Re: [PATCH v2] iommu/arm-smmu-v3-iommufd: Allow attaching nested domain for GBPA cases

[Pranjal Shrivastava <praan@google.com>]

On Mon, Nov 03, 2025 at 09:27:55AM -0800, Nicolin Chen wrote:
> A vDEVICE has been a hard requirement for attaching a nested domain to the
> device. This makes sense when installing a guest STE, since a vSID must be
> present and given to the kernel during the vDEVICE allocation.
> 
> But, when CR0.SMMUEN is disabled, VM doesn't really need a vSID to program
> the vSMMU behavior as GBPA will take effect, in which case the vSTE in the
> nested domain could have carried the bypass or abort configuration in GBPA
> register. Thus, having such a hard requirement doesn't work well for GBPA.
> 
> Skip vmaster allocation in arm_smmu_attach_prepare_vmaster() for an abort
> or bypass vSTE. Note that device on this attachment won't report vevents.
> 
> Update the uAPI doc accordingly.
> 
> Tested-by: Shameer Kolothum <skolothumtho@nvidia.com>
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> ---
> 
> Changelog
> v2
>  * Add Tested-by from Shameer
>  * Skip vmaster allocation instead of bypassing vsid=0
>  * Revise the uAPI doc to note a corner case when CR0.SMMUEN=1
> 
>  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-iommufd.c | 13 ++++++++++++-
>  include/uapi/linux/iommufd.h                        |  9 +++++++++
>  2 files changed, 21 insertions(+), 1 deletion(-)
> 
> 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 8cd8929bbfdf8..e5fbbdbdea242 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
> @@ -99,6 +99,8 @@ static void arm_smmu_make_nested_domain_ste(
>  int arm_smmu_attach_prepare_vmaster(struct arm_smmu_attach_state *state,
>  				    struct arm_smmu_nested_domain *nested_domain)
>  {
> +	unsigned int cfg =
> +		FIELD_GET(STRTAB_STE_0_CFG, le64_to_cpu(nested_domain->ste[0]));
>  	struct arm_smmu_vmaster *vmaster;
>  	unsigned long vsid;
>  	int ret;
> @@ -107,8 +109,17 @@ int arm_smmu_attach_prepare_vmaster(struct arm_smmu_attach_state *state,
>  
>  	ret = iommufd_viommu_get_vdev_id(&nested_domain->vsmmu->core,
>  					 state->master->dev, &vsid);
> -	if (ret)
> +	/*
> +	 * Attaching to a translate nested domain must allocate a vDEVICE prior,
> +	 * as CD/ATS invalidations and vevents require a vSID to work properly.
> +	 * A abort/bypass domain is allowed to attach w/o vmaster for GBPA case.
> +	 */
> +	if (ret) {
> +		if (cfg == STRTAB_STE_0_CFG_ABORT ||
> +		    cfg == STRTAB_STE_0_CFG_BYPASS)
> +			return 0;
>  		return ret;
> +	}
>

Skipping the vmaster allocation entirely for the GBPA-only case (when no
vdevice is found) is much cleaner. Thanks!

>  	vmaster = kzalloc(sizeof(*vmaster), GFP_KERNEL);
>  	if (!vmaster)
> diff --git a/include/uapi/linux/iommufd.h b/include/uapi/linux/iommufd.h
> index c218c89e0e2eb..225671603ade6 100644
> --- a/include/uapi/linux/iommufd.h
> +++ b/include/uapi/linux/iommufd.h
> @@ -450,6 +450,15 @@ struct iommu_hwpt_vtd_s1 {
>   * nested domain will translate the same as the nesting parent. The S1 will
>   * install a Context Descriptor Table pointing at userspace memory translated
>   * by the nesting parent.
> + *
> + * It's suggested to allocate a vDEVICE object carrying vSID and then re-attach
> + * the nested domain, as soon as the vSID is available in the VMM level:
> + * - when Cfg=translate, a vDEVICE must be allocated prior to attaching to the
> + *   allocated nested domain, as CD/ATS invalidations and vevents need a vSID.
> + * - when Cfg=bypass/abort, a vDEVICE is not enforced during the nested domain
> + *   attachment, to support a GBPA case where VM sets CR0.SMMUEN=0. However, if
> + *   VM sets CR0.SMMUEN=1 while missing a vDEVICE object, kernel would fail to
> + *   report events to the VM. E.g. F_TRANSLATION when guest STE.Cfg=abort.
>   */

Reviewed-by: Pranjal Shrivastava <praan@google.com>

Thanks,
Praan