- stunnel-users - stunnel.org

Unable to Stop Service - Windows XP SP3 x86
by Thomas Eifert 16 Oct '12

16 Oct '12

I just uninstalled Stunnel version 4.53 and replaced it with 4.54 on my Windows XP SP3 x86 machine. What I generally do is stop the service, uninstall the service, then uninstall Stunnel prior to installing the new version. When I attempted to stop the version 4.53 service via the start menu shortcut, I was unsuccessful; all that happened was that the Stunnel status/log window would launch. This was repeatable. Likewise, when I attempted to stop it from an elevated command prompt using stunnel.exe -stop, the result was the same. I am administrator on this machine. I launched services.msc, and was able to stop it from there without any difficulty. After I installed 4.54, just to see what would happen, I attempted to stop the service again using the same methods. Same result.. I just can't do it in the conventional manner. I checked to ensure that there was only one Stunnel service/process running, and that checks out okay. Not sure if others are seeing this issue, but thought I'd better let someone know about it. Thanks for the new version! Regards, Thomas -- Attention: This message and all attachments are private and may contain information that is confidential and privileged. If you received this message in error, please notify the sender by reply email and delete the message immediately.

1 0

X-Forwarded-For stunnel 4.54 patch
by Michael Wittig 16 Oct '12

16 Oct '12

Hi, I created the X-Forwarded-For stunnel 4.54 patch: https://gist.github.com/3899818 Could someone else check it? -- Michael

1 0

Client Authentication and CRL Verification
by Jean-Philippe Constant 16 Oct '12

16 Oct '12

Hi, I am using stunnel in server mode with mutual authentication. The PKI used to authenticate my client is the following : root CA -> Intermediate CA -> Client. My stunnel configuration is : CAfile = RootCA.pem CRLfile = IntermediateCACRL.pem verify = 2 RootCA.pem contains the Root CA certificate IntermediateCACRL.pem contains the CRL file of the Intermediate CA The client authentication with client certificate goes well. The problem occurs when a client certificate is revoked. After the Intermediate CA CRL updates, the client certificate is still accepted whereas it should be refused. With the following configuration the revoked certificate is refused : CAfile = IntermediateCA.pem CRLfile = IntermediateCACRL.pem verify = 2 but I would prefer using the first configuration. Everything happens like if stunnel checks the crl only for the CA certificate and not for the whole certification chain. Thank you for your answers, Jean-Philippe Constant

1 0

SSL proxy in Windows - IP Address Issue
by Brian McGinity 16 Oct '12

16 Oct '12

Thank you so much for stunnel. This program is amazing!!!! In Windows, I'd like to move SSL processing out of Apache and into stunnel. Is there a way for Apache to know the client's ip address? - In Windows, "transparent" mode is unavailable - "xforwardedfor" is for smtp and not https connections - "protocol = proxy" is for Haproxy only? Brian

2 3

Asterisk with stunnel
by manjula hettiarachchi 15 Oct '12

15 Oct '12

Hi List I am new to stunnel, Just tried today to send Asterisk SIP request to stunnel client and forward to stunnel-server --> asterisk on another machine. (server ip=x.x.x.x) (client ip=y.y.y.y) Client config: cert=/etc/stunnel/stunnel.pem client=yes pid=/tmp/stunnel.pid debug=7 output=/etc/stunnel/stunnel.log [sip] accept=16380 connect=x.x.x.x:17380 ================================ Server Config: cert=/etc/stunnel/stunnel.pem #client=no pid=/tmp/stunnel.pid debug=7 output=/etc/stunnel/stunel.log [sip-IN] accept=17380 connect=5060 When tried 'telnet localhost 16380 from client machine i could see packets coming to server. However when Asterisk sends a sip request to port 16380, exten => _X.,1,Dial(SIP/${EXTEN}@y.y.y.y:16380,20,rt) i dont see packets forwaded to the server. Pls advise if anyone has used this scenario Best Regards Manjula

2 1

Re: [stunnel-users] stunnel-users Digest, Vol 99, Issue 2
by Scott McKeown 10 Oct '12

10 Oct '12

Hi Michal, Thats great news and thank you for update. I think the 'renegotiation' option will be a great addition. -- With Kind Regards. Scott McKeown Loadbalancer.org http://www.loadbalancer.org

1 0

stunnel 4.54 released
by Michal Trojnara 09 Oct '12

09 Oct '12

Dear Users, I have released version 4.54 of stunnel. The ChangeLog entry: Version 4.54, 2012.10.09, urgency: MEDIUM: * New Win32 features - FIPS module updated to version 2.0. - OpenSSL DLLs updated to version 1.0.1c. - zlib DLL updated to version 1.2.7. - Engine DLLs added: 4758cca, aep, atalla, capi, chil, cswift, gmp, gost, nuron, padlock, sureware, ubsec. * Other new features - "session" option renamed to more readable "sessionCacheTimeout". The old name remains accepted for backward compatibility. - New service-level "sessionCacheSize" option to control session cache size. - New service-level option "reset" to control whether TCP RST flag is used to indicate errors. The default value is "reset = yes". - New service-level option "renegotiation" to disable SSL renegotiation. This feature is based on a public-domain patch by Janusz Dziemidowicz. - New FreeBSD socket options: IP_FREEBIND, IP_BINDANY, IPV6_BINDANY (thx to Janusz Dziemidowicz). - New parameters to configure TLS v1.1/v1.2 with OpenSSL version 1.0.1 or higher (thx to Henrik Riomar). * Bugfixes - Fixed "Application Failed to Initialize Properly (0xc0150002)" error. - Fixed missing SSL state debug log entries. - Fixed a race condition in libwrap code resulting in random stalls (thx to Andrew Skalski). - Session cache purged at configuration file reload to reduce memory leak. Remaining leak of a few kilobytes per section is yet to be fixed. - Fixed regression bug in "transparent = destination" functionality (thx to Stefan Lauterbach). This bug was introduced in stunnel 4.51. - "transparent = destination" is now a valid endpoint in inetd mode. - "delay = yes" fixed to work even if specified *after* "connect" option. - Multiple "connect" targets fixed to also work with delayed resolver. - The number of resolver retries of EAI_AGAIN error has been limited to 3 in order to prevent infinite loops.. Home page: https://www.stunnel.org/ <http://stunnel.mirt.net/> Download: https://www.stunnel.org/downloads.html <ftp://stunnel.mirt.net/stunnel/> SHA-256 hash for stunnel-4.54.tar.gz: b7e1b9e63569574dbdabee8af90b8ab88db3fe13dcb1268d59a1408c56e6de7b Best regards, Mike

1 0

How to forward client's IP address with stunnel on Windows
by Emmanuel Egensperger 08 Oct '12

08 Oct '12

Dear stunnel's users, I have search on google and try many things on stunnel's configuration to forward ip address, but nothing works ? I have found this option in the FAQ : transparent = source But it seems that it doesn't work on windows ? >Line 22: "transparent = source": Specified option name is not valid here >Server is down Is there any solution to my problem ? Have I done something wrong ? Best regards, Manu Egen

2 1

Timeouts and inetd mode
by Bruce Guenter 28 Sep '12

28 Sep '12

Hi. As far as I can tell, if stunnel is running in "inetd mode", at least TIMEOUTbusy does not work. In particularl, stunnel will never timeout waiting for the SSL handshake to complete. If I'm reading the sources right, this is because the code to set up the SSL session calls SSL_accept before checking for data on stdin, and stdin is never set to non-blocking mode (only newly created sockets and/or pipes are). Would it make sense to set stdin (and stdout) to non-blocking mode when timeouts are enabled? Something in init_local maybe? -- Bruce Guenter <bruce(a)untroubled.org> http://untroubled.org/

1 0

SNI with wildcard certificates
by Gustavo Machado 27 Sep '12

27 Sep '12

Hi, I would like to use SNI in combination with wildcard certificates. I currently have the following configuration: [https] accept = 443 connect = 80 TIMEOUTclose = 0 cert = C:\default.pem key = C:\default.key ;SNI Config ;========== [https_a] sni = https:one.org connect = 80 cert = C:\a.pem key = C:\a.key [https_a_wc] sni = https:*.one.org connect = 80 cert = C:\b.pem key = C:\b.key If I go to https://one.org then I can see a.pem in the browser. If I go to https://mysubdomain.one.org I see default.pem. What am I missing?? Thanks! Gustavo

2 1

Chain pem files
by Gustavo Machado 27 Sep '12

27 Sep '12

Hi guys, Sorry for the newbie question, but I have the following configuration for an HTTPS endpoint: [https_one] sni = https:one.org connect = 44300 cert = D:\node\https\default.crt key = D:\node\https\default.key It works wonders, however I need to provide to this configuration two more .pem files for "chain". How would go about doing that? Thanks, Gustavo

1 1

Stunnel build with openssl 1.0+
by Gustavo Machado 27 Sep '12

27 Sep '12

Hi guys, I would like to use SNI and looks like I need a build with openssl 1.0+ Can you guys point me to a download link to the installer for that build? Thanks, Gustavo

2 2

(no subject)
by larisaka＠yahoo.com 25 Sep '12

25 Sep '12

http://www.steam-n-gleam.com/wp-content/plugins/bookingbug/google.html?fs=z…

1 0

Stunnel blocks booting process when no network connectivity
by Muschi Stevenson 25 Sep '12

25 Sep '12

My stunnel implementation works fine when the network is plugged in but it does not start at all, which stops the whole boot process, when there is no network connected on the machine (as a client). As extra information: - I'm using "delay=yes" - I'm using an fqdn (e.g: stunnel.mydomain.com) for the connections - Using stunnel 4.53 - Using ubuntu 12.04 but this also happened with centos5 previously however after some time it did boot on centos which doesn't happen in ubuntu How can this be avoided or a timeout specified? Doing an strace as suggested by symcbean shows the following (including the last part where it hangs): [...] --- SIGCHLD (Child exited) @ 0 (0) --- rt_sigreturn(0x11) = 0 close(3) = 0 wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 6039 clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7ff9ce0c79d0) = 6046 wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 6046 --- SIGCHLD (Child exited) @ 0 (0) --- rt_sigreturn(0x11) = 6046 write(1, "[Started: /etc/stunnel/stunnel.c"..., 37) = 37 write(1, "stunnel.\n", 9) = 9 exit_group(0) = ? [...] stunnel hangs in this line: wait4(-1, and when i plug in the network cable it continues to show [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 6046 Any help would be appreciated as this is critically blocking our deployment http://serverfault.com/questions/429072/stunnel-delaying-boot

2 6

[patch] options-related-to-TLS-v1.1-v1.2
by Henrik Riomar 21 Sep '12

21 Sep '12

Hi, When building stunnel against OpenSSL 1.0.1 support for TLS v1.1/v1.2 gets included. This patch adds options to configure these protocols in stunnel.conf. The patch is against 4.53 and under public domain. Best Regards, Henrik

2 3

"Broken pipe" issue
by Kai M. Becker 20 Sep '12

20 Sep '12

Hi :) although several postings state that this issue has been fixed, I'm still getting this error message: Error detected on SSL (read) file descriptor: Broken pipe (32) OS: OpenSuSE 11.4 stunnel: 4.53 The original problem for taking a closer look though, are hanging incoming connections. I.e. a https request is not forwarded to the backend application, but hangs somewhere in limbo. My current workaround are TIMEOUT* settings to kill these connections. Does anyone have some suggestions how to solve these issues? Kai -- == Kai M. Becker == kmb(a)kai-m-becker.de == Bremen, Germany == "Any technology distinguishable from magic is insufficiently advanced"

2 2

SSL renegotiation patch
by Janusz Dziemidowicz 19 Sep '12

19 Sep '12

Hi, since I couldn't find a better place I'm sending a simple patch that allows to disable SSL renegotiation here. Possible reasons for this: - famous renegotiation SSL flaw, patched in OpenSSL a long time ago, but not everyone can or want to upgrade OpenSSL - renegotiation makes some DoS attacks much easier (see http://www.thc.org/thc-ssl-dos/) regardless of it being a secure one or not - it is really not needed in many cases The approach is based on what is being done in Apache. The default is to allow renegotation, so there should be no surprises for anyone after upgrade. Patch applies on latest (4.54b4) stunnel beta. Feel free to comment:) -- Janusz Dziemidowicz

3 13

Re: [stunnel-users] Binding to non-local sockets
by Janusz Dziemidowicz 19 Sep '12

19 Sep '12

2012/9/12 Willy Tarreau <w(a)1wt.eu>: > Janusz, > > I just saw your mail on stunnel-users. You don't need to use IP_FREEBIND > to bind to a non-existent address under Linux. You can simply enable > sys.net.ipv4.ip_nonlocal_bind to do this. This is how most users deal > with VRRP setups. It does not allow the backup server to receive the > traffic aimed at the missing address, but it allows the process to bind > even when the address is missing so that it becomes operational upon a > switchover. I am aware of this sysctl. However, this will allows any process on the system to bind to any address. In my case I prefer to enable this on per process basis to avoid any possible configuration errors in any other service running on my machines. Since stunnel already had low level socket options exposed I've simply added IP_FREEBIND to the list;) -- Janusz Dziemidowicz

1 0

Re: [stunnel-users] Trouble with smtp client connection and TLS inWindows
by David Schomaker 13 Sep '12

13 Sep '12

Adding "protocol = smtp" was the solution to my problem. Thank you Leonardo! Regards, David Schomaker ----- Original Message ----- From: "Leandro Avila" <leandro.avila(a)ymail.com> To: "David Schomaker" <david(a)schomaker.net>, "stunnel" <stunnel-users(a)stunnel.org> Date: Thu, 13 Sep 2012 09:26:00 -0700 Subject: Re: [stunnel-users] Trouble with smtp client connection and TLS inWindows David, Perhaps the issue is on the server side. A quick look to the SMTP server on smtp.smarshmail.com will show 220 smtp.smarshmail.com ESMTP Service is ready on Server 1 (EQ). EHLO 250-smtp.smarshmail.com Hello [123.456.790.101] 250-SIZE 36700160 250-PIPELINING 250-DSN 250-ENHANCEDSTATUSCODES 250-STARTTLS 250-AUTH GSSAPI NTLM 250-8BITMIME 250-BINARYMIME 250 CHUNKING QUIT 221 2.0.0 Service closing transmission channel I would try adding the following option to your stunnel configuration protocol = smtp I'm not certain but looks like you need to issue the STARTTLS to negociate the secure connection with this server. Check the stunnel manual for further explanation on the option http://www.stunnel.org/static/stunnel.html Hope this helps ----------------- Leandro Avila ----- Original Message ----- From: David Schomaker <david(a)schomaker.net> To: stunnel <stunnel-users(a)stunnel.org> Cc: Sent: Monday, September 10, 2012 10:32 PM Subject: [stunnel-users] Trouble with smtp client connection and TLS in Windows I need to configure stunnel to connect to smtp via TLS and am not able to get it to work either running 4.53 or 4.54beta. Perhaps I have the conf file created incorrectly. The stunnel.conf is as follows: +++++++++++++++++++++++++++++++++++++++ ; Debugging stuff (may useful for troubleshooting) debug = 7 output = c:\temp\stunnel.log ; Disable FIPS mode to allow non-approved protocols and algorithms ;fips = no ; Disable support for insecure SSLv2 protocol options = NO_SSLv2 [SMARSH-pop3] client = yes accept = 127.0.0.1:110 connect = pop.smarshmail.com:995 [SMARSH-smtp] client = yes SSLversion=TLSv1 accept = 127.0.0.1:25 connect = smtp.smarshmail.com:587 +++++++++++++++++++++++++ Pop works great. The log on an smtp session is as follows: +++++++++++++++++++++++++ 2012.09.10 12:48:31 LOG7[1984:300]: Service [SMARSH-smtp] accepted (FD=508) from 127.0.0.1:49517 2012.09.10 12:48:31 LOG7[1984:300]: Creating a new thread 2012.09.10 12:48:31 LOG7[1984:300]: New thread created 2012.09.10 12:48:31 LOG7[1984:2796]: Service [SMARSH-smtp] started 2012.09.10 12:48:31 LOG5[1984:2796]: Service [SMARSH-smtp] accepted connection from 127.0.0.1:49517 2012.09.10 12:48:31 LOG6[1984:2796]: connect_blocking: connecting 199.47.168.58:587 2012.09.10 12:48:31 LOG7[1984:2796]: connect_blocking: s_poll_wait 199.47.168.58:587: waiting 10 seconds 2012.09.10 12:48:31 LOG5[1984:2796]: connect_blocking: connected 199.47.168.58:587 2012.09.10 12:48:31 LOG5[1984:2796]: Service [SMARSH-smtp] connected remote server from 192.168.108.158:49518 2012.09.10 12:48:31 LOG7[1984:2796]: Remote socket (FD=528) initialized 2012.09.10 12:48:31 LOG7[1984:2796]: SNI: host name: smtp.smarshmail.com 2012.09.10 12:48:31 LOG7[1984:2796]: SSL state (connect): before/connect initialization 2012.09.10 12:48:31 LOG7[1984:2796]: SSL state (connect): SSLv3 write client hello A 2012.09.10 12:48:31 LOG7[1984:2796]: SSL alert (write): fatal: protocol version 2012.09.10 12:48:31 LOG3[1984:2796]: SSL_connect: 1408F10B: error:1408F10B:SSL routines:SSL3_GET_RECORD:wrong version number 2012.09.10 12:48:31 LOG5[1984:2796]: Connection reset: 0 byte(s) sent to SSL, 0 byte(s) sent to socket 2012.09.10 12:48:31 LOG7[1984:2796]: Remote socket (FD=528) closed 2012.09.10 12:48:31 LOG7[1984:2796]: Local socket (FD=508) closed 2012.09.10 12:48:31 LOG7[1984:2796]: Service [SMARSH-smtp] finished (0 left) 2012.09.10 12:48:40 LOG7[1984:300]: Dispatching signals from the signal pipe 2012.09.10 12:48:43 LOG7[1984:300]: Processing SIGNAL_TERMINATE 2012.09.10 12:48:43 LOG5[1984:300]: Terminated ++++++++++++++++++++++++++ Is sTunnel using SSLv3 rather than TLSv1? If so how do I force TLS? Thanks... _______________________________________________ stunnel-users mailing list stunnel-users(a)stunnel.org https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users _______________________________________________ stunnel-users mailing list stunnel-users(a)stunnel.org https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users

1 0

Trouble with smtp client connection and TLS in Windows
by David Schomaker 13 Sep '12

13 Sep '12

I need to configure stunnel to connect to smtp via TLS and am not able to get it to work either running 4.53 or 4.54beta. Perhaps I have the conf file created incorrectly. The stunnel.conf is as follows: +++++++++++++++++++++++++++++++++++++++ ; Debugging stuff (may useful for troubleshooting) debug = 7 output = c:\temp\stunnel.log ; Disable FIPS mode to allow non-approved protocols and algorithms ;fips = no ; Disable support for insecure SSLv2 protocol options = NO_SSLv2 [SMARSH-pop3] client = yes accept = 127.0.0.1:110 connect = pop.smarshmail.com:995 [SMARSH-smtp] client = yes SSLversion=TLSv1 accept = 127.0.0.1:25 connect = smtp.smarshmail.com:587 +++++++++++++++++++++++++ Pop works great. The log on an smtp session is as follows: +++++++++++++++++++++++++ 2012.09.10 12:48:31 LOG7[1984:300]: Service [SMARSH-smtp] accepted (FD=508) from 127.0.0.1:49517 2012.09.10 12:48:31 LOG7[1984:300]: Creating a new thread 2012.09.10 12:48:31 LOG7[1984:300]: New thread created 2012.09.10 12:48:31 LOG7[1984:2796]: Service [SMARSH-smtp] started 2012.09.10 12:48:31 LOG5[1984:2796]: Service [SMARSH-smtp] accepted connection from 127.0.0.1:49517 2012.09.10 12:48:31 LOG6[1984:2796]: connect_blocking: connecting 199.47.168.58:587 2012.09.10 12:48:31 LOG7[1984:2796]: connect_blocking: s_poll_wait 199.47.168.58:587: waiting 10 seconds 2012.09.10 12:48:31 LOG5[1984:2796]: connect_blocking: connected 199.47.168.58:587 2012.09.10 12:48:31 LOG5[1984:2796]: Service [SMARSH-smtp] connected remote server from 192.168.108.158:49518 2012.09.10 12:48:31 LOG7[1984:2796]: Remote socket (FD=528) initialized 2012.09.10 12:48:31 LOG7[1984:2796]: SNI: host name: smtp.smarshmail.com 2012.09.10 12:48:31 LOG7[1984:2796]: SSL state (connect): before/connect initialization 2012.09.10 12:48:31 LOG7[1984:2796]: SSL state (connect): SSLv3 write client hello A 2012.09.10 12:48:31 LOG7[1984:2796]: SSL alert (write): fatal: protocol version 2012.09.10 12:48:31 LOG3[1984:2796]: SSL_connect: 1408F10B: error:1408F10B:SSL routines:SSL3_GET_RECORD:wrong version number 2012.09.10 12:48:31 LOG5[1984:2796]: Connection reset: 0 byte(s) sent to SSL, 0 byte(s) sent to socket 2012.09.10 12:48:31 LOG7[1984:2796]: Remote socket (FD=528) closed 2012.09.10 12:48:31 LOG7[1984:2796]: Local socket (FD=508) closed 2012.09.10 12:48:31 LOG7[1984:2796]: Service [SMARSH-smtp] finished (0 left) 2012.09.10 12:48:40 LOG7[1984:300]: Dispatching signals from the signal pipe 2012.09.10 12:48:43 LOG7[1984:300]: Processing SIGNAL_TERMINATE 2012.09.10 12:48:43 LOG5[1984:300]: Terminated ++++++++++++++++++++++++++ Is sTunnel using SSLv3 rather than TLSv1? If so how do I force TLS? Thanks...

2 1

Stunnel message
by Philip Wong 13 Sep '12

13 Sep '12

Hi all, I'm still learning stunnel for my company support. l feel difficult to understand what's the meaning of some stunnel log message. Is there any doc listed the log messages with some explanation? Thanks, Philip

1 0

stunnel conf alternatives
by John A. Wallace 11 Sep '12

11 Sep '12

I have stunnel 4.53 on x86-pc-mingw32-gnu platform on Win7 OS. I use it minimally at this point but I would like to see whether I can expand my use of it if possible. When I installed it, I simply saved all the program's files into one folder without actually using the Windows installer program, and then I made some entries in the "stunnel.conf" to make it serve my purpose. Right now to start the program I can go to a Windows command prompt in the stunnel folder and type "stunnel stunnel.conf". Then a little icon appears in the notification area of Windows Taskbar. What would happen, I mean how would it differ, if I just typed "stunnel" instead of having to specify the "stunnel.conf" file too? And how would I be able to make use of it in that case? Thanks.

2 1

Binding to non-local sockets
by Janusz Dziemidowicz 09 Sep '12

09 Sep '12

Hi, attached is very simple patch that allows to set IP_FREEBIND under Linux (IP_BINDANY/IPV6_BINDANY under FreeBSD) on an accept socket. I've stumbled upon this need due to the following situation (which I believe might be quite common): - I've already had stunnel instance happily serving one website on port 443, single domain - a need occurred to set up another domain - SNI cannot be used, because Windows XP doesn't support it (and it is still a very large user base) - no problem, I thought, I have a few public IPs available, just set stunnel up on another IP - and here comes the problem, I have two frontend servers with VRRP configured to migrate public IP between them, only one server has the public IP at a time - under normal circumstances it is impossible to bind to an address that does not exist on the server (in case of one domain I've simply bound stunnel to 0.0.0.0) - starting stunnel manually when the address migrates is not an option (VRRP daemon probably can start it, but I was not happy with that solution, I want all services working constantly on both servers so there are no surprises in case of a failover) - setting IP_FREEBIND on accept socket solves the problem, stunnel is working on both machines and I can migrate the public IPs however I want As stunnel already exposes low level socket options, adding support for IP_FREEBIND is very simple. I've also added similar FreeBSD options (tested on FreeBSD 9.0). Oh, and the patch is public domain;) -- Janusz Dziemidowicz

1 0

SSL3_GET_RECORD:wrong version number
by Bob Tentinger 06 Sep '12

06 Sep '12

I have a POS application (SWIFTLINK) that uses Stunnel to send email on Windows XP. This is the error**************** SSL_connect: 1408F10B: error:1408F10B:SSL routines:SSL3_GET_RECORD:wrong version number ***************** I have searched the Internet and none of the resolutions help.

2 1

Race condition in libwrap_auth can hang threads waiting on libwrap process
by Andrew Skalski 05 Sep '12

05 Sep '12

I ran into an issue with stunnel (verified with the latest version) where under certain conditions, client connections will hang in the "Waiting for a libwrap process" state. The problem is a race condition in src/libwrap.c: 170 --num_busy; /* the child process has been released */ 171 if(num_busy==num_processes-1) { /* need to wake up a thread */ 172 retval=pthread_cond_signal(&cond); /* signal waiting threads */ 173 if(retval) { 174 errno=retval; 175 ioerror("pthread_cond_signal"); 176 longjmp(c->err, 1); 177 } 178 } Removing the statement "if (num_busy==num_processes-1)" is one way to fix the problem. The conditions under which the race condition occur are: - All libwrap processes are busy, and multiple threads are waiting [num_busy == num_processes] - One thread releases its libwrap process (this signals one of the waiting threads) - Another thread releases its libwrap process before the signaled thread has a chance to acquire and increment num_busy. Once that happens, all queued threads are stuck waiting on the condition variable, until stunnel is hit with enough concurrent connections to max out num_busy again. (Depending on traffic patterns, this may not happen for a long time.) Andy Skalski

2 1