Fix long-standing bug where ReadWriteLock could drop logging contex…

…ts (#10993)

Use `PreserveLoggingContext()` to ensure that logging contexts are not
lost when exiting a read/write lock.

When exiting a read/write lock, callbacks on a `Deferred` are triggered
as a signal to any waiting coroutines. Any waiting coroutine that
becomes runnable is likely to follow the Synapse logging context rules
and will restore its own logging context, then either run to completion
or await another `Deferred`, resetting the logging context in the
process.

changelog.d/10993.misc

@@ -0,0 +1 @@
+Fix a long-standing bug where `ReadWriteLock`s could drop logging contexts on exit.

synapse/util/async_helpers.py

@@ -438,7 +438,8 @@ def _ctx_manager():
             try:
                 yield
             finally:
-                new_defer.callback(None)
+                with PreserveLoggingContext():
+                    new_defer.callback(None)
                 self.key_to_current_readers.get(key, set()).discard(new_defer)
 
         return _ctx_manager()
@@ -466,7 +467,8 @@ def _ctx_manager():
             try:
                 yield
             finally:
-                new_defer.callback(None)
+                with PreserveLoggingContext():
+                    new_defer.callback(None)
                 if self.key_to_current_writer[key] == new_defer:
                     self.key_to_current_writer.pop(key)