Re: [PATCH] perf, nmi: Move LVT un-masking into irq handlers

From: Don Zickus
Date: Wed Apr 20 2011 - 10:58:16 EST


On Wed, Apr 20, 2011 at 06:37:10PM +0400, Cyrill Gorcunov wrote:
> Don, sorry for delay, I somehow missed this patch at first place :/ And probably
> the patch below could help with "stess" problem?

The patch I am using for the 'stress' problem is different than what I
posted because I wanted to keep it simple for 2.6.39. But yeah I already
moved the unmasking to the enable_pmus_all(0) line.

The thing is that cuts down on some of these in-flight NMIs but not all of
them. I came up with some logic that catches all the in-flight NMIs on
Nehalem and AMDs but my core2quad still fails. I'll try to post that
today while requesting help.

Still not sure you need the double APIC writes (as your patch) shows. I
haven't seen a case in my testing where a single APIC write fails.

Cheers,
Don
--
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/