[JarryShaw]

Part of #2460.

What

LinuxSLL/LinuxSLL1/LinuxSLL2 fell back to Raw for the IPGRE (778)
and IP6GRE (823) arphrd_type values. This adds
pcapkit.protocols.link.gre.GRE, dissecting the Generic Routing
Encapsulation base header (RFC 2784) plus the optional Key and Sequence
Number extensions (RFC 2890), and wires LinuxSLL's shared
_import_next_layer to dispatch both ARPHRD_IPGRE and ARPHRD_IP6GRE
to this one class.

Why one class, two bindings

IPGRE and IP6GRE both land on the same GRE header — the "IP" vs "IP6" in
the arphrd_type name is about which IP version the Linux kernel's GRE
tunnel interface carries on its outer layer (i.e. which arphrd_type
fired), not a different GRE header shape. The GRE framing itself is
identical either way, so this is one dissector bound at two arphrd_type
values, exactly as #2460 scoped it.

What the dissector covers

The base header (RFC 2784 §2.1): a C/Reserved0/Ver flags-and-version
word, then a 16-bit Protocol Type (an EtherType). When C is set, a
Checksum/Reserved1 pair follows (§2.5-2.6). RFC 2890 §2 carves K (Key
Present) and S (Sequence Number Present) out of Reserved0, each gating
its own 4-octet field — so the header is 4, 8, 12 or 16 octets depending on
which of C/K/S are set.

The Protocol Type field dispatches onward as an EtherType through
Link.__proto__, the same family registry Ethernet, VLAN and LinuxSLL
already share — exactly as VLAN resolves its own type field. Unlike
FRAD (keyed by NLPID, not an EtherType), GRE therefore keeps no registry
of its own.

A nonzero Version Number (RFC 2637's "Enhanced GRE"/PPTP, version 1, a
different and incompatible header layout) is refused rather than misread.

Not implemented: per-family interpretation of the Key/Sequence Number
values — recorded but not interpreted further, since their meaning is a
matter for the tunnel endpoints, not something a passive dissector can
resolve from the header alone.

Registration

Registered in all 8 places: pcapkit/__init__.py, pcapkit/all.py,
pcapkit/protocols/__init__.py, pcapkit/protocols/data/__init__.py,
pcapkit/protocols/data/link/__init__.py, pcapkit/protocols/link/__init__.py,
pcapkit/protocols/schema/__init__.py, pcapkit/protocols/schema/link/__init__.py.

Tests

• tests/protocols/link/test_gre_unit.py (new): 17 tests / 42 subtests —
  base header dispatch (IPv4/IPv6), each optional field individually and
  together (longest 16-octet form), reserved-bit round trips, the nonzero-
  version refusal, make()/read() round trips, and make() validation.
• test_linux_sll1_unit.py/test_linux_sll2_unit.py: moved IPGRE/IP6GRE
  out of the "kept raw" sweep, added an explicit GRE-dispatch test to each.
• test_short_read_roundtrip_unit.py: a GRE case (longest form, all three
  optional fields) added to the cut-at-every-offset sweep.
• tests/project/test_rfc_constants_one_home_1739_unit.py: four entries
  added for literals that collide with homed constants by value but mean
  something unrelated (GRE's own header-length/field-width/bit-span
  literals), following the FRAD/Netlink precedent.

Counts: new test file 17 passed / 42 subtests; tests/protocols/link/
170 passed / 922 subtests; tests/protocols/test_short_read_roundtrip_unit.py
9 passed / 1390 subtests; tests/project/ full run 2132 passed, 80 skipped,
10619 subtests, 0 failed. pylint 10.00/10 and mypy clean on every
touched/new pcapkit/ file (verified: 110 pre-existing pcapkit/vendor/
mypy errors are identical on bare main, unrelated to this change).

Changelog & docs

Entry in docs/source/changelog/1.6.0.rst, CHANGELOG.md regenerated
(util/changelog_md.py --check clean), new docs page
docs/source/pcapkit/protocols/link/gre.rst indexed in the link-layer
toctree.

Scope

Two of #2460's five arphrd_type follow-ups remain: IEEE80211_RADIOTAP
(#2502) and FRAD (#2499, already its own PR). NETLINK has its own PR,
#2507. This PR does not close #2460.

[JarryShaw]

Cross-review (Sonnet) at aac4e2714: GOOD TO GO. Header layout confirmed against RFC 2890 §2 directly, including the conditional Checksum/Key/Sequence Number gating -- a real C+K+S probe round-tripped correctly, and a bogus version-1 header (the PPTP variant with a different layout) is correctly refused. EtherType dispatch confirmed to mirror VLAN/Ethernet's existing pattern via the shared Link.__proto__, no new registry. Both ARPHRD_IPGRE and ARPHRD_IP6GRE confirmed dispatching to the same class. All 8 registrations present. Test counts reproduced exactly: 179/2312 subtests on the link suite, 2132/10619 on the full tests/project/ run, 0 failures. PR body correctly says 'Part of #2460', not 'Closes'. No conflict markers. Ready for merge -- your call.

[github-actions]

Coverage: 94.48% (unit tier, Python 3.15, aac4e2714, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19738	133	2642	120	98.85%
pcapkit/corekit	2605	74	818	37	96.52%
pcapkit/dumpkit	281	3	100	3	98.43%
pcapkit/foundation	3183	27	1110	31	98.65%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	18574	199	4892	184	98.36%
pcapkit/toolkit	924	36	334	10	95.55%
pcapkit/utilities	502	7	144	5	98.14%
pcapkit/vendor	4510	1888	1034	169	54.89%

Per-file detail: the coverage-html artifact of this run.