xref: /openbmc/linux/fs/xfs/libxfs/xfs_sb.c (revision 4a44a19b)
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_ag.h"
27 #include "xfs_mount.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_dinode.h"
37 #include "xfs_bmap_btree.h"
38 #include "xfs_alloc_btree.h"
39 #include "xfs_ialloc_btree.h"
40 
41 /*
42  * Physical superblock buffer manipulations. Shared with libxfs in userspace.
43  */
44 
45 static const struct {
46 	short offset;
47 	short type;	/* 0 = integer
48 			 * 1 = binary / string (no translation)
49 			 */
50 } xfs_sb_info[] = {
51 	{ offsetof(xfs_sb_t, sb_magicnum),	0 },
52 	{ offsetof(xfs_sb_t, sb_blocksize),	0 },
53 	{ offsetof(xfs_sb_t, sb_dblocks),	0 },
54 	{ offsetof(xfs_sb_t, sb_rblocks),	0 },
55 	{ offsetof(xfs_sb_t, sb_rextents),	0 },
56 	{ offsetof(xfs_sb_t, sb_uuid),		1 },
57 	{ offsetof(xfs_sb_t, sb_logstart),	0 },
58 	{ offsetof(xfs_sb_t, sb_rootino),	0 },
59 	{ offsetof(xfs_sb_t, sb_rbmino),	0 },
60 	{ offsetof(xfs_sb_t, sb_rsumino),	0 },
61 	{ offsetof(xfs_sb_t, sb_rextsize),	0 },
62 	{ offsetof(xfs_sb_t, sb_agblocks),	0 },
63 	{ offsetof(xfs_sb_t, sb_agcount),	0 },
64 	{ offsetof(xfs_sb_t, sb_rbmblocks),	0 },
65 	{ offsetof(xfs_sb_t, sb_logblocks),	0 },
66 	{ offsetof(xfs_sb_t, sb_versionnum),	0 },
67 	{ offsetof(xfs_sb_t, sb_sectsize),	0 },
68 	{ offsetof(xfs_sb_t, sb_inodesize),	0 },
69 	{ offsetof(xfs_sb_t, sb_inopblock),	0 },
70 	{ offsetof(xfs_sb_t, sb_fname[0]),	1 },
71 	{ offsetof(xfs_sb_t, sb_blocklog),	0 },
72 	{ offsetof(xfs_sb_t, sb_sectlog),	0 },
73 	{ offsetof(xfs_sb_t, sb_inodelog),	0 },
74 	{ offsetof(xfs_sb_t, sb_inopblog),	0 },
75 	{ offsetof(xfs_sb_t, sb_agblklog),	0 },
76 	{ offsetof(xfs_sb_t, sb_rextslog),	0 },
77 	{ offsetof(xfs_sb_t, sb_inprogress),	0 },
78 	{ offsetof(xfs_sb_t, sb_imax_pct),	0 },
79 	{ offsetof(xfs_sb_t, sb_icount),	0 },
80 	{ offsetof(xfs_sb_t, sb_ifree),		0 },
81 	{ offsetof(xfs_sb_t, sb_fdblocks),	0 },
82 	{ offsetof(xfs_sb_t, sb_frextents),	0 },
83 	{ offsetof(xfs_sb_t, sb_uquotino),	0 },
84 	{ offsetof(xfs_sb_t, sb_gquotino),	0 },
85 	{ offsetof(xfs_sb_t, sb_qflags),	0 },
86 	{ offsetof(xfs_sb_t, sb_flags),		0 },
87 	{ offsetof(xfs_sb_t, sb_shared_vn),	0 },
88 	{ offsetof(xfs_sb_t, sb_inoalignmt),	0 },
89 	{ offsetof(xfs_sb_t, sb_unit),		0 },
90 	{ offsetof(xfs_sb_t, sb_width),		0 },
91 	{ offsetof(xfs_sb_t, sb_dirblklog),	0 },
92 	{ offsetof(xfs_sb_t, sb_logsectlog),	0 },
93 	{ offsetof(xfs_sb_t, sb_logsectsize),	0 },
94 	{ offsetof(xfs_sb_t, sb_logsunit),	0 },
95 	{ offsetof(xfs_sb_t, sb_features2),	0 },
96 	{ offsetof(xfs_sb_t, sb_bad_features2),	0 },
97 	{ offsetof(xfs_sb_t, sb_features_compat),	0 },
98 	{ offsetof(xfs_sb_t, sb_features_ro_compat),	0 },
99 	{ offsetof(xfs_sb_t, sb_features_incompat),	0 },
100 	{ offsetof(xfs_sb_t, sb_features_log_incompat),	0 },
101 	{ offsetof(xfs_sb_t, sb_crc),		0 },
102 	{ offsetof(xfs_sb_t, sb_pad),		0 },
103 	{ offsetof(xfs_sb_t, sb_pquotino),	0 },
104 	{ offsetof(xfs_sb_t, sb_lsn),		0 },
105 	{ sizeof(xfs_sb_t),			0 }
106 };
107 
108 /*
109  * Reference counting access wrappers to the perag structures.
110  * Because we never free per-ag structures, the only thing we
111  * have to protect against changes is the tree structure itself.
112  */
113 struct xfs_perag *
114 xfs_perag_get(
115 	struct xfs_mount	*mp,
116 	xfs_agnumber_t		agno)
117 {
118 	struct xfs_perag	*pag;
119 	int			ref = 0;
120 
121 	rcu_read_lock();
122 	pag = radix_tree_lookup(&mp->m_perag_tree, agno);
123 	if (pag) {
124 		ASSERT(atomic_read(&pag->pag_ref) >= 0);
125 		ref = atomic_inc_return(&pag->pag_ref);
126 	}
127 	rcu_read_unlock();
128 	trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
129 	return pag;
130 }
131 
132 /*
133  * search from @first to find the next perag with the given tag set.
134  */
135 struct xfs_perag *
136 xfs_perag_get_tag(
137 	struct xfs_mount	*mp,
138 	xfs_agnumber_t		first,
139 	int			tag)
140 {
141 	struct xfs_perag	*pag;
142 	int			found;
143 	int			ref;
144 
145 	rcu_read_lock();
146 	found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
147 					(void **)&pag, first, 1, tag);
148 	if (found <= 0) {
149 		rcu_read_unlock();
150 		return NULL;
151 	}
152 	ref = atomic_inc_return(&pag->pag_ref);
153 	rcu_read_unlock();
154 	trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
155 	return pag;
156 }
157 
158 void
159 xfs_perag_put(
160 	struct xfs_perag	*pag)
161 {
162 	int	ref;
163 
164 	ASSERT(atomic_read(&pag->pag_ref) > 0);
165 	ref = atomic_dec_return(&pag->pag_ref);
166 	trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
167 }
168 
169 /*
170  * Check the validity of the SB found.
171  */
172 STATIC int
173 xfs_mount_validate_sb(
174 	xfs_mount_t	*mp,
175 	xfs_sb_t	*sbp,
176 	bool		check_inprogress,
177 	bool		check_version)
178 {
179 
180 	/*
181 	 * If the log device and data device have the
182 	 * same device number, the log is internal.
183 	 * Consequently, the sb_logstart should be non-zero.  If
184 	 * we have a zero sb_logstart in this case, we may be trying to mount
185 	 * a volume filesystem in a non-volume manner.
186 	 */
187 	if (sbp->sb_magicnum != XFS_SB_MAGIC) {
188 		xfs_warn(mp, "bad magic number");
189 		return -EWRONGFS;
190 	}
191 
192 
193 	if (!xfs_sb_good_version(sbp)) {
194 		xfs_warn(mp, "bad version");
195 		return -EWRONGFS;
196 	}
197 
198 	/*
199 	 * Version 5 superblock feature mask validation. Reject combinations the
200 	 * kernel cannot support up front before checking anything else. For
201 	 * write validation, we don't need to check feature masks.
202 	 */
203 	if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) {
204 		if (xfs_sb_has_compat_feature(sbp,
205 					XFS_SB_FEAT_COMPAT_UNKNOWN)) {
206 			xfs_warn(mp,
207 "Superblock has unknown compatible features (0x%x) enabled.\n"
208 "Using a more recent kernel is recommended.",
209 				(sbp->sb_features_compat &
210 						XFS_SB_FEAT_COMPAT_UNKNOWN));
211 		}
212 
213 		if (xfs_sb_has_ro_compat_feature(sbp,
214 					XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
215 			xfs_alert(mp,
216 "Superblock has unknown read-only compatible features (0x%x) enabled.",
217 				(sbp->sb_features_ro_compat &
218 						XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
219 			if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
220 				xfs_warn(mp,
221 "Attempted to mount read-only compatible filesystem read-write.\n"
222 "Filesystem can only be safely mounted read only.");
223 				return -EINVAL;
224 			}
225 		}
226 		if (xfs_sb_has_incompat_feature(sbp,
227 					XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
228 			xfs_warn(mp,
229 "Superblock has unknown incompatible features (0x%x) enabled.\n"
230 "Filesystem can not be safely mounted by this kernel.",
231 				(sbp->sb_features_incompat &
232 						XFS_SB_FEAT_INCOMPAT_UNKNOWN));
233 			return -EINVAL;
234 		}
235 	}
236 
237 	if (xfs_sb_version_has_pquotino(sbp)) {
238 		if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
239 			xfs_notice(mp,
240 			   "Version 5 of Super block has XFS_OQUOTA bits.");
241 			return -EFSCORRUPTED;
242 		}
243 	} else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
244 				XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
245 			xfs_notice(mp,
246 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
247 			return -EFSCORRUPTED;
248 	}
249 
250 	if (unlikely(
251 	    sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
252 		xfs_warn(mp,
253 		"filesystem is marked as having an external log; "
254 		"specify logdev on the mount command line.");
255 		return -EINVAL;
256 	}
257 
258 	if (unlikely(
259 	    sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
260 		xfs_warn(mp,
261 		"filesystem is marked as having an internal log; "
262 		"do not specify logdev on the mount command line.");
263 		return -EINVAL;
264 	}
265 
266 	/*
267 	 * More sanity checking.  Most of these were stolen directly from
268 	 * xfs_repair.
269 	 */
270 	if (unlikely(
271 	    sbp->sb_agcount <= 0					||
272 	    sbp->sb_sectsize < XFS_MIN_SECTORSIZE			||
273 	    sbp->sb_sectsize > XFS_MAX_SECTORSIZE			||
274 	    sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG			||
275 	    sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG			||
276 	    sbp->sb_sectsize != (1 << sbp->sb_sectlog)			||
277 	    sbp->sb_blocksize < XFS_MIN_BLOCKSIZE			||
278 	    sbp->sb_blocksize > XFS_MAX_BLOCKSIZE			||
279 	    sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG			||
280 	    sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG			||
281 	    sbp->sb_blocksize != (1 << sbp->sb_blocklog)		||
282 	    sbp->sb_dirblklog > XFS_MAX_BLOCKSIZE_LOG			||
283 	    sbp->sb_inodesize < XFS_DINODE_MIN_SIZE			||
284 	    sbp->sb_inodesize > XFS_DINODE_MAX_SIZE			||
285 	    sbp->sb_inodelog < XFS_DINODE_MIN_LOG			||
286 	    sbp->sb_inodelog > XFS_DINODE_MAX_LOG			||
287 	    sbp->sb_inodesize != (1 << sbp->sb_inodelog)		||
288 	    sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE			||
289 	    sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
290 	    (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog)	||
291 	    (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE)	||
292 	    (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)	||
293 	    (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */)	||
294 	    sbp->sb_dblocks == 0					||
295 	    sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp)			||
296 	    sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp)			||
297 	    sbp->sb_shared_vn != 0)) {
298 		xfs_notice(mp, "SB sanity check failed");
299 		return -EFSCORRUPTED;
300 	}
301 
302 	/*
303 	 * Until this is fixed only page-sized or smaller data blocks work.
304 	 */
305 	if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
306 		xfs_warn(mp,
307 		"File system with blocksize %d bytes. "
308 		"Only pagesize (%ld) or less will currently work.",
309 				sbp->sb_blocksize, PAGE_SIZE);
310 		return -ENOSYS;
311 	}
312 
313 	/*
314 	 * Currently only very few inode sizes are supported.
315 	 */
316 	switch (sbp->sb_inodesize) {
317 	case 256:
318 	case 512:
319 	case 1024:
320 	case 2048:
321 		break;
322 	default:
323 		xfs_warn(mp, "inode size of %d bytes not supported",
324 				sbp->sb_inodesize);
325 		return -ENOSYS;
326 	}
327 
328 	if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
329 	    xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
330 		xfs_warn(mp,
331 		"file system too large to be mounted on this system.");
332 		return -EFBIG;
333 	}
334 
335 	if (check_inprogress && sbp->sb_inprogress) {
336 		xfs_warn(mp, "Offline file system operation in progress!");
337 		return -EFSCORRUPTED;
338 	}
339 	return 0;
340 }
341 
342 void
343 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
344 {
345 	/*
346 	 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
347 	 * leads to in-core values having two different values for a quota
348 	 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
349 	 * NULLFSINO.
350 	 *
351 	 * Note that this change affect only the in-core values. These
352 	 * values are not written back to disk unless any quota information
353 	 * is written to the disk. Even in that case, sb_pquotino field is
354 	 * not written to disk unless the superblock supports pquotino.
355 	 */
356 	if (sbp->sb_uquotino == 0)
357 		sbp->sb_uquotino = NULLFSINO;
358 	if (sbp->sb_gquotino == 0)
359 		sbp->sb_gquotino = NULLFSINO;
360 	if (sbp->sb_pquotino == 0)
361 		sbp->sb_pquotino = NULLFSINO;
362 
363 	/*
364 	 * We need to do these manipilations only if we are working
365 	 * with an older version of on-disk superblock.
366 	 */
367 	if (xfs_sb_version_has_pquotino(sbp))
368 		return;
369 
370 	if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
371 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
372 					XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
373 	if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
374 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
375 					XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
376 	sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
377 
378 	if (sbp->sb_qflags & XFS_PQUOTA_ACCT)  {
379 		/*
380 		 * In older version of superblock, on-disk superblock only
381 		 * has sb_gquotino, and in-core superblock has both sb_gquotino
382 		 * and sb_pquotino. But, only one of them is supported at any
383 		 * point of time. So, if PQUOTA is set in disk superblock,
384 		 * copy over sb_gquotino to sb_pquotino.
385 		 */
386 		sbp->sb_pquotino = sbp->sb_gquotino;
387 		sbp->sb_gquotino = NULLFSINO;
388 	}
389 }
390 
391 static void
392 __xfs_sb_from_disk(
393 	struct xfs_sb	*to,
394 	xfs_dsb_t	*from,
395 	bool		convert_xquota)
396 {
397 	to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
398 	to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
399 	to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
400 	to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
401 	to->sb_rextents = be64_to_cpu(from->sb_rextents);
402 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
403 	to->sb_logstart = be64_to_cpu(from->sb_logstart);
404 	to->sb_rootino = be64_to_cpu(from->sb_rootino);
405 	to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
406 	to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
407 	to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
408 	to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
409 	to->sb_agcount = be32_to_cpu(from->sb_agcount);
410 	to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
411 	to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
412 	to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
413 	to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
414 	to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
415 	to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
416 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
417 	to->sb_blocklog = from->sb_blocklog;
418 	to->sb_sectlog = from->sb_sectlog;
419 	to->sb_inodelog = from->sb_inodelog;
420 	to->sb_inopblog = from->sb_inopblog;
421 	to->sb_agblklog = from->sb_agblklog;
422 	to->sb_rextslog = from->sb_rextslog;
423 	to->sb_inprogress = from->sb_inprogress;
424 	to->sb_imax_pct = from->sb_imax_pct;
425 	to->sb_icount = be64_to_cpu(from->sb_icount);
426 	to->sb_ifree = be64_to_cpu(from->sb_ifree);
427 	to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
428 	to->sb_frextents = be64_to_cpu(from->sb_frextents);
429 	to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
430 	to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
431 	to->sb_qflags = be16_to_cpu(from->sb_qflags);
432 	to->sb_flags = from->sb_flags;
433 	to->sb_shared_vn = from->sb_shared_vn;
434 	to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
435 	to->sb_unit = be32_to_cpu(from->sb_unit);
436 	to->sb_width = be32_to_cpu(from->sb_width);
437 	to->sb_dirblklog = from->sb_dirblklog;
438 	to->sb_logsectlog = from->sb_logsectlog;
439 	to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
440 	to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
441 	to->sb_features2 = be32_to_cpu(from->sb_features2);
442 	to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
443 	to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
444 	to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
445 	to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
446 	to->sb_features_log_incompat =
447 				be32_to_cpu(from->sb_features_log_incompat);
448 	/* crc is only used on disk, not in memory; just init to 0 here. */
449 	to->sb_crc = 0;
450 	to->sb_pad = 0;
451 	to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
452 	to->sb_lsn = be64_to_cpu(from->sb_lsn);
453 	/* Convert on-disk flags to in-memory flags? */
454 	if (convert_xquota)
455 		xfs_sb_quota_from_disk(to);
456 }
457 
458 void
459 xfs_sb_from_disk(
460 	struct xfs_sb	*to,
461 	xfs_dsb_t	*from)
462 {
463 	__xfs_sb_from_disk(to, from, true);
464 }
465 
466 static inline void
467 xfs_sb_quota_to_disk(
468 	xfs_dsb_t	*to,
469 	xfs_sb_t	*from,
470 	__int64_t	*fields)
471 {
472 	__uint16_t	qflags = from->sb_qflags;
473 
474 	/*
475 	 * We need to do these manipilations only if we are working
476 	 * with an older version of on-disk superblock.
477 	 */
478 	if (xfs_sb_version_has_pquotino(from))
479 		return;
480 
481 	if (*fields & XFS_SB_QFLAGS) {
482 		/*
483 		 * The in-core version of sb_qflags do not have
484 		 * XFS_OQUOTA_* flags, whereas the on-disk version
485 		 * does.  So, convert incore XFS_{PG}QUOTA_* flags
486 		 * to on-disk XFS_OQUOTA_* flags.
487 		 */
488 		qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
489 				XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
490 
491 		if (from->sb_qflags &
492 				(XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
493 			qflags |= XFS_OQUOTA_ENFD;
494 		if (from->sb_qflags &
495 				(XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
496 			qflags |= XFS_OQUOTA_CHKD;
497 		to->sb_qflags = cpu_to_be16(qflags);
498 		*fields &= ~XFS_SB_QFLAGS;
499 	}
500 
501 	/*
502 	 * GQUOTINO and PQUOTINO cannot be used together in versions of
503 	 * superblock that do not have pquotino. from->sb_flags tells us which
504 	 * quota is active and should be copied to disk. If neither are active,
505 	 * make sure we write NULLFSINO to the sb_gquotino field as a quota
506 	 * inode value of "0" is invalid when the XFS_SB_VERSION_QUOTA feature
507 	 * bit is set.
508 	 *
509 	 * Note that we don't need to handle the sb_uquotino or sb_pquotino here
510 	 * as they do not require any translation. Hence the main sb field loop
511 	 * will write them appropriately from the in-core superblock.
512 	 */
513 	if ((*fields & XFS_SB_GQUOTINO) &&
514 				(from->sb_qflags & XFS_GQUOTA_ACCT))
515 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
516 	else if ((*fields & XFS_SB_PQUOTINO) &&
517 				(from->sb_qflags & XFS_PQUOTA_ACCT))
518 		to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
519 	else {
520 		/*
521 		 * We can't rely on just the fields being logged to tell us
522 		 * that it is safe to write NULLFSINO - we should only do that
523 		 * if quotas are not actually enabled. Hence only write
524 		 * NULLFSINO if both in-core quota inodes are NULL.
525 		 */
526 		if (from->sb_gquotino == NULLFSINO &&
527 		    from->sb_pquotino == NULLFSINO)
528 			to->sb_gquotino = cpu_to_be64(NULLFSINO);
529 	}
530 
531 	*fields &= ~(XFS_SB_PQUOTINO | XFS_SB_GQUOTINO);
532 }
533 
534 /*
535  * Copy in core superblock to ondisk one.
536  *
537  * The fields argument is mask of superblock fields to copy.
538  */
539 void
540 xfs_sb_to_disk(
541 	xfs_dsb_t	*to,
542 	xfs_sb_t	*from,
543 	__int64_t	fields)
544 {
545 	xfs_caddr_t	to_ptr = (xfs_caddr_t)to;
546 	xfs_caddr_t	from_ptr = (xfs_caddr_t)from;
547 	xfs_sb_field_t	f;
548 	int		first;
549 	int		size;
550 
551 	ASSERT(fields);
552 	if (!fields)
553 		return;
554 
555 	/* We should never write the crc here, it's updated in the IO path */
556 	fields &= ~XFS_SB_CRC;
557 
558 	xfs_sb_quota_to_disk(to, from, &fields);
559 	while (fields) {
560 		f = (xfs_sb_field_t)xfs_lowbit64((__uint64_t)fields);
561 		first = xfs_sb_info[f].offset;
562 		size = xfs_sb_info[f + 1].offset - first;
563 
564 		ASSERT(xfs_sb_info[f].type == 0 || xfs_sb_info[f].type == 1);
565 
566 		if (size == 1 || xfs_sb_info[f].type == 1) {
567 			memcpy(to_ptr + first, from_ptr + first, size);
568 		} else {
569 			switch (size) {
570 			case 2:
571 				*(__be16 *)(to_ptr + first) =
572 				      cpu_to_be16(*(__u16 *)(from_ptr + first));
573 				break;
574 			case 4:
575 				*(__be32 *)(to_ptr + first) =
576 				      cpu_to_be32(*(__u32 *)(from_ptr + first));
577 				break;
578 			case 8:
579 				*(__be64 *)(to_ptr + first) =
580 				      cpu_to_be64(*(__u64 *)(from_ptr + first));
581 				break;
582 			default:
583 				ASSERT(0);
584 			}
585 		}
586 
587 		fields &= ~(1LL << f);
588 	}
589 }
590 
591 static int
592 xfs_sb_verify(
593 	struct xfs_buf	*bp,
594 	bool		check_version)
595 {
596 	struct xfs_mount *mp = bp->b_target->bt_mount;
597 	struct xfs_sb	sb;
598 
599 	/*
600 	 * Use call variant which doesn't convert quota flags from disk
601 	 * format, because xfs_mount_validate_sb checks the on-disk flags.
602 	 */
603 	__xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
604 
605 	/*
606 	 * Only check the in progress field for the primary superblock as
607 	 * mkfs.xfs doesn't clear it from secondary superblocks.
608 	 */
609 	return xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR,
610 				     check_version);
611 }
612 
613 /*
614  * If the superblock has the CRC feature bit set or the CRC field is non-null,
615  * check that the CRC is valid.  We check the CRC field is non-null because a
616  * single bit error could clear the feature bit and unused parts of the
617  * superblock are supposed to be zero. Hence a non-null crc field indicates that
618  * we've potentially lost a feature bit and we should check it anyway.
619  *
620  * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
621  * last field in V4 secondary superblocks.  So for secondary superblocks,
622  * we are more forgiving, and ignore CRC failures if the primary doesn't
623  * indicate that the fs version is V5.
624  */
625 static void
626 xfs_sb_read_verify(
627 	struct xfs_buf	*bp)
628 {
629 	struct xfs_mount *mp = bp->b_target->bt_mount;
630 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
631 	int		error;
632 
633 	/*
634 	 * open code the version check to avoid needing to convert the entire
635 	 * superblock from disk order just to check the version number
636 	 */
637 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
638 	    (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
639 						XFS_SB_VERSION_5) ||
640 	     dsb->sb_crc != 0)) {
641 
642 		if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
643 			/* Only fail bad secondaries on a known V5 filesystem */
644 			if (bp->b_bn == XFS_SB_DADDR ||
645 			    xfs_sb_version_hascrc(&mp->m_sb)) {
646 				error = -EFSBADCRC;
647 				goto out_error;
648 			}
649 		}
650 	}
651 	error = xfs_sb_verify(bp, true);
652 
653 out_error:
654 	if (error) {
655 		xfs_buf_ioerror(bp, error);
656 		if (error == -EFSCORRUPTED || error == -EFSBADCRC)
657 			xfs_verifier_error(bp);
658 	}
659 }
660 
661 /*
662  * We may be probed for a filesystem match, so we may not want to emit
663  * messages when the superblock buffer is not actually an XFS superblock.
664  * If we find an XFS superblock, then run a normal, noisy mount because we are
665  * really going to mount it and want to know about errors.
666  */
667 static void
668 xfs_sb_quiet_read_verify(
669 	struct xfs_buf	*bp)
670 {
671 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
672 
673 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
674 		/* XFS filesystem, verify noisily! */
675 		xfs_sb_read_verify(bp);
676 		return;
677 	}
678 	/* quietly fail */
679 	xfs_buf_ioerror(bp, -EWRONGFS);
680 }
681 
682 static void
683 xfs_sb_write_verify(
684 	struct xfs_buf		*bp)
685 {
686 	struct xfs_mount	*mp = bp->b_target->bt_mount;
687 	struct xfs_buf_log_item	*bip = bp->b_fspriv;
688 	int			error;
689 
690 	error = xfs_sb_verify(bp, false);
691 	if (error) {
692 		xfs_buf_ioerror(bp, error);
693 		xfs_verifier_error(bp);
694 		return;
695 	}
696 
697 	if (!xfs_sb_version_hascrc(&mp->m_sb))
698 		return;
699 
700 	if (bip)
701 		XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
702 
703 	xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
704 }
705 
706 const struct xfs_buf_ops xfs_sb_buf_ops = {
707 	.verify_read = xfs_sb_read_verify,
708 	.verify_write = xfs_sb_write_verify,
709 };
710 
711 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
712 	.verify_read = xfs_sb_quiet_read_verify,
713 	.verify_write = xfs_sb_write_verify,
714 };
715 
716 /*
717  * xfs_mount_common
718  *
719  * Mount initialization code establishing various mount
720  * fields from the superblock associated with the given
721  * mount structure
722  */
723 void
724 xfs_sb_mount_common(
725 	struct xfs_mount *mp,
726 	struct xfs_sb	*sbp)
727 {
728 	mp->m_agfrotor = mp->m_agirotor = 0;
729 	spin_lock_init(&mp->m_agirotor_lock);
730 	mp->m_maxagi = mp->m_sb.sb_agcount;
731 	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
732 	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
733 	mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
734 	mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
735 	mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
736 	mp->m_blockmask = sbp->sb_blocksize - 1;
737 	mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
738 	mp->m_blockwmask = mp->m_blockwsize - 1;
739 
740 	mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
741 	mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
742 	mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
743 	mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
744 
745 	mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1);
746 	mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0);
747 	mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2;
748 	mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2;
749 
750 	mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
751 	mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
752 	mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
753 	mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
754 
755 	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
756 	mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
757 					sbp->sb_inopblock);
758 	mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
759 }
760 
761 /*
762  * xfs_initialize_perag_data
763  *
764  * Read in each per-ag structure so we can count up the number of
765  * allocated inodes, free inodes and used filesystem blocks as this
766  * information is no longer persistent in the superblock. Once we have
767  * this information, write it into the in-core superblock structure.
768  */
769 int
770 xfs_initialize_perag_data(
771 	struct xfs_mount *mp,
772 	xfs_agnumber_t	agcount)
773 {
774 	xfs_agnumber_t	index;
775 	xfs_perag_t	*pag;
776 	xfs_sb_t	*sbp = &mp->m_sb;
777 	uint64_t	ifree = 0;
778 	uint64_t	ialloc = 0;
779 	uint64_t	bfree = 0;
780 	uint64_t	bfreelst = 0;
781 	uint64_t	btree = 0;
782 	int		error;
783 
784 	for (index = 0; index < agcount; index++) {
785 		/*
786 		 * read the agf, then the agi. This gets us
787 		 * all the information we need and populates the
788 		 * per-ag structures for us.
789 		 */
790 		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
791 		if (error)
792 			return error;
793 
794 		error = xfs_ialloc_pagi_init(mp, NULL, index);
795 		if (error)
796 			return error;
797 		pag = xfs_perag_get(mp, index);
798 		ifree += pag->pagi_freecount;
799 		ialloc += pag->pagi_count;
800 		bfree += pag->pagf_freeblks;
801 		bfreelst += pag->pagf_flcount;
802 		btree += pag->pagf_btreeblks;
803 		xfs_perag_put(pag);
804 	}
805 	/*
806 	 * Overwrite incore superblock counters with just-read data
807 	 */
808 	spin_lock(&mp->m_sb_lock);
809 	sbp->sb_ifree = ifree;
810 	sbp->sb_icount = ialloc;
811 	sbp->sb_fdblocks = bfree + bfreelst + btree;
812 	spin_unlock(&mp->m_sb_lock);
813 
814 	/* Fixup the per-cpu counters as well. */
815 	xfs_icsb_reinit_counters(mp);
816 
817 	return 0;
818 }
819 
820 /*
821  * xfs_mod_sb() can be used to copy arbitrary changes to the
822  * in-core superblock into the superblock buffer to be logged.
823  * It does not provide the higher level of locking that is
824  * needed to protect the in-core superblock from concurrent
825  * access.
826  */
827 void
828 xfs_mod_sb(xfs_trans_t *tp, __int64_t fields)
829 {
830 	xfs_buf_t	*bp;
831 	int		first;
832 	int		last;
833 	xfs_mount_t	*mp;
834 	xfs_sb_field_t	f;
835 
836 	ASSERT(fields);
837 	if (!fields)
838 		return;
839 	mp = tp->t_mountp;
840 	bp = xfs_trans_getsb(tp, mp, 0);
841 	first = sizeof(xfs_sb_t);
842 	last = 0;
843 
844 	/* translate/copy */
845 
846 	xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, fields);
847 
848 	/* find modified range */
849 	f = (xfs_sb_field_t)xfs_highbit64((__uint64_t)fields);
850 	ASSERT((1LL << f) & XFS_SB_MOD_BITS);
851 	last = xfs_sb_info[f + 1].offset - 1;
852 
853 	f = (xfs_sb_field_t)xfs_lowbit64((__uint64_t)fields);
854 	ASSERT((1LL << f) & XFS_SB_MOD_BITS);
855 	first = xfs_sb_info[f].offset;
856 
857 	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
858 	xfs_trans_log_buf(tp, bp, first, last);
859 }
860