|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.

## WARNINGS

**Warning 1: Missing release notes update**

This patch fixes a regression in the csum forwarding engine where `--vxlan-gpe-port` and `--geneve-parsed-port` options no longer take effect. This is a functional bug fix that restores important user-facing functionality. A release notes entry should document this fix.

**Suggested fix:** Add an entry to `doc/guides/rel_notes/release_26_11.rst` (or the appropriate current release file) under "Fixed Issues":

```rst
* **Fixed VXLAN-GPE and GENEVE custom port parsing in testpmd csum mode.**

  Restored functionality of ``--vxlan-gpe-port`` and ``--geneve-parsed-port``
  options in the csum forwarding engine. These options were non-functional
  after the switch to ``rte_net_get_ptype()``.
```

**Warning 2: Missing test coverage**

This patch fixes a regression and adds non-trivial logic (port swapping, inner header validation). There is no indication of functional test coverage for the fixed behavior. The patch should either:
- Add test cases to verify the custom port parsing works correctly
- Document manual testing steps in the commit message
- Reference existing tests that validate this functionality

## INFO

**Info 1: Consider helper function naming clarity**

The function `parse_tunnel_udp_custom_port()` modifies the mbuf (temporarily changes UDP port), reparses, then restores. The name suggests a pure parsing function but the implementation is actually a "parse-with-port-substitution" pattern. This is not wrong, but a comment emphasizing the temporary modification and restoration might help future readers.

Current implementation is correct with the restoration in place, but the side effect is non-obvious from the function name alone.

---

# Code Quality Assessment

**Correctness:** The patch correctly addresses the regression. Key correctness aspects verified:

1. **UDP port modification is properly scoped:** The port is saved, modified, used for parsing, and restored within the same function. The mbuf is not freed between modification and restoration.

2. **Bounds checking is appropriate:** 
   - `rte_pktmbuf_data_len(m) < off + sizeof(*uh)` check before accessing UDP header
   - `can_process_inner_headers()` validates that all declared headers fit in the packet before accessing inner headers
   - Uses `rte_pktmbuf_pkt_len()` for total packet length check (correct for potentially segmented packets)

3. **No resource leaks:** No allocations occur; only stack variables used.

4. **Logic correctness:**
   - Only processes custom ports when `txp->parse_tunnel` is enabled
   - Only applies when outer layer is UDP and not already classified as tunnel
   - Properly checks that new ptype indicates tunnel before accepting it
   - Correctly gates inner header processing on presence of inner L3

5. **Preserve existing behavior:** Plain VXLAN remains unchanged (stays on default port), which matches documented behavior.

**Style compliance:** Code follows DPDK style guidelines for indentation, bracing, and naming.

**API usage:** Uses standard DPDK APIs appropriately (`rte_net_get_ptype()`, mbuf accessors, byte order conversions).

Review-Result: WARNING