On Wed, Jul 17, 2024 at 06:32:05PM +0200, David Hildenbrand wrote: > On 17.07.24 18:30, Peter Xu wrote: > > On Wed, Jul 17, 2024 at 04:17:12PM +0200, David Hildenbrand wrote: > > > I'd be curious how a previous truncation on a file can tear down a PFNMAP > > > mapping -- and if there are simple ways to forbid that (if possible). > > > > Unfortunately, forbiding it may not work for vfio, as vfio would like to > > allow none (or partial) mapping on the bars.. at least so far that's the > > plan. > > I think vfio should handle that memtype reservation manually, as you > proposed. So that shouldn't block that IIUC. Oh, so maybe I misunderstood here. As long as we can live with VFIO's vma (with the pfn tracking on top, but without pgtable installed) I think it should be fine. Thanks, -- Peter Xu
