[PATCH v4] iommu/amd: Add support for fast IOTLB flushing

[PATCH v4] iommu/amd: Add support for fast IOTLB flushing

[Suravee Suthikulpanit]

Since AMD IOMMU driver currently flushes all TLB entries
when page size is more than one, use the same interface
for both iommu_ops.flush_iotlb_all() and iommu_ops.iotlb_sync().

Cc: Joerg Roedel <joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org>
Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit-5C7GfCeVMHo@public.gmane.org>
---
Changes from v3 (https://patchwork.kernel.org/patch/10193235)
 * Change amd_iommu_iotlb_range_add() to no-op and iotlb_sync()
   to full domain flush for now since we currently flush all entries
   when the page size is more than one.
 * Fine-grained invalidation will be introduced in subsequent
   patch series.

 drivers/iommu/amd_iommu.c | 19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

diff --git a/drivers/iommu/amd_iommu.c b/drivers/iommu/amd_iommu.c
index fed8059..6061a8d 100644
--- a/drivers/iommu/amd_iommu.c
+++ b/drivers/iommu/amd_iommu.c
@@ -3043,9 +3043,6 @@ static size_t amd_iommu_unmap(struct iommu_domain *dom, unsigned long iova,
 	unmap_size = iommu_unmap_page(domain, iova, page_size);
 	mutex_unlock(&domain->api_lock);
 
-	domain_flush_tlb_pde(domain);
-	domain_flush_complete(domain);
-
 	return unmap_size;
 }
 
@@ -3163,6 +3160,19 @@ static bool amd_iommu_is_attach_deferred(struct iommu_domain *domain,
 	return dev_data->defer_attach;
 }
 
+static void amd_iommu_flush_iotlb_all(struct iommu_domain *domain)
+{
+	struct protection_domain *dom = to_pdomain(domain);
+
+	domain_flush_tlb_pde(dom);
+	domain_flush_complete(dom);
+}
+
+static void amd_iommu_iotlb_range_add(struct iommu_domain *domain,
+				      unsigned long iova, size_t size)
+{
+}
+
 const struct iommu_ops amd_iommu_ops = {
 	.capable = amd_iommu_capable,
 	.domain_alloc = amd_iommu_domain_alloc,
@@ -3181,6 +3191,9 @@ static bool amd_iommu_is_attach_deferred(struct iommu_domain *domain,
 	.apply_resv_region = amd_iommu_apply_resv_region,
 	.is_attach_deferred = amd_iommu_is_attach_deferred,
 	.pgsize_bitmap	= AMD_IOMMU_PGSIZES,
+	.flush_iotlb_all = amd_iommu_flush_iotlb_all,
+	.iotlb_range_add = amd_iommu_iotlb_range_add,
+	.iotlb_sync = amd_iommu_flush_iotlb_all,
 };
 
 /*****************************************************************************
-- 
1.8.3.1

Re: [PATCH v4] iommu/amd: Add support for fast IOTLB flushing

[Suravee Suthikulpanit]

Ping..

Joerg, when you get a chance, would you please let me know if you have any other concerns for this v4.

Thanks,
Suravee

On 2/21/18 2:19 PM, Suravee Suthikulpanit wrote:
> Since AMD IOMMU driver currently flushes all TLB entries
> when page size is more than one, use the same interface
> for both iommu_ops.flush_iotlb_all() and iommu_ops.iotlb_sync().
> 
> Cc: Joerg Roedel <joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org>
> Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit-5C7GfCeVMHo@public.gmane.org>
> ---
> Changes from v3 (https://patchwork.kernel.org/patch/10193235)
>   * Change amd_iommu_iotlb_range_add() to no-op and iotlb_sync()
>     to full domain flush for now since we currently flush all entries
>     when the page size is more than one.
>   * Fine-grained invalidation will be introduced in subsequent
>     patch series.
> 
>   drivers/iommu/amd_iommu.c | 19 ++++++++++++++++---
>   1 file changed, 16 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/iommu/amd_iommu.c b/drivers/iommu/amd_iommu.c
> index fed8059..6061a8d 100644
> --- a/drivers/iommu/amd_iommu.c
> +++ b/drivers/iommu/amd_iommu.c
> @@ -3043,9 +3043,6 @@ static size_t amd_iommu_unmap(struct iommu_domain *dom, unsigned long iova,
>   	unmap_size = iommu_unmap_page(domain, iova, page_size);
>   	mutex_unlock(&domain->api_lock);
>   
> -	domain_flush_tlb_pde(domain);
> -	domain_flush_complete(domain);
> -
>   	return unmap_size;
>   }
>   
> @@ -3163,6 +3160,19 @@ static bool amd_iommu_is_attach_deferred(struct iommu_domain *domain,
>   	return dev_data->defer_attach;
>   }
>   
> +static void amd_iommu_flush_iotlb_all(struct iommu_domain *domain)
> +{
> +	struct protection_domain *dom = to_pdomain(domain);
> +
> +	domain_flush_tlb_pde(dom);
> +	domain_flush_complete(dom);
> +}
> +
> +static void amd_iommu_iotlb_range_add(struct iommu_domain *domain,
> +				      unsigned long iova, size_t size)
> +{
> +}
> +
>   const struct iommu_ops amd_iommu_ops = {
>   	.capable = amd_iommu_capable,
>   	.domain_alloc = amd_iommu_domain_alloc,
> @@ -3181,6 +3191,9 @@ static bool amd_iommu_is_attach_deferred(struct iommu_domain *domain,
>   	.apply_resv_region = amd_iommu_apply_resv_region,
>   	.is_attach_deferred = amd_iommu_is_attach_deferred,
>   	.pgsize_bitmap	= AMD_IOMMU_PGSIZES,
> +	.flush_iotlb_all = amd_iommu_flush_iotlb_all,
> +	.iotlb_range_add = amd_iommu_iotlb_range_add,
> +	.iotlb_sync = amd_iommu_flush_iotlb_all,
>   };
>   
>   /*****************************************************************************
> 

Re: [PATCH v4] iommu/amd: Add support for fast IOTLB flushing

[Joerg Roedel]

Hi Suravee,

On Mon, Mar 05, 2018 at 08:08:21AM +0700, Suravee Suthikulpanit wrote:
> Ping..
> 
> Joerg, when you get a chance, would you please let me know if you have any other concerns for this v4.

Sorry for the long delay. The patch is fine, I applied it to the
iommu-tree.

Thanks,

	Joerg