On Monday 20 December 2004 15:04, Sven Luther wrote:
> On Mon, Dec 20, 2004 at 02:51:23PM +0100, Philippe Guyot wrote:
> > On Monday 20 December 2004 13:56, Sven Luther wrote:
> > > On Fri, Dec 17, 2004 at 05:15:22PM +0100, Philippe Guyot wrote:
> > > > Hi all!
> > > >
> > > > I am trying to install a Debian on a powerpc 604 machine from Bull
> > >
> > > This is probably a prep box, do you know if it uses the prep or ppcbug
> > > boot mechanism ?
> > >
> > > > It has firmware 03.03.09 copyright Firmworks.
> > > >
> > > > If it doesn't lie, it is a prep machine.
> > >
> > > Cool.
> > >
> > > > I made all the stuff for netbooting (dhcd,tftp and so on).Seems OK.
> > > >
> > > > I am loading vmlinuz-prep.initrd ("load net") and the firmware
> > > > answer:
> > > >
> > > > ### : 0 d e boot-file = vmlinuz-prep.initrd
> > >
> > > Is this the rc2 or daily build, 2.4 or 2.6 d-i kernel ?
> >
> > /ftp.debian.org/debian/dists/testing/main/installer-powerpc/rc2/images/po
> >werpc/netboot/
>
> Ok, so rc2, 2.6 netboot image.
>
> Can you try also the netboot/2.4 image ?
I tried, same result....
>
> > > > I examined the memory, the image seems well loaded at load-base
> > > > (comparing hex display of memory at load-base and at load-base+size
> > > > of vm...with the output of hexdump vm....
> > > > (only first and last blocs of memory, of course)
> > > >
> > > > then I typed "go" to start execution. firmware answer : ### : 0 d f
> > > > and the machine is frozen.
> > > >
> > > > (of course, "boot net" gives the same result).
> > > >
> > > >
> > > > Can somebody point me to a right direction ?
> > >
> > > Not much, but i suppose that it may be that you run into some kernel
> > > size limit. Where does load-base point to ?
> >
> > load-base = 0x 400000
>
> This is 4MB, right ? The 2.6 kernel is smaller, but the intird used is big.
> Does the whole d-i image fit in the 6MB between load-base and OF ?
In fact, I don't know where OF is exactly (real-base = -1 and I can't find the
default value for OF 3). The size problems I red about exists with OF 1 or 2
which have load-base 100000 and real-base 400000 by default. Truly a kernel
greater than 3 MB is a problem in this case but remenber we are in OF 3 which
should be able to deal with large kernels.
I am searching about this, but really the very following blocks of memory at
[load-base + size of vmlinuz...] are only zeroes. So I don't think OF is
overwritten. Doing only "load" and you can do whatever you want whith OF
commands.
>
> > I made some experiences, not always pure orthodoxy:
> >
> > First of all, when booting disk ("boot disk"), the OF displays 0 d f then
> > boots proceed, so I think this 0df is not an error message.
> >
> > As far as I can see by dumping memory from place to place, the vmlinuz
> > loaded does not overwrite parts of code from OF (which should be at 0x
> > 100000). There were only zeroes before loading.
>
> Mmm, 0x100000 is 1MB, right ? Thisindeed cause problems.
>
> > So I tried to load...yaboot. (yes, I know, not the right way for a prep
> > machine but...)
>
> Cool thing to try :)
>
> > It gave strange things: It runs but you can enter only 2 chars at the
> > prompt. I realize also (tcpdump) that when loaded and running, it request
> > "yaboot" to the ftp server and not "yaboot.conf". I went around this by
> > loading first
>
> Known bug of the IBM OF implementation.
OK, I don't mind at the moment.
>
> > yaboot, then moving a yaboot..conf on yaboot then "go" . Same result
> > however (may be because this is *not* a chrp machine) but the fact is
> > that yaboot is loaded and running. May be some clue arising from that? I
> > red from IEEE standards that .ELF must be annotated for running on IBM
> > machines (but saying the truth, I do not understand yet what doest that
> > mean...).
>
> You have to run addnote on them, but this should be done for the yaboot on
> the rc2 CD.
I agree, seen by dumping the 1rst sector of chrp ELF image.
BTW netboot prep image is not an ELF ??
Friendly.
Felipe
