> 	If there's no way to make this work with the bus_dma architecure we
> have then it's time for bus_dma v2. This isn't all that strange a
> situation, and having to go uncached would be a real shame.

Either I misunderstood the original problem or this situation should be
realy strange. If we were talking about i386 here, everyone would yell
"broken PC hardware design".

Where cache hardware isn't able to cooperate with a foreign DMA master, it
has to pay it's price by not caching the DMA target memory, IMHO.

But, of course, if there is a clean way (maybe by extending bus_dma) to
provide the necessary hints to the driver and not breaking or slow down
other archs using that driver, why not do that. All I said was "I don't see
a way", which means absolutely nothing about the possible existance of such
a way in this context!


Martin