Re: NMI received for unknown reason, 2.6.38-rc6 regression?

[denys@visp.net.lb]

 On Wed, 2 Mar 2011 08:59:31 +0100, Ingo Molnar wrote:
> * denys@visp.net.lb <denys@visp.net.lb> wrote:
>
>> On Tue, 01 Mar 2011 19:08:43 +0300, Cyrill Gorcunov wrote:
>> >On 03/01/2011 06:03 PM, denys@visp.net.lb wrote:
>> >>I upgrade around 140 hosts (from 2.6.33 till 2.6.37), and got on
>> >>many of them error/warining, flooding kernel log. Here is short
>> >>snapshot:
>> >>
>> >>[ 1882.057474] Uhhuh. NMI received for unknown reason 3c on CPU 0.
>> >>[ 1882.057576] Do you have a strange power saving mode enabled?
>> >>[ 1882.057672] Dazed and confused, but trying to continue
>> >>[ 2421.419732] Uhhuh. NMI received for unknown reason 3c on CPU 0.
>> >>[ 2421.419835] Do you have a strange power saving mode enabled?
>> >>[ 2421.419930] Dazed and confused, but trying to continue
>> >>[ 2636.016831] Uhhuh. NMI received for unknown reason 2c on CPU 1.
>> >>[ 2636.016934] Do you have a strange power saving mode enabled?
>> >>[ 2636.017003] Dazed and confused, but trying to continue
>> >>
>> >>Full dmesg from 2 machines:
>> >>http://www.nuclearcat.com/dmesg1.txt
>> >>http://www.nuclearcat.com/dmesg2.txt
>> >>I can provide more, if required.
>> >>
>> >>It seems nmi_watchdog is enabled by default, and it is causing
>> >>issue. I am checking now with nmi_watchdog=0, but i need more
>> >>time to confirm that.
>> >>Also i am experiencing some problem with ppp users(all of them
>> >>is pppoe servers), but i am not sure it is related to that, so
>> >>maybe this NMI warning is just cosmetic regression.
>> >>
>> >>All systems is x86, same kernel config.
>> >>If you need more information - let me know.
>> >>
>> >
>> >nmi_watchdog=0 should help here, actually a nit was fixed by
>> >https://patchwork.kernel.org/patch/566611/
>> >which is not in 2.6.38-rc6 but I rather suspect it'll be in -rc7 or
>> >final .38. If you have an ability
>> >to pickup it and test -- this would be great!
>> I test it, and it seems helps. At least on one host, and yes, seems
>> all of them P4.
>
> Mind checking -rc7, does it work 'out of box', without requiring any
> workarounds?
> -rc7 already has this fix included:
>
>  7d44ec193d95: perf, x86: P4 PMU: Fix spurious NMI messages
>
> -rc6 did not have it yet.

 Yes, rc7 fine too, tested it now.

>
> Thanks,
>
> 	Ingo