#
#asahi-dev · 2026-01-12 12:52
fl0_idintegegralpilot shouldn't be aic3 support on main m1n1? In my logs it says sth like AIC: Version 3 @ 0x292400000, AIC: AIC3 with 1/2 dies, 2528/4096 IRQs, reg_size:40004, config:10000, extintrcfg_stride:04a00, intmaskset_stride:04a00, intmaskclear_stride:04a00, AIC: Configuring 1 external interrupts
#asahi-dev · 2026-01-12 09:27
nickchanm3 should not be any worse than m1/m2 but alas it looks like it's not super straightfoward either
#asahi-dev · 2026-01-12 09:23
nickchanBut m5 is even worse
#asahi-dev · 2026-01-12 09:23
nickchanwell m4/a18 is diffcult in general
#asahi-dev · 2026-01-12 09:23
nickchanit's m5 that's difficult due to exclaves
#asahi-dev · 2026-01-12 09:22
nickchanno, it should be similar to m4
#asahi-dev · 2026-01-12 09:22
chaos_princessis there something crazy about a18 that would make it difficult?
#asahi-dev · 2026-01-12 09:22
nickchanwe will see if asahi got to deal with a18 soon enough
#asahi-dev · 2026-01-12 09:20
nickchanfollowed by x86_64 mac
#asahi-dev · 2026-01-12 09:20
nickchanfollowed by m series pro/max/ultra
#asahi-dev · 2026-01-12 09:20
nickchanfollowed by base m series
#asahi-dev · 2026-01-12 09:20
nickchanso a series get all the security love
#asahi-dev · 2026-01-12 09:18
Shizwhat i find interesting is how late m3 got it
#asahi-dev · 2026-01-12 09:17
nickchannow of course later m3 ipad became a thing anyways
#asahi-dev · 2026-01-12 09:17
nickchanand apple care more about ios than macos
#asahi-dev · 2026-01-12 09:17
nickchanthe reason is probably m2 is already in ipads by then
#asahi-dev · 2026-01-12 09:13
LarstiQchaos_princess: huh, weird
#asahi-dev · 2026-01-12 09:09
integralpilotthanks for taking the time to look through it jannau. :)
#asahi-dev · 2026-01-12 09:09
integralpilotooh interesting sleep and DRAM research nickchan!
#asahi-dev · 2026-01-12 09:09
chaos_princessLarstiQ: for some reason vanilla m2 was the first thing that got it, pro/max m2 only got it in 26
#asahi-dev · 2026-01-12 07:23
nickchanin that **public KDK**'s there's many many sptm strings inside macOSKDKInternal.pkg/Payload/System/Library/Kernels/kernel.development.vminternal4
#asahi-dev · 2026-01-12 07:23
jannauintegralpilot: I'd like to take a look, please hold off. I'll take some time tonight (Central european time) for it
#asahi-dev · 2026-01-12 07:19
nickchanaccording to that one 26.0 beta 4 kdk that apple publicly released there's also a vminternal4 and the sptm on that must also run from M1-M5
#asahi-dev · 2026-01-12 07:17
nickchanalso note that A15/A16/A17 all got sptm treatment in iOS 17 (macOS 14) so whatever differences in gxf did not affect the release schedule
#asahi-dev · 2026-01-12 07:09
nickchanand for any gxf differences between SoCs it must be similar enough for one sptm vresearch1 image to run from M1-M5
#asahi-dev · 2026-01-12 07:07
LarstiQM2 got it before M3?
#asahi-dev · 2026-01-12 07:07
nickchanM3 on macOS 26
#asahi-dev · 2026-01-12 07:06
nickchanM2 and M4 on macOS 15
#asahi-dev · 2026-01-12 07:06
nickchansven: sptm rollout:
#asahi-dev · 2026-01-12 06:11
nickchanwith the wake on network connection all disabled
#asahi-dev · 2026-01-12 06:11
nickchandid any observe macOS in sleep state for longer amounts of time
#asahi-dev · 2026-01-12 06:11
nickchani don't think macOS does that immediately when the lid is closed but
#asahi-dev · 2026-01-12 06:10
nickchanthe deepest sleep state on this platform is very much suspend to RAM
#asahi-dev · 2026-01-12 06:02
nickchani think pmgr-misc on t6000 actually does some of that?
#asahi-dev · 2026-01-12 06:02
nickchanit should be permissible to down the dram frequency / power state in MCC/PMGR
#asahi-dev · 2026-01-12 06:02
nickchanlike
#asahi-dev · 2026-01-12 05:57
nickchannot enough stuff is powered down / set to low power state
#asahi-dev · 2026-01-12 05:56
nickchanI think there's some power savings still to be had there
#asahi-dev · 2026-01-12 05:56
nickchanBut assuming the deepest sleep state is not entered then the peripheral state is entirely up to the discretion of the OS
#asahi-dev · 2026-01-12 05:53
nickchanof course, linux's cpuidle driver doesn't support that and especially that on some socs it means it would end up in EL3 after deep wfi
#asahi-dev · 2026-01-12 05:52
nickchanbefore that all deep wfi is nonretention without setting that disable retention bit
#asahi-dev · 2026-01-12 05:50
nickchanretention deep wfi is added in A12 where most state are kept but at least GPR and FP state is lost
#asahi-dev · 2026-01-12 05:50
nickchanand on recent SoC when waking up from that it will go through reconfig engine with bytecode in AOP SRAM to reinitialize some system state and cpu ends up at the previous IORVBAR
#asahi-dev · 2026-01-12 05:49
nickchanand then there's a sleep state where most of the system powers down not sure how to enter that one yet
#asahi-dev · 2026-01-12 05:47
nickchanthis should be sufficient to enter non-retention wfi and wake up on an interrupt
#asahi-dev · 2026-01-12 05:47
nickchanhttps://github.com/AsahiLinux/m1n1/blob/29fa5411196e8c87af72eb119a496c34fe1f9c18/src/utils.c#L276C9-L276C17
#asahi-dev · 2026-01-12 03:28
integralpilotoh also just thought fl0_id, can you confirm you are on a branch with AIC3 support and you have the AIC config option enabled. not having interrupts might cause mysterious linux boot issues like you are having.
#asahi-dev · 2026-01-12 03:27
integralpilotif no-one has any issues with the MTP/keyboard stuff or wants more time to look through it (thanks if you do!), i might send it upstream tonight :)
#asahi-dev · 2026-01-12 01:59
Shiza slightly less hacky (but still hacky) way might look like this https://txt.shiz.me/YmVjYmVmZT
#asahi-dev · 2026-01-12 01:53
Shizyou might have already guessed that on the m3, pmgr is nowhere near that address so that's just hooking into the breeze

plus ›