#
#asahi-re · 2026-05-10 07:00
StanfordAppliedCyber[m]re: txm, right now basically everything there is stubbed to return txm_success
#asahi-re · 2026-05-10 06:55
JamesCalligeros[m]It’s SME v0.0.00001 pre release early access alpha :p
#asahi-re · 2026-05-10 06:50
StanfordAppliedCyber[m]ah just read about it, extra instructions yeah
#asahi-re · 2026-05-10 06:49
StanfordAppliedCyber[m]apologies, i'm not super familiar with apple platforms
#asahi-re · 2026-05-10 06:49
StanfordAppliedCyber[m]like is it just extra instructions?
#asahi-re · 2026-05-10 06:48
StanfordAppliedCyber[m]how does amx work
#asahi-re · 2026-05-10 06:48
StanfordAppliedCyber[m]interesting, I'll look into that
#asahi-re · 2026-05-10 06:48
StanfordAppliedCyber[m]the annoying part will be is if they have arguments-- ie genter is trivial search code for exact hex, replace with other exact hex for hvc
#asahi-re · 2026-05-10 06:48
svenuserspace is also going to hit AMX but you may be able to disable that
#asahi-re · 2026-05-10 06:48
svenyeah, which is fine because code segments are rx and not w anyway
#asahi-re · 2026-05-10 06:47
StanfordAppliedCyber[m]and triggers some macos exception handler, which obviously won't work, but cross that bridge when we get there
#asahi-re · 2026-05-10 06:47
StanfordAppliedCyber[m]the odds are those instructions will have to be mechanically pached out also because otherwise the illegal instruction traps to el1
#asahi-re · 2026-05-10 06:46
StanfordAppliedCyber[m]bunch of 0xaaaaaaa pac keys and ignore all the registers
#asahi-re · 2026-05-10 06:46
StanfordAppliedCyber[m]that's what we're already doing
#asahi-re · 2026-05-10 06:46
svenalso lots of sysregs for stuff like apple pac but most of that can be ignored I think
#asahi-re · 2026-05-10 06:46
StanfordAppliedCyber[m]ack, yeah i remember reading about those from one of marcan's old blog posts
#asahi-re · 2026-05-10 06:46
StanfordAppliedCyber[m]catching the first genter took forever
#asahi-re · 2026-05-10 06:45
StanfordAppliedCyber[m]my progress was actually really slow to start, because this whole thing is very sensitive, to put it mildly
#asahi-re · 2026-05-10 06:45
svenI know about some compress/uncompress instructions
#asahi-re · 2026-05-10 06:45
svenYeah, instructions
#asahi-re · 2026-05-10 06:44
svenim not, there’s always fast progress at the beginning and then the second 80% take as much time as the first one :D
#asahi-re · 2026-05-10 06:44
StanfordAppliedCyber[m]like you don't mean registers, you mean arm instructions
#asahi-re · 2026-05-10 06:44
StanfordAppliedCyber[m]what are some of the instructions that you know of?
#asahi-re · 2026-05-10 06:44
svenand getting that instruction emulation upstream would be good anyway
#asahi-re · 2026-05-10 06:44
StanfordAppliedCyber[m]i'm surprised any of this shit works at all
#asahi-re · 2026-05-10 06:43
StanfordAppliedCyber[m]for context i started working on this 3 days ago
#asahi-re · 2026-05-10 06:43
StanfordAppliedCyber[m]yeah i'm expecting problems
#asahi-re · 2026-05-10 06:43
svenit’s gonna happen and emulating those is gonna be useful either way
#asahi-re · 2026-05-10 06:43
svenright, probably cause you’re not even in bsd init
#asahi-re · 2026-05-10 06:43
StanfordAppliedCyber[m]only GENTER which was mechanically patched out
#asahi-re · 2026-05-10 06:42
StanfordAppliedCyber[m]no actually, haven't hit that at all
#asahi-re · 2026-05-10 06:42
svenand did you run into other undef instructions? because a lot of Apple ISAs are disabled and I don’t know which are used by xnu
#asahi-re · 2026-05-10 06:42
StanfordAppliedCyber[m]not even in bsd init
#asahi-re · 2026-05-10 06:42
StanfordAppliedCyber[m]that's my first goal
#asahi-re · 2026-05-10 06:42
StanfordAppliedCyber[m]absolutely not in userspace yet
#asahi-re · 2026-05-10 06:42
svenhow far does xnu boot? do you reach userspace?
#asahi-re · 2026-05-10 06:41
StanfordAppliedCyber[m]i'm applying brute force to the problem
#asahi-re · 2026-05-10 06:41
StanfordAppliedCyber[m]yes
#asahi-re · 2026-05-10 06:41
StanfordAppliedCyber[m]iirc that was actually pretty simple
#asahi-re · 2026-05-10 06:41
svenso TXM is just emulated as well?
#asahi-re · 2026-05-10 06:41
StanfordAppliedCyber[m]yes i am, i have 4 endpoints wired up, i'm also tainted there
#asahi-re · 2026-05-10 06:41
svenbecause that might be more complicated and is required by xnu as well
#asahi-re · 2026-05-10 06:41
svenare you dealing with TXM yet?
#asahi-re · 2026-05-10 06:41
StanfordAppliedCyber[m]yeah i agree
#asahi-re · 2026-05-10 06:40
sventhat doesn’t sound too bad, we might still want to just have documentation for those few complicated functions and have someone else reimplement them just to be safe
#asahi-re · 2026-05-10 06:34
StanfordAppliedCyber[m]we have to populate a giant sptm args struct, but this is from the public xnu dump and thus explictly allowed by the RE policy
#asahi-re · 2026-05-10 06:33
StanfordAppliedCyber[m]i think the actual sptm emulation logic is sub 100 lines of C
#asahi-re · 2026-05-10 06:32
StanfordAppliedCyber[m]like looking at this, as a non expert in RE, there really is NOT that much code
#asahi-re · 2026-05-10 06:31
StanfordAppliedCyber[m]there are some endpoints that have real copyrightable code, notably, everything around {UN}MAP_PAGE
#asahi-re · 2026-05-10 06:31
StanfordAppliedCyber[m]i'm sure you've seen the modern ios security paper-- i have somewhere around 10 endpoints that are literally just return SPTM_SUCCESS

plus ›