On 20.02.20 17:31, Christoph Hellwig wrote:

On Thu, Feb 20, 2020 at 05:23:20PM +0100, Christian Borntraeger wrote:

quoted

quoted

From a users perspective it makes absolutely perfect sense to use the

bounce buffers when they are NEEDED. 
Forcing the user to specify iommu_platform just because you need bounce buffers
really feels wrong. And obviously we have a severe performance issue
because of the indirections.

The point is that the user should not have to specify iommu_platform.

I (and Halil) agree on that. From a user perspective we want to
have the right thing in the guest without any command option. The iommu_plaform
thing was indeed a first step to make things work. 

I might mis-read Halils patch, but isnt one effect of this patch set 
that things WILL work without iommu_platform?

We need to make sure any new hypervisor (especially one that might require
bounce buffering) always sets it, as was a rather bogus legacy hack
that isn't extensibe for cases that for example require bounce buffering.