On 25/04/2022 09:30, Jan Beulich wrote:
> Recent changes (likely 5fafa6cf529a ["AMD/IOMMU: have callers specify
> the target level for page table walks"]) have made Coverity notice a
> shift count in iommu_pde_from_dfn() which might in theory grow too
> large. While this isn't a problem in practice, address the concern
> nevertheless to not leave dangling breakage in case very large
> superpages would be enabled at some point.
>
> Coverity ID: 1504264
>
> While there also address a similar issue in set_iommu_ptes_present().
> It's not clear to me why Coverity hasn't spotted that one.
>
> Signed-off-by: Jan Beulich <jbeul...@suse.com>
> ---
> v4: New.
>
> --- a/xen/drivers/passthrough/amd/iommu_map.c
> +++ b/xen/drivers/passthrough/amd/iommu_map.c
> @@ -89,11 +89,11 @@ static unsigned int set_iommu_ptes_prese
> bool iw, bool ir)
> {
> union amd_iommu_pte *table, *pde;
> - unsigned int page_sz, flush_flags = 0;
> + unsigned long page_sz = 1UL << (PTE_PER_TABLE_SHIFT * (pde_level - 1));
There's an off-by-12 error somewhere here.
Judging by it's use, it should be named mapping_frames (or similar) instead.
With that fixed, Reviewed-by: Andrew Cooper <andrew.coop...@citrix.com>
