On 2024/11/5 11:30, Baolu Lu wrote:
On 11/4/24 21:18, Yi Liu wrote:
There is no known usage that will attach SVA domain or detach SVA domain
by replacing PASID to or from SVA domain. It is supposed to use the
iommu_sva_{un}bind_device() which invoke the
iommu_{at|de}tach_device_pasid().
So Intel iommu driver decides to fail the domain replacement if the old
domain or new domain is SVA type.
I would suggest dropping this patch.
The iommu driver now supports switching SVA domains to and from other
types of domains. The current limitation comes from the iommu core,
where the SVA interface doesn't use replace.
I also noticed other iommu drivers (like ARM SMMUv3 and AMD iommu) would
support SVA replacement. So may we also make intel_svm_set_dev_pasid() be
able to handle the case in which @old==non-null? At first, it looks to be
dead code when SVA interface does not use replace. But this may make all
drivers aligned.
Looking forward, iommu_detach_device_pasid() will be converted to switch
an SVA domain to a blocking domain. Once this is implemented, perhaps
the check and failure scenario will no longer be relevant.
you are right.
--
Regards,
Yi Liu
