When falling back to standard SMMU CMDQ, impl_ops must be set to NULL, so
tegra241_cmdqv_remove() will not be re-entered by arm_smmu_impl_remove(),
which will trigger a UAF.

Fixes: 483e0bd8883a ("iommu/tegra241-cmdqv: Do not allocate vcmdq until dma_set_mask_and_coherent")
Cc: stable@vger.kernel.org
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
---
 drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
index ce7be8eeb43c..8249ef087af3 100644
--- a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
+++ b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
@@ -803,6 +803,7 @@ static int tegra241_cmdqv_init_structures(struct arm_smmu_device *smmu)
 	dev_info(smmu->impl_dev, "Falling back to standard SMMU CMDQ\n");
 	smmu->options &= ~ARM_SMMU_OPT_TEGRA241_CMDQV;
 	tegra241_cmdqv_remove(smmu);
+	smmu->impl_ops = NULL;
 	return 0;
 }
 
-- 
2.43.0

On Mon, Apr 07, 2025 at 01:35:00AM -0700, Nicolin Chen wrote:
> When falling back to standard SMMU CMDQ, impl_ops must be set to NULL, so
> tegra241_cmdqv_remove() will not be re-entered by arm_smmu_impl_remove(),
> which will trigger a UAF.
> 
> Fixes: 483e0bd8883a ("iommu/tegra241-cmdqv: Do not allocate vcmdq until dma_set_mask_and_coherent")
> Cc: stable@vger.kernel.org
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> ---
>  drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> index ce7be8eeb43c..8249ef087af3 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> +++ b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> @@ -803,6 +803,7 @@ static int tegra241_cmdqv_init_structures(struct arm_smmu_device *smmu)
>  	dev_info(smmu->impl_dev, "Falling back to standard SMMU CMDQ\n");
>  	smmu->options &= ~ARM_SMMU_OPT_TEGRA241_CMDQV;
>  	tegra241_cmdqv_remove(smmu);
> +	smmu->impl_ops = NULL;
>  	return 0;

Per discussion in the other patch, with a v2 this whole rewind
part would be dropped to ask SMMU driver to fail and then call 
tegra241_cmdqv_remove() via smmu->impl_ops->device_remove().

So we wouldn't need this patch after all.

Thanks
Nicolin

On Mon, Apr 07, 2025 at 01:35:00AM -0700, Nicolin Chen wrote:
> When falling back to standard SMMU CMDQ, impl_ops must be set to NULL, so
> tegra241_cmdqv_remove() will not be re-entered by arm_smmu_impl_remove(),
> which will trigger a UAF.
> 
> Fixes: 483e0bd8883a ("iommu/tegra241-cmdqv: Do not allocate vcmdq until dma_set_mask_and_coherent")
> Cc: stable@vger.kernel.org
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> ---
>  drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c | 1 +
>  1 file changed, 1 insertion(+)

Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>

Jason