[linux-dvb] [RFC] Hybrid tuner refactoring, phase 1

Michael Krufky mkrufky at linuxtv.org
Mon Aug 20 20:25:51 CEST 2007


Mauro Carvalho Chehab wrote:
> Em Seg, 2007-08-20 às 11:22 -0400, Michael Krufky escreveu:
>   
>> Manu Abraham wrote:
>>     
>>> You will need to handle the IOCTL calls for analog operations some
>>> place, whatever you remove.
>>>
>>> I don't see at any place you are handling the IOCTL's directly in the
>>> drivers. So you will be using callbacks from the place where you
>>> are applying the ioctl's
>>>       
>> You are getting ahead of things, Manu.  The tuner.ko module will 
>> continue to remain as the v4l2 / i2c_client interface to the tuning 
>> functionality.
>>     
>
> For now, at a first glance, the way it was mapped seems fine. The way
> the struct is described, a frontend can handle digital and/or analog
> tuning. This seems OK. 
>
> Using void* priv will just create some troubles, since this will avoid
> gcc to do typecast checkings for no gain.
>
>   
>>   We do not (yet) have to add these ioctl calls to the 
>> card-level drivers, as they will continue to interact with the various 
>> tuner modules via tuner.ko's i2c_client interface.
>>
>> I do have plans to remove this tuner.ko interface in the future, but 
>> that is way way in the future, after most of my planned tuner 
>> refactoring work is complete.  There will be a completely separate RFC 
>> issued at that time.  When that happens, the card level drivers will 
>> include dvb_frontend.h, just as the dvb card drivers do, and access the 
>> tuning methods that way, through the dvb_frontend structure.
>>     
>
> Removing the common ioctl handling code done by tuner.ko doesn't seem to
> be nice, since this would generate overhead on other drivers that will
> need to absorb the common tuner handling and probing. However, as
> pointed, this should be discussed later, after finishing the tuner
> refactoring.
>   
The plans that I have for this is an overall cleanup, and will be a 
significant improvement in code readability and handling.  Meanwhile, I 
do not plan on introducing this idea for many months, probably for 
2.6.25 or 2.6.26, maybe 2.6.27... so I would rather not discuss this now 
-- it is completely irrelevant to the matter at hand.
> Michael,
>
> I didn't reviewed yet the tea5767 changes. I expect to do it later this
> week (maybe today night).
Thank you -- I appreciate that.

When I did my testing of tuner-simple and tda8290 modules, I ran into 
some OOPS as a result of some simple issue, easily resolved.  As I hit 
those issues, I was able to take care that the problem would be 
prevented in all new modules...  If you get an OOPS in your testing of 
tea5767, just send me the dmesg output -- I'll probably be able to fix 
it immediately, although I doubt there will be any problem.  Thanks in 
advance for the testing.

Cheers,

Mike





More information about the linux-dvb mailing list