08:18 <mort_> I'm sometimes seeing qualcomm venus bugs like this: https://p.mort.coffee/kMG

08:18 <mort_> Deterministically, with certain phones

08:18 <mort_> Any idea?

09:35 <mort_> Okay, I figured it out: the Venus decoder really really didn't like that the frame rate was expressed as the fraction 60000000/2142857

21:05 <matthewcroughan_> ping

21:05 <matthewcroughan_> If I hardcode `dr_mode = "host"` my issue with https://github.com/sc7280-mainline/linux/blob/2a3775318eb42e97c970b6e41d1f290f0a76542a/arch/arm64/boot/dts/qcom/qcm6490-shift-otter.dts#L1041C2-L1041C18 goes away.

21:06 <matthewcroughan_> In summary, my issue is that when otg is set instead of hardcoding to host, only the root hub shows up, and not the device I plug into the hub. When otg is set instead of host, `lsusb` prints empty, whereas when host is set, it prints the root hubs properly.

21:07 <matthewcroughan_> Without anything plugged in: https://termbin.com/nqjpx

21:07 <matthewcroughan_> With something plugged in: https://termbin.com/9d71

21:08 <matthewcroughan_> When otg is enabled, lsusb prints empty, until you plug something in. Then when you plug something in, only the root hub shows up. With host mode, everything works properly. However, hardcoding host prevents display port over Type C alt mode from working.

21:10 <matthewcroughan_> When otg is set, it is as if the act of plugging them in, enables the dwc controller and root hub, but it fails to enumerate and discover the rest of the USB devices on the bus in the glasses, perhaps a power issue?

21:11 <matthewcroughan_> does that imply an issue in the role switching driver? Or maybe even a missing device tree property that informs the driver correctly? https://www.kernel.org/doc/Documentation/devicetree/bindings/usb/generic.txt Like maybe role-switch-default-mode should be host, rather than peripheral which is its default if not set