This project demonstrates how HID over I2C can be used as a flexible protocol between a Linux SoC and an STM32 MCU. When enumerated by the kernel, the MCU will be treated as 3 separate virtual devices: a keyboard, a mouse and a raw HID device. The number and purpose of the virtual devices is defined by the descriptors sent by the MCU. The demo code sends CAPS_LOCK key events when the button on the development board is pressed, and lights up an LED when the CAPS_LOCK LED is activated from the host.
This code is built for and tested on an 32F072BDISCOVERY kit, as this is a low cost MCU line, and the board has a separate connector for I2C communication.
Porting this code within the STM32 product line is guaranteed to be minimal effort, since the implementation interacts with the hardware via the HAL library. All you need to do is replace the current STM32CubeMX configuration with your choice and generate the new cmake project with it.
Porting this code to another MCU type will require porting hal_i2c_slave.cpp to interact with the I2C and GPIO
FW of your choice of silicon.
You can easily extend the HID functionality by modifying the report descriptor and adapting the app code. Either find the existing application and usages in the HID usage tables, or define your own vendor-specific usage page(s) - even by adding them to your hid-usage-tables fork. The hid-rp library allows you to perform compile-time verification of the report descriptor, helping you avoid some mistakes. This verification isn't all-encompassing, and it relies on C++ exceptions. That's why there is a separate cmake target, that compiles the related source file for verification with exceptions enabled. Exceptions are disabled in the firmware to be flashed itself.
This project is tested with a Raspberry Pi 400, please refer to this guide on how to configure a Linux kernel to use HID over I2C.