#asahi-dev · 2026-06-24 21:21
jannauno problem, hopefully easy to figure out via macos logs#asahi-dev · 2026-06-24 21:20
chaos_princessbut there is something weird about avd specifically, as there is a avd-specific pmp thing#asahi-dev · 2026-06-24 21:19
chaos_princessno idea, i don't have any multi-die devices#asahi-dev · 2026-06-24 21:18
jannauchaos_princess: do you have an idea idea how pmp works on multi-die? Not having to override power-domains for Ultra devices is a reason to look into it#asahi-dev · 2026-06-24 21:15
jannauboth avd instances on M2 Ultra work#asahi-dev · 2026-06-24 20:09
svenjust sounds like something where creating a simple reproducer is going to be rather annoying :D#asahi-dev · 2026-06-24 20:08
yukaBad: M4 E-Core secondary / M4 P-Core secondary#asahi-dev · 2026-06-24 20:07
yukaGood: A18 Pro boot core / A18 Pro E-Core secondary / A18 Pro P-Core secondary / M4 Boot Core#asahi-dev · 2026-06-24 20:07
yukaIf this behavior is indeed directly caused (or prevented) by chickens, we can compare the following:#asahi-dev · 2026-06-24 20:05
yukait's at least deterministic now#asahi-dev · 2026-06-24 20:05
yukaas in I need it? ;)#asahi-dev · 2026-06-24 20:04
svengood luck!#asahi-dev · 2026-06-24 20:04
yukaFirst I should probably create an isolated reproducer outside of linux#asahi-dev · 2026-06-24 20:02
svenif we complain about this we better be sure there's a difference or apple will just ignore us i guess#asahi-dev · 2026-06-24 20:02
svenwe should see the difference in chickens then, shouldn't we? we can still read sysregs, we just can't write them#asahi-dev · 2026-06-24 20:01
yukaI think A18 Pro is fine#asahi-dev · 2026-06-24 20:00
yukain raw boot mode#asahi-dev · 2026-06-24 20:00
yukabecause they messed up the chickens on the M4 secondary cores in particular?#asahi-dev · 2026-06-24 19:59
svenmaybe a stale entry in the tlb? but why would that only happen on m4 now? :/#asahi-dev · 2026-06-24 19:57
svenhuh, wtf, doesn't that pte even make the page writable? or did i decode that wrong?#asahi-dev · 2026-06-24 19:38
yukalog: https://x0.at/epMq.txt#asahi-dev · 2026-06-24 19:32
yukaOkay, I can actually reproduce the M4 crashes deterministically using jannau's switch_boot_cpu PR#asahi-dev · 2026-06-24 17:46
yukaindeed lol#asahi-dev · 2026-06-24 17:46
jnnone second for ever major version ;p#asahi-dev · 2026-06-24 17:46
yukaand 26.6 is already hard wired into my muscle memory#asahi-dev · 2026-06-24 17:45
yukatyping numbers is hard#asahi-dev · 2026-06-24 17:45
yuka*27.0b2#asahi-dev · 2026-06-24 17:45
yukaon 26.0b2#asahi-dev · 2026-06-24 17:45
yukasorry#asahi-dev · 2026-06-24 17:45
yukaMacbook neo on 26.6b2 takes 27(!) seconds until it executes m1n1#asahi-dev · 2026-06-24 17:44
yuka(which m1n1 also supports, but linux not yet)#asahi-dev · 2026-06-24 17:43
yukathey use the new dockchannel uart / debugusb#asahi-dev · 2026-06-24 17:43
yukaiboot and xnu don't do output on serial normally#asahi-dev · 2026-06-24 17:43
enverbalalicwould this give iboot output normally, or is this just m1n1/xnu/linux serial ?#asahi-dev · 2026-06-24 17:42
yukasoftware is unlikely#asahi-dev · 2026-06-24 17:42
enverbalalicok... i tried 2 more thick cables, still no serial. I'm inclined to say that something software-wise is not working here with the combo i'm using. I'm tempted to try to update the m2 to 26.6b2 as well#asahi-dev · 2026-06-24 17:41
yukalike the iboot part#asahi-dev · 2026-06-24 17:40
yukaA18 Pro takes significantly longer to boot with beta 2#asahi-dev · 2026-06-24 17:39
yukaWhat was fixed however is being unable to use bputil or csrutil on a non-dfu install#asahi-dev · 2026-06-24 17:38
yukaso basically unchanged from beta 1#asahi-dev · 2026-06-24 17:38
yukaand M4 I keep getting random iboot panics#asahi-dev · 2026-06-24 17:38
yuka27.0b2 does not have SYS_IMP_APL_VM_TMR_FIQ_ENA_EL2 unlocked on A18 Pro or M4#asahi-dev · 2026-06-24 17:12
sveniirc dp aux is required for link training so even 1080p wouldn't work without that#asahi-dev · 2026-06-24 17:11
enverbalalicyeah that's what i'm thinking as well... and it's not like i'm running a 1080p screen from it, it works for 4k@144#asahi-dev · 2026-06-24 17:11
svenbut if your display works with that cable i'm pretty sure it will have SBU connected because the DP AUX channels is on those lines#asahi-dev · 2026-06-24 17:10
svenyou could try to change the VDM to use D+/D- instead#asahi-dev · 2026-06-24 17:10
svenso there are cables that have the cc lines connected but no the sbu lines? huh#asahi-dev · 2026-06-24 17:06
yukaso having it work for display output is no guarantee that it is actually to-spec#asahi-dev · 2026-06-24 17:06
yukausb has quite a few fallback mechanisms to continue working if the cable is not to-spec#asahi-dev · 2026-06-24 17:05
enverbalalici'll be very surprised if this one is not a full-fat one