[PATCH 08/12] doc: binding: pwrseq-usb-generic: add binding doc for generic usb power sequence driver

[Rob Herring]

On Mon, Jun 20, 2016 at 07:26:51PM +0800, Peter Chen wrote:
> On Fri, Jun 17, 2016 at 12:16:48PM -0500, Rob Herring wrote:
> > On Fri, Jun 17, 2016 at 5:09 AM, Peter Chen <peter.chen at nxp.com> wrote:
> > > Add binding doc for generic usb power sequence driver, and update
> > > generic usb device binding-doc accordingly.
> > >
> > > Signed-off-by: Peter Chen <peter.chen at nxp.com>
> > > ---
> > >  .../bindings/power/pwrseq/pwrseq-usb-generic.txt   | 31 ++++++++++++++++++++++
> > >  .../devicetree/bindings/usb/usb-device.txt         |  2 ++
> > >  2 files changed, 33 insertions(+)
> > >  create mode 100644 Documentation/devicetree/bindings/power/pwrseq/pwrseq-usb-generic.txt
> > >
> > > diff --git a/Documentation/devicetree/bindings/power/pwrseq/pwrseq-usb-generic.txt b/Documentation/devicetree/bindings/power/pwrseq/pwrseq-usb-generic.txt
> > > new file mode 100644
> > > index 0000000..8ad98382
> > > --- /dev/null
> > > +++ b/Documentation/devicetree/bindings/power/pwrseq/pwrseq-usb-generic.txt
> > > @@ -0,0 +1,31 @@
> > > +The power sequence for generic USB Devices
> > > +
> > > +Some hard-wired USB devices need to do power sequence to let the
> > > +device work normally, the typical power sequence like: enable USB
> > > +PHY clock, toggle reset pin, etc. But current Linux USB driver
> > > +lacks of such code to do it, it may cause some hard-wired USB devices
> > > +works abnormal or can't be recognized by controller at all. The
> > > +power sequence will be done before this device can be found at USB
> > > +bus.
> > > +
> > > +Required properties:
> > > +- compatible : contains "usb-pwrseq-generic".
> > 
> > In case I have not been clear, no.
> > 
> > I am not going to accept anything along the lines of the current mmc
> > pwrseq. I am basically okay with Krzysztof's proposal as it is *only*
> > an added property and not a duplication of information. I'd suggest
> > you figure out how to make the kernel work with that rather than
> > trying to work-around whatever kernel limitations there are.
> > 
> 
> I see.
> 
> Would you agree with below:

In general, yes. There are some points being discussed in the other 
thread.

> &usbotg1 {
> 	vbus-supply = <&reg_usb_otg1_vbus>;
> 	pinctrl-names = "default";
> 	pinctrl-0 = <&pinctrl_usb_otg1_id>;
> 	status = "okay";
> 
> 	#address-cells = <1>;
> 	#size-cells = <0>;
> 	hub: genesys at 1 {
> 		compatible = "usb5e3,608";
> 		reg = <1>;
> 
> 		power-sequence;
> 		reset-gpios = <&gpio4 5 GPIO_ACTIVE_LOW>; /* hub reset pin */
> 		reset-duration-us = <10>;

I wonder if this really needs to be specified. A sufficiently long 
default should be good enough for 90 something % of devices.

> 		clocks = <&clks IMX6SX_CLK_CKO>;
> 
> 		#address-cells = <1>;
> 		#size-cells = <0>;
> 		ethernet: asix at 1 {
> 			compatible = "usbb95,1708";
> 			reg = <1>;
> 
> 			power-sequence;
> 			reset-gpios = <&gpio4 6 GPIO_ACTIVE_LOW>; /* ethernet_rst */
> 			reset-duration-us = <15>;
> 			clocks = <&clks IMX6SX_CLK_IPG>;
> 		};
> 	};
> };
> 
> If the node has property "power-sequence", the pwrseq core will create
> related platform device, and the driver under pwrseq driver will handle
> power sequence stuffs. 

This I have issue with. If you are creating a platform device here, you 
are trying to work-around limitations in the linux driver model. Either 
we need some sort of pre-probe hook to the drivers to call or each 
parent node driver is responsible for checking and calling pwr-seq 
functions for child nodes. e.g. The host controller calls pwr-seq for 
the hub, the hub driver calls the power seq for the asix chip. Soon as 
we have a case too complex for the generic pwr-seq, we're going to need 
the pre-probe hook as I don't want to see a continual expansion of 
generic pwr-seq binding for ever more complex cases.

> The property below "power-sequence" will be
> handled at pwrseq driver.

You cannot rely on order other than for logical readability. There are 
no guarantees of either the dtb or kernel ordering of properties.

Rob