common: add: a PACKET_TOO_BIG packet drop reason

[adamgeorge309]

PacketDropReason gains PACKET_TOO_BIG = 16, for a datagram that is larger than the Maximum Transmission Unit (MTU) of the next link and may not be fragmented. Ipv4 and Mpls report it where they reported OTHER_PACKET_DROP, and Ipv4 records it in a new statistic, packetDropPacketTooBig.

Risk: statistical results change. Ipv4's packetDropUndefined no longer counts these drops, and every Ipv4 module records two new scalars (packetDropPacketTooBig:count and :sum(packetBytes)). No fingerprint changes.

Closes #1359

One commit: common: add: a PACKET_TOO_BIG packet drop reason.

The problem

On master, Ipv4::fragmentAndSend() (src/inet/networklayer/ipv4/Ipv4.cc:980-989) drops a datagram with the Don't Fragment (DF) flag that does not fit the MTU, and emits packetDropped without setting a reason, so the reason is OTHER_PACKET_DROP. Mpls (Multiprotocol Label Switching, src/inet/networklayer/mpls/Mpls.cc:407-418) sets OTHER_PACKET_DROP explicitly for the same case with a label stack (Request for Comments (RFC) 3032 Section 3.4). Ipv4.ned counts these drops in packetDropUndefined, together with every drop that sets no reason.

The change

• src/inet/common/Simsignals.msg: PACKET_TOO_BIG = 16, appended, so no existing value is renumbered.
• src/inet/common/ResultFilters.cc: packetDropReasonIsPacketTooBig, registered next to the other reason filters.
• Ipv4::fragmentAndSend() sets PACKET_TOO_BIG; Mpls replaces OTHER_PACKET_DROP with it. No test covers the Mpls site, and Mpls.ned records no statistic on packetDropped.
• Ipv4.ned: @statistic[packetDropPacketTooBig], with the same record and interpolationmode as the other per-reason statistics.
• WHATSNEW: entry 50 under "Notable backward incompatible changes", because packetDropUndefined results change.

PacketDropVisualizer resolves the reason name through cEnum (PacketDropVisualizerBase.cc), so %s shows PACKET_TOO_BIG without a code change; %r shows 16.

No C++ interface, Network Description (NED) parameter, message field or feature descriptor changes; the enum gains one value, and Ipv4.ned one statistic.

Not addressed here: Mpls.ned declares neither @signal[packetDropped] nor a statistic, although Mpls.cc emits the signal; Ipv6 emits no packetDropped for this case on master (pull request #1284 adds it).

Verification

Release build (make MODE=release in the root and tests/module/lib), each suite run on master bba1db4c81 and with this change.

New module test tests/module/IPv4_dont_fragment_drop_reason.test: hostA --- router --- hostB, the router's link to hostB has an MTU of 576 bytes (**.router.eth[1].mac.mtu = 576B), and hostA's UdpBasicApp sends three datagrams with a 1000-byte payload with dontFragment = true. It checks that the router's ipv4.ip records packetDropPacketTooBig:count 3 and packetDropUndefined:count 0. On master's source the router records packetDropUndefined:count 3 and the test fails.

Command	Master	This change
tests/fingerprint: ./fingerprinttest -s -F tyf	1774 run, 5 failures, 62 errors, 1 expected error	the same failing and erroring rows, with the same computed fingerprints
tests/module: inet_run_module_tests -m release --no-build -l ERROR	657 run, 368 pass, 289 fail	658 run, 369 pass, 289 fail: the new test passes, and the same tests fail by name

Of the 289 module failures on master, 274 are tcp_* tests and 15 are ConvolutionalCoder12, ConvolutionalCoder34, DimensionalMediumAnalogModel_PrintAttenuation_1, EtherHost_lifecycle, ExternalProcess_3, Ieee80211BitDomain, Ieee80211BlockAckVariant_1, Ieee80211SymbolDomain, Ieee8021d-Rstp, Ieee8021d-Stp, IPv6_packet_too_big, LogNormalShadowing_Correlated_4, MIPv6_tcp_handover, PacketGate_1, UDPSocket_1. The 5 fingerprint failures are rows of mipv6-refactoring.csv whose recorded ~tNlb fingerprints are stale on master; the 62 errors are configurations whose optional features are not built.

Statistical baselines (statistics repository, origin/master): the routers of examples/inet/tcp_pmtud (PmtudEnabled 15, PmtudProbe 30) and examples/quic/dplpmtud (6 configurations, 1 to 18) record nonzero packetDropUndefined counts. With this change, all eight configurations record the same counts as packetDropPacketTooBig and packetDropUndefined 0. The other baselines with nonzero packetDropUndefined counts, examples/inet/multicast and examples/pim/iptv, record other drops and do not change. These baselines, and every baseline with an Ipv4 module (new scalars), need re-recording after merge.

Gates: check-pull-request.sh --base origin/master (commits, classification, includes, NED parameters, description), check-source-seals.sh --base origin/master and check-architecture.sh on src/inet/networklayer/ipv4 and src/inet/networklayer/mpls pass. check-architecture.sh src/inet/common reports include violations in ResultFilters.cc and other files; this commit adds no include.

Relationship to other pull requests

• ipv6: emit packetDropped where a router answers with an ICMPv6 error #1284 (for ipv6: a router emits no packetDropped when it discards a unicast datagram and answers with an ICMPv6 error #1282) makes Ipv6 emit packetDropped with OTHER_PACKET_DROP when it answers with Packet Too Big. Whichever merges later can switch that site to PACKET_TOO_BIG.
• ieee80211: repair Basic Block Ack agreement lifecycle #1343, ieee80211: complete Basic Block Ack transfers without queued DATA #1344, ieee80211: support immediate Compressed Block Ack #1345, ipv6: give each link a prefix of its own in Ipv6NetworkConfigurator #1351 and ipv6: advertise a default Router Lifetime of 3 * MaxRtrAdvInterval #1354 also add entries from number 50 under "Notable backward incompatible changes" in WHATSNEW; whichever merges later renumbers.