#asahi-dev · 2025-03-18 18:17
svenand I guess we’ll have to figure out how to fix the CI :/#asahi-dev · 2025-03-18 18:16
svenit looks like you just dropped that commit instead of squashing it into the first one #asahi-dev · 2025-03-18 13:55
sveni'm also not the only one who can review these things, even getting someone to test it on M1/M2/whatever and confirm nothing broke would help#asahi-dev · 2025-03-18 13:52
svenprobably not today since those probably need a more careful review since they also touch general code#asahi-dev · 2025-03-18 13:50
svenrest looks reasonable#asahi-dev · 2025-03-18 13:50
svenfrom a very brief glance i'd probably squash the last commit into the first one#asahi-dev · 2025-03-18 13:45
sven(the PR, not testing on t6030 ;))#asahi-dev · 2025-03-18 13:45
sveni can take a look at it this evening #asahi-dev · 2025-03-18 13:42
svenyeah, no need to have a perfect pull request. I’m happy as long as it doesn’t break anything and looks like it gets us closer to m3/m4 support #asahi-dev · 2025-03-16 15:58
svenalso much less awkward if all the probing and resource allocation, etc. happens directly inside mfd#asahi-dev · 2025-03-16 15:58
svenokay, this is looking much more manageable without the backend_ops indirection#asahi-dev · 2025-03-16 13:48
svenyet another reason to get this upstream #asahi-dev · 2025-03-16 11:45
sven*cares#asahi-dev · 2025-03-16 11:45
svenjust get the simplest possible driver upstream for now and then we can revisit that "what happens if we need a different way to access SMC" problem later if anyone actually care#asahi-dev · 2025-03-16 11:43
svenmaz: yeah, that's my thinking as well#asahi-dev · 2025-03-16 11:34
svenyeah, my 2011 mba's battery looked like a spicy pillow as well and iirc it didn't want to boot without a battery#asahi-dev · 2025-03-16 11:33
svenso might as well punt on all that until later#asahi-dev · 2025-03-16 11:33
svenafaict most sub-devices need (small) changes to work on pre-M1 anyway#asahi-dev · 2025-03-16 11:31
sventhe t2 people are aware of it but "haven't figured it out yet"#asahi-dev · 2025-03-16 11:30
svenand we can always add the backend abstraction back in later#asahi-dev · 2025-03-16 11:29
svenhonestly? they had >2 years and nothing happened#asahi-dev · 2025-03-16 11:29
svenouch#asahi-dev · 2025-03-16 11:28
svenany objections to just removing the entire backend stuff from SMC and making it M1/M2/etc. only for now? it's been >2 years and no one bothered to write a T2 or x86 mac backend so we might as well rip all the indirection and abstractions out for now and revisit that later once someone actually wants to write a backend for e.g. T2 macs#asahi-dev · 2025-03-16 10:39
sveni don't think there's anything in pmgr that has to be turned on to bring the other cores up#asahi-dev · 2025-03-15 14:52
sventhanks#asahi-dev · 2025-03-15 14:42
sven(i'm adding documentation to SMC before trying to upstream it once more)#asahi-dev · 2025-03-15 14:42
sveni found "SMC returns power in Watts with acceptable precision to scale to uW" in hwmon but dunno if that's f32 or ioft#asahi-dev · 2025-03-15 14:42
svenjannau: / chadmed: do you have an example for a f32 and a ioft / fixed point key we read from SMC that has to be scaled? #asahi-dev · 2025-03-14 21:25
sven(in m1n1 that is, kernel stuff should directly go upstream)#asahi-dev · 2025-03-14 21:25
svenyeah, even if it's not fully working yet i'm happy to merge anything that looks reasonable and gets us closer to m3+#asahi-dev · 2025-03-14 21:20
svenbut yeah, if you fix all that feel free to open a new one#asahi-dev · 2025-03-14 21:20
sveni think that pull request also suffered from a-ramses not squashing stuff like clang format fixes#asahi-dev · 2025-03-14 21:19
svenand honestly, it's code that's called once and lives in its own file. it's allowed to be ugly :D#asahi-dev · 2025-03-14 21:19
svenwhich brings us back to <sven> it shouldn't matter but we really have no way of knowing#asahi-dev · 2025-03-14 21:18
svenyup#asahi-dev · 2025-03-14 21:16
svenif the bits are wrong you can everything from an obvious explosion early in the kernel to very subtle misbehavior later on#asahi-dev · 2025-03-14 21:15
svenit shouldn't matter but we really have no way of knowing#asahi-dev · 2025-03-14 21:15
svenI wouldn't touch the order#asahi-dev · 2025-03-14 19:26
svenlooks like some implementation defined register. unless there’s a name in xnu sources or whatever it’s going to be very hard to figure out what it does if it’s part of the chicken bit sequence #asahi-dev · 2025-03-14 17:55
svenprobably to prevent a compiler warning when the argument of a function isn’t used anywhere #asahi-dev · 2025-03-14 16:14
svendunno, would have to take a closer look#asahi-dev · 2025-03-14 16:10
sventhere are also still two or three open MRs, maybe rebasing these is enough#asahi-dev · 2025-03-14 16:10
svenor maybe that's already upstream? dunno#asahi-dev · 2025-03-14 16:10
sveniirc at least the cpu start logic#asahi-dev · 2025-03-14 16:09
svenit'll probably need some minor changes at least#asahi-dev · 2025-03-14 16:03
svengood luck :) would be nice to get at least m1n1 + hv to work on M4#asahi-dev · 2025-03-14 15:58
svenM4 also has this weird SPTM thing which may or may not come with changes to GXF/SPRR and may or may not work with our current hv code#asahi-dev · 2025-03-14 15:58
sveniirc the hv did something with pmgr as well, you may to patch that out and hope everything works#asahi-dev · 2025-03-14 15:55
svenand i really wouldn't want to upload a full xnu kernel over that slow serial either :/#asahi-dev · 2025-03-14 15:55
svene.g. if they changed anything with gxf/sprr which they may have done on M4 to support that exclave stuff (though that may be ios only? dunno)