#asahi-dev · 2022-10-18 11:26
marcanmaz: btcoex? no, that's actually already working I think, but I think it needs a change on the *BT* side to mark audio connections are prioritized first#asahi-dev · 2022-10-18 11:25
marcanthe OTP and firmware selection is the only funny apple thing#asahi-dev · 2022-10-18 11:25
marcanexcept the OTP#asahi-dev · 2022-10-18 11:25
marcan*entire wifi series#asahi-dev · 2022-10-18 11:25
marcanit's all been "shit broadcom has been doing in bcmdhd"#asahi-dev · 2022-10-18 11:25
marcanso far I don't think we've implemented *any* apple-only features whatsover, in the entire pcie series#asahi-dev · 2022-10-18 11:24
marcanno, everything I implemented was already in bcmdhd#asahi-dev · 2022-10-18 11:23
marcan... I basically had to implement a pile of features that bcmdhd supports, because I guess the broadcom guys are too lazy to actually maintain the upstream Linux driver to modern features#asahi-dev · 2022-10-18 11:22
marcanwifi s2idle works \o/#asahi-dev · 2022-10-18 06:51
marcanI can't repro on m1air though#asahi-dev · 2022-10-18 06:51
marcanso I wouldn't be surprised if we need sleeps#asahi-dev · 2022-10-18 06:51
marcanre the tps, remember we already determined you can DoS them if you talk to them too fast#asahi-dev · 2022-10-18 06:49
marcanit's the usb3/tb buffer chips#asahi-dev · 2022-10-18 06:49
marcanjannau: atcrt = ATC retimer#asahi-dev · 2022-10-18 06:46
marcanthe new kernel has povik's patch to fix the channel count issue#asahi-dev · 2022-10-18 06:38
marcanthat said, please test since I intend to go into upstreaming mode again after this release is pushed to users#asahi-dev · 2022-10-18 06:36
marcanthere you go#asahi-dev · 2022-10-18 06:35
marcandid we really drop that many? or was it some large noisy thing?#asahi-dev · 2022-10-18 05:23
marcanso yeah, that tool is broken#asahi-dev · 2022-10-18 05:23
marcanI tried it, it actually should work as far as I can tell, other than not being able to specify 32 bits and the vast majority of MMIO being 32-bit only#asahi-dev · 2022-10-18 05:23
marcanlol, devmem2 only has byte/halfword/word options which map to 8/16/64 (!) bits#asahi-dev · 2022-10-18 05:17
marcanlooks like they left, but actually it looks like the devmem path does use nGnRnE? so it probably should work for real MMIO?#asahi-dev · 2022-10-18 05:09
marcanif you want to play around with device registers, that's what m1n1 is for (which using the hypervisor you can even do while linux is running)#asahi-dev · 2022-10-18 05:08
marcanand I'd say there is ~zero chance of that happening just so people can use devmem#asahi-dev · 2022-10-18 05:08
marcanI'm not entirely sure how the VM / devmem system picks IO mapping modes, but since these SoCs actually require different modes for different things (some PCIe stuff needs to be nGnRE), it would require a UAPI change to expose this via mmap and a change in devmem to let you specify that, to support it properly#asahi-dev · 2022-10-18 05:02
marcannot really, if you enable devmem its user is supposed to be able to kill the system by design#asahi-dev · 2022-10-18 05:00
marcanso basically "don't do that"#asahi-dev · 2022-10-18 05:00
marcanapple silicon platforms require device-nGnRnE mappings for IO, it cannot be done with mmap#asahi-dev · 2022-10-18 05:00
marcanrickybobby23: if you're trying to access registers, it's not supposed to work#asahi-dev · 2022-10-18 04:55
marcanpovik: let me know if you want to move the ucm repo under the org so we can both push :)#asahi-dev · 2022-10-18 04:54
marcannew kernel + pulseaudio packages pushed with the fixes#asahi-dev · 2022-10-18 04:34
marcanalso the UCM config for the Mac Studio needs a fix (ASI stuff should only be set inside the conditional for speakers present)#asahi-dev · 2022-10-18 04:33
marcanpovik: the PA thing was a PA bug as far as I can tell (regression as of dec 2021). there was also a cs42l84 bug I just fixed (see bits branch)#asahi-dev · 2022-10-17 14:31
marcancan you see what the old code does with the new DTs? if it just fails with something we might be okay#asahi-dev · 2022-10-17 14:30
marcanah, yeah, it does, because the model is closer to the old old cpufreq...#asahi-dev · 2022-10-17 14:17
marcanfixed in bits/ (and also something else)#asahi-dev · 2022-10-17 14:17
marcanj`ey: oh dur, that should just return 0 now#asahi-dev · 2022-10-17 14:03
marcanwith arch linux rust packages even, yay#asahi-dev · 2022-10-17 14:03
marcanwell, at least I got make rustavailable happy#asahi-dev · 2022-10-17 14:03
marcanoh, upstream rust support is x86-64 only, lol#asahi-dev · 2022-10-17 13:58
marcansure#asahi-dev · 2022-10-17 13:57
marcanI can try to debug this tomorrow if you want#asahi-dev · 2022-10-17 13:57
marcannot sure if UCM problem or PA bug?#asahi-dev · 2022-10-17 13:56
marcanyup#asahi-dev · 2022-10-17 13:55
marcanso it only works if I start pulse with the jack in#asahi-dev · 2022-10-17 13:55
marcanah, a hotplug breaks it too#asahi-dev · 2022-10-17 13:51
marcan"driver does not support hardware volume control" it claims#asahi-dev · 2022-10-17 13:51
marcanit doesn't work if pulse starts with the jack unplugged#asahi-dev · 2022-10-17 13:51
marcanohhh#asahi-dev · 2022-10-17 13:47
marcan(but I rebooted the whole machine...)