SE050 fixes and new simulator tests by LinuxJedi · Pull Request #10196 · wolfSSL/wolfssl

LinuxJedi · 2026-04-12T09:53:07Z

Description

This fixes bugs in the Ed25519 and RSA handling for the SE050 as well as introduces a new simulator so that we have regression testing from now on.

- se050_ed25519_verify_msg: initialize *res = 0 at entry so failures don't leak a stale res = 1 from a prior good verify. - Ed25519 import functions: reset keyIdSet / keyId under WOLFSSL_SE050 in wc_ed25519_import_private_key_ex, wc_ed25519_import_private_only, wc_ed25519_import_public_ex so overwriting host-side key material invalidates any prior SE050 object binding. - New workflow .github/workflows/se050-sim.yml: builds wolfSSL against the NXP Plug&Trust SDK and runs the wolfCrypt tests against the SE050Sim simulator. Patches the upstream Dockerfile to use the PR's wolfSSL source. - ed25519_test SE050 adjustments: - Cap the RFC 8032 loop at 5 iters — iter 5's 1023 B msg exceeds NXP SDK SE05X_TLV_BUF_SIZE_CMD = 900. - rareEd verifies and private-only sign: expect WC_HW_E (SE050 delegates malformed-input rejection to the secure element) instead of BAD_FUNC_ARG / SIG_VERIFY_E. - Skip ed25519ctx_test / ed25519ph_test — SE050 port drops the context/prehash params so RFC 8032 ctx/ph vectors can't byte-match.

- se050_rsa_verify: when the function uploads only the public part of the key (keyCreated == 1), erase the transient SE050 object and don't persist keyIdSet = 1. A subsequent sign on the same RsaKey was reusing the public-only SE050 object and failing. Pre-existing bindings (from wc_RsaUseKeyId or a prior sign that uploaded a keypair) are preserved untouched. - rsa_keygen_test: add WOLFSSL_SE050 to the existing WOLFSSL_CRYPTOCELL guard around the export-then-decode round-trip. SE050-generated keys keep their private components in the secure element, so wc_RsaKeyToDer + wc_RsaPrivateKeyDecode cannot complete. Matching guard on the idx declaration to avoid an unused-variable warning.

Copilot

Pull request overview

This PR fixes SE050-related crypto handling issues (RSA verify key binding and Ed25519 verify result handling / test expectations) and adds a GitHub Actions workflow to run wolfCrypt tests against an SE050 software simulator for regression coverage.

Changes:

Adjust SE050-specific behavior in wolfCrypt tests (skip/modify vectors and expected error codes where SE050 diverges from host-side validation).
Fix SE050 RSA verify to avoid persisting a public-only key binding that would break subsequent sign operations on the same RsaKey.
Add a CI workflow that builds and runs wolfCrypt tests against the external SE050 simulator.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 5 comments.

File	Description
`wolfcrypt/test/test.c`	Updates RSA/Ed25519 tests to handle SE050 secure-element constraints and differing error codes.
`wolfcrypt/src/port/nxp/se050_port.c`	Fixes RSA verify keyId binding behavior; initializes Ed25519 verify result.
`wolfcrypt/src/ed25519.c`	Resets SE050 key binding on import to avoid using stale SE050 objects after key material changes.
`.github/workflows/se050-sim.yml`	Adds CI job to build/run wolfCrypt tests against the SE050 simulator in Docker.

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

Copilot · 2026-04-12T09:58:41Z

wolfcrypt/src/port/nxp/se050_port.c


+    *res = 0;
+


se050_ed25519_verify_msg() now unconditionally writes *res = 0; before any argument validation. If the caller passes res == NULL this will segfault (and the SE050 path in wc_ed25519_verify_msg_ex() currently skips the host-side sanity checks). Add an explicit res != NULL check before dereferencing (and ideally mirror the host-side NULL checks for signature/msg/key on the SE050 path).

Suggested change

*res = 0;

if (res == NULL) {

return BAD_FUNC_ARG;

}

*res = 0;

if (signature == NULL || msg == NULL || key == NULL) {

return BAD_FUNC_ARG;

}

Copilot · 2026-04-12T09:58:41Z

wolfcrypt/src/ed25519.c

        return BAD_FUNC_ARG;

+#ifdef WOLFSSL_SE050
+    /* Importing new key material invalidates any prior SE050 object binding. */


In the SE050 build, clearing key->keyIdSet/key->keyId on public-key import drops the binding but does not erase the previously-uploaded SE050 object. Because se050_ed25519_free_key() only erases when keyIdSet==1, this change can leak transient key objects (and leave stale key material) when reusing an ed25519_key for multiple imports. Consider calling se050_ed25519_free_key(key) (or otherwise erasing the old object when keyIdSet==1) before resetting these fields.

Suggested change

/* Importing new key material invalidates any prior SE050 object binding. */

/* Importing new key material invalidates any prior SE050 object binding. */

if (key->keyIdSet == 1) {

se050_ed25519_free_key(key);

}

Copilot · 2026-04-12T09:58:41Z

wolfcrypt/src/ed25519.c

        return BAD_FUNC_ARG;

+#ifdef WOLFSSL_SE050
+    /* Importing new key material invalidates any prior SE050 object binding. */


In the SE050 build, clearing key->keyIdSet/key->keyId on private-only import drops the binding but does not erase the previously-uploaded SE050 object. Because se050_ed25519_free_key() only erases when keyIdSet==1, this can leak transient key objects (and leave stale key material) when reusing an ed25519_key for multiple imports. Consider calling se050_ed25519_free_key(key) (or otherwise erasing the old object when keyIdSet==1) before resetting these fields.

Suggested change

/* Importing new key material invalidates any prior SE050 object binding. */

/* Importing new key material invalidates any prior SE050 object binding. */

if (key->keyIdSet) {

se050_ed25519_free_key(key);

}

Copilot · 2026-04-12T09:58:41Z

wolfcrypt/src/ed25519.c

    }

+#ifdef WOLFSSL_SE050
+    /* Importing new key material invalidates any prior SE050 object binding. */


In the SE050 build, clearing key->keyIdSet/key->keyId on private+public import drops the binding but does not erase the previously-uploaded SE050 object. Because se050_ed25519_free_key() only erases when keyIdSet==1, this can leak transient key objects (and leave stale key material) when reusing an ed25519_key for multiple imports. Consider calling se050_ed25519_free_key(key) (or otherwise erasing the old object when keyIdSet==1) before resetting these fields.

Suggested change

/* Importing new key material invalidates any prior SE050 object binding. */

/* Importing new key material invalidates any prior SE050 object binding.

* If this key was previously uploaded to SE050, erase that object before

* dropping the binding metadata.

*/

if (key->keyIdSet == 1) {

ret = se050_ed25519_free_key(key);

if (ret != 0) {

return ret;

}

}

Copilot · 2026-04-12T09:58:42Z

.github/workflows/se050-sim.yml

+env:
+  # Pin the simulator to a known-good revision. Bump this deliberately after
+  # validating upstream changes in a standalone PR.
+  SE050SIM_REF: main


The workflow comment says the simulator is pinned to a "known-good revision", but SE050SIM_REF is set to main, which is a moving target and can make CI non-reproducible. Pin this to an immutable ref (tag or commit SHA), and update it intentionally when validating upstream changes.

Suggested change

SE050SIM_REF: main

SE050SIM_REF: "<FULL_40_CHAR_VALIDATED_SE050SIM_COMMIT_SHA>"

LinuxJedi added 2 commits April 12, 2026 07:51

Copilot AI review requested due to automatic review settings April 12, 2026 09:53

Copilot started reviewing on behalf of LinuxJedi April 12, 2026 09:55 View session

Copilot AI reviewed Apr 12, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

SE050 fixes and new simulator tests#10196

SE050 fixes and new simulator tests#10196
LinuxJedi wants to merge 2 commits intowolfSSL:masterfrom
LinuxJedi:se050-fixes3

LinuxJedi commented Apr 12, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 12, 2026

Uh oh!

Copilot AI Apr 12, 2026

Uh oh!

Copilot AI Apr 12, 2026

Uh oh!

Copilot AI Apr 12, 2026

Uh oh!

Copilot AI Apr 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

-    /* Importing new key material invalidates any prior SE050 object binding. */
+    /* Importing new key material invalidates any prior SE050 object binding.
+     * If this key was previously uploaded to SE050, erase that object before
+     * dropping the binding metadata.
+     */
+    if (key->keyIdSet == 1) {
+        ret = se050_ed25519_free_key(key);
+        if (ret != 0) {
+            return ret;
+        }
+    }

	SE050SIM_REF: main
	SE050SIM_REF: "<FULL_40_CHAR_VALIDATED_SE050SIM_COMMIT_SHA>"

Conversation

LinuxJedi commented Apr 12, 2026

Description

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 12, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 12, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 12, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 12, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 12, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants