dpp: handle protocol errors in ROC timeout

This is a standing TODO of properly handling these timeouts but
for now just treat any ROC timeout as an error if authenticating
or configuring.

src/dpp.c

@@ -1076,10 +1076,34 @@ static void dpp_presence_timeout(int error, void *user_data)
 {
 	struct dpp_sm *dpp = user_data;
 
-	if (dpp->state != DPP_STATE_PRESENCE) {
+	/*
-		l_debug("DPP state changed, stopping presence announcements");
+	 * If cancelled this is likely due to netdev going down or from Stop().
-		dpp->freqs_idx = 0;
+	 * Otherwise there was some other problem which is probably not
+	 * recoverable.
+	 */
+	if (error == -ECANCELED)
 		return;
+	else if (error < 0)
+		goto protocol_failed;
+
+	switch (dpp->state) {
+	case DPP_STATE_PRESENCE:
+		break;
+	case DPP_STATE_NOTHING:
+		/* Protocol already terminated */
+		return;
+	case DPP_STATE_AUTHENTICATING:
+	case DPP_STATE_CONFIGURING:
+		/*
+		 * TODO: If either the auth or config protocol is running we
+		 * need to stay on channel until the specified timeouts.
+		 * Unfortunately the kernel makes this very inconvenient since
+		 * there is no way to stay on channel indefinitely or any way
+		 * of knowing what duration the kernel/card actually chooses.
+		 *
+		 * For now just treat this as a failure.
+		 */
+		goto protocol_failed;
 	}
 
 	dpp->freqs_idx++;
@@ -1097,6 +1121,11 @@ static void dpp_presence_timeout(int error, void *user_data)
 	dpp->offchannel_id = offchannel_start(netdev_get_wdev_id(dpp->netdev),
 			dpp->current_freq, dpp->dwell, dpp_roc_started,
 			dpp, dpp_presence_timeout);
+	return;
+
+protocol_failed:
+	dpp_reset(dpp);
+	return;
 }
 
 /*