[libc] ensure tls dtors are called in main thread

libc/src/__support/threads/thread.cpp

@@ -163,6 +163,8 @@ void call_atexit_callbacks(ThreadAttributes *attrib) {
   }
 }
 
+extern "C" void __cxa_thread_finalize() { call_atexit_callbacks(self.attrib); }
+
 } // namespace internal
 
 cpp::optional<unsigned int> new_tss_key(TSSDtor *dtor) {

libc/src/__support/threads/thread.h

@@ -110,7 +110,7 @@ struct alignas(STACK_ALIGNMENT) ThreadAttributes {
   ThreadAtExitCallbackMgr *atexit_callback_mgr;
   void *platform_data;
 
-  constexpr ThreadAttributes()
+  LIBC_INLINE constexpr ThreadAttributes()
       : detach_state(uint32_t(DetachState::DETACHED)), stack(nullptr),
         stacksize(0), guardsize(0), tls(0), tls_size(0), owned_stack(false),
         tid(-1), style(ThreadStyle::POSIX), retval(),

libc/src/stdlib/exit.cpp

@@ -15,7 +15,17 @@ namespace LIBC_NAMESPACE_DECL {
 
 extern "C" void __cxa_finalize(void *);
 
+// exit needs to clean up TLS and call associated destructors.
+// TODO: Strictly speaking, it is not valid to call exit in overlay mode
+//       as we have no way to ensure system libc will call the TLS destructors.
+//       We should run exit related tests in hermetic mode but this is currently
+//       blocked by https://github.com/llvm/llvm-project/issues/133925.
+extern "C" [[gnu::weak]] void __cxa_thread_finalize();
+
+// TODO: use recursive mutex to protect this routine.
 [[noreturn]] LLVM_LIBC_FUNCTION(void, exit, (int status)) {
+  if (__cxa_thread_finalize)
+    __cxa_thread_finalize();
   __cxa_finalize(nullptr);
   internal::exit(status);
 }

libc/test/integration/src/__support/threads/CMakeLists.txt

@@ -25,3 +25,24 @@ add_integration_test(
   DEPENDS
     libc.src.__support.threads.thread
 )
+
+add_integration_test(
+  main_exit_test
+  SUITE
+    libc-support-threads-integration-tests
+  SRCS
+    main_exit_test.cpp
+  DEPENDS
+    libc.src.__support.threads.thread
+)
+
+add_integration_test(
+  double_exit_test
+  SUITE
+    libc-support-threads-integration-tests
+  SRCS
+    double_exit_test.cpp
+  DEPENDS
+    libc.src.__support.threads.thread
+    libc.src.stdlib.exit
+)

libc/test/integration/src/__support/threads/double_exit_test.cpp

@@ -0,0 +1,33 @@
+//===-- Test handling of thread local data --------------------------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/__support/threads/thread.h"
+#include "src/stdlib/exit.h"
+#include "test/IntegrationTest/test.h"
+
+extern "C" {
+[[gnu::weak]]
+void *__dso_handle = nullptr;
+int __cxa_thread_atexit_impl(void (*func)(void *), void *arg, void *dso);
+}
+
+int call_num = 0;
+
+[[gnu::destructor]]
+void check() {
+  // This destructor should be called only once.
+  if (call_num != 1)
+    __builtin_trap();
+}
+
+TEST_MAIN() {
+  __cxa_thread_atexit_impl([](void *) { LIBC_NAMESPACE::exit(0); }, nullptr,
+                           __dso_handle);
+  __cxa_thread_atexit_impl([](void *) { ++call_num; }, nullptr, __dso_handle);
+  LIBC_NAMESPACE::exit(1);
+}

libc/test/integration/src/__support/threads/main_exit_test.cpp

@@ -0,0 +1,30 @@
+//===-- Test handling of thread local data --------------------------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/__support/threads/thread.h"
+#include "test/IntegrationTest/test.h"
+
+bool called = false;
+
+extern "C" {
+[[gnu::weak]]
+void *__dso_handle = nullptr;
+int __cxa_thread_atexit_impl(void (*func)(void *), void *arg, void *dso);
+}
+
+[[gnu::destructor]]
+void destructor() {
+  if (!called)
+    __builtin_trap();
+}
+
+TEST_MAIN() {
+  __cxa_thread_atexit_impl([](void *) { called = true; }, nullptr,
+                           __dso_handle);
+  return 0;
+}