Re: [PATCH v5 04/13] iommu/vt-d: Add pasid replace helpers

[Baolu Lu <baolu.lu@linux.intel.com>]

On 11/6/24 23:45, Yi Liu wrote:
> +int intel_pasid_replace_first_level(struct intel_iommu *iommu,
> +				    struct device *dev, pgd_t *pgd,
> +				    u32 pasid, u16 did, u16 old_did,
> +				    int flags)
> +{
> +	struct pasid_entry *pte;
> +
> +	if (!ecap_flts(iommu->ecap)) {
> +		pr_err("No first level translation support on %s\n",
> +		       iommu->name);
> +		return -EINVAL;
> +	}
> +
> +	if ((flags & PASID_FLAG_FL5LP) && !cap_fl5lp_support(iommu->cap)) {
> +		pr_err("No 5-level paging support for first-level on %s\n",
> +		       iommu->name);
> +		return -EINVAL;
> +	}
> +
> +	spin_lock(&iommu->lock);
> +	pte = intel_pasid_get_entry(dev, pasid);
> +	if (!pte) {
> +		spin_unlock(&iommu->lock);
> +		return -ENODEV;
> +	}
> +
> +	if (!pasid_pte_is_present(pte)) {
> +		spin_unlock(&iommu->lock);
> +		return -EINVAL;
> +	}
> +
> +	WARN_ON(old_did != pasid_get_domain_id(pte));
> +
> +	pasid_pte_config_first_level(iommu, pte, pgd, did, flags);
> +	spin_unlock(&iommu->lock);
> +
> +	intel_pasid_flush_present(iommu, dev, pasid, old_did, pte);
> +	intel_iommu_drain_pasid_prq(dev, pasid);
> +
> +	return 0;
> +}

pasid_pte_config_first_level() causes the pasid entry to transition from
present to non-present and then to present. In this case, calling
intel_pasid_flush_present() is not accurate, as it is only intended for
pasid entries transitioning from present to present, according to the
specification.

It's recommended to move pasid_clear_entry(pte) and
pasid_set_present(pte) out to the caller, so ...

For setup case (pasid from non-present to present):

- pasid_clear_entry(pte)
- pasid_pte_config_first_level(pte)
- pasid_set_present(pte)
- cache invalidations

For replace case (pasid from present to present)

- pasid_pte_config_first_level(pte)
- cache invalidations

The same applies to other types of setup and replace.

--
baolu