#asahi-dev · 2025-12-07 12:23
jannauit should be the same on plain M3#asahi-dev · 2025-12-07 12:22
integralpiloti can try and do some work on that if you'd like, or i can switch my host & target so that my target is M3 plain (which doesn't have these I believe) and i can get the hypervisor working on that. which would be most useful?#asahi-dev · 2025-12-07 12:21
jannausigh: "Linux 6.19 Delivers Working USB3 Support For Apple Silicon Devices" https://www.phoronix.com/news/Linux-6.19-USB3-Apple-M1#asahi-dev · 2025-12-07 12:18
jannauI think we need to handle those CNT registers and possibly update m1n1 idea of what the counter frequency is#asahi-dev · 2025-12-07 12:18
integralpilotoh didn't notice those numbers matched with the ones nickchan told me before, thanks for pointing it out sorry!#asahi-dev · 2025-12-07 12:15
jannauintegralpilot: https://oftc.catirclogs.org/asahi-dev/search?q=s3_1_c15_c1_5##asahi-dev · 2025-12-07 12:09
integralpilotand both of these are the only 2 custom registers in the M3 pro only chickens#asahi-dev · 2025-12-07 12:06
integralpilotfor both of these i added a printf to the start of hv_handle_msr_unlocked and it's never even being called, just straight away causing an exception that it's too early for XNU to handle without ever going through the hypervisor#asahi-dev · 2025-12-07 12:06
integralpilotS3_1_C15_C1_5 is also caushing a bit of trouble#asahi-dev · 2025-12-07 12:02
integralpilothmmm adding it as SYSREG_PASS in hv_handle_msr_unlocked isn't helping#asahi-dev · 2025-12-07 12:02
integralpilotoh interesting, thank you#asahi-dev · 2025-12-07 11:53
jannauif I'm not up on the numbers S3_4_C15_C14_6 is related to apple's implementation specific counter scaling. the physical counter is still 24MHZ but Arm mandates a 1GHz counter for ARMv8.6-a#asahi-dev · 2025-12-07 11:47
integralpilotis there a way to find out what these mystery registers do?#asahi-dev · 2025-12-07 11:47
integralpilotcommit message that added it says "Everest settings introduce some changes to unknown registers" so ig we still don't know what it does#asahi-dev · 2025-12-07 11:46
integralpilotit's mentioned ONCE in the m3 pro only section of the everest chickens: https://github.com/AsahiLinux/m1n1/blob/64762a702b66062e545a92bf8aca04372ab2dc89/src/chickens_everest.c#L85#asahi-dev · 2025-12-07 11:45
integralpilotwait no actually it doesn't exist#asahi-dev · 2025-12-07 11:45
integralpilotwait sorry i missed it, it's late#asahi-dev · 2025-12-07 11:44
integralpilot*has#asahi-dev · 2025-12-07 11:44
integralpilotwhat is this register, i guess i'm the first person who as encountered it so i'll have to find out#asahi-dev · 2025-12-07 11:43
integralpilothuh https://github.com/AsahiLinux/m1n1/blob/main/tools/apple_regs.json has no [3, 4, 15, 14, 6]?#asahi-dev · 2025-12-07 11:34
integralpilottrying to fix it up now#asahi-dev · 2025-12-07 11:34
integralpilotjust did some more debugging and now turns out S3_4_C15_C14_6 is also a problem for m1n1 hypervising XNU on M3 #asahi-dev · 2025-12-07 10:46
sven*one other#asahi-dev · 2025-12-07 10:46
sventhere's at least one person working in the infrastructure team#asahi-dev · 2025-12-07 10:46
sveni think the preferred way is to write to helpdesk@kernel.org#asahi-dev · 2025-12-07 10:45
jannaudoesn't kernel.org use geoloacated mirrors? maybe ask https://social.kernel.org/monsieuricon about mirror status for australia/asia#asahi-dev · 2025-12-07 10:36
integralpilotoooh that would actually be quite handy, thanks for the idea! i'll try and see if i can get it to do that#asahi-dev · 2025-12-07 10:35
chaos_princessbut a simple check should also work#asahi-dev · 2025-12-07 10:35
chaos_princessif you want to be extra fancy, you can match to real addresses via adt#asahi-dev · 2025-12-07 10:34
svensure!#asahi-dev · 2025-12-07 10:34
integralpilotdisassemble_at in the M1N1 hypervisor python REPL has been annoying me because sometimes i accidentally use a virtual address and then the entire thing crashes and i lose m debugging state. locally i've made a change that if you use a higher half address it warns "are you sure this might be a virtual address by mistake?". its quite helpful for me but would others like this / should i submit it as a PR?#asahi-dev · 2025-12-07 10:34
sventhat’s annoying :(#asahi-dev · 2025-12-07 10:33
chadmedbut the web interface is basically a crapshoot on whether i can actually reach it or not#asahi-dev · 2025-12-07 10:33
svenhuh#asahi-dev · 2025-12-07 10:33
chadmedwhen it does work pulling from it is incredibly slow #asahi-dev · 2025-12-07 10:33
chadmedit just doesnt work from australia at all#asahi-dev · 2025-12-07 10:33
svenbut at least it doesn’t try to shove copilot down my throat every chance it gets ;)#asahi-dev · 2025-12-07 10:32
svenyeah, I wish kernel.org had a better web interface #asahi-dev · 2025-12-07 10:32
chaos_princessclassic#asahi-dev · 2025-12-07 10:31
chadmedasoc series have to be based on a specific for-next or i get yelled at :p #asahi-dev · 2025-12-07 10:31
chaos_princess(yes, msft, but the kernel.org git is worse)#asahi-dev · 2025-12-07 10:31
chaos_princesshttps://github.com/torvalds/linux/#asahi-dev · 2025-12-07 10:31
chadmedguess im not submitting this tonight then #asahi-dev · 2025-12-07 10:31
chadmedwhy is the kernel.org cgit so bad #asahi-dev · 2025-12-07 09:55
chadmedit was mesa for fex#asahi-dev · 2025-12-07 09:55
chadmedsee 4c86ab250a5#asahi-dev · 2025-12-07 09:54
chadmedit was/is a bdep of something#asahi-dev · 2025-12-07 09:53
chadmediirc it was a bdep of mesa for the fex rootfs at one point #asahi-dev · 2025-12-07 09:51
jannauwhy does asahi-sources in the overlay have "amd64" in keywords?#asahi-dev · 2025-12-07 09:40
chadmedokie doke