The wolfSSL library is a small, fast, portable implementation of TLS/SSL for embedded devices to the cloud. wolfSSL supports up to TLS 1.3 and DTLS 1.3!
wolfSSL Entering ClientSessionToSession
Setting session for wrong role
It uses wolfSSL_get_session which gets a reference to the ssl->session pointer but that gets free'd with wolfSSL_free.
The example as is won't work with NO_SESSION_CACHE_REF defined.
To make things more interesting: without NO_SESSION_CACHE_REF wolfSSL_get_session returns a pointer to the ClientCache. That is why it works without --enable-all .
--enable-all turns on NO_SESSION_CACHE_REF
If we have changed behavior for some of our existing API’s especially ones marked with WOLFSSL_ABI we need to review this more closely. Forcing our examples to use get1 vs get seems wrong. If there is a build case (like with NO_SESSION_CACHE_REF that needs to use get1 then the example should have macro logic for each supported case.
wolfSSL_get_session is marked with WOLFSSL_ABI. In that case it will need different logic based on NO_SESSION_CACHE_REF .
dgarske
commented
1 year ago
Contact Details
eric@wolfssl.com
Version
master
Description
wolfSSL_get_session is marked with WOLFSSL_ABI. In that case it will need different logic based on NO_SESSION_CACHE_REF
The resume example fails with --enable-all, but works with only --enable-session-ticket. Is this because of session cache? Is it expected not to work? https://github.com/wolfSSL/wolfssl-examples/blob/master/tls/client-tls-resume.c Failure:
It uses wolfSSL_get_session which gets a reference to the ssl->session pointer but that gets free'd with wolfSSL_free.
The example as is won't work with NO_SESSION_CACHE_REF defined.
To make things more interesting: without NO_SESSION_CACHE_REF wolfSSL_get_session returns a pointer to the ClientCache. That is why it works without --enable-all .
--enable-all turns on NO_SESSION_CACHE_REF
If we have changed behavior for some of our existing API’s especially ones marked with WOLFSSL_ABI we need to review this more closely. Forcing our examples to use get1 vs get seems wrong. If there is a build case (like with NO_SESSION_CACHE_REF that needs to use get1 then the example should have macro logic for each supported case.
wolfSSL_get_session is marked with WOLFSSL_ABI. In that case it will need different logic based on NO_SESSION_CACHE_REF .
Reproduction steps
No response
Relevant log output
No response