[PATCH 0/2] iommu/arm-smmu-v3: Two vsmmu impl_ops cleanups

[PATCH 0/2] iommu/arm-smmu-v3: Two vsmmu impl_ops cleanups

[Nicolin Chen]

Since the vsmmu and cmdqv patches were accepeted via the iommufd tree,
this is based on Jason's for-next tree.

Per request from Will following the accepted latest vcmdq series to clean
up the impl_ops:

https://lore.kernel.org/linux-iommu/aHE4Y-fbucm-j-yi@willie-the-truck/
"
 It would be nice to avoid adding data members to the ops structure, if
 at all possible. The easiest thing would probably be to add a function
 for getting the vsmmu size and then pushing the two checks against
 'vsmmu_type' down into the impl_ops callbacks so that: 

   1. If the type is IOMMU_VIOMMU_TYPE_ARM_SMMUV3, we don't bother with
      the impl_ops at all in arm_vsmmu_init() and arm_smmu_get_viommu_size()
 
   2. Otherwise, we pass the type into the impl_ops and they can check it

 Of course, that can be a patch on top of the series as there's no point
 respinning the whole just for this.
"

Add two clean up patches with the first one doing 1 and 2 to prioritize
IOMMU_VIOMMU_TYPE_ARM_SMMUV3 always, and the second one dropping static
vsmmu_type and vsmmu_size data members.

Nicolin Chen (2):
  iommu/arm-smmu-v3: Do not bother impl_ops if
    IOMMU_VIOMMU_TYPE_ARM_SMMUV3
  iommu/arm-smmu-v3: Replace vsmmu_size/type with get_viommu_size

 .../arm/arm-smmu-v3/arm-smmu-v3-iommufd.c     | 26 +++++++++++--------
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h   |  3 +--
 .../iommu/arm/arm-smmu-v3/tegra241-cmdqv.c    | 14 ++++++++--
 3 files changed, 28 insertions(+), 15 deletions(-)


base-commit: ab6bc44159d8f0c4ee757e0ce041fa9033e0ead8
-- 
2.43.0

[PATCH 1/2] iommu/arm-smmu-v3: Do not bother impl_ops if IOMMU_VIOMMU_TYPE_ARM_SMMUV3

[Nicolin Chen]

When viommu type is IOMMU_VIOMMU_TYPE_ARM_SMMUV3, always return or init the
standard struct arm_vsmmu, instead of going through impl_ops that must have
its own viommu type than the standard IOMMU_VIOMMU_TYPE_ARM_SMMUV3.

Given that arm_vsmmu_init() is called after arm_smmu_get_viommu_size(), any
unsupported viommu->type must be a corruption. And missing a pairing impl's
vsmmu_init should be a driver bug too. Warn these two cases.

Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
---
 .../arm/arm-smmu-v3/arm-smmu-v3-iommufd.c     | 30 ++++++++++++-------
 1 file changed, 19 insertions(+), 11 deletions(-)

diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c
index d9bea8f1f636..0b2acb80f41b 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c
@@ -420,14 +420,13 @@ size_t arm_smmu_get_viommu_size(struct device *dev,
 	    !(smmu->features & ARM_SMMU_FEAT_S2FWB))
 		return 0;
 
-	if (smmu->impl_ops && smmu->impl_ops->vsmmu_size &&
-	    viommu_type == smmu->impl_ops->vsmmu_type)
-		return smmu->impl_ops->vsmmu_size;
+	if (viommu_type == IOMMU_VIOMMU_TYPE_ARM_SMMUV3)
+		return VIOMMU_STRUCT_SIZE(struct arm_vsmmu, core);
 
-	if (viommu_type != IOMMU_VIOMMU_TYPE_ARM_SMMUV3)
+	if (!smmu->impl_ops || !smmu->impl_ops->vsmmu_size ||
+	    viommu_type != smmu->impl_ops->vsmmu_type)
 		return 0;
-
-	return VIOMMU_STRUCT_SIZE(struct arm_vsmmu, core);
+	return smmu->impl_ops->vsmmu_size;
 }
 
 int arm_vsmmu_init(struct iommufd_viommu *viommu,
@@ -447,12 +446,21 @@ int arm_vsmmu_init(struct iommufd_viommu *viommu,
 	/* FIXME Move VMID allocation from the S2 domain allocation to here */
 	vsmmu->vmid = s2_parent->s2_cfg.vmid;
 
-	if (smmu->impl_ops && smmu->impl_ops->vsmmu_init &&
-	    viommu->type == smmu->impl_ops->vsmmu_type)
-		return smmu->impl_ops->vsmmu_init(vsmmu, user_data);
+	if (viommu->type == IOMMU_VIOMMU_TYPE_ARM_SMMUV3) {
+		viommu->ops = &arm_vsmmu_ops;
+		return 0;
+	}
 
-	viommu->ops = &arm_vsmmu_ops;
-	return 0;
+	/*
+	 * If a non standard type was supported in arm_smmu_get_viommu_size() by
+	 * an implementation, a pairing vsmmu_init op must exist.
+	 */
+	if (WARN_ON_ONCE(!smmu->impl_ops || !smmu->impl_ops->vsmmu_init))
+		return -EOPNOTSUPP;
+	/* Unsupported type was rejected in arm_smmu_get_viommu_size() */
+	if (WARN_ON(viommu->type != smmu->impl_ops->vsmmu_type))
+		return -EOPNOTSUPP;
+	return smmu->impl_ops->vsmmu_init(vsmmu, user_data);
 }
 
 int arm_vmaster_report_event(struct arm_smmu_vmaster *vmaster, u64 *evt)
-- 
2.43.0

Re: [PATCH 1/2] iommu/arm-smmu-v3: Do not bother impl_ops if IOMMU_VIOMMU_TYPE_ARM_SMMUV3

[Will Deacon]

On Fri, Jul 18, 2025 at 04:48:21PM -0700, Nicolin Chen wrote:
> When viommu type is IOMMU_VIOMMU_TYPE_ARM_SMMUV3, always return or init the
> standard struct arm_vsmmu, instead of going through impl_ops that must have
> its own viommu type than the standard IOMMU_VIOMMU_TYPE_ARM_SMMUV3.
> 
> Given that arm_vsmmu_init() is called after arm_smmu_get_viommu_size(), any
> unsupported viommu->type must be a corruption. And missing a pairing impl's
> vsmmu_init should be a driver bug too. Warn these two cases.
> 
> Suggested-by: Will Deacon <will@kernel.org>
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> ---
>  .../arm/arm-smmu-v3/arm-smmu-v3-iommufd.c     | 30 ++++++++++++-------
>  1 file changed, 19 insertions(+), 11 deletions(-)
> 
> diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c
> index d9bea8f1f636..0b2acb80f41b 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c
> +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c
> @@ -420,14 +420,13 @@ size_t arm_smmu_get_viommu_size(struct device *dev,
>  	    !(smmu->features & ARM_SMMU_FEAT_S2FWB))
>  		return 0;
>  
> -	if (smmu->impl_ops && smmu->impl_ops->vsmmu_size &&
> -	    viommu_type == smmu->impl_ops->vsmmu_type)
> -		return smmu->impl_ops->vsmmu_size;
> +	if (viommu_type == IOMMU_VIOMMU_TYPE_ARM_SMMUV3)
> +		return VIOMMU_STRUCT_SIZE(struct arm_vsmmu, core);
>  
> -	if (viommu_type != IOMMU_VIOMMU_TYPE_ARM_SMMUV3)
> +	if (!smmu->impl_ops || !smmu->impl_ops->vsmmu_size ||
> +	    viommu_type != smmu->impl_ops->vsmmu_type)
>  		return 0;
> -
> -	return VIOMMU_STRUCT_SIZE(struct arm_vsmmu, core);
> +	return smmu->impl_ops->vsmmu_size;
>  }
>  
>  int arm_vsmmu_init(struct iommufd_viommu *viommu,
> @@ -447,12 +446,21 @@ int arm_vsmmu_init(struct iommufd_viommu *viommu,
>  	/* FIXME Move VMID allocation from the S2 domain allocation to here */
>  	vsmmu->vmid = s2_parent->s2_cfg.vmid;
>  
> -	if (smmu->impl_ops && smmu->impl_ops->vsmmu_init &&
> -	    viommu->type == smmu->impl_ops->vsmmu_type)
> -		return smmu->impl_ops->vsmmu_init(vsmmu, user_data);
> +	if (viommu->type == IOMMU_VIOMMU_TYPE_ARM_SMMUV3) {
> +		viommu->ops = &arm_vsmmu_ops;
> +		return 0;
> +	}
>  
> -	viommu->ops = &arm_vsmmu_ops;
> -	return 0;
> +	/*
> +	 * If a non standard type was supported in arm_smmu_get_viommu_size() by
> +	 * an implementation, a pairing vsmmu_init op must exist.
> +	 */
> +	if (WARN_ON_ONCE(!smmu->impl_ops || !smmu->impl_ops->vsmmu_init))
> +		return -EOPNOTSUPP;

It might be cleaner to check the impl_ops in arm_smmu_impl_probe() where
we're in a slightly better situation for failing gracefully and
deterministically.

Otherwise, looks good. Thanks!

Will

[PATCH 2/2] iommu/arm-smmu-v3: Replace vsmmu_size/type with get_viommu_size

[Nicolin Chen]

It's more flexible to have a get_viommu_size op. Replace static vsmmu_size
and vsmmu_type with that.

Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
---
 .../iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c    |  8 ++------
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h        |  3 +--
 drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c     | 14 ++++++++++++--
 3 files changed, 15 insertions(+), 10 deletions(-)

diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c
index 0b2acb80f41b..d4b7affaa480 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c
@@ -423,10 +423,9 @@ size_t arm_smmu_get_viommu_size(struct device *dev,
 	if (viommu_type == IOMMU_VIOMMU_TYPE_ARM_SMMUV3)
 		return VIOMMU_STRUCT_SIZE(struct arm_vsmmu, core);
 
-	if (!smmu->impl_ops || !smmu->impl_ops->vsmmu_size ||
-	    viommu_type != smmu->impl_ops->vsmmu_type)
+	if (!smmu->impl_ops || !smmu->impl_ops->get_viommu_size)
 		return 0;
-	return smmu->impl_ops->vsmmu_size;
+	return smmu->impl_ops->get_viommu_size(viommu_type);
 }
 
 int arm_vsmmu_init(struct iommufd_viommu *viommu,
@@ -457,9 +456,6 @@ int arm_vsmmu_init(struct iommufd_viommu *viommu,
 	 */
 	if (WARN_ON_ONCE(!smmu->impl_ops || !smmu->impl_ops->vsmmu_init))
 		return -EOPNOTSUPP;
-	/* Unsupported type was rejected in arm_smmu_get_viommu_size() */
-	if (WARN_ON(viommu->type != smmu->impl_ops->vsmmu_type))
-		return -EOPNOTSUPP;
 	return smmu->impl_ops->vsmmu_init(vsmmu, user_data);
 }
 
diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
index 3fa02c51df9f..e332f5ba2f8a 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
@@ -728,8 +728,7 @@ struct arm_smmu_impl_ops {
 	 */
 	void *(*hw_info)(struct arm_smmu_device *smmu, u32 *length,
 			 enum iommu_hw_info_type *type);
-	const size_t vsmmu_size;
-	const enum iommu_viommu_type vsmmu_type;
+	size_t (*get_viommu_size)(enum iommu_viommu_type viommu_type);
 	int (*vsmmu_init)(struct arm_vsmmu *vsmmu,
 			  const struct iommu_user_data *user_data);
 };
diff --git a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
index 4c86eacd36b1..46005ed52bc2 100644
--- a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
+++ b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
@@ -832,6 +832,13 @@ static void *tegra241_cmdqv_hw_info(struct arm_smmu_device *smmu, u32 *length,
 	return info;
 }
 
+static size_t tegra241_cmdqv_get_vintf_size(enum iommu_viommu_type viommu_type)
+{
+	if (viommu_type != IOMMU_VIOMMU_TYPE_TEGRA241_CMDQV)
+		return 0;
+	return VIOMMU_STRUCT_SIZE(struct tegra241_vintf, vsmmu.core);
+}
+
 static struct arm_smmu_impl_ops tegra241_cmdqv_impl_ops = {
 	/* For in-kernel use */
 	.get_secondary_cmdq = tegra241_cmdqv_get_cmdq,
@@ -839,8 +846,7 @@ static struct arm_smmu_impl_ops tegra241_cmdqv_impl_ops = {
 	.device_remove = tegra241_cmdqv_remove,
 	/* For user-space use */
 	.hw_info = tegra241_cmdqv_hw_info,
-	.vsmmu_size = VIOMMU_STRUCT_SIZE(struct tegra241_vintf, vsmmu.core),
-	.vsmmu_type = IOMMU_VIOMMU_TYPE_TEGRA241_CMDQV,
+	.get_viommu_size = tegra241_cmdqv_get_vintf_size,
 	.vsmmu_init = tegra241_cmdqv_init_vintf_user,
 };
 
@@ -1273,6 +1279,10 @@ tegra241_cmdqv_init_vintf_user(struct arm_vsmmu *vsmmu,
 	phys_addr_t page0_base;
 	int ret;
 
+	/* Unsupported type was rejected in tegra241_cmdqv_get_vintf_size() */
+	if (WARN_ON(vsmmu->core.type != IOMMU_VIOMMU_TYPE_TEGRA241_CMDQV))
+		return -EOPNOTSUPP;
+
 	if (!user_data)
 		return -EINVAL;
 
-- 
2.43.0

Re: [PATCH 2/2] iommu/arm-smmu-v3: Replace vsmmu_size/type with get_viommu_size

[Will Deacon]

On Fri, Jul 18, 2025 at 04:48:22PM -0700, Nicolin Chen wrote:
> It's more flexible to have a get_viommu_size op. Replace static vsmmu_size
> and vsmmu_type with that.
> 
> Suggested-by: Will Deacon <will@kernel.org>
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> ---
>  .../iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c    |  8 ++------
>  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h        |  3 +--
>  drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c     | 14 ++++++++++++--
>  3 files changed, 15 insertions(+), 10 deletions(-)

Acked-by: Will Deacon <will@kernel.org>

Will