RE: [PATCH v4 06/12] iommu/vt-d: Set the nested domain to a device

[Tian, Kevin]

> From: Liu, Yi L <yi.l.liu@xxxxxxxxx>
> Sent: Monday, July 24, 2023 7:13 PM
> +
> +static int intel_nested_attach_dev(struct iommu_domain *domain,
> +				   struct device *dev)
> +{
> +	struct device_domain_info *info = dev_iommu_priv_get(dev);
> +	struct dmar_domain *dmar_domain = to_dmar_domain(domain);
> +	struct intel_iommu *iommu = info->iommu;
> +	unsigned long flags;
> +	int ret = 0;
> +
> +	if (info->domain)
> +		device_block_translation(dev);
> +
> +	if (iommu->agaw < dmar_domain->s2_domain->agaw) {
> +		dev_err_ratelimited(dev, "Adjusted guest address width not
> compatible\n");
> +		return -ENODEV;
> +	}

this is the check duplicated with patch04.

> +
> +	ret = domain_attach_iommu(dmar_domain, iommu);
> +	if (ret) {
> +		dev_err_ratelimited(dev, "Failed to attach domain to
> iommu\n");
> +		return ret;
> +	}
> +

[...]

> +	domain_update_iommu_cap(dmar_domain);

iommu_cap is already updated in domain_attach_iommu().

> 
>  static const struct iommu_domain_ops intel_nested_domain_ops = {
> +	.attach_dev		= intel_nested_attach_dev,
>  	.free			= intel_nested_domain_free,
> +	.enforce_cache_coherency = intel_iommu_enforce_cache_coherency,

this is not required. enforce_cache_coherency() will be called on parent
hwpt when it's being created. patch04 should check parent's force_snooping
to set pgsnp in the pasid entry.

As Jason explained it should be done only for kernel owned page table.