06:46 <brgl> steev: I'm on it but always seem to get preempted with something else

06:47 <brgl> steev: it's coming up, maybe not for 6.11 but surely for 6.12

14:27 <steev> brgl: oh i'm not in a rush for it, was just wondering. the pwrseq looks interesting and i look forward to testing it on x13s :)

14:44 <arnd> bamse: in e2ebc65958ef ("Merge branch '20240602114439.1611-1-quic_jkona@quicinc.com' into arm64-for-6.11"), did you intentionally pull in both the binding and the driver, or was that a mistake?

14:45 <arnd> I've pulled it in already, but last time this happened, torvalds noticed it and asked me about it, so I want to be sure I explain it correctly in my own pull request

14:46 <arnd> Usually we only pull in the binding if that needs to be shared with another branch, and this one looks like it would have been trivial to do that way, but I may be missing something

15:39 <marc|gonzalez> bamse: jhugo: Errrrr, about include/dt-bindings/clock/qcom,gcc-msm8998.h ... WTF is this file? There are a bunch of identifiers in no particular order associated with a sequentially increasing value. Then a second short "block" of identifiers (GDSCs apparently), then a 3rd block (BCRs only?) ... First block is only clocks?

16:17 <aka_[m]> marc|gonzalez: indexes for gcc-msm8998.c clock list

16:18 <marc|gonzalez> what is an index? why do we need to manage a list manually?

16:21 <marc|gonzalez> Right, I had found that code. But why are we managing the indices manually? Because they must be shared by different code? (driver & DTS)

16:33 <aka_[m]> I guess so

16:33 <aka_[m]> It's easier to reference

16:34 <aka_[m]> Not developer so cannot tell

16:34 <marc|gonzalez> Then why not an enum? Why do we have to manually select a number???

16:34 <marc|gonzalez> A sequentially increasing number, which is exactly what an enum is for

16:34 <marc|gonzalez> ... Because an enum can't be used in DTS, only a CPP definition

16:35 <marc|gonzalez> ugh

16:38 <aka_[m]> Icc drivers do enum for qnodes but still we have #defines in bindings

16:46 <bamse> arnd: that was a mistake :(

16:48 <bamse> marc|gonzalez: that header file defines the ABI between the dtb and the OS implementation...in some such cases constants are representing something in the underlying hardware, in this case it's just there to give human readable names to the indexes in a table that needs to match on both sides

16:53 <bamse> marc|gonzalez: it's not a enum, because it's not an enumeration of values where order might matter, but the values must stay constant - to keep compatibility between the different entities (dtb, different OS, bootloaders...)

17:07 <aka_[m]> 😑

17:10 <bamse> z3ntu: typically someone like Andrew picks those, but as it hasn't been picked up i did so now

17:59 <aka_[m]> krzk: can you check these when you get some time?

17:59 <aka_[m]> I don't want to send to once again be told that I'm not describing stuff as I should

17:59 <aka_[m]> https://github.com/Kiciuk/lkml-patches/tree/icc/qcom-icc

18:00 <aka_[m]> Any idea if there is any other irc where I can catch @srinik

18:01 <konradybcio> the lkml is where things get checked out.. don't be afraid to send things out

18:02 <aka_[m]> konradybcio: i don't want to waste time

18:02 <aka_[m]> Either his or mine

18:03 <aka_[m]> I guess you are maintainer of code part of icc

18:03 <aka_[m]> So

18:06 <konradybcio> it's really not a shame to make mistakes on the list

18:06 <konradybcio> i just sent a v15 of cpr, and the past couple revisions were due to me not keeping track of previous issues..

18:10 <aka_[m]> konradybcio: oow can i add second driver into other driver?

18:11 <aka_[m]> Iike put clock controller into machine driver of asoc

18:11 <aka_[m]> Then register clock provider early

18:11 <aka_[m]> Like early_initcall(clock_provider)

18:35 <konradybcio> Technically yes but that sounds sketchy

20:15 <aka_[m]> konradybcio: would setting fake fixed clock be better?

20:16 <aka_[m]> Unless we bake clock setting into wcd codec

20:16 <aka_[m]> For now it takes pointer to mclk but if mclk is empty we can detect AFE api version and set it up

20:19 <z3ntu> bamse Thanks!

20:19 <aka_[m]> srinik: do you remember discussion about q6afe clocks on legacy firmware and wcd mclk issue?

20:34 <konradybcio> aka_: i would need the bigger picture - what is that clock for, what provides it, why does it need setting

20:40 <aka_[m]> konradybcio: brave of you to assume I know much lol

20:42 <aka_[m]> I guess I have to go and look into pmi schematics

20:43 <aka_[m]> To see how wcd is built

21:22 <aka_[m]> Ngl I'm kinda lost on what wcd digital represent

21:23 <aka_[m]> Analog seems to represent device on pm89xx module

21:43 <aka_[m]> konradybcio: do you think there is anyone having access to ds lpass firmware being able to check what specific abstraction do to hardware?

21:44 <aka_[m]> I mean how lpass_set_clock is being resolved

23:36 <konradybcio> someone probably does