01:10 <sboyd> bamse, lumag: Are you happy with the rcg parking patch? https://lore.kernel.org/r/20240502224703.103150-1-swboyd@chromium.org I can pick it up into clk-fixes to fast track it for v6.10

01:41 <bamse> sboyd: i'm not confident that #3 will not cause problems...but i've not analysed it sufficiently

01:42 <bamse> sboyd: i'll dig up some time to think it through...

08:03 <mgonzalez> Yes, thanks!

08:10 <lumag> sboyd, bamse probably we should test it on several platforms.

08:10 <lumag> Give me the time until EoW

08:35 <strongtz[m]> narmstrong: Is primary i2s working for you now? https://git.codelinaro.org/neil.armstrong/linux/-/commit/3ac448a95755e558c9331eb20de76dc7e6c2d872

08:40 <narmstrong> strongtz[m]: no, no sound so far on hdmi

08:41 <strongtz[m]> hmm okay

11:30 <marc|gonzalez> lumag, bamse, sboyd: something I've never understood in driver probe: why are clk_names hard coded in the driver, instead of being read at run-time from the DT node?

11:30 <lumag> marc|gonzalez, that's what devm_clk_bulk_get_all is for.

11:31 <lumag> In some cases you know that different clocks have different meaning, so you want to handle them separately.

11:31 <lumag> And in such case you clk_get htem one by one.

14:04 <marc|gonzalez> lumag: thanks for the quick review!

17:00 <marc|gonzalez> lumag: there is no interconnect driver for 8998 in mainline at the moment

17:00 <lumag> marc|gonzalez, oh, my

17:01 <lumag> jhugo, do you know anything about MNOC_AHB_CLK / MISC_AHB_CLK? Should they be enabled for HDMI to work?

17:06 <abhinav__> lumag I do see them listed https://github.com/android-linux-stable/msm-4.4/blob/kernel.lnx.4.4.r40-rel/arch/arm/boot/dts/qcom/msm8998-sde.dtsi#L226

17:07 <jhugo> lumag: MNOC AHB, yes

17:07 <jhugo> lumag: MISC, probably

17:07 <lumag> abhinav__, jhugo thanks!

17:07 <lumag> marc|gonzalez, so they probably are to be enabled in DT

17:07 <abhinav__> for all msm8998 hdmi related queries, I think msm-4.4 is the working reference

17:08 <abhinav__> as i recall working on msm-4.4 for a hdmi based product

17:09 <lumag> abhinav__, yeah. The questions is whether they should really be enabled by the HDMI or whether it should be an ICC who enables those clocks

17:10 <jhugo> Hmm, usually the ABH clocks just need to be on

17:10 <jhugo> although I think we moved those to icc for sensors....

17:12 <abhinav__> lumag I think the clocks listed with "hpd" prefix are needed for hpd block handling. i dont think these go through ICC as we usually just toggle the enable/disable of those based on whether we need hpd enable/disable

18:48 <lumag> abhinav__, hmm, interesting point

18:48 <lumag> marc|gonzalez, ^^

22:19 <bamse> marc|gonzalez: "clock-names" and other such resources denotes the name of the input... devicetree is used to describe how the component is integrated in the system, i.e. what provider signal feeds the input of the given node/block/device

22:22 <bamse> marc|gonzalez: so clock-names doesn't specify "which clock", but rather just to define the ordering of the entries in the clocks = <> property

23:01 <lumag> arnd, as you complained regarding firmware-name, can I attract your attention to https://lore.kernel.org/linux-arm-msm/20240521-qcom-firmware-name-v1-0-99a6d32b1e5e@linaro.org/ ?

23:03 <lumag> It's that I hate both approaches, but my preference is biased towards existing firmware-name scheme.

23:04 <lumag> Another approach might be to construct firmware paths from the root node compats by parsing the string list and using SoC, vendor and platform and then using fallback if the file doesn't exist. bamse wdyt?