[PATCH 0/2] iommu/tegra241-cmdqv: Fix initialization and uAPI related to HYP_OWN

[PATCH 0/2] iommu/tegra241-cmdqv: Fix initialization and uAPI related to HYP_OWN

[Nicolin Chen]

Hi,

Eric and Shameer found a bug in kernel driver that supports_cmd is set in
an incorrect order related to the hyp_own finalization. This results in a
unnecessary command filter in the host environment that sends unsupported
(non-invalidation) commands to the SMMU CMDQ.

Also, Eric pointed it out that the requirement of VMM wiring HYP_OWN to 0
is unclear.

Thus, I am sending this small series to address these issues.

Nicolin Chen (2):
  iommu/tegra241-cmdqv: Set supports_cmd op in tegra241_vcmdq_hw_init()
  iommu/tegra241-cmdqv: Update uAPI to clarify HYP_OWN requirement

 include/uapi/linux/iommufd.h                   | 5 +++++
 drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c | 7 ++++---
 2 files changed, 9 insertions(+), 3 deletions(-)

-- 
2.43.0

[PATCH 1/2] iommu/tegra241-cmdqv: Set supports_cmd op in tegra241_vcmdq_hw_init()

[Nicolin Chen]

vintf->hyp_own is finalized in tegra241_vintf_hw_init(). On the other hand,
tegra241_vcmdq_alloc_smmu_cmdq() is called via an init_structures callback,
which is earlier than tegra241_vintf_hw_init().

This results in the supports_cmd op always being set to the guest function,
although this doesn't break any functionality nor have some noticeable perf
impact since non-invalidation commands are not issued in the perf sensitive
context.

Fix this by moving supports_cmd to tegra241_vcmdq_hw_init().

After this change,
 - For a guest kernel, this will be a status quo
 - For a host kernel, non-invalidation commands will be issued to VCMDQ(s)

Fixes: a9d40285bdef ("iommu/tegra241-cmdqv: Limit CMDs for VCMDQs of a guest owned VINTF")
Reported-by: Eric Auger <eric.auger@redhat.com>
Reported-by: Shameer Kolothum <skolothumtho@nvidia.com>
Closes: https://lore.kernel.org/qemu-devel/CH3PR12MB754836BEE54E39B30C7210C0AB44A@CH3PR12MB7548.namprd12.prod.outlook.com/
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
---
 drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
index 6fe5563eaf9eb..83f6e9f6c51d6 100644
--- a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
+++ b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
@@ -479,6 +479,10 @@ static int tegra241_vcmdq_hw_init(struct tegra241_vcmdq *vcmdq)
 	/* Reset VCMDQ */
 	tegra241_vcmdq_hw_deinit(vcmdq);
 
+	/* vintf->hyp_own is a HW state finalized in tegra241_vintf_hw_init() */
+	if (!vcmdq->vintf->hyp_own)
+		vcmdq->cmdq.supports_cmd = tegra241_guest_vcmdq_supports_cmd;
+
 	/* Configure and enable VCMDQ */
 	writeq_relaxed(vcmdq->cmdq.q.q_base, REG_VCMDQ_PAGE1(vcmdq, BASE));
 
@@ -639,9 +643,6 @@ static int tegra241_vcmdq_alloc_smmu_cmdq(struct tegra241_vcmdq *vcmdq)
 	q->q_base = q->base_dma & VCMDQ_ADDR;
 	q->q_base |= FIELD_PREP(VCMDQ_LOG2SIZE, q->llq.max_n_shift);
 
-	if (!vcmdq->vintf->hyp_own)
-		cmdq->supports_cmd = tegra241_guest_vcmdq_supports_cmd;
-
 	return arm_smmu_cmdq_init(smmu, cmdq);
 }
 
-- 
2.43.0

[PATCH 2/2] iommu/tegra241-cmdqv: Update uAPI to clarify HYP_OWN requirement

[Nicolin Chen]

From hardware implementation perspective, a guest tegra241-cmdqv hardware
is different than the host hardware:
 - Host HW is backed by a VINTF (HYP_OWN=1)
 - Guest HW is backed by a VINTF (HYP_OWN=0)

The kernel driver has an implementation requirement of the HYP_OWN bit in
the VM. So, VMM must follow that to allow the same copy of Linux to work.

Add this requirement to the uAPI, which is currently missing.

Fixes: 4dc0d12474f9 ("iommu/tegra241-cmdqv: Add user-space use support")
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
---
 include/uapi/linux/iommufd.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/include/uapi/linux/iommufd.h b/include/uapi/linux/iommufd.h
index 1dafbc552d37d..f63edbe71d542 100644
--- a/include/uapi/linux/iommufd.h
+++ b/include/uapi/linux/iommufd.h
@@ -1052,6 +1052,11 @@ struct iommu_fault_alloc {
 enum iommu_viommu_type {
 	IOMMU_VIOMMU_TYPE_DEFAULT = 0,
 	IOMMU_VIOMMU_TYPE_ARM_SMMUV3 = 1,
+	/*
+	 * TEGRA241_CMDQV requirements (otherwise, VCMDQs will not work)
+	 * - Kernel will allocate a VINTF (HYP_OWN=0) to back this VIOMMU. So,
+	 *   VMM must wire the HYP_OWN bit to 0 in guest VINTF_CONFIG register
+	 */
 	IOMMU_VIOMMU_TYPE_TEGRA241_CMDQV = 2,
 };
 
-- 
2.43.0

RE: [PATCH 1/2] iommu/tegra241-cmdqv: Set supports_cmd op in tegra241_vcmdq_hw_init()

[Shameer Kolothum Thodi]

> -----Original Message-----
> From: Nicolin Chen <nicolinc@nvidia.com>
> Sent: 13 March 2026 00:37
> To: will@kernel.org; robin.murphy@arm.com; Jason Gunthorpe
> <jgg@nvidia.com>; kevin.tian@intel.com
> Cc: joro@8bytes.org; linux-tegra@vger.kernel.org; linux-arm-
> kernel@lists.infradead.org; iommu@lists.linux.dev; linux-
> kernel@vger.kernel.org; eric.auger@redhat.com; Shameer Kolothum Thodi
> <skolothumtho@nvidia.com>
> Subject: [PATCH 1/2] iommu/tegra241-cmdqv: Set supports_cmd op in
> tegra241_vcmdq_hw_init()
> 
> vintf->hyp_own is finalized in tegra241_vintf_hw_init(). On the other hand,
> tegra241_vcmdq_alloc_smmu_cmdq() is called via an init_structures callback,
> which is earlier than tegra241_vintf_hw_init().
> 
> This results in the supports_cmd op always being set to the guest function,
> although this doesn't break any functionality nor have some noticeable perf
> impact since non-invalidation commands are not issued in the perf sensitive
> context.
> 
> Fix this by moving supports_cmd to tegra241_vcmdq_hw_init().
> 
> After this change,
>  - For a guest kernel, this will be a status quo
>  - For a host kernel, non-invalidation commands will be issued to VCMDQ(s)
> 
> Fixes: a9d40285bdef ("iommu/tegra241-cmdqv: Limit CMDs for VCMDQs of a
> guest owned VINTF")
> Reported-by: Eric Auger <eric.auger@redhat.com>
> Reported-by: Shameer Kolothum <skolothumtho@nvidia.com>
> Closes: https://lore.kernel.org/qemu-
> devel/CH3PR12MB754836BEE54E39B30C7210C0AB44A@CH3PR12MB7548
> .namprd12.prod.outlook.com/
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> ---

Ran few tests with QEMU toggling HYP_OWN bit, and this now enforces
that QEMU sets HYP_OWN for VCMDQ to work properly.

Tested-by: Shameer Kolothum <skolothumtho@nvidia.com>

Thanks,
Shameer

>  drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> index 6fe5563eaf9eb..83f6e9f6c51d6 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> +++ b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> @@ -479,6 +479,10 @@ static int tegra241_vcmdq_hw_init(struct
> tegra241_vcmdq *vcmdq)
>  	/* Reset VCMDQ */
>  	tegra241_vcmdq_hw_deinit(vcmdq);
> 
> +	/* vintf->hyp_own is a HW state finalized in tegra241_vintf_hw_init()
> */
> +	if (!vcmdq->vintf->hyp_own)
> +		vcmdq->cmdq.supports_cmd =
> tegra241_guest_vcmdq_supports_cmd;
> +
>  	/* Configure and enable VCMDQ */
>  	writeq_relaxed(vcmdq->cmdq.q.q_base,
> REG_VCMDQ_PAGE1(vcmdq, BASE));
> 
> @@ -639,9 +643,6 @@ static int tegra241_vcmdq_alloc_smmu_cmdq(struct
> tegra241_vcmdq *vcmdq)
>  	q->q_base = q->base_dma & VCMDQ_ADDR;
>  	q->q_base |= FIELD_PREP(VCMDQ_LOG2SIZE, q->llq.max_n_shift);
> 
> -	if (!vcmdq->vintf->hyp_own)
> -		cmdq->supports_cmd =
> tegra241_guest_vcmdq_supports_cmd;
> -
>  	return arm_smmu_cmdq_init(smmu, cmdq);
>  }
> 
> --
> 2.43.0

Re: [PATCH 2/2] iommu/tegra241-cmdqv: Update uAPI to clarify HYP_OWN requirement

[Will Deacon]

On Thu, Mar 12, 2026 at 05:36:35PM -0700, Nicolin Chen wrote:
> From hardware implementation perspective, a guest tegra241-cmdqv hardware
> is different than the host hardware:
>  - Host HW is backed by a VINTF (HYP_OWN=1)
>  - Guest HW is backed by a VINTF (HYP_OWN=0)
> 
> The kernel driver has an implementation requirement of the HYP_OWN bit in
> the VM. So, VMM must follow that to allow the same copy of Linux to work.
> 
> Add this requirement to the uAPI, which is currently missing.
> 
> Fixes: 4dc0d12474f9 ("iommu/tegra241-cmdqv: Add user-space use support")
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> ---
>  include/uapi/linux/iommufd.h | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/include/uapi/linux/iommufd.h b/include/uapi/linux/iommufd.h
> index 1dafbc552d37d..f63edbe71d542 100644
> --- a/include/uapi/linux/iommufd.h
> +++ b/include/uapi/linux/iommufd.h
> @@ -1052,6 +1052,11 @@ struct iommu_fault_alloc {
>  enum iommu_viommu_type {
>  	IOMMU_VIOMMU_TYPE_DEFAULT = 0,
>  	IOMMU_VIOMMU_TYPE_ARM_SMMUV3 = 1,
> +	/*
> +	 * TEGRA241_CMDQV requirements (otherwise, VCMDQs will not work)
> +	 * - Kernel will allocate a VINTF (HYP_OWN=0) to back this VIOMMU. So,
> +	 *   VMM must wire the HYP_OWN bit to 0 in guest VINTF_CONFIG register
> +	 */
>  	IOMMU_VIOMMU_TYPE_TEGRA241_CMDQV = 2,
>  };

Jason, are you happy with this one?

Will

Re: [PATCH 2/2] iommu/tegra241-cmdqv: Update uAPI to clarify HYP_OWN requirement

[Jason Gunthorpe]

On Thu, Mar 19, 2026 at 02:47:13PM +0000, Will Deacon wrote:
> On Thu, Mar 12, 2026 at 05:36:35PM -0700, Nicolin Chen wrote:
> > From hardware implementation perspective, a guest tegra241-cmdqv hardware
> > is different than the host hardware:
> >  - Host HW is backed by a VINTF (HYP_OWN=1)
> >  - Guest HW is backed by a VINTF (HYP_OWN=0)
> > 
> > The kernel driver has an implementation requirement of the HYP_OWN bit in
> > the VM. So, VMM must follow that to allow the same copy of Linux to work.
> > 
> > Add this requirement to the uAPI, which is currently missing.
> > 
> > Fixes: 4dc0d12474f9 ("iommu/tegra241-cmdqv: Add user-space use support")
> > Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> > ---
> >  include/uapi/linux/iommufd.h | 5 +++++
> >  1 file changed, 5 insertions(+)
> > 
> > diff --git a/include/uapi/linux/iommufd.h b/include/uapi/linux/iommufd.h
> > index 1dafbc552d37d..f63edbe71d542 100644
> > --- a/include/uapi/linux/iommufd.h
> > +++ b/include/uapi/linux/iommufd.h
> > @@ -1052,6 +1052,11 @@ struct iommu_fault_alloc {
> >  enum iommu_viommu_type {
> >  	IOMMU_VIOMMU_TYPE_DEFAULT = 0,
> >  	IOMMU_VIOMMU_TYPE_ARM_SMMUV3 = 1,
> > +	/*
> > +	 * TEGRA241_CMDQV requirements (otherwise, VCMDQs will not work)
> > +	 * - Kernel will allocate a VINTF (HYP_OWN=0) to back this VIOMMU. So,
> > +	 *   VMM must wire the HYP_OWN bit to 0 in guest VINTF_CONFIG register
> > +	 */
> >  	IOMMU_VIOMMU_TYPE_TEGRA241_CMDQV = 2,
> >  };
> 
> Jason, are you happy with this one?

Yes, please pick up the series

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

Jason

Re: [PATCH 1/2] iommu/tegra241-cmdqv: Set supports_cmd op in tegra241_vcmdq_hw_init()

[Eric Auger]

Hi Nicolin,

On 3/13/26 1:36 AM, Nicolin Chen wrote:
> vintf->hyp_own is finalized in tegra241_vintf_hw_init(). On the other hand,
> tegra241_vcmdq_alloc_smmu_cmdq() is called via an init_structures callback,
> which is earlier than tegra241_vintf_hw_init().
>
> This results in the supports_cmd op always being set to the guest function,
> although this doesn't break any functionality nor have some noticeable perf
> impact since non-invalidation commands are not issued in the perf sensitive
> context.
>
> Fix this by moving supports_cmd to tegra241_vcmdq_hw_init().
>
> After this change,
>  - For a guest kernel, this will be a status quo
>  - For a host kernel, non-invalidation commands will be issued to VCMDQ(s)
>
> Fixes: a9d40285bdef ("iommu/tegra241-cmdqv: Limit CMDs for VCMDQs of a guest owned VINTF")
> Reported-by: Eric Auger <eric.auger@redhat.com>
> Reported-by: Shameer Kolothum <skolothumtho@nvidia.com>
> Closes: https://lore.kernel.org/qemu-devel/CH3PR12MB754836BEE54E39B30C7210C0AB44A@CH3PR12MB7548.namprd12.prod.outlook.com/
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>

Reviewed-by: Eric Auger <eric.auger@redhat.com>

Eric
> ---
>  drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> index 6fe5563eaf9eb..83f6e9f6c51d6 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> +++ b/drivers/iommu/arm/arm-smmu-v3/tegra241-cmdqv.c
> @@ -479,6 +479,10 @@ static int tegra241_vcmdq_hw_init(struct tegra241_vcmdq *vcmdq)
>  	/* Reset VCMDQ */
>  	tegra241_vcmdq_hw_deinit(vcmdq);
>  
> +	/* vintf->hyp_own is a HW state finalized in tegra241_vintf_hw_init() */
> +	if (!vcmdq->vintf->hyp_own)
> +		vcmdq->cmdq.supports_cmd = tegra241_guest_vcmdq_supports_cmd;
> +
>  	/* Configure and enable VCMDQ */
>  	writeq_relaxed(vcmdq->cmdq.q.q_base, REG_VCMDQ_PAGE1(vcmdq, BASE));
>  
> @@ -639,9 +643,6 @@ static int tegra241_vcmdq_alloc_smmu_cmdq(struct tegra241_vcmdq *vcmdq)
>  	q->q_base = q->base_dma & VCMDQ_ADDR;
>  	q->q_base |= FIELD_PREP(VCMDQ_LOG2SIZE, q->llq.max_n_shift);
>  
> -	if (!vcmdq->vintf->hyp_own)
> -		cmdq->supports_cmd = tegra241_guest_vcmdq_supports_cmd;
> -
>  	return arm_smmu_cmdq_init(smmu, cmdq);
>  }
>  

Re: [PATCH 2/2] iommu/tegra241-cmdqv: Update uAPI to clarify HYP_OWN requirement

[Eric Auger]

On 3/13/26 1:36 AM, Nicolin Chen wrote:
> From hardware implementation perspective, a guest tegra241-cmdqv hardware
> is different than the host hardware:
>  - Host HW is backed by a VINTF (HYP_OWN=1)
>  - Guest HW is backed by a VINTF (HYP_OWN=0)
>
> The kernel driver has an implementation requirement of the HYP_OWN bit in
> the VM. So, VMM must follow that to allow the same copy of Linux to work.
>
> Add this requirement to the uAPI, which is currently missing.
>
> Fixes: 4dc0d12474f9 ("iommu/tegra241-cmdqv: Add user-space use support")
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>

Eric
> ---
>  include/uapi/linux/iommufd.h | 5 +++++
>  1 file changed, 5 insertions(+)
>
> diff --git a/include/uapi/linux/iommufd.h b/include/uapi/linux/iommufd.h
> index 1dafbc552d37d..f63edbe71d542 100644
> --- a/include/uapi/linux/iommufd.h
> +++ b/include/uapi/linux/iommufd.h
> @@ -1052,6 +1052,11 @@ struct iommu_fault_alloc {
>  enum iommu_viommu_type {
>  	IOMMU_VIOMMU_TYPE_DEFAULT = 0,
>  	IOMMU_VIOMMU_TYPE_ARM_SMMUV3 = 1,
> +	/*
> +	 * TEGRA241_CMDQV requirements (otherwise, VCMDQs will not work)
> +	 * - Kernel will allocate a VINTF (HYP_OWN=0) to back this VIOMMU. So,
> +	 *   VMM must wire the HYP_OWN bit to 0 in guest VINTF_CONFIG register
> +	 */
>  	IOMMU_VIOMMU_TYPE_TEGRA241_CMDQV = 2,
>  };
>  

Re: [PATCH 0/2] iommu/tegra241-cmdqv: Fix initialization and uAPI related to HYP_OWN

[Will Deacon]

On Thu, 12 Mar 2026 17:36:33 -0700, Nicolin Chen wrote:
> Eric and Shameer found a bug in kernel driver that supports_cmd is set in
> an incorrect order related to the hyp_own finalization. This results in a
> unnecessary command filter in the host environment that sends unsupported
> (non-invalidation) commands to the SMMU CMDQ.
> 
> Also, Eric pointed it out that the requirement of VMM wiring HYP_OWN to 0
> is unclear.
> 
> [...]

Applied to iommu (arm/smmu/updates), thanks!

[1/2] iommu/tegra241-cmdqv: Set supports_cmd op in tegra241_vcmdq_hw_init()
      https://git.kernel.org/iommu/c/803e41f36d22
[2/2] iommu/tegra241-cmdqv: Update uAPI to clarify HYP_OWN requirement
      https://git.kernel.org/iommu/c/9dcef98dbee3

Cheers,
-- 
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev