ftp: code shrink
[oweals/busybox.git] / networking / traceroute.c
index a9cad4baf64c69b113ea4a188a4cafc06eafa70f..7284f0022d51facd417c4cd71db3ed07d2b603f8 100644 (file)
-/*-
- * Copyright (c) 1990, 1993
+/* vi: set sw=4 ts=4: */
+/*
+ * Copyright (c) 1988, 1989, 1991, 1994, 1995, 1996, 1997, 1998, 1999, 2000
  *      The Regents of the University of California.  All rights reserved.
  *
- * This code is derived from software contributed to Berkeley by
- * Van Jacobson.
+ * Busybox port by Vladimir Oleynik (C) 2005 <dzo@simtreas.ru>
  *
- * Special for busybox ported by Vladimir Oleynik <dzo@simtreas.ru> 2001
  * 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 name of the University 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 BY THE REGENTS AND CONTRIBUTORS ``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 REGENTS OR CONTRIBUTORS 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.
+ * modification, are permitted provided that: (1) source code distributions
+ * retain the above copyright notice and this paragraph in its entirety, (2)
+ * distributions including binary code include the above copyright notice and
+ * this paragraph in its entirety in the documentation or other materials
+ * provided with the distribution, and (3) all advertising materials mentioning
+ * features or use of this software display the following acknowledgement:
+ * ``This product includes software developed by the University of California,
+ * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of
+ * the University 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 WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
  */
 
 /*
  * traceroute host  - trace the route ip packets follow going to "host".
+ *
+ * Attempt to trace the route an ip packet would follow to some
+ * internet host.  We find out intermediate hops by launching probe
+ * packets with a small ttl (time to live) then listening for an
+ * icmp "time exceeded" reply from a gateway.  We start our probes
+ * with a ttl of one and increase by one until we get an icmp "port
+ * unreachable" (which means we got to "host") or hit a max (which
+ * defaults to 30 hops & can be changed with the -m flag).  Three
+ * probes (change with -q flag) are sent at each ttl setting and a
+ * line is printed showing the ttl, address of the gateway and
+ * round trip time of each probe.  If the probe answers come from
+ * different gateways, the address of each responding system will
+ * be printed.  If there is no response within a 5 sec. timeout
+ * interval (changed with the -w flag), a "*" is printed for that
+ * probe.
+ *
+ * Probe packets are UDP format.  We don't want the destination
+ * host to process them so the destination port is set to an
+ * unlikely value (if some clod on the destination is using that
+ * value, it can be changed with the -p flag).
+ *
+ * A sample use might be:
+ *
+ *     [yak 71]% traceroute nis.nsf.net.
+ *     traceroute to nis.nsf.net (35.1.1.48), 30 hops max, 56 byte packet
+ *      1  helios.ee.lbl.gov (128.3.112.1)  19 ms  19 ms  0 ms
+ *      2  lilac-dmc.Berkeley.EDU (128.32.216.1)  39 ms  39 ms  19 ms
+ *      3  lilac-dmc.Berkeley.EDU (128.32.216.1)  39 ms  39 ms  19 ms
+ *      4  ccngw-ner-cc.Berkeley.EDU (128.32.136.23)  39 ms  40 ms  39 ms
+ *      5  ccn-nerif22.Berkeley.EDU (128.32.168.22)  39 ms  39 ms  39 ms
+ *      6  128.32.197.4 (128.32.197.4)  40 ms  59 ms  59 ms
+ *      7  131.119.2.5 (131.119.2.5)  59 ms  59 ms  59 ms
+ *      8  129.140.70.13 (129.140.70.13)  99 ms  99 ms  80 ms
+ *      9  129.140.71.6 (129.140.71.6)  139 ms  239 ms  319 ms
+ *     10  129.140.81.7 (129.140.81.7)  220 ms  199 ms  199 ms
+ *     11  nic.merit.edu (35.1.1.48)  239 ms  239 ms  239 ms
+ *
+ * Note that lines 2 & 3 are the same.  This is due to a buggy
+ * kernel on the 2nd hop system -- lbl-csam.arpa -- that forwards
+ * packets with a zero ttl.
+ *
+ * A more interesting example is:
+ *
+ *     [yak 72]% traceroute allspice.lcs.mit.edu.
+ *     traceroute to allspice.lcs.mit.edu (18.26.0.115), 30 hops max
+ *      1  helios.ee.lbl.gov (128.3.112.1)  0 ms  0 ms  0 ms
+ *      2  lilac-dmc.Berkeley.EDU (128.32.216.1)  19 ms  19 ms  19 ms
+ *      3  lilac-dmc.Berkeley.EDU (128.32.216.1)  39 ms  19 ms  19 ms
+ *      4  ccngw-ner-cc.Berkeley.EDU (128.32.136.23)  19 ms  39 ms  39 ms
+ *      5  ccn-nerif22.Berkeley.EDU (128.32.168.22)  20 ms  39 ms  39 ms
+ *      6  128.32.197.4 (128.32.197.4)  59 ms  119 ms  39 ms
+ *      7  131.119.2.5 (131.119.2.5)  59 ms  59 ms  39 ms
+ *      8  129.140.70.13 (129.140.70.13)  80 ms  79 ms  99 ms
+ *      9  129.140.71.6 (129.140.71.6)  139 ms  139 ms  159 ms
+ *     10  129.140.81.7 (129.140.81.7)  199 ms  180 ms  300 ms
+ *     11  129.140.72.17 (129.140.72.17)  300 ms  239 ms  239 ms
+ *     12  * * *
+ *     13  128.121.54.72 (128.121.54.72)  259 ms  499 ms  279 ms
+ *     14  * * *
+ *     15  * * *
+ *     16  * * *
+ *     17  * * *
+ *     18  ALLSPICE.LCS.MIT.EDU (18.26.0.115)  339 ms  279 ms  279 ms
+ *
+ * (I start to see why I'm having so much trouble with mail to
+ * MIT.)  Note that the gateways 12, 14, 15, 16 & 17 hops away
+ * either don't send ICMP "time exceeded" messages or send them
+ * with a ttl too small to reach us.  14 - 17 are running the
+ * MIT C Gateway code that doesn't send "time exceeded"s.  God
+ * only knows what's going on with 12.
+ *
+ * The silent gateway 12 in the above may be the result of a bug in
+ * the 4.[23]BSD network code (and its derivatives):  4.x (x <= 3)
+ * sends an unreachable message using whatever ttl remains in the
+ * original datagram.  Since, for gateways, the remaining ttl is
+ * zero, the icmp "time exceeded" is guaranteed to not make it back
+ * to us.  The behavior of this bug is slightly more interesting
+ * when it appears on the destination system:
+ *
+ *      1  helios.ee.lbl.gov (128.3.112.1)  0 ms  0 ms  0 ms
+ *      2  lilac-dmc.Berkeley.EDU (128.32.216.1)  39 ms  19 ms  39 ms
+ *      3  lilac-dmc.Berkeley.EDU (128.32.216.1)  19 ms  39 ms  19 ms
+ *      4  ccngw-ner-cc.Berkeley.EDU (128.32.136.23)  39 ms  40 ms  19 ms
+ *      5  ccn-nerif35.Berkeley.EDU (128.32.168.35)  39 ms  39 ms  39 ms
+ *      6  csgw.Berkeley.EDU (128.32.133.254)  39 ms  59 ms  39 ms
+ *      7  * * *
+ *      8  * * *
+ *      9  * * *
+ *     10  * * *
+ *     11  * * *
+ *     12  * * *
+ *     13  rip.Berkeley.EDU (128.32.131.22)  59 ms !  39 ms !  39 ms !
+ *
+ * Notice that there are 12 "gateways" (13 is the final
+ * destination) and exactly the last half of them are "missing".
+ * What's really happening is that rip (a Sun-3 running Sun OS3.5)
+ * is using the ttl from our arriving datagram as the ttl in its
+ * icmp reply.  So, the reply will time out on the return path
+ * (with no notice sent to anyone since icmp's aren't sent for
+ * icmp's) until we probe with a ttl that's at least twice the path
+ * length.  I.e., rip is really only 7 hops away.  A reply that
+ * returns with a ttl of 1 is a clue this problem exists.
+ * Traceroute prints a "!" after the time if the ttl is <= 1.
+ * Since vendors ship a lot of obsolete (DEC's Ultrix, Sun 3.x) or
+ * non-standard (HPUX) software, expect to see this problem
+ * frequently and/or take care picking the target host of your
+ * probes.
+ *
+ * Other possible annotations after the time are !H, !N, !P (got a host,
+ * network or protocol unreachable, respectively), !S or !F (source
+ * route failed or fragmentation needed -- neither of these should
+ * ever occur and the associated gateway is busted if you see one).  If
+ * almost all the probes result in some kind of unreachable, traceroute
+ * will give up and exit.
+ *
  * Notes
  * -----
  * This program must be run by root or be setuid.  (I suggest that
  * you *don't* make it setuid -- casual use could result in a lot
  * of unnecessary traffic on our poor, congested nets.)
  *
+ * This program requires a kernel mod that does not appear in any
+ * system available from Berkeley:  A raw ip socket using proto
+ * IPPROTO_RAW must interpret the data sent as an ip datagram (as
+ * opposed to data to be wrapped in a ip datagram).  See the README
+ * file that came with the source to this program for a description
+ * of the mods I made to /sys/netinet/raw_ip.c.  Your mileage may
+ * vary.  But, again, ANY 4.x (x < 4) BSD KERNEL WILL HAVE TO BE
+ * MODIFIED TO RUN THIS PROGRAM.
+ *
+ * The udp port usage may appear bizarre (well, ok, it is bizarre).
+ * The problem is that an icmp message only contains 8 bytes of
+ * data from the original datagram.  8 bytes is the size of a udp
+ * header so, if we want to associate replies with the original
+ * datagram, the necessary information must be encoded into the
+ * udp header (the ip id could be used but there's no way to
+ * interlock with the kernel's assignment of ip id's and, anyway,
+ * it would have taken a lot more kernel hacking to allow this
+ * code to set the ip id).  So, to allow two or more users to
+ * use traceroute simultaneously, we use this task's pid as the
+ * source port (the high bit is set to move the port number out
+ * of the "likely" range).  To keep track of which probe is being
+ * replied to (so times and/or hop counts don't get confused by a
+ * reply that was delayed in transit), we increment the destination
+ * port number before each probe.
+ *
+ * Don't use this as a coding example.  I was trying to find a
+ * routing problem and this code sort-of popped out after 48 hours
+ * without sleep.  I was amazed it ever compiled, much less ran.
+ *
  * I stole the idea for this program from Steve Deering.  Since
  * the first release, I've learned that had I attended the right
  * IETF working group meetings, I also could have stolen it from Guy
  * back to yourself.  Unfortunately, SO many gateways botch source
  * routing, the thing is almost worthless.  Maybe one day...
  *
- *  -- Van Jacobson (van@helios.ee.lbl.gov)
+ *  -- Van Jacobson (van@ee.lbl.gov)
  *     Tue Dec 20 03:50:13 PST 1988
  */
 
-#undef CONFIG_FEATURE_TRACEROUTE_VERBOSE
+#define TRACEROUTE_SO_DEBUG 0
+
+/* TODO: undefs were uncommented - ??! we have config system for that! */
+/* probably ok to remove altogether */
+//#undef CONFIG_FEATURE_TRACEROUTE_VERBOSE
 //#define CONFIG_FEATURE_TRACEROUTE_VERBOSE
-#undef CONFIG_FEATURE_TRACEROUTE_SO_DEBUG   /* not in documentation man */
-
-#include <stdio.h>
-#include <errno.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
-#include <sys/time.h>
-#include "inet_common.h"
-#include <netdb.h>
-#include <endian.h>
+//#undef CONFIG_FEATURE_TRACEROUTE_SOURCE_ROUTE
+//#define CONFIG_FEATURE_TRACEROUTE_SOURCE_ROUTE
+//#undef CONFIG_FEATURE_TRACEROUTE_USE_ICMP
+//#define CONFIG_FEATURE_TRACEROUTE_USE_ICMP
+
+
+#include <net/if.h>
+#include <arpa/inet.h>
+#include <netinet/in.h>
 #include <netinet/udp.h>
 #include <netinet/ip.h>
 #include <netinet/ip_icmp.h>
 
- /* It turns out that libc5 doesn't have proper icmp support
- * built into it header files, so we have to supplement it */
-#if __GNU_LIBRARY__ < 5
-static const int ICMP_MINLEN = 8;                              /* abs minimum */
+#include "libbb.h"
+#include "inet_common.h"
 
-struct icmp_ra_addr
-{
-  u_int32_t ira_addr;
-  u_int32_t ira_preference;
+#ifndef IPPROTO_ICMP
+# define IPPROTO_ICMP 1
+#endif
+#ifndef IPPROTO_IP
+# define IPPROTO_IP 0
+#endif
+
+/* Keep in sync with getopt32 call! */
+enum {
+       OPT_DONT_FRAGMNT = (1 << 0),    /* F */
+       OPT_USE_ICMP     = (1 << 1) * ENABLE_FEATURE_TRACEROUTE_USE_ICMP, /* I */
+       OPT_TTL_FLAG     = (1 << 2),    /* l */
+       OPT_ADDR_NUM     = (1 << 3),    /* n */
+       OPT_BYPASS_ROUTE = (1 << 4),    /* r */
+       OPT_DEBUG        = (1 << 5),    /* d */
+       OPT_VERBOSE      = (1 << 6) * ENABLE_FEATURE_TRACEROUTE_VERBOSE, /* v */
+       OPT_IP_CHKSUM    = (1 << 7),    /* x */
+       OPT_TOS          = (1 << 8),    /* t */
+       OPT_DEVICE       = (1 << 9),    /* i */
+       OPT_MAX_TTL      = (1 << 10),   /* m */
+       OPT_PORT         = (1 << 11),   /* p */
+       OPT_NPROBES      = (1 << 12),   /* q */
+       OPT_SOURCE       = (1 << 13),   /* s */
+       OPT_WAITTIME     = (1 << 14),   /* w */
+       OPT_PAUSE_MS     = (1 << 15),   /* z */
+       OPT_FIRST_TTL    = (1 << 16),   /* f */
 };
+#define verbose (option_mask32 & OPT_VERBOSE)
 
+enum {
+       SIZEOF_ICMP_HDR = 8,
+       rcvsock = 3, /* receive (icmp) socket file descriptor */
+       sndsock = 4, /* send (udp/icmp) socket file descriptor */
+};
 
-struct icmp
-{
-  u_int8_t  icmp_type; /* type of message, see below */
-  u_int8_t  icmp_code; /* type sub code */
-  u_int16_t icmp_cksum;        /* ones complement checksum of struct */
-  union
-  {
-    u_char ih_pptr;            /* ICMP_PARAMPROB */
-    struct in_addr ih_gwaddr;  /* gateway address */
-    struct ih_idseq            /* echo datagram */
-    {
-      u_int16_t icd_id;
-      u_int16_t icd_seq;
-    } ih_idseq;
-    u_int32_t ih_void;
-
-    /* ICMP_UNREACH_NEEDFRAG -- Path MTU Discovery (RFC1191) */
-    struct ih_pmtu
-    {
-      u_int16_t ipm_void;
-      u_int16_t ipm_nextmtu;
-    } ih_pmtu;
-
-    struct ih_rtradv
-    {
-      u_int8_t irt_num_addrs;
-      u_int8_t irt_wpa;
-      u_int16_t irt_lifetime;
-    } ih_rtradv;
-  } icmp_hun;
-#define        icmp_pptr       icmp_hun.ih_pptr
-#define        icmp_gwaddr     icmp_hun.ih_gwaddr
-#define        icmp_id         icmp_hun.ih_idseq.icd_id
-#define        icmp_seq        icmp_hun.ih_idseq.icd_seq
-#define        icmp_void       icmp_hun.ih_void
-#define        icmp_pmvoid     icmp_hun.ih_pmtu.ipm_void
-#define        icmp_nextmtu    icmp_hun.ih_pmtu.ipm_nextmtu
-#define        icmp_num_addrs  icmp_hun.ih_rtradv.irt_num_addrs
-#define        icmp_wpa        icmp_hun.ih_rtradv.irt_wpa
-#define        icmp_lifetime   icmp_hun.ih_rtradv.irt_lifetime
-  union
-  {
-    struct
-    {
-      u_int32_t its_otime;
-      u_int32_t its_rtime;
-      u_int32_t its_ttime;
-    } id_ts;
-    struct
-    {
-      struct ip idi_ip;
-      /* options and then 64 bits of data */
-    } id_ip;
-    struct icmp_ra_addr id_radv;
-    u_int32_t   id_mask;
-    u_int8_t    id_data[1];
-  } icmp_dun;
-#define        icmp_otime      icmp_dun.id_ts.its_otime
-#define        icmp_rtime      icmp_dun.id_ts.its_rtime
-#define        icmp_ttime      icmp_dun.id_ts.its_ttime
-#define        icmp_ip         icmp_dun.id_ip.idi_ip
-#define        icmp_radv       icmp_dun.id_radv
-#define        icmp_mask       icmp_dun.id_mask
-#define        icmp_data       icmp_dun.id_data
+/* Data section of the probe packet */
+struct outdata_t {
+       unsigned char seq;             /* sequence number of this packet */
+       unsigned char ttl;             /* ttl packet left with */
+// UNUSED. Retaining to have the same packet size.
+       struct timeval tv_UNUSED PACKED; /* time packet left */
 };
 
-#define        ICMP_MINLEN     8                               /* abs minimum */
-#define        ICMP_UNREACH            3               /* dest unreachable, codes: */
-#define        ICMP_TIMXCEED           11              /* time exceeded, code: */
-#define        ICMP_TIMXCEED_INTRANS   0               /* ttl==0 in transit */
-#define        ICMP_UNREACH_NET                0       /* bad net */
-#define        ICMP_UNREACH_HOST               1       /* bad host */
-#define        ICMP_UNREACH_PROTOCOL           2       /* bad protocol */
-#define        ICMP_UNREACH_PORT               3       /* bad port */
-#define        ICMP_UNREACH_NEEDFRAG           4       /* IP_DF caused drop */
-#define        ICMP_UNREACH_SRCFAIL            5       /* src route failed */
+struct globals {
+       struct ip *outip;
+       struct outdata_t *outdata;
+       len_and_sockaddr *dest_lsa;
+       int packlen;                    /* total length of packet */
+       int pmtu;                       /* Path MTU Discovery (RFC1191) */
+       uint16_t ident;
+       uint16_t port; // 32768 + 666;  /* start udp dest port # for probe packets */
+       int waittime; // 5;             /* time to wait for response (in seconds) */
+#if ENABLE_FEATURE_TRACEROUTE_SOURCE_ROUTE
+       int optlen;                     /* length of ip options */
+#else
+#define optlen 0
 #endif
-
-
-#define MAXPACKET       65535   /* max ip packet size */
-#ifndef MAXHOSTNAMELEN
-#define MAXHOSTNAMELEN  64
+       unsigned char recv_pkt[512];    /* last inbound (icmp) packet */
+#if ENABLE_FEATURE_TRACEROUTE_SOURCE_ROUTE
+       /* Maximum number of gateways (include room for one noop) */
+#define NGATEWAYS ((int)((MAX_IPOPTLEN - IPOPT_MINOFF - 1) / sizeof(uint32_t)))
+       /* loose source route gateway list (including room for final destination) */
+       uint32_t gwlist[NGATEWAYS + 1];
 #endif
-
-/*
- * format of a (udp) probe packet.
- */
-struct opacket {
-       struct ip ip;
-       struct udphdr udp;
-       u_char seq;             /* sequence number of this packet */
-       u_char ttl;             /* ttl packet left with */
-       struct timeval tv;      /* time packet left */
 };
 
-/*
- * Definitions for internet protocol version 4.
- * Per RFC 791, September 1981.
- */
-#define IPVERSION       4
-
-
-#include "busybox.h"
-
-static u_char  packet[512];            /* last inbound (icmp) packet */
-static struct opacket  *outpacket;     /* last output (udp) packet */
-
-static int s;                          /* receive (icmp) socket file descriptor */
-static int sndsock;                    /* send (udp) socket file descriptor */
-
-static struct sockaddr whereto;        /* Who to try to reach */
-static int datalen;                    /* How much data */
+#define G (*ptr_to_globals)
+#define outip     (G.outip    )
+#define outdata   (G.outdata  )
+#define dest_lsa  (G.dest_lsa )
+#define packlen   (G.packlen  )
+#define pmtu      (G.pmtu     )
+#define ident     (G.ident    )
+#define port      (G.port     )
+#define waittime  (G.waittime )
+#if ENABLE_FEATURE_TRACEROUTE_SOURCE_ROUTE
+# define optlen   (G.optlen   )
+#endif
+#define recv_pkt  (G.recv_pkt )
+#define gwlist    (G.gwlist   )
+#define INIT_G() do { \
+       SET_PTR_TO_GLOBALS(xzalloc(sizeof(G))); \
+       port = 32768 + 666; \
+       waittime = 5; \
+} while (0)
 
-static char *hostname;
+#define outicmp ((struct icmp *)(outip + 1))
+#define outudp  ((struct udphdr *)(outip + 1))
 
-static int max_ttl = 30;
-static u_short ident;
-static u_short port = 32768+666;       /* start udp dest port # for probe packets */
 
-#ifdef CONFIG_FEATURE_TRACEROUTE_VERBOSE
-static int verbose;
-#endif
-static int waittime = 5;               /* time to wait for response (in seconds) */
-static int nflag;                      /* print addresses numerically */
+static int
+wait_for_reply(struct sockaddr_in *fromp)
+{
+       struct pollfd pfd[1];
+       int cc = 0;
+       socklen_t fromlen = sizeof(*fromp);
+
+       pfd[0].fd = rcvsock;
+       pfd[0].events = POLLIN;
+       if (safe_poll(pfd, 1, waittime * 1000) > 0)
+               cc = recvfrom(rcvsock, recv_pkt, sizeof(recv_pkt), 0,
+                           (struct sockaddr *)fromp, &fromlen);
+       return cc;
+}
 
 /*
- * Construct an Internet address representation.
- * If the nflag has been supplied, give
- * numeric value, otherwise try for symbolic name.
+ * Checksum routine for Internet Protocol family headers (C Version)
  */
-static inline void
-inetname(struct sockaddr_in *from)
+static uint16_t
+in_cksum(uint16_t *addr, int len)
 {
-       char *cp;
-       static char domain[MAXHOSTNAMELEN + 1];
-       char name[MAXHOSTNAMELEN + 1];
-       static int first = 1;
-       const char *ina;
-
-       if (first && !nflag) {
-               first = 0;
-               if (getdomainname(domain, MAXHOSTNAMELEN) != 0)
-                       domain[0] = 0;
-       }
-       cp = 0;
-       if (!nflag && from->sin_addr.s_addr != INADDR_ANY) {
-               if(INET_rresolve(name, sizeof(name), from, 0x4000, 0xffffffff) >= 0) {
-                       if ((cp = strchr(name, '.')) &&
-                           !strcmp(cp + 1, domain))
-                               *cp = 0;
-                       cp = (char *)name;
-               }
+       int nleft = len;
+       uint16_t *w = addr;
+       uint16_t answer;
+       int sum = 0;
+
+       /*
+        * Our algorithm is simple, using a 32 bit accumulator (sum),
+        * we add sequential 16 bit words to it, and at the end, fold
+        * back all the carry bits from the top 16 bits into the lower
+        * 16 bits.
+        */
+       while (nleft > 1) {
+               sum += *w++;
+               nleft -= 2;
        }
-       ina = inet_ntoa(from->sin_addr);
-       if (nflag)
-               printf(" %s", ina);
-       else
-               printf(" %s (%s)", (cp ? cp : ina), ina);
-}
-
-static inline void
-print(u_char *buf, int cc, struct sockaddr_in *from)
-{
-       struct ip *ip;
-       int hlen;
-
-       ip = (struct ip *) buf;
-       hlen = ip->ip_hl << 2;
-       cc -= hlen;
-
-       inetname(from);
-#ifdef CONFIG_FEATURE_TRACEROUTE_VERBOSE
-       if (verbose)
-               printf (" %d bytes to %s", cc, inet_ntoa (ip->ip_dst));
-#endif
-}
 
-static inline double
-deltaT(struct timeval *t1p, struct timeval *t2p)
-{
-       double dt;
+       /* mop up an odd byte, if necessary */
+       if (nleft == 1)
+               sum += *(unsigned char *)w;
 
-       dt = (double)(t2p->tv_sec - t1p->tv_sec) * 1000.0 +
-            (double)(t2p->tv_usec - t1p->tv_usec) / 1000.0;
-       return (dt);
+       /* add back carry outs from top 16 bits to low 16 bits */
+       sum = (sum >> 16) + (sum & 0xffff);     /* add hi 16 to low 16 */
+       sum += (sum >> 16);                     /* add carry */
+       answer = ~sum;                          /* truncate to 16 bits */
+       return answer;
 }
 
-static inline int
-wait_for_reply(int sock, struct sockaddr_in *from, int reset_timer)
+static void
+send_probe(int seq, int ttl)
 {
-       fd_set fds;
-       static struct timeval wait;
-       int cc = 0;
-       int fromlen = sizeof (*from);
+       int len, res;
+       void *out;
+
+       /* Payload */
+       outdata->seq = seq;
+       outdata->ttl = ttl;
+// UNUSED: was storing gettimeofday's result there, but never ever checked it
+       /*memcpy(&outdata->tv, tp, sizeof(outdata->tv));*/
+
+       if (option_mask32 & OPT_USE_ICMP) {
+               outicmp->icmp_seq = htons(seq);
+
+               /* Always calculate checksum for icmp packets */
+               outicmp->icmp_cksum = 0;
+               outicmp->icmp_cksum = in_cksum((uint16_t *)outicmp,
+                                       packlen - (sizeof(*outip) + optlen));
+               if (outicmp->icmp_cksum == 0)
+                       outicmp->icmp_cksum = 0xffff;
+       }
 
-       FD_ZERO(&fds);
-       FD_SET(sock, &fds);
-       if (reset_timer) {
-               /*
-                * traceroute could hang if someone else has a ping
-                * running and our ICMP reply gets dropped but we don't
-                * realize it because we keep waking up to handle those
-                * other ICMP packets that keep coming in.  To fix this,
-                * "reset_timer" will only be true if the last packet that
-                * came in was for us or if this is the first time we're
-                * waiting for a reply since sending out a probe.  Note
-                * that this takes advantage of the select() feature on
-                * Linux where the remaining timeout is written to the
-                * struct timeval area.
-                */
-               wait.tv_sec = waittime;
-               wait.tv_usec = 0;
+//BUG! verbose is (x & OPT_VERBOSE), not a counter!
+#if 0 //ENABLE_FEATURE_TRACEROUTE_VERBOSE
+       /* XXX undocumented debugging hack */
+       if (verbose > 1) {
+               const uint16_t *sp;
+               int nshorts, i;
+
+               sp = (uint16_t *)outip;
+               nshorts = (unsigned)packlen / sizeof(uint16_t);
+               i = 0;
+               printf("[ %d bytes", packlen);
+               while (--nshorts >= 0) {
+                       if ((i++ % 8) == 0)
+                               printf("\n\t");
+                       printf(" %04x", ntohs(*sp));
+                       sp++;
+               }
+               if (packlen & 1) {
+                       if ((i % 8) == 0)
+                               printf("\n\t");
+                       printf(" %02x", *(unsigned char *)sp);
+               }
+               printf("]\n");
        }
+#endif
 
-       if (select(sock+1, &fds, (fd_set *)0, (fd_set *)0, &wait) > 0)
-               cc=recvfrom(s, (char *)packet, sizeof(packet), 0,
-                           (struct sockaddr *)from, &fromlen);
+#if defined(IP_TTL)
+       if (setsockopt(sndsock, IPPROTO_IP, IP_TTL,
+                               (char *)&ttl, sizeof(ttl)) < 0) {
+               bb_perror_msg_and_die("setsockopt ttl %d", ttl);
+       }
+#endif
 
-       return(cc);
+       len = packlen - sizeof(*outip);
+       if (option_mask32 & OPT_USE_ICMP)
+               out = outicmp;
+       else {
+               out = outdata;
+               len -= sizeof(*outudp);
+               set_nport(dest_lsa, htons(port + seq));
+       }
+       res = xsendto(sndsock, out, len,
+                       (struct sockaddr *)&dest_lsa->u.sa, dest_lsa->len);
+       if (res != len) {
+               bb_info_msg("sent %d octets, ret=%d", len, res);
+       }
 }
 
-#ifdef CONFIG_FEATURE_TRACEROUTE_VERBOSE
+#if ENABLE_FEATURE_TRACEROUTE_VERBOSE
 /*
  * Convert an ICMP "type" field to a printable string.
  */
 static inline const char *
-pr_type(u_char t)
+pr_type(unsigned char t)
 {
-       static const char * const ttab[] = {
+       static const char *const ttab[] = {
        "Echo Reply",   "ICMP 1",       "ICMP 2",       "Dest Unreachable",
        "Source Quench", "Redirect",    "ICMP 6",       "ICMP 7",
-       "Echo",         "ICMP 9",       "ICMP 10",      "Time Exceeded",
+       "Echo",         "Router Advert", "Router Solicit", "Time Exceeded",
        "Param Problem", "Timestamp",   "Timestamp Reply", "Info Request",
-       "Info Reply"
+       "Info Reply",   "Mask Request", "Mask Reply"
        };
 
-       if(t > 16)
-               return("OUT-OF-RANGE");
+       if (t >= ARRAY_SIZE(ttab))
+               return "OUT-OF-RANGE";
 
-       return(ttab[t]);
+       return ttab[t];
 }
 #endif
 
-static inline int
-packet_ok(u_char *buf, int cc, struct sockaddr_in *from, int seq)
+#if !ENABLE_FEATURE_TRACEROUTE_VERBOSE
+#define packet_ok(cc, from, seq) \
+       packet_ok(cc, seq)
+#endif
+static int
+packet_ok(int cc, const struct sockaddr_in *from, int seq)
 {
-       struct icmp *icp;
-       u_char type, code;
+       const struct icmp *icp;
+       unsigned char type, code;
        int hlen;
-       struct ip *ip;
+       const struct ip *ip;
 
-       ip = (struct ip *) buf;
+       ip = (struct ip *) recv_pkt;
        hlen = ip->ip_hl << 2;
        if (cc < hlen + ICMP_MINLEN) {
-#ifdef CONFIG_FEATURE_TRACEROUTE_VERBOSE
+#if ENABLE_FEATURE_TRACEROUTE_VERBOSE
                if (verbose)
                        printf("packet too short (%d bytes) from %s\n", cc,
                                inet_ntoa(from->sin_addr));
 #endif
-               return (0);
+               return 0;
        }
        cc -= hlen;
-       icp = (struct icmp *)(buf + hlen);
-       type = icp->icmp_type; code = icp->icmp_code;
-       if ((type == ICMP_TIMXCEED && code == ICMP_TIMXCEED_INTRANS) ||
-           type == ICMP_UNREACH) {
-               struct ip *hip;
-               struct udphdr *up;
+       icp = (struct icmp *)(recv_pkt + hlen);
+       type = icp->icmp_type;
+       code = icp->icmp_code;
+       /* Path MTU Discovery (RFC1191) */
+       pmtu = 0;
+       if (code == ICMP_UNREACH_NEEDFRAG)
+               pmtu = ntohs(icp->icmp_nextmtu);
+
+       if ((type == ICMP_TIMXCEED && code == ICMP_TIMXCEED_INTRANS)
+        || type == ICMP_UNREACH
+        || type == ICMP_ECHOREPLY
+       ) {
+               const struct ip *hip;
+               const struct udphdr *up;
 
                hip = &icp->icmp_ip;
                hlen = hip->ip_hl << 2;
-               up = (struct udphdr *)((u_char *)hip + hlen);
-               if (hlen + 12 <= cc && hip->ip_p == IPPROTO_UDP &&
-                   up->source == htons(ident) &&
-                   up->dest == htons(port+seq))
-                       return (type == ICMP_TIMXCEED? -1 : code+1);
+               if (option_mask32 & OPT_USE_ICMP) {
+                       struct icmp *hicmp;
+
+                       /* XXX */
+                       if (type == ICMP_ECHOREPLY
+                        && icp->icmp_id == htons(ident)
+                        && icp->icmp_seq == htons(seq)
+                       ) {
+                               return -2;
+                       }
+
+                       hicmp = (struct icmp *)((unsigned char *)hip + hlen);
+                       if (hlen + SIZEOF_ICMP_HDR <= cc
+                        && hip->ip_p == IPPROTO_ICMP
+                        && hicmp->icmp_id == htons(ident)
+                        && hicmp->icmp_seq == htons(seq)
+                       ) {
+                               return (type == ICMP_TIMXCEED ? -1 : code + 1);
+                       }
+               } else {
+                       up = (struct udphdr *)((char *)hip + hlen);
+                       if (hlen + 12 <= cc
+                        && hip->ip_p == IPPROTO_UDP
+// Off: since we do not form the entire IP packet,
+// but defer it to kernel, we can't set source port,
+// and thus can't check it here in the reply
+                       /* && up->source == htons(ident) */
+                        && up->dest == htons(port + seq)
+                       ) {
+                               return (type == ICMP_TIMXCEED ? -1 : code + 1);
+                       }
+               }
        }
-#ifdef CONFIG_FEATURE_TRACEROUTE_VERBOSE
+#if ENABLE_FEATURE_TRACEROUTE_VERBOSE
        if (verbose) {
                int i;
-               u_long *lp = (u_long *)&icp->icmp_ip;
+               uint32_t *lp = (uint32_t *)&icp->icmp_ip;
 
-               printf("\n%d bytes from %s to %s: icmp type %d (%s) code %d\n",
-                       cc, inet_ntoa(from->sin_addr), inet_ntoa(ip->ip_dst),
+               printf("\n%d bytes from %s to "
+                      "%s: icmp type %d (%s) code %d\n",
+                       cc, inet_ntoa(from->sin_addr),
+                       inet_ntoa(ip->ip_dst),
                        type, pr_type(type), icp->icmp_code);
-               for (i = 4; i < cc ; i += sizeof(long))
-                       printf("%2d: x%8.8lx\n", i, *lp++);
+               for (i = 4; i < cc; i += sizeof(*lp))
+                       printf("%2d: x%8.8x\n", i, *lp++);
        }
 #endif
-       return(0);
+       return 0;
 }
 
-static void             /* not inline */
-send_probe(int seq, int ttl)
+/*
+ * Construct an Internet address representation.
+ * If the -n flag has been supplied, give
+ * numeric value, otherwise try for symbolic name.
+ */
+static void
+print_inetname(const struct sockaddr_in *from)
 {
-       struct opacket *op = outpacket;
-       struct ip *ip = &op->ip;
-       struct udphdr *up = &op->udp;
-       int i;
-       struct timezone tz;
-
-       ip->ip_off = 0;
-       ip->ip_hl = sizeof(*ip) >> 2;
-       ip->ip_p = IPPROTO_UDP;
-       ip->ip_len = datalen;
-       ip->ip_ttl = ttl;
-       ip->ip_v = IPVERSION;
-       ip->ip_id = htons(ident+seq);
-
-       up->source = htons(ident);
-       up->dest = htons(port+seq);
-       up->len = htons((u_short)(datalen - sizeof(struct ip)));
-       up->check = 0;
-
-       op->seq = seq;
-       op->ttl = ttl;
-       (void) gettimeofday(&op->tv, &tz);
-
-       i = sendto(sndsock, (char *)outpacket, datalen, 0, &whereto,
-                  sizeof(struct sockaddr));
-       if (i < 0 || i != datalen)  {
-               if (i<0)
-                       perror("sendto");
-               printf("traceroute: wrote %s %d chars, ret=%d\n", hostname,
-                       datalen, i);
-               (void) fflush(stdout);
+       const char *ina;
+
+       ina = inet_ntoa(from->sin_addr);
+       if (option_mask32 & OPT_ADDR_NUM)
+               printf("  %s", ina);
+       else {
+               char *n = NULL;
+               if (from->sin_addr.s_addr != INADDR_ANY)
+                       n = xmalloc_sockaddr2host_noport((struct sockaddr*)from);
+               printf("  %s (%s)", (n ? n : ina), ina);
+               free(n);
        }
 }
 
+static void
+print(int cc, const struct sockaddr_in *from)
+{
+       print_inetname(from);
+       if (verbose) {
+               const struct ip *ip;
+               int hlen;
+
+               ip = (struct ip *) recv_pkt;
+               hlen = ip->ip_hl << 2;
+               cc -= hlen;
+               printf(" %d bytes to %s", cc, inet_ntoa(ip->ip_dst));
+       }
+}
 
-int
-#ifndef CONFIG_TRACEROUTE
-main(int argc, char *argv[])
-#else
-traceroute_main(int argc, char *argv[])
-#endif
+static void
+print_delta_ms(unsigned t1p, unsigned t2p)
 {
-       extern char *optarg;
-       extern int optind;
-       struct hostent *hp;
-       struct sockaddr_in from, *to;
-       int ch, i, on, probe, seq, tos, ttl;
-
-       int options = 0;                /* socket options */
-       char *source = 0;
+       unsigned tt = t2p - t1p;
+       printf("  %u.%03u ms", tt / 1000, tt % 1000);
+}
+
+/*
+Usage: [-dFIlnrvx] [-g gateway] [-i iface] [-f first_ttl]
+[-m max_ttl] [ -p port] [-q nqueries] [-s src_addr] [-t tos]
+[-w waittime] [-z pausemsecs] host [packetlen]"
+*/
+
+int traceroute_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
+int traceroute_main(int argc, char **argv)
+{
+       int minpacket;
+       int ttl, i;
+       int seq = 0;
+       int tos = 0;
+       int max_ttl = 30;
        int nprobes = 3;
+       int first_ttl = 1;
+       unsigned pausemsecs = 0;
+       unsigned op;
+       char *source;
+       char *device;
+       char *tos_str;
+       char *max_ttl_str;
+       char *port_str;
+       char *nprobes_str;
+       char *waittime_str;
+       char *pausemsecs_str;
+       char *first_ttl_str;
+#if ENABLE_FEATURE_TRACEROUTE_SOURCE_ROUTE
+       llist_t *source_route_list = NULL;
+       int lsrr = 0;
+#endif
+
+       INIT_G();
 
-       on = 1;
-       seq = tos = 0;
-       to = (struct sockaddr_in *)&whereto;
-       while ((ch = getopt(argc, argv, "dm:np:q:rs:t:w:v")) != EOF)
-               switch(ch) {
-               case 'd':
-#ifdef CONFIG_FEATURE_TRACEROUTE_SO_DEBUG
-                       options |= SO_DEBUG;
+#if ENABLE_FEATURE_TRACEROUTE_SOURCE_ROUTE
+       opt_complementary = "x-x:g::";
+#else
+       opt_complementary = "x-x";
 #endif
-                       break;
-               case 'm':
-                       max_ttl = atoi(optarg);
-                       if (max_ttl <= 1)
-                               error_msg_and_die("max ttl must be >1.");
-                       break;
-               case 'n':
-                       nflag++;
-                       break;
-               case 'p':
-                       port = atoi(optarg);
-                       if (port < 1)
-                               error_msg_and_die("port must be >0.");
-                       break;
-               case 'q':
-                       nprobes = atoi(optarg);
-                       if (nprobes < 1)
-                               error_msg_and_die("nprobes must be >0.");
-                       break;
-               case 'r':
-                       options |= SO_DONTROUTE;
-                       break;
-               case 's':
-                       /*
-                        * set the ip source address of the outbound
-                        * probe (e.g., on a multi-homed host).
-                        */
-                       source = optarg;
-                       break;
-               case 't':
-                       tos = atoi(optarg);
-                       if (tos < 0 || tos > 255)
-                               error_msg_and_die("tos must be 0 to 255.");
-                       break;
-               case 'v':
-#ifdef CONFIG_FEATURE_TRACEROUTE_VERBOSE
-                       verbose++;
+
+       op = getopt32(argv, "FIlnrdvxt:i:m:p:q:s:w:z:f:"
+#if ENABLE_FEATURE_TRACEROUTE_SOURCE_ROUTE
+                                       "g:"
 #endif
-                       break;
-               case 'w':
-                       waittime = atoi(optarg);
-                       if (waittime <= 1)
-                               error_msg_and_die("wait must be >1 sec.");
-                       break;
-               default:
-                       show_usage();
+               , &tos_str, &device, &max_ttl_str, &port_str, &nprobes_str
+               , &source, &waittime_str, &pausemsecs_str, &first_ttl_str
+#if ENABLE_FEATURE_TRACEROUTE_SOURCE_ROUTE
+               , &source_route_list
+#endif
+       );
+
+#if 0 /* IGNORED */
+       if (op & OPT_IP_CHKSUM)
+               bb_error_msg("warning: ip checksums disabled");
+#endif
+       if (op & OPT_TOS)
+               tos = xatou_range(tos_str, 0, 255);
+       if (op & OPT_MAX_TTL)
+               max_ttl = xatou_range(max_ttl_str, 1, 255);
+       if (op & OPT_PORT)
+               port = xatou16(port_str);
+       if (op & OPT_NPROBES)
+               nprobes = xatou_range(nprobes_str, 1, INT_MAX);
+       if (op & OPT_SOURCE) {
+               /*
+                * set the ip source address of the outbound
+                * probe (e.g., on a multi-homed host).
+                */
+               if (getuid() != 0)
+                       bb_error_msg_and_die("you must be root to use -s");
+       }
+       if (op & OPT_WAITTIME)
+               waittime = xatou_range(waittime_str, 1, 24 * 60 * 60);
+       if (op & OPT_PAUSE_MS)
+               pausemsecs = xatou_range(pausemsecs_str, 0, 60 * 60 * 1000);
+       if (op & OPT_FIRST_TTL)
+               first_ttl = xatou_range(first_ttl_str, 1, max_ttl);
+
+#if ENABLE_FEATURE_TRACEROUTE_SOURCE_ROUTE
+       if (source_route_list) {
+               while (source_route_list) {
+                       len_and_sockaddr *lsa;
+
+                       if (lsrr >= NGATEWAYS)
+                               bb_error_msg_and_die("no more than %d gateways", NGATEWAYS);
+                       lsa = xhost_and_af2sockaddr(llist_pop(&source_route_list), 0, AF_INET);
+                       gwlist[lsrr] = lsa->u.sin.sin_addr.s_addr;
+                       free(lsa);
+                       ++lsrr;
                }
-       argc -= optind;
+               optlen = (lsrr + 1) * sizeof(gwlist[0]);
+       }
+#endif
+
+       minpacket = sizeof(*outip) + SIZEOF_ICMP_HDR + sizeof(*outdata) + optlen;
+       if (!(op & OPT_USE_ICMP))
+               minpacket += sizeof(*outudp) - SIZEOF_ICMP_HDR;
+       packlen = minpacket;
+
+       /* Process destination and optional packet size */
        argv += optind;
+       argc -= optind;
+       switch (argc) {
+       case 2:
+               packlen = xatoul_range(argv[1], minpacket, 32 * 1024);
+               /* Fall through */
+       case 1:
+               dest_lsa = xhost2sockaddr(argv[0], port);
+               break;
+       default:
+               bb_show_usage();
+       }
+
+       /* Ensure the socket fds won't be 0, 1 or 2 */
+       bb_sanitize_stdio();
 
-       if (argc < 1)
-               show_usage();
-
-       setlinebuf (stdout);
-
-       memset(&whereto, 0, sizeof(struct sockaddr));
-       hp = xgethostbyname(*argv);
-                       to->sin_family = hp->h_addrtype;
-       memcpy(&to->sin_addr, hp->h_addr, hp->h_length);
-                       hostname = (char *)hp->h_name;
-       if (*++argv)
-               datalen = atoi(*argv);
-       if (datalen < 0 || datalen >= MAXPACKET - sizeof(struct opacket))
-               error_msg_and_die("packet size must be 0 <= s < %d.",
-                   MAXPACKET - sizeof(struct opacket));
-       datalen += sizeof(struct opacket);
-       outpacket = (struct opacket *)xmalloc((unsigned)datalen);
-       memset(outpacket, 0, datalen);
-       outpacket->ip.ip_dst = to->sin_addr;
-       outpacket->ip.ip_tos = tos;
-       outpacket->ip.ip_v = IPVERSION;
-       outpacket->ip.ip_id = 0;
-
-       ident = (getpid() & 0xffff) | 0x8000;
-
-       if ((sndsock = socket(AF_INET, SOCK_RAW, IPPROTO_RAW)) < 0)
-               perror_msg_and_die(can_not_create_raw_socket);
-
-       s = create_icmp_socket();
-
-#ifdef CONFIG_FEATURE_TRACEROUTE_SO_DEBUG
-       if (options & SO_DEBUG)
-               (void) setsockopt(s, SOL_SOCKET, SO_DEBUG,
-                                 (char *)&on, sizeof(on));
+       xmove_fd(xsocket(AF_INET, SOCK_RAW, IPPROTO_ICMP), rcvsock);
+#if TRACEROUTE_SO_DEBUG
+       if (op & OPT_DEBUG)
+               setsockopt(rcvsock, SOL_SOCKET, SO_DEBUG,
+                               &const_int_1, sizeof(const_int_1));
 #endif
-       if (options & SO_DONTROUTE)
-               (void) setsockopt(s, SOL_SOCKET, SO_DONTROUTE,
-                                 (char *)&on, sizeof(on));
+       if (op & OPT_BYPASS_ROUTE)
+               setsockopt(rcvsock, SOL_SOCKET, SO_DONTROUTE,
+                               &const_int_1, sizeof(const_int_1));
+
+       if (op & OPT_USE_ICMP)
+               xmove_fd(xsocket(AF_INET, SOCK_RAW, IPPROTO_ICMP), sndsock);
+       else
+               xmove_fd(xsocket(AF_INET, SOCK_DGRAM, 0), sndsock);
+#if ENABLE_FEATURE_TRACEROUTE_SOURCE_ROUTE
+#if defined(IP_OPTIONS)
+       if (lsrr > 0) {
+               unsigned char optlist[MAX_IPOPTLEN];
+
+               /* final hop */
+               gwlist[lsrr] = dest_lsa->u.sin.sin_addr.s_addr;
+               ++lsrr;
+
+               /* force 4 byte alignment */
+               optlist[0] = IPOPT_NOP;
+               /* loose source route option */
+               optlist[1] = IPOPT_LSRR;
+               i = lsrr * sizeof(gwlist[0]);
+               optlist[2] = i + 3;
+               /* pointer to LSRR addresses */
+               optlist[3] = IPOPT_MINOFF;
+               memcpy(optlist + 4, gwlist, i);
+
+               if (setsockopt(sndsock, IPPROTO_IP, IP_OPTIONS,
+                               (char *)optlist, i + sizeof(gwlist[0])) < 0) {
+                       bb_perror_msg_and_die("IP_OPTIONS");
+               }
+       }
+#endif /* IP_OPTIONS */
+#endif /* CONFIG_FEATURE_TRACEROUTE_SOURCE_ROUTE */
 #ifdef SO_SNDBUF
-       if (setsockopt(sndsock, SOL_SOCKET, SO_SNDBUF, (char *)&datalen,
-                      sizeof(datalen)) < 0)
-               perror_msg_and_die("SO_SNDBUF");
+       if (setsockopt(sndsock, SOL_SOCKET, SO_SNDBUF, &packlen, sizeof(packlen)) < 0) {
+               bb_perror_msg_and_die("SO_SNDBUF");
+       }
 #endif
-#ifdef IP_HDRINCL
-       if (setsockopt(sndsock, IPPROTO_IP, IP_HDRINCL, (char *)&on,
-                      sizeof(on)) < 0)
-               perror_msg_and_die("IP_HDRINCL");
+#ifdef IP_TOS
+       if ((op & OPT_TOS) && setsockopt(sndsock, IPPROTO_IP, IP_TOS, &tos, sizeof(tos)) < 0) {
+               bb_perror_msg_and_die("setsockopt tos %d", tos);
+       }
 #endif
-#ifdef CONFIG_FEATURE_TRACEROUTE_SO_DEBUG
-       if (options & SO_DEBUG)
-               (void) setsockopt(sndsock, SOL_SOCKET, SO_DEBUG,
-                                 (char *)&on, sizeof(on));
+#ifdef IP_DONTFRAG
+       if (op & OPT_DONT_FRAGMNT)
+               setsockopt(sndsock, IPPROTO_IP, IP_DONTFRAG,
+                               &const_int_1, sizeof(const_int_1));
 #endif
-       if (options & SO_DONTROUTE)
-               (void) setsockopt(sndsock, SOL_SOCKET, SO_DONTROUTE,
-                                 (char *)&on, sizeof(on));
-
-       if (source) {
-               memset(&from, 0, sizeof(struct sockaddr));
-               from.sin_family = AF_INET;
-               from.sin_addr.s_addr = inet_addr(source);
-               if (from.sin_addr.s_addr == -1)
-                       error_msg_and_die("unknown host %s", source);
-               outpacket->ip.ip_src = from.sin_addr;
-#ifndef IP_HDRINCL
-               if (bind(sndsock, (struct sockaddr *)&from, sizeof(from)) < 0)
-                       perror_msg_and_die("bind");
+#if TRACEROUTE_SO_DEBUG
+       if (op & OPT_DEBUG)
+               setsockopt(sndsock, SOL_SOCKET, SO_DEBUG,
+                               &const_int_1, sizeof(const_int_1));
 #endif
+       if (op & OPT_BYPASS_ROUTE)
+               setsockopt(sndsock, SOL_SOCKET, SO_DONTROUTE,
+                               &const_int_1, sizeof(const_int_1));
+
+       outip = xzalloc(packlen);
+
+       if (op & OPT_USE_ICMP) {
+               ident = getpid() | 0x8000;
+               outicmp->icmp_type = ICMP_ECHO;
+               outicmp->icmp_id = htons(ident);
+               outdata = (struct outdata_t *)((char *)outicmp + SIZEOF_ICMP_HDR);
+       } else {
+               outdata = (struct outdata_t *)(outudp + 1);
        }
 
-       fprintf(stderr, "traceroute to %s (%s)", hostname,
-               inet_ntoa(to->sin_addr));
-       if (source)
-               fprintf(stderr, " from %s", source);
-       fprintf(stderr, ", %d hops max, %d byte packets\n", max_ttl, datalen);
+       if (op & OPT_DEVICE) /* hmm, do we need error check? */
+               setsockopt_bindtodevice(sndsock, device);
+
+       if (op & OPT_SOURCE) {
+               len_and_sockaddr *source_lsa = xdotted2sockaddr(source, 0);
+               /* Ping does this (why?) */
+               if (setsockopt(sndsock, IPPROTO_IP, IP_MULTICAST_IF,
+                               &source_lsa->u.sa, source_lsa->len))
+                       bb_error_msg_and_die("can't set multicast source interface");
+//TODO: we can query source port we bound to,
+// and check it in replies... if we care enough
+               xbind(sndsock, &source_lsa->u.sa, source_lsa->len);
+               free(source_lsa);
+       }
 
-       for (ttl = 1; ttl <= max_ttl; ++ttl) {
-               u_long lastaddr = 0;
+       /* Revert to non-privileged user after opening sockets */
+       xsetgid(getgid());
+       xsetuid(getuid());
+
+       printf("traceroute to %s (%s)", argv[0],
+                       xmalloc_sockaddr2dotted_noport(&dest_lsa->u.sa));
+       if (op & OPT_SOURCE)
+               printf(" from %s", source);
+       printf(", %d hops max, %d byte packets\n", max_ttl, packlen);
+
+       for (ttl = first_ttl; ttl <= max_ttl; ++ttl) {
+//TODO: make it protocol agnostic (get rid of sockaddr_in)
+               struct sockaddr_in from;
+               uint32_t lastaddr = 0;
+               int probe;
+               int unreachable = 0; /* counter */
+               int gotlastaddr = 0; /* flags */
                int got_there = 0;
-               int unreachable = 0;
+               int first = 1;
 
-               printf("%2d ", ttl);
+               printf("%2d", ttl);
                for (probe = 0; probe < nprobes; ++probe) {
-                       int cc, reset_timer;
-                       struct timeval t1, t2;
-                       struct timezone tz;
+                       int cc;
+                       unsigned t1;
+                       unsigned t2;
                        struct ip *ip;
 
-                       (void) gettimeofday(&t1, &tz);
+                       if (!first && pausemsecs > 0)
+                               usleep(pausemsecs * 1000);
+                       fflush(stdout);
+
+                       t1 = monotonic_us();
                        send_probe(++seq, ttl);
-                       reset_timer = 1;
-                       while ((cc = wait_for_reply(s, &from, reset_timer)) != 0) {
-                               (void) gettimeofday(&t2, &tz);
-                               if ((i = packet_ok(packet, cc, &from, seq))) {
-                                       reset_timer = 1;
-                                       if (from.sin_addr.s_addr != lastaddr) {
-                                               print(packet, cc, &from);
-                                               lastaddr = from.sin_addr.s_addr;
-                                       }
-                                       printf("  %g ms", deltaT(&t1, &t2));
-                                       switch(i - 1) {
-                                       case ICMP_UNREACH_PORT:
-                                               ip = (struct ip *)packet;
-                                               if (ip->ip_ttl <= 1)
-                                                       printf(" !");
-                                               ++got_there;
-                                               break;
-                                       case ICMP_UNREACH_NET:
-                                               ++unreachable;
-                                               printf(" !N");
-                                               break;
-                                       case ICMP_UNREACH_HOST:
-                                               ++unreachable;
-                                               printf(" !H");
-                                               break;
-                                       case ICMP_UNREACH_PROTOCOL:
-                                               ++got_there;
-                                               printf(" !P");
-                                               break;
-                                       case ICMP_UNREACH_NEEDFRAG:
-                                               ++unreachable;
-                                               printf(" !F");
-                                               break;
-                                       case ICMP_UNREACH_SRCFAIL:
-                                               ++unreachable;
-                                               printf(" !S");
-                                               break;
-                                       }
+                       first = 0;
+
+                       while ((cc = wait_for_reply(&from)) != 0) {
+                               t2 = monotonic_us();
+                               i = packet_ok(cc, &from, seq);
+                               /* Skip short packet */
+                               if (i == 0)
+                                       continue;
+                               if (!gotlastaddr
+                                || from.sin_addr.s_addr != lastaddr
+                               ) {
+                                       print(cc, &from);
+                                       lastaddr = from.sin_addr.s_addr;
+                                       gotlastaddr = 1;
+                               }
+                               print_delta_ms(t1, t2);
+                               ip = (struct ip *)recv_pkt;
+                               if (op & OPT_TTL_FLAG)
+                                       printf(" (%d)", ip->ip_ttl);
+                               if (i == -2) {
+                                       if (ip->ip_ttl <= 1)
+                                               printf(" !");
+                                       got_there = 1;
+                                       break;
+                               }
+                               /* time exceeded in transit */
+                               if (i == -1)
+                                       break;
+                               i--;
+                               switch (i) {
+                               case ICMP_UNREACH_PORT:
+                                       if (ip->ip_ttl <= 1)
+                                               printf(" !");
+                                       got_there = 1;
+                                       break;
+                               case ICMP_UNREACH_NET:
+                                       printf(" !N");
+                                       ++unreachable;
+                                       break;
+                               case ICMP_UNREACH_HOST:
+                                       printf(" !H");
+                                       ++unreachable;
+                                       break;
+                               case ICMP_UNREACH_PROTOCOL:
+                                       printf(" !P");
+                                       got_there = 1;
+                                       break;
+                               case ICMP_UNREACH_NEEDFRAG:
+                                       printf(" !F-%d", pmtu);
+                                       ++unreachable;
+                                       break;
+                               case ICMP_UNREACH_SRCFAIL:
+                                       printf(" !S");
+                                       ++unreachable;
+                                       break;
+                               case ICMP_UNREACH_FILTER_PROHIB:
+                               case ICMP_UNREACH_NET_PROHIB:   /* misuse */
+                                       printf(" !A");
+                                       ++unreachable;
+                                       break;
+                               case ICMP_UNREACH_HOST_PROHIB:
+                                       printf(" !C");
+                                       ++unreachable;
                                        break;
-                               } else
-                                       reset_timer = 0;
+                               case ICMP_UNREACH_HOST_PRECEDENCE:
+                                       printf(" !V");
+                                       ++unreachable;
+                                       break;
+                               case ICMP_UNREACH_PRECEDENCE_CUTOFF:
+                                       printf(" !C");
+                                       ++unreachable;
+                                       break;
+                               case ICMP_UNREACH_NET_UNKNOWN:
+                               case ICMP_UNREACH_HOST_UNKNOWN:
+                                       printf(" !U");
+                                       ++unreachable;
+                                       break;
+                               case ICMP_UNREACH_ISOLATED:
+                                       printf(" !I");
+                                       ++unreachable;
+                                       break;
+                               case ICMP_UNREACH_TOSNET:
+                               case ICMP_UNREACH_TOSHOST:
+                                       printf(" !T");
+                                       ++unreachable;
+                                       break;
+                               default:
+                                       printf(" !<%d>", i);
+                                       ++unreachable;
+                                       break;
+                               }
+                               break;
                        }
                        if (cc == 0)
-                               printf(" *");
-                       (void) fflush(stdout);
+                               printf("  *");
+               }
+               bb_putchar('\n');
+               if (got_there
+                || (unreachable > 0 && unreachable >= nprobes - 1)
+               ) {
+                       break;
                }
-               putchar('\n');
-               if (got_there || unreachable >= nprobes-1)
-                       return 0;
        }
-
        return 0;
 }