lorenzodonini / ocpp-go / 24061322700

83%

master: 83%

Pull #414

github

fix: prevent deadlock in websocket write during connection cleanup

When many connections disconnect simultaneously (e.g. cluster autoscaler
pod restart), a deadlock can occur between WriteManual/Close and cleanup:

1. WriteManual holds w.mutex.RLock and blocks on outQueue send (buffer full,
   writePump no longer reading because it exited to call cleanup)
2. cleanup waits for w.mutex.Lock, which requires all RLock holders to release
3. Neither can proceed -> deadlock

Additionally, server.Write held s.connMutex.RLock during the entire w.Write
call. If w.Write blocked, it prevented all handleDisconnect calls (which need
s.connMutex.Lock) across all connections, causing cascading "already exists"
errors when charge points attempted to reconnect.

Fix:
- Add a done channel to webSocket, closed at the start of cleanup
- Use select in WriteManual and Close to abort when done is signaled
- Release s.connMutex.RLock in server.Write before calling w.Write

12 of 16 new or added lines in 2 files covered. (75.0%)

3732 of 4504 relevant lines covered (82.86%)

18.18 hits per line