P33M

I think I've found the cause - there's a large number of gotchas when a URB is removed that is _ahead_ of the ring's current dequeue pointer, i.e. the xHC...

@Jorl17 @mhartoft can you try compiling from this branch: https://github.com/P33M/linux/tree/trb_leak_fix and testing with your hardware. Here, on either the vl805 or dwc3 controllers on a Pi 4, I no longer...

xhci_hcd is built-in, so /boot/kernelN.img has to be updated, and the machine has to be rebooted. It goes without saying that you shouldn't test this on a board you care...

The bug is exercised by a driver indirectly manipulating endpoint ring state, and is not limited to a particular class of driver. For example one that frequently issues queued commands...

The reset value of the GPIO pad output drive defaults to 0, thus selecting GPIO as output without first poking a 1 at GPIO_SET will result in the pin being...

It goes without saying that we don't have one of those devices. This is a full-speed device so is accessed behind a transaction translator. The very first transaction to the...

The disconnected mouse/keyboard are a side-effect of Linux trying to recover the device as evidenced by "attempt power cycle" messages in dmesg. What is the boot order of the system?...

The command response code is the same as is seen in Linux - split transaction error. The difference in behaviour between a replug on Linux and a replug with the...

In that case that pretty much exhausts all the tunable options for altering device probe behaviour. Does the device work when plugged into a Linux x86 PC with xHCI controllers?

Without representative hardware there is little chance of a diagnosis. If you have a unit spare that you can loan, then please email info > raspberrypi.com quoting this issue link...