[aidangarske]

F-10378 - 325fa7a - wolfPKCS11_Store_Name now declares its string unconditionally and guards the HOME and APPDIR lookup with WOLFPKCS11_NO_ENV, and a no_env CI job is added.

F-10379 - 5d0def0 - wp11_Token_Store no longer changes the token object count on a failed store or turns a TPM NV_SPACE error into success, and C_DeriveKey frees the derived object when AddObject fails.

F-10386 - b0ed4a4 - Store commit failures are now reported: a new wolfPKCS11_Store_CloseCommit returns rename and close errors and discards a failed write, and the user PIN, SO PIN and token reset paths restore their previous state when the store fails.

F-10384 - 617f7b4 - For the file backend, wp11_Token_Store no longer replaces the token record with an empty temporary file before the objects are written.

F-11505 - c79ea02 - WP11_Library_Final now returns the first slot store error after finishing cleanup, and C_Finalize maps it to CKR_FUNCTION_FAILED.

F-8645 - 5bb3614 - WP11_Library_Init now tracks the global lock, wolfCrypt, the global RNG and initialized slots and unwinds them when initialization fails.

F-10385 - 844537c - wp11_Object_New now frees only a non-NULL object, and WP11_Object_Free(NULL) does nothing.

F-13243 - fe5d444 - Destroying a token object now reports the first unstore error, ignoring NOT_AVAILABLE_E, and the file backend treats a missing file as success on remove.

F-2791 - ca3304f - wp11_Token_Final now clears the token with wc_ForceZero instead of XMEMSET.

F-11502 - 97b89ab - wp11_Token_Load now reads nextObjId only when the token flags read succeeded, and falls back only on BUFFER_E or a successful read of 0, so other errors fail the load.

F-10380 - 2711d6d - The TPM store now sizes ML-DSA, ML-KEM, HSS and XMSS records, and the handle type mask is widened from 0x0F to 0x1F so these types get distinct handles. Existing handles below 0x10 are unchanged.

[Copilot]

🟡 Changes recommended

A failed user-PIN change can leave committed object records encrypted under a discarded key, making them unreadable after restart.

3 open findings

• Atomically roll back PIN changes across committed objects · New
• Return 77 when feature-gated tests are skipped · New
• Return 77 when feature-gated tests are skipped · New

What changed in this PR

Hardens token persistence, lifecycle cleanup, stored-object validation, and PKCS#11 operation handling.

Changes:

• Propagates storage failures and improves rollback/cleanup behavior.
• Strengthens stored-object decoding, policy integrity, and session lifecycle handling.
• Adds targeted tests and a no-environment CI configuration.

File	Description
.github/​workflows/​unit-test.yml	Adds no-environment CI coverage.
src/​crypto.c	Improves operation cleanup and object failure handling.
src/​internal.c	Implements persistence, rollback, loading, and lifecycle changes.
src/​slot.c	Corrects token information and mechanism metadata.
src/​wolfpkcs11.c	Propagates finalization errors and pads information fields.
tests/​include.am	Registers the new test programs.
tests/​info_mech_table_test.c	Tests information and mechanism contracts.
tests/​login_pin_state_test.c	Tests PIN, login, and concurrency invariants.
tests/​object_policy_integrity_test.c	Tests persisted key-policy integrity.
tests/​object_store_decode_test.c	Tests malformed stored-record handling.
tests/​pkcs11mtt.c	Updates PKCS#11 v3 cancellation expectations.
tests/​pkcs11test.c	Updates clock and cancellation tests.
tests/​session_op_lifecycle_test.c	Tests operation cleanup and cancellation.
tests/​token_store_test.c	Tests storage failures and recovery.
wolfpkcs11/​internal.h	Adds internal limits and lifecycle APIs.
wolfpkcs11/​pkcs11.h	Defines the cancellation failure result code.

🧠 Review effort: Balanced

---

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

[Frauschi]

A few comments, inline.

[Frauschi]

Thanks, all addressed. The TPM half of my first comment (the positive RC from wolfPKCS11_Store_Remove) landed in #244 rather than here, which is fine as long as they merge in order. nss-test is only red here because of the #231 issue.

The merge-base changed after approval.