jsonrpc: Fix Python implementation of inactivity logic.

When a JSON-RPC session receives bytes, or when it successfully sends
queued bytes, then it should count that as activity.  However, the code
here was reversed, in that it used the wrong check in each place.  That is,
when it tried to receive data, it would check whether data had just been
sent, and when it tried to send data, it would check whether data had just
been received.  Neither one makes sense and doesn't work.

Bug #13214.
Reported-by: Luca Giraudo <[email protected]>
CC: James Schmidt <[email protected]>
Signed-off-by: Ben Pfaff <[email protected]>

python/ovs/jsonrpc.py

@@ -449,14 +449,16 @@ def run(self):
                 self.pstream = None
 
         if self.rpc:
-            received_bytes = self.rpc.get_received_bytes()
+            backlog = self.rpc.get_backlog()
             self.rpc.run()
-            if received_bytes != self.rpc.get_received_bytes():
-                # Data was successfully received.
+            if self.rpc.get_backlog() < backlog:
+                # Data previously caught in a queue was successfully sent (or
+                # there's an error, which we'll catch below).
                 #
-                # Previously we only counted receiving a full message as
-                # activity, but with large messages or a slow connection that
-                # policy could time out the session mid-message.
+                # We don't count data that is successfully sent immediately as
+                # activity, because there's a lot of queuing downstream from
+                # us, which means that we can push a lot of data into a
+                # connection that has stalled and won't ever recover.
                 self.reconnect.activity(ovs.timeval.msec())
 
             error = self.rpc.get_status()
@@ -516,16 +518,14 @@ def send(self, msg):
 
     def recv(self):
         if self.rpc is not None:
-            backlog = self.rpc.get_backlog()
+            received_bytes = self.rpc.get_received_bytes()
             error, msg = self.rpc.recv()
-            if self.rpc.get_backlog() < backlog:
-                # Data previously caught in a queue was successfully sent (or
-                # there's an error, which we'll catch below).
+            if received_bytes != self.rpc.get_received_bytes():
+                # Data was successfully received.
                 #
-                # We don't count data that is successfully sent immediately as
-                # activity, because there's a lot of queuing downstream from
-                # us, which means that we can push a lot of data into a
-                # connection that has stalled and won't ever recover.
+                # Previously we only counted receiving a full message as
+                # activity, but with large messages or a slow connection that
+                # policy could time out the session mid-message.
                 self.reconnect.activity(ovs.timeval.msec())
 
             if not error: