blob: 0e3d4f5ec33c6f945b30b5ad47c9f39c46139ac0 [file] [log] [blame]
Dave Chinnerff550682013-08-12 20:49:41 +10001/*
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"
Dave Chinner632b89e2013-10-29 22:11:58 +110020#include "xfs_shared.h"
Dave Chinnerff550682013-08-12 20:49:41 +100021#include "xfs_format.h"
Dave Chinner239880e2013-10-23 10:50:10 +110022#include "xfs_log_format.h"
23#include "xfs_trans_resv.h"
Dave Chinnerff550682013-08-12 20:49:41 +100024#include "xfs_bit.h"
Dave Chinnerff550682013-08-12 20:49:41 +100025#include "xfs_sb.h"
Dave Chinnerff550682013-08-12 20:49:41 +100026#include "xfs_mount.h"
Darrick J. Wong3ab78df2016-08-03 11:15:38 +100027#include "xfs_defer.h"
Dave Chinnerff550682013-08-12 20:49:41 +100028#include "xfs_inode.h"
Dave Chinnerff550682013-08-12 20:49:41 +100029#include "xfs_ialloc.h"
30#include "xfs_alloc.h"
Dave Chinnerff550682013-08-12 20:49:41 +100031#include "xfs_error.h"
Dave Chinnerff550682013-08-12 20:49:41 +100032#include "xfs_trace.h"
33#include "xfs_cksum.h"
Dave Chinner239880e2013-10-23 10:50:10 +110034#include "xfs_trans.h"
Dave Chinnerff550682013-08-12 20:49:41 +100035#include "xfs_buf_item.h"
Dave Chinnera4fbe6a2013-10-23 10:51:50 +110036#include "xfs_bmap_btree.h"
37#include "xfs_alloc_btree.h"
38#include "xfs_ialloc_btree.h"
Brian Fostera45086e2015-10-12 15:59:25 +110039#include "xfs_log.h"
Darrick J. Wong035e00a2016-08-03 11:36:07 +100040#include "xfs_rmap_btree.h"
Dave Chinnerff550682013-08-12 20:49:41 +100041
42/*
43 * Physical superblock buffer manipulations. Shared with libxfs in userspace.
44 */
45
Dave Chinnerff550682013-08-12 20:49:41 +100046/*
47 * Reference counting access wrappers to the perag structures.
48 * Because we never free per-ag structures, the only thing we
49 * have to protect against changes is the tree structure itself.
50 */
51struct xfs_perag *
52xfs_perag_get(
53 struct xfs_mount *mp,
54 xfs_agnumber_t agno)
55{
56 struct xfs_perag *pag;
57 int ref = 0;
58
59 rcu_read_lock();
60 pag = radix_tree_lookup(&mp->m_perag_tree, agno);
61 if (pag) {
62 ASSERT(atomic_read(&pag->pag_ref) >= 0);
63 ref = atomic_inc_return(&pag->pag_ref);
64 }
65 rcu_read_unlock();
66 trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
67 return pag;
68}
69
70/*
71 * search from @first to find the next perag with the given tag set.
72 */
73struct xfs_perag *
74xfs_perag_get_tag(
75 struct xfs_mount *mp,
76 xfs_agnumber_t first,
77 int tag)
78{
79 struct xfs_perag *pag;
80 int found;
81 int ref;
82
83 rcu_read_lock();
84 found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
85 (void **)&pag, first, 1, tag);
86 if (found <= 0) {
87 rcu_read_unlock();
88 return NULL;
89 }
90 ref = atomic_inc_return(&pag->pag_ref);
91 rcu_read_unlock();
92 trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
93 return pag;
94}
95
96void
97xfs_perag_put(
98 struct xfs_perag *pag)
99{
100 int ref;
101
102 ASSERT(atomic_read(&pag->pag_ref) > 0);
103 ref = atomic_dec_return(&pag->pag_ref);
104 trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
105}
106
107/*
108 * Check the validity of the SB found.
109 */
110STATIC int
111xfs_mount_validate_sb(
112 xfs_mount_t *mp,
113 xfs_sb_t *sbp,
114 bool check_inprogress,
115 bool check_version)
116{
Dave Chinnerff550682013-08-12 20:49:41 +1000117 if (sbp->sb_magicnum != XFS_SB_MAGIC) {
118 xfs_warn(mp, "bad magic number");
Dave Chinner24513372014-06-25 14:58:08 +1000119 return -EWRONGFS;
Dave Chinnerff550682013-08-12 20:49:41 +1000120 }
121
122
123 if (!xfs_sb_good_version(sbp)) {
124 xfs_warn(mp, "bad version");
Dave Chinner24513372014-06-25 14:58:08 +1000125 return -EWRONGFS;
Dave Chinnerff550682013-08-12 20:49:41 +1000126 }
127
128 /*
129 * Version 5 superblock feature mask validation. Reject combinations the
130 * kernel cannot support up front before checking anything else. For
131 * write validation, we don't need to check feature masks.
132 */
133 if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) {
Dave Chinnerff550682013-08-12 20:49:41 +1000134 if (xfs_sb_has_compat_feature(sbp,
135 XFS_SB_FEAT_COMPAT_UNKNOWN)) {
136 xfs_warn(mp,
Joe Perchesf41febd2015-07-29 11:52:04 +1000137"Superblock has unknown compatible features (0x%x) enabled.",
Dave Chinnerff550682013-08-12 20:49:41 +1000138 (sbp->sb_features_compat &
139 XFS_SB_FEAT_COMPAT_UNKNOWN));
Joe Perchesf41febd2015-07-29 11:52:04 +1000140 xfs_warn(mp,
141"Using a more recent kernel is recommended.");
Dave Chinnerff550682013-08-12 20:49:41 +1000142 }
143
144 if (xfs_sb_has_ro_compat_feature(sbp,
145 XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
146 xfs_alert(mp,
147"Superblock has unknown read-only compatible features (0x%x) enabled.",
148 (sbp->sb_features_ro_compat &
149 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
150 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
151 xfs_warn(mp,
Joe Perchesf41febd2015-07-29 11:52:04 +1000152"Attempted to mount read-only compatible filesystem read-write.");
153 xfs_warn(mp,
Dave Chinnerff550682013-08-12 20:49:41 +1000154"Filesystem can only be safely mounted read only.");
Joe Perchesf41febd2015-07-29 11:52:04 +1000155
Dave Chinner24513372014-06-25 14:58:08 +1000156 return -EINVAL;
Dave Chinnerff550682013-08-12 20:49:41 +1000157 }
158 }
159 if (xfs_sb_has_incompat_feature(sbp,
160 XFS_SB_FEAT_INCOMPAT_UNKNOWN)) {
161 xfs_warn(mp,
Joe Perchesf41febd2015-07-29 11:52:04 +1000162"Superblock has unknown incompatible features (0x%x) enabled.",
Dave Chinnerff550682013-08-12 20:49:41 +1000163 (sbp->sb_features_incompat &
164 XFS_SB_FEAT_INCOMPAT_UNKNOWN));
Joe Perchesf41febd2015-07-29 11:52:04 +1000165 xfs_warn(mp,
166"Filesystem can not be safely mounted by this kernel.");
Dave Chinner24513372014-06-25 14:58:08 +1000167 return -EINVAL;
Dave Chinnerff550682013-08-12 20:49:41 +1000168 }
Brian Fostera45086e2015-10-12 15:59:25 +1100169 } else if (xfs_sb_version_hascrc(sbp)) {
170 /*
171 * We can't read verify the sb LSN because the read verifier is
172 * called before the log is allocated and processed. We know the
173 * log is set up before write verifier (!check_version) calls,
174 * so just check it here.
175 */
176 if (!xfs_log_check_lsn(mp, sbp->sb_lsn))
177 return -EFSCORRUPTED;
Dave Chinnerff550682013-08-12 20:49:41 +1000178 }
179
180 if (xfs_sb_version_has_pquotino(sbp)) {
181 if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) {
182 xfs_notice(mp,
Eric Sandeen08e96e12013-10-11 20:59:05 -0500183 "Version 5 of Super block has XFS_OQUOTA bits.");
Dave Chinner24513372014-06-25 14:58:08 +1000184 return -EFSCORRUPTED;
Dave Chinnerff550682013-08-12 20:49:41 +1000185 }
186 } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD |
187 XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) {
188 xfs_notice(mp,
Eric Sandeen08e96e12013-10-11 20:59:05 -0500189"Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits.");
Dave Chinner24513372014-06-25 14:58:08 +1000190 return -EFSCORRUPTED;
Dave Chinnerff550682013-08-12 20:49:41 +1000191 }
192
Brian Fostere5376fc2015-05-29 08:57:27 +1000193 /*
194 * Full inode chunks must be aligned to inode chunk size when
195 * sparse inodes are enabled to support the sparse chunk
196 * allocation algorithm and prevent overlapping inode records.
197 */
198 if (xfs_sb_version_hassparseinodes(sbp)) {
199 uint32_t align;
200
Brian Fostere5376fc2015-05-29 08:57:27 +1000201 align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize
202 >> sbp->sb_blocklog;
203 if (sbp->sb_inoalignmt != align) {
204 xfs_warn(mp,
205"Inode block alignment (%u) must match chunk size (%u) for sparse inodes.",
206 sbp->sb_inoalignmt, align);
207 return -EINVAL;
208 }
209 }
210
Dave Chinnerff550682013-08-12 20:49:41 +1000211 if (unlikely(
212 sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
213 xfs_warn(mp,
214 "filesystem is marked as having an external log; "
215 "specify logdev on the mount command line.");
Dave Chinner24513372014-06-25 14:58:08 +1000216 return -EINVAL;
Dave Chinnerff550682013-08-12 20:49:41 +1000217 }
218
219 if (unlikely(
220 sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
221 xfs_warn(mp,
222 "filesystem is marked as having an internal log; "
223 "do not specify logdev on the mount command line.");
Dave Chinner24513372014-06-25 14:58:08 +1000224 return -EINVAL;
Dave Chinnerff550682013-08-12 20:49:41 +1000225 }
226
227 /*
228 * More sanity checking. Most of these were stolen directly from
229 * xfs_repair.
230 */
231 if (unlikely(
232 sbp->sb_agcount <= 0 ||
233 sbp->sb_sectsize < XFS_MIN_SECTORSIZE ||
234 sbp->sb_sectsize > XFS_MAX_SECTORSIZE ||
235 sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG ||
236 sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG ||
237 sbp->sb_sectsize != (1 << sbp->sb_sectlog) ||
238 sbp->sb_blocksize < XFS_MIN_BLOCKSIZE ||
239 sbp->sb_blocksize > XFS_MAX_BLOCKSIZE ||
240 sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG ||
241 sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
242 sbp->sb_blocksize != (1 << sbp->sb_blocklog) ||
Eric Sandeene1b05722014-09-09 11:47:24 +1000243 sbp->sb_dirblklog > XFS_MAX_BLOCKSIZE_LOG ||
Dave Chinnerff550682013-08-12 20:49:41 +1000244 sbp->sb_inodesize < XFS_DINODE_MIN_SIZE ||
245 sbp->sb_inodesize > XFS_DINODE_MAX_SIZE ||
246 sbp->sb_inodelog < XFS_DINODE_MIN_LOG ||
247 sbp->sb_inodelog > XFS_DINODE_MAX_LOG ||
248 sbp->sb_inodesize != (1 << sbp->sb_inodelog) ||
Eric Sandeene1b05722014-09-09 11:47:24 +1000249 sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE ||
Eric Sandeen392c6de2014-02-07 15:26:11 +1100250 sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) ||
Dave Chinnerff550682013-08-12 20:49:41 +1000251 (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) ||
252 (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) ||
253 (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) ||
254 (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) ||
255 sbp->sb_dblocks == 0 ||
256 sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) ||
Dave Chinnerab3e57b2014-05-20 07:47:05 +1000257 sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) ||
258 sbp->sb_shared_vn != 0)) {
Eric Sandeen5ef11eb2014-02-19 15:39:35 +1100259 xfs_notice(mp, "SB sanity check failed");
Dave Chinner24513372014-06-25 14:58:08 +1000260 return -EFSCORRUPTED;
Dave Chinnerff550682013-08-12 20:49:41 +1000261 }
262
263 /*
264 * Until this is fixed only page-sized or smaller data blocks work.
265 */
266 if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
267 xfs_warn(mp,
268 "File system with blocksize %d bytes. "
269 "Only pagesize (%ld) or less will currently work.",
270 sbp->sb_blocksize, PAGE_SIZE);
Dave Chinner24513372014-06-25 14:58:08 +1000271 return -ENOSYS;
Dave Chinnerff550682013-08-12 20:49:41 +1000272 }
273
274 /*
275 * Currently only very few inode sizes are supported.
276 */
277 switch (sbp->sb_inodesize) {
278 case 256:
279 case 512:
280 case 1024:
281 case 2048:
282 break;
283 default:
284 xfs_warn(mp, "inode size of %d bytes not supported",
285 sbp->sb_inodesize);
Dave Chinner24513372014-06-25 14:58:08 +1000286 return -ENOSYS;
Dave Chinnerff550682013-08-12 20:49:41 +1000287 }
288
289 if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
290 xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
291 xfs_warn(mp,
292 "file system too large to be mounted on this system.");
Dave Chinner24513372014-06-25 14:58:08 +1000293 return -EFBIG;
Dave Chinnerff550682013-08-12 20:49:41 +1000294 }
295
296 if (check_inprogress && sbp->sb_inprogress) {
297 xfs_warn(mp, "Offline file system operation in progress!");
Dave Chinner24513372014-06-25 14:58:08 +1000298 return -EFSCORRUPTED;
Dave Chinnerff550682013-08-12 20:49:41 +1000299 }
Dave Chinnerff550682013-08-12 20:49:41 +1000300 return 0;
301}
302
303void
304xfs_sb_quota_from_disk(struct xfs_sb *sbp)
305{
306 /*
307 * older mkfs doesn't initialize quota inodes to NULLFSINO. This
308 * leads to in-core values having two different values for a quota
309 * inode to be invalid: 0 and NULLFSINO. Change it to a single value
310 * NULLFSINO.
311 *
312 * Note that this change affect only the in-core values. These
313 * values are not written back to disk unless any quota information
314 * is written to the disk. Even in that case, sb_pquotino field is
315 * not written to disk unless the superblock supports pquotino.
316 */
317 if (sbp->sb_uquotino == 0)
318 sbp->sb_uquotino = NULLFSINO;
319 if (sbp->sb_gquotino == 0)
320 sbp->sb_gquotino = NULLFSINO;
321 if (sbp->sb_pquotino == 0)
322 sbp->sb_pquotino = NULLFSINO;
323
324 /*
325 * We need to do these manipilations only if we are working
326 * with an older version of on-disk superblock.
327 */
328 if (xfs_sb_version_has_pquotino(sbp))
329 return;
330
331 if (sbp->sb_qflags & XFS_OQUOTA_ENFD)
332 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
333 XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD;
334 if (sbp->sb_qflags & XFS_OQUOTA_CHKD)
335 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ?
336 XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD;
337 sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD);
338
339 if (sbp->sb_qflags & XFS_PQUOTA_ACCT) {
340 /*
341 * In older version of superblock, on-disk superblock only
342 * has sb_gquotino, and in-core superblock has both sb_gquotino
343 * and sb_pquotino. But, only one of them is supported at any
344 * point of time. So, if PQUOTA is set in disk superblock,
345 * copy over sb_gquotino to sb_pquotino.
346 */
347 sbp->sb_pquotino = sbp->sb_gquotino;
348 sbp->sb_gquotino = NULLFSINO;
349 }
350}
351
Eric Sandeen5ef828c2014-08-04 11:35:44 +1000352static void
353__xfs_sb_from_disk(
Dave Chinnerff550682013-08-12 20:49:41 +1000354 struct xfs_sb *to,
Eric Sandeen5ef828c2014-08-04 11:35:44 +1000355 xfs_dsb_t *from,
356 bool convert_xquota)
Dave Chinnerff550682013-08-12 20:49:41 +1000357{
358 to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
359 to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
360 to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
361 to->sb_rblocks = be64_to_cpu(from->sb_rblocks);
362 to->sb_rextents = be64_to_cpu(from->sb_rextents);
363 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
364 to->sb_logstart = be64_to_cpu(from->sb_logstart);
365 to->sb_rootino = be64_to_cpu(from->sb_rootino);
366 to->sb_rbmino = be64_to_cpu(from->sb_rbmino);
367 to->sb_rsumino = be64_to_cpu(from->sb_rsumino);
368 to->sb_rextsize = be32_to_cpu(from->sb_rextsize);
369 to->sb_agblocks = be32_to_cpu(from->sb_agblocks);
370 to->sb_agcount = be32_to_cpu(from->sb_agcount);
371 to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks);
372 to->sb_logblocks = be32_to_cpu(from->sb_logblocks);
373 to->sb_versionnum = be16_to_cpu(from->sb_versionnum);
374 to->sb_sectsize = be16_to_cpu(from->sb_sectsize);
375 to->sb_inodesize = be16_to_cpu(from->sb_inodesize);
376 to->sb_inopblock = be16_to_cpu(from->sb_inopblock);
377 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname));
378 to->sb_blocklog = from->sb_blocklog;
379 to->sb_sectlog = from->sb_sectlog;
380 to->sb_inodelog = from->sb_inodelog;
381 to->sb_inopblog = from->sb_inopblog;
382 to->sb_agblklog = from->sb_agblklog;
383 to->sb_rextslog = from->sb_rextslog;
384 to->sb_inprogress = from->sb_inprogress;
385 to->sb_imax_pct = from->sb_imax_pct;
386 to->sb_icount = be64_to_cpu(from->sb_icount);
387 to->sb_ifree = be64_to_cpu(from->sb_ifree);
388 to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks);
389 to->sb_frextents = be64_to_cpu(from->sb_frextents);
390 to->sb_uquotino = be64_to_cpu(from->sb_uquotino);
391 to->sb_gquotino = be64_to_cpu(from->sb_gquotino);
392 to->sb_qflags = be16_to_cpu(from->sb_qflags);
393 to->sb_flags = from->sb_flags;
394 to->sb_shared_vn = from->sb_shared_vn;
395 to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt);
396 to->sb_unit = be32_to_cpu(from->sb_unit);
397 to->sb_width = be32_to_cpu(from->sb_width);
398 to->sb_dirblklog = from->sb_dirblklog;
399 to->sb_logsectlog = from->sb_logsectlog;
400 to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize);
401 to->sb_logsunit = be32_to_cpu(from->sb_logsunit);
402 to->sb_features2 = be32_to_cpu(from->sb_features2);
403 to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2);
404 to->sb_features_compat = be32_to_cpu(from->sb_features_compat);
405 to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat);
406 to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat);
407 to->sb_features_log_incompat =
408 be32_to_cpu(from->sb_features_log_incompat);
Eric Sandeen04dd1a02014-10-02 09:24:11 +1000409 /* crc is only used on disk, not in memory; just init to 0 here. */
410 to->sb_crc = 0;
Brian Fosterfb4f2b42015-05-29 08:54:03 +1000411 to->sb_spino_align = be32_to_cpu(from->sb_spino_align);
Dave Chinnerff550682013-08-12 20:49:41 +1000412 to->sb_pquotino = be64_to_cpu(from->sb_pquotino);
413 to->sb_lsn = be64_to_cpu(from->sb_lsn);
Eric Sandeence748ea2015-07-29 11:53:31 +1000414 /*
415 * sb_meta_uuid is only on disk if it differs from sb_uuid and the
416 * feature flag is set; if not set we keep it only in memory.
417 */
418 if (xfs_sb_version_hasmetauuid(to))
419 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
420 else
421 uuid_copy(&to->sb_meta_uuid, &from->sb_uuid);
Eric Sandeen5ef828c2014-08-04 11:35:44 +1000422 /* Convert on-disk flags to in-memory flags? */
423 if (convert_xquota)
424 xfs_sb_quota_from_disk(to);
425}
426
427void
428xfs_sb_from_disk(
429 struct xfs_sb *to,
430 xfs_dsb_t *from)
431{
432 __xfs_sb_from_disk(to, from, true);
Dave Chinnerff550682013-08-12 20:49:41 +1000433}
434
Dave Chinner4d11a402015-01-22 09:10:26 +1100435static void
Dave Chinnerff550682013-08-12 20:49:41 +1000436xfs_sb_quota_to_disk(
Dave Chinner4d11a402015-01-22 09:10:26 +1100437 struct xfs_dsb *to,
438 struct xfs_sb *from)
Dave Chinnerff550682013-08-12 20:49:41 +1000439{
440 __uint16_t qflags = from->sb_qflags;
441
Dave Chinner4d11a402015-01-22 09:10:26 +1100442 to->sb_uquotino = cpu_to_be64(from->sb_uquotino);
443 if (xfs_sb_version_has_pquotino(from)) {
444 to->sb_qflags = cpu_to_be16(from->sb_qflags);
445 to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
446 to->sb_pquotino = cpu_to_be64(from->sb_pquotino);
Dave Chinnerff550682013-08-12 20:49:41 +1000447 return;
Dave Chinnerff550682013-08-12 20:49:41 +1000448 }
449
450 /*
Dave Chinner4d11a402015-01-22 09:10:26 +1100451 * The in-core version of sb_qflags do not have XFS_OQUOTA_*
452 * flags, whereas the on-disk version does. So, convert incore
453 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags.
Dave Chinnerff550682013-08-12 20:49:41 +1000454 */
Dave Chinner4d11a402015-01-22 09:10:26 +1100455 qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD |
456 XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD);
457
458 if (from->sb_qflags &
459 (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD))
460 qflags |= XFS_OQUOTA_ENFD;
461 if (from->sb_qflags &
462 (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD))
463 qflags |= XFS_OQUOTA_CHKD;
464 to->sb_qflags = cpu_to_be16(qflags);
465
466 /*
467 * GQUOTINO and PQUOTINO cannot be used together in versions
468 * of superblock that do not have pquotino. from->sb_flags
469 * tells us which quota is active and should be copied to
470 * disk. If neither are active, we should NULL the inode.
471 *
472 * In all cases, the separate pquotino must remain 0 because it
473 * it beyond the "end" of the valid non-pquotino superblock.
474 */
475 if (from->sb_qflags & XFS_GQUOTA_ACCT)
Dave Chinnerff550682013-08-12 20:49:41 +1000476 to->sb_gquotino = cpu_to_be64(from->sb_gquotino);
Dave Chinner4d11a402015-01-22 09:10:26 +1100477 else if (from->sb_qflags & XFS_PQUOTA_ACCT)
Dave Chinnerff550682013-08-12 20:49:41 +1000478 to->sb_gquotino = cpu_to_be64(from->sb_pquotino);
Dave Chinner03e01342014-07-15 07:28:41 +1000479 else {
480 /*
481 * We can't rely on just the fields being logged to tell us
482 * that it is safe to write NULLFSINO - we should only do that
483 * if quotas are not actually enabled. Hence only write
484 * NULLFSINO if both in-core quota inodes are NULL.
485 */
486 if (from->sb_gquotino == NULLFSINO &&
487 from->sb_pquotino == NULLFSINO)
488 to->sb_gquotino = cpu_to_be64(NULLFSINO);
489 }
Dave Chinnerff550682013-08-12 20:49:41 +1000490
Dave Chinner4d11a402015-01-22 09:10:26 +1100491 to->sb_pquotino = 0;
Dave Chinnerff550682013-08-12 20:49:41 +1000492}
493
Dave Chinnerff550682013-08-12 20:49:41 +1000494void
495xfs_sb_to_disk(
Dave Chinner4d11a402015-01-22 09:10:26 +1100496 struct xfs_dsb *to,
497 struct xfs_sb *from)
Dave Chinnerff550682013-08-12 20:49:41 +1000498{
Dave Chinner4d11a402015-01-22 09:10:26 +1100499 xfs_sb_quota_to_disk(to, from);
Dave Chinnerff550682013-08-12 20:49:41 +1000500
Dave Chinner4d11a402015-01-22 09:10:26 +1100501 to->sb_magicnum = cpu_to_be32(from->sb_magicnum);
502 to->sb_blocksize = cpu_to_be32(from->sb_blocksize);
503 to->sb_dblocks = cpu_to_be64(from->sb_dblocks);
504 to->sb_rblocks = cpu_to_be64(from->sb_rblocks);
505 to->sb_rextents = cpu_to_be64(from->sb_rextents);
506 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid));
507 to->sb_logstart = cpu_to_be64(from->sb_logstart);
508 to->sb_rootino = cpu_to_be64(from->sb_rootino);
509 to->sb_rbmino = cpu_to_be64(from->sb_rbmino);
510 to->sb_rsumino = cpu_to_be64(from->sb_rsumino);
511 to->sb_rextsize = cpu_to_be32(from->sb_rextsize);
512 to->sb_agblocks = cpu_to_be32(from->sb_agblocks);
513 to->sb_agcount = cpu_to_be32(from->sb_agcount);
514 to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks);
515 to->sb_logblocks = cpu_to_be32(from->sb_logblocks);
516 to->sb_versionnum = cpu_to_be16(from->sb_versionnum);
517 to->sb_sectsize = cpu_to_be16(from->sb_sectsize);
518 to->sb_inodesize = cpu_to_be16(from->sb_inodesize);
519 to->sb_inopblock = cpu_to_be16(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 = cpu_to_be64(from->sb_icount);
530 to->sb_ifree = cpu_to_be64(from->sb_ifree);
531 to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks);
532 to->sb_frextents = cpu_to_be64(from->sb_frextents);
Dave Chinnerff550682013-08-12 20:49:41 +1000533
Dave Chinner4d11a402015-01-22 09:10:26 +1100534 to->sb_flags = from->sb_flags;
535 to->sb_shared_vn = from->sb_shared_vn;
536 to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt);
537 to->sb_unit = cpu_to_be32(from->sb_unit);
538 to->sb_width = cpu_to_be32(from->sb_width);
539 to->sb_dirblklog = from->sb_dirblklog;
540 to->sb_logsectlog = from->sb_logsectlog;
541 to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize);
542 to->sb_logsunit = cpu_to_be32(from->sb_logsunit);
Dave Chinner074e4272015-01-22 09:10:33 +1100543
544 /*
545 * We need to ensure that bad_features2 always matches features2.
546 * Hence we enforce that here rather than having to remember to do it
547 * everywhere else that updates features2.
548 */
549 from->sb_bad_features2 = from->sb_features2;
Dave Chinner4d11a402015-01-22 09:10:26 +1100550 to->sb_features2 = cpu_to_be32(from->sb_features2);
551 to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2);
Dave Chinnerff550682013-08-12 20:49:41 +1000552
Dave Chinner4d11a402015-01-22 09:10:26 +1100553 if (xfs_sb_version_hascrc(from)) {
554 to->sb_features_compat = cpu_to_be32(from->sb_features_compat);
555 to->sb_features_ro_compat =
556 cpu_to_be32(from->sb_features_ro_compat);
557 to->sb_features_incompat =
558 cpu_to_be32(from->sb_features_incompat);
559 to->sb_features_log_incompat =
560 cpu_to_be32(from->sb_features_log_incompat);
Brian Fosterfb4f2b42015-05-29 08:54:03 +1000561 to->sb_spino_align = cpu_to_be32(from->sb_spino_align);
Dave Chinner4d11a402015-01-22 09:10:26 +1100562 to->sb_lsn = cpu_to_be64(from->sb_lsn);
Eric Sandeence748ea2015-07-29 11:53:31 +1000563 if (xfs_sb_version_hasmetauuid(from))
564 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid);
Dave Chinnerff550682013-08-12 20:49:41 +1000565 }
566}
567
568static int
569xfs_sb_verify(
570 struct xfs_buf *bp,
571 bool check_version)
572{
573 struct xfs_mount *mp = bp->b_target->bt_mount;
574 struct xfs_sb sb;
575
Eric Sandeen5ef828c2014-08-04 11:35:44 +1000576 /*
577 * Use call variant which doesn't convert quota flags from disk
578 * format, because xfs_mount_validate_sb checks the on-disk flags.
579 */
580 __xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false);
Dave Chinnerff550682013-08-12 20:49:41 +1000581
582 /*
583 * Only check the in progress field for the primary superblock as
584 * mkfs.xfs doesn't clear it from secondary superblocks.
585 */
586 return xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR,
587 check_version);
588}
589
590/*
591 * If the superblock has the CRC feature bit set or the CRC field is non-null,
592 * check that the CRC is valid. We check the CRC field is non-null because a
593 * single bit error could clear the feature bit and unused parts of the
594 * superblock are supposed to be zero. Hence a non-null crc field indicates that
595 * we've potentially lost a feature bit and we should check it anyway.
Eric Sandeen10e6e652013-09-09 15:33:29 -0500596 *
597 * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the
598 * last field in V4 secondary superblocks. So for secondary superblocks,
599 * we are more forgiving, and ignore CRC failures if the primary doesn't
600 * indicate that the fs version is V5.
Dave Chinnerff550682013-08-12 20:49:41 +1000601 */
602static void
603xfs_sb_read_verify(
604 struct xfs_buf *bp)
605{
606 struct xfs_mount *mp = bp->b_target->bt_mount;
607 struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp);
608 int error;
609
610 /*
611 * open code the version check to avoid needing to convert the entire
612 * superblock from disk order just to check the version number
613 */
614 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) &&
615 (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) ==
616 XFS_SB_VERSION_5) ||
617 dsb->sb_crc != 0)) {
618
Eric Sandeen51582172014-02-27 15:17:27 +1100619 if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) {
Eric Sandeen10e6e652013-09-09 15:33:29 -0500620 /* Only fail bad secondaries on a known V5 filesystem */
Eric Sandeen7a01e7072014-02-19 15:33:05 +1100621 if (bp->b_bn == XFS_SB_DADDR ||
Eric Sandeen10e6e652013-09-09 15:33:29 -0500622 xfs_sb_version_hascrc(&mp->m_sb)) {
Dave Chinner24513372014-06-25 14:58:08 +1000623 error = -EFSBADCRC;
Eric Sandeen10e6e652013-09-09 15:33:29 -0500624 goto out_error;
625 }
Dave Chinnerff550682013-08-12 20:49:41 +1000626 }
627 }
628 error = xfs_sb_verify(bp, true);
629
630out_error:
631 if (error) {
Dave Chinnerff550682013-08-12 20:49:41 +1000632 xfs_buf_ioerror(bp, error);
Dave Chinner24513372014-06-25 14:58:08 +1000633 if (error == -EFSCORRUPTED || error == -EFSBADCRC)
Eric Sandeence5028c2014-02-27 15:23:10 +1100634 xfs_verifier_error(bp);
Dave Chinnerff550682013-08-12 20:49:41 +1000635 }
636}
637
638/*
639 * We may be probed for a filesystem match, so we may not want to emit
640 * messages when the superblock buffer is not actually an XFS superblock.
Zhi Yong Wu25337872013-08-12 03:15:02 +0000641 * If we find an XFS superblock, then run a normal, noisy mount because we are
Dave Chinnerff550682013-08-12 20:49:41 +1000642 * really going to mount it and want to know about errors.
643 */
644static void
645xfs_sb_quiet_read_verify(
646 struct xfs_buf *bp)
647{
648 struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp);
649
Dave Chinnerff550682013-08-12 20:49:41 +1000650 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) {
651 /* XFS filesystem, verify noisily! */
652 xfs_sb_read_verify(bp);
653 return;
654 }
655 /* quietly fail */
Dave Chinner24513372014-06-25 14:58:08 +1000656 xfs_buf_ioerror(bp, -EWRONGFS);
Dave Chinnerff550682013-08-12 20:49:41 +1000657}
658
659static void
660xfs_sb_write_verify(
661 struct xfs_buf *bp)
662{
663 struct xfs_mount *mp = bp->b_target->bt_mount;
664 struct xfs_buf_log_item *bip = bp->b_fspriv;
665 int error;
666
667 error = xfs_sb_verify(bp, false);
668 if (error) {
Dave Chinnerff550682013-08-12 20:49:41 +1000669 xfs_buf_ioerror(bp, error);
Eric Sandeence5028c2014-02-27 15:23:10 +1100670 xfs_verifier_error(bp);
Dave Chinnerff550682013-08-12 20:49:41 +1000671 return;
672 }
673
674 if (!xfs_sb_version_hascrc(&mp->m_sb))
675 return;
676
677 if (bip)
678 XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
679
Eric Sandeenf1dbcd72014-02-27 15:18:23 +1100680 xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF);
Dave Chinnerff550682013-08-12 20:49:41 +1000681}
682
683const struct xfs_buf_ops xfs_sb_buf_ops = {
Eric Sandeen233135b2016-01-04 16:10:19 +1100684 .name = "xfs_sb",
Dave Chinnerff550682013-08-12 20:49:41 +1000685 .verify_read = xfs_sb_read_verify,
686 .verify_write = xfs_sb_write_verify,
687};
688
689const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
Eric Sandeen233135b2016-01-04 16:10:19 +1100690 .name = "xfs_sb_quiet",
Dave Chinnerff550682013-08-12 20:49:41 +1000691 .verify_read = xfs_sb_quiet_read_verify,
692 .verify_write = xfs_sb_write_verify,
693};
694
695/*
696 * xfs_mount_common
697 *
698 * Mount initialization code establishing various mount
699 * fields from the superblock associated with the given
700 * mount structure
701 */
702void
703xfs_sb_mount_common(
704 struct xfs_mount *mp,
705 struct xfs_sb *sbp)
706{
707 mp->m_agfrotor = mp->m_agirotor = 0;
708 spin_lock_init(&mp->m_agirotor_lock);
709 mp->m_maxagi = mp->m_sb.sb_agcount;
710 mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
711 mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
712 mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
713 mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
714 mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
715 mp->m_blockmask = sbp->sb_blocksize - 1;
716 mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
717 mp->m_blockwmask = mp->m_blockwsize - 1;
718
719 mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
720 mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
721 mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
722 mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
723
724 mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1);
725 mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0);
726 mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2;
727 mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2;
728
729 mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
730 mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
731 mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
732 mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
733
Darrick J. Wong035e00a2016-08-03 11:36:07 +1000734 mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 1);
735 mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 0);
736 mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2;
737 mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2;
738
Dave Chinnerff550682013-08-12 20:49:41 +1000739 mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
740 mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
741 sbp->sb_inopblock);
742 mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
Brian Foster066a1882015-05-29 08:55:20 +1000743
744 if (sbp->sb_spino_align)
745 mp->m_ialloc_min_blks = sbp->sb_spino_align;
746 else
747 mp->m_ialloc_min_blks = mp->m_ialloc_blks;
Darrick J. Wong52548852016-08-03 11:38:24 +1000748 mp->m_alloc_set_aside = xfs_alloc_set_aside(mp);
749 mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp);
Dave Chinnerff550682013-08-12 20:49:41 +1000750}
751
752/*
753 * xfs_initialize_perag_data
754 *
755 * Read in each per-ag structure so we can count up the number of
756 * allocated inodes, free inodes and used filesystem blocks as this
757 * information is no longer persistent in the superblock. Once we have
758 * this information, write it into the in-core superblock structure.
759 */
760int
761xfs_initialize_perag_data(
762 struct xfs_mount *mp,
763 xfs_agnumber_t agcount)
764{
765 xfs_agnumber_t index;
766 xfs_perag_t *pag;
767 xfs_sb_t *sbp = &mp->m_sb;
768 uint64_t ifree = 0;
769 uint64_t ialloc = 0;
770 uint64_t bfree = 0;
771 uint64_t bfreelst = 0;
772 uint64_t btree = 0;
773 int error;
774
775 for (index = 0; index < agcount; index++) {
776 /*
777 * read the agf, then the agi. This gets us
778 * all the information we need and populates the
779 * per-ag structures for us.
780 */
781 error = xfs_alloc_pagf_init(mp, NULL, index, 0);
782 if (error)
783 return error;
784
785 error = xfs_ialloc_pagi_init(mp, NULL, index);
786 if (error)
787 return error;
788 pag = xfs_perag_get(mp, index);
789 ifree += pag->pagi_freecount;
790 ialloc += pag->pagi_count;
791 bfree += pag->pagf_freeblks;
792 bfreelst += pag->pagf_flcount;
793 btree += pag->pagf_btreeblks;
794 xfs_perag_put(pag);
795 }
Dave Chinner5681ca42015-02-23 21:22:31 +1100796
797 /* Overwrite incore superblock counters with just-read data */
Dave Chinnerff550682013-08-12 20:49:41 +1000798 spin_lock(&mp->m_sb_lock);
799 sbp->sb_ifree = ifree;
800 sbp->sb_icount = ialloc;
801 sbp->sb_fdblocks = bfree + bfreelst + btree;
802 spin_unlock(&mp->m_sb_lock);
803
Dave Chinner5681ca42015-02-23 21:22:31 +1100804 xfs_reinit_percpu_counters(mp);
Dave Chinnerff550682013-08-12 20:49:41 +1000805
806 return 0;
807}
808
809/*
Dave Chinner61e63ec2015-01-22 09:10:31 +1100810 * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
811 * into the superblock buffer to be logged. It does not provide the higher
812 * level of locking that is needed to protect the in-core superblock from
813 * concurrent access.
Dave Chinnerff550682013-08-12 20:49:41 +1000814 */
815void
Dave Chinner61e63ec2015-01-22 09:10:31 +1100816xfs_log_sb(
Dave Chinner4d11a402015-01-22 09:10:26 +1100817 struct xfs_trans *tp)
Dave Chinnerff550682013-08-12 20:49:41 +1000818{
Dave Chinner4d11a402015-01-22 09:10:26 +1100819 struct xfs_mount *mp = tp->t_mountp;
820 struct xfs_buf *bp = xfs_trans_getsb(tp, mp, 0);
Dave Chinnerff550682013-08-12 20:49:41 +1000821
Dave Chinner501ab322015-02-23 21:19:28 +1100822 mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount);
Dave Chinnere88b64e2015-02-23 21:19:53 +1100823 mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree);
Dave Chinner0d485ad2015-02-23 21:22:03 +1100824 mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks);
Dave Chinner501ab322015-02-23 21:19:28 +1100825
Dave Chinner4d11a402015-01-22 09:10:26 +1100826 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb);
Dave Chinnerff550682013-08-12 20:49:41 +1000827 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
Dave Chinner4d11a402015-01-22 09:10:26 +1100828 xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb));
Dave Chinnerff550682013-08-12 20:49:41 +1000829}
Dave Chinner61e63ec2015-01-22 09:10:31 +1100830
831/*
832 * xfs_sync_sb
833 *
834 * Sync the superblock to disk.
835 *
836 * Note that the caller is responsible for checking the frozen state of the
837 * filesystem. This procedure uses the non-blocking transaction allocator and
838 * thus will allow modifications to a frozen fs. This is required because this
839 * code can be called during the process of freezing where use of the high-level
840 * allocator would deadlock.
841 */
842int
843xfs_sync_sb(
844 struct xfs_mount *mp,
845 bool wait)
846{
847 struct xfs_trans *tp;
848 int error;
849
Christoph Hellwig253f4912016-04-06 09:19:55 +1000850 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0,
851 XFS_TRANS_NO_WRITECOUNT, &tp);
852 if (error)
Dave Chinner61e63ec2015-01-22 09:10:31 +1100853 return error;
Dave Chinner61e63ec2015-01-22 09:10:31 +1100854
855 xfs_log_sb(tp);
856 if (wait)
857 xfs_trans_set_sync(tp);
Christoph Hellwig70393312015-06-04 13:48:08 +1000858 return xfs_trans_commit(tp);
Dave Chinner61e63ec2015-01-22 09:10:31 +1100859}