|WARNING| app/testpmd: fix VXLAN-GPE and GENEVE custom port parsing

[dpdklab at iol.unh.edu]

Test-Label: ai-code-review
Test-Status: WARNING
https://dpdk.org/patch/172651

_AI Code Review_

=== Patch Review: series.patch (via Claude) ===
Review date: 2026-11-06

# DPDK Patch Review

## ERRORS

None identified.

## WARNINGS

**Missing release notes update:**
This patch fixes a functional bug where `--vxlan-gpe-port` and `--geneve-parsed-port` options were broken since commit 76730c7b9b5a. The fix restores tunnel detection and processing on custom ports, which is a user-visible behavior change. This should be documented in the current release notes under "Fixed Issues" or "app/testpmd".

**Boundary check could miss inner L4 truncation:**
Line 633: `headers_len = hdr_lens->l2_len + hdr_lens->l3_len + hdr_lens->inner_l2_len + hdr_lens->inner_l3_len + hdr_lens->inner_l4_len;`

The check includes `inner_l4_len` but is intended to validate access to the inner IP header (line 791: `l3_hdr` calculation, line 806: `parse_l4_proto()` call). If the inner L4 header is truncated but the inner L3 header is complete, this function returns true and the code proceeds to parse the inner L4 protocol, potentially reading beyond packet end. Consider checking only through `inner_l3_len`, or verify that `parse_l4_proto()` has its own bounds check before accessing L4 data.

**Variable name `process_inner` could be clearer:**
Line 735: `bool process_inner = false;`

The variable guards both inner header bounds validation AND whether ethertype/l4_proto are set. The name suggests it controls whether to process inner headers, but it's actually "inner headers are present and complete". Consider `inner_headers_valid` or `has_valid_inner` to clarify that it's a validation result, not a configuration flag.

## INFO

**Code structure suggestion:**
Lines 647-648: The `saved_port` / port-swapping pattern is clever but adds complexity. An alternative would be to directly check the UDP payload's first byte for VXLAN-GPE or GENEVE magic numbers when `dst_port` matches a configured custom port. This would avoid modifying the packet. However, the current approach correctly reuses the library parser, so this is only a suggestion for future refactoring.

**Pre-existing: Outer checksum not validated before inner processing:**
Lines 767-806: The code processes tunnel packets without verifying the outer UDP checksum. If the outer headers are corrupt (wrong outer UDP dst_port due to bit flip), the parser might misidentify the tunnel type or miss it entirely. This is pre-existing behavior not introduced by this patch -- the csum engine has never validated outer checksums before processing inners.

Review-Result: WARNING