On 20/06/2023 18:31, Nícolas F. R. A. Prado wrote:

On Tue, Jun 20, 2023 at 03:00:00PM +0200, Krzysztof Kozlowski wrote:

quoted

On 20/06/2023 14:46, Nícolas F. R. A. Prado wrote:

quoted

On Tue, Jun 20, 2023 at 10:12:14AM +0200, Krzysztof Kozlowski wrote:

quoted

On 20/06/2023 02:03, Nícolas F. R. A. Prado wrote:

quoted

The binding expects the first register space to be VDEC_SYS. But on
mt8183, which uses the stateless decoders, this space is used only for
controlling clocks and resets, which are better described as separate
clock-controller and reset-controller nodes.

In fact, in mt8173's devicetree there are already such separate
clock-controller nodes, which cause duplicate addresses between the
vdecsys node and the vcodec node. But for this SoC, since the stateful
decoder code makes other uses of the VDEC_SYS register space, it's not
straightforward to remove it.

In order to avoid the same address conflict to happen on mt8183,
since the only current use of the VDEC_SYS register space in
the driver is to read the status of a hardware controlled clock, remove
the VDEC_SYS register space from the binding and describe an extra
syscon that will be used to directly check the hardware status.

Also add reg-names to be able to tell that this new register schema is
used, so the driver can keep backward compatibility.

Signed-off-by: Nícolas F. R. A. Prado <redacted>

---
I dropped the tags from this commit since a syscon is now used instead
of an extra clock.

Changes in v3:
- Removed the active clock
- Added a mediatek,vdecsys syscon property

Changes in v2:
- Merged with patch 1 (media: dt-bindings: mediatek,vcodec: Allow single
  clock for mt8183) to avoid changing number of clocks twice
- Added maxItems to reg-names
- Constrained clocks for each compatible
- Reordered properties for each compatible

 .../media/mediatek,vcodec-decoder.yaml        | 30 +++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/Documentation/devicetree/bindings/media/mediatek,vcodec-decoder.yaml b/Documentation/devicetree/bindings/media/mediatek,vcodec-decoder.yaml
index 1e56ece44aee..2f625c50bbfe 100644
--- a/Documentation/devicetree/bindings/media/mediatek,vcodec-decoder.yaml
+++ b/Documentation/devicetree/bindings/media/mediatek,vcodec-decoder.yaml

@@ -21,8 +21,13 @@ properties:
       - mediatek,mt8183-vcodec-dec
 
   reg:
+    minItems: 11
     maxItems: 12
 
+  reg-names:
+    minItems: 11
+    maxItems: 11

maxItems: 12

quoted

+
   interrupts:
     maxItems: 1
 

@@ -60,6 +65,10 @@ properties:
     description:
       Describes point to scp.
 
+  mediatek,vdecsys:
+    $ref: /schemas/types.yaml#/definitions/phandle
+    description: Phandle to the vdecsys syscon node.
+
 required:
   - compatible
   - reg

@@ -79,8 +88,26 @@ allOf:
     then:
       required:
         - mediatek,scp
+        - mediatek,vdecsys
 
       properties:
+        reg:
+          maxItems: 11
+
+        reg-names:
+          items:
+            - const: misc
+            - const: ld
+            - const: top
+            - const: cm
+            - const: ad
+            - const: av
+            - const: pp
+            - const: hwd
+            - const: hwq
+            - const: hwb
+            - const: hwg
+
         clocks:
           minItems: 1
           maxItems: 1

@@ -101,6 +128,9 @@ allOf:
         - mediatek,vpu
 
       properties:
+        reg:
+          minItems: 12

What about reg-names here? They should be also defined and in sync with
regs.

That's intentional. As described in the commit message, mt8173 will keep having
the VDEC_SYS iospace, while mt8183 won't. And we use the presence of reg-names
to tell them apart.

So, mt8173 has 12 regs, no reg-names and no syscon, while mt8183 has 11 regs,
with reg-names and the syscon.

reg-names is not the way to tell apart variants. Compatible is. If you
add reg-names for one variant, it's expected to have it defined for
other as well, because the order of items in reg is always fixed.

But differentiating with compatible in this case would be wrong, since it's not
not something inherent to the SoC. We really just want to be able to tell
whether the vdecsys iospace is supplied as a reg or as a syscon.

Wait, you should not have one device or even family of devices taking
their IO space with two different methods. It's exactly the same device,
the same bus.

This series focuses on getting the mt8183 decoder working, and as part of that
introduces the binding and DT node for mt8183 with vdecsys as a syscon instead
of a reg, to avoid introducing new 'duplicate unit-address' DT warnings.

I got patches 1, 2, 3 and 6, nothing more so I cannot comment on what
else you are trying to do here. Since you did not cc me, it's not relevant.

Your DTS change does nothing like switching from MMIO to syscon.

But anyway this variant comes with some set of regs and reg-names. Other
variant comes with different set. In all cases they should be defined,
even by "defined" means not allowed.

But in a separate series we could drop vdecsys from mt8173's reg as well,
passing it as a syscon instead, which would solve the warning on that platform,
though some more driver changes would be needed to be able to handle it for that
SoC. The newer SoCs like mt8192, mt8195, etc, should also get vdecsys dropped
from their regs to have a correct memory description.

Sure, but I don't understand how does it affect defining and making
specific regs/reg-names or keeping them loose.

Best regards,
Krzysztof


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel