sepfy
binding request timeout
两个设备属于对称型穿透，采集端的设备使用stun链接，接收端的设备使用turn链接，通过turn服务帮忙转发，发现测试一段时间后流就异常停止了，设备端打印“binding request timeout”，状态切换为closed，问题是必现的，不知是否有什么排查方法。
DEBUG libpeer/src/peer_connection.c 426 agent_recv 70 DEBUG libpeer/src/peer_connection.c 429 Got RTCP packet DEBUG libpeer/src/peer_connection.c 103 RTCP_PR DEBUG libpeer/src/peer_connection.c 119 RTCP_PSFB 15 DEBUG libpeer/src/peer_connection.c 426 agent_recv 70 DEBUG libpeer/src/peer_connection.c 429 Got RTCP packet DEBUG libpeer/src/peer_connection.c 103 RTCP_PR DEBUG libpeer/src/peer_connection.c 119 RTCP_PSFB 15 DEBUG libpeer/src/peer_connection.c 426 agent_recv 70 DEBUG libpeer/src/peer_connection.c 429 Got RTCP packet DEBUG libpeer/src/peer_connection.c 103 RTCP_PR DEBUG libpeer/src/peer_connection.c 119 RTCP_PSFB 15 DEBUG libpeer/src/peer_connection.c 426 agent_recv 61 DEBUG libpeer/src/dtls_srtp.c 485 DTLS content type: 23 DEBUG libpeer/src/peer_connection.c 435 Got DTLS data 32 DEBUG libpeer/src/sctp.c 451 Data of length 4 received on stream 0 with SSN 39, TSN 703256564, PPID 51 DEBUG libpeer/src/sctp.c 401 Got message (size = 4) on message: 0 ping, send pong DEBUG libpeer/src/peer_connection.c 426 agent_recv 70 DEBUG libpeer/src/peer_connection.c 429 Got RTCP packet DEBUG libpeer/src/peer_connection.c 103 RTCP_PR DEBUG libpeer/src/peer_connection.c 119 RTCP_PSFB 15 DEBUG libpeer/src/peer_connection.c 426 agent_recv 57 DEBUG libpeer/src/dtls_srtp.c 485 DTLS content type: 23 DEBUG libpeer/src/peer_connection.c 435 Got DTLS data 28 DEBUG libpeer/src/sctp.c 451 Data of length 12 received on stream 0 with SSN 0, TSN 0, PPID 0 DEBUG libpeer/src/peer_connection.c 426 agent_recv 70 DEBUG libpeer/src/peer_connection.c 429 Got RTCP packet DEBUG libpeer/src/peer_connection.c 103 RTCP_PR DEBUG libpeer/src/peer_connection.c 119 RTCP_PSFB 15 DEBUG libpeer/src/peer_connection.c 426 agent_recv 70 DEBUG libpeer/src/peer_connection.c 429 Got RTCP packet DEBUG libpeer/src/peer_connection.c 103 RTCP_PR DEBUG libpeer/src/peer_connection.c 119 RTCP_PSFB 15 INFO libpeer/src/peer_connection.c 459 binding request timeout state is changed: closed
Hi, I have a few suggestions maybe can help you figure out something.
First, you can modify the code like:
if ((pc->agent_ret = agent_recv(&pc->agent, pc->agent_buf, sizeof(pc->agent_buf))) > 0) { codes ... } else { // do that in else case only, maybe not a good idea. if (KEEPALIVE_CONNCHECK > 0 && (ports_get_epoch_time() - pc->agent.binding_request_time) > KEEPALIVE_CONNCHECK) { LOGI("binding request timeout"); STATE_CHANGED(pc, PEER_CONNECTION_CLOSED); } }
Second, you can set KEEPALIVE_CONNCHECK to a larger interval, like 90000. Third, you can see that 'agent->binding_request_time' is set in function 'agent_process_stun_request' when processing STUN_METHOD_BINDING, you can add logs before and after 'stun_msg_is_valid' like 'LOGD("stun request in STUN_METHOD_BINDING:%d", stun_msg->stunmethod);' to check if msg exists.
After these changes, does this still happen?
I think we need to update libpeer to take SCTP/RTP/RTCP traffic into account.
A remote peer doesn't need to send STUN Bindings if it has sent other traffic. libpeer is seeing no ICE traffic, and incorrectly assuming the PeerConnection is disconnected.
I will open a PR for this soon!
This can also happen for normal WebRTC connections from browsers, as the STUN bindings should be sent no less than every 15 seconds per RFC 8445, which is longer than the default 10 seconds for CONFIG_KEEPALIVE_TIMEOUT.
