On 11.03.2016 21:30, Sylwester Nawrocki wrote:

On 01/26/2016 01:41 AM, Krzysztof Kozlowski wrote:

quoted

In fimc_md_parse_port_node() endpoint node is get with of_get_next_child()
but it is not put on error path.

"is get" doesn't sound right to me, how about rephrasing this to:

"In fimc_md_parse_port_node() reference count of the endpoint node
"is incremented by of_get_next_child() but it is not decremented
 on error path."

quoted

Fixes: 56fa1a6a6a7d ("[media] s5p-fimc: Change the driver directory name to exynos4-is")
Signed-off-by: Krzysztof Kozlowski <redacted>
---
Not tested on hardware, only built+static checkers.
---
 drivers/media/platform/exynos4-is/media-dev.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/media/platform/exynos4-is/media-dev.c 

b/drivers/media/platform/exynos4-is/media-dev.c
index f3b2dd30ec77..de0977479327 100644

--- a/drivers/media/platform/exynos4-is/media-dev.c
+++ b/drivers/media/platform/exynos4-is/media-dev.c

@@ -339,8 +339,10 @@ static int fimc_md_parse_port_node(struct fimc_md *fmd,
 		return 0;
 
 	v4l2_of_parse_endpoint(ep, &endpoint);
-	if (WARN_ON(endpoint.base.port == 0) || index >= FIMC_MAX_SENSORS)
+	if (WARN_ON(endpoint.base.port == 0) || index >= FIMC_MAX_SENSORS) {
+		of_node_put(ep);
 		return -EINVAL;
+	}

Thanks for the patch, it looks correct but it doesn't apply cleanly
due to patches already in media master branch [1]. Could you refresh
this patch and resend?
Also I don't quite like multiple calls to of_node_put(), how about
doing something like this instead:

How about sending your patch then with my reported-by?

Best regards,
Krzysztof