Mailing List Archive

Help booting 2005.0 cdrom on headless sparc
I've got a headless Ultra Enterprise 2 that I'm trying to install Gentoo on.

There's no video card nor keyboard attached.

The serial console is on ttyb, because ttya is broken.

I get to eeprom just fine, and can boot to the silo prompt.

The system boots through single-user, then starts L3, and hangs.

INIT: Entering runlevel: 3 [ ok ]
* Starting syslog-ng... [ ok ]
* Hardware detection started... * ... [ ok ]
* Not Loading APM Bios support.... [ ok ]
* Coldplugging input devices... [ ok ]
* Coldplugging isapnp devices... [ ok ]
* Coldplugging pci devices...** can't synthesize pci hotplug events [ ok ]
* Coldplugging pnp devices... [ ok ]
* Coldplugging usb devices... [ ok ]
* Network device eth0 detected, DHCP broadcasting for IP.... [ ok ]
* Auto-scrambling root password for security... [ ok ]
* Starting local... [ ok ]


Well, it's not hanging, I'm sure, just not displaying a login prompt,
because it's trying to put one on ttyS0 -- which is broken.


How do I force login on ttyS1 ? I've tried using the console=ttyS1,9600n8
but that stops printing at "mem_init: calling free_all_bootmem" and just
'hangs' completely.

Help!
--
gentoo-sparc@gentoo.org mailing list
Re: Help booting 2005.0 cdrom on headless sparc [ In reply to ]
Leif Sawyer wrote:

> Well, it's not hanging, I'm sure, just not displaying a login prompt,
> because it's trying to put one on ttyS0 -- which is broken.
>
>
> How do I force login on ttyS1 ? I've tried using the console=ttyS1,9600n8
> but that stops printing at "mem_init: calling free_all_bootmem" and just
> 'hangs' completely.
>
> Help!

The LiveCD logic assumes that if you have a serial console it'll be on
ttya (ttyS0). It's not made to deal with broken machines or alternate
weird setups.
You'll either have to remake the livecd* if you want to change that or
make it a proud head-bearer.

(*) Rebuilding the livecd requires catalyst running on a sparc64 machine.
You can avoid this by doing some magic by looking at the scripts and
being able to modify the startup script involved in setting up the
serial console by extracting it from the squashfs loop, then rebuilding
said loop filesystem, and rebuilding the ISO with the specially crafted
mkisofs with the corresponding silo 1.2.6 version because newer silo
versions have broken isofs code that doesn't work on some machines.
In other words, it's easier to fix the machine somehow...

--
Gustavo Zacarias
Gentoo/SPARC monkey
--
gentoo-sparc@gentoo.org mailing list
Re: Help booting 2005.0 cdrom on headless sparc [ In reply to ]
Leif Sawyer wrote:
> I've got a headless Ultra Enterprise 2 that I'm trying to install Gentoo on.
...
> Well, it's not hanging, I'm sure, just not displaying a login prompt,
> because it's trying to put one on ttyS0 -- which is broken.

The problem is the kernel is booting up on ttyS1 fine, but on the livecd init is
not configured to start a getty process attached to ttyS1. That's why you see
the boot output, but no login prompt.

The only option I can think of trying is to swap the ttya ttyb prom values with
the devalias command. I'm not sure how the serial driver is setup, but if you're
lucky it may set ttyS0 to the node value of ttya (since the author of that
driver posts here he may have better insight into this). Just remember to write
down the original values if you have to switch back. ;)

Hardave
--
gentoo-sparc@gentoo.org mailing list
Re: Help booting 2005.0 cdrom on headless sparc [ In reply to ]
On Thu, 21 Apr 2005 18:07:17 -0300
Gustavo Zacarias <gustavoz@gentoo.org> wrote:

> The LiveCD logic assumes that if you have a serial console it'll be on
> ttya (ttyS0).

There is no reason to hardcode this, the exact serial device used
is stored in the 'output-device' prom environment variable, and
that is how the serial layer figures out which serial line to use.
--
gentoo-sparc@gentoo.org mailing list
Re: Help booting 2005.0 cdrom on headless sparc [ In reply to ]
David S. Miller wrote:

>>The LiveCD logic assumes that if you have a serial console it'll be on
>>ttya (ttyS0).
>
>
> There is no reason to hardcode this, the exact serial device used
> is stored in the 'output-device' prom environment variable, and
> that is how the serial layer figures out which serial line to use.

I did the script some time ago in a almost arch-agnostic fashion to
cover up some other architectures.
I'll probably revisit this for the next livecd release, given that other
ARCHes help in their turf or gimme free kit to play with ;-)

--
Gustavo Zacarias
Gentoo/SPARC monkey
--
gentoo-sparc@gentoo.org mailing list