hi,

On Sat, Oct 22, 2005 at 08:18:07PM +0100, Patrick Welche wrote:
> The info says - it's named, so not what you have been seeing:
> 
> Oct 22 19:23:02 quartz syslogd: Exiting on signal 15
> uvm_fault(0xc0537b80, 0xdeadb000, 0, 1) -> 0xe
> kernel: supervisor trap page fault, code=0
> Stopped in pid 317.1 (named) at 0xdeadbeef:     invalid address
> db{1}> bt
> acpi_softc(cdad564c,c32bf800,cdb13f9c,c02d5d8b,0) at 0xdeadbeef
> sa_switchcall(cdad564c,2b,2b,2b,2b) at netbsd:sa_switchcall+0x44
> db{1}> sync
> syncing disks... panic: TLB IPI rendezvous failed (mask 1)
> Stopped in pid 317.1 (named) at netbsd:breakpoint+0x4:  leave
> db{1}> sync
> 
> dump to dev 4,1 not possible
> rebooting...
> 
> This happened with today's cvs, while shutting down.  I remember
> restarting named when I had the freeze mentioned in the original
> email, but also had a build going at the same time, so it wasn't
> obviously the named restart...

ok, that does look like a bug in the sa change.  we're jumping through
a function pointer that has 0xdeadbeef as the value.  most likely the
sau is being freed before we try to use it, but I don't see where.

next time you see this, try "reboot 0x104" instead of "sync",
that's more likely to succeed in getting a dump.

...on second thought, the "not possible" message is because either
no dump device is configured or the device is too small to hold a dump.

-Chuck