20260720-linuxkm-CMAC-native

[douzzer]

Implement glue logic for Linux kernel registration of native cmac(aes) implementation, and fix gaps in several other struct shash_alg glue implementations:

linuxkm/lkcapi_aes_glue.c, linuxkm/lkcapi_glue.c, configure.ac: implement linuxkm cmac(aes) glue.

linuxkm/lkcapi_sha_glue.c: fix sha.W lifecycle management to prevent leaks:

• separate WC_LINUXKM_SHA_IMPLEMENT() into WC_LINUXKM_SHA1_IMPLEMENT() (no fixes
  needed) and WC_LINUXKM_SHA2_IMPLEMENT() (with associated new helpers
  WC_LINUXKM_SHA2_FREE_W(), WC_LINUXKM_SHA2_DECL_W(), WC_LINUXKM_SHA2_PUSH_W(),
  and WC_LINUXKM_SHA2_POP_W(), that move .W to a stack buffer).

• Reimplement SHA-2 one-shot digest callback to use only direct wolfCrypt calls
  rather than proxy to other callbacks.

linuxkm/linuxkm_wc_port.h, linuxkm/module_hooks.c: implement magic checks on kernel struct wolfSSL_Mutex in WOLFSSL_LINUXKM_VERBOSE_DEBUG builds.

linuxkm/lkcapi_sha_glue.c: refactor SHA-3 state around struct km_Sha3TfmCtx,
adding km_sha3_init_tfm(), km_sha3_exit_tfm(), km_sha3_alloc_tstate(), and
km_sha3_free_tstate(), implementing garbage collection of abandoned shash_desc
objects at .exit_tfm.

linuxkm/lkcapi_sha_glue.c: add struct km_sha3_export_state, and implement km_sha3_export() and per-alg _import() methods, and km_sha3_test_export_import().

linuxkm/lkcapi_sha_glue.c: fix export/import for HMAC:

• add to struct km_sha_hmac_pstate members desc_list_lock, desc_list, export_list, and export_list_len;
• add struct km_sha_hmac_export_state;
• add km_hmac_alloc_tstate(), km_hmac_export(), km_hmac_import(), km_hmac_test_export_import();
• and update km_hmac_*() and WC_LINUXKM_HMAC_IMPLEMENT() to implement.

test/:

• fixes for NO_DH;
• fixes in test_wc_ed448_import_public() and test_wc_Ed448DecisionCoverage() for FIPS v6;
• fixes in tests/api/test_sha3.c for KMAC keysize in FIPS builds.

tested with

wolfssl-multi-test.sh ...
check-source-text-fips-dev
quantum-safe-wolfssl-all-crypto-only-noasm-fips-dev-linuxkm-next-clang-tidy
'.*insmod.*'
all-crypto-openssl-extra-coexist-fips-dev

[github-actions]

MemBrowse Memory Report

linuxkm-pie

• Data: __patchable_function_entries +200 B (+0.8%, 26,160 B)

linuxkm-standard

• Data: __patchable_function_entries +176 B (+0.4%, 48,792 B)
  No memory changes detected for:
• gcc-arm-cortex-m0plus
• gcc-arm-cortex-m3
• gcc-arm-cortex-m4
• gcc-arm-cortex-m4-baremetal
• gcc-arm-cortex-m4-crypto-only
• gcc-arm-cortex-m4-dtls13
• gcc-arm-cortex-m4-min-ecc
• gcc-arm-cortex-m4-openssl-compat
• gcc-arm-cortex-m4-pkcs7
• gcc-arm-cortex-m4-pq
• gcc-arm-cortex-m4-rsa-only
• gcc-arm-cortex-m4-sp-math
• gcc-arm-cortex-m4-tls12
• gcc-arm-cortex-m4-tls13
• gcc-arm-cortex-m7
• gcc-arm-cortex-m7-pq
• gcc-arm-cortex-m7-tls13
• stm32-sim-stm32h753

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #10980

Scan targets checked: linuxkm-bugs, linuxkm-src, wolfcrypt-bugs, wolfcrypt-rs-bugs, wolfcrypt-src, wolfssl-bugs, wolfssl-src

Findings: 5
5 finding(s) posted as inline comments (see file-level comments below)

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

[philljj]

Retest this please.

FATAL: [ssh-agent] Unable to start agent
java.io.IOException: Unexpected EOF

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #10980

Scan targets checked: linuxkm-bugs, linuxkm-src, wolfcrypt-bugs, wolfcrypt-rs-bugs, wolfcrypt-src, wolfssl-bugs, wolfssl-src

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

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

[Frauschi]

🐺 Skoll Code Review

Overall recommendation: REQUEST_CHANGES
Findings: 4 total — 2 posted, 2 skipped

Posted findings

• [Medium] enabled_fips declaration/use guard mismatch can break -Werror kernel builds — linuxkm/lkcapi_glue.c:227
• [Medium] configure.ac lean-aesgcm-dev SHA512 logic inverted from disable to force-enable — configure.ac:7377-7379

Skipped findings

• [Medium] km_hmac_import() calls allocating wc_HmacCopy() while holding spinlock (sleep-in-atomic)
• [Low] cpuid_flag: reg[num] = 0 is dead code (overwritten by cpuid())

Review generated by Skoll via Claude/Codex

[Frauschi]

🟡 [Medium] enabled_fips declaration/use guard mismatch can break -Werror kernel builds
🚫 BLOCK bug

This PR renamed WC_LINUXKM_HAVE_SELFTEST to WC_LINUX_CONFIG_SELFTESTS and, at the two use sites of enabled_fips (lines 270 and 348/357), additionally added defined(CONFIG_CRYPTO_FIPS) to the preprocessor guard. However, the declaration of enabled_fips at line 227 was only renamed and did NOT get the CONFIG_CRYPTO_FIPS guard. As a result, when a target is built with HAVE_FIPS && WC_LINUX_CONFIG_SELFTESTS but WITHOUT CONFIG_CRYPTO_FIPS (a valid combination: wolfCrypt FIPS build on a kernel that has CONFIG_CRYPTO_MANAGER selftests but not CONFIG_CRYPTO_FIPS), static int enabled_fips is declared but never referenced. That yields a -Wunused-variable warning, which is fatal in the typical -Werror linuxkm build. Before this PR the declaration and use guards matched exactly, so this is a regression introduced by the diff.

Suggestion:

Suggested change

	#if defined(HAVE_FIPS) && defined(WC_LINUX_CONFIG_SELFTESTS)
	#if defined(HAVE_FIPS) && defined(CONFIG_CRYPTO_FIPS) && \
	defined(WC_LINUX_CONFIG_SELFTESTS)
	static int enabled_fips = 0;
	#endif

[Frauschi]

🟡 [Medium] configure.ac lean-aesgcm-dev SHA512 logic inverted from disable to force-enable
💡 SUGGEST question

The SHA512 block was inverted: the condition changed from test "$ENABLED_SHA512" != "no" (with the action DISABLING SHA512: -UWOLFSSL_SHA512) to test "$ENABLED_SHA512" = "no" (with the action force-ENABLING it: enable_sha512="yes"; ... -DWOLFSSL_SHA512 -DWOLFSSL_SHA384). This now mirrors the SHA256 block just above (line 7353), so SHA512 becomes mandatory in the lean-aesgcm-dev FIPS variant (presumably required by the SHA512-based DRBG). The surrounding SHA3/SHAKE/SHA224/SHA384 blocks still use the old disable pattern, so SHA512 is now the odd one out. The PR description does not mention this configure change. It looks intentional and consistent with SHA256, but it is a meaningful behavior flip in FIPS config selection worth an explicit confirmation.

Recommendation: Confirm SHA512 is meant to be force-enabled (not disabled) in the lean-aesgcm-dev FIPS variant, and consider noting it in the PR description since it is not part of the stated linuxkm CMAC scope.