Re: HPET (?) related hangs and breakage in 2.6.35,36

From: Andrew Lutomirski
Date: Tue Nov 09 2010 - 16:30:38 EST


On Tue, Nov 9, 2010 at 1:01 PM, Borislav Petkov <bp@xxxxxxxxx> wrote:
> On Tue, Nov 09, 2010 at 12:44:48PM -0500, Andrew Lutomirski wrote:
>> I may have just reproduced after a couple of days of uptime it in
>> 2.6.36 with hpet=disable (in any case, the system hung and netconsole
>> didn't work).
>>
>> I'm currently testing the hpet_min_tick stuff, backported to 2.6.36,
>> with hpet on.  Haven't had a problem yet, but I'll see what happens
>> after a few hours.  So far, it said:
>>
>> ACPI: HPET id: 0x8086a301 base: 0xfed00000 min tick: 128
>
> This doesn't say what your BIOS has set it to originally. For that, do
>
> hexdump -C /sys/firmware/acpi/tables/HPET
>
> and check what does the u16 little endian value at offset 0x35 say.

0x37ee, which people have seen before. In any case, even with the
patch applied my system hung, and I still got this off netconsole
sometime around when the hang happened:

Clocksource: tsc unstable (delta = -34355296774 ns)
Switching: to clocksource hpet

Another odd symptom: none of the SysRq hotkeys I could think of did
anything *except for SysRq-b*. I have no idea what could cause that.

My system appears to pass ingo's time-warp-test (when compiled with
-O0 -- it segfaults with -O2), so I think my TSC is okay.

--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/