Re: [PATCH v1 0/2] perf: Drop leaked kernel samples

[Stephane Eranian]

Hi,
On Fri, Jun 15, 2018 at 1:12 AM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> On Fri, Jun 15, 2018 at 04:01:45PM +0800, Jin, Yao wrote:
>
> > Bring more overhead to kernel if we zero the bits considering the number of
> > leaked samples may be not too small?
>
> Keeping the samples at least allows you to know how many samples
> happened and such things.
>
Yes, important point. You cannot just drop a sample. It may break some
calculations
done like num_samples x fixed_periods = total number of events observed. Some
tools do this.


> > And the skid information may be interesting (see example of hitting on
> > page_fault in previous mail). If we zero it, we will not know.
>
> If you throw them out the window you also don't know, do you?