#asahi-dev · 2023-01-31 11:00
sventrue#asahi-dev · 2023-01-31 10:09
svenah, no, i'm just dumb and accidentally reverted the hotplug fix while rebasing *sigh*#asahi-dev · 2023-01-31 09:12
sven... and here I thought thunderbolt was stable but I just got another obscure assert_id=whatever crash after the first connection :(#asahi-dev · 2023-01-30 18:45
svenshould get a m2#asahi-dev · 2023-01-30 17:21
svenlol :D#asahi-dev · 2023-01-28 13:06
svenyeah, the usb4 pcie thing require a different init sequence and doesn't even have that reset gpio e.g.#asahi-dev · 2023-01-28 13:06
svenlol :D#asahi-dev · 2023-01-28 13:05
svenmarcan: kettenis: so I guess for PCIe apple,usb4-pcie then? (plus the usual tXXXX-usb4-pcie)#asahi-dev · 2023-01-28 12:16
svenluckily we have mailing lists for that *sigh*#asahi-dev · 2023-01-28 12:08
svenit just happens to work for them because m1n1 already enables that and leaves it on#asahi-dev · 2023-01-28 12:08
sventhe normal pcie darts also depends on the port's APCIE_PORT_APPCLK_EN to be set fwiw#asahi-dev · 2023-01-28 12:02
svenyeah.. there's the thunderbolt core changes, the PCIe changes which will depend on them due to that OOB tunnel notification and the USB changes that probably also need the same OOB notification, probably also the PHY changes since it looks like I'll need a PHY_TYPE_USB4 used in both thunderbolt and the phy itself :/#asahi-dev · 2023-01-28 11:58
svenall the usb stuff is going to be fun anyway a cycle or two from now. with all its dependencies it'll probably be like 30 patches that all have to go through a single tree#asahi-dev · 2023-01-28 11:57
svennothing from my side#asahi-dev · 2023-01-28 11:54
svenlol :D#asahi-dev · 2023-01-28 11:54
svenjoro is the iommu maintainer, robin does almost all the reviews ;)#asahi-dev · 2023-01-28 11:50
svenall that's missing is those DPIN things which aren't that many MMIO pokes and some glue for dcpext#asahi-dev · 2023-01-28 11:49
sventhe NHI part already just works afaict#asahi-dev · 2023-01-28 11:49
svenbut DP-over-USB4 doesn't look too bad#asahi-dev · 2023-01-28 11:49
svenmostly separate#asahi-dev · 2023-01-28 11:49
svenand hopefully $work has also delivered my new dock by then so that I can also test usb tunneling#asahi-dev · 2023-01-28 11:48
sveni hope to remove all the annoying duct tape that still holds usb4 together sometime in the next weeks#asahi-dev · 2023-01-28 11:47
sventrue :)#asahi-dev · 2023-01-28 11:47
svendoesn't matter all that much since there's no way thunderbolt is ready this cycle :D#asahi-dev · 2023-01-28 11:47
svensure#asahi-dev · 2023-01-28 11:41
svenI haven't found anything in the DART_PARAMS that encode that sid count#asahi-dev · 2023-01-28 11:41
svenyeah, they use that sid-count#asahi-dev · 2023-01-28 11:21
svenlooks like I even have t8103-usb4-dart in the commit message *sigh*#asahi-dev · 2023-01-28 11:21
svendon't care either way :)#asahi-dev · 2023-01-28 10:58
svenand is it ever too early for a beer? ;)#asahi-dev · 2023-01-28 10:58
svennice!#asahi-dev · 2023-01-27 20:57
svenbut i'll eventually get to it#asahi-dev · 2023-01-27 20:57
svenI was hoping to volunteer someone else for that :D#asahi-dev · 2023-01-27 20:52
svenespecially with the altmode and usb4 transitions we have now#asahi-dev · 2023-01-27 20:52
svenI still think we should try to just write a custom dwc3 driver that mostly calls into the core driver but takes care of the shutdown/restart itself based on typec-mux instead of role-switch#asahi-dev · 2023-01-27 20:48
svenatcphy also changed a bit for thunderbolt fwiw#asahi-dev · 2023-01-27 20:48
svenah, nice! I just want to trace DPIN in a bit because I don't feel like getting rid of all the thunderbolt duct tape just yet ;)#asahi-dev · 2023-01-27 20:39
sveni just see a few register writes when plugging in a thunderbolt display#asahi-dev · 2023-01-27 20:38
svenalright, the DPIN thing dosn't look too bad#asahi-dev · 2023-01-26 19:24
sven~30 hotplug events and nothing broke :)#asahi-dev · 2023-01-26 19:24
svenhah, looks like it’s stable with that notification now!#asahi-dev · 2023-01-26 19:02
svenand there are no interrupts triggered in that state unfortunately #asahi-dev · 2023-01-26 19:02
svenif pcie isn’t kept in “tunnel reset” until the thunderbolt tunnel is ready it can crash the firmware #asahi-dev · 2023-01-26 19:01
svenhrm, looks like we’ll need a OOB notification from nhi to pcie after all to make complex tunnels (like the old Thunderbolt Display) reliable #asahi-dev · 2023-01-26 16:03
svenand it keeps working if I switch the cable type or orientation. so, uh, I guess that's actually a "write N times here to crash the co-processor" register? :P#asahi-dev · 2023-01-26 16:02
svenif I just don't do that more than once and write random parameters in there it doesn't seem to crash after hotplug anymore...#asahi-dev · 2023-01-26 16:02
svenokay, what the hell Apple. so there's a vendor specific "cable information" register that I was writing to on hotplug with bits having specific meaning (e.g. cable orientation, etc.)#asahi-dev · 2023-01-24 22:27
svenstill no idea why it occasionally still dies even on the first connection though