On Dec 15, 2004, at 11:32 PM, Dave Huang wrote:

> On Tue, Dec 07, 2004 at 04:16:40PM -0800, Matt Thomas wrote:
>> At 04:16 PM 12/7/2004, Dave Huang wrote:
>>> Which problem are we talking about? With a 2.0 kernel+rev 1.26 of
>>> locore_subr.S, I no longer get "panic: mp_save_fpu_proc", but I do 
>>> get
>>> "panic: pmap_pvo_enter: called recursively!"
>>
>> 1.26 of pmap.c will fixed that panic from pmap.c
>
> OK, I got 2.0 installed on all the machines I wanted running 2.0, so I
> upgraded my source tree to -current and built a new macppc kernel.
> It's looking pretty good so far--no panics, at least, so thanks to you
> and Tim :)
>
> However, programs still occasionally crash with SIGILL (which I think
> I've always seen with a MP kernel). For example, doing a build.sh -j3
> distribution will almost always crash at some point during the "make
> obj" stage, leaving either a sh.core or nbmake.core lying around in
> whichever directory it happened to be working in. As usual, building
> without -j seems to work fine.

Can you do 'sysctl -w machdep.printfataltraps=1'?
I'd be interested in knowing where these crash.  Can you look at the
cores?

-- 
Matt Thomas                     email: matt@3am-software.com
3am Software Foundry              www: http://3am-software.com/bio/matt/
Cupertino, CA              disclaimer: I avow all knowledge of this 
message.