xref: /openbmc/linux/fs/xfs/libxfs/xfs_sb.c (revision 0f4630f3)
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 #include "xfs_log.h"
39 
40 /*
41  * Physical superblock buffer manipulations. Shared with libxfs in userspace.
42  */
43 
44 /*
45  * Reference counting access wrappers to the perag structures.
46  * Because we never free per-ag structures, the only thing we
47  * have to protect against changes is the tree structure itself.
48  */
49 struct xfs_perag *
50 xfs_perag_get(
51 	struct xfs_mount	*mp,
52 	xfs_agnumber_t		agno)
53 {
54 	struct xfs_perag	*pag;
55 	int			ref = 0;
56 
57 	rcu_read_lock();
58 	pag = radix_tree_lookup(&mp->m_perag_tree, agno);
59 	if (pag) {
60 		ASSERT(atomic_read(&pag->pag_ref) >= 0);
61 		ref = atomic_inc_return(&pag->pag_ref);
62 	}
63 	rcu_read_unlock();
64 	trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
65 	return pag;
66 }
67 
68 /*
69  * search from @first to find the next perag with the given tag set.
70  */
71 struct xfs_perag *
72 xfs_perag_get_tag(
73 	struct xfs_mount	*mp,
74 	xfs_agnumber_t		first,
75 	int			tag)
76 {
77 	struct xfs_perag	*pag;
78 	int			found;
79 	int			ref;
80 
81 	rcu_read_lock();
82 	found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
83 					(void **)&pag, first, 1, tag);
84 	if (found <= 0) {
85 		rcu_read_unlock();
86 		return NULL;
87 	}
88 	ref = atomic_inc_return(&pag->pag_ref);
89 	rcu_read_unlock();
90 	trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
91 	return pag;
92 }
93 
94 void
95 xfs_perag_put(
96 	struct xfs_perag	*pag)
97 {
98 	int	ref;
99 
100 	ASSERT(atomic_read(&pag->pag_ref) > 0);
101 	ref = atomic_dec_return(&pag->pag_ref);
102 	trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
103 }
104 
105 /*
106  * Check the validity of the SB found.
107  */
108 STATIC int
109 xfs_mount_validate_sb(
110 	xfs_mount_t	*mp,
111 	xfs_sb_t	*sbp,
112 	bool		check_inprogress,
113 	bool		check_version)
114 {
115 	if (sbp->sb_magicnum != XFS_SB_MAGIC) {
116 		xfs_warn(mp, "bad magic number");
117 		return -EWRONGFS;
118 	}
119 
120 
121 	if (!xfs_sb_good_version(sbp)) {
122 		xfs_warn(mp, "bad version");
123 		return -EWRONGFS;
124 	}
125 
126 	/*
127 	 * Version 5 superblock feature mask validation. Reject combinations the
128 	 * kernel cannot support up front before checking anything else. For
129 	 * write validation, we don't need to check feature masks.
130 	 */
131 	if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) {
132 		if (xfs_sb_has_compat_feature(sbp,
133 					XFS_SB_FEAT_COMPAT_UNKNOWN)) {
134 			xfs_warn(mp,
135 "Superblock has unknown compatible features (0x%x) enabled.",
136 				(sbp->sb_features_compat &
137 						XFS_SB_FEAT_COMPAT_UNKNOWN));
138 			xfs_warn(mp,
139 "Using a more recent kernel is recommended.");
140 		}
141 
142 		if (xfs_sb_has_ro_compat_feature(sbp,
143 					XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
144 			xfs_alert(mp,
145 "Superblock has unknown read-only compatible features (0x%x) enabled.",
146 				(sbp->sb_features_ro_compat &
147 						XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
148 			if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
149 				xfs_warn(mp,
150 "Attempted to mount read-only compatible filesystem read-write.");
151 				xfs_warn(mp,
152 "Filesystem can only be safely mounted read only.");
153 
154 				return -EINVAL;
155 			}
156 		}
157 		if (xfs_sb_has_incompat_feature(sbp,
158 					XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
159 			xfs_warn(mp,
160 "Superblock has unknown incompatible features (0x%x) enabled.",
161 				(sbp->sb_features_incompat &
162 						XFS_SB_FEAT_INCOMPAT_UNKNOWN));
163 			xfs_warn(mp,
164 "Filesystem can not be safely mounted by this kernel.");
165 			return -EINVAL;
166 		}
167 	} else if (xfs_sb_version_hascrc(sbp)) {
168 		/*
169 		 * We can't read verify the sb LSN because the read verifier is
170 		 * called before the log is allocated and processed. We know the
171 		 * log is set up before write verifier (!check_version) calls,
172 		 * so just check it here.
173 		 */
174 		if (!xfs_log_check_lsn(mp, sbp->sb_lsn))
175 			return -EFSCORRUPTED;
176 	}
177 
178 	if (xfs_sb_version_has_pquotino(sbp)) {
179 		if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
180 			xfs_notice(mp,
181 			   "Version 5 of Super block has XFS_OQUOTA bits.");
182 			return -EFSCORRUPTED;
183 		}
184 	} else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
185 				XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
186 			xfs_notice(mp,
187 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
188 			return -EFSCORRUPTED;
189 	}
190 
191 	/*
192 	 * Full inode chunks must be aligned to inode chunk size when
193 	 * sparse inodes are enabled to support the sparse chunk
194 	 * allocation algorithm and prevent overlapping inode records.
195 	 */
196 	if (xfs_sb_version_hassparseinodes(sbp)) {
197 		uint32_t	align;
198 
199 		align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
200 				>> sbp->sb_blocklog;
201 		if (sbp->sb_inoalignmt != align) {
202 			xfs_warn(mp,
203 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
204 				 sbp->sb_inoalignmt, align);
205 			return -EINVAL;
206 		}
207 	}
208 
209 	if (unlikely(
210 	    sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
211 		xfs_warn(mp,
212 		"filesystem is marked as having an external log; "
213 		"specify logdev on the mount command line.");
214 		return -EINVAL;
215 	}
216 
217 	if (unlikely(
218 	    sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
219 		xfs_warn(mp,
220 		"filesystem is marked as having an internal log; "
221 		"do not specify logdev on the mount command line.");
222 		return -EINVAL;
223 	}
224 
225 	/*
226 	 * More sanity checking.  Most of these were stolen directly from
227 	 * xfs_repair.
228 	 */
229 	if (unlikely(
230 	    sbp->sb_agcount <= 0					||
231 	    sbp->sb_sectsize < XFS_MIN_SECTORSIZE			||
232 	    sbp->sb_sectsize > XFS_MAX_SECTORSIZE			||
233 	    sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG			||
234 	    sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG			||
235 	    sbp->sb_sectsize != (1 << sbp->sb_sectlog)			||
236 	    sbp->sb_blocksize < XFS_MIN_BLOCKSIZE			||
237 	    sbp->sb_blocksize > XFS_MAX_BLOCKSIZE			||
238 	    sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG			||
239 	    sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG			||
240 	    sbp->sb_blocksize != (1 << sbp->sb_blocklog)		||
241 	    sbp->sb_dirblklog > XFS_MAX_BLOCKSIZE_LOG			||
242 	    sbp->sb_inodesize < XFS_DINODE_MIN_SIZE			||
243 	    sbp->sb_inodesize > XFS_DINODE_MAX_SIZE			||
244 	    sbp->sb_inodelog < XFS_DINODE_MIN_LOG			||
245 	    sbp->sb_inodelog > XFS_DINODE_MAX_LOG			||
246 	    sbp->sb_inodesize != (1 << sbp->sb_inodelog)		||
247 	    sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE			||
248 	    sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
249 	    (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog)	||
250 	    (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE)	||
251 	    (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)	||
252 	    (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */)	||
253 	    sbp->sb_dblocks == 0					||
254 	    sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp)			||
255 	    sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp)			||
256 	    sbp->sb_shared_vn != 0)) {
257 		xfs_notice(mp, "SB sanity check failed");
258 		return -EFSCORRUPTED;
259 	}
260 
261 	/*
262 	 * Until this is fixed only page-sized or smaller data blocks work.
263 	 */
264 	if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
265 		xfs_warn(mp,
266 		"File system with blocksize %d bytes. "
267 		"Only pagesize (%ld) or less will currently work.",
268 				sbp->sb_blocksize, PAGE_SIZE);
269 		return -ENOSYS;
270 	}
271 
272 	/*
273 	 * Currently only very few inode sizes are supported.
274 	 */
275 	switch (sbp->sb_inodesize) {
276 	case 256:
277 	case 512:
278 	case 1024:
279 	case 2048:
280 		break;
281 	default:
282 		xfs_warn(mp, "inode size of %d bytes not supported",
283 				sbp->sb_inodesize);
284 		return -ENOSYS;
285 	}
286 
287 	if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
288 	    xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
289 		xfs_warn(mp,
290 		"file system too large to be mounted on this system.");
291 		return -EFBIG;
292 	}
293 
294 	if (check_inprogress && sbp->sb_inprogress) {
295 		xfs_warn(mp, "Offline file system operation in progress!");
296 		return -EFSCORRUPTED;
297 	}
298 	return 0;
299 }
300 
301 void
302 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
303 {
304 	/*
305 	 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
306 	 * leads to in-core values having two different values for a quota
307 	 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
308 	 * NULLFSINO.
309 	 *
310 	 * Note that this change affect only the in-core values. These
311 	 * values are not written back to disk unless any quota information
312 	 * is written to the disk. Even in that case, sb_pquotino field is
313 	 * not written to disk unless the superblock supports pquotino.
314 	 */
315 	if (sbp->sb_uquotino == 0)
316 		sbp->sb_uquotino = NULLFSINO;
317 	if (sbp->sb_gquotino == 0)
318 		sbp->sb_gquotino = NULLFSINO;
319 	if (sbp->sb_pquotino == 0)
320 		sbp->sb_pquotino = NULLFSINO;
321 
322 	/*
323 	 * We need to do these manipilations only if we are working
324 	 * with an older version of on-disk superblock.
325 	 */
326 	if (xfs_sb_version_has_pquotino(sbp))
327 		return;
328 
329 	if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
330 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
331 					XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
332 	if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
333 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
334 					XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
335 	sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
336 
337 	if (sbp->sb_qflags & XFS_PQUOTA_ACCT)  {
338 		/*
339 		 * In older version of superblock, on-disk superblock only
340 		 * has sb_gquotino, and in-core superblock has both sb_gquotino
341 		 * and sb_pquotino. But, only one of them is supported at any
342 		 * point of time. So, if PQUOTA is set in disk superblock,
343 		 * copy over sb_gquotino to sb_pquotino.
344 		 */
345 		sbp->sb_pquotino = sbp->sb_gquotino;
346 		sbp->sb_gquotino = NULLFSINO;
347 	}
348 }
349 
350 static void
351 __xfs_sb_from_disk(
352 	struct xfs_sb	*to,
353 	xfs_dsb_t	*from,
354 	bool		convert_xquota)
355 {
356 	to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
357 	to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
358 	to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
359 	to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
360 	to->sb_rextents = be64_to_cpu(from->sb_rextents);
361 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
362 	to->sb_logstart = be64_to_cpu(from->sb_logstart);
363 	to->sb_rootino = be64_to_cpu(from->sb_rootino);
364 	to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
365 	to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
366 	to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
367 	to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
368 	to->sb_agcount = be32_to_cpu(from->sb_agcount);
369 	to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
370 	to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
371 	to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
372 	to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
373 	to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
374 	to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
375 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
376 	to->sb_blocklog = from->sb_blocklog;
377 	to->sb_sectlog = from->sb_sectlog;
378 	to->sb_inodelog = from->sb_inodelog;
379 	to->sb_inopblog = from->sb_inopblog;
380 	to->sb_agblklog = from->sb_agblklog;
381 	to->sb_rextslog = from->sb_rextslog;
382 	to->sb_inprogress = from->sb_inprogress;
383 	to->sb_imax_pct = from->sb_imax_pct;
384 	to->sb_icount = be64_to_cpu(from->sb_icount);
385 	to->sb_ifree = be64_to_cpu(from->sb_ifree);
386 	to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
387 	to->sb_frextents = be64_to_cpu(from->sb_frextents);
388 	to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
389 	to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
390 	to->sb_qflags = be16_to_cpu(from->sb_qflags);
391 	to->sb_flags = from->sb_flags;
392 	to->sb_shared_vn = from->sb_shared_vn;
393 	to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
394 	to->sb_unit = be32_to_cpu(from->sb_unit);
395 	to->sb_width = be32_to_cpu(from->sb_width);
396 	to->sb_dirblklog = from->sb_dirblklog;
397 	to->sb_logsectlog = from->sb_logsectlog;
398 	to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
399 	to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
400 	to->sb_features2 = be32_to_cpu(from->sb_features2);
401 	to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
402 	to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
403 	to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
404 	to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
405 	to->sb_features_log_incompat =
406 				be32_to_cpu(from->sb_features_log_incompat);
407 	/* crc is only used on disk, not in memory; just init to 0 here. */
408 	to->sb_crc = 0;
409 	to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
410 	to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
411 	to->sb_lsn = be64_to_cpu(from->sb_lsn);
412 	/*
413 	 * sb_meta_uuid is only on disk if it differs from sb_uuid and the
414 	 * feature flag is set; if not set we keep it only in memory.
415 	 */
416 	if (xfs_sb_version_hasmetauuid(to))
417 		uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
418 	else
419 		uuid_copy(&to->sb_meta_uuid, &from->sb_uuid);
420 	/* Convert on-disk flags to in-memory flags? */
421 	if (convert_xquota)
422 		xfs_sb_quota_from_disk(to);
423 }
424 
425 void
426 xfs_sb_from_disk(
427 	struct xfs_sb	*to,
428 	xfs_dsb_t	*from)
429 {
430 	__xfs_sb_from_disk(to, from, true);
431 }
432 
433 static void
434 xfs_sb_quota_to_disk(
435 	struct xfs_dsb	*to,
436 	struct xfs_sb	*from)
437 {
438 	__uint16_t	qflags = from->sb_qflags;
439 
440 	to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
441 	if (xfs_sb_version_has_pquotino(from)) {
442 		to->sb_qflags = cpu_to_be16(from->sb_qflags);
443 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
444 		to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
445 		return;
446 	}
447 
448 	/*
449 	 * The in-core version of sb_qflags do not have XFS_OQUOTA_*
450 	 * flags, whereas the on-disk version does.  So, convert incore
451 	 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
452 	 */
453 	qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
454 			XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
455 
456 	if (from->sb_qflags &
457 			(XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
458 		qflags |= XFS_OQUOTA_ENFD;
459 	if (from->sb_qflags &
460 			(XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
461 		qflags |= XFS_OQUOTA_CHKD;
462 	to->sb_qflags = cpu_to_be16(qflags);
463 
464 	/*
465 	 * GQUOTINO and PQUOTINO cannot be used together in versions
466 	 * of superblock that do not have pquotino. from->sb_flags
467 	 * tells us which quota is active and should be copied to
468 	 * disk. If neither are active, we should NULL the inode.
469 	 *
470 	 * In all cases, the separate pquotino must remain 0 because it
471 	 * it beyond the "end" of the valid non-pquotino superblock.
472 	 */
473 	if (from->sb_qflags & XFS_GQUOTA_ACCT)
474 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
475 	else if (from->sb_qflags & XFS_PQUOTA_ACCT)
476 		to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
477 	else {
478 		/*
479 		 * We can't rely on just the fields being logged to tell us
480 		 * that it is safe to write NULLFSINO - we should only do that
481 		 * if quotas are not actually enabled. Hence only write
482 		 * NULLFSINO if both in-core quota inodes are NULL.
483 		 */
484 		if (from->sb_gquotino == NULLFSINO &&
485 		    from->sb_pquotino == NULLFSINO)
486 			to->sb_gquotino = cpu_to_be64(NULLFSINO);
487 	}
488 
489 	to->sb_pquotino = 0;
490 }
491 
492 void
493 xfs_sb_to_disk(
494 	struct xfs_dsb	*to,
495 	struct xfs_sb	*from)
496 {
497 	xfs_sb_quota_to_disk(to, from);
498 
499 	to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
500 	to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
501 	to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
502 	to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
503 	to->sb_rextents = cpu_to_be64(from->sb_rextents);
504 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
505 	to->sb_logstart = cpu_to_be64(from->sb_logstart);
506 	to->sb_rootino = cpu_to_be64(from->sb_rootino);
507 	to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
508 	to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
509 	to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
510 	to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
511 	to->sb_agcount = cpu_to_be32(from->sb_agcount);
512 	to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
513 	to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
514 	to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
515 	to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
516 	to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
517 	to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
518 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
519 	to->sb_blocklog = from->sb_blocklog;
520 	to->sb_sectlog = from->sb_sectlog;
521 	to->sb_inodelog = from->sb_inodelog;
522 	to->sb_inopblog = from->sb_inopblog;
523 	to->sb_agblklog = from->sb_agblklog;
524 	to->sb_rextslog = from->sb_rextslog;
525 	to->sb_inprogress = from->sb_inprogress;
526 	to->sb_imax_pct = from->sb_imax_pct;
527 	to->sb_icount = cpu_to_be64(from->sb_icount);
528 	to->sb_ifree = cpu_to_be64(from->sb_ifree);
529 	to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
530 	to->sb_frextents = cpu_to_be64(from->sb_frextents);
531 
532 	to->sb_flags = from->sb_flags;
533 	to->sb_shared_vn = from->sb_shared_vn;
534 	to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
535 	to->sb_unit = cpu_to_be32(from->sb_unit);
536 	to->sb_width = cpu_to_be32(from->sb_width);
537 	to->sb_dirblklog = from->sb_dirblklog;
538 	to->sb_logsectlog = from->sb_logsectlog;
539 	to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
540 	to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
541 
542 	/*
543 	 * We need to ensure that bad_features2 always matches features2.
544 	 * Hence we enforce that here rather than having to remember to do it
545 	 * everywhere else that updates features2.
546 	 */
547 	from->sb_bad_features2 = from->sb_features2;
548 	to->sb_features2 = cpu_to_be32(from->sb_features2);
549 	to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
550 
551 	if (xfs_sb_version_hascrc(from)) {
552 		to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
553 		to->sb_features_ro_compat =
554 				cpu_to_be32(from->sb_features_ro_compat);
555 		to->sb_features_incompat =
556 				cpu_to_be32(from->sb_features_incompat);
557 		to->sb_features_log_incompat =
558 				cpu_to_be32(from->sb_features_log_incompat);
559 		to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
560 		to->sb_lsn = cpu_to_be64(from->sb_lsn);
561 		if (xfs_sb_version_hasmetauuid(from))
562 			uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
563 	}
564 }
565 
566 static int
567 xfs_sb_verify(
568 	struct xfs_buf	*bp,
569 	bool		check_version)
570 {
571 	struct xfs_mount *mp = bp->b_target->bt_mount;
572 	struct xfs_sb	sb;
573 
574 	/*
575 	 * Use call variant which doesn't convert quota flags from disk
576 	 * format, because xfs_mount_validate_sb checks the on-disk flags.
577 	 */
578 	__xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
579 
580 	/*
581 	 * Only check the in progress field for the primary superblock as
582 	 * mkfs.xfs doesn't clear it from secondary superblocks.
583 	 */
584 	return xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR,
585 				     check_version);
586 }
587 
588 /*
589  * If the superblock has the CRC feature bit set or the CRC field is non-null,
590  * check that the CRC is valid.  We check the CRC field is non-null because a
591  * single bit error could clear the feature bit and unused parts of the
592  * superblock are supposed to be zero. Hence a non-null crc field indicates that
593  * we've potentially lost a feature bit and we should check it anyway.
594  *
595  * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
596  * last field in V4 secondary superblocks.  So for secondary superblocks,
597  * we are more forgiving, and ignore CRC failures if the primary doesn't
598  * indicate that the fs version is V5.
599  */
600 static void
601 xfs_sb_read_verify(
602 	struct xfs_buf	*bp)
603 {
604 	struct xfs_mount *mp = bp->b_target->bt_mount;
605 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
606 	int		error;
607 
608 	/*
609 	 * open code the version check to avoid needing to convert the entire
610 	 * superblock from disk order just to check the version number
611 	 */
612 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
613 	    (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
614 						XFS_SB_VERSION_5) ||
615 	     dsb->sb_crc != 0)) {
616 
617 		if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
618 			/* Only fail bad secondaries on a known V5 filesystem */
619 			if (bp->b_bn == XFS_SB_DADDR ||
620 			    xfs_sb_version_hascrc(&mp->m_sb)) {
621 				error = -EFSBADCRC;
622 				goto out_error;
623 			}
624 		}
625 	}
626 	error = xfs_sb_verify(bp, true);
627 
628 out_error:
629 	if (error) {
630 		xfs_buf_ioerror(bp, error);
631 		if (error == -EFSCORRUPTED || error == -EFSBADCRC)
632 			xfs_verifier_error(bp);
633 	}
634 }
635 
636 /*
637  * We may be probed for a filesystem match, so we may not want to emit
638  * messages when the superblock buffer is not actually an XFS superblock.
639  * If we find an XFS superblock, then run a normal, noisy mount because we are
640  * really going to mount it and want to know about errors.
641  */
642 static void
643 xfs_sb_quiet_read_verify(
644 	struct xfs_buf	*bp)
645 {
646 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
647 
648 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
649 		/* XFS filesystem, verify noisily! */
650 		xfs_sb_read_verify(bp);
651 		return;
652 	}
653 	/* quietly fail */
654 	xfs_buf_ioerror(bp, -EWRONGFS);
655 }
656 
657 static void
658 xfs_sb_write_verify(
659 	struct xfs_buf		*bp)
660 {
661 	struct xfs_mount	*mp = bp->b_target->bt_mount;
662 	struct xfs_buf_log_item	*bip = bp->b_fspriv;
663 	int			error;
664 
665 	error = xfs_sb_verify(bp, false);
666 	if (error) {
667 		xfs_buf_ioerror(bp, error);
668 		xfs_verifier_error(bp);
669 		return;
670 	}
671 
672 	if (!xfs_sb_version_hascrc(&mp->m_sb))
673 		return;
674 
675 	if (bip)
676 		XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
677 
678 	xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
679 }
680 
681 const struct xfs_buf_ops xfs_sb_buf_ops = {
682 	.verify_read = xfs_sb_read_verify,
683 	.verify_write = xfs_sb_write_verify,
684 };
685 
686 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
687 	.verify_read = xfs_sb_quiet_read_verify,
688 	.verify_write = xfs_sb_write_verify,
689 };
690 
691 /*
692  * xfs_mount_common
693  *
694  * Mount initialization code establishing various mount
695  * fields from the superblock associated with the given
696  * mount structure
697  */
698 void
699 xfs_sb_mount_common(
700 	struct xfs_mount *mp,
701 	struct xfs_sb	*sbp)
702 {
703 	mp->m_agfrotor = mp->m_agirotor = 0;
704 	spin_lock_init(&mp->m_agirotor_lock);
705 	mp->m_maxagi = mp->m_sb.sb_agcount;
706 	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
707 	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
708 	mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
709 	mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
710 	mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
711 	mp->m_blockmask = sbp->sb_blocksize - 1;
712 	mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
713 	mp->m_blockwmask = mp->m_blockwsize - 1;
714 
715 	mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
716 	mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
717 	mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
718 	mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
719 
720 	mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1);
721 	mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0);
722 	mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2;
723 	mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2;
724 
725 	mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
726 	mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
727 	mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
728 	mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
729 
730 	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
731 	mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
732 					sbp->sb_inopblock);
733 	mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
734 
735 	if (sbp->sb_spino_align)
736 		mp->m_ialloc_min_blks = sbp->sb_spino_align;
737 	else
738 		mp->m_ialloc_min_blks = mp->m_ialloc_blks;
739 }
740 
741 /*
742  * xfs_initialize_perag_data
743  *
744  * Read in each per-ag structure so we can count up the number of
745  * allocated inodes, free inodes and used filesystem blocks as this
746  * information is no longer persistent in the superblock. Once we have
747  * this information, write it into the in-core superblock structure.
748  */
749 int
750 xfs_initialize_perag_data(
751 	struct xfs_mount *mp,
752 	xfs_agnumber_t	agcount)
753 {
754 	xfs_agnumber_t	index;
755 	xfs_perag_t	*pag;
756 	xfs_sb_t	*sbp = &mp->m_sb;
757 	uint64_t	ifree = 0;
758 	uint64_t	ialloc = 0;
759 	uint64_t	bfree = 0;
760 	uint64_t	bfreelst = 0;
761 	uint64_t	btree = 0;
762 	int		error;
763 
764 	for (index = 0; index < agcount; index++) {
765 		/*
766 		 * read the agf, then the agi. This gets us
767 		 * all the information we need and populates the
768 		 * per-ag structures for us.
769 		 */
770 		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
771 		if (error)
772 			return error;
773 
774 		error = xfs_ialloc_pagi_init(mp, NULL, index);
775 		if (error)
776 			return error;
777 		pag = xfs_perag_get(mp, index);
778 		ifree += pag->pagi_freecount;
779 		ialloc += pag->pagi_count;
780 		bfree += pag->pagf_freeblks;
781 		bfreelst += pag->pagf_flcount;
782 		btree += pag->pagf_btreeblks;
783 		xfs_perag_put(pag);
784 	}
785 
786 	/* Overwrite incore superblock counters with just-read data */
787 	spin_lock(&mp->m_sb_lock);
788 	sbp->sb_ifree = ifree;
789 	sbp->sb_icount = ialloc;
790 	sbp->sb_fdblocks = bfree + bfreelst + btree;
791 	spin_unlock(&mp->m_sb_lock);
792 
793 	xfs_reinit_percpu_counters(mp);
794 
795 	return 0;
796 }
797 
798 /*
799  * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
800  * into the superblock buffer to be logged.  It does not provide the higher
801  * level of locking that is needed to protect the in-core superblock from
802  * concurrent access.
803  */
804 void
805 xfs_log_sb(
806 	struct xfs_trans	*tp)
807 {
808 	struct xfs_mount	*mp = tp->t_mountp;
809 	struct xfs_buf		*bp = xfs_trans_getsb(tp, mp, 0);
810 
811 	mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount);
812 	mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree);
813 	mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks);
814 
815 	xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
816 	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
817 	xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb));
818 }
819 
820 /*
821  * xfs_sync_sb
822  *
823  * Sync the superblock to disk.
824  *
825  * Note that the caller is responsible for checking the frozen state of the
826  * filesystem. This procedure uses the non-blocking transaction allocator and
827  * thus will allow modifications to a frozen fs. This is required because this
828  * code can be called during the process of freezing where use of the high-level
829  * allocator would deadlock.
830  */
831 int
832 xfs_sync_sb(
833 	struct xfs_mount	*mp,
834 	bool			wait)
835 {
836 	struct xfs_trans	*tp;
837 	int			error;
838 
839 	tp = _xfs_trans_alloc(mp, XFS_TRANS_SB_CHANGE, KM_SLEEP);
840 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0);
841 	if (error) {
842 		xfs_trans_cancel(tp);
843 		return error;
844 	}
845 
846 	xfs_log_sb(tp);
847 	if (wait)
848 		xfs_trans_set_sync(tp);
849 	return xfs_trans_commit(tp);
850 }
851