xref: /openbmc/linux/fs/xfs/libxfs/xfs_sb.c (revision 799a545b)
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_defer.h"
28 #include "xfs_inode.h"
29 #include "xfs_ialloc.h"
30 #include "xfs_alloc.h"
31 #include "xfs_error.h"
32 #include "xfs_trace.h"
33 #include "xfs_cksum.h"
34 #include "xfs_trans.h"
35 #include "xfs_buf_item.h"
36 #include "xfs_bmap_btree.h"
37 #include "xfs_alloc_btree.h"
38 #include "xfs_ialloc_btree.h"
39 #include "xfs_log.h"
40 #include "xfs_rmap_btree.h"
41 
42 /*
43  * Physical superblock buffer manipulations. Shared with libxfs in userspace.
44  */
45 
46 /*
47  * Reference counting access wrappers to the perag structures.
48  * Because we never free per-ag structures, the only thing we
49  * have to protect against changes is the tree structure itself.
50  */
51 struct xfs_perag *
52 xfs_perag_get(
53 	struct xfs_mount	*mp,
54 	xfs_agnumber_t		agno)
55 {
56 	struct xfs_perag	*pag;
57 	int			ref = 0;
58 
59 	rcu_read_lock();
60 	pag = radix_tree_lookup(&mp->m_perag_tree, agno);
61 	if (pag) {
62 		ASSERT(atomic_read(&pag->pag_ref) >= 0);
63 		ref = atomic_inc_return(&pag->pag_ref);
64 	}
65 	rcu_read_unlock();
66 	trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
67 	return pag;
68 }
69 
70 /*
71  * search from @first to find the next perag with the given tag set.
72  */
73 struct xfs_perag *
74 xfs_perag_get_tag(
75 	struct xfs_mount	*mp,
76 	xfs_agnumber_t		first,
77 	int			tag)
78 {
79 	struct xfs_perag	*pag;
80 	int			found;
81 	int			ref;
82 
83 	rcu_read_lock();
84 	found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
85 					(void **)&pag, first, 1, tag);
86 	if (found <= 0) {
87 		rcu_read_unlock();
88 		return NULL;
89 	}
90 	ref = atomic_inc_return(&pag->pag_ref);
91 	rcu_read_unlock();
92 	trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
93 	return pag;
94 }
95 
96 void
97 xfs_perag_put(
98 	struct xfs_perag	*pag)
99 {
100 	int	ref;
101 
102 	ASSERT(atomic_read(&pag->pag_ref) > 0);
103 	ref = atomic_dec_return(&pag->pag_ref);
104 	trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
105 }
106 
107 /*
108  * Check the validity of the SB found.
109  */
110 STATIC int
111 xfs_mount_validate_sb(
112 	xfs_mount_t	*mp,
113 	xfs_sb_t	*sbp,
114 	bool		check_inprogress,
115 	bool		check_version)
116 {
117 	if (sbp->sb_magicnum != XFS_SB_MAGIC) {
118 		xfs_warn(mp, "bad magic number");
119 		return -EWRONGFS;
120 	}
121 
122 
123 	if (!xfs_sb_good_version(sbp)) {
124 		xfs_warn(mp, "bad version");
125 		return -EWRONGFS;
126 	}
127 
128 	/*
129 	 * Version 5 superblock feature mask validation. Reject combinations the
130 	 * kernel cannot support up front before checking anything else. For
131 	 * write validation, we don't need to check feature masks.
132 	 */
133 	if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) {
134 		if (xfs_sb_has_compat_feature(sbp,
135 					XFS_SB_FEAT_COMPAT_UNKNOWN)) {
136 			xfs_warn(mp,
137 "Superblock has unknown compatible features (0x%x) enabled.",
138 				(sbp->sb_features_compat &
139 						XFS_SB_FEAT_COMPAT_UNKNOWN));
140 			xfs_warn(mp,
141 "Using a more recent kernel is recommended.");
142 		}
143 
144 		if (xfs_sb_has_ro_compat_feature(sbp,
145 					XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
146 			xfs_alert(mp,
147 "Superblock has unknown read-only compatible features (0x%x) enabled.",
148 				(sbp->sb_features_ro_compat &
149 						XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
150 			if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
151 				xfs_warn(mp,
152 "Attempted to mount read-only compatible filesystem read-write.");
153 				xfs_warn(mp,
154 "Filesystem can only be safely mounted read only.");
155 
156 				return -EINVAL;
157 			}
158 		}
159 		if (xfs_sb_has_incompat_feature(sbp,
160 					XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
161 			xfs_warn(mp,
162 "Superblock has unknown incompatible features (0x%x) enabled.",
163 				(sbp->sb_features_incompat &
164 						XFS_SB_FEAT_INCOMPAT_UNKNOWN));
165 			xfs_warn(mp,
166 "Filesystem can not be safely mounted by this kernel.");
167 			return -EINVAL;
168 		}
169 	} else if (xfs_sb_version_hascrc(sbp)) {
170 		/*
171 		 * We can't read verify the sb LSN because the read verifier is
172 		 * called before the log is allocated and processed. We know the
173 		 * log is set up before write verifier (!check_version) calls,
174 		 * so just check it here.
175 		 */
176 		if (!xfs_log_check_lsn(mp, sbp->sb_lsn))
177 			return -EFSCORRUPTED;
178 	}
179 
180 	if (xfs_sb_version_has_pquotino(sbp)) {
181 		if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
182 			xfs_notice(mp,
183 			   "Version 5 of Super block has XFS_OQUOTA bits.");
184 			return -EFSCORRUPTED;
185 		}
186 	} else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
187 				XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
188 			xfs_notice(mp,
189 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
190 			return -EFSCORRUPTED;
191 	}
192 
193 	/*
194 	 * Full inode chunks must be aligned to inode chunk size when
195 	 * sparse inodes are enabled to support the sparse chunk
196 	 * allocation algorithm and prevent overlapping inode records.
197 	 */
198 	if (xfs_sb_version_hassparseinodes(sbp)) {
199 		uint32_t	align;
200 
201 		align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
202 				>> sbp->sb_blocklog;
203 		if (sbp->sb_inoalignmt != align) {
204 			xfs_warn(mp,
205 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
206 				 sbp->sb_inoalignmt, align);
207 			return -EINVAL;
208 		}
209 	}
210 
211 	if (unlikely(
212 	    sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
213 		xfs_warn(mp,
214 		"filesystem is marked as having an external log; "
215 		"specify logdev on the mount command line.");
216 		return -EINVAL;
217 	}
218 
219 	if (unlikely(
220 	    sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
221 		xfs_warn(mp,
222 		"filesystem is marked as having an internal log; "
223 		"do not specify logdev on the mount command line.");
224 		return -EINVAL;
225 	}
226 
227 	/*
228 	 * More sanity checking.  Most of these were stolen directly from
229 	 * xfs_repair.
230 	 */
231 	if (unlikely(
232 	    sbp->sb_agcount <= 0					||
233 	    sbp->sb_sectsize < XFS_MIN_SECTORSIZE			||
234 	    sbp->sb_sectsize > XFS_MAX_SECTORSIZE			||
235 	    sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG			||
236 	    sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG			||
237 	    sbp->sb_sectsize != (1 << sbp->sb_sectlog)			||
238 	    sbp->sb_blocksize < XFS_MIN_BLOCKSIZE			||
239 	    sbp->sb_blocksize > XFS_MAX_BLOCKSIZE			||
240 	    sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG			||
241 	    sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG			||
242 	    sbp->sb_blocksize != (1 << sbp->sb_blocklog)		||
243 	    sbp->sb_dirblklog > XFS_MAX_BLOCKSIZE_LOG			||
244 	    sbp->sb_inodesize < XFS_DINODE_MIN_SIZE			||
245 	    sbp->sb_inodesize > XFS_DINODE_MAX_SIZE			||
246 	    sbp->sb_inodelog < XFS_DINODE_MIN_LOG			||
247 	    sbp->sb_inodelog > XFS_DINODE_MAX_LOG			||
248 	    sbp->sb_inodesize != (1 << sbp->sb_inodelog)		||
249 	    sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE			||
250 	    sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
251 	    (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog)	||
252 	    (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE)	||
253 	    (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)	||
254 	    (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */)	||
255 	    sbp->sb_dblocks == 0					||
256 	    sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp)			||
257 	    sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp)			||
258 	    sbp->sb_shared_vn != 0)) {
259 		xfs_notice(mp, "SB sanity check failed");
260 		return -EFSCORRUPTED;
261 	}
262 
263 	/*
264 	 * Until this is fixed only page-sized or smaller data blocks work.
265 	 */
266 	if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
267 		xfs_warn(mp,
268 		"File system with blocksize %d bytes. "
269 		"Only pagesize (%ld) or less will currently work.",
270 				sbp->sb_blocksize, PAGE_SIZE);
271 		return -ENOSYS;
272 	}
273 
274 	/*
275 	 * Currently only very few inode sizes are supported.
276 	 */
277 	switch (sbp->sb_inodesize) {
278 	case 256:
279 	case 512:
280 	case 1024:
281 	case 2048:
282 		break;
283 	default:
284 		xfs_warn(mp, "inode size of %d bytes not supported",
285 				sbp->sb_inodesize);
286 		return -ENOSYS;
287 	}
288 
289 	if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
290 	    xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
291 		xfs_warn(mp,
292 		"file system too large to be mounted on this system.");
293 		return -EFBIG;
294 	}
295 
296 	if (check_inprogress && sbp->sb_inprogress) {
297 		xfs_warn(mp, "Offline file system operation in progress!");
298 		return -EFSCORRUPTED;
299 	}
300 	return 0;
301 }
302 
303 void
304 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
305 {
306 	/*
307 	 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
308 	 * leads to in-core values having two different values for a quota
309 	 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
310 	 * NULLFSINO.
311 	 *
312 	 * Note that this change affect only the in-core values. These
313 	 * values are not written back to disk unless any quota information
314 	 * is written to the disk. Even in that case, sb_pquotino field is
315 	 * not written to disk unless the superblock supports pquotino.
316 	 */
317 	if (sbp->sb_uquotino == 0)
318 		sbp->sb_uquotino = NULLFSINO;
319 	if (sbp->sb_gquotino == 0)
320 		sbp->sb_gquotino = NULLFSINO;
321 	if (sbp->sb_pquotino == 0)
322 		sbp->sb_pquotino = NULLFSINO;
323 
324 	/*
325 	 * We need to do these manipilations only if we are working
326 	 * with an older version of on-disk superblock.
327 	 */
328 	if (xfs_sb_version_has_pquotino(sbp))
329 		return;
330 
331 	if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
332 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
333 					XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
334 	if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
335 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
336 					XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
337 	sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
338 
339 	if (sbp->sb_qflags & XFS_PQUOTA_ACCT)  {
340 		/*
341 		 * In older version of superblock, on-disk superblock only
342 		 * has sb_gquotino, and in-core superblock has both sb_gquotino
343 		 * and sb_pquotino. But, only one of them is supported at any
344 		 * point of time. So, if PQUOTA is set in disk superblock,
345 		 * copy over sb_gquotino to sb_pquotino.
346 		 */
347 		sbp->sb_pquotino = sbp->sb_gquotino;
348 		sbp->sb_gquotino = NULLFSINO;
349 	}
350 }
351 
352 static void
353 __xfs_sb_from_disk(
354 	struct xfs_sb	*to,
355 	xfs_dsb_t	*from,
356 	bool		convert_xquota)
357 {
358 	to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
359 	to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
360 	to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
361 	to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
362 	to->sb_rextents = be64_to_cpu(from->sb_rextents);
363 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
364 	to->sb_logstart = be64_to_cpu(from->sb_logstart);
365 	to->sb_rootino = be64_to_cpu(from->sb_rootino);
366 	to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
367 	to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
368 	to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
369 	to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
370 	to->sb_agcount = be32_to_cpu(from->sb_agcount);
371 	to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
372 	to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
373 	to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
374 	to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
375 	to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
376 	to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
377 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
378 	to->sb_blocklog = from->sb_blocklog;
379 	to->sb_sectlog = from->sb_sectlog;
380 	to->sb_inodelog = from->sb_inodelog;
381 	to->sb_inopblog = from->sb_inopblog;
382 	to->sb_agblklog = from->sb_agblklog;
383 	to->sb_rextslog = from->sb_rextslog;
384 	to->sb_inprogress = from->sb_inprogress;
385 	to->sb_imax_pct = from->sb_imax_pct;
386 	to->sb_icount = be64_to_cpu(from->sb_icount);
387 	to->sb_ifree = be64_to_cpu(from->sb_ifree);
388 	to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
389 	to->sb_frextents = be64_to_cpu(from->sb_frextents);
390 	to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
391 	to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
392 	to->sb_qflags = be16_to_cpu(from->sb_qflags);
393 	to->sb_flags = from->sb_flags;
394 	to->sb_shared_vn = from->sb_shared_vn;
395 	to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
396 	to->sb_unit = be32_to_cpu(from->sb_unit);
397 	to->sb_width = be32_to_cpu(from->sb_width);
398 	to->sb_dirblklog = from->sb_dirblklog;
399 	to->sb_logsectlog = from->sb_logsectlog;
400 	to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
401 	to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
402 	to->sb_features2 = be32_to_cpu(from->sb_features2);
403 	to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
404 	to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
405 	to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
406 	to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
407 	to->sb_features_log_incompat =
408 				be32_to_cpu(from->sb_features_log_incompat);
409 	/* crc is only used on disk, not in memory; just init to 0 here. */
410 	to->sb_crc = 0;
411 	to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
412 	to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
413 	to->sb_lsn = be64_to_cpu(from->sb_lsn);
414 	/*
415 	 * sb_meta_uuid is only on disk if it differs from sb_uuid and the
416 	 * feature flag is set; if not set we keep it only in memory.
417 	 */
418 	if (xfs_sb_version_hasmetauuid(to))
419 		uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
420 	else
421 		uuid_copy(&to->sb_meta_uuid, &from->sb_uuid);
422 	/* Convert on-disk flags to in-memory flags? */
423 	if (convert_xquota)
424 		xfs_sb_quota_from_disk(to);
425 }
426 
427 void
428 xfs_sb_from_disk(
429 	struct xfs_sb	*to,
430 	xfs_dsb_t	*from)
431 {
432 	__xfs_sb_from_disk(to, from, true);
433 }
434 
435 static void
436 xfs_sb_quota_to_disk(
437 	struct xfs_dsb	*to,
438 	struct xfs_sb	*from)
439 {
440 	__uint16_t	qflags = from->sb_qflags;
441 
442 	to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
443 	if (xfs_sb_version_has_pquotino(from)) {
444 		to->sb_qflags = cpu_to_be16(from->sb_qflags);
445 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
446 		to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
447 		return;
448 	}
449 
450 	/*
451 	 * The in-core version of sb_qflags do not have XFS_OQUOTA_*
452 	 * flags, whereas the on-disk version does.  So, convert incore
453 	 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
454 	 */
455 	qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
456 			XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
457 
458 	if (from->sb_qflags &
459 			(XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
460 		qflags |= XFS_OQUOTA_ENFD;
461 	if (from->sb_qflags &
462 			(XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
463 		qflags |= XFS_OQUOTA_CHKD;
464 	to->sb_qflags = cpu_to_be16(qflags);
465 
466 	/*
467 	 * GQUOTINO and PQUOTINO cannot be used together in versions
468 	 * of superblock that do not have pquotino. from->sb_flags
469 	 * tells us which quota is active and should be copied to
470 	 * disk. If neither are active, we should NULL the inode.
471 	 *
472 	 * In all cases, the separate pquotino must remain 0 because it
473 	 * it beyond the "end" of the valid non-pquotino superblock.
474 	 */
475 	if (from->sb_qflags & XFS_GQUOTA_ACCT)
476 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
477 	else if (from->sb_qflags & XFS_PQUOTA_ACCT)
478 		to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
479 	else {
480 		/*
481 		 * We can't rely on just the fields being logged to tell us
482 		 * that it is safe to write NULLFSINO - we should only do that
483 		 * if quotas are not actually enabled. Hence only write
484 		 * NULLFSINO if both in-core quota inodes are NULL.
485 		 */
486 		if (from->sb_gquotino == NULLFSINO &&
487 		    from->sb_pquotino == NULLFSINO)
488 			to->sb_gquotino = cpu_to_be64(NULLFSINO);
489 	}
490 
491 	to->sb_pquotino = 0;
492 }
493 
494 void
495 xfs_sb_to_disk(
496 	struct xfs_dsb	*to,
497 	struct xfs_sb	*from)
498 {
499 	xfs_sb_quota_to_disk(to, from);
500 
501 	to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
502 	to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
503 	to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
504 	to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
505 	to->sb_rextents = cpu_to_be64(from->sb_rextents);
506 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
507 	to->sb_logstart = cpu_to_be64(from->sb_logstart);
508 	to->sb_rootino = cpu_to_be64(from->sb_rootino);
509 	to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
510 	to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
511 	to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
512 	to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
513 	to->sb_agcount = cpu_to_be32(from->sb_agcount);
514 	to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
515 	to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
516 	to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
517 	to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
518 	to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
519 	to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
520 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
521 	to->sb_blocklog = from->sb_blocklog;
522 	to->sb_sectlog = from->sb_sectlog;
523 	to->sb_inodelog = from->sb_inodelog;
524 	to->sb_inopblog = from->sb_inopblog;
525 	to->sb_agblklog = from->sb_agblklog;
526 	to->sb_rextslog = from->sb_rextslog;
527 	to->sb_inprogress = from->sb_inprogress;
528 	to->sb_imax_pct = from->sb_imax_pct;
529 	to->sb_icount = cpu_to_be64(from->sb_icount);
530 	to->sb_ifree = cpu_to_be64(from->sb_ifree);
531 	to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
532 	to->sb_frextents = cpu_to_be64(from->sb_frextents);
533 
534 	to->sb_flags = from->sb_flags;
535 	to->sb_shared_vn = from->sb_shared_vn;
536 	to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
537 	to->sb_unit = cpu_to_be32(from->sb_unit);
538 	to->sb_width = cpu_to_be32(from->sb_width);
539 	to->sb_dirblklog = from->sb_dirblklog;
540 	to->sb_logsectlog = from->sb_logsectlog;
541 	to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
542 	to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
543 
544 	/*
545 	 * We need to ensure that bad_features2 always matches features2.
546 	 * Hence we enforce that here rather than having to remember to do it
547 	 * everywhere else that updates features2.
548 	 */
549 	from->sb_bad_features2 = from->sb_features2;
550 	to->sb_features2 = cpu_to_be32(from->sb_features2);
551 	to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
552 
553 	if (xfs_sb_version_hascrc(from)) {
554 		to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
555 		to->sb_features_ro_compat =
556 				cpu_to_be32(from->sb_features_ro_compat);
557 		to->sb_features_incompat =
558 				cpu_to_be32(from->sb_features_incompat);
559 		to->sb_features_log_incompat =
560 				cpu_to_be32(from->sb_features_log_incompat);
561 		to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
562 		to->sb_lsn = cpu_to_be64(from->sb_lsn);
563 		if (xfs_sb_version_hasmetauuid(from))
564 			uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
565 	}
566 }
567 
568 static int
569 xfs_sb_verify(
570 	struct xfs_buf	*bp,
571 	bool		check_version)
572 {
573 	struct xfs_mount *mp = bp->b_target->bt_mount;
574 	struct xfs_sb	sb;
575 
576 	/*
577 	 * Use call variant which doesn't convert quota flags from disk
578 	 * format, because xfs_mount_validate_sb checks the on-disk flags.
579 	 */
580 	__xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
581 
582 	/*
583 	 * Only check the in progress field for the primary superblock as
584 	 * mkfs.xfs doesn't clear it from secondary superblocks.
585 	 */
586 	return xfs_mount_validate_sb(mp, &sb,
587 				     bp->b_maps[0].bm_bn == XFS_SB_DADDR,
588 				     check_version);
589 }
590 
591 /*
592  * If the superblock has the CRC feature bit set or the CRC field is non-null,
593  * check that the CRC is valid.  We check the CRC field is non-null because a
594  * single bit error could clear the feature bit and unused parts of the
595  * superblock are supposed to be zero. Hence a non-null crc field indicates that
596  * we've potentially lost a feature bit and we should check it anyway.
597  *
598  * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
599  * last field in V4 secondary superblocks.  So for secondary superblocks,
600  * we are more forgiving, and ignore CRC failures if the primary doesn't
601  * indicate that the fs version is V5.
602  */
603 static void
604 xfs_sb_read_verify(
605 	struct xfs_buf	*bp)
606 {
607 	struct xfs_mount *mp = bp->b_target->bt_mount;
608 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
609 	int		error;
610 
611 	/*
612 	 * open code the version check to avoid needing to convert the entire
613 	 * superblock from disk order just to check the version number
614 	 */
615 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
616 	    (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
617 						XFS_SB_VERSION_5) ||
618 	     dsb->sb_crc != 0)) {
619 
620 		if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
621 			/* Only fail bad secondaries on a known V5 filesystem */
622 			if (bp->b_bn == XFS_SB_DADDR ||
623 			    xfs_sb_version_hascrc(&mp->m_sb)) {
624 				error = -EFSBADCRC;
625 				goto out_error;
626 			}
627 		}
628 	}
629 	error = xfs_sb_verify(bp, true);
630 
631 out_error:
632 	if (error) {
633 		xfs_buf_ioerror(bp, error);
634 		if (error == -EFSCORRUPTED || error == -EFSBADCRC)
635 			xfs_verifier_error(bp);
636 	}
637 }
638 
639 /*
640  * We may be probed for a filesystem match, so we may not want to emit
641  * messages when the superblock buffer is not actually an XFS superblock.
642  * If we find an XFS superblock, then run a normal, noisy mount because we are
643  * really going to mount it and want to know about errors.
644  */
645 static void
646 xfs_sb_quiet_read_verify(
647 	struct xfs_buf	*bp)
648 {
649 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
650 
651 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
652 		/* XFS filesystem, verify noisily! */
653 		xfs_sb_read_verify(bp);
654 		return;
655 	}
656 	/* quietly fail */
657 	xfs_buf_ioerror(bp, -EWRONGFS);
658 }
659 
660 static void
661 xfs_sb_write_verify(
662 	struct xfs_buf		*bp)
663 {
664 	struct xfs_mount	*mp = bp->b_target->bt_mount;
665 	struct xfs_buf_log_item	*bip = bp->b_fspriv;
666 	int			error;
667 
668 	error = xfs_sb_verify(bp, false);
669 	if (error) {
670 		xfs_buf_ioerror(bp, error);
671 		xfs_verifier_error(bp);
672 		return;
673 	}
674 
675 	if (!xfs_sb_version_hascrc(&mp->m_sb))
676 		return;
677 
678 	if (bip)
679 		XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
680 
681 	xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
682 }
683 
684 const struct xfs_buf_ops xfs_sb_buf_ops = {
685 	.name = "xfs_sb",
686 	.verify_read = xfs_sb_read_verify,
687 	.verify_write = xfs_sb_write_verify,
688 };
689 
690 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
691 	.name = "xfs_sb_quiet",
692 	.verify_read = xfs_sb_quiet_read_verify,
693 	.verify_write = xfs_sb_write_verify,
694 };
695 
696 /*
697  * xfs_mount_common
698  *
699  * Mount initialization code establishing various mount
700  * fields from the superblock associated with the given
701  * mount structure
702  */
703 void
704 xfs_sb_mount_common(
705 	struct xfs_mount *mp,
706 	struct xfs_sb	*sbp)
707 {
708 	mp->m_agfrotor = mp->m_agirotor = 0;
709 	spin_lock_init(&mp->m_agirotor_lock);
710 	mp->m_maxagi = mp->m_sb.sb_agcount;
711 	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
712 	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
713 	mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
714 	mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
715 	mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
716 	mp->m_blockmask = sbp->sb_blocksize - 1;
717 	mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
718 	mp->m_blockwmask = mp->m_blockwsize - 1;
719 
720 	mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
721 	mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
722 	mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
723 	mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
724 
725 	mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1);
726 	mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0);
727 	mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2;
728 	mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2;
729 
730 	mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
731 	mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
732 	mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
733 	mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
734 
735 	mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 1);
736 	mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 0);
737 	mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2;
738 	mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2;
739 
740 	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
741 	mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
742 					sbp->sb_inopblock);
743 	mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
744 
745 	if (sbp->sb_spino_align)
746 		mp->m_ialloc_min_blks = sbp->sb_spino_align;
747 	else
748 		mp->m_ialloc_min_blks = mp->m_ialloc_blks;
749 	mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
750 	mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp);
751 }
752 
753 /*
754  * xfs_initialize_perag_data
755  *
756  * Read in each per-ag structure so we can count up the number of
757  * allocated inodes, free inodes and used filesystem blocks as this
758  * information is no longer persistent in the superblock. Once we have
759  * this information, write it into the in-core superblock structure.
760  */
761 int
762 xfs_initialize_perag_data(
763 	struct xfs_mount *mp,
764 	xfs_agnumber_t	agcount)
765 {
766 	xfs_agnumber_t	index;
767 	xfs_perag_t	*pag;
768 	xfs_sb_t	*sbp = &mp->m_sb;
769 	uint64_t	ifree = 0;
770 	uint64_t	ialloc = 0;
771 	uint64_t	bfree = 0;
772 	uint64_t	bfreelst = 0;
773 	uint64_t	btree = 0;
774 	int		error;
775 
776 	for (index = 0; index < agcount; index++) {
777 		/*
778 		 * read the agf, then the agi. This gets us
779 		 * all the information we need and populates the
780 		 * per-ag structures for us.
781 		 */
782 		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
783 		if (error)
784 			return error;
785 
786 		error = xfs_ialloc_pagi_init(mp, NULL, index);
787 		if (error)
788 			return error;
789 		pag = xfs_perag_get(mp, index);
790 		ifree += pag->pagi_freecount;
791 		ialloc += pag->pagi_count;
792 		bfree += pag->pagf_freeblks;
793 		bfreelst += pag->pagf_flcount;
794 		btree += pag->pagf_btreeblks;
795 		xfs_perag_put(pag);
796 	}
797 
798 	/* Overwrite incore superblock counters with just-read data */
799 	spin_lock(&mp->m_sb_lock);
800 	sbp->sb_ifree = ifree;
801 	sbp->sb_icount = ialloc;
802 	sbp->sb_fdblocks = bfree + bfreelst + btree;
803 	spin_unlock(&mp->m_sb_lock);
804 
805 	xfs_reinit_percpu_counters(mp);
806 
807 	return 0;
808 }
809 
810 /*
811  * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
812  * into the superblock buffer to be logged.  It does not provide the higher
813  * level of locking that is needed to protect the in-core superblock from
814  * concurrent access.
815  */
816 void
817 xfs_log_sb(
818 	struct xfs_trans	*tp)
819 {
820 	struct xfs_mount	*mp = tp->t_mountp;
821 	struct xfs_buf		*bp = xfs_trans_getsb(tp, mp, 0);
822 
823 	mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount);
824 	mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree);
825 	mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks);
826 
827 	xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
828 	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
829 	xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb));
830 }
831 
832 /*
833  * xfs_sync_sb
834  *
835  * Sync the superblock to disk.
836  *
837  * Note that the caller is responsible for checking the frozen state of the
838  * filesystem. This procedure uses the non-blocking transaction allocator and
839  * thus will allow modifications to a frozen fs. This is required because this
840  * code can be called during the process of freezing where use of the high-level
841  * allocator would deadlock.
842  */
843 int
844 xfs_sync_sb(
845 	struct xfs_mount	*mp,
846 	bool			wait)
847 {
848 	struct xfs_trans	*tp;
849 	int			error;
850 
851 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0,
852 			XFS_TRANS_NO_WRITECOUNT, &tp);
853 	if (error)
854 		return error;
855 
856 	xfs_log_sb(tp);
857 	if (wait)
858 		xfs_trans_set_sync(tp);
859 	return xfs_trans_commit(tp);
860 }
861