Dev integration hein.v8

[htibosch]

Description

All sorts of tests have been performed:

This time I tested close to 100% of the protocols and situations.

Especially DNS and the local name-service protocols ( mDNS, LMNNR, NBNS ), both with IPv4 and IPv6 addresses.

Another important change is that UDPv6 packets will only be processed after the IP-address has been seen in a "Neighbor Advertisement".

ACK number in TCP RESET reply to SYN packet

ICMP/ping: active and passive, through the Internet and locally.

TCP active/passive, Internet and LAN

● FreeRTOS_UDP_IPv6.c : xProcessReceivedUDPPacket_IPv6(): removed test on ipLOCAL_IP_ADDRESS_POINTER

● FreeRTOS_UDP_IPv6.c : Received UDPv6 packets must also wait for address resolution, just like IPv4 packets: xCheckRequiresARPResolution(). For now, this will only be done for link-local addresses. Maybe rename xCheckRequiresARPResolution() to xCheckRequiresAddressResolution()?

● FreeRTOS_TCP_Transmission_IPV6.c : prvTCPSendSpecialPktHelper_IPV6() : applied the patch for sending correct seq in TCP RST packet.

● FreeRTOS_TCP_Transmission_IPV4.c : prvTCPSendSpecialPktHelper_IPV4() : applied the patch for sending correct seq in TCP RST

● FreeRTOS_TCP_IP.c : vTCPStateChange(): made important logging look nicer: both IPv4 and IPv6 addresses now shown correctly.

● FreeRTOS_Sockets.c : vTCPNetStat_TCPSocket(): corrected a FreeRTOS_printf() format.

● FreeRTOS_Routing.c : pxEasyFit() : made some changes after much testing

● FreeRTOS_ND.c : in various functions, allow that ppxEndPoint equals NULL

+    if( ppxEndPoint != NULL )
+    {
         *( ppxEndPoint ) = pxEndPoint;
+    }

● FreeRTOS_ND.c : vNDAgeCache(): when ucAge becomes zero, stop processing the record.

● FreeRTOS_ND.c : prvCheckWaitingBuffer() : when ND sollicitation is answered by an advertisement, process the packet in pxARPWaitingNetworkBuffer and clear it.

● FreeRTOS_ND.c : When ipICMP_NEIGHBOR_ADVERTISEMENT_IPv6 is received, call prvCheckWaitingBuffer()

● FreeRTOS_IP.c : Call xCheckRequiresARPResolution() for both IPv4 and IPv6 packets, before refreshing the MAC/address table

● FreeRTOS_DNS_Parser.c : DNS_ParseDNSReply() : instead of always using ipUDP_PAYLOAD_OFFSET_IPv4 as the UDP offset, look at the packets received.

+    uxUDPOffset = ( size_t ) ( pucUDPPayloadBuffer - pxNetworkBuffer->pucEthernetBuffer );
+    configASSERT( ( uxUDPOffset == ipUDP_PAYLOAD_OFFSET_IPv4 ) || ( uxUDPOffset == ipUDP_PAYLOAD_OFFSET_IPv6 ) );

● FreeRTOS_DNS_Parser.c : prepareReplyDNSMessage() : correct ucTimeToLive for mDNS packets.

● FreeRTOS_DNS_Networking.c: added a cast in a call to FreeRTOS_printf()

● FreeRTOS_ARP.c : vARPProcessPacketReply() : see if the IP-type of pxARPWaitingNetworkBuffer is IPv4

● FreeRTOS_ARP.c : xCheckRequiresARPResolution() : let it also check IPv6 frames, and have it call vNDSendNeighbourSolicitation() when necessary.

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

[moninom1]

Is this file duplicate in include folder?

[moninom1]

Why is ARP resolution needed here for IPv6?

[htibosch]

Yes maybe we should give that function a different name like: xCheckRequiresAddressResolution(). In this PR I added code to handle both IPv4 and IPv6 packets.

[shubnil]

Hi Hein, can you explain a little more on this, for which case do we need this

[htibosch]

As you know the SYN and the FIN packets carry a sort of pseudo data byte. It isn't there, but the counterparty checks our ACK to know that the SYN/FIN has been seen. In earlier versions, we just copied SEQ to ACK in this situation.

Now we get a SYN for a non-existing port number. We respond with a RESET packet.
In the above code, ulSequenceNumber is increased. Later on, in prvTCPReturnPacket(), the two fields SEQ and ACK will be swapped, so ACK will be as expected.

It was easy to test: try to connect to a non-existing port number e.g. using puTTY.