Respond to usb_control_msg

[meeuw]

Hi, I'd like to emulate Razer HID devices (keyboards, mice) to test OpenRazer (an open source linux driver).

I've adapted the hid example (changed the report descriptor and added 5 extra functionfs.HIDFunctions) so it matches the real hardware Razer keyboard I have. After that the kernel module want to communicate with it (through dummy_hcd) 🥳 .

Now it immediately fails on the call of usb_control_msg: https://github.com/openrazer/openrazer/blob/fe62bdadcfdb05fc9768cf3a38e954ae7cbb4991/driver/razercommon.c#L35

Can you point me in the right direction how to be able to handle/respond to an usb_control_msg call in a Python script using python-functionfs? Some pseudocode or related example would be great.

[vpelletier]

You are going to want to override the onSetup method.

In the code you linked, I see bmRequestType contains USB_RECIP_INTERFACE, which is good news: you can implement control requests targetted at this recipient at the function level. Requests targetting USB_RECIP_ENDPOINT are also possible (please check the onSetup implementation linked above). I am less sure about other recipients (just because I have not played with setup handling recently), especially USB_RECIP_DEVICE will likely be reserved by the kernel for its own handling, and it will only understand standard requests, and Microsoft's magic string descriptor stuff.

Here is a realistic implementation I did that may help explain what I mean in the docstring linked above. including the super call at the end.

[meeuw]

Thank you very much, for your answer and this great Python package.

I think I can just override setHIDReport in this case.

I wish I understood FunctionFS (and USB) a bit better, do you know of any more elaborate documentation than https://docs.kernel.org/usb/functionfs.html ?

[vpelletier]

I think I can just override setHIDReport in this case.

Oh, yes, this is even better.

About documentation on USB, I remember reading a nice simplification of the (very long, but thankfully publicly available) USB specification, and I belive it was BeyondLogic's USB in a NutShell. USB Protocols and USB Descriptors should be the most relevant (probably the latter more than the former).

About documentation on Linux's FunctionFS, I do not clearly remember what I read in order to make sense of the userland API the kernel exposes. I remember reading the ffs examples from the kernel's source, especially aio_simple.

[meeuw]

Thanks a lot, I guess that will do, I hope I can contribute soon to this project!