Re: [PATCH v3 06/10] iommu/vt-d: Set the nested domain to a device

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

On 5/24/23 3:22 PM, Tian, Kevin wrote:
>> From: Liu, Yi L <yi.l.liu@intel.com>
>> Sent: Thursday, May 11, 2023 10:51 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);
>> +
>> +	/* Is s2_domain compatible with this IOMMU? */
>> +	ret = prepare_domain_attach_device(&dmar_domain->s2_domain-
>>> domain, dev);
>> +	if (ret) {
>> +		dev_err_ratelimited(dev, "s2 domain is not compatible\n");
>> +		return ret;
>> +	}
> 
> this also includes logic to trim higher page levels:
> 
> 	/*
> 	 * Knock out extra levels of page tables if necessary
> 	 */
> 	while (iommu->agaw < dmar_domain->agaw) {
> 		struct dma_pte *pte;
> 
> 		pte = dmar_domain->pgd;
> 		if (dma_pte_present(pte)) {
> 			dmar_domain->pgd = phys_to_virt(dma_pte_addr(pte));
> 			free_pgtable_page(pte);
> 		}
> 		dmar_domain->agaw--;
> 	}
> 
> What's the background of doing such truncation instead of simply
> failing the request?

This code existed a long time ago. I'm not sure if it's still reasonable
so far.

> In any means it's probably fine before the domain includes any mapping
> but really unreasonable to apply it to an existing s2 when it's used as
> a parent.

But for the new nested translation, it is obviously unreasonable.

Let me revisit it.

Best regards,
baolu