Re: [PATCH v3 1/1] PCI: Add translated request only flag for pci_enable_pasid()

[Baolu Lu]

On 2023/2/1 8:14, Bjorn Helgaas wrote:
> > > > The AMD iommu driver allocates a new domain (called v2 domain) for the
> > > "v2 domain" needs to be something greppable -- an identifier,
> > > filename, etc.
> > The code reads,
> >
> > 2052         if (iommu_feature(iommu, FEATURE_GT) &&
> > 2053             iommu_feature(iommu, FEATURE_PPR)) {
> > 2054                 iommu->is_iommu_v2   = true;
> >
> > So, how about
> >
> > ..The AMD GPU has a private interface to its own AMD IOMMU, which could
> > be detected by the FEATURE_GT && FEATURE_PPR features. The AMD iommu
> > driver allocates a special domain for the GPU device ..
> Where is this special domain allocated?  I think the above tests for
> *IOMMU*  features (I assume "GTSup: Guest translations supported" and
> "PPRSup: Peripheral page request support" based on the AMD IOMMU
> spec).  It doesn't test that this is a GPU.

From the discussion, my understanding is that the IOMMU is a GPU
dedicated IOMMU. The translated-request-only feature is enumerated
through the IOMMU feature bits. However, I am not familiar with AMD
architecture. AMD guys may have a better explanation.

> This change doesn't feel safe for all possible devices that have a
> PASID Capability because we don't know whether they*always*  use
> Translated addresses with PASID TLPs.

I tend to think you are right. :-)

Best regards,
baolu