On Thu, May 28, 2009 at 3:32 PM, Victor Duchovni <victor.ducho...@morganstanley.com> wrote: > NO! You call > > n = SSL_write(myssl, buffer, len); > err = SSL_get_error(myssl, n); > > "err" may be SSL_ERROR_WANT_READ or SSL_ERROR_WANT_WRITE, in which case > you must retry the write again later, after there is room. > > On success (n>=0), or with WANT_READ/WANT_WRITE, arrange to move data from > the network bio to the peer and from the peer to the network bio. This > activity may be asynchronous if you are using non-blocking sockets with > select/epoll/... in an event loop. In that case you mark the socket for > read/write interest with appropriate callbacks.
So, in the case of WANT_READ or WANT_WRITE, I interact with the network BIO through BIO_read and BIO_write, and recv()'d data gets dumped into the BIO via BIO_write() and data to be sent should be retrieved via BIO_read()? In the case I get an epoll-triggered recv() from the socket, do I dump that data into the BIO via BIO_write, and follow that with a SSL_read()? Once this is all over, I'll write a little piece of example code that could be thrown into a document somewhere - I would be surprised if this wasn't a common bit of misunderstanding for developers. -Aaron ______________________________________________________________________ OpenSSL Project http://www.openssl.org User Support Mailing List openssl-users@openssl.org Automated List Manager majord...@openssl.org
