The OSSLCryptoFactory class destructor should not clean up the CRYPTO library state about softhsmv2 HOT 9 CLOSED

It is a bit more complex. There is no single call to initialize libcrypto:
https://wiki.openssl.org/index.php/Library_Initialization#libcrypto_Initialization
The application may only initialize a few algorithms it intends to use. This is safe with libcrypto, because libcrypto initialization (unlike libcrypto clean-up or PKCS#11 initialization) does not remove any accumulated internal state of the library. The initialization of every libcrypto component (for example an algorithm) has an internal check to prevent double initialization. It is also perfectly okay for the application to initialize the PKCS#11 library, and then initialize the libcrypto for the algorithms it wants to use directly.

As long as the application is still allowed to use libcrypto after PKCS#11 finalization, the only solution is to leave all the libcrypto clean-up to the application.

from softhsmv2.

And it's broken again after f1a8474

Please remember that SoftHSM2 is a library, and quite likely not the only code that uses OpenSSL in an application. Some other code my still need OpenSSL after SoftHSMv2 executed OSSLCryptoFactory::~OSSLCryptoFactory().

from softhsmv2.

Is there a way of checking if OpenSSL has already been initialized by the calling application?

Compare

from softhsmv2.

The order should be:

• Application initialize crypto library
• Application initialize PKCS#11 library
• PKCS#11 library initialize crypto library, if not initialized by the calling application
• Application finalize PKCS#11 library
• PKCS#11 library clean up crypto library, if it was initialized by the PKCS#11 library
• Application clean up crypto library

from softhsmv2.

It might be interesting to check how is PKCS#11 module provided by OpenSC (opensc-pkcs11.so) dealing with this. AFAIK it uses OpenSSL for general cryptographic tasks too so its authors were probably facing the same problem.

from softhsmv2.

OpenSC simply never invokes ENGINE_cleanup()...

from softhsmv2.

This is a critical issue with a very simple fix available as #184. Please merge it.

from softhsmv2.

Thank you for the fix, have merged the code.

from softhsmv2.

Thank you for merging it.

from softhsmv2.