#asahi-dev · 2026-10-06 11:05
MichaelReeves[m]yep, that's the problem this should fix, hopefully#asahi-dev · 2026-10-06 11:04
yukawith nested virt, the host also sometimes hangs#asahi-dev · 2026-10-06 11:03
yukaI know M3 Max is weirdly broken. The guest does start to some degree but then hangs (nixos tests never reach their backdoor target).#asahi-dev · 2026-10-06 11:03
yukaM3 Max is very easy for me to test, as it's my main device and already has everything installed. On M4, I'll need a few minutes to put together a setup where I can boot a more complete system with qemu and networking.#asahi-dev · 2026-10-06 10:56
MichaelReeves[m]yes please thanks#asahi-dev · 2026-10-06 10:56
myriaWant me to test the patch on M3 max too?#asahi-dev · 2026-10-06 10:54
MichaelReeves[m]thank you! can you try KVM without it and then what should happen is either EL2 SError or it hangs. and then apply the patch (will send soon) and it should work fine#asahi-dev · 2026-10-06 10:53
yukawhere can I find the patch?#asahi-dev · 2026-10-06 10:53
yukaMichaelReeves[m] would gladly test#asahi-dev · 2026-10-06 10:43
chadmedthey left again, hopefully they read the logs#asahi-dev · 2026-10-06 10:42
MichaelReeves[m]did keyboard just work with adding it to DT? there were no dchid changes?#asahi-dev · 2026-10-06 10:38
anon_Any comments?#asahi-dev · 2026-10-06 10:38
anon_Hello, I bet you guys have read my misplace post in #asahi about m5 nvme.#asahi-dev · 2026-10-06 10:17
myriaMichaelReeves[m]: I missed it too, dw. Probably related to some of the funky PMU stuff they do, and how we emulate PMUv3 on it for guests#asahi-dev · 2026-10-06 10:13
MichaelReeves[m]would you like to test? (if not cause ur busy with other stuff it's fine because its harmless to just send m4 later, it's tiny patch). if so, thanks & I'll wait before sending series.#asahi-dev · 2026-10-06 10:06
yukaI should test it to confirm#asahi-dev · 2026-10-06 10:06
yukaI expect it will be similar to M3#asahi-dev · 2026-10-06 10:06
yukasomeone has tested kvm on M4, but I'm afraid they're not in here#asahi-dev · 2026-10-06 10:05
MichaelReeves[m]from Myria's branch I don't think someone's done m4 yet so I'll just keep m3 for now I think#asahi-dev · 2026-10-06 10:03
MichaelReeves[m]so just m3 or has anyone tested m4 PMU/KVM (its all worked into the one generic apple cpus errata)#asahi-dev · 2026-10-06 10:03
MichaelReeves[m]ok i was just saying because myria mentioned it might be worth doing that before but probably good idea not to then#asahi-dev · 2026-10-06 10:02
jannau_MichaelReeves[m]: do not add SoCs nobody has tested#asahi-dev · 2026-10-06 10:00
MichaelReeves[m]anyway... at least I know now#asahi-dev · 2026-10-06 10:00
MichaelReeves[m]I didn't notice because that other thing was making the changes in a completely different place than I expected (and added my patch) and also its naming didnt make it sound like it included that thing as well#asahi-dev · 2026-10-06 09:59
MichaelReeves[m]thankfully I was doing some last checks before I send#asahi-dev · 2026-10-06 09:59
MichaelReeves[m]and while we're doing it might as well add m4, m5, m6 and the other MIDRs myria has ready#asahi-dev · 2026-10-06 09:58
MichaelReeves[m]it just needs to be adding M3 to that#asahi-dev · 2026-10-06 09:58
MichaelReeves[m]so#asahi-dev · 2026-10-06 09:58
MichaelReeves[m]ok so turns out the "new errata" I made in my patch was already part of a larger "apple CPUs broken SEIS" thing that all M1 and M2 are in#asahi-dev · 2026-10-06 09:58
MichaelReeves[m]arghh im dumb#asahi-dev · 2026-10-06 09:28
chaos_princessmaybe the vendors were onto something with "just put code into the firmware instead of having to upstream it" :P#asahi-dev · 2026-10-06 09:26
noopwafeloh I already said the latter :P but I cannot praise enough#asahi-dev · 2026-10-06 09:26
noopwafelLarstiQ: yes it's really not that huge a diff in the end on kernel side :) current diff (and I did end up fixing it btw, \o/) is <5k lines and most of that is basically churn like added ::ver. the refactor of much of initdata to m1n1 helps a lot.#asahi-dev · 2026-10-06 09:18
sofusoh wait sorry#asahi-dev · 2026-10-06 09:18
sofusnot really tho, it would also need to do the same in avd-drv#asahi-dev · 2026-10-06 09:17
chadmedyeah#asahi-dev · 2026-10-06 09:17
chaos_princessnp then, more commenting on the fact that 631 is technically a valid fix, just wrapped in layers of... whatever that is#asahi-dev · 2026-10-06 09:17
chadmedor just wait until we rebase on whatever kernel version ends up with avd in it upstream#asahi-dev · 2026-10-06 09:16
chadmedid rather merge the actual author's fix :P#asahi-dev · 2026-10-06 09:16
chadmedim not in favour of merging 631#asahi-dev · 2026-10-06 09:16
sofusits done totally differently in my ml version. I can update the whole thing to use kvzalloc instead but thats likely not necessary if 615 gets merged#asahi-dev · 2026-10-06 09:15
chadmedso of course when noobs get a hold of the robot and go "make apple go brrr make no mistakes" they dont know theyre supposed to get rid of the verbose slop comments /* Assign value to variable */#asahi-dev · 2026-10-06 09:15
chadmedand of course uni course material being waaaay too verbosely commented because the whole point is to teach noobs what things do#asahi-dev · 2026-10-06 09:14
chadmedthe problem is the terabytes and terabytes of babbys first hellorld program tutorials that the robot trains on#asahi-dev · 2026-10-06 09:12
chaos_princesslike, didn't the whole "literate programming" fad die a while ago, why are we bringing it back#asahi-dev · 2026-10-06 09:11
chaos_princessnuke the stupid comment and merge it#asahi-dev · 2026-10-06 09:09
chadmedcommit message longer than the patch #asahi-dev · 2026-10-06 09:09
chadmedmy eyes just glaze over when i see this bullshit#asahi-dev · 2026-10-06 09:08
MichaelReeves[m]idk if that's different though#asahi-dev · 2026-10-06 09:08
MichaelReeves[m]yep https://github.com/AsahiLinux/linux/pull/631