#asahi-dev · 2026-01-17 17:12
svenI suspect I messed up somewhere around there #asahi-dev · 2026-01-17 17:12
svenright now it finds the host router and establishes the domain but then it doesn’t see the connected device after I set that Apple specific VSE reg in the root switch #asahi-dev · 2026-01-17 17:10
svenyeah, don’t see anything suspicious either #asahi-dev · 2026-01-17 17:00
svenhrm, so either I messed something or something in drivers/thunderbolt changed between early 2025 or so and now :(#asahi-dev · 2026-01-17 16:49
svenhuh, I thought they all died #asahi-dev · 2026-01-17 16:29
svenand i'm not sure if stuff showed up when it still triggered that assert#asahi-dev · 2026-01-17 16:28
svenhm... i fixed the weird assert but now nothing shows up in boltctl anymore#asahi-dev · 2026-01-17 13:23
sven"With a sufficient number of users of an API, it does not matter what you promise in the contract: all observable behaviours of your system will be depended on by somebody. " ;)#asahi-dev · 2026-01-17 13:21
svenit's just that at least some subsystem (v4l? something like that) abuses the scatter-gatherlist map api which implicitly does that #asahi-dev · 2026-01-17 13:21
sveniirc there's no iommu api that guarantees continous maps#asahi-dev · 2026-01-17 13:07
sveni had a very old patchset that fixed that but dropped it because it didn't work for the gpu and we also had stuff like muvm at that point#asahi-dev · 2026-01-17 13:06
svenyeah, the 4k cpu / 16k iommu thing is just hard due to linux abstractions#asahi-dev · 2026-01-17 13:05
svenintegralpilot: i'd put it into the aic2 binding since there's no change except for the compatible#asahi-dev · 2026-01-17 12:05
sven-the#asahi-dev · 2026-01-17 12:05
svengreat, so just using the m3 compatible as baseline and hardcoded the values sounds viable then #asahi-dev · 2026-01-17 11:15
svenouch :D#asahi-dev · 2026-01-17 11:14
svenall that really matters is that the emails get sent to that asahi list anyway, and that happens due to MAINTAINERS#asahi-dev · 2026-01-17 11:14
sveneuh.. i guess we could change that while we're at it#asahi-dev · 2026-01-17 11:12
svenand the description adjusted i guess#asahi-dev · 2026-01-17 11:12
svenlooking at the dt-bindings if we do it like that it just needs the compatibles added to apple,aic2 (though it technically isn't aic2 anymore then :D)#asahi-dev · 2026-01-17 11:10
sveneven if it's different i'd just hardcode the values for M3 max and just not use the fallback M3 base compatible there then #asahi-dev · 2026-01-17 11:07
svenand then just remove those device tree properties#asahi-dev · 2026-01-17 11:07
svenif it is i'd just use the base M3 compatible as the fallback for all of these and hardcode the offsets in the driver#asahi-dev · 2026-01-17 10:53
svenmaybe there's aic4 and aic5 now :P#asahi-dev · 2026-01-17 10:53
svenstill a good idea to look at m4 and m5 adts as well#asahi-dev · 2026-01-17 10:52
svennickchan: and they need different compatibles anyway, so at least right now I don't see a reason for those apple,config-offset, etc. properties in the FDT#asahi-dev · 2026-01-17 10:52
sventhen come up with dt-bindings based on that which just use "apple,tXXXX-aic3" or something like that#asahi-dev · 2026-01-17 10:51
svenwhat i'd do is look at all M3 ADT and collect config-offset,cap0-offset and maxnumirq-offset and see if and how they differ#asahi-dev · 2026-01-17 10:50
svenneeds dt-bindings still I think#asahi-dev · 2026-01-17 10:50
svenit can be submitted on its own#asahi-dev · 2026-01-17 10:50
svenand if we remove the generic compatible anyway we can also hardcode the 3(?) new fdt properties i guess#asahi-dev · 2026-01-17 10:49
svenand the current thing at least needs the generic compatible removed before that's possible#asahi-dev · 2026-01-17 10:49
svensure, that's part of that#asahi-dev · 2026-01-17 10:47
svenfirst step would be to get some minimal M3 device tree on LKML#asahi-dev · 2026-01-17 10:47
svenbut still... much more important things to do before that#asahi-dev · 2026-01-17 10:47
svenoh, because it's a MIDR list and not just enabled on any apple machine#asahi-dev · 2026-01-17 10:43
svenwhy before release?#asahi-dev · 2026-01-17 10:40
svenmight be fun to know, but not very important imho #asahi-dev · 2026-01-17 10:02
svenyeah#asahi-dev · 2026-01-17 10:00
svenit' just weird that it really doesn't like multiple cpus to be active#asahi-dev · 2026-01-17 09:23
svenyup :D#asahi-dev · 2026-01-17 09:22
svenlol#asahi-dev · 2026-01-17 09:11
svennvme probably just needs someone with the hv to run run_guest.py with -m trace_nvme.py (maybe set verbose=3 in there) and see what changed#asahi-dev · 2026-01-17 09:10
svennoopwafel: afaik there are issues with nvme, integralpilot has managed to boot a linux kernel to userspace but other people ran into issues so far#asahi-dev · 2026-01-17 08:59
svenit’s gonna suck if this some long standing locking or whatever issue that we only hit now because m3 speculates even harder than m1/2 or something equally cursed 🙈#asahi-dev · 2026-01-17 08:52
svennice, will take another look either this afternoon or tomorrow probably #asahi-dev · 2026-01-17 08:45
svenso you’re booting with two cores and that works but any more and it’s unstable again? Weird #asahi-dev · 2026-01-16 22:48
svenso you get about 5 more seconds before the SoC resets :D#asahi-dev · 2026-01-16 22:48
svenso the problem is that un-gracefully shutting down the co-processor by plugging in something for the second time into that port also triggers a nasty watchdog#asahi-dev · 2026-01-16 22:48
svenlol