On Tue, 6 Dec 2022 13:55:46 -0800 Steve Sistare <steven.sistare@xxxxxxxxxx> wrote: > Delete the interfaces that allow an iova range to be re-mapped in a new > address space. They allow userland to indefinitely block vfio mediated > device kernel threads, and do not propagate the locked_vm count to a > new mm. > > - disable the VFIO_UPDATE_VADDR extension > - delete VFIO_DMA_UNMAP_FLAG_VADDR > - delete most of VFIO_DMA_MAP_FLAG_VADDR (but keep some for use in a > new implementation in a subsequent patch). > > Revert most of the code of these commits: > > 441e810 ("vfio: interfaces to update vaddr") > c3cbab2 ("vfio/type1: implement interfaces to update vaddr") > 898b9ea ("vfio/type1: block on invalid vaddr") > > Revert these commits. They are harmless, but no longer used after the > above are reverted, and this kind of functionality is better handled by > adding new methods to vfio_iommu_driver_ops. > > ec5e329 ("vfio: iommu driver notify callback") > 487ace1 ("vfio/type1: implement notify callback") > > Signed-off-by: Steve Sistare <steven.sistare@xxxxxxxxxx> > --- > drivers/vfio/container.c | 5 -- > drivers/vfio/vfio.h | 7 -- > drivers/vfio/vfio_iommu_type1.c | 144 ++-------------------------------------- > include/uapi/linux/vfio.h | 17 +---- > 4 files changed, 8 insertions(+), 165 deletions(-) Picked just this patch and applied to the vfio next branch w/ my follow-up patch to clean out the remainder of the VADDR code and mark feature and flags as deprecated. Added stable cc for both. Thanks, Alex