Re: ir-core multi-protocol decode and mceusb

[Mauro Carvalho Chehab <mchehab@redhat.com>]

Hi Andy,

Em 31-05-2010 09:15, Andy Walls escreveu:
>> I've now got an ir-lirc-codec bridge passing data over to a completely
>> unmodified lirc_dev, with the data then making its way out to the
>> lircd userspace where its even getting properly decoded. I don't have
>> the transmit side of things in ir-lirc-codec wired up just yet, but
>> I'd like to submit what I've got (after some cleanup) tomorrow, and
>> will then incrementally work on transmit. I'm pretty sure wiring up
>> transmit is going to require some of the bits we'd be using for native
>> transmit as well, so there may be some discussion required. Will give
>> a look at setting enabled/disabled decoders tomorrow too, hopefully.
> 
> 
> Since you're looking at Tx, please take a look at the v4l2_subdev
> interface for ir devices.  See 
> 
> linux/include/media/v4l2-subdev.h: struct v4l2_subdev_ir_ops 
> 
> I was wondering how this interface could be modified to interface nicely
> to lirc (or I guess ir-lirc-codec) for transmit functionality.

> Right now, only the cx23885 driver uses it:
> 
> linux/drivers/media/video/cx23885/cx23888-ir.[ch]
> 
> I have the skeleton of transmit for the device implemented (it does need
> some fixing up).
> 
> (The CX23888 hardware is nice in that it only deals with raw pulses so
> one can decode any protocol and transmit any protocols.  The hardware
> provides hardware counter/timers for measuring incoming pulses and
> sending outgoing pulses.)

This interface is bound to V4L needs. As the Remote Controller subsystem
is meant to support not only V4L or DVB IR's, but also other kinds of remote
controllers that aren't associated to media devices, it makes no sense on
binding TX to this interface. 

The biggest advantage of V4L subdev interface is that a command, like VIDIOC_S_STD
could be sent to several devices that may need to know what's the current standard,
in order to configure audio, video, etc. It also provides a nice way to access
devices on a device-internal bus. In the case of RC, I don't see any similar
need. So, IMO, the better is to use an in interface similar to RX for TX, e. g.,
something like:
	rc_register_tx()
	rc_unregister_tx()
	rc_send_code()

Of course, inside the V4L drivers, it may actually make sense to keep using the 
v4l2-subdev if the IR is accessed, for example, via I2C bus.

Cheers,
Mauro.