From 2d434675e5d5e2d111d7ca7ee53d9a4d367189dc Mon Sep 17 00:00:00 2001
From: Felix Weibach <felixweilbach@meta.com>
Date: Tue, 30 Dec 2025 07:43:29 -0800
Subject: [PATCH] Mitigate deadlock on DLL unload (#16416)

Summary:

ThreadPool gets stored in a static variable here extension/threadpool/threadpool.cpp:146

This means the destructor of ThreadPool will be run when the process exits or a DLL containing this code unloads.

While working with ExecuTorch I experienced a deadlock during unloading our DLL (which contained ExecuTorch) at runtime. This was caused by the pthreadpool_destroy function pthreadpool/src/windows.c:366 waiting forever on the worker threads.

Why this is happening exactly is unclear to me. It is likely a race condition inside Windows Parallel Loader (https://blogs.blackberry.com/en/2017/10/windows-10-parallel-loading-breakdown) as I could see its functions in the stack trace of the stuck worker threads after they returned from their main function.

The issue was mitigated on my side by calling `executorch::extension::threadpool::get_threadpool()->_unsafe_reset_threadpool(0);` before unloading the DLL.

This is just a workaround. I think a proper fix would be to rework the ThreadPool singleton and allow for explicit termination of it.

Differential Revision: D89889628
---
 extension/threadpool/threadpool.cpp | 6 ++++++
 extension/threadpool/threadpool.h   | 9 +++++++++
 2 files changed, 15 insertions(+)

diff --git a/extension/threadpool/threadpool.cpp b/extension/threadpool/threadpool.cpp
index e244954de6e..26acc9fb81b 100644
--- a/extension/threadpool/threadpool.cpp
+++ b/extension/threadpool/threadpool.cpp
@@ -75,6 +75,12 @@ bool ThreadPool::_unsafe_reset_threadpool(uint32_t new_thread_count) {
   return true;
 }
 
+void ThreadPool::_unsafe_destroy_threadpool() {
+  std::lock_guard<std::mutex> lock{mutex_};
+  ET_LOG(Info, "Destroying threadpool.");
+  threadpool_.reset();
+}
+
 void ThreadPool::run(
     runtime::FunctionRef<void(size_t)> fn,
     const size_t range) {
diff --git a/extension/threadpool/threadpool.h b/extension/threadpool/threadpool.h
index b4fe7e4be96..43957473a26 100644
--- a/extension/threadpool/threadpool.h
+++ b/extension/threadpool/threadpool.h
@@ -69,6 +69,15 @@ class ThreadPool final {
       "This API is experimental and may change without notice. Consider using UseNThreadsThreadPoolGuard")]]
   bool _unsafe_reset_threadpool(uint32_t num_threads);
 
+  /**
+   * INTERNAL: Destroys the threadpool. This is not a thread safe call. When calling this method,
+   * threads of the threadpool might be doing some work. Some other code may
+   * also be holding on to the threadpool pointer, that is no longer valid.
+   */
+  [[deprecated(
+      "This API is experimental and may change without notice.")]]
+  void _unsafe_destroy_threadpool();
+
   /**
    * Run, in parallel, function fn(task_id) over task_id in range [0, range).
    * This function is blocking.  All input is processed by the time it returns.