OSDN Git Service

x86/kvm/vmx: do not use vm-exit instruction length for fast MMIO when running nested
authorVitaly Kuznetsov <vkuznets@redhat.com>
Thu, 25 Jan 2018 15:37:07 +0000 (16:37 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 13 Jan 2019 09:05:31 +0000 (10:05 +0100)
commit0c53038267a9883e4d0d591dc620fc7f0da4c584
tree4a28da9ba5fa715d43bfe6fc1a58fc517007966a
parentc78a5d4a64811b9f5d9d3b9dfa88ab786d949c1b
x86/kvm/vmx: do not use vm-exit instruction length for fast MMIO when running nested

commit d391f1207067268261add0485f0f34503539c5b0 upstream.

I was investigating an issue with seabios >= 1.10 which stopped working
for nested KVM on Hyper-V. The problem appears to be in
handle_ept_violation() function: when we do fast mmio we need to skip
the instruction so we do kvm_skip_emulated_instruction(). This, however,
depends on VM_EXIT_INSTRUCTION_LEN field being set correctly in VMCS.
However, this is not the case.

Intel's manual doesn't mandate VM_EXIT_INSTRUCTION_LEN to be set when
EPT MISCONFIG occurs. While on real hardware it was observed to be set,
some hypervisors follow the spec and don't set it; we end up advancing
IP with some random value.

I checked with Microsoft and they confirmed they don't fill
VM_EXIT_INSTRUCTION_LEN on EPT MISCONFIG.

Fix the issue by doing instruction skip through emulator when running
nested.

Fixes: 68c3b4d1676d870f0453c31d5a52e7e65c7448ae
Suggested-by: Radim Krčmář <rkrcmar@redhat.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
[mhaboustak: backport to 4.9.y]
Signed-off-by: Mike Haboustak <haboustak@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arch/x86/kvm/vmx.c
arch/x86/kvm/x86.c