Re: [PATCH v2 2/4] vfio: Move the Intel no-snoop control off of IOMMU_CACHE

[Jason Gunthorpe <jgg@nvidia.com>]

On Tue, Apr 12, 2022 at 09:13:27PM +0800, Lu Baolu wrote:

> > > > btw as discussed in last version it is not necessarily to recalculate
> > > > snoop control globally with this new approach. Will follow up to
> > > > clean it up after this series is merged.
> > > Agreed. But it also requires the enforce_cache_coherency() to be called
> > > only after domain being attached to a device just as VFIO is doing.
> > that actually makes sense, right? w/o device attached it's pointless to
> > call that interface on a domain...
> 
> Agreed. Return -EOPNOTSUPP or -EINVAL to tell the caller that this
> operation is invalid before any device attachment.

That is backwards. enforce_cache_coherency() succeeds on an empty
domain and attach of an incompatible device must fail.

Meaning you check the force_snoop flag in the domain when attaching
and refuse to attach the device if it cannot support it. This will
trigger vfio to create a new iommu_domain for that device and then
enforce_cache_coherency() will fail on that domain due to the
incompatible attached device.

Same scenario if it is the Nth device to be attached to a domain.

Jason