On Tue, Nov 27, 2018 at 10:36:19AM +0100, Maxime Ripard wrote:
> On Mon, Nov 26, 2018 at 05:27:47PM +0200, Priit Laes wrote:
> > From: Olliver Schinagl <oliver@xxxxxxxxxxx>
> >
> > In the past, there have been words on various lists that if LDO3 is
> > disabled in u-boot, but enabled in the DTS, the axp209 driver would
> > fail to continue/hang. Several enable/disable patches have been
> > issues to devicetree's in both the kernel and u-boot to address
> > this issue.
> >
> > What really happened however, was that the AXP209 shuts down without
> > a notice and without setting an interrupt. This is caused when LDO3
> > gets overloaded, for example with large capacitors on the LDO3 output.
> >
> > Normally, we would expect that AXP209 would source 200 mA as per
> > datasheet and set and trigger an interrupt when being overloaded.
> > For some reason however, this does not happen.
> >
> > As a work-around, we use the soft-start constraint of the regulator
> > node to first bring up the LDO3 to the lowest possible voltage and
> > then enable the LDO. After that, we can set the requested voltage
> > as usual.
> >
> > Combining this setting with the regulator-ramp-delay allows LDO3 to
> > enable voltage slowly and staggered, potentially reducing overall
> > inrush current.
> >
> > Signed-off-by: Olliver Schinagl <oliver@xxxxxxxxxxx>
> > Signed-off-by: Priit Laes <plaes@xxxxxxxxx>
> > ---
> > drivers/regulator/axp20x-regulator.c | 57 ++++++++++++++++++++++++++++-
> > 1 file changed, 56 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/regulator/axp20x-regulator.c b/drivers/regulator/axp20x-regulator.c
> > index 1d9fa62..e8a895b 100644
> > --- a/drivers/regulator/axp20x-regulator.c
> > +++ b/drivers/regulator/axp20x-regulator.c
> > @@ -14,6 +14,7 @@
> > */
> >
> > #include <linux/bitops.h>
> > +#include <linux/delay.h>
> > #include <linux/err.h>
> > #include <linux/init.h>
> > #include <linux/mfd/axp20x.h>
> > @@ -23,6 +24,7 @@
> > #include <linux/platform_device.h>
> > #include <linux/regmap.h>
> > #include <linux/regulator/driver.h>
> > +#include <linux/regulator/machine.h>
> > #include <linux/regulator/of_regulator.h>
> >
> > #define AXP20X_GPIO0_FUNC_MASK GENMASK(3, 0)
> > @@ -430,6 +432,59 @@ static int axp20x_set_ramp_delay(struct regulator_dev *rdev, int ramp)
> > return regmap_update_bits(axp20x->regmap, reg, mask, cfg);
> > }
> >
> > +static int axp20x_regulator_enable_regmap(struct regulator_dev *rdev)
> > +{
> > + struct axp20x_dev *axp20x = rdev_get_drvdata(rdev);
> > + const struct regulator_desc *desc = rdev->desc;
> > +
> > + if (!rdev)
> > + return -EINVAL;
> > +
> > + switch (axp20x->variant) {
> > + case AXP209_ID:
> > + if ((desc->id == AXP20X_LDO3) &&
> > + rdev->constraints && rdev->constraints->soft_start) {
> > + int v_out;
> > + int ret;
> > +
> > + /*
> > + * On some boards, the LDO3 can be overloaded when
> > + * turning on, causing the entire PMIC to shutdown
> > + * without warning. Turning it on at the minimal voltage
> > + * and then setting the voltage to the requested value
> > + * works reliably.
> > + */
> > + if (regulator_is_enabled_regmap(rdev))
> > + break;
> > +
> > + v_out = regulator_get_voltage_sel_regmap(rdev);
> > + if (v_out < 0)
> > + return v_out;
> > +
> > + if (v_out == 0)
> > + break;
> > +
> > + ret = regulator_set_voltage_sel_regmap(rdev, 0x00);
> > + /*
> > + * A small pause is needed between
> > + * setting the voltage and enabling the LDO to give the
> > + * internal state machine time to process the request.
> > + */
> > + usleep_range(1000, 5000);
> > + ret |= regulator_enable_regmap(rdev);
> > + ret |= regulator_set_voltage_sel_regmap(rdev, v_out);
> > +
> > + return ret;
> > + }
> > + break;
> > + default:
> > + /* No quirks */
> > + break;
> > + }
> > +
> > + return regulator_enable_regmap(rdev);
> > +};
> > +
>
> This is some pretty generic code, and could be useful to some other
> users. I guess a generic function would be better for this.
Yes, makes sense. Although, should we then also distinguish between
regulators which support soft-start in hardware and devices which emulate
it by delay, like in this case?
Päikest,
Priit
