Worked example: the charger refuses the TCP connection
The tcp-refused demo read step by step: SLAC matches and SDP is answered, then each of four TCP SYN to the announced port gets a reset within a millisecond, and no V2G message is ever sent.
What you see
- Load the demo TCP refused (tcp-refused): SLAC, one SDP exchange and four short TCP attempts, 4.6 s in all, without a Control Pilot log.
- The protocol reads unknown: no V2G message was exchanged, so the app handshake never chose one.
- Stage strip: SLAC and SDP passed, TCP / TLS failed, everything after it not reached.
The verdict
Never charged, TCP failure, likely party EVSE, confidence high. The explanation reads: the connection was reset at 1.6 s before any V2G message was exchanged. The chain of evidence is TIM-04, TC-01 and TC-02.
Reading the capture step by step
- EV: CM_SLAC_PARM.REQ, 0.000 s
- EVSE: CM_SLAC_MATCH.CNF, 0.696 s: matched, 23.0 dB
- EV: SDP request, 1.500 s: no TLS, TCP
- EVSE: SDP response, 1.531 s: SECC at [fe80::2]:15118, no TLS, TCP
- EV: TCP SYN, 1.600 s: from port 49152 to [fe80::2]:15118
- EVSE: TCP RST, less than a millisecond later
- EV: TCP SYN, 2.600 s, 3.600 s and 4.600 s: from ports 49153, 49154 and 49155
- EVSE: TCP RST, each time, at once
- SLAC and discovery work: SL-01 to SL-08 pass, SD-01 finds the SDP answer after 31 ms, and SD-03 finds the car going to exactly the address and port the charger announced.
- The car does what it should with a refused connection: it waits a second and tries again from a new source port. Four attempts, four resets.
- The reset comes from the charger address and within a millisecond. That is the TCP stack of the charger answering for a port on which no program listens, not a timeout and not a lost packet.
- TC-01 reports that the charger never accepted the connection: a SYN to [fe80::2]:15118 at 1.600 s and no SYN-ACK. TC-02 reports the reset before any SessionStop. TIM-04 reports that SDP was followed by no SessionSetupRes at all.
What the cause is
- The charger announces a port in its SDP response on which nothing accepts connections. The host is up, since it answers SDP and resets the SYN, but the V2G server behind the port is not there.
- Common reasons: the SDP responder and the V2G server are configured with different ports; the V2G server has not started yet or has crashed after a previous session; it listens on another address or interface than the one announced; a firewall on the charger rejects the port.
- A charger that announced TLS but listens only for plain TCP, or the other way round, fails differently: the connection opens and the handshake breaks. Here the SDP response says no TLS, and the car asked for none.
What to fix
- Compare the port in the SDP response with the port the V2G server of the charger binds. They have to be the same, here 15118.
- Check on the charger that a process listens on that port and address while the car is plugged in, and that it comes back after a session ends or fails.
- Check the firewall rules of the charger for the link local IPv6 address of the PLC interface.
- SLAC matched, but no SDP answer or TCP connection covers the neighbouring failures: an SDP request that is never answered, and a connection that is never answered at all.
How to confirm the fix
- The first SYN gets a SYN-ACK and TC-01 passes with the time from SYN to SYN-ACK.
- supportedAppProtocolReq follows on the new connection, and TIM-04 passes once SessionSetupRes arrives.
- Try several plug-ins in a row: a server that dies after one session shows up only on the second.
Sources
- ISO 15118-2:2014, SECC Discovery Protocol and the TCP connection to the announced port
- RFC 9293, TCP, reset of a connection request to a port nobody listens on
- SPEC.md of this tool, FR-ENC-05, the failing demo sessions