linux-kernel - [PATCH v2] perf/core: Wake up tasks for failing pinned events

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-Id: <20210129075757.1720900-1-namhyung@kernel.org>
Date:   Fri, 29 Jan 2021 16:57:57 +0900
From:   Namhyung Kim <namhyung@...nel.org>
To:     Arnaldo Carvalho de Melo <acme@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>
Cc:     Jiri Olsa <jolsa@...hat.com>, Ingo Molnar <mingo@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        LKML <linux-kernel@...r.kernel.org>,
        Stephane Eranian <eranian@...gle.com>,
        Andi Kleen <ak@...ux.intel.com>,
        Ian Rogers <irogers@...gle.com>
Subject: [PATCH v2] perf/core: Wake up tasks for failing pinned events

As of now we don't get any notice for pinned events when it's failed
to be scheduled and make it in an error state not try to schedule it
again.  That means we won't get any samples for the event.

It's possible we can detect it by reading the file, but usually we
only monitor it via mmap-ed ring buffers.  Let's poke the tasks
waiting for poll(2) so that they can respond to the event.

Signed-off-by: Namhyung Kim <namhyung@...nel.org>
---
 include/linux/perf_event.h |  1 +
 kernel/events/core.c       | 10 ++++++++++
 2 files changed, 11 insertions(+)

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 9a38f579bc76..0b3b3e97243b 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -733,6 +733,7 @@ struct perf_event {
 	int				pending_wakeup;
 	int				pending_kill;
 	int				pending_disable;
+	int				pending_pin_error;
 	struct irq_work			pending;
 
 	atomic_t			event_limit;
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 55d18791a72d..f8e9db30a573 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -3675,6 +3675,8 @@ static int merge_sched_in(struct perf_event *event, void *data)
 		if (event->attr.pinned) {
 			perf_cgroup_event_disable(event, ctx);
 			perf_event_set_state(event, PERF_EVENT_STATE_ERROR);
+			event->pending_pin_error = 1;
+			irq_work_queue(&event->pending);
 		}
 
 		*can_add_hw = 0;
@@ -5288,6 +5290,9 @@ static __poll_t perf_poll(struct file *file, poll_table *wait)
 	if (is_event_hup(event))
 		return events;
 
+	if (event->attr.pinned && event->state == PERF_EVENT_STATE_ERROR)
+		return EPOLLERR;
+
 	/*
 	 * Pin the event->rb by taking event->mmap_mutex; otherwise
 	 * perf_event_set_output() can swizzle our rb and make us miss wakeups.
@@ -6333,6 +6338,11 @@ static void perf_pending_event(struct irq_work *entry)
 		perf_event_wakeup(event);
 	}
 
+	if (event->pending_pin_error) {
+		event->pending_pin_error = 0;
+		wake_up_all(&event->waitq);
+	}
+
 	if (rctx >= 0)
 		perf_swevent_put_recursion_context(rctx);
 }
-- 
2.30.0.365.g02bc693789-goog