On 21/08/2024 09:06, sunyeal.hong wrote: >>>>>>> + .clk_name = "dout_clkcmu_peric0_noc", >>>>>> >>>>>> same question. >>>>>> Isn't it "noc"? >>>>>> https://lore.kernel.org/linux-samsung- >>>>>> soc/58dfae564a4a624e464c7803a309f1f07b5ae83d.camel@xxxxxxxxxxx/ >>>>>> >>>>>> In my case(autov9), if put wrong clk_name dmesg will show that, >>>>>> exynos_arm64_register_cmu: could not enable bus clock ...; err = -2 >>>>>> >>>>>> Kwang. >>>>>> >>>>>> >>>>> >>>>> clk_name follows the guide document provided by hw. v9 is bus, but >>>>> v920 >>>> uses noc. >>>> >>>> What I mean, >>>> >>>> .clk_name = "dout_clkcmu_peric0_noc", // wrong >>>> .clk_name = "noc", // correct >>>> >>>> Because there is no clock-names "dout_clkcmu_peric0_noc" in >>>> exynos/exynosautov920.dtsi. >>>> >>> >>> The clk_name written here has nothing to do with the device tree. Please >> look at the code carefully. >> >> Hm? I see in the code clearly: >> >> clk_get(dev, cmu->clk_name); >> >> Where cmu is the discussed struct. >> >> If you claim it does not have anything to do with DT, then what is it for? >> >> Best regards, >> Krzysztof > > In general, clk_get is used via the clk_name declared in the DT. > > However, the question asked here is the parent clock name of peric0_noc, so it is unrelated to the device tree. No. The question was about clk_name entry in cmu info used directly for clk_get. Best regards, Krzysztof
