#
✕ nick:sven
#asahi-dev · 2023-03-14 12:41
sveneven the initial submission I did had that
#asahi-dev · 2023-03-14 12:41
svenyeah
#asahi-dev · 2023-03-14 12:38
sventhough I guess that helps both versions :)
#asahi-dev · 2023-03-14 12:38
sveniirc it was threaded because the original rtkit implementation used that as a hack
#asahi-dev · 2023-03-14 12:37
svenyou could also try to convert that threaded recv irq to a real irq handler
#asahi-dev · 2023-03-14 12:35
svenhuh
#asahi-dev · 2023-03-14 12:32
svendcp will full debug spew probably also won't overflow it anymore now
#asahi-dev · 2023-03-14 12:24
sveni would've expected those two or three context switches for each send to matter more but I guess the CPU is just too fast
#asahi-dev · 2023-03-14 12:24
svenstill a pity that you didn't get much performance improvements
#asahi-dev · 2023-03-11 13:44
svenhas been busy with $work that actually pays my bills as well :D
#asahi-dev · 2023-03-11 13:40
svenlovely :(
#asahi-dev · 2023-03-10 14:41
svennvme should actually work without DART. not that that helps much
#asahi-dev · 2023-03-08 10:11
svenAnd dptx is only present on dcpext there
#asahi-dev · 2023-03-08 10:11
svenmy comments only refer to m1 fwiw
#asahi-dev · 2023-03-08 09:17
svenI tried to debug that for days and then when I switched to the real iomfb it worked almost immediately
#asahi-dev · 2023-03-08 09:17
sveniirc I either never got modes or hotplug just didn’t go through
#asahi-dev · 2023-03-08 09:16
svenoh, dptx never worked correctly with the iboot interface for me fwiw
#asahi-dev · 2023-03-08 09:16
svenmakes sense since it needs to be able to route to atcphy as well
#asahi-dev · 2023-03-08 09:11
svenjust guessing here though since I only have this m1 ;)
#asahi-dev · 2023-03-08 09:11
svenand at least on m1 the dptx endpoints for DCP just don’t exist and that connectTo call sounds like the proper place to setup the routing
#asahi-dev · 2023-03-08 09:10
svenyeah, I just assumed that you set that output mux to „dp2hdmi“ and then it just works
#asahi-dev · 2023-03-08 08:54
svendoes DCP have the dptx endpoints? There’s that init connection call or however it’s called that also takes the target ATC as arguments
#asahi-dev · 2023-03-08 07:02
svenbut it does on M2 so there’s certainly something different
#asahi-dev · 2023-03-08 07:01
svenDCP doesn’t go through the crossbar on M1
#asahi-dev · 2023-03-04 18:34
svenI love how just everyone runs into things that are pretty broken :p
#asahi-dev · 2023-03-04 15:24
sven:D
#asahi-dev · 2023-03-04 15:23
svenif only everything they did wasn’t so broken
#asahi-dev · 2023-03-04 15:22
sven<3 synopsys
#asahi-dev · 2023-03-04 14:55
svenso you finally admit to writing DCP again? :>
#asahi-dev · 2023-03-04 13:50
svenI’m probably gonna bypass that typec altmode subsystem mess and just do what chrome-ec or however it’s called does and manually set the typec muxes
#asahi-dev · 2023-03-04 13:25
svenreally needs to finish display over type c support so that people don’t have to use displaylink :D
#asahi-dev · 2023-03-04 12:36
svensounds like a joke but that’s literally how acio can be reset
#asahi-dev · 2023-03-04 12:36
svenhave you tried turning the power on and off? :D
#asahi-dev · 2023-03-03 16:12
sventhough I dunno which part of that is specific to the usb4 controller
#asahi-dev · 2023-03-03 16:08
svenhttps://github.com/AsahiLinux/linux/blob/atcphy-20230219/drivers/pci/controller/pcie-apple.c#L541
#asahi-dev · 2023-03-03 16:06
svenjust write 0 to that ltssm reg and then either sleep for a bit or poll a bunch of ltssm status regs
#asahi-dev · 2023-03-03 14:51
sveniirc the power domain has to be on for the reset to work
#asahi-dev · 2023-03-03 14:45
svenhuh, for the usb4 controller it actually resets everything
#asahi-dev · 2023-03-03 14:41
sven*but
#asahi-dev · 2023-03-03 14:41
sventhere’s always a full pmgr reset as the last resort bht I’m fairly sure that does require tunables again
#asahi-dev · 2023-03-02 09:42
sven*clear
#asahi-dev · 2023-03-02 09:41
svenyeah, if you clean it more stuff seems to be reset (at least on usb4)
#asahi-dev · 2023-03-02 09:40
sveniirc that then disables the app clk as well which also disables the pcie dart
#asahi-dev · 2023-03-02 09:39
svenif you set/clear that bit in the MMIO at least on the usb4 one additional stuff seems to be reset
#asahi-dev · 2023-03-02 09:38
sveniirc port setup doesn’t actually set the bit in the MMIO, it just uses that gpio
#asahi-dev · 2023-03-02 09:27
sven(at least it’s possible to recover from that for the usb4 pcie)
#asahi-dev · 2023-03-02 09:27
svenit should be possible to recover from that by going through a perst cycle but that might need tunable to be reapplied :/
#asahi-dev · 2023-03-01 21:52
svenI guess no one could predict we’d be running pcie over very thin wires with crazy topologies when that was originally written ;)
#asahi-dev · 2023-03-01 21:51
sventhere’s also a big fat “rescan, remove or add pci device” lock
#asahi-dev · 2023-03-01 21:51
svenhotplug seems to have mostly been written with plugging in a single card into a slot

plus ›