linux-kernel - [tip:perfcounters/core] perf_counter: Do not throttle single swcounter 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: <tip-850bc73ffcc99cddfb52bc23217c60810c508853@git.kernel.org>
Date:	Thu, 17 Sep 2009 20:12:54 GMT
From:	tip-bot for Peter Zijlstra <a.p.zijlstra@...llo.nl>
To:	linux-tip-commits@...r.kernel.org
Cc:	linux-kernel@...r.kernel.org, hpa@...or.com, mingo@...hat.com,
	a.p.zijlstra@...llo.nl, tglx@...utronix.de, mingo@...e.hu
Subject: [tip:perfcounters/core] perf_counter: Do not throttle single swcounter events

Commit-ID:  850bc73ffcc99cddfb52bc23217c60810c508853
Gitweb:     http://git.kernel.org/tip/850bc73ffcc99cddfb52bc23217c60810c508853
Author:     Peter Zijlstra <a.p.zijlstra@...llo.nl>
AuthorDate: Thu, 17 Sep 2009 18:47:11 +0200
Committer:  Ingo Molnar <mingo@...e.hu>
CommitDate: Thu, 17 Sep 2009 22:08:25 +0200

perf_counter: Do not throttle single swcounter events

We can have swcounter events that contribute more than a single
count per event, when used with a non-zero period, those can
generate multiple events, which is when we need throttling.

However, swcounter that contribute only a single count per event
can only come as fast as we can run code, hence don't throttle
them.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@...llo.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@...e.hu>


---
 kernel/perf_counter.c |   17 +++++++++++++----
 1 files changed, 13 insertions(+), 4 deletions(-)

diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index 667ab25..fe0d1ad 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -3494,14 +3494,15 @@ static void perf_log_throttle(struct perf_counter *counter, int enable)
  * Generic counter overflow handling, sampling.
  */
 
-int perf_counter_overflow(struct perf_counter *counter, int nmi,
-			  struct perf_sample_data *data)
+static int __perf_counter_overflow(struct perf_counter *counter, int nmi,
+				   int throttle, struct perf_sample_data *data)
 {
 	int events = atomic_read(&counter->event_limit);
-	int throttle = counter->pmu->unthrottle != NULL;
 	struct hw_perf_counter *hwc = &counter->hw;
 	int ret = 0;
 
+	throttle = (throttle && counter->pmu->unthrottle != NULL);
+
 	if (!throttle) {
 		hwc->interrupts++;
 	} else {
@@ -3554,6 +3555,12 @@ int perf_counter_overflow(struct perf_counter *counter, int nmi,
 	return ret;
 }
 
+int perf_counter_overflow(struct perf_counter *counter, int nmi,
+			  struct perf_sample_data *data)
+{
+	return __perf_counter_overflow(counter, nmi, 1, data);
+}
+
 /*
  * Generic software counter infrastructure
  */
@@ -3592,6 +3599,7 @@ static void perf_swcounter_overflow(struct perf_counter *counter,
 				    int nmi, struct perf_sample_data *data)
 {
 	struct hw_perf_counter *hwc = &counter->hw;
+	int throttle = 0;
 	u64 overflow;
 
 	data->period = counter->hw.last_period;
@@ -3601,13 +3609,14 @@ static void perf_swcounter_overflow(struct perf_counter *counter,
 		return;
 
 	for (; overflow; overflow--) {
-		if (perf_counter_overflow(counter, nmi, data)) {
+		if (__perf_counter_overflow(counter, nmi, throttle, data)) {
 			/*
 			 * We inhibit the overflow from happening when
 			 * hwc->interrupts == MAX_INTERRUPTS.
 			 */
 			break;
 		}
+		throttle = 0;
 	}
 }
 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/