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

Bug#248047: Cobalt install

On Wed, May 19, 2004 at 01:40:15AM +0100, Martin Michlmayr wrote:

> > The LCD could be more verbose, i.e. after "loading debian-installer"
> > it could say "connect to the console now".
> I will consider doing this unless we get SSH support.

With ssh, it should display "now telnet to xx.xx.xx.xx", like the
(inofficial?) woody installer

> > >  - support installs via SSH, w/o serial console
> > Definitely. And an option to set the serial baudrate (from CoLo
> > onwards) to other speeds, i.e. 9600 like god intended.
> d-i and colo should just support any baud rate... however, I have no
> idea how to change the Cobalt to something else.  Do you know?

I don't think you can change the baud rate used by the cobalt firmware
-- my idea would have been to let colo and the following steps use a
different speed than 115200, but I've got no idea how to tell colo
about it.

> > The partitioning tool doesn't set the swap partition type on the swap
> > partition (it works, so it's a cosmetic error).
> I've seen this too recently.  I've to check if this is still there.

It does the right thing now.

> > inittab apparently has entries for virtual terminals, which those
> > messages:
> > INIT: Id "2" respawning too fast: disabled for 5 minutes
> > INIT: Id "3" respawning too fast: disabled for 5 minutes
> This should be gone now (in unstable).

Yup, inittab looks good.

installation report revealing a show-stopper:

> Do you perhaps have time to test an installation from unstable?  You
> can use the initrd from
> http://people.debian.org/~tbm/d-i/images/mipsel/daily/cobalt/netboot
> (same kernel as before), and the install unstable rather than testing.

OK; I re-fetched the whole setup from that URL into /nfsroot on
bauer.tahina.priv.at (, which is NFS-exported and has the
TFTP server:

# ls -lR
total 4162
lrwxrwxrwx    1 root     root            4 May 20 17:02 boot -> ./
-rw-rw-r--    1 root     root          213 May 19 02:22 default.colo
-rw-rw-r--    1 root     root      1448550 May 19 02:22 initrd.gz
-rw-rw-r--    1 root     root      2652775 May 19 02:23 vmlinux-2.4.25-r5k-cobalt
-rw-rw-r--    1 root     root        44836 May 19 02:23 vmlinux.gz
-rw-rw-r--    1 root     root        44836 May 19 02:23 vmlinux_RAQ.gz
-rw-rw-r--    1 root     root        44836 May 19 02:23 vmlinux_raq-2800.gz

dhcpd.conf (2.0pl5-11):

host remailer {
    hardware ethernet 00:10:E0:00:30:F5;
    fixed-address remailer.tahina.priv.at;
    option routers;
    option root-path "/nfsroot";
    filename "/boot/default.colo";

Boots fine, menu comes up on LCD, select "TFTP", installer boots.
Interesting points:

- when the installer comes up on the serial console, the LCD
  still shows "loading debian-installer".

- it asks for an IP address and other network config data -- it should
  just use the data it already has gotten via DHCP (ISTR the woody
  netinst does just that)

- it mis-configures the IP address -- I entered
  "", but it ifconfig's Yes, I tried that
  twice to make sure I didn't mistype the address :-) 

~ # ifconfig
eth0      Link encap:Ethernet  HWaddr 00:10:E0:00:30:F5  
          inet addr:  Bcast:  Mask:
          RX packets:55 errors:1 dropped:0 overruns:0 frame:0
          TX packets:22 errors:4 dropped:0 overruns:0 carrier:4
          collisions:0 txqueuelen:1000 
          RX bytes:4221 (4.1 KiB)  TX bytes:968 (968.0 B)

  I went thru the process of configuring debian mirror and proxy,
  "execute a shell" afterwards and ifconfig manually with the correct
  values; exit from the shell, choose "Choose a mirror...", continue.

- The first time when asking for a debian mirror, it didn't ask if I
  wanted http or ftp; the second time around, it does. Also, it
  doesn't remember the country I selected first time around.

- it does configure /etc/resolv.conf with the nameservers I entered,
  but it doesn't add "search domain.name" or "domain domain.name";
  therefore I need to enter the FQDN of the proxy.

- select "unstable"

- in the "load installer components" screen, I selected "continue",
  all of the offered modules were unselected by default and left that

- partitioning: the disk already had partitions and file systems from
  the last test; choose "partitioning method: erase entire disk".

- partitioning scheme: "all files in one partition"; looks OK:

  |           IDE1 master (hda) - 20.4 GB ST320423A             #           |   
  |           >      #1 primary   98.6 MB B f ext2r0     /boot  #           |   
  |           >      #2 primary   19.8 GB   f ext3       /      #           |   
  |           >      #5 logical  518.1 MB   f swap       swap   #           |   

[many hours later... boy, is that box sloooow!]

- reboot works well, machine comes up fine.

- in base-config, there's no visual indication of the active button --
  both look exactely like this (i.e. no bold or inverse or stuff,
  cursor is somewhere else):

 . Is the hardware clock set to GMT?                                         .  
 .                                                                           .  
 .                    <Yes>                       <No>                       .  

- selected list items _are_ visible, in bold.

- the timezone selector starts in the US submenu, instead of on the
  top level.

- line editing characters (^U, ^A, ...) don't work in text entry fields.

- base-config asks for the system hostname; it should take the value
  entered during the network configuration at the start of the
  install; apt-source configuration is the same, I already have
  entered the data before, it starts with default settings, except the

- network config is broken, it doesn't even send out ARP packets when
  trying to apt-get update; I select "edit sources.list by hand" and
  leave it empty to get ahead.

  after logging in, neither eth0 nor lo are configured/up,
  /etc/network/interfaces has _only_ the following content:

--- cut ---
# Used by ifup(8) and ifdown(8). See the interfaces(5) manpage or
# /usr/share/doc/ifupdown/examples for more information.
--- cut ---

I'll be mighty busy the next two weeks, and maybe I'll not have a RAQ
available afterwards... maybe I could do another test install next



Wildcard NS records, while theoretically possible, have sufficiently
bizarre semantics that it is probably best to limit their use to
torture-tests of DNS software.
-- IAB Commentary: Architectural Concerns on the use of DNS Wildcards

Reply to: