#asahi-re · 2021-02-18 17:13
modwizcodeI'm very very interested in GL1/2#asahi-re · 2021-02-18 17:13
never_releasedin which case here be dragons#asahi-re · 2021-02-18 17:13
never_releasedyou might get yourself into GL1 or GL2#asahi-re · 2021-02-18 17:13
never_releasedand if you do the probing from EL2 or EL1#asahi-re · 2021-02-18 17:13
never_releasedmarcan: yeah, tons#asahi-re · 2021-02-18 16:24
marcanyeah, AMX is off as one would expect (AFAICT), but... there are other things in here#asahi-re · 2021-02-18 16:23
j`eymarcan: heh, mark got you wondering? :P#asahi-re · 2021-02-18 16:23
marcanfun things to do: exhaustive probe of instruction set#asahi-re · 2021-02-18 15:07
modwizcodethe names come from xnu headers iirc plus some I think were named after discovery of their uses but mostly from xnu I think#asahi-re · 2021-02-18 10:39
Shizsad i didn't notice it#asahi-re · 2021-02-18 10:39
Shizsven: damn it, someone wrote a usable devtree parser around the same time i did then :p#asahi-re · 2021-02-18 07:29
bgb_are the register bit definitions of https://github.com/AsahiLinux/docs/wiki/HW:ARM-System-Registers extracted from xnu kernel ?#asahi-re · 2021-02-14 17:30
jannaudocumentaton updated#asahi-re · 2021-02-14 17:17
marcanI've noticed several things that need an isb to work#asahi-re · 2021-02-14 17:16
jannau"RCC Disable Power Save Prf (performance?) Clock Off" is unrelated it was just a missing isb#asahi-re · 2021-02-14 15:50
modwizcode /mark#asahi-re · 2021-02-14 14:58
jannauargh, I've missed osfmk/arm64/monotonic_arm64.c. I only looked at osfmk/arm64/kpc.c#asahi-re · 2021-02-14 14:51
marcanjannau: XNU says PMC0 is cycles and PMC1 is instructions#asahi-re · 2021-02-14 14:37
jannauit's working now after a reboot. PMC0 seems to be CPU cycles, at least it reports roughly the same value as the configurable CPU_CYCLES counter#asahi-re · 2021-02-14 14:18
jannauthe configureable performance counters still don't count#asahi-re · 2021-02-14 14:16
jannaunot sure what they count though, PMC0 updates at 2.11 GHz and PMC1 updates at 4.158MHz#asahi-re · 2021-02-14 14:14
jannau"RCC Disable Power Save Prf (performance?) Clock Off" needs to be set in EHID for PMC 0/1 to count#asahi-re · 2021-02-14 12:03
svenyeah, the 0x300 bits are some kind of "something changed" flag related to the 0b1100 bits#asahi-re · 2021-02-14 11:59
sven0x300 seems to behave like a "was disabled" bit#asahi-re · 2021-02-14 11:55
sven0b0010 and 0b0001 can only be enabled when 0b1000 is set I think#asahi-re · 2021-02-14 11:54
svenso the lower 8 bits are [actual][wanted]. 0b1000 and 0b0100 can be enabled independently.#asahi-re · 2021-02-14 11:47
svenhrm, I think I actually need to write to the register once more before that 0x300 disappears#asahi-re · 2021-02-14 11:46
marcanpossibly related to clamping/"ready"#asahi-re · 2021-02-14 11:40
svenbit 0x300 might be "power is stable" or something like that. after reboot the value might e.g. be 2300. i then set 0xf and the register then reads back as 23ff and eventually 20ff after more time passes#asahi-re · 2021-02-14 11:37
svenbut I'm curious about e.g. "Higher bits exist, but for most their meaning is unknown and might change between chip generations."#asahi-re · 2021-02-14 11:36
svenso.. has anyone looked into this power management controller yet? I found https://github.com/Siguza/dt which is already helpful#asahi-re · 2021-02-10 18:39
davidrysk[m]marcan: yeah, I believe it all traces back to the same code#asahi-re · 2021-02-10 18:39
marcangoing to sleep now, I'll get the firestorm stuff tomorrow#asahi-re · 2021-02-10 18:38
marcanmodwizcode: feel free to write the docs!#asahi-re · 2021-02-10 18:38
marcandavidrysk[m]: IIRC there was something like that elsewhere too#asahi-re · 2021-02-10 18:35
davidrysk[m]directhw is probably a good place to start, just rip out all the x86-specific code and use your own asm :)#asahi-re · 2021-02-10 18:15
modwizcodemarcan: can I get the same dump on a firestorm core?#asahi-re · 2021-02-10 18:13
davidrysk[m]I think flashrom's version is the most current, but the closest thing to upstream is coreboot's repo#asahi-re · 2021-02-10 18:11
davidrysk[m]lemme find upstream, it's kinda buried#asahi-re · 2021-02-10 18:10
davidrysk[m]marcan: hm do you think DirectHW.kext would be of any use? :P#asahi-re · 2021-02-10 16:42
modwizcodeheh the best part of proxyclient is it's lack of documentation#asahi-re · 2021-02-10 15:57
marcanI need to clean up proxyclient/ some day...#asahi-re · 2021-02-10 15:56
marcanpushed the tool to m1n1.git, and also a register database of all actually documented ARM regs#asahi-re · 2021-02-10 15:40
jn__kextec :P#asahi-re · 2021-02-10 15:38
sventime for m1n1.kext? :-P#asahi-re · 2021-02-10 15:36
marcanideally you want it to be interactive :)#asahi-re · 2021-02-10 15:36
marcangood point, though the problem is if you just log you need to reboot to load a new kext#asahi-re · 2021-02-10 15:16
davidrysk[m]if you're writing drivers that have to interact with userspace it's more work :p#asahi-re · 2021-02-10 15:16
davidrysk[m]if all you're doing is logging crap to the kernel log, you don't need to do that much#asahi-re · 2021-02-10 15:15
davidrysk[m]if you want to avoid xcode projects you can use https://github.com/droe/example.kext