#
✕ nick:sven
#asahi-dev · 2023-01-21 08:41
svenonly difference to the working bars in Linux seems to be that the broken one is marked as prefetchable
#asahi-dev · 2023-01-21 08:41
svenI guess there’s still some magic init missing somewhere
#asahi-dev · 2023-01-21 08:41
svenhm, so it seems like that bar that only reads as -1 in Linux works fine in macOS
#asahi-dev · 2023-01-20 14:52
svenfrom what I understand the NHI driver establishes the tunnel and then usually the pcie just gets a hotplug interrupt on x86
#asahi-dev · 2023-01-20 14:52
svenhah. sounds like some firmware somewhere does all the heavy lifting then
#asahi-dev · 2023-01-20 14:51
svendoes hotplug also work?
#asahi-dev · 2023-01-20 14:50
svenshows how clueless I am still about thunderbolt ;)
#asahi-dev · 2023-01-20 14:49
svenyeah, there are firmware connection managers especially on older hardware afaik
#asahi-dev · 2023-01-20 14:47
svendoes openbsd already have support for the "normal" Thunderbolt NHI?
#asahi-dev · 2023-01-20 14:45
svenit uses the same chip as the internal one
#asahi-dev · 2023-01-20 14:44
svenyes
#asahi-dev · 2023-01-20 13:55
svenI think I'm just missing some magic poke somewhere. the adapter works fine under macos
#asahi-dev · 2023-01-20 13:53
svenI don't think I want to know ;)
#asahi-dev · 2023-01-20 13:52
sventhe ohci/ehci controllers in this old thunderbolt display work fine e.g.
#asahi-dev · 2023-01-20 13:52
svenit only happens with that bcm device so far
#asahi-dev · 2023-01-20 13:51
svenyeah, it sounds like something is still a bit broken. no idea what exactly though
#asahi-dev · 2023-01-20 13:49
sven(tg3 is the 15000+ lines broadcom driver)
#asahi-dev · 2023-01-20 13:49
sventhe bigger issue is that everything in that tg3 bar reads back as 0xffffffff
#asahi-dev · 2023-01-20 13:49
svenmaz: I figured out the latency issues, it was mostly me being dumb. I had to enable some quirks I didn't understand in tg3 to make the adapter work. one of them replaced every single register access with an indirect one through config space
#asahi-dev · 2023-01-20 13:05
svenI need to figure out how to that properly
#asahi-dev · 2023-01-20 13:05
svenwhich also completely breaks bus rescanning. I had to rescan like 5 times before everything was up :/
#asahi-dev · 2023-01-20 13:04
sventhat's via a USB adapter via my old apple thunderbolt display
#asahi-dev · 2023-01-20 13:04
sven64 bytes from fra24s04-in-f3.1e100.net (142.250.186.35): icmp_seq=2 ttl=115 time=17.0 ms
#asahi-dev · 2023-01-20 13:04
svenwell... maybe I'm just gonna blame Broadcom for this
#asahi-dev · 2023-01-20 12:39
svenI wouldn't know how
#asahi-dev · 2023-01-20 12:35
sventhe dma comment was just to say that the iommu finally works, I'm not sure what exactly causes that latency
#asahi-dev · 2023-01-20 12:35
svenMSIs
#asahi-dev · 2023-01-20 12:16
svenmy internet is sometimes pretty bad but it's not _that_ bad :D
#asahi-dev · 2023-01-20 12:15
sven64 bytes from fra16s52-in-f3.1e100.net (142.250.185.195): icmp_seq=1 ttl=115 time=141 ms
#asahi-dev · 2023-01-20 12:15
svenPING google.de (142.250.185.195) 56(84) bytes of data.
#asahi-dev · 2023-01-20 12:15
sven# ping google.de
#asahi-dev · 2023-01-20 12:15
svenping over thunderbolt with dma. for some reason there's a pretty high latency though:
#asahi-dev · 2023-01-20 11:05
svenand I thought I tracked that down to the probe dependency order
#asahi-dev · 2023-01-20 11:05
svenI vaguely remember making it working during boot by adding a sleep somewhere in tipd
#asahi-dev · 2023-01-20 11:04
svenhuh, true. never saw that here though
#asahi-dev · 2023-01-20 10:49
svenif you manage to have dcp up and running before tipd it'll work even when connected during boot iirc
#asahi-dev · 2023-01-20 10:48
svencompare how e.g. usb-role-switch or the typec-mux works to how that OOB event works
#asahi-dev · 2023-01-20 10:48
svenyeah, that's a probe dependency issue
#asahi-dev · 2023-01-19 21:13
svenyeah, I remember now
#asahi-dev · 2023-01-19 20:59
svenI don’t think it’s required if we just always configure it as usb3/DP just like macOS
#asahi-dev · 2023-01-19 20:59
svenI still want to look into preventing the phy from resetting when switching to altmode
#asahi-dev · 2023-01-19 20:58
sventhat should trigger that change
#asahi-dev · 2023-01-19 20:58
svenuh, do you have that DisplayPort altmode commit?
#asahi-dev · 2023-01-19 19:00
svenuhhhh… that shouldn’t happen
#asahi-dev · 2023-01-19 18:48
svenstill took two or three tries sometimes though
#asahi-dev · 2023-01-19 18:48
svenyeah, I think that was pretty stable for me as well
#asahi-dev · 2023-01-19 18:45
svenat least until I had a framebuffer
#asahi-dev · 2023-01-19 18:44
svenI remember it being very yanky
#asahi-dev · 2023-01-19 16:28
svenhuh
#asahi-dev · 2023-01-19 16:08
svenhuh, dp-altmode isn't work anymore?

plus ›