|WARNING| net/mana: fix Tx stall from send queue free-space unit mismatch

[dpdklab at iol.unh.edu]

Test-Label: ai-code-review
Test-Status: WARNING
https://dpdk.org/patch/169521

_AI Code Review_

=== Patch Review: series.patch (via Claude) ===
Review date: 2026-09-14

# DPDK Patch Review: net/mana: fix Tx stall from send queue free-space unit mismatch

## Summary
This patch fixes a unit mismatch bug in the MANA driver's send queue capacity check that caused premature Tx stalls. The fix is correct and the patch is clean.

---

## Errors
None.

---

## Warnings

### 1. Multi-line comment style (Warning)
The comment uses `/* ... */` block-comment style but is formatted as multiple lines of comment text. DPDK style prefers the opening `/*` alone on the first line for multi-line comments.

**Current:**
```c
/* head/tail count WQE alignment units, so the capacity they are compared
 * against must too: queue->count is in entries and undercounts the
 * queue, stalling Tx well below capacity.
 */
```

**Preferred DPDK style:**
```c
/*
 * head/tail count WQE alignment units, so the capacity they are compared
 * against must too: queue->count is in entries and undercounts the
 * queue, stalling Tx well below capacity.
 */
```

---

## Info

### Patch Quality
- **Correctness**: The fix correctly resolves a unit mismatch where `queue->count` (in entries) was compared against `head - tail` (in WQE alignment units), causing the capacity check to underestimate available space by 8x on typical queue configurations.
- **Clarity**: The comment explains the bug and why the fix uses `queue->size / GDMA_WQE_ALIGNMENT_UNIT_SIZE`.
- **Scope**: Appropriately limited to Tx path only (Rx unaffected due to 1:1 entry-to-unit mapping).
- **Metadata**: Proper `Fixes:` tag, `Cc: stable at dpdk.org`, and independence note for backporting.

### Code Review
- Variable naming is clear (`queue_free_units` accurately describes what it now measures).
- The calculation matches the ring-wrap logic in `gdma_get_wqe_pointer()`, ensuring consistency.
- No resource leaks, race conditions, or forbidden tokens introduced.
- No ABI changes or new API.

---

## Conclusion
The patch is functionally correct and addresses a real correctness bug (premature Tx stall due to incorrect capacity calculation). The only improvement needed is adjusting the multi-line comment style to match DPDK conventions.