Re: [PATCH v2] iommu/vt-d: Fix a bug in intel_iommu_iova_to_phys() for huge page

From: Lu Baolu
Date: Tue Feb 25 2020 - 00:53:15 EST


Hi,

On 2020/2/21 3:44, Yonghyun Hwang wrote:
intel_iommu_iova_to_phys() has a bug when it translates an IOVA for a huge
page onto its corresponding physical address. This commit fixes the bug by
accomodating the level of page entry for the IOVA and adds IOVA's lower
address to the physical address.

Signed-off-by: Yonghyun Hwang <yonghyun@xxxxxxxxxx>
---

Changes from v1:
- level cannot be 0. So, the condition, "if (level > 1)", is removed, which results in a simple code.
- a macro, BIT_MASK, is used to have a bit mask

---
drivers/iommu/intel-iommu.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index 932267f49f9a..4fd5c6287b6d 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -5554,7 +5554,9 @@ static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
pte = pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT, &level);
if (pte)

Need to check whether the pte is present here. Otherwise, the returned
level makes no sense.

Increased change looks like:

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index 61c7ef2f33b4..33593fea0250 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -5700,7 +5700,7 @@ static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
u64 phys = 0;

pte = pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT, &level);
- if (pte)
+ if (pte && dma_pte_present(pte))
phys = dma_pte_addr(pte) +
(iova & (BIT_MASK(level_to_offset_bits(level) +
VTD_PAGE_SHIFT) - 1));

- phys = dma_pte_addr(pte);
+ phys = dma_pte_addr(pte) +
+ (iova & (BIT_MASK(level_to_offset_bits(level) +
+ VTD_PAGE_SHIFT) - 1));
return phys;
}


Best regards,
baolu