Hi Kyle, thanks for the response (although I missed it at first). I do get a SSL_ERROR_SYSCALL although a subsequent call to ERR_error_string gives me nothing. I'm now not sure that I'm on the correct track though as I'll explain.
* We use OpenSSL to secure a "ping" from a mobile device to a server so that we can then receive data later. * We are seeing what we thought was a memory leak with each ping, although checking with Entrek CodeSnitch shows that we are getting a leak of 2.6MB whether we send a ping, once, ten or a hundred times. I believe that this is linked to the SSL_library_init method (which we only call once per application) and the fact that we don't call EVP_cleanup. * Both the SSL_CTX and SSL objects are cleared after each use using SSL_CTX_free and SSL_free and both of their reference counts are set to 0 after the freeing. * We get a steady 200k "leak" for each call. I read somewhere that OpenSSL keeps an internal cache and am wondering if this could be the cause. A bit more history is that as always the original developer has long flown the nest so I'm learning the library here. Thanks for any help you can give. Andy > -----Original Message----- > From: owner-openssl-us...@openssl.org [mailto:owner-openssl- > us...@openssl.org] On Behalf Of Kyle Hamilton > Sent: 14 May 2009 18:10 > To: openssl-users@openssl.org > Subject: Re: SSL_shutdown never returns 1 > > If you expect no more data from the client, you can simply destroy the > SSL context at the point you first call it (even if it returns 0). It > will only return 1 if it has received the peer's close_notify, and > it's possible that those peers failed to send it. (Do you get an > SSL_ERROR_SYSCALL with EINVAL or EPIPE?) > > -Kyle H > > On Thu, May 14, 2009 at 7:31 AM, Andy Murphy <an...@tbsmobility.com> > wrote: > > Hi, we are trying to track down a memory leak that occurs when we run > secure > > comms using the OpenSSL library. Now we’ve had this before where we > thought > > it was the library but I tracked it down to us not using the word > virtual on > > our base class’ destructor, so I’m not about to start blaming OpenSSL > again > > J. > > > > > > > > I have however seen from some other threads that if SSL_shutdown does > not > > return 1 then there could be a leak. Our logs show that two attempts > to > > call SSL_shutdown do not result in 1 coming back and I’ve just tried > it in a > > never-ending while loop and well, it never ended, suggesting that we > are > > doing something wrong. > > > > > > > > Do we know what SSL_shutdown depends on in order to be able to > shutdown > > correctly and thus free the memory used for the connection? > > > > > > > > Cheers > > > > > > > > Andrew Murphy > > > > Senior Software Developer > > > > Ironstone House, Kedleston Close, Belper, > > > > Derbyshire, England DE56 1TZ > > > > Office Reception: +44 1773 596900 > > Fax: +44 1773 596901 > > > > Skype: andyleemurphy > > Web: www.tbsmobility.com > > Email: andrew.mur...@tbsmobility.com > > > > > ______________________________________________________________________ > OpenSSL Project http://www.openssl.org > User Support Mailing List openssl-users@openssl.org > Automated List Manager majord...@openssl.org
