fdisk and native partition formats (was: Re: Problems with SILO (Sparc slink))
Edmund GRIMLEY EVANS wrote:
> Eric Delaunay wrote:
> > Hmm, this thread is too repetitive. I propose to fix the installer to enfore a
> > Sun disklabel on the disk.
> > I guess it could be fixed in either 2 ways:
> > - add some obscure option to fdisk telling it to treat non Sun disklabel the
> > same way as blank disk
> > - or check for a Sun disklabel before calling fdisk, zeroes the partition table
> > if not found, then call fdisk.
> I don't like either of these, because it is possible that someone
> really does want to keep the old partition table. People use
> installation media for all sorts of things. The other day I saw a
> colleague here rescuing an NT machine with a broken partition table by
> using a Red Hat installation CD for example ...
> I think:
> - the 's' option should always be visible in fdisk
In fact it is already documented in potato fdisk menu.
> - dbootstrap should check for a Sun disk label after calling fdisk
> and issue a warning if it doesn't find one, something like:
> "/dev/whatever appears not to have a Sun disk label. This means you
> will not be able to boot directly from this disk. You can create a Sun
> disk label with the option 's' in fdisk." with options to continue or
> run fdisk again. (But I can't remember at the moment how these
> messages are usually phrased; I'd have to look at the context again ...)
I like this idea more than my proposal.
However, I'd also want to check for valid Sun disk label before running fdisk,
then issuing the warning here. This will give the user more informations on
how to proceed instead of doing wrong the first time then forced to partition
Of course, if the disk does not contains a disk label at all or one which is
already of Sun type, this message will not be displayed and fdisk will be run
It will be checked again after fdisk has existed, like you describe, to be sure
the user is knowing what it does.
Well, I would implement it asap.
Btw, how other arches (alpha, ppc, m68k, ...) are supposed to work with non
native partition formats ? Is there a need for such a scheme for them ?
Or is it just an artifact of sparc fdisk that also support i386 partitions?
Eric Delaunay | "La guerre justifie l'existence des militaires.
email@example.com | En les supprimant." Henri Jeanson (1900-1970)