[Linux-stm32] [PATCH 1/3] mmc: Prepare all code for mmc_set_signal_voltage() returning > 0

[Ulf Hansson]

On Thu, 16 Apr 2020 at 16:45, Marek Vasut <marex at denx.de> wrote:
>
> On 4/16/20 2:07 PM, Ulf Hansson wrote:
> > On Thu, 16 Apr 2020 at 12:29, Marek Vasut <marex at denx.de> wrote:
> >>
> >> On 4/15/20 10:40 AM, Ulf Hansson wrote:
> >>> On Wed, 1 Apr 2020 at 21:57, Marek Vasut <marex at denx.de> wrote:
> >>>>
> >>>> Patch all drivers and core code which uses mmc_set_signal_voltage()
> >>>> and prepare it for the fact that mmc_set_signal_voltage() can return
> >>>> a value > 0, which would happen if the signal voltage switch did NOT
> >>>> happen, because the voltage was already set correctly.
> >>>
> >>> I am not sure why you want to change mmc_set_signal_voltage(), can you
> >>> elaborate on that?
> >>>
> >>> I thought we discussed changing mmc_regulator_set_vqmmc(), what am I missing?
> >>
> >> Because mmc_set_signal_voltage() optionally calls
> >> host->ops_>start_signal_voltage_switch() , which can now return value >
> >> 0 , so the rest of the core needs to be patched to cater for that.
> >
> > The issue that you wanted to solve (at least by looking at the
> > original patch) was to understand whether the vqmmc regulator changes
> > voltage level and then take different actions based on that in the
> > mmci host driver.
> >
> > You don't need to change anything related to mmc_set_signal_voltage()
> > to accomplish that, do you? Moreover, I am worried that it may affect
> > the host driver's expectations from when
> > ->start_signal_voltage_switch() may be called.
>
> So, shall I just patch all the sites which use mmc_regulator_set_vqmmc()
> with something like
>
> return mmc_regulator_set_vqmmc();
> becomes
> ret = mmc_regulator_set_vqmmc();
> if (ret > 0)
>         ret = 0;
> return ret;

Yes, something along the lines of that. Or if you think it's better to
make mmc_regulator_set_vqmmc() to pass an out parameter. Whatever you
think looks best, I am fine with whatever.

Kind regards
Uffe