#asahi-dev · 2026-10-04 13:00
myriaAh true, I guess if only the trap 1 fails...#asahi-dev · 2026-10-04 13:00
integralpilotim trying https://pastebin.com/4fhjyLnG now (which, given the kernel cmdline thing fixes it, i think will fix it)#asahi-dev · 2026-10-04 12:59
integralpilotmyria: thanks, i'll try that#asahi-dev · 2026-10-04 12:59
myriahttps://github.com/sempiternal-aurora/linux/tree/fix-vgic-m3 <- this might fix it? Might need to also disable nv1, but let me know what happens#asahi-dev · 2026-10-04 12:53
myriaMichaelReeves[m] if I send you a patch for the host, can you test if it works?#asahi-dev · 2026-10-04 12:40
MichaelReeves[m]oh btw i sent v2 of regmap: https://lore.kernel.org/asahi/20261004-regcache-lifecycle-submit-v2-0-976d405e89a7@gmail.com/#asahi-dev · 2026-10-04 12:38
MichaelReeves[m]s/lvm/kvm#asahi-dev · 2026-10-04 12:38
MichaelReeves[m]lvm-arm.vgic_v3_group1_trap=1 fixes it#asahi-dev · 2026-10-04 12:27
mmediouni[m]But filtering by EL would also be wrong. What's conventionally done by other platforms is to filter by PSCI SoC ID but not an option here...#asahi-dev · 2026-10-04 12:26
mmediouni[m]so hvf just zeroes it out (and that's why kernel-irqchip=off,gic-version=3,its=on works fine on a Mac host)#asahi-dev · 2026-10-04 12:25
mmediouni[m]but that's more linked to being heterogenous and macOS not supporting core pinning#asahi-dev · 2026-10-04 12:25
mmediouni[m]Apple deliberately only exposes CPU model 0x0#asahi-dev · 2026-10-04 12:13
myriaHmmmm, then yeah, probably the midr based workarounds#asahi-dev · 2026-10-04 12:13
MichaelReeves[m]jannau: thanks. I will look into it#asahi-dev · 2026-10-04 12:13
MichaelReeves[m]software emulation doesn't run into the issue#asahi-dev · 2026-10-04 12:13
MichaelReeves[m]and yes with KVM. #asahi-dev · 2026-10-04 12:12
MichaelReeves[m]myria: its linux guest#asahi-dev · 2026-10-04 12:12
JamesCalligeros[m]sofus: thanks, will send with only the typo fix then :)#asahi-dev · 2026-10-04 12:11
myriaWindows ARM may run into pmuv3 issues#asahi-dev · 2026-10-04 12:11
myriaqemu_kvm I'm assuming? What is the guest?#asahi-dev · 2026-10-04 12:11
jannauMichaelReeves[m]: we have MIDR based workarounds, for example e910baa9c1efdf7634519c135c6723b0fd499683 and possible impdef_pmuv3_cpus in arch/arm64/kernel/cpu_errata.c#asahi-dev · 2026-10-04 12:09
sofuschadmed: description is still correct, the issue was bpp#asahi-dev · 2026-10-04 12:06
MichaelReeves[m]someone reported it on reddit, and I can reproduce. but if I set its=off, then it works, so I think it's something wit virtual PCI interrupts#asahi-dev · 2026-10-04 12:06
MichaelReeves[m]*qemu#asahi-dev · 2026-10-04 12:06
MichaelReeves[m]would aicv3 have changed anything that breaks emu?#asahi-dev · 2026-10-04 12:00
yukaack#asahi-dev · 2026-10-04 12:00
jannaudrop the m4 max mac studio, that should use dcp it supports up to 5 displays#asahi-dev · 2026-10-04 11:58
jannaufinding out that dcp/disp0 is not used. The j773sap template ADT from the ipsw has those nodes but the adt retrieved via `python3 -m m1n1.adt -r ...` when connected to j773s does not#asahi-dev · 2026-10-04 11:56
yukaI don't think I understand what requires device ADT#asahi-dev · 2026-10-04 11:56
yuka> annoying if that requires device ADT#asahi-dev · 2026-10-04 11:54
jannauno, just the dcp related device node, the power-domains are still there#asahi-dev · 2026-10-04 11:53
yukaDISP_CPU no_ps = True#asahi-dev · 2026-10-04 11:53
yukaindeed#asahi-dev · 2026-10-04 11:53
yukaAha#asahi-dev · 2026-10-04 11:52
yukacan easily check that#asahi-dev · 2026-10-04 11:51
yukayou mean that iboot removes the power domains from the adt pmgr devices?#asahi-dev · 2026-10-04 11:50
jannauthe macOS 27 template adt has dcp. annoying if that requires device ADT#asahi-dev · 2026-10-04 11:45
jannauI'd be surprised if mbp did not use dcp0 so just remove it for the mac mini/studio#asahi-dev · 2026-10-04 11:37
yukado I delete the disp domains in t6040.dtsi or in the Mac mini / Mac Studio dts?#asahi-dev · 2026-10-04 11:36
yukais that also the case on the macbooks?#asahi-dev · 2026-10-04 11:36
yuka> the M4 Pro mini uses only dcpext* for its 3 supported displays#asahi-dev · 2026-10-04 11:34
allusivedevok sweet thanks!#asahi-dev · 2026-10-04 11:33
chadmedits not important, the production kernel is easier to get#asahi-dev · 2026-10-04 11:33
allusivedevor does it not matter?#asahi-dev · 2026-10-04 11:33
allusivedevHey guys for the purpose of trying to help with the M4 support for asahi should I get the production kernel or KDK?#asahi-dev · 2026-10-04 11:29
chadmedi am going to resend the interchange modifier patch but if the description needs to change to be accurate for 10-bit formats lmk before i do so#asahi-dev · 2026-10-04 11:29
chadmedsofus: were 32x32 tiles *not* the right size for 10-bit luma interchange? #asahi-dev · 2026-10-04 11:27
allusivedevoh sick#asahi-dev · 2026-10-04 11:26
MichaelReeves[m]https://github.com/ninxsoft/Mist#asahi-dev · 2026-10-04 11:26
MichaelReeves[m]you can use Mist if it's taking too long, it's what I use