[libc++] fix counting_semaphore lost wakeups

Author: huixie90

libcxx/include/__atomic/atomic_sync.h

@@ -46,12 +46,12 @@ __libcpp_atomic_wait(__cxx_atomic_contention_t const volatile*, __cxx_contention
 template <class _Atp, class _Fn>
 struct __libcpp_atomic_wait_backoff_impl {
   _Atp* __a;
-  _Fn __test_fn;
+  _Fn __test_with_old;
   _LIBCPP_AVAILABILITY_SYNC
   _LIBCPP_HIDE_FROM_ABI bool operator()(chrono::nanoseconds __elapsed) const {
     if (__elapsed > chrono::microseconds(64)) {
-      auto const __monitor = std::__libcpp_atomic_monitor(__a);
-      if (__test_fn())
+      auto __monitor = std::__libcpp_atomic_monitor(__a);
+      if (__test_with_old(__monitor))
         return true;
       std::__libcpp_atomic_wait(__a, __monitor);
     } else if (__elapsed > chrono::microseconds(4))
@@ -62,9 +62,17 @@ struct __libcpp_atomic_wait_backoff_impl {
   }
 };
 
+template <class _Atp, class _Fn, class _Fn2>
+_LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI bool
+__cxx_atomic_wait(_Atp* __a, _Fn&& __test_fn, _Fn2&& __test_with_old) {
+  __libcpp_atomic_wait_backoff_impl<_Atp, __decay_t<_Fn2> > __backoff_fn = {__a, __test_with_old};
+  return std::__libcpp_thread_poll_with_backoff(__test_fn, __backoff_fn);
+}
+
 template <class _Atp, class _Fn>
 _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI bool __cxx_atomic_wait(_Atp* __a, _Fn&& __test_fn) {
-  __libcpp_atomic_wait_backoff_impl<_Atp, __decay_t<_Fn> > __backoff_fn = {__a, __test_fn};
+  auto __test_with_old = [&](auto&) { return __test_fn(); };
+  __libcpp_atomic_wait_backoff_impl<_Atp, decltype(__test_with_old)> __backoff_fn{__a, __test_with_old};
   return std::__libcpp_thread_poll_with_backoff(__test_fn, __backoff_fn);
 }
 

libcxx/include/semaphore

@@ -54,6 +54,7 @@ using binary_semaphore = counting_semaphore<1>;
 #include <__assert> // all public C++ headers provide the assertion handler
 #include <__atomic/atomic_base.h>
 #include <__atomic/atomic_sync.h>
+#include <__atomic/contention_t.h>
 #include <__atomic/memory_order.h>
 #include <__availability>
 #include <__chrono/time_point.h>
@@ -95,19 +96,22 @@ public:
     _LIBCPP_ASSERT_ARGUMENT_WITHIN_DOMAIN(
         __update <= _LIBCPP_SEMAPHORE_MAX - __old, "update is greater than the expected value");
 
-    if (__old > 0) {
-      // Nothing to do
-    } else if (__update > 1)
+    if (__old == 0) {
       __a_.notify_all();
-    else
-      __a_.notify_one();
+    }
   }
   _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void acquire() {
     auto const __test_fn = [this]() -> bool {
       auto __old = __a_.load(memory_order_relaxed);
       return (__old != 0) && __a_.compare_exchange_strong(__old, __old - 1, memory_order_acquire, memory_order_relaxed);
     };
-    __cxx_atomic_wait(&__a_.__a_, __test_fn);
+    auto const __test_with_old = [this](__cxx_contention_t& __monitor) -> bool {
+      ptrdiff_t __old = __monitor;
+      bool __r        = __try_acquire_impl(__old);
+      __monitor       = static_cast<__cxx_contention_t>(__old);
+      return __r;
+    };
+    __cxx_atomic_wait(&__a_.__a_, __test_fn, __test_with_old);
   }
   template <class _Rep, class _Period>
   _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI bool
@@ -119,6 +123,11 @@ public:
   }
   _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI bool try_acquire() {
     auto __old = __a_.load(memory_order_acquire);
+    return __try_acquire_impl(__old);
+  }
+
+private:
+  _LIBCPP_AVAILABILITY_SYNC _LIBCPP_INLINE_VISIBILITY bool __try_acquire_impl(ptrdiff_t& __old) {
     while (true) {
       if (__old == 0)
         return false;

libcxx/test/std/thread/thread.semaphore/lost_wakeup.pass.cpp

@@ -0,0 +1,65 @@
+//===----------------------------------------------------------------------===//
+//
+// 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}}
+
+// This is a regression test for https://llvm.org/PR47013.
+
+// <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**) {
+  for (int run = 0; run < 20; ++run) {
+    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;
+}