2.4.20-pre1-ac1: Filesystem panic attempting to mount ext3
Hello,
the problem appeared on the first partition of an ide IBM-DHEA-36481 with one fat partition on it. I repartioned the device (4 primaries) and "mke2fs -j" three of them.
Than i tried to mount the newly created filesystems and got this in syslog:
Sep 1 08:47:32 steel kernel: FAT: Did not find valid FSINFO signature. Sep 1 08:47:32 steel kernel: Found signature1 0x0 signature2 0x0 sector=1. Sep 1 08:47:32 steel kernel: Directory 1: bad FAT Sep 1 08:47:32 steel kernel: attempt to access beyond end of device Sep 1 08:47:32 steel kernel: 16:41: rw=0, want=902238098, limit=3903763 Sep 1 08:47:32 steel kernel: attempt to access beyond end of device Sep 1 08:47:32 steel kernel: 16:41: rw=0, want=902238099, limit=3903763 Sep 1 08:47:32 steel kernel: attempt to access beyond end of device Sep 1 08:47:32 steel kernel: 16:41: rw=0, want=902238099, limit=3903763 Sep 1 08:47:32 steel kernel: attempt to access beyond end of device Sep 1 08:47:32 steel kernel: 16:41: rw=0, want=902238100, limit=3903763 Sep 1 08:47:32 steel kernel: attempt to access beyond end of device Sep 1 08:47:32 steel kernel: 16:41: rw=0, want=902238100, limit=3903763 Sep 1 08:47:32 steel kernel: attempt to access beyond end of device Sep 1 08:47:32 steel kernel: 16:41: rw=0, want=902238101, limit=3903763 Sep 1 08:47:32 steel kernel: attempt to access beyond end of device Sep 1 08:47:32 steel kernel: 16:41: rw=0, want=902238101, limit=3903763 Sep 1 08:47:32 steel kernel: attempt to access beyond end of device Sep 1 08:47:32 steel kernel: 16:41: rw=0, want=902238102, limit=3903763 Sep 1 08:47:32 steel kernel: Filesystem panic (dev 16:41). Sep 1 08:47:32 steel kernel: FAT error Sep 1 08:47:32 steel kernel: File system has been set read-only
Umount produced something as well:
Sep 1 08:47:54 steel kernel: FAT: Did not find valid FSINFO signature. Sep 1 08:47:54 steel kernel: Found signature1 0x0 signature2 0x0 sector=1.
Assuming that some garbage was left on the disk event after mke2fs, i did "dd if=/dev/zero of=/dev/hdd1 bs=512", which cured the problem, after being followed by mke2fs.
# fdisk -l /dev/hdd
Disk /dev/hdd: 255 heads, 63 sectors, 790 cylinders Units = cylinders of 16065 * 512 bytes
Device Boot Start End Blocks Id System /dev/hdd1 * 1 486 3903763+ 83 Linux /dev/hdd2 487 608 979965 83 Linux /dev/hdd3 609 632 192780 82 Linux swap /dev/hdd4 633 790 1269135 83 Linux
Quoted 5 lines · Alex Riesen
Hello,
the problem appeared on the first partition of an ide IBM-DHEA-36481 with one fat partition on it. I repartioned the device (4 primaries) and "mke2fs -j" three of them.
[...]
Quoted 9 lines
Umount produced something as well:
Sep 1 08:47:54 steel kernel: FAT: Did not find valid FSINFO signature. Sep 1 08:47:54 steel kernel: Found signature1 0x0 signature2 0x0 sector=1.Assuming that some garbage was left on the disk event after mke2fs, i did "dd if=/dev/zero of=/dev/hdd1 bs=512", which cured the problem, after being followed by mke2fs.
It's problem of fatfs, and I think it's fixed in 2.5 series. I'll try to back porting the part of fatfs of 2.5 series.
Signature · 1 line
OGAWA Hirofumi <[email protected]>
OGAWA Hirofumi, Sun, Sep 01, 2002 19:02:59 +0200:
Quoted 16 lines
Quoted 5 lines · Alex Riesen
Hello,
the problem appeared on the first partition of an ide IBM-DHEA-36481 with one fat partition on it. I repartioned the device (4 primaries) and "mke2fs -j" three of them.
[...]
Quoted 6 lines
Umount produced something as well:
Sep 1 08:47:54 steel kernel: FAT: Did not find valid FSINFO signature. Sep 1 08:47:54 steel kernel: Found signature1 0x0 signature2 0x0 sector=1.
Quoted 6 lines
Quoted 3 lines
Assuming that some garbage was left on the disk event after mke2fs, i did "dd if=/dev/zero of=/dev/hdd1 bs=512", which cured the problem, after being followed by mke2fs.
It's problem of fatfs, and I think it's fixed in 2.5 series. I'll try to back porting the part of fatfs of 2.5 series.
Thanks.
Does mounting somehow depend on order of initialization of filesystems? The fatfs is a module here, and it was inserted last (ext3 is compiled in).
-alex
Quoted 21 lines · Alex Riesen
OGAWA Hirofumi, Sun, Sep 01, 2002 19:02:59 +0200:
Quoted 20 lines
Quoted 5 lines · Alex Riesen
Hello,
the problem appeared on the first partition of an ide IBM-DHEA-36481 with one fat partition on it. I repartioned the device (4 primaries) and "mke2fs -j" three of them.
[...]
Quoted 9 lines
Umount produced something as well:
Sep 1 08:47:54 steel kernel: FAT: Did not find valid FSINFO signature. Sep 1 08:47:54 steel kernel: Found signature1 0x0 signature2 0x0 sector=1.Assuming that some garbage was left on the disk event after mke2fs, i did "dd if=/dev/zero of=/dev/hdd1 bs=512", which cured the problem, after being followed by mke2fs.
Quoted 6 lines
Quoted 2 lines
It's problem of fatfs, and I think it's fixed in 2.5 series. I'll try to back porting the part of fatfs of 2.5 series.
Thanks.
Does mounting somehow depend on order of initialization of filesystems? The fatfs is a module here, and it was inserted last (ext3 is compiled in).
Well, the filesystem is detected by the mount command, except the root filesystem. And mount command detects FAT before EXT3.
Signature · 1 line
OGAWA Hirofumi <[email protected]>
OGAWA Hirofumi, Sun, Sep 01, 2002 21:02:04 +0200:
Quoted 2 lines
Well, the filesystem is detected by the mount command, except the root filesystem. And mount command detects FAT before EXT3.
A. That explains everything.
-alex
Hi,
Quoted 8 lines · Alex Riesen
the problem appeared on the first partition of an ide IBM-DHEA-36481 with one fat partition on it. I repartioned the device (4 primaries) and "mke2fs -j" three of them.
Than i tried to mount the newly created filesystems and got this in syslog:
Sep 1 08:47:32 steel kernel: FAT: Did not find valid FSINFO signature.
Which version of e2fsprogs?
Quoted 3 lines
Assuming that some garbage was left on the disk event after mke2fs, i did "dd if=/dev/zero of=/dev/hdd1 bs=512", which cured the problem, after being followed by mke2fs.
mke2fs from older versions of e2fsprogs didn't clear out all the filesystem and md signatures on a new filesystem.
The right way to avoid this is to tell the kernel to mount the fs as ext2 or ext3 explicitly, not to rely on the fs-type autodetection in mount().
Cheers,
Stephen
Stephen C. Tweedie, Mon, Sep 02, 2002 12:38:34 +0200:
Quoted 8 lines
Hi,
Quoted 4 lines · Alex Riesen
the problem appeared on the first partition of an ide IBM-DHEA-36481 with one fat partition on it. I repartioned the device (4 primaries) and "mke2fs -j" three of them.
Quoted 6 lines
Quoted 4 lines
Than i tried to mount the newly created filesystems and got this in syslog:
Sep 1 08:47:32 steel kernel: FAT: Did not find valid FSINFO signature.Which version of e2fsprogs?
mke2fs 1.27 (8-Mar-2002)
Quoted 10 lines
Quoted 3 lines
Assuming that some garbage was left on the disk event after mke2fs, i did "dd if=/dev/zero of=/dev/hdd1 bs=512", which cured the problem, after being followed by mke2fs.
mke2fs from older versions of e2fsprogs didn't clear out all the filesystem and md signatures on a new filesystem.
The right way to avoid this is to tell the kernel to mount the fs as ext2 or ext3 explicitly, not to rely on the fs-type autodetection in
mount().
This works, too. Anyway, not an issue anymore