ChecksumType is missing two CRC-32 configurations that Checksum.h can already compute

[torokati44]

Type: enhancement
Component: src/inet/common/checksum, src/inet/protocolelement/checksum
Version: INET 4.7.0 (dfe270b21f, 2026-07-07)

Summary

ChecksumType exposes an arbitrary subset of the CRC-32 configurations Checksum.h supports. Two are missing:

• CRC-32/ISO-HDLC - has a named function, crc32_iso_hdlc(), but no enum entry; only its byte-swapped sibling CHECKSUM_ETHERNET_FCS has one.
• CRC-32/MPEG-2 (poly=0x04c11db7 init=0xffffffff refin=false refout=false xorout=0x00000000 check=0x0376e6e7) - no named function and no enum entry, although generic_crc32() computes it correctly today.

Since the enum is the only way into the checksum module machinery, a protocol using either one cannot use that machinery at all.

Current state

enum ChecksumType
{
    CHECKSUM_TYPE_UNDEFINED = -1;
    CHECKSUM_INTERNET = 1;
    CHECKSUM_CRC16_IBM = 3;
    CHECKSUM_CRC16_CCITT = 4;
    CHECKSUM_ETHERNET_FCS = 5;
    CHECKSUM_CRC32C = 6;
}

computeChecksum() and getChecksumSizeInBytes() throw "Unknown checksum type" outside that set, and parseChecksumType() accepts only internet, ethernet-fcs, crc32c, crc16-ibm, crc16-ccitt.

(Aside: value 2 is unused - was an entry removed? Reusing it would break result comparability against older .sca/.vec files, so new values look safer either way.)

Consequence

ChecksumInserterBase.ned:25 and ChecksumCheckerBase.ned:25 hardcode the list:

string checksumType @enum("internet","ethernet-fcs","crc32c","crc16-ibm","crc16-ccitt") = default("ethernet-fcs");

so ChecksumInserter / ChecksumChecker and everything built on them are unavailable to such a protocol, which then has to hand-roll its trailer chunk, serializer and checker, and ends up handling checksumMode differently from every other INET protocol.

Substituting the nearest available value is not a way out: crc32_iso_hdlc is reflected with xorout=0xffffffff and ethernetFcs is that byte-swapped, so either puts wrong bytes on the wire - silently, since the same wrong function on both ends still validates.

Suggested fix

Only the surface is missing; the arithmetic exists.

1. Named wrappers in Checksum.h following the existing crc32c / crc32_iso_hdlc pattern - crc32_mpeg2(), the std::vector<uint8_t> overload, and a crc32_mpeg2_bitwise() delegating to generic_crc32(buf, bufsize, 0x04c11db7, 0xffffffff, false, false, 0x00000000).
2. CHECKSUM_CRC32_MPEG2 and CHECKSUM_CRC32_ISO_HDLC in ChecksumType, with new values rather than reusing 2.
3. Extend parseChecksumType() ("crc32-mpeg2", "crc32-iso-hdlc"), getChecksumSizeInBytes() (4 bytes each) and computeChecksum().
4. Extend the @enum(...) lists in ChecksumInserterBase.ned and ChecksumCheckerBase.ned.

Check values for the test alongside Checksum_1.test, with crc32_iso_hdlc shown for contrast:

Message	CRC-32/MPEG-2	CRC-32/ISO-HDLC
"123456789"	0x0376E6E7	0xCBF43926
"Lorem ipsum dolor sit amet"	0x1C2A1845	0x5F29D461

Both first-row values are the published CRC RevEng check values, so the configurations are identifiable independently of any implementation. The second row reuses the string already in Checksum_1.test.