[PATCH v10 03/15] ARM: sunxi: Add driver for SD/MMC hosts found on Allwinner sunxi SoCs

[Maxime Ripard]

On Thu, May 08, 2014 at 02:30:55PM +0200, Hans de Goede wrote:
> Hi,
> 
> On 05/08/2014 02:17 PM, Ulf Hansson wrote:
> > On 8 May 2014 13:26, Hans de Goede <hdegoede at redhat.com> wrote:
> >> Hi,
> >>
> >> On 05/05/2014 10:33 PM, Ulf Hansson wrote:
> >>> [snip]
> >>>
> >>>> On 05/05/2014 02:41 PM, Ulf Hansson wrote:
> >>>>>> +struct sunxi_mmc_host {
> >>>>>> +       struct mmc_host *mmc;
> >>>>>> +       struct regulator *vmmc;
> >>>>>
> >>>>> Instead of having a specific regulator for this driver, please use the
> >>>>> mmc_regulator_get_supply API.
> >>>>
> >>>> We cannot use mmc_regulator_get_supply because for the sunxi mmc controller
> >>>> not only vqmmc but also vmmc itself is optional, and mmc_regulator_get_supply
> >>>> calls devm_regulator_get rather then devm_regulator_get_optional for vmmc.
> >>>
> >>> Is that because the mmc controller handle the power to the card or
> >>> because you have a fixed supply?
> >>>
> >>> Having a fixed regulator supply could easily be set up in DT, which
> >>> then also dynamically gives you the ocr mask instead of having a them
> >>> "hard coded".
> >>
> >> It is because the sdcard slot power tends to be hooked directly to the 3.3V
> >> of the board. So in a sense this is a fixed regulator, but I really, REALLY
> >> don't want to add fixed regulator boilerplate to all sunxi dts files for this.
> > 
> > So, how would you then distinguish between let's say a 3.1V and 3.3V
> > fixed regulator? That is something that is board specific, thus I
> > don't think you can get away from not adding them to DT.
> 
> All boards I've seen sofar use 3.3V which seems sensible since that is
> what the spec says you must supply to SDSC cards. I agree that if a board
> differs from the standard 3.3V, a fixed regulator node specifying the
> voltage should be added.

I think we introduced the sunxi-regulators DTSI just for this :)

I'd really like to start stabilizing a bit the DT and at least
consider being able to use an older DT on a newer kernel. If we take
such approach, I'm afraid it will break at some point.

Maxime

-- 
Maxime Ripard, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20140508/5011128a/attachment.sig>