mDNS hosts fail to resolve with getaddrinfo when in APSTA mode

[kevinevans]

I'm trying to use getaddrinfo to resolve an mdns hostname. This seems to work fine when I'm running in AP mode but when I start using APSTA, getaddrinfo occasionally fails. I'm wondering if I'm initializing something incorrectly or if there's a timing issue here.

I have two ESP32-S3s that are both in APSTA mode. One of the devices (n2kwg-xxx) is a STA to the other device's soft AP (tcm-xxx). On the AP device, I'm calling getaddrinfo with the hostname of the n2kwg device:

Code: Select all

    const struct addrinfo hints = {
        .ai_family = AF_INET,
        .ai_socktype = SOCK_STREAM,
        .ai_flags = AI_ADDRCONFIG,
    };
    struct addrinfo* resolved = NULL;

    // [...]

        struct sockaddr_in dest_addr = {
            .sin_len = 0,
            .sin_family = AF_INET,
            .sin_port = (in_port_t)htons(s_settings->port),
            .sin_addr = {0},
        };
        int32_t error = getaddrinfo(s_settings->address, NULL, &hints, &resolved);

        if (error != 0 || resolved == NULL) {
            ESP_LOGE(TAG, "getaddrinfo failed for '%s', returned %" PRId32, s_settings->address, error);

            // TODO: freeaddrinfo if not null
            
            vTaskDelay(pdMS_TO_TICKS(10000));
            continue;
        }

In my logs, with mdns debugging enabled, I'm seeing that the mdns packets are arriving to the mdns service just fine.

Code: Select all

RX[1][0]: From: 192.168.5.2:5353, To: 224.0.0.251, Packet[9950]: AUTHORITATIVE
    A: _services._dns-sd._udp.local. PTR IN 4500[17] ._xxx._tcp.local.
    A: _xxx._tcp.local. PTR IN 4500[21] n2kwg-348518454ac0._xxx._tcp.local.
    A: n2kwg-348518454ac0._xxx._tcp.local. SRV IN FLUSH 120[27] 0 0 9 n2kwg-348518454ac0.local.
    A: n2kwg-348518454ac0._xxx._tcp.local. TXT IN FLUSH 4500[174] device_name=xxxxxx; services=_http._tcp, _n2k._tcp; device_type=n2kwg
    A: n2kwg-348518454ac0.local. A IN FLUSH 120[4] 192.168.5.2
    A: _services._dns-sd._udp.local. PTR IN 4500[7] ._n2k._tcp.local.
    A: _n2k._tcp.local. PTR IN 4500[21] n2kwg-348518454ac0._n2k._tcp.local.
    A: n2kwg-348518454ac0._n2k._tcp.local. SRV IN FLUSH 120[8] 0 0 2000 n2kwg-348518454ac0.local.
    A: n2kwg-348518454ac0._n2k._tcp.local. TXT IN FLUSH 4500[1] 
    A: _services._dns-sd._udp.local. PTR IN 4500[8] ._http._tcp.local.
    A: _http._tcp.local. PTR IN 4500[21] n2kwg-348518454ac0._http._tcp.local.
    A: n2kwg-348518454ac0._http._tcp.local. SRV IN FLUSH 120[8] 0 0 80 n2kwg-348518454ac0.local.
    A: n2kwg-348518454ac0._http._tcp.local. TXT IN FLUSH 4500[29] api_path=/api/v1/; web_path=/

E (26355) n2kwg: getaddrinfo failed for 'n2kwg-348518454ac0.local', returned 202

I can also see the mdns packets in Wireshark and it seems to be correct, but getaddrinfo is still failing. Any ideas at what's wrong here?

[kevinevans]

For reference, here's the order of what I'm doing on the soft AP device:

1. esp_netif_init
2. esp_event_loop_create_default
3. init wifi in soft ap mode
3. init mdns services
4. start the tcp client (which calls getaddrinfo)

[kevinevans]

My current shim is using `mdns_query_a` when getaddrinfo fails, lol