Re: [PATCH V5 1/1] bpf: control events stored in PERF_EVENT_ARRAY maps trace data output when perf sampling

[Peter Zijlstra <peterz@infradead.org>]

On Thu, Oct 22, 2015 at 11:12:16AM +0800, Wangnan (F) wrote:
> On 2015/10/22 11:09, Alexei Starovoitov wrote:
> >On 10/21/15 6:56 PM, Wangnan (F) wrote:
> >>>One alternative solution I can image is to attach a BPF program
> >>>at sampling like kprobe, and return 0 if we don't want sampling
> >>>take action. Thought?
> >>
> >>Do you think attaching BPF programs to sampling is an acceptable idea?
> >
> >If you mean to extend 'filter' concept to sampling events?
> >So instead of soft_disable of non-local events, you'll attach bpf
> >program to sampling events and use map lookup to decide whether
> >to filter out or not such sampling event?
> 
> Yes.

One could overload or stack the overflow handler I suppose. But this
would be in line with the software/tracepoint events calling eBPF muck
on trigger, right?

> >What pt_regs would be in such case?

> Sampling is based on interruption. We can use pt_reg captured by the IRQ
> handler,

s/IRQ/NMI/

Also, we 'edit' the pt_regs on the way down to the overflow handler as
sometimes 'better' information can be had from PMU state. But a pt_regs
is available there.