#asahi-dev · 2022-10-20 13:17
marcanso either it's a hardware thing or hardware that only exists on the laptops#asahi-dev · 2022-10-20 13:16
marcanquick test on M1 Studio says all CPUs are idle and no weird MMIO traffic is happening#asahi-dev · 2022-10-20 13:13
marcanunless it's an obvious "something eating cpu or poking MMIO" problem (easy to see from HV)#asahi-dev · 2022-10-20 13:13
marcanthat sounds like some dumb bug, though we might need hwmon to debug it#asahi-dev · 2022-10-20 08:43
marcanyup, let me take care of that today#asahi-dev · 2022-10-20 08:31
marcan:D#asahi-dev · 2022-10-20 08:29
marcanbut the split userspace support is quite another story#asahi-dev · 2022-10-20 08:29
marcanthat's the first thing macOS did, yeah, they turned that into an expression instead of a constant#asahi-dev · 2022-10-20 08:18
marcanfor something so fundamental, it could just be a quirk based on CPU IDs (if apple -> 16K) I suspect#asahi-dev · 2022-10-19 05:36
marcanI wonder if SEP is involved...#asahi-dev · 2022-10-19 05:36
marcanbut I didn't see anything that clearly said link keys#asahi-dev · 2022-10-19 05:36
marcannvram has some stuff, possibly for a subset of devices only? (maybe HID only?)#asahi-dev · 2022-10-19 05:34
marcanthen how do they sync between macOS installs? I assume they have to do that#asahi-dev · 2022-10-19 05:34
marcanohh, keychain. huh.#asahi-dev · 2022-10-19 04:22
marcanI guess in practice these things can be disabled via cmdline options, but I feel like switching packages is more user-friendly than telling people to put stuff in /etc/default/grub to fix regressions...#asahi-dev · 2022-10-19 04:20
marcanI could probably have the same PKGBUILD make both variants, and just do it as a .config difference#asahi-dev · 2022-10-19 04:19
marcanthoughts?#asahi-dev · 2022-10-19 04:18
marcanthe idea being that users can install it without switching wholesale to asahi-dev, and can go back if it doesn't work for them#asahi-dev · 2022-10-19 04:18
marcanit would basically be the same thing as linux-asahi in asahi-dev, but pushed less often but enable more bleeding edge stuff (like sleep mode and eventually DCP and AGX)#asahi-dev · 2022-10-19 04:17
marcanhm, I'm considering having a linux-asahi-dev package#asahi-dev · 2022-10-19 02:49
marcannot sure...#asahi-dev · 2022-10-19 02:49
marcannvram has a bunch of bluetooth related stuff, but I didn't see anything that clearly looked like keys?#asahi-dev · 2022-10-19 02:49
marcanI tried following the Arch instructions for that but the plist mentioned no longer has the keys as far as I can tell#asahi-dev · 2022-10-19 02:48
marcansven: do you know how bluetooth link keys are stored/shared? we're going to have to figure out how to sync that with macos at some point#asahi-dev · 2022-10-19 00:32
marcanno idea about the SPI ones though#asahi-dev · 2022-10-19 00:32
marcanshould be easy on the M2, just suspend rtkit#asahi-dev · 2022-10-19 00:32
marcanjannau: I expected that, yes we need to add proper PM to the keyboard stuff :)#asahi-dev · 2022-10-18 13:39
marcanthere are much bigger things to fix first that will make the overall machines and s2idle more efficient#asahi-dev · 2022-10-18 13:38
marcanthere is "real" full sleep mode on these machines, but I think it won't save that much power, not enough for us to bother with it for a while#asahi-dev · 2022-10-18 13:38
marcanit's just that s2idle forces everything to be idle#asahi-dev · 2022-10-18 13:38
marcanand honestly that will always be the main benefit, because s2idle by *definition* provides zero power saving over an idle system, because an idle system is already supposed to save power#asahi-dev · 2022-10-18 13:36
marcanthe main benefit right now is knowing that the system is actually forced to be idle, instead of hoping apps aren't eating CPU#asahi-dev · 2022-10-18 13:36
marcanhaven't tested, honestly I don't expect much on laptops because I think the biggest power hogs are probably CPU and display scanout (which don't change)#asahi-dev · 2022-10-18 13:28
marcanjoske: ?#asahi-dev · 2022-10-18 13:28
marcansince upstream net apparently doesn't like reviewing "big" patch series, where "big" refers exclusively to the number in the subject line#asahi-dev · 2022-10-18 13:27
marcanbtw, the wifi stuff is 8 new patches because I broke it up, but given my previous experience upstreaming this stuff, I'm tempted to squash them into 2 or so...#asahi-dev · 2022-10-18 13:26
marcanI think that came with the same update that broke things in uboot#asahi-dev · 2022-10-18 13:25
marcanyes, it's been there at least since ventura I think#asahi-dev · 2022-10-18 13:24
marcanpovik: I suspect the mca errors are because the clocks are off? we might have to add code to mca to bring up clocks (dummy ones) around rpm resume so the power states can change properly, or otherwise change the pmgr code so it doesn't complain about this particular case#asahi-dev · 2022-10-18 13:23
marcanhttps://gist.github.com/marcan/bdeb06fd3048481f88490870bf361ff5 <- how it looks on an m1 mini#asahi-dev · 2022-10-18 13:23
marcanright now I think the only functional wakeup source is the SMC button/lid#asahi-dev · 2022-10-18 13:22
marcanpushed the wifi S3 fixes to asahi-wip and the bits branch. I'm not going to enable suspend in the arch kernel yet, but feel free to try it if you're building your own kernels.#asahi-dev · 2022-10-18 11:30
marcanbut that may not quite be necessary for a first shot#asahi-dev · 2022-10-18 11:30
marcanapple actually sends wifi commands when the BT side starts streaming, which is going to be... fun to wire up#asahi-dev · 2022-10-18 11:30
marcanit exists in brcmfmac already, somewhat, but done differently to how apple does it#asahi-dev · 2022-10-18 11:27
marcan(cc sven)#asahi-dev · 2022-10-18 11:27
marcanthe funny btcoex stuff does seem to be apple-exclusive, but when I tried to implement it it didn't help much, so I think we really need the BT side flag first#asahi-dev · 2022-10-18 11:26
marcan(or my SD card is borked)#asahi-dev · 2022-10-18 11:26
marcanjannau: might be a regression#asahi-dev · 2022-10-18 11:26
marcanand then the wifi side will want some changes to play well with that