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

Re: Some comments on 2001-05-17 boot-floppies



On Mon, May 21, 2001 at 03:11:07PM -0400, David Huggins-Daines wrote:
>   1) I used the 'idepci' flavour, hoping it would have a good set of
>      network drivers in its kernel so I wouldn't have to install from
>      floppies.  Unfortunately it lacks a driver for the rtl8139.
>      These cards are so incredibly common nowadays that it might be a
>      good idea to include this one.

I guess you'd need to ask the idepci kernel-image maintainer about
this.  No idea who that is, though.

>      Of course, a better idea would be allowing the user to load an
>      ethernet module of their choice from the drivers disk, but
>      knowing dbootstrap, I'm sure that feature would be rather hard to
>      implement.  I suppose I might potentially be able to install the
>      driver disks, run modconf, and then go back, but that's not
>      exactly obvious.

If a module is provided for your network card, you should only
need disks for the root, rescue, and driver(s) floppies.  After
loading the module for your NIC, you should be able to configure
the network and download everything else via it.  Where exactly do
you need to go back?

>   2) The "Select Partition" screen is confused about what partitions
>      actually exist on my system.  I have Linux partitions on
>      /dev/hda1 and /dev/hda6-/dev/hda10 (in /dev/hda3) and a FreeBSD
>      slice (containing, um, five partitions, I think) in /dev/hda2.
>      However, the listing on that screen shows Linux partitions in
>      /dev/hda1 and /dev/hda10-/dev/hda15.
> 
>      This confused me for a while until I chose "View Partition Table"
>      and realized that libfdisk had put the BSD partitions *before*
>      the extended partitions in the numbering, whereas the kernel had
>      put them *after*.  As it turns out, the kernel makes two passes
>      over the partition table - the first one only expands genuine
>      MS-DOG extended partitions, and the second one expands
>      BSD/Solaris/whatever disklabel-in-partition schemes.  Thus it is
>      guaranteed that the partitions are numbered as such:
> 
>         1 to 4             primary partitions
>         5 to N             logical partitions
>         (N+1) to infinity  BSD partitions .. and beyond!
> 
>      This is actually a pretty serious bug which might cause people to
>      nuke their BSD partitions, but it shouldn't be hard to fix.
>      Bother me for a patch later if you can't figure it out.

Indeed, unlike the sun partitions which are renumbered depending
on the kernel version, BSD disklabel partitions appear to have
always come after other partition flavors.  I just committed a
patch to this effect to CVS.

Thanks for testing the boot-floppies.

Matt



Reply to: