#asahi-dev · 2026-11-17 16:18
svenIt’s all in the docs #asahi-dev · 2026-11-17 16:07
svenyou need the Linux vm to build the Linux kernel, that’s why it was suggested #asahi-dev · 2026-11-17 15:59
svenso, uh, do you want to just continue then? As I said, I haven’t written any code yet and just thought about what I was going to do #asahi-dev · 2026-11-17 15:54
svenoh well.. if we need to do that matching anyway we may as well do everything that way then #asahi-dev · 2026-11-17 15:49
svenoh, lol, or just construct the alias rvalue in m1n1 if we want to hack this :D#asahi-dev · 2026-11-17 15:41
svenso my plan was iterate over nodes by compatible, then extend dt_get_address to dt_get_address_by_index (or maybe even dt_get_address_by_name) to match against the right adt reg index#asahi-dev · 2026-11-17 15:38
svenand then just look for the NHI in its children#asahi-dev · 2026-11-17 15:38
svenfor tbt I'd match the "rc" reg to adt reg[2] to find the parent node#asahi-dev · 2026-11-17 15:37
svenI haven't written much code yet though so feel free to take over :D#asahi-dev · 2026-11-17 15:37
sven*match#asahi-dev · 2026-11-17 15:37
svenfor atcphy I think we can patch the usb2phy regs to adt reg[2] #asahi-dev · 2026-11-17 15:36
svenah... I was kinda working on atcphy and tbt/nhi in parallel#asahi-dev · 2026-11-17 13:21
sven*sigh*#asahi-dev · 2026-11-17 13:05
svenI’d rather just write code in m1n1 to find the right nodes without aliases 😅#asahi-dev · 2026-11-17 13:00
svenwe can just walk the dt in m1n1 instead so I’m just very confused now #asahi-dev · 2026-11-17 12:59
svenI also don’t get how the aliases are related to the dt-binding review he wants to retract now #asahi-dev · 2026-11-17 12:58
svenyeah, same *sigh*#asahi-dev · 2026-11-16 18:25
svendo you configure the crossbar correctly?#asahi-dev · 2026-11-16 18:25
svenit probably doesn't, I know it matters for pcie because rtkit will crash if you don't do that so if you don't see any crashes this should be fine#asahi-dev · 2026-11-16 18:24
sveniirc xnu only brings DPIN up after the tunnel has been established in the NHI but I'm not sure that matters#asahi-dev · 2026-11-16 18:24
svenhm... that looks reasonable at first glance#asahi-dev · 2026-11-14 13:49
svennice!#asahi-dev · 2026-11-14 13:49
sven\o/#asahi-dev · 2026-11-13 17:38
sventhe m5 series is ready as well, isn't it? will pick that up next week or so then#asahi-dev · 2026-11-13 17:18
svenMichaelReeves[m]: done#asahi-dev · 2026-11-13 13:00
svenMichaelReeves[m]: I’ll pick it up this evening #asahi-dev · 2026-11-13 07:46
svenI’ll just pick that one up then #asahi-dev · 2026-11-13 07:46
svenah, the patch to rtkit is just independent#asahi-dev · 2026-11-13 07:45
svenbut hid makes most sense imho (plus dts changes through soc as usual)#asahi-dev · 2026-11-13 07:44
svenIsn’t that series soc+mailbox+hid tree?#asahi-dev · 2026-11-13 07:40
svenor we do the silly immutable branch that’s merged into multiple trees thing but ugh.. that’s the most effort for everyone #asahi-dev · 2026-11-13 07:40
svenif they don’t want it I can take it with their acks as well ofc ;)#asahi-dev · 2026-11-13 07:30
svenand I think it probably makes most sense if it all goes through the hid tree. #asahi-dev · 2026-11-13 07:29
svenno, I can’t take any patches touching other trees without acks #asahi-dev · 2026-11-11 17:24
svendidn't realize they released it afterwards#asahi-dev · 2026-11-11 17:24
svenyeah, that's what I hope as well#asahi-dev · 2026-11-11 17:14
sventhis looks different enough such that if/then in the dt-binding would get annoying#asahi-dev · 2026-11-11 17:14
svenwe should probably rename apple,atcphy.yaml to apple,t8103-atcphy.yaml as well since I'll probably create apple,t6040-atcphy.yaml (or whatever m3 ultra is if it's really the same there)#asahi-dev · 2026-11-11 17:11
svenat least XNU is very verbose again, still need to map MMIO pokes to register names though#asahi-dev · 2026-11-11 17:10
svenugh... let's see then :d#asahi-dev · 2026-11-11 17:09
svenand the co-processor seems to be very useless, you start it with the usual rtkit dance and then never interact with it again before shutdown afaict#asahi-dev · 2026-11-11 17:09
sventunables look very different fwiw#asahi-dev · 2026-11-11 17:08
svenneed to stare a bit more at traces anyway#asahi-dev · 2026-11-11 17:08
svenokay, nice, I'll wait for that then#asahi-dev · 2026-11-11 17:03
svenchaos_princess: do you plan to split atc.c for v2? asking because I'll definitely need to do that for m4 pro I think#asahi-dev · 2026-11-10 20:36
svenI was worried they changed that part#asahi-dev · 2026-11-10 20:36
svenall the sequencing with dwc3, etc. seems to be just the same#asahi-dev · 2026-11-10 20:36
svenm4 pro atcphy really doesn't look too bad. different MMIO pokes and you need to start a random co-processor in the middle and then just not interact with it :D#asahi-dev · 2026-11-10 17:52
svenif it’s upstream just follow the usual Linux stuff #asahi-dev · 2026-11-10 17:21
svenYou’ll also need a notification from tbt tunnel up to dpin but feel free to just hack that for now