Interoperability issue when BBL acts as a PPPoE client against an L2TP LAC/LNS

[alexk99]

Test topology:

• BBL = PPPoE client
• BR1 = LAC
• BR2 = LNS

Flow:

1. BBL establishes PPPoE with the LAC.
2. The LAC completes the initial PPP LCP phase locally in order to receive PAP username/password.
3. After receiving the username, the LAC establishes an L2TP session to the LNS.
4. The LNS then starts a fresh PPP negotiation on the L2TP session with full LCP negotiation and no LCP proxying.

Observed BBL behavior:
After PAP, when the LNS starts sending new LCP Configure-Request packets, BBL responds with Configure-Ack, but does not restart its own LCP negotiation and does not send a new Configure-Request.

As a result, LCP on the L2TP side never converges: the LNS keeps retransmitting its Configure-Request, while BBL only ACKs it and never starts the peer-to-LNS direction of negotiation.

[GIC-de]

I will work on this, but you need to verify before releasing.

[alexk99]

Sure, I will test it.

So far, I've been using this patch, which works fine. But I guess it could be a naive fix that breaks other things

static void
bbl_access_lcp_restart(bbl_session_s *session)
{
    /* Stop activity from the previous PPP phase. */
    timer_del(session->timer_auth);
    timer_del(session->timer_lcp_echo);
    timer_del(session->timer_ipcp);
    timer_del(session->timer_ip6cp);

    session->lcp_state = BBL_PPP_PEER_ACK;
    session->lcp_request_code = PPP_CODE_CONF_REQUEST;
    session->lcp_retries = 0;
    session->auth_retries = 0;

    if (session->ipcp_state > BBL_PPP_DISABLED)
        session->ipcp_state = BBL_PPP_CLOSED;
    
    if (session->ip6cp_state > BBL_PPP_DISABLED)
        session->ip6cp_state = BBL_PPP_CLOSED;

    session->send_requests &= ~(BBL_SEND_PAP_REQUEST |
                                BBL_SEND_CHAP_RESPONSE |
                                BBL_SEND_IPCP_REQUEST |
                                BBL_SEND_IPCP_RESPONSE |
                                BBL_SEND_IP6CP_REQUEST |
                                BBL_SEND_IP6CP_RESPONSE);

    bbl_session_update_state(session, BBL_PPP_LINK);
    session->send_requests |= BBL_SEND_LCP_REQUEST;
}

bbl_access_lcp_opened()

case BBL_PPP_OPENED:
    /*
        * LAC/LNS handoff case:
        * peer starts a fresh LCP negotiation after an earlier
        * PPP phase already reached OPENED on the PPPoE side.
        */
    bbl_access_lcp_restart(session);
    break;

[GIC-de]

Please verify with latest dev branch (c13c4b4).

To test this, I also added a new command to trigger BNG Blaster LNS to restart LCP by sending LCP configuration request.

sudo bngblaster-cli run.sock l2tp-lcp-restart session-id 1

[alexk99]

I did not use L2TP on the BBL side in my initial test. I am using BBL as a PPPoE client, so I am not sure this command is appropriate for my case: sudo bngblaster-cli run.sock l2tp-lcp-restart session-id 1

[GIC-de]

This command was only needed for me to test this without having a "real" LNS doing this. So yes, you do not need this command.

[alexk99]

It works! Thank you.