[PATCH 2/4] pinctrl: Add skew-delay pin config and bindings

On Wed, Nov 1, 2017 at 10:26 PM, Rob Herring [off-list ref] wrote:

On Sat, Oct 28, 2017 at 03:37:17PM +0200, Linus Walleij wrote:

quoted

Some pin controllers (such as the Gemini) can control the
expected clock skew and output delay on certain pins with a
sub-nanosecond granularity. This is typically done by shunting
in a number of double inverters in front of or behind the pin.
Make it possible to configure this with a generic binding.

Cc: devicetree at vger.kernel.org
Signed-off-by: Linus Walleij <redacted>
---
 Documentation/devicetree/bindings/pinctrl/pinctrl-bindings.txt | 4 ++++
 drivers/pinctrl/pinconf-generic.c                              | 2 ++
 include/linux/pinctrl/pinconf-generic.h                        | 5 +++++
 3 files changed, 11 insertions(+)

diff --git a/Documentation/devicetree/bindings/pinctrl/pinctrl-bindings.txt b/Documentation/devicetree/bindings/pinctrl/pinctrl-bindings.txt
index 4483cc31e531..ad9bbbba36e9 100644
--- a/Documentation/devicetree/bindings/pinctrl/pinctrl-bindings.txt
+++ b/Documentation/devicetree/bindings/pinctrl/pinctrl-bindings.txt

@@ -271,6 +271,10 @@ output-high              - set the pin to output mode with high level
 sleep-hardware-state - indicate this is sleep related state which will be programmed
                        into the registers for the sleep state.
 slew-rate            - set the slew rate
+skew-delay           - this affects the expected clock skew on input pins
+                       and the delay before latching a value to an output
+                       pin. Typically indicates how many double-inverters are
+                       used to delay the signal.

Units?

Number of double-inverters I guess, just as it says.

skew-delay = <1>; // number of inverter pairs

Like with other bindings of this type they are custom.

The Gemini datasheet says "every unit represents about 0.2 ns".

So if you set it to 1, that means (as far as I can guess) this
is shunted in:

   |\    |\
--+ >o--+ >o---
   |/    |/

The signal will take ~0.2 ns to propagate through this.
For 2, there will be four inverters etc.

It is dependent of production technology, and I do not know
if everyone is using something like double-inverters.

We could try to define a value in picoseconds instead so I
would need to set this to 200 in that case if you think that is better:

skew-delay = <200>; // picoseconds

I think it would be kludgy since with increasing clock speeds
maybe picoseconds would be too big even. Nanoseconds is
apparently already a to coarse unit.

Yours,
Linus Walleij