Re: [PATCH 00/10] perf/uprobe: Optimize uprobes

[Google]

On Tue, 9 Jul 2024 12:16:34 +0200
Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:

> On Tue, Jul 09, 2024 at 12:01:03PM +0200, Jiri Olsa wrote:
> > On Tue, Jul 09, 2024 at 11:03:04AM +0200, Peter Zijlstra wrote:
> > > On Mon, Jul 08, 2024 at 05:25:14PM -0700, Andrii Nakryiko wrote:
> > > 
> > > > Ramping this up to 16 threads shows that mmap_rwsem is getting more
> > > > costly, up to 45% of CPU. SRCU is also growing a bit slower to 19% of
> > > > CPU. Is this expected? (I'm not familiar with the implementation
> > > > details)
> > > 
> > > SRCU getting more expensive is a bit unexpected, it's just a per-cpu
> > > inc/dec and a full barrier.
> > > 
> > > > P.S. Would you be able to rebase your patches on top of latest
> > > > probes/for-next, which include Jiri's sys_uretprobe changes. Right now
> > > > uretprobe benchmarks are quite unrepresentative because of that.
> > > 
> > > What branch is that? kernel/events/ stuff usually goes through tip, no?
> > 
> > it went through the trace tree:
> > 
> > https://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace.git probes/for-next
> > 
> > and it's in linux-next/master already
> 
> FFS :-/ That touches all sorts and doesn't have any perf ack on. Masami
> what gives?

This is managing *probes and related dynamic trace-events. Those has been
moved from tip. Could you also add linux-trace-kernel@vger ML to CC?

Thank you,

-- 
Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>