On Mon, Oct 14, 2013 at 04:17:51PM +0100, Antonios Motakis wrote:
> On Mon, Oct 14, 2013 at 2:48 PM, Will Deacon <will.deacon@xxxxxxx> wrote:
> > On Fri, Oct 11, 2013 at 02:24:47PM +0100, Antonios Motakis wrote:
> >> The return value of arm_smmu_iova_to_phys is directly passed to the
> >> user of the IOMMU API via iommu_iova_to_phys; however the ARM SMMU
> >> driver returns -EINVAL on error, which is not consistent with the
> >> rest of the drivers implementing the IOMMU API. VFIO also relies on
> >> the call returning NULL when a page has not been mapped already.
> >>
> >> Signed-off-by: Antonios Motakis <a.motakis@xxxxxxxxxxxxxxxxxxxxxx>
> >> ---
> >> drivers/iommu/arm-smmu.c | 5 +----
> >> 1 file changed, 1 insertion(+), 4 deletions(-)
> >>
> >> diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
> >> index 8b71332..fe81b20 100644
> >> --- a/drivers/iommu/arm-smmu.c
> >> +++ b/drivers/iommu/arm-smmu.c
> >> @@ -1480,10 +1480,7 @@ static phys_addr_t arm_smmu_iova_to_phys(struct iommu_domain *domain,
> >>
> >> err_unlock:
> >> spin_unlock(&smmu_domain->lock);
> >> - dev_warn(smmu->dev,
> >> - "invalid (corrupt?) page tables detected for iova 0x%llx\n",
> >> - (unsigned long long)iova);
> >> - return -EINVAL;
> >> + return NULL;
> >
> > Why are you removing the warning message?
>
> VFIO will exercise this code path every time when mapping DMA memory.
> This is normal and VFIO *expects* the function to fail - it is only if
> the function succeeds that VFIO needs to back down from the DMA
> mapping and fail.
>
> This means that there would be a warning every time a VFIO user maps
> some memory for DMA use, even though nothing went wrong.
Ok, in which case it might be worth reworking arm_smmu_iova_to_phys to treat
{pgd,pud,pmd,pte}_none different from {pgd,pud,pmd,pte}_bad.
Will
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
