AMD/VIA: v4l and adi conflict update

[Scott Bronson <bronson@rinspin.com>]

Last week I described the conflict between v4l and the adi module
I was seeing.  Since then, Oliver Neukum and I have traced it to the
K7 optimizations--if I optimize the stock 2.4.7 kernel for PII, the
problem goes away.  This is 100% reproducible.

Can this provide any insight into the ongoing VIA problems?  If
anyone has anything they'd like me to test (Alan?), please tell me.

I'm running a VIA Apollo KX133-based motherboard (700 MHz Athlon on
an Epox EP-7KXA).  The problem that I am tracking, real briefly, is if
if I use v4l (either a bttv-based TV tuner or a webcam) before using
the joystick after I boot the computer, the joystick fails to respond.
However, if I use the joystick first, then everything works as it should.

	- Scott


For completeness:
In a different converstaion, Oliver Neukum wrote:
>changing the compiler is unlikely to work, as most K7 optimisations are 
>assembly.
>
>However these problems are probably related to VIA chipsets. Your symptoms 
>are new. There is a chance that a report to the kernel mailing list and Alan 
>Cox might be valuable to those developing a workaround.
>
>There is a good chance that there's a bug in the DMA handling in the VIA 
>chipsets. It might be triggered by the picture transfers.
>Please report this.