On Tue, Oct 27, 2020 at 06:38:40PM +0100, Jean-Philippe Brucker wrote: > On Fri, Oct 23, 2020 at 12:47:02PM -0400, Peter Xu wrote: > > On Fri, Oct 23, 2020 at 09:48:58AM +0200, Jean-Philippe Brucker wrote: > > > Arm CPUs and SMMU support 4k, 16k and 64k page sizes. I don't think 16k is > > > used anywhere but some distributions chose 64k (RHEL, I think?), others > > > 4k, so we need to support both. > > > > > > Unfortunately as noted above host64k-guest4k is not possible without > > > adding a negotiation mechanism to virtio-iommu, host VFIO and IOMMU > > > driver. > > > > I see. Then it seems we would still need to support host4k-guest64k. > > > > Maybe for assigned case, we can simply AND all the psize_masks of all the > > vfio > > containers that supported to replace the default psize mask > > (TARGET_PAGE_SIZE) > > without caring about whether it's shrinking or not? Note that current patch > > only update config.psize_mask to the new one, but I think we need to > > calculate > > the subset of all containers rather than a simply update. > > Yes I think an AND is the right operation. We'll return an error if the > resulting mask is 0. Then for hotplug, I think I'll keep the current "best > effort" code from this patch. If necessary we could later add a parameter > to set a default mask and guarantee hotplug success. > > Thanks, > Jean
So I should expect a new version with that? > > Then with the help > > of 39b3b3c9cac1 imho we'll gracefully fail the probe if the psize is not > > suitable anyway, e.g., host64k-guest4k. > > > > Thanks, > > > > -- > > Peter Xu > >
