On Wed, Oct 16, 2024 at 11:16:08PM -0700, Christoph Hellwig wrote: > On Wed, Oct 16, 2024 at 01:41:51PM -0400, Michael S. Tsirkin wrote: > > It's basically because vfio does, so we have to follow suit. > > That's a very bold argument, especially without any rationale of > > a) why you need to match the feature set
Because people want to move from some vendor specific solution with vfio to a standard vdpa compatible one with vdpa. We could just block them since we don't support tainted kernels anyway, but it's a step in the right direction since it allows moving to virtio, and the kernel is tained so no big support costs (although qe costs do exist, as with any code). > b) how even adding it to vfio was agood idea That ship has sailed. -- MST
