Linux Perf Users
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Ian Rogers" <irogers@google.com>
Cc: linux-perf-users@vger.kernel.org
Subject: Re: [PATCH v3 03/16] perf pmu: Add perf_pmu__probe() for basic open, counting, and sampling
Date: Fri, 20 Nov 2026 22:39:25 +0000	[thread overview]
Message-ID: <sashiko-outbox-226286@kernel.org> (raw)
In-Reply-To: <20261120222615.2291168-4-irogers@google.com>

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Low] perf pmu: try_open_attr_variants() closes tracepoint fd before
  keepalive
--

commit 000f6d2820c05383b074beef8f5d9879e6d9d78a
Author: Ian Rogers <irogers@google.com>

perf pmu: Add perf_pmu__probe() for basic open, counting, and sampling

This commit adds perf_pmu__probe() and struct perf_pmu::probed to probe a
live PMU via perf_event_open() and record whether opening, counting, and
sampling are supported.

> diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
> index 1d3cfa0823475..b43afb8b155b3 100644
> --- a/tools/perf/util/pmu.c
> +++ b/tools/perf/util/pmu.c
[ ... ]
> @@ -2850,9 +2866,570 @@ void perf_pmu__delete(struct perf_pmu *pmu)
[ ... ]
> +static bool try_open_attr_variants(struct perf_event_attr *attr,
> +				   bool is_uncore, int pmu_cpu,
> +				   unsigned long flags,
> +				   pid_t *out_pid, int *out_cpu,
> +				   int *last_err)
> +{
[ ... ]
> +				fd = sys_perf_event_open(&try_attr, pid, pmu_cpu, -1, flags);
> +				if (fd >= 0) {
> +					close(fd);
> +					*attr = try_attr;
> +					*out_pid = pid;
> +					*out_cpu = pmu_cpu;
> +					return true;
> +				}
[ ... ]
> +static void __perf_pmu__probe(struct perf_pmu *pmu, int pmu_cpu)
> +{
[ ... ]
> +	/*
> +	 * For tracepoint events, closing the last fd for a tracepoint triggers
> +	 * tracepoint_synchronize_unregister() (an RCU grace period) in the
> +	 * kernel. Hold one disabled fd open across the probe sequence so
> +	 * tp_event->perf_refcount stays >= 1 and each probe open/close is
> +	 * instantaneous.
> +	 */
> +	if (pmu->type == PERF_TYPE_TRACEPOINT) {
> +		struct perf_event_attr ka_attr = base_attr;
> +
> +		ka_attr.disabled = 1;
> +		keepalive_fd = sys_perf_event_open(&ka_attr, pid, cpu, -1, flags);
> +	}

[Severity: Low]
Does closing fd inside try_open_attr_variants() defeat the tracepoint
keepalive optimization?

When try_open_attr_variants() closes the initial tracepoint fd before
keepalive_fd is opened, tp_event->perf_refcount drops to 0 in
perf_trace_event_unreg(), paying the tracepoint_synchronize_unregister()
RCU grace period and freeing the per-CPU tracepoint buffers before
__perf_pmu__probe() re-opens the same disabled attr as keepalive_fd.
Returning the open fd from try_open_attr_variants() to use as
keepalive_fd (or closing it when pmu->type != PERF_TYPE_TRACEPOINT) would
avoid the extra RCU grace period and redundant perf_event_open() call.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261120222615.2291168-1-irogers@google.com?part=3

  reply	other threads:[~2026-11-20 22:39 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-20 22:25 [PATCH v3 00/16] perf sysinfo: Add subcommand to probe and summarize PMU capabilities Ian Rogers
2026-11-20 22:26 ` [PATCH v3 01/16] perf pmus: Expose duplicate uncore PMU scanning helpers Ian Rogers
2026-11-20 22:26 ` [PATCH v3 02/16] perf mem-events: Expose PMU init and name formatting helpers Ian Rogers
2026-11-20 22:26 ` [PATCH v3 03/16] perf pmu: Add perf_pmu__probe() for basic open, counting, and sampling Ian Rogers
2026-11-20 22:39   ` sashiko-bot [this message]
2026-11-20 22:26 ` [PATCH v3 04/16] perf pmu: Probe supported sample_type flags Ian Rogers
2026-11-20 22:26 ` [PATCH v3 05/16] perf pmu: Probe precise_ip levels Ian Rogers
2026-11-20 22:26 ` [PATCH v3 06/16] perf pmu: Probe privilege exclusion modifiers Ian Rogers
2026-11-20 22:26 ` [PATCH v3 07/16] perf pmu: Probe execution and AUX area features Ian Rogers
2026-11-20 22:26 ` [PATCH v3 08/16] perf pmu: Probe event group size limits Ian Rogers
2026-11-20 22:26 ` [PATCH v3 09/16] perf pmu: Probe user and interrupt sample register masks Ian Rogers
2026-11-20 22:26 ` [PATCH v3 10/16] perf pmu: Probe interrupt registers that require precise sampling Ian Rogers
2026-11-20 22:26 ` [PATCH v3 11/16] perf pmu: Probe memory profiling events and sample formats Ian Rogers
2026-11-20 22:26 ` [PATCH v3 12/16] perf pmu: Probe user-space mmap capabilities Ian Rogers
2026-11-20 22:26 ` [PATCH v3 13/16] perf pmu: Probe branch instruction skid Ian Rogers
2026-11-20 22:26 ` [PATCH v3 14/16] perf sysinfo: Add sysinfo subcommand Ian Rogers
2026-11-20 22:26 ` [PATCH v3 15/16] perf docs: Add perf-sysinfo man page Ian Rogers
2026-11-20 22:26 ` [PATCH v3 16/16] perf test: Add perf sysinfo shell test Ian Rogers
2026-11-20 22:45 ` [PATCH v3 00/16] perf sysinfo: Add subcommand to probe and summarize PMU capabilities sashiko-bot

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-226286@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=irogers@google.com \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox