xref: /linux/fs/xfs/libxfs/xfs_sb.c (revision 827634added7f38b7d724cab1dccdb2b004c13c3)
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 	if (sbp->sb_magicnum != XFS_SB_MAGIC) {
115 		xfs_warn(mp, "bad magic number");
116 		return -EWRONGFS;
117 	}
118 
119 
120 	if (!xfs_sb_good_version(sbp)) {
121 		xfs_warn(mp, "bad version");
122 		return -EWRONGFS;
123 	}
124 
125 	/*
126 	 * Version 5 superblock feature mask validation. Reject combinations the
127 	 * kernel cannot support up front before checking anything else. For
128 	 * write validation, we don't need to check feature masks.
129 	 */
130 	if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) {
131 		if (xfs_sb_has_compat_feature(sbp,
132 					XFS_SB_FEAT_COMPAT_UNKNOWN)) {
133 			xfs_warn(mp,
134 "Superblock has unknown compatible features (0x%x) enabled.\n"
135 "Using a more recent kernel is recommended.",
136 				(sbp->sb_features_compat &
137 						XFS_SB_FEAT_COMPAT_UNKNOWN));
138 		}
139 
140 		if (xfs_sb_has_ro_compat_feature(sbp,
141 					XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
142 			xfs_alert(mp,
143 "Superblock has unknown read-only compatible features (0x%x) enabled.",
144 				(sbp->sb_features_ro_compat &
145 						XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
146 			if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
147 				xfs_warn(mp,
148 "Attempted to mount read-only compatible filesystem read-write.\n"
149 "Filesystem can only be safely mounted read only.");
150 				return -EINVAL;
151 			}
152 		}
153 		if (xfs_sb_has_incompat_feature(sbp,
154 					XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
155 			xfs_warn(mp,
156 "Superblock has unknown incompatible features (0x%x) enabled.\n"
157 "Filesystem can not be safely mounted by this kernel.",
158 				(sbp->sb_features_incompat &
159 						XFS_SB_FEAT_INCOMPAT_UNKNOWN));
160 			return -EINVAL;
161 		}
162 	}
163 
164 	if (xfs_sb_version_has_pquotino(sbp)) {
165 		if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
166 			xfs_notice(mp,
167 			   "Version 5 of Super block has XFS_OQUOTA bits.");
168 			return -EFSCORRUPTED;
169 		}
170 	} else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
171 				XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
172 			xfs_notice(mp,
173 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
174 			return -EFSCORRUPTED;
175 	}
176 
177 	if (unlikely(
178 	    sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
179 		xfs_warn(mp,
180 		"filesystem is marked as having an external log; "
181 		"specify logdev on the mount command line.");
182 		return -EINVAL;
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 internal log; "
189 		"do not specify logdev on the mount command line.");
190 		return -EINVAL;
191 	}
192 
193 	/*
194 	 * More sanity checking.  Most of these were stolen directly from
195 	 * xfs_repair.
196 	 */
197 	if (unlikely(
198 	    sbp->sb_agcount <= 0					||
199 	    sbp->sb_sectsize < XFS_MIN_SECTORSIZE			||
200 	    sbp->sb_sectsize > XFS_MAX_SECTORSIZE			||
201 	    sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG			||
202 	    sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG			||
203 	    sbp->sb_sectsize != (1 << sbp->sb_sectlog)			||
204 	    sbp->sb_blocksize < XFS_MIN_BLOCKSIZE			||
205 	    sbp->sb_blocksize > XFS_MAX_BLOCKSIZE			||
206 	    sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG			||
207 	    sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG			||
208 	    sbp->sb_blocksize != (1 << sbp->sb_blocklog)		||
209 	    sbp->sb_dirblklog > XFS_MAX_BLOCKSIZE_LOG			||
210 	    sbp->sb_inodesize < XFS_DINODE_MIN_SIZE			||
211 	    sbp->sb_inodesize > XFS_DINODE_MAX_SIZE			||
212 	    sbp->sb_inodelog < XFS_DINODE_MIN_LOG			||
213 	    sbp->sb_inodelog > XFS_DINODE_MAX_LOG			||
214 	    sbp->sb_inodesize != (1 << sbp->sb_inodelog)		||
215 	    sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE			||
216 	    sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
217 	    (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog)	||
218 	    (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE)	||
219 	    (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)	||
220 	    (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */)	||
221 	    sbp->sb_dblocks == 0					||
222 	    sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp)			||
223 	    sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp)			||
224 	    sbp->sb_shared_vn != 0)) {
225 		xfs_notice(mp, "SB sanity check failed");
226 		return -EFSCORRUPTED;
227 	}
228 
229 	/*
230 	 * Until this is fixed only page-sized or smaller data blocks work.
231 	 */
232 	if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
233 		xfs_warn(mp,
234 		"File system with blocksize %d bytes. "
235 		"Only pagesize (%ld) or less will currently work.",
236 				sbp->sb_blocksize, PAGE_SIZE);
237 		return -ENOSYS;
238 	}
239 
240 	/*
241 	 * Currently only very few inode sizes are supported.
242 	 */
243 	switch (sbp->sb_inodesize) {
244 	case 256:
245 	case 512:
246 	case 1024:
247 	case 2048:
248 		break;
249 	default:
250 		xfs_warn(mp, "inode size of %d bytes not supported",
251 				sbp->sb_inodesize);
252 		return -ENOSYS;
253 	}
254 
255 	if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
256 	    xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
257 		xfs_warn(mp,
258 		"file system too large to be mounted on this system.");
259 		return -EFBIG;
260 	}
261 
262 	if (check_inprogress && sbp->sb_inprogress) {
263 		xfs_warn(mp, "Offline file system operation in progress!");
264 		return -EFSCORRUPTED;
265 	}
266 	return 0;
267 }
268 
269 void
270 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
271 {
272 	/*
273 	 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
274 	 * leads to in-core values having two different values for a quota
275 	 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
276 	 * NULLFSINO.
277 	 *
278 	 * Note that this change affect only the in-core values. These
279 	 * values are not written back to disk unless any quota information
280 	 * is written to the disk. Even in that case, sb_pquotino field is
281 	 * not written to disk unless the superblock supports pquotino.
282 	 */
283 	if (sbp->sb_uquotino == 0)
284 		sbp->sb_uquotino = NULLFSINO;
285 	if (sbp->sb_gquotino == 0)
286 		sbp->sb_gquotino = NULLFSINO;
287 	if (sbp->sb_pquotino == 0)
288 		sbp->sb_pquotino = NULLFSINO;
289 
290 	/*
291 	 * We need to do these manipilations only if we are working
292 	 * with an older version of on-disk superblock.
293 	 */
294 	if (xfs_sb_version_has_pquotino(sbp))
295 		return;
296 
297 	if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
298 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
299 					XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
300 	if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
301 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
302 					XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
303 	sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
304 
305 	if (sbp->sb_qflags & XFS_PQUOTA_ACCT)  {
306 		/*
307 		 * In older version of superblock, on-disk superblock only
308 		 * has sb_gquotino, and in-core superblock has both sb_gquotino
309 		 * and sb_pquotino. But, only one of them is supported at any
310 		 * point of time. So, if PQUOTA is set in disk superblock,
311 		 * copy over sb_gquotino to sb_pquotino.
312 		 */
313 		sbp->sb_pquotino = sbp->sb_gquotino;
314 		sbp->sb_gquotino = NULLFSINO;
315 	}
316 }
317 
318 static void
319 __xfs_sb_from_disk(
320 	struct xfs_sb	*to,
321 	xfs_dsb_t	*from,
322 	bool		convert_xquota)
323 {
324 	to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
325 	to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
326 	to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
327 	to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
328 	to->sb_rextents = be64_to_cpu(from->sb_rextents);
329 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
330 	to->sb_logstart = be64_to_cpu(from->sb_logstart);
331 	to->sb_rootino = be64_to_cpu(from->sb_rootino);
332 	to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
333 	to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
334 	to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
335 	to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
336 	to->sb_agcount = be32_to_cpu(from->sb_agcount);
337 	to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
338 	to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
339 	to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
340 	to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
341 	to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
342 	to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
343 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
344 	to->sb_blocklog = from->sb_blocklog;
345 	to->sb_sectlog = from->sb_sectlog;
346 	to->sb_inodelog = from->sb_inodelog;
347 	to->sb_inopblog = from->sb_inopblog;
348 	to->sb_agblklog = from->sb_agblklog;
349 	to->sb_rextslog = from->sb_rextslog;
350 	to->sb_inprogress = from->sb_inprogress;
351 	to->sb_imax_pct = from->sb_imax_pct;
352 	to->sb_icount = be64_to_cpu(from->sb_icount);
353 	to->sb_ifree = be64_to_cpu(from->sb_ifree);
354 	to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
355 	to->sb_frextents = be64_to_cpu(from->sb_frextents);
356 	to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
357 	to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
358 	to->sb_qflags = be16_to_cpu(from->sb_qflags);
359 	to->sb_flags = from->sb_flags;
360 	to->sb_shared_vn = from->sb_shared_vn;
361 	to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
362 	to->sb_unit = be32_to_cpu(from->sb_unit);
363 	to->sb_width = be32_to_cpu(from->sb_width);
364 	to->sb_dirblklog = from->sb_dirblklog;
365 	to->sb_logsectlog = from->sb_logsectlog;
366 	to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
367 	to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
368 	to->sb_features2 = be32_to_cpu(from->sb_features2);
369 	to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
370 	to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
371 	to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
372 	to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
373 	to->sb_features_log_incompat =
374 				be32_to_cpu(from->sb_features_log_incompat);
375 	/* crc is only used on disk, not in memory; just init to 0 here. */
376 	to->sb_crc = 0;
377 	to->sb_pad = 0;
378 	to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
379 	to->sb_lsn = be64_to_cpu(from->sb_lsn);
380 	/* Convert on-disk flags to in-memory flags? */
381 	if (convert_xquota)
382 		xfs_sb_quota_from_disk(to);
383 }
384 
385 void
386 xfs_sb_from_disk(
387 	struct xfs_sb	*to,
388 	xfs_dsb_t	*from)
389 {
390 	__xfs_sb_from_disk(to, from, true);
391 }
392 
393 static void
394 xfs_sb_quota_to_disk(
395 	struct xfs_dsb	*to,
396 	struct xfs_sb	*from)
397 {
398 	__uint16_t	qflags = from->sb_qflags;
399 
400 	to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
401 	if (xfs_sb_version_has_pquotino(from)) {
402 		to->sb_qflags = cpu_to_be16(from->sb_qflags);
403 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
404 		to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
405 		return;
406 	}
407 
408 	/*
409 	 * The in-core version of sb_qflags do not have XFS_OQUOTA_*
410 	 * flags, whereas the on-disk version does.  So, convert incore
411 	 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
412 	 */
413 	qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
414 			XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
415 
416 	if (from->sb_qflags &
417 			(XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
418 		qflags |= XFS_OQUOTA_ENFD;
419 	if (from->sb_qflags &
420 			(XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
421 		qflags |= XFS_OQUOTA_CHKD;
422 	to->sb_qflags = cpu_to_be16(qflags);
423 
424 	/*
425 	 * GQUOTINO and PQUOTINO cannot be used together in versions
426 	 * of superblock that do not have pquotino. from->sb_flags
427 	 * tells us which quota is active and should be copied to
428 	 * disk. If neither are active, we should NULL the inode.
429 	 *
430 	 * In all cases, the separate pquotino must remain 0 because it
431 	 * it beyond the "end" of the valid non-pquotino superblock.
432 	 */
433 	if (from->sb_qflags & XFS_GQUOTA_ACCT)
434 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
435 	else if (from->sb_qflags & XFS_PQUOTA_ACCT)
436 		to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
437 	else {
438 		/*
439 		 * We can't rely on just the fields being logged to tell us
440 		 * that it is safe to write NULLFSINO - we should only do that
441 		 * if quotas are not actually enabled. Hence only write
442 		 * NULLFSINO if both in-core quota inodes are NULL.
443 		 */
444 		if (from->sb_gquotino == NULLFSINO &&
445 		    from->sb_pquotino == NULLFSINO)
446 			to->sb_gquotino = cpu_to_be64(NULLFSINO);
447 	}
448 
449 	to->sb_pquotino = 0;
450 }
451 
452 void
453 xfs_sb_to_disk(
454 	struct xfs_dsb	*to,
455 	struct xfs_sb	*from)
456 {
457 	xfs_sb_quota_to_disk(to, from);
458 
459 	to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
460 	to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
461 	to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
462 	to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
463 	to->sb_rextents = cpu_to_be64(from->sb_rextents);
464 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
465 	to->sb_logstart = cpu_to_be64(from->sb_logstart);
466 	to->sb_rootino = cpu_to_be64(from->sb_rootino);
467 	to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
468 	to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
469 	to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
470 	to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
471 	to->sb_agcount = cpu_to_be32(from->sb_agcount);
472 	to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
473 	to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
474 	to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
475 	to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
476 	to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
477 	to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
478 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
479 	to->sb_blocklog = from->sb_blocklog;
480 	to->sb_sectlog = from->sb_sectlog;
481 	to->sb_inodelog = from->sb_inodelog;
482 	to->sb_inopblog = from->sb_inopblog;
483 	to->sb_agblklog = from->sb_agblklog;
484 	to->sb_rextslog = from->sb_rextslog;
485 	to->sb_inprogress = from->sb_inprogress;
486 	to->sb_imax_pct = from->sb_imax_pct;
487 	to->sb_icount = cpu_to_be64(from->sb_icount);
488 	to->sb_ifree = cpu_to_be64(from->sb_ifree);
489 	to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
490 	to->sb_frextents = cpu_to_be64(from->sb_frextents);
491 
492 	to->sb_flags = from->sb_flags;
493 	to->sb_shared_vn = from->sb_shared_vn;
494 	to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
495 	to->sb_unit = cpu_to_be32(from->sb_unit);
496 	to->sb_width = cpu_to_be32(from->sb_width);
497 	to->sb_dirblklog = from->sb_dirblklog;
498 	to->sb_logsectlog = from->sb_logsectlog;
499 	to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
500 	to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
501 
502 	/*
503 	 * We need to ensure that bad_features2 always matches features2.
504 	 * Hence we enforce that here rather than having to remember to do it
505 	 * everywhere else that updates features2.
506 	 */
507 	from->sb_bad_features2 = from->sb_features2;
508 	to->sb_features2 = cpu_to_be32(from->sb_features2);
509 	to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
510 
511 	if (xfs_sb_version_hascrc(from)) {
512 		to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
513 		to->sb_features_ro_compat =
514 				cpu_to_be32(from->sb_features_ro_compat);
515 		to->sb_features_incompat =
516 				cpu_to_be32(from->sb_features_incompat);
517 		to->sb_features_log_incompat =
518 				cpu_to_be32(from->sb_features_log_incompat);
519 		to->sb_pad = 0;
520 		to->sb_lsn = cpu_to_be64(from->sb_lsn);
521 	}
522 }
523 
524 static int
525 xfs_sb_verify(
526 	struct xfs_buf	*bp,
527 	bool		check_version)
528 {
529 	struct xfs_mount *mp = bp->b_target->bt_mount;
530 	struct xfs_sb	sb;
531 
532 	/*
533 	 * Use call variant which doesn't convert quota flags from disk
534 	 * format, because xfs_mount_validate_sb checks the on-disk flags.
535 	 */
536 	__xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
537 
538 	/*
539 	 * Only check the in progress field for the primary superblock as
540 	 * mkfs.xfs doesn't clear it from secondary superblocks.
541 	 */
542 	return xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR,
543 				     check_version);
544 }
545 
546 /*
547  * If the superblock has the CRC feature bit set or the CRC field is non-null,
548  * check that the CRC is valid.  We check the CRC field is non-null because a
549  * single bit error could clear the feature bit and unused parts of the
550  * superblock are supposed to be zero. Hence a non-null crc field indicates that
551  * we've potentially lost a feature bit and we should check it anyway.
552  *
553  * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
554  * last field in V4 secondary superblocks.  So for secondary superblocks,
555  * we are more forgiving, and ignore CRC failures if the primary doesn't
556  * indicate that the fs version is V5.
557  */
558 static void
559 xfs_sb_read_verify(
560 	struct xfs_buf	*bp)
561 {
562 	struct xfs_mount *mp = bp->b_target->bt_mount;
563 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
564 	int		error;
565 
566 	/*
567 	 * open code the version check to avoid needing to convert the entire
568 	 * superblock from disk order just to check the version number
569 	 */
570 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
571 	    (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
572 						XFS_SB_VERSION_5) ||
573 	     dsb->sb_crc != 0)) {
574 
575 		if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
576 			/* Only fail bad secondaries on a known V5 filesystem */
577 			if (bp->b_bn == XFS_SB_DADDR ||
578 			    xfs_sb_version_hascrc(&mp->m_sb)) {
579 				error = -EFSBADCRC;
580 				goto out_error;
581 			}
582 		}
583 	}
584 	error = xfs_sb_verify(bp, true);
585 
586 out_error:
587 	if (error) {
588 		xfs_buf_ioerror(bp, error);
589 		if (error == -EFSCORRUPTED || error == -EFSBADCRC)
590 			xfs_verifier_error(bp);
591 	}
592 }
593 
594 /*
595  * We may be probed for a filesystem match, so we may not want to emit
596  * messages when the superblock buffer is not actually an XFS superblock.
597  * If we find an XFS superblock, then run a normal, noisy mount because we are
598  * really going to mount it and want to know about errors.
599  */
600 static void
601 xfs_sb_quiet_read_verify(
602 	struct xfs_buf	*bp)
603 {
604 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
605 
606 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
607 		/* XFS filesystem, verify noisily! */
608 		xfs_sb_read_verify(bp);
609 		return;
610 	}
611 	/* quietly fail */
612 	xfs_buf_ioerror(bp, -EWRONGFS);
613 }
614 
615 static void
616 xfs_sb_write_verify(
617 	struct xfs_buf		*bp)
618 {
619 	struct xfs_mount	*mp = bp->b_target->bt_mount;
620 	struct xfs_buf_log_item	*bip = bp->b_fspriv;
621 	int			error;
622 
623 	error = xfs_sb_verify(bp, false);
624 	if (error) {
625 		xfs_buf_ioerror(bp, error);
626 		xfs_verifier_error(bp);
627 		return;
628 	}
629 
630 	if (!xfs_sb_version_hascrc(&mp->m_sb))
631 		return;
632 
633 	if (bip)
634 		XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
635 
636 	xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
637 }
638 
639 const struct xfs_buf_ops xfs_sb_buf_ops = {
640 	.verify_read = xfs_sb_read_verify,
641 	.verify_write = xfs_sb_write_verify,
642 };
643 
644 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
645 	.verify_read = xfs_sb_quiet_read_verify,
646 	.verify_write = xfs_sb_write_verify,
647 };
648 
649 /*
650  * xfs_mount_common
651  *
652  * Mount initialization code establishing various mount
653  * fields from the superblock associated with the given
654  * mount structure
655  */
656 void
657 xfs_sb_mount_common(
658 	struct xfs_mount *mp,
659 	struct xfs_sb	*sbp)
660 {
661 	mp->m_agfrotor = mp->m_agirotor = 0;
662 	spin_lock_init(&mp->m_agirotor_lock);
663 	mp->m_maxagi = mp->m_sb.sb_agcount;
664 	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
665 	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
666 	mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
667 	mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
668 	mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
669 	mp->m_blockmask = sbp->sb_blocksize - 1;
670 	mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
671 	mp->m_blockwmask = mp->m_blockwsize - 1;
672 
673 	mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
674 	mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
675 	mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
676 	mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
677 
678 	mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1);
679 	mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0);
680 	mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2;
681 	mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2;
682 
683 	mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
684 	mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
685 	mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
686 	mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
687 
688 	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
689 	mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
690 					sbp->sb_inopblock);
691 	mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
692 }
693 
694 /*
695  * xfs_initialize_perag_data
696  *
697  * Read in each per-ag structure so we can count up the number of
698  * allocated inodes, free inodes and used filesystem blocks as this
699  * information is no longer persistent in the superblock. Once we have
700  * this information, write it into the in-core superblock structure.
701  */
702 int
703 xfs_initialize_perag_data(
704 	struct xfs_mount *mp,
705 	xfs_agnumber_t	agcount)
706 {
707 	xfs_agnumber_t	index;
708 	xfs_perag_t	*pag;
709 	xfs_sb_t	*sbp = &mp->m_sb;
710 	uint64_t	ifree = 0;
711 	uint64_t	ialloc = 0;
712 	uint64_t	bfree = 0;
713 	uint64_t	bfreelst = 0;
714 	uint64_t	btree = 0;
715 	int		error;
716 
717 	for (index = 0; index < agcount; index++) {
718 		/*
719 		 * read the agf, then the agi. This gets us
720 		 * all the information we need and populates the
721 		 * per-ag structures for us.
722 		 */
723 		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
724 		if (error)
725 			return error;
726 
727 		error = xfs_ialloc_pagi_init(mp, NULL, index);
728 		if (error)
729 			return error;
730 		pag = xfs_perag_get(mp, index);
731 		ifree += pag->pagi_freecount;
732 		ialloc += pag->pagi_count;
733 		bfree += pag->pagf_freeblks;
734 		bfreelst += pag->pagf_flcount;
735 		btree += pag->pagf_btreeblks;
736 		xfs_perag_put(pag);
737 	}
738 
739 	/* Overwrite incore superblock counters with just-read data */
740 	spin_lock(&mp->m_sb_lock);
741 	sbp->sb_ifree = ifree;
742 	sbp->sb_icount = ialloc;
743 	sbp->sb_fdblocks = bfree + bfreelst + btree;
744 	spin_unlock(&mp->m_sb_lock);
745 
746 	xfs_reinit_percpu_counters(mp);
747 
748 	return 0;
749 }
750 
751 /*
752  * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
753  * into the superblock buffer to be logged.  It does not provide the higher
754  * level of locking that is needed to protect the in-core superblock from
755  * concurrent access.
756  */
757 void
758 xfs_log_sb(
759 	struct xfs_trans	*tp)
760 {
761 	struct xfs_mount	*mp = tp->t_mountp;
762 	struct xfs_buf		*bp = xfs_trans_getsb(tp, mp, 0);
763 
764 	mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount);
765 	mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree);
766 	mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks);
767 
768 	xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
769 	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
770 	xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb));
771 }
772 
773 /*
774  * xfs_sync_sb
775  *
776  * Sync the superblock to disk.
777  *
778  * Note that the caller is responsible for checking the frozen state of the
779  * filesystem. This procedure uses the non-blocking transaction allocator and
780  * thus will allow modifications to a frozen fs. This is required because this
781  * code can be called during the process of freezing where use of the high-level
782  * allocator would deadlock.
783  */
784 int
785 xfs_sync_sb(
786 	struct xfs_mount	*mp,
787 	bool			wait)
788 {
789 	struct xfs_trans	*tp;
790 	int			error;
791 
792 	tp = _xfs_trans_alloc(mp, XFS_TRANS_SB_CHANGE, KM_SLEEP);
793 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0);
794 	if (error) {
795 		xfs_trans_cancel(tp, 0);
796 		return error;
797 	}
798 
799 	xfs_log_sb(tp);
800 	if (wait)
801 		xfs_trans_set_sync(tp);
802 	return xfs_trans_commit(tp, 0);
803 }
804