Raw_hid corrupted or enumerated improperly on STM32F072 when used with a TI 8020B USB hub controller (bounty available)

[TerryMathews]

VIA freezes on a keyboard designed around the STM32F072 CPU (TKC GodSpeed75) when used in conjunction with the USB hub designed to fit it, our custom TI 8020B USB 3.0 hub.

Describe the Bug

• TKC GodSpeed75 is a keyboard PCB that utilizes the STM32F072 CPU running QMK and has a custom-designed USB hub based around the TI 8020B 2-port USB hub
• This combination causes VIA to hang upon connection
• The hub works correctly when used in conjunction with the TKC M0lly PCB, a board designed around the AT90USB1286 MCU
• The GodSpeed75 works in VIA when connected directly via USB cable
• The GodSpeed75 works correctly absent VIA support when connected via its USB hub; RGB etc works as expected Conclusion: There is an issue in the STM32 codebase causing an issue with VIA support in conjunction with that USB hub; likely exists in the USB enumeration code. There have been a few other similar STM32 issues related to sleep and wake states
• We have eliminated QMK version as a factor - I have tested that this bug exists clear back to May 2020.

Additional Context

There is a $500 bug bounty available, see attached. Approved in advance by @jackhumbert.

GodSpeed75 Bug Bounty.pdf

[TerryMathews]

After a lot of experimentation, I have determined that this problem is intertwined with the AMD Ryzen USB issues that have been getting more and more press (see: https://www.anandtech.com/show/16554/amd-set-to-roll-out-agesa-firmware-usb-fix-for-ryzen-coming-april)

I have determined that the problem does not occur when using a powered USB hub, at least with the one I have. I have ordered another to test with. I still believe that this issue also lies within the QMK code for STM32 as the behavior does not occur on AVR-based boards however this firmware update by AMD referenced in the attached article may resolve the issue as well.