#asahi-dev · 2022-03-20 19:35
svencan you show the full dmesg preferably with that trace_event stuff enabed?#asahi-dev · 2022-03-20 19:35
svenweird#asahi-dev · 2022-03-20 19:30
svenit should just fall back to usb2-only#asahi-dev · 2022-03-20 19:29
svenok. it still shouldn't fail with "failed to reinitialize core" though#asahi-dev · 2022-03-20 19:28
svenso far i've only been using the one that doesn't have the USB PD debug crap#asahi-dev · 2022-03-20 19:28
svenand did you try both ports?#asahi-dev · 2022-03-20 19:27
svenas bootargs#asahi-dev · 2022-03-20 19:27
sventp_printk trace_event=appletypecphy:*#asahi-dev · 2022-03-20 19:26
svenand can you run it with tracing enabled and show me the dmesg#asahi-dev · 2022-03-20 19:26
sveni assume you merged the mini patch?#asahi-dev · 2022-03-20 19:26
sventhat's my machine as well where it just works :D#asahi-dev · 2022-03-20 19:26
svenhm :/#asahi-dev · 2022-03-20 19:26
svenm1 or m1 pro/max?#asahi-dev · 2022-03-20 17:44
svenlol#asahi-dev · 2022-03-20 17:30
svenohh… right, there’s a disp and a DCP dart#asahi-dev · 2022-03-20 17:25
svenbut that’s still weird that they don’t map it#asahi-dev · 2022-03-20 17:25
sveni guess xnu can use that property in the dcpext node to map the firmware data section#asahi-dev · 2022-03-20 17:25
svenis it still locked?#asahi-dev · 2022-03-20 17:25
svenhuh#asahi-dev · 2022-03-20 16:48
svenso if you do that for the dpphy part as well it’ll probably work#asahi-dev · 2022-03-20 16:48
svenwhat I started with was literally the output of macOS debug serial + the MMIO trace copy and pasted into the driver :D#asahi-dev · 2022-03-20 16:48
svenand then there are some missing pokes that macOS does to bring up what it calls “dpphy” but they are obvious from a hv trace #asahi-dev · 2022-03-20 16:47
svenso all you need to do is set atcphy->target_mode to usb3_dp#asahi-dev · 2022-03-20 16:47
svenjannau: let’s maybe move this here#asahi-dev · 2022-03-20 13:48
svenjannau: nice!#asahi-dev · 2022-03-20 13:35
sveni'm still confused why this makes a difference :/#asahi-dev · 2022-03-20 13:33
svenyeah, readl_relaxed should be much cheaper than disabling irqs locally. but i doubt we get more iops#asahi-dev · 2022-03-20 13:31
svenhttps://github.com/AsahiLinux/docs/wiki/HW%3AARM-System-Registers#asahi-dev · 2022-03-20 13:31
svennot just me :-)#asahi-dev · 2022-03-20 13:29
sveni just don't know _why_ that matters#asahi-dev · 2022-03-20 13:29
svenno, the writel should ensure that the memcpy is visible to the device. the readl_relaxed afterwards just ensures that the writel is visible#asahi-dev · 2022-03-20 13:27
svenmaz: ^-- looks like you were correct, it's somehow ordering related. i'm still as confused as before though. #asahi-dev · 2022-03-20 13:23
svenyup#asahi-dev · 2022-03-20 13:21
sveninteresting. can you try readl_relaxed instead of readl as well?#asahi-dev · 2022-03-20 13:15
svenmissing a ; at the end but yes#asahi-dev · 2022-03-20 13:13
svenyes#asahi-dev · 2022-03-20 13:12
svencan you check what happens if you add that readl I mentioned above?#asahi-dev · 2022-03-20 13:12
svenSo that’s just without the lock, right?#asahi-dev · 2022-03-20 12:37
svenGlanzmann: can you try what happens when you have no spinlock but add a readl(q->sq_db); just after the writel(tag, q->sq_db);?#asahi-dev · 2022-03-20 12:32
svenyeah, I can’t reproduce that on the mini#asahi-dev · 2022-03-20 12:32
svenah. no. I couldn’t even reproduce it manually inside m1n1#asahi-dev · 2022-03-20 12:31
sven*days#asahi-dev · 2022-03-20 12:31
svenwaaait…. maybe I could never reproduce it because I’m always running under the hv these daya#asahi-dev · 2022-03-20 12:30
sventhough I fear that might slow everything down too much#asahi-dev · 2022-03-20 12:30
svenwould be interesting to see if it still happens when running under the m1n1 hv and trace_nvme.py#asahi-dev · 2022-03-20 12:29
svenweird. I can’t reproduce it on my mini at all#asahi-dev · 2022-03-20 12:26
svenis that on a m1 or a m1 max/pro?#asahi-dev · 2022-03-20 12:08
svenshould be just passed along#asahi-dev · 2022-03-20 12:07
svenyes