[PATCH] gpu: host1x: Fix of_node reference leak in host1x_device_add()

[Wentao Liang]

When host1x_device_parse_dt() fails, host1x_device_add() only frees the
host1x_device and leaves the subdevs collected by host1x_subdev_add()
on device->subdevs. Each of those subdevs still holds an OF node
reference acquired with of_node_get(), which is normally dropped by
host1x_subdev_del().

Remove all stale subdevs, dropping their device node references, before
freeing the device.

Fixes: 776dc3840367 ("drm/tegra: Move subdevice infrastructure to host1x")
Cc: stable@xxxxxxxxxxxxxxx
Signed-off-by: Wentao Liang <vulab@xxxxxxxxxxx>
---
 drivers/gpu/host1x/bus.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/host1x/bus.c b/drivers/gpu/host1x/bus.c
index f814eb4941c0..8dbab96b7f5a 100644
--- a/drivers/gpu/host1x/bus.c
+++ b/drivers/gpu/host1x/bus.c
@@ -455,7 +455,7 @@ static int host1x_device_add(struct host1x *host1x,
 			     struct host1x_driver *driver)
 {
 	struct host1x_client *client, *tmp;
-	struct host1x_subdev *subdev;
+	struct host1x_subdev *subdev, *sd;
 	struct host1x_device *device;
 	int err;
 
@@ -485,6 +485,10 @@ static int host1x_device_add(struct host1x *host1x,
 
 	err = host1x_device_parse_dt(device, driver);
 	if (err < 0) {
+		mutex_lock(&device->subdevs_lock);
+		list_for_each_entry_safe(subdev, sd, &device->subdevs, list)
+			host1x_subdev_del(subdev);
+		mutex_unlock(&device->subdevs_lock);
 		kfree(device);
 		return err;
 	}
-- 
2.34.1