https://bugzilla.novell.com/show_bug.cgi?id=579932
https://bugzilla.novell.com/show_bug.cgi?id=579932#c21
--- Comment #21 from Andreas Nordal 2010-09-02 21:39:03 UTC ---
So here comes the test results.
Judging by ´dmesg | grep clock´, hpet is my default clocksource in all my 4
kernels tested, namely 2.6.34-12-vanilla, 2.6.34-12-desktop, 2.6.32.20 and
2.6.35.3. They all concluded with "Switching to clocksource hpet".
The affected kernels hang if and only if hpet is the clocksource and
nolapic_timer is not set!!! To sum up, the 3 known workarounds so far are:
clocksource=jiffies, clocksource=tsc and nolapic_timer.
2.6.34-12-desktop also hangs when trying clocksource=tsc, but that was
apparently because it fell back to hpet:
[ 0.000000] hpet clockevent registered
[ 0.384023] Switching to clocksource tsc
[ 3.683563] Override clocksource tsc is not HRT compatible. Cannot switch
while in HRT/NOHZ mode
[ 3.683566] Switching to clocksource hpet
Not all is good with workarounds:
With 2.6.35.5, all 3 workarounds inhibited Powertop from giving detailed info
about each task, and caused it to complain that I should enable CONFIG_NOHZ in
the kernel (which I in fact had). Same with 2.6.34-12-desktop, with the strange
exception (hope I'm not mistaken), that Powertop worked fine with
clocksource=jiffies. I did not test Powertop with 2.6.34-12-vanilla.
The unaffected 2.6.32.20 is of course impossible to hang, yet Powertop works
flawless. Since this combination is inachievable with new kernels, this bug is
a regression, not a feature.
--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.