"So Many AI-Fueled Fixes" Means No New ARM64 KVM Features For Linux 7.2

The KVM virtualization-related changes were merged a few days ago for the ongoing Linux 7.2 kernel merge window. While there are a number of features/improvements for AMD and Intel virtualization as well as the likes of s390 and RISC-V, there aren't any new features on ARM64. The lack of ARM64 feature work this cycle is being attributed to "so many AI-fueled fixes" swamping the ARM Linux developers.

As I covered a week ago when it came to the ARM64 architecture changes for Linux 7.2 or there the lack of, AI/LLM patch craziness is impacting ARM64 Linux kernel development. That was in relation to the common ARM64 kernel code while it turns out to be a similar story too over in the KVM/virtualization space. The KVM pull request for Linux 7.2 noted:

"This is a bit of an odd merge window on the KVM/arm64 front. There is absolutely no new feature in the pull request. It is purely fixes, because it is simply becoming too hard to review new stuff when so many AI-fuelled fixes hit the list".

Besides ARM64 KVM fixes, over on the RISC-V side there is now support to batch G-stage TLB flushes, s390 now supports 2G hugepages, and there are a number of common KVM x86 code improvements.

Intel-specific work for Linux 7.2 KVM includes adding support for nested Mode-Based Execute Control (MBEC), various Trust Domain Extensions (TDX) improvements, and adapting to use the safe WRMSR API. On the AMD side there is now support for nested Guest Mode Execution Trap (GMET), support for virtualizing gPAT, support for virtualizing Host-Only/Guest-Only bits in the mediated PMU, and other improvements.

The full list of merged KVM changes for the Linux 7.2 kernel can be found via this pull request.