[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Ext3 partition not appearing in df output

On Thu, Nov 01, 2001 at 02:01:21PM +0000, Stephen C. Tweedie wrote:
> For reliable operation of standard mount points, we want to have fs types
> specified in /etc/fstab as accurately as possible.  However, there is
> still good reason to want user-space to be able to autodetect other
> cases: udf versus iso9660 depending on media type for /dev/dvd, for
> example, or ext3 vs ext2 depending on kernel capabilities.

Ok, so why not extend fstab(5) fs_vfstype field (and mount(8) -t switch) to
be able to have multiple filesystem specified ?

so, you would have /etc/fstab

/dev/cdrom  /mnt/cdrom  udf,iso9660  owner,noauto,ro  0 0
/dev/hda1   /usr        ext3,ext2    defaults         1 2

mount (and fsck) would then try first ext3, and if that fails, the ext2.
It will probably be little slower than current magic-string guessing, but I
don't think either mount or fsck are *that* much speed-critical.

> Supporting such mechanisms does *not* mean that we have to guess every
> mount point.  The danger of doing the guessing by default does not
> imply that the functionality should be removed entirely.

Maybe not... But it means that mount and fsck will always have to up-to-date
with kernel (or users will whine that it does not work correctly for their
setup), and that every change will have to be done in both, and some error
(or not keeping them in sync) could have grave results.

With the idea above, this responsibility to specify exactly what is wanted
(including fallback alternatives) is shifted to user, which I think is good

Opinions above are GNU-copylefted.

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]