On Fri, 2021-07-16 at 11:33 -0600, Rob Herring wrote:

On Wed, Jun 30, 2021 at 10:49:33AM +0800, Jianjun Wang wrote:

quoted

Add property to disable dvfsrc voltage request, if this property
is presented, we assume that the requested voltage is always
higher enough to keep the PCIe controller active.

Signed-off-by: Jianjun Wang <redacted>
---
 .../devicetree/bindings/pci/mediatek-pcie-gen3.yaml       | 8
++++++++
 1 file changed, 8 insertions(+)

diff --git a/Documentation/devicetree/bindings/pci/mediatek-pcie-

gen3.yaml b/Documentation/devicetree/bindings/pci/mediatek-pcie-
gen3.yaml
index e7b1f9892da4..3e26c032cea9 100644

--- a/Documentation/devicetree/bindings/pci/mediatek-pcie-gen3.yaml
+++ b/Documentation/devicetree/bindings/pci/mediatek-pcie-gen3.yaml

@@ -96,6 +96,12 @@ properties:
   phys:
     maxItems: 1
 
+  disable-dvfsrc-vlt-req:
+    description: Disable dvfsrc voltage request, if this property

is presented,
+      we assume that the requested voltage is always higher enough
to keep
+      the PCIe controller active.
+    type: boolean

What determines setting this property? Can it be implied by the 
compatible (which should be SoC specific).

Is this property specific to PCIe controller? 

Wouldn't the request be harmless to make the voltage request even if
not 
needed?

I think this probably should be addressed in a common way as part of 
other QoS, devfreq, etc. requirements for devices.

Rob

Hi Rob,

Thanks for your review and sorry for the late response.

We have internal discussion and we agree with that this feature should
not be specific to the PCIe controller, we need to find a common way to
do this. 

But as the driver of dvfsrc is not finished the upstream, we don't have
a better solution for now, so we would like to pull back this patch and
will send another patch to disable dvfsrc by default until we find a
common solution to enable it.

Thanks.