[ofa-general] Re: mthca use of dma_sync_single is bogus

[Lukas Hejtmanek]

On Tue, Jul 10, 2007 at 12:08:43PM -0700, Roland Dreier wrote:
>  > I think he spoke about coherent, didn't he? On i386/x86_64, the consistent and
>  > coherent are the same but on some architectures they are not and I think that
>  > using consistent (in particular pci_alloc_consistent) is exactly what should
>  > be used. Keir also recommended to use this one.
> 
> coherent and consistent are synonyms.  It's confusing because there is
> pci_alloc_consistent(), which is in general just a wrapper for
> dma_alloc_coherent().

According to DMA-mapping.txt they are not. Alpha, M68000 wihtout MMU, PPC,
Sparc, Sparc64, V850 have own implementation of pci_alloc_consistent().

Yes, on i386, the pci_alloc_consistent() is just wrapper for
dma_alloc_coherent().

>  > And moreover, it avoids using swiotlb and bounce buffers, I think. Am I right,
>  > Keir?
> 
> Yes, but I'm not really willing to make things worse for standard i386
> just to make Xen work a little better.

So, what about some #ifdefs ? E.g., allow config option - Xen optimizations?

-- 
Lukáš Hejtmánek