stunnel-users

stunnel-users@stunnel.org

5 participants
3293 discussions

stunnel-4.11 patch for multiple certificates
by Nick Tolomiczenko 14 Sep '05

14 Sep '05

Hi stunnel users, I placed a patch to stunnel-4.11 that will enable it to handle multiple certificates. It is currently awaiting moderator approval. For those who can't wait, please contact myself, Nick Tolomiczenko, or Shem Ali below. Here's the description of what it does: This patch addresses the need for service-level ssl contexts. In particular, the following options which were only available as global options may now be used at the service level: cert, key, CApath, CAfile, ciphers, CRLpath, CRLfile, options, client and verify. If any of these options are used in a service section, stunnel will override its global setting -- if it exists -- and initialize a separate ssl context for the service. On the other hand, for those services that do not specify any of these options, stunnel will fall back on what was specified in the global section of the configuration file and intialize a "common" ssl context for those services. Note: If all services each specified at least one the the above ssl options, then the no "common" ssl context will be initialized since each service will have its own. In this case, it's still a good idea to put common default ssl options in the global section as each service will inherit these settings even if the service initializes its own ssl context. Brought to you by: Nick Tolomiczenko: nick(a)renderquest.com or nick(a)neikos.com Shem Ali: shem(a)renderquest.com

1 0

stunnel-4.11 patch for multiple certificates
by Nick Tolomiczenko 14 Sep '05

14 Sep '05

This patch addresses the need for service-level ssl contexts. In particular, the following options which were only available as global options may now be used at the service level: cert, key, CApath, CAfile, ciphers, CRLpath, CRLfile, options, client and verify. If any of these options are used in a service section, stunnel will override its global setting -- if it exists -- and initialize a separate ssl context for the service. On the other hand, for those services that do not specify any of these options, stunnel will fall back on what was specified in the global section of the configuration file and intialize a "common" ssl context for those services. Note: If all services each specified at least one the the above ssl options, then the no "common" ssl context will be initialized since each service will have its own. In this case, it's still a good idea to put common default ssl options in the global section as each service will inherit these settings even if the service initializes its own ssl context. Brought to you by: Nick Tolomiczenko: nick(a)renderquest.com or nick(a)neikos.com Shem Ali: shem(a)renderquest.com

1 0

Access to a daemon that has SSL abilities
by Amiram Oren 14 Sep '05

14 Sep '05

Hi. Can I use stunnel only on a client side, in case on the other side, the server already supports SSL? Will this work for SPOP3 and SMTP\S? 10X! Oren Amiram

2 1

stunnel-4.11 stable?
by Serge Marigold 13 Sep '05

13 Sep '05

Hi, Is anyone using stunnel-4.11 in a heavily trafficed production environment?

1 0

4.10 stable?
by conover＠rahul.net 09 Sep '05

09 Sep '05

Stunnel 4.10 has gzip compression, right? Is it stable? Thanks, John -- John Conover, conover(a)rahul.net, http://www.johncon.com/

1 0

[BUG] Stunnel 4.10 + xinetd : Connection reset by peer
by Eric Piel 02 Sep '05

02 Sep '05

Hello, I've upgraded to stunnel 4.10 yesterday and since this time I can't managed to get stunnel working with xinetd. I try to access a CVS server with ssl encryption. If stunnel is started as stand-alone, it works as before, but when it's started by xinetd all I get is: cvs [update aborted]: reading from server: Connection reset by peer I've attached the connection logs with 4.09 and 4.10 as well as my configuration file (which used to work for more than a year). My xinetd is version 2.3.13 (and I haven't changed it for several months). Please, let me know if I'm overlooking at something or if I can try anything to debug more in depth... Thanks in advance, Eric PS: actually the CVS is accesible from outside, so you might want to try yourself! # Sample stunnel configuration file # Copyright by Michal Trojnara 2002 setuid = root setgid = root # Authentication stuff verify = 2 CAfile = /etc/ssl/stunnel/lifl-cvs.pem # Use it for client mode client = yes #debug = 7 #output = /home/eric/busy/stunnel.log connect = cvs.lifl.fr:2405

4 5

Re: Flyspray task #7 - CRLfile is never refreshed
by Kartik Subbarao 27 Aug '05

27 Aug '05

I wanted to follow up on this. Flyspray did not let me add a comment since the bug is closed, so I'm posting to the mailing list. 1. I tried using CRLpath (and setting up the symbolic link properly to the hashed value), but that had the same behavior as CRLfile. It only reads the actual CRL once. If I update the file on disk, it never re-reads the file. So I don't see how this is a workaround. Am I missing something? 2. As I mentioned in my initial report, it is a significant disruption for existing stunnel processes to be killed. If it were possible just to kill the main daemon without also killing the child processes (i.e. like sshd), that could be a workaround. But currently, killing the main stunnel process also kills all child stunnel processes. I guess I don't understand what else one could do in this situation, apart from having the application properly reread the CRL on a regular basis. (OCSP is not yet available). When you say that "that's how openssl works", are you saying that the API does not allow you to reread/reload the CRL file? That seems rather odd. It seems like there should be some kind of way to refresh the OpenSSL data structures that store the CRL. Any thoughts/suggestions would be appreciated. Thanks, -Kartik Michal.Trojnara(a)mirt.net wrote: > Notice from stunnel > > Michal Trojnara (mtrojnar) has closed the following task. You are > receiving this because you are on the notification list. > > Task #7: CRLfile is never refreshed > The reason for closing is: Not a bug > That's how OpenSSL work. There are two workarounds: > 1. Use CRLpath instead of CRLfile (I recommend it). > 2. Restart stunnel after the CRL file modification. > > You can get more information about this task at the following URL: > http://stunnel.mirt.net/flyspray/index.php?do=details&id=7 > >

1 0

[stunnel patch] updated transparent proxy on linux 2.6 using cttproxy patch
by MAtteo HCE Valsasna 26 Aug '05

26 Aug '05

hello folks, here comes an improved version of the transparent proxy patch, basically solving the previous TODO issues. please try it if you are interested in transparent stunnel. any feedback il welcome. the code is heavly based on the examples distributed along with the cttproxy patch, many thanks to Balzs Scheidler for its great work and to Lennert Buytenhek for the skaidrus example program. this code is (re)distributed under the GPL version 2 MAtteo ---README--- cttproxy patch to stunnel MAtteo HCE Valsasna (matteo dot valsasna at uninsubria dot it) based on example code in cttproxy patch ports stunnel transparent proxy to cttproxy kernel patch allow stunnel server to connect to the application server faking the stunnel client's IP source address. the server sees connections coming from the stunnel client's IP address instead of that of the stunnel server, and thus can apply admission control, bandwidth control, logging policies as if the client connected directly. should work 2.4 and 2.6 kernels and 2.0.x cttproxy patch tested with kernel 2.6.12.5 and cttproxy-2.6.12-2.0.2 patch setup: * application server(s): here run the application(s) you want to access trough stunnel must route all traffic, or at least all traffic for transproxyed connections, trough the stunnel server * stunnel server: must not be the same host as the application server grab recent kernel source patch your kernel source with the appropriate cttproxy patch (from http://www.balabit.com/products/oss/tproxy/) configure kernel with: CONFIG_IP_NF_TPROXY=m CONFIG_IP_NF_MATCH_TPROXY=m CONFIG_IP_NF_TARGET_TPROXY=m make and install kernel reboot and choose patched kernel load tproxy kernel modules # modprobe iptable_tproxy # modprobe ipt_tproxy enable ip forwarding (this must act as a router) # echo 1 > /proc/sys/net/ipv4/ip_forward grab stunnel source and extract it $ tar xzf stunnel-4.11.tar.gz patch it $ patch -p0 < cttproxy.patch configure selecting tproxy support $ ./configure --enable-tproxy make and install $ make # make install run stunnel in server mode with transparent = yes remote = <application server>:<port> in conf file * stunnel client run plain stunnel, client = yes remote = stunnel server * application client may or may not be the same host as the stunnel client connect to stunnel client to talk securely to application server ---PATCH---diff -crb /tmp/stunnel-4.11/configure stunnel-4.11/configure *** /tmp/stunnel-4.11/configure 2005-07-09 23:00:44.000000000 +0200 --- stunnel-4.11/configure 2005-08-25 16:59:57.000000000 +0200 *************** *** 463,469 **** # include <unistd.h> #endif" ! ac_subst_vars='SHELL PATH_SEPARATOR PACKAGE_NAME PACKAGE_TARNAME PACKAGE_VERSION PACKAGE_STRING PACKAGE_BUGREPORT exec_prefix prefix program_transform_name bindir sbindir libexecdir datadir sysconfdir sharedstatedir localstatedir libdir includedir oldincludedir infodir mandir build_alias host_alias target_alias DEFS ECHO_C ECHO_N ECHO_T LIBS INSTALL_PROGRAM INSTALL_SCRIPT INSTALL_DATA CYGPATH_W PACKAGE VERSION ACLOCAL AUTOCONF AUTOMAKE AUTOHEADER MAKEINFO install_sh STRIP ac_ct_STRIP INSTALL_STRIP_PROGRAM mkdir_p AWK SET_MAKE am__leading_dot AMTAR am__tar am__untar build build_cpu build_vendor build_os host host_cpu host_vendor host_os CC CFLAGS LDFLAGS CPPFLAGS ac_ct_CC EXEEXT OBJEXT DEPDIR am__include am__quote AMDEP_TRUE AMDEP_FALSE AMDEPBACKSLASH CCDEPMODE am__fastdepCC_TRUE am__fastdepCC_FALSE EGREP LN_S ECHO AR ac_ct_AR RANLIB ac_ct_RANLIB CPP CXX CXXFLAGS ac_ct_CXX CXXDEPMODE am__fastdepCXX_TRUE am__fastdepCXX_FALSE CXXCPP F77 FFLAGS ac_ct_F77 LIBTOOL ssldir RANDOM_FILE USE_DH USE_IPv6 LIBOBJS LTLIBOBJS' ac_subst_files='' # Initialize some variables set by options. --- 463,469 ---- # include <unistd.h> #endif" ! ac_subst_vars='SHELL PATH_SEPARATOR PACKAGE_NAME PACKAGE_TARNAME PACKAGE_VERSION PACKAGE_STRING PACKAGE_BUGREPORT exec_prefix prefix program_transform_name bindir sbindir libexecdir datadir sysconfdir sharedstatedir localstatedir libdir includedir oldincludedir infodir mandir build_alias host_alias target_alias DEFS ECHO_C ECHO_N ECHO_T LIBS INSTALL_PROGRAM INSTALL_SCRIPT INSTALL_DATA CYGPATH_W PACKAGE VERSION ACLOCAL AUTOCONF AUTOMAKE AUTOHEADER MAKEINFO install_sh STRIP ac_ct_STRIP INSTALL_STRIP_PROGRAM mkdir_p AWK SET_MAKE am__leading_dot AMTAR am__tar am__untar build build_cpu build_vendor build_os host host_cpu host_vendor host_os CC CFLAGS LDFLAGS CPPFLAGS ac_ct_CC EXEEXT OBJEXT DEPDIR am__include am__quote AMDEP_TRUE AMDEP_FALSE AMDEPBACKSLASH CCDEPMODE am__fastdepCC_TRUE am__fastdepCC_FALSE CPP LN_S RANLIB ac_ct_RANLIB EGREP ECHO AR ac_ct_AR CXX CXXFLAGS ac_ct_CXX CXXDEPMODE am__fastdepCXX_TRUE am__fastdepCXX_FALSE CXXCPP F77 FFLAGS ac_ct_F77 LIBTOOL ssldir RANDOM_FILE USE_DH USE_IPv6 USE_TPROXY LIBOBJS LTLIBOBJS' ac_subst_files='' # Initialize some variables set by options. *************** *** 1037,1042 **** --- 1037,1043 ---- --disable-rsa Disable RSA support --enable-dh Enable DH support --enable-ipv6 Enable IPv6 support + --enable-tproxy Enable linux 2.6 tproxy patch support --disable-libwrap Disable TCP wrappers library support Optional Packages: *************** *** 23166,23171 **** --- 23344,23367 ---- fi; + # use linux 2.[46] tproxy patch + echo "$as_me:$LINENO: checking whether to enable transparent proxyng using linux 2.46 tproxy patch" >&5 + echo $ECHO_N "checking whether to enable transparent proxyng using linux 2.46 tproxy patch... $ECHO_C" >&6 + # Check whether --enable-tproxy or --disable-tproxy was given. + if test "${enable_tproxy+set}" = set; then + enableval="$enable_tproxy" + echo "$as_me:$LINENO: result: yes" >&5 + echo "${ECHO_T}yes" >&6; cat >>confdefs.h <<\_ACEOF + #define USE_TPROXY 1 + _ACEOF + + else + echo "$as_me:$LINENO: result: no" >&5 + echo "${ECHO_T}no" >&6 + + fi; + + # Disable use of libwrap (TCP wrappers) # it should be the last check! echo "$as_me:$LINENO: checking whether to disable TCP wrappers library support" >&5 *************** *** 23987,23992 **** --- 24184,24190 ---- s,@RANDOM_FILE@,$RANDOM_FILE,;t t s,@USE_DH@,$USE_DH,;t t s,@USE_IPv6@,$USE_IPv6,;t t + s,@USE_TPROXY@,$USE_TPROXY,;t t s,@LIBOBJS@,$LIBOBJS,;t t s,@LTLIBOBJS@,$LTLIBOBJS,;t t CEOF diff -crb /tmp/stunnel-4.11/configure.ac stunnel-4.11/configure.ac *** /tmp/stunnel-4.11/configure.ac 2005-07-09 22:58:05.000000000 +0200 --- stunnel-4.11/configure.ac 2005-08-25 16:56:33.000000000 +0200 *************** *** 14,21 **** --- 14,24 ---- if test "$GCC" = "yes" then CFLAGS="$CFLAGS -Wall -Wshadow -Wcast-align -Wpointer-arith" fi + AC_PROG_CPP AC_PROG_INSTALL + AC_PROG_LN_S AC_PROG_MAKE_SET + AC_PROG_RANLIB # Checks for typedefs, structures, and compiler characteristics # AC_C_CONST *************** *** 254,259 **** --- 257,271 ---- ) AC_SUBST(USE_IPv6) + # use linux 2.[46] tproxy patch + AC_MSG_CHECKING([whether to enable transparent proxyng using linux 2.[46] tproxy patch]) + AC_ARG_ENABLE(tproxy, + [ --enable-tproxy Enable linux 2.6 tproxy patch support], + [AC_MSG_RESULT([yes]); AC_DEFINE(USE_TPROXY)], + [AC_MSG_RESULT([no])] + ) + AC_SUBST(USE_TPROXY) + # Disable use of libwrap (TCP wrappers) # it should be the last check! AC_MSG_CHECKING([whether to disable TCP wrappers library support]) diff -crb /tmp/stunnel-4.11/src/client.c stunnel-4.11/src/client.c *** /tmp/stunnel-4.11/src/client.c 2005-07-02 08:55:58.000000000 +0200 --- stunnel-4.11/src/client.c 2005-08-26 16:01:34.000000000 +0200 *************** *** 44,49 **** --- 44,65 ---- #define SHUT_RDWR 2 #endif + #ifdef USE_TPROXY + #define LINUX_NETFILTER + #define LINUX_TPROXY + #define TPROXY_MAJOR_VERSION(x) ((x >> 24) & 0xff) + #define TPROXY_MINOR_VERSION(x) ((x >> 16) & 0xff) + #define TPROXY_PATCH_VERSION(x) (x & 0xffff) + #endif + + /* transparent proxy */ + #ifdef LINUX_NETFILTER + #include <linux/netfilter_ipv4.h> + #endif + #ifdef LINUX_TPROXY + #include <linux/netfilter_ipv4/ip_tproxy.h> + #endif + /* TCP wrapper */ #ifdef USE_LIBWRAP #include <tcpd.h> *************** *** 77,82 **** --- 93,103 ---- static int connect_wait(CLI *, int, int); static void reset(int, char *); + #ifdef USE_TPROXY + static int determine_source_address(struct sockaddr_in *destination, + struct sockaddr_in *source); + #endif + int max_clients; #ifndef USE_WIN32 int max_fds; *************** *** 921,926 **** --- 942,954 ---- int s; /* destination socket */ u16 i; + // struct in_addr *local=NULL; + #ifdef USE_TPROXY + int f; + struct in_tproxy itp; + struct sockaddr_in tproxy_sin; + #endif + /* setup address_list */ if(c->opt->option.delayed_lookup) { resolved_list.num=0; *************** *** 945,959 **** if(alloc_fd(s)) return -1; ! if(c->bind_addr.num) { /* explicit local bind or transparent proxy */ memcpy(&bind_addr, &c->bind_addr.addr[0], sizeof(SOCKADDR_UNION)); if(bind(s, &bind_addr.sa, addr_len(bind_addr))<0) { sockerror("bind transparent"); closesocket(s); return -1; } } /* try to connect for the 1st time */ s_ntop(c->connecting_address, &addr); s_log(LOG_DEBUG, "%s connecting %s", --- 973,1066 ---- if(alloc_fd(s)) return -1; ! if(c->bind_addr.num ! #ifdef USE_TPROXY ! /* if we're using the tproxy patch, transparent proxy is ! handled below */ ! && !c->opt->option.transparent ! #endif ! ) { ! /* explicit local bind, not tproxy transparent mode */ ! memcpy(&bind_addr, &c->bind_addr.addr[0], sizeof(SOCKADDR_UNION)); if(bind(s, &bind_addr.sa, addr_len(bind_addr))<0) { sockerror("bind transparent"); closesocket(s); return -1; } + sockerror("local bind succeeded"); } + #ifdef USE_TPROXY + if (c->opt->option.transparent){ + /* transparent proxy support on 2.[46] linux kernels */ + /* added by MAtteo HCE Valsasna (matteo.valsasna(a)uninsubria.it), 20050826 */ + /* requires tproxy kernel patch (http://www.balabit.com/products/oss/tproxy */ + /* based on examples in cttproxy patch */ + /* this patch is released under the GNU General Public Licence, Version 2 */ + + memcpy(&bind_addr, &c->bind_addr.addr[0], sizeof(SOCKADDR_UNION)); + + /* check tproxy version */ + itp.op = TPROXY_VERSION; + itp.v.version = 0x00000000; + + getsockopt(s, SOL_IP, IP_TPROXY, &itp, &f); + s_log(LOG_NOTICE, "TPROXY_VERSION: %d.%d.%d", + TPROXY_MAJOR_VERSION(itp.v.version), + TPROXY_MINOR_VERSION(itp.v.version), + TPROXY_PATCH_VERSION(itp.v.version)); + + /* conservative: accepting only tproxy version 2.0.x */ + if (TPROXY_MAJOR_VERSION(itp.v.version) != 2 || + TPROXY_MINOR_VERSION(itp.v.version) != 0) + { + s_log(LOG_ERR, "unsupported tproxy version: %x (2.0.x required)", itp.v.version); + return -1; + } + + tproxy_sin.sin_family = AF_INET; + if (determine_source_address(&addr.in, &tproxy_sin) < 0){ + sockerror("determine_source_address"); + return -1; + } + + /* do not request a specific source port */ + tproxy_sin.sin_port = htons(0); + + /* bind to local source address */ + /* needed for tproxy magic */ + if (bind(s, (struct sockaddr *) &tproxy_sin, sizeof(tproxy_sin)) == -1) + { + sockerror("bind"); + return -1; + } + s_log(LOG_NOTICE, "bound to local: %s", inet_ntoa(tproxy_sin.sin_addr)); + + /* assign foreign address */ + itp.op = TPROXY_ASSIGN; + itp.v.addr.faddr = bind_addr.in.sin_addr; + itp.v.addr.fport = 0; + + if (setsockopt(s, SOL_IP, IP_TPROXY, &itp, sizeof(itp)) == -1) + { + s_log(LOG_NOTICE, "error assigning foreign address: %s", inet_ntoa(bind_addr.in.sin_addr)); + sockerror("setsockopt(SOL_IP, IP_TPROXY, TPROXY_ASSIGN)"); + return -1; + } + s_log(LOG_NOTICE, "assigned foreign address: %s", inet_ntoa(bind_addr.in.sin_addr)); + + /* set connect flag on socket */ + itp.op = TPROXY_FLAGS; + itp.v.flags = ITP_CONNECT|ITP_ONCE; + if (setsockopt(s, SOL_IP, IP_TPROXY, &itp, sizeof(itp)) == -1) + { + sockerror("setsockopt(SOL_IP, IP_TPROXY, TPROXY_FLAGS)"); + return -1; + } + } + #endif + /* try to connect for the 1st time */ s_ntop(c->connecting_address, &addr); s_log(LOG_DEBUG, "%s connecting %s", *************** *** 1020,1023 **** --- 1127,1175 ---- log_error(LOG_DEBUG, get_last_socket_error(), txt); } + + #ifdef USE_TPROXY + + /* from skaidrus */ + static int determine_source_address(struct sockaddr_in *destination, + struct sockaddr_in *source) + { + struct sockaddr_in addr; + socklen_t addrlen; + char textaddr[32]; + int sock; + + sock = socket(AF_INET, SOCK_DGRAM, 0); + if (sock < 0) { + syslog(LOG_CRIT, "determine_source_address: error " + "%d while creating socket", errno); + abort(); + } + + addr = *destination; + addr.sin_port = htons(80); + if (connect(sock, (struct sockaddr *)&addr, sizeof(addr)) < 0) { + syslog(LOG_ALERT, "determine_source_address: error %d " + "while looking up route for %d", errno, + inet_aton(textaddr, &(addr.sin_addr))); + close(sock); + return -1; + } + + addrlen = sizeof(*source); + if (getsockname(sock, (struct sockaddr *)source, &addrlen) < 0) { + syslog(LOG_CRIT, "determine_source_address: error %d " + "while retrieving source address", errno); + abort(); + } + source->sin_port = 0; + + close(sock); + + return 0; + } + + #endif + + /* End of client.c */

1 0

RE: stunnel and OpenSSL v0.9.8 - tunneling the UDP. Possible?
by rz1a＠mail.ru 22 Aug '05

22 Aug '05

Hello Michal, Just one thought: - if the stunnel on LAN1 joins a group to receive a multicasted stream does it really depend on DTLS's [in]ability to handle the multicasting? I see it this way - on LAN1 a machine running the (first instance of) stunnel joins a group, i.e. stunnel is told to open a socket with all the relevant socket options. Then, another machine on LAN2 (behind NAT + firewall + internet + firewall + NAT) connects to the first stunnel and tells it's own instance of stunnel to open a socket with all the relevant socket options to emit the (apparently) multicasted stream. Please comment. -- Best regards, Tony

2 2

[stunnel patch] transparent proxy on linux 2.6 using cttproxy patch
by MAtteo HCE Valsasna 22 Aug '05

22 Aug '05

hi folks: as suggested in http://www.stunnel.org/faq/transparent.html#ToC3, I modified the stunnel code to use the transparent proxy funcionality on 2.6 (and possibly 2.4) linux kernels with the cttproxy patch (http://www.balabit.com/products/oss/tproxy/, I used )cttproxy-2.6.12-2.0.2. on a clean 2.6.12.5 kernel) the patch is higly incomplete, but it works (see TODO). please try it if you are interested in transparent stunnel. any feedback il welcome. MAtteo the code is heavly based on the examples distributed along with the cttproxy patch, many thanks to Balázs Scheidler for its great work and for making it GPL. with this patch, I was able to run a stunnel in server mode with the transparent option, and the application server I connected to saw the IP address of the stunnel client. of course, this code il also redistributed under the GPL. conditions for running stunnel in transparent mode: if you want to try this patch, you MUST modify the hardcoded IP address in client.c to the IP address of the stunnel server (look for DOIT! in the source code). any improvements wellcome. the stunnel server must run a linux kernel patched with the cttproxy patch the stunnel server and the application server cannot be the same host the application server must believe that it has to route all packets of transparently tunneled connections via the stunnel server. thus the two server should be on the same subnet. I was also able to use linux policy routing to avoid modifying the general routing on the application server: transparently tunneled connections are directed to a different IP on the application server, where a policy rule applies a specific routing table to packets with that source address (see at the end of this message). TODO: stunnel must explicitly bind to an *existing* local address (not loopback) before calling setsockopt(SOL_IP, IP_TPROXY, TPROXY_ASSIGN)"). now the address is hardcoded in the patch, it should automatically find a local address to bind to (or get it from the config file) now the code will only run with tproxy patch version 2.0.2 it should be less strict about tproxy version (make it a configurable option in stunnel?). which tproxy version will it run with? auto detect or configure kernel version (2.2 vs 2.4 or 2.6 with cttproxy), or better add it in .configure ----begin patch----- *** /tmp/stunnel-4.11/src/client.c 2005-07-02 08:55:58.000000000 +0200 --- stunnel-4.11/src/client.c 2005-08-22 12:19:57.000000000 +0200 *************** *** 44,49 **** --- 44,60 ---- #define SHUT_RDWR 2 #endif + #define LINUX_NETFILTER + #define LINUX_TPROXY + + /* transparent proxy */ + #ifdef LINUX_NETFILTER + #include <linux/netfilter_ipv4.h> + #endif + #ifdef LINUX_TPROXY + #include <linux/netfilter_ipv4/ip_tproxy.h> + #endif + /* TCP wrapper */ #ifdef USE_LIBWRAP #include <tcpd.h> *************** *** 921,926 **** --- 932,944 ---- int s; /* destination socket */ u16 i; + // struct in_addr *local=NULL; + #ifdef LINUX_TPROXY + int f; + struct in_tproxy itp; + struct sockaddr_in tproxy_sin; + #endif + /* setup address_list */ if(c->opt->option.delayed_lookup) { resolved_list.num=0; *************** *** 945,960 **** if(alloc_fd(s)) return -1; ! if(c->bind_addr.num) { /* explicit local bind or transparent proxy */ ! memcpy(&bind_addr, &c->bind_addr.addr[0], sizeof(SOCKADDR_UNION)); ! if(bind(s, &bind_addr.sa, addr_len(bind_addr))<0) { ! sockerror("bind transparent"); ! closesocket(s); ! return -1; ! } ! } ! ! /* try to connect for the 1st time */ s_ntop(c->connecting_address, &addr); s_log(LOG_DEBUG, "%s connecting %s", c->opt->servname, c->connecting_address); --- 963,1053 ---- if(alloc_fd(s)) return -1; ! if(c->bind_addr.num && !c->opt->option.transparent) { ! /* explicit local bind, not transparent mode */ ! ! memcpy(&bind_addr, &c->bind_addr.addr[0], sizeof(SOCKADDR_UNION)); ! if(bind(s, &bind_addr.sa, addr_len(bind_addr))<0) { ! sockerror("bind transparent"); ! closesocket(s); ! return -1; ! } ! sockerror("local bind succeeded"); ! } ! ! if (c->opt->option.transparent){ ! ! /* transparent proxy support on 2.[46] linux kernels */ ! /* added by MAtteo HCE Valsasna (matteo.valsasna(a)uninsubria.it), 20050822 */ ! /* requires tproxy kernel patch (http://www.balabit.com/products/oss/tproxy */ ! /* based on examples in cttproxy patch */ ! ! /* FIXME: should detect 2.2 kernel or 2.[46] + cttproxy patch and use appropriate code */ ! ! memcpy(&bind_addr, &c->bind_addr.addr[0], sizeof(SOCKADDR_UNION)); ! ! /* check tproxy version */ ! itp.op = TPROXY_VERSION; ! itp.v.version = 0x00000000; ! ! getsockopt(s, SOL_IP, IP_TPROXY, &itp, &f); ! s_log(LOG_NOTICE, "pre getsockopt(s, SOL_IP, IP_TPROXY): %x", itp.v.version); ! ! /* FIXME: should accept a range of tproxy versions - which? */ ! ! if (itp.v.version != 0x02000002) { ! itp.v.version = 0x02000002; ! if (setsockopt(s, SOL_IP, IP_TPROXY, &itp, sizeof(itp)) == -1) ! { ! sockerror("setsockopt(SOL_IP, IP_TPROXY, TPROXY_VERSION)"); ! getsockopt(s, SOL_IP, IP_TPROXY, &itp, &f); ! s_log(LOG_NOTICE, "err getsockopt(s, SOL_IP, IP_TPROXY): %x", itp.v.version); ! return -1; ! } ! } ! else{ ! s_log(LOG_NOTICE, "getsockopt(SOL_IP, IP_TPROXY, TPROXY_VERSION) returned right version"); ! } ! ! /* bind to local address */ ! /* FIXME: must become dynamic */ ! tproxy_sin.sin_family = AF_INET; ! /* DOIT! add a local IP address here */ ! inet_aton("192.168.179.71", &tproxy_sin.sin_addr); ! /* do not request a specific source port */ ! tproxy_sin.sin_port = htons(0); ! ! if (bind(s, (struct sockaddr *) &tproxy_sin, sizeof(tproxy_sin)) == -1) ! { ! sockerror("bind"); ! return -1; ! } ! s_log(LOG_NOTICE, "bound to local: %s", inet_ntoa(tproxy_sin.sin_addr)); ! ! /* assign foreign address */ ! itp.op = TPROXY_ASSIGN; ! itp.v.addr.faddr = bind_addr.in.sin_addr; ! itp.v.addr.fport = 0; ! ! if (setsockopt(s, SOL_IP, IP_TPROXY, &itp, sizeof(itp)) == -1) ! { ! s_log(LOG_NOTICE, "error assigning foreign address: %s", inet_ntoa(bind_addr.in.sin_addr)); ! sockerror("setsockopt(SOL_IP, IP_TPROXY, TPROXY_ASSIGN)"); ! return -1; ! } ! s_log(LOG_NOTICE, "assigned foreign address: %s", inet_ntoa(bind_addr.in.sin_addr)); ! ! /* set connect flag on socket */ ! itp.op = TPROXY_FLAGS; ! itp.v.flags = ITP_CONNECT|ITP_ONCE; ! if (setsockopt(s, SOL_IP, IP_TPROXY, &itp, sizeof(itp)) == -1) ! { ! sockerror("setsockopt(SOL_IP, IP_TPROXY, TPROXY_FLAGS)"); ! return -1; ! } ! } ! ! /* try to connect for the 1st time */ s_ntop(c->connecting_address, &addr); s_log(LOG_DEBUG, "%s connecting %s", c->opt->servname, c->connecting_address); ----end patch----- EXTRAS source routing on the application server intended usage: the application server is a squid proxy with pubblic adresses. its default gw is an HW-based router (not the transparent stunnel server). we want normal traffic (i.e. not handled by transparent stunnel) to continue to use normal routing. only traffic handled by stunnel should be routed back to the stunnel server. a new IP address (10.0.0.1) is added on the application server specifically for this purpose. stunnel server will forward connections to this address, normal clients will use another address. then, only traffic originating from this address needs to be routed back to the stunnel server. the stunnel server has IP 10.0.0.254 (but it can have more, of course) in /etc/iproute2/rt_tables add a line defining a new routing table: 10 inr.tproxy # add a rule to the new routing table ip r a table inr.tproxy default via 10.0.0.254 # create ip policy routing rule: packets with source address = # 10.0.0.1 should be routed according to the inr.tproxy table ip rule add type unicast from 10.0.0.1 table inr.tproxy priority 220 # apply policy routing changes ip route flush cache

1 0

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

stunnel-users