Messages in this thread | | | Date | Thu, 29 Nov 2007 12:45:48 -0500 | From | Dave Jones <> | Subject | Re: [BUG] jiffies counter leaps in 2.6.24-rc3 |
| |
On Thu, Nov 29, 2007 at 11:35:11AM +0100, Ingo Molnar wrote:
> * Stefano Brivio <stefano.brivio@polimi.it> wrote: > > > Ok, I found out how to reliably reproduce this bug. The root session issue > > was a bit weird, but I noticed I usually switch to root only when I need to > > change the CPU frequency. And here's what happens: > > > > # cpuspeedy min > > [after some time...] > > Nov 29 09:42:31 morte [ 4408.722597] b43-phy0 ERROR: PHY transmission error > > > > # cpuspeedy max > > Nov 29 09:42:41 morte [ 1177.238599] wmaster0: STA 00:14:c1:35:8d:eb Average rate: 540 (2160/4) > > > > It looks like jiffies get scaled exactly by a 3.75 factor, which is > > <max_frequency>/<min_frequency> here. I use acpi-cpufreq and this is my > > ah, the cpufreq/acpi/acpi-cpufreq angle is indeed a big clue. I've > Cc:-ed various interested parties.
Hmm, what's in /sys/devices/system/clocksource/clocksource0/current_clocksource ?
adjust_jiffies() always gets called on a speed transition. I'm wondering if perhaps we shouldn't do that if we're using something other than the tsc for timekeeping.
Dave
-- http://www.codemonkey.org.uk - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |