xref: /openbmc/linux/fs/xfs/libxfs/xfs_sb.c (revision 867a0e05)
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 static const struct {
44 	short offset;
45 	short type;	/* 0 = integer
46 			 * 1 = binary / string (no translation)
47 			 */
48 } xfs_sb_info[] = {
49 	{ offsetof(xfs_sb_t, sb_magicnum),	0 },
50 	{ offsetof(xfs_sb_t, sb_blocksize),	0 },
51 	{ offsetof(xfs_sb_t, sb_dblocks),	0 },
52 	{ offsetof(xfs_sb_t, sb_rblocks),	0 },
53 	{ offsetof(xfs_sb_t, sb_rextents),	0 },
54 	{ offsetof(xfs_sb_t, sb_uuid),		1 },
55 	{ offsetof(xfs_sb_t, sb_logstart),	0 },
56 	{ offsetof(xfs_sb_t, sb_rootino),	0 },
57 	{ offsetof(xfs_sb_t, sb_rbmino),	0 },
58 	{ offsetof(xfs_sb_t, sb_rsumino),	0 },
59 	{ offsetof(xfs_sb_t, sb_rextsize),	0 },
60 	{ offsetof(xfs_sb_t, sb_agblocks),	0 },
61 	{ offsetof(xfs_sb_t, sb_agcount),	0 },
62 	{ offsetof(xfs_sb_t, sb_rbmblocks),	0 },
63 	{ offsetof(xfs_sb_t, sb_logblocks),	0 },
64 	{ offsetof(xfs_sb_t, sb_versionnum),	0 },
65 	{ offsetof(xfs_sb_t, sb_sectsize),	0 },
66 	{ offsetof(xfs_sb_t, sb_inodesize),	0 },
67 	{ offsetof(xfs_sb_t, sb_inopblock),	0 },
68 	{ offsetof(xfs_sb_t, sb_fname[0]),	1 },
69 	{ offsetof(xfs_sb_t, sb_blocklog),	0 },
70 	{ offsetof(xfs_sb_t, sb_sectlog),	0 },
71 	{ offsetof(xfs_sb_t, sb_inodelog),	0 },
72 	{ offsetof(xfs_sb_t, sb_inopblog),	0 },
73 	{ offsetof(xfs_sb_t, sb_agblklog),	0 },
74 	{ offsetof(xfs_sb_t, sb_rextslog),	0 },
75 	{ offsetof(xfs_sb_t, sb_inprogress),	0 },
76 	{ offsetof(xfs_sb_t, sb_imax_pct),	0 },
77 	{ offsetof(xfs_sb_t, sb_icount),	0 },
78 	{ offsetof(xfs_sb_t, sb_ifree),		0 },
79 	{ offsetof(xfs_sb_t, sb_fdblocks),	0 },
80 	{ offsetof(xfs_sb_t, sb_frextents),	0 },
81 	{ offsetof(xfs_sb_t, sb_uquotino),	0 },
82 	{ offsetof(xfs_sb_t, sb_gquotino),	0 },
83 	{ offsetof(xfs_sb_t, sb_qflags),	0 },
84 	{ offsetof(xfs_sb_t, sb_flags),		0 },
85 	{ offsetof(xfs_sb_t, sb_shared_vn),	0 },
86 	{ offsetof(xfs_sb_t, sb_inoalignmt),	0 },
87 	{ offsetof(xfs_sb_t, sb_unit),		0 },
88 	{ offsetof(xfs_sb_t, sb_width),		0 },
89 	{ offsetof(xfs_sb_t, sb_dirblklog),	0 },
90 	{ offsetof(xfs_sb_t, sb_logsectlog),	0 },
91 	{ offsetof(xfs_sb_t, sb_logsectsize),	0 },
92 	{ offsetof(xfs_sb_t, sb_logsunit),	0 },
93 	{ offsetof(xfs_sb_t, sb_features2),	0 },
94 	{ offsetof(xfs_sb_t, sb_bad_features2),	0 },
95 	{ offsetof(xfs_sb_t, sb_features_compat),	0 },
96 	{ offsetof(xfs_sb_t, sb_features_ro_compat),	0 },
97 	{ offsetof(xfs_sb_t, sb_features_incompat),	0 },
98 	{ offsetof(xfs_sb_t, sb_features_log_incompat),	0 },
99 	{ offsetof(xfs_sb_t, sb_crc),		0 },
100 	{ offsetof(xfs_sb_t, sb_pad),		0 },
101 	{ offsetof(xfs_sb_t, sb_pquotino),	0 },
102 	{ offsetof(xfs_sb_t, sb_lsn),		0 },
103 	{ sizeof(xfs_sb_t),			0 }
104 };
105 
106 /*
107  * Reference counting access wrappers to the perag structures.
108  * Because we never free per-ag structures, the only thing we
109  * have to protect against changes is the tree structure itself.
110  */
111 struct xfs_perag *
112 xfs_perag_get(
113 	struct xfs_mount	*mp,
114 	xfs_agnumber_t		agno)
115 {
116 	struct xfs_perag	*pag;
117 	int			ref = 0;
118 
119 	rcu_read_lock();
120 	pag = radix_tree_lookup(&mp->m_perag_tree, agno);
121 	if (pag) {
122 		ASSERT(atomic_read(&pag->pag_ref) >= 0);
123 		ref = atomic_inc_return(&pag->pag_ref);
124 	}
125 	rcu_read_unlock();
126 	trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
127 	return pag;
128 }
129 
130 /*
131  * search from @first to find the next perag with the given tag set.
132  */
133 struct xfs_perag *
134 xfs_perag_get_tag(
135 	struct xfs_mount	*mp,
136 	xfs_agnumber_t		first,
137 	int			tag)
138 {
139 	struct xfs_perag	*pag;
140 	int			found;
141 	int			ref;
142 
143 	rcu_read_lock();
144 	found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
145 					(void **)&pag, first, 1, tag);
146 	if (found <= 0) {
147 		rcu_read_unlock();
148 		return NULL;
149 	}
150 	ref = atomic_inc_return(&pag->pag_ref);
151 	rcu_read_unlock();
152 	trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
153 	return pag;
154 }
155 
156 void
157 xfs_perag_put(
158 	struct xfs_perag	*pag)
159 {
160 	int	ref;
161 
162 	ASSERT(atomic_read(&pag->pag_ref) > 0);
163 	ref = atomic_dec_return(&pag->pag_ref);
164 	trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
165 }
166 
167 /*
168  * Check the validity of the SB found.
169  */
170 STATIC int
171 xfs_mount_validate_sb(
172 	xfs_mount_t	*mp,
173 	xfs_sb_t	*sbp,
174 	bool		check_inprogress,
175 	bool		check_version)
176 {
177 
178 	/*
179 	 * If the log device and data device have the
180 	 * same device number, the log is internal.
181 	 * Consequently, the sb_logstart should be non-zero.  If
182 	 * we have a zero sb_logstart in this case, we may be trying to mount
183 	 * a volume filesystem in a non-volume manner.
184 	 */
185 	if (sbp->sb_magicnum != XFS_SB_MAGIC) {
186 		xfs_warn(mp, "bad magic number");
187 		return -EWRONGFS;
188 	}
189 
190 
191 	if (!xfs_sb_good_version(sbp)) {
192 		xfs_warn(mp, "bad version");
193 		return -EWRONGFS;
194 	}
195 
196 	/*
197 	 * Version 5 superblock feature mask validation. Reject combinations the
198 	 * kernel cannot support up front before checking anything else. For
199 	 * write validation, we don't need to check feature masks.
200 	 */
201 	if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) {
202 		if (xfs_sb_has_compat_feature(sbp,
203 					XFS_SB_FEAT_COMPAT_UNKNOWN)) {
204 			xfs_warn(mp,
205 "Superblock has unknown compatible features (0x%x) enabled.\n"
206 "Using a more recent kernel is recommended.",
207 				(sbp->sb_features_compat &
208 						XFS_SB_FEAT_COMPAT_UNKNOWN));
209 		}
210 
211 		if (xfs_sb_has_ro_compat_feature(sbp,
212 					XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
213 			xfs_alert(mp,
214 "Superblock has unknown read-only compatible features (0x%x) enabled.",
215 				(sbp->sb_features_ro_compat &
216 						XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
217 			if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
218 				xfs_warn(mp,
219 "Attempted to mount read-only compatible filesystem read-write.\n"
220 "Filesystem can only be safely mounted read only.");
221 				return -EINVAL;
222 			}
223 		}
224 		if (xfs_sb_has_incompat_feature(sbp,
225 					XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
226 			xfs_warn(mp,
227 "Superblock has unknown incompatible features (0x%x) enabled.\n"
228 "Filesystem can not be safely mounted by this kernel.",
229 				(sbp->sb_features_incompat &
230 						XFS_SB_FEAT_INCOMPAT_UNKNOWN));
231 			return -EINVAL;
232 		}
233 	}
234 
235 	if (xfs_sb_version_has_pquotino(sbp)) {
236 		if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
237 			xfs_notice(mp,
238 			   "Version 5 of Super block has XFS_OQUOTA bits.");
239 			return -EFSCORRUPTED;
240 		}
241 	} else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
242 				XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
243 			xfs_notice(mp,
244 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
245 			return -EFSCORRUPTED;
246 	}
247 
248 	if (unlikely(
249 	    sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
250 		xfs_warn(mp,
251 		"filesystem is marked as having an external log; "
252 		"specify logdev on the mount command line.");
253 		return -EINVAL;
254 	}
255 
256 	if (unlikely(
257 	    sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
258 		xfs_warn(mp,
259 		"filesystem is marked as having an internal log; "
260 		"do not specify logdev on the mount command line.");
261 		return -EINVAL;
262 	}
263 
264 	/*
265 	 * More sanity checking.  Most of these were stolen directly from
266 	 * xfs_repair.
267 	 */
268 	if (unlikely(
269 	    sbp->sb_agcount <= 0					||
270 	    sbp->sb_sectsize < XFS_MIN_SECTORSIZE			||
271 	    sbp->sb_sectsize > XFS_MAX_SECTORSIZE			||
272 	    sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG			||
273 	    sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG			||
274 	    sbp->sb_sectsize != (1 << sbp->sb_sectlog)			||
275 	    sbp->sb_blocksize < XFS_MIN_BLOCKSIZE			||
276 	    sbp->sb_blocksize > XFS_MAX_BLOCKSIZE			||
277 	    sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG			||
278 	    sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG			||
279 	    sbp->sb_blocksize != (1 << sbp->sb_blocklog)		||
280 	    sbp->sb_dirblklog > XFS_MAX_BLOCKSIZE_LOG			||
281 	    sbp->sb_inodesize < XFS_DINODE_MIN_SIZE			||
282 	    sbp->sb_inodesize > XFS_DINODE_MAX_SIZE			||
283 	    sbp->sb_inodelog < XFS_DINODE_MIN_LOG			||
284 	    sbp->sb_inodelog > XFS_DINODE_MAX_LOG			||
285 	    sbp->sb_inodesize != (1 << sbp->sb_inodelog)		||
286 	    sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE			||
287 	    sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
288 	    (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog)	||
289 	    (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE)	||
290 	    (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)	||
291 	    (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */)	||
292 	    sbp->sb_dblocks == 0					||
293 	    sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp)			||
294 	    sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp)			||
295 	    sbp->sb_shared_vn != 0)) {
296 		xfs_notice(mp, "SB sanity check failed");
297 		return -EFSCORRUPTED;
298 	}
299 
300 	/*
301 	 * Until this is fixed only page-sized or smaller data blocks work.
302 	 */
303 	if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
304 		xfs_warn(mp,
305 		"File system with blocksize %d bytes. "
306 		"Only pagesize (%ld) or less will currently work.",
307 				sbp->sb_blocksize, PAGE_SIZE);
308 		return -ENOSYS;
309 	}
310 
311 	/*
312 	 * Currently only very few inode sizes are supported.
313 	 */
314 	switch (sbp->sb_inodesize) {
315 	case 256:
316 	case 512:
317 	case 1024:
318 	case 2048:
319 		break;
320 	default:
321 		xfs_warn(mp, "inode size of %d bytes not supported",
322 				sbp->sb_inodesize);
323 		return -ENOSYS;
324 	}
325 
326 	if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
327 	    xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
328 		xfs_warn(mp,
329 		"file system too large to be mounted on this system.");
330 		return -EFBIG;
331 	}
332 
333 	if (check_inprogress && sbp->sb_inprogress) {
334 		xfs_warn(mp, "Offline file system operation in progress!");
335 		return -EFSCORRUPTED;
336 	}
337 	return 0;
338 }
339 
340 void
341 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
342 {
343 	/*
344 	 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
345 	 * leads to in-core values having two different values for a quota
346 	 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
347 	 * NULLFSINO.
348 	 *
349 	 * Note that this change affect only the in-core values. These
350 	 * values are not written back to disk unless any quota information
351 	 * is written to the disk. Even in that case, sb_pquotino field is
352 	 * not written to disk unless the superblock supports pquotino.
353 	 */
354 	if (sbp->sb_uquotino == 0)
355 		sbp->sb_uquotino = NULLFSINO;
356 	if (sbp->sb_gquotino == 0)
357 		sbp->sb_gquotino = NULLFSINO;
358 	if (sbp->sb_pquotino == 0)
359 		sbp->sb_pquotino = NULLFSINO;
360 
361 	/*
362 	 * We need to do these manipilations only if we are working
363 	 * with an older version of on-disk superblock.
364 	 */
365 	if (xfs_sb_version_has_pquotino(sbp))
366 		return;
367 
368 	if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
369 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
370 					XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
371 	if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
372 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
373 					XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
374 	sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
375 
376 	if (sbp->sb_qflags & XFS_PQUOTA_ACCT)  {
377 		/*
378 		 * In older version of superblock, on-disk superblock only
379 		 * has sb_gquotino, and in-core superblock has both sb_gquotino
380 		 * and sb_pquotino. But, only one of them is supported at any
381 		 * point of time. So, if PQUOTA is set in disk superblock,
382 		 * copy over sb_gquotino to sb_pquotino.
383 		 */
384 		sbp->sb_pquotino = sbp->sb_gquotino;
385 		sbp->sb_gquotino = NULLFSINO;
386 	}
387 }
388 
389 static void
390 __xfs_sb_from_disk(
391 	struct xfs_sb	*to,
392 	xfs_dsb_t	*from,
393 	bool		convert_xquota)
394 {
395 	to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
396 	to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
397 	to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
398 	to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
399 	to->sb_rextents = be64_to_cpu(from->sb_rextents);
400 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
401 	to->sb_logstart = be64_to_cpu(from->sb_logstart);
402 	to->sb_rootino = be64_to_cpu(from->sb_rootino);
403 	to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
404 	to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
405 	to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
406 	to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
407 	to->sb_agcount = be32_to_cpu(from->sb_agcount);
408 	to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
409 	to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
410 	to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
411 	to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
412 	to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
413 	to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
414 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
415 	to->sb_blocklog = from->sb_blocklog;
416 	to->sb_sectlog = from->sb_sectlog;
417 	to->sb_inodelog = from->sb_inodelog;
418 	to->sb_inopblog = from->sb_inopblog;
419 	to->sb_agblklog = from->sb_agblklog;
420 	to->sb_rextslog = from->sb_rextslog;
421 	to->sb_inprogress = from->sb_inprogress;
422 	to->sb_imax_pct = from->sb_imax_pct;
423 	to->sb_icount = be64_to_cpu(from->sb_icount);
424 	to->sb_ifree = be64_to_cpu(from->sb_ifree);
425 	to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
426 	to->sb_frextents = be64_to_cpu(from->sb_frextents);
427 	to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
428 	to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
429 	to->sb_qflags = be16_to_cpu(from->sb_qflags);
430 	to->sb_flags = from->sb_flags;
431 	to->sb_shared_vn = from->sb_shared_vn;
432 	to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
433 	to->sb_unit = be32_to_cpu(from->sb_unit);
434 	to->sb_width = be32_to_cpu(from->sb_width);
435 	to->sb_dirblklog = from->sb_dirblklog;
436 	to->sb_logsectlog = from->sb_logsectlog;
437 	to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
438 	to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
439 	to->sb_features2 = be32_to_cpu(from->sb_features2);
440 	to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
441 	to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
442 	to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
443 	to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
444 	to->sb_features_log_incompat =
445 				be32_to_cpu(from->sb_features_log_incompat);
446 	/* crc is only used on disk, not in memory; just init to 0 here. */
447 	to->sb_crc = 0;
448 	to->sb_pad = 0;
449 	to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
450 	to->sb_lsn = be64_to_cpu(from->sb_lsn);
451 	/* Convert on-disk flags to in-memory flags? */
452 	if (convert_xquota)
453 		xfs_sb_quota_from_disk(to);
454 }
455 
456 void
457 xfs_sb_from_disk(
458 	struct xfs_sb	*to,
459 	xfs_dsb_t	*from)
460 {
461 	__xfs_sb_from_disk(to, from, true);
462 }
463 
464 static inline void
465 xfs_sb_quota_to_disk(
466 	xfs_dsb_t	*to,
467 	xfs_sb_t	*from,
468 	__int64_t	*fields)
469 {
470 	__uint16_t	qflags = from->sb_qflags;
471 
472 	/*
473 	 * We need to do these manipilations only if we are working
474 	 * with an older version of on-disk superblock.
475 	 */
476 	if (xfs_sb_version_has_pquotino(from))
477 		return;
478 
479 	if (*fields & XFS_SB_QFLAGS) {
480 		/*
481 		 * The in-core version of sb_qflags do not have
482 		 * XFS_OQUOTA_* flags, whereas the on-disk version
483 		 * does.  So, convert incore XFS_{PG}QUOTA_* flags
484 		 * to on-disk XFS_OQUOTA_* flags.
485 		 */
486 		qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
487 				XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
488 
489 		if (from->sb_qflags &
490 				(XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
491 			qflags |= XFS_OQUOTA_ENFD;
492 		if (from->sb_qflags &
493 				(XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
494 			qflags |= XFS_OQUOTA_CHKD;
495 		to->sb_qflags = cpu_to_be16(qflags);
496 		*fields &= ~XFS_SB_QFLAGS;
497 	}
498 
499 	/*
500 	 * GQUOTINO and PQUOTINO cannot be used together in versions of
501 	 * superblock that do not have pquotino. from->sb_flags tells us which
502 	 * quota is active and should be copied to disk. If neither are active,
503 	 * make sure we write NULLFSINO to the sb_gquotino field as a quota
504 	 * inode value of "0" is invalid when the XFS_SB_VERSION_QUOTA feature
505 	 * bit is set.
506 	 *
507 	 * Note that we don't need to handle the sb_uquotino or sb_pquotino here
508 	 * as they do not require any translation. Hence the main sb field loop
509 	 * will write them appropriately from the in-core superblock.
510 	 */
511 	if ((*fields & XFS_SB_GQUOTINO) &&
512 				(from->sb_qflags & XFS_GQUOTA_ACCT))
513 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
514 	else if ((*fields & XFS_SB_PQUOTINO) &&
515 				(from->sb_qflags & XFS_PQUOTA_ACCT))
516 		to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
517 	else {
518 		/*
519 		 * We can't rely on just the fields being logged to tell us
520 		 * that it is safe to write NULLFSINO - we should only do that
521 		 * if quotas are not actually enabled. Hence only write
522 		 * NULLFSINO if both in-core quota inodes are NULL.
523 		 */
524 		if (from->sb_gquotino == NULLFSINO &&
525 		    from->sb_pquotino == NULLFSINO)
526 			to->sb_gquotino = cpu_to_be64(NULLFSINO);
527 	}
528 
529 	*fields &= ~(XFS_SB_PQUOTINO | XFS_SB_GQUOTINO);
530 }
531 
532 /*
533  * Copy in core superblock to ondisk one.
534  *
535  * The fields argument is mask of superblock fields to copy.
536  */
537 void
538 xfs_sb_to_disk(
539 	xfs_dsb_t	*to,
540 	xfs_sb_t	*from,
541 	__int64_t	fields)
542 {
543 	xfs_caddr_t	to_ptr = (xfs_caddr_t)to;
544 	xfs_caddr_t	from_ptr = (xfs_caddr_t)from;
545 	xfs_sb_field_t	f;
546 	int		first;
547 	int		size;
548 
549 	ASSERT(fields);
550 	if (!fields)
551 		return;
552 
553 	/* We should never write the crc here, it's updated in the IO path */
554 	fields &= ~XFS_SB_CRC;
555 
556 	xfs_sb_quota_to_disk(to, from, &fields);
557 	while (fields) {
558 		f = (xfs_sb_field_t)xfs_lowbit64((__uint64_t)fields);
559 		first = xfs_sb_info[f].offset;
560 		size = xfs_sb_info[f + 1].offset - first;
561 
562 		ASSERT(xfs_sb_info[f].type == 0 || xfs_sb_info[f].type == 1);
563 
564 		if (size == 1 || xfs_sb_info[f].type == 1) {
565 			memcpy(to_ptr + first, from_ptr + first, size);
566 		} else {
567 			switch (size) {
568 			case 2:
569 				*(__be16 *)(to_ptr + first) =
570 				      cpu_to_be16(*(__u16 *)(from_ptr + first));
571 				break;
572 			case 4:
573 				*(__be32 *)(to_ptr + first) =
574 				      cpu_to_be32(*(__u32 *)(from_ptr + first));
575 				break;
576 			case 8:
577 				*(__be64 *)(to_ptr + first) =
578 				      cpu_to_be64(*(__u64 *)(from_ptr + first));
579 				break;
580 			default:
581 				ASSERT(0);
582 			}
583 		}
584 
585 		fields &= ~(1LL << f);
586 	}
587 }
588 
589 static int
590 xfs_sb_verify(
591 	struct xfs_buf	*bp,
592 	bool		check_version)
593 {
594 	struct xfs_mount *mp = bp->b_target->bt_mount;
595 	struct xfs_sb	sb;
596 
597 	/*
598 	 * Use call variant which doesn't convert quota flags from disk
599 	 * format, because xfs_mount_validate_sb checks the on-disk flags.
600 	 */
601 	__xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
602 
603 	/*
604 	 * Only check the in progress field for the primary superblock as
605 	 * mkfs.xfs doesn't clear it from secondary superblocks.
606 	 */
607 	return xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR,
608 				     check_version);
609 }
610 
611 /*
612  * If the superblock has the CRC feature bit set or the CRC field is non-null,
613  * check that the CRC is valid.  We check the CRC field is non-null because a
614  * single bit error could clear the feature bit and unused parts of the
615  * superblock are supposed to be zero. Hence a non-null crc field indicates that
616  * we've potentially lost a feature bit and we should check it anyway.
617  *
618  * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
619  * last field in V4 secondary superblocks.  So for secondary superblocks,
620  * we are more forgiving, and ignore CRC failures if the primary doesn't
621  * indicate that the fs version is V5.
622  */
623 static void
624 xfs_sb_read_verify(
625 	struct xfs_buf	*bp)
626 {
627 	struct xfs_mount *mp = bp->b_target->bt_mount;
628 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
629 	int		error;
630 
631 	/*
632 	 * open code the version check to avoid needing to convert the entire
633 	 * superblock from disk order just to check the version number
634 	 */
635 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
636 	    (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
637 						XFS_SB_VERSION_5) ||
638 	     dsb->sb_crc != 0)) {
639 
640 		if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
641 			/* Only fail bad secondaries on a known V5 filesystem */
642 			if (bp->b_bn == XFS_SB_DADDR ||
643 			    xfs_sb_version_hascrc(&mp->m_sb)) {
644 				error = -EFSBADCRC;
645 				goto out_error;
646 			}
647 		}
648 	}
649 	error = xfs_sb_verify(bp, true);
650 
651 out_error:
652 	if (error) {
653 		xfs_buf_ioerror(bp, error);
654 		if (error == -EFSCORRUPTED || error == -EFSBADCRC)
655 			xfs_verifier_error(bp);
656 	}
657 }
658 
659 /*
660  * We may be probed for a filesystem match, so we may not want to emit
661  * messages when the superblock buffer is not actually an XFS superblock.
662  * If we find an XFS superblock, then run a normal, noisy mount because we are
663  * really going to mount it and want to know about errors.
664  */
665 static void
666 xfs_sb_quiet_read_verify(
667 	struct xfs_buf	*bp)
668 {
669 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
670 
671 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
672 		/* XFS filesystem, verify noisily! */
673 		xfs_sb_read_verify(bp);
674 		return;
675 	}
676 	/* quietly fail */
677 	xfs_buf_ioerror(bp, -EWRONGFS);
678 }
679 
680 static void
681 xfs_sb_write_verify(
682 	struct xfs_buf		*bp)
683 {
684 	struct xfs_mount	*mp = bp->b_target->bt_mount;
685 	struct xfs_buf_log_item	*bip = bp->b_fspriv;
686 	int			error;
687 
688 	error = xfs_sb_verify(bp, false);
689 	if (error) {
690 		xfs_buf_ioerror(bp, error);
691 		xfs_verifier_error(bp);
692 		return;
693 	}
694 
695 	if (!xfs_sb_version_hascrc(&mp->m_sb))
696 		return;
697 
698 	if (bip)
699 		XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
700 
701 	xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
702 }
703 
704 const struct xfs_buf_ops xfs_sb_buf_ops = {
705 	.verify_read = xfs_sb_read_verify,
706 	.verify_write = xfs_sb_write_verify,
707 };
708 
709 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
710 	.verify_read = xfs_sb_quiet_read_verify,
711 	.verify_write = xfs_sb_write_verify,
712 };
713 
714 /*
715  * xfs_mount_common
716  *
717  * Mount initialization code establishing various mount
718  * fields from the superblock associated with the given
719  * mount structure
720  */
721 void
722 xfs_sb_mount_common(
723 	struct xfs_mount *mp,
724 	struct xfs_sb	*sbp)
725 {
726 	mp->m_agfrotor = mp->m_agirotor = 0;
727 	spin_lock_init(&mp->m_agirotor_lock);
728 	mp->m_maxagi = mp->m_sb.sb_agcount;
729 	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
730 	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
731 	mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
732 	mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
733 	mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
734 	mp->m_blockmask = sbp->sb_blocksize - 1;
735 	mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
736 	mp->m_blockwmask = mp->m_blockwsize - 1;
737 
738 	mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
739 	mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
740 	mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
741 	mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
742 
743 	mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1);
744 	mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0);
745 	mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2;
746 	mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2;
747 
748 	mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
749 	mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
750 	mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
751 	mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
752 
753 	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
754 	mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
755 					sbp->sb_inopblock);
756 	mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
757 }
758 
759 /*
760  * xfs_initialize_perag_data
761  *
762  * Read in each per-ag structure so we can count up the number of
763  * allocated inodes, free inodes and used filesystem blocks as this
764  * information is no longer persistent in the superblock. Once we have
765  * this information, write it into the in-core superblock structure.
766  */
767 int
768 xfs_initialize_perag_data(
769 	struct xfs_mount *mp,
770 	xfs_agnumber_t	agcount)
771 {
772 	xfs_agnumber_t	index;
773 	xfs_perag_t	*pag;
774 	xfs_sb_t	*sbp = &mp->m_sb;
775 	uint64_t	ifree = 0;
776 	uint64_t	ialloc = 0;
777 	uint64_t	bfree = 0;
778 	uint64_t	bfreelst = 0;
779 	uint64_t	btree = 0;
780 	int		error;
781 
782 	for (index = 0; index < agcount; index++) {
783 		/*
784 		 * read the agf, then the agi. This gets us
785 		 * all the information we need and populates the
786 		 * per-ag structures for us.
787 		 */
788 		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
789 		if (error)
790 			return error;
791 
792 		error = xfs_ialloc_pagi_init(mp, NULL, index);
793 		if (error)
794 			return error;
795 		pag = xfs_perag_get(mp, index);
796 		ifree += pag->pagi_freecount;
797 		ialloc += pag->pagi_count;
798 		bfree += pag->pagf_freeblks;
799 		bfreelst += pag->pagf_flcount;
800 		btree += pag->pagf_btreeblks;
801 		xfs_perag_put(pag);
802 	}
803 	/*
804 	 * Overwrite incore superblock counters with just-read data
805 	 */
806 	spin_lock(&mp->m_sb_lock);
807 	sbp->sb_ifree = ifree;
808 	sbp->sb_icount = ialloc;
809 	sbp->sb_fdblocks = bfree + bfreelst + btree;
810 	spin_unlock(&mp->m_sb_lock);
811 
812 	/* Fixup the per-cpu counters as well. */
813 	xfs_icsb_reinit_counters(mp);
814 
815 	return 0;
816 }
817 
818 /*
819  * xfs_mod_sb() can be used to copy arbitrary changes to the
820  * in-core superblock into the superblock buffer to be logged.
821  * It does not provide the higher level of locking that is
822  * needed to protect the in-core superblock from concurrent
823  * access.
824  */
825 void
826 xfs_mod_sb(xfs_trans_t *tp, __int64_t fields)
827 {
828 	xfs_buf_t	*bp;
829 	int		first;
830 	int		last;
831 	xfs_mount_t	*mp;
832 	xfs_sb_field_t	f;
833 
834 	ASSERT(fields);
835 	if (!fields)
836 		return;
837 	mp = tp->t_mountp;
838 	bp = xfs_trans_getsb(tp, mp, 0);
839 	first = sizeof(xfs_sb_t);
840 	last = 0;
841 
842 	/* translate/copy */
843 
844 	xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, fields);
845 
846 	/* find modified range */
847 	f = (xfs_sb_field_t)xfs_highbit64((__uint64_t)fields);
848 	ASSERT((1LL << f) & XFS_SB_MOD_BITS);
849 	last = xfs_sb_info[f + 1].offset - 1;
850 
851 	f = (xfs_sb_field_t)xfs_lowbit64((__uint64_t)fields);
852 	ASSERT((1LL << f) & XFS_SB_MOD_BITS);
853 	first = xfs_sb_info[f].offset;
854 
855 	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
856 	xfs_trans_log_buf(tp, bp, first, last);
857 }
858