- tproxy - lists.balabit.hu

2.6.12 patch on 2.6.11 kernel
by Tudor Alexandru Dragos 02 Nov '05

02 Nov '05

I have applied the tproxy 2.6.12-2.0.2. patch on 2.6.11-hardened-r15 (Gentoo SElinux default kernel). The patch applied almost clean I have to fix some Make files but no other major problem. When the kernel is compiled with SMP support the kernel freeze when the j TPROXY rule is hit by any package more exactly is crash on a writelock. The last line is executed is in ip_nat_core.c if (have_to_hash) { unsigned int srchash = hash_by_src(&conntrack->tuplehash[IP_CT_DIR_ORIGINAL] .tuple); WRITE_LOCK(&ip_nat_lock); <-- here is hangs list_add(&info->bysource, &bysource[srchash]); WRITE_UNLOCK(&ip_nat_lock); } I have tried the compiling the kernel without SMP and is almost working I get some kernel panic even with no tproxy rule at all But I didnt have time to investigate. Anybody manage to apply the patch on 2.6.11 kernel ? I have spent a lot of time applying a lot of patched on this kernel (I intended to use it in a extremely experimental network) so I am not too happy to change it because until now I never have problems with it. Do I have any chance to make it work or should I use a clean vanilla kernel sources. Sorry if is sounds kind of stupid applying a patch for kernel 2.6.12 on a 2.6.11 that is already extremely blotted with other patches. P.S: The problems manifests even with the stock hardened Gentoo kernel with no others patched but the one from the portage. (Not in enforce mode) __________________________________ Yahoo! FareChase: Search multiple travel sites in one click. http://farechase.yahoo.com

2 1

RE: RE: [tproxy] Fail to REDIRECT (of nat table)
by wckwon 11 Oct '05

11 Oct '05

Thanks for your answer. As soon as receiving your reply, I tried those and results are below. OS version : linux-2.6.12.6 with cttproxy-2.6.10-2.0.1.tar.gz test status ======================================================== bridging - on (not working) bridging - off (not working) using TPROXY rule instead of REDIRECT (not working, halting system) REDIRECT Rule without Bridging : iptables -t nat -A PREROUTING -i eth1 -d 192.168.0.107 -p tcp --dport 9999 -j REDIRECT --to-port 9216 TPROXY target Rule : iptables -m physdev -t tproxy -A PREROUTING --physdev-in eth1 -d A.B.C.D tcp --dport 80 -j TPROXY --on-port 2101 My poor daemon couldn't read any data in any contditions. )-;

1 0

Fail to REDIRECT (of nat table)
by wckwon 10 Oct '05

10 Oct '05

Hi people, My network daemon program can't receive any packet from iptables's REDIRECT after patching cttproxy-2.6.12-2.0.2.tar.gz. Of course, I checked it out that REDIRECT worked before patching in the same-kernel. (So, I tried it in Fedora kernel(2.6.12-1.1376_FC3). It was worse. The system was halting without any message.) <<My system configuration>> Linux-kernel version : 2.6.12 original (no warning, no error when patching) Gcc version : gcc (GCC) 3.4.4 20050721 (Red Hat 3.4.4-2) Modules : Module Size Used by ipt_REDIRECT 2688 1 ipt_physdev 2832 1 iptable_filter 3584 0 iptable_tproxy 19780 0 iptable_nat 27976 3 ipt_REDIRECT,iptable_tproxy ip_conntrack 44040 2 iptable_tproxy,iptable_nat ipt_TPROXY 2688 0 ip_tables 24448 6 ipt_REDIRECT,ipt_physdev,iptable_filter,iptable_tproxy,iptable_nat,ipt_TPROX Y video 16388 0 Device Network state : Bridge bridge name bridge id STP enabled interfaces br0 8000.00d06809c827 no eth3 eth1 IP table Rule : (iptables -m physdev -t nat -A PREROUTING --physdev-in eth1 -d A.B.C.D -p tcp --dport 80 -j REDIRECT --to-port 2101) Chain PREROUTING (policy ACCEPT) target prot opt source destination REDIRECT tcp -- anywhere A.B.C.D PHYSDEV match --physdev-in eth1 tcp dpt:http redir ports 2101 CPU info: 2 x Intel(R) Xeon(TM) CPU 3.00GHz (Hiper-Threading : On) netstat info: (telnet A.B.C.D 80) Active Internet connections (w/o servers) Proto Recv-Q Send-Q Local Address Foreign Address State tcp 0 0 192.168.0.107:2101 192.168.0.101:1147 SYN_RECV Thanks, in advance.

3 2

'ip_conntrack_put': function body not available
by Jörgen Bilander 03 Oct '05

03 Oct '05

hi! Hope someone can help me out here, tried compiling a kernel-source-2.6.12.6-2tr patched with cttproxy-2.6.12-2.0.2. When compiling I get this error message. Does anyone now a workaround for this problem? tia, Jörgen CC [M] net/ipv4/netfilter/iptable_filter.o CC [M] net/ipv4/netfilter/iptable_mangle.o LD [M] net/ipv4/netfilter/iptable_nat.o CC [M] net/ipv4/netfilter/iptable_tproxy.o net/ipv4/netfilter/iptable_tproxy.c: In function `ip_tproxy_kill_related': include/linux/netfilter_ipv4/ip_conntrack.h:261: sorry, unimplemented: inlining failed in call to 'ip_conntrack_put': function body not available net/ipv4/netfilter/iptable_tproxy.c:458: sorry, unimplemented: called from here make[3]: *** [net/ipv4/netfilter/iptable_tproxy.o] Error 1 make[2]: *** [net/ipv4/netfilter] Error 2 make[1]: *** [net/ipv4] Error 2 make: *** [net] Error 2

2 1

[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

Problem with tproxy and active-FTP
by tomerl lekach 24 Aug '05

24 Aug '05

Hello all, I'm using tproxy for a transparent active FTP proxy application. Clients use active-ftp towards the server, so my proxy gets "active" data connections initiated by the server from source port 20 (ftp-data) destined to the client to some high destination port. My application is using iptables (v1.2.8) rules to forward the incoming packets into an internal IP/Port used by the application. Then the application initiates a new connection to the client, using the same IP/Port of the connection coming from the FTP server. I face a problem that after a number of such successful connections with the right NAT, I see SYN packets leaving towards the client with an "internal" source Port, instead of source port 20 (as was used by the server). I'm using 2.4.18-24.8.0 linux kernel version, and the tproxy patch of 2.4.21-23 (cttproxy-2.4.21-23.tar.gz). Did you ever see such behavior and do you know how this can be fixed? Thanks in advance, Tomer.

2 1

Squid 3 TPROXY patch
by Jan Engelhardt 22 Aug '05

22 Aug '05

Hi list, for reference, I'll post my Squid3 patch here. Chunk 7 (@@ -1473,6) of comm.cc is the 'most important' part. diff -Pdpru squid-3.0-PRE3-20050524~/src/cf.data.pre squid-3.0-PRE3-20050524/src/cf.data.pre --- squid-3.0-PRE3-20050524~/src/cf.data.pre 2005-08-14 17:00:06.000000000 +0200 +++ squid-3.0-PRE3-20050524/src/cf.data.pre 2005-08-14 20:17:20.000000000 +0200 @@ -2800,6 +2800,16 @@ DOC_START the correct result. DOC_END +NAME: tproxy +TYPE: onoff +DEFAULT: off +LOC: Config.onoff.tproxy +DOC_START + If you have Linux with iptables and TPROXY support, you can enable + this option to have SQUID make outgoing connections using the original + IP address of the client. +DOC_END + NAME: tcp_outgoing_tos tcp_outgoing_ds tcp_outgoing_dscp TYPE: acl_tos DEFAULT: none diff -Pdpru squid-3.0-PRE3-20050524~/src/comm.cc squid-3.0-PRE3-20050524/src/comm.cc --- squid-3.0-PRE3-20050524~/src/comm.cc 2005-05-14 04:39:40.000000000 +0200 +++ squid-3.0-PRE3-20050524/src/comm.cc 2005-08-14 20:46:02.000000000 +0200 @@ -41,6 +41,7 @@ #include "fde.h" #include "CommIO.h" #include "ConnectionDetail.h" +#include "HttpRequest.h" #if defined(_SQUID_CYGWIN_) #include <sys/ioctl.h> @@ -49,6 +50,7 @@ #include <netinet/tcp.h> #endif +#include "ip_tproxy.h" class ConnectStateData { @@ -63,7 +65,7 @@ public: char *host; u_short port; - struct sockaddr_in S; + struct sockaddr_in S, src_addr; CallBack<CNCB> callback; struct IN_ADDR in_addr; @@ -1198,6 +1200,25 @@ ConnectStateData::operator delete (void cbdataFree(address); } +void commConnectStart2(int fd, const char *host, u_short port, CNCB *callback, + FwdState *fs) +{ + ConnectStateData *cs; + cs = new ConnectStateData; + cs->fd = fd; + cs->host = xstrdup(host); + cs->port = port; + cs->callback = CallBack<CNCB>(callback, fs); + if(fs->request != NULL) { + cs->src_addr.sin_addr = fs->request->client_addr; + cs->src_addr.sin_port = fs->request->client_port; + } else { + memset(&cs->src_addr, 0, sizeof(cs->src_addr)); + } + comm_add_close_handler(fd, commConnectFree, cs); + ipcache_nbgethostbyname(host, commConnectDnsHandle, cs); +} + void commConnectStart(int fd, const char *host, u_short port, CNCB * callback, void *data) { @@ -1401,7 +1422,7 @@ ConnectStateData::connect() if (S.sin_addr.s_addr == 0) defaults(); - switch (comm_connect_addr(fd, &S)) { + switch (comm_connect_addr(fd, &S, &src_addr)) { case COMM_INPROGRESS: debug(5, 5) ("commConnectHandle: FD %d: COMM_INPROGRESS\n", fd); @@ -1455,8 +1476,8 @@ commSetTimeout(int fd, int timeout, PF * } int - -comm_connect_addr(int sock, const struct sockaddr_in *address) +comm_connect_addr(int sock, const struct sockaddr_in *address, + const struct sockaddr_in *src) { comm_err_t status = COMM_OK; fde *F = &fd_table[sock]; @@ -1473,6 +1494,21 @@ comm_connect_addr(int sock, const struct F->flags.called_connect = 1; statCounter.syscalls.sock.connects++; + if(Config.onoff.tproxy && src != NULL && src->sin_addr.s_addr != 0 && + (ntohl(src->sin_addr.s_addr) & 0xFF000000) != 0x7F000000) { + struct in_tproxy itp; + memset(&itp, 0, sizeof(itp)); + itp.v.addr.faddr = src->sin_addr; + itp.v.addr.fport = src->sin_port; + itp.op = TPROXY_ASSIGN; + if((x = setsockopt(sock, SOL_IP, IP_TPROXY, &itp, sizeof(itp))) == 0) { + memset(&itp, 0, sizeof(itp)); + itp.v.flags = ITP_CONNECT; + itp.op = TPROXY_FLAGS; + setsockopt(sock, SOL_IP, IP_TPROXY, &itp, sizeof(itp)); + } + } + x = connect(sock, (struct sockaddr *) address, sizeof(*address)); if (x < 0) diff -Pdpru squid-3.0-PRE3-20050524~/src/forward.cc squid-3.0-PRE3-20050524/src/forward.cc --- squid-3.0-PRE3-20050524~/src/forward.cc 2005-04-18 23:52:42.000000000 +0200 +++ squid-3.0-PRE3-20050524/src/forward.cc 2005-08-14 20:01:25.000000000 +0200 @@ -663,7 +663,7 @@ fwdConnectStart(void *data) fwdConnectTimeout, fwdState); - commConnectStart(fd, host, port, fwdConnectDone, fwdState); + commConnectStart2(fd, host, port, fwdConnectDone, fwdState); } static void ##diff -Pdpru squid-3.0-PRE3-20050524~/src/ip_tproxy.h squid-3.0-PRE3-20050524/src/ip_tproxy.h ##--- squid-3.0-PRE3-20050524~/src/ip_tproxy.h 1970-01-01 01:00:00.000000000 +0100 ##+++ squid-3.0-PRE3-20050524/src/ip_tproxy.h 2005-08-14 20:01:25.000000000 +0200 ##@@ -0,0 +1,78 @@ ## Omitted, copy ip_tproxy.h here diff -Pdpru squid-3.0-PRE3-20050524~/src/protos.h squid-3.0-PRE3-20050524/src/protos.h --- squid-3.0-PRE3-20050524~/src/protos.h 2005-04-18 23:52:43.000000000 +0200 +++ squid-3.0-PRE3-20050524/src/protos.h 2005-08-14 20:01:25.000000000 +0200 @@ -126,8 +126,9 @@ SQUIDCEXTERN void comm_reset_close(int f SQUIDCEXTERN void comm_lingering_close(int fd); #endif SQUIDCEXTERN void commConnectStart(int fd, const char *, u_short, CNCB *, void *); +SQUIDCEXTERN void commConnectStart2(int fd, const char *, u_short, CNCB *, FwdState *); -SQUIDCEXTERN int comm_connect_addr(int sock, const struct sockaddr_in *); +SQUIDCEXTERN int comm_connect_addr(int sock, const struct sockaddr_in *, const struct sockaddr_in * = NULL); SQUIDCEXTERN void comm_init(void); SQUIDCEXTERN int comm_open(int, int, struct IN_ADDR, u_short port, int, const char *note); diff -Pdpru squid-3.0-PRE3-20050524~/src/structs.h squid-3.0-PRE3-20050524/src/structs.h --- squid-3.0-PRE3-20050524~/src/structs.h 2005-05-05 17:44:45.000000000 +0200 +++ squid-3.0-PRE3-20050524/src/structs.h 2005-08-14 20:17:53.000000000 +0200 @@ -580,6 +580,7 @@ struct _SquidConfig int via; int emailErrData; int httpd_suppress_version_string; + int tproxy; } onoff; ## eof Jan Engelhardt -- | Alphagate Systems, http://alphagate.hopto.org/

1 0

[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

2 1

TProxy version 2.0.2 released
by KOVACS Krisztian 25 Jul '05

25 Jul '05

Hi, TProxy version 2.0.2 has been released for Linux 2.4.31 and Linux 2.6.12. Finally, it's possible to run TProxy on the latest 2.6 kernels. Apart from sigificant code updates to work with the new 2.6 NAT code, this release has serious connection tracking entry reference counting changes. These should fix all TCP conntrack entry leakage problems experienced so far. The release tarballs are available here: http://www.balabit.com/downloads/tproxy/ MD5 checksums of the release tarballs: 58a43e1ac3c31a24e9eea2b7fbd7ea13 cttproxy-2.4.31-2.0.2.tar.gz 0f5380e9cbc1d432e392bfb6a6d1c7d3 cttproxy-2.6.12-2.0.2.tar.gz -- Regards, Krisztian Kovacs

1 0

Conntrack entries not decreasing
by Mohammed Riyaz 07 Jul '05

07 Jul '05

Hi, > Do you have any patches applied on 2.6.10 apart from tproxy? Vanilla >2.6.10 had a TCP connection tracking bug which caused some TCP >connections linger in the conntrack table for way too much time. Please >take a look at the original tproxy for 2.6.10 announcement in the >mailing list archives: ... > Nothing suspicious here, so I'd wait for your experience with the >2.6.10 TCP conntrack patch. Things seem to be running fine now with the conntrack patch. During day time the maximun value is reaching 3000 entries and in the nights it reduces below 700. :) Thanks a lot. Mohammed Riyaz P. -- http://www.fastmail.fm - And now for something completely different

1 0