- stunnel-users - stunnel.org

certificate authentications
by John A. Wallace 21 Jan '12

21 Jan '12

I have two questions, which I think may be related, regarding how to use the information from stunnel log. I use stunnel to connect to an SMTP server on the internet from my home network, and in particular from my Windows laptop. My stunnel version is this: stunnel 4.50 on x86-pc-mingw32-gnu platform Compiled/running with OpenSSL 0.9.8r-fips 8 Feb 2011 It works well for my purposes, and I can see, by using a program for monitoring process and network connections, that the connections are now secured as expected. However, I believe it can be made more secure if I can utilize the certificate that is offered by the server, but I am not sure how to make that happen. In my stunnel log for the connection, I get this message: Client-mode smtp protocol negotiations started Client-mode smtp protocol negotiations succeeded No peer certificate received SSL connected: new session negotiated Negotiated ciphers: ADH-AES256-SHA SSLv3 Kx=DH Au=None Enc=AES(256) Mac=SHA1 My first question is, how should I go about getting that "No peer certificate received" issue corrected and how do I install it? Secondly, when I issue this command at the cmd shell prompt: openssl s_client -starttls smtp -connect host.server:port The output is lengthy and it includes, among other things, clearly what is identified as a certificate. I have been told that this is a good certificate, and one that I should utilize for an authenticated connection. So, my question is, is this the same certificate that I saw referenced in the log as the "peer certificate", and how do I go about putting this certificate where it belongs in my directory? I know how to copy it and save it as a file, but where do I put it and should it have a special name? If someone wants to direct me to the correct instruction for doing this, that would be fine too. I am just looking for some pointers for assistance. Thanks.

3 5

Re: [stunnel-users] ver 4.5 VS 4.2: running single process now, how to close when all client apps closed?
by Jose Alf. 21 Jan '12

21 Jan '12

All right. I just recompiled against the latest OpenSSL 0.9.8t. Still not enought to tempt you? :) Regards, Jose ________________________________ From: Larisa <larisaka(a)yahoo.com> To: Jose Alf. <josealf(a)rocketmail.com> Sent: Friday, January 20, 2012 5:20 PM Subject: Re: [stunnel-users] ver 4.5 VS 4.2: running single process now, how to close when all client apps closed? Jose, I am thinking to stay with the old version for a little bit longer, though hoping to get newer one 4.5_ soon. Thank you for the offer. Larisa --- On Thu, 1/19/12, Jose Alf. <josealf(a)rocketmail.com> wrote: >From: Jose Alf. <josealf(a)rocketmail.com> >Subject: Re: [stunnel-users] ver 4.5 VS 4.2: running single process now, how to close when all client apps closed? >To: "Stunnel Users Mailing List" <stunnel-users(a)mirt.net>, "larisa_y" <larisaka(a)yahoo.com> >Date: Thursday, January 19, 2012, 8:01 AM > > >Larisa, >I compiled stunnel-4.39 against OpenSSL 0.9.8s. Let me know if you are interested in the binaries. > >Regards >Jose > >________________________________ >From: Michal Trojnara <Michal.Trojnara(a)mirt.net> >To: Stunnel Users Mailing List <stunnel-users(a)mirt.net> >Sent: Tuesday, January 17, 2012 5:08 PM >Subject: Re: [stunnel-users] ver 4.5 VS 4.2: running single process now, how to close when all client apps closed? >larisa_y wrote:> Could you tell me starting which version the new behaviour was implemented?> (single process architecture) I tried the changelog but could not really> identify the item.The implementation of this change started with stunnel 4.40 with new "stunnel -exit" option. The last version fully supporting multiple processes is 4.39.Please be aware of security risks related to using obsolete versions of OpenSSL. Some Win32 GUI features are also unstable until version 4.50.Mike_______________________________________________stunnel-users mailing liststunnel-users@stunnel.orghttp://stunnel.mirt.net/mailman/listinfo/stunnel-users

1 0

Re: [stunnel-users] STARTTLS
by James Devine 20 Jan '12

20 Jan '12

http and smtp are two different birds, https runs on a different port, whereas an smtp connection running on port 25 can be upgraded to secure with starttls On Fri, Jan 20, 2012 at 12:29 AM, Michal Trojnara <Michal.Trojnara(a)mirt.net>wrote: > James Devine <fxmulder(a)gmail.com> wrote: > > >I've setup stunnel to tunnel port 25 to a few backend smtp servers, it > >appears that it only works with starttls, is there a way to make it > >work if > >no starttls is issued also? > >_______________________________________________ > >stunnel-users mailing list > >stunnel-users(a)stunnel.org > >http://stunnel.mirt.net/mailman/listinfo/stunnel-users > > Do you also have problems with https on port 80? > > Mike >

1 0

STARTTLS
by James Devine 19 Jan '12

19 Jan '12

I've setup stunnel to tunnel port 25 to a few backend smtp servers, it appears that it only works with starttls, is there a way to make it work if no starttls is issued also?

1 0

ver 4.5 VS 4.2: running single process now, how to close when all client apps closed?
by larisa_y 19 Jan '12

19 Jan '12

I have several client applications running on win machine, connecting to the same server, or sometimes to different servers. ver. 4.2 allowed me to do the following: modify .conf file to [my_section] accept=5566 connect 123.22.33.44:5005; then run stunnel.exe; modify .conf file to [my_section] accept=5567 connect 456.22.33.44:5006; then run stunnel.exe. That would start 2 stunnel.exe processes, each one handling its own connection; which were easy to close when each respective client app closed - each one would close the one it started. Now with 4.5 version: the single stunnel.exe process runs and is able to handle all communications perfectly, provided I prepare .conf file for all possible connections in advance, before starting stunnel. The problem I cannot solve easily is how not to leave stunnel running after client logged out of all the applications. Without building something to keep track of "stunnels users" count, how could I do it? Is there some way to switch to multiple instances mode? Please advise. Larisa -- View this message in context: http://old.nabble.com/ver-4.5-VS-4.2%3A-running-single-process-now%2C-how-t… Sent from the Stunnel - Users mailing list archive at Nabble.com.

3 7

Windows Live Mail issue ?
by Victor 19 Jan '12

19 Jan '12

I'm new with stunnel and I'm experiencing a issue. Windows Live Mail is closing the connection with the SSL tunnel without send the close down packet, resulting in connection reset by peer. Follow the output debug of stunnel: 2012.01.18 17:13:25 LOG7[16954:34377572800]: Socket closed on read 2012.01.18 17:13:25 LOG7[16954:34377572800]: Sending SSL write shutdown 2012.01.18 17:13:25 LOG7[16954:34377572800]: SSL alert (write): warning: close notify 2012.01.18 17:13:25 LOG6[16954:34377572800]: SSL_shutdown successfully sent close_notify 2012.01.18 17:13:25 LOG3[16954:34377572800]: SSL_read: Connection reset by peer (54) 2012.01.18 17:13:25 LOG5[16954:34377572800]: Connection reset: 293 bytes sent to SSL, 1293 bytes sent to socket 2012.01.18 17:13:25 LOG7[16954:34377572800]: Service stunnel finished (0 left) 2012.01.18 17:13:25 LOG7[16954:34377572800]: str_stats: 28 block(s), 988 byte(s) The e-mail message is sent successfully but Im very intrigued with the line "Connection reset: 293 bytes sent to SSL, 1293 bytes sent to socket", that make me think that some data was lost in the middle of the connection. I need to fix something or this is a normal behavior ? qmail 1.03 with SPAMCONTROL stunnel 4.43 running on FreeBSD 8.0-RELEASE Config file: ; SMTP SSL (SMTP-AUTH) debug = 7 foreground = yes syslog = no cert = /usr/local/etc/stunnel/stunnel.pem exec = /var/qmail/bin/qmail-smtpd execargs = qmail-smtpd /mailhome/bin/vchkpw /bin/true Thanks in advance !

2 1

client side SNI
by yyy 14 Jan '12

14 Jan '12

It looks like server name for client side SNI is defined as name to connect to in connect option. According to RFC, this name must be FQDN. It is the only place where it is defined and server name for SNI cannot be defined outside connect option. Is that right? Server side SNI has some good examples. Searching for "client side SNI" did not yield anything useful. I am trying to run multiple independent services over the same port there is no DNS infrastructure in place, so those server names would be random strings not refering to anything. Both ends would be stunnel. If it is as it seems, it will not work, because defined server name (at client) must be working DNS name (FQDN).

2 7

Verify = 3 and Giganews
by Thomas Eifert 13 Jan '12

13 Jan '12

Hello, I've tried to use verify = 3 with Giganews on several occasions now, but it always fails. It works fine with other servers, however. Here is the debug output for the Giganews scenario: 2012.01.12 14:05:01 LOG4[292:3840]: CERT: Verification error: unable to get local issuer certificate 2012.01.12 14:05:01 LOG4[292:3840]: Certificate check failed: depth=1, /C=US/O=GeoTrust Inc./OU=Domain Validated SSL/CN=GeoTrust DV SSL CA 2012.01.12 14:05:01 LOG3[292:3840]: SSL_connect: 14090086: error:14090086:SSL routines:SSL3_GET_SERVER_CERTIFICATE:certificate verify failed And the following is from my stunnel.conf: ~~~~~~~~~~~~~~~~~~~~~~~~~~~ debug = 6 delay = yes [nntps.3] client = yes sslVersion = TLSv1 ciphers = AES256-SHA cafile = peer-nntps.3.pem verify = 3 accept = 127.0.0.1:119 connect = news.giganews.com:443 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ As I have with other connections, I am saving peer-nntps3.pem from the log window menu. Can anyone tell me why this is failing on this server only? Regards; Thomas

2 2

stunnel 4.52 released
by Michal Trojnara 12 Jan '12

12 Jan '12

Dear Users, I have released version 4.52 of stunnel. This is a bugfix release. Upgrade is highly recommended. The ChangeLog entry: Version 4.52, 2012.01.12, urgency: MEDIUM: * Bugfixes - Fixed write closure notification for non-socket file descriptors. - Removed a line logged to stderr in inetd mode. - Fixed "Socket operation on non-socket" error in inetd mode on Mac OS X platform. - Removed direct access to the fields of the X509_STORE_CTX data structure. Home page: http://www.stunnel.org/ Download: ftp://ftp.stunnel.org/stunnel/ SHA-256 hash for stunnel-4.52.tar.gz: 7c78c178074e9b96331518a9c309d2e95ca9ad6e0338a96d5ab8ad47fde4347c Best regards, Mike

1 0

fips=no and slow performance problem
by Owen Ching 12 Jan '12

12 Jan '12

I have a production system that uses stunnel and it's been working pretty well. Mike, thanks for all your hard work. But there has been a weird issue that I ran into a while ago and now it's happening again. we're using a rackspace cloud machine to run stunnel and haproxy. we're using the x-forwarded-for stunnel patch for now with plans to upgrade to send-proxy method once haproxy 1.5 is considered the stable branch. So I built one machine and ran into the "FIPS_mode_set: 2D06C06E: error:2D06C06E:FIPS routines:FIPS_mode_set:fingerprint does not match" error message. So I changed the config to fips=no and stunnel started up but the https seems really slow (multiple browsers). We run with significant https volume. So the fips=no option didn't work for us. I kept trying different things but nothing worked. I decided to start clean and built a new machine. This time stunnel didn't throw the FIPS error and everything performed great. So used the new machine instead. Now after some time (over a year), we had some performance problems. we rebooted the machine and now we have the FIPS error again. I've tried multiple versions of stunnel (whatever I could find working patches for) and also tried a clean 4.51 with no patches. all of them throw the FIPS error now on this machine. I'm in the process of building a new machine to see if it magically works again. Any help or insight would be greatly appreciated. Thanks, Owen

2 2