Hardware error reporting, error handling and debugging

This BOF is intended to cover the current gaps in handling hardware error conditions.

Many hardware failures result in a NMI or Machine Check. Although they rarely occur, when they do, it can be devastating in some environments if they go unchecked or ignored. The kernel currently logs these conditions and moves on. Data integrity issues can arise when this is the only method employed. Instead, the kernel should attempt to handle these conditions intelligently. A presentation on current error handling in some architectures will be followed with a discussion.

The goals of the discussion are to uncover ways to expand the error handling to other architectures, increase its robustness and determine what features are available in hardware that are not yet utilized.