diff mbox

iommu: arm-smmu: set a more appropriate DMA mask

Message ID 2ad6f0a7d198d607e252c420fef52d95dbe048c6.1425581413.git.robin.murphy@arm.com (mailing list archive)
State New, archived
Headers show

Commit Message

Robin Murphy March 5, 2015, 6:56 p.m. UTC
Since we use dma_map_page() as an architecture-independent means of
making page table updates visible to non-coherent SMMUs, we need to
have a suitable DMA mask set to discourage the DMA mapping layer from
creating bounce buffers and flushing those instead, if said page tables
happen to lie outside the default 32-bit mask.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
---
 drivers/iommu/arm-smmu.c | 7 +++++++
 1 file changed, 7 insertions(+)

Comments

Mitchel Humpherys March 5, 2015, 9:20 p.m. UTC | #1
On Thu, Mar 05 2015 at 10:56:36 AM, Robin Murphy <robin.murphy@arm.com> wrote:
> Since we use dma_map_page() as an architecture-independent means of
> making page table updates visible to non-coherent SMMUs, we need to
> have a suitable DMA mask set to discourage the DMA mapping layer from
> creating bounce buffers and flushing those instead, if said page tables
> happen to lie outside the default 32-bit mask.
>
> Signed-off-by: Robin Murphy <robin.murphy@arm.com>

Tested-by: Mitchel Humpherys <mitchelh@codeaurora.org>

Thanks Robin!


-Mitch
Arnd Bergmann March 9, 2015, 9:27 p.m. UTC | #2
On Thursday 05 March 2015 18:56:36 Robin Murphy wrote:
> @@ -1630,6 +1630,13 @@ static int arm_smmu_device_cfg_probe(struct arm_smmu_device *smmu)
>         size = arm_smmu_id_size_to_bits((id >> ID2_OAS_SHIFT) & ID2_OAS_MASK);
>         smmu->pa_size = size;
>  
> +       /*
> +        * What the page table walker can address actually depends on which
> +        * descriptor format is in use, but since a) we don't know that yet,
> +        * and b) it can vary per context bank, this will have to do...
> +        */
> +       dma_set_mask_and_coherent(smmu->dev, DMA_BIT_MASK(size));
> +
>         if (smmu->version == ARM_SMMU_V1) {
>                 smmu->va_size = smmu->ipa_size;
>                 size = SZ_4K | SZ_2M | SZ_1G;
> 

dma_set_mask_and_coherent() can fail if the parent bus cannot access all
of RAM. While that would be serious misdesign for an IOMMU, you should not
just ignore that failure and at least print a message once.

Why not just attempt to set a 64-bit mask?

	Arnd
diff mbox

Patch

diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
index fc13dd5..dc0ae62 100644
--- a/drivers/iommu/arm-smmu.c
+++ b/drivers/iommu/arm-smmu.c
@@ -1630,6 +1630,13 @@  static int arm_smmu_device_cfg_probe(struct arm_smmu_device *smmu)
 	size = arm_smmu_id_size_to_bits((id >> ID2_OAS_SHIFT) & ID2_OAS_MASK);
 	smmu->pa_size = size;
 
+	/*
+	 * What the page table walker can address actually depends on which
+	 * descriptor format is in use, but since a) we don't know that yet,
+	 * and b) it can vary per context bank, this will have to do...
+	 */
+	dma_set_mask_and_coherent(smmu->dev, DMA_BIT_MASK(size));
+
 	if (smmu->version == ARM_SMMU_V1) {
 		smmu->va_size = smmu->ipa_size;
 		size = SZ_4K | SZ_2M | SZ_1G;