From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from mx1.redhat.com ([209.132.183.28]:65176 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752444AbaDIJGB (ORCPT ); Wed, 9 Apr 2014 05:06:01 -0400 Message-ID: <53450D76.2010405@redhat.com> Date: Wed, 09 Apr 2014 11:05:58 +0200 From: Hans de Goede MIME-Version: 1.0 To: alexander@xxor.de, linux-media@vger.kernel.org Subject: Re: gspca second isoc endpoint / kinect depth References: <53443C5D.9000607@xxor.de> In-Reply-To: <53443C5D.9000607@xxor.de> Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit Sender: linux-media-owner@vger.kernel.org List-ID: Hi, On 04/08/2014 08:13 PM, Alexander Sosna wrote: > Hi, > > I took drivers/media/usb/gspca/kinect.c as skeleton to build a depth > driver for the kinect camera. > > I needed to implement this feature because libfreenect performs so badly > on the raspberry pi that you can't get a single frame. > > The kinecet has two isoc endpoints but gspca only uses the first. > To get it running I made a dirty hack to drivers/media/usb/gspca/gspca.c > I changed usb_host_endpoint *alt_xfer(...) so that it always returns the > second endpoint, which is not really good for everyone. > > > My driver is not ready for upstream now, it can not coexist with the > current gspca_kinect so you have to decide if you want to load the video > or the depth driver. Would be better to have one driver to do it all. > > But in the meantime I would like to ask for ideas about a more clean > solution to get other isoc endpoints. > > There was already a little discussion about this when kinect.c was > written by Antonio Ospite: > http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/26194 > > http://article.gmane.org/gmane.linux.drivers.video-input-infrastructure/26213 > > Has something changed? No. > Is there a point against making multiple endpoints available? No. > Better solution? Not that I know of. Regards, Hans