Re: SCSI breakage on non-cache coherent architectures

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Mon, 2007-11-19 at 13:43 -0800, Roland Dreier wrote:
> > I've been debugging various issues on the PowerPC 44x embedded
>  > architecture which happens to have non-coherent PCI DMA.
>  > 
>  > One of the problem I'm hitting is that one really need to enforce
>  > kmalloc alignement to cache lines or bad things will happen (among
>  > others with USB), for some reasons, powerpc failed to do so, I fixed it.
> 
> Heh... I hit the same problem literally 5 years ago:
>     http://lwn.net/Articles/1783/
> 
> I implemented the __dma_buffer annotation:
>     http://lwn.net/Articles/2269/
> 
> But DaveM said we should just use the PCI pool code instead:
>     http://lwn.net/Articles/2270/

Heh, well... 

In this case, DaveM just proposed something akin to your
__dma_buffer :-)

On the other hand, after discussing with James, it looks like we'll just
be reverting the patch that removed the kmalloc of the sense buffer
since non cache-coherent archs are supposed to enforce kmalloc alignment
to cache lines.

__dma_buffer still seems like a good thing to have if too many other
drivers are hitting that but for this specific problem, it's not the
approach that we'll be taking.

Cheers,
Ben.