kern/79117: [iavc] iavc(4) for AVM B1 PCI does not attach

[Hans Petter Selasky]

On Saturday 13 May 2006 17:15, Hellmuth Michaelis wrote:
> Synopsis: [iavc] iavc(4) for AVM B1 PCI does not attach
>
> State-Changed-From-To: open->suspended
> State-Changed-By: hm
> State-Changed-When: Sat May 13 17:14:05 CEST 2006
> State-Changed-Why:
> i cannot reproduce this bug here, and due to the fact
> that c4b is close to hitting the tree i will not put
> too much work into this anymore.

Some comments:

I think that the 'iavc' approach is right. You should not replace it with C4B.  
The CAPI to I4B bridge is straight forward to implement. If you need more 
features, then upgrade I4B. I cannot see why this is any problem at all. 
Maybe Thomas does not understand I4B so good?

This is what I think is best:

  [HW CAPI] <-> [I4B*] <-> [SW CAPI]

*new I4B

Then one can allow the [SW CAPI] interface to be developed, while keeping the 
[HW CAPI] interface compatible with the connected hardware.

Personally I am not convinced about the framework that C4B provides in its 
capi20.h. It does not support recursive CAPI structures, and is therefore not 
very useable, hence many CAPI structures are recursive. That means decoding 
and encoding CAPI structures in CAPI structures is nearly impossible, and 
must be done using very bad coding techniques that can easily lead to buffer 
overflows, and is never going to be upgradeable. Now there is something 
called liblinuxcapi, but that is at first very inefficient, and it does not 
provide any methods to encode and decode custom CAPI structures.

--HPS