Line data Source code
1 : // SPDX-License-Identifier: GPL-2.0
2 : /*
3 : * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4 : * All Rights Reserved.
5 : */
6 : #include "xfs.h"
7 : #include "xfs_fs.h"
8 : #include "xfs_shared.h"
9 : #include "xfs_format.h"
10 : #include "xfs_log_format.h"
11 : #include "xfs_trans_resv.h"
12 : #include "xfs_bit.h"
13 : #include "xfs_sb.h"
14 : #include "xfs_mount.h"
15 : #include "xfs_ialloc.h"
16 : #include "xfs_alloc.h"
17 : #include "xfs_error.h"
18 : #include "xfs_trans.h"
19 : #include "xfs_buf_item.h"
20 : #include "xfs_bmap_btree.h"
21 : #include "xfs_alloc_btree.h"
22 : #include "xfs_log.h"
23 : #include "xfs_rmap_btree.h"
24 : #include "xfs_refcount_btree.h"
25 : #include "xfs_da_format.h"
26 : #include "xfs_health.h"
27 : #include "xfs_ag.h"
28 : #include "xfs_swapext.h"
29 :
30 : /*
31 : * Physical superblock buffer manipulations. Shared with libxfs in userspace.
32 : */
33 :
34 : /*
35 : * Check that all the V4 feature bits that the V5 filesystem format requires are
36 : * correctly set.
37 : */
38 : static bool
39 430542 : xfs_sb_validate_v5_features(
40 : struct xfs_sb *sbp)
41 : {
42 : /* We must not have any unknown V4 feature bits set */
43 430542 : if (sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS)
44 : return false;
45 :
46 : /*
47 : * The CRC bit is considered an invalid V4 flag, so we have to add it
48 : * manually to the OKBITS mask.
49 : */
50 430542 : if (sbp->sb_features2 & ~(XFS_SB_VERSION2_OKBITS |
51 : XFS_SB_VERSION2_CRCBIT))
52 : return false;
53 :
54 : /* Now check all the required V4 feature flags are set. */
55 :
56 : #define V5_VERS_FLAGS (XFS_SB_VERSION_NLINKBIT | \
57 : XFS_SB_VERSION_ALIGNBIT | \
58 : XFS_SB_VERSION_LOGV2BIT | \
59 : XFS_SB_VERSION_EXTFLGBIT | \
60 : XFS_SB_VERSION_DIRV2BIT | \
61 : XFS_SB_VERSION_MOREBITSBIT)
62 :
63 : #define V5_FEAT_FLAGS (XFS_SB_VERSION2_LAZYSBCOUNTBIT | \
64 : XFS_SB_VERSION2_ATTR2BIT | \
65 : XFS_SB_VERSION2_PROJID32BIT | \
66 : XFS_SB_VERSION2_CRCBIT)
67 :
68 430542 : if ((sbp->sb_versionnum & V5_VERS_FLAGS) != V5_VERS_FLAGS)
69 : return false;
70 430542 : if ((sbp->sb_features2 & V5_FEAT_FLAGS) != V5_FEAT_FLAGS)
71 0 : return false;
72 : return true;
73 : }
74 :
75 : /*
76 : * We current support XFS v5 formats with known features and v4 superblocks with
77 : * at least V2 directories.
78 : */
79 : bool
80 435252 : xfs_sb_good_version(
81 : struct xfs_sb *sbp)
82 : {
83 : /*
84 : * All v5 filesystems are supported, but we must check that all the
85 : * required v4 feature flags are enabled correctly as the code checks
86 : * those flags and not for v5 support.
87 : */
88 435252 : if (xfs_sb_is_v5(sbp))
89 430542 : return xfs_sb_validate_v5_features(sbp);
90 :
91 : /* versions prior to v4 are not supported */
92 4710 : if (XFS_SB_VERSION_NUM(sbp) != XFS_SB_VERSION_4)
93 : return false;
94 :
95 : /* We must not have any unknown v4 feature bits set */
96 4710 : if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) ||
97 4710 : ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) &&
98 4710 : (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS)))
99 : return false;
100 :
101 : /* V4 filesystems need v2 directories and unwritten extents */
102 4710 : if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT))
103 : return false;
104 4710 : if (!(sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT))
105 0 : return false;
106 :
107 : /* It's a supported v4 filesystem */
108 : return true;
109 : }
110 :
111 : uint64_t
112 35424 : xfs_sb_version_to_features(
113 : struct xfs_sb *sbp)
114 : {
115 35424 : uint64_t features = 0;
116 :
117 : /* optional V4 features */
118 35424 : if (sbp->sb_rblocks > 0)
119 181 : features |= XFS_FEAT_REALTIME;
120 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_NLINKBIT)
121 35424 : features |= XFS_FEAT_NLINK;
122 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT)
123 35424 : features |= XFS_FEAT_ATTR;
124 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT)
125 28911 : features |= XFS_FEAT_QUOTA;
126 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT)
127 35424 : features |= XFS_FEAT_ALIGN;
128 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT)
129 35414 : features |= XFS_FEAT_LOGV2;
130 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT)
131 114 : features |= XFS_FEAT_DALIGN;
132 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT)
133 35424 : features |= XFS_FEAT_EXTFLG;
134 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT)
135 35100 : features |= XFS_FEAT_SECTOR;
136 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT)
137 26 : features |= XFS_FEAT_ASCIICI;
138 35424 : if (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) {
139 35424 : if (sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT)
140 35418 : features |= XFS_FEAT_LAZYSBCOUNT;
141 35424 : if (sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT)
142 35418 : features |= XFS_FEAT_ATTR2;
143 35424 : if (sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT)
144 35418 : features |= XFS_FEAT_PROJID32;
145 35424 : if (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE)
146 30 : features |= XFS_FEAT_FTYPE;
147 : }
148 :
149 35424 : if (!xfs_sb_is_v5(sbp))
150 : return features;
151 :
152 : /* Always on V5 features */
153 35372 : features |= XFS_FEAT_ALIGN | XFS_FEAT_LOGV2 | XFS_FEAT_EXTFLG |
154 : XFS_FEAT_LAZYSBCOUNT | XFS_FEAT_ATTR2 | XFS_FEAT_PROJID32 |
155 : XFS_FEAT_V3INODES | XFS_FEAT_CRC | XFS_FEAT_PQUOTINO;
156 :
157 : /* Optional V5 features */
158 35372 : if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT)
159 35366 : features |= XFS_FEAT_FINOBT;
160 35372 : if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT)
161 35163 : features |= XFS_FEAT_RMAPBT;
162 35372 : if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK)
163 35158 : features |= XFS_FEAT_REFLINK;
164 35372 : if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_INOBTCNT)
165 35343 : features |= XFS_FEAT_INOBTCNT;
166 35372 : if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_FTYPE)
167 35372 : features |= XFS_FEAT_FTYPE;
168 35372 : if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES)
169 35372 : features |= XFS_FEAT_SPINODES;
170 35372 : if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID)
171 18 : features |= XFS_FEAT_META_UUID;
172 35372 : if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_BIGTIME)
173 35337 : features |= XFS_FEAT_BIGTIME;
174 35372 : if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR)
175 4 : features |= XFS_FEAT_NEEDSREPAIR;
176 35372 : if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_NREXT64)
177 35370 : features |= XFS_FEAT_NREXT64;
178 35372 : if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_PARENT)
179 34736 : features |= XFS_FEAT_PARENT;
180 :
181 : return features;
182 : }
183 :
184 : /* Check all the superblock fields we care about when reading one in. */
185 : STATIC int
186 195020 : xfs_validate_sb_read(
187 : struct xfs_mount *mp,
188 : struct xfs_sb *sbp)
189 : {
190 195020 : if (!xfs_sb_is_v5(sbp))
191 : return 0;
192 :
193 : /*
194 : * Version 5 superblock feature mask validation. Reject combinations
195 : * the kernel cannot support up front before checking anything else.
196 : */
197 194966 : if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) {
198 0 : xfs_warn(mp,
199 : "Superblock has unknown compatible features (0x%x) enabled.",
200 : (sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN));
201 0 : xfs_warn(mp,
202 : "Using a more recent kernel is recommended.");
203 : }
204 :
205 194966 : if (xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
206 4 : xfs_alert(mp,
207 : "Superblock has unknown read-only compatible features (0x%x) enabled.",
208 : (sbp->sb_features_ro_compat &
209 : XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
210 8 : if (!xfs_is_readonly(mp)) {
211 2 : xfs_warn(mp,
212 : "Attempted to mount read-only compatible filesystem read-write.");
213 2 : xfs_warn(mp,
214 : "Filesystem can only be safely mounted read only.");
215 :
216 2 : return -EINVAL;
217 : }
218 : }
219 194964 : if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
220 0 : xfs_warn(mp,
221 : "Superblock has unknown incompatible features (0x%x) enabled.",
222 : (sbp->sb_features_incompat &
223 : XFS_SB_FEAT_INCOMPAT_UNKNOWN));
224 0 : xfs_warn(mp,
225 : "Filesystem cannot be safely mounted by this kernel.");
226 0 : return -EINVAL;
227 : }
228 :
229 : return 0;
230 : }
231 :
232 : /* Check all the superblock fields we care about when writing one out. */
233 : STATIC int
234 240208 : xfs_validate_sb_write(
235 : struct xfs_mount *mp,
236 : struct xfs_buf *bp,
237 : struct xfs_sb *sbp)
238 : {
239 : /*
240 : * Carry out additional sb summary counter sanity checks when we write
241 : * the superblock. We skip this in the read validator because there
242 : * could be newer superblocks in the log and if the values are garbage
243 : * even after replay we'll recalculate them at the end of log mount.
244 : *
245 : * mkfs has traditionally written zeroed counters to inprogress and
246 : * secondary superblocks, so allow this usage to continue because
247 : * we never read counters from such superblocks.
248 : */
249 240208 : if (xfs_buf_daddr(bp) == XFS_SB_DADDR && !sbp->sb_inprogress &&
250 437176 : (sbp->sb_fdblocks > sbp->sb_dblocks ||
251 218587 : !xfs_verify_icount(mp, sbp->sb_icount) ||
252 218587 : sbp->sb_ifree > sbp->sb_icount)) {
253 2 : xfs_warn(mp, "SB summary counter sanity check failed");
254 2 : return -EFSCORRUPTED;
255 : }
256 :
257 240206 : if (!xfs_sb_is_v5(sbp))
258 : return 0;
259 :
260 : /*
261 : * Version 5 superblock feature mask validation. Reject combinations
262 : * the kernel cannot support since we checked for unsupported bits in
263 : * the read verifier, which means that memory is corrupt.
264 : */
265 235550 : if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) {
266 0 : xfs_warn(mp,
267 : "Corruption detected in superblock compatible features (0x%x)!",
268 : (sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN));
269 0 : return -EFSCORRUPTED;
270 : }
271 :
272 235550 : if (xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
273 2 : xfs_alert(mp,
274 : "Corruption detected in superblock read-only compatible features (0x%x)!",
275 : (sbp->sb_features_ro_compat &
276 : XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
277 2 : return -EFSCORRUPTED;
278 : }
279 235548 : if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
280 0 : xfs_warn(mp,
281 : "Corruption detected in superblock incompatible features (0x%x)!",
282 : (sbp->sb_features_incompat &
283 : XFS_SB_FEAT_INCOMPAT_UNKNOWN));
284 0 : return -EFSCORRUPTED;
285 : }
286 235548 : if (xfs_sb_has_incompat_log_feature(sbp,
287 : XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN)) {
288 0 : xfs_warn(mp,
289 : "Corruption detected in superblock incompatible log features (0x%x)!",
290 : (sbp->sb_features_log_incompat &
291 : XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN));
292 0 : return -EFSCORRUPTED;
293 : }
294 :
295 : /*
296 : * We can't read verify the sb LSN because the read verifier is called
297 : * before the log is allocated and processed. We know the log is set up
298 : * before write verifier calls, so check it here.
299 : */
300 235548 : if (!xfs_log_check_lsn(mp, sbp->sb_lsn))
301 0 : return -EFSCORRUPTED;
302 :
303 : return 0;
304 : }
305 :
306 : /* Check the validity of the SB. */
307 : STATIC int
308 435254 : xfs_validate_sb_common(
309 : struct xfs_mount *mp,
310 : struct xfs_buf *bp,
311 : struct xfs_sb *sbp)
312 : {
313 435254 : struct xfs_dsb *dsb = bp->b_addr;
314 435254 : uint32_t agcount = 0;
315 435254 : uint32_t rem;
316 435254 : bool has_dalign;
317 :
318 435254 : if (!xfs_verify_magic(bp, dsb->sb_magicnum)) {
319 2 : xfs_warn(mp,
320 : "Superblock has bad magic number 0x%x. Not an XFS filesystem?",
321 : be32_to_cpu(dsb->sb_magicnum));
322 2 : return -EWRONGFS;
323 : }
324 :
325 435252 : if (!xfs_sb_good_version(sbp)) {
326 0 : xfs_warn(mp,
327 : "Superblock has unknown features enabled or corrupted feature masks.");
328 0 : return -EWRONGFS;
329 : }
330 :
331 : /*
332 : * Validate feature flags and state
333 : */
334 435252 : if (xfs_sb_is_v5(sbp)) {
335 430542 : if (sbp->sb_blocksize < XFS_MIN_CRC_BLOCKSIZE) {
336 0 : xfs_notice(mp,
337 : "Block size (%u bytes) too small for Version 5 superblock (minimum %d bytes)",
338 : sbp->sb_blocksize, XFS_MIN_CRC_BLOCKSIZE);
339 0 : return -EFSCORRUPTED;
340 : }
341 :
342 : /* V5 has a separate project quota inode */
343 430542 : if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
344 0 : xfs_notice(mp,
345 : "Version 5 of Super block has XFS_OQUOTA bits.");
346 0 : return -EFSCORRUPTED;
347 : }
348 :
349 : /*
350 : * Full inode chunks must be aligned to inode chunk size when
351 : * sparse inodes are enabled to support the sparse chunk
352 : * allocation algorithm and prevent overlapping inode records.
353 : */
354 430542 : if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES) {
355 430542 : uint32_t align;
356 :
357 430542 : align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
358 430542 : >> sbp->sb_blocklog;
359 430542 : if (sbp->sb_inoalignmt != align) {
360 0 : xfs_warn(mp,
361 : "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
362 : sbp->sb_inoalignmt, align);
363 0 : return -EINVAL;
364 : }
365 : }
366 4710 : } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
367 : XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
368 0 : xfs_notice(mp,
369 : "Superblock earlier than Version 5 has XFS_{P|G}QUOTA_{ENFD|CHKD} bits.");
370 0 : return -EFSCORRUPTED;
371 : }
372 :
373 435252 : if (unlikely(
374 : sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
375 2 : xfs_warn(mp,
376 : "filesystem is marked as having an external log; "
377 : "specify logdev on the mount command line.");
378 2 : return -EINVAL;
379 : }
380 :
381 435250 : if (unlikely(
382 : sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
383 2 : xfs_warn(mp,
384 : "filesystem is marked as having an internal log; "
385 : "do not specify logdev on the mount command line.");
386 2 : return -EINVAL;
387 : }
388 :
389 : /* Compute agcount for this number of dblocks and agblocks */
390 435248 : if (sbp->sb_agblocks) {
391 435248 : agcount = div_u64_rem(sbp->sb_dblocks, sbp->sb_agblocks, &rem);
392 435248 : if (rem)
393 14540 : agcount++;
394 : }
395 :
396 : /*
397 : * More sanity checking. Most of these were stolen directly from
398 : * xfs_repair.
399 : */
400 870494 : if (unlikely(
401 : sbp->sb_agcount <= 0 ||
402 : sbp->sb_sectsize < XFS_MIN_SECTORSIZE ||
403 : sbp->sb_sectsize > XFS_MAX_SECTORSIZE ||
404 : sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG ||
405 : sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG ||
406 : sbp->sb_sectsize != (1 << sbp->sb_sectlog) ||
407 : sbp->sb_blocksize < XFS_MIN_BLOCKSIZE ||
408 : sbp->sb_blocksize > XFS_MAX_BLOCKSIZE ||
409 : sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG ||
410 : sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
411 : sbp->sb_blocksize != (1 << sbp->sb_blocklog) ||
412 : sbp->sb_dirblklog + sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
413 : sbp->sb_inodesize < XFS_DINODE_MIN_SIZE ||
414 : sbp->sb_inodesize > XFS_DINODE_MAX_SIZE ||
415 : sbp->sb_inodelog < XFS_DINODE_MIN_LOG ||
416 : sbp->sb_inodelog > XFS_DINODE_MAX_LOG ||
417 : sbp->sb_inodesize != (1 << sbp->sb_inodelog) ||
418 : sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
419 : XFS_FSB_TO_B(mp, sbp->sb_agblocks) < XFS_MIN_AG_BYTES ||
420 : XFS_FSB_TO_B(mp, sbp->sb_agblocks) > XFS_MAX_AG_BYTES ||
421 : sbp->sb_agblklog != xfs_highbit32(sbp->sb_agblocks - 1) + 1 ||
422 : agcount == 0 || agcount != sbp->sb_agcount ||
423 : (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) ||
424 : (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) ||
425 : (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) ||
426 : (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) ||
427 : sbp->sb_dblocks == 0 ||
428 : sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) ||
429 : sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) ||
430 : sbp->sb_shared_vn != 0)) {
431 18 : xfs_notice(mp, "SB sanity check failed");
432 18 : return -EFSCORRUPTED;
433 : }
434 :
435 : /*
436 : * Logs that are too large are not supported at all. Reject them
437 : * outright. Logs that are too small are tolerated on v4 filesystems,
438 : * but we can only check that when mounting the log. Hence we skip
439 : * those checks here.
440 : */
441 435230 : if (sbp->sb_logblocks > XFS_MAX_LOG_BLOCKS) {
442 0 : xfs_notice(mp,
443 : "Log size 0x%x blocks too large, maximum size is 0x%llx blocks",
444 : sbp->sb_logblocks, XFS_MAX_LOG_BLOCKS);
445 0 : return -EFSCORRUPTED;
446 : }
447 :
448 435230 : if (XFS_FSB_TO_B(mp, sbp->sb_logblocks) > XFS_MAX_LOG_BYTES) {
449 0 : xfs_warn(mp,
450 : "log size 0x%llx bytes too large, maximum size is 0x%llx bytes",
451 : XFS_FSB_TO_B(mp, sbp->sb_logblocks),
452 : XFS_MAX_LOG_BYTES);
453 0 : return -EFSCORRUPTED;
454 : }
455 :
456 : /*
457 : * Do not allow filesystems with corrupted log sector or stripe units to
458 : * be mounted. We cannot safely size the iclogs or write to the log if
459 : * the log stripe unit is not valid.
460 : */
461 435230 : if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT) {
462 432392 : if (sbp->sb_logsectsize != (1U << sbp->sb_logsectlog)) {
463 0 : xfs_notice(mp,
464 : "log sector size in bytes/log2 (0x%x/0x%x) must match",
465 : sbp->sb_logsectsize, 1U << sbp->sb_logsectlog);
466 0 : return -EFSCORRUPTED;
467 : }
468 2838 : } else if (sbp->sb_logsectsize || sbp->sb_logsectlog) {
469 0 : xfs_notice(mp,
470 : "log sector size in bytes/log2 (0x%x/0x%x) are not zero",
471 : sbp->sb_logsectsize, sbp->sb_logsectlog);
472 0 : return -EFSCORRUPTED;
473 : }
474 :
475 435230 : if (sbp->sb_logsunit > 1) {
476 432640 : if (sbp->sb_logsunit % sbp->sb_blocksize) {
477 2 : xfs_notice(mp,
478 : "log stripe unit 0x%x bytes must be a multiple of block size",
479 : sbp->sb_logsunit);
480 2 : return -EFSCORRUPTED;
481 : }
482 432638 : if (sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE) {
483 0 : xfs_notice(mp,
484 : "log stripe unit 0x%x bytes over maximum size (0x%x bytes)",
485 : sbp->sb_logsunit, XLOG_MAX_RECORD_BSIZE);
486 0 : return -EFSCORRUPTED;
487 : }
488 : }
489 :
490 : /* Validate the realtime geometry; stolen from xfs_repair */
491 435228 : if (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE ||
492 : sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) {
493 0 : xfs_notice(mp,
494 : "realtime extent sanity check failed");
495 0 : return -EFSCORRUPTED;
496 : }
497 :
498 435228 : if (sbp->sb_rblocks == 0) {
499 306304 : if (sbp->sb_rextents != 0 || sbp->sb_rbmblocks != 0 ||
500 306304 : sbp->sb_rextslog != 0 || sbp->sb_frextents != 0) {
501 0 : xfs_notice(mp,
502 : "realtime zeroed geometry check failed");
503 0 : return -EFSCORRUPTED;
504 : }
505 : } else {
506 128924 : uint64_t rexts;
507 128924 : uint64_t rbmblocks;
508 :
509 128924 : rexts = div_u64(sbp->sb_rblocks, sbp->sb_rextsize);
510 128924 : rbmblocks = howmany_64(sbp->sb_rextents,
511 : NBBY * sbp->sb_blocksize);
512 :
513 257848 : if (sbp->sb_rextents != rexts ||
514 128924 : sbp->sb_rextslog != xfs_highbit32(sbp->sb_rextents) ||
515 128924 : sbp->sb_rbmblocks != rbmblocks) {
516 0 : xfs_notice(mp,
517 : "realtime geometry sanity check failed");
518 0 : return -EFSCORRUPTED;
519 : }
520 : }
521 :
522 : /*
523 : * Either (sb_unit and !hasdalign) or (!sb_unit and hasdalign)
524 : * would imply the image is corrupted.
525 : */
526 435228 : has_dalign = sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT;
527 435228 : if (!!sbp->sb_unit ^ has_dalign) {
528 0 : xfs_notice(mp, "SB stripe alignment sanity check failed");
529 0 : return -EFSCORRUPTED;
530 : }
531 :
532 435228 : if (!xfs_validate_stripe_geometry(mp, XFS_FSB_TO_B(mp, sbp->sb_unit),
533 435228 : XFS_FSB_TO_B(mp, sbp->sb_width), 0, false))
534 : return -EFSCORRUPTED;
535 :
536 : /*
537 : * Currently only very few inode sizes are supported.
538 : */
539 435228 : switch (sbp->sb_inodesize) {
540 : case 256:
541 : case 512:
542 : case 1024:
543 : case 2048:
544 : break;
545 0 : default:
546 0 : xfs_warn(mp, "inode size of %d bytes not supported",
547 : sbp->sb_inodesize);
548 0 : return -ENOSYS;
549 : }
550 :
551 : return 0;
552 : }
553 :
554 : void
555 59579 : xfs_sb_quota_from_disk(struct xfs_sb *sbp)
556 : {
557 : /*
558 : * older mkfs doesn't initialize quota inodes to NULLFSINO. This
559 : * leads to in-core values having two different values for a quota
560 : * inode to be invalid: 0 and NULLFSINO. Change it to a single value
561 : * NULLFSINO.
562 : *
563 : * Note that this change affect only the in-core values. These
564 : * values are not written back to disk unless any quota information
565 : * is written to the disk. Even in that case, sb_pquotino field is
566 : * not written to disk unless the superblock supports pquotino.
567 : */
568 59579 : if (sbp->sb_uquotino == 0)
569 6626 : sbp->sb_uquotino = NULLFSINO;
570 59579 : if (sbp->sb_gquotino == 0)
571 6626 : sbp->sb_gquotino = NULLFSINO;
572 59579 : if (sbp->sb_pquotino == 0)
573 6690 : sbp->sb_pquotino = NULLFSINO;
574 :
575 : /*
576 : * We need to do these manipilations only if we are working
577 : * with an older version of on-disk superblock.
578 : */
579 59579 : if (xfs_sb_is_v5(sbp))
580 : return;
581 :
582 104 : if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
583 32 : sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
584 : XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
585 104 : if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
586 32 : sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
587 : XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
588 104 : sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
589 :
590 104 : if (sbp->sb_qflags & XFS_PQUOTA_ACCT &&
591 16 : sbp->sb_gquotino != NULLFSINO) {
592 : /*
593 : * In older version of superblock, on-disk superblock only
594 : * has sb_gquotino, and in-core superblock has both sb_gquotino
595 : * and sb_pquotino. But, only one of them is supported at any
596 : * point of time. So, if PQUOTA is set in disk superblock,
597 : * copy over sb_gquotino to sb_pquotino. The NULLFSINO test
598 : * above is to make sure we don't do this twice and wipe them
599 : * both out!
600 : */
601 16 : sbp->sb_pquotino = sbp->sb_gquotino;
602 16 : sbp->sb_gquotino = NULLFSINO;
603 : }
604 : }
605 :
606 : static void
607 494833 : __xfs_sb_from_disk(
608 : struct xfs_sb *to,
609 : struct xfs_dsb *from,
610 : bool convert_xquota)
611 : {
612 494833 : to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
613 494833 : to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
614 494833 : to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
615 494833 : to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
616 494833 : to->sb_rextents = be64_to_cpu(from->sb_rextents);
617 989666 : memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
618 494833 : to->sb_logstart = be64_to_cpu(from->sb_logstart);
619 494833 : to->sb_rootino = be64_to_cpu(from->sb_rootino);
620 494833 : to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
621 494833 : to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
622 494833 : to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
623 494833 : to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
624 494833 : to->sb_agcount = be32_to_cpu(from->sb_agcount);
625 494833 : to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
626 494833 : to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
627 494833 : to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
628 494833 : to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
629 494833 : to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
630 494833 : to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
631 989666 : memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
632 494833 : to->sb_blocklog = from->sb_blocklog;
633 494833 : to->sb_sectlog = from->sb_sectlog;
634 494833 : to->sb_inodelog = from->sb_inodelog;
635 494833 : to->sb_inopblog = from->sb_inopblog;
636 494833 : to->sb_agblklog = from->sb_agblklog;
637 494833 : to->sb_rextslog = from->sb_rextslog;
638 494833 : to->sb_inprogress = from->sb_inprogress;
639 494833 : to->sb_imax_pct = from->sb_imax_pct;
640 494833 : to->sb_icount = be64_to_cpu(from->sb_icount);
641 494833 : to->sb_ifree = be64_to_cpu(from->sb_ifree);
642 494833 : to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
643 494833 : to->sb_frextents = be64_to_cpu(from->sb_frextents);
644 494833 : to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
645 494833 : to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
646 494833 : to->sb_qflags = be16_to_cpu(from->sb_qflags);
647 494833 : to->sb_flags = from->sb_flags;
648 494833 : to->sb_shared_vn = from->sb_shared_vn;
649 494833 : to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
650 494833 : to->sb_unit = be32_to_cpu(from->sb_unit);
651 494833 : to->sb_width = be32_to_cpu(from->sb_width);
652 494833 : to->sb_dirblklog = from->sb_dirblklog;
653 494833 : to->sb_logsectlog = from->sb_logsectlog;
654 494833 : to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
655 494833 : to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
656 494833 : to->sb_features2 = be32_to_cpu(from->sb_features2);
657 494833 : to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
658 494833 : to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
659 494833 : to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
660 494833 : to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
661 989666 : to->sb_features_log_incompat =
662 494833 : be32_to_cpu(from->sb_features_log_incompat);
663 : /* crc is only used on disk, not in memory; just init to 0 here. */
664 494833 : to->sb_crc = 0;
665 494833 : to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
666 494833 : to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
667 494833 : to->sb_lsn = be64_to_cpu(from->sb_lsn);
668 : /*
669 : * sb_meta_uuid is only on disk if it differs from sb_uuid and the
670 : * feature flag is set; if not set we keep it only in memory.
671 : */
672 494833 : if (xfs_sb_is_v5(to) &&
673 490019 : (to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID))
674 90 : uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
675 : else
676 494743 : uuid_copy(&to->sb_meta_uuid, &from->sb_uuid);
677 : /* Convert on-disk flags to in-memory flags? */
678 494833 : if (convert_xquota)
679 59579 : xfs_sb_quota_from_disk(to);
680 494833 : }
681 :
682 : void
683 59579 : xfs_sb_from_disk(
684 : struct xfs_sb *to,
685 : struct xfs_dsb *from)
686 : {
687 59579 : __xfs_sb_from_disk(to, from, true);
688 59579 : }
689 :
690 : static void
691 425190 : xfs_sb_quota_to_disk(
692 : struct xfs_dsb *to,
693 : struct xfs_sb *from)
694 : {
695 425190 : uint16_t qflags = from->sb_qflags;
696 :
697 425190 : to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
698 :
699 : /*
700 : * The in-memory superblock quota state matches the v5 on-disk format so
701 : * just write them out and return
702 : */
703 425190 : if (xfs_sb_is_v5(from)) {
704 420514 : to->sb_qflags = cpu_to_be16(from->sb_qflags);
705 420514 : to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
706 420514 : to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
707 420514 : return;
708 : }
709 :
710 : /*
711 : * For older superblocks (v4), the in-core version of sb_qflags do not
712 : * have XFS_OQUOTA_* flags, whereas the on-disk version does. So,
713 : * convert incore XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
714 : */
715 4676 : qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
716 : XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
717 :
718 4676 : if (from->sb_qflags &
719 : (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
720 4304 : qflags |= XFS_OQUOTA_ENFD;
721 4676 : if (from->sb_qflags &
722 : (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
723 4304 : qflags |= XFS_OQUOTA_CHKD;
724 4676 : to->sb_qflags = cpu_to_be16(qflags);
725 :
726 : /*
727 : * GQUOTINO and PQUOTINO cannot be used together in versions
728 : * of superblock that do not have pquotino. from->sb_flags
729 : * tells us which quota is active and should be copied to
730 : * disk. If neither are active, we should NULL the inode.
731 : *
732 : * In all cases, the separate pquotino must remain 0 because it
733 : * is beyond the "end" of the valid non-pquotino superblock.
734 : */
735 4676 : if (from->sb_qflags & XFS_GQUOTA_ACCT)
736 12 : to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
737 4664 : else if (from->sb_qflags & XFS_PQUOTA_ACCT)
738 4304 : to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
739 : else {
740 : /*
741 : * We can't rely on just the fields being logged to tell us
742 : * that it is safe to write NULLFSINO - we should only do that
743 : * if quotas are not actually enabled. Hence only write
744 : * NULLFSINO if both in-core quota inodes are NULL.
745 : */
746 360 : if (from->sb_gquotino == NULLFSINO &&
747 358 : from->sb_pquotino == NULLFSINO)
748 358 : to->sb_gquotino = cpu_to_be64(NULLFSINO);
749 : }
750 :
751 4676 : to->sb_pquotino = 0;
752 : }
753 :
754 : void
755 425186 : xfs_sb_to_disk(
756 : struct xfs_dsb *to,
757 : struct xfs_sb *from)
758 : {
759 425186 : xfs_sb_quota_to_disk(to, from);
760 :
761 425190 : to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
762 425190 : to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
763 425190 : to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
764 425190 : to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
765 425190 : to->sb_rextents = cpu_to_be64(from->sb_rextents);
766 850380 : memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
767 425190 : to->sb_logstart = cpu_to_be64(from->sb_logstart);
768 425190 : to->sb_rootino = cpu_to_be64(from->sb_rootino);
769 425190 : to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
770 425190 : to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
771 425190 : to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
772 425190 : to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
773 425190 : to->sb_agcount = cpu_to_be32(from->sb_agcount);
774 425190 : to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
775 425190 : to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
776 425190 : to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
777 425190 : to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
778 425190 : to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
779 425190 : to->sb_inopblock = cpu_to_be16(from->sb_inopblock);
780 850380 : memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
781 425190 : to->sb_blocklog = from->sb_blocklog;
782 425190 : to->sb_sectlog = from->sb_sectlog;
783 425190 : to->sb_inodelog = from->sb_inodelog;
784 425190 : to->sb_inopblog = from->sb_inopblog;
785 425190 : to->sb_agblklog = from->sb_agblklog;
786 425190 : to->sb_rextslog = from->sb_rextslog;
787 425190 : to->sb_inprogress = from->sb_inprogress;
788 425190 : to->sb_imax_pct = from->sb_imax_pct;
789 425190 : to->sb_icount = cpu_to_be64(from->sb_icount);
790 425190 : to->sb_ifree = cpu_to_be64(from->sb_ifree);
791 425190 : to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
792 425190 : to->sb_frextents = cpu_to_be64(from->sb_frextents);
793 :
794 425190 : to->sb_flags = from->sb_flags;
795 425190 : to->sb_shared_vn = from->sb_shared_vn;
796 425190 : to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
797 425190 : to->sb_unit = cpu_to_be32(from->sb_unit);
798 425190 : to->sb_width = cpu_to_be32(from->sb_width);
799 425190 : to->sb_dirblklog = from->sb_dirblklog;
800 425190 : to->sb_logsectlog = from->sb_logsectlog;
801 425190 : to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
802 425190 : to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
803 :
804 : /*
805 : * We need to ensure that bad_features2 always matches features2.
806 : * Hence we enforce that here rather than having to remember to do it
807 : * everywhere else that updates features2.
808 : */
809 425190 : from->sb_bad_features2 = from->sb_features2;
810 425190 : to->sb_features2 = cpu_to_be32(from->sb_features2);
811 425190 : to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
812 :
813 425190 : if (!xfs_sb_is_v5(from))
814 : return;
815 :
816 420513 : to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
817 841026 : to->sb_features_ro_compat =
818 420513 : cpu_to_be32(from->sb_features_ro_compat);
819 841026 : to->sb_features_incompat =
820 420513 : cpu_to_be32(from->sb_features_incompat);
821 841026 : to->sb_features_log_incompat =
822 420513 : cpu_to_be32(from->sb_features_log_incompat);
823 420513 : to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
824 420513 : to->sb_lsn = cpu_to_be64(from->sb_lsn);
825 420513 : if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID)
826 30 : uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
827 : }
828 :
829 : /*
830 : * If the superblock has the CRC feature bit set or the CRC field is non-null,
831 : * check that the CRC is valid. We check the CRC field is non-null because a
832 : * single bit error could clear the feature bit and unused parts of the
833 : * superblock are supposed to be zero. Hence a non-null crc field indicates that
834 : * we've potentially lost a feature bit and we should check it anyway.
835 : *
836 : * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
837 : * last field in V4 secondary superblocks. So for secondary superblocks,
838 : * we are more forgiving, and ignore CRC failures if the primary doesn't
839 : * indicate that the fs version is V5.
840 : */
841 : static void
842 195050 : xfs_sb_read_verify(
843 : struct xfs_buf *bp)
844 : {
845 195050 : struct xfs_sb sb;
846 195050 : struct xfs_mount *mp = bp->b_mount;
847 195050 : struct xfs_dsb *dsb = bp->b_addr;
848 195050 : int error;
849 :
850 : /*
851 : * open code the version check to avoid needing to convert the entire
852 : * superblock from disk order just to check the version number
853 : */
854 195050 : if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
855 195048 : (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
856 54 : XFS_SB_VERSION_5) ||
857 54 : dsb->sb_crc != 0)) {
858 :
859 194994 : if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
860 : /* Only fail bad secondaries on a known V5 filesystem */
861 4 : if (xfs_buf_daddr(bp) == XFS_SB_DADDR ||
862 : xfs_has_crc(mp)) {
863 4 : error = -EFSBADCRC;
864 4 : goto out_error;
865 : }
866 : }
867 : }
868 :
869 : /*
870 : * Check all the superblock fields. Don't byteswap the xquota flags
871 : * because _verify_common checks the on-disk values.
872 : */
873 195046 : __xfs_sb_from_disk(&sb, dsb, false);
874 195046 : error = xfs_validate_sb_common(mp, bp, &sb);
875 195046 : if (error)
876 26 : goto out_error;
877 195020 : error = xfs_validate_sb_read(mp, &sb);
878 :
879 195050 : out_error:
880 195050 : if (error == -EFSCORRUPTED || error == -EFSBADCRC)
881 24 : xfs_verifier_error(bp, error, __this_address);
882 195026 : else if (error)
883 8 : xfs_buf_ioerror(bp, error);
884 195050 : }
885 :
886 : /*
887 : * We may be probed for a filesystem match, so we may not want to emit
888 : * messages when the superblock buffer is not actually an XFS superblock.
889 : * If we find an XFS superblock, then run a normal, noisy mount because we are
890 : * really going to mount it and want to know about errors.
891 : */
892 : static void
893 0 : xfs_sb_quiet_read_verify(
894 : struct xfs_buf *bp)
895 : {
896 0 : struct xfs_dsb *dsb = bp->b_addr;
897 :
898 0 : if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
899 : /* XFS filesystem, verify noisily! */
900 0 : xfs_sb_read_verify(bp);
901 0 : return;
902 : }
903 : /* quietly fail */
904 0 : xfs_buf_ioerror(bp, -EWRONGFS);
905 : }
906 :
907 : static void
908 240208 : xfs_sb_write_verify(
909 : struct xfs_buf *bp)
910 : {
911 240208 : struct xfs_sb sb;
912 240208 : struct xfs_mount *mp = bp->b_mount;
913 240208 : struct xfs_buf_log_item *bip = bp->b_log_item;
914 240208 : struct xfs_dsb *dsb = bp->b_addr;
915 240208 : int error;
916 :
917 : /*
918 : * Check all the superblock fields. Don't byteswap the xquota flags
919 : * because _verify_common checks the on-disk values.
920 : */
921 240208 : __xfs_sb_from_disk(&sb, dsb, false);
922 240208 : error = xfs_validate_sb_common(mp, bp, &sb);
923 240208 : if (error)
924 0 : goto out_error;
925 240208 : error = xfs_validate_sb_write(mp, bp, &sb);
926 240208 : if (error)
927 4 : goto out_error;
928 :
929 240204 : if (!xfs_sb_is_v5(&sb))
930 240204 : return;
931 :
932 235548 : if (bip)
933 179723 : dsb->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
934 :
935 235548 : xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
936 : return;
937 :
938 4 : out_error:
939 4 : xfs_verifier_error(bp, error, __this_address);
940 : }
941 :
942 : const struct xfs_buf_ops xfs_sb_buf_ops = {
943 : .name = "xfs_sb",
944 : .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
945 : .verify_read = xfs_sb_read_verify,
946 : .verify_write = xfs_sb_write_verify,
947 : };
948 :
949 : const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
950 : .name = "xfs_sb_quiet",
951 : .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) },
952 : .verify_read = xfs_sb_quiet_read_verify,
953 : .verify_write = xfs_sb_write_verify,
954 : };
955 :
956 : /*
957 : * xfs_mount_common
958 : *
959 : * Mount initialization code establishing various mount
960 : * fields from the superblock associated with the given
961 : * mount structure.
962 : *
963 : * Inode geometry are calculated in xfs_ialloc_setup_geometry.
964 : */
965 : void
966 24125 : xfs_sb_mount_common(
967 : struct xfs_mount *mp,
968 : struct xfs_sb *sbp)
969 : {
970 24125 : mp->m_agfrotor = 0;
971 24125 : atomic_set(&mp->m_agirotor, 0);
972 24125 : mp->m_maxagi = mp->m_sb.sb_agcount;
973 24125 : mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
974 24125 : mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
975 24125 : mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
976 24125 : mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
977 24125 : mp->m_blockmask = sbp->sb_blocksize - 1;
978 24125 : mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
979 24125 : mp->m_blockwmask = mp->m_blockwsize - 1;
980 :
981 24125 : mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
982 24125 : mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
983 24125 : mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
984 24125 : mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
985 :
986 24125 : mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
987 24125 : mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
988 24125 : mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
989 24125 : mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
990 :
991 24125 : mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(sbp->sb_blocksize, 1);
992 24125 : mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(sbp->sb_blocksize, 0);
993 24125 : mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2;
994 24125 : mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2;
995 :
996 24125 : mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(sbp->sb_blocksize, true);
997 24125 : mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(sbp->sb_blocksize, false);
998 24125 : mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2;
999 24125 : mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2;
1000 :
1001 24125 : mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
1002 24125 : mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
1003 24125 : mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp);
1004 24125 : }
1005 :
1006 : /*
1007 : * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
1008 : * into the superblock buffer to be logged. It does not provide the higher
1009 : * level of locking that is needed to protect the in-core superblock from
1010 : * concurrent access.
1011 : */
1012 : void
1013 208000 : xfs_log_sb(
1014 : struct xfs_trans *tp)
1015 : {
1016 208000 : struct xfs_mount *mp = tp->t_mountp;
1017 208000 : struct xfs_buf *bp = xfs_trans_getsb(tp);
1018 :
1019 : /*
1020 : * Lazy sb counters don't update the in-core superblock so do that now.
1021 : * If this is at unmount, the counters will be exactly correct, but at
1022 : * any other time they will only be ballpark correct because of
1023 : * reservations that have been taken out percpu counters. If we have an
1024 : * unclean shutdown, this will be corrected by log recovery rebuilding
1025 : * the counters from the AGF block counts.
1026 : *
1027 : * Do not update sb_frextents here because it is not part of the lazy
1028 : * sb counters, despite having a percpu counter. It is always kept
1029 : * consistent with the ondisk rtbitmap by xfs_trans_apply_sb_deltas()
1030 : * and hence we don't need have to update it here.
1031 : */
1032 208000 : if (xfs_has_lazysbcount(mp)) {
1033 207978 : mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount);
1034 207978 : mp->m_sb.sb_ifree = min_t(uint64_t,
1035 : percpu_counter_sum(&mp->m_ifree),
1036 : mp->m_sb.sb_icount);
1037 207978 : mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks);
1038 : }
1039 :
1040 208000 : xfs_sb_to_disk(bp->b_addr, &mp->m_sb);
1041 208000 : xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
1042 208000 : xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb) - 1);
1043 208000 : }
1044 :
1045 : /*
1046 : * xfs_sync_sb
1047 : *
1048 : * Sync the superblock to disk.
1049 : *
1050 : * Note that the caller is responsible for checking the frozen state of the
1051 : * filesystem. This procedure uses the non-blocking transaction allocator and
1052 : * thus will allow modifications to a frozen fs. This is required because this
1053 : * code can be called during the process of freezing where use of the high-level
1054 : * allocator would deadlock.
1055 : */
1056 : int
1057 199558 : xfs_sync_sb(
1058 : struct xfs_mount *mp,
1059 : bool wait)
1060 : {
1061 199558 : struct xfs_trans *tp;
1062 199558 : int error;
1063 :
1064 199558 : error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0,
1065 : XFS_TRANS_NO_WRITECOUNT, &tp);
1066 199558 : if (error)
1067 : return error;
1068 :
1069 199525 : xfs_log_sb(tp);
1070 199525 : if (wait)
1071 96551 : xfs_trans_set_sync(tp);
1072 199525 : return xfs_trans_commit(tp);
1073 : }
1074 :
1075 : /*
1076 : * Update all the secondary superblocks to match the new state of the primary.
1077 : * Because we are completely overwriting all the existing fields in the
1078 : * secondary superblock buffers, there is no need to read them in from disk.
1079 : * Just get a new buffer, stamp it and write it.
1080 : *
1081 : * The sb buffers need to be cached here so that we serialise against other
1082 : * operations that access the secondary superblocks, but we don't want to keep
1083 : * them in memory once it is written so we mark it as a one-shot buffer.
1084 : */
1085 : int
1086 180 : xfs_update_secondary_sbs(
1087 : struct xfs_mount *mp)
1088 : {
1089 180 : struct xfs_perag *pag;
1090 180 : xfs_agnumber_t agno = 1;
1091 180 : int saved_error = 0;
1092 180 : int error = 0;
1093 180 : LIST_HEAD (buffer_list);
1094 :
1095 : /* update secondary superblocks. */
1096 8600 : for_each_perag_from(mp, agno, pag) {
1097 8420 : struct xfs_buf *bp;
1098 :
1099 8420 : error = xfs_buf_get(mp->m_ddev_targp,
1100 8420 : XFS_AG_DADDR(mp, pag->pag_agno, XFS_SB_DADDR),
1101 8420 : XFS_FSS_TO_BB(mp, 1), &bp);
1102 : /*
1103 : * If we get an error reading or writing alternate superblocks,
1104 : * continue. xfs_repair chooses the "best" superblock based
1105 : * on most matches; if we break early, we'll leave more
1106 : * superblocks un-updated than updated, and xfs_repair may
1107 : * pick them over the properly-updated primary.
1108 : */
1109 8420 : if (error) {
1110 0 : xfs_warn(mp,
1111 : "error allocating secondary superblock for ag %d",
1112 : pag->pag_agno);
1113 0 : if (!saved_error)
1114 0 : saved_error = error;
1115 7944 : continue;
1116 : }
1117 :
1118 8420 : bp->b_ops = &xfs_sb_buf_ops;
1119 8420 : xfs_buf_oneshot(bp);
1120 8420 : xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
1121 8420 : xfs_sb_to_disk(bp->b_addr, &mp->m_sb);
1122 8420 : xfs_buf_delwri_queue(bp, &buffer_list);
1123 8420 : xfs_buf_relse(bp);
1124 :
1125 : /* don't hold too many buffers at once */
1126 8420 : if (agno % 16)
1127 7944 : continue;
1128 :
1129 476 : error = xfs_buf_delwri_submit(&buffer_list);
1130 476 : if (error) {
1131 0 : xfs_warn(mp,
1132 : "write error %d updating a secondary superblock near ag %d",
1133 : error, pag->pag_agno);
1134 0 : if (!saved_error)
1135 0 : saved_error = error;
1136 0 : continue;
1137 : }
1138 : }
1139 180 : error = xfs_buf_delwri_submit(&buffer_list);
1140 180 : if (error) {
1141 0 : xfs_warn(mp,
1142 : "write error %d updating a secondary superblock near ag %d",
1143 : error, agno);
1144 : }
1145 :
1146 180 : return saved_error ? saved_error : error;
1147 : }
1148 :
1149 : /*
1150 : * Same behavior as xfs_sync_sb, except that it is always synchronous and it
1151 : * also writes the superblock buffer to disk sector 0 immediately.
1152 : */
1153 : int
1154 52 : xfs_sync_sb_buf(
1155 : struct xfs_mount *mp)
1156 : {
1157 52 : struct xfs_trans *tp;
1158 52 : struct xfs_buf *bp;
1159 52 : int error;
1160 :
1161 52 : error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0, 0, &tp);
1162 52 : if (error)
1163 : return error;
1164 :
1165 52 : bp = xfs_trans_getsb(tp);
1166 52 : xfs_log_sb(tp);
1167 52 : xfs_trans_bhold(tp, bp);
1168 52 : xfs_trans_set_sync(tp);
1169 52 : error = xfs_trans_commit(tp);
1170 52 : if (error)
1171 0 : goto out;
1172 : /*
1173 : * write out the sb buffer to get the changes to disk
1174 : */
1175 52 : error = xfs_bwrite(bp);
1176 52 : out:
1177 52 : xfs_buf_relse(bp);
1178 52 : return error;
1179 : }
1180 :
1181 : void
1182 2004025 : xfs_fs_geometry(
1183 : struct xfs_mount *mp,
1184 : struct xfs_fsop_geom *geo,
1185 : int struct_version)
1186 : {
1187 2004025 : struct xfs_sb *sbp = &mp->m_sb;
1188 :
1189 2004025 : memset(geo, 0, sizeof(struct xfs_fsop_geom));
1190 :
1191 2004025 : geo->blocksize = sbp->sb_blocksize;
1192 2004025 : geo->rtextsize = sbp->sb_rextsize;
1193 2004025 : geo->agblocks = sbp->sb_agblocks;
1194 2004025 : geo->agcount = sbp->sb_agcount;
1195 2004025 : geo->logblocks = sbp->sb_logblocks;
1196 2004025 : geo->sectsize = sbp->sb_sectsize;
1197 2004025 : geo->inodesize = sbp->sb_inodesize;
1198 2004025 : geo->imaxpct = sbp->sb_imax_pct;
1199 2004025 : geo->datablocks = sbp->sb_dblocks;
1200 2004025 : geo->rtblocks = sbp->sb_rblocks;
1201 2004025 : geo->rtextents = sbp->sb_rextents;
1202 2004025 : geo->logstart = sbp->sb_logstart;
1203 2004025 : BUILD_BUG_ON(sizeof(geo->uuid) != sizeof(sbp->sb_uuid));
1204 4008050 : memcpy(geo->uuid, &sbp->sb_uuid, sizeof(sbp->sb_uuid));
1205 :
1206 2004025 : if (struct_version < 2)
1207 : return;
1208 :
1209 2004025 : geo->sunit = sbp->sb_unit;
1210 2004025 : geo->swidth = sbp->sb_width;
1211 :
1212 2004025 : if (struct_version < 3)
1213 : return;
1214 :
1215 2004025 : geo->version = XFS_FSOP_GEOM_VERSION;
1216 2004025 : geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
1217 : XFS_FSOP_GEOM_FLAGS_DIRV2 |
1218 : XFS_FSOP_GEOM_FLAGS_EXTFLG;
1219 2004025 : if (xfs_has_attr(mp))
1220 2019749 : geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR;
1221 2004025 : if (xfs_has_quota(mp))
1222 1633594 : geo->flags |= XFS_FSOP_GEOM_FLAGS_QUOTA;
1223 2004025 : if (xfs_has_align(mp))
1224 2016897 : geo->flags |= XFS_FSOP_GEOM_FLAGS_IALIGN;
1225 2004025 : if (xfs_has_dalign(mp))
1226 596 : geo->flags |= XFS_FSOP_GEOM_FLAGS_DALIGN;
1227 2004025 : if (xfs_has_asciici(mp))
1228 52 : geo->flags |= XFS_FSOP_GEOM_FLAGS_DIRV2CI;
1229 2004025 : if (xfs_has_lazysbcount(mp))
1230 2015140 : geo->flags |= XFS_FSOP_GEOM_FLAGS_LAZYSB;
1231 2004025 : if (xfs_has_attr2(mp))
1232 2016295 : geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR2;
1233 2004025 : if (xfs_has_projid32(mp))
1234 2016302 : geo->flags |= XFS_FSOP_GEOM_FLAGS_PROJID32;
1235 2004025 : if (xfs_has_crc(mp))
1236 2015182 : geo->flags |= XFS_FSOP_GEOM_FLAGS_V5SB;
1237 2004025 : if (xfs_has_ftype(mp))
1238 2015192 : geo->flags |= XFS_FSOP_GEOM_FLAGS_FTYPE;
1239 2004025 : if (xfs_has_finobt(mp))
1240 2015689 : geo->flags |= XFS_FSOP_GEOM_FLAGS_FINOBT;
1241 2004025 : if (xfs_has_sparseinodes(mp))
1242 2016015 : geo->flags |= XFS_FSOP_GEOM_FLAGS_SPINODES;
1243 2004025 : if (xfs_has_rmapbt(mp))
1244 1638788 : geo->flags |= XFS_FSOP_GEOM_FLAGS_RMAPBT;
1245 2004025 : if (xfs_has_reflink(mp))
1246 1639286 : geo->flags |= XFS_FSOP_GEOM_FLAGS_REFLINK;
1247 2004025 : if (xfs_has_bigtime(mp))
1248 2008931 : geo->flags |= XFS_FSOP_GEOM_FLAGS_BIGTIME;
1249 2004025 : if (xfs_has_inobtcounts(mp))
1250 2010075 : geo->flags |= XFS_FSOP_GEOM_FLAGS_INOBTCNT;
1251 2004025 : if (xfs_has_parent(mp))
1252 2004292 : geo->flags |= XFS_FSOP_GEOM_FLAGS_PARENT;
1253 2004025 : if (xfs_has_sector(mp)) {
1254 1995043 : geo->flags |= XFS_FSOP_GEOM_FLAGS_SECTOR;
1255 1995043 : geo->logsectsize = sbp->sb_logsectsize;
1256 : } else {
1257 8982 : geo->logsectsize = BBSIZE;
1258 : }
1259 2004025 : if (xfs_has_large_extent_counts(mp))
1260 1993166 : geo->flags |= XFS_FSOP_GEOM_FLAGS_NREXT64;
1261 2377322 : if (xfs_swapext_supported(mp))
1262 1621271 : geo->flags |= XFS_FSOP_GEOM_FLAGS_ATOMIC_SWAP;
1263 2004025 : geo->rtsectsize = sbp->sb_blocksize;
1264 2004025 : geo->dirblocksize = xfs_dir2_dirblock_bytes(sbp);
1265 :
1266 2004025 : if (struct_version < 4)
1267 : return;
1268 :
1269 2003881 : if (xfs_has_logv2(mp))
1270 1994406 : geo->flags |= XFS_FSOP_GEOM_FLAGS_LOGV2;
1271 :
1272 2003881 : geo->logsunit = sbp->sb_logsunit;
1273 :
1274 2003881 : if (struct_version < 5)
1275 : return;
1276 :
1277 2003881 : geo->version = XFS_FSOP_GEOM_VERSION_V5;
1278 : }
1279 :
1280 : /* Read a secondary superblock. */
1281 : int
1282 444701 : xfs_sb_read_secondary(
1283 : struct xfs_mount *mp,
1284 : struct xfs_trans *tp,
1285 : xfs_agnumber_t agno,
1286 : struct xfs_buf **bpp)
1287 : {
1288 444701 : struct xfs_buf *bp;
1289 444701 : int error;
1290 :
1291 444701 : ASSERT(agno != 0 && agno != NULLAGNUMBER);
1292 444701 : error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
1293 444701 : XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1294 444701 : XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_sb_buf_ops);
1295 444703 : if (xfs_metadata_is_sick(error))
1296 16 : xfs_agno_mark_sick(mp, agno, XFS_SICK_AG_SB);
1297 444703 : if (error)
1298 : return error;
1299 444685 : xfs_buf_set_ref(bp, XFS_SSB_REF);
1300 444684 : *bpp = bp;
1301 444684 : return 0;
1302 : }
1303 :
1304 : /* Get an uninitialised secondary superblock buffer. */
1305 : int
1306 146017 : xfs_sb_get_secondary(
1307 : struct xfs_mount *mp,
1308 : struct xfs_trans *tp,
1309 : xfs_agnumber_t agno,
1310 : struct xfs_buf **bpp)
1311 : {
1312 146017 : struct xfs_buf *bp;
1313 146017 : int error;
1314 :
1315 146017 : ASSERT(agno != 0 && agno != NULLAGNUMBER);
1316 146017 : error = xfs_trans_get_buf(tp, mp->m_ddev_targp,
1317 146017 : XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
1318 146017 : XFS_FSS_TO_BB(mp, 1), 0, &bp);
1319 146016 : if (error)
1320 : return error;
1321 146016 : bp->b_ops = &xfs_sb_buf_ops;
1322 146016 : xfs_buf_oneshot(bp);
1323 146016 : *bpp = bp;
1324 146016 : return 0;
1325 : }
1326 :
1327 : /*
1328 : * sunit, swidth, sectorsize(optional with 0) should be all in bytes,
1329 : * so users won't be confused by values in error messages.
1330 : */
1331 : bool
1332 435228 : xfs_validate_stripe_geometry(
1333 : struct xfs_mount *mp,
1334 : __s64 sunit,
1335 : __s64 swidth,
1336 : int sectorsize,
1337 : bool silent)
1338 : {
1339 435228 : if (swidth > INT_MAX) {
1340 0 : if (!silent)
1341 0 : xfs_notice(mp,
1342 : "stripe width (%lld) is too large", swidth);
1343 0 : return false;
1344 : }
1345 :
1346 435228 : if (sunit > swidth) {
1347 0 : if (!silent)
1348 0 : xfs_notice(mp,
1349 : "stripe unit (%lld) is larger than the stripe width (%lld)", sunit, swidth);
1350 0 : return false;
1351 : }
1352 :
1353 435228 : if (sectorsize && (int)sunit % sectorsize) {
1354 0 : if (!silent)
1355 0 : xfs_notice(mp,
1356 : "stripe unit (%lld) must be a multiple of the sector size (%d)",
1357 : sunit, sectorsize);
1358 0 : return false;
1359 : }
1360 :
1361 435228 : if (sunit && !swidth) {
1362 0 : if (!silent)
1363 0 : xfs_notice(mp,
1364 : "invalid stripe unit (%lld) and stripe width of 0", sunit);
1365 0 : return false;
1366 : }
1367 :
1368 435228 : if (!sunit && swidth) {
1369 0 : if (!silent)
1370 0 : xfs_notice(mp,
1371 : "invalid stripe width (%lld) and stripe unit of 0", swidth);
1372 0 : return false;
1373 : }
1374 :
1375 435228 : if (sunit && (int)swidth % (int)sunit) {
1376 0 : if (!silent)
1377 0 : xfs_notice(mp,
1378 : "stripe width (%lld) must be a multiple of the stripe unit (%lld)",
1379 : swidth, sunit);
1380 0 : return false;
1381 : }
1382 : return true;
1383 : }
|
