Contribution description

When implementing ICMPv6 for GNRC I also added build and send functions for ICMPv6 error messages, but I never actually used them while implementing IPv6. Reviewing that sub-module I found it quite buggy:

• Packets in receive order were assumed in send order
• Packets were sent out without an IP header attached
• Packets were just sent to the PID of the IP thread instead of searching the registry
• etc.

This PR fixes those issues. Additionally a call of the send functions is added were appropriate, so GNRC is now able to correctly (but optionally) report to a sender what went wrong. To prevent further degeneration of that module, I activated it for gnrc_networking (but also some tests might be nice, which is why I set the WIP label, though the tests can also come as a follow-up IMHO).

Testing procedures

I tested this using examples/gnrc_networking and scapy. If you get problems receiving you might need to merge #10356 first. The tests are required to be executed in order as commands previously executed in other tests might be a precondition:

1. Compile and start gnrc_networking (make sure gnrc_icmpv6_error is included; should be automatically as per this PR) for native

   make -C examples/gnrc_networking all term

2. Get the link-local IPv6 address and hardware address of the native instance:

   > ifconfig
   Iface  6  HWaddr: 36:54:72:4F:8B:1F 
             MTU:1500  HL:64  RTR  
             RTR_ADV  
             Source address length: 6
             Link type: wired
             inet6 addr: fe80::3454:72ff:fe4f:8b1f  scope: local  VAL
             inet6 group: ff02::2
             inet6 group: ff02::1
             inet6 group: ff02::1:ff4f:8b1f
             inet6 group: ff02::1a
             
             Statistics for Layer 2
               RX packets 32  bytes 6964
               TX packets 28 (Multicast: 6)  bytes 6980
               TX succeeded 28 errors 0
             Statistics for IPv6
               RX packets 32  bytes 6516
               TX packets 28 (Multicast: 6)  bytes 6588
               TX succeeded 28 errors 0
   

3. Get the link-local IPv6 address of your TAP interface/bridge
4. Start scapy with root privileges (required to send ethernet frames)
5. Set some variables:

   DST_ADDR = "<native link-local IPv6 address>"
   DST_HWADDR = "<native hardware address>"
   SRC_ADDR = "<TAP interface link-local IPv6 address>"

6. Test UDP "port unreachable" error:
   1. Port not registered:
      scapy:

      p = srp1(Ether(dst=DST_HWADDR) / IPv6(dst=DST_ADDR, src=SRC_ADDR) / UDP(),
                iface="tapbr0")
      assert(ICMPv6DestUnreach in p)
      assert(p[ICMPv6DestUnreach].code == 4) # Port unreachable

   2. Port registered:
      RIOT:

      udp server start 1337

      scapy:

      p = srp1(Ether(dst=DST_HWADDR) / IPv6(dst=DST_ADDR, src=SRC_ADDR) / UDP(dport=1337),
                iface="tapbr0", timeout=1)
      assert(p == None)

7. Check if large packets are carried correctly:

   # 1452 = 1500 (Ethernet payload length) - 40 (IPv6 header size) - 8 (UDP header size)
   p = srp1(Ether(dst=DST_HWADDR) / IPv6(dst=DST_ADDR, src=SRC_ADDR) / UDP() / (1452 * "x"),
                  iface="tapbr0")
   assert(ICMPv6DestUnreach in p)
   assert(p[ICMPv6DestUnreach].code == 4) # Port unreachable
   assert(len(p[Raw]) < 1452)

8. Test "address unreachable" (local address):

   p = srp1(Ether(dst=DST_HWADDR) / IPv6(dst="fe80::1", src=SRC_ADDR) / UDP(), iface="tapbr0", timeout=1)
   assert(ICMPv6DestUnreach in p)
   assert(p[ICMPv6DestUnreach].code == 3) # Address unreachable

9. Test "beyond scope of source address":

   p = srp1(Ether(dst=DST_HWADDR) / IPv6(dst="affe::1", src=SRC_ADDR) / UDP(), iface="tapbr0", timeout=1)
   assert(ICMPv6DestUnreach in p)
   assert(p[ICMPv6DestUnreach].code == 2) # Beyond scope of source address

10. Test "erroneous header field encountered":

    p = srp1(Ether(dst=DST_HWADDR) / IPv6(dst=DST_ADDR, src=SRC_ADDR, plen=20) / UDP(),
                   iface="tapbr0")
    assert(ICMPv6ParamProblem in p)
    assert(p[ICMPv6ParamProblem].code == 0) # erroneous header field encountered
    assert(p[ICMPv6ParamProblem].ptr == 4)  # IPv6 length field

11. Test "No route to destination"
    Linux: Configure route to native instance and set a global source address:

    sudo ip a a beef::1 dev tapbr0
    sudo ip r a affe::/64 via "<native link-local IPv6 address>" dev tapbr0

    RIOT: Configure route back to Linux:

    nib route add 6 beef::/64 "<TAP interface link-local IPv6 address>"
    

    scapy:

    p = srp1(Ether(dst=DST_HWADDR) / IPv6(dst="affe::1", src="beef::1") / UDP(), iface="tapbr0")
    assert(ICMPv6DestUnreach in p)
    assert(p[ICMPv6DestUnreach].code == 0) # No route to destination

12. Test "Address unreachable" (neighboring node):
    RIOT: Configure route to bogus neighbor:

    nib route add 6 affe::/64 fe80::1
    

    scapy:

    sendp(Ether(dst=DST_HWADDR) / IPv6(dst="affe::1", src="beef::1") / UDP(), iface="tapbr0")
    # srp1 would not return, for some reason
    # just check with wireshark or a second `scapy` instance using
    ps = sniff(iface="tapbr0", count=20)
    p = [p for p in ps if ICMPv6DestUnreach in p]
    assert(len(p) == 1)
    assert(p[0][ICMPv6DestUnreach].code == 3) # Address unreachable

13. Test "Hop limit exceeded in transit"

    p = srp1(Ether(dst=DST_HWADDR) / IPv6(dst="affe::1", src="beef::1", hlim=1) / UDP(), iface="tapbr0")
    assert(ICMPv6TimeExceeded in p)
    assert(p[ICMPv6TimeExceeded].code == 0) # Hop limit exceeded in transit

Issues/PRs references

None