|SUCCESS| [v3] net/vhost: fix SW checksum when zero

[dpdklab at iol.unh.edu]

Test-Label: ai-code-review
Test-Status: SUCCESS
https://dpdk.org/patch/171608

_AI Code Review_

=== Patch Review: series.patch (via Claude) ===
Review date: 2026-10-18

# Review of DPDK Patch: net/vhost: fix SW checksum when zero

## ERRORS

None identified.

## WARNINGS

None identified.

## INFO

**1. Consider bounds check ordering for clarity**

The bounds check at lines 353 and 396:
```c
if (rte_pktmbuf_data_len(mbuf) >= csum_offset + sizeof(uint16_t))
```

While this fix is correct (changing from `+ 1` to `+ sizeof(uint16_t)`), consider whether the checksum write should happen unconditionally or whether the bounds check failure should return an error. Currently, if the mbuf is too short, the checksum is silently not written, which may hide bugs.

However, this is pre-existing behavior that the patch does not change, so it does not block this fix.

**2. Verify csum_offset is properly validated upstream**

The patch assumes `csum_offset` is within reasonable bounds (no integer overflow when adding `sizeof(uint16_t)`). The calling code should validate this, but that validation is outside the scope of this patch.

**3. Alternative approach for UDP zero checksum check**

The patch uses:
```c
if (unlikely(csum == 0) &&
    (mbuf->ol_flags & RTE_MBUF_F_TX_L4_MASK) == RTE_MBUF_F_TX_UDP_CKSUM)
```

This is correct and follows the note in the commit message about early branching. The use of `unlikely()` on the compound condition would be semantically equivalent but the current form is acceptable as it optimizes for the common case where `csum != 0`.

Review-Result: CLEAN