Skip to content

Pq fips fenrir - #11359

Open
jackctj117 wants to merge 13 commits into
wolfSSL:masterfrom
jackctj117:PQ-FIPS-Fenrir
Open

jackctj117 wants to merge 13 commits into
wolfSSL:masterfrom
jackctj117:PQ-FIPS-Fenrir

Conversation

@jackctj117

Copy link
Copy Markdown
Contributor

This pull request introduces several FIPS 140-3 and SP 800-131A Rev. 2 compliance checks and test updates for cryptographic operations, mainly affecting ECC, HMAC, RNG/DRBG, and RSA code paths. The changes ensure that only CAVP-tested and standards-approved parameters, key sizes, and algorithms are permitted in approved mode, and that test suites properly verify these constraints.

FIPS 140-3 and SP 800-131A Compliance Enforcement

  • ECC Algorithm Restrictions:

    • Added checks to restrict ECC shared secret computation and key generation to only approved curves (P-256, P-384, P-521) and key sizes (≥224 bits) in FIPS v7+ mode, per SP 800-131A and FIPS 140-3 IG C.B. ECDSA signing is also gated to disallow curves with fewer than 224 bits. [1] [2] [3] [4] [5]
  • HMAC Key Length Enforcement:

    • Enforced a maximum HMAC key length of 1024 bits in FIPS v7+ mode, matching the module's CAVP-tested range, unless an explicit override is provided.
  • RNG/DRBG Entropy Source Enforcement:

    • Modified DRBG reseeding in FIPS v7+ mode to always obtain entropy from the module's internal seed source, per SP 800-90A, treating caller-supplied bytes as additional input only. [1] [2] [3]

Test Suite Updates for Compliance

  • HMAC Test Enhancements:

    • Updated HMAC test cases to verify rejection of overlong keys in FIPS v7+ mode and acceptance only when explicitly allowed, across all SHA and SHA3 variants. [1] [2] [3] [4] [5] [6]
  • RSA Test Adjustments:

    • Updated RSA test cases to expect and handle new FIPS_WRONG_API_E errors for disallowed PKCS#1 v1.5 key transport in FIPS v7+ mode, ensuring tests reflect current compliance rules. [1] [2] [3] [4] [5] [6]

These changes collectively ensure that cryptographic operations are restricted to approved parameters and algorithms in FIPS 140-3 approved mode and that the test suite robustly verifies these restrictions.

@jackctj117 jackctj117 self-assigned this Sep 2, 2026
Copilot AI lite review requested due to automatic review settings September 2, 2026 20:05

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Changes recommended

The new ECC signing gate returns a misleading error code (SIG_TYPE_E) and the DRBG reseed helper should better align with existing SMALL_STACK_CACHE behavior and optional additional_input semantics.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR tightens FIPS 140-3 / SP 800-131A Rev. 2 enforcement in wolfCrypt (notably for ECC KAS/KeyGen/Sign, HMAC key sizing, and DRBG reseed entropy sourcing) and updates the wolfCrypt test suite to assert the new approved-mode behaviors and expected error codes.

Changes:

  • Enforces FIPS v7+ constraints: ECC KAS limited to P-256/P-384/P-521; ECC keygen/signing disallows <224-bit curves; HMAC rejects keys >1024 bits unless explicitly overridden.
  • Updates DRBG reseed in FIPS v7+ to always pull entropy from the module seed source and treat caller bytes as additional input.
  • Extends test coverage to validate new rejection/acceptance behaviors (HMAC long keys, RSA PKCS#1 v1.5 transport refusal, PBKDF2/HKDF/ECDSA “not approved” signaling).
File summaries
File Description
wolfssl/wolfcrypt/hmac.h Adds HMAC_FIPS_MAX_KEY constant for v7+ (1024-bit max) used to enforce approved-mode key-size limits.
wolfcrypt/src/hmac.c Enforces v7+ maximum HMAC key length when allowFlag is not set.
wolfcrypt/src/random.c Alters v7+ DRBG reseed to draw entropy internally and treat caller bytes as additional input; adds helper to reseed from seed source.
wolfcrypt/src/ecc.c Adds v7+ ECC curve gating for shared secret, shared secret (ex), key generation, and signing restrictions.
wolfcrypt/test/test.c Updates/extends tests for HMAC long-key rejection + override, RSA PKCS#1 v1.5 transport refusal behavior, and other v7+ non-approved indicators.
Review details
  • Files reviewed: 5/5 changed files
  • Comments generated: 3
  • Review effort level: Lite

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

Comment thread wolfcrypt/src/ecc.c Resolved
Comment thread wolfcrypt/src/random.c Resolved
Comment thread wolfcrypt/src/random.c Resolved
@github-actions

github-actions Bot commented Sep 3, 2026 •

Copy link
Copy Markdown

MemBrowse Memory Report · Explore what changed →

gcc-arm-cortex-m4-openssl-compat

  • FLASH: .text +64 B (+0.0%, 802,180 B / 1,048,576 B, total: 77% used)

stm32-sim-stm32h753

  • FLASH: .text +440 B (+0.2%, 204,908 B / 2,097,152 B, total: 10% used)

zephyr-frdm_rw612

  • FLASH: datas +16 B, log_const_area -8 B, rodata +36 B, text +11,728 B (+4.2%, 294,236 B / 67,108,864 B, total: 0% used)
  • RAM: bss +49 B, datas +16 B (+0.1%, 58,423 B / 983,040 B, total: 6% used)

No memory changes detected for:


Catch memory regressions in CI with MemBrowse. Free for open source.

@jackctj117

Copy link
Copy Markdown
Contributor Author

Jenkins retest this please.

@jackctj117

jackctj117 commented Oct 21, 2026 •

Copy link
Copy Markdown
Contributor Author

@claude review

Comment thread wolfcrypt/src/kdf.c Resolved
Comment thread wolfcrypt/src/hmac.c Resolved
Comment thread wolfcrypt/src/ecc.c Resolved
@jackctj117

Copy link
Copy Markdown
Contributor Author

Jenkins retest this please.

@jackctj117

Copy link
Copy Markdown
Contributor Author

https://github.com/wolfSSL/fips/pull/420 Has to merge first for this PR to pass Jenkins testing.

@jackctj117 jackctj117 added the For FIPS v7 Module Related to FIPS v7.0.0 module prep for submission label Nov 2, 2026
kaleb-himes added a commit to kaleb-himes/wolfssl that referenced this pull request Nov 8, 2026
kaleb-himes added a commit to kaleb-himes/wolfssl that referenced this pull request Nov 8, 2026
@kaleb-himes kaleb-himes mentioned this pull request Nov 9, 2026
1 of 4 tasks
kaleb-himes added a commit to kaleb-himes/wolfssl that referenced this pull request Nov 9, 2026
kaleb-himes added a commit to kaleb-himes/wolfssl that referenced this pull request Nov 9, 2026
kaleb-himes added a commit to kaleb-himes/wolfssl that referenced this pull request Nov 9, 2026
kaleb-himes added a commit to kaleb-himes/wolfssl that referenced this pull request Nov 9, 2026
kaleb-himes added a commit to kaleb-himes/wolfssl that referenced this pull request Nov 10, 2026
kaleb-himes added a commit to kaleb-himes/wolfssl that referenced this pull request Nov 10, 2026
@jackctj117
jackctj117 force-pushed the PQ-FIPS-Fenrir branch 2 times, most recently from 6fb9078 to 10b95fa Compare November 11, 2026 19:38
…since FIPS 198-1 Section 4 folds them to K0 = H(K) but they fall outside the module's CAVP-tested key range (Fenrir F-9546).
…elow 224 bits under FIPS v7 per SP 800-131A Rev. 2 Table 2, leaving signature verification with those curves available as legacy use (Fenrir F-7359).
…tyle test vectors and keep them covered through the wc_HmacSetKey_ex allowFlag escape (Fenrir F-9546).
The PKCS#1 v1.5 refusal, the HKDF SHA-1 refusal and the WC_FIPS_NOT_APPROVED
indicator come from the fips.c wrappers, which --enable-fips=dev-no-post
builds without (and without fips.h, so WC_FIPS_NOT_APPROVED is undeclared).
Matches what the fips.c keygen wrappers return, which test_wc_ecc_sig_size_calc
expects; dev-no-post builds reach the library check directly.
wc_PRF keys HMAC with the pre-master secret, which is 256 bytes for a 2048-bit
DHE group, so the 1024-bit cap broke TLS 1.2 DHE handshakes. Like HKDF and
PBKDF2, the PRF's HMAC key is covered by its KDF validation. Also expect the
OpenSSL-compat HMAC_Init long-key (160 byte) case to be refused under v7.
FIPS_ALLOW_SHORT lifts the HMAC_FIPS_MAX_KEY cap but also the
HMAC_FIPS_MIN_KEY floor, so wc_PRF accepted secrets under 14 bytes.
The kernel's testmgr HMAC vectors use 131-byte keys, which the
HMAC_FIPS_MAX_KEY cap rejected on 6.0+ kernels.

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Changes recommended

Critical active findings show that alternate hardware and callback implementations can bypass the new ECC curve-size and HMAC key-length restrictions, leaving the approved-mode enforcement incomplete.

3 open issues
3 resolved since last review

Review effort: Lite

Comment thread wolfcrypt/src/ecc.c Resolved
Comment thread wolfcrypt/src/hmac.c Resolved
Comment thread wolfssl/wolfcrypt/hmac.h Resolved

@wolfSSL-Fenrir-bot wolfSSL-Fenrir-bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fenrir Automated Review — PR #11359

Scan targets checked: linuxkm-src, linuxkm-bugs, wolfcrypt-src, wolfcrypt-bugs, wolfssl-src, wolfssl-bugs
Coverage: 7 of 7 in-scope changed file(s) opened by the reviewer

Findings: 3
3 finding(s) posted as inline comments (see file-level comments below)

This review was generated automatically by Fenrir. Reported findings require changes before merge.

Review tier: Max

Comment thread linuxkm/lkcapi_sha_glue.c Resolved
Comment thread wolfcrypt/src/kdf.c Resolved
Comment thread wolfcrypt/src/ecc.c Resolved
…k and hardware dispatch

The gate sat only in the software wc_ecc_sign_hash_ex(), so wc_ecc_sign_hash() could hand a sub-224-bit key to a crypto callback or hardware backend first, and the STM32 PKA wc_ecc_sign_hash_ex() had no check at all. All three entry points now share one check.

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

Approval is blocked by two critical correctness issues: the HMAC key-length exemption applies to ordinary setkey calls, and curve validation accepts non-approved curves based solely on byte size.

No open issues 🎉

2 previously missed issues

In code that hasn't changed since last review

High severity Self-test build bypasses the HMAC maximum for normal callers

linuxkm/​lkcapi_sha_glue.c:1420

WC_LINUX_CONFIG_SELFTESTS is a compile-time configuration flag, so this path is also used for ordinary crypto_shash_setkey calls after the kernel self-tests finish. In FIPS v7, any key longer than HMAC_FIPS_MAX_KEY is therefore sent through allowFlag = 1, which bypasses the new maximum and lets normal callers use an overlong HMAC key whenever kernel self-tests are enabled. The self-test exemption needs to be scoped to the kernel's self-test invocation (or normal setkey calls must be validated before this override) rather than applied to every call in a self-test-enabled build.

High severity Key-size check permits non-approved ECC curves

wolfcrypt/​src/​ecc.c:6548

Checking only the byte size is not sufficient for the stated FIPS v7 approved-curve restriction. When a non-NIST curve such as ECC_SECP256K1 is compiled, wc_ecc_get_curve_size_from_id() returns 32, so this check passes and _ecc_make_key_ex() generates a key on that non-approved curve. Please gate explicit curve IDs with the module's approved-curve allowlist (while preserving the default-curve resolution).

3 resolved since last review

Review effort: Lite

@wolfSSL-Fenrir-bot wolfSSL-Fenrir-bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fenrir Automated Review — PR #11359

Scan targets checked: linuxkm-src, linuxkm-bugs, wolfcrypt-src, wolfcrypt-bugs, wolfssl-src, wolfssl-bugs
Failed targets: wolfcrypt-port-bugs
Coverage: 3 of 4 in-scope changed file(s) opened by the reviewer; not opened: tests/api/test_ecc.c

Findings: 1
1 finding(s) posted as inline comments (see file-level comments below)

This review was generated automatically by Fenrir. Reported findings require changes before merge.

Review tier: Lite

Comment thread linuxkm/lkcapi_sha_glue.c
/* The kernel's own HMAC self-tests use keys longer than the v7
* HMAC_FIPS_MAX_KEY cap, so let those through. Without the kernel
* self-tests nothing needs the exemption and the cap applies. */
if (length > HMAC_FIPS_MAX_KEY)

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

linuxkm HMAC setkey skips the FIPS v7 key-length cap for every kernel caller, not only testmgr · Cryptographic correctness

WC_LINUX_CONFIG_SELFTESTS is a compile-time kernel config, and the kernel's CRYPTO_FIPS option requires it. On every FIPS-capable kernel, all in-kernel HMAC users (IPsec, dm-verity and others) can set keys longer than HMAC_FIPS_MAX_KEY with allowFlag=1. That undoes the v7 1024-bit cap this PR adds for F-9546.

Suggested fix: Fix needs a design decision: allow the long key only while the kernel self-test runs (as in the registration-time test window), or mark the result non-approved, per the module's FIPS 140-3 security policy.

Related known findings (similar but distinct; listed for context, not part of this finding)

  • Earlier finding on this pull request: (1) File/function: both are in linuxkm_hmac_setkey_common in lkcapi_sha_glue.c, so this part matches. (2) Faulting operation: both involve the wc_HmacSetKey_ex(..., 1) exemption for keys longer than HMAC_FIPS_MAX_KEY, so this part also matches. (3) Root cause: the historical issue is that the exemption applied in builds without WC_LINUX_CONFIG_SELFTESTS, where testmgr never runs. The candidate is the residual case. In builds with WC_LINUX_CONFIG_SELFTESTS, which every FIPS-capable kernel needs,

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

For FIPS v7 Module Related to FIPS v7.0.0 module prep for submission

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants