mdns: fixed crashes on network changes

1) two events AP_STOP, AP_START shortly after each other may cause IGMP config on already stopped netif
2) not properly locked sending packets to queue from timer task

closes #2580


* Original commit: espressif/esp-idf@097282a

Author: david-cermak

components/mdns/mdns.c

@@ -3794,13 +3794,14 @@ static esp_err_t _mdns_send_search_action(mdns_action_type_t type, mdns_search_o
  */
 static void _mdns_scheduler_run()
 {
+    MDNS_SERVICE_LOCK();
     mdns_tx_packet_t * p = _mdns_server->tx_queue_head;
     mdns_action_t * action = NULL;
 
     if (!p) {
+        MDNS_SERVICE_UNLOCK();
         return;
     }
-    MDNS_SERVICE_LOCK();
     if ((int32_t)(p->send_at - (xTaskGetTickCount() * portTICK_PERIOD_MS)) < 0) {
         action = (mdns_action_t *)malloc(sizeof(mdns_action_t));
         if (action) {

components/mdns/mdns_networking.c

@@ -62,6 +62,12 @@ static esp_err_t _udp_join_group(tcpip_adapter_if_t tcpip_if, mdns_ip_protocol_t
 {
     struct netif * netif = NULL;
     void * nif = NULL;
+
+    if (!tcpip_adapter_is_netif_up(tcpip_if)) {
+        // Network interface went down before event propagated, skipping IGMP config
+        return ESP_ERR_INVALID_STATE;
+    }
+
     esp_err_t err = tcpip_adapter_get_netif(tcpip_if, &nif);
     if (err) {
         return ESP_ERR_INVALID_ARG;