Hi Krzysztof, On Fri, Jun 23, 2023 at 02:18:32PM +0200, Krzysztof Kozlowski wrote: > On 23/06/2023 13:48, Andi Shyti wrote: > > On Fri, Jun 23, 2023 at 12:03:56PM +0200, Wolfram Sang wrote: > >> > >>> - clk_unprepare(i2c->clk); > >>> - clk_unprepare(i2c->pclk); > >> > >> Are you sure we can use devm_clk_get_enabled() here which calls > >> clk_disable_unprepare() on remove and not clk_unprepare()? > > > > Unless I am missing something, I think so. This is what the > > driver does, gets the clock and enables it. > > > > I don't know why there is just unprepare() and not > > disable_unprepare() in this function. > > Your code is not equivalent and does not explain why. Pure > devm_clk_get_enabled() conversion should be equivalent. > > If original code was correct, your patch will cause double clk disable. > If original code was not correct, your patch silently fixes it without > explaining that there was a bug. > > Did you test the patch, including the unbind path? you are right! My code, indeed, doesn't do a 1to1 conversion. Will prepare it again, maybe in two patches, one for the conversion and one for clearing the removal path. Thanks for dropping in, Andi
