15:01 <ichernev[m]> sboyd: any comments on the v4 sm6115 gcc patchset: https://lkml.org/lkml/2021/8/5/765? I'm waiting to submit my DT patches ;-)

19:05 <lumag> bamse, minecrell : if I remember correctly what somebody (ndec?) has told me, the 1034.2.1 was released separately to us under redistribution-permitted license, but we never got 1036 under such license. I think we'd better send 1034 ATM and check whether we can get permission for the 1036.

19:32 <minecrell> lumag: Ah, thanks for clarifying! If you have permission for the 1034 it's indeed better to send that first. As far as I noticed FastRPC is the only advantage of 1036 so it's not really worth delaying all other firmware only because of that

22:23 <Marijn[m]> lumag: Bamse and me were discussing the removal of global clock name dependencies (`.parent_names = .. "xo"`) from DSI PHY PLLs, in favour of `.fw_name = "xo"` and specifying this in the DT. It's problematic on many platforms which don't define their xo to be named exactly "xo" (xo_board or bi_tcxo nowadays) and we have a hacky `fixed_factor` clock in gcc just to provide a global clock with the `"xo"` name. We'll switch the driver to use

22:23 <Marijn[m]> `.fw_name` and update DT to provide this clock. Do you by any chance have changes in progress that change clock handling one way or another, that might conflict?

22:25 <bamse> minecrell: afaict there are a couple of platforms that rely on "xo" and don't have a "xo" clock...so it would be good to fix that up

22:25 <bamse> not minecrell, Marijn[m]

22:27 <Marijn[m]> lumag: In addition I can't help but notice that these drivers are repeatedly `snprintf`-ing the names of the clocks for use as parent(s); any thoughts on adding a few more helper functions to `clk-provider.h` and switch them all over to using `hw` pointers too like the `*cc` drivers?

22:28 <Marijn[m]> bamse: Correct, there are quite a few (clock-controller) drivers still relying on a global "xo" clock (grep for `{ "xo"` or `.name = "xo"`), should probably be fixed in a large cleanup-spree but it's a lot of work :(

22:29 <Marijn[m]> bamse: We even came across a driver (`hfpll.c`) that doesn't use parent names at all but `.index = 0` :P

22:31 <Marijn[m]> Works, but we might want to be more explicit since the DT for this driver (qcs404 and the dt-bindings) already explicitly specify the `"xo"` clock name for that...

22:32 <bamse> Marijn[m]: doesn't .index=0 mean "the first clock specified in the dt node"?

22:32 <Marijn[m]> bamse: Pretty sure, yes

22:33 <bamse> Marijn[m]: that's perfectly fine, if there's a single clock for a device clk_get(dev, NULL) is preferable

22:33 <Marijn[m]> Afaik it even makes `clock-names = "whatever"` obsolete

22:33 <bamse> right

22:33 <Marijn[m]> Okay, let's stick with that then :)