#asahi-dev · 2026-10-08 17:53
yukafirst finding: dispext2 needs to be removed as well#asahi-dev · 2026-10-08 17:32
yukaI am in front of an M4 Pro MacBook now, yes#asahi-dev · 2026-10-08 17:23
nickchanwriting incorrect things into the vgic then immediately exiting the vm might reproduce the situation I said#asahi-dev · 2026-10-08 17:21
Hel3ziyuka: Hey yuka, did you get the M4 Pro MacBook?#asahi-dev · 2026-10-08 17:21
nickchanI think M2 and later do route the vGIC SError to the guest in most cases#asahi-dev · 2026-10-08 17:20
nickchananother thing is SErrors are asynchronous so in principle by the time the SError is ready to be delievered, HCR_EL2.TGE=1 may already be set and so the only place to deliver it to is... the host#asahi-dev · 2026-10-08 17:04
yukamaz: can you share the reproducer for the seis erratum?#asahi-dev · 2026-10-08 13:51
mmediouni[m]nickchan: sent over DM, they're using an impdef bit in ESR as far as I can tell#asahi-dev · 2026-10-08 13:36
nickchanwhat's the ESR in that case?#asahi-dev · 2026-10-08 13:25
mazbut you don't *know* what triggered the error. there is no syndrome information.#asahi-dev · 2026-10-08 13:25
mmediouni[m]Timers are where things shift outright to crazy though...#asahi-dev · 2026-10-08 13:24
mmediouni[m]They do this thing where if it's vGIC it's always from the guest (tm) - because they only target apple hardware and don't care about anything else#asahi-dev · 2026-10-08 13:22
mazmmediouni[m]: except that they demonstratively fire on the host with the correct timing, and are not synchronised by ESB.#asahi-dev · 2026-10-08 13:16
mmediouni[m]macOS doesn't do this but considers those SErrors as always from the guest#asahi-dev · 2026-10-08 13:12
myriahttps://termbin.com/19ofh is what I was going to use, applied atop asahi-wip#asahi-dev · 2026-10-08 13:00
yuka> if you send the patch with the same diff I sent [...]#asahi-dev · 2026-10-08 13:00
yukathis is why I said earlier:#asahi-dev · 2026-10-08 13:00
yukaif there were relevant changes in between#asahi-dev · 2026-10-08 12:59
yukaI think I should re-test whatever is your current proposal based on linux-next or at least 7.3-rc6#asahi-dev · 2026-10-08 12:59
yukabut you said they are now one?#asahi-dev · 2026-10-08 12:59
yukato broken_seis I guess#asahi-dev · 2026-10-08 12:59
yukaah sorry#asahi-dev · 2026-10-08 12:58
yukaI *only* successfully tested on 7.1 / asahi-wip *with* adding the parts to has_vgic_v3 list: https://termbin.com/44r8m#asahi-dev · 2026-10-08 12:57
yukaWhat exactly would I test?#asahi-dev · 2026-10-08 12:49
MichaelReeves[m]so its just v1 and then add all 'em to has_vgic_v3 too#asahi-dev · 2026-10-08 12:49
MichaelReeves[m]thank you!#asahi-dev · 2026-10-08 12:48
myriaI'll test on m3 max again#asahi-dev · 2026-10-08 12:43
MichaelReeves[m]actually I think before has_vgic_v3 (in additional to the others) is added it needs to be checked on M4 (I'll do M3 now). yuka could you do that (whenever you've got time) and then after it's confirmed, then I'll send v2? thank you!#asahi-dev · 2026-10-08 12:32
jannauor reply with an additional patch to your series#asahi-dev · 2026-10-08 12:31
MichaelReeves[m]or for a patch this small is v2 OK now or tomorrow morning?#asahi-dev · 2026-10-08 12:30
MichaelReeves[m]ill send v2 in a few days#asahi-dev · 2026-10-08 12:30
MichaelReeves[m]oh well sorry #asahi-dev · 2026-10-08 12:30
MichaelReeves[m]but then there's also has_vgic_v3 needed to stop software emulation (which doesn't break it, which is why our tests worked, but it makes it slow)#asahi-dev · 2026-10-08 12:29
MichaelReeves[m]the two lists in the termbin link (impdef_pmuv3_cpus, broken_seis) were merged into 1 list now (which I added)#asahi-dev · 2026-10-08 12:28
MichaelReeves[m]crap#asahi-dev · 2026-10-08 12:27
MichaelReeves[m]wait#asahi-dev · 2026-10-08 12:27
MichaelReeves[m]sashiko sent a hallucination #asahi-dev · 2026-10-08 12:25
MichaelReeves[m]sorry I shouldnt have said 7.2 I should have said <7.3#asahi-dev · 2026-10-08 12:25
MichaelReeves[m]but in 7.3 they are 1 thing now#asahi-dev · 2026-10-08 12:25
MichaelReeves[m]they were 2 seperate things#asahi-dev · 2026-10-08 12:25
MichaelReeves[m]in before 7.3#asahi-dev · 2026-10-08 12:25
yukaasahi-wip#asahi-dev · 2026-10-08 12:25
MichaelReeves[m]yes#asahi-dev · 2026-10-08 12:25
yukaI tested based on 7.1#asahi-dev · 2026-10-08 12:24
yukaUhm#asahi-dev · 2026-10-08 12:23
MichaelReeves[m]thats ok!!#asahi-dev · 2026-10-08 12:23
myriathank you, sorry I was confused#asahi-dev · 2026-10-08 12:23
MichaelReeves[m]s/is/in#asahi-dev · 2026-10-08 12:23
MichaelReeves[m]is 7.3 they are 1 thing now