iommu/mediatek: Flush IOTLB completely only if domain has been attached

[PATCH RFC] iommu/mediatek: Flush IOTLB completely only if domain has been attached

Posted by Chen-Yu Tsai 10 months, 4 weeks ago

If an IOMMU domain was never attached, it lacks any linkage to the
actual IOMMU hardware. Attempting to do flush_iotlb_all() on it will
result in a NULL pointer dereference. This seems to happen after the
recent IOMMU core rework in v6.4-rc1.

    Unable to handle kernel read from unreadable memory at virtual address 0000000000000018
    Call trace:
     mtk_iommu_flush_iotlb_all+0x20/0x80
     iommu_create_device_direct_mappings.part.0+0x13c/0x230
     iommu_setup_default_domain+0x29c/0x4d0
     iommu_probe_device+0x12c/0x190
     of_iommu_configure+0x140/0x208
     of_dma_configure_id+0x19c/0x3c0
     platform_dma_configure+0x38/0x88
     really_probe+0x78/0x2c0

Check if the "bank" field has been filled in before actually attempting
the IOTLB flush to avoid it. The IOTLB is also flushed when the device
comes out of runtime suspend, so it should have a clean initial state.

Fixes: 08500c43d4f7 ("iommu/mediatek: Adjust the structure")
Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
---

I think this is a valid fix, but I'm not very familiar with the hardware
or the design of the driver. The ARM SMMU drivers seem to do this as well.

 drivers/iommu/mtk_iommu.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
index aecc7d154f28..e93906d6e112 100644
--- a/drivers/iommu/mtk_iommu.c
+++ b/drivers/iommu/mtk_iommu.c
@@ -781,7 +781,8 @@ static void mtk_iommu_flush_iotlb_all(struct iommu_domain *domain)
 {
 	struct mtk_iommu_domain *dom = to_mtk_domain(domain);
 
-	mtk_iommu_tlb_flush_all(dom->bank->parent_data);
+	if (dom->bank)
+		mtk_iommu_tlb_flush_all(dom->bank->parent_data);
 }
 
 static void mtk_iommu_iotlb_sync(struct iommu_domain *domain,
-- 
2.41.0.rc0.172.g3f132b7071-goog

Re: [PATCH RFC] iommu/mediatek: Flush IOTLB completely only if domain has been attached

Posted by Joerg Roedel 10 months, 3 weeks ago

On Fri, May 26, 2023 at 04:53:59PM +0800, Chen-Yu Tsai wrote:
> Fixes: 08500c43d4f7 ("iommu/mediatek: Adjust the structure")
> Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>

Applied for 6.4, thanks.

Re: [PATCH RFC] iommu/mediatek: Flush IOTLB completely only if domain has been attached

Posted by Yong Wu (吴勇) 10 months, 3 weeks ago

On Fri, 2023-05-26 at 16:53 +0800, Chen-Yu Tsai wrote:
>  	 
>  If an IOMMU domain was never attached, it lacks any linkage to the
> actual IOMMU hardware. Attempting to do flush_iotlb_all() on it will
> result in a NULL pointer dereference. This seems to happen after the
> recent IOMMU core rework in v6.4-rc1.
> 
>     Unable to handle kernel read from unreadable memory at virtual
> address 0000000000000018
>     Call trace:
>      mtk_iommu_flush_iotlb_all+0x20/0x80
>      iommu_create_device_direct_mappings.part.0+0x13c/0x230
>      iommu_setup_default_domain+0x29c/0x4d0
>      iommu_probe_device+0x12c/0x190
>      of_iommu_configure+0x140/0x208
>      of_dma_configure_id+0x19c/0x3c0
>      platform_dma_configure+0x38/0x88
>      really_probe+0x78/0x2c0
> 
> Check if the "bank" field has been filled in before actually
> attempting
> the IOTLB flush to avoid it. The IOTLB is also flushed when the
> device
> comes out of runtime suspend, so it should have a clean initial
> state.
> 
> Fixes: 08500c43d4f7 ("iommu/mediatek: Adjust the structure")

The interface "iommu_setup_default_domain" doesn't exist in v6.4-rc1.

This is a fixes for linux-next. And the Fixes tag should be:

152431e4fe7f ("iommu: Do iommu_group_create_direct_mappings() before
attach")

then: 
Reviewed-by: Yong Wu <yong.wu@mediatek.com>

Thanks very much.

> Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>
> ---
> 
> I think this is a valid fix, but I'm not very familiar with the
> hardware
> or the design of the driver. The ARM SMMU drivers seem to do this as
> well.
> 
>  drivers/iommu/mtk_iommu.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
> index aecc7d154f28..e93906d6e112 100644
> --- a/drivers/iommu/mtk_iommu.c
> +++ b/drivers/iommu/mtk_iommu.c
> @@ -781,7 +781,8 @@ static void mtk_iommu_flush_iotlb_all(struct
> iommu_domain *domain)
>  {
>  	struct mtk_iommu_domain *dom = to_mtk_domain(domain);
>  
> -	mtk_iommu_tlb_flush_all(dom->bank->parent_data);
> +	if (dom->bank)
> +		mtk_iommu_tlb_flush_all(dom->bank->parent_data);
>  }
>  
>  static void mtk_iommu_iotlb_sync(struct iommu_domain *domain,
> -- 
> 2.41.0.rc0.172.g3f132b7071-goog

Re: [PATCH RFC] iommu/mediatek: Flush IOTLB completely only if domain has been attached

Posted by AngeloGioacchino Del Regno 10 months, 3 weeks ago

Il 26/05/23 10:53, Chen-Yu Tsai ha scritto:
> If an IOMMU domain was never attached, it lacks any linkage to the
> actual IOMMU hardware. Attempting to do flush_iotlb_all() on it will
> result in a NULL pointer dereference. This seems to happen after the
> recent IOMMU core rework in v6.4-rc1.
> 
>      Unable to handle kernel read from unreadable memory at virtual address 0000000000000018
>      Call trace:
>       mtk_iommu_flush_iotlb_all+0x20/0x80
>       iommu_create_device_direct_mappings.part.0+0x13c/0x230
>       iommu_setup_default_domain+0x29c/0x4d0
>       iommu_probe_device+0x12c/0x190
>       of_iommu_configure+0x140/0x208
>       of_dma_configure_id+0x19c/0x3c0
>       platform_dma_configure+0x38/0x88
>       really_probe+0x78/0x2c0
> 
> Check if the "bank" field has been filled in before actually attempting
> the IOTLB flush to avoid it. The IOTLB is also flushed when the device
> comes out of runtime suspend, so it should have a clean initial state.
> 
> Fixes: 08500c43d4f7 ("iommu/mediatek: Adjust the structure")
> Signed-off-by: Chen-Yu Tsai <wenst@chromium.org>

Not only ARM SMMU does this, others are doing the same, some in a different
form (walking a list)... So I agree with this being a valid fix.

Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>

> ---
> 
> I think this is a valid fix, but I'm not very familiar with the hardware
> or the design of the driver. The ARM SMMU drivers seem to do this as well.
> 
>   drivers/iommu/mtk_iommu.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
> index aecc7d154f28..e93906d6e112 100644
> --- a/drivers/iommu/mtk_iommu.c
> +++ b/drivers/iommu/mtk_iommu.c
> @@ -781,7 +781,8 @@ static void mtk_iommu_flush_iotlb_all(struct iommu_domain *domain)
>   {
>   	struct mtk_iommu_domain *dom = to_mtk_domain(domain);
>   
> -	mtk_iommu_tlb_flush_all(dom->bank->parent_data);
> +	if (dom->bank)
> +		mtk_iommu_tlb_flush_all(dom->bank->parent_data);
>   }
>   
>   static void mtk_iommu_iotlb_sync(struct iommu_domain *domain,