KVM: SVM: Stuff next_rip on emulated INT3 injection if NRIPS is supported
authorSean Christopherson <seanjc@google.com>
Sun, 1 May 2022 22:07:28 +0000 (00:07 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 17 Aug 2022 12:23:40 +0000 (14:23 +0200)
[ Upstream commit 3741aec4c38fa4123ab08ae552f05366d4fd05d8 ]

If NRIPS is supported in hardware but disabled in KVM, set next_rip to
the next RIP when advancing RIP as part of emulating INT3 injection.
There is no flag to tell the CPU that KVM isn't using next_rip, and so
leaving next_rip is left as is will result in the CPU pushing garbage
onto the stack when vectoring the injected event.

Reviewed-by: Maxim Levitsky <mlevitsk@redhat.com>
Fixes: 66b7138f9136 ("KVM: SVM: Emulate nRIP feature when reinjecting INT3")
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Message-Id: <cd328309a3b88604daa2359ad56f36cb565ce2d4.1651440202.git.maciej.szmigiero@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
arch/x86/kvm/svm/svm.c

index 05d76832362dcd2f4d49cbc16dd3e51f56700a65..2947e3c965e329a6c362cc0c4e7f2bc81f83dc80 100644 (file)
@@ -394,6 +394,10 @@ static void svm_queue_exception(struct kvm_vcpu *vcpu)
                 */
                (void)skip_emulated_instruction(vcpu);
                rip = kvm_rip_read(vcpu);
+
+               if (boot_cpu_has(X86_FEATURE_NRIPS))
+                       svm->vmcb->control.next_rip = rip;
+
                svm->int3_rip = rip + svm->vmcb->save.cs.base;
                svm->int3_injected = rip - old_rip;
        }
@@ -3683,7 +3687,7 @@ static void svm_complete_interrupts(struct kvm_vcpu *vcpu)
        /*
         * If NextRIP isn't enabled, KVM must manually advance RIP prior to
         * injecting the soft exception/interrupt.  That advancement needs to
-        * be unwound if vectoring didn't complete.  Note, the _new_ event may
+        * be unwound if vectoring didn't complete.  Note, the new event may
         * not be the injected event, e.g. if KVM injected an INTn, the INTn
         * hit a #NP in the guest, and the #NP encountered a #PF, the #NP will
         * be the reported vectored event, but RIP still needs to be unwound.