[linux-dvb] Rationalisation of /dev/adapterX/caY devices

Georg Acher acher at in.tum.de
Tue Apr 11 15:44:32 CEST 2006

On Tue, Apr 11, 2006 at 02:46:45PM +0200, Johannes Stezenbach wrote:
> > The existing ones are fine, since (the only ones that are there now does 
> > chain the slots, no special controller (CPLD) also in some cases) But 
> > cards will be coming out with multiple frontends and ca devices with 
> > routable logic on one PCI adapter (a CPLD which does the stream control).

There is already HW out, but it's no PCI adapter. The Reelbox uses such a HW
scheme. There is a 14x14 data crossbar/matrix to connect any data source (4
frontends, 4 pid filters, 4 CAs, TS from the CPU) to any data sink
(pid-filter, CA, DMA). Looks a bit oversized, but the FPGA implementation is
not much more complicated than fixed paths with synchronisation stages.

> > In this case, we can route the TS from one frontend to the another CA 
> > device etc, functionality depends on the complexity of the CPLD/FPGA.
> So what you need is an API to route the TS between
> frontend, CI and demux.

The RB-driver uses a separate matrix control ioctl to work around this missing
feature in the API. It is assumed in the driver that the
frontend/pid-filter/DMA path is fixed. Only the CAs are inserted or removed,
they can also be chained.

Unfortunately, vdr also has no knowledge of the freely usable CAMs, so the
matrix settings are done via an additional plugin. It works like softcam in
finding out the decoding capabilities and then inserts the CAM in the TS
stream if needed.

         Georg Acher, acher at in.tum.de
         "Oh no, not again !" The bowl of petunias

More information about the linux-dvb mailing list