Messages in this thread | | | Subject | Re: [RFC][PATCH] perf, intel: Don't touch MSR_IA32_DEBUGCTLMSR from NMI context | From | Peter Zijlstra <> | Date | Thu, 13 Sep 2012 14:11:05 +0200 |
| |
On Thu, 2012-09-13 at 13:49 +0200, Stephane Eranian wrote: > Should be, though it is pretty ugly to stash all of this in the > put/get constraints.
Agreed, I almost added two extra functions for it but when I went to look at where to call them I ended up next to get/put constraints.
> I will run some tests. > > I wonder what this does when you come in to get/put with a fake cpuc. You don't > want to perturb the local lbr which may be in use.
Fake cpu will never set ->alloc and thus intel_pmu_has_lbr() should fail and we don't do anything.
| |