#asahi-dev · 2026-06-04 06:36
StanfordAppliedCyber[m]may also be able to mess with commpage#asahi-dev · 2026-06-04 06:22
NickChan[m]evident from the "advertise Apple TSO but it's broken" in some of the logs#asahi-dev · 2026-06-04 06:21
NickChan[m]hacr_el2 is probably raz/wi; apple at least managed to have that behavior on actlr_el1#asahi-dev · 2026-06-04 06:20
jannauthose == the apple counter sysregs#asahi-dev · 2026-06-04 06:20
StanfordAppliedCyber[m]ofc not#asahi-dev · 2026-06-04 06:20
JamesCalligeros[m]Dear god i hope you did not talk to them to inform your development work…#asahi-dev · 2026-06-04 06:20
StanfordAppliedCyber[m]see if we can escalate some of the radars#asahi-dev · 2026-06-04 06:20
jannauyou could try if we can trap those. there are unknown bits in HACR_EL2 and AHCR_EL2 might be another hyp config reg#asahi-dev · 2026-06-04 06:20
StanfordAppliedCyber[m]one of my friends is a very senior engineer who lead a lot of the sptm work and controls a lot of boot, may talk to him about it (obviously he hasn't told me anything that isn't public)#asahi-dev · 2026-06-04 06:19
NickChan[m]as is the current state windows guests will also be broken even after all of that is fixed#asahi-dev · 2026-06-04 06:19
jannauwe probably need to fill a radar about CYC_OVRD as well#asahi-dev · 2026-06-04 06:18
NickChan[m]doesn't have much to do with the actual aic it's however handled in linux's aic driver#asahi-dev · 2026-06-04 06:18
StanfordAppliedCyber[m]oh, i hit that, we have some horrible fiq handling code in my branch currently#asahi-dev · 2026-06-04 06:18
JamesCalligeros[m]But as sven said, they wont “fix” what they dont consider broken (which is anything that is not required for strict arm spec compliance)#asahi-dev · 2026-06-04 06:18
NickChan[m]the fiq virt timer#asahi-dev · 2026-06-04 06:18
StanfordAppliedCyber[m]what's "that aic one" btw? not familiar with that#asahi-dev · 2026-06-04 06:18
JamesCalligeros[m]iboot does it before jumping to its boot payload so yes apple will “fix” what they want to fix with iboot updates#asahi-dev · 2026-06-04 06:17
StanfordAppliedCyber[m]you may be able to patch all the register accesses to svc, add a new syscall to xnu that just calls hvc (or find some way of making a hypercall directly from el0) and then software emulate the access#asahi-dev · 2026-06-04 06:17
Sobek[m]Hmm, how can this be fixed ? Chicken bits / iBoot code change are enough ?#asahi-dev · 2026-06-04 06:17
JamesCalligeros[m]I mean we’re already stubbing and emulating sptm, in for a penny…#asahi-dev · 2026-06-04 06:16
StanfordAppliedCyber[m]i think i'm going to do something horrible#asahi-dev · 2026-06-04 06:16
sventhe timer regs will likely remain disabled because they’re not required for eg linux #asahi-dev · 2026-06-04 06:16
svenso far I’ve only filed a radar about that aic one#asahi-dev · 2026-06-04 06:16
StanfordAppliedCyber[m]amazing, time to think of something interesting#asahi-dev · 2026-06-04 06:15
svenand they won’t fix all registers #asahi-dev · 2026-06-04 06:15
JamesCalligeros[m]Nope#asahi-dev · 2026-06-04 06:15
svennope #asahi-dev · 2026-06-04 06:15
StanfordAppliedCyber[m]are there any known solutions for this other than waiting for apple to fix them?#asahi-dev · 2026-06-04 06:15
svenyeah, it’s the same mechanism that locks out gxf#asahi-dev · 2026-06-04 06:15
StanfordAppliedCyber[m]unfortunate#asahi-dev · 2026-06-04 06:15
StanfordAppliedCyber[m]oh so this is expected?#asahi-dev · 2026-06-04 06:14
JamesCalligeros[m]Some of these will be bugs that apple is aware of and is fixing#asahi-dev · 2026-06-04 06:14
JamesCalligeros[m]Raw boot disables/locks a bunch of stuff including some instructions and some sysregs#asahi-dev · 2026-06-04 06:14
StanfordAppliedCyber[m]actually yeah, does anyone know if there's a precedent for apple impdef registers behaving differently in macho boot vs raw boot? launchd is faulting trying to read ACNTVCT_EL0, the mrs x8 ACNTVCT_EL0 is being decoded as an illegal instruction and launchd is getting killed#asahi-dev · 2026-06-04 04:46
StanfordAppliedCyber[m]is there precedent for that?#asahi-dev · 2026-06-04 04:46
StanfordAppliedCyber[m]i find it hard to believe it's a raw vs macho boot difference#asahi-dev · 2026-06-04 04:46
StanfordAppliedCyber[m]...............but xnu writes them just fine?#asahi-dev · 2026-06-04 04:46
NickChan[m]StanfordAppliedCyber[m]: they are probably all locked#asahi-dev · 2026-06-04 04:45
StanfordAppliedCyber[m]i'm not sure why but writing to some of the new timer registers in el2 fault, despite me doing exactly what xnu is doing#asahi-dev · 2026-06-04 04:38
StanfordAppliedCyber[m]are there any resources from the m3 learnings?#asahi-dev · 2026-06-04 04:26
StanfordAppliedCyber[m]launchd is now exiting for a different reason which i need to figure out#asahi-dev · 2026-06-04 04:26
StanfordAppliedCyber[m]fixed this particular error though#asahi-dev · 2026-06-04 04:26
StanfordAppliedCyber[m]yeah i found all of them in this hypervisor work#asahi-dev · 2026-06-04 04:26
NickChan[m]erm, A16#asahi-dev · 2026-06-04 04:26
NickChan[m]it's first added in M3/A15#asahi-dev · 2026-06-04 04:26
NickChan[m]we know from m3 that there's a whole set of parallel registers to arm's arch timer registers#asahi-dev · 2026-06-04 04:15
StanfordAppliedCyber[m]AND IT'S TIMER BULLSHIT#asahi-dev · 2026-06-04 04:11
StanfordAppliedCyber[m]oh ffs i'm hitting SIGILL on a fucking sysreg#asahi-dev · 2026-06-04 04:11
StanfordAppliedCyber[m]if this is true that means if the kernel works fine there's a way of turning off all the apple bs in userspace#asahi-dev · 2026-06-04 04:10
StanfordAppliedCyber[m]ive heard from multiple people that the userspace in vm images and bare metal boot images is the same, the only difference is the kernel