[REVIEW] common/cnxk: fix duplicate branch compiler warning

[Stephen Hemminger <stephen@networkplumber.org>]

AI-generated review of bundle-1704-common-cnxk.mbox
Reviewed using Claude (claude-opus-4-5-20251101)

This is an automated review. Please verify all suggestions.

---

# DPDK Patch Review: bundle-1704-common-cnxk.mbox

## Overview
This is a patch series of 18 patches for DPDK common/cnxk and net/cnxk drivers, primarily dealing with inline IPsec improvements for CN20K platform.

---

## Patch 1/18: common/cnxk: fix engine capabilities fetch logic

### Commit Message
**Warning**: The commit message body lacks detail. The description "This patch fixes the same" is vague and should explain what the actual fix is.

**Info**: The `Fixes:` tag is correctly formatted with 12-character SHA.

### Code Review
**Error**: In `cnxk_security.c`, lines 1400 and 1409 - The change from `ROC_IE_OT_SA_LIFE_UNIT_*` to `ROC_IE_OW_SA_LIFE_UNIT_*` macros appears correct for the OW (cn20k) context.

**Error**: In `roc_nix_inl.c`, line 1231 - The loop condition change from `while (lmt_status != 0)` to `while (lmt_status == 0)` is a significant logic change. The commit message should explain why the previous condition was wrong (it was causing an infinite loop as stated).

---

## Patch 2/18: common/cnxk: remove dependency on cryptodev for RXC

### Commit Message
**Info**: Good descriptive subject line.

### Code Review
**Info**: The code correctly moves the cryptodev check inside the `roc_model_is_cn10k()` block, allowing cn20k to proceed without this dependency.

---

## Patch 3/18: common/cnxk: support inbound pdb configuration

### Commit Message
**Info**: Subject and body are acceptable.

### Code Review
**Info**: Simple addition of `pdb_ena` field usage in mbox configuration.

---

## Patch 4/18: common/cnxk: update CPT RXC structures

### Commit Message
**Warning**: The subject says "update" but should be more specific about what was updated (byte order/endianness handling).

### Code Review
**Info**: The structures are being reorganized with proper unions for different platforms (cn10k vs generic). The byte order appears to be reversed in the new `cpt_frag_info_s` compared to `cpt_cn10k_frag_info_s`.

---

## Patch 5/18: common/cnxk: update inline profile ID for cn20k

### Commit Message
**Info**: Acceptable.

### Code Review
**Info**: Correctly adds cn20k-specific profile ID handling.

---

## Patch 6/18: common/cnxk: update inline RQ mask

### Commit Message
**Info**: Acceptable.

### Code Review
**Warning**: Significant code removal - the `nix_inl_rq_mask_cfg` function removes the SPB setup via `nix_rx_inl_lf_cfg` mbox call. This should be validated that it's not needed for cn20k.

---

## Patch 7/18: net/cnxk: avoid security flag for custom inbound SA

### Commit Message
**Info**: Good descriptive subject.

### Code Review
**Info**: Simple conditional change that looks correct.

---

## Patch 8/18: net/cnxk: add CPT code check for soft expiry

### Commit Message
**Info**: Acceptable.

### Code Review
**Info**: Adding `ROC_IE_OW_UCC_SUCCESS_SA_SOFTEXP_AGAIN` case is a valid addition.

---

## Patch 9/18: net/cnxk: skip write SA for cn20k

### Commit Message
**Info**: Acceptable but could explain why write SA should be skipped.

### Code Review
**Info**: Simple flag change.

---

## Patch 10/18: net/cnxk: update NIX reassembly fast path

### Commit Message
**Info**: Acceptable.

### Code Review
**Warning**: The function `nix_cqe_xtract_mseg` now has an `#if defined(RTE_ARCH_ARM64)` guard with an empty stub for non-ARM64. This may break functionality on other architectures silently.

**Info**: The use of NEON intrinsics (`vreinterpret_u16_u64`, `vrev64_u16`, etc.) is appropriate for ARM64 optimization.

---

## Patch 11/18: net/cnxk: update aura batch free

### Commit Message
**Info**: Subject could be more descriptive about what was updated (mask change from 0x1 to 0x3).

### Code Review
**Info**: The mask change from `0x1` to `0x3` appears consistent across all occurrences.

---

## Patch 12/18: net/cnxk: update fastpath function for OOP

### Commit Message
**Info**: OOP should be expanded (Out-Of-Place) in the commit body for clarity.

### Code Review
**Info**: Adds OOP (out-of-place) processing support. The `nix_sec_oop_process` function is well-implemented.

**Info**: The function correctly handles the alternate encrypted-decrypted pointer pattern in the gather list.

---

## Patch 13/18: event/cnxk: update fastpath function for OOP

### Commit Message
**Info**: Acceptable.

### Code Review
**Info**: Consistent OOP handling with the net/cnxk changes.

---

## Patch 14/18: common/cnxk: flow rule config for non-inplace

### Commit Message
**Info**: Acceptable.

### Code Review
**Info**: Adds `is_non_inp` flag handling in flow rules.

---

## Patch 15/18: net/cnxk: enable PDB in IPsec outbound path

### Commit Message
**Info**: PDB should be expanded in the commit body.

### Code Review
**Info**: Simple bit flag additions for PDB enablement.

---

## Patch 16/18: common/cnxk: initialize CPT LF for CQ config

### Commit Message
**Info**: Acceptable.

### Code Review
**Info**: Properly initializes CPT LF for CQ configurations with appropriate cleanup in the error path and fini function.

---

## Patch 17/18: common/cnxk: fix CPT CQ roll over handling

### Commit Message
**Info**: Good - includes proper `Fixes:` tag.

### Code Review
**Info**: The fix correctly handles ring rollover by tracking `cq_head` separately and using modulo arithmetic.

**Warning**: Line 102-103 adds error logging with `plt_err` but this could be noisy in production. Consider using `plt_warn` or adding rate limiting.

---

## Patch 18/18: common/cnxk: fix duplicate branch compiler warning

### Commit Message
**Info**: Good - includes Bugzilla ID and proper `Fixes:` tags.

### Code Review
**Info**: The `PLT_STATIC_ASSERT` additions verify that SA sizes are equal across platforms, which justifies using a common macro.

**Warning**: In `roc_nix_priv.h`, lines 26 and 29 - changing integer division to floating point division (`100` to `100.0`) changes the type of the result. Verify this doesn't cause issues with consumers of these macros.

```c
#define NIX_CQ_LBP_THRESH_FRAC_REF1	(80 * 16 / 100.0)
#define NIX_CQ_LBP_THRESH_FRAC	(75 * 16 / 100.0)
```

---

## General Issues Across All Patches

### License Compliance
**Info**: All modified files appear to have proper SPDX license identifiers already present.

### Missing Items
**Warning**: No release notes updates found in this patch series despite API/behavior changes. Consider adding entries to `doc/guides/rel_notes/`.

**Warning**: Several patches lack `Cc: stable@dpdk.org` tag despite being bug fixes (patches 1, 17, 18).

### Coding Style
**Info**: Code generally follows DPDK coding standards with proper indentation and naming conventions.

---

## Summary

| Severity | Count |
|----------|-------|
| Error | 0 |
| Warning | 8 |
| Info | Multiple |

The patch series is generally well-structured. Main concerns are:
1. Missing `Cc: stable@dpdk.org` on fix patches
2. Missing release notes
3. Some commit messages could be more descriptive
4. Floating point division change in macros should be verified