Re: [PATCH v3 2/2] vfio/iommu_type1: Fix some sanity checks in detach group

[Alex Williamson]

On Fri, 22 Jan 2021 17:26:35 +0800
Keqian Zhu <zhukeqian1@xxxxxxxxxx> wrote:

> vfio_sanity_check_pfn_list() is used to check whether pfn_list and
> notifier are empty when remove the external domain, so it makes a
> wrong assumption that only external domain will use the pinning
> interface.
> 
> Now we apply the pfn_list check when a vfio_dma is removed and apply
> the notifier check when all domains are removed.
> 
> Fixes: a54eb55045ae ("vfio iommu type1: Add support for mediated devices")
> Signed-off-by: Keqian Zhu <zhukeqian1@xxxxxxxxxx>
> ---
>  drivers/vfio/vfio_iommu_type1.c | 33 ++++++++++-----------------------
>  1 file changed, 10 insertions(+), 23 deletions(-)
> 
> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> index 161725395f2f..d8c10f508321 100644
> --- a/drivers/vfio/vfio_iommu_type1.c
> +++ b/drivers/vfio/vfio_iommu_type1.c
> @@ -957,6 +957,7 @@ static long vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma,
>  
>  static void vfio_remove_dma(struct vfio_iommu *iommu, struct vfio_dma *dma)
>  {
> +	WARN_ON(!RB_EMPTY_ROOT(&dma->pfn_list));
>  	vfio_unmap_unpin(iommu, dma, true);
>  	vfio_unlink_dma(iommu, dma);
>  	put_task_struct(dma->task);
> @@ -2250,23 +2251,6 @@ static void vfio_iommu_unmap_unpin_reaccount(struct vfio_iommu *iommu)
>  	}
>  }
>  
> -static void vfio_sanity_check_pfn_list(struct vfio_iommu *iommu)
> -{
> -	struct rb_node *n;
> -
> -	n = rb_first(&iommu->dma_list);
> -	for (; n; n = rb_next(n)) {
> -		struct vfio_dma *dma;
> -
> -		dma = rb_entry(n, struct vfio_dma, node);
> -
> -		if (WARN_ON(!RB_EMPTY_ROOT(&dma->pfn_list)))
> -			break;
> -	}
> -	/* mdev vendor driver must unregister notifier */
> -	WARN_ON(iommu->notifier.head);
> -}
> -
>  /*
>   * Called when a domain is removed in detach. It is possible that
>   * the removed domain decided the iova aperture window. Modify the
> @@ -2366,10 +2350,10 @@ static void vfio_iommu_type1_detach_group(void *iommu_data,
>  			kfree(group);
>  
>  			if (list_empty(&iommu->external_domain->group_list)) {
> -				vfio_sanity_check_pfn_list(iommu);
> -
> -				if (!IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu))
> +				if (!IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu)) {
> +					WARN_ON(iommu->notifier.head);
>  					vfio_iommu_unmap_unpin_all(iommu);
> +				}
>  
>  				kfree(iommu->external_domain);
>  				iommu->external_domain = NULL;
> @@ -2403,10 +2387,12 @@ static void vfio_iommu_type1_detach_group(void *iommu_data,
>  		 */
>  		if (list_empty(&domain->group_list)) {
>  			if (list_is_singular(&iommu->domain_list)) {
> -				if (!iommu->external_domain)
> +				if (!iommu->external_domain) {
> +					WARN_ON(iommu->notifier.head);
>  					vfio_iommu_unmap_unpin_all(iommu);
> -				else
> +				} else {
>  					vfio_iommu_unmap_unpin_reaccount(iommu);
> +				}
>  			}
>  			iommu_domain_free(domain->domain);
>  			list_del(&domain->next);
> @@ -2488,9 +2474,10 @@ static void vfio_iommu_type1_release(void *iommu_data)
>  	struct vfio_iommu *iommu = iommu_data;
>  	struct vfio_domain *domain, *domain_tmp;
>  
> +	WARN_ON(iommu->notifier.head);

I don't see that this does any harm, but isn't it actually redundant?
It seems vfio-core only calls the iommu backend release function after
removing all the groups, so the tests in _detach_group should catch all
cases.  We're expecting the vfio bus/mdev driver to remove the notifier
when a device is closed, which necessarily occurs before detaching the
group.  Thanks,

Alex

> +
>  	if (iommu->external_domain) {
>  		vfio_release_domain(iommu->external_domain, true);
> -		vfio_sanity_check_pfn_list(iommu);
>  		kfree(iommu->external_domain);
>  	}
>