Re: [PATCH v2 18/19] PCI: Replace untrusted state with device trust

From: Jason Gunthorpe

Date: Tue Nov 10 2026 - 09:58:58 EST


On Tue, Nov 10, 2026 at 02:22:37PM +0200, Leon Romanovsky wrote:
> On Mon, Nov 09, 2026 at 01:53:02PM +0000, Robin Murphy wrote:
> > On 09/11/2026 7:33 am, Leon Romanovsky wrote:
> > [...]
> > > diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
> > > index b486b8bbd1fcd..8be2a246d4903 100644
> > > --- a/drivers/iommu/iommu.c
> > > +++ b/drivers/iommu/iommu.c
> > > @@ -1879,7 +1879,7 @@ static int iommu_get_def_domain_type(struct iommu_group *group,
> > > static int iommu_get_default_domain_type(struct iommu_group *group,
> > > int target_type)
> > > {
> > > - struct device *untrusted = NULL;
> > > + struct device *restricted = NULL;
> > > struct group_device *gdev;
> > > int driver_type = 0;
> > > @@ -1901,14 +1901,15 @@ static int iommu_get_default_domain_type(struct iommu_group *group,
> > > driver_type = iommu_get_def_domain_type(group, gdev->dev,
> > > driver_type);
> > > - if (dev_is_pci(gdev->dev) && to_pci_dev(gdev->dev)->untrusted) {
> > > + if (dev_is_pci(gdev->dev) &&
> > > + pci_dev_default_is_restricted(to_pci_dev(gdev->dev))) {
> >
> > I think this should be the generic device_is_adversarial(gdev->dev) now - in
> > fact probably all the dev_is_pci()/to_pci_dev() stanzas aren't conceptually
> > PCI-specific, just an artefact of the current "untrusted" flag being so.
>
> I wanted to do the same, but iommu_get_default_domain_type() may be called
> before the PCI device's trust level has been set. At that point, we are
> still in "disabled" mode, whereas device_is_adversarial() checks for
> "adversarial" mode.
>
> The below is from AI, I asked to present me the path for that:
> --------------------------------------------------------------------------------
> For a newly discovered PCI device with an available IOMMU and a new IOMMU group:
> pci_scan_single_device()
> pci_device_add()
> device_add()
> bus_notify(dev, BUS_NOTIFY_ADD_DEVICE)
> blocking_notifier_call_chain()
> notifier_call_chain()
> iommu_bus_notifier()
> iommu_probe_device(dev)
> __iommu_probe_device(dev, NULL)
> iommu_setup_default_domain(group, 0)
> iommu_get_default_domain_type(group, 0)
>
> For a device whose initial policy is adversary, its state during that call is:
> dev->trust_policy = DEVICE_TRUST_POLICY_ADVERSARY
> dev->trust_level = DEVICE_TRUST_DISABLED
> dev->trust_committed_level = DEVICE_TRUST_DISABLED
> dev->driver = NULL
>
> device_is_adversarial(dev) = false
>
> The active level changes later, during binding:
> really_probe()
> device_trust_prepare()
> __device_trust_prepare()
> WRITE_ONCE(dev->trust_level, DEVICE_TRUST_ADVERSARY)
> dev->bus->dma_configure()
> call_driver_probe()

The ordering here is wrong, this is a consequence of splitting this
into different series. I don't think we need to fix it in this series,
but as a followup. For now I would aim to make 'no functional change'
and still follow the ACPI selection during default domain creation.

The long term goal is to detect that the trust level is disabled
during iommu_probe_device() (ie we are running in a secure booting
mode), attach the device to BLOCKED (which should be a NOP) and defer
allocating the default domain.

This would preserve the DRTM iommu & ATS blocked behavior through
until driver binding, giving userspace a chance to do an attestation,
turn on link ide and so on before exposing the system to DMA from the
device.

Once userspace sets the trust level and binds a driver then we can
allocate a default domain using the user set trust level and avoid the
above problem.

For the CC VM use case today there won't be a vIOMMU, and there is an
indepenent control for DMA ENABLE in the TSM driver so this doesn't
matter right now for CC.

Probably put this detail in the commit message

Jason