#
✕ nick:sven
#asahi-dev · 2026-08-08 07:58
svenit'll cause some confusion though because even "normal" tbt devices will show up but just not do anything useful
#asahi-dev · 2026-08-08 07:57
svenyeah, i also think it would be funny
#asahi-dev · 2026-08-08 07:57
svenalso, lol, should we add thunderbolt host2host connections only to the release kernel? :D
#asahi-dev · 2026-08-08 07:51
svenbut very nice work on fixing it, i truly gave up because i had no idea what the issue was for months now!
#asahi-dev · 2026-08-08 07:51
svendamnit :D
#asahi-dev · 2026-08-08 07:48
svenit also needs more work for m4
#asahi-dev · 2026-08-08 07:48
svenI based my code on the thesis and can’t release it until he’s handed in the thesis at least
#asahi-dev · 2026-08-08 07:48
svenno
#asahi-dev · 2026-08-08 06:49
sven🙃
#asahi-dev · 2026-08-08 06:44
svenat least hotplug is also pretty broken on x86
#asahi-dev · 2026-08-08 06:39
svenand then there are also issues with the PCIe hotplug code because it wasn’t written with the craziness you can do with tbt in mind but maybe some of that is fixed now
#asahi-dev · 2026-08-08 06:38
svenso when it sees a change there it walks the tbt topology together with the PCIe topology and injects the interrupt manually 🫠
#asahi-dev · 2026-08-08 06:38
svenXnu doesn’t use it but if I read the logs right back then does something very annoying: every switch also shows up on the tbt/nhi side and you have to setup tunnels there anyway
#asahi-dev · 2026-08-08 06:37
svenso if you eg connect a network adapter to a dock it sometimes takes seconds for the interrupt to fire and sometimes it just never fires
#asahi-dev · 2026-08-08 06:36
svenwanna do PCIe hotplug next? the basics are simple but then it gets very cursed: the normal hotplug interrupt is unfortunately not very reliable here
#asahi-dev · 2026-08-08 06:14
svenNice find!
#asahi-dev · 2026-08-08 06:13
svenoh the pinctrl stuff, duh.
#asahi-dev · 2026-08-08 04:05
svennice! What was the issue?
#asahi-dev · 2026-08-07 17:43
svenso usually i send the PRs out the weekend before the merge window opens
#asahi-dev · 2026-08-07 17:43
sven*i can send some really late
#asahi-dev · 2026-08-07 17:42
svenand the patches have to be in -next for at least a day before i can send a PR
#asahi-dev · 2026-08-07 17:42
sveni can't send some really late PR but i'd like to avoid that
#asahi-dev · 2026-08-07 17:42
svenyuka: i don't sent PRs directly to linus but to the soc team so i have to send them before the merge window ideally
#asahi-dev · 2026-08-07 16:26
svenlol :D
#asahi-dev · 2026-08-07 16:25
sven*inevitable and pointless
#asahi-dev · 2026-08-07 16:25
sven(also, anyone wanna bet how long until the ) and pointless phoronix article? :P
#asahi-dev · 2026-08-07 16:23
svenalright, PRs sent. guess we'll do the M4 device trees after the merge window
#asahi-dev · 2026-08-06 17:18
svennice!
#asahi-dev · 2026-08-06 16:17
svenchaos_princess: it didn't make a difference and i don't understand why that's required
#asahi-dev · 2026-08-06 16:17
svenkettenis: called the v2 due to the quirk :D
#asahi-dev · 2026-08-06 16:01
svenit's currently stuck because there's some very weird bug that makes the co-processor not see the connected cable
#asahi-dev · 2026-08-06 15:32
svennow can someone please make thunderbolt work? i'd really like to use my dock with the external screen :(
#asahi-dev · 2026-08-06 15:27
sveni'm just gonna send the patches to the ML, it's probably gonna require a v2 due to that quirk anyway
#asahi-dev · 2026-08-06 15:27
svennice!
#asahi-dev · 2026-08-06 15:11
svenand here's the m1n1 side https://github.com/AsahiLinux/m1n1/pull/651
#asahi-dev · 2026-08-06 15:08
svennew version pushed to github and kernel.org
#asahi-dev · 2026-08-06 15:04
svenokay, confirmed that it's really not required for the io queue for some reason
#asahi-dev · 2026-08-06 13:47
sveni think i can get away with a quirk because there seems to already be some dma_alignment thingy in the core that i can use
#asahi-dev · 2026-08-06 13:47
sveni'll figure out how to fix that alignment for the admin queue and sent them to the ML then
#asahi-dev · 2026-08-06 13:45
sventhe good news is that we won't need that weird -fwabi compatible and apple,firmware-abi with these :)
#asahi-dev · 2026-08-06 13:44
svenkettenis: yeah, these are all required for m4+ or if we ever install macos >15 firmware for m1/m2
#asahi-dev · 2026-08-06 13:41
svenapparently this only applies to the admin queue though and not the io queue which makes it extra weird
#asahi-dev · 2026-08-06 13:40
sven🙃
#asahi-dev · 2026-08-06 12:48
sventhat's gonna be a fun quirk
#asahi-dev · 2026-08-06 12:47
svenso now it's gonna reject anything that's not 4 KiB aligned i think
#asahi-dev · 2026-08-06 12:46
svenso what i think that chicken bit does is skip quite a bunch of DMA validation crap
#asahi-dev · 2026-08-06 12:46
sven*sigh*
#asahi-dev · 2026-08-06 12:44
svencan you try https://f.svpe.de/005141b962949cf20379044f282ecce257301ee9e99f2509d85a99c773c456d6_sigh.patch
#asahi-dev · 2026-08-06 12:31
svenhrm, i don't see macos ever issuing that command
#asahi-dev · 2026-08-06 12:24
svennvme_admin_get_log_page is the thing that fails

plus ›