mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [RESEND PATCH v3] coresight: etm-perf: Fix reference count leak in etm_setup_aux
@ 2026-06-09  8:02 Ma Ke
  0 siblings, 0 replies; 4+ messages in thread
From: Ma Ke @ 2026-06-09  8:02 UTC (permalink / raw)
  To: suzuki.poulose, mike.leach, james.clark, leo.yan,
	alexander.shishkin, mathieu.poirier
  Cc: coresight, linux-arm-kernel, linux-kernel, akpm, Ma Ke, stable

bus_find_device() returns a device with its reference count
incremented. When a user-selected sink is obtained through
coresight_get_sink_by_id(), etm_setup_aux() keeps using the returned
sink while building the path and allocating the sink buffer.

Therefore the lookup reference must remain valid while etm_setup_aux()
is still using the sink, otherwise the sink could be removed under the
caller. Drop the lookup reference on the common exit path, after
etm_setup_aux() no longer directly uses the user-selected sink.

The CoreSight path code takes the references it needs for built paths,
so the initial lookup reference from coresight_get_sink_by_id() is no
longer needed after setup_aux finishes.

Found by code review.

Signed-off-by: Ma Ke <make24@iscas.ac.cn>
Cc: stable@vger.kernel.org
Fixes: 0e6c20517596 ("coresight: etm-perf: Allow an event to use different sinks")
---
Changes in v3:
- do not drop the lookup reference in coresight_get_sink_by_id(), as 
that would return a sink pointer without keeping the device reference 
while etm_setup_aux() is still using it.
- dropped the lookup reference in etm_setup_aux on the common exit path, 
as suggested by Suzuki.
- updated the commit message to describe why the reference is kept 
until etm_setup_aux() finishes using the sink.
Changes in v2:
- modified the patch as suggestions.
---
 drivers/hwtracing/coresight/coresight-etm-perf.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c
index f85dedf89a3f..d5116177c1b9 100644
--- a/drivers/hwtracing/coresight/coresight-etm-perf.c
+++ b/drivers/hwtracing/coresight/coresight-etm-perf.c
@@ -456,6 +456,11 @@ static void *etm_setup_aux(struct perf_event *event, void **pages,
 		goto err;
 
 out:
+	if (user_sink) {
+		put_device(&user_sink->dev);
+		user_sink = NULL;
+	}
+
 	return event_data;
 
 err:
-- 
2.43.0


^ permalink raw reply	[flat|nested] 4+ messages in thread
* [RESEND PATCH v3] coresight: etm-perf: Fix reference count leak in etm_setup_aux
@ 2026-06-22  7:10 Ma Ke
  0 siblings, 0 replies; 4+ messages in thread
From: Ma Ke @ 2026-06-22  7:10 UTC (permalink / raw)
  To: suzuki.poulose, mike.leach, james.clark, leo.yan,
	alexander.shishkin, mathieu.poirier
  Cc: coresight, linux-arm-kernel, linux-kernel, akpm, Ma Ke, stable

bus_find_device() returns a device with its reference count
incremented. When a user-selected sink is obtained through
coresight_get_sink_by_id(), etm_setup_aux() keeps using the returned
sink while building the path and allocating the sink buffer.

Therefore the lookup reference must remain valid while etm_setup_aux()
is still using the sink, otherwise the sink could be removed under the
caller. Drop the lookup reference on the common exit path, after
etm_setup_aux() no longer directly uses the user-selected sink.

The CoreSight path code takes the references it needs for built paths,
so the initial lookup reference from coresight_get_sink_by_id() is no
longer needed after setup_aux finishes.

Found by code review.

Signed-off-by: Ma Ke <make_ruc2021@163.com>
Cc: stable@vger.kernel.org
Fixes: 0e6c20517596 ("coresight: etm-perf: Allow an event to use different sinks")
---
Changes in v3:
- do not drop the lookup reference in coresight_get_sink_by_id(), as 
that would return a sink pointer without keeping the device reference 
while etm_setup_aux() is still using it.
- dropped the lookup reference in etm_setup_aux on the common exit path, 
as suggested by Suzuki.
- updated the commit message to describe why the reference is kept 
until etm_setup_aux() finishes using the sink.
Changes in v2:
- modified the patch as suggestions.
---
 drivers/hwtracing/coresight/coresight-etm-perf.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c
index f85dedf89a3f..d5116177c1b9 100644
--- a/drivers/hwtracing/coresight/coresight-etm-perf.c
+++ b/drivers/hwtracing/coresight/coresight-etm-perf.c
@@ -456,6 +456,11 @@ static void *etm_setup_aux(struct perf_event *event, void **pages,
 		goto err;
 
 out:
+	if (user_sink) {
+		put_device(&user_sink->dev);
+		user_sink = NULL;
+	}
+
 	return event_data;
 
 err:
-- 
2.43.0


^ permalink raw reply	[flat|nested] 4+ messages in thread
* [RESEND PATCH v3] coresight: etm-perf: Fix reference count leak in etm_setup_aux
@ 2026-06-30  9:11 Ma Ke
  0 siblings, 0 replies; 4+ messages in thread
From: Ma Ke @ 2026-06-30  9:11 UTC (permalink / raw)
  To: suzuki.poulose, mike.leach, james.clark, leo.yan,
	alexander.shishkin, mathieu.poirier
  Cc: coresight, linux-arm-kernel, linux-kernel, akpm, Ma Ke, stable

bus_find_device() returns a device with its reference count
incremented. When a user-selected sink is obtained through
coresight_get_sink_by_id(), etm_setup_aux() keeps using the returned
sink while building the path and allocating the sink buffer.

Therefore the lookup reference must remain valid while etm_setup_aux()
is still using the sink, otherwise the sink could be removed under the
caller. Drop the lookup reference on the common exit path, after
etm_setup_aux() no longer directly uses the user-selected sink.

The CoreSight path code takes the references it needs for built paths,
so the initial lookup reference from coresight_get_sink_by_id() is no
longer needed after setup_aux finishes.

Found by code review.

Signed-off-by: Ma Ke <make_ruc2021@163.com>
Cc: stable@vger.kernel.org
Fixes: 0e6c20517596 ("coresight: etm-perf: Allow an event to use different sinks")
---
Changes in v3:
- do not drop the lookup reference in coresight_get_sink_by_id(), as 
that would return a sink pointer without keeping the device reference 
while etm_setup_aux() is still using it.
- dropped the lookup reference in etm_setup_aux on the common exit path, 
as suggested by Suzuki.
- updated the commit message to describe why the reference is kept 
until etm_setup_aux() finishes using the sink.
Changes in v2:
- modified the patch as suggestions.
---
 drivers/hwtracing/coresight/coresight-etm-perf.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c
index 09b21a711a87..a25985743b7f 100644
--- a/drivers/hwtracing/coresight/coresight-etm-perf.c
+++ b/drivers/hwtracing/coresight/coresight-etm-perf.c
@@ -509,6 +509,11 @@ static void *etm_setup_aux(struct perf_event *event, void **pages,
 		goto err;
 
 out:
+	if (user_sink) {
+		put_device(&user_sink->dev);
+		user_sink = NULL;
+	}
+
 	return event_data;
 
 err:
-- 
2.43.0


^ permalink raw reply	[flat|nested] 4+ messages in thread
* [RESEND PATCH v3] coresight: etm-perf: Fix reference count leak in etm_setup_aux
@ 2026-07-22  7:31 Ma Ke
  0 siblings, 0 replies; 4+ messages in thread
From: Ma Ke @ 2026-07-22  7:31 UTC (permalink / raw)
  To: suzuki.poulose, mike.leach, james.clark, leo.yan,
	alexander.shishkin, mathieu.poirier
  Cc: coresight, linux-arm-kernel, linux-kernel, akpm, Ma Ke, stable

bus_find_device() returns a device with its reference count
incremented. When a user-selected sink is obtained through
coresight_get_sink_by_id(), etm_setup_aux() keeps using the returned
sink while building the path and allocating the sink buffer.

Therefore the lookup reference must remain valid while etm_setup_aux()
is still using the sink, otherwise the sink could be removed under the
caller. Drop the lookup reference on the common exit path, after
etm_setup_aux() no longer directly uses the user-selected sink.

The CoreSight path code takes the references it needs for built paths,
so the initial lookup reference from coresight_get_sink_by_id() is no
longer needed after setup_aux finishes.

Found by code review.

Signed-off-by: Ma Ke <make_ruc2021@163.com>
Cc: stable@vger.kernel.org
Fixes: 0e6c20517596 ("coresight: etm-perf: Allow an event to use different sinks")
---
Changes in v3:
- do not drop the lookup reference in coresight_get_sink_by_id(), as 
that would return a sink pointer without keeping the device reference 
while etm_setup_aux() is still using it.
- dropped the lookup reference in etm_setup_aux on the common exit path, 
as suggested by Suzuki.
- updated the commit message to describe why the reference is kept 
until etm_setup_aux() finishes using the sink.
Changes in v2:
- modified the patch as suggestions.
---
 drivers/hwtracing/coresight/coresight-etm-perf.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c
index 09b21a711a87..a25985743b7f 100644
--- a/drivers/hwtracing/coresight/coresight-etm-perf.c
+++ b/drivers/hwtracing/coresight/coresight-etm-perf.c
@@ -509,6 +509,11 @@ static void *etm_setup_aux(struct perf_event *event, void **pages,
 		goto err;
 
 out:
+	if (user_sink) {
+		put_device(&user_sink->dev);
+		user_sink = NULL;
+	}
+
 	return event_data;
 
 err:
-- 
2.43.0


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2026-07-22  7:32 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-06-09  8:02 [RESEND PATCH v3] coresight: etm-perf: Fix reference count leak in etm_setup_aux Ma Ke
2026-06-22  7:10 Ma Ke
2026-06-30  9:11 Ma Ke
2026-07-22  7:31 Ma Ke

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®