diff --git a/lib/libc/gen/Symbol.map b/lib/libc/gen/Symbol.map
--- a/lib/libc/gen/Symbol.map
+++ b/lib/libc/gen/Symbol.map
@@ -464,6 +464,7 @@
 	/* needed by thread libraries */
 	__thr_jtable;
 
+	_dl_iterate_phdr_locked;
 	_pthread_atfork;
 	_pthread_attr_destroy;
 	_pthread_attr_getdetachstate;
diff --git a/lib/libc/gen/dlfcn.c b/lib/libc/gen/dlfcn.c
--- a/lib/libc/gen/dlfcn.c
+++ b/lib/libc/gen/dlfcn.c
@@ -202,9 +202,10 @@
 }
 #endif
 
-#pragma weak dl_iterate_phdr
+#pragma weak _dl_iterate_phdr_locked
 int
-dl_iterate_phdr(int (*callback)(struct dl_phdr_info *, size_t, void *) __unused,
+_dl_iterate_phdr_locked(
+    int (*callback)(struct dl_phdr_info *, size_t, void *) __unused,
     void *data __unused)
 {
 #if defined IN_LIBDL
@@ -226,14 +227,29 @@
 	_once(&dl_phdr_info_once, dl_init_phdr_info);
 	ti.ti_module = 1;
 	ti.ti_offset = 0;
-	mutex_lock(&dl_phdr_info_lock);
 	phdr_info.dlpi_tls_data = __tls_get_addr(&ti);
 	ret = callback(&phdr_info, sizeof(phdr_info), data);
-	mutex_unlock(&dl_phdr_info_lock);
 	return (ret);
 #endif
 }
 
+#pragma weak dl_iterate_phdr
+int
+dl_iterate_phdr(int (*callback)(struct dl_phdr_info *, size_t, void *) __unused,
+    void *data __unused)
+{
+	int error;
+
+#if !defined(IN_LIBDL) && !defined(PIC)
+	mutex_lock(&dl_phdr_info_lock);
+#endif
+	error = _dl_iterate_phdr_locked(callback, data);
+#if !defined(IN_LIBDL) && !defined(PIC)
+	mutex_unlock(&dl_phdr_info_lock);
+#endif
+	return (error);
+}
+
 #pragma weak fdlopen
 void *
 fdlopen(int fd __unused, int mode __unused)
diff --git a/libexec/rtld-elf/Symbol.map b/libexec/rtld-elf/Symbol.map
--- a/libexec/rtld-elf/Symbol.map
+++ b/libexec/rtld-elf/Symbol.map
@@ -22,6 +22,7 @@
 };
 
 FBSDprivate_1.0 {
+    _dl_iterate_phdr_locked;
     _rtld_thread_init;
     _rtld_allocate_tls;
     _rtld_free_tls;
diff --git a/libexec/rtld-elf/rtld.c b/libexec/rtld-elf/rtld.c
--- a/libexec/rtld-elf/rtld.c
+++ b/libexec/rtld-elf/rtld.c
@@ -252,6 +252,7 @@
 void dllockinit(void *, void *(*)(void *), void (*)(void *), void (*)(void *),
     void (*)(void *), void (*)(void *), void (*)(void *)) __exported;
 int dlinfo(void *, int , void *) __exported;
+int _dl_iterate_phdr_locked(__dl_iterate_hdr_callback, void *) __exported;
 int dl_iterate_phdr(__dl_iterate_hdr_callback, void *) __exported;
 int _rtld_addr_phdr(const void *, struct dl_phdr_info *) __exported;
 int _rtld_get_stack_prot(void) __exported;
@@ -4192,43 +4193,71 @@
 	phdr_info->dlpi_subs = obj_loads - obj_count;
 }
 
-int
-dl_iterate_phdr(__dl_iterate_hdr_callback callback, void *param)
+/*
+ * Returns with the bind lock dropped.
+ */
+static int
+_dl_iterate_phdr_impl(__dl_iterate_hdr_callback callback, void *param,
+    RtldLockState *bind_lockstate)
 {
 	struct dl_phdr_info phdr_info;
 	Obj_Entry *obj, marker;
-	RtldLockState bind_lockstate, phdr_lockstate;
 	int error;
 
-	init_marker(&marker);
 	error = 0;
+	init_marker(&marker);
 
-	wlock_acquire(rtld_phdr_lock, &phdr_lockstate);
-	wlock_acquire(rtld_bind_lock, &bind_lockstate);
 	for (obj = globallist_curr(TAILQ_FIRST(&obj_list)); obj != NULL;) {
 		TAILQ_INSERT_AFTER(&obj_list, obj, &marker, next);
 		rtld_fill_dl_phdr_info(obj, &phdr_info);
 		hold_object(obj);
-		lock_release(rtld_bind_lock, &bind_lockstate);
+
+		if (bind_lockstate != NULL)
+			lock_release(rtld_bind_lock, bind_lockstate);
 
 		error = callback(&phdr_info, sizeof phdr_info, param);
 
-		wlock_acquire(rtld_bind_lock, &bind_lockstate);
+		if (bind_lockstate != NULL)
+			wlock_acquire(rtld_bind_lock, bind_lockstate);
 		unhold_object(obj);
 		obj = globallist_next(&marker);
 		TAILQ_REMOVE(&obj_list, &marker, next);
 		if (error != 0) {
-			lock_release(rtld_bind_lock, &bind_lockstate);
-			lock_release(rtld_phdr_lock, &phdr_lockstate);
+			if (bind_lockstate != NULL)
+				lock_release(rtld_bind_lock, bind_lockstate);
 			return (error);
 		}
 	}
 
-	if (error == 0) {
-		rtld_fill_dl_phdr_info(&obj_rtld, &phdr_info);
-		lock_release(rtld_bind_lock, &bind_lockstate);
-		error = callback(&phdr_info, sizeof(phdr_info), param);
-	}
+	rtld_fill_dl_phdr_info(&obj_rtld, &phdr_info);
+	if (bind_lockstate != NULL)
+		lock_release(rtld_bind_lock, bind_lockstate);
+
+	return (callback(&phdr_info, sizeof(phdr_info), param));
+}
+
+/*
+ * It's completely UB to actually use this, so extreme caution is advised.  It's
+ * probably not what you want.
+ */
+int
+_dl_iterate_phdr_locked(__dl_iterate_hdr_callback callback, void *param)
+{
+
+	return (_dl_iterate_phdr_impl(callback, param, NULL));
+}
+
+int
+dl_iterate_phdr(__dl_iterate_hdr_callback callback, void *param)
+{
+	RtldLockState bind_lockstate, phdr_lockstate;
+	int error;
+
+	wlock_acquire(rtld_phdr_lock, &phdr_lockstate);
+	wlock_acquire(rtld_bind_lock, &bind_lockstate);
+
+	error = _dl_iterate_phdr_impl(callback, param, &phdr_lockstate);
+
 	lock_release(rtld_phdr_lock, &phdr_lockstate);
 	return (error);
 }