#asahi · 2022-08-20 09:50
marcanI guess I should make a new branch for this#asahi · 2022-08-20 09:39
marcanthat was literally just a test merge waiting for povik to rebase the audio stuff :)#asahi · 2022-08-20 09:38
marcanand yeah, I make no promises about 6.0#asahi · 2022-08-20 09:38
marcanjannau: pushed to -dev#asahi · 2022-08-20 09:23
marcanthat only has the dockchannel fix so it should be safe#asahi · 2022-08-20 09:23
marcan(forgot I forgot :))#asahi · 2022-08-20 09:22
marcanI am doing that now#asahi · 2022-08-20 09:19
marcanspeaking of, I think it's time to push dev to prod, nobody has complained I think#asahi · 2022-08-20 09:18
marcanthese days the rules are, whatever we push to the asahi repo is OK for end users, whatever the latest asahi tag is is what ends up in asahi-dev (and *usually* works but can also break horribly; if it does I usually find out and fix it quickly though), and the main branch *usually* tracks that but not always (as you found out)#asahi · 2022-08-20 09:16
marcan(although apparently Linus rolled with it anyway and it works; I hadn't even tested it myself)#asahi · 2022-08-20 09:16
marcanfwiw the 6.0-rc1 merge on github is *not* ready for consumption#asahi · 2022-08-18 05:53
marcanM1 Pro/Max have more#asahi · 2022-08-18 05:53
marcanM2 has the same AS as M1#asahi · 2022-08-18 05:53
marcankeegans: 0x200000000 definitely works, that's where MMIO is on bare metal here. you can use all 36 bits of address space for sure, at least as far as hardware is concerned.#asahi · 2022-08-17 03:38
marcanif you get an M1 Pro/Max you can have more bits :)#asahi · 2022-08-17 03:38
marcan(this broke a bunch of stuff that assumed larger address spaces, but is perfectly spec-compliant)#asahi · 2022-08-17 03:38
marcanthat's how that CPU is configured#asahi · 2022-08-17 03:37
marcankeegans: no, the hardware has a 36-bit address space#asahi · 2022-08-16 13:47
marcan(this is on '87, it might work on older chips for all I know)#asahi · 2022-08-16 13:45
marcan(since obviously this does work on macOS)#asahi · 2022-08-16 13:45
marcanThere is at least one command that does not work (firmware returns an error) involved in AP setup, that's also something I want to fix with the WiFi tracer#asahi · 2022-08-16 13:40
marcanthat's on my list to fix this month#asahi · 2022-08-16 13:40
marcanMuonBluon: hostapd does not work#asahi · 2022-08-16 02:09
marcanM1 does not#asahi · 2022-08-16 02:09
marcanM2 does#asahi · 2022-08-16 02:08
marcanmilek7: no, that means you're missing a driver#asahi · 2022-08-15 19:50
marcananyway, I should really get some sleep :)#asahi · 2022-08-15 19:50
marcan*convince#asahi · 2022-08-15 19:49
marcanby the time I got to tty I knew this was going to be "fun"#asahi · 2022-08-15 19:49
marcanwhat you didn't see before that (because it happened while I was on that trip) was another 4h or so of me starting out debugging the m1n1 hypervisor, then the m1n1 USB CDC code, then the cdc-acm driver, until I got to tty, then taking a detour to conving myself it wasn't aic's fault#asahi · 2022-08-15 19:48
marcanbeing in the TTY layer already was deep in the woods :p#asahi · 2022-08-15 16:56
marcanArch Linux does not support installing packages without updating your system. You are required to keep your system up to date, otherwise you will get version conflicts.#asahi · 2022-08-15 16:55
marcandid you actually reboot with the updated kernel?#asahi · 2022-08-15 15:46
marcanscary#asahi · 2022-08-15 15:46
marcanI mean you'd *think* nobody would implement it like that, but...#asahi · 2022-08-15 15:46
marcanif the insane "atomic writes aren't atomic with atomics" thing I found were real, we'd be in serious trouble#asahi · 2022-08-15 15:45
marcanthankfully :-)#asahi · 2022-08-15 15:45
marcanI was still using the spec from back when the M1 came out, hence why I was finding those corner cases still#asahi · 2022-08-15 15:44
marcanatomic_long_set was also broken, that one got fixed in the spec too#asahi · 2022-08-15 15:43
marcanit's fixed now, not because it changed, but because ARM re-defined the spec to make it not broken#asahi · 2022-08-15 15:43
marcantest_and_set_bit() was broken when the M1 came out#asahi · 2022-08-15 15:43
marcanI'm writing the patch#asahi · 2022-08-15 15:43
marcanno, the workqueue code is broken and needs a barrier#asahi · 2022-08-15 05:06
marcanstarting shortly#asahi · 2022-08-14 17:02
marcanI just tested it with my combo dock, confirmed it fixes it#asahi · 2022-08-14 17:02
marcanit applies to any usb mass storage device that reports not ready#asahi · 2022-08-14 17:02
marcanif nothing regressed I'll probably push to stable soonish#asahi · 2022-08-14 17:02
marcantest please, this all also includes the prior u-boot keyboard support for M2 plus the race condition fix for the kernel#asahi · 2022-08-14 17:01
marcanthat's the usb card reader fix, a very minor thing for m1n1, and a minor .config change for the kernel#asahi · 2022-08-14 17:00
marcanm1n1/u-boot/linux updated in asahi-dev (cc jannau)