Re: [PATCH 3/3] mfd: bd9571mwv: Add support for BD9574MWF

From: Geert Uytterhoeven
Date: Thu Dec 10 2020 - 03:20:50 EST


Hi Matti, Shimoda-san,

On Thu, Dec 10, 2020 at 8:33 AM Vaittinen, Matti
<Matti.Vaittinen@xxxxxxxxxxxxxxxxx> wrote:
> On Thu, 2020-12-10 at 04:44 +0000, Yoshihiro Shimoda wrote:
> > > From: Geert Uytterhoeven, Sent: Wednesday, December 9, 2020 10:30
> > > PM
> > <snip>
> > > > --- a/drivers/mfd/bd9571mwv.c
> > > > +++ b/drivers/mfd/bd9571mwv.c
> > > >
> > > > @@ -182,6 +272,8 @@ static int bd9571mwv_probe(struct i2c_client
> > > > *client,
> > > > product_code = (unsigned int)ret;
> > > > if (product_code == BD9571MWV_PRODUCT_CODE_VAL)
> > > > bd->data = &bd9571mwv_data;
> > > > + else if (product_code == BD9574MWF_PRODUCT_CODE_VAL)
> > > > + bd->data = &bd9574mwf_data;
> > > >
> > > > if (!bd->data) {
> > > > dev_err(bd->dev, "No found supported device
> > > > %d\n",
> > >
> > > While BD9571MWV and BD9574MWF can be distinguished at runtime,
> > > I think it would still be a good idea to document a
> > > "rohm,bd9574mwf"
> > > compatible value in the DT bindings, and let the driver match on
> > > that.
> >
> > In this driver point of view, we can use such the DT bindings,
> > however, in the board point of view, it's difficult to describe
> > which chip is installed on r8a77990-ebisu.dts. So, I'd like to
> > keep this runtime detection.

To clarify: I meant to document and add the compatible value, but
treat both compatible values the same in the driver, and still do runtime
probing.

> First of all - I don't want to insist changes here so my comment can be
> ignored. I would definitely like seeing the support for BD9574 in-tree!
>
> But as a 'nit':
> I don't know what are the difficulties you are referring to so it's
> hard for me to comment this. Without better understanding of board dts
> files - I think BD9574MWF should really have own compatible as I
> understood it is different from the BD9571MWV. Relying on product code
> probing sounds like something that may easily break when/if new
> variants are produced. ( I've seen new HW variants using the same
> ID information being produced in previous companies I've worked. Sure

Yes, this happens from time to time, unfortunately.

> ROHM wouldn't do this but still... :] ). And producing boards where DTS
> does not allow describing the correct components sounds like asking for
> a nose-bleed to me... If probing of IC type fails AND there is devices
> with wrong PMIC information burned in DT - then fixing it can be a
> nightmare. So I would really try make DTS files such that they can be

The issue we're facing is that older Ebisu-4D boards have BD9571, while
newer boards have BD9574. The schematics say "BD9574MWF-M (tentative
ver:BD9571TL1_E3)", so it looks like both parts are pin-compatible
(ignoring missing pins for AVS, LDO1, LDO2, and LDO6 on BD9574).
Hence arch/arm64/boot/dts/renesas/r8a77990-ebisu.dts has a device node
compatible with "rohm,bd9571mwv". If we have runtime probing, we can
keep on using that for both variants.

> changed to match the actual board. (Perhaps introduce the compatible
> for BD9574MWF - make this driver to match both of the PMICs - leave the
> runtime probing here for now - and in parallel work with the DTS files
> so that eventually the probing can be removed(?) That was my 10 cents
> on this topic :] )

Exactly. Thanks!

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds