#asahi-dev · 2022-01-22 05:02
marcanalyssa: re nvme, not a priority but in the mid term I've been leaning towards having m1n1 chainload itself from the ESP as a way to allow upgrades without having to go into 1TR#asahi-dev · 2022-01-21 17:34
marcanmaybe that *is* the max?#asahi-dev · 2022-01-21 17:34
marcanactually I only saw max 3 when we had ~that~ Signal conversation#asahi-dev · 2022-01-21 17:34
marcanclearly we need to go deeper#asahi-dev · 2022-01-21 17:23
marcanthat way we can ship proxyclient to everyone, which can come in handy for debugging#asahi-dev · 2022-01-21 17:22
marcan(once we have nvram support)#asahi-dev · 2022-01-21 17:22
marcanthen -v is a softer gate you can flip from linux#asahi-dev · 2022-01-21 17:22
marcanflipping bootargs restriction requires the same access as installing m1n1, so that is the hard gate#asahi-dev · 2022-01-21 17:22
marcansven: the idea with the proxyclient/hv stuff is that if you have bootargs restriction off and -v in bootargs we'll enable verbose boot and USB proxy (with timeout to normal boot if no enumeration)#asahi-dev · 2022-01-21 17:16
marcanj`ey: we live in a strange world#asahi-dev · 2022-01-21 17:16
marcanj`ey: it isn't actually#asahi-dev · 2022-01-21 17:15
marcanI need to get some sleep but I'll catch up on email tomorrow and send it properly#asahi-dev · 2022-01-21 17:15
marcanpovik: a-b and all that :p#asahi-dev · 2022-01-21 17:15
marcan01:10:42 < alyssa> no i'm sven <- I thought you were sven's girlfriend#asahi-dev · 2022-01-21 13:09
marcanit's part of the dcpep buffer#asahi-dev · 2022-01-21 13:09
marcanIIRC there is also some shared memory with a swap number, not sure if that is used for communication between GFX and DCP#asahi-dev · 2022-01-21 08:00
marcanof course#asahi-dev · 2022-01-21 07:59
marcanwhat's the point once you're running fuOS anyway#asahi-dev · 2022-01-21 07:59
marcanthey really should just give us a bootpolicy knob to turn all the lockdowns off#asahi-dev · 2022-01-21 07:59
marcanyeah, my only reasons are debugging patches and security audits#asahi-dev · 2022-01-21 07:58
marcanthe only thing I want to know is what its physmem map is, and what it can access#asahi-dev · 2022-01-21 07:58
marcantbh though, I don't think there's... actually anything *interesting* to do from DCP? I mean we have access to DISP0 anyway#asahi-dev · 2022-01-21 07:57
marcanand just run stuff on DCP :D#asahi-dev · 2022-01-21 07:57
marcanmake it look like a normal proxyclient object, thank you duck typing#asahi-dev · 2022-01-21 07:57
marcanahahahaha#asahi-dev · 2022-01-21 07:57
marcanyeah, that might be in there#asahi-dev · 2022-01-21 07:56
marcanunless they're in some internal SRAM...#asahi-dev · 2022-01-21 07:56
marcan... it can't be in the CTRR region so...#asahi-dev · 2022-01-21 07:56
marcanha, good point#asahi-dev · 2022-01-21 07:55
marcanno ptrauth though#asahi-dev · 2022-01-21 07:55
marcanyeah but I'm sure it's W^X at least, so you need to ROP it if nothing else#asahi-dev · 2022-01-21 07:54
marcanyeah but like, snapshot it while it's asleep, wake it up, crash it, figure out how to stop the CPU, restore .data, wake it up again#asahi-dev · 2022-01-21 07:54
marcanall we need then is a way to shut down the CPU and start it again as if from hibernation#asahi-dev · 2022-01-21 07:54
marcanand then restore it :)#asahi-dev · 2022-01-21 07:54
marcanwhat if we just snapshot .data#asahi-dev · 2022-01-21 07:54
marcanactually I was thinking about that!#asahi-dev · 2022-01-21 07:53
marcanyou'd have to ROP your way in but...#asahi-dev · 2022-01-21 07:53
marcanyeah#asahi-dev · 2022-01-21 07:53
marcanhah, cute#asahi-dev · 2022-01-21 07:52
marcanthough given that I think all .data was writable? I wouldn't be surprised if that's doable without any security flaws involved#asahi-dev · 2022-01-21 07:51
marcanwould be nice to engineer a DCP takeover just to validate that it doesn't have any backdoors :)#asahi-dev · 2022-01-21 07:51
marcanmaybe I can disclose now :p#asahi-dev · 2022-01-21 07:51
marcanthat reminds me, I wonder if 90 days have passed since that security report...#asahi-dev · 2022-01-21 07:51
marcanright#asahi-dev · 2022-01-21 07:50
marcansanity checks are always good to have#asahi-dev · 2022-01-21 07:50
marcanbut not necessary for the kernel#asahi-dev · 2022-01-21 07:50
marcanthat might be a good sanity check to have in m1n1, just for kicks#asahi-dev · 2022-01-21 07:50
marcannot stuff the bootloader sets up#asahi-dev · 2022-01-21 07:50
marcanas far as I'm concerned the only thing that matters is that stuff coming *from* IOPs is validated#asahi-dev · 2022-01-21 07:50
marcansince we need to trust locked DARTs/DAPFs and such anyway