[PATCH] KVM: Return the actual unmapped size in intel_iommu_unmap()

["Wu, Feng" <feng.wu@xxxxxxxxx>]

Actual unmapped size should be returned by intel_iommu_unmap(), because
iommu_map() which calls this function depends on the real unmapped size.
However, in the current logic, the return value of intel_iommu_unmap()
is far smaller than the actual unmapped size, which leads a lot of
redundant calls to intel_iommu_unmap() in iommu_map(). Since dma_pte_clear_range()
can always unmap the space from "start_pfn" to "last_pfn" successfully,
it is okay to return "size" for intel_iommu_unmap().

Signed-off-by: Feng Wu <feng.wu@xxxxxxxxx>
---
 drivers/iommu/intel-iommu.c |    5 ++---
 1 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index 15e9b57..bb795d5 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -4113,15 +4113,14 @@ static size_t intel_iommu_unmap(struct iommu_domain *domain,
                             unsigned long iova, size_t size)
 {
        struct dmar_domain *dmar_domain = domain->priv;
-       int order;

-       order = dma_pte_clear_range(dmar_domain, iova >> VTD_PAGE_SHIFT,
+       dma_pte_clear_range(dmar_domain, iova >> VTD_PAGE_SHIFT,
                            (iova + size - 1) >> VTD_PAGE_SHIFT);

        if (dmar_domain->max_addr == iova + size)
                dmar_domain->max_addr = iova;

-       return PAGE_SIZE << order;
+       return size;
 }

 static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
--
1.7.1

BTW: Here is the only place where the return value of dma_pte_clear_range() is used, if we don't use it here neither, maybe we can make it a void function.

Attachment: 0001-KVM-Return-the-actual-unmapped-size-in-intel_iommu_u.patch
Description: 0001-KVM-Return-the-actual-unmapped-size-in-intel_iommu_u.patch