KVM Changes Land In Linux 5.16: RISC-V Hypervisor Support, AMD PSF Control Bit
Last week the Kernel-based Virtual Machine (KVM) feature patches were sent out and subsequently merged for Linux 5.16.
Arguably most notable for KVM with Linux 5.16 is the introduction of the RISC-V architecture code for the Kernel-based Virtual Machine. The KVM RISC-V hypervisor support depends upon the RISC-V ISA's hypervisor extension, which was recently frozen. Now to wait for RISC-V processors to market that are performant and supporting the extension...
Over on the x86 (x86_64) side, there are various minor improvements to the code. One user-facing change for AMD SVM is support for turning off AMD PSF to KVM guests. PSF is Predictive Store Forwarding with Zen 3 and what was disclosed earlier this year as possible security implications. The AMD PSF toggle patch to outright disable it has yet to be merged to the Linux kernel while for 5.16 is now the patch to at least allowing the PSF control bit to be toggled by guest kernels.
KVM on the Arm front has continued its enablement around protected virtual machines. There are also workarounds for the Apple M1 with a broke VGIC implementation. The KVM s390 code has also seen various improvements, including work around handling for lazy destroying of secure VMs.
The full list of KVM patches so far for Linux 5.16 can be found via this mailing list post. A second batch of KVM changes for this new kernel are likely to be posted this week.
Arguably most notable for KVM with Linux 5.16 is the introduction of the RISC-V architecture code for the Kernel-based Virtual Machine. The KVM RISC-V hypervisor support depends upon the RISC-V ISA's hypervisor extension, which was recently frozen. Now to wait for RISC-V processors to market that are performant and supporting the extension...
Over on the x86 (x86_64) side, there are various minor improvements to the code. One user-facing change for AMD SVM is support for turning off AMD PSF to KVM guests. PSF is Predictive Store Forwarding with Zen 3 and what was disclosed earlier this year as possible security implications. The AMD PSF toggle patch to outright disable it has yet to be merged to the Linux kernel while for 5.16 is now the patch to at least allowing the PSF control bit to be toggled by guest kernels.
KVM on the Arm front has continued its enablement around protected virtual machines. There are also workarounds for the Apple M1 with a broke VGIC implementation. The KVM s390 code has also seen various improvements, including work around handling for lazy destroying of secure VMs.
The full list of KVM patches so far for Linux 5.16 can be found via this mailing list post. A second batch of KVM changes for this new kernel are likely to be posted this week.
1 Comment