regulator: regulator-fixed: device is never ready

[GrygiriiS]

It seems regulator-fixed regulator device is never become ready (device_is_ready(reg_dev)) since one of the commits

• 539d4aa ("drivers: regulator: add common init enable API"
• b3fdb74 ("drivers: regulator: fixed: Removed pin state setting from init")

and subsequent commits in drivers/regulator/regulator_common.c which added regulator_common_init() call to the regulator-fixed itself, and non valid (for regulator-fixed) regulator_set_voltage() calls in regulator_common_init().

Is it know issue?

Hence, regulator_set_voltage() is not valid for regulator-fixed, and, actually, most of steps in regulator_common_init() - it seems reasonable to just call regulator_enable() for regulator-fixed on init to update regulator state and ref_cnt.

Smth. like below:

--- a/drivers/regulator/regulator_fixed.c
+++ b/drivers/regulator/regulator_fixed.c
@@ -84,6 +84,7 @@ static const struct regulator_driver_api regulator_fixed_api = {
 static int regulator_fixed_init(const struct device *dev)
 {
        const struct regulator_fixed_config *cfg = dev->config;
+       int ret;
 
        regulator_common_data_init(dev);
 
@@ -93,14 +94,17 @@ static int regulator_fixed_init(const struct device *dev)
                        return -ENODEV;
                }
 
-               int ret = gpio_pin_configure_dt(&cfg->enable, GPIO_OUTPUT);
+               ret = gpio_pin_configure_dt(&cfg->enable, GPIO_OUTPUT);
 
                if (ret < 0) {
                        return ret;
                }
        }
 
-       return regulator_common_init(dev, false);
+       if (regulator_common_is_init_enabled(dev)) {
+               ret = regulator_enable(dev);
+       }
+       return ret;
 }

[edt-alex]

I'm looking at Zephyr v4.3.0-rc2 (commit e438b57) and I think the problem is that regulator_fixed.c does not provide a .get_voltage function in the DEVICE_API.

When the device tree sets min/max voltage of a fixed regulator, regulator_common_init() tries to read the voltage setting and adjust it to a valid range before possibly enabling the regulator.

It's that regulator_get_voltage() call which is failing for me. I've locally patched regulator_fixed.c to add a get_voltage function (simply returns the min-millivolts value, like voltage list does). This seems to fix it for me.

[edt-alex]

#99339

[nandojve]

fixed by #99339