[ohki-saito]

Description

Fenrir Finding F-10621

wolfSSL_OBJ_txt2nid() and wolfSSL_OBJ_obj2nid() identified an OID only by its 32-bit wc_oid_sum() value, without comparing the encoded OID bytes. OID sums are not unique: 2.5.29.19.1.1.1.1.1.1.1.1 has the same sum as 2.5.29.19, because the eight added bytes cancel out in the XOR lanes. Both functions therefore returned NID_basic_constraints for the longer OID.

Certificate parsing already checks the OID bytes in GetOID(), so this PR only changes the two OBJ functions, which match by sum alone.

Changes:

• Add wolfssl_object_info_oid_match(), which compares the length and bytes of an OID with the encoding from OidFromId() for a candidate table entry.
• Split oid2nid() into oid2nid_ex(oid, grp, der, derSz), which also checks the encoding when it is given, and keep oid2nid(oid, grp) as a sum-only wrapper. The existing callers of oid2nid() are unchanged; its comment now states that it is for OIDs whose encoding was already checked when parsed.
• In wolfSSL_OBJ_obj2nid(), get the OID content bytes from the object and call oid2nid_ex().
• In wolfSSL_OBJ_txt2nid(), check the encoding before returning a NID.

Remaining limitations

These are not addressed in this PR, and I would appreciate guidance on whether to handle them separately:

• Table entries with no encoding in OidFromId() (38 of 143 with --enable-all, e.g. CN, O, some curves) are still matched by sum only, so OIDs colliding with them can still be misidentified.
• oid2nid() is still sum-only. Its fallback search across all groups could map an unknown OID to an entry of another type

Testing

Added tests for OBJ_txt2nid() and OBJ_obj2nid():

• 2.5.29.19 is still identified as NID_basic_constraints.
• 2.5.29.19.1.1.1.1.1.1.1.1 is no longer identified as NID_basic_constraints.

Checklist

• added tests
• updated/added doxygen
• updated appropriate READMEs
• Updated manual and documentation

[wolfSSL-Bot]

Can one of the admins verify this patch?

[Copilot]

🟡 Changes recommended

The regression test covers only a different-length OID collision, so it would still pass if the byte comparison were removed and same-length collisions were accepted.

1 open issue

• Add regression test for same-length OID collision · New

What changed in this PR

This PR addresses Fenrir F-10621 in wolfSSL’s OpenSSL-compatible object lookup functions by checking encoded OID bytes, rather than relying solely on OID sums, when a reference encoding is available.

Changes:

• Add shared OID length-and-byte validation to OBJ_txt2nid() and OBJ_obj2nid().
• Preserve the sum-only oid2nid() wrapper and document its intended use.
• Add regression tests for basic constraints and a colliding longer OID.

File	Description
tests/​api/​test_ossl_obj.c	Adds valid-OID and collision regression tests for both OBJ lookup functions.
src/​ssl_asn1.c	Adds encoding-aware lookup while preserving existing sum-only callers.

_{Review effort: Balanced}

[philljj]

Okay to test.