Fix linearizer cancellation on twisted < 18.7

Turns out that cancellation of inlineDeferreds didn't really work properly
until Twisted 18.7. This commit refactors Linearizer.queue to avoid
inlineCallbacks.

1 files changed, 68 insertions, 43 deletions

diff --git a/synapse/util/async.py b/synapse/util/async.py
index a7094e2fb4..9b3f2f4b96 100644
--- a/synapse/util/async.py
+++ b/synapse/util/async.py
@@ -188,62 +188,30 @@ class Linearizer(object):
         # things blocked from executing.
         self.key_to_defer = {}
 
-    @defer.inlineCallbacks
     def queue(self, key):
+        # we avoid doing defer.inlineCallbacks here, so that cancellation works correctly.
+        # (https://twistedmatrix.com/trac/ticket/4632 meant that cancellations were not
+        # propagated inside inlineCallbacks until Twisted 18.7)
         entry = self.key_to_defer.setdefault(key, [0, collections.OrderedDict()])
 
         # If the number of things executing is greater than the maximum
         # then add a deferred to the list of blocked items
-        # When on of the things currently executing finishes it will callback
+        # When one of the things currently executing finishes it will callback
         # this item so that it can continue executing.
         if entry[0] >= self.max_count:
-            new_defer = defer.Deferred()
-            entry[1][new_defer] = 1
-
-            logger.info(
-                "Waiting to acquire linearizer lock %r for key %r", self.name, key,
-            )
-            try:
-                yield make_deferred_yieldable(new_defer)
-            except Exception as e:
-                if isinstance(e, CancelledError):
-                    logger.info(
-                        "Cancelling wait for linearizer lock %r for key %r",
-                        self.name, key,
-                    )
-                else:
-                    logger.warn(
-                        "Unexpected exception waiting for linearizer lock %r for key %r",
-                        self.name, key,
-                    )
-
-                # we just have to take ourselves back out of the queue.
-                del entry[1][new_defer]
-                raise
-
-            logger.info("Acquired linearizer lock %r for key %r", self.name, key)
-            entry[0] += 1
-
-            # if the code holding the lock completes synchronously, then it
-            # will recursively run the next claimant on the list. That can
-            # relatively rapidly lead to stack exhaustion. This is essentially
-            # the same problem as http://twistedmatrix.com/trac/ticket/9304.
-            #
-            # In order to break the cycle, we add a cheeky sleep(0) here to
-            # ensure that we fall back to the reactor between each iteration.
-            #
-            # (This needs to happen while we hold the lock, and the context manager's exit
-            # code must be synchronous, so this is the only sensible place.)
-            yield self._clock.sleep(0)
-
+            res = self._await_lock(key)
         else:
             logger.info(
                 "Acquired uncontended linearizer lock %r for key %r", self.name, key,
             )
             entry[0] += 1
+            res = defer.succeed(None)
+
+        # once we successfully get the lock, we need to return a context manager which
+        # will release the lock.
 
         @contextmanager
-        def _ctx_manager():
+        def _ctx_manager(_):
             try:
                 yield
             finally:
@@ -264,7 +232,64 @@ class Linearizer(object):
                     # map.
                     del self.key_to_defer[key]
 
-        defer.returnValue(_ctx_manager())
+        res.addCallback(_ctx_manager)
+        return res
+
+    def _await_lock(self, key):
+        """Helper for queue: adds a deferred to the queue
+
+        Assumes that we've already checked that we've reached the limit of the number
+        of lock-holders we allow. Creates a new deferred which is added to the list, and
+        adds some management around cancellations.
+
+        Returns the deferred, which will callback once we have secured the lock.
+
+        """
+        entry = self.key_to_defer[key]
+
+        logger.info(
+            "Waiting to acquire linearizer lock %r for key %r", self.name, key,
+        )
+
+        new_defer = make_deferred_yieldable(defer.Deferred())
+        entry[1][new_defer] = 1
+
+        def cb(_r):
+            logger.info("Acquired linearizer lock %r for key %r", self.name, key)
+            entry[0] += 1
+
+            # if the code holding the lock completes synchronously, then it
+            # will recursively run the next claimant on the list. That can
+            # relatively rapidly lead to stack exhaustion. This is essentially
+            # the same problem as http://twistedmatrix.com/trac/ticket/9304.
+            #
+            # In order to break the cycle, we add a cheeky sleep(0) here to
+            # ensure that we fall back to the reactor between each iteration.
+            #
+            # (This needs to happen while we hold the lock, and the context manager's exit
+            # code must be synchronous, so this is the only sensible place.)
+            return self._clock.sleep(0)
+
+        def eb(e):
+            logger.info("defer %r got err %r", new_defer, e)
+            if isinstance(e, CancelledError):
+                logger.info(
+                    "Cancelling wait for linearizer lock %r for key %r",
+                    self.name, key,
+                )
+
+            else:
+                logger.warn(
+                    "Unexpected exception waiting for linearizer lock %r for key %r",
+                    self.name, key,
+                )
+
+            # we just have to take ourselves back out of the queue.
+            del entry[1][new_defer]
+            return e
+
+        new_defer.addCallbacks(cb, eb)
+        return new_defer
 
 
 class ReadWriteLock(object):