On Friday 15 September 2006 03:52, Steven Toth wrote: > Hi, > > I'm working on drivers for three products that share a single transport > bus with two or more demodulators or other pieces of silicon. > > A good example of this is the hvr1300. > http://www.linuxtv.org/hg/~stoth/hvr1300. > > Other examples are the HVR3000 (which has a cx22702 and cx24123 sharing) > and HVR4000 products which has a cx22702 and cx24118 sharing. > > With the hvr1300 device, the transport bus is shared between a MPEG2 > hardware encoder and a cx22702. Rather than add specific code to the > cx22702 demodulator to use a series of callbacks back into the cx88 > framework, to toggle GPIOs and request hardware access, it started to > make more sense to add this as part of the generic frontend framework > and not pollute the demodulator drivers. > > The following patch to dvb-frontend.c|h uses a callback during open and > release in dvb_frontend to determine whether the request should succeed. > > I'm also showing you a typical implementation within the cx88 framework. > We override the fe->ops function with a generic framework callback in > the cx88 source to deal with the specifics for this device. (much of the > hardware arbitration is not shown here but can be seen in > hg/~stoth/hvr1300 for those who are interested). > > A minor patch like this to dvb-frontend.c|h would allow dvb frameworks > (cx88/budget etc) to determine whether hardware was technically free for > use, or make it available if required. > > Is this type of changing likely to cause any unforeseen problems? > > Does anyone object to this approach? Sounds good to me; I was wondering how to go about this myself for a saa7134 based card actually. _______________________________________________ linux-dvb mailing list linux-dvb@xxxxxxxxxxx http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
