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

Bug#241231: DECstation 5000/133 (mipsel) install with some problems



On Wed, Mar 31, 2004 at 02:21:49PM +0100, Martin Michlmayr wrote:

> Package: installation-reports

> Debian-installer-version: daily r3k-kn02/boot.img snapshot, 2003-03-30
> uname -a: Linux decstation 2.4.19-r3k-kn02 #1 Tue Dec 16 08:42:55 UTC 2003 mips GNU/Linux
> Method: TFTP via network; CD boot failed
> Machine: DECstation 5000/133
> Processor: R3K
> Memory: 32 MB
> Root Device: SCSI
> Output of lspci: no PCI

> I first tried an installation via TFTP.  I started it like this, but there
> for only those 4 lines of output:
> 
> >>boot 3/tftp/boot.img-r3k
> 4247552+0+4096
> t-rex-loader V0.1 Copyright 2003  Thiemo Seufer <seufer@csv.ica.uni-stuttgart.de>
> Loading ... ok
> This is a DECstation 5000/1xx
> 
> Kernel booted correctly, but didn't use the serial console... so I tried
> again with:
> 
> >>boot 3/tftp/boot.img-r3k console=ttyS2

This is a general problem with Linux/MIPS - the kernel cannot
autodetect the current PROM console setting and always assumes
framebuffer console (or dummy console if no framebuffer is
available) unless booted with console=ttySx.

> I first chose the language, but it didn't ask me for the country but
> instead dropped me into main menu on "Select a keyboard layout".   Maybe
> this was because of 32 MB memory only, but I didn't get a message about
> low-mem installs.
> 
> "Select a keyboard layout": doesn't do anything (I didn't have a
> keyboard conncted, but anyway).

This is the major problem we have on mipsel currently. The LK201
keymaps have been splitted out into a seperate udeb and kbd-chooser
needs to handle this correctly. I had preliminary code for this but
due to several issues it does not work properly. Another thing in
kbd-chooser is that for some reason the keyboard-handling routines
seem to be passed a subarch name (as in -r4k-kn04) instead of the 
keymap name on mipsel. This needs further inspection.

> Hardware detection: it tries to load all kind of IDE modules, but then
> skips it because the modules are not available.  In fact, it also shows me
> a menu and the IDE modules are selected by default?  Why?  Surely this
> shouldn't be selected by default on this machine.  I'm talking about this:

AFAIK IDE modules are tried unconditionally on all architectures.

> partman: it feels very slow, it asks many questions: many windows are
> opened which is slow on a 9600 serial console.  I think it's possible to
> run the DECstation at a higher speed.  How can this be configured and can
> this be documented?  Also, it would be good if partman would not ask that
> many questions and open so many new windows.

In theory, it should be possible to run the Linux serial console at
a higher speed than 9600 baud, but AFAIK the PROM console speed
cannot be changed. As they need to match for practical use (in
particular with kernels >=2.4.24 which contain "early console
support" and use promcalls for output of kernel messages), this
effectively limits the console speed to 9600 baud.

> In partman, I first set up a root partition, swap and home, and one for
> LVM.  When I went to "configure LVM", it formated all the partitions, but
> then went back to the "main menu on "Select a keyboard layout".  When I

Going back to the keyboard layout is probably due to the fact that
kbd-chooser is not yet successfully configured.

> start partman again, I see "Starting up the partitioner" but then it just
> goes back to the main menu.  Ugh, everything lost after so much time.  How
> can I find out what went wrong?

> The install was fine, but no boot loader was installed!  I only got
> "Continue without boot loader".  I don't know why delo-installer was not
> installed at all!

That is strange. I had done a test install some days ago and delo was
properly installed there, but I did not use LVM. Could you retry
without LVM?

> Also, kernel-image-2.4.19-r4k-kn04 was installed instead of the -r3k-kn02
> variant.  From looking at the code, I have no idea why.

I had a similar effect some time ago, see Bug #235260. Interestingly
I got an r3k-kn02 kernel on an r4k-kn04 machine...
On the test install some days ago, the correct kernel was installed
for me; on the other hand that might have been pure luck.

> Also, no ttyS2 entry was added to /etc/inittab

A possible fix for that has been posted on debian-boot in 
<[🔎] 20040328153655.GH17081@p12n.org> but needs testing before it
can be committed.

> During boot, I get:
>     sed: can't read MOTD: No such file or directory
> I have no idea where this is from.

This seems to be caused by this fragment in /etc/init.d/bootmisc.sh:

if [ "$EDITMOTD" != no ]
then
        MOTD="eadlink -f /etc/motd || :"
        if [ "$MOTD" != "" ]
        then
                uname -a > $MOTD.tmp
                sed 1d MOTD >> $MOTD.tmp
                       ^ missing $
 
                mv $MOTD.tmp $MOTD
        fi
fi

[booting from CD]

> >>cnfg 3
>  3: KN02-BA  DEC      V5.7j    TCF0  ( 32 MB)
>                                      (enet: 08-00-2b-34-ee-42)
>                                      (SCSI = 7)
>             ---------------------------------------------------
>             DEV   PID                VID        REV    SCSI DEV
>             ===== ================== ========== ====== ========
>             rz1   RZ29B    (C) DEC   DEC        0016   DIR
>             rz5   CD-ROM CR-506      MATSHITA   8S05   CD-ROM
> 
> So I did this in order to boot from CD, but it simply hangs and does not
> (not even read from the CD afaict; I can mount the CD from Linux, though.)

Does the CDROM support a blocksize of 512 bytes? DECstations cannot boot
from "PC-style" CDROM drives with 2048 bytes/sector. The Linux kernel
does not have this limitation, so once Linux is booted, "PC-style" drives
work without problems.

Regards,
Karsten
-- 
#include <standard_disclaimer>
Nach Paragraph 28 Abs. 3 Bundesdatenschutzgesetz widerspreche ich der Nutzung
oder Uebermittlung meiner Daten fuer Werbezwecke oder fuer die Markt- oder
Meinungsforschung.



Reply to: