[linux-dvb] [RFC-final] videobuf tree

Michael Krufky mkrufky at linuxtv.org
Mon Oct 8 14:47:57 CEST 2007


Mauro Carvalho Chehab wrote:
>> Mauro,
>>
>> This new patch fixed the problem.  CX23885 functionality is restored!  :-)
> Good! If you send your reviewed-by, I'll add at the proper changesets
> touching videobuf.

3762b92e232a - V4L/DVB (6287) - Fix DMA Scatter/Gather constructor

Reviewed-by: Michael Krufky <mkrufky at linuxtv.org>

>> side note:  If we had left a single header, video-buf.h, we could have
>> avoided this problem.  When we rename files like this, we run the risk
>> of building against the wrong headers, if errors are not caught &
>> corrected quickly enough.
>>
>> Are you open to merging the videobuf-*.h into a single video-buf.h
>> header file, to match the filename in previous kernel versions so that
>> we can avoid this issue from recurring?  Either that, or including the
>> current headers into a new video-buf.h would do the same job.
>>
>> What do you think?
> 
> I don't like to create a video-buf.h header. This will make non-pci
> devices dependent on PCI, or will require some additional logic for
> checking kernel Kconfig symbols. I also expect that other newer videobuf
> methods to be created. So, this header will just generate undesirable
> mess.

This would not create dependencies of non-pci devices on pci -- a simple header inclusion is optimized by the c compiler -- only needed methods are considered and are actually included in the build.

> What we might do is to rename the generic abstract method to another
> name. This will generate compilation errors, making easier for people to
> realize what happened.

If we rename it to video-buf.h, it would cover the issue that I have in mind.

> I'm not sure if this is valuable enough, since I don't know any other
> DMA S/G driver using videobuf being developed for their inclusion at
> Kernel.

Maybe an out of tree driver uses it?  Maybe there is an in-tree driver that still might have old methods being used but we didnt hit those bugs yet due to incomplete testing methods?

Cheers,

Mike



More information about the linux-dvb mailing list