eth: cancel DAO challenge on peer drop (annoying log) #2833
Conversation
Updated: Tue Jul 19 09:43:07 UTC 2016 |
| @@ -295,6 +295,13 @@ func (pm *ProtocolManager) handle(p *peer) error { | |||
| glog.V(logger.Warn).Infof("%v: timed out DAO fork-check, dropping", p) | |||
| pm.removePeer(p.id) | |||
| }) | |||
| // Make sure it's cleaned up if the peer dies off | |||
| defer func() { | |||
| if p.forkDrop != nil { | |||
There was a problem hiding this comment.
What may set this to nil other than the line below?
There was a problem hiding this comment.
The reply to the challenge in:
There was a problem hiding this comment.
PS: They set it to nil so those parts of the code that check headers for challenges is completely disabled after the challenge passed.
There was a problem hiding this comment.
Could this therefor not be racey?
There was a problem hiding this comment.
It's all in a single thread: the defer is in the same method that calls handleMsg for every network packet.
|
The frequency in which the messages were churned out before has been halted to a minimum. I still get it occasionally but assume that it actually does timeout for those instances. |
|
We can hide the message completely after a while (i.e. debug log) but I'd keep it around for a bit just to see that everything behaves well. |
When peers are disconnected for some reason other than the DAO challenge (either locally or remotely), but the challenge is still running, after the 15 sec timeout a log will appear warning of a disconnect (of the already disconnected peer). Although it's fully harmless, it's annoying. This PR makes sure the challenge timer is aborted when the peer drops.