#asahi · 2026-01-08 08:44
chaos_princesswill save us both time explaining how things work#asahi · 2026-01-08 08:44
freetolgaOk i will read thanks#asahi · 2026-01-08 08:44
freetolgaI guess resigning needed so that it only boots on your machinr?#asahi · 2026-01-08 08:44
chaos_princessi would _highly_ recommend reading through various pages here https://asahilinux.org/docs/#asahi · 2026-01-08 08:44
chadmedto even add a kext to the kernelcache you need to rebuild the kernelcache and let SEP recompute the hash of the new boot object#asahi · 2026-01-08 08:43
chadmedapple wants third parties out of kernelspace so pluggable kexts are gone#asahi · 2026-01-08 08:43
chaos_princesswhen you start macos you load a blob consisting of kernel and kexts glued together#asahi · 2026-01-08 08:42
chaos_princess"load kexts" is not really something that makes sense on modern macos, as kexts are already in the kernelcache#asahi · 2026-01-08 08:42
freetolgaor you mean macos partially working#asahi · 2026-01-08 08:42
freetolgaby boot you mean not instantly crash?#asahi · 2026-01-08 08:41
chaos_princessno, the goal is to boot it#asahi · 2026-01-08 08:41
freetolgakexts from m4#asahi · 2026-01-08 08:41
freetolgaso the goal is to make vmapple load kexts#asahi · 2026-01-08 08:41
chaos_princesswell, vGIC is technically in drivers, but whatever, you get the idea#asahi · 2026-01-08 08:41
chadmedthe vmapple kernel is designed to work with hypervisor.framework so its going to do things slightly differently like use vGIC instead of AIC#asahi · 2026-01-08 08:40
chaos_princessthat is why the vmapple kernel exists, it uses generic interrupt controller, doesn't do sptm, doesnt do apple page table format and so on#asahi · 2026-01-08 08:40
chaos_princessit emulates "generic arm"#asahi · 2026-01-08 08:40
chaos_princessno#asahi · 2026-01-08 08:40
freetolgaOk but utm at the end can emulate macos arm#asahi · 2026-01-08 08:39
chaos_princessyou can't pass enough devices to utm#asahi · 2026-01-08 08:39
chaos_princess(i did briefly try booting a vmapple kernel on m1, it even did something, but then m1n1 got very confused once xnu turned on paging)#asahi · 2026-01-08 08:39
freetolgai got an idea if we can pass through devices to emulate macos in utm cant we test vmapple with that easily?#asahi · 2026-01-08 08:39
chadmedwell yes at some point praxis is required, its just that all of us are preoccupied with other things like upstreaming and m3 and making display output suck a bit less (impossible challenge) #asahi · 2026-01-08 08:38
freetolgaLemme share some gdb backtraces firdt#asahi · 2026-01-08 08:38
freetolgaIm not home right now but i dont think we can theorize only#asahi · 2026-01-08 08:38
chaos_princessand you will have a problem with the gpu as the gpu iommu driver is in the kernel, not drivers, but i _think_ i have a way around that too#asahi · 2026-01-08 08:37
chadmedoh yeah thats it#asahi · 2026-01-08 08:37
chaos_princesschadmed: no, you need to emulate some xnu hypercalls, but those are almost all are pac-related #asahi · 2026-01-08 08:37
freetolgaif i have a debugger its not impossible#asahi · 2026-01-08 08:37
freetolgaok i can try maybe#asahi · 2026-01-08 08:37
chadmedif you have teh kdk#asahi · 2026-01-08 08:37
chadmedno its just in the preboot volume with all the other kernels #asahi · 2026-01-08 08:37
chaos_princessit comes with the kdk#asahi · 2026-01-08 08:36
freetolgaand vmapple is compiled by me right#asahi · 2026-01-08 08:36
chadmedbut i thought we needed to do some crazy live patching thing for that to work #asahi · 2026-01-08 08:36
chaos_princessno, you just run kmutil and instead of feeding the t6030 kernel to it, do the vmapple one#asahi · 2026-01-08 08:36
chadmedthe kernel image and kexts get bundled into an im4p so you can theoretically use the vmapple kernel binary and bundle it with the bare metal m4 kexts#asahi · 2026-01-08 08:36
freetolgathat doesnt cry without sptm#asahi · 2026-01-08 08:36
freetolgaHmmmmmm we just need a kext loader right#asahi · 2026-01-08 08:35
chaos_princessalso, another crazy sptm idea - build a frankenkernel out of vmapple kernel and m4 drivers and load that#asahi · 2026-01-08 08:35
freetolgaThen i write some dts and see how far i can go#asahi · 2026-01-08 08:35
chadmedi may have a bad habit of smushing my main macos install with m1n1 instead of installing secondary macoses :P#asahi · 2026-01-08 08:35
freetolgaSo i get m1n1 running first#asahi · 2026-01-08 08:34
chaos_princessok#asahi · 2026-01-08 08:34
chadmednah -f takes you back to fully trusted #asahi · 2026-01-08 08:34
chaos_princessisn't it "-np"?#asahi · 2026-01-08 08:34
chadmedyou cant really permanently brick the thing just by changing the boot object#asahi · 2026-01-08 08:33
chadmedif m1n1 doesnt work just reboot into 1TR and do bputil -f and it will switch back to the signed kernelcache #asahi · 2026-01-08 08:33
freetolgaoh perfect#asahi · 2026-01-08 08:33
chaos_princessyou will not even be able to power on the nvme controller