|WARNING| common/mlx5: fix mkey ordering enum

[dpdklab at iol.unh.edu]

Test-Label: ai-code-review
Test-Status: WARNING
https://dpdk.org/patch/171825

_AI Code Review_

=== Patch Review: series.patch (via Claude) ===
Review date: 2026-10-22

# DPDK Patch Review

## Patch Summary
This patch updates the `mlx5_mkc_raw_ordering` enum in the mlx5 common driver to align with updated hardware specification. It removes the `MLX5_MKC_RAW_ORDERING_SAO` value (0x1) and changes `MLX5_MKC_RAW_ORDERING_RO` from 0x2 to 0x1.

## Errors

**1. Potential breaking change without usage verification**

The patch removes `MLX5_MKC_RAW_ORDERING_SAO` entirely and changes the value of `MLX5_MKC_RAW_ORDERING_RO` from 0x2 to 0x1. If any code in the tree currently uses `MLX5_MKC_RAW_ORDERING_SAO`, this removal will cause a compilation failure. If any code compares against the literal value 0x2 expecting relaxed ordering, the changed enum value will cause incorrect behavior at runtime.

```c
/* Removed enum value - may break existing users */
- MLX5_MKC_RAW_ORDERING_SAO = 0x1,
```

The patch should verify that:
- No code references `MLX5_MKC_RAW_ORDERING_SAO`
- No code depends on `MLX5_MKC_RAW_ORDERING_RO` having value 0x2
- The hardware specification change is backward compatible

If existing deployed hardware interprets 0x2 as relaxed ordering and new code sends 0x1, the mismatch could cause memory ordering violations leading to data corruption or crashes.

**Suggested fix:** Before accepting this patch, grep the codebase for all uses of these enum values and verify the change is safe:
```bash
git grep -n "MLX5_MKC_RAW_ORDERING_SAO\|MLX5_MKC_RAW_ORDERING_RO" drivers/
```

If any uses exist, they must be updated in the same patch or a preceding patch in the series.

## Warnings

None.

## Info

**1. Release notes required**

This is a change to driver behavior that affects hardware configuration. If this changes the memory ordering semantics sent to the device, it should be documented in the release notes under the mlx5 driver section.

**2. Consider ABI implications**

If `mlx5_mkc_raw_ordering` is used in any structure exposed in an installed header (checked by reviewing `drivers/common/mlx5/meson.build` for `headers` or `indirect_headers`), changing enum values is an ABI break. The enum values are part of the ABI contract. However, PRM headers are typically internal to the driver, so this is likely not an issue - but worth verifying.

**3. Hardware specification reference**

The commit message references "updated device specification" but doesn't cite which version or document. For hardware-related changes, including a specification version or reference helps reviewers verify correctness.

Review-Result: ERROR