system stalls when wallclock/TOD nudged

[Adrian Chadd]

On 20 November 2010 16:14, Colin Percival <cperciva at freebsd.org> wrote:
> On 11/19/10 23:40, Adrian Chadd wrote:
>> I bet if you did some slicing of -head versions, you'll find an
>> earlier version that doesn't do that. :-)
>
> Sure, all I need to do is go back a couple of days, and I find a FreeBSD
> Xen kernel which doesn't do this... because it panics before the clock
> gets nudged. :-)
>

Well, try a -head kernel from 6 months ago but patch with your recent
changes? :-)


Adrian

>> On 20 November 2010 15:38, Colin Percival <cperciva at freebsd.org> wrote:
>>> It seems that this bug is related somehow to the idle thread sleeping -- when
>>> I comment out the call to idle_block from cpu_idle_hlt in i386/machdep.c the
>>> stalling goes away.
>>>
>>>> On Fri, 2010-11-19 at 06:42 -0800, Colin Percival wrote:
>>>>> During the boot process, in the middle of running rc.d scripts, my FreeBSD/Xen
>>>>> instance stalls.  It does nothing until I send a shutdown signal, at which point
>>>>>> [XEN] hypervisor wallclock nudged; nudging TOD.
>>>>> and a couple buffered lines of rc.d script output are printed.  From there the
>>>>> rc.shutdown script starts running, but it too stalls after a short time; then I
>>>>> see
>>>>>> [XEN] hypervisor wallclock nudged; nudging TOD.
>>>>>> Nov 19 14:08:33 init: timeout expired for /bin/sh on /etc/rc.shutdown: Interrupted system call; going to single user mode
>>>>> printed, suggesting that the shutdown watchdog timer at least managed to not get
>>>>> stuck even though everything else did.
>
> --
> Colin Percival
> Security Officer, FreeBSD | freebsd.org | The power to serve
> Founder / author, Tarsnap | tarsnap.com | Online backups for the truly paranoid
>