issue #414: reinitialize service pool lock on fork

docs/changelog.rst

@@ -298,6 +298,10 @@ Core Library
   was configured. This would lead to connection timeouts due to the hung
   response.
 
+* `#414 <https://github.com/dw/mitogen/issues/414>`_,
+  `#425 <https://github.com/dw/mitogen/issues/425>`_: avoid deadlock of forked
+  children by reinitializing the :mod:`mitogen.service` pool lock.
+
 * `#416 <https://github.com/dw/mitogen/issues/416>`_: around 1.4KiB of memory
   was leaked on every RPC, due to a list of strong references keeping alive any
   handler ever registered for disconnect notification.

mitogen/fork.py

@@ -85,6 +85,10 @@ def on_fork():
     mitogen.core.Latch._on_fork()
     mitogen.core.Side._on_fork()
 
+    mitogen__service = sys.modules.get('mitogen.service')
+    if mitogen__service:
+        mitogen__service._pool_lock = threading.Lock()
+
 
 def handle_child_crash():
     """