- stunnel-users - stunnel.org

SSL3_READ_BYTES:tlsv1 alert internal error
by Andrew Heuneman 07 Sep '11

07 Sep '11

Hello, We recently implemented Stunnel on Centos 5.6 for ssl offloading for our java application. The application has applets to communicate java objects over https to a tomcat server on the server side. We have it setup in front of our Alteon/Radware load balancer. This hardware load balancer is capable of ssl load balancing, but has produced a very specific packet reset that only presents itself in ssl processing. We decided to implement Stunnel in front of this load balancer to fix this problem. Ssl offloading was working great with Stunnel until we ran into Java 7. if I run any version of our applets on java 6 they work. If i run java 7 they do not work. I have tried googling and looking for this error but I have only found some references to SNI... is this correct? Is there anything I can do. Please forgive me if i have omitted any details I will be more than happy to include a packet capture or other details if needed. I compiled stunnel with the following options ./configure --disable-libwrap --bindir=/usr/sbin --sbindir=/usr/sbin --sysconfdir=/etc --with-ssl=/usr/local/ssl also i compiled OpenSSL 1.0.0d with the following ./Configure threads shared linux-generic64 stunnel -version No limit detected for the number of clients signal_pipe: FD=3 allocated (non-blocking mode) signal_pipe: FD=4 allocated (non-blocking mode) stunnel 4.42 on x86_64-unknown-linux-gnu platform Compiled/running with OpenSSL 1.0.0d 8 Feb 2011 Threading:PTHREAD SSL:ENGINE Auth:none Sockets:POLL,IPv6 stunnel 4.42 on x86_64-unknown-linux-gnu platform Compiled/running with OpenSSL 1.0.0d 8 Feb 2011 Threading:PTHREAD SSL:ENGINE Auth:none Sockets:POLL,IPv6 Global option defaults debug = daemon.notice pid = /usr/local/var/run/stunnel/stunnel.pid RNDbytes = 64 RNDfile = /dev/urandom RNDoverwrite = yes Service-level option defaults ciphers = ALL:!SSLv2:!aNULL:!EXP:!LOW:-MEDIUM:RC4:+HIGH curve = prime256v1 session = 300 seconds sslVersion = TLSv1 for client, all for server stack = 65536 bytes TIMEOUTbusy = 300 seconds TIMEOUTclose = 60 seconds TIMEOUTconnect = 10 seconds TIMEOUTidle = 43200 seconds verify = none str_stats: 112 block(s), 4046 byte(s) stunnel.conf cert=/etc/stunnel/stunnel.pem debug=7 output=/var/log/stunnel.log socket=l:TCP_NODELAY=1 socket=r:TCP_NODELAY=1 [https] accept=0.0.0.0:443 connect=172.16.18.100:80 session = 300 TIMEOUTbusy = 300 TIMEOUTconnect = 10 TIMEOUTidle = 43200 client = no stunnel.log 2011.08.22 16:58:49 LOG7[438154:47689394220768]: Service https accepted FD=2 from 10.0.11.27:46830 2011.08.22 16:58:49 LOG7[438154:1104877888]: Service https started 2011.08.22 16:58:49 LOG7[438154:1104877888]: Option TCP_NODELAY set on local socket 2011.08.22 16:58:49 LOG5[438154:1104877888]: Service https accepted connection from 10.0.11.27:46830 2011.08.22 16:58:49 LOG7[438154:1104877888]: SSL state (accept): before/accept initialization 2011.08.22 16:58:49 LOG7[438154:1104877888]: SSL state (accept): SSLv3 read client hello A 2011.08.22 16:58:49 LOG7[438154:1104877888]: SSL state (accept): SSLv3 write server hello A 2011.08.22 16:58:49 LOG7[438154:1104877888]: SSL state (accept): SSLv3 write certificate A 2011.08.22 16:58:49 LOG7[438154:1104877888]: SSL state (accept): SSLv3 write key exchange A 2011.08.22 16:58:49 LOG7[438154:1104877888]: SSL state (accept): SSLv3 write certificate request A 2011.08.22 16:58:49 LOG7[438154:1104877888]: SSL state (accept): SSLv3 flush data 2011.08.22 16:58:49 LOG7[438154:1104877888]: SSL alert (read): fatal: internal error 2011.08.22 16:58:49 LOG3[438154:1104877888]: SSL_accept: 14094438: error:14094438:SSL routines:SSL3_READ_BYTES:tlsv1 alert internal error 2011.08.22 16:58:49 LOG5[438154:1104877888]: Connection reset: 0 bytes sent to SSL, 0 bytes sent to socket 2011.08.22 16:58:49 LOG7[438154:1104877888]: Service https finished (1 left) 2011.08.22 16:58:49 LOG7[438154:1104877888]: str_stats: 0 block(s), 0 byte(s) 2011.08.22 16:59:01 LOG7[438154:1104947520]: Socket closed on read 2011.08.22 16:59:01 LOG7[438154:1104947520]: Sending SSL write shutdown 2011.08.22 16:59:01 LOG7[438154:1104947520]: SSL alert (write): warning: close notify 2011.08.22 16:59:01 LOG6[438154:1104947520]: SSL_shutdown successfully sent close_notify 2011.08.22 16:59:01 LOG7[438154:1104947520]: SSL alert (read): warning: close notify 2011.08.22 16:59:01 LOG7[438154:1104947520]: SSL closed on SSL_read 2011.08.22 16:59:01 LOG7[438154:1104947520]: Sending socket write shutdown 2011.08.22 16:59:01 LOG5[438154:1104947520]: Connection closed: 49445 bytes sent to SSL, 8175 bytes sent to socket -- Thank You, Andrew Heuneman Senior Systems Administrator Reading Plus®/Taylor Associates Helping students become proficient silent readers. <http://twitter.com/readingplus> <http://www.facebook.com/pages/Reading-Plus/165970877038>

2 1

Internal error after reload modified server mode section
by Javier 07 Sep '11

07 Sep '11

Hi, I didn't found this question, so I hope isn't duplicate. I'm using the latest stable 4.42 version and I'm running Windows 2000 and I want to ask is this is a normal behavior. What is happening is that, after change the server section in the config file (the rest are all client mode), if it has been any connection to that server (and only if it has been a connection), when I reload the modified conf, it gives...: INTERNAL ERROR: get_alloc_list_ptr: Bad magic str_stats: 380 block(s), 96684 byte(s) This happens after no errors in the load (except for DH being hardcoded), SSL context initialized Configuration successful after open every service (clients and server) and then, after the error, closes every service. Trying to reload the conf, doesn't nothing, just stunnel hangs (is not responding to reloads). The only way to restart all is to close stunnel and reopen. Then, the configuration is accepted without problems and works perfectly. This happens, as I said, after modify the conf file and if it was a connection. The section is related to a local test server, where I modify the IP and/or ports where to accept and where to connect. As general options I have: debug = debug delay = yes And the server section: [mySSL] client = no cert = mytestpem.pem accept = 127.0.0.1:443 connect = 127.0.0.1:80 What I'm doing wrong? Is the normal behavior? Is it a new bug? Regards, Javier.

3 5

Compiling Stunnel For Windows
by Albert 05 Sep '11

05 Sep '11

Hi, I want to compile stunnel for windows, to include the XMPP patch (that I managed to apply cleanly to the current stunnel code) that I saw on the mailing list earlier as I need to get it to connect to a SSL only server in client mode. I followed the instructions in the Install.w32 file 1) apt-get install gcc-mingw32 - Worked fine with no issues 2) Download the recent zlib from http://www.zlib.net/ , Update the following definitions in win32/Makefile.gcc file: etc *snip* Worked fine as far as I can tell 3) Download the recent OpenSSL in unpack it to /usr/src/ directory., cd /usr/src && tar zvxf ~/openssl-(version).tar.gz I got OpenSSL 1.0.0d 4) Build OpenSSL with cross_mingw32.sh script. I ran instead ./Configure --cross-compile-prefix=i586-mingw32msvc- mingw shared zlib-dynamic && make it seems to have worked fine 5) Download and unpack stunnel-(version).tar.gz. -got stunnel 4.42 6) Configure stunnel. cd stunnel-(version) && ./configure --with-ssl=/path/to/openssl-(version) This step worked fine 7) Build windows executable. cd src && make stunnel.exe The first few compile steps work fine, but I get stuck on: ...i586-mingw32msvc/bin/ld: cannot find -lcrypto.dll I did a few hours of troubleshooting and trying to move a few files around but couldn't find crypto.dll anywhere I'm running a fresh install of Ubuntu 11.04 in a VM if that makes any difference. Help would be greatly appreciated :) Thanks, Albert

2 1

Changing the secure tunnel 'encryption' method ....
by VANDOOREN, Pascal 02 Sep '11

02 Sep '11

Hello Everyone, I have a question, Why use you a certificate and not a hash AES, or other crypting algorithm to make a secure tunnel ? Because i use not a validate cert and i can only generate a cert valid for maximum 1year ... In the config file you can let the choice to use the cert or hash .. It's only a idea Kind regards Pascal -- This e-mail is confidential. If you are not the addressee or an authorized recipient of this message, any distribution, copying, publication or use of this information for any purpose is prohibited. Please notify the sender immediately by e-mail and then delete this message. Ce message est confidentiel. Si vous n'etes pas le destinataire designe de ce message ou une personne autorisee a l'utiliser, toute distribution, copie, publication ou usage a quelques fins que ce soit des informations contenues dans ce message sont interdits. Merci d'informer immediatement l'expediteur par messagerie electronique et d'ensuite detruire ce message. --

1 0

Windows version not starting when Run
by Dayo Akande 31 Aug '11

31 Aug '11

The main thing that I'm seeing as the error is: "Could not load DH parameters from stunnel.pem" Below is the entire log file, I am so stumped here. I appreciate any help. No limit detected for the number of clients make_sockets: s_socket#1: FD=272 allocated (blocking mode) make_sockets: s_socket#2: FD=276 allocated (blocking mode) make_sockets: s_accept: FD=292 allocated (non-blocking mode) stunnel 4.42 on x86-pc-mingw32-gnu platform Compiled/running with OpenSSL 1.0.0d 8 Feb 2011 Threading:WIN32 SSL:ENGINE Auth:none Sockets:SELECT,IPv6 Reading configuration from file stunnel.conf Snagged 64 random bytes from C:/.rnd Wrote 1024 new random bytes to C:/.rnd PRNG seeded successfully Initializing SSL context for service pop3s Certificate: stunnel.pem Certificate loaded Key file: stunnel.pem Private key loaded Could not load DH parameters from stunnel.pem Using hardcoded DH parameters DH initialized with 2048-bit key ECDH initialized with curve prime256v1 SSL options set: 0x01180004 SSL context initialized Initializing SSL context for service imaps Certificate: stunnel.pem Certificate loaded Key file: stunnel.pem Private key loaded Could not load DH parameters from stunnel.pem Using hardcoded DH parameters DH initialized with 2048-bit key ECDH initialized with curve prime256v1 SSL options set: 0x01180004 SSL context initialized Line 91: "socket = l:TCP_NODELAY=1 ": Specified option name is not valid here str_stats: 152 block(s), 11942 byte(s) Server is down

2 2

Re: [stunnel-users] CRL
by Uldis Biks 31 Aug '11

31 Aug '11

1 0

CRL
by Uldis Biks 31 Aug '11

31 Aug '11

2 2

stunnel : binding in ipv6+ipv4 : problem
by Guylhem Aznar 27 Aug '11

27 Aug '11

Hello I have a problem with stunnel : I just don't know how to specify 2 different port forwarding, one for all ipv4 connections to port 443, the other for all ipv6 connections. Either one seem to work, but I can't have both at the same time. In my /etc/hosts : 127.0.0.1 localhost ::1 localhost6 0.0.0.0 listenall4 :: listenall6 In my stunnel.conf: [sslh] accept = listenall4:443 connect = 127.0.0.1:444 TIMEOUTclose = 0 [ssl6] accept = listenall6:443 connect = 6.guylhem.org:80 TIMEOUTclose = 0 I have read that linux can block ipv6 sockets if the ipv4 socket is created first, but here the order doesn't change anything. I though that stunnel was ipv6 capable, but apparently it doesn't play well with ipv4 on the same port. So I tried to remove one of the listenall : the listenall4 is important, but listenall6 is not. Yet giving a AAAA only host like 6.guylhem.org does not help - even if it goes before the listenall4. So I'm stuck. Has anyone patched stunnel for such a situation? Guylhem

1 0

FW: Mysql doesnt run with stunnel
by Freeh Sophia 26 Aug '11

26 Aug '11

Craig Retief <stunnel <at> rsw.co.za> writes: > > I see that on your server you have specified your config for mysqls like > this: > Hi Craig, That fixed it for me. Thanks! Freeh Sophia Marketing GmbH Emanuelstr. 3, 10317 Berlin Deutschland Telefon: +49 (33) 5310967 Email: freehsophia(a)gmail.com Site: http://flug.airego.de/

1 0

stunnel4 tcp_nodelay
by Madhu R 21 Aug '11

21 Aug '11

Hi, We are using stunnel in order to tunnel http proxy traffic. Is it better in this case to turn on Nagle's algo? I have tried to set socket=l:TCP_NODELAY=0 and socket=r:TCP_NODELAY=0 in the global config, but when I look at debug log it keeps saying option tcp_nodelay is set on local and remote socket. Is this a bug, or is there a different way of turning off tcp_nodelay (since I understand this is on by default now) Regds

1 0