Re: [reSIProcate] OpenSSLInit is not successfully ensuring the thread-safety of OpenSSL
Ok, good. At least this stuff isn't completely broken on Windows. I
will be fixing the *nix code today.
Best regards,
Byron Campen
CRYPTO_thread_id is implemented as follows, so we do not need to
specify a custom implementation for windows:
unsigned long CRYPTO_thread_id(void)
{
unsigned long ret=0;
if (id_callback == NULL)
{
#ifdef OPENSSL_SYS_WIN16
ret=(unsigned long)GetCurrentTask();
#elif defined(OPENSSL_SYS_WIN32)
ret=(unsigned long)GetCurrentThreadId();
#elif defined(GETPID_IS_MEANINGLESS)
ret=1L;
#else
ret=(unsigned long)getpid();
#endif
}
else
ret=id_callback();
return(ret);
}
On Wed, Nov 5, 2008 at 11:51 PM, Byron Campen <bcampen@xxxxxxxxxxxx>
wrote:
Here's the offending code:
*snip*
#if !defined(WIN32) && defined(PTHREADS)
CRYPTO_set_id_callback(OpenSSLInit::threadIdFunction);
#endif
*snip*
There is no such preprocessor macro as "PTHREADS". The correct
thing to be
checking for here, AFAICT, is "_POSIX_THREADS". We should change
this code
to something like
#if !defined(WIN32)
#if defined(_POSIX_THREADS)
CRYPTO_set_id_callback(OpenSSLInit::threadIdFunction);
#else
#error Can't set OpenSSL up to be threadsafe!
#endif
#endif
This assumes CRYPTO_thread_id() is set to something sensible by
default on
Windows; is this true Scott/Derek?
Also, this code would need to use _POSIX_THREADS
*snip*
unsigned long
OpenSSLInit::threadIdFunction()
{
#if defined(WIN32)
assert(0);
#else
#ifndef PTHREADS
assert(0);
#endif
unsigned long ret;
ret= (unsigned long)pthread_self();
return ret;
#endif
return 0;
}
*snip*
Best regards,
Byron Campen
_______________________________________________
resiprocate-devel mailing list
resiprocate-devel@xxxxxxxxxxxxxxx
https://list.resiprocate.org/mailman/listinfo/resiprocate-devel