#asahi-dev · 2026-08-14 09:18
svennope, didn't know about that. looks useful#asahi-dev · 2026-08-14 09:01
svenack#asahi-dev · 2026-08-14 09:00
sveni've also started with the M2 pro/max nodes#asahi-dev · 2026-08-14 09:00
svenbut thanks!#asahi-dev · 2026-08-14 08:59
svenah.. i've also done those locally already#asahi-dev · 2026-08-14 08:36
svenno idea tbh#asahi-dev · 2026-08-14 07:38
sventhe aliases trigger warnings right now i think and one of them looks very wrong#asahi-dev · 2026-08-14 07:38
sveni need to clean up the dt anyway a bit#asahi-dev · 2026-08-14 07:38
svensure#asahi-dev · 2026-08-14 04:46
svenbringing up dwc3 is probably very wrong but I don’t get why the full phy and dwc reset isn’t enough to fix it again #asahi-dev · 2026-08-13 21:09
sventhat’s with bonded lanes, right?#asahi-dev · 2026-08-13 21:06
svennice!#asahi-dev · 2026-08-13 21:05
sventhat’s surprising since we go through a full phy reset there #asahi-dev · 2026-08-13 21:04
svenI don’t care tbh, I’m just not going to take part in that discussion #asahi-dev · 2026-08-13 21:02
svenand it might be hard to argue for that when adding multiple iommus already works #asahi-dev · 2026-08-13 21:02
svenit would require an additional device tree property #asahi-dev · 2026-08-13 21:01
sven*tried#asahi-dev · 2026-08-13 21:00
svenI‘m kinda glad I obsessively tired to figure out how dart works back when m1 was released and still remembered this #asahi-dev · 2026-08-13 20:58
svenfuck you Apple, was that really necessary :D#asahi-dev · 2026-08-13 20:57
sven🥳🥳🥳#asahi-dev · 2026-08-13 20:55
svencan you try those two in the fdt?#asahi-dev · 2026-08-13 20:54
svenSo that’s sid 1 and 2 both mapped to 1#asahi-dev · 2026-08-13 20:39
sventrue, that also works#asahi-dev · 2026-08-13 20:39
svenotherwise on the first acio register access#asahi-dev · 2026-08-13 20:38
svenmaybe hook that pmgr register? though i don't know how quickly the dart comes up after that#asahi-dev · 2026-08-13 20:38
svenhm.. right, we'd need to capture the early setup#asahi-dev · 2026-08-13 20:25
svenwould still be good to trace the dart with verbose=3 and just hack the linux driver to reproduce as much of that as possible to rule it out#asahi-dev · 2026-08-13 20:25
svenso maybe this is just another dead end :(#asahi-dev · 2026-08-13 20:25
svenhm... okay, that's odd#asahi-dev · 2026-08-13 20:23
svenyes, so it'll just stay however iboot programmed it#asahi-dev · 2026-08-13 20:22
svenisn't the dcp dart locked?#asahi-dev · 2026-08-13 20:15
svenjannau: huh, did i look in the wrong device tree then? i thought base M2 also had no remap#asahi-dev · 2026-08-13 20:13
svenyes#asahi-dev · 2026-08-13 20:12
svenmax verbose trace on the dart and also iommus = <.. 0>, <.. 1>; would be nice#asahi-dev · 2026-08-13 20:11
svenit's usually 0102030405060708090a0b0c0d0e0f iirc#asahi-dev · 2026-08-13 20:10
svenhttps://github.com/AsahiLinux/m1n1/blob/06a4601a351ebfd1abb6abba9a44c34e40d94776/src/dart.c#L36 that thing#asahi-dev · 2026-08-13 20:09
svenyes#asahi-dev · 2026-08-13 20:08
svenchaos_princess: i vaguely remember that t8103 had a feature somewhere where you could map the hardware stream id to the stream id you actually program#asahi-dev · 2026-08-13 20:07
svennope, just noticed that as well#asahi-dev · 2026-08-13 20:03
svenso the t600x ADT has sids = 32771 (0x8003, so streams 0,1 and 15) and fucking remap = 258 which is 0x0102#asahi-dev · 2026-08-13 19:57
svenor do iommus = <&DIE_NODE(usb4_1_dart) 0>, iommus = <&DIE_NODE(usb4_1_dart) 1>; in the dt?#asahi-dev · 2026-08-13 19:56
svenwait. can you specifically enable stream 0 as well?#asahi-dev · 2026-08-13 19:56
svenjust double checked i got the correct dart and stream with the adt. also no issue there :(