#
✕ nick:Sobek[m]
#asahi · 2022-05-20 17:52
Sobek[m](Is there anything remotely like SEP that has linux drivers btw or will that likely be a significant amount of work simply becuase such component don't have any subsystem in the kernel and interface ?)
#asahi · 2022-05-20 17:50
Sobek[m]How SEP requires APFS support exactly ?
#asahi · 2022-05-20 17:49
Sobek[m]Random thought, is it technically possible to write drivers to support Touch ID under linux ? What are the moving parts required ?
#asahi · 2022-05-20 12:18
Sobek[m]Also didn't marcan ask people to actually provide feedback about which keyboard layout matched hardware best and what bugs there were ?
#asahi · 2022-05-17 13:49
Sobek[m]Just updated my macOS. I should test the asahi reboot at some point (and perhaps run some update in there)
#asahi · 2022-05-16 11:31
Sobek[m](Well, you could use macro black magic to write code naturally and have it re-written at compile time to add those parameters, however that would be pretty heavy black magic)
#asahi · 2022-05-16 11:02
Sobek[m]Now developping a plugin to track all these things requires some R&D work, as would a type system solution, but it could probably fit within the rust language
#asahi · 2022-05-16 10:59
Sobek[m]There isn't an rfc for that apparently. Rust compiler plugins are a thing on nightly though. The API isn't quite stable but it does exist.
#asahi · 2022-05-16 10:54
Sobek[m]j`ey: What's exactly the no_panic you are referring to ? I've seen a crate that exploits the linker to make the compilation fail if certain functions cannot be proven not to panic.
#asahi · 2022-05-16 10:48
Sobek[m](Then you could annotate functions to indicate their locking expectations and let the compiler shout at you until your annotation and the code match)
#asahi · 2022-05-16 10:46
Sobek[m]Another way one could try out something like auditing logs and friends could be using a compiler plugin implementing a lint https://doc.rust-lang.org/1.5.0/book/compiler-plugins.html (And add flags in your config to make the lint warnings into compile errors)
#asahi · 2022-05-16 10:42
Sobek[m]I don't know the state of the art in kernel development / system programming research but I feel like what marcan is looking for probably is kind of close to the research bleeding edge.
#asahi · 2022-05-16 10:40
Sobek[m]Yep, that's the drawback at that point, perhaps macros could help though >:) Rust macros are powerful
#asahi · 2022-05-16 10:39
Sobek[m]Also, it might be worth looking at what the redow folks did btw.
#asahi · 2022-05-16 10:38
Sobek[m]I'd have to spend time trying out writing these sort of stuff to see if it works out and how heavy the syntax may be.
#asahi · 2022-05-16 10:38
Sobek[m]the context, where those context encode the interesting info in their types. The interrupt disabling / enabling routine would also use these and require a context type that's incompatible with the one returned by locks.
#asahi · 2022-05-16 10:38
Sobek[m]marcan: I saw some of your can rust be used to prevent people from writing code that may dedalock (especially around the interupt disabling logic). Rust isn't design to solve that problem, but I wonder if one could do some things using fancy type system stuff and using custom lock that use these. When you acquire a lock you have to give it the current locking context, and it returns a new context, and when you release the lock it gves you back
#asahi · 2022-05-14 14:31
Sobek[m](Well, not the m1n1 that boot Asahi Linux, but the hypervisor logic that is used for reverse engineering)
#asahi · 2022-05-14 14:30
Sobek[m]Are you aware most of m1n1 is written in python ?
#asahi · 2022-05-11 21:09
Sobek[m]Have they ?
#asahi · 2022-05-10 06:11
Sobek[m]Isn't there also box64 ?
#asahi · 2022-05-06 19:02
Sobek[m]Is scheduling on asymmetric core an active research area ?
#asahi · 2022-05-06 18:16
Sobek[m]+E
#asahi · 2022-05-06 18:16
Sobek[m]But now it has a nice room icon matrix sid
#asahi · 2022-04-30 20:12
Sobek[m]I'd be curious if anyone has the lkml link fro the NVME v3 to see what changed.
#asahi · 2022-04-30 20:11
Sobek[m]I was more mentioning the lack of a single message in say a day (which is unusual given the traffic teh channel had all week long). I wonder if this could be a IRC - matrix bridge or federation bug.
#asahi · 2022-04-30 18:50
Sobek[m]Things are suprisingly silent here
#asahi · 2022-04-29 18:22
Sobek[m]It look like the branch for A15 support and M2 support was like mixed up in a way it shouldn't have. Perhaps next time round they'll do stuff in a way such that they can release A16 first and then push the M3-A16 diff later ?
#asahi · 2022-04-29 18:20
Sobek[m]M1 Max / Ultra included or some M1Pro/Max/Ultra bits are missing too ?
#asahi · 2022-04-29 18:19
Sobek[m]Or even apple LLVM doesn't have support for A15 ?
#asahi · 2022-04-29 18:19
Sobek[m]wait, that's in spite of A15 being out and a supported target for Apple LLVM ?
#asahi · 2022-04-29 18:18
Sobek[m]A15 got released in to teh general public in september 2021, so roughly a year before it got out, they were updating compilers to support the target. probably a point where engineering samples for A15 might have been used for prototypes and testing stuff ?
#asahi · 2022-04-29 18:16
Sobek[m]Well yes, but them lurking hear doesn't necessarily mean they snitched and got that commit removed.
#asahi · 2022-04-29 18:16
Sobek[m]Perhaps one apple person lurking here saw it and got it yanked within an hour. Does it seem that bad ? Perhaps apple has its opsec team keeping an eye on us ?
#asahi · 2022-04-29 18:13
Sobek[m]I wonder if a. You were very lucky to be in a very short time frame where this think had been mistakenly pushed and yanked or b. The thing got pulled just after you saw it by sheer unluck after being there for a while ?
#asahi · 2022-04-29 18:07
Sobek[m]Ah. Hmm but why H = A-1 ? Did they start with H0 ? Or did those code name start at another point ?
#asahi · 2022-04-29 18:06
Sobek[m]Staten is the code name for M2 isn't it ?
#asahi · 2022-04-29 18:06
Sobek[m]H14 ? What's that ?
#asahi · 2022-04-29 17:56
Sobek[m]Has anyone managed to archive it ?
#asahi · 2022-04-29 12:29
Sobek[m]Same way you got here, but with -offtopic added to the name of the channel right after #asahi ;)
#asahi · 2022-04-29 12:26
Sobek[m]mps: Usually there are way too many ways we humans fail at opsec, that they don't need to actually go break the boot security to get where they want, but at least that's going to make them work more than stupid Lenovo rootkit certificates etc.
#asahi · 2022-04-29 12:24
Sobek[m]<m1n1m4[m]> "in the wiki?" <- Yes in the Wiki, where he describes the Apple Silicon boot process.
#asahi · 2022-04-29 12:24
Sobek[m]Well, if they hit me with a wrench, I know I've been compromised at least
#asahi · 2022-04-29 12:23
Sobek[m]Security and usability have always been kind of at odds. Yet apple made something where the proper owner (and only them), can pretty much boot any code they want. It requires them to do a few things to get there, but it is completely doable, that's what marcan has been doing here actually. m1n1 is a big example of running code kind of unfettered.
#asahi · 2022-04-29 12:21
Sobek[m]password.
#asahi · 2022-04-29 12:21
Sobek[m]Your usual evil maid from the SDECE or whatever secret service is after you won't be able to install a rootkit simply with physical access. Managing boot security actually requires both physical access (physical button press to get in 1TR), and an owner password. I haven't done any external boot of standard macOS stuff so I don't known if that can be set-up without password, but anything that involves downgrading security will require a
#asahi · 2022-04-29 12:17
Sobek[m]The whole design is pretty much with security in mind. See what marcan said about its security
#asahi · 2022-04-29 12:15
Sobek[m]drive, having the second stages be on the drive two doesn't change much the equation, and allows configs to have their integrity checked pretty easily, and the first iBoot stage is simpler too.
#asahi · 2022-04-29 12:15
Sobek[m]The lack of straight from external device boot make the bootloader code extremely simpler. For what it is worth, a mac mini / studio whose internal drive has been wiped is only able to communicate with the external world using the led colour and blink (Amber + morse SOS blinking), and getting a recovery software blob (signed by apple) through USB recovery. The attack surface is quite small. And once the bootloader first stage is on the internal
#asahi · 2022-04-29 12:05
Sobek[m]m1n1m4[m]: Not only did they not lock the bootloader, they actually designed it so that something like asahi be possible (extra efforts), and even added small features that help asahi (raw image after making breaking change to the Mach-O boot protocol, bless support, etc).

plus ›