Index: libcxx/include/semaphore
===================================================================
--- libcxx/include/semaphore
+++ libcxx/include/semaphore
@@ -87,10 +87,9 @@
     {
         if(0 < __a.fetch_add(__update, memory_order_release))
             ;
-        else if(__update > 1)
-            __a.notify_all();
         else
-            __a.notify_one();
+            // Always notify all, regardless of the value of __update (see PR47013)
+            __a.notify_all();
     }
     _LIBCPP_AVAILABILITY_SYNC _LIBCPP_INLINE_VISIBILITY
     void acquire()
Index: libcxx/test/std/thread/thread.semaphore/lost_wakeup.pass.cpp
===================================================================
--- /dev/null
+++ libcxx/test/std/thread/thread.semaphore/lost_wakeup.pass.cpp
@@ -0,0 +1,67 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// UNSUPPORTED: libcpp-has-no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// This test requires the dylib support introduced in D68480, which shipped in macOS 11.0.
+// XFAIL: use_system_cxx_lib && target={{.+}}-apple-macosx10.{{9|10|11|12|13|14|15}}
+
+// TODO(ldionne): This test fails on Ubuntu Focal on our CI nodes (and only there), in 32 bit mode.
+// UNSUPPORTED: linux && 32bits-on-64bits
+
+// This is a regression test for PR#47013.
+
+// <semaphore>
+
+#include <barrier>
+#include <semaphore>
+#include <thread>
+#include <vector>
+
+#include "make_test_thread.h"
+
+static std::counting_semaphore s(0);
+static std::barrier b(8 + 1);
+
+void acquire() {
+  for (int i = 0; i < 10'000; ++i) {
+    s.acquire();
+    b.arrive_and_wait();
+  }
+}
+
+void release() {
+  for (int i = 0; i < 10'000; ++i) {
+    s.release(1);
+    s.release(1);
+    s.release(1);
+    s.release(1);
+
+    s.release(1);
+    s.release(1);
+    s.release(1);
+    s.release(1);
+
+    b.arrive_and_wait();
+  }
+}
+
+int main(int, char**) {
+  std::vector<std::thread> threads;
+
+  for (int i = 0; i < 8; ++i)
+    threads.push_back(support::make_test_thread(acquire));
+
+  threads.push_back(support::make_test_thread(release));
+
+  for (auto& thread : threads)
+    thread.join();
+
+  return 0;
+}