]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
kvm: x86: IA32_ARCH_CAPABILITIES is always supported
authorJim Mattson <jmattson@google.com>
Wed, 9 May 2018 21:29:35 +0000 (14:29 -0700)
committerBen Hutchings <ben@decadent.org.uk>
Tue, 13 Aug 2019 11:38:56 +0000 (12:38 +0100)
commit0824675ce7b01210b2e35b6f10b690c3f6ca0d93
tree6ff997c2e68f75aa523dc6cb3b381a0e8e30233d
parente62b3b79691dc02ebf32b2326f0d9509c7702365
kvm: x86: IA32_ARCH_CAPABILITIES is always supported

commit 1eaafe91a0df4157521b6417b3dd8430bf5f52f0 upstream.

If there is a possibility that a VM may migrate to a Skylake host,
then the hypervisor should report IA32_ARCH_CAPABILITIES.RSBA[bit 2]
as being set (future work, of course). This implies that
CPUID.(EAX=7,ECX=0):EDX.ARCH_CAPABILITIES[bit 29] should be
set. Therefore, kvm should report this CPUID bit as being supported
whether or not the host supports it.  Userspace is still free to clear
the bit if it chooses.

For more information on RSBA, see Intel's white paper, "Retpoline: A
Branch Target Injection Mitigation" (Document Number 337131-001),
currently available at https://bugzilla.kernel.org/show_bug.cgi?id=199511.

Since the IA32_ARCH_CAPABILITIES MSR is emulated in kvm, there is no
dependency on hardware support for this feature.

Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Fixes: 28c1c9fabf48 ("KVM/VMX: Emulate MSR_IA32_ARCH_CAPABILITIES")
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
[bwh: Backported to 3.16: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
arch/x86/kvm/cpuid.c