Hi Andy, On Mon, Jan 22, 2018 at 08:03:23AM +0000, Yeh, Andy wrote: > Hi Sakari, Tomasz, > > As below discussion that other drivers are with this pattern, I would prefer to defer to address the concern in later discussion with you and owners of other sensors. > > Thanks a lot. I thought of taking a look into the problem area and one sensor driver which doesn't appear to have the problem in this respect is imx258. This is because the v4l2_ctrl_handler_setup() isn't called in a runtime PM callback, but through V4L2 sub-dev s_stream callback instead. The runtime PM transition has already taken place by then. This isn't entirely optimal but works. Other sensor drivers will still need to be fixed. -- Sakari Ailus sakari.ailus@xxxxxxxxxxxxxxx