#asahi-dev · 2026-06-08 09:26
StanfordAppliedCyber[m]because i see this "AppleSEPLoggerService"#asahi-dev · 2026-06-08 09:25
svenand there’s ~nothing known about its peripherals #asahi-dev · 2026-06-08 09:25
chaos_princessnot to us#asahi-dev · 2026-06-08 09:25
svensep also has its keys hidden in hardware engines#asahi-dev · 2026-06-08 09:24
StanfordAppliedCyber[m]is there any sort of log sep gives us?#asahi-dev · 2026-06-08 09:24
StanfordAppliedCyber[m]good to know#asahi-dev · 2026-06-08 09:22
chaos_princessbut yes, they have to be in sync#asahi-dev · 2026-06-08 09:22
chaos_princessopposite actually i think, gigalocker is the bulk storage, secure element is anti-replay#asahi-dev · 2026-06-08 09:22
chadmedsep stores a bunch of stuff in a partition on the ssd which it uses for anti replay tokens. if you fake sep but not this you will explode a lot of things and you will need to factory restore the machine #asahi-dev · 2026-06-08 09:22
chaos_princesshas a single file, {uuid}.gl - that is sep external storage#asahi-dev · 2026-06-08 09:21
chaos_princessthere is a partition on isc preboot, usually mounted at /System/Volumes/xarts#asahi-dev · 2026-06-08 09:21
StanfordAppliedCyber[m]for context i started looking at macs less than a month ago --> I don't know much about these devices#asahi-dev · 2026-06-08 09:20
chaos_princesssep "memory" is split between two locations - secure element and a file on disk#asahi-dev · 2026-06-08 09:20
StanfordAppliedCyber[m]whats gigalocker?#asahi-dev · 2026-06-08 09:20
chaos_princessmacos will write to gigalocker, desyncing it from secure element#asahi-dev · 2026-06-08 09:20
chaos_princessStanfordAppliedCyber[m]: no, no, absolutely not. #asahi-dev · 2026-06-08 09:19
StanfordAppliedCyber[m]obviously this isn't a serious proposal#asahi-dev · 2026-06-08 09:19
nickchani can't tell if hacr is accidential#asahi-dev · 2026-06-08 09:19
StanfordAppliedCyber[m]chaos_princess: yeah but i just wouldn't load the sep firmware and then just try to emulate the mmio stuff-- like, we see the mmio write, we read it, emulate the internal sep stuff, then write the right outbox address#asahi-dev · 2026-06-08 09:19
chadmedbut tso, fiq thing, hypervisor control regs, etc are all most likely accidental#asahi-dev · 2026-06-08 09:19
nickchanbut it has kexts that provided functionailites that is otherwise provided by sep kexts#asahi-dev · 2026-06-08 09:18
chadmedsprr/gxf/txm/sptm/whatever will never ever ever be exposed to raw boot mode because the whole point was to lock these things out #asahi-dev · 2026-06-08 09:18
nickchanvirtualmac2,1 doesn't have virtual sep#asahi-dev · 2026-06-08 09:18
Sobek[m](Fair, it's not needed for linux, only for the RE Hypervisor)#asahi-dev · 2026-06-08 09:18
chadmedall we know is that apple basically chose for us to do their testing for them, so if anything is broken that we think shouldnt be we file bug reports for it#asahi-dev · 2026-06-08 09:18
chaos_princesssep has a secure element with anti-replay counters, if you desync those - i think that is a trip to dfu-city#asahi-dev · 2026-06-08 09:17
chadmedSobek[m]: maybe, not worth thinking about right now and not really important either way#asahi-dev · 2026-06-08 09:17
chaos_princessdon't do that#asahi-dev · 2026-06-08 09:17
StanfordAppliedCyber[m](i'm not actually doing this)#asahi-dev · 2026-06-08 09:17
StanfordAppliedCyber[m]might as well add sep to the mix :)#asahi-dev · 2026-06-08 09:17
chaos_princessSobek[m]: the chance is non zero#asahi-dev · 2026-06-08 09:17
StanfordAppliedCyber[m]i'm emulating all of sptm#asahi-dev · 2026-06-08 09:17
StanfordAppliedCyber[m]but it would be done in the hypervisor#asahi-dev · 2026-06-08 09:17
StanfordAppliedCyber[m]this was mostly a joke#asahi-dev · 2026-06-08 09:17
chaos_princessisn't that a different kext#asahi-dev · 2026-06-08 09:16
StanfordAppliedCyber[m]maybe i could attach a virtual SEP :)#asahi-dev · 2026-06-08 09:16
chaos_princessvms have virtual sep#asahi-dev · 2026-06-08 09:16
chaos_princessyou need it to be able to log in#asahi-dev · 2026-06-08 09:16
StanfordAppliedCyber[m]like, we're trying to convince userspace this is a VM anyways#asahi-dev · 2026-06-08 09:16
chaos_princessno, can't do that#asahi-dev · 2026-06-08 09:16
chaos_princessit should be fine#asahi-dev · 2026-06-08 09:16
StanfordAppliedCyber[m]it may be worth just disabling sep broadcase entirely#asahi-dev · 2026-06-08 09:16
Sobek[m]chaos_princess: Do you think there's a world were newer Apple CPUs will use that thing instead of custom Apple stuff ?#asahi-dev · 2026-06-08 09:16
StanfordAppliedCyber[m]i feel like that would just crash another codepath#asahi-dev · 2026-06-08 09:16
StanfordAppliedCyber[m]i did indeed have that idea#asahi-dev · 2026-06-08 09:15
chaos_princessStanfordAppliedCyber[m]: another bad idea, edit the discovery endpoint table, and kill the stac endpoint#asahi-dev · 2026-06-08 09:15
chaos_princessSobek[m]: if it makes you feel better, there is some new thing coming down the aarch64 extension pipeline that seems very gxf inspired#asahi-dev · 2026-06-08 09:15
StanfordAppliedCyber[m]the firmware is still clearly alive in there#asahi-dev · 2026-06-08 09:14
StanfordAppliedCyber[m]but you can then send known good commands to the same endpoint and it still responds with the right values#asahi-dev · 2026-06-08 09:14
StanfordAppliedCyber[m]yeah i can confirm we issue the known failing command --> nothing happens