On Mon, Jun 21, 2021 at 08:46:54AM +0200, Krzysztof Kozlowski wrote:

On 18/06/2021 21:47, Rob Herring wrote:

quoted

On Thu, Jun 3, 2021 at 10:49 AM Thierry Reding [off-list ref] wrote:

quoted

From: Thierry Reding <redacted>

The ARM SMMU instantiations found on Tegra186 and later need inter-
operation with the memory controller in order to correctly program
stream ID overrides.

Furthermore, on Tegra194 multiple instances of the SMMU can gang up
to achieve higher throughput. In order to do this, they have to be
programmed identically so that the memory controller can interleave
memory accesses between them.

Add the Tegra186 compatible string to make sure the interoperation
with the memory controller can be enabled on that SoC generation.

Signed-off-by: Thierry Reding <redacted>
---
 Documentation/devicetree/bindings/iommu/arm,smmu.yaml | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/Documentation/devicetree/bindings/iommu/arm,smmu.yaml b/Documentation/devicetree/bindings/iommu/arm,smmu.yaml
index 9d27aa5111d4..1181b590db71 100644
--- a/Documentation/devicetree/bindings/iommu/arm,smmu.yaml
+++ b/Documentation/devicetree/bindings/iommu/arm,smmu.yaml

@@ -54,8 +54,14 @@ properties:
           - const: arm,mmu-500
       - description: NVIDIA SoCs that program two ARM MMU-500s identically
         items:
+      - description: NVIDIA SoCs that require memory controller interaction

This is not valid jsonschema:

/builds/robherring/linux-dt/Documentation/devicetree/bindings/iommu/arm,smmu.yaml:
properties:compatible:oneOf:4:items: 'anyOf' conditional failed, one
must be fixed:
None is not of type 'object', 'boolean'
None is not of type 'array'
from schema $id: http://json-schema.org/draft-07/schema#
/builds/robherring/linux-dt/Documentation/devicetree/bindings/iommu/arm,smmu.yaml:
properties:compatible:oneOf:4:items: 'oneOf' conditional failed, one
must be fixed:
None is not of type 'object'
None is not of type 'array'
from schema $id: http://devicetree.org/meta-schemas/keywords.yaml#
/builds/robherring/linux-dt/Documentation/devicetree/bindings/iommu/arm,smmu.yaml:
properties:compatible:oneOf:4:items: 'oneOf' conditional failed, one
must be fixed:
None is not of type 'object'
None is not of type 'array'
from schema $id: http://devicetree.org/meta-schemas/string-array.yaml#
/builds/robherring/linux-dt/Documentation/devicetree/bindings/iommu/arm,smmu.yaml:
properties:compatible:oneOf:5:items: 'oneOf' conditional failed, one
must be fixed:
[{'enum': [{'const': 'nvidia,tegra194-smmu'}, {'const':
'nvidia,tegra186-smmu'}]}, {'const': 'nvidia,smmu-500'}] is not of
type 'object'
{'const': 'nvidia,tegra194-smmu'} is not of type 'string'
{'const': 'nvidia,tegra186-smmu'} is not of type 'string'
from schema $id: http://devicetree.org/meta-schemas/string-array.yaml#


This was not reviewed nor tested since the DT list was not Cc'ed.

Ugh, I see now weird empty item on a list... and not only DT list was
skipped - Thierry did not Cc you either.

This seemed like a too trivial addition to waste Rob's time on, so I
didn't add him (or the DT list for that matter) on Cc. The ARM SMMU
maintainers had reviewed this, which seemed like it was enough for what
the DT bindings change was doing.

In any case, I clearly should've checked the DT binding check output
more carefully. It's rather messy for Tegra because there's quite a few
that we haven't converted yet. I'll have to resume my effort to convert
the remaining ones and fixup the device trees so that we can actually
run the DT binding and DTB validation checks more usefully.

My bad, I did not check that patch thoroughly before applying.

Thierry, please Cc folks mentioned by get_maintainer.pl. Either sent a
fix or a revert, if fix needs more time.

I've sent out a follow-up fix that removes the two bogus lines. It looks
like that was the result of a bad conflict resolution on my part.

Additionally, why the patch changes reg to "minItems: 1" for
nvidia,tegra194-smmu?

This is because originally the Tegra194 SMMU nodes were supposed to only
represent a "dual" instance. However, on Tegra194 there are three SMMU
instances in total, with the third instance (dedicated for isochronous
traffic) being completely separate and having only a single range of
registers.

That third instance was previously supposed to be covered by the normal
"arm,mmu-500" compatible string, but given that we really need that
interoperation between SMMU and memory controller for SID override
programming, we need the Tegra-specific compatible for the ISO instance
of the SMMU as well. And since that uses only one set of registers,
minItems had to become 1.

Thierry