Re: [PATCH 1/3] dt-bindings: mmc: Add 'fixed-emmc-driver-type-hs{200,400}'

From: Wolfram Sang <wsa@the-dreams.de>
To: Rob Herring <robh@kernel.org>
Cc: Eugeniu Rosca <erosca@de.adit-jv.com>,
	Ulf Hansson <ulf.hansson@linaro.org>,
	Adrian Hunter <adrian.hunter@intel.com>,
	Wolfram Sang <wsa+renesas@sang-engineering.com>,
	linux-mmc@vger.kernel.org,
	Linus Walleij <linus.walleij@linaro.org>,
	Mathieu Malaterre <malat@debian.org>, Pavel Machek <pavel@ucw.cz>,
	devicetree@vger.kernel.org, linux-kernel@vger.kernel.org,
	Eugeniu Rosca <roscaeugeniu@gmail.com>
Subject: Re: [PATCH 1/3] dt-bindings: mmc: Add 'fixed-emmc-driver-type-hs{200,400}'
Date: Tue, 12 Nov 2019 22:19:50 +0100	[thread overview]
Message-ID: <20191112211950.GB3402@kunai> (raw)
In-Reply-To: <20191107003907.GA22634@bogus>

[-- Attachment #1: Type: text/plain, Size: 2811 bytes --]

Hi everyone,

> > The first question which pops up in my mind is related to the meaning
> > of each item. The option which I envision based on your proposal is:
> > 
> >   * minItems: 1
> >   * maxItems: 2
> >   * Item[0]: Presumably equivalent to the current
> >     "fixed-emmc-driver-type", i.e. the strength value applied in both
> >     HS200 and HS400 modes.
> >   * Item[1] (optional): Presumably equivalent to
> >     "fixed-emmc-driver-type-hs400" proposed in this series. If this
> >     element is provided, the first one should likely change its role
> >     and become an equivalent of "fixed-emmc-driver-type-hs200" from
> >     this series.
> >   + Pro: Full backward compatibility. No need to touch the existing
> >     users of "fixed-emmc-driver-type".
> >   - Con: Not sure we have such DT bindings which dynamically change
> >     their semantics based on the usage pattern.
> >   - Con: Can't easily achieve the same flexibility as accomplished in
> >     this series. For example, current implementation allows users to
> >     define each of the three parameters (i.e. HSx00-agnostic drive
> >     strength, HS200 and HS400 specific drive strengths) individually,
> >     as well as in all possible combinations. This might be needed if,
> >     in certain HSx00 mode, users still need to rely on the
> >     RAW/unmodified drive strength. I am unsure if/how this can be
> >     achieved with an array OF property with a constant or variable
> >     number of elements (I try to sketch one solution below).
> > 
> > One option to achieve a similar degree of flexibility by using an array
> > OF property (instead of several u32 properties) would be to agree on a
> > convention based on magic values, i.e. below DT one-liner could be an
> > example of providing solely the "fixed-emmc-driver-type-hs200" value
> > (based on the agreement that 0xFF values are discarded by the driver):
> > 
> >     fixed-emmc-driver-type = <0xFF 0x1 0xFF>;
> 
> I don't understand why you have 3 values instead of 2.

Because he sketches maximum flexibility here. Have a non-HS (default)
value, one for HS200, and one for HS400:

	fixed-emmc-driver-type = <[default] [HS200] [HS400]>;

> I would just use -1 if you want to ignore an entry. If that's the common 

'-1' sounds good to me, too.

> case, then I'd stick with what you originally proposed. If rare, then I 
> think an array is the better route.

What I have seen so far: setting drive strength alone is more on the
rare side. Setting specific values for default and HS200/400 seems even
more rare to me. With this patchset, it is the first time I hear about
it.

Yet, my experience might be a bit limited, maybe others (Hi Ulf! ;)) can add
their views, too?

Regards,

   Wolfram

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]