On 11/02/2011 12:56 PM, Alexander Graf wrote:
Michael S. Tsirkin wrote:
On Tue, Nov 01, 2011 at 03:24:28PM -0500, Anthony Liguori wrote:
What does it get us, applying it before freeze?
It's an api change without new functionality.
Seems better on -next branch.
It gets us that downstreams can convert to the API for 1.0. It just
feels a lot more right to change APIs for a 1.0 release than a 1.1 release.
Also, it gives it more exposure and if anyone is inclined to have a
small patch to add IOMMU support for a specific platform downstream,
they can do so.
Overall, I think it's a good idea to pull it in for 1.0. It certainly
makes David's life easier :). Since we also have PCI support in -M
pseries now we could even declare the non-usage of an IOMMU there as a
bug and fix it for 1.0.1.
I was in violent agreement up until this last sentence ;-)
Having new bidirectional memory APIs is a really big positive infrastructure
change for us. It's been something we've needed for a long time.
Regards,
Anthony Liguori
