[linux-dvb] Re: refactoring

[Holger Waechtler <holger@qanu.de>]

Ralph Metzler wrote:

> Holger Waechtler writes:
> > Gerd Knorr wrote:
> > 
> > >
> > >>I have left it using the kernel i2c_adapter stuff, and i2c_transfer() 
> > >>functions as before.
> > >>    
> > >>
> > >
> > >Which is a bad idea.
> > >
> > 
> > why? please elaborate.
> > 
> > >  Have you actually looked at my cx22702 version?
> > >There is much generic, reusable code in there.  And it is generic
> > >*because* it makes heavy use of the kernl i2c infrastructure.
> > >
> > 
> > no. that's just the reason why it's not portable, not testable in a 
> > userspace mockup and hardly portable to other OSes like MacOS-X and BSD. 
> > The new approach without dependencies to kernel APIs is much nicer.
>
> You could completely rewrite lots of drivers (and not for the better)
> with this argument. Of course Linux kernel APIs primarily make it
> easier to write drivers just for Linux. So, I should not use any
> Linux kernel APIs because they are not in userspace, MacOS-X and BSD?
>  
You should try to encapsulate them at lease so that porting is made 
easy. You know that there have been several approaches in the past to 
port the linux-dvb code to BSD, MacOS-X and even the GPL'd Windows BDA 
drivers on sourceforge mention the linux driver as codebase:

http://cvs.sourceforge.net/viewcvs.py/bdadev/NovaTDriver/saa7146.cpp?rev=1.11&view=auto

btw: the BDA drivers are a very nice example how to encapsulate your 
code away from OS-specific APIs: please take a look on the l64781 or 
SAA7146 implementations, they are much better organized, simpler and 
easier to maintain that their linux parents:

I think it's a good thing to inherit some of the ideas and maybe we even 
reach one day a point where a common 'hardware library' codebase can 
used for all platforms and the OS-specific PCI/USB/BDA/linux-DVB API 
registration functions are collected in a single file as small and 
compact as possible.

> > Not allocating dynamic memory but using static data 
> > instead is always a good thing.
>
> Why?
> Now each and every driver has to take care of it instead of just the
> demodulator driver.
>  
no. Please have a look:

struct budget_with_stv0299 {
   struct pci_device pcidev;
   struct dma_related_stuff dma;
   struct stv0299_state_and_config stv0299;
   /* ans so on... */
};

this card struct gets allocated once in the pci attach() can and freed 
in the release() callback. No need for additional maintenance, kmalloc() 
and kfree() anywhere.

Holger