tcpreplay/src/bridge.c

/* $Id: bridge.c 2195 2009-02-04 21:17:45Z aturner $ */

/*
 * Copyright (c) 2001-2005 Aaron Turner.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. Neither the names of the copyright owners nor the names of its
 *    contributors may be used to endorse or promote products derived from
 *    this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
 * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE
 * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER
 * IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

#include "config.h"
#include "defines.h"
#include "common.h"

#include <sys/time.h>
#include <signal.h>
#include <string.h>
#include <netinet/in.h>
#include <time.h>
#include <errno.h>
#include <stdlib.h>

#ifdef HAVE_BPF
#include <sys/select.h> /* necessary for using select() for BPF devices */
#endif

#include "tcpbridge.h"
#include "bridge.h"
#include "send_packets.h"
#include "tcpedit/tcpedit.h"

extern tcpbridge_opt_t options;
extern struct timeval begin, end;
extern COUNTER bytes_sent, failed, pkts_sent;
extern volatile int didsig;

#ifdef DEBUG
extern int debug;
#endif

static int live_callback(struct live_data_t *,
                         struct pcap_pkthdr *, const u_char *);

/**
 * First, prep our RB Tree which tracks where each (source)
 * MAC really lives so we don't create really nasty network
 * storms.  
 */
static struct macsrc_t *new_node(void);

RB_HEAD(macsrc_tree, macsrc_t) macsrc_root;

static int
rbmacsrc_comp(struct macsrc_t *a, struct macsrc_t *b)
{
    return (memcmp(a->key, b->key, ETHER_ADDR_LEN));
}

RB_PROTOTYPE(macsrc_tree, macsrc_t, node, rbmacsrc_comp)
RB_GENERATE(macsrc_tree, macsrc_t, node, rbmacsrc_comp)

/**
 * redblack init
 */
void
rbinit(void)
{
    RB_INIT(&macsrc_root);
}

/**
 * create a new node... Malloc's memory
 */
struct macsrc_t *
new_node(void)
{
    struct macsrc_t *node;

    node = (struct macsrc_t *)safe_malloc(sizeof(struct macsrc_t));
    
    memset(node, '\0', sizeof(struct macsrc_t));
    return (node);
}


/**
 * main loop for bridging in only one direction
 * optimized to not use poll(), but rather libpcap's builtin pcap_loop()
 */
static void
do_bridge_unidirectional(tcpbridge_opt_t *options, tcpedit_t *tcpedit)
{
    struct live_data_t livedata;
    int retcode;

    assert(options);
    assert(tcpedit);
    
    livedata.tcpedit = tcpedit;
    livedata.source = PCAP_INT1;
    livedata.pcap = options->pcap1;
    livedata.options = options;

    if ((retcode = pcap_loop(options->pcap1, options->limit_send, 
            (pcap_handler)live_callback, (u_char *) &livedata)) < 0) {
        warnx("Error in pcap_loop(): %s", pcap_geterr(options->pcap1));
    }
    
}

#ifndef HAVE_BPF
/**
 * main loop for bridging in both directions.  Since we dealing with two handles
 * we need to poll() on them which isn't the most efficent. 
 *
 * Note that this function is only used on systems which do not have a BPF
 * device because poll() behaves poorly with /dev/bpf
 */
static void
do_bridge_bidirectional(tcpbridge_opt_t *options, tcpedit_t *tcpedit)
{
    struct pollfd polls[2];     /* one for left & right pcap */
    int pollresult, pollcount, timeout;
    struct live_data_t livedata;
    
    assert(options);
    assert(tcpedit);

    livedata.tcpedit = tcpedit;
    livedata.options = options;


    /* 
     * loop until ctrl-C or we've sent enough packets
     * note that if -L wasn't specified, limit_send is
     * set to 0 so this will loop infinately
     */
    while ((options->limit_send == 0) || (options->limit_send > pkts_sent)) {
        if (didsig)
            break;

        dbgx(3, "limit_send: " COUNTER_SPEC " \t pkts_sent: " COUNTER_SPEC, 
            options->limit_send, pkts_sent);

        /* reset the result codes */
        polls[PCAP_INT1].revents = 0;
        polls[PCAP_INT1].events = POLLIN;
        polls[PCAP_INT1].fd = pcap_fileno(options->pcap1);
        
        polls[PCAP_INT2].revents = 0;
        polls[PCAP_INT2].events = POLLIN;
        polls[PCAP_INT2].fd = pcap_fileno(options->pcap2);

        timeout = options->poll_timeout;
        pollcount = 2;

        /* poll for a packet on the two interfaces */
        pollresult = poll(polls, pollcount, timeout);

        /* poll has returned, process the result */
        if (pollresult > 0) {
            dbgx(3, "pollresult: %d", pollresult);
            
            /* success, got one or more packets */
            if (polls[PCAP_INT1].revents > 0) {
                dbg(5, "Processing first interface");
                livedata.source = PCAP_INT1;
                livedata.pcap = options->pcap1;
                pcap_dispatch(options->pcap1, -1, (pcap_handler) live_callback,
                              (u_char *) &livedata);
            }

            /* check the other interface?? */
            if (polls[PCAP_INT2].revents > 0) {
                dbg(5, "Processing second interface");
                livedata.source = PCAP_INT2;
                livedata.pcap = options->pcap2;
                pcap_dispatch(options->pcap2, -1, (pcap_handler) live_callback,
                              (u_char *) &livedata);
            }

        }
        else if (pollresult == 0) {
            dbg(3, "poll timeout exceeded...");
            /* do something here? */
        }
        else {
            /* poll error, probably a Ctrl-C */
            warnx("poll() error: %s", strerror(errno));
        }

        /* go back to the top of the loop */
    }

} /* do_bridge_bidirectional() */

#elif defined HAVE_BPF && defined HAVE_PCAP_SETNONBLOCK 
/**
 * main loop for bridging in both directions with BPF.  We'll be using
 * select() because that works better on older *BSD and OSX
 *
 * See this for details behind this maddness:
 * http://article.gmane.org/gmane.network.tcpdump.devel/3581
 */
static void
do_bridge_bidirectional(tcpbridge_opt_t *options, tcpedit_t *tcpedit)
{
    fd_set readfds, writefds, errorfds;
    struct live_data_t livedata;
    int fd, nfds, ret;
    struct timeval timeout = { 0, 100 }; /* default to 100ms timeout */
    char ebuf[PCAP_ERRBUF_SIZE];
    
    assert(options);
    assert(tcpedit);

    livedata.tcpedit = tcpedit;
    livedata.options = options;

    /* 
     * loop until ctrl-C or we've sent enough packets
     * note that if -L wasn't specified, limit_send is
     * set to 0 so this will loop infinately
     */
    while ((options->limit_send == 0) || (options->limit_send > pkts_sent)) {
        if (didsig)
            break;

        dbgx(3, "limit_send: " COUNTER_SPEC " \t pkts_sent: " COUNTER_SPEC, 
            options->limit_send, pkts_sent);

        /* reset the result codes */
        FD_ZERO(&readfds);
        FD_ZERO(&writefds);
        FD_ZERO(&errorfds);

        /* set for reading */
#ifdef HAVE_PCAP_GET_SELECTABLE_FD
        fd = pcap_get_selectable_fd(options->pcap1);
#else
        fd = pcap_fileno(options->pcap1);
#endif
        if ((pcap_setnonblock(options->pcap1, 1, ebuf)) < 0)
            errx(1, "Unable to set %s into nonblocking mode: %s", options->intf1, ebuf);
        FD_SET(fd, &readfds);
            
#ifdef HAVE_PCAP_GET_SELECTABLE_FD
        fd = pcap_get_selectable_fd(options->pcap2);
#else
        fd = pcap_fileno(options->pcap2);
#endif
        if ((pcap_setnonblock(options->pcap2, 1, ebuf)) < 0)
            errx(1, "Unable to set %s into nonblocking mode: %s", options->intf2, ebuf);
        FD_SET(fd, &readfds);
        
        nfds = 2;

        /* wait for a packet on the two interfaces */
        ret = select(nfds, &readfds, &writefds, &errorfds, &timeout);

        /* 
         * There is a problem with OS X and certian *BSD's when using
         * select() on a character device like /dev/bpf.  Hence we always
         * must attempt to read off each fd after the timeout.  This is why
         * we put the fd's in nonblocking mode above!
         */
         
        dbg(5, "Processing first interface");
        livedata.source = PCAP_INT1;
        livedata.pcap = options->pcap1;
        pcap_dispatch(options->pcap1, -1, (pcap_handler) live_callback,
                      (u_char *) &livedata);
         
        dbg(5, "Processing second interface");
        livedata.source = PCAP_INT2;
        livedata.pcap = options->pcap2;
        pcap_dispatch(options->pcap2, -1, (pcap_handler) live_callback,
                      (u_char *) &livedata);

        /* go back to the top of the loop */
    }    
} 
#else
#error "Your system needs a libpcap with pcap_setnonblock().  Please upgrade libpcap."
#endif

/**
 * Main entry point to bridging.  Does some initial setup and then calls the 
 * correct loop (unidirectional or bidirectional)
 */
void
do_bridge(tcpbridge_opt_t *options, tcpedit_t *tcpedit)
{   
    /* do we apply a bpf filter? */
    if (options->bpf.filter != NULL) {
        /* compile filter */
        dbgx(2, "Try to compile pcap bpf filter: %s", options->bpf.filter);
        if (pcap_compile(options->pcap1, &options->bpf.program, options->bpf.filter, options->bpf.optimize, 0) != 0) {
            errx(-1, "Error compiling BPF filter: %s", pcap_geterr(options->pcap1));
        }
        
        /* apply filter */
        pcap_setfilter(options->pcap1, &options->bpf.program);

        /* same for other interface if applicable */
        if (options->unidir == 0) {
            /* compile filter */
            dbgx(2, "Try to compile pcap bpf filter: %s", options->bpf.filter);
            if (pcap_compile(options->pcap2, &options->bpf.program, options->bpf.filter, options->bpf.optimize, 0) != 0) {
                errx(-1, "Error compiling BPF filter: %s", pcap_geterr(options->pcap2));
            }
        
            /* apply filter */
            pcap_setfilter(options->pcap2, &options->bpf.program);
        }
    }

    /* register signals */
    didsig = 0;
    (void)signal(SIGINT, catcher);


    if (options->unidir == 1) {
        do_bridge_unidirectional(options, tcpedit);
    } else {
        do_bridge_bidirectional(options, tcpedit);
    }
            
    packet_stats(&begin, &end, bytes_sent, pkts_sent, failed);
}


/**
 * This is the callback we use with pcap_dispatch to process
 * each packet recieved by libpcap on the two interfaces.
 * Need to return > 0 to denote success
 */
static int
live_callback(struct live_data_t *livedata, struct pcap_pkthdr *pkthdr,
              const u_char * nextpkt)
{
    ipv4_hdr_t *ip_hdr = NULL;
    pcap_t *send = NULL;
    static u_char *pktdata = NULL;     /* full packet buffer */
    int cache_mode, retcode;
    static unsigned long packetnum = 0;
    struct macsrc_t *node, finder;  /* rb tree nodes */
#ifdef DEBUG
    u_char dstmac[ETHER_ADDR_LEN];
#endif
    u_int16_t l2proto;

    packetnum++;
    dbgx(2, "packet %lu caplen %d", packetnum, pkthdr->caplen);

    /* only malloc the first time */
    if (pktdata == NULL) {
        /* create packet buffers */
        pktdata = (u_char *)safe_malloc(MAXPACKET);
    } else {
        /* zero out the old packet info */
        memset(pktdata, '\0', MAXPACKET);
    }

    /* copy the packet to our buffer */
    memcpy(pktdata, nextpkt, pkthdr->caplen);


#ifdef ENABLE_VERBOSE
    /* decode packet? */
    if (livedata->options->verbose)
        tcpdump_print(livedata->options->tcpdump, pkthdr, nextpkt);
#endif


    /* lookup our source MAC in the tree */
    memcpy(&finder.key, &pktdata[ETHER_ADDR_LEN], ETHER_ADDR_LEN);
#ifdef DEBUG
    memcpy(&dstmac, pktdata, ETHER_ADDR_LEN);
    dbgx(1, "SRC MAC: " MAC_FORMAT "\tDST MAC: " MAC_FORMAT,
        MAC_STR(finder.key), MAC_STR(dstmac));
#endif

    /* first, is this a packet sent locally?  If so, ignore it */
    if ((memcmp(livedata->options->intf1_mac, &finder.key, ETHER_ADDR_LEN)) == 0) {
        dbgx(1, "Packet matches the MAC of %s, skipping.", livedata->options->intf1);
        return (1);
    }
    else if ((memcmp(livedata->options->intf2_mac, &finder.key, ETHER_ADDR_LEN)) == 0) {
        dbgx(1, "Packet matches the MAC of %s, skipping.", livedata->options->intf2);
        return (1);
    }

    node = RB_FIND(macsrc_tree, &macsrc_root, &finder);
    
    /* if we can't find the node, build a new one */
    if (node == NULL) {
        dbg(1, "Unable to find MAC in the tree");
        node = new_node();
        node->source = livedata->source;
        memcpy(&node->key, &finder.key, ETHER_ADDR_LEN);
        RB_INSERT(macsrc_tree, &macsrc_root, node);
    }
    
    /* otherwise compare sources */
    else if (node->source != livedata->source) {
        dbg(1, "Found the dest MAC in the tree and it doesn't match this source NIC... skipping packet");
        /*
         * IMPORTANT!!!
         * Never send a packet out the same interface we sourced it on!
         */
        return (1);
    }

    /* what is our cache mode? */
    cache_mode = livedata->source == PCAP_INT1 ? TCPR_DIR_C2S : TCPR_DIR_S2C;

    l2proto = tcpedit_l3proto(livedata->tcpedit, BEFORE_PROCESS, pktdata, pkthdr->len);
    dbgx(2, "Packet protocol: %04hx", l2proto);
    
    /* should we skip this packet based on CIDR match? */
    if (l2proto == ETHERTYPE_IP) {
        dbg(3, "Packet is IP");
        ip_hdr = (ipv4_hdr_t *)tcpedit_l3data(livedata->tcpedit, BEFORE_PROCESS, pktdata, pkthdr->len);

        /* look for include or exclude CIDR match */
        if (livedata->options->xX.cidr != NULL) {
            if (!process_xX_by_cidr(livedata->options->xX.mode, livedata->options->xX.cidr, ip_hdr)) {
                dbg(2, "Skipping packet due to CIDR match");
                return (1);
            }
        }

    }

    if ((retcode = tcpedit_packet(livedata->tcpedit, &pkthdr, &pktdata, cache_mode)) < 0) {
        if (retcode == TCPEDIT_SOFT_ERROR) {
            return 1;
        } else { /* TCPEDIT_ERROR */
            return -1;
        }
    }

    /* 
     * send packets out the OTHER interface
     * and update the dst mac if necessary
     */
    switch(node->source) {
        case PCAP_INT1:
            dbgx(2, "Packet source was %s... sending out on %s", livedata->options->intf1, 
                livedata->options->intf2);
            send = livedata->options->pcap2;
            break;

        case PCAP_INT2:
            dbgx(2, "Packet source was %s... sending out on %s", livedata->options->intf2, 
                livedata->options->intf1);
            send = livedata->options->pcap1;
            break;
        
        default:
            errx(-1, "wtf?  our node->source != PCAP_INT1 and != PCAP_INT2: %c", 
                 node->source);        
    }

    /*
     * write packet out on the network 
     */
     if (pcap_sendpacket(send, pktdata, pkthdr->caplen) < 0)
         errx(-1, "Unable to send packet out %s: %s", 
            send == livedata->options->pcap1 ? livedata->options->intf1 : livedata->options->intf2, pcap_geterr(send));

    bytes_sent += pkthdr->caplen;
    pkts_sent++;

    dbgx(1, "Sent packet " COUNTER_SPEC, pkts_sent);


    return (1);
} /* live_callback() */