Index: sys/kern/subr_smp.c
===================================================================
--- sys/kern/subr_smp.c
+++ sys/kern/subr_smp.c
@@ -56,7 +56,8 @@
 MALLOC_DEFINE(M_TOPO, "toponodes", "SMP topology data");
 
 volatile cpuset_t stopped_cpus;
-volatile cpuset_t started_cpus;
+/* Align to attempt to keep started_cpus within a cache line for MWAIT. */
+volatile cpuset_t started_cpus __aligned(CACHE_LINE_SIZE);
 volatile cpuset_t suspended_cpus;
 cpuset_t hlt_cpus_mask;
 cpuset_t logical_cpus_mask;
Index: sys/x86/x86/mp_x86.c
===================================================================
--- sys/x86/x86/mp_x86.c
+++ sys/x86/x86/mp_x86.c
@@ -161,6 +161,10 @@
 
 unsigned int boot_address;
 
+static bool stop_mwait = false;
+SYSCTL_BOOL(_machdep, OID_AUTO, stop_mwait, CTLFLAG_RWTUN, &stop_mwait, 0,
+    "Use MONITOR/MWAIT when stopping CPU");
+
 #define MiB(v)	(v ## ULL << 20)
 
 void
@@ -1389,6 +1393,17 @@
 		cpustop_handler_post(cpu);
 }
 
+static bool
+range_can_be_monitored(volatile void *start, size_t length)
+{
+	uint64_t addr, preceding;
+
+	addr = (uintptr_t)start;
+	preceding = addr & (cpu_mon_min_size - 1);
+
+	return (preceding + length <= cpu_mon_min_size);
+}
+
 /*
  * Handle an IPI_STOP by saving our current context and spinning until we
  * are resumed.
@@ -1397,6 +1412,7 @@
 cpustop_handler(void)
 {
 	u_int cpu;
+	bool use_mwait;
 
 	cpu = PCPU_GET(cpuid);
 
@@ -1405,8 +1421,18 @@
 	/* Indicate that we are stopped */
 	CPU_SET_ATOMIC(cpu, &stopped_cpus);
 
+	use_mwait = (stop_mwait && (cpu_feature2 & CPUID2_MON) &&
+	    range_can_be_monitored(&started_cpus, sizeof(started_cpus)));
+
 	/* Wait for restart */
 	while (!CPU_ISSET(cpu, &started_cpus)) {
+		if (use_mwait) {
+			cpu_monitor(__DEVOLATILE(void *, &started_cpus), 0, 0);
+			if (!CPU_ISSET(cpu, &started_cpus))
+				cpu_mwait(0, MWAIT_C1);
+			continue;
+		}
+
 		ia32_pause();
 
 		/*