[PATCH v4 07/13] iommu/arm-smmu-v3: Keep track of attached ssids

[Michael Shavit]

On Sat, Jul 15, 2023 at 8:35 AM Nicolin Chen <nicolinc at nvidia.com> wrote:
> Well, firstly it's kinda odd to have this optimization hidden in
> a big rework patch. And I am not sure if it alone would work for
> all use cases, as it impacts the arm_smmu_atc_inv_domain() that
> passes in a zero ssid, in which case the affected function is not
> used by a pasid case all the time:

To clarify, we haven't changed anything for the existing flow. Before
the latter set_dev_pasid patch is introduced, there's only ever a
single {master,domain} pair in the attached_domain list. When
arm_smmu_atc_inv_domain() is called on a non-pasid domain, it issues a
single atc inv command with substream_valid=false, same as before.
When SVA wants to invalidate a domain, it calls
arm_smmu_atc_inv_domain_ssid to issue a single atc inv command with
that specific PASID, same as before. At no point does the driver ever
rely on the fact that invalidating with substream_valid=false would
invalidate all PASIDs.

This patch series simply follows the precedent set by SVA: when a
domain is attached with a pasid, invalidates those specific pasids.
It'd feel kinda hacky to rely on the fact that invalidating with
substream_valid=false would invalidate all PASIDs.