#asahi-re · 2022-10-18 14:04
svenso?#asahi-re · 2022-10-18 14:04
svenand you can just dump the NOR from linux#asahi-re · 2022-10-18 14:04
svenit's encrypted anyway#asahi-re · 2022-10-18 14:03
svenactually wouldn't surprise me if it is#asahi-re · 2022-10-18 14:03
svenand the firmware might actually be inside the NOR flash depending on how early iboot wants to bring SEP up#asahi-re · 2022-10-18 14:01
svenhttps://github.com/AsahiLinux/m1n1/blob/main/proxyclient/m1n1/hw/sep.py dunno if it still works#asahi-re · 2022-10-18 14:00
sventhere's code somewhere in m1n1 to boot it#asahi-re · 2022-10-18 14:00
sveniboot puts the firmware in memory#asahi-re · 2022-10-18 14:00
svenwe probably only care about a subset of what it does anyway. right now I can only think about touchid and secure key store/filevault#asahi-re · 2022-10-18 13:55
sven26: "generic_operation", <-- that call is that other IPC protocol iirc#asahi-re · 2022-10-18 13:54
svenyeah#asahi-re · 2022-10-18 13:54
svenbut there's also at least a secondary IPC on top of the first one#asahi-re · 2022-10-18 13:53
svenI think it's at least a bit more sane though and they shouldn't be able to break the API all the time#asahi-re · 2022-10-18 13:53
svenSEP could well be DCP-level insanity :-P#asahi-re · 2022-10-18 13:47
svenI don't think i'll have enough time to look at them#asahi-re · 2022-10-18 13:43
svenit won't even expose the sks endpoint before it did some xARTs init dance#asahi-re · 2022-10-18 13:43
sventhat shouldn't be very spammy and it's also the first thing required to get SEP up#asahi-re · 2022-10-18 13:43
sveni'd start with xarts anyway#asahi-re · 2022-10-16 15:09
sventhe kernel driver is probably just going to expose a few ioctls to talk to the various endpoints #asahi-re · 2022-10-16 15:08
svenmost of sep will be a userspace daemon fwiw#asahi-re · 2022-10-16 14:32
svenit's been a while#asahi-re · 2022-10-16 14:32
sveni through it was xart, but dunno#asahi-re · 2022-10-16 14:29
svenand it should be relatively simple#asahi-re · 2022-10-16 14:29
sventhat's the one that uses the gigalocker file and if you don't reply to its requests the key store won't even appear#asahi-re · 2022-10-16 14:29
svenat first you'll have to figure out the xart endpoint anyway#asahi-re · 2022-10-16 14:28
sveni'd certainly start there#asahi-re · 2022-10-16 14:24
svenit's already pretty verbose#asahi-re · 2022-10-16 14:23
svennot sure how that would help really#asahi-re · 2022-10-16 14:21
svenit's more or less explained in that apple security whitepaper iirc#asahi-re · 2022-10-16 14:20
svenand SEP will send requests like get(UUID), set(UUID, payload), delete(UUID)#asahi-re · 2022-10-16 14:19
svengigalocker is literally a rather simple key-value store for SEP#asahi-re · 2022-10-16 14:19
svenyeah. the UUID is just the key#asahi-re · 2022-10-16 14:11
svenah, lol, now I remember. I tried to start learning rust with that gigalocker thing :D https://f.svpe.de/b5f8e413e3ae6d2c5ea3d1b5635e6bd329a53ca1bc5c8d208248f396463726e1_gigalocker.rs.txt#asahi-re · 2022-10-16 14:08
svenand the crc is CRC-32/ISO-HDLC#asahi-re · 2022-10-16 14:08
svenamarioguy: ^--#asahi-re · 2022-10-16 14:08
svenand it's essentially a key-value store that SEP updates#asahi-re · 2022-10-16 14:07
sventhe file always contains 170 entries#asahi-re · 2022-10-16 14:07
svenand the last 0x6000 bytes of the file are padding#asahi-re · 2022-10-16 14:07
svenmax payload size is 0x8000 and each header+payload is padded to 0x9000#asahi-re · 2022-10-16 14:07
svener. no#asahi-re · 2022-10-16 14:06
svenmax payload size is 0x9000#asahi-re · 2022-10-16 14:06
sven^-- that should be a single entry in the file#asahi-re · 2022-10-16 14:06
sven0x22 - end: data#asahi-re · 2022-10-16 14:06
sven0x1a - 0x22: unknown#asahi-re · 2022-10-16 14:06
sven0x16 - 0x1a: CRC#asahi-re · 2022-10-16 14:06
sven0x01 - 0x12: UUID#asahi-re · 2022-10-16 14:06
sven0x12 - 0x16: length#asahi-re · 2022-10-16 14:06
sven0x00 - 0x01: always zero#asahi-re · 2022-10-16 14:05
svenhm... can't find the code anymore#asahi-re · 2022-10-16 13:59
svenoh, sure