Fix ECX/ECC get_params to size export buffer from data_size

[yosuke-wolfssl]

Summary

Fixes finding f_2236: ECX and ECC get_params used p->return_size instead of
p->data_size as the export buffer size.

Per the OSSL_PARAM contract, data_size is the caller's buffer capacity while
return_size is the provider's output field for the actual data length. Three
get_params helpers seeded the wolfSSL export length (outLen, an in/out
parameter) from return_size. When a param is reused — e.g. it carries a stale
or zero return_size from a prior size-query call — the export function receives
a wrong (often 0) buffer size and truncates or fails. It only worked for
freshly-constructed params because OpenSSL initializes return_size to an
unmodified sentinel rather than a small value.

Changes

• src/wp_ecx_kmgmt.c
  • wp_ecx_get_params_enc_pub_key: outLen now seeded from p->data_size.
  • wp_ecx_get_params_priv_key: same fix.
• src/wp_ecc_kmgmt.c
  • wp_ecc_get_params_enc_pub_key: same fix.
• test/test_ecx.c (+ test/unit.c, test/unit.h)
  • New test_ecx_x25519_get_params_stale_ret: constructs params with a valid
    32-byte buffer but forces return_size = 0 (the reused-param condition),
    then calls EVP_PKEY_get_params for both the private and public key and
    verifies the returned bytes (against the RFC 7748 §6.1 Alice key pair) and
    return_size.

The p->return_size = outLen output assignments at the end of each helper are
unchanged — those are correct. A repo-wide sweep of return_size/data_size
uses confirmed no other helper (ML-DSA, ML-KEM, ML-X, DH, RSA) misused
return_size as an input; those already size from data_size.

Testing

• New test passes; reverting the fix makes it fail (priv key get_params failed with stale return_size), confirming it catches the bug.
• All 34 ECX/ECC/ECDH/eckeygen unit tests pass.

[Copilot]

Pull request overview

This PR fixes ECX/ECC get_params export behavior to follow the OSSL_PARAM contract by sizing the export buffer from the caller-provided data_size (capacity) rather than return_size (provider-filled output length). This prevents failures/truncation when callers reuse an OSSL_PARAM that carries a stale (e.g., zero) return_size from a prior size-query call.

Changes:

• Update ECX get_params helpers to seed outLen from p->data_size for encoded public key and private key exports.
• Update ECC encoded public key get_params helper to seed outLen from p->data_size.
• Add a regression test that forces return_size = 0 while providing a valid 32-byte buffer, verifying correct returned bytes and updated return_size for X25519 private/public key params.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
src/wp_ecx_kmgmt.c	Fixes ECX export buffer sizing in get_params to use data_size as input capacity.
src/wp_ecc_kmgmt.c	Fixes ECC encoded public key export buffer sizing in get_params to use data_size as input capacity.
test/test_ecx.c	Adds a regression test covering reused-parameter/stale-return_size behavior for X25519 EVP_PKEY_get_params.
test/unit.c	Registers the new X25519 stale-return_size unit test.
test/unit.h	Declares the new X25519 stale-return_size unit test.

---

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

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #447

Scan targets checked: wolfprovider-bugs, wolfprovider-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.

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #447

Scan targets checked: wolfprovider-bugs, wolfprovider-src

No new issues found in the changed files. ✅