#
…

2026-01-11

00:12jmax has quit [Remote host closed the connection]
00:13jmax has joined #asahi-dev
00:13mixi has quit [Ping timeout: 480 seconds]
00:19hdbngr has joined #asahi-dev
00:23mixi has joined #asahi-dev
00:26slowcache has joined #asahi-dev
00:27hdbngr has quit [Ping timeout: 480 seconds]
00:50okb__ has quit [Ping timeout: 480 seconds]
00:50hdbngr has joined #asahi-dev
00:59hdbngr has quit [Ping timeout: 480 seconds]
01:12rany has quit [Read error: Connection reset by peer]
01:12rany has joined #asahi-dev
01:14rany has quit [Remote host closed the connection]
01:20hdbngr has joined #asahi-dev
01:23rany has joined #asahi-dev
01:23rany has quit [Remote host closed the connection]
01:28hdbngr has quit [Ping timeout: 480 seconds]
01:33rany has joined #asahi-dev
01:55linuxgemini has quit [Quit: getting (hopefully fresh) air o/]
01:55linuxgemini has joined #asahi-dev
02:01chadmed_ has joined #asahi-dev
02:01chadmed has quit [Read error: Connection reset by peer]
02:08slowcache has quit [Quit: Konversation terminated!]
02:19rany has quit [Remote host closed the connection]
02:21hdbngr has joined #asahi-dev
02:23rany has joined #asahi-dev
02:24rany has quit [Remote host closed the connection]
02:27ciara is now known as Guest35597
02:27ciara has joined #asahi-dev
02:29Guest35597 has quit [Read error: Connection reset by peer]
02:29hdbngr has quit [Ping timeout: 480 seconds]
02:30rany has joined #asahi-dev
02:37john-cabaj has joined #asahi-dev
02:55midou has quit [Remote host closed the connection]
03:04cremfuelled has joined #asahi-dev
03:06rany has quit [Remote host closed the connection]
03:09icanc has quit [Remote host closed the connection]
03:11chadmed_ has quit [Quit: Konversation terminated!]
03:11chadmed_ has joined #asahi-dev
03:11rany has joined #asahi-dev
03:12cremfuelled has quit [Quit: cremfuelled]
03:21hexdump01 has joined #asahi-dev
03:23hexdump0815 has quit [Ping timeout: 480 seconds]
03:35tobhe has joined #asahi-dev
03:42tobhe_ has quit [Ping timeout: 480 seconds]
03:51MikeM64 has quit [Ping timeout: 480 seconds]
03:54midou has joined #asahi-dev
03:58MikeM64 has joined #asahi-dev
04:09hdbngr has joined #asahi-dev
04:10jvoisin has quit [Ping timeout: 480 seconds]
04:18hdbngr has quit [Ping timeout: 480 seconds]
04:18jvoisin has joined #asahi-dev
04:26nora has joined #asahi-dev
04:28nora_ has quit [Ping timeout: 480 seconds]
05:06jvoisin has quit [Remote host closed the connection]
05:08john-cabaj has quit [Ping timeout: 480 seconds]
05:36jvoisin has joined #asahi-dev
06:03nickchansven: context was trying to see if smc-reboot is really only introduced in later devices
06:04nickchanespecially i did post on the list saying that macsmc-reboot was getting probed without a node
06:04nickchannow T2 still doesn't have mbse but that had always been a weird device
06:05nickchannot sure if off0 would work on mac
06:11hdbngr has joined #asahi-dev
06:12nickchanokay writing off0 to MBSE without doing anything else reboots the machine
06:20hdbngr has quit [Ping timeout: 480 seconds]
06:47nickchanset shutdown flag and write off0 then j314s still reboots
06:48nickchannot an expert in actual silicon mac stuff
08:12hdbngr has joined #asahi-dev
08:19okb has joined #asahi-dev
08:21hdbngr has quit [Ping timeout: 480 seconds]
08:31hdbngr has joined #asahi-dev
08:32svenhrm, had my lid closed last night and couldn't get my macbook to turn on this morning anymore
08:32svenJan 11 02:28:24 sven-macbookair kernel: PM: hibernation: hibernation entry
08:32svenlast dmesg entry, has anyone seen that before?
08:33svennot sure why it even tries to hibernate since we don't really support that
08:35nickchansven: how much battery does it have when closing the lid
08:35svenit's connected to power
08:35svenso 100%
08:35nickchanokay so not any attempt at hybrid sleep
08:37svenwe shouldn't ever try to enter that tbh
08:39hdbngr has quit [Ping timeout: 480 seconds]
08:44Glanzmannsven: I had recently two crashes while trying dp altmode usb/c with the wrong m1n1 version where I had press the powerbutton for a long time to turn off the m1 mini and than it would not start immediatly, but after lets say 30 seconds I could start it again.
08:45sven*sigh*
08:45svendo i need to mention again that that branch is meant for developers who can debug issues?
08:45Glanzmannsven: It works for me, I debugged it by myself by reading the atc code. However I was just answering you questions if someone had problems turning on the m1. And that was that.
08:46GlanzmannBut it is probably unrelated to the problem you're seeing.
08:47svenyes, it's definitely unrelated
08:47violette has quit [Ping timeout: 480 seconds]
08:47integralpilot has joined #asahi-dev
08:48integralpilotive had the same thing, on M3 if i leave alpine running and shut the lid and forget about it, when i open it again its just black and unresposive and wont turn on. had to hold power button for a while and it fixed itself up
08:48integralpilotmight be unrelated or could also be a similar failed hibernation thing
08:48sventhat sounds like my issue, i suspect that happens when it tries to enter hibernate for some reason
08:49svenwhich we just don't support on these machines
08:49integralpiloti didnt think alpine was smart enough to try and hibernate but not sure
08:49svenyeah, not sure either. but dmesg looks like it tried to hibernate
08:49sveni just wonder what makes the system think it can do that
08:50integralpilotsadly no NVMe yet so cant see the dmesgs from when that happened, forever lost in RAM...
08:50svenoh right, do you still need that nvme trace?
08:50integralpilotthat would be helpful if you can get around to it, but not urgent
08:50svenok, will try to grab one later
08:51hdbngr has joined #asahi-dev
08:54svendo you know why nvme doesn't work on m3?
08:54svendoes it just not come up?
08:59integralpilotwriting to a chicken register fails
08:59integralpilotbut the exact issue the chicken fixed will occur if we skip that right
09:00hdbngr has quit [Ping timeout: 480 seconds]
09:01integralpilotwait, i think it's actually the reading of the register (so we apply the bitmask) that fails, but for some reason the write succeeds
09:02svenbut the issue still happens even with writing to it only?
09:02integralpilotbut i don't think it's actually going through
09:03integralpiloti think so, though don't remember exactly
09:13hdbngr has joined #asahi-dev
09:18okb has quit [Ping timeout: 480 seconds]
09:22Glanzmannsven: When I try to echo disk > /sys/power/state on m1 air, it tries to hibernate, but fails, if I boot with the kernel parameter hibernate=no than it says 'operation not permitted' https://pbot.rmdir.de/Dqk3h_oiBPaOoeZ10oVXSg
09:22svenyes, that's obvious. what i want to figure out is how the kernel determines if hibernation is supported without a boot arg
09:22hdbngr has quit [Ping timeout: 480 seconds]
09:38___nick___ has joined #asahi-dev
09:42hdbngr has joined #asahi-dev
09:44okb has joined #asahi-dev
09:50hdbngr has quit [Ping timeout: 480 seconds]
09:55hdbngr has joined #asahi-dev
09:56midou has quit [Remote host closed the connection]
09:59integralpilot has quit [Ping timeout: 480 seconds]
10:06midou has joined #asahi-dev
10:08aditya has joined #asahi-dev
10:14hdbngr has quit [Ping timeout: 480 seconds]
10:15hdbngr has joined #asahi-dev
10:23hdbngr has quit [Ping timeout: 480 seconds]
10:28integralpilot has joined #asahi-dev
10:29hdbngr has joined #asahi-dev
10:33___nick___ has quit []
10:36___nick___ has joined #asahi-dev
10:36___nick___ has quit []
10:38___nick___ has joined #asahi-dev
10:48hdbngr has quit [Ping timeout: 480 seconds]
10:55hdbngr has joined #asahi-dev
11:09xsploit has joined #asahi-dev
11:15xsploit1 has joined #asahi-dev
11:17hdbngr has quit [Ping timeout: 480 seconds]
11:20hdbngr has joined #asahi-dev
11:21xsploit has quit [Ping timeout: 480 seconds]
11:27pb17 has quit [Ping timeout: 480 seconds]
11:30okb has quit [Ping timeout: 480 seconds]
11:40hdbngr has quit [Ping timeout: 480 seconds]
11:46pb17 has joined #asahi-dev
11:50hdbngr has joined #asahi-dev
11:54integralpilot has quit [Remote host closed the connection]
11:59xsploit1 has quit [Quit: Leaving.]
12:02okb has joined #asahi-dev
12:08hdbngr has quit [Ping timeout: 480 seconds]
12:08hdbngr has joined #asahi-dev
12:17aditya has quit [Quit: Connection closed for inactivity]
12:19hdbngr has quit [Ping timeout: 480 seconds]
12:21hdbngr has joined #asahi-dev
12:23T4ForFun has joined #asahi-dev
12:31okb has quit [Ping timeout: 480 seconds]
12:31hdbngr has quit [Ping timeout: 480 seconds]
12:37midou has quit [Read error: No route to host]
12:41hdbngr has joined #asahi-dev
12:50hdbngr has quit [Ping timeout: 480 seconds]
12:57midou has joined #asahi-dev
13:17okb has joined #asahi-dev
13:22___nick___ has quit [Ping timeout: 480 seconds]
13:42hdbngr has joined #asahi-dev
13:45jannauI thought hibernation would be blocked by pinned GPU memory without evictions. hibernate didn't do anything when I tried, could of course have other reasons
13:47svenit's been trying to hibernate for ~2 hours last night before it finally died somewhere i think
13:47svenJan 11 01:34:49 sven-macbookair kernel: Error taking CPU7 down: -95 <-- i think that's what stops it in the end
13:47svenand then it brings everything back up
13:47svenand then immediately tries again
13:48jannauno, it fails due missing CPU offline support: "Error taking CPU7 down: -95", "Non-boot CPUs are not disabled"
13:49svenmaybe userland doesn't notice about that failure and then immediately tries again?
13:50svenand then there's a bug in some driver that only rarely happens and gets hit when we try to constantly hibernate
13:51hdbngr has quit [Ping timeout: 480 seconds]
13:53svenpower management was set to hybrid sleep which is what triggered hibernation i think
14:06jannauthere doesn't seem to be straight forward way to disable hibernation. We could register platform_hibernation_ops that always fail
14:09sventhat would at least prevent all devices from being put to sleep but i'm not sure if that would stop the entry->exit loop
14:10T4ForFun has quit [Remote host closed the connection]
14:11svenat least downstream we could just add a hack to hibernation_available that makes it always return false on apple silicon
14:14svenof_machine_compatible_match against apple,arm-platform should work i guess
14:15KongrooParadox has joined #asahi-dev
14:15svenof_machine_is_compatible is even better
14:15violette has joined #asahi-dev
14:17jannauas downstream hack that's better. platform_hibernation_ops gets called quite late
14:17[KongrooParadox] has quit [Ping timeout: 480 seconds]
14:18T4ForFun has joined #asahi-dev
14:27hdbngr has joined #asahi-dev
14:28Shizmore m3 stuff today i guess
14:32svenyay! also el1_call fix PR would be nice ;)
14:37mattia013 has quit [Quit: mattia013]
14:37mattia013 has joined #asahi-dev
14:38hdbngr has quit [Ping timeout: 480 seconds]
14:40hdbngr has joined #asahi-dev
14:42jannaupushed to asahi-wip. "disk" vanishes from /sys/power/state and returns EPERM when written to anyway
14:53svenworks here as well
15:05hdbngr has quit [Ping timeout: 480 seconds]
15:13hdbngr has joined #asahi-dev
15:15dhewg has quit [Remote host closed the connection]
15:25sven has quit [Ping timeout: 480 seconds]
15:28nickchanapple spent sigificant enginnering effort to explicitly make hibernation possible on apple silicon macos but's it's complicated
15:29nickchanin the meantime the workaround should be well
15:29LarstiQmore so than elsewhere?
15:29nickchanLarstiQ: iOS doesn't support hibernation
15:29chaos_princessarm macs are in some sense oversized iphones and those don't hibernate
15:30nickchanand apple has to make secureboot and stuff still work with hibernation and apple made it work with SEP
15:31rany has quit [Remote host closed the connection]
15:35rany has joined #asahi-dev
15:35svenpeter[m]yeah, hibernation is gonna be complicated as well due to state of co-processors like gpu or DCP
15:37svenpeter[m]And we also don’t even support the basics like taking CPUs completely offline
15:43T4ForFun has quit [Remote host closed the connection]
15:43jannauI think dcp shouldn't be an issue there isn't persistent state and it's in DPMS anyway. GPU is a larger problem as we don't have eviction/memory shrinker. not sure how much fw state there is which needs to be restored in the exact same way as long as jobs are either completed or canceled
15:45svenpeter[m]so maybe it's just gpu that is going to be fun then. and SEP i guess if we ever support that
15:47jannauCPU hotplug in its self is an issue
15:47svenpeter[m]oh yeah, there's a bunch of other issues. i was just talking about co-processors there
15:47svenpeter[m]we still need to figure out some PSCI conduit for proper cpu sleep
15:48chaos_princesssep is probably not that bad depending on what we do with it. the only thing that looks hard is transparent disk encryption if we ever do that
15:49svenpeter[m]that's gonna be extra spicy anyway because iirc you have to stuff something that comes from SEP into that weird nvme iommu
15:49nickchansep isn't a requirement for hibernation... it just wouldn't be... secure
15:49svenpeter[m]not looking forward to finding a sane way to do that and then get it upstream 😄
15:50svenpeter[m]noopwafel / Shiz: nice! even some drive-by readability fixes
15:50chaos_princessi also don't think that the SED ecosystem is very well supported on linux, as nobody uses those for pretty good reasons
15:50nickchanis el1_call in EL1 something that needs to be supported
15:50nickchannow what's SED
15:52noopwafelI decided not to touch more than the minimum required to make this work, for now
15:52chaos_princessself-encrypting drive
15:53noopwafel(the readability fixes are because Shiz is sitting next to me and wanted to understand what the heck the code was doing)
15:53hdbngr has quit [Ping timeout: 480 seconds]
15:53nickchanI run m1n1 in EL1 most of the time but then I just use p.call
15:53svenpeter[m]if you run make format and just squash whatever it fixes i'll merge it
15:54nickchanhave not ran into any complicated scripts that really wants el1_call to be working
15:54Shizmake format is giving me some changes unrelated to the code we changed :D
15:55nickchanlet me guess it's src/smp.c and src/exception.c
15:55svenpeter[m]nickchan: the PR improves "el1_call crashes m1n1" to "el1_call works again", that's good enough for me. feel free to remove el1_call from el1 support if you care
15:55Shiznickchan: yes
15:55Shizto be fair, we did change src/exception.c
15:56nickchanthose are the two files that i keep having formatting issues in
15:56chaos_princessaren't smp and exception because we are using ancient clang-format in ci?
15:57Shizah yeah local clang-format is 21
15:57ShizCI is 18
15:57Shizanyway, I cherry-picked the ones that matched the CI log
15:57okb has quit [Quit: Leaving]
15:58svenpeter[m]maybe we should update clang-format inside the CI then
16:01hdbngr has joined #asahi-dev
16:02svenpeter[m]lol, did we really get stuck in a brk -> EL2 does whatever but doesn't adjust ELR -> back to brk -> EL2 does whatever again loop? 😄
16:02Shizyeah haha
16:02ShizI wonder a bit how the commit ever worked
16:02noopwafelI mean, I guess it didn't
16:03noopwafelthis is one of the reasons the PR took a while, I was kind of second-guessing myself a lot because .. what
16:03noopwafel(and with that in mind, if anyone wants to sanity-check the commits then please do)
16:05svenpeter[m]they look good to me, will wait a bit in case anyone else wants to triple-check before merging
16:11nickchanim going to test el3
16:11nickchani saw they are touched
16:12nickchantime to do it most broken cpu which is cyclone
16:16xsploit has joined #asahi-dev
16:19ravikant_ has joined #asahi-dev
16:19nickchanokay i need to call a symbol in EL3
16:26nickchanand load_inline_c don't work on macos i think
16:27nickchanso i should just hack my tests into m1n1_main
16:31nickchanokay it works fine
16:33nickchansince rust is required either way i think CHAINLOADING build option doesn't need to exist anymore
16:34hdbngr has quit [Ping timeout: 480 seconds]
16:36svenpeter[m]that has a few more side effects, doesn't it?
16:37nickchanwell, it enables chainload support and some workarounds to fix chainload related bugs in apple/previous m1n1
16:38nickchanit also adds /chosen/asahi,m1n1-stage1-version into FDT
16:38nickchanDon't think there's ever a reason to build with chainload off now
16:45hdbngr has joined #asahi-dev
16:54hdbngr has quit [Ping timeout: 480 seconds]
16:55hdbngr has joined #asahi-dev
17:03hdbngr has quit [Ping timeout: 480 seconds]
17:13hdbngr has joined #asahi-dev
17:18xsploit1 has joined #asahi-dev
17:19xsploit1 has quit []
17:21xsploit has quit [Read error: Connection reset by peer]
17:23hdbngr has quit [Ping timeout: 480 seconds]
17:23svenpeter[m]https://github.com/AsahiLinux/m1n1/blob/fa822be70783bd721589a47c4355e76d6ed9c356/src/exception.c#L357 that's broken as well, isn't it? we always overwrite GL's ELR and the normal ELR there
17:28ravikant_ has quit [Ping timeout: 480 seconds]
17:36Shizsvenpeter[m]: can that ever occur? GL3 doesn't exist
17:37Shizoh, it's about the last conditional
17:37svenpeter[m]yuuup. it'll update both GL1_ELR and EL1_ELR when the exception is taken from EL1
17:37kettenisnickchan: we do want to make sure that "asahi,m1n1-stage1-version" has the version number of the stage1 that was actually used and not the version number of the stage2 m1n1 ;)
17:38svenpeter[m]*taken from GL1
17:46hdbngr has joined #asahi-dev
17:49nickchanoops I broke the exception handling code
17:55hdbngr has quit [Ping timeout: 480 seconds]
18:01svenpeter[m]pity that can't explain the weird m3 corruption with the hv though ☹️
18:08noopwafelif only we had a hypervisor to trap writes
18:09hdbngr has joined #asahi-dev
18:11pb17 has quit [Ping timeout: 480 seconds]
18:14pb17 has joined #asahi-dev
18:15hdbngr has quit [Remote host closed the connection]
18:15hdbngr has joined #asahi-dev
18:19svenpeter[m]https://github.com/AsahiLinux/m1n1/issues/463 is actually the first bug in 4 years where i wish i had jtag
18:19alicela1n has joined #asahi-dev
18:22noopwafel(are you trying to nerdsnipe us with other cursed bugs)
18:22svenpeter[m]you told me you like cursed things! 😛
18:22svenpeter[m]i'm just showing you more cursed things!
18:24alicela1n has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
18:24Shizalyssa is like a curse magnet
18:27hdbngr has quit [Ping timeout: 480 seconds]
18:31nickchani mean jtag surely aren't hard to find but there's reasons to not touch them
18:32svenpeter[m]we know where SWD is, we can even mux it to some pins on the type-c port. it's unfortunately fused off though and doesn't do much
18:33nickchanafter all they are the only plausible source of all the firmware keys
18:33svenpeter[m]oh, you meant devices with jtag enabled.
18:33svenpeter[m]yeah, i'd rather not touch them 🙂
18:33alicela1n has joined #asahi-dev
18:37nickchanthis probably doesn't help but i think hypervisor framework is capable of exposing vGL2 on M3+
18:38alicela1n has quit []
18:38nickchanneed to run with internal isa level and have allow-research-guest csr turned on though
18:39nickchanso in theory the hypervisor can be debugged from there
18:39caszue has joined #asahi-dev
18:39nickchanthe massive problem is ofc it would need a lot of upfront porting
18:43nickchanand then praying that the bug can be reproduced in the drastically different environment
18:45nickchanas for fw vresearch1 would run in GL1
18:46alicela1n has joined #asahi-dev
18:47alicela1n has quit []
18:47svenpeter[m]that bug isn't that bad tbh. if it was actually happening regularly and impacting everything we'd probably figure out what's going on in a week or two of actually debugging that thing every day
18:47svenpeter[m]right now i take look at it whenever i remember it exists for maybe 30 minutes
18:52hdbngr has joined #asahi-dev
18:52nickchanwatch the bug to be some sort of erratum
18:55nickchanwouldn't be the first cache related erratum workaround in m1n1
18:56alicela1n has joined #asahi-dev
19:01hdbngr has quit [Ping timeout: 480 seconds]
19:11alicela1n has quit [Ping timeout: 480 seconds]
19:15hdbngr has joined #asahi-dev
19:22bjarne has joined #asahi-dev
19:24hdbngr has quit [Ping timeout: 480 seconds]
19:41ravikant_ has joined #asahi-dev
19:42ravikant_ has quit []
19:48pb17 has quit [Ping timeout: 480 seconds]
19:59pb17 has joined #asahi-dev
20:00hdbngr has joined #asahi-dev
20:04freetolga has joined #asahi-dev
20:10abrender has joined #asahi-dev
20:11abrenderHi Folks - I am considering spending time working to improve the power management during sleep on my M2 Macbook Air. Before I start down that rabbit hole, I was wondering if there is anyone who has already started to tackle that, so that i don't step on their toes? Thanks in advance!
20:21tbodt_ has quit []
20:23tbodt has joined #asahi-dev
20:23tbodt has quit [Remote host closed the connection]
20:26tbodt has joined #asahi-dev
20:27hdbngr has quit [Ping timeout: 480 seconds]
20:28sven has joined #asahi-dev
20:29svenpeter[m]i think chaos_princess has looked into power management in general
20:29chaos_princesssomething like that, yes
20:30svenpeter[m]one thing that i'm aware of that's missing is putting the cpu in their lowest power state, we only very roughly know how that works and there's also quite some effort into getting that upstreamable
20:31chaos_princesswait, don't we have the cpuidle driver that does it?
20:31chaos_princessi thought we did
20:31svenpeter[m]yeah, but not into the lowest state
20:32svenpeter[m]it's that deep sleep state where you can turn off the entire cluster and once it comes back up it lost every state and jumps back to RVBAR (i.e. wherever m1n1 was during boot)
20:32chaos_princesshuh, i thought we knew that one, i was more looking into how to disable entire clusters, or lower fabric states
20:33svenpeter[m]we roughly know how to enter it, but i don't think we ever looked into recovering from that deep sleep
20:33svenpeter[m]i know the core is supposed to come up at RVBAR again but i never saw that happen iirc
20:34chaos_princessdo we have a m1n1 thing for that, or somewhere where it is written? might look into that too
20:35svenpeter[m]there's some information in open source xnu at its entry point
20:35hdbngr has joined #asahi-dev
20:36svenpeter[m]then there's that wfi control register we mess with and iirc it has a lower state you can configure
20:36svenpeter[m]but i'm not sure if that's enough to properly enter that state with a wfi
20:37svenpeter[m]maybe nickchan knows more from the iphone world
20:40svenpeter[m]oh, and you can't observe it from the hypervisor because systemwide sleep breaks that one and results in a reboot
20:46hdbngr has quit [Ping timeout: 480 seconds]
20:52cyberheadi wanted to chime in and to give information that I strongly suspect that the base M4 has 1 dcp and 2 dcpext
20:53jannauthe cpuidle driver only uses the deeper wfi state which only retains program counter and stack pointer
20:53cyberheadmy config is a 2025 MacBook Air, using the built-in monitor and two external monitors, one over usb-c and another over official Apple HDMI adapter
20:54jannaucyberhead: that's accurate but known since apple published the devicetree for the M4 ipad pro
20:55cyberheadjannau: i was looking around in the wiki
20:56hdbngr has joined #asahi-dev
21:03abrenderokay, thanks @svenpeters and @chaos_princess. So it sounds like if I were to dig further into all of the above I wouldn't be stepping on anyones toes? You're not aware of anyone that would come back to me saying "oh, we already have it working and are just productionizing it, you've just wasted your time", right? :)
21:04chaos_princessdepends on what you are going for
21:05hdbngr has quit [Ping timeout: 480 seconds]
21:11Shizthe 14.3.1 installer automatically updated to 14.8.3 while installing :')
21:11Shizdoing dfu restore now
21:14hdbngr has joined #asahi-dev
21:16tbodt has quit []
21:17tbodt has joined #asahi-dev
21:20chaos_princesssvenpeter[m]: isn't the thing you are describing already implemented in m1n1 as deep_sleep/SLEEP_GLOBAL?
21:22svenpeter[m]i don't think so
21:22svenpeter[m]i think that uses the msr that controls at least part of it
21:23svenpeter[m]but i don't think we ever managed to enter the deepest sleep state and then wake up from it at RVBAR again
21:27cow123 has quit [Ping timeout: 480 seconds]
21:29hdbngr has quit [Ping timeout: 480 seconds]
21:30amarioguy has joined #asahi-dev
21:36hdbngr has joined #asahi-dev
21:38sventhat maybe enters the deep sleep state, but i haven't seen it exit from that one and show up at RVBAR again
21:39svenhuh.. actually, looks like it's implemented for the secondary cores at least
21:39svenso maybe we know how it works
21:40sventhe linux driver definitely doesn't implement any of that though
21:46dhewg has joined #asahi-dev
21:47chaos_princessi am a bit confused about what oss xnu is trying to do, as it claims that deep sleep registers are global, but configures it on every cpu before doing a wfi
21:50natalie has quit [Quit: Connection closed for inactivity]
21:53hdbngr has quit [Ping timeout: 480 seconds]
21:54hdbngr has joined #asahi-dev
21:56kettenischaos_princess: maybe they're trying to say that deep sleep won't be reached unless it is enabled on all cores?
22:03hdbngr has quit [Ping timeout: 480 seconds]
22:05tbodt has quit []
22:05hdbngr has joined #asahi-dev
22:05tbodt has joined #asahi-dev
22:13tbodt has quit []
22:16tbodt has joined #asahi-dev
22:19hdbngr has quit [Ping timeout: 480 seconds]
22:21hdbngr has joined #asahi-dev
22:30hdbngr has quit [Ping timeout: 480 seconds]
22:39tbodt has quit []
22:41tbodt has joined #asahi-dev
22:44Shizso, dumb question: what's the mcc specifically? it seems to be a memory controller of some kind, but its exact function is a bit unclear to me
22:46Shizjust cache?
22:46chaos_princessmemory, cache, other magic
22:47Shizseems maybe important
22:47noopwafelthat last 'c' is not just 'curses'?
22:51chaos_princessi've spent like 15 seconds trying to work in the last c :P
22:52JamesCalligeros[m]I saw Memory Cache Controller somewhere
22:52JamesCalligeros[m]Cant remember where
22:58svenwe used to boot linux and even macos in the hv without mcc.c iirc, it just makes things faster
23:00noopwafelwe wasted some time trying to downgrade macos and now we are back to staring at the broken remains of macos boots
23:01chaos_princessnoopwafel: install 14.latest via recovery, then use the installassistant to make a separate install of 14.desired
23:01chaos_princessmust use specifically installassistant from macos for that trick to work
23:02noopwafelwe wanted 14.3.X based on your advice the other day, but apparently you _need_ 23D2057
23:07svenso i'm still wondering what would cause all that brokeness on m3 but not on m1/2
23:08svenif it was some new feature xnu exercises we'd also see all that crap with macos 26+ on m1/2. has anyone ever tried to run that in the hv?
23:09noopwafelthat is probably a good next step if no-one has. we have a bunch of ???s which we're chasing at random for now..
23:11alxdrl has quit [Quit: Textual IRC Client: www.textualapp.com]
23:12svennow if that works i can only think of hardware differences and then we're getting into really obscure stuff. like maybe m3 speculates harder and we're missing barriers somewhere? i dunno
23:13noopwafelis it permissible/desired to add comments to explain what a bunch of the hv stuff is doing (especially in the proxyclient)?
23:13chadmed_all documentation improvements are welcome :)
23:14svendon't think anyone's going to complain if you make the code more accessible :-)
23:14chadmed_usual caveats though, no '# Assign variable' tier first year bachelor of react and node type stuff :P
23:14chadmed_ is now known as chadmed
23:16noopwafel"this is a silly comment" is an easy PR comment to make anyway I guess :p
23:16noopwafeljust wanted to make sure there was not a desire to keep some of it as magic. thaks.
23:16svenoh no, absolutely not
23:16chadmedundocumented magic is absolutely unintentional
23:18Shizit seems the pmgr hack addresses are completely not even near where the pmgr on the m3 is :p
23:18Shiz(is there an archive of adt files somewhere?)
23:20chadmedyou can dump the adt out with m1n1 or are you looking to compare with a machine you dont ow
23:20svenfyi: be careful when posting adt files dumped with m1n1, they contain stuff like bluetooth and wifi keys in not-so-obvious places
23:20chaos_princesswhat j-code is your box
23:20sventhe ipsws contain the adt templates though and you can extract them, they should have stuff like MMIO addresses
23:22hdbngr has joined #asahi-dev
23:22Shizoh nice
23:23Shizchaos_princess: current box we're working on is J615, I got the ADT for it here
23:23Shizsecondary box (my laptop) is J516AP (not confusing at all)
23:23Shizso disabling the pmgr hacks gives us a semi-consistent 'PMGR failed to come online in 30 secs', which is very much what you'd expect
23:23Shizconsidering the clock thing
23:26ShizPMP failed*
23:28noopwafel(and it fails later in the boot process than it often gets..)
23:29chaos_princessyou don't really need pmp to work (at least for linux)
23:31hdbngr has quit [Ping timeout: 480 seconds]
23:33hdbngr has joined #asahi-dev
23:35svenSPTM was only rolled out for M3 and M4, wasn't it?
23:37chaos_princessat least the im4p is there in ipsw
23:39svenokay, i was wondering if they maybe changed something subtle about gxf/sprr that SPTM needs with m3 which now breaks the hv in strange ways
23:39svenbut that's probably not it then
23:44hdbngr has quit [Ping timeout: 480 seconds]
23:56hdbngr has joined #asahi-dev
23:57nela has joined #asahi-dev
…