#asahi-dev · 2022-10-05 15:12
sventhen some configuration inside what apple calls “dpin” needs to be done to hook everything up#asahi-dev · 2022-10-05 15:12
svenalright, looks like the thunderbolt display port tunnel connected signal actually comes from the NHI#asahi-dev · 2022-10-05 14:08
svensure#asahi-dev · 2022-10-05 14:06
svenso if fastipi on a11 crashes with undef instr because it tries to access a the sysreg that doesn't exist or maybe just hangs because it uses the other one incorrectly it's probably that#asahi-dev · 2022-10-05 14:03
sventhere's a #if defined(APPLEMONSOON) || HAS_CLUSTER somewhere and monsoon is one of the A11 cores, isn't it?#asahi-dev · 2022-10-05 13:58
svenkonradybcio: https://github.com/apple/darwin-xnu/blob/main/osfmk/arm64/machine_routines.c#L176 i'm willing to bet A11 just doesn't have the global/local split#asahi-dev · 2022-10-03 15:23
svenIm very glad I kept atcphy stashed away in a side branch. Looks like it’ll already have to change quite a bit just for DP altmode #asahi-dev · 2022-10-03 14:02
sventhat makes two of us ;)#asahi-dev · 2022-10-03 13:54
svenjannau: fyi, https://lore.kernel.org/linux-usb/20220927155332.10762-1-andriy.shevchenko@linux.intel.com/ also fixes that phy/dwc3 error that you run into when rebasing atcphy#asahi-dev · 2022-10-02 14:35
svenno worries #asahi-dev · 2022-10-01 15:59
svenso that all makes sense now#asahi-dev · 2022-10-01 15:58
svenand then dfp-switch-ufp-endpoint encodes the index inside the dpxbar#asahi-dev · 2022-10-01 15:58
svenwhich is zero for everything on the first die and one for everything on the second die#asahi-dev · 2022-10-01 15:58
svenokay, i think apple collects the valid dispext<->dpxbar connections inside a "domain" and the ADT then has dp-switch-ufp-domain#asahi-dev · 2022-10-01 15:27
svensure#asahi-dev · 2022-10-01 15:26
svenyeah, I guess I could do that. probably a nice test case for thunderbolt anyway#asahi-dev · 2022-10-01 15:25
svenofc it also has firmware updates#asahi-dev · 2022-10-01 15:23
svenlol, apparently it essentially runs ios#asahi-dev · 2022-10-01 15:21
svenI assume it also works on the m1 non-pro/max?#asahi-dev · 2022-10-01 15:20
svenwhat is that display anyway? just thunderbolt with some special usb devices?#asahi-dev · 2022-10-01 15:18
sveni feel like i've already signed up for this cursed mess when I wanted thunderbolt anyway :D#asahi-dev · 2022-10-01 15:18
svenbut on the trace from yesterday it looked very similar to the situation when just a single screen was connected#asahi-dev · 2022-10-01 15:17
svenno idea what iomfb itself does. i'll probably annoy dottedmag for more traces once I get there#asahi-dev · 2022-10-01 15:17
svenso far I just have an idea how to convince it to connect two streams and probably how to route them to separate atcs#asahi-dev · 2022-10-01 15:17
svenno idea yet, it's still all very vague#asahi-dev · 2022-10-01 15:16
sventhe actual implementation will be pretty cursed if it works out, yes#asahi-dev · 2022-10-01 15:15
svendon't know yet#asahi-dev · 2022-10-01 15:14
svenlayers?#asahi-dev · 2022-10-01 15:13
svenI'm pretty sure that they could also support two external screens on the normal M1 if they modified their DCP firmware a bit. my guess is marketing ;)#asahi-dev · 2022-10-01 15:12
svenyeah, I guess in their "fake crossbar kext" they just hardcoded that it will only match a dpxbar and a dcpext from the same die#asahi-dev · 2022-10-01 15:11
svendpxbar is actual HW and there's one of those per ATC#asahi-dev · 2022-10-01 15:10
svencrossbar is their fake top-level node that decides routing#asahi-dev · 2022-10-01 15:10
svennote that there is "crossbar" and "dpxbar" in the ADT#asahi-dev · 2022-10-01 15:09
svenbut then again, the ADT is full of hacks anyway#asahi-dev · 2022-10-01 15:09
svenyeah, I would've just expected to see that represented in the ADT#asahi-dev · 2022-10-01 15:08
sven*routing#asahi-dev · 2022-10-01 15:08
svenso I guess they hardcoded that rouing can only happen within a single die somewhere#asahi-dev · 2022-10-01 15:07
svenah, lol#asahi-dev · 2022-10-01 15:07
svenStreamCodec?#asahi-dev · 2022-10-01 14:53
sveni assumed there should be 16 of those per crossbar but there seem to be 9#asahi-dev · 2022-10-01 14:53
sventhe weird part is the dcpextN,M input side#asahi-dev · 2022-10-01 14:53
sventhere is a single xbar for each atcphy#asahi-dev · 2022-10-01 14:52
svendisplay-crossbar0 in the ADT claims 16 UFP which matches 8 x dcpext and 2 streams/dcpext#asahi-dev · 2022-10-01 14:51
svenstill, the crossbar is a bit weird unless you can actually only route within a single die and they hardcoded that inside some kext or something#asahi-dev · 2022-10-01 14:50
svenokay, that might make sense#asahi-dev · 2022-10-01 14:49
svenor do you need two dcps for that? :s#asahi-dev · 2022-10-01 14:48
svena single dcp should be able to drive a 6k display, shouldn't it?#asahi-dev · 2022-10-01 14:44
svenand yet the ADT only has a single display crossbar top-level node that seems to list everything#asahi-dev · 2022-10-01 14:42
svenhrm.. so that's at most 9 inputs for the crossbar but each dcpext should be 2 separate streams