On Fri, Dec 25, 2020 at 11:33:21AM +0000, Timon Baetz wrote: > On Thu, 24 Dec 2020 15:00:38 +0100, Krzysztof Kozlowski wrote: > > On Thu, Dec 24, 2020 at 02:37:06PM +0100, Krzysztof Kozlowski wrote: > > > On Thu, Dec 24, 2020 at 01:13:02PM +0000, Timon Baetz wrote: > > > > On Thu, 24 Dec 2020 10:55:59 +0100, Krzysztof Kozlowski wrote: > > > > > > @@ -170,6 +237,28 @@ static int max8997_battery_probe(struct platform_device *pdev) > > > > > > return PTR_ERR(charger->battery); > > > > > > } > > > > > > > > > > > > + charger->reg = devm_regulator_get(&pdev->dev, "charger"); > > > > > > > > > > Since you do not use get_optional, you will always get a dummy > > > > > regulator. In case of error, you should either print it or entirely fail > > > > > the probe. Silently continuing makes it difficult to spot errors. > > > > > > > > > > Since the driver could operate in case of extcon/regulator error, just > > > > > dev_err() so failure will be spotted with dmesg. > > > > > > > > I will switch to devm_regulator_get_optional() and print an error on > > > > failure, thanks. > > > > > > > > > It will complain on older DTBs because you are introducing incompatible > > > > > change, but that's expected. Just correct all other in-tree DTS. > > > > > > > > The other 2 in-tree DTS don't have CHARGER regulators. Not sure > > > > how to correct those. Should I add muic and charger nodes without a > > > > charger-supply? It will still complain in that case. > > > > > > +Cc Marek, > > > > > > This is why leaving the code as is - devm_regulator_get(), not optional > > > - makes sense. Core would provide dummy regulator, so you only have to > > > provide MUIC node. > > > > > > If you change the code to devm_regulator_get_optional(), you need to add > > > everything: the charger regulator, the charger node and MUIC node. > > > > > > For Trats, the configuration should be similar as i9100, although I > > > don't know the exact values of chargign voltage. > > > > > > For Origen, there is no battery, so the power supply should not bind. > > > Maybe this could be achieved with "status disabled" for charger node? It > > > depends whether MFD will respect such field... If it disables the > > > charger, you're done. > > > > I just looked at the MFD code and tested it - it nicely skips disabled > > devices. Therefore, for Origen I propose to add disabled nodes for > > charger and MUIC because these pins are not connected. No need to add > > regulators in such case. > > With a dummy regulator regulator_set_current_limit() fails with -EINVAL. Good point. > Isn't it better to just skip charging control (and dev_info()) when there > is no extcon or regulator? The charger driver would still probe > without those 2 properties and work as before. Yes, makes sense. > > Adding disabled nodes for Origen would probably still makes sense. > > I also noticed that adding nodes for those MFD cells prints "DMA mask > not set" which seems to be related to https://lkml.org/lkml/2020/4/23/873. > Any suggestions on how to handle that? I don't think it is your problem to solve. It affects other MFD devices as well. Best regards, Krzysztof
