[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAP-5=fVp48KUB5sawBp-6t5DyzGqDfxtXReSSTQ-LDXFKMAvzQ@mail.gmail.com>
Date: Mon, 17 Mar 2025 10:41:28 -0700
From: Ian Rogers <irogers@...gle.com>
To: Arnaldo Carvalho de Melo <acme@...nel.org>
Cc: Namhyung Kim <namhyung@...nel.org>, Ingo Molnar <mingo@...nel.org>,
Thomas Gleixner <tglx@...utronix.de>, James Clark <james.clark@...aro.org>,
Jiri Olsa <jolsa@...nel.org>, Adrian Hunter <adrian.hunter@...el.com>,
Kan Liang <kan.liang@...ux.intel.com>, Clark Williams <williams@...hat.com>,
linux-kernel@...r.kernel.org, linux-perf-users@...r.kernel.org,
Arnaldo Carvalho de Melo <acme@...hat.com>
Subject: Re: [PATCH v2 0/6] perf python binding fixes
On Wed, Mar 12, 2025 at 1:31 PM Arnaldo Carvalho de Melo
<acme@...nel.org> wrote:
>
> From: Arnaldo Carvalho de Melo <acme@...hat.com>
>
> Hi Namhyung,
>
> So this simplifies it greatly, it almost ends up as a one-liner,
> but there is an improvement as well to mark the event as consumed to
> then parse its sample, because the copy of the whole event was done all
> along.
>
> This is brittle, as the header size can be bigger, than the
> space we use and in that case we fail to parse the event by bounds
> checking it.
>
> Supporting larger event payloads can be done on top of this,
> possibly by deferring consuming the event in the ring buffer by parsing
> it all instead of having pre-allocated space, measurements need to be
> made to see whats best. I'd say leave this for when it proves necessary.
>
> With this series I managed to run it for a long time without
> crashes and 'top' says it doesn't seem to be leaking anything, as its
> memory usage stays the same for as long as I looked.
>
> Please consider applying to perf-tools-next,
>
> Best regards
>
> P.S.: In other news, the syscalltbl series from Ian built on all my
> containers, I'm now trying to go over it patch by patch.
>
> Arnaldo Carvalho de Melo (6):
> perf python: Fixup description of sample.id event member
> perf python: Remove some unused macros (_PyUnicode_FromString(arg), etc)
> perf python tracepoint.py: Change the COMM using setproctitle if available
> perf python: Decrement the refcount of just created event on failure
> perf python: Don't keep a raw_data pointer to consumed ring buffer space
> perf python: Check if there is space to copy all the event
Reviewed-by: Ian Rogers <irogers@...gle.com>
Thanks,
Ian
Powered by blists - more mailing lists