mDNS: when a reply do not repeat the questions, set authoritative flag

[htibosch]

Description

Both mDNS and LLMNR work fine, except whan it receives a request from a Linux machine: we found that an mDNS server must set the authoritative flag, and also that it shall not repeat the Questions from the requester.
Other OSs like Windows are more forgiving.

1. Set a new variable xSet.uxSkipCount as soon as the full name in Questions has been parsed. Why: because the Questions section will become the Answers section, containing:

   Name: 'full_name.local', Type A, Class IN, TTL, Addr length, Address.
   Or the name followed by 16 or 28 bytes.

2. When replying to an mDNS request, do not include the Questions section: 'xSet.usQuestions' will become zero.

3. With thanks to @ActoryOu : mDNS sets the authoritative flag when replying. See PR Set MDNS AA bit while answering MDNS request. #1223

4. Do not handle type A when IPv4 is not enabled
   Do not handle type AAAA when IPv6 is not enabled

5. Make sure that 'pxNewBuffer' will be released just before exiting the function DNS_ParseDNSReply()

Test Steps

I tested while enabling either ipconfigUSE_MDNS or ipconfigUSE_LLMNR. The former replies with an Answers section only, the latter repeats the questions and then sends the Answers.

Checklist:

• I have tested my changes. No regression in existing tests.
• I have modified and/or added unit-tests to cover the code changes in this Pull Request.

Related Issue

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[htibosch]

/bot run formatting

[moninom1]

/bot run formatting

[moninom1]

@htibosch Seems like UTs are also failing, can you please help in checking on that. Thanks

[bjsowa]

@htibosch I tried your changes and it fixed mDNS for me. I found, however, two small issues that still persist:

1. When IPv4 endpoint receives AAAA question, it answers with empty IP (::)
2. When a single query contains two or more questions (for example one A and one AAAA type), FreeRTOS+TCP will only answer the last one.

[htibosch]

/bot run formatting

[htibosch]

/bot run formatting

[htibosch]

@bjsowa wrote:

1. When IPv4 endpoint receives AAAA question, it answers with empty IP (::)

I just checked these commands from a laptop:

ping -6 zynq.local
ping -4 zynq.local

and I had these expected results, see this PCAP file: samples_mdns.zip

My configuration is:

IP-address : 192.168.2.127
End-point  : up = yes method static
Net mask   : 255.255.255.0
GW         : 192.168.2.1
DNS-0      : 118.98.44.10
DNS-1      : 118.98.44.100
Broadcast  : 192.168.2.255
MAC address: 00-11-11-11-11-42
 
IP-address : 2600:70ff:c040:0:6a05:0:811:cc55
Default IP : 2600:70ff:c040::2001
End-point  : up = yes method RA
Prefix     : 2600:70ff:c040::/64
GW         : fe80::ba27:ebff:fe5a:d751
DNS-0      : fe80::1
DNS-1      : 2001:4860:4860::8888
MAC address: 00-11-11-11-11-42
 
IP-address : fe80::7004
End-point  : up = yes method static
Prefix     : fe80::/10
GW         : ::
MAC address: 00-11-11-11-11-42

[htibosch]

Please find attached some code that I use for xApplicationDNSQueryHook_Multi():

dns_query_hook.zip

It will check the IP-type that is desired by the requester. Maybe that will solve the problem of the empty IPv6 address.

[bjsowa]

Please find attached some code that I use for xApplicationDNSQueryHook_Multi():

dns_query_hook.zip

It will check the IP-type that is desired by the requester. Maybe that will solve the problem of the empty IPv6 address.

This feels like a workaround rather than an actual solution. Why not add this logic to FreeRTOS+TCP?

[htibosch]

This feels like a workaround rather than an actual solution. Why not add this logic to FreeRTOS+TCP?

I can not agree more! I used mentioned code within a private project, and I am thinking of making this automatic, and downward-compatible.
I would love to hear a proposal.