KVM: nSVM: always intercept VMLOAD/VMSAVE when nested (CVE-2021-3656)
authorMaxim Levitsky <mlevitsk@redhat.com>
Mon, 19 Jul 2021 13:05:00 +0000 (16:05 +0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 18 Aug 2021 07:07:08 +0000 (09:07 +0200)
commit639a033fd765ed473dfee27028df5ccbe1038a2e
treeafc01ec52a932960df5e9a4a2c9892ee69040098
parenta0949ee63cf95408870a564ccad163018b1a9e6b
KVM: nSVM: always intercept VMLOAD/VMSAVE when nested (CVE-2021-3656)

commit c7dfa4009965a9b2d7b329ee970eb8da0d32f0bc upstream.

If L1 disables VMLOAD/VMSAVE intercepts, and doesn't enable
Virtual VMLOAD/VMSAVE (currently not supported for the nested hypervisor),
then VMLOAD/VMSAVE must operate on the L1 physical memory, which is only
possible by making L0 intercept these instructions.

Failure to do so allowed the nested guest to run VMLOAD/VMSAVE unintercepted,
and thus read/write portions of the host physical memory.

Fixes: 89c8a4984fc9 ("KVM: SVM: Enable Virtual VMLOAD VMSAVE feature")
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arch/x86/kvm/svm/nested.c