#
✕ nick:marcan
#asahi · 2021-01-28 16:46
marcanI don't even think I ever had to go through a v2 for any of those, unless it was something I missed myself
#asahi · 2021-01-28 16:46
marcancorollary: none of the kernel code I've upstreamed took longer to write than to upstream, but that's because it was patches to existing drivers, so there was no real way to "do it wrong" in a way that wouldn't be immediately accepted
#asahi · 2021-01-28 16:46
marcanit's much more efficient to involve upstream in the process from the start than to fix stuff later
#asahi · 2021-01-28 16:45
marcannever_released: it depends on whether you write for upstreamability, or not
#asahi · 2021-01-28 15:49
marcanI'll make it work like `cat m1n1.macho kernel > m1n1-kernel.macho` to boot a built-in kernel a la corellium (for people without serial who are desperate to try stuff)
#asahi · 2021-01-28 15:49
marcanpretty much ready to throw in a proper Linux boot routine that deals with memory maps, initramfs, SMP spintables, etc properly
#asahi · 2021-01-28 15:48
marcanpushed a bunch of m1n1 stuff to spin up smp, and also start using the ADT properly
#asahi · 2021-01-28 11:37
marcannever_released: I know, I've just been using static devicetrees for testing
#asahi · 2021-01-28 06:42
marcanmaximus64: haven't done linux yet, shaving the yak of ADT->FDT stuff because I can't do the spin-table without actual dynamism in the fdt any more
#asahi · 2021-01-28 06:15
marcanhttps://mrcn.st/p/7C78xHgs love it when things work first try :)
#asahi · 2021-01-27 16:43
marcanyeah I'm setting them before the mmu turns on
#asahi · 2021-01-27 16:00
marcanj`ey: I don't think they should be
#asahi · 2021-01-27 15:59
marcanI'm looking at the source
#asahi · 2021-01-27 15:59
marcandavidrysk[m]: their preloader?
#asahi · 2021-01-27 15:59
marcanit is at *best* a register trace, at worst based on decompilation
#asahi · 2021-01-27 15:59
marcanif they told you they did that, they lied
#asahi · 2021-01-27 15:58
marcanthe order of operations is blatantly the macos order
#asahi · 2021-01-27 15:58
marcannever_released: it is absolutely not doing that
#asahi · 2021-01-27 15:58
marcandavidrysk[m]: where? I don't see bit 4
#asahi · 2021-01-27 12:37
marcanthat's mediawiki, but similar type docs
#asahi · 2021-01-27 12:37
marcanjn__: you can look at https://wiibrew.org/wiki/Hardware for inspiration
#asahi · 2021-01-27 12:37
marcanjn__: yeah there should be a hardware index page
#asahi · 2021-01-27 12:37
marcanit's an IOMMU
#asahi · 2021-01-27 12:36
marcanterm comes from ppc-era macs
#asahi · 2021-01-27 12:36
marcanDevice Address Resolution Table IIRC
#asahi · 2021-01-27 10:35
marcanbut more like a sensible list of handy pages or entry points
#asahi · 2021-01-27 10:35
marcanjust making it a page list wouldn't be useful of course, github already does that for us
#asahi · 2021-01-27 10:35
marcanfeel free to propose/just do something about it
#asahi · 2021-01-27 10:34
marcanjn__: no particular policy, more that the sidebar should contain useful stuff, and then maybe some index pages with more info?
#asahi · 2021-01-27 09:17
marcanand they could've known this by cross-referencing with public xnu code like I did :)
#asahi · 2021-01-27 09:17
marcanwhich again, is completely unnecessary unless you use apple proprietary pointer auth features
#asahi · 2021-01-27 09:17
marcanthey also copied pointer auth init stuff
#asahi · 2021-01-27 09:16
marcanbut you wouldn't know it from the corellium file because they just copied a larger chunk of the startup sequence than just the chicken bits routine
#asahi · 2021-01-27 09:16
marcanthat part of the code isn't even in the chicken bits section of macOS
#asahi · 2021-01-27 09:16
marcanthose don't look like chicken bits, they look like a proprietary CPU feature register containing a table of access permissions or similar
#asahi · 2021-01-27 09:15
marcan(if apple doesn't open source it in xnu first that is)
#asahi · 2021-01-27 09:15
marcanI'm not going to throw it in until I can prove it fixes a problem, and if it does I'll reverse engineer exactly what those bits do and document it
#asahi · 2021-01-27 09:15
marcanthat 0x2020a505f020f0f0 business in particular stinks of being related to APRR or something related
#asahi · 2021-01-27 09:07
marcanI'm 99% sure parts of that are just copyright violations of the macOS startup sequence that have nothing to do with fixing CPU bugs
#asahi · 2021-01-27 09:06
marcanI would never ship something like what corellium shipped there :)
#asahi · 2021-01-27 09:06
marcanI spent 5 hours on that last stream, but now I can say my code *knows* what it does, does *less* than the corellium code, and what it does fixes real bugs and issues with the chips
#asahi · 2021-01-27 09:06
marcanvs https://github.com/AsahiLinux/m1n1/blob/main/src/chickens.c#L66
#asahi · 2021-01-27 09:06
marcanand so, compare: https://github.com/corellium/preloader-m1/blob/main/preboot.S#L104
#asahi · 2021-01-27 09:05
marcanbecause if you include unnecessary stuff, then that's grounds for losing on any copyright defense
#asahi · 2021-01-27 09:05
marcanbut then, you need to at least attempt to understand it so that you can *prove* what is necessary and what is not
#asahi · 2021-01-27 09:05
marcansometimes you *have* to do binary RE, or do certain things one way because otherwise the hardware won't run
#asahi · 2021-01-27 09:03
marcan*these
#asahi · 2021-01-27 09:03
marcanbut there's a reason I wrote alx.sh/re
#asahi · 2021-01-27 09:03
marcandifferent projects might care more/less about this things
#asahi · 2021-01-27 09:03
marcanbut it seems things got really bad with DDR4 and especially LPDDR4

plus ›