#
✕ nick:marcan
#asahi-dev · 2023-05-09 10:46
marcanpovik: ivsense fun stuff in https://github.com/AsahiLinux/linux/tree/sound/ivsense-improvements
#asahi-dev · 2023-05-09 04:41
marcan(or maybe once we get that Rust USB stack into m1n1 we should replace u-boot's stack with that too :p)
#asahi-dev · 2023-05-09 04:39
marcankettenis: is there some way we can skip u-boot USB init on laptops by default on internal installs? that will both speed up boot time and insulate us from USB stack brokenness a bit (which people keep running into...)
#asahi-dev · 2023-05-09 00:31
marcanwhat I meant is that ASoC at least seems to have a reasonable model to cover the combinations, while so many other niche SoC subsystems in the kernel kind of utterly fail at abstracting things out usefully
#asahi-dev · 2023-05-09 00:30
marcanpovik: oh yeah, docs are a different issue and it's not perfect
#asahi-dev · 2023-05-08 14:12
marcanmore importantly we'll need hub support
#asahi-dev · 2023-05-08 14:12
marcanUSBMS (the old style) is pretty trivial IIRC
#asahi-dev · 2023-05-08 13:23
marcanwell there's a collab idea for you, port that to m1n1 ;)
#asahi-dev · 2023-05-08 13:23
marcanand it's even MIT licensed :D
#asahi-dev · 2023-05-08 13:23
marcannice lol
#asahi-dev · 2023-05-08 13:02
marcan*how
#asahi-dev · 2023-05-08 13:02
marcanI was just thinking about howe we need xHCI support in m1n1 and I remembered that...
#asahi-dev · 2023-05-08 13:02
marcanlina: hey, didn't you do a collab with someone writing an OS in Rust?
#asahi-dev · 2023-05-08 09:20
marcanhttps://youtube.com/live/bh7KbJIq_iM
#asahi-dev · 2023-05-08 08:50
marcanah it's a separate bit for that one, yeah
#asahi-dev · 2023-05-08 08:43
marcan(which maps to apple's IMPDEF ACTLR_EL12 from EL2)
#asahi-dev · 2023-05-08 08:43
marcanI mean ACTLR_EL1 from a guest should trap I think
#asahi-dev · 2023-05-08 08:43
marcanoh a guest wouldn't even see that one
#asahi-dev · 2023-05-08 08:42
marcanbut that shouldn't affect KVM much, since I think TIDCP makes it trap anyway
#asahi-dev · 2023-05-08 08:42
marcanapple actually has ACTLR_EL12 as an impdef, so the guest wouldn't even see the host's ACTLR I think (it got the VHE treatment)
#asahi-dev · 2023-05-08 08:40
marcanas in save/restore the host's value, right?
#asahi-dev · 2023-05-08 08:40
marcanso you're safe for now ;)
#asahi-dev · 2023-05-08 08:39
marcanbut that future may never come
#asahi-dev · 2023-05-08 08:39
marcanin a hypothetical future where FEX runs as a 4K guest we can revisit that idea, though for that it would be sufficient to enable TSO guest-wide for most purposes
#asahi-dev · 2023-05-08 08:39
marcanoh yeah I don't care about KVM, at least not yet
#asahi-dev · 2023-05-08 08:39
marcanOTOH it's not exactly the ugliest thing ever, it's flipping a bit in a documented, IMPDEF-defined register (ACTLR_EL1)
#asahi-dev · 2023-05-08 08:37
marcan(you know I'm not a fan of having to do this, but the benefits are too big to ignore)
#asahi-dev · 2023-05-08 08:37
marcanhence annoy you
#asahi-dev · 2023-05-08 08:37
marcanmaz: well obviously we'll be pushing this upstream because it's a *BIG* performance deal
#asahi-dev · 2023-05-08 08:36
marcanI thought you didn't want impdef arch stuff ;)
#asahi-dev · 2023-05-08 08:35
marcan(TSO is confirmed to be a *big* deal for FEX so we are definitely implementing it, sorry maz)
#asahi-dev · 2023-05-08 08:35
marcanI'm sure that will greatly annoy maz and make the FEX folks very happy :p
#asahi-dev · 2023-05-08 08:35
marcanI was going to stream speakers today but maybe I should try to add a prctl for TSO?
#asahi-dev · 2023-05-08 08:33
marcanespecially for stuff that we really care about
#asahi-dev · 2023-05-08 08:33
marcan*Apple versions
#asahi-dev · 2023-05-08 08:33
marcanI do wonder, once official versions of these things get supported, how likely it is we can convince upstream to take alternate codepaths for the pre-standard ARM versions
#asahi-dev · 2023-05-08 08:32
marcanooh nice :D
#asahi-dev · 2023-05-08 08:30
marcanall the extra SPRR stuff is curious but likely nobody cares enough to figure out what exactly it all does, since it doesn't really affect us
#asahi-dev · 2023-05-08 08:29
marcanwas nice confirming basically everything we already knew though, I didn't find anything unexpected in there other than a pile of official register names for registers we'd already roughly identified
#asahi-dev · 2023-05-08 08:28
marcan(and then of course it calls into pmap/gl to actually switch to the guest etc)
#asahi-dev · 2023-05-08 08:28
marcanARM HVs are boring enough a significant part of the entire HV is touching secret stuff
#asahi-dev · 2023-05-08 08:27
marcanbut I can sort of understand why, because they save/restore all the secret registers in there lol
#asahi-dev · 2023-05-08 08:27
marcanself-compiled XNU = no hv
#asahi-dev · 2023-05-08 08:27
marcanI'm still annoyed the hypervisor is sekrit, like not even distributed as object files
#asahi-dev · 2023-05-08 08:26
marcanand then there's that raspberry pi target :D
#asahi-dev · 2023-05-08 08:26
marcanyeah of course
#asahi-dev · 2023-05-08 08:26
marcanmac13g, mac13j, mac14g, mac14j
#asahi-dev · 2023-05-08 08:26
marcanyeah
#asahi-dev · 2023-05-08 08:26
marcanthey haven't figured out how to stop using #ifdef everywhere :p
#asahi-dev · 2023-05-08 08:25
marcanwell 5 with vmapple

plus ›