#asahi-dev · 2026-08-21 18:38
svenjannau: yeah, pretty sure that commit is the issue#asahi-dev · 2026-08-21 16:30
sventhe monitor will probably show up in boltctl and that's it if everything works#asahi-dev · 2026-08-21 16:29
svenonly thing that works is XDomain (machine to machine) and usb3-via-usb4 tunneling#asahi-dev · 2026-08-21 16:28
svenyou're also aware the monitor won't work because that branch doesn't implement dp (or pcie) tunneling, right?#asahi-dev · 2026-08-21 16:27
svenyou do have thunderbolt devices or another thunderbolt-capable machine, right?#asahi-dev · 2026-08-21 16:23
svenespecially for hardware bringup that touches multiple subsystems #asahi-dev · 2026-08-21 16:23
svenlinux with its dozens (hundreds?) of subsystem trees also results in extra fun#asahi-dev · 2026-08-21 16:22
sven*quite bad #asahi-dev · 2026-08-21 16:22
svengit _is_ hard and the UX is quite and imho. I’ve just gotten used to the pain #asahi-dev · 2026-08-21 16:19
svenit needs some tbt patches that are in 7.2#asahi-dev · 2026-08-21 16:19
svenyeah, that’s not gonna be fun #asahi-dev · 2026-08-21 16:07
svenbah… so much for let someone else take care of that :D#asahi-dev · 2026-08-21 16:00
svenoh, maybe it needs that driver unbind or root domain teardown the apple one does :/#asahi-dev · 2026-08-21 16:00
svenhuh#asahi-dev · 2026-08-21 15:37
svennice!#asahi-dev · 2026-08-21 15:35
sventhis sounds tricky to track down and fix #asahi-dev · 2026-08-21 15:34
svenif it happens on the x86 machine I’d probably just report it to the Thunderbolt maintainers #asahi-dev · 2026-08-21 15:32
svenyeah, same. looking at git log I suspect a8937f35cf39c39c64325aa84d0463d866850857#asahi-dev · 2026-08-21 15:29
svenyou have a x86 machine as well, don'y you? can you try to connect a dock and then another host and then unplug the dock?#asahi-dev · 2026-08-21 15:22
sventhe switch is already gone but somehow the xdomain teardown still wants to do something with it#asahi-dev · 2026-08-21 15:21
sveni got it wrong from staring at the ADT, janne just dumped the REMAP register where it was more obvious#asahi-dev · 2026-08-21 15:20
svenlooks like something has been freed already by the time the xdomain is torn down#asahi-dev · 2026-08-21 15:16
svenyeah, this smells like something with xdomain goes wrong, there's tb_xdomain_unregister in the stack tace#asahi-dev · 2026-08-21 15:06
sveni think i just reproduced it with just a single dock and tbnet behind it#asahi-dev · 2026-08-21 15:02
svendrivers/thunderbolt RiiR wen? :-P#asahi-dev · 2026-08-21 14:59
svenand maybe run with KASAN or however it's called#asahi-dev · 2026-08-21 14:59
svencan you show me the other kernel panic?#asahi-dev · 2026-08-21 14:58
svenmaybe you're just hitting this now because the usb3 tunnels are actually up now#asahi-dev · 2026-08-21 14:56
sventhat sounds like pcie hotplug hell!#asahi-dev · 2026-08-21 14:51
svenoh, i mean your kernel panic#asahi-dev · 2026-08-21 14:47
svenhm, i wonder if it's related to the usb3 quirk i had to add. don't see how though#asahi-dev · 2026-08-21 14:41
sveni did some cleanup in thunderbolt/apple.c though, maybe that broke something#asahi-dev · 2026-08-21 14:40
sveni don't think it does, i just rebased on that because i wanted to get rid of the nvme commits#asahi-dev · 2026-08-21 14:40
svenlet me check#asahi-dev · 2026-08-21 14:39
svenhuh, don't think so. i just based it on torvalds/master#asahi-dev · 2026-08-21 13:52
svenalso stumbled across a register called "AUSBC_PWR_CTRL" which we just completely ignore so far which seems to maybe be used in suspend/resume of atcphy#asahi-dev · 2026-08-21 13:18
svenoh, and dt-bindings#asahi-dev · 2026-08-21 13:18
sventhat only leaves the rest of the owl: pcie and dp tunnels :D#asahi-dev · 2026-08-21 13:17
svennow with usb3-via-4 tunnel support: https://github.com/AsahiLinux/linux/commits/tbt-reset-wip/#asahi-dev · 2026-08-21 07:32
svenbecause i always get the "lock not acked" there#asahi-dev · 2026-08-21 07:32
sveni wonder if we should just skip the pipehandler lock when switching to dummy on teardown#asahi-dev · 2026-08-20 12:12
sveni wonder if they ever tested a setup like that :D#asahi-dev · 2026-08-20 12:11
svenlol#asahi-dev · 2026-08-20 12:10
svenand for usb4 mode the pipehandler will fall back to dummy#asahi-dev · 2026-08-20 12:09
sventhat makes sense too because D+/- is still separate :D#asahi-dev · 2026-08-20 12:06
svenanything that needs a pcie, dp or usb3 tunnel won't