|SUCCESS| net/mlx5: fix external egress group translation

[dpdklab at iol.unh.edu]

Test-Label: ai-code-review
Test-Status: SUCCESS
https://dpdk.org/patch/172738

_AI Code Review_

=== Patch Review: series.patch (via Claude) ===
Review date: 2026-11-09

# DPDK Patch Review

## Analysis

This patch modifies group translation logic in the mlx5 PMD's hardware flow (HWS) implementation. The change removes a condition that gated external egress group shifting on extended metadata mode, making the shift apply whenever E-Switch is enabled.

I have traced the code path and verified the logic against the evidence rule. The patch touches only the condition in `__translate_group()` that controls when external egress groups are shifted by +1.

---

## ERRORS

None identified.

---

## WARNINGS

None identified.

---

## INFO

**Documentation consistency**

The comment block at lines 5596-5603 is updated to reflect the new logic, which is good. However, the comment still says "Representor matching is always enabled" without explaining *why* removing the `dv_xmeta_en` check is correct.

The original condition required both:
1. `config->dv_esw_en` (E-Switch enabled)
2. `config->dv_xmeta_en == MLX5_XMETA_MODE_META32_HWS` (extended metadata mode)

The patch removes condition 2, applying the group shift whenever E-Switch is enabled, regardless of metadata mode. The commit message explains that "Representor matching is always enabled on an E-Switch port" and that the previous behavior (conditioned on repr_matching_en) shifted all external egress groups. This patch restores that behavior by making the shift unconditional when E-Switch is on.

The updated comment could more explicitly state that the shift is now unconditional when E-Switch is enabled, not just in extended metadata mode. This is an observation about clarity, not correctness -- the code change itself is clear.

---

Review-Result: CLEAN