On Thu, Jan 8, 2026 at 7:45 AM Alex Williamson <[email protected]> wrote: > > On Thu, 8 Jan 2026 10:10:44 -0400 > Jason Gunthorpe <[email protected]> wrote: > > > On Wed, Jan 07, 2026 at 06:41:10PM -0800, Alex Mastro wrote: > > > On Wed, Jan 07, 2026 at 08:54:06PM -0400, Jason Gunthorpe wrote: > > > > On Wed, Jan 07, 2026 at 11:54:09PM +0000, David Matlack wrote: > > > > > On 2026-01-07 02:13 PM, Alex Mastro wrote: > > > > > > Test MMIO-backed DMA mappings by iommu_map()-ing mmap'ed BAR > > > > > > regions. > > > > > > > > > > Thanks for adding this! > > > > > > > > > > > Also update vfio_pci_bar_map() to align BAR mmaps for efficient huge > > > > > > page mappings. > > > > > > > > > > > > Only vfio_type1 variants are tested; iommufd variants can be added > > > > > > once kernel support lands. > > > > > > > > > > Are there plans to support mapping BARs via virtual address in > > > > > iommufd? > > > > > I thought the plan was only to support via dma-bufs. Maybe Jason can > > > > > confirm. > > > > > > > > Only dmabuf. > > > > > > Ack. I got confused. I had thought iommufd's vfio container compatibility > > > mode > > > was going to support this, but realized that doesn't make sense given past > > > discussions about the pitfalls of achieving these mappings the legacy way. > > > > Oh, I was thinking about a compatability only flow only in the type 1 > > emulation that internally magically converts a VMA to a dmabuf, but I > > haven't written anything.. It is a bit tricky and the type 1 emulation > > has not been as popular as I expected?? > > In part because of this gap, I'd guess. Thanks,
Lack of huge mappings in the IOMMU when using VFIO_TYPE1_IOMMU is another gap I'm aware of. vfio_dma_mapping_test.vfio_type1_iommu_anonymous_hugetlb_1gb.dma_map_unmap fails when IOMMUFD_VFIO_CONTAINER is enabled. Is the plan to address all the gaps so IOMMUFD_VFIO_CONTAINER can be made the default and the type1 code can be dropped from the upstream kernel?
