|SUCCESS| common/cnxk: handle increased count eot width for cn20k

[dpdklab at iol.unh.edu]

Test-Label: ai-code-review
Test-Status: SUCCESS
https://dpdk.org/patch/172272

_AI Code Review_

=== Patch Review: series.patch (via Claude) ===
Review date: 2026-10-30

# DPDK Patch Review

## ERRORS

None identified.

## WARNINGS

None identified.

## INFO

### 1. Bit manipulation comment could be clearer

**File:** `drivers/common/cnxk/roc_npa.h`  
**Lines:** Comments at lines ~517 and ~555

The ASCII art diagrams show the COUNT_EOT field spanning bits 33-32, but the description suggests it's 2 bits wide. The diagram format could more clearly indicate which bits are part of COUNT_EOT:

```
 * 63   63 62  34 33    32 31  20 19    0
 *        vs
 * 63   63 62  34 33  32 31  20 19    0
```

Consider updating the diagram to show:
```
 * 63      62..34 33..32  31..20  19..0
 * -----------------------------------------
 * | FABS | Rsvd | COUNT_EOT | Rsvd | AURA |
 * -----------------------------------------
```

This is a minor documentation clarity improvement and does not affect correctness.

### 2. Magic number 0x3 could use a symbolic constant

**File:** `drivers/common/cnxk/roc_npa.h`  
**Lines:** ~524, ~566

The patch introduces `0x3` as the mask value for the 2-bit COUNT_EOT field. While the comments explain this adequately, a named constant would improve readability:

```c
#define NPA_BATCH_FREE_COUNT_EOT_MASK 0x3
```

This is a style preference rather than a correctness issue. The existing code pattern in this file uses inline hex literals, so the current approach is consistent with the surrounding code.

---

**Summary:**

The patch correctly handles the hardware difference between CN10K (1-bit COUNT_EOT) and CN20K (2-bit COUNT_EOT) by using a 2-bit mask that works on both platforms. The logic is sound: the upper bit (bit 33) is reserved and ignored on CN10K, so writing `0x3` instead of `0x1` is backward-compatible.

The code changes are minimal, well-commented, and preserve the existing behavior while adding CN20K support. No correctness bugs, resource leaks, or style violations were found.

Review-Result: CLEAN