Re: [PATCH v4 2/4] mmc: core: Add mmc_regulator_set_vqmmc()

[Doug Anderson]

Hi,

On Fri, Mar 20, 2015 at 4:28 AM, Mark Brown <broonie@xxxxxxxxxx> wrote:
> As previously discussed the problem is that there can be a *lot* of
> voltages on a modern regulator with fine grained voltage steps and
> tolerances are also used for things like cpufreq where we care about
> performance.  We need something that doesn't require a linear scan of
> possible values.

Finally getting back to this (sorry for the delay!).  I tried
modifying my patches to keep using the simple implementation of
regulator_set_voltage_tol() but to take two tolerances (lower and
upper).  ...but when I thought about it I decided it wasn't enough.  I
think that doing a proper implementation of
regulator_set_voltage_tol() is going to be a requirement for getting
the MMC core changes posted.

Specifically, I think the right implementation for the MMC core's 3.3V
signaling is something like this:

  /*
   * Bus operating conditions say that card should accept input
   * between (0.625 * VDD) and (VDD + 0.3), so we'll use those
   * as tolerances.
   */
  return mmc_regulator_set_voltage_if_supported(mmc->supply.vqmmc,
       vmmc_voltage, vmmc_voltage * 375 / 1000, 300000);

Ulf says he has a board where vmmc is 3.4V and the max vqmmc is 2.9V.
If we think about that board, we'll end up calling
regulator_set_voltage_tol() with a lower/upper tolerance of 1.275V and
.3V.

Extending the current simple implementation of
regulator_set_voltage_tol() to take an upper and lower, that will
translate to first trying to set the voltage to 3.4V - 3.7V, which
will fail.  We'll then try to set the voltage to 2.125V - 3.7V.
Presumably that will end up picking 2.125V, which is really non-ideal
compared to 2.9V and it seems likely to cause some cards to start
failing.

Mark: I know you said you were considering writing a better
regulator_set_voltage_tol() yourself, but I don't know if you've
already started work on it.

I'm expecting to maybe have time to take a crack at it in a few weeks
if you haven't already done it by then.

Thanks!

-Doug
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/