Hi, its me again,

Sorry, but I went dark for a while trying (vainly) to set up my Qube
2700
with 1.6.  I've not had luck with Dennis C.'s cool BSD recovery CD (even
the one with he made for me with HD DMA disabled) and I was equally
unlucky
in genning up an HD from a surrogate i386 box then transplanting it to
the
Qube.

I just happened across a posting in the Debian MIPs group talking about
these pesky headless Qubes and it mentioned the issue of a lack of a
serial
module actually causing the kernel to panic. 

snip ...
The kernel must not (if I remember correctly) access the serial console.
As
this chip is missing on the 2700 platform, the server crashes when it
tries
to access the serial port (and the serial console). That's why it dies
right after the "Starting up" message. It loads the kernel and when the
first char is written out the serial port, it dies. I'd try
CONFIG_COBALT_SERIAL=n and CONFIG_SERIAL_PCI=n (on the 2.0 tree, it
might
be different on the CVS linux/mips tree. Haven't checked).
....

I searched the cobalt archives here and didn't see any mention of this
but I
was wondering if anybody could corroborate this behavior under NetBSD? 
I
notice that the config file does include a console option, but would the
kernel panic, or just not use that code?

Also, per a previous posting from S. Jacobsen re PUC(4) has anybody just
plugged a serial card into the spare PCI slot to see if this magically
becomes the console port?

I'd appreciate any thoughts or experiences pertaining to getting this OS
running on the 2700.

Regardless, I'll try a PCI serial card and tear apart both qube and i386
boxes to try out a "comless"
kernel to see what happens.

Thanks for any advice


Michael Francini <francini@mindspring.com>