Hi,

On Fri, Aug 16, 2024 at 05:40:37PM GMT, Krzysztof Kozlowski wrote:

On 16/08/2024 16:25, Artur Weber wrote:

quoted

On 16.08.2024 11:54, Krzysztof Kozlowski wrote:

quoted

On 16/08/2024 10:19, Artur Weber wrote:

quoted

@@ -732,6 +794,15 @@ static int max77693_charger_probe(struct platform_device *pdev)
  	chg->dev = &pdev->dev;
  	chg->max77693 = max77693;
  
+	psy_cfg.drv_data = chg;
+
+	chg->charger = devm_power_supply_register(&pdev->dev,
+						  &max77693_charger_desc,
+						  &psy_cfg);
+	if (IS_ERR(chg->charger))
+		return dev_err_probe(&pdev->dev, PTR_ERR(chg->charger),
+				     "failed: power supply register\n");

This code move is not explained in the commit msg. At least I could not
find it. Please explain why you need it in the commit msg.

This is done because the call to power_supply_get_battery_info in
max77693_dt_init requires chg->charger to be set. (I was considering
putting this in the commit message, can't remember why I didn't do it.
I'll add it in the next version.)

I think that's wrong. Power supply is being available to the system
before it is being configured.

It's a known limitation of the power_supply_get_battery_info API.
I think it would be best to add an register_init() hook to struct
power_supply_desc, which would be called from __power_supply_register()
directly before it calls device_add(). At that point the power_supply
struct is initialized far enough for getting the battery info, but not
yet exposed to the remaining system.

As a nice side effect the register writes happen after checking the
supplies, so the registers are not written if the probe errors out
with a probe defer anyways.

Greetings,

-- Sebastian