Fix ML-DSA key cache size check for verify-only builds

[miyazakh]

Summary

When WOLFSSL_DILITHIUM_VERIFY_ONLY is defined, the server only needs to
cache the public key DER, not the full keypair. The previous static assert
unconditionally required the cache buffer to fit a full keypair DER, causing
unnecessary failures in verify-only configurations.

This change introduces WH_SERVER_MLDSA_MAX_CACHE_DER_SIZE to select the
correct maximum size at compile time:

• WOLFSSL_DILITHIUM_VERIFY_ONLY defined: MLDSA_MAX_PUB_KEY_DER_SIZE
• Otherwise: MLDSA_MAX_BOTH_KEY_DER_SIZE

The static assert is updated to use this macro so it reflects the actual
requirement for each build configuration.

Testing

• Build with and without WOLFSSL_DILITHIUM_VERIFY_ONLY and confirm the
  static assert passes in both cases.

[Copilot]

Pull request overview

Fixes an ML-DSA server key-cache sizing check to accommodate WOLFSSL_DILITHIUM_VERIFY_ONLY builds where only the public key DER needs to be cached, avoiding unnecessary compile-time failures in verify-only configurations.

Changes:

• Adds a configuration-dependent ML-DSA maximum cache DER size selection in wh_server_crypto.c.
• Updates the key-cache static assert logic for verify-only vs full-keypair builds.

Comments suppressed due to low confidence (1)

src/wh_server_crypto.c:787

• In WOLFSSL_DILITHIUM_VERIFY_ONLY builds the static assert now allows WOLFHSM_CFG_SERVER_KEYCACHE_BIG_BUFSIZE to be sized for only the public-key DER, but wh_Server_MlDsaKeyCacheImport() still hard-codes MLDSA_MAX_BOTH_KEY_DER_SIZE when acquiring a cache slot and serializing. That can make verify-only builds pass the compile-time check but fail key cache import at runtime due to the larger requested size. Use WH_SERVER_MLDSA_MAX_CACHE_DER_SIZE for the cache-slot check and serialization sizing.

#endif

int wh_Server_MlDsaKeyCacheImport(whServerContext* ctx, wc_MlDsaKey* key,
                                  whKeyId keyId, whNvmFlags flags,
                                  uint16_t label_len, uint8_t* label)

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[padelsbach]

Looks good, but can you please add simple test cases to build-and-test.yml and build-and-test-refactor.yml

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #452

Scan targets checked: wolfhsm-crypto-bugs, wolfhsm-src
Findings: 1

Medium (1)

wh_Server_MlDsaKeyCacheImport hardcodes MLDSA_MAX_BOTH_KEY_DER_SIZE after static assert is relaxed

File: src/wh_server_crypto.c:800
Function: wh_Server_MlDsaKeyCacheImport
Category: Logic errors

wh_Server_MlDsaKeyCacheImport() still requests MLDSA_MAX_BOTH_KEY_DER_SIZE from wh_Server_KeystoreGetCacheSlotChecked(), but the PR now allows WOLFHSM_CFG_SERVER_KEYCACHE_BIG_BUFSIZE >= MLDSA_MAX_PUB_KEY_DER_SIZE in verify-only mode, so key import fails at runtime in any minimally-configured verify-only build.

Recommendation: Replace both MLDSA_MAX_BOTH_KEY_DER_SIZE occurrences in wh_Server_MlDsaKeyCacheImport() with WH_SERVER_MLDSA_MAX_CACHE_DER_SIZE.

Referenced code: src/wh_server_crypto.c:800-804 (5 lines)

---

This review was generated automatically by Fenrir. Findings are non-blocking.

[padelsbach]

Agree with the Fenrir finding. Please correct uses of WH_SERVER_MLDSA_MAX_CACHE_DER_SIZE.

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #452

Scan targets checked: wolfhsm-crypto-bugs, wolfhsm-src

Findings: 2
2 finding(s) posted as inline comments (see file-level comments below)

This review was generated automatically by Fenrir. Findings are non-blocking.

[padelsbach]

@miyazakh there was a github outage earlier which caused test failures. Can you push again to re-trigger CI?

[miyazakh]

@padelsbach
It seems that all the tests passed after manual re-testing.

[padelsbach]

LGTM, will give @bigbrett a chance for a final review.

[bigbrett]

overall looks good, I think one of the test cases should be removed as it is overkill.

[bigbrett]

I appreciate the spirit of adding test cases for both of these but IMO its pretty overkill to add two separate builds and expose two internal parameters just to essentially verify at compile time.I'd prefer to skip this second one that verifies the static assert hits just for the sake of conserving CI time.IMO something like this is better suited for a nightly (which we don't have yet.