#asahi · 2026-03-11 16:10
chaos_princessah, no, alx.sh now does too#asahi · 2026-03-11 16:09
chaos_princessuse this installer: https://fedora-asahi-remix.org/ i don't think the one at alx.sh offers you a choice#asahi · 2026-03-11 16:09
jxkc[m]perfect, im going to go with gnome later then, thank you#asahi · 2026-03-11 16:08
chaos_princessyes, but you can pick gnome, or nothing at all (server) and install whatever you want#asahi · 2026-03-11 16:08
jxkc[m]hey there, im thinking of moving my m1 to ashai :). is KDE the default?#asahi · 2026-03-11 16:07
bgtlover[mds]hmm, gonna look into that, I forgot much of how that stuff works#asahi · 2026-03-11 16:05
chaos_princessbgtlover[mds]: you are describing snp-sev#asahi · 2026-03-11 16:05
chaos_princessweirdtreething: m4+ can boot either in "macos mode", or in "linux mode". in "linux mode" we get full access to the cpu, but apple-specific instructions are locked out, and sptm uses them. in "macos mode" apple instructions are available, but we only launch after sptm. so, the idea is to use "linux mode" and emulate the missing instructions#asahi · 2026-03-11 16:03
bgtlover[mds]that would also tie in with object capabilities and stuff like that, but yeah, I'll stop because this is getting really offtopic haha#asahi · 2026-03-11 16:02
bgtlover[mds]I was imagining some kind of MMU that also encrypts the pages, and then the memory ownership question becomes a cryptographic one, a program can only access the memory regions it has a key for, so either something it created, or something it was given by another program, like a shared memory pool. But yeah, that wouldn't really be that useful when it can be done with code instead of this convoluted thing#asahi · 2026-03-11 16:02
weirdtreethingwhat prevents sptm from being virtualized?#asahi · 2026-03-11 15:58
bgtlover[mds]true#asahi · 2026-03-11 15:58
chaos_princessright, but if you don't care about the keys, you can just pretend it does not exist. Or rip it out if it is socketed#asahi · 2026-03-11 15:57
bgtlover[mds]chaos_princess: it prevents you from tampering with it from outside the OS and firmware and such to get the key material, but yeah#asahi · 2026-03-11 15:55
chaos_princessyes. it _just_ stores key material and can operate on it. it does not enforce any policy, or prevent you from doing anything#asahi · 2026-03-11 15:55
bgtlover[mds]chaos_princess: it's a hardware store for key material, no?#asahi · 2026-03-11 15:54
chaos_princessa tpm does something completely different#asahi · 2026-03-11 15:53
chaos_princess"prodding memory addresses it shouldn't" - preventing that is literally what virtualization is about#asahi · 2026-03-11 15:53
chaos_princessall those hypervisors openly advertise their presence via a standardized cpuid leaf. they also expose hypercalls to the vm to make some things faster#asahi · 2026-03-11 15:51
bgtlover[mds]the user doesn't want it to, but I was reaching for the wrong tool I suppose#asahi · 2026-03-11 15:51
bgtlover[mds]chaos_princess: really? hmm weird, then it's some trickery systemd does when I run systemd-detect-virt , so like for virtualbox, it says oracle, same for kvm/qemu. Maybe that can be disabled from qemu and it's there just for debugging? but yeah, I thought the guest is prodding memory addresses it shouldn't or running specific instructions and comparing timing to see if it's in a vm, and maybe memory TPM would prevent it from doing that if#asahi · 2026-03-11 15:46
chaos_princessand a fully stealth vm is a bit of a fool's errand, an os can just attempt to launch a vm of it's own and measure how long it takes. #asahi · 2026-03-11 15:44
chaos_princessnot on arm, you need to enable nested virt for that, but on x86, an os isn't supposed to know whether it is in vm or not#asahi · 2026-03-11 15:43
chaos_princesshow is that relevant at all to the "tpm for memory". normal virtualization already kinda does that#asahi · 2026-03-11 15:43
bgtlover[mds]chaos_princess: I mean, isolation as in making the OS unable to know it's in a vm, while also not emulating the whole cpu, so still using stuff like KVM#asahi · 2026-03-11 15:40
chaos_princessif you want to hide from hypervisor, that is no longer vm isolation, but drm#asahi · 2026-03-11 15:40
chaos_princessvm isolation can be achieved by normal virtualization#asahi · 2026-03-11 15:39
bgtlover[mds]sure, that's one of the uses, but I was thinking more vm isolation#asahi · 2026-03-11 15:38
chaos_princessnot really, all those technologies are only useful for drm#asahi · 2026-03-11 15:38
bgtlover[mds]chaos_princess: I know, I thought apple implemented it on arm, that would have been interesting#asahi · 2026-03-11 15:37
chaos_princessif you want tpm for memory - look into SNP-SEV on amd, or SGX on intel #asahi · 2026-03-11 15:34
bgtlover[mds]ah, sad#asahi · 2026-03-11 15:34
chaos_princessno, it is a piece of code that runs in a separate privilege level#asahi · 2026-03-11 15:34
bgtlover[mds]interesting what apple is doing, but also darn that's annoying as hell#asahi · 2026-03-11 15:33
bgtlover[mds]and that's a piece of hardware, right? if so, that sounds kinda like the TPM but for memory :p#asahi · 2026-03-11 15:33
chaos_princessno, it means that xnu has to call into sptm when it wants to change page tables#asahi · 2026-03-11 15:32
bgtlover[mds]chaos_princess: wouldn't that mean that memory allocation would be limited?#asahi · 2026-03-11 15:31
chaos_princesspretty much that, page tables can no longer be edited by most of xnu, only by sptm, which has a much smaller attack surface#asahi · 2026-03-11 15:30
bgtlover[mds]that sounds like something that makes page tables immutable somehow#asahi · 2026-03-11 15:30
jbechaos_princess: yeah, it definitely looks less banana in the real life photos I see now#asahi · 2026-03-11 15:30
jbebgtlover[mds]: secure page table monitor#asahi · 2026-03-11 15:29
chaos_princessjbe: i don't think that bananas should be that color#asahi · 2026-03-11 15:29
bgtlover[mds]chaos_princess: what's SPTM?#asahi · 2026-03-11 15:29
jbethe banana laptop...#asahi · 2026-03-11 15:26
sveni think so, yeah #asahi · 2026-03-11 15:26
jbethat sounds about right#asahi · 2026-03-11 15:26
chaos_princesssomeone should probably check if kmutil works#asahi · 2026-03-11 15:26
chaos_princessneo should be out today, right?#asahi · 2026-03-11 15:25
chaos_princessthe issue is not about what is different, but rather what is similar. it has the same issue as m4, in that all available macos versions want to run under sptm, and we can't virtualize that yet, meaning our RE tools do not work#asahi · 2026-03-11 15:25
svensimilar to m4 probably with some additional work for usb and wifi/bt maybe