diff --git a/lib/core/context.c b/lib/core/context.c
index 91bfd1701..ea1f1936b 100644
--- a/lib/core/context.c
+++ b/lib/core/context.c
@@ -996,7 +996,7 @@ lws_create_vhost(struct lws_context *context,
 		lwsl_err("%s: lws_context_init_client_ssl failed\n", __func__);
 		goto bail1;
 	}
-	lws_context_lock(context);
+	lws_context_lock(context, "create_vhost");
 	n = _lws_vhost_init_server(info, vh);
 	lws_context_unlock(context);
 	if (n < 0) {
@@ -1215,7 +1215,7 @@ lws_create_context(const struct lws_context_creation_info *info)
 #endif
 
 #if LWS_MAX_SMP > 1
-	pthread_mutex_init(&context->lock, NULL);
+	lws_mutex_refcount_init(&context->mr);
 #endif
 
 #if defined(LWS_WITH_ESP32)
@@ -1844,7 +1844,7 @@ lws_check_deferred_free(struct lws_context *context, int tsi, int force)
 	 * because there is nothing left using the vhost to conflict.
 	 */
 
-	lws_context_lock(context); /* ------------------- context { */
+	lws_context_lock(context, "check deferred free"); /* ------ context { */
 
 	lws_start_foreach_ll(struct lws_vhost *, v, context->vhost_list) {
 		if (v->being_destroyed
@@ -2034,7 +2034,7 @@ lws_context_destroy2(struct lws_context *context)
 	lws_check_deferred_free(context, 0, 1);
 
 #if LWS_MAX_SMP > 1
-	pthread_mutex_destroy(&context->lock);
+	lws_mutex_refcount_destroy(&context->mr);
 #endif
 
 	if (context->event_loop_ops->destroy_context2)
diff --git a/lib/core/libwebsockets.c b/lib/core/libwebsockets.c
index c42b3eac3..38f13a47a 100644
--- a/lib/core/libwebsockets.c
+++ b/lib/core/libwebsockets.c
@@ -2888,6 +2888,65 @@ lws_strncpy(char *dest, const char *src, size_t size)
 	return dest;
 }
 
+#if LWS_MAX_SMP > 1
+
+void
+lws_mutex_refcount_init(struct lws_mutex_refcount *mr)
+{
+	pthread_mutex_init(&mr->lock, NULL);
+	mr->last_lock_reason = NULL;
+	mr->lock_depth = 0;
+	mr->metadata = 0;
+	mr->lock_owner = 0;
+}
+
+void
+lws_mutex_refcount_destroy(struct lws_mutex_refcount *mr)
+{
+	pthread_mutex_destroy(&mr->lock);
+}
+
+void
+lws_mutex_refcount_lock(struct lws_mutex_refcount *mr, const char *reason)
+{
+	/* if true, this sequence is atomic because our thread has the lock
+	 *
+	 *  - if true, only guy who can race to make it untrue is our thread,
+	 *    and we are here.
+	 *
+	 *  - if false, only guy who could race to make it true is our thread,
+	 *    and we are here
+	 *
+	 *  - it can be false and change to a different tid that is also false
+	 */
+	if (mr->lock_owner == pthread_self()) {
+		/* atomic because we only change it if we own the lock */
+		mr->lock_depth++;
+		return;
+	}
+
+	pthread_mutex_lock(&mr->lock);
+	/* atomic because only we can have the lock */
+	mr->last_lock_reason = reason;
+	mr->lock_owner = pthread_self();
+	mr->lock_depth = 1;
+	//lwsl_notice("tid %d: lock %s\n", mr->tid, reason);
+}
+
+void
+lws_mutex_refcount_unlock(struct lws_mutex_refcount *mr)
+{
+	if (--mr->lock_depth)
+		/* atomic because only thread that has the lock can unlock */
+		return;
+
+	mr->last_lock_reason = "free";
+	mr->lock_owner = 0;
+	//lwsl_notice("tid %d: unlock %s\n", mr->tid, mr->last_lock_reason);
+	pthread_mutex_unlock(&mr->lock);
+}
+
+#endif /* SMP */
 
 LWS_VISIBLE LWS_EXTERN int
 lws_is_cgi(struct lws *wsi) {
diff --git a/lib/core/private.h b/lib/core/private.h
index 0d0da2be2..dd1080421 100644
--- a/lib/core/private.h
+++ b/lib/core/private.h
@@ -303,6 +303,28 @@ struct lws_foreign_thread_pollfd {
 	int _or;
 };
 
+#if LWS_MAX_SMP > 1
+
+struct lws_mutex_refcount {
+	pthread_mutex_t lock;
+	pthread_t lock_owner;
+	const char *last_lock_reason;
+	char lock_depth;
+	char metadata;
+};
+
+void
+lws_mutex_refcount_init(struct lws_mutex_refcount *mr);
+
+void
+lws_mutex_refcount_destroy(struct lws_mutex_refcount *mr);
+
+void
+lws_mutex_refcount_lock(struct lws_mutex_refcount *mr, const char *reason);
+
+void
+lws_mutex_refcount_unlock(struct lws_mutex_refcount *mr);
+#endif
 
 #define LWS_HRTIMER_NOWAIT (0x7fffffffffffffffll)
 
@@ -313,10 +335,8 @@ struct lws_foreign_thread_pollfd {
 
 struct lws_context_per_thread {
 #if LWS_MAX_SMP > 1
-	pthread_mutex_t lock;
 	pthread_mutex_t lock_stats;
-	pthread_t lock_owner;
-	const char *last_lock_reason;
+	struct lws_mutex_refcount mr;
 #endif
 
 	struct lws_context *context;
@@ -379,7 +399,6 @@ struct lws_context_per_thread {
 
 	unsigned char tid;
 
-	unsigned char lock_depth;
 	unsigned char inside_service:1;
 	unsigned char event_loop_foreign:1;
 	unsigned char event_loop_destroy_processing_done:1;
@@ -575,8 +594,7 @@ struct lws_context {
 	struct lws_context_per_thread pt[LWS_MAX_SMP];
 	struct lws_conn_stats conn_stats;
 #if LWS_MAX_SMP > 1
-	pthread_mutex_t lock;
-	int lock_depth;
+	struct lws_mutex_refcount mr;
 #endif
 #ifdef _WIN32
 /* different implementation between unix and windows */
@@ -1197,7 +1215,7 @@ LWS_EXTERN void lwsl_emit_stderr(int level, const char *line);
 static LWS_INLINE void
 lws_pt_mutex_init(struct lws_context_per_thread *pt)
 {
-	pthread_mutex_init(&pt->lock, NULL);
+	lws_mutex_refcount_init(&pt->mr);
 	pthread_mutex_init(&pt->lock_stats, NULL);
 }
 
@@ -1205,34 +1223,11 @@ static LWS_INLINE void
 lws_pt_mutex_destroy(struct lws_context_per_thread *pt)
 {
 	pthread_mutex_destroy(&pt->lock_stats);
-	pthread_mutex_destroy(&pt->lock);
+	lws_mutex_refcount_destroy(&pt->mr);
 }
 
-static LWS_INLINE void
-lws_pt_lock(struct lws_context_per_thread *pt, const char *reason)
-{
-	if (pt->lock_owner == pthread_self()) {
-		pt->lock_depth++;
-		return;
-	}
-	pthread_mutex_lock(&pt->lock);
-	pt->last_lock_reason = reason;
-	pt->lock_owner = pthread_self();
-	//lwsl_notice("tid %d: lock %s\n", pt->tid, reason);
-}
-
-static LWS_INLINE void
-lws_pt_unlock(struct lws_context_per_thread *pt)
-{
-	if (pt->lock_depth) {
-		pt->lock_depth--;
-		return;
-	}
-	pt->last_lock_reason = "free";
-	pt->lock_owner = 0;
-	//lwsl_notice("tid %d: unlock %s\n", pt->tid, pt->last_lock_reason);
-	pthread_mutex_unlock(&pt->lock);
-}
+#define lws_pt_lock(pt, reason) lws_mutex_refcount_lock(&pt->mr, reason)
+#define lws_pt_unlock(pt) lws_mutex_refcount_unlock(&pt->mr)
 
 static LWS_INLINE void
 lws_pt_stats_lock(struct lws_context_per_thread *pt)
@@ -1246,17 +1241,8 @@ lws_pt_stats_unlock(struct lws_context_per_thread *pt)
 	pthread_mutex_unlock(&pt->lock_stats);
 }
 
-static LWS_INLINE void
-lws_context_lock(struct lws_context *context)
-{
-	pthread_mutex_lock(&context->lock);
-}
-
-static LWS_INLINE void
-lws_context_unlock(struct lws_context *context)
-{
-	pthread_mutex_unlock(&context->lock);
-}
+#define lws_context_lock(c, reason) lws_mutex_refcount_lock(&c->mr, reason)
+#define lws_context_unlock(c) lws_mutex_refcount_unlock(&c->mr)
 
 static LWS_INLINE void
 lws_vhost_lock(struct lws_vhost *vhost)
@@ -1276,7 +1262,7 @@ lws_vhost_unlock(struct lws_vhost *vhost)
 #define lws_pt_mutex_destroy(_a) (void)(_a)
 #define lws_pt_lock(_a, b) (void)(_a)
 #define lws_pt_unlock(_a) (void)(_a)
-#define lws_context_lock(_a) (void)(_a)
+#define lws_context_lock(_a, _b) (void)(_a)
 #define lws_context_unlock(_a) (void)(_a)
 #define lws_vhost_lock(_a) (void)(_a)
 #define lws_vhost_unlock(_a) (void)(_a)
diff --git a/lib/core/service.c b/lib/core/service.c
index 5cde4d2e8..2f6656bcf 100644
--- a/lib/core/service.c
+++ b/lib/core/service.c
@@ -772,7 +772,7 @@ lws_service_periodic_checks(struct lws_context *context,
 	 *	    interface missing before
 	 */
 
-	lws_context_lock(context);
+	lws_context_lock(context, "periodic checks");
 	lws_start_foreach_llp(struct lws_vhost **, pv,
 			      context->no_listener_vhost_list) {
 		struct lws_vhost *v = *pv;
diff --git a/lib/misc/peer-limits.c b/lib/misc/peer-limits.c
index 66deef49d..1f1ef4d3e 100644
--- a/lib/misc/peer-limits.c
+++ b/lib/misc/peer-limits.c
@@ -95,7 +95,7 @@ lws_get_or_create_peer(struct lws_vhost *vhost, lws_sockfd_type sockfd)
 
 	hash = hash % context->pl_hash_elements;
 
-	lws_context_lock(context); /* <====================================== */
+	lws_context_lock(context, "peer search"); /* <======================= */
 
 	lws_start_foreach_ll(struct lws_peer *, peerx,
 			     context->pl_hash_table[hash]) {
@@ -163,7 +163,7 @@ lws_peer_cull_peer_wait_list(struct lws_context *context)
 	if (context->next_cull && t < context->next_cull)
 		return;
 
-	lws_context_lock(context); /* <====================================== */
+	lws_context_lock(context, "peer cull"); /* <========================= */
 
 	context->next_cull = t + 5;
 
@@ -190,7 +190,7 @@ lws_peer_add_wsi(struct lws_context *context, struct lws_peer *peer,
 	if (!peer)
 		return;
 
-	lws_context_lock(context); /* <====================================== */
+	lws_context_lock(context, "peer add"); /* <========================== */
 
 	peer->count_wsi++;
 	wsi->peer = peer;
@@ -225,7 +225,7 @@ lws_peer_track_wsi_close(struct lws_context *context, struct lws_peer *peer)
 	if (!peer)
 		return;
 
-	lws_context_lock(context); /* <====================================== */
+	lws_context_lock(context, "peer wsi close"); /* <==================== */
 
 	assert(peer->count_wsi);
 	peer->count_wsi--;
@@ -273,7 +273,7 @@ lws_peer_track_ah_detach(struct lws_context *context, struct lws_peer *peer)
 	if (!peer)
 		return;
 
-	lws_context_lock(context); /* <====================================== */
+	lws_context_lock(context, "peer ah detach"); /* <==================== */
 	assert(peer->http.count_ah);
 	peer->http.count_ah--;
 	lws_context_unlock(context); /* ====================================> */
diff --git a/lib/roles/http/server/parsers.c b/lib/roles/http/server/parsers.c
index a023d8d82..82cf5ba76 100644
--- a/lib/roles/http/server/parsers.c
+++ b/lib/roles/http/server/parsers.c
@@ -241,7 +241,7 @@ lws_header_table_attach(struct lws *wsi, int autoservice)
 	pt->http.ah_count_in_use++;
 
 #if defined(LWS_WITH_PEER_LIMITS) && (defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2))
-	lws_context_lock(context); /* <====================================== */
+	lws_context_lock(context, "ah attach"); /* <========================= */
 	if (wsi->peer)
 		wsi->peer->http.count_ah++;
 	lws_context_unlock(context); /* ====================================> */
@@ -360,7 +360,7 @@ int __lws_header_table_detach(struct lws *wsi, int autoservice)
 
 	__lws_header_table_reset(wsi, autoservice);
 #if defined(LWS_WITH_PEER_LIMITS) && (defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2))
-	lws_context_lock(context); /* <====================================== */
+	lws_context_lock(context, "ah detach"); /* <========================= */
 	if (wsi->peer)
 		wsi->peer->http.count_ah++;
 	lws_context_unlock(context); /* ====================================> */