Re: Sudden reboot

[Nick Hudson <skrll%netbsd.org@localhost>]

On 12/03/13 07:49, David Holland wrote:
> On Tue, Dec 03, 2013 at 08:13:29AM +0100, Jan Danielsson wrote:
>   >    I'm running netbsd-6 (a month old or so) on a Soekris net6501
>   > (NetBSD/amd64), and tonight I had an unexpected reboot.
>   > /var/log/messages contains:
>   >
>   > Dec  3 03:15:46 aria syslogd[188]: restart
>   > Dec  3 03:15:46 aria /netbsd: uvm_fault(0xffffffff8071b4d8, 0x0, 1) -> e
>   > Dec  3 03:15:46 aria /netbsd: fatal page fault in supervisor mode
>   > Dec  3 03:15:46 aria /netbsd: trap type 6 code 0 rip ffffffff8031162f cs
>   > 8 rflags 10283 cr2  60 cpl 4 rsp fffffe8004cf9b98
>   > Dec  3 03:15:46 aria /netbsd: panic: trap
>   > Dec  3 03:15:46 aria /netbsd: cpu1: Begin traceback...
>   >
>   >    Thinking back, I recall noticing that I have had unexpected reboots
>   > once or twice before, but it's very rare -- been running this system for
>   > a year or so.
>   >
>   >    The entry before syslogd restarting is simply a dhcp client renewal
>   > (againt my ISP), which occured 03:14:04.
>   >
>   >    Trap 6 = illegal op-code? Should I start worrying about broken RAM
>   > modules?
>
> from sys/arch/x86/include/trap.h:
> #define T_PAGEFLT        6      /* page fault */
>
> and %cr2 contains 60, so in all probability it's just a null pointer
> dereference.

Definitely null pointer deref - the second arg to uvm_fault is the 
virtual address.

        uvm_fault(0xffffffff8071b4d8, 0x0, 1) -> e


> If you use objdump -d or nm -n or gdb or whatever to find the code
> that's at 0xffffffff8031162f in your kernel, you'll probably get a
> fairly good idea of what broke.
backtrace is always useful as well.

Nick