> The kernel itself can't include anything requiring floating point maths, This sounds like a limitation derived from ancient architectures and platforms. What

> The kernel itself can't include anything requiring floating point maths, This sounds like a limitation derived from ancient architectures and platforms. What's the barrier to allowing FP in kernel modules for platforms that are guaranteed to have FP support? Is it too much work for the kernel to save and restore FP registers and modes?