Commit | Line | Data |
---|---|---|
0b61f8a4 | 1 | // SPDX-License-Identifier: GPL-2.0 |
ff55068c DC |
2 | /* |
3 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. | |
4 | * All Rights Reserved. | |
ff55068c DC |
5 | */ |
6 | #include "xfs.h" | |
7 | #include "xfs_fs.h" | |
632b89e8 | 8 | #include "xfs_shared.h" |
ff55068c | 9 | #include "xfs_format.h" |
239880ef DC |
10 | #include "xfs_log_format.h" |
11 | #include "xfs_trans_resv.h" | |
ff55068c | 12 | #include "xfs_bit.h" |
5f213ddb | 13 | #include "xfs_sb.h" |
ff55068c | 14 | #include "xfs_mount.h" |
ff55068c DC |
15 | #include "xfs_ialloc.h" |
16 | #include "xfs_alloc.h" | |
ff55068c | 17 | #include "xfs_error.h" |
239880ef | 18 | #include "xfs_trans.h" |
ff55068c | 19 | #include "xfs_buf_item.h" |
a4fbe6ab DC |
20 | #include "xfs_bmap_btree.h" |
21 | #include "xfs_alloc_btree.h" | |
a45086e2 | 22 | #include "xfs_log.h" |
035e00ac | 23 | #include "xfs_rmap_btree.h" |
1946b91c | 24 | #include "xfs_refcount_btree.h" |
c368ebcd | 25 | #include "xfs_da_format.h" |
39353ff6 | 26 | #include "xfs_health.h" |
7f8d3b3c | 27 | #include "xfs_ag.h" |
ff55068c DC |
28 | |
29 | /* | |
30 | * Physical superblock buffer manipulations. Shared with libxfs in userspace. | |
31 | */ | |
32 | ||
fe08cc50 DC |
33 | /* |
34 | * We support all XFS versions newer than a v4 superblock with V2 directories. | |
35 | */ | |
36 | bool | |
37 | xfs_sb_good_version( | |
38 | struct xfs_sb *sbp) | |
39 | { | |
40 | /* all v5 filesystems are supported */ | |
d6837c1a | 41 | if (xfs_sb_is_v5(sbp)) |
fe08cc50 DC |
42 | return true; |
43 | ||
44 | /* versions prior to v4 are not supported */ | |
45 | if (XFS_SB_VERSION_NUM(sbp) < XFS_SB_VERSION_4) | |
46 | return false; | |
47 | ||
48 | /* V4 filesystems need v2 directories and unwritten extents */ | |
49 | if (!(sbp->sb_versionnum & XFS_SB_VERSION_DIRV2BIT)) | |
50 | return false; | |
51 | if (!(sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT)) | |
52 | return false; | |
53 | ||
54 | /* And must not have any unknown v4 feature bits set */ | |
55 | if ((sbp->sb_versionnum & ~XFS_SB_VERSION_OKBITS) || | |
56 | ((sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) && | |
57 | (sbp->sb_features2 & ~XFS_SB_VERSION2_OKBITS))) | |
58 | return false; | |
59 | ||
60 | /* It's a supported v4 filesystem */ | |
61 | return true; | |
62 | } | |
63 | ||
a1d86e8d DC |
64 | uint64_t |
65 | xfs_sb_version_to_features( | |
66 | struct xfs_sb *sbp) | |
67 | { | |
68 | uint64_t features = 0; | |
69 | ||
70 | /* optional V4 features */ | |
71 | if (sbp->sb_rblocks > 0) | |
72 | features |= XFS_FEAT_REALTIME; | |
73 | if (sbp->sb_versionnum & XFS_SB_VERSION_ATTRBIT) | |
74 | features |= XFS_FEAT_ATTR; | |
75 | if (sbp->sb_versionnum & XFS_SB_VERSION_QUOTABIT) | |
76 | features |= XFS_FEAT_QUOTA; | |
77 | if (sbp->sb_versionnum & XFS_SB_VERSION_ALIGNBIT) | |
78 | features |= XFS_FEAT_ALIGN; | |
79 | if (sbp->sb_versionnum & XFS_SB_VERSION_LOGV2BIT) | |
80 | features |= XFS_FEAT_LOGV2; | |
81 | if (sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT) | |
82 | features |= XFS_FEAT_DALIGN; | |
83 | if (sbp->sb_versionnum & XFS_SB_VERSION_EXTFLGBIT) | |
84 | features |= XFS_FEAT_EXTFLG; | |
85 | if (sbp->sb_versionnum & XFS_SB_VERSION_SECTORBIT) | |
86 | features |= XFS_FEAT_SECTOR; | |
87 | if (sbp->sb_versionnum & XFS_SB_VERSION_BORGBIT) | |
88 | features |= XFS_FEAT_ASCIICI; | |
89 | if (sbp->sb_versionnum & XFS_SB_VERSION_MOREBITSBIT) { | |
90 | if (sbp->sb_features2 & XFS_SB_VERSION2_LAZYSBCOUNTBIT) | |
91 | features |= XFS_FEAT_LAZYSBCOUNT; | |
92 | if (sbp->sb_features2 & XFS_SB_VERSION2_ATTR2BIT) | |
93 | features |= XFS_FEAT_ATTR2; | |
94 | if (sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT) | |
95 | features |= XFS_FEAT_PROJID32; | |
96 | if (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE) | |
97 | features |= XFS_FEAT_FTYPE; | |
98 | } | |
99 | ||
d6837c1a | 100 | if (!xfs_sb_is_v5(sbp)) |
a1d86e8d DC |
101 | return features; |
102 | ||
103 | /* Always on V5 features */ | |
104 | features |= XFS_FEAT_ALIGN | XFS_FEAT_LOGV2 | XFS_FEAT_EXTFLG | | |
105 | XFS_FEAT_LAZYSBCOUNT | XFS_FEAT_ATTR2 | XFS_FEAT_PROJID32 | | |
106 | XFS_FEAT_V3INODES | XFS_FEAT_CRC | XFS_FEAT_PQUOTINO; | |
107 | ||
108 | /* Optional V5 features */ | |
109 | if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT) | |
110 | features |= XFS_FEAT_FINOBT; | |
111 | if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT) | |
112 | features |= XFS_FEAT_RMAPBT; | |
113 | if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK) | |
114 | features |= XFS_FEAT_REFLINK; | |
115 | if (sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_INOBTCNT) | |
116 | features |= XFS_FEAT_INOBTCNT; | |
117 | if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_FTYPE) | |
118 | features |= XFS_FEAT_FTYPE; | |
119 | if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES) | |
120 | features |= XFS_FEAT_SPINODES; | |
121 | if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID) | |
122 | features |= XFS_FEAT_META_UUID; | |
123 | if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_BIGTIME) | |
124 | features |= XFS_FEAT_BIGTIME; | |
125 | if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_NEEDSREPAIR) | |
126 | features |= XFS_FEAT_NEEDSREPAIR; | |
127 | return features; | |
128 | } | |
129 | ||
eca383fc | 130 | /* Check all the superblock fields we care about when reading one in. */ |
ff55068c | 131 | STATIC int |
eca383fc DW |
132 | xfs_validate_sb_read( |
133 | struct xfs_mount *mp, | |
134 | struct xfs_sb *sbp) | |
ff55068c | 135 | { |
d6837c1a | 136 | if (!xfs_sb_is_v5(sbp)) |
eca383fc | 137 | return 0; |
ff55068c DC |
138 | |
139 | /* | |
eca383fc DW |
140 | * Version 5 superblock feature mask validation. Reject combinations |
141 | * the kernel cannot support up front before checking anything else. | |
ff55068c | 142 | */ |
eca383fc DW |
143 | if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) { |
144 | xfs_warn(mp, | |
f41febd2 | 145 | "Superblock has unknown compatible features (0x%x) enabled.", |
eca383fc DW |
146 | (sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN)); |
147 | xfs_warn(mp, | |
f41febd2 | 148 | "Using a more recent kernel is recommended."); |
eca383fc | 149 | } |
ff55068c | 150 | |
eca383fc DW |
151 | if (xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { |
152 | xfs_alert(mp, | |
ff55068c | 153 | "Superblock has unknown read-only compatible features (0x%x) enabled.", |
eca383fc DW |
154 | (sbp->sb_features_ro_compat & |
155 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); | |
2e973b2c | 156 | if (!xfs_is_readonly(mp)) { |
eca383fc | 157 | xfs_warn(mp, |
f41febd2 | 158 | "Attempted to mount read-only compatible filesystem read-write."); |
eca383fc | 159 | xfs_warn(mp, |
ff55068c | 160 | "Filesystem can only be safely mounted read only."); |
f41febd2 | 161 | |
2451337d | 162 | return -EINVAL; |
ff55068c | 163 | } |
eca383fc DW |
164 | } |
165 | if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) { | |
166 | xfs_warn(mp, | |
167 | "Superblock has unknown incompatible features (0x%x) enabled.", | |
168 | (sbp->sb_features_incompat & | |
169 | XFS_SB_FEAT_INCOMPAT_UNKNOWN)); | |
170 | xfs_warn(mp, | |
171 | "Filesystem cannot be safely mounted by this kernel."); | |
172 | return -EINVAL; | |
173 | } | |
174 | ||
175 | return 0; | |
176 | } | |
177 | ||
178 | /* Check all the superblock fields we care about when writing one out. */ | |
179 | STATIC int | |
180 | xfs_validate_sb_write( | |
181 | struct xfs_mount *mp, | |
1f31c98d | 182 | struct xfs_buf *bp, |
eca383fc DW |
183 | struct xfs_sb *sbp) |
184 | { | |
8756a5af BD |
185 | /* |
186 | * Carry out additional sb summary counter sanity checks when we write | |
187 | * the superblock. We skip this in the read validator because there | |
188 | * could be newer superblocks in the log and if the values are garbage | |
69775fd1 | 189 | * even after replay we'll recalculate them at the end of log mount. |
1f31c98d DW |
190 | * |
191 | * mkfs has traditionally written zeroed counters to inprogress and | |
192 | * secondary superblocks, so allow this usage to continue because | |
193 | * we never read counters from such superblocks. | |
8756a5af | 194 | */ |
04fcad80 | 195 | if (xfs_buf_daddr(bp) == XFS_SB_DADDR && !sbp->sb_inprogress && |
1f31c98d DW |
196 | (sbp->sb_fdblocks > sbp->sb_dblocks || |
197 | !xfs_verify_icount(mp, sbp->sb_icount) || | |
198 | sbp->sb_ifree > sbp->sb_icount)) { | |
8756a5af BD |
199 | xfs_warn(mp, "SB summary counter sanity check failed"); |
200 | return -EFSCORRUPTED; | |
201 | } | |
202 | ||
d6837c1a | 203 | if (!xfs_sb_is_v5(sbp)) |
eca383fc DW |
204 | return 0; |
205 | ||
9e037cb7 DW |
206 | /* |
207 | * Version 5 superblock feature mask validation. Reject combinations | |
208 | * the kernel cannot support since we checked for unsupported bits in | |
209 | * the read verifier, which means that memory is corrupt. | |
210 | */ | |
211 | if (xfs_sb_has_compat_feature(sbp, XFS_SB_FEAT_COMPAT_UNKNOWN)) { | |
212 | xfs_warn(mp, | |
213 | "Corruption detected in superblock compatible features (0x%x)!", | |
214 | (sbp->sb_features_compat & XFS_SB_FEAT_COMPAT_UNKNOWN)); | |
215 | return -EFSCORRUPTED; | |
216 | } | |
217 | ||
218 | if (xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { | |
219 | xfs_alert(mp, | |
220 | "Corruption detected in superblock read-only compatible features (0x%x)!", | |
221 | (sbp->sb_features_ro_compat & | |
222 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); | |
223 | return -EFSCORRUPTED; | |
224 | } | |
225 | if (xfs_sb_has_incompat_feature(sbp, XFS_SB_FEAT_INCOMPAT_UNKNOWN)) { | |
226 | xfs_warn(mp, | |
227 | "Corruption detected in superblock incompatible features (0x%x)!", | |
228 | (sbp->sb_features_incompat & | |
229 | XFS_SB_FEAT_INCOMPAT_UNKNOWN)); | |
230 | return -EFSCORRUPTED; | |
231 | } | |
232 | if (xfs_sb_has_incompat_log_feature(sbp, | |
233 | XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN)) { | |
234 | xfs_warn(mp, | |
235 | "Corruption detected in superblock incompatible log features (0x%x)!", | |
236 | (sbp->sb_features_log_incompat & | |
237 | XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN)); | |
238 | return -EFSCORRUPTED; | |
239 | } | |
eca383fc DW |
240 | |
241 | /* | |
242 | * We can't read verify the sb LSN because the read verifier is called | |
243 | * before the log is allocated and processed. We know the log is set up | |
244 | * before write verifier calls, so check it here. | |
245 | */ | |
246 | if (!xfs_log_check_lsn(mp, sbp->sb_lsn)) | |
247 | return -EFSCORRUPTED; | |
248 | ||
249 | return 0; | |
250 | } | |
251 | ||
252 | /* Check the validity of the SB. */ | |
253 | STATIC int | |
254 | xfs_validate_sb_common( | |
255 | struct xfs_mount *mp, | |
256 | struct xfs_buf *bp, | |
257 | struct xfs_sb *sbp) | |
258 | { | |
3e6e8afd | 259 | struct xfs_dsb *dsb = bp->b_addr; |
eca383fc DW |
260 | uint32_t agcount = 0; |
261 | uint32_t rem; | |
fe08cc50 | 262 | bool has_dalign; |
eca383fc | 263 | |
39708c20 | 264 | if (!xfs_verify_magic(bp, dsb->sb_magicnum)) { |
eca383fc DW |
265 | xfs_warn(mp, "bad magic number"); |
266 | return -EWRONGFS; | |
267 | } | |
268 | ||
269 | if (!xfs_sb_good_version(sbp)) { | |
270 | xfs_warn(mp, "bad version"); | |
271 | return -EWRONGFS; | |
ff55068c DC |
272 | } |
273 | ||
fe08cc50 DC |
274 | /* |
275 | * Validate feature flags and state | |
276 | */ | |
d6837c1a | 277 | if (xfs_sb_is_v5(sbp)) { |
fe08cc50 DC |
278 | if (sbp->sb_blocksize < XFS_MIN_CRC_BLOCKSIZE) { |
279 | xfs_notice(mp, | |
280 | "Block size (%u bytes) too small for Version 5 superblock (minimum %d bytes)", | |
281 | sbp->sb_blocksize, XFS_MIN_CRC_BLOCKSIZE); | |
282 | return -EFSCORRUPTED; | |
283 | } | |
284 | ||
285 | /* V5 has a separate project quota inode */ | |
ff55068c DC |
286 | if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) { |
287 | xfs_notice(mp, | |
08e96e1a | 288 | "Version 5 of Super block has XFS_OQUOTA bits."); |
2451337d | 289 | return -EFSCORRUPTED; |
ff55068c | 290 | } |
fe08cc50 DC |
291 | |
292 | /* | |
293 | * Full inode chunks must be aligned to inode chunk size when | |
294 | * sparse inodes are enabled to support the sparse chunk | |
295 | * allocation algorithm and prevent overlapping inode records. | |
296 | */ | |
297 | if (sbp->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_SPINODES) { | |
298 | uint32_t align; | |
299 | ||
300 | align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize | |
301 | >> sbp->sb_blocklog; | |
302 | if (sbp->sb_inoalignmt != align) { | |
303 | xfs_warn(mp, | |
304 | "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.", | |
305 | sbp->sb_inoalignmt, align); | |
306 | return -EINVAL; | |
307 | } | |
308 | } | |
ff55068c DC |
309 | } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD | |
310 | XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) { | |
311 | xfs_notice(mp, | |
14506f7a | 312 | "Superblock earlier than Version 5 has XFS_{P|G}QUOTA_{ENFD|CHKD} bits."); |
2451337d | 313 | return -EFSCORRUPTED; |
ff55068c DC |
314 | } |
315 | ||
316 | if (unlikely( | |
317 | sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) { | |
318 | xfs_warn(mp, | |
319 | "filesystem is marked as having an external log; " | |
320 | "specify logdev on the mount command line."); | |
2451337d | 321 | return -EINVAL; |
ff55068c DC |
322 | } |
323 | ||
324 | if (unlikely( | |
325 | sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) { | |
326 | xfs_warn(mp, | |
327 | "filesystem is marked as having an internal log; " | |
328 | "do not specify logdev on the mount command line."); | |
2451337d | 329 | return -EINVAL; |
ff55068c DC |
330 | } |
331 | ||
4bb73d01 DW |
332 | /* Compute agcount for this number of dblocks and agblocks */ |
333 | if (sbp->sb_agblocks) { | |
334 | agcount = div_u64_rem(sbp->sb_dblocks, sbp->sb_agblocks, &rem); | |
335 | if (rem) | |
336 | agcount++; | |
337 | } | |
338 | ||
ff55068c DC |
339 | /* |
340 | * More sanity checking. Most of these were stolen directly from | |
341 | * xfs_repair. | |
342 | */ | |
343 | if (unlikely( | |
344 | sbp->sb_agcount <= 0 || | |
345 | sbp->sb_sectsize < XFS_MIN_SECTORSIZE || | |
346 | sbp->sb_sectsize > XFS_MAX_SECTORSIZE || | |
347 | sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG || | |
348 | sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG || | |
349 | sbp->sb_sectsize != (1 << sbp->sb_sectlog) || | |
350 | sbp->sb_blocksize < XFS_MIN_BLOCKSIZE || | |
351 | sbp->sb_blocksize > XFS_MAX_BLOCKSIZE || | |
352 | sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG || | |
353 | sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG || | |
354 | sbp->sb_blocksize != (1 << sbp->sb_blocklog) || | |
83d230eb | 355 | sbp->sb_dirblklog + sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG || |
ff55068c DC |
356 | sbp->sb_inodesize < XFS_DINODE_MIN_SIZE || |
357 | sbp->sb_inodesize > XFS_DINODE_MAX_SIZE || | |
358 | sbp->sb_inodelog < XFS_DINODE_MIN_LOG || | |
359 | sbp->sb_inodelog > XFS_DINODE_MAX_LOG || | |
360 | sbp->sb_inodesize != (1 << sbp->sb_inodelog) || | |
e1b05723 | 361 | sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE || |
392c6de9 | 362 | sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) || |
4bb73d01 DW |
363 | XFS_FSB_TO_B(mp, sbp->sb_agblocks) < XFS_MIN_AG_BYTES || |
364 | XFS_FSB_TO_B(mp, sbp->sb_agblocks) > XFS_MAX_AG_BYTES || | |
365 | sbp->sb_agblklog != xfs_highbit32(sbp->sb_agblocks - 1) + 1 || | |
366 | agcount == 0 || agcount != sbp->sb_agcount || | |
ff55068c DC |
367 | (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) || |
368 | (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) || | |
369 | (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) || | |
370 | (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) || | |
371 | sbp->sb_dblocks == 0 || | |
372 | sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) || | |
ab3e57b5 DC |
373 | sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) || |
374 | sbp->sb_shared_vn != 0)) { | |
5ef11eb0 | 375 | xfs_notice(mp, "SB sanity check failed"); |
2451337d | 376 | return -EFSCORRUPTED; |
bec9d48d DW |
377 | } |
378 | ||
f8e566c0 DW |
379 | /* Validate the realtime geometry; stolen from xfs_repair */ |
380 | if (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE || | |
381 | sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) { | |
382 | xfs_notice(mp, | |
383 | "realtime extent sanity check failed"); | |
384 | return -EFSCORRUPTED; | |
385 | } | |
386 | ||
387 | if (sbp->sb_rblocks == 0) { | |
388 | if (sbp->sb_rextents != 0 || sbp->sb_rbmblocks != 0 || | |
389 | sbp->sb_rextslog != 0 || sbp->sb_frextents != 0) { | |
390 | xfs_notice(mp, | |
391 | "realtime zeroed geometry check failed"); | |
392 | return -EFSCORRUPTED; | |
393 | } | |
394 | } else { | |
395 | uint64_t rexts; | |
396 | uint64_t rbmblocks; | |
397 | ||
398 | rexts = div_u64(sbp->sb_rblocks, sbp->sb_rextsize); | |
399 | rbmblocks = howmany_64(sbp->sb_rextents, | |
400 | NBBY * sbp->sb_blocksize); | |
401 | ||
402 | if (sbp->sb_rextents != rexts || | |
403 | sbp->sb_rextslog != xfs_highbit32(sbp->sb_rextents) || | |
404 | sbp->sb_rbmblocks != rbmblocks) { | |
405 | xfs_notice(mp, | |
406 | "realtime geometry sanity check failed"); | |
407 | return -EFSCORRUPTED; | |
408 | } | |
409 | } | |
410 | ||
7bc1fea9 GX |
411 | /* |
412 | * Either (sb_unit and !hasdalign) or (!sb_unit and hasdalign) | |
413 | * would imply the image is corrupted. | |
414 | */ | |
fe08cc50 DC |
415 | has_dalign = sbp->sb_versionnum & XFS_SB_VERSION_DALIGNBIT; |
416 | if (!!sbp->sb_unit ^ has_dalign) { | |
fa4ca9c5 DC |
417 | xfs_notice(mp, "SB stripe alignment sanity check failed"); |
418 | return -EFSCORRUPTED; | |
fa4ca9c5 DC |
419 | } |
420 | ||
7bc1fea9 GX |
421 | if (!xfs_validate_stripe_geometry(mp, XFS_FSB_TO_B(mp, sbp->sb_unit), |
422 | XFS_FSB_TO_B(mp, sbp->sb_width), 0, false)) | |
423 | return -EFSCORRUPTED; | |
fa4ca9c5 | 424 | |
ff55068c DC |
425 | /* |
426 | * Currently only very few inode sizes are supported. | |
427 | */ | |
428 | switch (sbp->sb_inodesize) { | |
429 | case 256: | |
430 | case 512: | |
431 | case 1024: | |
432 | case 2048: | |
433 | break; | |
434 | default: | |
435 | xfs_warn(mp, "inode size of %d bytes not supported", | |
436 | sbp->sb_inodesize); | |
2451337d | 437 | return -ENOSYS; |
ff55068c DC |
438 | } |
439 | ||
ff55068c DC |
440 | return 0; |
441 | } | |
442 | ||
443 | void | |
444 | xfs_sb_quota_from_disk(struct xfs_sb *sbp) | |
445 | { | |
446 | /* | |
447 | * older mkfs doesn't initialize quota inodes to NULLFSINO. This | |
448 | * leads to in-core values having two different values for a quota | |
449 | * inode to be invalid: 0 and NULLFSINO. Change it to a single value | |
450 | * NULLFSINO. | |
451 | * | |
452 | * Note that this change affect only the in-core values. These | |
453 | * values are not written back to disk unless any quota information | |
454 | * is written to the disk. Even in that case, sb_pquotino field is | |
455 | * not written to disk unless the superblock supports pquotino. | |
456 | */ | |
457 | if (sbp->sb_uquotino == 0) | |
458 | sbp->sb_uquotino = NULLFSINO; | |
459 | if (sbp->sb_gquotino == 0) | |
460 | sbp->sb_gquotino = NULLFSINO; | |
461 | if (sbp->sb_pquotino == 0) | |
462 | sbp->sb_pquotino = NULLFSINO; | |
463 | ||
464 | /* | |
465 | * We need to do these manipilations only if we are working | |
466 | * with an older version of on-disk superblock. | |
467 | */ | |
d6837c1a | 468 | if (xfs_sb_is_v5(sbp)) |
ff55068c DC |
469 | return; |
470 | ||
471 | if (sbp->sb_qflags & XFS_OQUOTA_ENFD) | |
472 | sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? | |
473 | XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD; | |
474 | if (sbp->sb_qflags & XFS_OQUOTA_CHKD) | |
475 | sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? | |
476 | XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD; | |
477 | sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD); | |
478 | ||
e6fc6fcf ES |
479 | if (sbp->sb_qflags & XFS_PQUOTA_ACCT && |
480 | sbp->sb_gquotino != NULLFSINO) { | |
ff55068c DC |
481 | /* |
482 | * In older version of superblock, on-disk superblock only | |
483 | * has sb_gquotino, and in-core superblock has both sb_gquotino | |
484 | * and sb_pquotino. But, only one of them is supported at any | |
485 | * point of time. So, if PQUOTA is set in disk superblock, | |
e6fc6fcf ES |
486 | * copy over sb_gquotino to sb_pquotino. The NULLFSINO test |
487 | * above is to make sure we don't do this twice and wipe them | |
488 | * both out! | |
ff55068c DC |
489 | */ |
490 | sbp->sb_pquotino = sbp->sb_gquotino; | |
491 | sbp->sb_gquotino = NULLFSINO; | |
492 | } | |
493 | } | |
494 | ||
5ef828c4 ES |
495 | static void |
496 | __xfs_sb_from_disk( | |
ff55068c | 497 | struct xfs_sb *to, |
5ef828c4 ES |
498 | xfs_dsb_t *from, |
499 | bool convert_xquota) | |
ff55068c DC |
500 | { |
501 | to->sb_magicnum = be32_to_cpu(from->sb_magicnum); | |
502 | to->sb_blocksize = be32_to_cpu(from->sb_blocksize); | |
503 | to->sb_dblocks = be64_to_cpu(from->sb_dblocks); | |
504 | to->sb_rblocks = be64_to_cpu(from->sb_rblocks); | |
505 | to->sb_rextents = be64_to_cpu(from->sb_rextents); | |
506 | memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid)); | |
507 | to->sb_logstart = be64_to_cpu(from->sb_logstart); | |
508 | to->sb_rootino = be64_to_cpu(from->sb_rootino); | |
509 | to->sb_rbmino = be64_to_cpu(from->sb_rbmino); | |
510 | to->sb_rsumino = be64_to_cpu(from->sb_rsumino); | |
511 | to->sb_rextsize = be32_to_cpu(from->sb_rextsize); | |
512 | to->sb_agblocks = be32_to_cpu(from->sb_agblocks); | |
513 | to->sb_agcount = be32_to_cpu(from->sb_agcount); | |
514 | to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks); | |
515 | to->sb_logblocks = be32_to_cpu(from->sb_logblocks); | |
516 | to->sb_versionnum = be16_to_cpu(from->sb_versionnum); | |
517 | to->sb_sectsize = be16_to_cpu(from->sb_sectsize); | |
518 | to->sb_inodesize = be16_to_cpu(from->sb_inodesize); | |
519 | to->sb_inopblock = be16_to_cpu(from->sb_inopblock); | |
520 | memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname)); | |
521 | to->sb_blocklog = from->sb_blocklog; | |
522 | to->sb_sectlog = from->sb_sectlog; | |
523 | to->sb_inodelog = from->sb_inodelog; | |
524 | to->sb_inopblog = from->sb_inopblog; | |
525 | to->sb_agblklog = from->sb_agblklog; | |
526 | to->sb_rextslog = from->sb_rextslog; | |
527 | to->sb_inprogress = from->sb_inprogress; | |
528 | to->sb_imax_pct = from->sb_imax_pct; | |
529 | to->sb_icount = be64_to_cpu(from->sb_icount); | |
530 | to->sb_ifree = be64_to_cpu(from->sb_ifree); | |
531 | to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks); | |
532 | to->sb_frextents = be64_to_cpu(from->sb_frextents); | |
533 | to->sb_uquotino = be64_to_cpu(from->sb_uquotino); | |
534 | to->sb_gquotino = be64_to_cpu(from->sb_gquotino); | |
535 | to->sb_qflags = be16_to_cpu(from->sb_qflags); | |
536 | to->sb_flags = from->sb_flags; | |
537 | to->sb_shared_vn = from->sb_shared_vn; | |
538 | to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt); | |
539 | to->sb_unit = be32_to_cpu(from->sb_unit); | |
540 | to->sb_width = be32_to_cpu(from->sb_width); | |
541 | to->sb_dirblklog = from->sb_dirblklog; | |
542 | to->sb_logsectlog = from->sb_logsectlog; | |
543 | to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize); | |
544 | to->sb_logsunit = be32_to_cpu(from->sb_logsunit); | |
545 | to->sb_features2 = be32_to_cpu(from->sb_features2); | |
546 | to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2); | |
547 | to->sb_features_compat = be32_to_cpu(from->sb_features_compat); | |
548 | to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat); | |
549 | to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat); | |
550 | to->sb_features_log_incompat = | |
551 | be32_to_cpu(from->sb_features_log_incompat); | |
04dd1a0d ES |
552 | /* crc is only used on disk, not in memory; just init to 0 here. */ |
553 | to->sb_crc = 0; | |
fb4f2b4e | 554 | to->sb_spino_align = be32_to_cpu(from->sb_spino_align); |
ff55068c DC |
555 | to->sb_pquotino = be64_to_cpu(from->sb_pquotino); |
556 | to->sb_lsn = be64_to_cpu(from->sb_lsn); | |
ce748eaa ES |
557 | /* |
558 | * sb_meta_uuid is only on disk if it differs from sb_uuid and the | |
559 | * feature flag is set; if not set we keep it only in memory. | |
560 | */ | |
d6837c1a | 561 | if (xfs_sb_is_v5(to) && |
fe08cc50 | 562 | (to->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID)) |
ce748eaa ES |
563 | uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid); |
564 | else | |
565 | uuid_copy(&to->sb_meta_uuid, &from->sb_uuid); | |
5ef828c4 ES |
566 | /* Convert on-disk flags to in-memory flags? */ |
567 | if (convert_xquota) | |
568 | xfs_sb_quota_from_disk(to); | |
569 | } | |
570 | ||
571 | void | |
572 | xfs_sb_from_disk( | |
573 | struct xfs_sb *to, | |
574 | xfs_dsb_t *from) | |
575 | { | |
576 | __xfs_sb_from_disk(to, from, true); | |
ff55068c DC |
577 | } |
578 | ||
4d11a402 | 579 | static void |
ff55068c | 580 | xfs_sb_quota_to_disk( |
4d11a402 DC |
581 | struct xfs_dsb *to, |
582 | struct xfs_sb *from) | |
ff55068c | 583 | { |
c8ce540d | 584 | uint16_t qflags = from->sb_qflags; |
ff55068c | 585 | |
4d11a402 | 586 | to->sb_uquotino = cpu_to_be64(from->sb_uquotino); |
fe08cc50 DC |
587 | |
588 | /* | |
589 | * The in-memory superblock quota state matches the v5 on-disk format so | |
590 | * just write them out and return | |
591 | */ | |
d6837c1a | 592 | if (xfs_sb_is_v5(from)) { |
4d11a402 DC |
593 | to->sb_qflags = cpu_to_be16(from->sb_qflags); |
594 | to->sb_gquotino = cpu_to_be64(from->sb_gquotino); | |
595 | to->sb_pquotino = cpu_to_be64(from->sb_pquotino); | |
596 | return; | |
597 | } | |
598 | ||
ff55068c | 599 | /* |
fe08cc50 DC |
600 | * For older superblocks (v4), the in-core version of sb_qflags do not |
601 | * have XFS_OQUOTA_* flags, whereas the on-disk version does. So, | |
602 | * convert incore XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags. | |
ff55068c | 603 | */ |
4d11a402 DC |
604 | qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD | |
605 | XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD); | |
ff55068c | 606 | |
4d11a402 DC |
607 | if (from->sb_qflags & |
608 | (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD)) | |
609 | qflags |= XFS_OQUOTA_ENFD; | |
610 | if (from->sb_qflags & | |
611 | (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) | |
612 | qflags |= XFS_OQUOTA_CHKD; | |
613 | to->sb_qflags = cpu_to_be16(qflags); | |
ff55068c DC |
614 | |
615 | /* | |
4d11a402 DC |
616 | * GQUOTINO and PQUOTINO cannot be used together in versions |
617 | * of superblock that do not have pquotino. from->sb_flags | |
618 | * tells us which quota is active and should be copied to | |
619 | * disk. If neither are active, we should NULL the inode. | |
03e01349 | 620 | * |
4d11a402 | 621 | * In all cases, the separate pquotino must remain 0 because it |
b63da6c8 | 622 | * is beyond the "end" of the valid non-pquotino superblock. |
ff55068c | 623 | */ |
4d11a402 | 624 | if (from->sb_qflags & XFS_GQUOTA_ACCT) |
ff55068c | 625 | to->sb_gquotino = cpu_to_be64(from->sb_gquotino); |
4d11a402 | 626 | else if (from->sb_qflags & XFS_PQUOTA_ACCT) |
ff55068c | 627 | to->sb_gquotino = cpu_to_be64(from->sb_pquotino); |
03e01349 DC |
628 | else { |
629 | /* | |
630 | * We can't rely on just the fields being logged to tell us | |
631 | * that it is safe to write NULLFSINO - we should only do that | |
632 | * if quotas are not actually enabled. Hence only write | |
633 | * NULLFSINO if both in-core quota inodes are NULL. | |
634 | */ | |
635 | if (from->sb_gquotino == NULLFSINO && | |
636 | from->sb_pquotino == NULLFSINO) | |
637 | to->sb_gquotino = cpu_to_be64(NULLFSINO); | |
638 | } | |
ff55068c | 639 | |
4d11a402 | 640 | to->sb_pquotino = 0; |
ff55068c DC |
641 | } |
642 | ||
ff55068c DC |
643 | void |
644 | xfs_sb_to_disk( | |
4d11a402 DC |
645 | struct xfs_dsb *to, |
646 | struct xfs_sb *from) | |
ff55068c | 647 | { |
4d11a402 DC |
648 | xfs_sb_quota_to_disk(to, from); |
649 | ||
650 | to->sb_magicnum = cpu_to_be32(from->sb_magicnum); | |
651 | to->sb_blocksize = cpu_to_be32(from->sb_blocksize); | |
652 | to->sb_dblocks = cpu_to_be64(from->sb_dblocks); | |
653 | to->sb_rblocks = cpu_to_be64(from->sb_rblocks); | |
654 | to->sb_rextents = cpu_to_be64(from->sb_rextents); | |
655 | memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid)); | |
656 | to->sb_logstart = cpu_to_be64(from->sb_logstart); | |
657 | to->sb_rootino = cpu_to_be64(from->sb_rootino); | |
658 | to->sb_rbmino = cpu_to_be64(from->sb_rbmino); | |
659 | to->sb_rsumino = cpu_to_be64(from->sb_rsumino); | |
660 | to->sb_rextsize = cpu_to_be32(from->sb_rextsize); | |
661 | to->sb_agblocks = cpu_to_be32(from->sb_agblocks); | |
662 | to->sb_agcount = cpu_to_be32(from->sb_agcount); | |
663 | to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks); | |
664 | to->sb_logblocks = cpu_to_be32(from->sb_logblocks); | |
665 | to->sb_versionnum = cpu_to_be16(from->sb_versionnum); | |
666 | to->sb_sectsize = cpu_to_be16(from->sb_sectsize); | |
667 | to->sb_inodesize = cpu_to_be16(from->sb_inodesize); | |
668 | to->sb_inopblock = cpu_to_be16(from->sb_inopblock); | |
669 | memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname)); | |
670 | to->sb_blocklog = from->sb_blocklog; | |
671 | to->sb_sectlog = from->sb_sectlog; | |
672 | to->sb_inodelog = from->sb_inodelog; | |
673 | to->sb_inopblog = from->sb_inopblog; | |
674 | to->sb_agblklog = from->sb_agblklog; | |
675 | to->sb_rextslog = from->sb_rextslog; | |
676 | to->sb_inprogress = from->sb_inprogress; | |
677 | to->sb_imax_pct = from->sb_imax_pct; | |
678 | to->sb_icount = cpu_to_be64(from->sb_icount); | |
679 | to->sb_ifree = cpu_to_be64(from->sb_ifree); | |
680 | to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks); | |
681 | to->sb_frextents = cpu_to_be64(from->sb_frextents); | |
ff55068c | 682 | |
4d11a402 DC |
683 | to->sb_flags = from->sb_flags; |
684 | to->sb_shared_vn = from->sb_shared_vn; | |
685 | to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt); | |
686 | to->sb_unit = cpu_to_be32(from->sb_unit); | |
687 | to->sb_width = cpu_to_be32(from->sb_width); | |
688 | to->sb_dirblklog = from->sb_dirblklog; | |
689 | to->sb_logsectlog = from->sb_logsectlog; | |
690 | to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize); | |
691 | to->sb_logsunit = cpu_to_be32(from->sb_logsunit); | |
074e427b DC |
692 | |
693 | /* | |
694 | * We need to ensure that bad_features2 always matches features2. | |
695 | * Hence we enforce that here rather than having to remember to do it | |
696 | * everywhere else that updates features2. | |
697 | */ | |
698 | from->sb_bad_features2 = from->sb_features2; | |
4d11a402 DC |
699 | to->sb_features2 = cpu_to_be32(from->sb_features2); |
700 | to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2); | |
701 | ||
d6837c1a DC |
702 | if (!xfs_sb_is_v5(from)) |
703 | return; | |
704 | ||
705 | to->sb_features_compat = cpu_to_be32(from->sb_features_compat); | |
706 | to->sb_features_ro_compat = | |
707 | cpu_to_be32(from->sb_features_ro_compat); | |
708 | to->sb_features_incompat = | |
709 | cpu_to_be32(from->sb_features_incompat); | |
710 | to->sb_features_log_incompat = | |
711 | cpu_to_be32(from->sb_features_log_incompat); | |
712 | to->sb_spino_align = cpu_to_be32(from->sb_spino_align); | |
713 | to->sb_lsn = cpu_to_be64(from->sb_lsn); | |
714 | if (from->sb_features_incompat & XFS_SB_FEAT_INCOMPAT_META_UUID) | |
715 | uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid); | |
ff55068c DC |
716 | } |
717 | ||
ff55068c DC |
718 | /* |
719 | * If the superblock has the CRC feature bit set or the CRC field is non-null, | |
720 | * check that the CRC is valid. We check the CRC field is non-null because a | |
721 | * single bit error could clear the feature bit and unused parts of the | |
722 | * superblock are supposed to be zero. Hence a non-null crc field indicates that | |
723 | * we've potentially lost a feature bit and we should check it anyway. | |
10e6e65d ES |
724 | * |
725 | * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the | |
726 | * last field in V4 secondary superblocks. So for secondary superblocks, | |
727 | * we are more forgiving, and ignore CRC failures if the primary doesn't | |
728 | * indicate that the fs version is V5. | |
ff55068c DC |
729 | */ |
730 | static void | |
731 | xfs_sb_read_verify( | |
eca383fc | 732 | struct xfs_buf *bp) |
ff55068c | 733 | { |
eca383fc | 734 | struct xfs_sb sb; |
dbd329f1 | 735 | struct xfs_mount *mp = bp->b_mount; |
3e6e8afd | 736 | struct xfs_dsb *dsb = bp->b_addr; |
eca383fc | 737 | int error; |
ff55068c DC |
738 | |
739 | /* | |
740 | * open code the version check to avoid needing to convert the entire | |
741 | * superblock from disk order just to check the version number | |
742 | */ | |
743 | if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) && | |
744 | (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) == | |
745 | XFS_SB_VERSION_5) || | |
746 | dsb->sb_crc != 0)) { | |
747 | ||
51582170 | 748 | if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) { |
10e6e65d | 749 | /* Only fail bad secondaries on a known V5 filesystem */ |
9343ee76 | 750 | if (xfs_buf_daddr(bp) == XFS_SB_DADDR || |
fe08cc50 | 751 | xfs_has_crc(mp)) { |
2451337d | 752 | error = -EFSBADCRC; |
10e6e65d ES |
753 | goto out_error; |
754 | } | |
ff55068c DC |
755 | } |
756 | } | |
eca383fc DW |
757 | |
758 | /* | |
759 | * Check all the superblock fields. Don't byteswap the xquota flags | |
760 | * because _verify_common checks the on-disk values. | |
761 | */ | |
3e6e8afd | 762 | __xfs_sb_from_disk(&sb, dsb, false); |
eca383fc DW |
763 | error = xfs_validate_sb_common(mp, bp, &sb); |
764 | if (error) | |
765 | goto out_error; | |
766 | error = xfs_validate_sb_read(mp, &sb); | |
ff55068c DC |
767 | |
768 | out_error: | |
31ca03c9 | 769 | if (error == -EFSCORRUPTED || error == -EFSBADCRC) |
bc1a09b8 | 770 | xfs_verifier_error(bp, error, __this_address); |
31ca03c9 | 771 | else if (error) |
ff55068c | 772 | xfs_buf_ioerror(bp, error); |
ff55068c DC |
773 | } |
774 | ||
775 | /* | |
776 | * We may be probed for a filesystem match, so we may not want to emit | |
777 | * messages when the superblock buffer is not actually an XFS superblock. | |
2533787a | 778 | * If we find an XFS superblock, then run a normal, noisy mount because we are |
ff55068c DC |
779 | * really going to mount it and want to know about errors. |
780 | */ | |
781 | static void | |
782 | xfs_sb_quiet_read_verify( | |
783 | struct xfs_buf *bp) | |
784 | { | |
3e6e8afd | 785 | struct xfs_dsb *dsb = bp->b_addr; |
ff55068c | 786 | |
ff55068c DC |
787 | if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) { |
788 | /* XFS filesystem, verify noisily! */ | |
789 | xfs_sb_read_verify(bp); | |
790 | return; | |
791 | } | |
792 | /* quietly fail */ | |
2451337d | 793 | xfs_buf_ioerror(bp, -EWRONGFS); |
ff55068c DC |
794 | } |
795 | ||
796 | static void | |
797 | xfs_sb_write_verify( | |
798 | struct xfs_buf *bp) | |
799 | { | |
eca383fc | 800 | struct xfs_sb sb; |
dbd329f1 | 801 | struct xfs_mount *mp = bp->b_mount; |
fb1755a6 | 802 | struct xfs_buf_log_item *bip = bp->b_log_item; |
3e6e8afd | 803 | struct xfs_dsb *dsb = bp->b_addr; |
ff55068c DC |
804 | int error; |
805 | ||
eca383fc DW |
806 | /* |
807 | * Check all the superblock fields. Don't byteswap the xquota flags | |
808 | * because _verify_common checks the on-disk values. | |
809 | */ | |
3e6e8afd | 810 | __xfs_sb_from_disk(&sb, dsb, false); |
eca383fc DW |
811 | error = xfs_validate_sb_common(mp, bp, &sb); |
812 | if (error) | |
813 | goto out_error; | |
1f31c98d | 814 | error = xfs_validate_sb_write(mp, bp, &sb); |
eca383fc DW |
815 | if (error) |
816 | goto out_error; | |
ff55068c | 817 | |
d6837c1a | 818 | if (!xfs_sb_is_v5(&sb)) |
ff55068c DC |
819 | return; |
820 | ||
821 | if (bip) | |
3e6e8afd | 822 | dsb->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn); |
ff55068c | 823 | |
f1dbcd7e | 824 | xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF); |
eca383fc DW |
825 | return; |
826 | ||
827 | out_error: | |
828 | xfs_verifier_error(bp, error, __this_address); | |
ff55068c DC |
829 | } |
830 | ||
831 | const struct xfs_buf_ops xfs_sb_buf_ops = { | |
233135b7 | 832 | .name = "xfs_sb", |
39708c20 | 833 | .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) }, |
ff55068c DC |
834 | .verify_read = xfs_sb_read_verify, |
835 | .verify_write = xfs_sb_write_verify, | |
836 | }; | |
837 | ||
838 | const struct xfs_buf_ops xfs_sb_quiet_buf_ops = { | |
233135b7 | 839 | .name = "xfs_sb_quiet", |
39708c20 | 840 | .magic = { cpu_to_be32(XFS_SB_MAGIC), cpu_to_be32(XFS_SB_MAGIC) }, |
ff55068c DC |
841 | .verify_read = xfs_sb_quiet_read_verify, |
842 | .verify_write = xfs_sb_write_verify, | |
843 | }; | |
844 | ||
845 | /* | |
846 | * xfs_mount_common | |
847 | * | |
848 | * Mount initialization code establishing various mount | |
849 | * fields from the superblock associated with the given | |
494dba7b DW |
850 | * mount structure. |
851 | * | |
852 | * Inode geometry are calculated in xfs_ialloc_setup_geometry. | |
ff55068c DC |
853 | */ |
854 | void | |
855 | xfs_sb_mount_common( | |
ef325959 DW |
856 | struct xfs_mount *mp, |
857 | struct xfs_sb *sbp) | |
ff55068c DC |
858 | { |
859 | mp->m_agfrotor = mp->m_agirotor = 0; | |
ff55068c DC |
860 | mp->m_maxagi = mp->m_sb.sb_agcount; |
861 | mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG; | |
862 | mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT; | |
863 | mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT; | |
864 | mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1; | |
ff55068c DC |
865 | mp->m_blockmask = sbp->sb_blocksize - 1; |
866 | mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG; | |
867 | mp->m_blockwmask = mp->m_blockwsize - 1; | |
868 | ||
869 | mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1); | |
870 | mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0); | |
871 | mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2; | |
872 | mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2; | |
873 | ||
ff55068c DC |
874 | mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1); |
875 | mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0); | |
876 | mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2; | |
877 | mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2; | |
878 | ||
a1f69417 ES |
879 | mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(sbp->sb_blocksize, 1); |
880 | mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(sbp->sb_blocksize, 0); | |
035e00ac DW |
881 | mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2; |
882 | mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2; | |
883 | ||
a1f69417 ES |
884 | mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(sbp->sb_blocksize, true); |
885 | mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(sbp->sb_blocksize, false); | |
1946b91c DW |
886 | mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2; |
887 | mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2; | |
888 | ||
ff55068c | 889 | mp->m_bsize = XFS_FSB_TO_BB(mp, 1); |
52548852 DW |
890 | mp->m_alloc_set_aside = xfs_alloc_set_aside(mp); |
891 | mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp); | |
ff55068c DC |
892 | } |
893 | ||
ff55068c | 894 | /* |
61e63ecb DC |
895 | * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock |
896 | * into the superblock buffer to be logged. It does not provide the higher | |
897 | * level of locking that is needed to protect the in-core superblock from | |
898 | * concurrent access. | |
ff55068c DC |
899 | */ |
900 | void | |
61e63ecb | 901 | xfs_log_sb( |
4d11a402 | 902 | struct xfs_trans *tp) |
ff55068c | 903 | { |
4d11a402 | 904 | struct xfs_mount *mp = tp->t_mountp; |
cead0b10 | 905 | struct xfs_buf *bp = xfs_trans_getsb(tp); |
ff55068c | 906 | |
6543990a DC |
907 | /* |
908 | * Lazy sb counters don't update the in-core superblock so do that now. | |
909 | * If this is at unmount, the counters will be exactly correct, but at | |
910 | * any other time they will only be ballpark correct because of | |
911 | * reservations that have been taken out percpu counters. If we have an | |
912 | * unclean shutdown, this will be corrected by log recovery rebuilding | |
913 | * the counters from the AGF block counts. | |
914 | */ | |
ebd9027d | 915 | if (xfs_has_lazysbcount(mp)) { |
6543990a DC |
916 | mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount); |
917 | mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree); | |
918 | mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks); | |
919 | } | |
501ab323 | 920 | |
3e6e8afd | 921 | xfs_sb_to_disk(bp->b_addr, &mp->m_sb); |
ff55068c | 922 | xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF); |
6f4ff81a | 923 | xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb) - 1); |
ff55068c | 924 | } |
61e63ecb DC |
925 | |
926 | /* | |
927 | * xfs_sync_sb | |
928 | * | |
929 | * Sync the superblock to disk. | |
930 | * | |
931 | * Note that the caller is responsible for checking the frozen state of the | |
932 | * filesystem. This procedure uses the non-blocking transaction allocator and | |
933 | * thus will allow modifications to a frozen fs. This is required because this | |
934 | * code can be called during the process of freezing where use of the high-level | |
935 | * allocator would deadlock. | |
936 | */ | |
937 | int | |
938 | xfs_sync_sb( | |
939 | struct xfs_mount *mp, | |
940 | bool wait) | |
941 | { | |
942 | struct xfs_trans *tp; | |
943 | int error; | |
944 | ||
253f4911 CH |
945 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0, |
946 | XFS_TRANS_NO_WRITECOUNT, &tp); | |
947 | if (error) | |
61e63ecb | 948 | return error; |
61e63ecb DC |
949 | |
950 | xfs_log_sb(tp); | |
951 | if (wait) | |
952 | xfs_trans_set_sync(tp); | |
70393313 | 953 | return xfs_trans_commit(tp); |
61e63ecb | 954 | } |
c368ebcd | 955 | |
b16817b6 DC |
956 | /* |
957 | * Update all the secondary superblocks to match the new state of the primary. | |
958 | * Because we are completely overwriting all the existing fields in the | |
959 | * secondary superblock buffers, there is no need to read them in from disk. | |
960 | * Just get a new buffer, stamp it and write it. | |
961 | * | |
962 | * The sb buffers need to be cached here so that we serialise against other | |
963 | * operations that access the secondary superblocks, but we don't want to keep | |
964 | * them in memory once it is written so we mark it as a one-shot buffer. | |
965 | */ | |
966 | int | |
967 | xfs_update_secondary_sbs( | |
968 | struct xfs_mount *mp) | |
969 | { | |
7f8d3b3c DC |
970 | struct xfs_perag *pag; |
971 | xfs_agnumber_t agno = 1; | |
b16817b6 DC |
972 | int saved_error = 0; |
973 | int error = 0; | |
974 | LIST_HEAD (buffer_list); | |
975 | ||
976 | /* update secondary superblocks. */ | |
7f8d3b3c | 977 | for_each_perag_from(mp, agno, pag) { |
b16817b6 DC |
978 | struct xfs_buf *bp; |
979 | ||
841263e9 | 980 | error = xfs_buf_get(mp->m_ddev_targp, |
7f8d3b3c | 981 | XFS_AG_DADDR(mp, pag->pag_agno, XFS_SB_DADDR), |
841263e9 | 982 | XFS_FSS_TO_BB(mp, 1), &bp); |
b16817b6 DC |
983 | /* |
984 | * If we get an error reading or writing alternate superblocks, | |
985 | * continue. xfs_repair chooses the "best" superblock based | |
986 | * on most matches; if we break early, we'll leave more | |
987 | * superblocks un-updated than updated, and xfs_repair may | |
988 | * pick them over the properly-updated primary. | |
989 | */ | |
841263e9 | 990 | if (error) { |
b16817b6 DC |
991 | xfs_warn(mp, |
992 | "error allocating secondary superblock for ag %d", | |
7f8d3b3c | 993 | pag->pag_agno); |
b16817b6 | 994 | if (!saved_error) |
841263e9 | 995 | saved_error = error; |
b16817b6 DC |
996 | continue; |
997 | } | |
998 | ||
999 | bp->b_ops = &xfs_sb_buf_ops; | |
1000 | xfs_buf_oneshot(bp); | |
1001 | xfs_buf_zero(bp, 0, BBTOB(bp->b_length)); | |
3e6e8afd | 1002 | xfs_sb_to_disk(bp->b_addr, &mp->m_sb); |
b16817b6 DC |
1003 | xfs_buf_delwri_queue(bp, &buffer_list); |
1004 | xfs_buf_relse(bp); | |
1005 | ||
1006 | /* don't hold too many buffers at once */ | |
1007 | if (agno % 16) | |
1008 | continue; | |
1009 | ||
1010 | error = xfs_buf_delwri_submit(&buffer_list); | |
1011 | if (error) { | |
1012 | xfs_warn(mp, | |
1013 | "write error %d updating a secondary superblock near ag %d", | |
7f8d3b3c | 1014 | error, pag->pag_agno); |
b16817b6 DC |
1015 | if (!saved_error) |
1016 | saved_error = error; | |
1017 | continue; | |
1018 | } | |
1019 | } | |
1020 | error = xfs_buf_delwri_submit(&buffer_list); | |
1021 | if (error) { | |
1022 | xfs_warn(mp, | |
1023 | "write error %d updating a secondary superblock near ag %d", | |
1024 | error, agno); | |
1025 | } | |
1026 | ||
1027 | return saved_error ? saved_error : error; | |
1028 | } | |
1029 | ||
f7664b31 ES |
1030 | /* |
1031 | * Same behavior as xfs_sync_sb, except that it is always synchronous and it | |
1032 | * also writes the superblock buffer to disk sector 0 immediately. | |
1033 | */ | |
1034 | int | |
1035 | xfs_sync_sb_buf( | |
1036 | struct xfs_mount *mp) | |
1037 | { | |
1038 | struct xfs_trans *tp; | |
89c2e711 | 1039 | struct xfs_buf *bp; |
f7664b31 ES |
1040 | int error; |
1041 | ||
1042 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0, 0, &tp); | |
1043 | if (error) | |
1044 | return error; | |
1045 | ||
cead0b10 | 1046 | bp = xfs_trans_getsb(tp); |
f7664b31 | 1047 | xfs_log_sb(tp); |
89c2e711 | 1048 | xfs_trans_bhold(tp, bp); |
f7664b31 ES |
1049 | xfs_trans_set_sync(tp); |
1050 | error = xfs_trans_commit(tp); | |
1051 | if (error) | |
1052 | goto out; | |
1053 | /* | |
1054 | * write out the sb buffer to get the changes to disk | |
1055 | */ | |
89c2e711 | 1056 | error = xfs_bwrite(bp); |
f7664b31 | 1057 | out: |
89c2e711 | 1058 | xfs_buf_relse(bp); |
f7664b31 ES |
1059 | return error; |
1060 | } | |
1061 | ||
91083269 | 1062 | void |
c368ebcd | 1063 | xfs_fs_geometry( |
03288b19 | 1064 | struct xfs_mount *mp, |
ac503a4c DW |
1065 | struct xfs_fsop_geom *geo, |
1066 | int struct_version) | |
c368ebcd | 1067 | { |
03288b19 DC |
1068 | struct xfs_sb *sbp = &mp->m_sb; |
1069 | ||
ac503a4c DW |
1070 | memset(geo, 0, sizeof(struct xfs_fsop_geom)); |
1071 | ||
1072 | geo->blocksize = sbp->sb_blocksize; | |
1073 | geo->rtextsize = sbp->sb_rextsize; | |
1074 | geo->agblocks = sbp->sb_agblocks; | |
1075 | geo->agcount = sbp->sb_agcount; | |
1076 | geo->logblocks = sbp->sb_logblocks; | |
1077 | geo->sectsize = sbp->sb_sectsize; | |
1078 | geo->inodesize = sbp->sb_inodesize; | |
1079 | geo->imaxpct = sbp->sb_imax_pct; | |
1080 | geo->datablocks = sbp->sb_dblocks; | |
1081 | geo->rtblocks = sbp->sb_rblocks; | |
1082 | geo->rtextents = sbp->sb_rextents; | |
1083 | geo->logstart = sbp->sb_logstart; | |
1084 | BUILD_BUG_ON(sizeof(geo->uuid) != sizeof(sbp->sb_uuid)); | |
1085 | memcpy(geo->uuid, &sbp->sb_uuid, sizeof(sbp->sb_uuid)); | |
1086 | ||
1087 | if (struct_version < 2) | |
91083269 | 1088 | return; |
ac503a4c DW |
1089 | |
1090 | geo->sunit = sbp->sb_unit; | |
1091 | geo->swidth = sbp->sb_width; | |
1092 | ||
1093 | if (struct_version < 3) | |
91083269 | 1094 | return; |
ac503a4c DW |
1095 | |
1096 | geo->version = XFS_FSOP_GEOM_VERSION; | |
1097 | geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK | | |
daa79bae CH |
1098 | XFS_FSOP_GEOM_FLAGS_DIRV2 | |
1099 | XFS_FSOP_GEOM_FLAGS_EXTFLG; | |
03288b19 | 1100 | if (xfs_has_attr(mp)) |
ac503a4c | 1101 | geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR; |
03288b19 | 1102 | if (xfs_has_quota(mp)) |
ac503a4c | 1103 | geo->flags |= XFS_FSOP_GEOM_FLAGS_QUOTA; |
03288b19 | 1104 | if (xfs_has_align(mp)) |
ac503a4c | 1105 | geo->flags |= XFS_FSOP_GEOM_FLAGS_IALIGN; |
03288b19 | 1106 | if (xfs_has_dalign(mp)) |
ac503a4c | 1107 | geo->flags |= XFS_FSOP_GEOM_FLAGS_DALIGN; |
03288b19 | 1108 | if (xfs_has_asciici(mp)) |
ac503a4c | 1109 | geo->flags |= XFS_FSOP_GEOM_FLAGS_DIRV2CI; |
03288b19 | 1110 | if (xfs_has_lazysbcount(mp)) |
ac503a4c | 1111 | geo->flags |= XFS_FSOP_GEOM_FLAGS_LAZYSB; |
03288b19 | 1112 | if (xfs_has_attr2(mp)) |
ac503a4c | 1113 | geo->flags |= XFS_FSOP_GEOM_FLAGS_ATTR2; |
03288b19 | 1114 | if (xfs_has_projid32(mp)) |
ac503a4c | 1115 | geo->flags |= XFS_FSOP_GEOM_FLAGS_PROJID32; |
03288b19 | 1116 | if (xfs_has_crc(mp)) |
ac503a4c | 1117 | geo->flags |= XFS_FSOP_GEOM_FLAGS_V5SB; |
03288b19 | 1118 | if (xfs_has_ftype(mp)) |
ac503a4c | 1119 | geo->flags |= XFS_FSOP_GEOM_FLAGS_FTYPE; |
03288b19 | 1120 | if (xfs_has_finobt(mp)) |
ac503a4c | 1121 | geo->flags |= XFS_FSOP_GEOM_FLAGS_FINOBT; |
03288b19 | 1122 | if (xfs_has_sparseinodes(mp)) |
ac503a4c | 1123 | geo->flags |= XFS_FSOP_GEOM_FLAGS_SPINODES; |
03288b19 | 1124 | if (xfs_has_rmapbt(mp)) |
ac503a4c | 1125 | geo->flags |= XFS_FSOP_GEOM_FLAGS_RMAPBT; |
03288b19 | 1126 | if (xfs_has_reflink(mp)) |
ac503a4c | 1127 | geo->flags |= XFS_FSOP_GEOM_FLAGS_REFLINK; |
03288b19 | 1128 | if (xfs_has_bigtime(mp)) |
f93e5436 | 1129 | geo->flags |= XFS_FSOP_GEOM_FLAGS_BIGTIME; |
03288b19 | 1130 | if (xfs_has_inobtcounts(mp)) |
bc41fa53 | 1131 | geo->flags |= XFS_FSOP_GEOM_FLAGS_INOBTCNT; |
03288b19 DC |
1132 | if (xfs_has_sector(mp)) { |
1133 | geo->flags |= XFS_FSOP_GEOM_FLAGS_SECTOR; | |
ac503a4c | 1134 | geo->logsectsize = sbp->sb_logsectsize; |
03288b19 | 1135 | } else { |
ac503a4c | 1136 | geo->logsectsize = BBSIZE; |
03288b19 | 1137 | } |
ac503a4c DW |
1138 | geo->rtsectsize = sbp->sb_blocksize; |
1139 | geo->dirblocksize = xfs_dir2_dirblock_bytes(sbp); | |
1140 | ||
a8789a5a | 1141 | if (struct_version < 4) |
91083269 | 1142 | return; |
ac503a4c | 1143 | |
03288b19 | 1144 | if (xfs_has_logv2(mp)) |
ac503a4c DW |
1145 | geo->flags |= XFS_FSOP_GEOM_FLAGS_LOGV2; |
1146 | ||
1147 | geo->logsunit = sbp->sb_logsunit; | |
c368ebcd | 1148 | |
1b6d968d | 1149 | if (struct_version < 5) |
91083269 | 1150 | return; |
1b6d968d DC |
1151 | |
1152 | geo->version = XFS_FSOP_GEOM_VERSION_V5; | |
c368ebcd | 1153 | } |
689e11c8 DW |
1154 | |
1155 | /* Read a secondary superblock. */ | |
1156 | int | |
1157 | xfs_sb_read_secondary( | |
1158 | struct xfs_mount *mp, | |
1159 | struct xfs_trans *tp, | |
1160 | xfs_agnumber_t agno, | |
1161 | struct xfs_buf **bpp) | |
1162 | { | |
1163 | struct xfs_buf *bp; | |
1164 | int error; | |
1165 | ||
1166 | ASSERT(agno != 0 && agno != NULLAGNUMBER); | |
1167 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, | |
1168 | XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)), | |
1169 | XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_sb_buf_ops); | |
1170 | if (error) | |
1171 | return error; | |
1172 | xfs_buf_set_ref(bp, XFS_SSB_REF); | |
1173 | *bpp = bp; | |
1174 | return 0; | |
1175 | } | |
d25522f1 DW |
1176 | |
1177 | /* Get an uninitialised secondary superblock buffer. */ | |
1178 | int | |
1179 | xfs_sb_get_secondary( | |
1180 | struct xfs_mount *mp, | |
1181 | struct xfs_trans *tp, | |
1182 | xfs_agnumber_t agno, | |
1183 | struct xfs_buf **bpp) | |
1184 | { | |
1185 | struct xfs_buf *bp; | |
ce92464c | 1186 | int error; |
d25522f1 DW |
1187 | |
1188 | ASSERT(agno != 0 && agno != NULLAGNUMBER); | |
ce92464c | 1189 | error = xfs_trans_get_buf(tp, mp->m_ddev_targp, |
d25522f1 | 1190 | XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)), |
ce92464c DW |
1191 | XFS_FSS_TO_BB(mp, 1), 0, &bp); |
1192 | if (error) | |
1193 | return error; | |
d25522f1 DW |
1194 | bp->b_ops = &xfs_sb_buf_ops; |
1195 | xfs_buf_oneshot(bp); | |
1196 | *bpp = bp; | |
1197 | return 0; | |
1198 | } | |
7bc1fea9 GX |
1199 | |
1200 | /* | |
1201 | * sunit, swidth, sectorsize(optional with 0) should be all in bytes, | |
1202 | * so users won't be confused by values in error messages. | |
1203 | */ | |
1204 | bool | |
1205 | xfs_validate_stripe_geometry( | |
1206 | struct xfs_mount *mp, | |
1207 | __s64 sunit, | |
1208 | __s64 swidth, | |
1209 | int sectorsize, | |
1210 | bool silent) | |
1211 | { | |
1212 | if (swidth > INT_MAX) { | |
1213 | if (!silent) | |
1214 | xfs_notice(mp, | |
1215 | "stripe width (%lld) is too large", swidth); | |
1216 | return false; | |
1217 | } | |
1218 | ||
1219 | if (sunit > swidth) { | |
1220 | if (!silent) | |
1221 | xfs_notice(mp, | |
1222 | "stripe unit (%lld) is larger than the stripe width (%lld)", sunit, swidth); | |
1223 | return false; | |
1224 | } | |
1225 | ||
1226 | if (sectorsize && (int)sunit % sectorsize) { | |
1227 | if (!silent) | |
1228 | xfs_notice(mp, | |
1229 | "stripe unit (%lld) must be a multiple of the sector size (%d)", | |
1230 | sunit, sectorsize); | |
1231 | return false; | |
1232 | } | |
1233 | ||
1234 | if (sunit && !swidth) { | |
1235 | if (!silent) | |
1236 | xfs_notice(mp, | |
1237 | "invalid stripe unit (%lld) and stripe width of 0", sunit); | |
1238 | return false; | |
1239 | } | |
1240 | ||
1241 | if (!sunit && swidth) { | |
1242 | if (!silent) | |
1243 | xfs_notice(mp, | |
1244 | "invalid stripe width (%lld) and stripe unit of 0", swidth); | |
1245 | return false; | |
1246 | } | |
1247 | ||
1248 | if (sunit && (int)swidth % (int)sunit) { | |
1249 | if (!silent) | |
1250 | xfs_notice(mp, | |
1251 | "stripe width (%lld) must be a multiple of the stripe unit (%lld)", | |
1252 | swidth, sunit); | |
1253 | return false; | |
1254 | } | |
1255 | return true; | |
1256 | } |