Hi Sylvain,
On Wed, Apr 30, 2025 at 04:19:14PM +0200, Sylvain Petinot wrote:
+static int vd56g3_subdev_init(struct vd56g3 *sensor)
+{
+ int ret;
+
+ /* Init remaining sub device ops */
+ sensor->sd.internal_ops = &vd56g3_internal_ops;
+ sensor->sd.flags |= V4L2_SUBDEV_FL_HAS_DEVNODE;
+ sensor->sd.entity.ops = &vd56g3_subdev_entity_ops;
+
+ /* Init source pad */
+ sensor->pad.flags = MEDIA_PAD_FL_SOURCE;
+ sensor->sd.entity.function = MEDIA_ENT_F_CAM_SENSOR;
+ ret = media_entity_pads_init(&sensor->sd.entity, 1, &sensor->pad);
+ if (ret) {
+ dev_err(sensor->dev, "Failed to init media entity : %d", ret);
+ return ret;
+ }
+
+ /* Init controls */
+ ret = vd56g3_init_controls(sensor);
+ if (ret) {
+ dev_err(sensor->dev, "Controls initialization failed %d", ret);
+ goto err_media;
+ }
+
+ /* Init vd56g3 struct : default resolution + raw8 */
+ sensor->sd.state_lock = sensor->ctrl_handler.lock;
+ ret = v4l2_subdev_init_finalize(&sensor->sd);
+ if (ret) {
+ dev_err(sensor->dev, "subdev init error: %d", ret);
+ goto err_ctrls;
+ }
+
+ return vd56g3_update_controls(sensor);
You're not holding the control handler's lock in the above call.
If your comment is related to the fact that 'vd56g3_update_controls() can
fail and that we do not free control handler not cleanup media entity, it's
fixed in V7.
It's not. The access to the control handler is serialised by a mutex and
you're not holding that mutex whilst calling vd56g3_update_controls() here.
The same issue exists in accessing sub-device state.
...
+ ret = vd56g3_parse_dt(sensor);
+ if (ret)
+ return dev_err_probe(dev, ret, "Failed to parse Device Tree.");
No need for the trailing dot in these messages. Same elsewhere.
You're right, Fixed in V7.
I took the opportunity to harmonize the error messages (Capital letter, no
space before ':', no trailing dot, missing '\n').
Ack, sounds good!