KVM: nVMX: Skip IBPB when switching between vmcs01 and vmcs02
authorSean Christopherson <sean.j.christopherson@intel.com>
Fri, 1 May 2020 16:31:17 +0000 (09:31 -0700)
committerPaolo Bonzini <pbonzini@redhat.com>
Wed, 13 May 2020 16:15:02 +0000 (12:15 -0400)
commit5c911beff20aa8639e7a1f28988736c13e03ed54
tree517e9191ca552bbc01125fba3ede2f8b9f1096c4
parentf27ad73a6ef791731fa75ca183d5c64a930b9cc1
KVM: nVMX: Skip IBPB when switching between vmcs01 and vmcs02

Skip the Indirect Branch Prediction Barrier that is triggered on a VMCS
switch when running with spectre_v2_user=on/auto if the switch is
between two VMCSes in the same guest, i.e. between vmcs01 and vmcs02.
The IBPB is intended to prevent one guest from attacking another, which
is unnecessary in the nested case as it's the same guest from KVM's
perspective.

This all but eliminates the overhead observed for nested VMX transitions
when running with CONFIG_RETPOLINE=y and spectre_v2_user=on/auto, which
can be significant, e.g. roughly 3x on current systems.

Reported-by: Alexander Graf <graf@amazon.com>
Cc: KarimAllah Raslan <karahmed@amazon.de>
Cc: stable@vger.kernel.org
Fixes: 15d45071523d ("KVM/x86: Add IBPB support")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20200501163117.4655-1-sean.j.christopherson@intel.com>
[Invert direction of bool argument. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/vmx/nested.c
arch/x86/kvm/vmx/vmx.c
arch/x86/kvm/vmx/vmx.h