remi_zara@mac.com wrote:

> >> I run -current (NetBSD qube2.remi-zara.org 4.99.17 NetBSD 4.99.17
> >> (GENERIC) #0: Thu Apr 19 22:36:19 PDT 2007  builds@wb37:/home/builds/
> >> ab/HEAD/cobalt/200704200002Z-obj/home/builds/ab/HEAD/src/sys/arch/
> >> cobalt/compile/GENERIC cobalt) from the daily builds.
> >
> >> May 27 12:22:42 qube2 /netbsd: trap: TLB miss (load or instr. fetch)
> >> in kernel mode
> >> May 27 12:22:42 qube2 /netbsd: status=0x3, cause=0x8, epc=0x801d3fdc,
> >> vaddr=0x7ff96de2
> >> May 27 12:22:42 qube2 /netbsd: pid=23154 cmd=netstat usp=0x7fffd4a0
> >> ksp=0xcc72bc28
> >
> > Could you put your kernel on somewhere?
> > (it's no longer available on releng server)
> 
> I've uploaded it here:
> 
> http://remi.zara.free.fr/netbsd/netbsd
> 
> > Or could you check in which function the panic happened?
> > (by checking address 0x801d3fdc in outputs of
> >  "nm /netbsd | sort" or "objdump --disassemble /netbsd" etc.)
> 
> 801d3a84 T ktrgenio
> 801d3cc4 T ksyms_getname
> 801d409c T ksyms_getval
> 
> What I'm not sure to understand is how these messages get there. They  
> appear on reboot (and do not prevent it), which mean after a powercycle.
> I'm not sure if they are related to the "crash" of the machine.

Hmm, it looks your kernel got TLB miss panic during trace after
the first panic, so the address didn't point the real problem.

I've just create a kernel with fixed trace command and 20070420
source tree, so could you try this one?
http://www.ceres.dti.ne.jp/~tsutsui/netbsd/netbsd-cobalt-tracefix-20070420.gz

Of cource, it's much better to get real panic message and trace output
on serial console..
---
Izumi Tsutsui