On Dec 8, 2007 2:42 PM, Ingo Molnar <[EMAIL PROTECTED]> wrote: > > * Parag Warudkar <[EMAIL PROTECTED]> wrote: > > > Even on 100% idle I get variations that are approx in the same range > > when not idle. Clocksource is hpet if that matters. Next I think I > > will disable CPU_IDLE, Tickless > > also try the hpet=disable boot option.
With hpet=disable (clocksource is acpi_pm ) the usleeps are somewhat more closer to correct - under idle as well as under kernel compilation load. But there are still fluctuations under 100% idle - Kernel compile -------------------------- real 0m1.112s user 0m0.004s sys 0m0.000s real 0m1.112s user 0m0.000s sys 0m0.000s real 0m1.115s user 0m0.000s sys 0m0.000s real 0m1.112s user 0m0.000s sys 0m0.000s real 0m1.112s user 0m0.000s sys 0m0.000s ------------------------------ IDLE real 0m1.112s user 0m0.000s sys 0m0.000s real 0m1.131s user 0m0.000s sys 0m0.000s real 0m1.112s user 0m0.000s sys 0m0.000s real 0m1.112s user 0m0.000s sys 0m0.000s real 0m1.139s user 0m0.000s sys 0m0.000s > > > My ssh connection just died - another lockup in dmesg - > > > > BUG: soft lockup - CPU#1 stuck for 14s! [sshd:509] > > this would suggest that the softlockup watchdog is correct and indeed > there was some lockup of sorts? Yep, I was about to say that there are indeed real soft-lockups going on as opposed to the psuedo ones that we thought may be due to timer inconsistencies. BTW this is no-preempt config and 2.6.22 and latest Centos-5.1 kernels ran fine with HPET. Parag -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
