diff --git a/sys/kern/subr_xcall.c b/sys/kern/subr_xcall.c
index e82a26b56143..310b2e882b09 100644
--- a/sys/kern/subr_xcall.c
+++ b/sys/kern/subr_xcall.c
@@ -78,6 +78,7 @@ __KERNEL_RCSID(0, "$NetBSD: subr_xcall.c,v 1.28 2019/11/11 09:50:11 maxv Exp $")
 
 #include <sys/types.h>
 #include <sys/param.h>
+#include <sys/atomic.h>
 #include <sys/xcall.h>
 #include <sys/mutex.h>
 #include <sys/condvar.h>
@@ -334,7 +335,14 @@ xc_wait(uint64_t where)
 		xc = &xc_low_pri;
 	}
 
-	/* Block until awoken. */
+#ifdef __HAVE_ATOMIC64_LOADSTORE
+	/* Fast path, if already done. */
+	if (atomic_load_acquire(&xc->xc_donep) >= where) {
+		return;
+	}
+#endif
+
+	/* Slow path: block until awoken. */
 	mutex_enter(&xc->xc_lock);
 	while (xc->xc_donep < where) {
 		cv_wait(&xc->xc_busy, &xc->xc_lock);
@@ -417,7 +425,11 @@ xc_thread(void *cookie)
 		(*func)(arg1, arg2);
 
 		mutex_enter(&xc->xc_lock);
+#ifdef __HAVE_ATOMIC64_LOADSTORE
+		atomic_store_release(&xc->xc_donep, xc->xc_donep + 1);
+#else
 		xc->xc_donep++;
+#endif
 	}
 	/* NOTREACHED */
 }
@@ -470,7 +482,12 @@ xc__highpri_intr(void *dummy)
 	 */
 	mutex_enter(&xc->xc_lock);
 	KASSERT(xc->xc_donep < xc->xc_headp);
-	if (++xc->xc_donep == xc->xc_headp) {
+#ifdef __HAVE_ATOMIC64_LOADSTORE
+	atomic_store_release(&xc->xc_donep, xc->xc_donep + 1);
+#else
+	xc->xc_donep++;
+#endif
+	if (xc->xc_donep == xc->xc_headp) {
 		cv_broadcast(&xc->xc_busy);
 	}
 	mutex_exit(&xc->xc_lock);