Fix: Delete allocation on TCP broken pipe #336
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
It seems many TURN/TCP clients out there implement a rather brute-force way to end TURN/TCP allocations: they spontaneously close the TCP connection underlying the allocation rather than properly closing it down by refreshing with zero lifetime. The server must delete the TURN allocation when it sees the client TCP connection has gone, otherwise the allocation will linger on and next time the client try to allocate from the same IP:port pair they receive a
relay already allocated for 5-TUPLE
error. Our users are actively being affected by this.This PR fixes this issue and adds a test. The easiest way to understand the issue is through the test:
The issue is fixed by explicitly deleting the allocation in the
Server.readListener
when thereadLoop
returns.