Install hangs on SILO using a LDom on a T5220

[O_Possum]

Binary package hint: debian-installer

Hello.

I'm trying to install Ubuntu on a LDom on a Sun Enterprise T5220.

I'm using the tftp boot image founded at ftp://ftp.ubuntu.com/ubuntu/dists/hardy/main/installer-sparc/current/images/sparc64/netboot/2.6/boot.img because it's impossible to use the CD-ROM for a LDom.

Everything is OK until the SILO installation. It's seems to wait for a user input according to logs (Attached)

The hard disk is a 10 GB image made with mkfile, in a ZFS container, exported to LDom as usual. (ldm add-vdsdev path-to-img vol-name@primary-vds0 and ldm add-vdisk vdisk0 vol-name@primary-vds0 my-ldom)

I'm connected to my ldom telneting it from the primary where a SSH is opened from my box. Once in the OBP, I use boot vnet0 debconf/priority=low DEBIAN_FRONTEND=text BOOT_DEBUG=2 to initiate installation.

The whole Syslog file is joined to this bug report. Hope it coul'd help.

I'm using hardy but same problem with gusty.

PS: forgive my bad English ;)

[Piero Ottuzzi (ottuzzi)]

Hi there,

I'm seeing the same problem here with Ubuntu Hardy Heron ALPHA-5 using the netboot image found in [1] on a old SUN Ultra5.
Please ask for any other useful info I can provide.

Bye
Piero

[1]http://archive.ubuntu.com/ubuntu/dists/hardy/main/installer-sparc/current/images/sparc64/netboot/2.6/boot.img

[Piero Ottuzzi (ottuzzi)]

Here is the workaround:

- go to console 2 (ALT-F2)
- press ENTER
- /target/usr/sbin/siloconfig -r /target
- return to installation screen (ALT-F1)
- press CTRL-C
- continue without installing the bootloader
- finish and reboot

Bye
Piero

[O_Possum (frederic-passerin)]

This workaround can't work in my case as I telnet the LDom.

And, If I do a [CTRL-ALT-F2], I'll be on the second tty of my machine and [ALT-F2] opens the run-command dialog of KDE ;)

[Piero Ottuzzi (ottuzzi)]

Hi there,

my workaround leaves the system bootable but unusable after reboot: it cannot start any service saying
Warning: Fake start-stop-daemon called, doing nothing.

Bye
Piero

[Colin Watson (cjwatson)]

Looks like silo-installer needs to write out /boot/silo.conf before installing the silo package into /target. At a guess, anyway.

[Piero Ottuzzi (ottuzzi)]

Hi again,

just to report that with the image from 03-Mar-2008 13:04 the problem still exist.
I also found another problem during installation... going to open another bug and report here the new bug id.

Bye
Piero

[Piero Ottuzzi (ottuzzi)]

I opened a new bug for the SPARC netboot installer as previously reported: bug 198307.

Bye
Piero

[Jeff Balderson (jbalders)]

Since my report (bug 203947) has been marked a duplicate of this one, I thought I'd chime in:

I netbooted, chrooted into my mostly installed system and performed the following:

apt-get --purge remove silo
then:
root@suntest:/# apt-get install silo
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following NEW packages will be installed:
  silo
0 upgraded, 1 newly installed, 0 to remove and 0 not upgraded.
Need to get 0B/179kB of archives.
After this operation, 393kB of additional disk space will be used.
Selecting previously deselected package silo.
(Reading database ... 15512 files and directories currently installed.)
Unpacking silo (from .../silo_1.4.13a+git20070930-1ubuntu1_sparc.deb) ...
Setting up silo (1.4.13a+git20070930-1ubuntu1) ...
SILO, the Sparc Improved LOader, sets up your system to boot Linux
directly from your hard disk, without the need for a boot floppy or a net
boot.
Install a partition boot record to boot Linux from UUID=3cb47072-33de-49fd-aed6-d77ccea87723? [Yes]
Creating small silo.conf and running silo.
Config file error: Syntax error near line 1 in file /etc/silo.conf
Error parsing silo config file /etc/silo.conf

root@suntest:/# cat /etc/silo.conf
partition=87723
root=UUID=3cb47072-33de-49fd-aed6-d77ccea87723
timeout=20

image=/vmlinuz
        label=Linux
        read-only

image=/vmlinuz.old
        label=LinuxOLD
        read-only

Note that SILO counts the lines in the config file starting at 0, so it looks like the problems with the auto-generated file are:

1) the "root=" line uses a UUID, and SILO doesn't understand UUIDs
2) vmlinuz lives in /boot/ if you don't have a separate /boot partition. So the correct path in this case would be /boot/vmlinuz

If I again netboot/chroot, edit /etc/silo.conf, and change it to :
partition=1
root=/dev/md0
timeout=20

image=/boot/vmlinuz
        label=Linux
        read-only

image=/boot/vmlinuz.old
        label=LinuxOLD
        read-only

My system will then boot.

[Jeff Balderson (jbalders)]

I noticed an omission in that last bit (should have waited to cut-n-paste instead of manually making the changes):

partition=1
root=/dev/md0
timeout=20

image=/boot/vmlinuz
        label=Linux
        read-only
        initrd=/boot/initrd.img

image=/boot/vmlinuz.old
        label=LinuxOLD
        read-only
        initrd=/boot/initrd.img.old

[Piero Ottuzzi (ottuzzi)]

Any news on this topic?
We are few weeks from the final release and I'm trying new net-installers as they appear but none of them worked.
Please let me know if you need any help testing new images... I'll be very glad to help you.

Bye
Piero

[Piero Ottuzzi (ottuzzi)]

As anyone can read on [1]:

The officially released architectures for Ubuntu 8.04 will be i386 and amd64. The SPARC port will continue to be provided with build infrastructure, and Ubuntu 6.06 LTS, 7.04 and 7.10 will continue to enable SPARC deployments well into the future, but there will not be an official Ubuntu 8.04 release for SPARC.

I think this bug will never be fixed :(

Bye
Piero
[1]https://lists.ubuntu.com/archives/ubuntu-devel-announce/2008-March/000400.html

[Claudio Luck (cluck)]

The problem sits in the postinst script, which calls siloconfig.

siloconfig is interactive and asks questions in the background:

 - Install a boot block using your current SILO configuration?
 - Wipe out your old SILO configuration and make a new one?
 - Install a partition boot record to boot Linux from $device?

The right way to fix it would be to add command line options to siloconfig and wire them up to debconf.

The attached patch simply pipes in `yes' to siloconfig in the postinst to make it work. As I understand it, it is safe to say yes to all three questions above, even in multi-boot environments.

[Claudio Luck (cluck)]

This bug was fixed in the package silo - 1.4.14-0ubuntu3

---------------
silo (1.4.14-0ubuntu3) karmic; urgency=low

  * Disable siloconfig from ever running automatically
    - siloconfig doesn't properly handle UUIDs, and its default configuration
      doesn't set important things like initrd. In almost all cases it will be
      a silo -f, which we can simply do ourselves. We'll leave it in case the user
      ever needs it. This also puts silo more in line with the current behavior of
      lilo. (LP: #431106)

 -- Michael Casadevall <email address hidden> Wed, 16 Sep 2009 23:37:30 -0400