Linux Microcode Loading For x86 32-bit CPUs Being Cleaned Up & Corrected

Continuing to support x86 32-bit processors with the mainline Linux kernel continues to be a maintenance burden and uncovering ugly bits of code that are seldom touched. The latest work is on fixing up the 32-bit early microcode loading code so that it's more robust and actually correct.

Thomas Gleixner of Intel-owned Linutronix recently uncovered that the 32-bit early loading mechanism for the CPU microcode was home to some problems: "Aside of being a source of code horror and debug trouble, it also turned out that this is fundamentally broken vs. stackprotector and tracing."

The early Intel Atom N270 32-bit days with netbooks...

Part of this 32-bit microcode loading mess was due to erratum of early Intel Atom CPUs. In any event Gleixner tracked down the technical reasons leading to this mess and has now cleaned up the 32-bit early loading microcode handling.

He's sent out this patch series for making the x86 microcode 32-bit early loading more robust and correct. As he's already queued it into TIP.git's x86/microcode branch it's likely he will be submitting these fixes/improvements for the upcoming Linux 6.6 merge window.

It was just last month that Gleixner had suggested dropping x86 32-bit support from the Linux kernel given that they are "just museum pieces" though it doesn't look like that will happen in the very near future.