On 20/12/2021 15:55, Sam Protsenko wrote: > On Mon, 20 Dec 2021 at 11:31, Krzysztof Kozlowski > <krzysztof.kozlowski@xxxxxxxxxxxxx> wrote: >> >> On 19/12/2021 23:29, Sylwester Nawrocki wrote: >>> On 17.12.2021 17:15, Sam Protsenko wrote: >>>> System Register is used to configure system behavior, like USI protocol, >>>> etc. SYSREG clocks should be provided to corresponding syscon nodes, to >>>> make it possible to modify SYSREG registers. >>>> >>>> While at it, add also missing PMU and GPIO clocks, which looks necessary >>>> and might be needed for corresponding Exynos850 features soon. >>>> >>>> Reviewed-by: Krzysztof Kozlowski<krzysztof.kozlowski@xxxxxxxxxxxxx> >>>> Acked-by: Rob Herring<robh@xxxxxxxxxx> >>>> Acked-by: Chanwoo Choi<cw00.choi@xxxxxxxxxxx> >>>> Signed-off-by: Sam Protsenko<semen.protsenko@xxxxxxxxxx> >>> >>> Apologies for late reply, this patch is applied now. >>> >> >> Sam, >> >> The clock is used in the DTSI, so since this was applied, there are only >> two choices now: >> 1. wait for next cycle with DTSI and DTS, >> 2. Resubmit with replacing the newly added clocks in DTSI/DTS with >> numbers and a TODO note. >> > > But why? I thought because Sylwester applied my clock patches, those > will get into v5.17, and so DTSI/DTS might rely on those clocks? If I > get it wrong, please let me know why, and I'll go with item (2) you > suggested. If I apply the DTSI+DTS, all my builds will start failing. The linux-next (since Sylwester's tree is included) should build fine, but my tree won't be buildable anymore. Then arm-soc pulls my tree and gets said because it does not build. Later, Linus will be unhappy if he pulls arm-soc (thus mine) before clock tree. Other solution, instead of using raw numbers, is to copy-paste the clock macros you use directly in DTSI and do not include the clock header. This actually might be cleaner choice - changes will be limited to one place in DTSI. Best regards, Krzysztof
