On Fri, 2018-06-15 at 02:16 -0700, Christoph Hellwig wrote:

On Wed, Jun 13, 2018 at 11:11:01PM +1000, Benjamin Herrenschmidt wrote:

quoted

Actually ... the stuff in lib/dma-direct.c seems to be just it, no ?

There's no cache flushing and there's no architecture hooks that I can
see other than the AMD security stuff which is probably fine.

Or am I missing something ?

You are missing the __phys_to_dma arch hook that allows architectures
to adjust the dma address.  Various systems have offsets, or even
multiple banks with different offsets there.  Most of them don't
use the dma-direct code yet (working on it), but there are a few
examples in the tree already.

Ok and on those systems, qemu will bypass said offset ? Maybe we could
just create a device DMA "flag" or (or use the attributes) to instruct
dma-direct to not use that for legacy virtio ?

Cheers,
Ben.