#asahi-dev · 2021-03-04 13:52
marcanand then the exception code has some magic to forward exceptions from EL1 to EL2 when doing one-off EL1 function calls (since there is no way to do that with a setting, globally - that's what TGE is for but that only works for EL0)#asahi-dev · 2021-03-04 13:51
marcanI changed all the memory and exception code to use _EL1 registers, since those are redirected to _EL2 with VHE anyway#asahi-dev · 2021-03-04 13:50
marcanit's not a hypervisor, there is no stage2 translation, but it's good for seeing what's different in EL1#asahi-dev · 2021-03-04 13:50
marcananyway, pushed some stuff to call functions at EL1, and also you can chainload.py --el1 now#asahi-dev · 2021-03-04 13:49
marcanyeah...#asahi-dev · 2021-03-04 13:49
marcanit's not hard to have a better explanation than what's in the arm manual...#asahi-dev · 2021-03-04 13:48
marcanand then add 'is' to make it global across all CPUs#asahi-dev · 2021-03-04 13:48
marcanbut since we have VHE, that gets redefined to do the same thing as alle2 in EL2#asahi-dev · 2021-03-04 13:47
marcanvmalle1 is the "normal" invalidate everything version. in a VM (EL1), it invalidates everything in that VM#asahi-dev · 2021-03-04 13:47
marcanALLE* never works from EL1#asahi-dev · 2021-03-04 13:46
marcanyou then changed it to alle2, which is fine, though CPU-local, but only works from EL2#asahi-dev · 2021-03-04 13:45
marcanyou previously had vmalls12e1is, which is for invalidating *VM* TLBs - EL1&0 regime only, VMID match (no special behavior with VHE). that's why that one didn't work.#asahi-dev · 2021-03-04 13:45
marcanyes :D#asahi-dev · 2021-03-04 13:41
marcansven: so it turns out `tlbi vmalle1is` is what we want :-)#asahi-dev · 2021-02-20 09:30
marcanmore than one even, because this wasn't it: https://github.com/osresearch/rwmem#asahi-dev · 2021-02-20 09:29
marcanand promptly forgot to wtrite down#asahi-dev · 2021-02-20 09:29
marcanthere was one thing I found at some point#asahi-dev · 2021-02-19 18:42
marcannot sure if anything implements it properly though :)#asahi-dev · 2021-02-19 18:42
marcanmodwizcode: there is some stuff, yeah#asahi-dev · 2021-02-19 18:40
marcanI would not expect USB to be passed through#asahi-dev · 2021-02-19 18:40
marcan... like graphics/display#asahi-dev · 2021-02-19 18:40
marcanyeah, so maybe just things you might reasonably want to pass to a guest#asahi-dev · 2021-02-19 18:40
marcanthat would make sense if it really is a thing meant for VM guests#asahi-dev · 2021-02-19 18:40
marcanmaybe there's a lock/enable in the DART side?#asahi-dev · 2021-02-19 18:39
marcanmuch like this thing has a vGIC for VMs...#asahi-dev · 2021-02-19 18:38
marcansven: shot in the dark: something something virtualization?#asahi-dev · 2021-02-19 17:29
marcanare either of them already configured by the time we run?#asahi-dev · 2021-02-19 17:29
marcanlike I think most xHCI controllers run firmware and stuff, so maybe there's a carve-out for that#asahi-dev · 2021-02-19 17:29
marcansven: USB3 is... complicated, maybe there are separate DARTs for e.g. host/device mode, or for data and controller shenanigans, or stuff like that?#asahi-dev · 2021-02-18 07:09
marcanit calls that "VRAM" somewhere IIRC#asahi-dev · 2021-02-18 04:31
marcanbut it's consistent between boots and such#asahi-dev · 2021-02-18 04:31
marcanIIRC it changed with an iBoot update#asahi-dev · 2021-02-18 04:31
marcanfwiw I think it's dynamic to some extent on real HW#asahi-dev · 2021-02-18 00:30
marcanI have the same exact address?#asahi-dev · 2021-02-18 00:30
marcanFB at: 0x9e0df8000#asahi-dev · 2021-02-17 17:21
marcanmodwizcode: distribution of any kind triggers the GPL rules, and putting something up on github absolutely counts as distribution#asahi-dev · 2021-02-17 11:00
marcanlike wtf#asahi-dev · 2021-02-17 11:00
marcanI'm sorry, I can't trust someone who plasters *decompiler output* all over their repo in a *target repo*#asahi-dev · 2021-02-17 11:00
marcanhttps://github.com/TrungNguyen1909/qemu-t8030/blob/master/include/hw/intc/apple-aic.h#asahi-dev · 2021-02-17 11:00
marcanthere is hex-rays output in that file#asahi-dev · 2021-02-17 10:59
marcanalso, uh#asahi-dev · 2021-02-17 10:59
marcanI need to test those extra registers again though#asahi-dev · 2021-02-17 10:57
marcanwhoever wrote that QEMU code copied the register defines from the corellium patchset#asahi-dev · 2021-02-17 10:56
marcanthat's corellium terminology, and i don't agree with it#asahi-dev · 2021-02-17 10:56
marcanmodwizcode: that "defer" doesn't mean what I called "deferred IPIs" for non-AIC#asahi-dev · 2021-02-16 07:33
marcancool!#asahi-dev · 2021-02-16 07:29
marcanI'll put up a scheduled stream a bit later#asahi-dev · 2021-02-16 07:29
marcanI'm going to (finally) start on the usb-pd/serial interface design tomorrow, if anyone wants to join on the stream#asahi-dev · 2021-02-16 07:23
marcannice :D#asahi-dev · 2021-02-15 19:29
marcanand supposedly you can provision secureboot yourself for the latter, i.e. make it only run code you sign