#asahi · 2021-04-11 13:30
marcansee https://github.com/torvalds/linux/blob/master/Documentation/devicetree/bindings/clock/clock-bindings.txt#asahi · 2021-04-11 13:30
marcanI think that can specify parent relationships within a single clock controller#asahi · 2021-04-11 13:29
marcansven: look at assigned-clock-parents and assigned-clocks#asahi · 2021-04-11 13:25
marcanis any other soc doing what you do here?#asahi · 2021-04-11 13:23
marcanshouldn't it be one clock controller node with clock IDs corresponding to register indices? or perhaps one clock controller with subclock nodes that have reg offsets relative to the parent?#asahi · 2021-04-11 13:22
marcansven: isn't this ioremapping a whole page for every 32-bit clock index? this seems sub-optimal#asahi · 2021-04-10 20:54
marcanthere's a github workflow that should do a clang-format check on all PRs#asahi · 2021-04-10 20:51
marcanI do enforce the clang-format thing, even when it does the "wrong" thing it's better to leave the formatting to a tool than agonize over doing things manually#asahi · 2021-04-10 20:50
marcanthere's no formal coding style other than the rules in that .clang-format file (which I update sometimes) and basically following the style of the rest of the code#asahi · 2021-04-10 20:50
marcanfor C? we have the clang-format thing, just run "make format"#asahi · 2021-04-10 20:49
marcanwhich will allow "proper" in-place chainloads#asahi · 2021-04-10 20:49
marcanbut then again all this ties in with the xnu loading stuff for mmiotrace, I need to just write code that can rebuild the adt/bootargs/stuff anyway#asahi · 2021-04-10 20:49
marcanheh#asahi · 2021-04-10 20:48
marcandue to memory layout#asahi · 2021-04-10 20:48
marcanthough that's only going to work after the first chainload, properly at least#asahi · 2021-04-10 20:48
marcanstill need to make chainload do in-place updates somehow#asahi · 2021-04-10 20:48
marcan(and reboot takes 7 seconds so whatever)#asahi · 2021-04-10 20:48
marcanI always crash it before that happens#asahi · 2021-04-10 20:48
marcanheh#asahi · 2021-04-10 20:47
marcanI've never managed that :D#asahi · 2021-04-10 20:47
marcanamazing#asahi · 2021-04-10 20:47
marcanwow, you actually ran out of RAM?#asahi · 2021-04-10 20:46
marcanyeah, I think it's up there? I know the highest part of RAM faults anyway#asahi · 2021-04-10 20:45
marcanthat's probably going to annoy us when we least expect it#asahi · 2021-04-10 20:45
marcanbecause right now the CPU is allowed to speculate into a carveout there that blows up IIRC#asahi · 2021-04-10 20:44
marcansven: this reminds me, on paper we prooobably need to unmap some high RAM at some point, need to figure out what the sane way of determining what the hole is#asahi · 2021-04-10 20:43
marcanthe display hardware is cache coherent as far as I can tell#asahi · 2021-04-10 20:43
marcanwe map it as normal memory, so there is no difference#asahi · 2021-04-10 20:41
marcan*glue#asahi · 2021-04-10 20:41
marcanwhich is fine as long as the underlying primitives are OK for each bit#asahi · 2021-04-10 20:41
marcanI expect to do some munging over the blue between fb / usb / serial / proxy after this is all merged anyway, it needs a rethink#asahi · 2021-04-10 20:41
marcanwill review the FB stuff tomorrow too, but go over the diff a bit, there's at least one random blank line added and also some of the fix commits could use squashing#asahi · 2021-04-10 20:40
marcanyeah, definitely#asahi · 2021-04-10 20:38
marcan(e.g. your framebuffer scrolling thing is definitely weird enough to get complaints; stuff like this is fine :p)#asahi · 2021-04-10 20:38
marcan(honestly, I don't really care too much about details with m1n1; anything broken we can fix, as long as the basic approach is fine)#asahi · 2021-04-10 20:37
marcanthere's your amazing review :p#asahi · 2021-04-10 20:37
marcanbut also, eh, whatever *click*#asahi · 2021-04-10 20:36
marcanI actually have a draft review on that one fwiw :p#asahi · 2021-04-10 20:35
marcansven: speaking of, what's the state of the USB one?#asahi · 2021-04-10 20:34
marcanthe setup that --el1 does is correct, you just need to enable the second-stage page tables (VM mode) like you would with E2H=0#asahi · 2021-04-10 20:33
marcanbut yeah, you don't care about E2H=1. it just means EL2 can pretend to be EL1.#asahi · 2021-04-10 20:33
marcanI have a fairly well cooked idea in my head, just don't want to end up overlapping and ending up with wasted time :)#asahi · 2021-04-10 20:28
marcanroxfan: come on, I finally did my taxes today, I need to avoid getting deported first :p#asahi · 2021-04-10 20:27
marcankettenis_: merged :p#asahi · 2021-04-10 20:26
marcanI was going to start on monday!#asahi · 2021-04-10 20:24
marcanjeffmiw: you're describing what I've been planning on doing all along :p#asahi · 2021-04-10 19:30
marcanit doesn't actually hypervise anything yet though, making it do that is the next step :)#asahi · 2021-04-10 19:30
marcanand then boot linux under that (only works in single-cpu mode because there's no real virtualization of the SMP spinup stuff, so the other cores come up in EL2 and that breaks)#asahi · 2021-04-10 19:29
marcanjeffmiw: m1n1 can already be a "hypervisor", you can boot m1n1 as a "guest" with chainload.py --el1#asahi · 2021-04-10 19:29
marcanawesome!