avahi-daemon rejects response packet if it contains the original question record

[avillacis]

Version: avahi-0.8-9.fc34.x86_64 on Fedora 34 x86_64

If a MDNS response packet received through port 5353 contains the original MDNS question record, avahi-daemon rejects further processing and logs "Invalid response packet from host ..." if running in verbose mode. Even if the response packet contains a valid answer record in addition to the question record.

I consider this behavior to be incorrect and needlessly strict. For reference, this is the relevant section from RFC 6762 section 6. Responding:

Multicast DNS responses MUST NOT contain any questions in the
Question Section. Any questions in the Question Section of a
received Multicast DNS response MUST be silently ignored. Multicast
DNS queriers receiving Multicast DNS responses do not care what
question elicited the response; they care only that the information
in the response is true and accurate.

I take this to mean that whether the response contains an echoed question record should be irrelevant to whether the rest of the packet is taken as valid or not. So, if the packet contains a response to an actual issued query, any echoed question in the packet should be ignored and the packet should be considered valid.

I regularly work with ESP32 microcontrollers which use the Espressif MDNS implementation. The Espressif MDNS implementation echoes the question in the response packet. This causes the .local name previously broadcasted by the device to be no longer useful for locating the device in the network, since the device actually responds the query, but avahi-daemon is rejecting the response.

[avillacis]

esp32-mdns-query-dropped.zip
This file contains a wireshark capture of the queries and (ignored) responses from the ESP32 device.

[d8ahazard]

Version: avahi-0.8-9.fc34.x86_64 on Fedora 34 x86_64

If a MDNS response packet received through port 5353 contains the original MDNS question record, avahi-daemon rejects further processing and logs "Invalid response packet from host ..." if running in verbose mode. Even if the response packet contains a valid answer record in addition to the question record.

I consider this behavior to be incorrect and needlessly strict. For reference, this is the relevant section from RFC 6762 section 6. Responding:

Multicast DNS responses MUST NOT contain any questions in the
Question Section. Any questions in the Question Section of a
received Multicast DNS response MUST be silently ignored. Multicast
DNS queriers receiving Multicast DNS responses do not care what
question elicited the response; they care only that the information
in the response is true and accurate.

I take this to mean that whether the response contains an echoed question record should be irrelevant to whether the rest of the packet is taken as valid or not. So, if the packet contains a response to an actual issued query, any echoed question in the packet should be ignored and the packet should be considered valid.

I regularly work with ESP32 microcontrollers which use the Espressif MDNS implementation. The Espressif MDNS implementation echoes the question in the response packet. This causes the .local name previously broadcasted by the device to be no longer useful for locating the device in the network, since the device actually responds the query, but avahi-daemon is rejecting the response.

Doing a quick grep of the code, the only place I find that error message is in /avahi-core/server.c:

`
if (avahi_dns_packet_get_field(p, AVAHI_DNS_FIELD_QDCOUNT) != 0 ||
avahi_dns_packet_get_field(p, AVAHI_DNS_FIELD_ANCOUNT) == 0 ||
avahi_dns_packet_get_field(p, AVAHI_DNS_FIELD_NSCOUNT) != 0) {

        avahi_log_debug("Invalid response packet from host %s.", avahi_address_snprint(t, sizeof(t), src_address));
        return;
    }

`

So, a bit of experimentation would show which of these checks need to be removed...but I'd guess it would be the first one for AVAHI_DNS_FIELD_QDCOUNT.