wp_drbg: gate wc_RNG_DRBG_Reseed on FIPS version

[ColtonWilley]

wp_drbg_reseed references wc_RNG_DRBG_Reseed, but that is not always an exported symbol. It is WOLFSSL_API in non-FIPS >= 5.7.2 and in FIPS v6+, while FIPS v5.x commercial bundles are inconsistent -- the same v5.2.4 cert is WOLFSSL_LOCAL on a 5.8.4 wrapper but WOLFSSL_API on a 5.9.1 wrapper. Linking it fails (undefined reference to wc_RNG_DRBG_Reseed) on the bundles that keep it LOCAL, which the previous version-only gate did not account for.

• settings.h: gate WP_HAVE_DRBG_RESEED conservatively. Use the native in-place reseed only where the symbol is reliably exported (non-FIPS >= 5.7.2, FIPS v6+) and fall back to DRBG re-instantiation for all FIPS v5.x (the fallback links on every build). WP_NO_DRBG_RESEED forces the fallback for any other bundle that keeps the symbol WOLFSSL_LOCAL despite its version.
• wp_drbg.c: native in-place reseed when available, else re-instantiate the DRBG in place via wc_FreeRng()+wc_InitRngNonce() (with error-state reporting). Handle reseed with no caller entropy on both paths, and keep reseed seccomp-safe (native draws from the cached /dev/urandom fd via wp_urandom_read; fallback self-seeds via wc_InitRngNonce).
• utils-wolfssl.sh: map --fips-check=v5.2.4 to --enable-fips=v5.2.4 (was collapsing to v5, which mismatched the SP-math FIPS module).

Validated: build + full unit suite pass (136/0) against the 5.8.4 and 5.9.1 wolfSSL wrappers x v5.2.1 and v5.2.4 commercial FIPS bundles, plus non-FIPS.

Note on the FIPS v5.x fallback path: caller-supplied entropy/addIn is folded into wc_InitRngNonce as the nonce, not mixed as DRBG entropy_input, and predResist is not honored -- wolfCrypt re-seeds internally via its seed source. This is the available behavior where the module does not export wc_RNG_DRBG_Reseed, and differs from the native path where caller entropy is reseed input.

[aidangarske]

Skoll Multi-Scan Review

Modes: review + review-security + bugsOverall recommendation: COMMENT
Findings: 5 total — 4 posted, 1 skipped
4 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [Medium] [review+review-security] FIPS fallback reseed re-instantiates from OS entropy, bypassing the parent-seed/seccomp design and treating caller entropy as nonce — src/wp_drbg.c:440-485
• [Medium] [review] New error-state and explicit-entropy reseed paths have no test coverage — src/wp_drbg.c:328-333,440-485,612-644
• [Low] [review-security+bugs] wp_drbg_get_seed does not check the new rngError state, operating on a de-instantiated RNG after a failed fallback reseed — src/wp_drbg.c:734-756
• [Low] [review] wc_FreeRng() return code ignored in fallback reseed — src/wp_drbg.c:468

Skipped findings

• [Low] utils-wolfssl.sh adds untested v5.2.3 - --enable-fips=v5.2.3 mapping that changes prior behavior

Review generated by Skoll

[aidangarske]

Skoll Multi-Scan Review

Modes: review + review-securityOverall recommendation: COMMENT
Findings: 6 total — 4 posted, 2 skipped
4 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [Medium] [review] Untested v5.2.3 FIPS tag mapping added without validation — scripts/utils-wolfssl.sh:186-188
• [Medium] [review] Bare scope block introduced in wp_drbg_reseed native path — src/wp_drbg.c:400-455
• [Low] [review] Non-ASCII em-dash in shell comment — scripts/utils-wolfssl.sh:183
• [Low] [review-security] Bare scope block introduced in wp_drbg_reseed native reseed path — src/wp_drbg.c:400-455

Skipped findings

• [Info] FIPS v5.x fallback reseed deviates from SP 800-90A reseed semantics
• [Low] Missing unit coverage for fallback reseed error-state path

Review generated by Skoll

[aidangarske]

🟠 [Medium] Untested v5.2.3 FIPS tag mapping added without validation

The PR adds a new case mapping v5.2.3|linuxv5.2.3 to --enable-fips=v5.2.3. The PR description and validation only cover v5.2.1 and v5.2.4 bundles; v5.2.3 is neither mentioned nor tested, and unlike the v5.2.4 case it carries no comment explaining why it is a distinct module. If wolfSSL's configure does not accept --enable-fips=v5.2.3 as a recognized value, builds invoked with that tag will now fail at configure time, whereas before this change they fell through to the working v5 mapping. This silently broadens behavior for an unvalidated tag.

Fix: Confirm --enable-fips=v5.2.3 is a valid configure value for the targeted wolfSSL versions. If it is not validated, drop the v5.2.3 case so it falls through to v5 (the prior, working behavior), or add a comment documenting which cert/module it targets like the v5.2.4 case does.

[aidangarske]

🟠 [Medium] Bare scope block introduced in wp_drbg_reseed native path

The #ifdef WP_HAVE_DRBG_RESEED branch opens a standalone { ... } block solely to scope the local declarations unsigned char* seed and size_t seedLen. This is a bare scope block that does not belong to a function, control statement, switch/case, or type definition. The wolfSSL/wolfProvider C style explicitly bans introducing bare scope blocks in new code to make C89 declarations work — declarations should move to the top of the enclosing function (or the logic split into a helper). Note the sibling #else fallback branch correctly declares its locals inside an if (ok) { control block, so the two halves of the same function are inconsistent.

Fix: Remove the bare scope block to comply with the no-bare-scope-block convention; the empty-brace-scan skill flags exactly this pattern.

[aidangarske]

🔵 [Low] Non-ASCII em-dash in shell comment

The added comment uses a Unicode em-dash (—) in an otherwise ASCII source tree. For consistency with the rest of the C/shell sources and to avoid encoding surprises in tooling/diffs, prefer a plain ASCII hyphen.

Fix: Replace the em-dash with an ASCII hyphen.

[aidangarske]

🔵 [Low] Bare scope block introduced in wp_drbg_reseed native reseed path · Logic

The new native (WP_HAVE_DRBG_RESEED) branch opens a standalone { ... } scope solely to declare unsigned char* seed and size_t seedLen mid-function. This is a bare scope block, which the wolfSSL/wolfProvider C coding standard prohibits in new code (scope blocks are only allowed for function/control/switch/type bodies). The parallel fallback branch correctly uses a control-statement block (if (ok) { ... }), so only the native path violates the rule. BEFORE: these two locals were declared at the top of the function. AFTER: they were moved into a new bare { } block under the #ifdef. This is a style/maintainability issue, not a security defect.

Fix: Declare seed/seedLen at the top of wp_drbg_reseed (guarded as needed) and remove the bare { } scope, or move the native logic into a small helper function, to comply with the no-bare-scope coding standard. Run the empty-brace-scan to confirm no other new scope blocks were introduced.