Due to xfs_repair uses direct IO, sometimes it can't read superblock
from an image file has smaller sector size than host filesystem.
Especially that superblock doesn't align with host filesystem's
sector size.
To avoid this, when direct read returns EINVAL, turn off direct IO,
then try to read again.
Signed-off-by: Zorro Lang <zlang@xxxxxxxxxx>
---
Hi,
I found this bug when I try to modify xfstests' xfs/078 on s390x,
manually reproduce this bug by below steps:
[root@ibm-z-32 ~]# blockdev --getss --getpbsz --getbsz /dev/dasda1
4096
4096
4096
[root@ibm-z-32 ~]# truncate -s $((168024*1024)) fsfile
[root@ibm-z-32 ~]# echo $((168024*1024))
172056576
[root@ibm-z-32 ~]# losetup /dev/loop0 fsfile
[root@ibm-z-32 ~]# mkfs.xfs -f -b size=1k /dev/loop0
meta-data=/dev/loop0 isize=512 agcount=4, agsize=42006 blks
= sectsz=512 attr=2, projid32bit=1
= crc=1 finobt=0, sparse=0
data = bsize=1024 blocks=168024, imaxpct=25
= sunit=0 swidth=0 blks
naming =version 2 bsize=4096 ascii-ci=0 ftype=1
log =internal log bsize=1024 blocks=2573, version=2
= sectsz=512 sunit=0 blks, lazy-count=1
realtime =none extsz=4096 blocks=0, rtextents=0
[root@ibm-z-32 ~]# losetup -d /dev/loop0
[root@ibm-z-32 ~]# xfs_io -c "pwrite 688230400 1024" fsfile
wrote 1024/1024 bytes at offset 688230400
1 KiB, 1 ops; 0.0000 sec (13.563 MiB/sec and 13888.8889 ops/sec)
[root@ibm-z-32 ~]# losetup /dev/loop0 fsfile
[root@ibm-z-32 ~]# mount /dev/loop0 /mnt/test
[root@ibm-z-32 ~]# xfs_growfs /mnt/test
meta-data=/dev/loop0 isize=512 agcount=4, agsize=42006 blks
= sectsz=512 attr=2, projid32bit=1
= crc=1 finobt=0 spinodes=0
data = bsize=1024 blocks=168024, imaxpct=25
= sunit=0 swidth=0 blks
naming =version 2 bsize=4096 ascii-ci=0 ftype=1
log =internal bsize=1024 blocks=2573, version=2
= sectsz=512 sunit=0 blks, lazy-count=1
realtime =none extsz=4096 blocks=0, rtextents=0
data blocks changed from 168024 to 672096
[root@ibm-z-32 ~]# umount -d /mnt/test
[root@ibm-z-32 ~]# losetup -a
[root@ibm-z-32 ~]# xfs_repair -f -n fsfile
Phase 1 - find and verify superblock...
superblock read failed, offset 43014144, size 131072, ag 1, rval -1
fatal error -- Invalid argument
To avoid this problem, I use the same way as Dave did in:
f63fd26 repair: handle repair of image files on large sector size filesystems
So there're some duplicate code in "fcntl" part, I want to pick up
this part to be a common function in xfsprogs or xfsprogs/repair,
but I don't know where's the proper place and if that's necessary?
Thanks,
Zorro
repair/sb.c | 25 +++++++++++++++++++++++--
1 file changed, 23 insertions(+), 2 deletions(-)
diff --git a/repair/sb.c b/repair/sb.c
index 77e5154..617ad98 100644
--- a/repair/sb.c
+++ b/repair/sb.c
@@ -567,11 +567,32 @@ get_sb(xfs_sb_t *sbp, xfs_off_t off, int size, xfs_agnumber_t agno)
}
if ((rval = read(x.dfd, buf, size)) != size) {
+ /*
+ * If file image sector is smaller than the host filesystem
+ * sector, this O_DIRECT read will return EINVAL. So turn
+ * off O_DIRECT and try to buffer read.
+ */
+ if (errno == EINVAL) {
+ long old_flags;
+
+ old_flags = fcntl(x.dfd, F_GETFL, 0);
+ if (fcntl(x.dfd, F_SETFL, old_flags & ~O_DIRECT) < 0) {
+ do_warn(
+ _("Sector size on host filesystem larger than image sector size.\n"
+ "Cannot turn off direct IO, so exiting.\n"));
+ exit(1);
+ } else if ((rval = read(x.dfd, buf, size)) == size) {
+ errno = 0;
+ }
+ }
error = errno;
- do_warn(
+ if (error != 0) {
+ do_warn(
_("superblock read failed, offset %" PRId64 ", size %d, ag %u, rval %d\n"),
off, size, agno, rval);
- do_error("%s\n", strerror(error));
+ do_error("%s\n", strerror(error));
+ }
+
}
libxfs_sb_from_disk(sbp, buf);
--
2.7.4
--
To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
