AM6548: Build Floating point driver - Linux Kernel v4.19

Hi Johni,

Here's the response from our expert. In summary it's easy and not right to enable floating operations in Linux kernel.

For the current 64-bit kernel we don't know if there is a 'good' way to do this.

The basic issue is kernel space typically doesn't maintain a floating point context. Often its only added to a user process in a lazy way so as to keep context switch costs down.  Only user processes who use floats pay the cost.  The mechanism to do the lazy usage typically has the float feature disabled until first use.  At first float use an exception happens where the kernel then allocates and assigns resources.

For something in the kernel to use floats, it needs to save and restore any floating point context it touches.  It may have to enable and disable the unit. Further, some worry needs to be given to preemption of system code (kernel entities and interrupt handlers).  Even if a context were created for the kernel how would multiple interrupt handlers leveraged by different processes share this."

As told earlier we do not recommend float usage in Linux kernel drivers. Hope you can close the issue

Best regards,

Keerthy