Re: [PATCH 1/2] vfio: Introduce bus driver to IOMMU invalidation interface

[Jason Gunthorpe]

On Thu, May 14, 2020 at 10:51:58AM -0600, Alex Williamson wrote:
> @@ -1450,6 +1467,10 @@ static int vfio_pci_zap_and_vma_lock(struct vfio_pci_device *vdev, bool try)
>  
>  				zap_vma_ptes(vma, vma->vm_start,
>  					     vma->vm_end - vma->vm_start);
> +				mutex_unlock(&vdev->vma_lock);
> +				up_read(&mm->mmap_sem);
> +				vfio_invalidate_pfnmap_vma(vdev->group, vma);
> +				goto again;

The vma pointer can't leave the read side of the mmap_sem

> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> index 4a4cb7cd86b2..62ba6bd8a486 100644
> +++ b/drivers/vfio/vfio_iommu_type1.c
> @@ -91,6 +91,7 @@ struct vfio_dma {
>  	bool			lock_cap;	/* capable(CAP_IPC_LOCK) */
>  	struct task_struct	*task;
>  	struct rb_root		pfn_list;	/* Ex-user pinned pfn list */
> +	struct vm_area_struct	*pfnmap_vma;

This is also confusing, how does it prevent pfnmap_vma from becoming
freed?

Jason