Re: [RFC v1 1/2] scmi: Introduce pinctrl SCMI protocol driver

From: Cristian Marussi <cristian.marussi@arm.com>
To: Linus Walleij <linus.walleij@linaro.org>
Cc: Oleksii Moisieiev <Oleksii_Moisieiev@epam.com>,
	Peng Fan <peng.fan@oss.nxp.com>,
	"sudeep.holla@arm.com" <sudeep.holla@arm.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-arm-kernel@lists.infradead.org" 
	<linux-arm-kernel@lists.infradead.org>,
	"linux-gpio@vger.kernel.org" <linux-gpio@vger.kernel.org>,
	"michal.simek@amd.com" <michal.simek@amd.com>,
	"vincent.guittot@linaro.org" <vincent.guittot@linaro.org>,
	"souvik.chakravarty@arm.com" <souvik.chakravarty@arm.com>
Subject: Re: [RFC v1 1/2] scmi: Introduce pinctrl SCMI protocol driver
Date: Fri, 21 Apr 2023 10:47:49 +0100	[thread overview]
Message-ID: <ZEJbxTzv6tlBkExS@e120937-lin> (raw)
In-Reply-To: <CACRpkdY1eSdWX6+azn43MO77urVf_t25wgZtuyTwKHw4v+QYNg@mail.gmail.com>

On Fri, Apr 21, 2023 at 11:28:38AM +0200, Linus Walleij wrote:
> On Fri, Apr 21, 2023 at 10:40 AM Oleksii Moisieiev
> <Oleksii_Moisieiev@epam.com> wrote:
> > On 17.04.23 05:55, Peng Fan wrote:
> > > On 4/13/2023 6:04 AM, Cristian Marussi wrote:
> 
> > > Is it possible to extend the spec to support multilple uint32_t for PIN
> > > CONFIG SET?
> > >
> > > With only one uint32_t could not satisfy i.MX requirement.
> > >
> > > Thanks,
> > > Peng.
> > >
> > IIUC you are expecting to have an ability to set some kind of array of
> > uint32_t config values to some specific ConfigType?
> >
> > I'm not sure if it's supported by pintctrl subsystem right now. I was
> > unable to find an example in the existing device-tree pinctrl bindings.
> > This makes me think that this kind of binding is OEM specific.
> >
> > Maybe it can be implemented by adding new IDs to OEM specific range
> > (192-255) which is reserved for OEM specific units (See Table 23 of
> > DEN0056E).
> 

Hi Linus,

> From a pinctrl point of view I do not understand this requirement.
> 
> The pinctrl subsystem in the Linux kernel certainly does not support
> an array of u32 for the pin config, we only support passing a single
> u32 value along with the enumerator (config type), or well it is
> actually 24 bits in Linux, the uppermost 8 bits is for the config type:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/linux/pinctrl/pinconf-generic.h
> 
> /*
>  * Helpful configuration macro to be used in tables etc.
>  */
> #define PIN_CONF_PACKED(p, a) ((a << 8) | ((unsigned long) p & 0xffUL))
> 
> p = parameter (PIN_CONFIG_DRIVE_STRENGTH etc)
> a = argument (value such as in mA)
> 

My (possibly wrong) reasoning on the other reply, is based on the
(possibly equally wrong :D) understanding that what Peng wants is just
the possibility at the spec and the SCMI protocol layer (exposed in
protocol operations) to issue PINCTRL_SET requests containing optionally
an array of multiple ConfigType/Value pairs (which is anyway not supported
by PinCtrl as I understand) instead of a single pair.

... but I can divine (:D)....that soon a new SCMI spec review/comment/amend
cycle will be coming for people reading this...

Thanks,
Cristian