Re: [PATCH v2 1/2] dt-bindings: Update iProc GPIO bindings

[Rob Herring]

On Wed, May 04, 2016 at 09:29:04AM -0700, Ray Jui wrote:
> Hi Rob,
> 
> On 5/4/2016 6:20 AM, Rob Herring wrote:
> >On Mon, May 02, 2016 at 01:51:47PM -0700, Ray Jui wrote:
> >>Update the iProc GPIO binding document to add new compatible strings
> >>"brcm,iproc-gpio-v2" and "brcm,iproc-gpio-v3" for the 2nd and 3rd
> >>generation of the iProc GPIO controllers
> >>
> >>Signed-off-by: Ray Jui <ray.jui@xxxxxxxxxxxx>
> >>---
> >> Documentation/devicetree/bindings/pinctrl/brcm,iproc-gpio.txt | 11 ++++++++++-
> >> 1 file changed, 10 insertions(+), 1 deletion(-)
> >>
> >>diff --git a/Documentation/devicetree/bindings/pinctrl/brcm,iproc-gpio.txt b/Documentation/devicetree/bindings/pinctrl/brcm,iproc-gpio.txt
> >>index e427792..3a56649 100644
> >>--- a/Documentation/devicetree/bindings/pinctrl/brcm,iproc-gpio.txt
> >>+++ b/Documentation/devicetree/bindings/pinctrl/brcm,iproc-gpio.txt
> >>@@ -4,7 +4,16 @@ Required properties:
> >>
> >> - compatible:
> >>     Must be "brcm,cygnus-ccm-gpio", "brcm,cygnus-asiu-gpio",
> >>-    "brcm,cygnus-crmu-gpio" or "brcm,iproc-gpio"
> >>+    or "brcm,cygnus-crmu-gpio" for Cygnus SoCs
> >>+
> >>+    "brcm,iproc-gpio" for the first generation of the GPIO controller that
> >>+    supports full-featured pinctrl and GPIO functions used in iProc based SoCs
> >>+
> >>+    "brcm,iproc-gpio-v2" for the second generation of the GPIO controller that
> >>+    has the drive strength pinctrl support disabled, e.g., in the iProc NSP SoC
> >>+
> >>+    "brcm,iproc-gpio-v3" for the third generation of the GPIO controller that
> >>+    has the general pinctrl support completely disabled
> >
> >You can have these for driver matching, but you still need SoC specific
> >compatible strings.
> >
> >Rob
> >
> 
> I think I'm missing something and hope you can help to clarify here. It
> looks like the notion of v2, v3 should only be used if there's an indeed an
> revision update on the controller IP itself, correct?

Yes, but only if v2, v3 is actually a meaningful number rather than 
something made up. You have to have a well defined process of IP 
revisions which I have not seen during my time in chip companies, so I'm 
generally suspicious of version numbers. The exception is FPGA IP 
blocks.

> In our case, if the same revision of GPIO controller is used but instead
> synthesized differently with different SoCs, then you are suggesting here
> that it should be dealt with SoC specific compatible string, correct?

Yes, that and/or integration differences is why you need SoC specific 
compatible strings. A block could be "identical" but have different max 
frequency for example.
 
> For example, for the GPIO controller on NSP where drive strength is
> disabled, the compatible string in DT should look like this?
> 
> compatible = "brcm,iproc-gpio", "brcm,iproc-gpio-nsp";

Yes, but reverse the order. Most specific to least specific.

> 
> For the GPIO controller on Stingray where pinconf is completely disabled,
> the compatible string in DT should look like:
> 
> compatible = "brcm,iproc-gpio", "brcm,iproc-gpio-stingray";
> 
> Is that correct?
> 
> Thanks,
> 
> Ray