On Mon, Mar 1, 2010 at 4:58 AM, Hans Verkuil <hverkuil@xxxxxxxxx> wrote: > New private controls should not use V4L2_CID_PRIVATE_BASE at all. That > mechanism was a really bad idea. Instead a new control should be added to > the appropriate control class and with a offset >= 0x1000. See for example > the CX2341X private controls in videodev2.h. So, you're suggesting that the following patch then is going to be NAK'd and that I'm going to have to go back and convert saa7115 to support the extended controls API, extend the em28xx driver to support the extended controls API, and retest with all the possible applications given how they might potentially be attempting to implement the rather poorly documented interface? http://kernellabs.com/hg/~dheitmueller/em28xx-test/rev/a7d50db75420 And exactly what "class" of extended controls should I use for video decoders? It would seem that a video decoder doesn't really fall into any of the existing categories - "Old-style user controls", "MPEG compression controls", "Camera controls", or "FM modulator controls". Are we saying that now I'm also going to be introducing a whole new class of control too? > The EXT_G/S_CTRLS ioctls do not accept PRIVATE_BASE because I want to > force driver developers to stop using PRIVATE_BASE. So only G/S_CTRL will > support controls in that range for backwards compatibility. While we're on the topic, do you see any problem with the proposed fix for the regression you introduced?: http://kernellabs.com/hg/~dheitmueller/em28xx-test/rev/142ae5aa9e8e Between trying to figure out what the expected behavior is supposed to be (given the complete lack of documentation on how private controls are expected to be implemented in the extended controls API) and isolating and fixing the regression, it's hard not to be a little irritated at this situation. This was supposed to be a very small change - a single private control to a mature driver. And now it seems like I'm going to have to extend the basic infrastructure in the decoder driver, the bridge driver, add a new class of controls, all so I can poke one register? Devin -- Devin J. Heitmueller - Kernel Labs http://www.kernellabs.com -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html