#asahi-dev · 2026-01-17 10:49
integralpilotupstream aicv3 would actually be useful because currently it's one of the patches i need to apply ontop of upstream whenever testing something i'm working on#asahi-dev · 2026-01-17 10:49
svenand the current thing at least needs the generic compatible removed before that's possible#asahi-dev · 2026-01-17 10:49
svensure, that's part of that#asahi-dev · 2026-01-17 10:48
nickchanerm, aicv3#asahi-dev · 2026-01-17 10:48
nickchanthe step before that would be upstream gicv3#asahi-dev · 2026-01-17 10:47
svenfirst step would be to get some minimal M3 device tree on LKML#asahi-dev · 2026-01-17 10:47
nickchanright and testing this stuff would be easier when there's like a working desktop so not now#asahi-dev · 2026-01-17 10:47
svenbut still... much more important things to do before that#asahi-dev · 2026-01-17 10:47
svenoh, because it's a MIDR list and not just enabled on any apple machine#asahi-dev · 2026-01-17 10:43
nickchanwell, a broken SEI generation can take down the whole system#asahi-dev · 2026-01-17 10:43
svenwhy before release?#asahi-dev · 2026-01-17 10:42
nickchanit's something that needs to be figured out before release but not neccessarily now#asahi-dev · 2026-01-17 10:40
svenmight be fun to know, but not very important imho #asahi-dev · 2026-01-17 10:37
integralpilotor do you not think it's worth it?#asahi-dev · 2026-01-17 10:30
integralpilothaven't got KVM enabled so need to rebuild, might take a while (very old hardware for my x86 kernel buildbox), sorry#asahi-dev · 2026-01-17 10:29
nickchangiven the changes to kvm over time#asahi-dev · 2026-01-17 10:29
nickchani figured it wouldn't help#asahi-dev · 2026-01-17 10:28
integralpilotcan grab it quickly, don't think i've saved that.#asahi-dev · 2026-01-17 10:27
nickchankvm has some logs#asahi-dev · 2026-01-17 10:27
nickchanright#asahi-dev · 2026-01-17 10:27
integralpilotas in, dmesg output?#asahi-dev · 2026-01-17 10:27
nickchanintegralpilot: do you have linux boot logs?#asahi-dev · 2026-01-17 10:26
integralpiloti wonder why apple felt the need to even implement this feature?#asahi-dev · 2026-01-17 10:23
nickchanso it was changed to a list of midrs#asahi-dev · 2026-01-17 10:23
nickchanand they did find another systems that implemented the under-specified feature in a way that makes sense#asahi-dev · 2026-01-17 10:21
nickchanhttps://github.com/AsahiLinux/linux/blob/f1879b3d78dcf69a8ff9be3b504c1bd6d5fa2f47/arch/arm64/kvm/vgic/vgic-v3.c#L632 #asahi-dev · 2026-01-17 10:21
integralpiloti've been doing research and think i understand more now, should be able to test if fixed in M3 when i've got free time. unfortunetly am quite busy for the next few days, if anyone else with M3 wants to test it before me feel free.#asahi-dev · 2026-01-17 10:20
nickchan but given that there's 10 lines in the spec there's no asserting which behaivor is correct even though delivering to guest makes more sense#asahi-dev · 2026-01-17 10:20
nickchanmaybe on m3 the serror is delivered to the guest or something#asahi-dev · 2026-01-17 10:19
integralpilotoh that's quite strange#asahi-dev · 2026-01-17 10:18
nickchanhttps://patchwork.kernel.org/project/linux-arm-kernel/patch/20210924082542.2766170-3-maz@kernel.org/ not super sure either#asahi-dev · 2026-01-17 10:14
nickchanand was very under-specified#asahi-dev · 2026-01-17 10:13
nickchanokay it's not exactly an errata it's just a specified feature that absolutely nobody else implemented and also deprecated#asahi-dev · 2026-01-17 10:11
integralpilotsorry, don't know much about vGIC / arm interrupts in general.#asahi-dev · 2026-01-17 10:10
integralpilothow do i make a SError generating access?#asahi-dev · 2026-01-17 10:10
integralpilotoh cool, hopefully it is.#asahi-dev · 2026-01-17 10:09
nickchanBut supposedly that's fixed in m3#asahi-dev · 2026-01-17 10:09
nickchanWhich would of course crash the host were it not for a workaround in kvm#asahi-dev · 2026-01-17 10:09
nickchanon m1 there's an issue if the guest write something that isn't right into the vGIC, it sends an Serror to the host#asahi-dev · 2026-01-17 10:08
integralpilotoh, that's interesting!#asahi-dev · 2026-01-17 10:08
nickchanm1/a14+ does have vGIC without the GIC#asahi-dev · 2026-01-17 10:08
nickchanvirtual generic interrupt controller and GIC is the standard arm interrupt controller#asahi-dev · 2026-01-17 10:07
integralpilotwhat is vGIC, sorry?#asahi-dev · 2026-01-17 10:07
nickchanon m1 that would would generate an SError... to EL2#asahi-dev · 2026-01-17 10:06
nickchanon m3#asahi-dev · 2026-01-17 10:06
nickchanintegralpilot: have you tested what happens if a EL1 guest performs SError-generating accesses on the vGIC#asahi-dev · 2026-01-17 10:02
svenyeah#asahi-dev · 2026-01-17 10:01
integralpilotand mulitple CPUs works totally fine on hypervised linux on M3#asahi-dev · 2026-01-17 10:00
svenit' just weird that it really doesn't like multiple cpus to be active#asahi-dev · 2026-01-17 09:59
integralpilotat least its semi-deterministic :D