Re: [PATCH v3 1/4] mfd: mxs-lradc: Add support for mxs-lradc MFD

[Stefan Wahren]

Hi,

> Marek Vasut <marex@xxxxxxx> hat am 14. Juli 2016 um 17:38 geschrieben:
> 
> 
> On 07/13/2016 02:49 PM, Lee Jones wrote:
> > On Fri, 01 Jul 2016, Harald Geyer wrote:
> >
> >> Hi Ksenija!
> >>
> >> Ksenija StanojeviÄ writes:
> >>> On Tue, Jun 28, 2016 at 6:28 PM, Lee Jones <lee.jones@xxxxxxxxxx> wrote:
> >>>>> +static int mxs_lradc_add_device(struct platform_device *pdev,
> >>>>> +                             struct mxs_lradc *lradc, char *name, int
> >>>>> i)
> >>>>> +{
> >>>>> +     struct mfd_cell *cell;
> >>>>> +
> >>>>> +     cell = &lradc->cells[i];
> >>>>> +     cell->name = name;
> >>>>> +     cell->platform_data = lradc;
> >>>>> +     cell->pdata_size = sizeof(*lradc);
> >>>>> +
> >>>>> +     return devm_mfd_add_devices(&pdev->dev, -1, cell, 1, NULL, 0,
> >>>>> NULL);
> >>>>> +}
> >>>>
> >>>> Please don't roll your own API.
> >>>>
> >>>> Use 'struct mfd_cell' like everyone else does.
> >>>
> >>> It has been suggested in previous reviews to use separate function to
> >>> register mfd device, and to make mfd_cell allocate dynamically because
> >>> struc mxs-lradc is allocated dynamically.
> >>> But I can revrse changes and make mfd_cells allocate staticaly
> >>> wthout separate function.
> >>
> >> I think making mfd_cells members of struct mxs-lradc will address all
> >> review comments.
> >
> > No, please don't do that either.
> >
> It'd be nice if you explained in detail why not. Otherwise this is just 
> empty splat.

since there is no reply, here is my guess:

static const struct mfd_cell mxs_lradc_devs[] = {
	{
		.name = DRIVER_NAME_ADC,
	},
	{
		.name = DRIVER_NAME_TS,
	},
};

But i'm not sure if we need of_compatible defined here. The intension of this
patch series is to keep the DT binding.

@Lee: Could you please give us a feedback?

@Ksenija: Still motivated for next round?

Regards
Stefan

> 
> -- 
> Best regards,
> Marek Vasut