x32 Shadow Stacks, Locking Optimizations, Intel VFM & Other x86 Changes For Linux 6.10

Today marks the first official day of the Linux 6.10 merge window. Among the horde of pull requests sent out today were the numerous x86 pull requests of material that's been queuing up in TIP.git of which there are many changes benefiting both Intel and AMD.

First up, the x86/shstk pull is enabling shadow stacks for the x32 architecture. The x32 ABI as a reminder provides the benefits of x86_64 (and requires x86_64) while using 32-bit pointers. Back in Linux 6.6 the x86_64 Shadow Stack support was merged while more recently has been enabling of Shadow Stacks for x32 and that is now being merged for Linux 6.10 for those still using x32. That pull request does acknowledge it's rare seeing new 32-bit Linux kernel features:

"While we normally don't do such feature-enabling on 32-bit kernels anymore, this change is small, straightforward & tested on upstream glibc."

The x86/cpu changes were also sent out today by Ingo Molnar. That pull does bring the rework for Intel CPU model handling with future Intel CPUs moving past the "Family 6" that's long been used for Intel processors. Now with Linux 6.10+, the Intel CPUs rely on a "VFM" value for the Vendor/Family/Model combined into a single value to make it easier introducing new Intel CPU models and families.

The x86/cpu pull for Linux 6.10 also brings improved AMD CPU topology parsing for CCD/CCX details. There is also optimizing the NUMA allocation layout of more per-CPU data structures, improved the AMD Erratum 1386 workaround, clearing Total Memory Encryption (TME) for the /proc/cpuinfo output when disabled by the firmware, and other fixes.

With the locking changes submitted for Linux 6.10 there are more than one dozen micro-optimizations to the atomic and spinlock code benefiting x86-based processors.

The other x86 pull requests sent out so far today for Linux 6.10 are mostly bug fixes and other smaller changes. Linux 6.10 should be another feature-packed kernel cycle.