Fix global ref lifecycle for verify and missing-CRL callbacks

[cconlon]

This PR fixes a JNI callback global-reference lifecycle bug (F-1024) where setVerify() / setCRLCb() on one WolfSSLContext or WolfSSLSession could delete a callback ref another thread was about to dereference inside the corresponding native callback.

Verify callback

• Moves the verify callback jobject from a process-wide static to per-WOLFSSL_CTX ex_data, allocated in Java_com_wolfssl_WolfSSL_init() after wolfSSL_Init(). NativeVerifyCallback() resolves the cb via store -> ssl -> ctx -> ex_data, so each context fires its own user callback.
• Adds g_verifyCbMutex plus NativeVerifyCbLock/Unlock helpers.

Missing-CRL callbacks

• Keeps both CRL jobjects as process-globals (wolfSSL CbMissingCRL has no SSL/CTX argument), but adds g_crlCbMutex shared across g_crlCtxCbIfaceObj and g_crlCbIfaceObj to close the same race for setCRLCb() on both WolfSSLContext and WolfSSLSession.
• setCRLCb(null) now also calls wolfSSL_(CTX_)SetCRL_Cb(..., NULL) so the native cb is unregistered when the Java side clears it.
• Stops deleting CRL globals in freeContext() / freeSSL(); only setCRLCb() overwrites them and JNI_OnUnload() releases them via new NativeWolfSSLContextCleanup() / NativeWolfSSLSessionCleanup() helpers.

Behavior changes

• NativeVerifyCallback now fails closed (returns 0, handshake fails) when no Java callback is registered on the connection's WOLFSSL_CTX. Previously the process-global verify cb was invoked across all contexts.
• NativeMissingCRLCallback / NativeCtxMissingCRLCallback now drop silently when no Java cb is registered. Previously a WolfSSLException / WolfSSLJNIException would be thrown back into native wolfSSL.

Tests

• test_WolfSSLContext_setVerifyMultipleContexts - install / replace / unregister / interleaved-free lifecycle across two contexts.
• test_WolfSSLContext_setVerifyHandshakeRouting - handshake on two client CTXs asserts each cb fires only for its own handshake.
• test_WolfSSLContext_setCRLCbMultipleContexts - parallel lifecycle test for both WolfSSLContext.setCRLCb and WolfSSLSession.setCRLCb, asserts no spurious cb fires across contexts.

Fixes F-1024.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Fixes JNI global-ref lifecycle issues by making the verify callback reference per-WOLFSSL_CTX (via ex_data) and adjusting missing-CRL callback cleanup to only happen at library unload.

Changes:

• Store per-context verify callback jobject in WOLFSSL_CTX ex_data and resolve it in NativeVerifyCallback().
• Move missing-CRL callback global-ref cleanup to JNI_OnUnload() via new cleanup helpers.
• Add a JUnit test covering multi-context setVerify() install/replace/unregister/free flows.

Reviewed changes

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

Show a summary per file

File	Description
src/test/com/wolfssl/test/WolfSSLContextTest.java	Adds a multi-context setVerify() lifecycle test intended to surface global-ref mishandling.
native/com_wolfssl_globals.h	Declares the ex_data index global and new cleanup helper prototypes.
native/com_wolfssl_WolfSSLSession.c	Adds JNI_OnUnload() cleanup for the process-global session missing-CRL callback; stops freeing it in freeSSL().
native/com_wolfssl_WolfSSLContext.c	Moves verify callback storage to per-CTX ex_data, adds CTX-level missing-CRL cleanup helper, and adjusts freeContext() cleanup.
native/com_wolfssl_WolfSSL.c	Allocates the ex_data index during WolfSSL_init() and calls cleanup helpers in JNI_OnUnload().

---

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

[Copilot]

Pull request overview

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

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #369

Scan targets checked: wolfssljni-bugs, wolfssljni-src

No new issues found in the changed files. ✅

[Copilot]

Pull request overview

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

[cconlon]

Force pushed to add a mutex around the missing CRL callback.