On Fri, Jun 17, 2022 at 01:42:42AM -0700, Christoph Hellwig wrote:

On Thu, Jun 16, 2022 at 04:52:09PM -0700, Nicolin Chen wrote:

...

• ret = vfio_unpin_pages(&vgpu->vfio_device, gfn << PAGE_SHIFT, npage);
• drm_WARN_ON(&i915->drm, ret != npage);

The shifting of gfn seems to happen bother here and in the callers.

Also this is the only caller that does anything withthe vfio_unpin_pages return value. Given that you touch the API here we might as well not return any value, and turn the debug checks that can return errors into WARN_ON_ONCE calls the vfio/iommu_type1 code.

Thanks for the suggestion. I will do that.

...

+extern int vfio_pin_pages(struct vfio_device *device, dma_addr_t iova, int npage, int prot, unsigned long *phys_pfn); -extern int vfio_unpin_pages(struct vfio_device *device, unsigned long *user_pfn, +extern int vfio_unpin_pages(struct vfio_device *device, dma_addr_t iova, int npage);

This will clash with the extern removal patch that Alex has sent.

And I will rebase on top of Alex's change.