05:36 <hanetzer> hrm.

10:27 <hanetzer> an interesting catch-22 situation. I need to use the platform_device form of registration to register certain kinds of clocks in the crg, but I need to use the CLK_OF_DECLARE form of registration to get the clocks in the sysctrl running early; problem is, those clocks depend on the crg clocks lol

11:12 <apritzel> broonie: is there some appetite (or even precedence?) for a generic clock gate via regmap?

11:13 <apritzel> I have that simple "a bit in an I2C register toggles the clock" case

11:13 <apritzel> and there is some wrapper for the same thing via MMIO: clk_register_gate(), which nicely hides all this CCF boilerplate

11:14 <broonie> apritzel: Seems like a sensible enough thing for the clock API to implement, we've got the equivalents for other APIs like regulator - I'd not do it in regmap.

11:14 <broonie> I'd put it in the clock code.

11:14 <apritzel> yeah, I was just wondering if that should be generic, or live in my driver

11:14 <broonie> I'd say generic.

11:15 <apritzel> broonie: cool, thanks, will give it a try!

11:32 <hanetzer> apritzel: tbh, the clock setup in drivers/clk/hisilicon is pretty robust for that kind of thing.

11:35 <broonie> seems like a clear indication it should be in the framework/library code somewhere then!

11:39 <hanetzer> maybe. a bit over my pay grade to be messing around too much with kernel internals. Already felt a bit eh to carve some soc specific drivers in that dir into something I can use :)

11:44 <hanetzer> but yeah. for mux or gate clocks, you define a few struct arrays that give you a id number you can use in dts, the register block that has the bits you need, the bits themselves, flags, and such.

12:53 <apritzel> hanetzer: ah, thanks, I started to look around for prior art already

12:56 <hanetzer> blerg. apparently the clk-hi3559a.c pll driver never got updated to the new api :P

13:31 <hanetzer> I'll prolly have to skip the pll clocks for now, then :/

13:50 <hanetzer> is there a tool for debugging/decoding kasan dumps?

13:58 <ukleinek> hanetzer: brain?

14:04 <hanetzer> ukleinek: I haz teh dum

14:04 <hanetzer> but yeah. I can't see why its borking where it is tbh

14:31 <hanetzer> ARRAY_SIZE is compile-time only yeh?

14:39 <hanetzer> ah. allocated not enough space. now to figure out why this is -EBUSY

14:45 <hanetzer> basically, 0x12040000 thru 0x120400bc is crg regs. 0x1204008c is a gmac interface control register. I pass 0x12040000 to my crg driver, and try to pass 0x1204008c into another node (see arch/arm/boot/dts/hisi-x5hd2.dtsi:439 for what I mean) and it EBUSY's on me.