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

From: Google
Date: Thu Jul 11 2024 - 11:18:05 EST


On Thu, 11 Jul 2024 10:51:18 +0200
Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:

> On Wed, Jul 10, 2024 at 11:40:17AM -0700, Andrii Nakryiko wrote:
> > On Wed, Jul 10, 2024 at 7:56 AM Masami Hiramatsu <mhiramat@xxxxxxxxxx> wrote:
> > >
> > > On Wed, 10 Jul 2024 12:10:03 +0200
> > > Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> > >
> > > > On Wed, Jul 10, 2024 at 07:10:46AM +0900, Masami Hiramatsu wrote:
> > > >
> > > > > > 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?
> > > >
> > > > ./scripts/get_maintainer.pl -f kernel/events/uprobes.c
> > > >
> > > > disagrees with that, also things like:
> > > >
> > > > https://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace.git/commit/?h=probes/for-next&id=4a365eb8a6d9940e838739935f1ce21f1ec8e33f
> > > >
> > > > touch common perf stuff, and very much would require at least an ack
> > > > from the perf folks.
> > >
> > > Hmm, indeed. I'm OK to pass those patches (except for trace_uprobe things)
> > > to -tip if you can.
> > >
> > > >
> > > > Not cool.
> > >
> >
> > You were aware of this patch and cc'ed personally (just like
> > linux-perf-users@xxxxxxxxxxxxxxx) on all revisions of it. I addressed
> > your concerns in [0], you went silent after that and patches were
> > sitting idle for more than a month.
>
> Yeah, I remember seeing it. But I was surprised it got applied. If I'm
> tardy -- this can happen, more so of late since I'm still recovering
> from injury and I get far more email than I could hope to process in a
> work day -- please ping.

I also forgot to ping you. I'll ask you next time.

>
> (also, being 'forced' into using a split keyboard means I'm also
> re-learning how to type, further slowing me down -- training muscle
> memory takes a while)
>
> Taking patches that touch other trees is fairly common, but in all those
> cases an ACK is 'required'.

OK, should I wait for your Ack for other patches on probes/for-next?

>
> (also also, I'm not the only maintainer there)
>
> > But regardless, if you'd like me to do any adjustments, please let me know.
> >
> > [0] https://lore.kernel.org/all/CAEf4Bzazi7YMz9n0V46BU7xthQjNdQL_zma5vzgCm_7C-_CvmQ@xxxxxxxxxxxxxx/
> >
>
> I'll check, it might be fine, its just the surprise of having it show up
> in some random tree that set me off.
>
> > > Yeah, the probe things are boundary.
> > > BTW, IMHO, there could be dependency issues on *probes. Those are usually used
> > > by ftrace/perf/bpf, which are managed by different trees. This means a series
> > > can span multiple trees. Mutually reviewing is the solution?
> > >
> >
> > I agree, there is no one best tree for stuff like this. So as long as
> > relevant people and mailing lists are CC'ed we hopefully should be
> > fine?
>
> Typically, yeah, that should work just fine.
>
> But if Masami wants to do uprobes, then it might be prudent to add a
> MAINTAINERS entry for it.
>
> A solution might be to add a UPROBES entry and add masami, oleg (if he
> wants) and myself as maintainers -- did I forget anyone? Git seems to
> suggest it's mostly been Oleg carrying this thing.

That sounds good for me. Like this?