xref: /openbmc/linux/fs/xfs/libxfs/xfs_sb.c (revision 5e29a910)
1 /*
2  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU General Public License as
7  * published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope that it would be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12  * GNU General Public License for more details.
13  *
14  * You should have received a copy of the GNU General Public License
15  * along with this program; if not, write the Free Software Foundation,
16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17  */
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
24 #include "xfs_bit.h"
25 #include "xfs_sb.h"
26 #include "xfs_mount.h"
27 #include "xfs_inode.h"
28 #include "xfs_ialloc.h"
29 #include "xfs_alloc.h"
30 #include "xfs_error.h"
31 #include "xfs_trace.h"
32 #include "xfs_cksum.h"
33 #include "xfs_trans.h"
34 #include "xfs_buf_item.h"
35 #include "xfs_bmap_btree.h"
36 #include "xfs_alloc_btree.h"
37 #include "xfs_ialloc_btree.h"
38 
39 /*
40  * Physical superblock buffer manipulations. Shared with libxfs in userspace.
41  */
42 
43 /*
44  * Reference counting access wrappers to the perag structures.
45  * Because we never free per-ag structures, the only thing we
46  * have to protect against changes is the tree structure itself.
47  */
48 struct xfs_perag *
49 xfs_perag_get(
50 	struct xfs_mount	*mp,
51 	xfs_agnumber_t		agno)
52 {
53 	struct xfs_perag	*pag;
54 	int			ref = 0;
55 
56 	rcu_read_lock();
57 	pag = radix_tree_lookup(&mp->m_perag_tree, agno);
58 	if (pag) {
59 		ASSERT(atomic_read(&pag->pag_ref) >= 0);
60 		ref = atomic_inc_return(&pag->pag_ref);
61 	}
62 	rcu_read_unlock();
63 	trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
64 	return pag;
65 }
66 
67 /*
68  * search from @first to find the next perag with the given tag set.
69  */
70 struct xfs_perag *
71 xfs_perag_get_tag(
72 	struct xfs_mount	*mp,
73 	xfs_agnumber_t		first,
74 	int			tag)
75 {
76 	struct xfs_perag	*pag;
77 	int			found;
78 	int			ref;
79 
80 	rcu_read_lock();
81 	found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
82 					(void **)&pag, first, 1, tag);
83 	if (found <= 0) {
84 		rcu_read_unlock();
85 		return NULL;
86 	}
87 	ref = atomic_inc_return(&pag->pag_ref);
88 	rcu_read_unlock();
89 	trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
90 	return pag;
91 }
92 
93 void
94 xfs_perag_put(
95 	struct xfs_perag	*pag)
96 {
97 	int	ref;
98 
99 	ASSERT(atomic_read(&pag->pag_ref) > 0);
100 	ref = atomic_dec_return(&pag->pag_ref);
101 	trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
102 }
103 
104 /*
105  * Check the validity of the SB found.
106  */
107 STATIC int
108 xfs_mount_validate_sb(
109 	xfs_mount_t	*mp,
110 	xfs_sb_t	*sbp,
111 	bool		check_inprogress,
112 	bool		check_version)
113 {
114 
115 	/*
116 	 * If the log device and data device have the
117 	 * same device number, the log is internal.
118 	 * Consequently, the sb_logstart should be non-zero.  If
119 	 * we have a zero sb_logstart in this case, we may be trying to mount
120 	 * a volume filesystem in a non-volume manner.
121 	 */
122 	if (sbp->sb_magicnum != XFS_SB_MAGIC) {
123 		xfs_warn(mp, "bad magic number");
124 		return -EWRONGFS;
125 	}
126 
127 
128 	if (!xfs_sb_good_version(sbp)) {
129 		xfs_warn(mp, "bad version");
130 		return -EWRONGFS;
131 	}
132 
133 	/*
134 	 * Version 5 superblock feature mask validation. Reject combinations the
135 	 * kernel cannot support up front before checking anything else. For
136 	 * write validation, we don't need to check feature masks.
137 	 */
138 	if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) {
139 		if (xfs_sb_has_compat_feature(sbp,
140 					XFS_SB_FEAT_COMPAT_UNKNOWN)) {
141 			xfs_warn(mp,
142 "Superblock has unknown compatible features (0x%x) enabled.\n"
143 "Using a more recent kernel is recommended.",
144 				(sbp->sb_features_compat &
145 						XFS_SB_FEAT_COMPAT_UNKNOWN));
146 		}
147 
148 		if (xfs_sb_has_ro_compat_feature(sbp,
149 					XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
150 			xfs_alert(mp,
151 "Superblock has unknown read-only compatible features (0x%x) enabled.",
152 				(sbp->sb_features_ro_compat &
153 						XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
154 			if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
155 				xfs_warn(mp,
156 "Attempted to mount read-only compatible filesystem read-write.\n"
157 "Filesystem can only be safely mounted read only.");
158 				return -EINVAL;
159 			}
160 		}
161 		if (xfs_sb_has_incompat_feature(sbp,
162 					XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
163 			xfs_warn(mp,
164 "Superblock has unknown incompatible features (0x%x) enabled.\n"
165 "Filesystem can not be safely mounted by this kernel.",
166 				(sbp->sb_features_incompat &
167 						XFS_SB_FEAT_INCOMPAT_UNKNOWN));
168 			return -EINVAL;
169 		}
170 	}
171 
172 	if (xfs_sb_version_has_pquotino(sbp)) {
173 		if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
174 			xfs_notice(mp,
175 			   "Version 5 of Super block has XFS_OQUOTA bits.");
176 			return -EFSCORRUPTED;
177 		}
178 	} else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
179 				XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
180 			xfs_notice(mp,
181 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
182 			return -EFSCORRUPTED;
183 	}
184 
185 	if (unlikely(
186 	    sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
187 		xfs_warn(mp,
188 		"filesystem is marked as having an external log; "
189 		"specify logdev on the mount command line.");
190 		return -EINVAL;
191 	}
192 
193 	if (unlikely(
194 	    sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
195 		xfs_warn(mp,
196 		"filesystem is marked as having an internal log; "
197 		"do not specify logdev on the mount command line.");
198 		return -EINVAL;
199 	}
200 
201 	/*
202 	 * More sanity checking.  Most of these were stolen directly from
203 	 * xfs_repair.
204 	 */
205 	if (unlikely(
206 	    sbp->sb_agcount <= 0					||
207 	    sbp->sb_sectsize < XFS_MIN_SECTORSIZE			||
208 	    sbp->sb_sectsize > XFS_MAX_SECTORSIZE			||
209 	    sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG			||
210 	    sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG			||
211 	    sbp->sb_sectsize != (1 << sbp->sb_sectlog)			||
212 	    sbp->sb_blocksize < XFS_MIN_BLOCKSIZE			||
213 	    sbp->sb_blocksize > XFS_MAX_BLOCKSIZE			||
214 	    sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG			||
215 	    sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG			||
216 	    sbp->sb_blocksize != (1 << sbp->sb_blocklog)		||
217 	    sbp->sb_dirblklog > XFS_MAX_BLOCKSIZE_LOG			||
218 	    sbp->sb_inodesize < XFS_DINODE_MIN_SIZE			||
219 	    sbp->sb_inodesize > XFS_DINODE_MAX_SIZE			||
220 	    sbp->sb_inodelog < XFS_DINODE_MIN_LOG			||
221 	    sbp->sb_inodelog > XFS_DINODE_MAX_LOG			||
222 	    sbp->sb_inodesize != (1 << sbp->sb_inodelog)		||
223 	    sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE			||
224 	    sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
225 	    (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog)	||
226 	    (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE)	||
227 	    (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)	||
228 	    (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */)	||
229 	    sbp->sb_dblocks == 0					||
230 	    sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp)			||
231 	    sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp)			||
232 	    sbp->sb_shared_vn != 0)) {
233 		xfs_notice(mp, "SB sanity check failed");
234 		return -EFSCORRUPTED;
235 	}
236 
237 	/*
238 	 * Until this is fixed only page-sized or smaller data blocks work.
239 	 */
240 	if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
241 		xfs_warn(mp,
242 		"File system with blocksize %d bytes. "
243 		"Only pagesize (%ld) or less will currently work.",
244 				sbp->sb_blocksize, PAGE_SIZE);
245 		return -ENOSYS;
246 	}
247 
248 	/*
249 	 * Currently only very few inode sizes are supported.
250 	 */
251 	switch (sbp->sb_inodesize) {
252 	case 256:
253 	case 512:
254 	case 1024:
255 	case 2048:
256 		break;
257 	default:
258 		xfs_warn(mp, "inode size of %d bytes not supported",
259 				sbp->sb_inodesize);
260 		return -ENOSYS;
261 	}
262 
263 	if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
264 	    xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
265 		xfs_warn(mp,
266 		"file system too large to be mounted on this system.");
267 		return -EFBIG;
268 	}
269 
270 	if (check_inprogress && sbp->sb_inprogress) {
271 		xfs_warn(mp, "Offline file system operation in progress!");
272 		return -EFSCORRUPTED;
273 	}
274 	return 0;
275 }
276 
277 void
278 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
279 {
280 	/*
281 	 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
282 	 * leads to in-core values having two different values for a quota
283 	 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
284 	 * NULLFSINO.
285 	 *
286 	 * Note that this change affect only the in-core values. These
287 	 * values are not written back to disk unless any quota information
288 	 * is written to the disk. Even in that case, sb_pquotino field is
289 	 * not written to disk unless the superblock supports pquotino.
290 	 */
291 	if (sbp->sb_uquotino == 0)
292 		sbp->sb_uquotino = NULLFSINO;
293 	if (sbp->sb_gquotino == 0)
294 		sbp->sb_gquotino = NULLFSINO;
295 	if (sbp->sb_pquotino == 0)
296 		sbp->sb_pquotino = NULLFSINO;
297 
298 	/*
299 	 * We need to do these manipilations only if we are working
300 	 * with an older version of on-disk superblock.
301 	 */
302 	if (xfs_sb_version_has_pquotino(sbp))
303 		return;
304 
305 	if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
306 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
307 					XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
308 	if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
309 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
310 					XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
311 	sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
312 
313 	if (sbp->sb_qflags & XFS_PQUOTA_ACCT)  {
314 		/*
315 		 * In older version of superblock, on-disk superblock only
316 		 * has sb_gquotino, and in-core superblock has both sb_gquotino
317 		 * and sb_pquotino. But, only one of them is supported at any
318 		 * point of time. So, if PQUOTA is set in disk superblock,
319 		 * copy over sb_gquotino to sb_pquotino.
320 		 */
321 		sbp->sb_pquotino = sbp->sb_gquotino;
322 		sbp->sb_gquotino = NULLFSINO;
323 	}
324 }
325 
326 static void
327 __xfs_sb_from_disk(
328 	struct xfs_sb	*to,
329 	xfs_dsb_t	*from,
330 	bool		convert_xquota)
331 {
332 	to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
333 	to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
334 	to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
335 	to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
336 	to->sb_rextents = be64_to_cpu(from->sb_rextents);
337 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
338 	to->sb_logstart = be64_to_cpu(from->sb_logstart);
339 	to->sb_rootino = be64_to_cpu(from->sb_rootino);
340 	to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
341 	to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
342 	to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
343 	to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
344 	to->sb_agcount = be32_to_cpu(from->sb_agcount);
345 	to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
346 	to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
347 	to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
348 	to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
349 	to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
350 	to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
351 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
352 	to->sb_blocklog = from->sb_blocklog;
353 	to->sb_sectlog = from->sb_sectlog;
354 	to->sb_inodelog = from->sb_inodelog;
355 	to->sb_inopblog = from->sb_inopblog;
356 	to->sb_agblklog = from->sb_agblklog;
357 	to->sb_rextslog = from->sb_rextslog;
358 	to->sb_inprogress = from->sb_inprogress;
359 	to->sb_imax_pct = from->sb_imax_pct;
360 	to->sb_icount = be64_to_cpu(from->sb_icount);
361 	to->sb_ifree = be64_to_cpu(from->sb_ifree);
362 	to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
363 	to->sb_frextents = be64_to_cpu(from->sb_frextents);
364 	to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
365 	to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
366 	to->sb_qflags = be16_to_cpu(from->sb_qflags);
367 	to->sb_flags = from->sb_flags;
368 	to->sb_shared_vn = from->sb_shared_vn;
369 	to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
370 	to->sb_unit = be32_to_cpu(from->sb_unit);
371 	to->sb_width = be32_to_cpu(from->sb_width);
372 	to->sb_dirblklog = from->sb_dirblklog;
373 	to->sb_logsectlog = from->sb_logsectlog;
374 	to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
375 	to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
376 	to->sb_features2 = be32_to_cpu(from->sb_features2);
377 	to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
378 	to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
379 	to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
380 	to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
381 	to->sb_features_log_incompat =
382 				be32_to_cpu(from->sb_features_log_incompat);
383 	/* crc is only used on disk, not in memory; just init to 0 here. */
384 	to->sb_crc = 0;
385 	to->sb_pad = 0;
386 	to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
387 	to->sb_lsn = be64_to_cpu(from->sb_lsn);
388 	/* Convert on-disk flags to in-memory flags? */
389 	if (convert_xquota)
390 		xfs_sb_quota_from_disk(to);
391 }
392 
393 void
394 xfs_sb_from_disk(
395 	struct xfs_sb	*to,
396 	xfs_dsb_t	*from)
397 {
398 	__xfs_sb_from_disk(to, from, true);
399 }
400 
401 static void
402 xfs_sb_quota_to_disk(
403 	struct xfs_dsb	*to,
404 	struct xfs_sb	*from)
405 {
406 	__uint16_t	qflags = from->sb_qflags;
407 
408 	to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
409 	if (xfs_sb_version_has_pquotino(from)) {
410 		to->sb_qflags = cpu_to_be16(from->sb_qflags);
411 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
412 		to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
413 		return;
414 	}
415 
416 	/*
417 	 * The in-core version of sb_qflags do not have XFS_OQUOTA_*
418 	 * flags, whereas the on-disk version does.  So, convert incore
419 	 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
420 	 */
421 	qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
422 			XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
423 
424 	if (from->sb_qflags &
425 			(XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
426 		qflags |= XFS_OQUOTA_ENFD;
427 	if (from->sb_qflags &
428 			(XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
429 		qflags |= XFS_OQUOTA_CHKD;
430 	to->sb_qflags = cpu_to_be16(qflags);
431 
432 	/*
433 	 * GQUOTINO and PQUOTINO cannot be used together in versions
434 	 * of superblock that do not have pquotino. from->sb_flags
435 	 * tells us which quota is active and should be copied to
436 	 * disk. If neither are active, we should NULL the inode.
437 	 *
438 	 * In all cases, the separate pquotino must remain 0 because it
439 	 * it beyond the "end" of the valid non-pquotino superblock.
440 	 */
441 	if (from->sb_qflags & XFS_GQUOTA_ACCT)
442 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
443 	else if (from->sb_qflags & XFS_PQUOTA_ACCT)
444 		to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
445 	else {
446 		/*
447 		 * We can't rely on just the fields being logged to tell us
448 		 * that it is safe to write NULLFSINO - we should only do that
449 		 * if quotas are not actually enabled. Hence only write
450 		 * NULLFSINO if both in-core quota inodes are NULL.
451 		 */
452 		if (from->sb_gquotino == NULLFSINO &&
453 		    from->sb_pquotino == NULLFSINO)
454 			to->sb_gquotino = cpu_to_be64(NULLFSINO);
455 	}
456 
457 	to->sb_pquotino = 0;
458 }
459 
460 void
461 xfs_sb_to_disk(
462 	struct xfs_dsb	*to,
463 	struct xfs_sb	*from)
464 {
465 	xfs_sb_quota_to_disk(to, from);
466 
467 	to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
468 	to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
469 	to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
470 	to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
471 	to->sb_rextents = cpu_to_be64(from->sb_rextents);
472 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
473 	to->sb_logstart = cpu_to_be64(from->sb_logstart);
474 	to->sb_rootino = cpu_to_be64(from->sb_rootino);
475 	to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
476 	to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
477 	to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
478 	to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
479 	to->sb_agcount = cpu_to_be32(from->sb_agcount);
480 	to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
481 	to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
482 	to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
483 	to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
484 	to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
485 	to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
486 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
487 	to->sb_blocklog = from->sb_blocklog;
488 	to->sb_sectlog = from->sb_sectlog;
489 	to->sb_inodelog = from->sb_inodelog;
490 	to->sb_inopblog = from->sb_inopblog;
491 	to->sb_agblklog = from->sb_agblklog;
492 	to->sb_rextslog = from->sb_rextslog;
493 	to->sb_inprogress = from->sb_inprogress;
494 	to->sb_imax_pct = from->sb_imax_pct;
495 	to->sb_icount = cpu_to_be64(from->sb_icount);
496 	to->sb_ifree = cpu_to_be64(from->sb_ifree);
497 	to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
498 	to->sb_frextents = cpu_to_be64(from->sb_frextents);
499 
500 	to->sb_flags = from->sb_flags;
501 	to->sb_shared_vn = from->sb_shared_vn;
502 	to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
503 	to->sb_unit = cpu_to_be32(from->sb_unit);
504 	to->sb_width = cpu_to_be32(from->sb_width);
505 	to->sb_dirblklog = from->sb_dirblklog;
506 	to->sb_logsectlog = from->sb_logsectlog;
507 	to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
508 	to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
509 
510 	/*
511 	 * We need to ensure that bad_features2 always matches features2.
512 	 * Hence we enforce that here rather than having to remember to do it
513 	 * everywhere else that updates features2.
514 	 */
515 	from->sb_bad_features2 = from->sb_features2;
516 	to->sb_features2 = cpu_to_be32(from->sb_features2);
517 	to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
518 
519 	if (xfs_sb_version_hascrc(from)) {
520 		to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
521 		to->sb_features_ro_compat =
522 				cpu_to_be32(from->sb_features_ro_compat);
523 		to->sb_features_incompat =
524 				cpu_to_be32(from->sb_features_incompat);
525 		to->sb_features_log_incompat =
526 				cpu_to_be32(from->sb_features_log_incompat);
527 		to->sb_pad = 0;
528 		to->sb_lsn = cpu_to_be64(from->sb_lsn);
529 	}
530 }
531 
532 static int
533 xfs_sb_verify(
534 	struct xfs_buf	*bp,
535 	bool		check_version)
536 {
537 	struct xfs_mount *mp = bp->b_target->bt_mount;
538 	struct xfs_sb	sb;
539 
540 	/*
541 	 * Use call variant which doesn't convert quota flags from disk
542 	 * format, because xfs_mount_validate_sb checks the on-disk flags.
543 	 */
544 	__xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
545 
546 	/*
547 	 * Only check the in progress field for the primary superblock as
548 	 * mkfs.xfs doesn't clear it from secondary superblocks.
549 	 */
550 	return xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR,
551 				     check_version);
552 }
553 
554 /*
555  * If the superblock has the CRC feature bit set or the CRC field is non-null,
556  * check that the CRC is valid.  We check the CRC field is non-null because a
557  * single bit error could clear the feature bit and unused parts of the
558  * superblock are supposed to be zero. Hence a non-null crc field indicates that
559  * we've potentially lost a feature bit and we should check it anyway.
560  *
561  * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
562  * last field in V4 secondary superblocks.  So for secondary superblocks,
563  * we are more forgiving, and ignore CRC failures if the primary doesn't
564  * indicate that the fs version is V5.
565  */
566 static void
567 xfs_sb_read_verify(
568 	struct xfs_buf	*bp)
569 {
570 	struct xfs_mount *mp = bp->b_target->bt_mount;
571 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
572 	int		error;
573 
574 	/*
575 	 * open code the version check to avoid needing to convert the entire
576 	 * superblock from disk order just to check the version number
577 	 */
578 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
579 	    (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
580 						XFS_SB_VERSION_5) ||
581 	     dsb->sb_crc != 0)) {
582 
583 		if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
584 			/* Only fail bad secondaries on a known V5 filesystem */
585 			if (bp->b_bn == XFS_SB_DADDR ||
586 			    xfs_sb_version_hascrc(&mp->m_sb)) {
587 				error = -EFSBADCRC;
588 				goto out_error;
589 			}
590 		}
591 	}
592 	error = xfs_sb_verify(bp, true);
593 
594 out_error:
595 	if (error) {
596 		xfs_buf_ioerror(bp, error);
597 		if (error == -EFSCORRUPTED || error == -EFSBADCRC)
598 			xfs_verifier_error(bp);
599 	}
600 }
601 
602 /*
603  * We may be probed for a filesystem match, so we may not want to emit
604  * messages when the superblock buffer is not actually an XFS superblock.
605  * If we find an XFS superblock, then run a normal, noisy mount because we are
606  * really going to mount it and want to know about errors.
607  */
608 static void
609 xfs_sb_quiet_read_verify(
610 	struct xfs_buf	*bp)
611 {
612 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
613 
614 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
615 		/* XFS filesystem, verify noisily! */
616 		xfs_sb_read_verify(bp);
617 		return;
618 	}
619 	/* quietly fail */
620 	xfs_buf_ioerror(bp, -EWRONGFS);
621 }
622 
623 static void
624 xfs_sb_write_verify(
625 	struct xfs_buf		*bp)
626 {
627 	struct xfs_mount	*mp = bp->b_target->bt_mount;
628 	struct xfs_buf_log_item	*bip = bp->b_fspriv;
629 	int			error;
630 
631 	error = xfs_sb_verify(bp, false);
632 	if (error) {
633 		xfs_buf_ioerror(bp, error);
634 		xfs_verifier_error(bp);
635 		return;
636 	}
637 
638 	if (!xfs_sb_version_hascrc(&mp->m_sb))
639 		return;
640 
641 	if (bip)
642 		XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
643 
644 	xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
645 }
646 
647 const struct xfs_buf_ops xfs_sb_buf_ops = {
648 	.verify_read = xfs_sb_read_verify,
649 	.verify_write = xfs_sb_write_verify,
650 };
651 
652 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
653 	.verify_read = xfs_sb_quiet_read_verify,
654 	.verify_write = xfs_sb_write_verify,
655 };
656 
657 /*
658  * xfs_mount_common
659  *
660  * Mount initialization code establishing various mount
661  * fields from the superblock associated with the given
662  * mount structure
663  */
664 void
665 xfs_sb_mount_common(
666 	struct xfs_mount *mp,
667 	struct xfs_sb	*sbp)
668 {
669 	mp->m_agfrotor = mp->m_agirotor = 0;
670 	spin_lock_init(&mp->m_agirotor_lock);
671 	mp->m_maxagi = mp->m_sb.sb_agcount;
672 	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
673 	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
674 	mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
675 	mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
676 	mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
677 	mp->m_blockmask = sbp->sb_blocksize - 1;
678 	mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
679 	mp->m_blockwmask = mp->m_blockwsize - 1;
680 
681 	mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
682 	mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
683 	mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
684 	mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
685 
686 	mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1);
687 	mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0);
688 	mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2;
689 	mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2;
690 
691 	mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
692 	mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
693 	mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
694 	mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
695 
696 	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
697 	mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
698 					sbp->sb_inopblock);
699 	mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
700 }
701 
702 /*
703  * xfs_initialize_perag_data
704  *
705  * Read in each per-ag structure so we can count up the number of
706  * allocated inodes, free inodes and used filesystem blocks as this
707  * information is no longer persistent in the superblock. Once we have
708  * this information, write it into the in-core superblock structure.
709  */
710 int
711 xfs_initialize_perag_data(
712 	struct xfs_mount *mp,
713 	xfs_agnumber_t	agcount)
714 {
715 	xfs_agnumber_t	index;
716 	xfs_perag_t	*pag;
717 	xfs_sb_t	*sbp = &mp->m_sb;
718 	uint64_t	ifree = 0;
719 	uint64_t	ialloc = 0;
720 	uint64_t	bfree = 0;
721 	uint64_t	bfreelst = 0;
722 	uint64_t	btree = 0;
723 	int		error;
724 
725 	for (index = 0; index < agcount; index++) {
726 		/*
727 		 * read the agf, then the agi. This gets us
728 		 * all the information we need and populates the
729 		 * per-ag structures for us.
730 		 */
731 		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
732 		if (error)
733 			return error;
734 
735 		error = xfs_ialloc_pagi_init(mp, NULL, index);
736 		if (error)
737 			return error;
738 		pag = xfs_perag_get(mp, index);
739 		ifree += pag->pagi_freecount;
740 		ialloc += pag->pagi_count;
741 		bfree += pag->pagf_freeblks;
742 		bfreelst += pag->pagf_flcount;
743 		btree += pag->pagf_btreeblks;
744 		xfs_perag_put(pag);
745 	}
746 	/*
747 	 * Overwrite incore superblock counters with just-read data
748 	 */
749 	spin_lock(&mp->m_sb_lock);
750 	sbp->sb_ifree = ifree;
751 	sbp->sb_icount = ialloc;
752 	sbp->sb_fdblocks = bfree + bfreelst + btree;
753 	spin_unlock(&mp->m_sb_lock);
754 
755 	/* Fixup the per-cpu counters as well. */
756 	xfs_icsb_reinit_counters(mp);
757 
758 	return 0;
759 }
760 
761 /*
762  * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
763  * into the superblock buffer to be logged.  It does not provide the higher
764  * level of locking that is needed to protect the in-core superblock from
765  * concurrent access.
766  */
767 void
768 xfs_log_sb(
769 	struct xfs_trans	*tp)
770 {
771 	struct xfs_mount	*mp = tp->t_mountp;
772 	struct xfs_buf		*bp = xfs_trans_getsb(tp, mp, 0);
773 
774 	xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
775 	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
776 	xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb));
777 }
778 
779 /*
780  * xfs_sync_sb
781  *
782  * Sync the superblock to disk.
783  *
784  * Note that the caller is responsible for checking the frozen state of the
785  * filesystem. This procedure uses the non-blocking transaction allocator and
786  * thus will allow modifications to a frozen fs. This is required because this
787  * code can be called during the process of freezing where use of the high-level
788  * allocator would deadlock.
789  */
790 int
791 xfs_sync_sb(
792 	struct xfs_mount	*mp,
793 	bool			wait)
794 {
795 	struct xfs_trans	*tp;
796 	int			error;
797 
798 	tp = _xfs_trans_alloc(mp, XFS_TRANS_SB_CHANGE, KM_SLEEP);
799 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0);
800 	if (error) {
801 		xfs_trans_cancel(tp, 0);
802 		return error;
803 	}
804 
805 	xfs_log_sb(tp);
806 	if (wait)
807 		xfs_trans_set_sync(tp);
808 	return xfs_trans_commit(tp, 0);
809 }
810