[PATCHv16 08/13] DocBook/media: add CEC documentation

[Dmitry Torokhov]

On Fri, Jun 17, 2016 at 08:37:38AM -0300, Mauro Carvalho Chehab wrote:
> Em Fri, 17 Jun 2016 13:09:10 +0200
> Hans Verkuil <hverkuil at xs4all.nl> escreveu:
> 
> > On 06/17/2016 11:50 AM, Mauro Carvalho Chehab wrote:
> > One area where I am uncertain is when remote control messages are received and
> > passed on by the framework to the RC input device.
> > 
> > Suppose the application is the one receiving a password, then that password appears
> > both in the input device and the cec device. What I think will be useful is if the
> > application can prevent the use of an input device to pass on remote control messages.
> > 
> > CEC_ADAP_S_LOG_ADDRS has a flags field that I intended for just that purpose.
> > 
> > Note that RC messages are always passed on to CEC followers even if there is an
> > input device since some RC messages have additional arguments that the rc subsystem
> > can't handle. Also I think that it is often easier to handle all messages from the
> > same CEC device instead of having to read from two devices (cec and input). I
> > actually considered removing the input support, but it turned out to be useful in
> > existing video streaming apps since they don't need to add special cec support to
> > handle remote control presses.
> > 
> > Question: is there a way for applications to get exclusive access to an input device?
> > Or can anyone always read from it?
> 
> That's a very good question. I did a quick test to check how this is
> currently protected, by running:
> 
> $ strace ir-keytable -t
> ...
> open("/dev/input/event12", O_RDONLY)    = -1 EACCES (Permission denied)
> ...
> 
> It turns that the input device was created by udev with those
> permissions:
> 
> crw-rw---- 1 root input 13, 76 Jun 17 08:26 /dev/input/event12
> 
> Changing access to 666 allowed to run ir-keytable -t without the
> need of being root.
> 
> Yet, maybe there's a way to get exclusive access to input/event
> device, but I never needed to go that deep at the input subsystem.
> Maybe Dmitry could shed some light on that. Adding him in the loop.

EVIOCGRAB ioctl will do what you want.

Thanks.

-- 
Dmitry