KVM: VMX: Do not allow reexecute_instruction() when skipping MMIO instr
authorSean Christopherson <sean.j.christopherson@intel.com>
Thu, 23 Aug 2018 20:56:46 +0000 (13:56 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 2 Apr 2020 17:02:34 +0000 (19:02 +0200)
[ Upstream commit c4409905cd6eb42cfd06126e9226b0150e05a715 ]

Re-execution after an emulation decode failure is only intended to
handle a case where two or vCPUs race to write a shadowed page, i.e.
we should never re-execute an instruction as part of MMIO emulation.
As handle_ept_misconfig() is only used for MMIO emulation, it should
pass EMULTYPE_NO_REEXECUTE when using the emulator to skip an instr
in the fast-MMIO case where VM_EXIT_INSTRUCTION_LEN is invalid.

And because the cr2 value passed to x86_emulate_instruction() is only
destined for use when retrying or reexecuting, we can simply call
emulate_instruction().

Fixes: d391f1207067 ("x86/kvm/vmx: do not use vm-exit instruction length
                      for fast MMIO when running nested")
Cc: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
arch/x86/kvm/vmx.c

index 78daf891abec8adbf81a37486b7e060c14292f15..2634b455620266bd218ad950e1c3fe87c06e3038 100644 (file)
@@ -6187,8 +6187,8 @@ static int handle_ept_misconfig(struct kvm_vcpu *vcpu)
                        return 1;
                }
                else
-                       return x86_emulate_instruction(vcpu, gpa, EMULTYPE_SKIP,
-                                                      NULL, 0) == EMULATE_DONE;
+                       return emulate_instruction(vcpu, EMULTYPE_SKIP) ==
+                                                               EMULATE_DONE;
        }
 
        ret = handle_mmio_page_fault(vcpu, gpa, true);