#asahi-dev · 2022-11-22 10:50
svenbut there's no good way to prevent that :(#asahi-dev · 2022-11-22 10:50
svendwc3 will fail to probe if it sees a phy = <...> reference and there's no phy available#asahi-dev · 2022-11-22 10:50
svenunfortunately not inside linux#asahi-dev · 2022-11-22 10:27
svenyeah, it's much more likely xnu is just less efficient somewhere along the rest of the way#asahi-dev · 2022-11-22 10:25
svenI do configure atcphy to usb3-only while xnu always uses usb3+dp unless you coerce it but I really doubt that makes a difference#asahi-dev · 2022-11-22 10:24
svenunfortunately atcphy probably can't take credit for that since it's likely just something more efficient inside dwc3 or vfs or whatever :D#asahi-dev · 2022-11-22 10:13
svensounds good to me#asahi-dev · 2022-11-22 08:29
svenI’ve switched to exponential increase, so *2 #asahi-dev · 2022-11-22 08:28
sven^— exhibit a why I don’t publish it #asahi-dev · 2022-11-22 08:17
sven*havent #asahi-dev · 2022-11-22 08:17
svenit’s quite neat actually. I just haben rebased it because people kept using it and preparing it for end users despite me telling them not so #asahi-dev · 2022-11-22 08:16
svenphire: the 4K patch is limited to dma-iommu #asahi-dev · 2022-11-22 08:16
svenI guess the gpu mmu only does 16k as well?#asahi-dev · 2022-11-22 08:09
svenfor me it was mostly in those cases but I haven’t even used this with anything more than a tty yet#asahi-dev · 2022-11-22 08:03
svenand you get to fix lots of bugs#asahi-dev · 2022-11-22 08:03
svenit’s gonna be quite horrible fwiw#asahi-dev · 2022-11-22 07:53
svenand I’m kinda afraid that someone will be jannau and me #asahi-dev · 2022-11-22 07:53
svensomeone needs to start daily driving that thing to get it stable#asahi-dev · 2022-11-22 07:52
svenyeah, I’d just remove that for now #asahi-dev · 2022-11-22 07:49
svenmarcan: uh, yes. dcpext is still highly unstable #asahi-dev · 2022-11-21 21:03
svenit also doesn’t matter much yet since the only issue right now is that you just can’t have the display connected at boot time if you want it to work reliably #asahi-dev · 2022-11-21 21:03
svenhaven’t really thought about it#asahi-dev · 2022-11-21 21:01
svenand caching the callback if drm probes first so that it just doesn’t matter #asahi-dev · 2022-11-21 21:01
svenor maybe just caching the state if tipd probes first #asahi-dev · 2022-11-21 21:00
svenprobably similar to usb-role-switch #asahi-dev · 2022-11-21 20:58
svenI still have to check and see if I can hack dual display support but I actually don’t have two displays around right now#asahi-dev · 2022-11-21 20:57
svenyeah, that case is always going to be an issue #asahi-dev · 2022-11-21 20:52
svenwe’ll also have to rewrite the oob hpd thingy so that there’s no race between tipd and drm probe#asahi-dev · 2022-11-21 20:50
svenif we get a hotplug event can we just assign a connector to a dcpext and bring it up enough to get the edid and only then let drm know?#asahi-dev · 2022-11-21 20:50
svenhrm, so we’ll have three connectors per dcpext #asahi-dev · 2022-11-21 20:12
svenalright, just the lockdep issue remaining but otherwise altmode now correctly works with the typec subsystem :)#asahi-dev · 2022-11-21 18:55
svenapparently the supported svids need to listed because there can be different muxes for different altmodes?!#asahi-dev · 2022-11-21 18:54
svenhuh, so this explains why I didn’t see any mux set events when relying on typec to trigger to switch to altmode, https://github.com/torvalds/linux/blob/eb7081409f94a9a8608593d0fb63a1aa3d6f95d8/drivers/usb/typec/mux.c#L296#asahi-dev · 2022-11-21 16:36
svenit's probably not that bad anyway once tbt is up (famous last words and all that)#asahi-dev · 2022-11-21 16:35
svenI mostly block on dp-over-tbt because I don't want to switch setups between $work and my macbook#asahi-dev · 2022-11-21 16:30
sveni might just suffer for a while replugging cables and use this usb3+dp hub with my macbook air#asahi-dev · 2022-11-21 16:23
svenit kinda works. just need to ignore lockdep complaining about a possible deadlock and the missing calls to the usb role switch which makes atcphy WARN#asahi-dev · 2022-11-21 16:22
sven[C]#asahi-dev · 2022-11-21 16:22
sven$ cat ./devices/platform/soc/235010000.i2c/i2c-0/0-003f/typec/port1/port1-partner/port1-partner.0/displayport/pin_assignment#asahi-dev · 2022-11-21 16:21
svenjust need to finish this small adventure into USB PD and drivers/usb/typec and clean the DP altmode stuff up before#asahi-dev · 2022-11-21 16:20
svenbut I need tbt itself up before that anyway#asahi-dev · 2022-11-21 16:20
svendp-over-tbt is probably the biggest part that I still have no idea about#asahi-dev · 2022-11-21 16:20
svenjust need to find some annoying bug somewhere that makes part of the MMIO still SError#asahi-dev · 2022-11-21 16:19
svenI think I know how to bring TB up#asahi-dev · 2022-11-21 16:19
svensame situation here tbh#asahi-dev · 2022-11-21 16:19
svenno worries, i'm on it :D#asahi-dev · 2022-11-21 16:06
sven*any#asahi-dev · 2022-11-21 16:06
svenshould be fine. anything updates I have are unstable at best#asahi-dev · 2022-11-21 14:52
svenand apparently the author never followed up anyway *shrug*#asahi-dev · 2022-11-21 14:52
sventhey both do the same thing, my patch is just the minimal possible change because the more complex rework confused people on the ML