Linux Virtualization - Jun 2018 - [RFC V2] virtio: Add platform specific DMA API translation for virito devices

On Mon, Jun 04, 2018 at 03:43:09PM +0300, Michael S. Tsirkin
wrote:
> Another is that given the basic functionality is in there, optimizations
> can possibly wait until per-device quirks in DMA API are supported.
We have had per-device dma_ops for quite a while.

On Mon, 2018-06-04 at 05:55 -0700, Christoph Hellwig
wrote:
> On Mon, Jun 04, 2018 at 03:43:09PM +0300, Michael S. Tsirkin wrote:
> > Another is that given the basic functionality is in there,
optimizations
> > can possibly wait until per-device quirks in DMA API are supported.
> 
> We have had per-device dma_ops for quite a while.
I've asked Ansuman to start with a patch that converts virtio to use
DMA ops always, along with an init quirk to hookup "direct" ops when
the IOMMU flag isn't set.

This will at least remove that horrid duplication of code path we have
in there.

Then we can just involve the arch in that init quirk so we can chose an
alternate set of ops when running a secure VM.

This is completely orthogonal to whether an iommu exist qemu side or
not, and should be entirely solved on the Linux side.

Cheers,
Ben.

On Mon, Jun 04, 2018 at 11:14:36PM +1000, Benjamin Herrenschmidt
wrote:
> On Mon, 2018-06-04 at 05:55 -0700, Christoph Hellwig wrote:
> > On Mon, Jun 04, 2018 at 03:43:09PM +0300, Michael S. Tsirkin wrote:
> > > Another is that given the basic functionality is in there,
optimizations
> > > can possibly wait until per-device quirks in DMA API are
supported.
> > 
> > We have had per-device dma_ops for quite a while.
> 
> I've asked Ansuman to start with a patch that converts virtio to use
> DMA ops always, along with an init quirk to hookup "direct" ops
when
> the IOMMU flag isn't set.
> 
> This will at least remove that horrid duplication of code path we have
> in there.
> 
> Then we can just involve the arch in that init quirk so we can chose an
> alternate set of ops when running a secure VM.
> 
> This is completely orthogonal to whether an iommu exist qemu side or
> not, and should be entirely solved on the Linux side.
> 
> Cheers,
> Ben.
Sounds good to me.

-- 
MST