Commit | Line | Data |
---|---|---|
ff55068c DC |
1 | /* |
2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. | |
3 | * All Rights Reserved. | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or | |
6 | * modify it under the terms of the GNU General Public License as | |
7 | * published by the Free Software Foundation. | |
8 | * | |
9 | * This program is distributed in the hope that it would be useful, | |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
13 | * | |
14 | * You should have received a copy of the GNU General Public License | |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
17 | */ | |
18 | #include "xfs.h" | |
19 | #include "xfs_fs.h" | |
632b89e8 | 20 | #include "xfs_shared.h" |
ff55068c | 21 | #include "xfs_format.h" |
239880ef DC |
22 | #include "xfs_log_format.h" |
23 | #include "xfs_trans_resv.h" | |
ff55068c | 24 | #include "xfs_bit.h" |
ff55068c | 25 | #include "xfs_sb.h" |
ff55068c | 26 | #include "xfs_mount.h" |
3ab78df2 | 27 | #include "xfs_defer.h" |
ff55068c | 28 | #include "xfs_inode.h" |
ff55068c DC |
29 | #include "xfs_ialloc.h" |
30 | #include "xfs_alloc.h" | |
ff55068c | 31 | #include "xfs_error.h" |
ff55068c DC |
32 | #include "xfs_trace.h" |
33 | #include "xfs_cksum.h" | |
239880ef | 34 | #include "xfs_trans.h" |
ff55068c | 35 | #include "xfs_buf_item.h" |
a4fbe6ab DC |
36 | #include "xfs_bmap_btree.h" |
37 | #include "xfs_alloc_btree.h" | |
38 | #include "xfs_ialloc_btree.h" | |
a45086e2 | 39 | #include "xfs_log.h" |
035e00ac | 40 | #include "xfs_rmap_btree.h" |
1946b91c DW |
41 | #include "xfs_bmap.h" |
42 | #include "xfs_refcount_btree.h" | |
ff55068c DC |
43 | |
44 | /* | |
45 | * Physical superblock buffer manipulations. Shared with libxfs in userspace. | |
46 | */ | |
47 | ||
ff55068c DC |
48 | /* |
49 | * Reference counting access wrappers to the perag structures. | |
50 | * Because we never free per-ag structures, the only thing we | |
51 | * have to protect against changes is the tree structure itself. | |
52 | */ | |
53 | struct xfs_perag * | |
54 | xfs_perag_get( | |
55 | struct xfs_mount *mp, | |
56 | xfs_agnumber_t agno) | |
57 | { | |
58 | struct xfs_perag *pag; | |
59 | int ref = 0; | |
60 | ||
61 | rcu_read_lock(); | |
62 | pag = radix_tree_lookup(&mp->m_perag_tree, agno); | |
63 | if (pag) { | |
64 | ASSERT(atomic_read(&pag->pag_ref) >= 0); | |
65 | ref = atomic_inc_return(&pag->pag_ref); | |
66 | } | |
67 | rcu_read_unlock(); | |
68 | trace_xfs_perag_get(mp, agno, ref, _RET_IP_); | |
69 | return pag; | |
70 | } | |
71 | ||
72 | /* | |
73 | * search from @first to find the next perag with the given tag set. | |
74 | */ | |
75 | struct xfs_perag * | |
76 | xfs_perag_get_tag( | |
77 | struct xfs_mount *mp, | |
78 | xfs_agnumber_t first, | |
79 | int tag) | |
80 | { | |
81 | struct xfs_perag *pag; | |
82 | int found; | |
83 | int ref; | |
84 | ||
85 | rcu_read_lock(); | |
86 | found = radix_tree_gang_lookup_tag(&mp->m_perag_tree, | |
87 | (void **)&pag, first, 1, tag); | |
88 | if (found <= 0) { | |
89 | rcu_read_unlock(); | |
90 | return NULL; | |
91 | } | |
92 | ref = atomic_inc_return(&pag->pag_ref); | |
93 | rcu_read_unlock(); | |
94 | trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_); | |
95 | return pag; | |
96 | } | |
97 | ||
98 | void | |
99 | xfs_perag_put( | |
100 | struct xfs_perag *pag) | |
101 | { | |
102 | int ref; | |
103 | ||
104 | ASSERT(atomic_read(&pag->pag_ref) > 0); | |
105 | ref = atomic_dec_return(&pag->pag_ref); | |
106 | trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_); | |
107 | } | |
108 | ||
109 | /* | |
110 | * Check the validity of the SB found. | |
111 | */ | |
112 | STATIC int | |
113 | xfs_mount_validate_sb( | |
114 | xfs_mount_t *mp, | |
115 | xfs_sb_t *sbp, | |
116 | bool check_inprogress, | |
117 | bool check_version) | |
118 | { | |
ff55068c DC |
119 | if (sbp->sb_magicnum != XFS_SB_MAGIC) { |
120 | xfs_warn(mp, "bad magic number"); | |
2451337d | 121 | return -EWRONGFS; |
ff55068c DC |
122 | } |
123 | ||
124 | ||
125 | if (!xfs_sb_good_version(sbp)) { | |
126 | xfs_warn(mp, "bad version"); | |
2451337d | 127 | return -EWRONGFS; |
ff55068c DC |
128 | } |
129 | ||
130 | /* | |
131 | * Version 5 superblock feature mask validation. Reject combinations the | |
132 | * kernel cannot support up front before checking anything else. For | |
133 | * write validation, we don't need to check feature masks. | |
134 | */ | |
135 | if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) { | |
ff55068c DC |
136 | if (xfs_sb_has_compat_feature(sbp, |
137 | XFS_SB_FEAT_COMPAT_UNKNOWN)) { | |
138 | xfs_warn(mp, | |
f41febd2 | 139 | "Superblock has unknown compatible features (0x%x) enabled.", |
ff55068c DC |
140 | (sbp->sb_features_compat & |
141 | XFS_SB_FEAT_COMPAT_UNKNOWN)); | |
f41febd2 JP |
142 | xfs_warn(mp, |
143 | "Using a more recent kernel is recommended."); | |
ff55068c DC |
144 | } |
145 | ||
146 | if (xfs_sb_has_ro_compat_feature(sbp, | |
147 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { | |
148 | xfs_alert(mp, | |
149 | "Superblock has unknown read-only compatible features (0x%x) enabled.", | |
150 | (sbp->sb_features_ro_compat & | |
151 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); | |
152 | if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { | |
153 | xfs_warn(mp, | |
f41febd2 JP |
154 | "Attempted to mount read-only compatible filesystem read-write."); |
155 | xfs_warn(mp, | |
ff55068c | 156 | "Filesystem can only be safely mounted read only."); |
f41febd2 | 157 | |
2451337d | 158 | return -EINVAL; |
ff55068c DC |
159 | } |
160 | } | |
161 | if (xfs_sb_has_incompat_feature(sbp, | |
162 | XFS_SB_FEAT_INCOMPAT_UNKNOWN)) { | |
163 | xfs_warn(mp, | |
f41febd2 | 164 | "Superblock has unknown incompatible features (0x%x) enabled.", |
ff55068c DC |
165 | (sbp->sb_features_incompat & |
166 | XFS_SB_FEAT_INCOMPAT_UNKNOWN)); | |
f41febd2 JP |
167 | xfs_warn(mp, |
168 | "Filesystem can not be safely mounted by this kernel."); | |
2451337d | 169 | return -EINVAL; |
ff55068c | 170 | } |
a45086e2 BF |
171 | } else if (xfs_sb_version_hascrc(sbp)) { |
172 | /* | |
173 | * We can't read verify the sb LSN because the read verifier is | |
174 | * called before the log is allocated and processed. We know the | |
175 | * log is set up before write verifier (!check_version) calls, | |
176 | * so just check it here. | |
177 | */ | |
178 | if (!xfs_log_check_lsn(mp, sbp->sb_lsn)) | |
179 | return -EFSCORRUPTED; | |
ff55068c DC |
180 | } |
181 | ||
182 | if (xfs_sb_version_has_pquotino(sbp)) { | |
183 | if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) { | |
184 | xfs_notice(mp, | |
08e96e1a | 185 | "Version 5 of Super block has XFS_OQUOTA bits."); |
2451337d | 186 | return -EFSCORRUPTED; |
ff55068c DC |
187 | } |
188 | } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD | | |
189 | XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) { | |
190 | xfs_notice(mp, | |
08e96e1a | 191 | "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits."); |
2451337d | 192 | return -EFSCORRUPTED; |
ff55068c DC |
193 | } |
194 | ||
e5376fc1 BF |
195 | /* |
196 | * Full inode chunks must be aligned to inode chunk size when | |
197 | * sparse inodes are enabled to support the sparse chunk | |
198 | * allocation algorithm and prevent overlapping inode records. | |
199 | */ | |
200 | if (xfs_sb_version_hassparseinodes(sbp)) { | |
201 | uint32_t align; | |
202 | ||
e5376fc1 BF |
203 | align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize |
204 | >> sbp->sb_blocklog; | |
205 | if (sbp->sb_inoalignmt != align) { | |
206 | xfs_warn(mp, | |
207 | "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.", | |
208 | sbp->sb_inoalignmt, align); | |
209 | return -EINVAL; | |
210 | } | |
211 | } | |
212 | ||
ff55068c DC |
213 | if (unlikely( |
214 | sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) { | |
215 | xfs_warn(mp, | |
216 | "filesystem is marked as having an external log; " | |
217 | "specify logdev on the mount command line."); | |
2451337d | 218 | return -EINVAL; |
ff55068c DC |
219 | } |
220 | ||
221 | if (unlikely( | |
222 | sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) { | |
223 | xfs_warn(mp, | |
224 | "filesystem is marked as having an internal log; " | |
225 | "do not specify logdev on the mount command line."); | |
2451337d | 226 | return -EINVAL; |
ff55068c DC |
227 | } |
228 | ||
229 | /* | |
230 | * More sanity checking. Most of these were stolen directly from | |
231 | * xfs_repair. | |
232 | */ | |
233 | if (unlikely( | |
234 | sbp->sb_agcount <= 0 || | |
235 | sbp->sb_sectsize < XFS_MIN_SECTORSIZE || | |
236 | sbp->sb_sectsize > XFS_MAX_SECTORSIZE || | |
237 | sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG || | |
238 | sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG || | |
239 | sbp->sb_sectsize != (1 << sbp->sb_sectlog) || | |
240 | sbp->sb_blocksize < XFS_MIN_BLOCKSIZE || | |
241 | sbp->sb_blocksize > XFS_MAX_BLOCKSIZE || | |
242 | sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG || | |
243 | sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG || | |
244 | sbp->sb_blocksize != (1 << sbp->sb_blocklog) || | |
e1b05723 | 245 | sbp->sb_dirblklog > XFS_MAX_BLOCKSIZE_LOG || |
ff55068c DC |
246 | sbp->sb_inodesize < XFS_DINODE_MIN_SIZE || |
247 | sbp->sb_inodesize > XFS_DINODE_MAX_SIZE || | |
248 | sbp->sb_inodelog < XFS_DINODE_MIN_LOG || | |
249 | sbp->sb_inodelog > XFS_DINODE_MAX_LOG || | |
250 | sbp->sb_inodesize != (1 << sbp->sb_inodelog) || | |
e1b05723 | 251 | sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE || |
392c6de9 | 252 | sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) || |
ff55068c DC |
253 | (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) || |
254 | (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) || | |
255 | (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) || | |
256 | (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) || | |
257 | sbp->sb_dblocks == 0 || | |
258 | sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) || | |
ab3e57b5 DC |
259 | sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) || |
260 | sbp->sb_shared_vn != 0)) { | |
5ef11eb0 | 261 | xfs_notice(mp, "SB sanity check failed"); |
2451337d | 262 | return -EFSCORRUPTED; |
ff55068c DC |
263 | } |
264 | ||
265 | /* | |
266 | * Until this is fixed only page-sized or smaller data blocks work. | |
267 | */ | |
268 | if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) { | |
269 | xfs_warn(mp, | |
270 | "File system with blocksize %d bytes. " | |
271 | "Only pagesize (%ld) or less will currently work.", | |
272 | sbp->sb_blocksize, PAGE_SIZE); | |
2451337d | 273 | return -ENOSYS; |
ff55068c DC |
274 | } |
275 | ||
276 | /* | |
277 | * Currently only very few inode sizes are supported. | |
278 | */ | |
279 | switch (sbp->sb_inodesize) { | |
280 | case 256: | |
281 | case 512: | |
282 | case 1024: | |
283 | case 2048: | |
284 | break; | |
285 | default: | |
286 | xfs_warn(mp, "inode size of %d bytes not supported", | |
287 | sbp->sb_inodesize); | |
2451337d | 288 | return -ENOSYS; |
ff55068c DC |
289 | } |
290 | ||
291 | if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) || | |
292 | xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) { | |
293 | xfs_warn(mp, | |
294 | "file system too large to be mounted on this system."); | |
2451337d | 295 | return -EFBIG; |
ff55068c DC |
296 | } |
297 | ||
298 | if (check_inprogress && sbp->sb_inprogress) { | |
299 | xfs_warn(mp, "Offline file system operation in progress!"); | |
2451337d | 300 | return -EFSCORRUPTED; |
ff55068c | 301 | } |
ff55068c DC |
302 | return 0; |
303 | } | |
304 | ||
305 | void | |
306 | xfs_sb_quota_from_disk(struct xfs_sb *sbp) | |
307 | { | |
308 | /* | |
309 | * older mkfs doesn't initialize quota inodes to NULLFSINO. This | |
310 | * leads to in-core values having two different values for a quota | |
311 | * inode to be invalid: 0 and NULLFSINO. Change it to a single value | |
312 | * NULLFSINO. | |
313 | * | |
314 | * Note that this change affect only the in-core values. These | |
315 | * values are not written back to disk unless any quota information | |
316 | * is written to the disk. Even in that case, sb_pquotino field is | |
317 | * not written to disk unless the superblock supports pquotino. | |
318 | */ | |
319 | if (sbp->sb_uquotino == 0) | |
320 | sbp->sb_uquotino = NULLFSINO; | |
321 | if (sbp->sb_gquotino == 0) | |
322 | sbp->sb_gquotino = NULLFSINO; | |
323 | if (sbp->sb_pquotino == 0) | |
324 | sbp->sb_pquotino = NULLFSINO; | |
325 | ||
326 | /* | |
327 | * We need to do these manipilations only if we are working | |
328 | * with an older version of on-disk superblock. | |
329 | */ | |
330 | if (xfs_sb_version_has_pquotino(sbp)) | |
331 | return; | |
332 | ||
333 | if (sbp->sb_qflags & XFS_OQUOTA_ENFD) | |
334 | sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? | |
335 | XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD; | |
336 | if (sbp->sb_qflags & XFS_OQUOTA_CHKD) | |
337 | sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? | |
338 | XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD; | |
339 | sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD); | |
340 | ||
341 | if (sbp->sb_qflags & XFS_PQUOTA_ACCT) { | |
342 | /* | |
343 | * In older version of superblock, on-disk superblock only | |
344 | * has sb_gquotino, and in-core superblock has both sb_gquotino | |
345 | * and sb_pquotino. But, only one of them is supported at any | |
346 | * point of time. So, if PQUOTA is set in disk superblock, | |
347 | * copy over sb_gquotino to sb_pquotino. | |
348 | */ | |
349 | sbp->sb_pquotino = sbp->sb_gquotino; | |
350 | sbp->sb_gquotino = NULLFSINO; | |
351 | } | |
352 | } | |
353 | ||
5ef828c4 ES |
354 | static void |
355 | __xfs_sb_from_disk( | |
ff55068c | 356 | struct xfs_sb *to, |
5ef828c4 ES |
357 | xfs_dsb_t *from, |
358 | bool convert_xquota) | |
ff55068c DC |
359 | { |
360 | to->sb_magicnum = be32_to_cpu(from->sb_magicnum); | |
361 | to->sb_blocksize = be32_to_cpu(from->sb_blocksize); | |
362 | to->sb_dblocks = be64_to_cpu(from->sb_dblocks); | |
363 | to->sb_rblocks = be64_to_cpu(from->sb_rblocks); | |
364 | to->sb_rextents = be64_to_cpu(from->sb_rextents); | |
365 | memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid)); | |
366 | to->sb_logstart = be64_to_cpu(from->sb_logstart); | |
367 | to->sb_rootino = be64_to_cpu(from->sb_rootino); | |
368 | to->sb_rbmino = be64_to_cpu(from->sb_rbmino); | |
369 | to->sb_rsumino = be64_to_cpu(from->sb_rsumino); | |
370 | to->sb_rextsize = be32_to_cpu(from->sb_rextsize); | |
371 | to->sb_agblocks = be32_to_cpu(from->sb_agblocks); | |
372 | to->sb_agcount = be32_to_cpu(from->sb_agcount); | |
373 | to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks); | |
374 | to->sb_logblocks = be32_to_cpu(from->sb_logblocks); | |
375 | to->sb_versionnum = be16_to_cpu(from->sb_versionnum); | |
376 | to->sb_sectsize = be16_to_cpu(from->sb_sectsize); | |
377 | to->sb_inodesize = be16_to_cpu(from->sb_inodesize); | |
378 | to->sb_inopblock = be16_to_cpu(from->sb_inopblock); | |
379 | memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname)); | |
380 | to->sb_blocklog = from->sb_blocklog; | |
381 | to->sb_sectlog = from->sb_sectlog; | |
382 | to->sb_inodelog = from->sb_inodelog; | |
383 | to->sb_inopblog = from->sb_inopblog; | |
384 | to->sb_agblklog = from->sb_agblklog; | |
385 | to->sb_rextslog = from->sb_rextslog; | |
386 | to->sb_inprogress = from->sb_inprogress; | |
387 | to->sb_imax_pct = from->sb_imax_pct; | |
388 | to->sb_icount = be64_to_cpu(from->sb_icount); | |
389 | to->sb_ifree = be64_to_cpu(from->sb_ifree); | |
390 | to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks); | |
391 | to->sb_frextents = be64_to_cpu(from->sb_frextents); | |
392 | to->sb_uquotino = be64_to_cpu(from->sb_uquotino); | |
393 | to->sb_gquotino = be64_to_cpu(from->sb_gquotino); | |
394 | to->sb_qflags = be16_to_cpu(from->sb_qflags); | |
395 | to->sb_flags = from->sb_flags; | |
396 | to->sb_shared_vn = from->sb_shared_vn; | |
397 | to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt); | |
398 | to->sb_unit = be32_to_cpu(from->sb_unit); | |
399 | to->sb_width = be32_to_cpu(from->sb_width); | |
400 | to->sb_dirblklog = from->sb_dirblklog; | |
401 | to->sb_logsectlog = from->sb_logsectlog; | |
402 | to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize); | |
403 | to->sb_logsunit = be32_to_cpu(from->sb_logsunit); | |
404 | to->sb_features2 = be32_to_cpu(from->sb_features2); | |
405 | to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2); | |
406 | to->sb_features_compat = be32_to_cpu(from->sb_features_compat); | |
407 | to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat); | |
408 | to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat); | |
409 | to->sb_features_log_incompat = | |
410 | be32_to_cpu(from->sb_features_log_incompat); | |
04dd1a0d ES |
411 | /* crc is only used on disk, not in memory; just init to 0 here. */ |
412 | to->sb_crc = 0; | |
fb4f2b4e | 413 | to->sb_spino_align = be32_to_cpu(from->sb_spino_align); |
ff55068c DC |
414 | to->sb_pquotino = be64_to_cpu(from->sb_pquotino); |
415 | to->sb_lsn = be64_to_cpu(from->sb_lsn); | |
ce748eaa ES |
416 | /* |
417 | * sb_meta_uuid is only on disk if it differs from sb_uuid and the | |
418 | * feature flag is set; if not set we keep it only in memory. | |
419 | */ | |
420 | if (xfs_sb_version_hasmetauuid(to)) | |
421 | uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid); | |
422 | else | |
423 | uuid_copy(&to->sb_meta_uuid, &from->sb_uuid); | |
5ef828c4 ES |
424 | /* Convert on-disk flags to in-memory flags? */ |
425 | if (convert_xquota) | |
426 | xfs_sb_quota_from_disk(to); | |
427 | } | |
428 | ||
429 | void | |
430 | xfs_sb_from_disk( | |
431 | struct xfs_sb *to, | |
432 | xfs_dsb_t *from) | |
433 | { | |
434 | __xfs_sb_from_disk(to, from, true); | |
ff55068c DC |
435 | } |
436 | ||
4d11a402 | 437 | static void |
ff55068c | 438 | xfs_sb_quota_to_disk( |
4d11a402 DC |
439 | struct xfs_dsb *to, |
440 | struct xfs_sb *from) | |
ff55068c DC |
441 | { |
442 | __uint16_t qflags = from->sb_qflags; | |
443 | ||
4d11a402 DC |
444 | to->sb_uquotino = cpu_to_be64(from->sb_uquotino); |
445 | if (xfs_sb_version_has_pquotino(from)) { | |
446 | to->sb_qflags = cpu_to_be16(from->sb_qflags); | |
447 | to->sb_gquotino = cpu_to_be64(from->sb_gquotino); | |
448 | to->sb_pquotino = cpu_to_be64(from->sb_pquotino); | |
449 | return; | |
450 | } | |
451 | ||
ff55068c | 452 | /* |
4d11a402 DC |
453 | * The in-core version of sb_qflags do not have XFS_OQUOTA_* |
454 | * flags, whereas the on-disk version does. So, convert incore | |
455 | * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags. | |
ff55068c | 456 | */ |
4d11a402 DC |
457 | qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD | |
458 | XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD); | |
ff55068c | 459 | |
4d11a402 DC |
460 | if (from->sb_qflags & |
461 | (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD)) | |
462 | qflags |= XFS_OQUOTA_ENFD; | |
463 | if (from->sb_qflags & | |
464 | (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) | |
465 | qflags |= XFS_OQUOTA_CHKD; | |
466 | to->sb_qflags = cpu_to_be16(qflags); | |
ff55068c DC |
467 | |
468 | /* | |
4d11a402 DC |
469 | * GQUOTINO and PQUOTINO cannot be used together in versions |
470 | * of superblock that do not have pquotino. from->sb_flags | |
471 | * tells us which quota is active and should be copied to | |
472 | * disk. If neither are active, we should NULL the inode. | |
03e01349 | 473 | * |
4d11a402 DC |
474 | * In all cases, the separate pquotino must remain 0 because it |
475 | * it beyond the "end" of the valid non-pquotino superblock. | |
ff55068c | 476 | */ |
4d11a402 | 477 | if (from->sb_qflags & XFS_GQUOTA_ACCT) |
ff55068c | 478 | to->sb_gquotino = cpu_to_be64(from->sb_gquotino); |
4d11a402 | 479 | else if (from->sb_qflags & XFS_PQUOTA_ACCT) |
ff55068c | 480 | to->sb_gquotino = cpu_to_be64(from->sb_pquotino); |
03e01349 DC |
481 | else { |
482 | /* | |
483 | * We can't rely on just the fields being logged to tell us | |
484 | * that it is safe to write NULLFSINO - we should only do that | |
485 | * if quotas are not actually enabled. Hence only write | |
486 | * NULLFSINO if both in-core quota inodes are NULL. | |
487 | */ | |
488 | if (from->sb_gquotino == NULLFSINO && | |
489 | from->sb_pquotino == NULLFSINO) | |
490 | to->sb_gquotino = cpu_to_be64(NULLFSINO); | |
491 | } | |
ff55068c | 492 | |
4d11a402 | 493 | to->sb_pquotino = 0; |
ff55068c DC |
494 | } |
495 | ||
ff55068c DC |
496 | void |
497 | xfs_sb_to_disk( | |
4d11a402 DC |
498 | struct xfs_dsb *to, |
499 | struct xfs_sb *from) | |
ff55068c | 500 | { |
4d11a402 DC |
501 | xfs_sb_quota_to_disk(to, from); |
502 | ||
503 | to->sb_magicnum = cpu_to_be32(from->sb_magicnum); | |
504 | to->sb_blocksize = cpu_to_be32(from->sb_blocksize); | |
505 | to->sb_dblocks = cpu_to_be64(from->sb_dblocks); | |
506 | to->sb_rblocks = cpu_to_be64(from->sb_rblocks); | |
507 | to->sb_rextents = cpu_to_be64(from->sb_rextents); | |
508 | memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid)); | |
509 | to->sb_logstart = cpu_to_be64(from->sb_logstart); | |
510 | to->sb_rootino = cpu_to_be64(from->sb_rootino); | |
511 | to->sb_rbmino = cpu_to_be64(from->sb_rbmino); | |
512 | to->sb_rsumino = cpu_to_be64(from->sb_rsumino); | |
513 | to->sb_rextsize = cpu_to_be32(from->sb_rextsize); | |
514 | to->sb_agblocks = cpu_to_be32(from->sb_agblocks); | |
515 | to->sb_agcount = cpu_to_be32(from->sb_agcount); | |
516 | to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks); | |
517 | to->sb_logblocks = cpu_to_be32(from->sb_logblocks); | |
518 | to->sb_versionnum = cpu_to_be16(from->sb_versionnum); | |
519 | to->sb_sectsize = cpu_to_be16(from->sb_sectsize); | |
520 | to->sb_inodesize = cpu_to_be16(from->sb_inodesize); | |
521 | to->sb_inopblock = cpu_to_be16(from->sb_inopblock); | |
522 | memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname)); | |
523 | to->sb_blocklog = from->sb_blocklog; | |
524 | to->sb_sectlog = from->sb_sectlog; | |
525 | to->sb_inodelog = from->sb_inodelog; | |
526 | to->sb_inopblog = from->sb_inopblog; | |
527 | to->sb_agblklog = from->sb_agblklog; | |
528 | to->sb_rextslog = from->sb_rextslog; | |
529 | to->sb_inprogress = from->sb_inprogress; | |
530 | to->sb_imax_pct = from->sb_imax_pct; | |
531 | to->sb_icount = cpu_to_be64(from->sb_icount); | |
532 | to->sb_ifree = cpu_to_be64(from->sb_ifree); | |
533 | to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks); | |
534 | to->sb_frextents = cpu_to_be64(from->sb_frextents); | |
ff55068c | 535 | |
4d11a402 DC |
536 | to->sb_flags = from->sb_flags; |
537 | to->sb_shared_vn = from->sb_shared_vn; | |
538 | to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt); | |
539 | to->sb_unit = cpu_to_be32(from->sb_unit); | |
540 | to->sb_width = cpu_to_be32(from->sb_width); | |
541 | to->sb_dirblklog = from->sb_dirblklog; | |
542 | to->sb_logsectlog = from->sb_logsectlog; | |
543 | to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize); | |
544 | to->sb_logsunit = cpu_to_be32(from->sb_logsunit); | |
074e427b DC |
545 | |
546 | /* | |
547 | * We need to ensure that bad_features2 always matches features2. | |
548 | * Hence we enforce that here rather than having to remember to do it | |
549 | * everywhere else that updates features2. | |
550 | */ | |
551 | from->sb_bad_features2 = from->sb_features2; | |
4d11a402 DC |
552 | to->sb_features2 = cpu_to_be32(from->sb_features2); |
553 | to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2); | |
554 | ||
555 | if (xfs_sb_version_hascrc(from)) { | |
556 | to->sb_features_compat = cpu_to_be32(from->sb_features_compat); | |
557 | to->sb_features_ro_compat = | |
558 | cpu_to_be32(from->sb_features_ro_compat); | |
559 | to->sb_features_incompat = | |
560 | cpu_to_be32(from->sb_features_incompat); | |
561 | to->sb_features_log_incompat = | |
562 | cpu_to_be32(from->sb_features_log_incompat); | |
fb4f2b4e | 563 | to->sb_spino_align = cpu_to_be32(from->sb_spino_align); |
4d11a402 | 564 | to->sb_lsn = cpu_to_be64(from->sb_lsn); |
ce748eaa ES |
565 | if (xfs_sb_version_hasmetauuid(from)) |
566 | uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid); | |
ff55068c DC |
567 | } |
568 | } | |
569 | ||
570 | static int | |
571 | xfs_sb_verify( | |
572 | struct xfs_buf *bp, | |
573 | bool check_version) | |
574 | { | |
575 | struct xfs_mount *mp = bp->b_target->bt_mount; | |
576 | struct xfs_sb sb; | |
577 | ||
5ef828c4 ES |
578 | /* |
579 | * Use call variant which doesn't convert quota flags from disk | |
580 | * format, because xfs_mount_validate_sb checks the on-disk flags. | |
581 | */ | |
582 | __xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false); | |
ff55068c DC |
583 | |
584 | /* | |
585 | * Only check the in progress field for the primary superblock as | |
586 | * mkfs.xfs doesn't clear it from secondary superblocks. | |
587 | */ | |
f3d7ebde DC |
588 | return xfs_mount_validate_sb(mp, &sb, |
589 | bp->b_maps[0].bm_bn == XFS_SB_DADDR, | |
ff55068c DC |
590 | check_version); |
591 | } | |
592 | ||
593 | /* | |
594 | * If the superblock has the CRC feature bit set or the CRC field is non-null, | |
595 | * check that the CRC is valid. We check the CRC field is non-null because a | |
596 | * single bit error could clear the feature bit and unused parts of the | |
597 | * superblock are supposed to be zero. Hence a non-null crc field indicates that | |
598 | * we've potentially lost a feature bit and we should check it anyway. | |
10e6e65d ES |
599 | * |
600 | * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the | |
601 | * last field in V4 secondary superblocks. So for secondary superblocks, | |
602 | * we are more forgiving, and ignore CRC failures if the primary doesn't | |
603 | * indicate that the fs version is V5. | |
ff55068c DC |
604 | */ |
605 | static void | |
606 | xfs_sb_read_verify( | |
607 | struct xfs_buf *bp) | |
608 | { | |
609 | struct xfs_mount *mp = bp->b_target->bt_mount; | |
610 | struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp); | |
611 | int error; | |
612 | ||
613 | /* | |
614 | * open code the version check to avoid needing to convert the entire | |
615 | * superblock from disk order just to check the version number | |
616 | */ | |
617 | if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) && | |
618 | (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) == | |
619 | XFS_SB_VERSION_5) || | |
620 | dsb->sb_crc != 0)) { | |
621 | ||
51582170 | 622 | if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) { |
10e6e65d | 623 | /* Only fail bad secondaries on a known V5 filesystem */ |
7a01e707 | 624 | if (bp->b_bn == XFS_SB_DADDR || |
10e6e65d | 625 | xfs_sb_version_hascrc(&mp->m_sb)) { |
2451337d | 626 | error = -EFSBADCRC; |
10e6e65d ES |
627 | goto out_error; |
628 | } | |
ff55068c DC |
629 | } |
630 | } | |
631 | error = xfs_sb_verify(bp, true); | |
632 | ||
633 | out_error: | |
634 | if (error) { | |
ff55068c | 635 | xfs_buf_ioerror(bp, error); |
2451337d | 636 | if (error == -EFSCORRUPTED || error == -EFSBADCRC) |
ce5028cf | 637 | xfs_verifier_error(bp); |
ff55068c DC |
638 | } |
639 | } | |
640 | ||
641 | /* | |
642 | * We may be probed for a filesystem match, so we may not want to emit | |
643 | * messages when the superblock buffer is not actually an XFS superblock. | |
2533787a | 644 | * If we find an XFS superblock, then run a normal, noisy mount because we are |
ff55068c DC |
645 | * really going to mount it and want to know about errors. |
646 | */ | |
647 | static void | |
648 | xfs_sb_quiet_read_verify( | |
649 | struct xfs_buf *bp) | |
650 | { | |
651 | struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp); | |
652 | ||
ff55068c DC |
653 | if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) { |
654 | /* XFS filesystem, verify noisily! */ | |
655 | xfs_sb_read_verify(bp); | |
656 | return; | |
657 | } | |
658 | /* quietly fail */ | |
2451337d | 659 | xfs_buf_ioerror(bp, -EWRONGFS); |
ff55068c DC |
660 | } |
661 | ||
662 | static void | |
663 | xfs_sb_write_verify( | |
664 | struct xfs_buf *bp) | |
665 | { | |
666 | struct xfs_mount *mp = bp->b_target->bt_mount; | |
667 | struct xfs_buf_log_item *bip = bp->b_fspriv; | |
668 | int error; | |
669 | ||
670 | error = xfs_sb_verify(bp, false); | |
671 | if (error) { | |
ff55068c | 672 | xfs_buf_ioerror(bp, error); |
ce5028cf | 673 | xfs_verifier_error(bp); |
ff55068c DC |
674 | return; |
675 | } | |
676 | ||
677 | if (!xfs_sb_version_hascrc(&mp->m_sb)) | |
678 | return; | |
679 | ||
680 | if (bip) | |
681 | XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn); | |
682 | ||
f1dbcd7e | 683 | xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF); |
ff55068c DC |
684 | } |
685 | ||
686 | const struct xfs_buf_ops xfs_sb_buf_ops = { | |
233135b7 | 687 | .name = "xfs_sb", |
ff55068c DC |
688 | .verify_read = xfs_sb_read_verify, |
689 | .verify_write = xfs_sb_write_verify, | |
690 | }; | |
691 | ||
692 | const struct xfs_buf_ops xfs_sb_quiet_buf_ops = { | |
233135b7 | 693 | .name = "xfs_sb_quiet", |
ff55068c DC |
694 | .verify_read = xfs_sb_quiet_read_verify, |
695 | .verify_write = xfs_sb_write_verify, | |
696 | }; | |
697 | ||
698 | /* | |
699 | * xfs_mount_common | |
700 | * | |
701 | * Mount initialization code establishing various mount | |
702 | * fields from the superblock associated with the given | |
703 | * mount structure | |
704 | */ | |
705 | void | |
706 | xfs_sb_mount_common( | |
707 | struct xfs_mount *mp, | |
708 | struct xfs_sb *sbp) | |
709 | { | |
710 | mp->m_agfrotor = mp->m_agirotor = 0; | |
711 | spin_lock_init(&mp->m_agirotor_lock); | |
712 | mp->m_maxagi = mp->m_sb.sb_agcount; | |
713 | mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG; | |
714 | mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT; | |
715 | mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT; | |
716 | mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1; | |
717 | mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog; | |
718 | mp->m_blockmask = sbp->sb_blocksize - 1; | |
719 | mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG; | |
720 | mp->m_blockwmask = mp->m_blockwsize - 1; | |
721 | ||
722 | mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1); | |
723 | mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0); | |
724 | mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2; | |
725 | mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2; | |
726 | ||
727 | mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1); | |
728 | mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0); | |
729 | mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2; | |
730 | mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2; | |
731 | ||
732 | mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1); | |
733 | mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0); | |
734 | mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2; | |
735 | mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2; | |
736 | ||
035e00ac DW |
737 | mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 1); |
738 | mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 0); | |
739 | mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2; | |
740 | mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2; | |
741 | ||
1946b91c DW |
742 | mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, |
743 | true); | |
744 | mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, | |
745 | false); | |
746 | mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2; | |
747 | mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2; | |
748 | ||
ff55068c DC |
749 | mp->m_bsize = XFS_FSB_TO_BB(mp, 1); |
750 | mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK, | |
751 | sbp->sb_inopblock); | |
752 | mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog; | |
066a1884 BF |
753 | |
754 | if (sbp->sb_spino_align) | |
755 | mp->m_ialloc_min_blks = sbp->sb_spino_align; | |
756 | else | |
757 | mp->m_ialloc_min_blks = mp->m_ialloc_blks; | |
52548852 DW |
758 | mp->m_alloc_set_aside = xfs_alloc_set_aside(mp); |
759 | mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp); | |
ff55068c DC |
760 | } |
761 | ||
762 | /* | |
763 | * xfs_initialize_perag_data | |
764 | * | |
765 | * Read in each per-ag structure so we can count up the number of | |
766 | * allocated inodes, free inodes and used filesystem blocks as this | |
767 | * information is no longer persistent in the superblock. Once we have | |
768 | * this information, write it into the in-core superblock structure. | |
769 | */ | |
770 | int | |
771 | xfs_initialize_perag_data( | |
772 | struct xfs_mount *mp, | |
773 | xfs_agnumber_t agcount) | |
774 | { | |
775 | xfs_agnumber_t index; | |
776 | xfs_perag_t *pag; | |
777 | xfs_sb_t *sbp = &mp->m_sb; | |
778 | uint64_t ifree = 0; | |
779 | uint64_t ialloc = 0; | |
780 | uint64_t bfree = 0; | |
781 | uint64_t bfreelst = 0; | |
782 | uint64_t btree = 0; | |
783 | int error; | |
784 | ||
785 | for (index = 0; index < agcount; index++) { | |
786 | /* | |
787 | * read the agf, then the agi. This gets us | |
788 | * all the information we need and populates the | |
789 | * per-ag structures for us. | |
790 | */ | |
791 | error = xfs_alloc_pagf_init(mp, NULL, index, 0); | |
792 | if (error) | |
793 | return error; | |
794 | ||
795 | error = xfs_ialloc_pagi_init(mp, NULL, index); | |
796 | if (error) | |
797 | return error; | |
798 | pag = xfs_perag_get(mp, index); | |
799 | ifree += pag->pagi_freecount; | |
800 | ialloc += pag->pagi_count; | |
801 | bfree += pag->pagf_freeblks; | |
802 | bfreelst += pag->pagf_flcount; | |
803 | btree += pag->pagf_btreeblks; | |
804 | xfs_perag_put(pag); | |
805 | } | |
5681ca40 DC |
806 | |
807 | /* Overwrite incore superblock counters with just-read data */ | |
ff55068c DC |
808 | spin_lock(&mp->m_sb_lock); |
809 | sbp->sb_ifree = ifree; | |
810 | sbp->sb_icount = ialloc; | |
811 | sbp->sb_fdblocks = bfree + bfreelst + btree; | |
812 | spin_unlock(&mp->m_sb_lock); | |
813 | ||
5681ca40 | 814 | xfs_reinit_percpu_counters(mp); |
ff55068c DC |
815 | |
816 | return 0; | |
817 | } | |
818 | ||
819 | /* | |
61e63ecb DC |
820 | * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock |
821 | * into the superblock buffer to be logged. It does not provide the higher | |
822 | * level of locking that is needed to protect the in-core superblock from | |
823 | * concurrent access. | |
ff55068c DC |
824 | */ |
825 | void | |
61e63ecb | 826 | xfs_log_sb( |
4d11a402 | 827 | struct xfs_trans *tp) |
ff55068c | 828 | { |
4d11a402 DC |
829 | struct xfs_mount *mp = tp->t_mountp; |
830 | struct xfs_buf *bp = xfs_trans_getsb(tp, mp, 0); | |
ff55068c | 831 | |
501ab323 | 832 | mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount); |
e88b64ea | 833 | mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree); |
0d485ada | 834 | mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks); |
501ab323 | 835 | |
4d11a402 | 836 | xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb); |
ff55068c | 837 | xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF); |
4d11a402 | 838 | xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb)); |
ff55068c | 839 | } |
61e63ecb DC |
840 | |
841 | /* | |
842 | * xfs_sync_sb | |
843 | * | |
844 | * Sync the superblock to disk. | |
845 | * | |
846 | * Note that the caller is responsible for checking the frozen state of the | |
847 | * filesystem. This procedure uses the non-blocking transaction allocator and | |
848 | * thus will allow modifications to a frozen fs. This is required because this | |
849 | * code can be called during the process of freezing where use of the high-level | |
850 | * allocator would deadlock. | |
851 | */ | |
852 | int | |
853 | xfs_sync_sb( | |
854 | struct xfs_mount *mp, | |
855 | bool wait) | |
856 | { | |
857 | struct xfs_trans *tp; | |
858 | int error; | |
859 | ||
253f4911 CH |
860 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0, |
861 | XFS_TRANS_NO_WRITECOUNT, &tp); | |
862 | if (error) | |
61e63ecb | 863 | return error; |
61e63ecb DC |
864 | |
865 | xfs_log_sb(tp); | |
866 | if (wait) | |
867 | xfs_trans_set_sync(tp); | |
70393313 | 868 | return xfs_trans_commit(tp); |
61e63ecb | 869 | } |