Skip to contentSkip to conversation

perf/core: do not set cpuctx->cgrp for unscheduled cgroups

PATCH Superseded David Carrillo-Cisneros 2 participants 3 messages Oct 11, 2016 – Nov 16, 2016 perf events

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 perf/core: do not set cpuctx->cgrp for unscheduled cgroups

commit db4a835601b7
("perf/core: Set cgroup in CPU contexts for new cgroup events")

fails to verify that event->cgrp is actually the scheduled cgroup in a CPU before setting cpuctx->cgrp. This patch fixes that. The bug only impacts the first cgroup that is created in a CPU.

Now that there is a different path for scheduled and unscheduled cgroups, add a warning to catch when cpuctx->cgrp is still set after the last cgroup event has been unsheduled.

---
 kernel/events/core.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
kernel/events/core.c +11 -0
index 7c0d263..62f70a3 100644
@@ -902,6 +902,17 @@ list_update_cgroup_event(struct perf_event *event,
 	 * this will always be called from the right CPU.
 	 */
 	cpuctx = __get_cpu_context(ctx);
+
+	/* Only set/clear cpuctx->cgrp if current task uses event->cgrp. */
+	if (perf_cgroup_from_task(current, ctx) != event->cgrp) {
+		/*
+		 * We are removing the last cpu event in this context.
+		 * If that event is not active in this cpu, cpuctx->cgrp
+		 * should've been cleared by perf_cgroup_switch.
+		 */
+		WARN_ON_ONCE(!add && cpuctx->cgrp);
+		return;
+	}
 	cpuctx->cgrp = add ? event->cgrp : NULL;
 }
 
Signature · 1 line
2.8.0.rc3.226.g39d4020
#1 lore raw

PATCH v2 1/1 perf/core: do not set cpuctx->cgrp for unscheduled cgroups

(v2 only extends this changelog.)

commit db4a835601b7
("perf/core: Set cgroup in CPU contexts for new cgroup events")

failed to verify that event->cgrp is actually the scheduled cgroup in a CPU before setting cpuctx->cgrp. This patch fixes that.

Now that there is a different path for scheduled and unscheduled cgroup, add a warning to catch when cpuctx->cgrp is still set after the last cgroup event has been unsheduled.

To verify the bug:

# Create 2 cgroups.
mkdir /dev/cgroups/devices/g1
mkdir /dev/cgroups/devices/g2

  # launch a task, bind it to a cpu and move it to g1
  CPU=2
  while :; do : ; done &
  P=$!

taskset -pc $CPU $P
echo $P > /dev/cgroups/devices/g1/tasks

  # monitor g2 (it runs no tasks) and observe output
  perf stat -e cycles -I 1000 -C $CPU -G g2

# time counts unit events

     1.000091408          7,579,527      cycles                    g2
     2.000350111      <not counted>      cycles                    g2
     3.000589181      <not counted>      cycles                    g2
     4.000771428      <not counted>      cycles                    g2
  # note first line that displays that a task run in g2, despite
  # g2 having no tasks. This is because cpuctx->cgrp was wrongly
  # set when context of new event was installed.
  # After applying the fix we obtain the right output:

perf stat -e cycles -I 1000 -C $CPU -G g2
# time counts unit events

     1.000119615      <not counted>      cycles                    g2
     2.000389430      <not counted>      cycles                    g2
     3.000590962      <not counted>      cycles                    g2
---
 kernel/events/core.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
kernel/events/core.c +11 -0
index c6e47e9..55953db 100644
@@ -902,6 +902,17 @@ list_update_cgroup_event(struct perf_event *event,
 	 * this will always be called from the right CPU.
 	 */
 	cpuctx = __get_cpu_context(ctx);
+
+	/* Only set/clear cpuctx->cgrp if current task uses event->cgrp. */
+	if (perf_cgroup_from_task(current, ctx) != event->cgrp) {
+		/*
+		 * We are removing the last cpu event in this context.
+		 * If that event is not active in this cpu, cpuctx->cgrp
+		 * should've been cleared by perf_cgroup_switch.
+		 */
+		WARN_ON_ONCE(!add && cpuctx->cgrp);
+		return;
+	}
 	cpuctx->cgrp = add ? event->cgrp : NULL;
 }
 
Signature · 1 line
2.8.0.rc3.226.g39d4020
#2 lore raw

[tip:perf/urgent] perf/core: Do not set cpuctx->cgrp for unscheduled cgroups

perf/core: Do not set cpuctx->cgrp for unscheduled cgroups
Commit-ID:  864c2357ca898c6171fe5284f5ecc795c8ce27a8
Gitweb:     http://git.kernel.org/tip/864c2357ca898c6171fe5284f5ecc795c8ce27a8
Author:     David Carrillo-Cisneros <[email protected]>
AuthorDate: Tue, 1 Nov 2016 11:52:58 -0700
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 15 Nov 2016 14:18:22 +0100

perf/core: Do not set cpuctx->cgrp for unscheduled cgroups

Commit:

db4a835601b7 ("perf/core: Set cgroup in CPU contexts for new cgroup events")

failed to verify that event->cgrp is actually the scheduled cgroup in a CPU before setting cpuctx->cgrp. This patch fixes that.

Now that there is a different path for scheduled and unscheduled cgroup, add a warning to catch when cpuctx->cgrp is still set after the last cgroup event has been unsheduled.

To verify the bug:

# Create 2 cgroups.
mkdir /dev/cgroups/devices/g1
mkdir /dev/cgroups/devices/g2

  # launch a task, bind it to a cpu and move it to g1
  CPU=2
  while :; do : ; done &
  P=$!

taskset -pc $CPU $P
echo $P > /dev/cgroups/devices/g1/tasks

  # monitor g2 (it runs no tasks) and observe output
  perf stat -e cycles -I 1000 -C $CPU -G g2

# time counts unit events

     1.000091408          7,579,527      cycles                    g2
     2.000350111      <not counted>      cycles                    g2
     3.000589181      <not counted>      cycles                    g2
     4.000771428      <not counted>      cycles                    g2
  # note first line that displays that a task run in g2, despite
  # g2 having no tasks. This is because cpuctx->cgrp was wrongly
  # set when context of new event was installed.
  # After applying the fix we obtain the right output:

perf stat -e cycles -I 1000 -C $CPU -G g2
# time counts unit events

     1.000119615      <not counted>      cycles                    g2
     2.000389430      <not counted>      cycles                    g2
     3.000590962      <not counted>      cycles                    g2
Cc Alexander Shishkin <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Kan Liang <[email protected]>
Cc Linus Torvalds <[email protected]>
Cc Nilay Vaish <[email protected]>
Cc Paul Turner <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Vegard Nossum <[email protected]>
---
 kernel/events/core.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
kernel/events/core.c +11 -0
index 0e29213..ff230bb 100644
@@ -902,6 +902,17 @@ list_update_cgroup_event(struct perf_event *event,
 	 * this will always be called from the right CPU.
 	 */
 	cpuctx = __get_cpu_context(ctx);
+
+	/* Only set/clear cpuctx->cgrp if current task uses event->cgrp. */
+	if (perf_cgroup_from_task(current, ctx) != event->cgrp) {
+		/*
+		 * We are removing the last cpu event in this context.
+		 * If that event is not active in this cpu, cpuctx->cgrp
+		 * should've been cleared by perf_cgroup_switch.
+		 */
+		WARN_ON_ONCE(!add && cpuctx->cgrp);
+		return;
+	}
 	cpuctx->cgrp = add ? event->cgrp : NULL;
 }