First the basics, I am running openssl-0.9.7d, compiled on a WBEL system,
using ./config (linux-elf was the result), with -DOPENSSL_THREADS &
-DPURIFY, so my dev library plays nice with valgrind.
The problem arises when I try to multi-thread my application. Valgrind
produces this output :
==27982== Thread 7:
==27982== Invalid write of size 4
==27982== at 0x80765A9: lh_insert (in
/home/ladar/Projects/program/src/program)
==27982== by 0xB7370505: thread_wrapper (vg_libpthread.c:667)
==27982== by 0xB741083A: do__quit (vg_scheduler.c:2146)
==27982== Address 0xB608D090 is 308 bytes inside a block of size 1024 free'd
==27982== at 0xB74D3C2C: realloc (vg_replace_malloc.c:310)
==27982== by 0x806D132: default_realloc_ex (in
/home/ladar/Projects/program/src/program)
Can anyone tell me why my threads aren't playing nicely with each other?
None of the openssl related code is hit until after the threads are spawned.
In other words, each thread has its own : SSL_CTX, SSL, SSL_METHOD, & BIO.
They do share the same socket, but I use accept to create a specific socket
which is then passed to SSL_accept :
// Lock the accept function, and get a new connection.
pthread_mutex_lock(&ssl_accept_mutex);
temp_sock_descriptor = accept(*sock_descriptor, (struct sockaddr
*)&pin,
(socklen_t *)&address_size);
if (temp_sock_descriptor == -1) {
quiet_log(0, NULL, "accept returned an error code, in
waiter_ssl");
printf("Error while accepting connection.\n");
exit(1);
}
pthread_mutex_unlock(&ssl_accept_mutex);
// Setup the SSL object.
sbio = BIO_new_socket(temp_sock_descriptor, BIO_NOCLOSE);
ssl = SSL_new(ctx);
state = SSL_accept(ssl);
SSL_set_bio(ssl, sbio, sbio);
Can someone help me out here?
Thanks,
L~
______________________________________________________________________
OpenSSL Project http://www.openssl.org
User Support Mailing List [EMAIL PROTECTED]
Automated List Manager [EMAIL PROTECTED]
