As soon as the kernel will support the XSAVE extensions in Skylake processors, we will want both userspace and the hypervisor to run guests without showing any trace of the new features (because support for them in the hypervisor will come later). This series does exactly this. Patches 1 and 3 ensures that userspace does not get access to features not supported in the hypervisor. Patch 2 does the same for the guest. The effect of these patches is already visible together with the corresponding QEMU patches and a command line such as "-cpu SandyBridge,-avx". Previously, the guest could still enable AVX via xsetbv, and state would be sent by QEMU during migration. These patches make sure that disabling AVX will really prevent the guest from using it, which helps if you have to deal with migration from newer versions of QEMU and the kernel to older versions. The QEMU patches set up the guest's 0xd CPUID leaf, which also ensures that the kernel has the necessary information. Paolo v2->v3: rename supported_xcr0 field to guest_supported_xcr0 rename xstate_size field to guest_xstate_size do not hardcode XSTATE_FPSSE in guest_supported_xcr0 handle missing 0xd leaf Paolo Bonzini (3): KVM: x86: mask unsupported XSAVE entries from leaf 0Dh index 0 KVM: x86: prevent setting unsupported XSAVE states KVM: x86: only copy XSAVE state for the supported features arch/x86/include/asm/kvm_host.h | 2 ++ arch/x86/kvm/cpuid.c | 36 +++++++++++++++++++++++++++++++++++- arch/x86/kvm/x86.c | 29 ++++++++++++++++++++++------- arch/x86/kvm/x86.h | 1 + 4 files changed, 60 insertions(+), 8 deletions(-) -- 1.8.3.1 -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
