tunnel: report which stage failed instead of guessing the cause

5 files changed, 93 insertions, 15 deletions

diff --git a/src/btfrv.cpp b/src/btfrv.cpp
index 133eb89..864fe3d 100644
--- a/src/btfrv.cpp
+++ b/src/btfrv.cpp
@@ -334,17 +334,29 @@ bool RvServiceWaitPaired(RvSocket s, int nTimeoutSecs)
 }
 
 RvSocket RvClientConnect(const char* relay_host, unsigned short port,
-                         const unsigned char target_pubkey[32])
+                         const unsigned char target_pubkey[32],
+                         RvClientFail* pFail)
 {
+    if (pFail) *pFail = RVC_FAIL_DIAL;
     if (!RvInit()) return RV_INVALID;
     SOCKET s = ConnectTo(relay_host, port);
     if (s == INVALID_SOCKET) return RV_INVALID;
+    if (pFail) *pFail = RVC_FAIL_PAIR;
     unsigned char hdr[33];
     hdr[0] = (unsigned char)'C';
     memcpy(hdr + 1, target_pubkey, 32);
     if (!WriteN(s, hdr, 33)) { CLOSESOCK(s); return RV_INVALID; }
     unsigned char paired = 0;
-    if (!ReadN(s, &paired, 1) || paired != 0x01) { CLOSESOCK(s); return RV_INVALID; }
+    if (!ReadN(s, &paired, 1)) { CLOSESOCK(s); return RV_INVALID; }
+    if (paired != 0x01)
+    {
+        // The relay answered and declined: HandleIncoming writes 0x00 and
+        // closes when no service is parked under this pubkey right now.
+        if (pFail) *pFail = RVC_FAIL_REFUSED;
+        CLOSESOCK(s);
+        return RV_INVALID;
+    }
+    if (pFail) *pFail = RVC_FAIL_NONE;
     return (RvSocket)s;
 }
 
diff --git a/src/btfrv.h b/src/btfrv.h
index 9cd7476..1e85cb0 100644
--- a/src/btfrv.h
+++ b/src/btfrv.h
@@ -57,10 +57,25 @@ RvSocket RvServiceRegister(const char* relay_host, unsigned short port,
 bool RvServiceWaitPaired(RvSocket s, int nTimeoutSecs = 0);
 
 // ---- client side (the node dialing a .btf address) ----
+// Where a failed RvClientConnect stopped. RV_INVALID alone collapses three
+// different situations -- relay unreachable, relay reachable but nobody
+// registered, connection died mid-pairing -- and they point at different
+// components. The caller can only report the right one if this side says
+// which it saw.
+enum RvClientFail
+{
+    RVC_FAIL_NONE = 0,
+    RVC_FAIL_DIAL,    // TCP connect to the relay itself failed
+    RVC_FAIL_REFUSED, // relay answered 0x00: no service registered under this pubkey
+    RVC_FAIL_PAIR,    // relay connection died before pairing completed
+};
+
 // Connect through the relay to the service registered under `target_pubkey`.
-// Returns the paired socket, or RV_INVALID on error.
+// Returns the paired socket, or RV_INVALID on error -- with `pFail`, if
+// given, naming the stage that failed.
 RvSocket RvClientConnect(const char* relay_host, unsigned short port,
-                         const unsigned char target_pubkey[32]);
+                         const unsigned char target_pubkey[32],
+                         RvClientFail* pFail = 0);
 
 // Blocking read/write of exactly n bytes over a rendezvous socket. Return false
 // on error/EOF.
diff --git a/src/btftunnel.cpp b/src/btftunnel.cpp
index 63d178b..408476e 100644
--- a/src/btftunnel.cpp
+++ b/src/btftunnel.cpp
@@ -136,18 +136,47 @@ static btf_socket_t StartTunnel(RvSocket rv, const unsigned char shared[32])
     return appEnd;
 }
 
+const char* TunnelFailText(TunnelFail f)
+{
+    switch (f)
+    {
+    case TUNNEL_FAIL_NONE:      return "no failure";
+    case TUNNEL_FAIL_INIT:      return "local socket/crypto init failed";
+    case TUNNEL_FAIL_DIAL:      return "meeting node unreachable (tcp dial failed)";
+    case TUNNEL_FAIL_REFUSED:   return "relay reachable, but no service registered there for this address";
+    case TUNNEL_FAIL_PAIR:      return "relay connection died before pairing completed";
+    case TUNNEL_FAIL_HANDSHAKE: return "end-to-end handshake failed";
+    case TUNNEL_FAIL_LOCAL:     return "local tunnel setup failed";
+    }
+    return "unknown failure";
+}
+
 btf_socket_t BtfClientTunnel(const char* meeting_host, unsigned short port,
                              const unsigned char target_pubkey[32],
-                             const unsigned char service_enc_pub[32])
+                             const unsigned char service_enc_pub[32],
+                             TunnelFail* pFail)
 {
+    if (pFail) *pFail = TUNNEL_FAIL_INIT;
     if (!TunnelInit()) return INVALID_SOCKET;
-    RvSocket rv = RvClientConnect(meeting_host, port, target_pubkey);
-    if (rv == RV_INVALID) return INVALID_SOCKET;
+    RvClientFail rvFail = RVC_FAIL_NONE;
+    RvSocket rv = RvClientConnect(meeting_host, port, target_pubkey, &rvFail);
+    if (rv == RV_INVALID)
+    {
+        if (pFail)
+            *pFail = rvFail == RVC_FAIL_DIAL    ? TUNNEL_FAIL_DIAL
+                   : rvFail == RVC_FAIL_REFUSED ? TUNNEL_FAIL_REFUSED
+                                                : TUNNEL_FAIL_PAIR;
+        return INVALID_SOCKET;
+    }
 
     unsigned char eph_pub[32], shared[32];
+    if (pFail) *pFail = TUNNEL_FAIL_HANDSHAKE;
     if (!ChanClientHandshake(service_enc_pub, eph_pub, shared)) { RvClose(rv); return INVALID_SOCKET; }
     if (!RvWriteN(rv, eph_pub, 32)) { RvClose(rv); return INVALID_SOCKET; }
-    return StartTunnel(rv, shared);
+    if (pFail) *pFail = TUNNEL_FAIL_LOCAL;
+    btf_socket_t app = StartTunnel(rv, shared);
+    if (app != INVALID_SOCKET && pFail) *pFail = TUNNEL_FAIL_NONE;
+    return app;
 }
 
 btf_socket_t BtfServiceWrap(RvSocket rv, const unsigned char my_enc_sk[32])
diff --git a/src/btftunnel.h b/src/btftunnel.h
index 24a7509..3cc87cd 100644
--- a/src/btftunnel.h
+++ b/src/btftunnel.h
@@ -36,12 +36,33 @@ namespace btf
 // Initialize (sockets + libsodium). Idempotent.
 bool TunnelInit();
 
+// Which stage of BtfClientTunnel failed. INVALID_SOCKET alone says only "no
+// tunnel"; a caller that then guesses at causes in its log line sends whoever
+// reads it to the wrong component (issue #17: "relay unreachable or pool not
+// listening" blamed two healthy parts). The stage is what this side actually
+// knows -- report that.
+enum TunnelFail
+{
+    TUNNEL_FAIL_NONE = 0,
+    TUNNEL_FAIL_INIT,      // local socket/crypto init failed
+    TUNNEL_FAIL_DIAL,      // TCP connect to the meeting node failed
+    TUNNEL_FAIL_REFUSED,   // relay reachable, but no service registered under the pubkey
+    TUNNEL_FAIL_PAIR,      // relay connection died before pairing completed
+    TUNNEL_FAIL_HANDSHAKE, // end-to-end channel handshake failed
+    TUNNEL_FAIL_LOCAL,     // local pump plumbing (socket pair / threads) failed
+};
+
+// Short description of a TunnelFail for log lines.
+const char* TunnelFailText(TunnelFail f);
+
 // Client: dial the service registered under `target_pubkey` at the meeting node
 // (host:port), box to its static x25519 `service_enc_pub`, and return a local
-// plaintext SOCKET tunneled end-to-end to the service. INVALID_SOCKET on error.
+// plaintext SOCKET tunneled end-to-end to the service. INVALID_SOCKET on error
+// -- with `pFail`, if given, naming the stage that failed.
 btf_socket_t BtfClientTunnel(const char* meeting_host, unsigned short port,
                              const unsigned char target_pubkey[32],
-                             const unsigned char service_enc_pub[32]);
+                             const unsigned char service_enc_pub[32],
+                             TunnelFail* pFail = 0);
 
 // Service: wrap an already-paired rendezvous socket (RvServiceRegister followed
 // by RvServiceWaitPaired -- the pairing byte MUST already be consumed) with
diff --git a/src/main.cpp b/src/main.cpp
index 7b3e555..c0a8825 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -3758,13 +3758,14 @@ static bool PoolParticipantMiner()
     // Step 2b: open encrypted tunnel to pool
     // ------------------------------------------------------------------
     SetParticipantMiningStatus("opening tunnel");
+    btf::TunnelFail tunnelFail = btf::TUNNEL_FAIL_NONE;
     s = btf::BtfClientTunnel(host.c_str(), (unsigned short)port,
-                             target_pubkey, service_enc_pub);
+                             target_pubkey, service_enc_pub, &tunnelFail);
     if (s == INVALID_SOCKET) {
-        LogPrint("worker", "[worker] tunnel connect FAILED via %s"
-                 " (relay unreachable or pool not listening)\n",
-                 meetingHostPort.c_str());
-        SetParticipantMiningStatus("tunnel connect failed");
+        LogPrint("worker", "[worker] tunnel connect FAILED via %s: %s\n",
+                 meetingHostPort.c_str(), btf::TunnelFailText(tunnelFail));
+        SetParticipantMiningStatus(std::string("tunnel connect failed: ")
+                                   + btf::TunnelFailText(tunnelFail));
         return false;
     }
     LogPrint("worker", "[worker] tunnel open via %s\n", meetingHostPort.c_str());