KVM: nVMX: track NMI blocking state separately for each VMCS
authorPaolo Bonzini <pbonzini@redhat.com>
Fri, 14 Jul 2017 11:36:11 +0000 (13:36 +0200)
committerRadim Krčmář <rkrcmar@redhat.com>
Wed, 19 Jul 2017 14:05:41 +0000 (16:05 +0200)
commit4c4a6f790ee862ee9f0dc8b35c71f55bcf792b71
tree827f8e6a071255cb2a82645d4adf84691631d64c
parentf85c758dbee54cc3612a6e873ef7cecdb66ebee5
KVM: nVMX: track NMI blocking state separately for each VMCS

vmx_recover_nmi_blocking is using a cached value of the guest
interruptibility info, which is stored in vmx->nmi_known_unmasked.
vmx_recover_nmi_blocking is run for both normal and nested guests,
so the cached value must be per-VMCS.

This fixes eventinj.flat in a nested non-EPT environment.  With EPT it
works, because the EPT violation handler doesn't have the
vmx->nmi_known_unmasked optimization (it is unnecessary because, unlike
vmx_recover_nmi_blocking, it can just look at the exit qualification).

Thanks to Wanpeng Li for debugging the testcase and providing an initial
patch.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
arch/x86/kvm/vmx.c