s390x/kvm: always ignore empty vcpu interrupt state

kvm_s390_vcpu_interrupt_pre_save() and
kvm_s390_vcpu_interrupt_post_load() are essentially no-ops on hosts
without KVM_CAP_S390_IRQ_STATE. Move the capability check after the
check for saved IRQ state in kvm_s390_vcpu_interrupt_post_load() so that
migration between hosts without KVM_CAP_S390_IRQ_STATE (including save /
restore on the same host) continues to work.

Fixes: 3cda44f ("s390x/kvm: migrate vcpu interrupt state")
Signed-off-by: Sascha Silbe <silbe@linux.vnet.ibm.com>
Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Reviewed-by: Jens Freimann <jfrei@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>

target-s390x/kvm.c

@@ -2175,13 +2175,14 @@ int kvm_s390_vcpu_interrupt_post_load(S390CPU *cpu)
     struct kvm_s390_irq_state irq_state;
     int r;
 
+    if (cpu->irqstate_saved_size == 0) {
+        return 0;
+    }
+
     if (!kvm_check_extension(kvm_state, KVM_CAP_S390_IRQ_STATE)) {
         return -ENOSYS;
     }
 
-    if (cpu->irqstate_saved_size == 0) {
-        return 0;
-    }
     irq_state.buf = (uint64_t) cpu->irqstate;
     irq_state.len = cpu->irqstate_saved_size;