|WARNING| crypto/cnxk: fix AE table get on unprobed device

[dpdklab at iol.unh.edu]

Test-Label: ai-code-review
Test-Status: WARNING
https://dpdk.org/patch/170969

_AI Code Review_

=== Patch Review: series.patch (via Claude) ===
Review date: 2026-10-09

# DPDK Patch Review

## Summary

This patch fixes NULL pointer dereference bugs in two cnxk cryptodev PMD functions by replacing incorrect validation checks. The changes are correct and address real correctness issues.

---

## Correctness Issues

### Fixed Issues (no further action needed)

The patch correctly addresses:

1. **Invalid device validation** - `rte_cryptodev_pmd_get_dev()` returns a pointer into a static array even for unprobed devices, so checking its return against NULL is ineffective. The patch correctly switches to `rte_cryptodev_is_valid_dev()`.

2. **NULL pointer parameter check added** - `rte_pmd_cnxk_ae_ec_grp_table_get()` now validates `nb_max_entries` is not NULL before dereferencing it.

### Remaining Issue

**Error** - Missing error check on second `dev->data` dereference:

After validating the device ID and calling `rte_cryptodev_pmd_get_dev()`, both functions access `dev->data->dev_private` without checking if `dev->data` is NULL.

While `rte_cryptodev_is_valid_dev()` ensures the device is probed, there is a theoretical window where `dev->data` could be NULL during device teardown or in error states. The defensive check should be added:

```c
dev = rte_cryptodev_pmd_get_dev(dev_id);
if (dev->data == NULL) {
    plt_err("Device data is not initialized");
    return NULL;
}

vf = dev->data->dev_private;
```

This pattern appears in both `rte_pmd_cnxk_ae_fpm_table_get()` and `rte_pmd_cnxk_ae_ec_grp_table_get()`.

**Note**: If the DPDK guarantees that `dev->data` is never NULL when `rte_cryptodev_is_valid_dev()` returns true, this can be ignored. However, the existing check on `vf == NULL` suggests the code is being defensive about internal state, so checking `dev->data` would be consistent.

---

## Style and Code Quality

**Warning** - Implicit comparison on pointer:

Line 1163 (and similar pattern at line 1141):
```c
if (vf == NULL) {
```

This is correct per DPDK style (explicit NULL comparison). No issue here.

---

## Process and Documentation

**Info** - Missing information for reviewers:

The commit message could benefit from a brief note about why the original check was ineffective (that `rte_cryptodev_pmd_get_dev()` returns a static array pointer). This would help reviewers understand the severity of the bug.

---

## No Issues Found

- Code style follows DPDK conventions
- Explicit NULL comparisons are used correctly
- Error messages are clear and actionable
- Fixes tags are present and correctly formatted
- No use-after-free or resource leaks
- No forbidden tokens or deprecated API usage

---

## Recommendation

**Accept with minor enhancement**: Add the `dev->data` NULL check for complete defensive programming, or confirm with maintainers that it's guaranteed non-NULL after `rte_cryptodev_is_valid_dev()` validation.