The Kernel-based Virtual Machine changes were recently merged for the Linux 7.1 merge window for further enhancing KVM as this important piece of the open-source virtualization stack.

KVM on the Arm side adds "very experimental" support for pKVM protected guests. With pKVM protected guests, pages are unmapped from the host as they are faulted into the guest and then shared back from the guest using pKVM hypercalls. This pKVM protected guests support is working toward full isolation from the host. For now the support is highly experimental (it will taint your kernel when launching such a guest) and requires the kernel be built with the new CONFIG_ARM_PVKM_GUEST option and launching with the "kvm-arm.mode=protected" option.

KVM on s390 meanwhile added support for ESA 31-bit guests inside nested hypervisors.

KVM on x86 now supports [1]advertising support for AVX-512 BMM for the Bit Matrix Multiply instructions coming with upcoming AMD Zen 6 processors.

Plus there are some minor AMD SEV-SNP changes, various hardening improvements, and other minor changes.

The full list of KVM changes merged for Linux 7.1 can be found via [2]this pull .



[1] https://www.phoronix.com/news/Linux-7.1-KVM-AVX-512-BMM

[2] https://lore.kernel.org/lkml/20260416164526.171467-1-pbonzini@redhat.com/