Hi Dave,

> So I
> guess I'm saying not to take that comment at face value... while
> reducing the number of buffers to 2 may have fixed the problem, it's
> worth investigating _why_ it fixed the problem. It may turn out that
> the problem has nothing to do with noncontiguous DMA buffers.

To be honest, I was concerned/afraid someone would say something like
this. Given that it appears to be linked to pages, it would not surprise
me that it ends up pointing to something deeper being problematic. I was
just hoping that it was simple...

tim