On 4/24/2024 10:00 PM, Bartosz Golaszewski wrote: > On Wed, 24 Apr 2024 at 15:53, quic_zijuhu <quic_zijuhu@xxxxxxxxxxx> wrote: >> >>>>> >>>>> Please slow down here. Zijun's patch works and Bartosz's patch does not. >>>>> I don't think Zijun means any ill intent. I am replying to Bartosz's >>>>> patch right now. >>>> >>>> Ok, that is great feedback, so I might be picking up the Zijun v7 set >>>> if we don't find any major problems with it. >>>> >>> >>> Luiz, >>> >>> Please consider my alternative[1] also tested by Wren. Zijun's usage >>> of GPIO API is wrong. >>> >> why is it wrong ? >> > > I have already told you that at least three times. But whatever, let > me repeat again: gpiod_get_optional() returns NULL if the given GPIO > is not assigned to the device in question OR a pointer to a valid GPIO > descriptor. Anything else returned by it is an error and the driver > must abort probe(). > notice that i talked many times for you. the only different between my fix and present kernel code is that how to handle NULL case. for QCA6390. the GPIO is not marked as required by DTS binding spec. so, we don't need to take the case the gpio is not configured(return NULL) as error. i don't need to care about how to handle gpiod_get_optional() returning error case since my change don't change current handle logic for it. i currently only care about the issue reported. > Bart
