On Thu 30 Nov, Ben Harris wrote:
> On Thu, 30 Nov 2000, Kjetil B. Thomassen wrote:
> 
> > A couple of days ago, I sent this email to Ben, but so far I have not
> > received a reply to it.
> 
> *grin*  Don't expect fast responses from me -- I've got a real life, and a
> show in Birmingham to prepare for...

In my previous reply to this question, I forgot to ask you if there is
anything I can do, so hereby that question has been asked.

> > I decided to try to compile my own kernel on my SPARCstation
> > 2 running NetBSD/sparc, and this worked, but the kernel I made did not
> > get very far. Now I am building a new FOURMEG kernel to see if that
> > works.
> > 
> > The sources I am building from are from yesterday, so this will also be
> > a test of whether Ben's patches to the PCF8583 probing works on my R140.
> 
> Be careful: the UBC integration broke things rather.  You'll need
> sys/arch/arm26/include/vmparam.h to be at least revision 1.2 to get beyond
> the copyright notice.

OK, I made it past that (with revision 1.2), but it panics in ubc_init
from main from start. The error message is ubc_init failed to map
ubc_object.

So, it seems like the UBC integration definitely broke things. I am now
fetching the sources from 05 October to see if I can get any further.
Should I send-pr this problem?

And, BTW, the PCF8583 (RTC chip) reported the frequency this time, so
that seems like it has been fixed.

I am definitely building arm26 kernels now on my ss2. The make depend
takes around 15 minutes from scratch, and the compile a bit less than an
hour, also from scratch.

> > nfs_boot: gateway=192.168.169.168
> > root on 192.168.169.180:/home/export/r140
> > panic: Reset handler called. Branch through zero?
> > system halted
> > 
> > Any ideas why?
> 
> I don't know.  I don't think I've ever seen that panic happen before.
> 
> Do you know if it's started init by then?  If you haven't got DDB built in
> to the kernel, build it in, then you'll get dropped into it at startup and
> can type "t" to get a stack backtrace (and "ps" to get a process listing).  
> I think it should correctly trace back through the trap.

As previously announced, I was using the FOURMEG kernel you made without
ddb. But, I'm fetching the sources from 05 Oct now and will make a
kernel with ddb and see what I can find out.

And, BTW, if anyone wants any custom kernels made, I hereby offer my
services. My ss2 isn't doing much else at the moment, so it may as well
be compiling kernels for people.

The fetching is still progressing, so I probably won't get any further
tonight. I will keep you posted.

Kjetil B.
mailto:kjetil@thomassen.priv.no
http://www.thomassen.priv.no/