@@ -360,7 +360,8 @@ int xen_smp_call_function_mask(cpumask_t
 			       void *info, int wait)
 {
 	struct call_data_struct data;
-	int cpus;
+	int cpus, cpu;
+	bool yield;
 
 	/* Holding any lock stops cpus from going down. */
 	spin_lock(&call_lock);

@@ -389,9 +390,14 @@ int xen_smp_call_function_mask(cpumask_t
 	/* Send a message to other CPUs and wait for them to respond */
 	xen_send_IPI_mask(mask, XEN_CALL_FUNCTION_VECTOR);
 
-	/* Make sure other vcpus get a chance to run.
-	   XXX too severe?  Maybe we should check the other CPU's states? */
-	HYPERVISOR_sched_op(SCHEDOP_yield, 0);
+	/* Make sure other vcpus get a chance to run if they need to. */
+	yield = false;
+	for_each_cpu_mask(cpu, mask)
+		if (xen_vcpu_stolen(cpu))
+			yield = true;
+
+	if (yield)
+		HYPERVISOR_sched_op(SCHEDOP_yield, 0);
 
 	/* Wait for response */
 	while (atomic_read(&data.started) != cpus ||

@@ -103,6 +103,12 @@ static void get_runstate_snapshot(struct
 		*res = *state;
 		barrier();
 	} while (get64(&state->state_entry_time) != state_time);
+}
+
+/* return true when a vcpu could run but has no real cpu to run on */
+bool xen_vcpu_stolen(int vcpu)
+{
+	return per_cpu(runstate, vcpu).state == RUNSTATE_runnable;
 }
 
 static void setup_runstate_info(int cpu)

@@ -27,6 +27,8 @@ int xen_set_wallclock(unsigned long time
 int xen_set_wallclock(unsigned long time);
 unsigned long long xen_sched_clock(void);
 
+bool xen_vcpu_stolen(int vcpu);
+
 void xen_mark_init_mm_pinned(void);
 
 DECLARE_PER_CPU(enum paravirt_lazy_mode, xen_lazy_mode);