xref: /openbmc/linux/fs/xfs/libxfs/xfs_sb.c (revision e3b9f1e8)
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 #include "xfs_bmap.h"
42 #include "xfs_refcount_btree.h"
43 #include "xfs_da_format.h"
44 #include "xfs_da_btree.h"
45 
46 /*
47  * Physical superblock buffer manipulations. Shared with libxfs in userspace.
48  */
49 
50 /*
51  * Reference counting access wrappers to the perag structures.
52  * Because we never free per-ag structures, the only thing we
53  * have to protect against changes is the tree structure itself.
54  */
55 struct xfs_perag *
56 xfs_perag_get(
57 	struct xfs_mount	*mp,
58 	xfs_agnumber_t		agno)
59 {
60 	struct xfs_perag	*pag;
61 	int			ref = 0;
62 
63 	rcu_read_lock();
64 	pag = radix_tree_lookup(&mp->m_perag_tree, agno);
65 	if (pag) {
66 		ASSERT(atomic_read(&pag->pag_ref) >= 0);
67 		ref = atomic_inc_return(&pag->pag_ref);
68 	}
69 	rcu_read_unlock();
70 	trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
71 	return pag;
72 }
73 
74 /*
75  * search from @first to find the next perag with the given tag set.
76  */
77 struct xfs_perag *
78 xfs_perag_get_tag(
79 	struct xfs_mount	*mp,
80 	xfs_agnumber_t		first,
81 	int			tag)
82 {
83 	struct xfs_perag	*pag;
84 	int			found;
85 	int			ref;
86 
87 	rcu_read_lock();
88 	found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
89 					(void **)&pag, first, 1, tag);
90 	if (found <= 0) {
91 		rcu_read_unlock();
92 		return NULL;
93 	}
94 	ref = atomic_inc_return(&pag->pag_ref);
95 	rcu_read_unlock();
96 	trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
97 	return pag;
98 }
99 
100 void
101 xfs_perag_put(
102 	struct xfs_perag	*pag)
103 {
104 	int	ref;
105 
106 	ASSERT(atomic_read(&pag->pag_ref) > 0);
107 	ref = atomic_dec_return(&pag->pag_ref);
108 	trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
109 }
110 
111 /*
112  * Check the validity of the SB found.
113  */
114 STATIC int
115 xfs_mount_validate_sb(
116 	xfs_mount_t	*mp,
117 	xfs_sb_t	*sbp,
118 	bool		check_inprogress,
119 	bool		check_version)
120 {
121 	uint32_t	agcount = 0;
122 	uint32_t	rem;
123 
124 	if (sbp->sb_magicnum != XFS_SB_MAGIC) {
125 		xfs_warn(mp, "bad magic number");
126 		return -EWRONGFS;
127 	}
128 
129 
130 	if (!xfs_sb_good_version(sbp)) {
131 		xfs_warn(mp, "bad version");
132 		return -EWRONGFS;
133 	}
134 
135 	/*
136 	 * Version 5 superblock feature mask validation. Reject combinations the
137 	 * kernel cannot support up front before checking anything else. For
138 	 * write validation, we don't need to check feature masks.
139 	 */
140 	if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) {
141 		if (xfs_sb_has_compat_feature(sbp,
142 					XFS_SB_FEAT_COMPAT_UNKNOWN)) {
143 			xfs_warn(mp,
144 "Superblock has unknown compatible features (0x%x) enabled.",
145 				(sbp->sb_features_compat &
146 						XFS_SB_FEAT_COMPAT_UNKNOWN));
147 			xfs_warn(mp,
148 "Using a more recent kernel is recommended.");
149 		}
150 
151 		if (xfs_sb_has_ro_compat_feature(sbp,
152 					XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
153 			xfs_alert(mp,
154 "Superblock has unknown read-only compatible features (0x%x) enabled.",
155 				(sbp->sb_features_ro_compat &
156 						XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
157 			if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
158 				xfs_warn(mp,
159 "Attempted to mount read-only compatible filesystem read-write.");
160 				xfs_warn(mp,
161 "Filesystem can only be safely mounted read only.");
162 
163 				return -EINVAL;
164 			}
165 		}
166 		if (xfs_sb_has_incompat_feature(sbp,
167 					XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
168 			xfs_warn(mp,
169 "Superblock has unknown incompatible features (0x%x) enabled.",
170 				(sbp->sb_features_incompat &
171 						XFS_SB_FEAT_INCOMPAT_UNKNOWN));
172 			xfs_warn(mp,
173 "Filesystem can not be safely mounted by this kernel.");
174 			return -EINVAL;
175 		}
176 	} else if (xfs_sb_version_hascrc(sbp)) {
177 		/*
178 		 * We can't read verify the sb LSN because the read verifier is
179 		 * called before the log is allocated and processed. We know the
180 		 * log is set up before write verifier (!check_version) calls,
181 		 * so just check it here.
182 		 */
183 		if (!xfs_log_check_lsn(mp, sbp->sb_lsn))
184 			return -EFSCORRUPTED;
185 	}
186 
187 	if (xfs_sb_version_has_pquotino(sbp)) {
188 		if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
189 			xfs_notice(mp,
190 			   "Version 5 of Super block has XFS_OQUOTA bits.");
191 			return -EFSCORRUPTED;
192 		}
193 	} else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
194 				XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
195 			xfs_notice(mp,
196 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
197 			return -EFSCORRUPTED;
198 	}
199 
200 	/*
201 	 * Full inode chunks must be aligned to inode chunk size when
202 	 * sparse inodes are enabled to support the sparse chunk
203 	 * allocation algorithm and prevent overlapping inode records.
204 	 */
205 	if (xfs_sb_version_hassparseinodes(sbp)) {
206 		uint32_t	align;
207 
208 		align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
209 				>> sbp->sb_blocklog;
210 		if (sbp->sb_inoalignmt != align) {
211 			xfs_warn(mp,
212 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
213 				 sbp->sb_inoalignmt, align);
214 			return -EINVAL;
215 		}
216 	}
217 
218 	if (unlikely(
219 	    sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
220 		xfs_warn(mp,
221 		"filesystem is marked as having an external log; "
222 		"specify logdev on the mount command line.");
223 		return -EINVAL;
224 	}
225 
226 	if (unlikely(
227 	    sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
228 		xfs_warn(mp,
229 		"filesystem is marked as having an internal log; "
230 		"do not specify logdev on the mount command line.");
231 		return -EINVAL;
232 	}
233 
234 	/* Compute agcount for this number of dblocks and agblocks */
235 	if (sbp->sb_agblocks) {
236 		agcount = div_u64_rem(sbp->sb_dblocks, sbp->sb_agblocks, &rem);
237 		if (rem)
238 			agcount++;
239 	}
240 
241 	/*
242 	 * More sanity checking.  Most of these were stolen directly from
243 	 * xfs_repair.
244 	 */
245 	if (unlikely(
246 	    sbp->sb_agcount <= 0					||
247 	    sbp->sb_sectsize < XFS_MIN_SECTORSIZE			||
248 	    sbp->sb_sectsize > XFS_MAX_SECTORSIZE			||
249 	    sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG			||
250 	    sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG			||
251 	    sbp->sb_sectsize != (1 << sbp->sb_sectlog)			||
252 	    sbp->sb_blocksize < XFS_MIN_BLOCKSIZE			||
253 	    sbp->sb_blocksize > XFS_MAX_BLOCKSIZE			||
254 	    sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG			||
255 	    sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG			||
256 	    sbp->sb_blocksize != (1 << sbp->sb_blocklog)		||
257 	    sbp->sb_dirblklog + sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
258 	    sbp->sb_inodesize < XFS_DINODE_MIN_SIZE			||
259 	    sbp->sb_inodesize > XFS_DINODE_MAX_SIZE			||
260 	    sbp->sb_inodelog < XFS_DINODE_MIN_LOG			||
261 	    sbp->sb_inodelog > XFS_DINODE_MAX_LOG			||
262 	    sbp->sb_inodesize != (1 << sbp->sb_inodelog)		||
263 	    sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE			||
264 	    sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
265 	    XFS_FSB_TO_B(mp, sbp->sb_agblocks) < XFS_MIN_AG_BYTES	||
266 	    XFS_FSB_TO_B(mp, sbp->sb_agblocks) > XFS_MAX_AG_BYTES	||
267 	    sbp->sb_agblklog != xfs_highbit32(sbp->sb_agblocks - 1) + 1	||
268 	    agcount == 0 || agcount != sbp->sb_agcount			||
269 	    (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog)	||
270 	    (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE)	||
271 	    (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE)	||
272 	    (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */)	||
273 	    sbp->sb_dblocks == 0					||
274 	    sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp)			||
275 	    sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp)			||
276 	    sbp->sb_shared_vn != 0)) {
277 		xfs_notice(mp, "SB sanity check failed");
278 		return -EFSCORRUPTED;
279 	}
280 
281 	if (xfs_sb_version_hascrc(&mp->m_sb) &&
282 	    sbp->sb_blocksize < XFS_MIN_CRC_BLOCKSIZE) {
283 		xfs_notice(mp, "v5 SB sanity check failed");
284 		return -EFSCORRUPTED;
285 	}
286 
287 	/*
288 	 * Until this is fixed only page-sized or smaller data blocks work.
289 	 */
290 	if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
291 		xfs_warn(mp,
292 		"File system with blocksize %d bytes. "
293 		"Only pagesize (%ld) or less will currently work.",
294 				sbp->sb_blocksize, PAGE_SIZE);
295 		return -ENOSYS;
296 	}
297 
298 	/*
299 	 * Currently only very few inode sizes are supported.
300 	 */
301 	switch (sbp->sb_inodesize) {
302 	case 256:
303 	case 512:
304 	case 1024:
305 	case 2048:
306 		break;
307 	default:
308 		xfs_warn(mp, "inode size of %d bytes not supported",
309 				sbp->sb_inodesize);
310 		return -ENOSYS;
311 	}
312 
313 	if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
314 	    xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
315 		xfs_warn(mp,
316 		"file system too large to be mounted on this system.");
317 		return -EFBIG;
318 	}
319 
320 	if (check_inprogress && sbp->sb_inprogress) {
321 		xfs_warn(mp, "Offline file system operation in progress!");
322 		return -EFSCORRUPTED;
323 	}
324 	return 0;
325 }
326 
327 void
328 xfs_sb_quota_from_disk(struct xfs_sb *sbp)
329 {
330 	/*
331 	 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
332 	 * leads to in-core values having two different values for a quota
333 	 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
334 	 * NULLFSINO.
335 	 *
336 	 * Note that this change affect only the in-core values. These
337 	 * values are not written back to disk unless any quota information
338 	 * is written to the disk. Even in that case, sb_pquotino field is
339 	 * not written to disk unless the superblock supports pquotino.
340 	 */
341 	if (sbp->sb_uquotino == 0)
342 		sbp->sb_uquotino = NULLFSINO;
343 	if (sbp->sb_gquotino == 0)
344 		sbp->sb_gquotino = NULLFSINO;
345 	if (sbp->sb_pquotino == 0)
346 		sbp->sb_pquotino = NULLFSINO;
347 
348 	/*
349 	 * We need to do these manipilations only if we are working
350 	 * with an older version of on-disk superblock.
351 	 */
352 	if (xfs_sb_version_has_pquotino(sbp))
353 		return;
354 
355 	if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
356 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
357 					XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
358 	if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
359 		sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
360 					XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
361 	sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
362 
363 	if (sbp->sb_qflags & XFS_PQUOTA_ACCT &&
364 	    sbp->sb_gquotino != NULLFSINO)  {
365 		/*
366 		 * In older version of superblock, on-disk superblock only
367 		 * has sb_gquotino, and in-core superblock has both sb_gquotino
368 		 * and sb_pquotino. But, only one of them is supported at any
369 		 * point of time. So, if PQUOTA is set in disk superblock,
370 		 * copy over sb_gquotino to sb_pquotino.  The NULLFSINO test
371 		 * above is to make sure we don't do this twice and wipe them
372 		 * both out!
373 		 */
374 		sbp->sb_pquotino = sbp->sb_gquotino;
375 		sbp->sb_gquotino = NULLFSINO;
376 	}
377 }
378 
379 static void
380 __xfs_sb_from_disk(
381 	struct xfs_sb	*to,
382 	xfs_dsb_t	*from,
383 	bool		convert_xquota)
384 {
385 	to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
386 	to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
387 	to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
388 	to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
389 	to->sb_rextents = be64_to_cpu(from->sb_rextents);
390 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
391 	to->sb_logstart = be64_to_cpu(from->sb_logstart);
392 	to->sb_rootino = be64_to_cpu(from->sb_rootino);
393 	to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
394 	to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
395 	to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
396 	to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
397 	to->sb_agcount = be32_to_cpu(from->sb_agcount);
398 	to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
399 	to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
400 	to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
401 	to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
402 	to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
403 	to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
404 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
405 	to->sb_blocklog = from->sb_blocklog;
406 	to->sb_sectlog = from->sb_sectlog;
407 	to->sb_inodelog = from->sb_inodelog;
408 	to->sb_inopblog = from->sb_inopblog;
409 	to->sb_agblklog = from->sb_agblklog;
410 	to->sb_rextslog = from->sb_rextslog;
411 	to->sb_inprogress = from->sb_inprogress;
412 	to->sb_imax_pct = from->sb_imax_pct;
413 	to->sb_icount = be64_to_cpu(from->sb_icount);
414 	to->sb_ifree = be64_to_cpu(from->sb_ifree);
415 	to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
416 	to->sb_frextents = be64_to_cpu(from->sb_frextents);
417 	to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
418 	to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
419 	to->sb_qflags = be16_to_cpu(from->sb_qflags);
420 	to->sb_flags = from->sb_flags;
421 	to->sb_shared_vn = from->sb_shared_vn;
422 	to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
423 	to->sb_unit = be32_to_cpu(from->sb_unit);
424 	to->sb_width = be32_to_cpu(from->sb_width);
425 	to->sb_dirblklog = from->sb_dirblklog;
426 	to->sb_logsectlog = from->sb_logsectlog;
427 	to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
428 	to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
429 	to->sb_features2 = be32_to_cpu(from->sb_features2);
430 	to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
431 	to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
432 	to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
433 	to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
434 	to->sb_features_log_incompat =
435 				be32_to_cpu(from->sb_features_log_incompat);
436 	/* crc is only used on disk, not in memory; just init to 0 here. */
437 	to->sb_crc = 0;
438 	to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
439 	to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
440 	to->sb_lsn = be64_to_cpu(from->sb_lsn);
441 	/*
442 	 * sb_meta_uuid is only on disk if it differs from sb_uuid and the
443 	 * feature flag is set; if not set we keep it only in memory.
444 	 */
445 	if (xfs_sb_version_hasmetauuid(to))
446 		uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
447 	else
448 		uuid_copy(&to->sb_meta_uuid, &from->sb_uuid);
449 	/* Convert on-disk flags to in-memory flags? */
450 	if (convert_xquota)
451 		xfs_sb_quota_from_disk(to);
452 }
453 
454 void
455 xfs_sb_from_disk(
456 	struct xfs_sb	*to,
457 	xfs_dsb_t	*from)
458 {
459 	__xfs_sb_from_disk(to, from, true);
460 }
461 
462 static void
463 xfs_sb_quota_to_disk(
464 	struct xfs_dsb	*to,
465 	struct xfs_sb	*from)
466 {
467 	uint16_t	qflags = from->sb_qflags;
468 
469 	to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
470 	if (xfs_sb_version_has_pquotino(from)) {
471 		to->sb_qflags = cpu_to_be16(from->sb_qflags);
472 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
473 		to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
474 		return;
475 	}
476 
477 	/*
478 	 * The in-core version of sb_qflags do not have XFS_OQUOTA_*
479 	 * flags, whereas the on-disk version does.  So, convert incore
480 	 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
481 	 */
482 	qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
483 			XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
484 
485 	if (from->sb_qflags &
486 			(XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
487 		qflags |= XFS_OQUOTA_ENFD;
488 	if (from->sb_qflags &
489 			(XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
490 		qflags |= XFS_OQUOTA_CHKD;
491 	to->sb_qflags = cpu_to_be16(qflags);
492 
493 	/*
494 	 * GQUOTINO and PQUOTINO cannot be used together in versions
495 	 * of superblock that do not have pquotino. from->sb_flags
496 	 * tells us which quota is active and should be copied to
497 	 * disk. If neither are active, we should NULL the inode.
498 	 *
499 	 * In all cases, the separate pquotino must remain 0 because it
500 	 * it beyond the "end" of the valid non-pquotino superblock.
501 	 */
502 	if (from->sb_qflags & XFS_GQUOTA_ACCT)
503 		to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
504 	else if (from->sb_qflags & XFS_PQUOTA_ACCT)
505 		to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
506 	else {
507 		/*
508 		 * We can't rely on just the fields being logged to tell us
509 		 * that it is safe to write NULLFSINO - we should only do that
510 		 * if quotas are not actually enabled. Hence only write
511 		 * NULLFSINO if both in-core quota inodes are NULL.
512 		 */
513 		if (from->sb_gquotino == NULLFSINO &&
514 		    from->sb_pquotino == NULLFSINO)
515 			to->sb_gquotino = cpu_to_be64(NULLFSINO);
516 	}
517 
518 	to->sb_pquotino = 0;
519 }
520 
521 void
522 xfs_sb_to_disk(
523 	struct xfs_dsb	*to,
524 	struct xfs_sb	*from)
525 {
526 	xfs_sb_quota_to_disk(to, from);
527 
528 	to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
529 	to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
530 	to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
531 	to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
532 	to->sb_rextents = cpu_to_be64(from->sb_rextents);
533 	memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
534 	to->sb_logstart = cpu_to_be64(from->sb_logstart);
535 	to->sb_rootino = cpu_to_be64(from->sb_rootino);
536 	to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
537 	to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
538 	to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
539 	to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
540 	to->sb_agcount = cpu_to_be32(from->sb_agcount);
541 	to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
542 	to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
543 	to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
544 	to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
545 	to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
546 	to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
547 	memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
548 	to->sb_blocklog = from->sb_blocklog;
549 	to->sb_sectlog = from->sb_sectlog;
550 	to->sb_inodelog = from->sb_inodelog;
551 	to->sb_inopblog = from->sb_inopblog;
552 	to->sb_agblklog = from->sb_agblklog;
553 	to->sb_rextslog = from->sb_rextslog;
554 	to->sb_inprogress = from->sb_inprogress;
555 	to->sb_imax_pct = from->sb_imax_pct;
556 	to->sb_icount = cpu_to_be64(from->sb_icount);
557 	to->sb_ifree = cpu_to_be64(from->sb_ifree);
558 	to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
559 	to->sb_frextents = cpu_to_be64(from->sb_frextents);
560 
561 	to->sb_flags = from->sb_flags;
562 	to->sb_shared_vn = from->sb_shared_vn;
563 	to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
564 	to->sb_unit = cpu_to_be32(from->sb_unit);
565 	to->sb_width = cpu_to_be32(from->sb_width);
566 	to->sb_dirblklog = from->sb_dirblklog;
567 	to->sb_logsectlog = from->sb_logsectlog;
568 	to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
569 	to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
570 
571 	/*
572 	 * We need to ensure that bad_features2 always matches features2.
573 	 * Hence we enforce that here rather than having to remember to do it
574 	 * everywhere else that updates features2.
575 	 */
576 	from->sb_bad_features2 = from->sb_features2;
577 	to->sb_features2 = cpu_to_be32(from->sb_features2);
578 	to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
579 
580 	if (xfs_sb_version_hascrc(from)) {
581 		to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
582 		to->sb_features_ro_compat =
583 				cpu_to_be32(from->sb_features_ro_compat);
584 		to->sb_features_incompat =
585 				cpu_to_be32(from->sb_features_incompat);
586 		to->sb_features_log_incompat =
587 				cpu_to_be32(from->sb_features_log_incompat);
588 		to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
589 		to->sb_lsn = cpu_to_be64(from->sb_lsn);
590 		if (xfs_sb_version_hasmetauuid(from))
591 			uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
592 	}
593 }
594 
595 static int
596 xfs_sb_verify(
597 	struct xfs_buf	*bp,
598 	bool		check_version)
599 {
600 	struct xfs_mount *mp = bp->b_target->bt_mount;
601 	struct xfs_sb	sb;
602 
603 	/*
604 	 * Use call variant which doesn't convert quota flags from disk
605 	 * format, because xfs_mount_validate_sb checks the on-disk flags.
606 	 */
607 	__xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
608 
609 	/*
610 	 * Only check the in progress field for the primary superblock as
611 	 * mkfs.xfs doesn't clear it from secondary superblocks.
612 	 */
613 	return xfs_mount_validate_sb(mp, &sb,
614 				     bp->b_maps[0].bm_bn == XFS_SB_DADDR,
615 				     check_version);
616 }
617 
618 /*
619  * If the superblock has the CRC feature bit set or the CRC field is non-null,
620  * check that the CRC is valid.  We check the CRC field is non-null because a
621  * single bit error could clear the feature bit and unused parts of the
622  * superblock are supposed to be zero. Hence a non-null crc field indicates that
623  * we've potentially lost a feature bit and we should check it anyway.
624  *
625  * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
626  * last field in V4 secondary superblocks.  So for secondary superblocks,
627  * we are more forgiving, and ignore CRC failures if the primary doesn't
628  * indicate that the fs version is V5.
629  */
630 static void
631 xfs_sb_read_verify(
632 	struct xfs_buf	*bp)
633 {
634 	struct xfs_mount *mp = bp->b_target->bt_mount;
635 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
636 	int		error;
637 
638 	/*
639 	 * open code the version check to avoid needing to convert the entire
640 	 * superblock from disk order just to check the version number
641 	 */
642 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
643 	    (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
644 						XFS_SB_VERSION_5) ||
645 	     dsb->sb_crc != 0)) {
646 
647 		if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
648 			/* Only fail bad secondaries on a known V5 filesystem */
649 			if (bp->b_bn == XFS_SB_DADDR ||
650 			    xfs_sb_version_hascrc(&mp->m_sb)) {
651 				error = -EFSBADCRC;
652 				goto out_error;
653 			}
654 		}
655 	}
656 	error = xfs_sb_verify(bp, true);
657 
658 out_error:
659 	if (error == -EFSCORRUPTED || error == -EFSBADCRC)
660 		xfs_verifier_error(bp, error, __this_address);
661 	else if (error)
662 		xfs_buf_ioerror(bp, error);
663 }
664 
665 /*
666  * We may be probed for a filesystem match, so we may not want to emit
667  * messages when the superblock buffer is not actually an XFS superblock.
668  * If we find an XFS superblock, then run a normal, noisy mount because we are
669  * really going to mount it and want to know about errors.
670  */
671 static void
672 xfs_sb_quiet_read_verify(
673 	struct xfs_buf	*bp)
674 {
675 	struct xfs_dsb	*dsb = XFS_BUF_TO_SBP(bp);
676 
677 	if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
678 		/* XFS filesystem, verify noisily! */
679 		xfs_sb_read_verify(bp);
680 		return;
681 	}
682 	/* quietly fail */
683 	xfs_buf_ioerror(bp, -EWRONGFS);
684 }
685 
686 static void
687 xfs_sb_write_verify(
688 	struct xfs_buf		*bp)
689 {
690 	struct xfs_mount	*mp = bp->b_target->bt_mount;
691 	struct xfs_buf_log_item	*bip = bp->b_log_item;
692 	int			error;
693 
694 	error = xfs_sb_verify(bp, false);
695 	if (error) {
696 		xfs_verifier_error(bp, error, __this_address);
697 		return;
698 	}
699 
700 	if (!xfs_sb_version_hascrc(&mp->m_sb))
701 		return;
702 
703 	if (bip)
704 		XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
705 
706 	xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
707 }
708 
709 const struct xfs_buf_ops xfs_sb_buf_ops = {
710 	.name = "xfs_sb",
711 	.verify_read = xfs_sb_read_verify,
712 	.verify_write = xfs_sb_write_verify,
713 };
714 
715 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
716 	.name = "xfs_sb_quiet",
717 	.verify_read = xfs_sb_quiet_read_verify,
718 	.verify_write = xfs_sb_write_verify,
719 };
720 
721 /*
722  * xfs_mount_common
723  *
724  * Mount initialization code establishing various mount
725  * fields from the superblock associated with the given
726  * mount structure
727  */
728 void
729 xfs_sb_mount_common(
730 	struct xfs_mount *mp,
731 	struct xfs_sb	*sbp)
732 {
733 	mp->m_agfrotor = mp->m_agirotor = 0;
734 	spin_lock_init(&mp->m_agirotor_lock);
735 	mp->m_maxagi = mp->m_sb.sb_agcount;
736 	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
737 	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
738 	mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
739 	mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
740 	mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
741 	mp->m_blockmask = sbp->sb_blocksize - 1;
742 	mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
743 	mp->m_blockwmask = mp->m_blockwsize - 1;
744 
745 	mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
746 	mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
747 	mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
748 	mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
749 
750 	mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1);
751 	mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0);
752 	mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2;
753 	mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2;
754 
755 	mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
756 	mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
757 	mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
758 	mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
759 
760 	mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 1);
761 	mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 0);
762 	mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2;
763 	mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2;
764 
765 	mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize,
766 			true);
767 	mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize,
768 			false);
769 	mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2;
770 	mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2;
771 
772 	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
773 	mp->m_ialloc_inos = (int)MAX((uint16_t)XFS_INODES_PER_CHUNK,
774 					sbp->sb_inopblock);
775 	mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
776 
777 	if (sbp->sb_spino_align)
778 		mp->m_ialloc_min_blks = sbp->sb_spino_align;
779 	else
780 		mp->m_ialloc_min_blks = mp->m_ialloc_blks;
781 	mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
782 	mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp);
783 }
784 
785 /*
786  * xfs_initialize_perag_data
787  *
788  * Read in each per-ag structure so we can count up the number of
789  * allocated inodes, free inodes and used filesystem blocks as this
790  * information is no longer persistent in the superblock. Once we have
791  * this information, write it into the in-core superblock structure.
792  */
793 int
794 xfs_initialize_perag_data(
795 	struct xfs_mount *mp,
796 	xfs_agnumber_t	agcount)
797 {
798 	xfs_agnumber_t	index;
799 	xfs_perag_t	*pag;
800 	xfs_sb_t	*sbp = &mp->m_sb;
801 	uint64_t	ifree = 0;
802 	uint64_t	ialloc = 0;
803 	uint64_t	bfree = 0;
804 	uint64_t	bfreelst = 0;
805 	uint64_t	btree = 0;
806 	int		error;
807 
808 	for (index = 0; index < agcount; index++) {
809 		/*
810 		 * read the agf, then the agi. This gets us
811 		 * all the information we need and populates the
812 		 * per-ag structures for us.
813 		 */
814 		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
815 		if (error)
816 			return error;
817 
818 		error = xfs_ialloc_pagi_init(mp, NULL, index);
819 		if (error)
820 			return error;
821 		pag = xfs_perag_get(mp, index);
822 		ifree += pag->pagi_freecount;
823 		ialloc += pag->pagi_count;
824 		bfree += pag->pagf_freeblks;
825 		bfreelst += pag->pagf_flcount;
826 		btree += pag->pagf_btreeblks;
827 		xfs_perag_put(pag);
828 	}
829 
830 	/* Overwrite incore superblock counters with just-read data */
831 	spin_lock(&mp->m_sb_lock);
832 	sbp->sb_ifree = ifree;
833 	sbp->sb_icount = ialloc;
834 	sbp->sb_fdblocks = bfree + bfreelst + btree;
835 	spin_unlock(&mp->m_sb_lock);
836 
837 	xfs_reinit_percpu_counters(mp);
838 
839 	return 0;
840 }
841 
842 /*
843  * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
844  * into the superblock buffer to be logged.  It does not provide the higher
845  * level of locking that is needed to protect the in-core superblock from
846  * concurrent access.
847  */
848 void
849 xfs_log_sb(
850 	struct xfs_trans	*tp)
851 {
852 	struct xfs_mount	*mp = tp->t_mountp;
853 	struct xfs_buf		*bp = xfs_trans_getsb(tp, mp, 0);
854 
855 	mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount);
856 	mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree);
857 	mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks);
858 
859 	xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
860 	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
861 	xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb));
862 }
863 
864 /*
865  * xfs_sync_sb
866  *
867  * Sync the superblock to disk.
868  *
869  * Note that the caller is responsible for checking the frozen state of the
870  * filesystem. This procedure uses the non-blocking transaction allocator and
871  * thus will allow modifications to a frozen fs. This is required because this
872  * code can be called during the process of freezing where use of the high-level
873  * allocator would deadlock.
874  */
875 int
876 xfs_sync_sb(
877 	struct xfs_mount	*mp,
878 	bool			wait)
879 {
880 	struct xfs_trans	*tp;
881 	int			error;
882 
883 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0,
884 			XFS_TRANS_NO_WRITECOUNT, &tp);
885 	if (error)
886 		return error;
887 
888 	xfs_log_sb(tp);
889 	if (wait)
890 		xfs_trans_set_sync(tp);
891 	return xfs_trans_commit(tp);
892 }
893 
894 int
895 xfs_fs_geometry(
896 	struct xfs_sb		*sbp,
897 	struct xfs_fsop_geom	*geo,
898 	int			struct_version)
899 {
900 	memset(geo, 0, sizeof(struct xfs_fsop_geom));
901 
902 	geo->blocksize = sbp->sb_blocksize;
903 	geo->rtextsize = sbp->sb_rextsize;
904 	geo->agblocks = sbp->sb_agblocks;
905 	geo->agcount = sbp->sb_agcount;
906 	geo->logblocks = sbp->sb_logblocks;
907 	geo->sectsize = sbp->sb_sectsize;
908 	geo->inodesize = sbp->sb_inodesize;
909 	geo->imaxpct = sbp->sb_imax_pct;
910 	geo->datablocks = sbp->sb_dblocks;
911 	geo->rtblocks = sbp->sb_rblocks;
912 	geo->rtextents = sbp->sb_rextents;
913 	geo->logstart = sbp->sb_logstart;
914 	BUILD_BUG_ON(sizeof(geo->uuid) != sizeof(sbp->sb_uuid));
915 	memcpy(geo->uuid, &sbp->sb_uuid, sizeof(sbp->sb_uuid));
916 
917 	if (struct_version < 2)
918 		return 0;
919 
920 	geo->sunit = sbp->sb_unit;
921 	geo->swidth = sbp->sb_width;
922 
923 	if (struct_version < 3)
924 		return 0;
925 
926 	geo->version = XFS_FSOP_GEOM_VERSION;
927 	geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
928 		     XFS_FSOP_GEOM_FLAGS_DIRV2;
929 	if (xfs_sb_version_hasattr(sbp))
930 		geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR;
931 	if (xfs_sb_version_hasquota(sbp))
932 		geo->flags |= XFS_FSOP_GEOM_FLAGS_QUOTA;
933 	if (xfs_sb_version_hasalign(sbp))
934 		geo->flags |= XFS_FSOP_GEOM_FLAGS_IALIGN;
935 	if (xfs_sb_version_hasdalign(sbp))
936 		geo->flags |= XFS_FSOP_GEOM_FLAGS_DALIGN;
937 	if (xfs_sb_version_hasextflgbit(sbp))
938 		geo->flags |= XFS_FSOP_GEOM_FLAGS_EXTFLG;
939 	if (xfs_sb_version_hassector(sbp))
940 		geo->flags |= XFS_FSOP_GEOM_FLAGS_SECTOR;
941 	if (xfs_sb_version_hasasciici(sbp))
942 		geo->flags |= XFS_FSOP_GEOM_FLAGS_DIRV2CI;
943 	if (xfs_sb_version_haslazysbcount(sbp))
944 		geo->flags |= XFS_FSOP_GEOM_FLAGS_LAZYSB;
945 	if (xfs_sb_version_hasattr2(sbp))
946 		geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR2;
947 	if (xfs_sb_version_hasprojid32bit(sbp))
948 		geo->flags |= XFS_FSOP_GEOM_FLAGS_PROJID32;
949 	if (xfs_sb_version_hascrc(sbp))
950 		geo->flags |= XFS_FSOP_GEOM_FLAGS_V5SB;
951 	if (xfs_sb_version_hasftype(sbp))
952 		geo->flags |= XFS_FSOP_GEOM_FLAGS_FTYPE;
953 	if (xfs_sb_version_hasfinobt(sbp))
954 		geo->flags |= XFS_FSOP_GEOM_FLAGS_FINOBT;
955 	if (xfs_sb_version_hassparseinodes(sbp))
956 		geo->flags |= XFS_FSOP_GEOM_FLAGS_SPINODES;
957 	if (xfs_sb_version_hasrmapbt(sbp))
958 		geo->flags |= XFS_FSOP_GEOM_FLAGS_RMAPBT;
959 	if (xfs_sb_version_hasreflink(sbp))
960 		geo->flags |= XFS_FSOP_GEOM_FLAGS_REFLINK;
961 	if (xfs_sb_version_hassector(sbp))
962 		geo->logsectsize = sbp->sb_logsectsize;
963 	else
964 		geo->logsectsize = BBSIZE;
965 	geo->rtsectsize = sbp->sb_blocksize;
966 	geo->dirblocksize = xfs_dir2_dirblock_bytes(sbp);
967 
968 	if (struct_version < 4)
969 		return 0;
970 
971 	if (xfs_sb_version_haslogv2(sbp))
972 		geo->flags |= XFS_FSOP_GEOM_FLAGS_LOGV2;
973 
974 	geo->logsunit = sbp->sb_logsunit;
975 
976 	return 0;
977 }
978