Re: [PATCH 1/4] iommu/vt-d: Implement set device pasid op for default domain

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

On 3/7/23 1:41 AM, Jason Gunthorpe wrote:
> On Mon, Mar 06, 2023 at 09:36:26AM -0800, Jacob Pan wrote:
> 
>>> It seems like all this is doing is flushing the PRI queue.
>>> A domain should have a dedicated flag unrelated to the type if it is
>>> using PRI and all PRI using domains should have the PRI queue flushed
>>> here, using the same code as flushing the PRI for a RID attachment.
>> Yes, or if the teardown op is domain-specific, then it works too?
> That could only sense if we end up creating a PRI domain type too..
> 
> Right now PRI is messy because it doesn't really work with unmanaged
> domains and that is something that will have to get fixed sort of
> soonish
> 
> Once PRI is a proper thing then "SVA" is just a PRI domain that has a
> non-managed from-the-mm page table pointer.
> 
> Most of the driver code marked svm should entirely disappear.

Agreed.

There is also another code consolidation we discussed earlier, that is
moving mmu_notifier_register/unregister to drivers/iommu/iommu-sva.c.
It's common for all SVA-capable iommu drivers.

With PRI domain and mm_notifier addressed, we could safely remove the
switch here. SVA domain will be nothing special. :-)

Best regards,
baolu