KVM: nVMX: Ignore limit checks on VMX instructions using flat segments

commit 34333cc upstream.

Regarding segments with a limit==0xffffffff, the SDM officially states:

    When the effective limit is FFFFFFFFH (4 GBytes), these accesses may
    or may not cause the indicated exceptions.  Behavior is
    implementation-specific and may vary from one execution to another.

In practice, all CPUs that support VMX ignore limit checks for "flat
segments", i.e. an expand-up data or code segment with base=0 and
limit=0xffffffff.  This is subtly different than wrapping the effective
address calculation based on the address size, as the flat segment
behavior also applies to accesses that would wrap the 4g boundary, e.g.
a 4-byte access starting at 0xffffffff will access linear addresses
0xffffffff, 0x0, 0x1 and 0x2.

Fixes: f9eb4af ("KVM: nVMX: VMX instructions: add checks for #GP/#SS exceptions")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

arch/x86/kvm/vmx.c

@@ -6702,10 +6702,16 @@ static int get_vmx_mem_address(struct kvm_vcpu *vcpu,
 		/* Protected mode: #GP(0)/#SS(0) if the segment is unusable.
 		 */
 		exn = (s.unusable != 0);
-		/* Protected mode: #GP(0)/#SS(0) if the memory
-		 * operand is outside the segment limit.
+
+		/*
+		 * Protected mode: #GP(0)/#SS(0) if the memory operand is
+		 * outside the segment limit.  All CPUs that support VMX ignore
+		 * limit checks for flat segments, i.e. segments with base==0,
+		 * limit==0xffffffff and of type expand-up data or code.
 		 */
-		exn = exn || (off + sizeof(u64) > s.limit);
+		if (!(s.base == 0 && s.limit == 0xffffffff &&
+		     ((s.type & 8) || !(s.type & 4))))
+			exn = exn || (off + sizeof(u64) > s.limit);
 	}
 	if (exn) {
 		kvm_queue_exception_e(vcpu,