On Wed, Jan 12, 2005 at 07:39:52AM +0100, Gert Doering wrote:
> 
> Machine crashed at 05:24 tonight, right in the middle of an amanda dump
> (so it was not "idle" today while crashing).
> 
> Console showed the following messages:
> --------------------------------------------------------------------
> data fault: pc=11b0684 addr=0
> kernel fault 30: data access exception
> Stopped in pid 7266,1 (dump) at netbsd:unsleep:0x34:    ldx  [%g1 + %g0], %g1
> db>

This appears to be a NULL pointer dereference.  

> db> bt
> kpsignal2(c8c64e0, cbcfc50, 1, 0, 0, 0) at netbsd:kpsignal2+0x334
> sys_kill(c8c6270, 1f, cbcfdc0, 0, cbcfdd0, 40530ff8) at netbsd:sys_kill+0x114
> syscall(cbcfed0, 25, 40531304, cbcfdd0, 40531304, 40531308) at netbsd:syscall+0x
> d4
> ?(2f6d, 1f, 0, 20e000, 2, 21fc00) at 0x1008cb8

It looks like some process was sending a signal to another process.

> so, what to try next?

I would recommend debugging the problem.

1) Dump the registers.  Make sure %g1 is zero.

2) Disassemble unsleep and see where the contents of %g1 came from.
Make sure that is also zero.

Or you could try disabling the data cache and see if that solves your problem.

Eduardo