Is it safe to do SSL_shutdown again if socket returns SSL_ERROR_WANT_READ or SSL_ERROR_WANT_WRITE

Question

I am trying perform SSL_shutdown on a non-blocking socket. In the code below, i am trying poll after trying SSL_shutdown.

do {
    err = SSL_shutdown();
    if (err == 0) {
       // unidirectional shutdown success, will try SSL_shutdown once more and exit
    } else if (err == 1) {
       //shutdown complete. Exit
    } else {
       ec = SSL_get_error(ctx, err);
       if (ec == SSL_ERROR_WANT_READ) {
           rc = poll(fd, POLLIN|POLLPRI, timeout);
           if (rc > 0) {
               continue;
           } else {
               break;
           }
       } else if (ec == SSL_ERROR_WANT_WRITE) {
           rc = poll(fd, POLLOUT, timeout);
           if (rc > 0) {
               continue;
           } else {
               break;
           }
       } else {
           break;
       }
    }
} while(1);

After a successful poll is it possible that the socket is writeable or readable and triggering the shutdown will WANT_WRITE or WANT_READ again. Can the above code get stuck in an infinite loop?

Answer 1

From the official documentation :

If the underlying BIO is nonblocking, SSL_shutdown() will also return when the underlying BIO could not satisfy the needs of SSL_shutdown() to continue the handshake. In this case a call to SSL_get_error() with the return value of SSL_shutdown() will yield SSL_ERROR_WANT_READ or SSL_ERROR_WANT_WRITE. The calling process then must repeat the call after taking appropriate action to satisfy the needs of SSL_shutdown() .

In other words, it is not only safe but it is expected.

But note the "... after taking appropriate action to satisfy the needs of SSL_shutdown()..." . If your SSL object is implicitly backed by a file descriptor then it will automatically do the necessary reads and writes. If it is only backed by a memory BIO these read and writes must be done in your code. It is unclear from your code if it is backed by a memory BIO or not though.