KVM: SVM: move MSR_IA32_SPEC_CTRL save/restore to assembly
authorPaolo Bonzini <pbonzini@redhat.com>
Fri, 30 Sep 2022 18:24:40 +0000 (14:24 -0400)
committerPaolo Bonzini <pbonzini@redhat.com>
Wed, 9 Nov 2022 17:25:53 +0000 (12:25 -0500)
commit9f2febf3f04daebdaaa5a43cfa20e3844905c0f9
tree17b9d3dff95f378c2d562a17d4a912365cc9fa4c
parente287bd005ad9d85dd6271dd795d3ecfb6bca46ad
KVM: SVM: move MSR_IA32_SPEC_CTRL save/restore to assembly

Restoration of the host IA32_SPEC_CTRL value is probably too late
with respect to the return thunk training sequence.

With respect to the user/kernel boundary, AMD says, "If software chooses
to toggle STIBP (e.g., set STIBP on kernel entry, and clear it on kernel
exit), software should set STIBP to 1 before executing the return thunk
training sequence." I assume the same requirements apply to the guest/host
boundary. The return thunk training sequence is in vmenter.S, quite close
to the VM-exit. On hosts without V_SPEC_CTRL, however, the host's
IA32_SPEC_CTRL value is not restored until much later.

To avoid this, move the restoration of host SPEC_CTRL to assembly and,
for consistency, move the restoration of the guest SPEC_CTRL as well.
This is not particularly difficult, apart from some care to cover both
32- and 64-bit, and to share code between SEV-ES and normal vmentry.

Cc: stable@vger.kernel.org
Fixes: a149180fbcf3 ("x86: Add magic AMD return-thunk")
Suggested-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kernel/cpu/bugs.c
arch/x86/kvm/kvm-asm-offsets.c
arch/x86/kvm/svm/svm.c
arch/x86/kvm/svm/svm.h
arch/x86/kvm/svm/vmenter.S