#asahi-dev · 2026-06-24 11:13
svenwould move the enum to a header and add yield like will suggested, but looks good otherwise #asahi-dev · 2026-06-24 11:03
svenack, gonna take a look on the weekend then #asahi-dev · 2026-06-24 11:01
svensure, just send me the latest code and I’ll take a look at it on the weekend #asahi-dev · 2026-06-24 08:32
svenit’s not as bad for us since the fw will be open source anyway imho#asahi-dev · 2026-06-24 06:20
svenGreat!#asahi-dev · 2026-06-24 06:09
svencan we just try UART and see what happens? we have kisd and all the infrastructure now and a simple UART driver that uses mbox shouldn’t be hard #asahi-dev · 2026-06-24 06:09
svenwe may need to access the regs directly for earlycon anyway #asahi-dev · 2026-06-23 17:53
svenfwiw, i'd just fold the MAINTAINERS update to the patches where you add new files#asahi-dev · 2026-06-23 17:53
svenintegralpilot: will take look at the weekend#asahi-dev · 2026-06-23 16:07
svenfrom what i can tell it's actually meant for weird stuff like apple socs #asahi-dev · 2026-06-23 16:06
svenno, i think it's instruction set architecture here#asahi-dev · 2026-06-23 16:03
svenhah, UEFI even supports EFI_MEMORY_ISA_MASK to encode arbitrary MAIR attributes. no need for that arm,apple-platform hack I used to have :)#asahi-dev · 2026-06-21 20:49
svenEpicMan: what are you trying to do? looks like we don't know what that reg does, otherwise it would usually be in there https://github.com/AsahiLinux/m1n1/blob/04bad8c9908e734dbe9d0389f594447520232a94/proxyclient/m1n1/apple_regs.json#L165#asahi-dev · 2026-06-21 17:54
svens/place/replace/#asahi-dev · 2026-06-21 17:54
svenhave you just not tried smp yet or is there something broken?#asahi-dev · 2026-06-21 17:54
svennohlt is probably fine for now, long-term we need whatever yuka comes up with that's accepted upstream to place wfi with nop or yield or whatever and that psci-via-efi thingy to actually get to reasonable power consumption#asahi-dev · 2026-06-21 17:42
sventhe reason i doubt it because xnu doesn't guard wfe with context save/restore#asahi-dev · 2026-06-21 17:42
svencan you reproduce that inside m1n1?#asahi-dev · 2026-06-21 17:41
svenuhh... i highly doubt that wfe thing#asahi-dev · 2026-06-20 10:50
svenyup#asahi-dev · 2026-06-20 10:06
sventhat dockchannel downstream drivers needs quite a bit of rework anyway #asahi-dev · 2026-06-19 10:00
svenwonder if this is another missing barrier somewhere that we accidentally introduce or work around with the hv then :/ #asahi-dev · 2026-06-19 07:41
svenbut we'll need something like that for proper deep sleep anyway#asahi-dev · 2026-06-19 07:41
svenwe just never cared to implement it because the downstream cpuidle driver was good enough#asahi-dev · 2026-06-19 07:40
sveni'm working on psci via efi runtime services right now for the proper solution and it's not that much of a hack tbh#asahi-dev · 2026-06-19 07:40
svenr.e. the wfi stuff: for now just patch wfi to nop, yuka has a patch on LKML for that#asahi-dev · 2026-06-18 19:27
svenunless you set some bit in some impdef sysreg to trap genter, then you'll end up in GL2 or EL2#asahi-dev · 2026-06-18 19:27
svenoh, and if a guest on M1-M3 calls genter from EL1 it'll end up in GL1 fwiw#asahi-dev · 2026-06-18 19:26
svenyeah, that should be fine#asahi-dev · 2026-06-18 19:14
svenif gxf isn't enabled just writing to them from EL2 is fine#asahi-dev · 2026-06-18 19:14
svenat least HCR_EL2 but i think there were a few more#asahi-dev · 2026-06-18 19:14
sventhere are some sysregs that can only be written from gl2 once its enabled#asahi-dev · 2026-06-18 18:59
svenproxyclient contains quite a bunch of hacks anyway since we don't ship that to end-users#asahi-dev · 2026-06-18 18:58
svenno idea :D#asahi-dev · 2026-06-18 14:58
sveneven better then#asahi-dev · 2026-06-18 14:58
sventhis will only exist for M1+ for now and you can fix it later for <M1 then#asahi-dev · 2026-06-18 14:57
svenwhy not?#asahi-dev · 2026-06-18 14:57
svenanything before that can always be a follow-up patch#asahi-dev · 2026-06-18 14:57
sveni only care about M1+ here#asahi-dev · 2026-06-18 14:53
svenneeds more invasive patches because that's not how efi runtime services are supposed to work#asahi-dev · 2026-06-18 14:50
sven+ prot = __pgprot(PROT_DEVICE_nGnRnE);#asahi-dev · 2026-06-18 14:50
sven+ if (of_machine_is_compatible("apple,arm-platform"))#asahi-dev · 2026-06-18 14:50
svenyeah, right now i have this hack here but that's something we can bikeshed on the mailing list#asahi-dev · 2026-06-18 14:48
svenso it's just the usual 1:1 mapping#asahi-dev · 2026-06-18 14:48
svenno, you can define special EFI runtime mappings that the kernel enables before jumping to m1n1 again#asahi-dev · 2026-06-18 14:30
svenrequires a rather hacky kernel patch because there's no way to express "we really need Device-nGnRnE" in the memory table#asahi-dev · 2026-06-18 14:30
svenyup#asahi-dev · 2026-06-18 14:16
svenit's just a nop for now but the kernel -> m1n1 -> kernel calls work now and the stub can still access all mmio