Re: ext3 message if FS is not ext3

Re: ext3 message if FS is not ext3

[Steve Dodd]

On Wed, May 23, 2001 at 01:06:16PM +0100, Stephen C. Tweedie wrote:
> On Wed, May 23, 2001 at 02:00:13PM +0200, Florian Lohoff wrote:

> > i think this message should be removed ;)
[..]
> > VFS: Can't find an ext3 filesystem on dev fd(2,0).

> mount(8) tried to get the kernel to mount /dev/fd0 as an ext3
> filesystem.  The kernel is entitled to emit an error in that case.
> ext2 will complain too.

Shouldn't it be doing the mount 'silently' when mount(8) is guessing the
filesystem type? I'm seeing this too (2.2.19 + ext3 0.0.6b):

lilith:tmp$ cat /etc/filesystems
ext3
vfat
lilith:tmp$ grep floppy /etc/fstab
/dev/fd0 /floppy auto noauto,nodev,nosuid,user 0 0
lilith:tmp$ # with a VFAT filesystem on /dev/fd0:
lilith:tmp$ mount /floppy
VFS: Can't find an ext3 filesystem on dev fd(2,0).

Looking at the kernel source, read_super (and hence ext3_read_super) are only
called with silent=1 when mounting the root filesystem. I believe mount(8)
checks for magic numbers at the start of the filesystem, and so avoids
attempting a mount for ext2 or various others.

As the kernel (2.2 or 2.4) doesn't seem to provide a way for userspace to
request a silent mount, I don't know whose (if anyone's) bug this is.

-- 
PGP signed or encrypted mail preferred, key ID 0x68383A73.
Please *do* Cc: me on mailing list replies.

Re: ext3 message if FS is not ext3

[Stephen C. Tweedie]

Hi,

On Sat, May 26, 2001 at 10:54:39AM +0100, Steve Dodd wrote:
> On Wed, May 23, 2001 at 01:06:16PM +0100, Stephen C. Tweedie wrote:
> > On Wed, May 23, 2001 at 02:00:13PM +0200, Florian Lohoff wrote:
> 
> > > i think this message should be removed ;)
> [..]
> > > VFS: Can't find an ext3 filesystem on dev fd(2,0).
> 
> > mount(8) tried to get the kernel to mount /dev/fd0 as an ext3
> > filesystem.  The kernel is entitled to emit an error in that case.
> > ext2 will complain too.
> 
> Shouldn't it be doing the mount 'silently' when mount(8) is guessing the
> filesystem type? I'm seeing this too (2.2.19 + ext3 0.0.6b):

That's possible.  There is actually a "silent" parameter passed to the
filesystem mount routines in the kernel.  Unfortunately, it's actually
encoded as a "mount-root" parameter, and it gets used when attempting
the initial root mount when we don't know the filesystem type in
advance.  Ext2/ext3 interpret the mount-root parameter as an
instruction to fail silently, but I'm not sure whether there are other
filesystems which actually use it as a hint to treat the root
differently.  If so, we can't overload that parameter to mean
go-silent.  That's no big deal, we can create a new MS_* flag for
it in that case.

> As the kernel (2.2 or 2.4) doesn't seem to provide a way for userspace to
> request a silent mount, I don't know whose (if anyone's) bug this is.

It's not a bug.  It is a missing feature, perhaps.

Cheers,
 Stephen