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
1287	mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
1288
1289	mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, true);
1290	mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, false);
1291	mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2;
1292	mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2;
1293
1294	mp->m_rtrmap_mxr[0] = xfs_rtrmapbt_maxrecs(mp, sbp->sb_blocksize, true);
1295	mp->m_rtrmap_mxr[1] = xfs_rtrmapbt_maxrecs(mp, sbp->sb_blocksize, false);
1296	mp->m_rtrmap_mnr[0] = mp->m_rtrmap_mxr[0] / 2;
1297	mp->m_rtrmap_mnr[1] = mp->m_rtrmap_mxr[1] / 2;
1298
1299	mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, true);
1300	mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, false);
1301	mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2;
1302	mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2;
1303
1304	mp->m_rtrefc_mxr[0] = xfs_rtrefcountbt_maxrecs(mp, sbp->sb_blocksize,
1305	true);
1306	mp->m_rtrefc_mxr[1] = xfs_rtrefcountbt_maxrecs(mp, sbp->sb_blocksize,
1307	false);
1308	mp->m_rtrefc_mnr[0] = mp->m_rtrefc_mxr[0] / 2;
1309	mp->m_rtrefc_mnr[1] = mp->m_rtrefc_mxr[1] / 2;
1310
1311	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
1312	mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
1313	mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp);
1314	}
1315
1316	/*
1317	* xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
1318	* into the superblock buffer to be logged. It does not provide the higher
1319	* level of locking that is needed to protect the in-core superblock from
1320	* concurrent access.
1321	*/
1322	void
1323	xfs_log_sb(
1324	struct xfs_trans *tp)
1325	{
1326	struct xfs_mount *mp = tp->t_mountp;
1327	struct xfs_buf *bp = xfs_trans_getsb(tp);
1328
1329	/*
1330	* Lazy sb counters don't update the in-core superblock so do that now.
1331	* If this is at unmount, the counters will be exactly correct, but at
1332	* any other time they will only be ballpark correct because of
1333	* reservations that have been taken out percpu counters. If we have an
1334	* unclean shutdown, this will be corrected by log recovery rebuilding
1335	* the counters from the AGF block counts.
1336	*/
1337	if (xfs_has_lazysbcount(mp)) {
1338	mp->m_sb.sb_icount = percpu_counter_sum_positive(&mp->m_icount);
1339	mp->m_sb.sb_ifree = min_t(uint64_t,
1340	percpu_counter_sum_positive(&mp->m_ifree),
1341	mp->m_sb.sb_icount);
1342	mp->m_sb.sb_fdblocks = xfs_sum_freecounter(mp, XC_FREE_BLOCKS);
1343	}
1344
1345	/*
1346	* sb_frextents was added to the lazy sb counters when the rt groups
1347	* feature was introduced. This counter can go negative due to the way
1348	* we handle nearly-lockless reservations, so we must use the _positive
1349	* variant here to avoid writing out nonsense frextents.
1350	*/
1351	if (xfs_has_rtgroups(mp) && !xfs_has_zoned(mp)) {
1352	mp->m_sb.sb_frextents =
1353	xfs_sum_freecounter(mp, XC_FREE_RTEXTENTS);
1354	}
1355
1356	xfs_sb_to_disk(to: bp->b_addr, from: &mp->m_sb);
1357	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
1358	xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb) - 1);
1359	}
1360
1361	/*
1362	* xfs_sync_sb
1363	*
1364	* Sync the superblock to disk.
1365	*
1366	* Note that the caller is responsible for checking the frozen state of the
1367	* filesystem. This procedure uses the non-blocking transaction allocator and
1368	* thus will allow modifications to a frozen fs. This is required because this
1369	* code can be called during the process of freezing where use of the high-level
1370	* allocator would deadlock.
1371	*/
1372	int
1373	xfs_sync_sb(
1374	struct xfs_mount *mp,
1375	bool wait)
1376	{
1377	struct xfs_trans *tp;
1378	int error;
1379
1380	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0,
1381	XFS_TRANS_NO_WRITECOUNT, &tp);
1382	if (error)
1383	return error;
1384
1385	xfs_log_sb(tp);
1386	if (wait)
1387	xfs_trans_set_sync(tp);
1388	return xfs_trans_commit(tp);
1389	}
1390
1391	/*
1392	* Update all the secondary superblocks to match the new state of the primary.
1393	* Because we are completely overwriting all the existing fields in the
1394	* secondary superblock buffers, there is no need to read them in from disk.
1395	* Just get a new buffer, stamp it and write it.
1396	*
1397	* The sb buffers need to be cached here so that we serialise against other
1398	* operations that access the secondary superblocks, but we don't want to keep
1399	* them in memory once it is written so we mark it as a one-shot buffer.
1400	*/
1401	int
1402	xfs_update_secondary_sbs(
1403	struct xfs_mount *mp)
1404	{
1405	struct xfs_perag *pag = NULL;
1406	int saved_error = 0;
1407	int error = 0;
1408	LIST_HEAD (buffer_list);
1409
1410	/* update secondary superblocks. */
1411	while ((pag = xfs_perag_next_from(mp, pag, 1))) {
1412	struct xfs_buf *bp;
1413
1414	error = xfs_buf_get(mp->m_ddev_targp,
1415	XFS_AG_DADDR(mp, pag_agno(pag), XFS_SB_DADDR),
1416	XFS_FSS_TO_BB(mp, 1), &bp);
1417	/*
1418	* If we get an error reading or writing alternate superblocks,
1419	* continue. xfs_repair chooses the "best" superblock based
1420	* on most matches; if we break early, we'll leave more
1421	* superblocks un-updated than updated, and xfs_repair may
1422	* pick them over the properly-updated primary.
1423	*/
1424	if (error) {
1425	xfs_warn(mp,
1426	"error allocating secondary superblock for ag %d",
1427	pag_agno(pag));
1428	if (!saved_error)
1429	saved_error = error;
1430	continue;
1431	}
1432
1433	bp->b_ops = &xfs_sb_buf_ops;
1434	xfs_buf_oneshot(bp);
1435	xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
1436	xfs_sb_to_disk(to: bp->b_addr, from: &mp->m_sb);
1437	xfs_buf_delwri_queue(bp, &buffer_list);
1438	xfs_buf_relse(bp);
1439
1440	/* don't hold too many buffers at once */
1441	if (pag_agno(pag) % 16)
1442	continue;
1443
1444	error = xfs_buf_delwri_submit(&buffer_list);
1445	if (error) {
1446	xfs_warn(mp,
1447	"write error %d updating a secondary superblock near ag %d",
1448	error, pag_agno(pag));
1449	if (!saved_error)
1450	saved_error = error;
1451	continue;
1452	}
1453	}
1454	error = xfs_buf_delwri_submit(&buffer_list);
1455	if (error)
1456	xfs_warn(mp, "error %d writing secondary superblocks", error);
1457	return saved_error ? saved_error : error;
1458	}
1459
1460	/*
1461	* Same behavior as xfs_sync_sb, except that it is always synchronous and it
1462	* also writes the superblock buffer to disk sector 0 immediately.
1463	*/
1464	int
1465	xfs_sync_sb_buf(
1466	struct xfs_mount *mp,
1467	bool update_rtsb)
1468	{
1469	struct xfs_trans *tp;
1470	struct xfs_buf *bp;
1471	struct xfs_buf *rtsb_bp = NULL;
1472	int error;
1473
1474	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0, 0, &tp);
1475	if (error)
1476	return error;
1477
1478	bp = xfs_trans_getsb(tp);
1479	xfs_log_sb(tp);
1480	xfs_trans_bhold(tp, bp);
1481	if (update_rtsb) {
1482	rtsb_bp = xfs_log_rtsb(tp, sb_bp: bp);
1483	if (rtsb_bp)
1484	xfs_trans_bhold(tp, rtsb_bp);
1485	}
1486	xfs_trans_set_sync(tp);
1487	error = xfs_trans_commit(tp);
1488	if (error)
1489	goto out;
1490	/*
1491	* write out the sb buffer to get the changes to disk
1492	*/
1493	error = xfs_bwrite(bp);
1494	if (!error && rtsb_bp)
1495	error = xfs_bwrite(rtsb_bp);
1496	out:
1497	if (rtsb_bp)
1498	xfs_buf_relse(rtsb_bp);
1499	xfs_buf_relse(bp);
1500	return error;
1501	}
1502
1503	void
1504	xfs_fs_geometry(
1505	struct xfs_mount *mp,
1506	struct xfs_fsop_geom *geo,
1507	int struct_version)
1508	{
1509	struct xfs_sb *sbp = &mp->m_sb;
1510
1511	memset(geo, 0, sizeof(struct xfs_fsop_geom));
1512
1513	geo->blocksize = sbp->sb_blocksize;
1514	geo->rtextsize = sbp->sb_rextsize;
1515	geo->agblocks = sbp->sb_agblocks;
1516	geo->agcount = sbp->sb_agcount;
1517	geo->logblocks = sbp->sb_logblocks;
1518	geo->sectsize = sbp->sb_sectsize;
1519	geo->inodesize = sbp->sb_inodesize;
1520	geo->imaxpct = sbp->sb_imax_pct;
1521	geo->datablocks = sbp->sb_dblocks;
1522	geo->rtblocks = sbp->sb_rblocks;
1523	geo->rtextents = sbp->sb_rextents;
1524	geo->logstart = sbp->sb_logstart;
1525	BUILD_BUG_ON(sizeof(geo->uuid) != sizeof(sbp->sb_uuid));
1526	memcpy(geo->uuid, &sbp->sb_uuid, sizeof(sbp->sb_uuid));
1527
1528	if (struct_version < 2)
1529	return;
1530
1531	geo->sunit = sbp->sb_unit;
1532	geo->swidth = sbp->sb_width;
1533
1534	if (struct_version < 3)
1535	return;
1536
1537	geo->version = XFS_FSOP_GEOM_VERSION;
1538	geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
1539	XFS_FSOP_GEOM_FLAGS_DIRV2 |
1540	XFS_FSOP_GEOM_FLAGS_EXTFLG |
1541	XFS_FSOP_GEOM_FLAGS_ATTR2;
1542	if (xfs_has_attr(mp))
1543	geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR;
1544	if (xfs_has_quota(mp))
1545	geo->flags |= XFS_FSOP_GEOM_FLAGS_QUOTA;
1546	if (xfs_has_align(mp))
1547	geo->flags |= XFS_FSOP_GEOM_FLAGS_IALIGN;
1548	if (xfs_has_dalign(mp))
1549	geo->flags |= XFS_FSOP_GEOM_FLAGS_DALIGN;
1550	if (xfs_has_asciici(mp))
1551	geo->flags |= XFS_FSOP_GEOM_FLAGS_DIRV2CI;
1552	if (xfs_has_lazysbcount(mp))
1553	geo->flags |= XFS_FSOP_GEOM_FLAGS_LAZYSB;
1554	if (xfs_has_projid32(mp))
1555	geo->flags |= XFS_FSOP_GEOM_FLAGS_PROJID32;
1556	if (xfs_has_crc(mp))
1557	geo->flags |= XFS_FSOP_GEOM_FLAGS_V5SB;
1558	if (xfs_has_ftype(mp))
1559	geo->flags |= XFS_FSOP_GEOM_FLAGS_FTYPE;
1560	if (xfs_has_finobt(mp))
1561	geo->flags |= XFS_FSOP_GEOM_FLAGS_FINOBT;
1562	if (xfs_has_sparseinodes(mp))
1563	geo->flags |= XFS_FSOP_GEOM_FLAGS_SPINODES;
1564	if (xfs_has_rmapbt(mp))
1565	geo->flags |= XFS_FSOP_GEOM_FLAGS_RMAPBT;
1566	if (xfs_has_reflink(mp))
1567	geo->flags |= XFS_FSOP_GEOM_FLAGS_REFLINK;
1568	if (xfs_has_bigtime(mp))
1569	geo->flags |= XFS_FSOP_GEOM_FLAGS_BIGTIME;
1570	if (xfs_has_inobtcounts(mp))
1571	geo->flags |= XFS_FSOP_GEOM_FLAGS_INOBTCNT;
1572	if (xfs_has_parent(mp))
1573	geo->flags |= XFS_FSOP_GEOM_FLAGS_PARENT;
1574	if (xfs_has_sector(mp)) {
1575	geo->flags |= XFS_FSOP_GEOM_FLAGS_SECTOR;
1576	geo->logsectsize = sbp->sb_logsectsize;
1577	} else {
1578	geo->logsectsize = BBSIZE;
1579	}
1580	if (xfs_has_large_extent_counts(mp))
1581	geo->flags |= XFS_FSOP_GEOM_FLAGS_NREXT64;
1582	if (xfs_has_exchange_range(mp))
1583	geo->flags |= XFS_FSOP_GEOM_FLAGS_EXCHANGE_RANGE;
1584	if (xfs_has_metadir(mp))
1585	geo->flags |= XFS_FSOP_GEOM_FLAGS_METADIR;
1586	if (xfs_has_zoned(mp))
1587	geo->flags |= XFS_FSOP_GEOM_FLAGS_ZONED;
1588	geo->rtsectsize = sbp->sb_blocksize;
1589	geo->dirblocksize = xfs_dir2_dirblock_bytes(sbp);
1590
1591	if (struct_version < 4)
1592	return;
1593
1594	if (xfs_has_logv2(mp))
1595	geo->flags |= XFS_FSOP_GEOM_FLAGS_LOGV2;
1596
1597	geo->logsunit = sbp->sb_logsunit;
1598
1599	if (struct_version < 5)
1600	return;
1601
1602	geo->version = XFS_FSOP_GEOM_VERSION_V5;
1603
1604	if (xfs_has_rtgroups(mp)) {
1605	geo->rgcount = sbp->sb_rgcount;
1606	geo->rgextents = sbp->sb_rgextents;
1607	}
1608	if (xfs_has_zoned(mp)) {
1609	geo->rtstart = sbp->sb_rtstart;
1610	geo->rtreserved = sbp->sb_rtreserved;
1611	}
1612	}
1613
1614	/* Read a secondary superblock. */
1615	int
1616	xfs_sb_read_secondary(
1617	struct xfs_mount *mp,
1618	struct xfs_trans *tp,
1619	xfs_agnumber_t agno,
1620	struct xfs_buf **bpp)
1621	{
1622	struct xfs_buf *bp;
1623	int error;
1624
1625	ASSERT(agno != 0 && agno != NULLAGNUMBER);
1626	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
1627	XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1628	XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_sb_buf_ops);
1629	if (xfs_metadata_is_sick(error))
1630	xfs_agno_mark_sick(mp, agno, XFS_SICK_AG_SB);
1631	if (error)
1632	return error;
1633	xfs_buf_set_ref(bp, XFS_SSB_REF);
1634	*bpp = bp;
1635	return 0;
1636	}
1637
1638	/* Get an uninitialised secondary superblock buffer. */
1639	int
1640	xfs_sb_get_secondary(
1641	struct xfs_mount *mp,
1642	struct xfs_trans *tp,
1643	xfs_agnumber_t agno,
1644	struct xfs_buf **bpp)
1645	{
1646	struct xfs_buf *bp;
1647	int error;
1648
1649	ASSERT(agno != 0 && agno != NULLAGNUMBER);
1650	error = xfs_trans_get_buf(tp, mp->m_ddev_targp,
1651	XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1652	XFS_FSS_TO_BB(mp, 1), 0, &bp);
1653	if (error)
1654	return error;
1655	bp->b_ops = &xfs_sb_buf_ops;
1656	xfs_buf_oneshot(bp);
1657	*bpp = bp;
1658	return 0;
1659	}
1660
1661	/*
1662	* sunit, swidth, sectorsize(optional with 0) should be all in bytes, so users
1663	* won't be confused by values in error messages. This function returns false
1664	* if the stripe geometry is invalid and the caller is unable to repair the
1665	* stripe configuration later in the mount process.
1666	*/
1667	bool
1668	xfs_validate_stripe_geometry(
1669	struct xfs_mount *mp,
1670	__s64 sunit,
1671	__s64 swidth,
1672	int sectorsize,
1673	bool may_repair,
1674	bool silent)
1675	{
1676	if (swidth > INT_MAX) {
1677	if (!silent)
1678	xfs_notice(mp,
1679	"stripe width (%lld) is too large", swidth);
1680	goto check_override;
1681	}
1682
1683	if (sunit > swidth) {
1684	if (!silent)
1685	xfs_notice(mp,
1686	"stripe unit (%lld) is larger than the stripe width (%lld)", sunit, swidth);
1687	goto check_override;
1688	}
1689
1690	if (sectorsize && (int)sunit % sectorsize) {
1691	if (!silent)
1692	xfs_notice(mp,
1693	"stripe unit (%lld) must be a multiple of the sector size (%d)",
1694	sunit, sectorsize);
1695	goto check_override;
1696	}
1697
1698	if (sunit && !swidth) {
1699	if (!silent)
1700	xfs_notice(mp,
1701	"invalid stripe unit (%lld) and stripe width of 0", sunit);
1702	goto check_override;
1703	}
1704
1705	if (!sunit && swidth) {
1706	if (!silent)
1707	xfs_notice(mp,
1708	"invalid stripe width (%lld) and stripe unit of 0", swidth);
1709	goto check_override;
1710	}
1711
1712	if (sunit && (int)swidth % (int)sunit) {
1713	if (!silent)
1714	xfs_notice(mp,
1715	"stripe width (%lld) must be a multiple of the stripe unit (%lld)",
1716	swidth, sunit);
1717	goto check_override;
1718	}
1719	return true;
1720
1721	check_override:
1722	if (!may_repair)
1723	return false;
1724	/*
1725	* During mount, mp->m_dalign will not be set unless the sunit mount
1726	* option was set. If it was set, ignore the bad stripe alignment values
1727	* and allow the validation and overwrite later in the mount process to
1728	* attempt to overwrite the bad stripe alignment values with the values
1729	* supplied by mount options.
1730	*/
1731	if (!mp->m_dalign)
1732	return false;
1733	if (!silent)
1734	xfs_notice(mp,
1735	"Will try to correct with specified mount options sunit (%d) and swidth (%d)",
1736	BBTOB(mp->m_dalign), BBTOB(mp->m_swidth));
1737	return true;
1738	}
1739
1740	/*
1741	* Compute the maximum level number of the realtime summary file, as defined by
1742	* mkfs. The historic use of highbit32 on a 64-bit quantity prohibited correct
1743	* use of rt volumes with more than 2^32 extents.
1744	*/
1745	uint8_t
1746	xfs_compute_rextslog(
1747	xfs_rtbxlen_t rtextents)
1748	{
1749	if (!rtextents)
1750	return 0;
1751	return xfs_highbit64(rtextents);
1752	}
1753