It is better for a device model to use IOVAs to read/write memory. And because the rw operations come from CPUs, it is not necessary to call vfio_pin_pages() to pin those pages. patch 1 introduces interface vfio_dma_rw in vfio to read/write IOVAs without pinning user space pages. patch 2 let gvt switch from kvm side rw interface to vfio_dma_rw. v2 changelog: - rename vfio_iova_rw to vfio_dma_rw, vfio iommu driver ops .iova_rw to .dma_rw. (Alex). - change iova and len from unsigned long to dma_addr_t and size_t, respectively. (Alex) - fix possible overflow in dma->vaddr + iova - dma->iova + offset (Alex) - split DMAs from on page boundary to on max available size to eliminate redundant searching of vfio_dma and switching mm. (Alex) - add a check for IOMMU_WRITE permission. Yan Zhao (2): vfio: introduce vfio_dma_rw to read/write a range of IOVAs drm/i915/gvt: subsitute kvm_read/write_guest with vfio_dma_rw drivers/gpu/drm/i915/gvt/kvmgt.c | 26 +++-------- drivers/vfio/vfio.c | 45 +++++++++++++++++++ drivers/vfio/vfio_iommu_type1.c | 76 ++++++++++++++++++++++++++++++++ include/linux/vfio.h | 5 +++ 4 files changed, 133 insertions(+), 19 deletions(-) -- 2.17.1
