The Linux Kernel Dev Staging Area Now Rejects AI-Generated Patches
Unless a kernel patch is a valid security fix, it will be rejected if it was created using AI.
Firstly, it's important to understand that the developer staging area is a place for new kernel developers to get involved in Linux kernel development. "It contains loads of 'low-hanging fruit' with regards to code cleanups and API changes, perfect for new developers to learn the process in a safe and friendly way," said Greg Kroah-Hartman.
Because of that specific purpose, Hartman has made it clear that anyone "attempting to use an LLM [Large Language Model] to clean up or 'fix' any code in drivers/staging/ is EXPLICITLY defeating the whole purpose of it existing in the first place." To that end, Kroah-Hartman will automatically reject any patches that are generated using an LLM in the drives/staging/subsystem -- with one exception.
Kroah-Hartman states that LLMs are quite good at locating suspected "security issues" in the kernel code. However, even with the improvements, he says that "at least 1/3 of the results they generate are flat out wrong or harmful." At the same time, Kroah-Hartman informs developers that if their LLM-found fix in a drivers/staging file is valid, it's okay to submit it. However, he warns, "you must have first tested it on the actual hardware for the driver and described how you have done so, in order for us to be willing to take the change."