xref: /titanic_50/usr/src/cmd/cmd-inet/usr.sbin/traceroute/traceroute.c (revision 7b0bedd42192a2f6bcd6fc4b637d23892303a962)
1 /*
2  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
3  * Use is subject to license terms.
4  */
5 
6 /*
7  * Copyright (c) 1988, 1989, 1991, 1994, 1995, 1996, 1997
8  *	The Regents of the University of California.  All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that: (1) source code distributions
12  * retain the above copyright notice and this paragraph in its entirety, (2)
13  * distributions including binary code include the above copyright notice and
14  * this paragraph in its entirety in the documentation or other materials
15  * provided with the distribution, and (3) all advertising materials mentioning
16  * features or use of this software display the following acknowledgement:
17  * ``This product includes software developed by the University of California,
18  * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of
19  * the University nor the names of its contributors may be used to endorse
20  * or promote products derived from this software without specific prior
21  * written permission.
22  * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED
23  * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
24  * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
25  *
26  *
27  * @(#)$Header: traceroute.c,v 1.49 97/06/13 02:30:23 leres Exp $ (LBL)
28  */
29 
30 #include <sys/param.h>
31 #include <sys/file.h>
32 #include <sys/ioctl.h>
33 #include <sys/socket.h>
34 #include <sys/time.h>
35 #include <sys/sysmacros.h>
36 
37 #include <netinet/in_systm.h>
38 #include <netinet/in.h>
39 #include <netinet/ip.h>
40 #include <netinet/ip_var.h>
41 #include <netinet/ip_icmp.h>
42 #include <netinet/udp.h>
43 #include <netinet/udp_var.h>
44 #include <netinet/ip6.h>
45 #include <netinet/icmp6.h>
46 
47 #include <arpa/inet.h>
48 
49 #include <ctype.h>
50 #include <errno.h>
51 #include <malloc.h>
52 #include <memory.h>
53 #include <netdb.h>
54 #include <stdio.h>
55 #include <stdlib.h>
56 #include <strings.h>
57 #include <unistd.h>
58 #include <libintl.h>
59 #include <locale.h>
60 #include <signal.h>
61 #include <setjmp.h>
62 #include <limits.h>
63 #include <zone.h>
64 
65 #include <priv_utils.h>
66 
67 #include <libinetutil.h>
68 #include "traceroute.h"
69 
70 #define	MAX_SEQ			65535	/* max sequence value for ICMP */
71 #define	MAX_TRAFFIC_CLASS	255	/* max traffic class for IPv6 */
72 #define	MAX_FLOW_LABEL		0xFFFFF	/* max flow label for IPv6 */
73 #define	MAX_TOS			255	/* max type-of-service for IPv4 */
74 #define	STR_LEN			30
75 
76 /* store the information about a host */
77 struct hostinfo {
78 	char *name;		/* hostname */
79 	int family;		/* address family of the IP addresses */
80 	int num_addr;			/* number of IP addresses */
81 	union any_in_addr *addrs;	/* list of IP addresses */
82 };
83 
84 /* used to store a bunch of protocol specific values */
85 struct pr_set {
86 	int family;		/* AF_INET or AF_INET6 */
87 	char name[STR_LEN];	/* "IPv4" or "IPv6" */
88 	char icmp[STR_LEN];	/* "icmp" or "ipv6-icmp" */
89 	int icmp_minlen;
90 	int addr_len;
91 	int ip_hdr_len;
92 	int packlen;
93 	int sock_size;		/* size of sockaddr_in or sockaddr_in6 */
94 	struct sockaddr *to;
95 	struct sockaddr *from;
96 	void *from_sin_addr;
97 	union any_in_addr *gwIPlist;
98 	/* pointers to v4/v6 functions */
99 	struct ip *(*set_buffers_fn) (int);
100 	int (*check_reply_fn)(struct msghdr *, int, int, uchar_t *, uchar_t *);
101 	boolean_t (*print_icmp_other_fn)(uchar_t, uchar_t);
102 	void (*print_addr_fn)(uchar_t *, int, struct sockaddr *);
103 
104 };
105 
106 /*
107  * LBNL bug fixed: in LBNL traceroute 'uchar_t packet[512];'
108  * Not sufficient to hold the complete packet for ECHO REPLY of a big probe.
109  * Packet size is reported incorrectly in such a case.
110  * Also this buffer needs to be 32 bit aligned. In the future the alignment
111  * requirement will be increased to 64 bit. So, let's use 64 bit alignment now.
112  */
113 static uint64_t packet[(IP_MAXPACKET + 1)/8];	/* received packet */
114 
115 static struct ip *outip4;	/* output buffer to send as an IPv4 datagram */
116 static struct ip *outip6;	/* output buffer to send as an IPv6 datagram */
117 
118 /* Used to store the ancillary data that comes with the received packets */
119 static uint64_t ancillary_data[(IP_MAXPACKET + 1)/8];
120 
121 /* first get the gw names, later you'll resolve them based on the family */
122 static char *gwlist[MAXMAX_GWS];		/* gateway names list */
123 static union any_in_addr gwIPlist[MAX_GWS];	/* gateway IPv4 address list */
124 static union any_in_addr gwIP6list[MAX_GWS6];	/* gateway IPv6 address list */
125 
126 static int family_input = AF_UNSPEC;	/* User supplied protocol family */
127 static int rcvsock4;		/* receive (icmp) socket file descriptor */
128 static int sndsock4;		/* send (udp/icmp) socket file descriptor */
129 static int rcvsock6;		/* receive (icmp6) socket file descriptor */
130 static int sndsock6;		/* send (udp6/icmp6) socket file descriptor */
131 int gw_count = 0;		/* number of gateways */
132 static struct sockaddr_in whereto;	/* Who to try to reach */
133 static struct sockaddr_in6 whereto6;
134 static struct sockaddr_in wherefrom;	/* Who we are */
135 static struct sockaddr_in6 wherefrom6;
136 static int packlen_input = 0;		/* user input for packlen */
137 
138 char *prog;
139 static char *source_input = NULL; /* this is user arg. source, doesn't change */
140 static char *source = NULL;	/* this gets modified after name lookup */
141 char *hostname;
142 static char *device = NULL;   	/* interface name */
143 static struct pr_set *pr4;	/* protocol info for IPv4 */
144 static struct pr_set *pr6;	/* protocol info for IPv6 */
145 static struct ifaddrlist *al4;	/* list of interfaces */
146 static struct ifaddrlist *al6;	/* list of interfaces */
147 static uint_t if_index = 0;	/* interface index */
148 static int num_v4 = 0;		/* count of IPv4 addresses */
149 static int num_v6 = 0;		/* count of IPv6 addresses */
150 static int num_ifs4 = 0;	/* count of local IPv4 interfaces */
151 static int num_ifs6 = 0;	/* count of local IPv6 interfaces */
152 
153 static int nprobes = 3;		/* number of probes */
154 static int max_ttl = 30;	/* max number of hops */
155 static int first_ttl = 1;	/* initial number of hops */
156 ushort_t ident;			/* used to authenticate replies */
157 ushort_t port = 32768 + 666;	/* start udp dest port # for probe packets */
158 
159 static int options = 0;		/* socket options */
160 boolean_t verbose = _B_FALSE;	/* verbose output */
161 static int waittime = 5;	/* time to wait for response (in seconds) */
162 static struct timeval delay = {0, 0}; /* delay between consecutive probe */
163 boolean_t nflag = _B_FALSE;	/* print addresses numerically */
164 static boolean_t showttl = _B_FALSE; /* print the ttl(hop limit) of recvd pkt */
165 boolean_t useicmp = _B_FALSE;  	/* use icmp echo instead of udp packets */
166 boolean_t docksum = _B_TRUE;	/* calculate checksums */
167 static boolean_t collect_stat = _B_FALSE;	/* print statistics */
168 boolean_t settos = _B_FALSE;   	/* set type-of-service field */
169 int dontfrag = 0;		/* IP*_DONTFRAG */
170 static int max_timeout = 5;	/* quit after this consecutive timeouts */
171 static boolean_t probe_all = _B_FALSE;	/* probe all the IFs of the target */
172 static boolean_t pick_src = _B_FALSE;	/* traceroute picks the src address */
173 
174 /*
175  * flow and class are specific to IPv6, tos and off are specific to IPv4.
176  * Each protocol uses the ones that are specific to itself, and ignores
177  * others.
178  */
179 static uint_t flow = 0;		/* IPv6 flow info */
180 static uint_t class = 0;	/* IPv6 class */
181 uchar_t tos = 0;		/* IPv4 type-of-service */
182 ushort_t off = 0;		/* set DF bit */
183 
184 static jmp_buf env;		/* stack environment for longjmp() */
185 boolean_t raw_req;		/* if sndsock for IPv4 must be raw */
186 
187 /* Forwards */
188 static uint_t calc_packetlen(int, struct pr_set *);
189 extern int check_reply(struct msghdr *, int, int, uchar_t *, uchar_t *);
190 extern int check_reply6(struct msghdr *, int, int, uchar_t *, uchar_t *);
191 static double deltaT(struct timeval *, struct timeval *);
192 static char *device_name(struct ifaddrlist *, int, union any_in_addr *,
193     struct pr_set *);
194 extern void *find_ancillary_data(struct msghdr *, int, int);
195 static boolean_t has_addr(struct addrinfo *, union any_in_addr *);
196 static struct ifaddrlist *find_device(struct ifaddrlist *, int, char *);
197 static struct ifaddrlist *find_ifaddr(struct ifaddrlist *, int,
198     union any_in_addr *, int);
199 static void get_gwaddrs(char **, int, union any_in_addr *,
200     union any_in_addr *, int *, int *);
201 static void get_hostinfo(char *, int, struct addrinfo **);
202 char *inet_name(union any_in_addr *, int);
203 ushort_t in_cksum(ushort_t *, int);
204 extern int ip_hdr_length_v6(ip6_t *, int, uint8_t *);
205 extern char *pr_type(uchar_t);
206 extern char *pr_type6(uchar_t);
207 extern void print_addr(uchar_t *, int, struct sockaddr *);
208 extern void print_addr6(uchar_t *, int, struct sockaddr *);
209 extern boolean_t print_icmp_other(uchar_t, uchar_t);
210 extern boolean_t print_icmp_other6(uchar_t, uchar_t);
211 static void print_stats(int, int, double, double, double, double);
212 static void print_unknown_host_msg(const char *, const char *);
213 static void record_stats(double, int *, double *, double *, double *, double *);
214 static void resolve_nodes(int *, struct addrinfo **);
215 static void select_src_addr(union any_in_addr *, union any_in_addr *, int);
216 extern void send_probe(int, struct sockaddr *, struct ip *, int, int,
217     struct timeval *, int);
218 extern void send_probe6(int, struct msghdr *, struct ip *, int, int,
219     struct timeval *, int);
220 extern void set_ancillary_data(struct msghdr *, int, union any_in_addr *, int,
221     uint_t);
222 extern struct ip *set_buffers(int);
223 extern struct ip *set_buffers6(int);
224 extern void set_IPv4opt_sourcerouting(int, union any_in_addr *,
225     union any_in_addr *);
226 static void set_sin(struct sockaddr *, union any_in_addr *, int);
227 static int set_src_addr(struct pr_set *, struct ifaddrlist **);
228 static void setup_protocol(struct pr_set *, int);
229 static void setup_socket(struct pr_set *, int);
230 static void sig_handler(int);
231 static int str2int(const char *, const char *, int, int);
232 static double str2dbl(const char *, const char *, double, double);
233 static void trace_it(struct addrinfo *);
234 static void traceroute(union any_in_addr *, struct msghdr *, struct pr_set *,
235     int, struct ifaddrlist *);
236 static void tv_sub(struct timeval *, struct timeval *);
237 static void usage(void);
238 static int wait_for_reply(int, struct msghdr *, struct timeval *);
239 static double xsqrt(double);
240 
241 /*
242  * main
243  */
244 int
245 main(int argc, char **argv)
246 {
247 	struct addrinfo *ai_dst = NULL;		/* destination host */
248 	/*
249 	 * "probing_successful" indicates if we could successfully send probes,
250 	 * not necessarily received reply from the target (this behavior is from
251 	 * the original traceroute). It's _B_FALSE if packlen is invalid, or no
252 	 * interfaces found.
253 	 */
254 	boolean_t probing_successful = _B_FALSE;
255 	int longjmp_return;			/* return value from longjump */
256 	int i = 0;
257 	char *cp;
258 	int op;
259 	char *ep;
260 	char temp_buf[INET6_ADDRSTRLEN];	/* use for inet_ntop() */
261 	double pause;
262 
263 	/*
264 	 * A raw socket will be used for IPv4 if there is sufficient
265 	 * privilege.
266 	 */
267 	raw_req = priv_ineffect(PRIV_NET_RAWACCESS);
268 
269 	/*
270 	 * We'll need the privilege only when we open the sockets; that's
271 	 * when we'll fail if the program has insufficient privileges.
272 	 */
273 	(void) __init_suid_priv(PU_CLEARLIMITSET, PRIV_NET_ICMPACCESS,
274 	    raw_req ? PRIV_NET_RAWACCESS : NULL, NULL);
275 
276 	(void) setlinebuf(stdout);
277 
278 	if ((cp = strrchr(argv[0], '/')) != NULL)
279 		prog = cp + 1;
280 	else
281 		prog = argv[0];
282 
283 	opterr = 0;
284 	while ((op = getopt(argc, argv, "adFIlnrSvxA:c:f:g:i:L:m:P:p:Q:q:s:"
285 	    "t:w:")) != EOF) {
286 		switch (op) {
287 		case 'A':
288 			if (strcmp(optarg, "inet") == 0) {
289 				family_input = AF_INET;
290 			} else if (strcmp(optarg, "inet6") == 0) {
291 				family_input = AF_INET6;
292 			} else {
293 				Fprintf(stderr,
294 				    "%s: unknown address family %s\n",
295 				    prog, optarg);
296 				exit(EXIT_FAILURE);
297 			}
298 			break;
299 
300 		case 'a':
301 			probe_all = _B_TRUE;
302 			break;
303 
304 		case 'c':
305 			class = str2int(optarg, "traffic class", 0,
306 			    MAX_TRAFFIC_CLASS);
307 			break;
308 
309 		case 'd':
310 			options |= SO_DEBUG;
311 			break;
312 
313 		case 'f':
314 			first_ttl = str2int(optarg, "first ttl", 1, MAXTTL);
315 			break;
316 
317 		case 'F':
318 			off = IP_DF;
319 			dontfrag = 1;
320 			break;
321 
322 		case 'g':
323 			if (!raw_req) {
324 				Fprintf(stderr,
325 				    "%s: privilege to specify a loose source "
326 				    "route gateway is unavailable\n",
327 				    prog);
328 				exit(EXIT_FAILURE);
329 			}
330 			if (gw_count >= MAXMAX_GWS) {
331 				Fprintf(stderr,
332 				    "%s: Too many gateways\n", prog);
333 				exit(EXIT_FAILURE);
334 			}
335 			gwlist[gw_count] = strdup(optarg);
336 			if (gwlist[gw_count] == NULL) {
337 				Fprintf(stderr, "%s: strdup %s\n", prog,
338 				    strerror(errno));
339 				exit(EXIT_FAILURE);
340 			}
341 
342 			++gw_count;
343 			break;
344 
345 		case 'l':
346 			showttl = _B_TRUE;
347 			break;
348 
349 		case 'i':
350 			/* this can be IF name or IF index */
351 			if_index = (uint_t)strtol(optarg, &ep, 10);
352 
353 			/* convert IF index <-->  IF name */
354 			if (errno != 0 || *ep != '\0') {
355 				device = optarg;
356 				if_index = if_nametoindex((const char *)device);
357 
358 				/*
359 				 * In case it fails, check to see if the problem
360 				 * is other than "IF not found".
361 				 */
362 				if (if_index == 0 && errno != ENXIO) {
363 					Fprintf(stderr, "%s: if_nametoindex:"
364 					    "%s\n", prog, strerror(errno));
365 					exit(EXIT_FAILURE);
366 				}
367 			} else {
368 				device = (char *)malloc(LIFNAMSIZ + 1);
369 				if (device == NULL) {
370 					Fprintf(stderr, "%s: malloc: %s\n",
371 					    prog, strerror(errno));
372 					exit(EXIT_FAILURE);
373 				}
374 
375 				device = if_indextoname(if_index, device);
376 				if (device != NULL) {
377 					device[LIFNAMSIZ] = '\0';
378 				} else if (errno != ENXIO) {
379 					/*
380 					 * The problem was other than "index
381 					 * not found".
382 					 */
383 					Fprintf(stderr, "%s: if_indextoname:"
384 					    "%s\n", prog, strerror(errno));
385 					exit(EXIT_FAILURE);
386 				}
387 			}
388 
389 			if (device == NULL || if_index == 0) {
390 				Fprintf(stderr, "%s: interface %s "
391 				    "doesn't match any actual interfaces\n",
392 				    prog, optarg);
393 				exit(EXIT_FAILURE);
394 			}
395 			break;
396 
397 		case 'I':
398 			useicmp = _B_TRUE;
399 			break;
400 
401 		case 'L':
402 			flow = str2int(optarg, "flow label", 0, MAX_FLOW_LABEL);
403 			break;
404 
405 		case 'm':
406 			max_ttl = str2int(optarg, "max ttl(hop limit)", 1,
407 			    MAXTTL);
408 			break;
409 
410 		case 'n':
411 			nflag = _B_TRUE;
412 			break;
413 
414 		case 'P':
415 			pause = str2dbl(optarg, "pause", 0, INT_MAX);
416 			delay.tv_sec = (time_t)pause;
417 			delay.tv_usec = (suseconds_t)((pause - delay.tv_sec) *
418 			    1000000);
419 			break;
420 
421 		case 'p':
422 			port = str2int(optarg, "port", 1, MAX_PORT);
423 			break;
424 
425 		case 'Q':
426 			max_timeout = str2int(optarg, "max timeout", 1, -1);
427 			break;
428 
429 		case 'q':
430 			nprobes = str2int(optarg, "nprobes", 1, -1);
431 			break;
432 
433 		case 'r':
434 			options |= SO_DONTROUTE;
435 			break;
436 
437 		case 'S':
438 			collect_stat = _B_TRUE;
439 			break;
440 
441 		case 's':
442 			/*
443 			 * set the ip source address of the outbound
444 			 * probe (e.g., on a multi-homed host).
445 			 */
446 			source_input = optarg;
447 			break;
448 
449 		case 't':
450 			tos = (uchar_t)str2int(optarg, "tos", 0, MAX_TOS);
451 			settos = _B_TRUE;
452 			break;
453 
454 		case 'v':
455 			verbose = _B_TRUE;
456 			break;
457 
458 		case 'x':
459 			docksum = _B_FALSE;
460 			break;
461 
462 		case 'w':
463 			waittime = str2int(optarg, "wait time", 2, -1);
464 			break;
465 
466 		default:
467 			usage();
468 			break;
469 		}
470 	}
471 
472 	/*
473 	 * If it's probe_all, SIGQUIT makes traceroute exit(). But we set the
474 	 * address to jump back to in traceroute(). Until then, we'll need to
475 	 * temporarily specify one.
476 	 */
477 	if (probe_all) {
478 		if ((longjmp_return = setjmp(env)) != 0) {
479 			if (longjmp_return == SIGQUIT) {
480 				Printf("(exiting)\n");
481 				exit(EXIT_SUCCESS);
482 			} else {		/* should never happen */
483 				exit(EXIT_FAILURE);
484 			}
485 		}
486 		(void) signal(SIGQUIT, sig_handler);
487 	}
488 
489 	if ((gw_count > 0) && (options & SO_DONTROUTE)) {
490 		Fprintf(stderr, "%s: loose source route gateways (-g)"
491 		    " cannot be specified when probe packets are sent"
492 		    " directly to a host on an attached network (-r)\n",
493 		    prog);
494 		exit(EXIT_FAILURE);
495 	}
496 
497 	i = argc - optind;
498 	if (i == 1 || i == 2) {
499 		hostname = argv[optind];
500 
501 		if (i == 2) {
502 			/* accept any length now, we'll check it later */
503 			packlen_input = str2int(argv[optind + 1],
504 			    "packet length", 0, -1);
505 		}
506 	} else {
507 		usage();
508 	}
509 
510 	if (first_ttl > max_ttl) {
511 		Fprintf(stderr,
512 		    "%s: first ttl(hop limit) (%d) may not be greater"
513 		    " than max ttl(hop limit) (%d)\n",
514 		    prog, first_ttl, max_ttl);
515 		exit(EXIT_FAILURE);
516 	}
517 
518 	/* resolve hostnames */
519 	resolve_nodes(&family_input, &ai_dst);
520 	if (ai_dst == NULL) {
521 		exit(EXIT_FAILURE);
522 	}
523 
524 	/*
525 	 * If it's probe_all, SIGINT makes traceroute skip to probing next IP
526 	 * address of the target. The new interrupt handler is assigned in
527 	 * traceroute() function. Until then let's ignore the signal.
528 	 */
529 	if (probe_all)
530 		(void) signal(SIGINT, SIG_IGN);
531 
532 	ident = (getpid() & 0xffff) | 0x8000;
533 
534 	/*
535 	 * We KNOW that probe_all == TRUE if family is AF_UNSPEC,
536 	 * since family is set to the specific AF found unless it's
537 	 * probe_all. So if family == AF_UNSPEC, we need to init pr4 and pr6.
538 	 */
539 	switch (family_input) {
540 	case AF_UNSPEC:
541 		pr4 = (struct pr_set *)malloc(sizeof (struct pr_set));
542 		if (pr4 == NULL) {
543 			Fprintf(stderr,
544 			    "%s: malloc %s\n", prog, strerror(errno));
545 			exit(EXIT_FAILURE);
546 		}
547 		pr6 = (struct pr_set *)malloc(sizeof (struct pr_set));
548 		if (pr6 == NULL) {
549 			Fprintf(stderr,
550 			    "%s: malloc %s\n", prog, strerror(errno));
551 			exit(EXIT_FAILURE);
552 		}
553 		setup_protocol(pr6, AF_INET6);
554 		setup_protocol(pr4, AF_INET);
555 		outip6 = (*pr6->set_buffers_fn)(pr6->packlen);
556 		setup_socket(pr6, pr6->packlen);
557 
558 		outip4 = (*pr4->set_buffers_fn)(pr4->packlen);
559 		setup_socket(pr4, pr4->packlen);
560 		num_ifs6 = set_src_addr(pr6, &al6);
561 		num_ifs4 = set_src_addr(pr4, &al4);
562 		break;
563 	case AF_INET6:
564 		pr6 = (struct pr_set *)malloc(sizeof (struct pr_set));
565 		if (pr6 == NULL) {
566 			Fprintf(stderr,
567 			    "%s: malloc %s\n", prog, strerror(errno));
568 			exit(EXIT_FAILURE);
569 		}
570 		setup_protocol(pr6, AF_INET6);
571 		outip6 = (*pr6->set_buffers_fn)(pr6->packlen);
572 		setup_socket(pr6, pr6->packlen);
573 		num_ifs6 = set_src_addr(pr6, &al6);
574 		break;
575 	case AF_INET:
576 		pr4 = (struct pr_set *)malloc(sizeof (struct pr_set));
577 		if (pr4 == NULL) {
578 			Fprintf(stderr,
579 			    "%s: malloc %s\n", prog, strerror(errno));
580 			exit(EXIT_FAILURE);
581 		}
582 		setup_protocol(pr4, AF_INET);
583 		outip4 = (*pr4->set_buffers_fn)(pr4->packlen);
584 		setup_socket(pr4, pr4->packlen);
585 		num_ifs4 = set_src_addr(pr4, &al4);
586 		break;
587 	default:
588 		Fprintf(stderr, "%s: unknow address family.\n", prog);
589 		exit(EXIT_FAILURE);
590 	}
591 
592 	if (num_v4 + num_v6 > 1 && !probe_all) {
593 		if (ai_dst->ai_family == AF_INET) {
594 			Fprintf(stderr,
595 			    "%s: Warning: %s has multiple addresses;"
596 			    " using %s\n", prog, hostname,
597 			    inet_ntop(AF_INET,
598 			    /* LINTED E_BAD_PTR_CAST_ALIGN */
599 			    (void *)&((struct sockaddr_in *)
600 			    ai_dst->ai_addr)->sin_addr,
601 			    temp_buf, sizeof (temp_buf)));
602 		} else {
603 			Fprintf(stderr,
604 			    "%s: Warning: %s has multiple addresses;"
605 			    " using %s\n", prog, hostname,
606 			    inet_ntop(AF_INET6,
607 			    /* LINTED E_BAD_PTR_CAST_ALIGN */
608 			    (void *)&((struct sockaddr_in6 *)
609 			    ai_dst->ai_addr)->sin6_addr,
610 			    temp_buf, sizeof (temp_buf)));
611 		}
612 	}
613 
614 	if (num_ifs4 + num_ifs6 > 0) {
615 		trace_it(ai_dst);
616 		probing_successful = _B_TRUE;
617 	}
618 
619 	(void) close(rcvsock4);
620 	(void) close(sndsock4);
621 	(void) close(rcvsock6);
622 	(void) close(sndsock6);
623 
624 	/*
625 	 * if we could probe any of the IP addresses of the target, that means
626 	 * this was a successful operation
627 	 */
628 	if (probing_successful)
629 		return (EXIT_SUCCESS);
630 	else
631 		return (EXIT_FAILURE);
632 }
633 
634 /*
635  * print "unknown host" message
636  */
637 static void
638 print_unknown_host_msg(const char *protocol, const char *host)
639 {
640 	Fprintf(stderr, "%s: unknown%s host %s\n", prog, protocol, host);
641 }
642 
643 /*
644  * resolve destination host and gateways
645  */
646 static void
647 resolve_nodes(int *family, struct addrinfo **ai_dstp)
648 {
649 	struct addrinfo *ai_dst = NULL;
650 	struct addrinfo *aip = NULL;
651 	int num_resolved_gw = 0;
652 	int num_resolved_gw6 = 0;
653 
654 	get_hostinfo(hostname, *family, &ai_dst);
655 	if (ai_dst == NULL) {
656 		print_unknown_host_msg("", hostname);
657 		exit(EXIT_FAILURE);
658 	}
659 	/* Get a count of the v4 & v6 addresses */
660 	for (aip = ai_dst; aip != NULL; aip = aip->ai_next) {
661 		switch (aip->ai_family) {
662 		case AF_INET:
663 			num_v4++;
664 			break;
665 		case AF_INET6:
666 			num_v6++;
667 			break;
668 		}
669 	}
670 
671 	if (*family == AF_UNSPEC && !probe_all) {
672 		*family = ai_dst->ai_family;
673 	}
674 
675 	/* resolve gateways */
676 	if (gw_count > 0) {
677 		get_gwaddrs(gwlist, *family, gwIPlist, gwIP6list,
678 		    &num_resolved_gw, &num_resolved_gw6);
679 
680 		/* we couldn't resolve a gateway as an IPv6 host */
681 		if (num_resolved_gw6 != gw_count && num_v6 != 0) {
682 			if (*family == AF_INET6 || *family == AF_UNSPEC)
683 				print_unknown_host_msg(" IPv6",
684 				    gwlist[num_resolved_gw6]);
685 			num_v6 = 0;
686 		}
687 
688 		/* we couldn't resolve a gateway as an IPv4 host */
689 		if (num_resolved_gw != gw_count && num_v4 != 0) {
690 			if (*family == AF_INET || *family == AF_UNSPEC)
691 				print_unknown_host_msg(" IPv4",
692 				    gwlist[num_resolved_gw]);
693 			num_v4 = 0;
694 		}
695 	}
696 
697 	*ai_dstp = (num_v4 + num_v6 > 0) ? ai_dst : NULL;
698 }
699 
700 /*
701  * Given IP address or hostname, return v4 and v6 hostinfo lists.
702  * Assumes that hostinfo ** ptrs are non-null.
703  */
704 static void
705 get_hostinfo(char *host, int family, struct addrinfo **aipp)
706 {
707 	struct addrinfo hints, *ai;
708 	struct in6_addr addr6;
709 	struct in_addr addr;
710 	char abuf[INET6_ADDRSTRLEN];	/* use for inet_ntop() */
711 	int rc;
712 
713 	/*
714 	 * Take care of v4-mapped addresses. It should run same as v4, after
715 	 * chopping off the prefix, leaving the IPv4 address
716 	 */
717 	if ((inet_pton(AF_INET6, host, &addr6) > 0) &&
718 	    IN6_IS_ADDR_V4MAPPED(&addr6)) {
719 		/* peel off the "mapping" stuff, leaving 32 bit IPv4 address */
720 		IN6_V4MAPPED_TO_INADDR(&addr6, &addr);
721 
722 		/* convert it back to a string */
723 		(void) inet_ntop(AF_INET, &addr, abuf, sizeof (abuf));
724 
725 		/* now the host is an IPv4 address */
726 		(void) strcpy(host, abuf);
727 
728 		/*
729 		 * If it's a mapped address, we convert it into IPv4
730 		 * address because traceroute will send and receive IPv4
731 		 * packets for that address. Therefore, it's a failure case to
732 		 * ask get_hostinfo() to treat a mapped address as an IPv6
733 		 * address.
734 		 */
735 		if (family == AF_INET6) {
736 			return;
737 		}
738 	}
739 
740 	(void) memset(&hints, 0, sizeof (hints));
741 	hints.ai_family = family;
742 	hints.ai_flags = AI_ADDRCONFIG | AI_CANONNAME;
743 	rc = getaddrinfo(host, NULL, &hints, &ai);
744 	if (rc != 0) {
745 		if (rc != EAI_NONAME)
746 			Fprintf(stderr, "%s: getaddrinfo: %s\n", prog,
747 			    gai_strerror(rc));
748 		*aipp = NULL;
749 		return;
750 	}
751 	*aipp = ai;
752 }
753 
754 /*
755  * Calculate the packet length to be used, and check against the valid range.
756  * Returns -1 if range check fails.
757  */
758 static uint_t
759 calc_packetlen(int plen_input, struct pr_set *pr)
760 {
761 	int minpacket;			/* min ip packet size */
762 	int optlen;			/* length of ip options */
763 	int plen;
764 
765 	/*
766 	 * LBNL bug fixed: miscalculation of optlen
767 	 */
768 	if (gw_count > 0) {
769 		/*
770 		 * IPv4:
771 		 * ----
772 		 * 5 (NO OPs) + 3 (code, len, ptr) + gateways
773 		 * IP options field can hold up to 9 gateways. But the API
774 		 * allows you to specify only 8, because the last one is the
775 		 * destination host. When this packet is sent, on the wire
776 		 * you see one gateway replaced by 4 NO OPs. The other 1 NO
777 		 * OP is for alignment
778 		 *
779 		 * IPv6:
780 		 * ----
781 		 * Well, formula is different, but the result is same.
782 		 * 8 byte fixed part for Type 0 Routing header, followed by
783 		 * gateway addresses
784 		 */
785 		optlen = 8 + gw_count * pr->addr_len;
786 	} else {
787 		optlen = 0;
788 	}
789 
790 	/* take care of the packet length calculations and checks */
791 	minpacket = pr->ip_hdr_len + sizeof (struct outdata) + optlen;
792 	if (useicmp)
793 		minpacket += pr->icmp_minlen;	/* minimum ICMP header size */
794 	else
795 		minpacket += sizeof (struct udphdr);
796 	plen = plen_input;
797 	if (plen == 0) {
798 		plen = minpacket;		/* minimum sized packet */
799 	} else if (minpacket > plen || plen > IP_MAXPACKET) {
800 		Fprintf(stderr, "%s: %s packet size must be >= %d and <= %d\n",
801 		    prog, pr->name, minpacket, IP_MAXPACKET);
802 		return (0);
803 	}
804 
805 	return (plen);
806 }
807 
808 /*
809  * Sets the source address by resolving -i and -s arguments, or if -i and -s
810  * don't dictate any, it sets the pick_src to make sure traceroute uses the
811  * kernel's pick of the source address.
812  * Returns number of interfaces configured on the source host, 0 on error or
813  * there's no interface which is up amd not a loopback.
814  */
815 static int
816 set_src_addr(struct pr_set *pr, struct ifaddrlist **alp)
817 {
818 	union any_in_addr *ap;
819 	struct ifaddrlist *al = NULL;
820 	struct ifaddrlist *tmp1_al = NULL;
821 	struct ifaddrlist *tmp2_al = NULL;
822 	/* LINTED E_BAD_PTR_CAST_ALIGN */
823 	struct sockaddr_in *sin_from = (struct sockaddr_in *)pr->from;
824 	/* LINTED E_BAD_PTR_CAST_ALIGN */
825 	struct sockaddr_in6 *sin6_from = (struct sockaddr_in6 *)pr->from;
826 	struct addrinfo *aip;
827 	char errbuf[ERRBUFSIZE];
828 	char abuf[INET6_ADDRSTRLEN];		/* use for inet_ntop() */
829 	int num_ifs;				/* all the interfaces  */
830 	int num_src_ifs;			/* exclude loopback and down */
831 	int i;
832 	uint_t ifaddrflags = 0;
833 
834 	source = source_input;
835 
836 	if (device != NULL)
837 		ifaddrflags |= LIFC_UNDER_IPMP;
838 
839 	/* get the interface address list */
840 	num_ifs = ifaddrlist(&al, pr->family, ifaddrflags, errbuf);
841 	if (num_ifs < 0) {
842 		Fprintf(stderr, "%s: ifaddrlist: %s\n", prog, errbuf);
843 		exit(EXIT_FAILURE);
844 	}
845 
846 	num_src_ifs = 0;
847 	for (i = 0; i < num_ifs; i++) {
848 		if (!(al[i].flags & IFF_LOOPBACK) && (al[i].flags & IFF_UP))
849 			num_src_ifs++;
850 	}
851 
852 	if (num_src_ifs == 0) {
853 		Fprintf(stderr, "%s: can't find any %s network interfaces\n",
854 		    prog, pr->name);
855 		return (0);
856 	}
857 
858 	/* verify the device */
859 	if (device != NULL) {
860 		tmp1_al = find_device(al, num_ifs, device);
861 
862 		if (tmp1_al == NULL) {
863 			Fprintf(stderr, "%s: %s (index %d) is an invalid %s"
864 			    " interface\n", prog, device, if_index, pr->name);
865 			free(al);
866 			return (0);
867 		}
868 	}
869 
870 	/* verify the source address */
871 	if (source != NULL) {
872 		get_hostinfo(source, pr->family, &aip);
873 		if (aip == NULL) {
874 			Fprintf(stderr,
875 			    "%s: %s is an invalid %s source address\n",
876 			    prog, source, pr->name);
877 
878 			free(al);
879 			return (0);
880 		}
881 
882 		source = aip->ai_canonname;
883 
884 		if (pr->family == AF_INET)
885 			ap = (union any_in_addr *)
886 			    /* LINTED E_BAD_PTR_CAST_ALIGN */
887 			    &((struct sockaddr_in *)aip->ai_addr)->sin_addr;
888 		else
889 			ap = (union any_in_addr *)
890 			    /* LINTED E_BAD_PTR_CAST_ALIGN */
891 			    &((struct sockaddr_in6 *)aip->ai_addr)->sin6_addr;
892 
893 		/*
894 		 * LBNL bug fixed: used to accept any src address
895 		 */
896 		tmp2_al = find_ifaddr(al, num_ifs, ap, pr->family);
897 		if (tmp2_al == NULL) {
898 			(void) inet_ntop(pr->family, ap, abuf, sizeof (abuf));
899 			Fprintf(stderr, "%s: %s is not a local %s address\n",
900 			    prog, abuf, pr->name);
901 			free(al);
902 			freeaddrinfo(aip);
903 			return (0);
904 		}
905 	}
906 
907 	pick_src = _B_FALSE;
908 
909 	if (source == NULL) {			/* no -s used */
910 		if (device == NULL) {		/* no -i used, no -s used */
911 			pick_src = _B_TRUE;
912 		} else {			/* -i used, no -s used */
913 			/*
914 			 * -i used, but not -s, and it's IPv4: set the source
915 			 * address to whatever the interface has configured on
916 			 * it.
917 			 */
918 			if (pr->family == AF_INET)
919 				set_sin(pr->from, &(tmp1_al->addr), pr->family);
920 			else
921 				pick_src = _B_TRUE;
922 		}
923 	} else {				/* -s used */
924 		if (device == NULL) {		/* no -i used, -s used */
925 			set_sin(pr->from, ap, pr->family);
926 
927 			if (aip->ai_next != NULL) {
928 				(void) inet_ntop(pr->family, pr->from_sin_addr,
929 				    abuf, sizeof (abuf));
930 				Fprintf(stderr, "%s: Warning: %s has multiple "
931 				    "addresses; using %s\n", prog, source,
932 				    abuf);
933 			}
934 		} else {			/* -i and -s used */
935 			/*
936 			 * Make sure the source specified matches the
937 			 * interface address. You only care about this for IPv4
938 			 * IPv6 can handle IF not matching src address
939 			 */
940 			if (pr->family == AF_INET) {
941 				if (!has_addr(aip, &tmp1_al->addr)) {
942 					Fprintf(stderr,
943 					    "%s: %s is not on interface %s\n",
944 					    prog, source, device);
945 					exit(EXIT_FAILURE);
946 				}
947 				/*
948 				 * make sure we use the one matching the
949 				 * interface's address
950 				 */
951 				*ap = tmp1_al->addr;
952 			}
953 
954 			set_sin(pr->from, ap, pr->family);
955 		}
956 	}
957 
958 	/*
959 	 * Binding at this point will set the source address to be used
960 	 * for both IPv4 (when raw IP datagrams are not required) and
961 	 * IPv6.  If the address being bound to is zero, then the kernel
962 	 * will end up choosing the source address when the datagram is
963 	 * sent.
964 	 *
965 	 * For raw IPv4 datagrams, the source address is initialized
966 	 * within traceroute() along with the outbound destination
967 	 * address.
968 	 */
969 	if (pr->family == AF_INET && !raw_req) {
970 		sin_from->sin_family = AF_INET;
971 		sin_from->sin_port = htons(ident);
972 		if (bind(sndsock4, (struct sockaddr *)pr->from,
973 			sizeof (struct sockaddr_in)) < 0) {
974 			Fprintf(stderr, "%s: bind: %s\n", prog,
975 			    strerror(errno));
976 			exit(EXIT_FAILURE);
977 		}
978 	} else if (pr->family == AF_INET6) {
979 		sin6_from->sin6_family = AF_INET6;
980 		sin6_from->sin6_port = htons(ident);
981 		if (bind(sndsock6, (struct sockaddr *)pr->from,
982 			sizeof (struct sockaddr_in6)) < 0) {
983 			Fprintf(stderr, "%s: bind: %s\n", prog,
984 			    strerror(errno));
985 			exit(EXIT_FAILURE);
986 		}
987 
988 		whereto6.sin6_flowinfo = htonl((class << 20) | flow);
989 	}
990 	*alp = al;
991 	return (num_ifs);
992 }
993 
994 /*
995  * Returns the complete ifaddrlist structure matching the desired interface
996  * address. Ignores interfaces which are either down or loopback.
997  */
998 static struct ifaddrlist *
999 find_ifaddr(struct ifaddrlist *al, int len, union any_in_addr *addr,
1000     int family)
1001 {
1002 	struct ifaddrlist *tmp_al = al;
1003 	int i;
1004 	size_t addr_len = (family == AF_INET) ? sizeof (struct in_addr) :
1005 	    sizeof (struct in6_addr);
1006 
1007 	for (i = 0; i < len; i++, tmp_al++) {
1008 		if ((!(tmp_al->flags & IFF_LOOPBACK) &&
1009 		    (tmp_al->flags & IFF_UP)) &&
1010 		    (memcmp(&tmp_al->addr, addr, addr_len) == 0))
1011 			break;
1012 	}
1013 
1014 	if (i < len) {
1015 		return (tmp_al);
1016 	} else {
1017 		return (NULL);
1018 	}
1019 }
1020 
1021 /*
1022  * Returns the complete ifaddrlist structure matching the desired interface name
1023  * Ignores interfaces which are either down or loopback.
1024  */
1025 static struct ifaddrlist *
1026 find_device(struct ifaddrlist *al, int len, char *device)
1027 {
1028 	struct ifaddrlist *tmp_al = al;
1029 	int i;
1030 
1031 	for (i = 0; i < len; i++, tmp_al++) {
1032 		if ((!(tmp_al->flags & IFF_LOOPBACK) &&
1033 		    (tmp_al->flags & IFF_UP)) &&
1034 		    (strcmp(tmp_al->device, device) == 0))
1035 			break;
1036 	}
1037 
1038 	if (i < len) {
1039 		return (tmp_al);
1040 	} else {
1041 		return (NULL);
1042 	}
1043 }
1044 
1045 /*
1046  * returns _B_TRUE if given hostinfo contains the given address
1047  */
1048 static boolean_t
1049 has_addr(struct addrinfo *ai, union any_in_addr *addr)
1050 {
1051 	struct addrinfo *ai_tmp = NULL;
1052 	union any_in_addr *ap;
1053 
1054 	for (ai_tmp = ai; ai_tmp != NULL; ai_tmp = ai_tmp->ai_next) {
1055 		if (ai_tmp->ai_family == AF_INET6)
1056 			continue;
1057 		ap = (union any_in_addr *)
1058 		    /* LINTED E_BAD_PTR_CAST_ALIGN */
1059 		    &((struct sockaddr_in *)ai_tmp->ai_addr)->sin_addr;
1060 		if (memcmp(ap, addr, sizeof (struct in_addr)) == 0)
1061 			break;
1062 	}
1063 
1064 	if (ai_tmp != NULL) {
1065 		return (_B_TRUE);
1066 	} else {
1067 		return (_B_FALSE);
1068 	}
1069 }
1070 
1071 /*
1072  * Resolve the gateway names, splitting results into v4 and v6 lists.
1073  * Gateway addresses are added to the appropriate passed-in array; the
1074  * number of resolved gateways for each af is returned in resolved[6].
1075  * Assumes that passed-in arrays are large enough for MAX_GWS[6] addrs
1076  * and resolved[6] ptrs are non-null; ignores array and counter if the
1077  * address family param makes them irrelevant.
1078  */
1079 static void
1080 get_gwaddrs(char **gwlist, int family, union any_in_addr *gwIPlist,
1081     union any_in_addr *gwIPlist6, int *resolved, int *resolved6)
1082 {
1083 	int i;
1084 	boolean_t check_v4 = _B_TRUE, check_v6 = _B_TRUE;
1085 	struct addrinfo *ai = NULL;
1086 	struct addrinfo *aip = NULL;
1087 
1088 	*resolved = *resolved6 = 0;
1089 	switch (family) {
1090 	case AF_UNSPEC:
1091 		break;
1092 	case AF_INET:
1093 		check_v6 = _B_FALSE;
1094 		break;
1095 	case AF_INET6:
1096 		check_v4 = _B_FALSE;
1097 		break;
1098 	default:
1099 		return;
1100 	}
1101 
1102 	if (check_v4 && gw_count >= MAX_GWS) {
1103 		check_v4 = _B_FALSE;
1104 		Fprintf(stderr, "%s: too many IPv4 gateways\n", prog);
1105 		num_v4 = 0;
1106 	}
1107 	if (check_v6 && gw_count >= MAX_GWS6) {
1108 		check_v6 = _B_FALSE;
1109 		Fprintf(stderr, "%s: too many IPv6 gateways\n", prog);
1110 		num_v6 = 0;
1111 	}
1112 
1113 	for (i = 0; i < gw_count; i++) {
1114 		if (!check_v4 && !check_v6)
1115 			return;
1116 		get_hostinfo(gwlist[i], family, &ai);
1117 		if (ai == NULL)
1118 			return;
1119 		if (check_v4 && num_v4 != 0) {
1120 			check_v4 = _B_FALSE;
1121 			for (aip = ai; aip != NULL; aip = aip->ai_next) {
1122 				if (aip->ai_family == AF_INET) {
1123 					/* LINTED E_BAD_PTR_CAST_ALIGN */
1124 					bcopy(&((struct sockaddr_in *)
1125 					    aip->ai_addr)->sin_addr,
1126 					    &gwIPlist[i].addr,
1127 					    aip->ai_addrlen);
1128 					(*resolved)++;
1129 					check_v4 = _B_TRUE;
1130 					break;
1131 				}
1132 			}
1133 		} else if (check_v4) {
1134 			check_v4 = _B_FALSE;
1135 		}
1136 		if (check_v6 && num_v6 != 0) {
1137 			check_v6 = _B_FALSE;
1138 			for (aip = ai; aip != NULL; aip = aip->ai_next) {
1139 				if (aip->ai_family == AF_INET6) {
1140 					/* LINTED E_BAD_PTR_CAST_ALIGN */
1141 					bcopy(&((struct sockaddr_in6 *)
1142 					    aip->ai_addr)->sin6_addr,
1143 					    &gwIPlist6[i].addr6,
1144 					    aip->ai_addrlen);
1145 					(*resolved6)++;
1146 					check_v6 = _B_TRUE;
1147 					break;
1148 				}
1149 			}
1150 		} else if (check_v6) {
1151 			check_v6 = _B_FALSE;
1152 		}
1153 	}
1154 	freeaddrinfo(ai);
1155 }
1156 
1157 /*
1158  * set protocol specific values here
1159  */
1160 static void
1161 setup_protocol(struct pr_set *pr, int family)
1162 {
1163 	/*
1164 	 * Set the global variables for each AF. This is going to save us lots
1165 	 * of "if (family == AF_INET)... else .."
1166 	 */
1167 	pr->family = family;
1168 
1169 	if (family == AF_INET) {
1170 		if (!docksum) {
1171 			Fprintf(stderr,
1172 			    "%s: Warning: checksums disabled\n", prog);
1173 		}
1174 		(void) strcpy(pr->name, "IPv4");
1175 		(void) strcpy(pr->icmp, "icmp");
1176 		pr->icmp_minlen = ICMP_MINLEN;
1177 		pr->addr_len = sizeof (struct in_addr);
1178 		pr->ip_hdr_len = sizeof (struct ip);
1179 		pr->sock_size = sizeof (struct sockaddr_in);
1180 		pr->to = (struct sockaddr *)&whereto;
1181 		pr->from = (struct sockaddr *)&wherefrom;
1182 		pr->from_sin_addr = (void *)&wherefrom.sin_addr;
1183 		pr->gwIPlist = gwIPlist;
1184 		pr->set_buffers_fn = set_buffers;
1185 		pr->check_reply_fn = check_reply;
1186 		pr->print_icmp_other_fn = print_icmp_other;
1187 		pr->print_addr_fn = print_addr;
1188 		pr->packlen = calc_packetlen(packlen_input, pr);
1189 	} else {
1190 		(void) strcpy(pr->name, "IPv6");
1191 		(void) strcpy(pr->icmp, "ipv6-icmp");
1192 		pr->icmp_minlen = ICMP6_MINLEN;
1193 		pr->addr_len = sizeof (struct in6_addr);
1194 		pr->ip_hdr_len = sizeof (struct ip6_hdr);
1195 		pr->sock_size = sizeof (struct sockaddr_in6);
1196 		pr->to = (struct sockaddr *)&whereto6;
1197 		pr->from = (struct sockaddr *)&wherefrom6;
1198 		pr->from_sin_addr = (void *)&wherefrom6.sin6_addr;
1199 		pr->gwIPlist = gwIP6list;
1200 		pr->set_buffers_fn = set_buffers6;
1201 		pr->check_reply_fn = check_reply6;
1202 		pr->print_icmp_other_fn = print_icmp_other6;
1203 		pr->print_addr_fn = print_addr6;
1204 		pr->packlen = calc_packetlen(packlen_input, pr);
1205 	}
1206 	if (pr->packlen == 0)
1207 		exit(EXIT_FAILURE);
1208 }
1209 
1210 /*
1211  * setup the sockets for the given protocol's address family
1212  */
1213 static void
1214 setup_socket(struct pr_set *pr, int packet_len)
1215 {
1216 	int on = 1;
1217 	struct protoent *pe;
1218 	int type;
1219 	int proto;
1220 	int int_op;
1221 	int rsock;
1222 	int ssock;
1223 
1224 	if ((pe = getprotobyname(pr->icmp)) == NULL) {
1225 		Fprintf(stderr, "%s: unknown protocol %s\n", prog, pr->icmp);
1226 		exit(EXIT_FAILURE);
1227 	}
1228 
1229 	/* privilege bracketing */
1230 	(void) __priv_bracket(PRIV_ON);
1231 
1232 	if ((rsock = socket(pr->family, SOCK_RAW, pe->p_proto)) < 0) {
1233 		Fprintf(stderr, "%s: icmp socket: %s\n", prog, strerror(errno));
1234 		exit(EXIT_FAILURE);
1235 	}
1236 
1237 	if (options & SO_DEBUG) {
1238 		if (setsockopt(rsock, SOL_SOCKET, SO_DEBUG, (char *)&on,
1239 		    sizeof (on)) < 0) {
1240 			Fprintf(stderr, "%s: SO_DEBUG: %s\n", prog,
1241 			    strerror(errno));
1242 			exit(EXIT_FAILURE);
1243 		}
1244 	}
1245 	if (options & SO_DONTROUTE) {
1246 		if (setsockopt(rsock, SOL_SOCKET, SO_DONTROUTE, (char *)&on,
1247 		    sizeof (on)) < 0) {
1248 			Fprintf(stderr, "%s: SO_DONTROUTE: %s\n", prog,
1249 			    strerror(errno));
1250 			exit(EXIT_FAILURE);
1251 		}
1252 	}
1253 
1254 	if (pr->family == AF_INET6) {
1255 		/* Enable receipt of destination address info */
1256 		if (setsockopt(rsock, IPPROTO_IPV6, IPV6_RECVPKTINFO,
1257 		    (char *)&on, sizeof (on)) < 0) {
1258 			Fprintf(stderr, "%s: IPV6_RECVPKTINFO: %s\n", prog,
1259 			    strerror(errno));
1260 			exit(EXIT_FAILURE);
1261 		}
1262 		/* Enable receipt of hoplimit info */
1263 		if (setsockopt(rsock, IPPROTO_IPV6, IPV6_RECVHOPLIMIT,
1264 		    (char *)&on, sizeof (on)) < 0) {
1265 			Fprintf(stderr, "%s: IPV6_RECVHOPLIMIT: %s\n", prog,
1266 			    strerror(errno));
1267 			exit(EXIT_FAILURE);
1268 		}
1269 
1270 	}
1271 
1272 	/*
1273 	 * Initialize the socket type and protocol based on the address
1274 	 * family, whether or not a raw IP socket is required (for IPv4)
1275 	 * or whether ICMP will be used instead of UDP.
1276 	 *
1277 	 * For historical reasons, the datagrams sent out by
1278 	 * traceroute(1M) do not have the "don't fragment" flag set.  For
1279 	 * this reason as well as the ability to set the Loose Source and
1280 	 * Record Route (LSRR) option, a raw IP socket will be used for
1281 	 * IPv4 when run in the global zone.  Otherwise, the actual
1282 	 * datagram that will be sent will be a regular UDP or ICMP echo
1283 	 * request packet.  However for convenience and for future options
1284 	 * when other IP header information may be specified using
1285 	 * traceroute, the buffer including the raw IP and UDP or ICMP
1286 	 * header is always filled in.  When the probe is actually sent,
1287 	 * the size of the request and the start of the packet is set
1288 	 * according to the type of datagram to send.
1289 	 */
1290 	if (pr->family == AF_INET && raw_req) {
1291 		type = SOCK_RAW;
1292 		proto = IPPROTO_RAW;
1293 	} else if (useicmp) {
1294 		type = SOCK_RAW;
1295 		if (pr->family == AF_INET)
1296 			proto = IPPROTO_ICMP;
1297 		else
1298 			proto = IPPROTO_ICMPV6;
1299 	} else {
1300 		type = SOCK_DGRAM;
1301 		proto = IPPROTO_UDP;
1302 	}
1303 	ssock = socket(pr->family, type, proto);
1304 
1305 	if (ssock < 0) {
1306 		if (proto == IPPROTO_RAW) {
1307 			Fprintf(stderr, "%s: raw socket: %s\n", prog,
1308 			    strerror(errno));
1309 		} else if (proto == IPPROTO_UDP) {
1310 			Fprintf(stderr, "%s: udp socket: %s\n", prog,
1311 			    strerror(errno));
1312 		} else {
1313 			Fprintf(stderr, "%s: icmp socket: %s\n", prog,
1314 			    strerror(errno));
1315 		}
1316 		exit(EXIT_FAILURE);
1317 	}
1318 
1319 	if (setsockopt(ssock, SOL_SOCKET, SO_SNDBUF, (char *)&packet_len,
1320 	    sizeof (packet_len)) < 0) {
1321 		Fprintf(stderr, "%s: SO_SNDBUF: %s\n", prog, strerror(errno));
1322 		exit(EXIT_FAILURE);
1323 	}
1324 
1325 	if (pr->family == AF_INET && raw_req) {
1326 		if (setsockopt(ssock, IPPROTO_IP, IP_HDRINCL, (char *)&on,
1327 		    sizeof (on)) < 0) {
1328 			Fprintf(stderr, "%s: IP_HDRINCL: %s\n", prog,
1329 			    strerror(errno));
1330 			exit(EXIT_FAILURE);
1331 		}
1332 	}
1333 
1334 	if (options & SO_DEBUG) {
1335 		if (setsockopt(ssock, SOL_SOCKET, SO_DEBUG, (char *)&on,
1336 		    sizeof (on)) < 0) {
1337 			Fprintf(stderr, "%s: SO_DEBUG: %s\n", prog,
1338 			    strerror(errno));
1339 			exit(EXIT_FAILURE);
1340 		}
1341 	}
1342 	if (options & SO_DONTROUTE) {
1343 		if (setsockopt(ssock, SOL_SOCKET, SO_DONTROUTE,
1344 		    (char *)&on, sizeof (on)) < 0) {
1345 			Fprintf(stderr, "%s: SO_DONTROUTE: %s\n", prog,
1346 			    strerror(errno));
1347 			exit(EXIT_FAILURE);
1348 		}
1349 	}
1350 
1351 	/*
1352 	 * If a raw IPv4 packet is going to be sent, the Type of Service
1353 	 * field in the packet will be initialized in set_buffers().
1354 	 * Otherwise, it is initialized here using the IPPROTO_IP level
1355 	 * socket option.
1356 	 */
1357 	if (settos && !raw_req) {
1358 		int_op = tos;
1359 		if (setsockopt(ssock, IPPROTO_IP, IP_TOS, (char *)&int_op,
1360 		    sizeof (int_op)) < 0) {
1361 			Fprintf(stderr, "%s: IP_TOS: %s\n", prog,
1362 			    strerror(errno));
1363 			exit(EXIT_FAILURE);
1364 		}
1365 	}
1366 
1367 	/* We enable or disable to not depend on the kernel default */
1368 	if (pr->family == AF_INET) {
1369 		if (setsockopt(ssock, IPPROTO_IP, IP_DONTFRAG,
1370 		    (char *)&dontfrag, sizeof (dontfrag)) == -1) {
1371 			Fprintf(stderr, "%s: IP_DONTFRAG %s\n", prog,
1372 			    strerror(errno));
1373 			exit(EXIT_FAILURE);
1374 		}
1375 	} else {
1376 		if (setsockopt(ssock, IPPROTO_IPV6, IPV6_DONTFRAG,
1377 		    (char *)&dontfrag, sizeof (dontfrag)) == -1) {
1378 			Fprintf(stderr, "%s: IPV6_DONTFRAG %s\n", prog,
1379 			    strerror(errno));
1380 			exit(EXIT_FAILURE);
1381 		}
1382 	}
1383 
1384 	if (pr->family == AF_INET) {
1385 		rcvsock4 = rsock;
1386 		sndsock4 = ssock;
1387 	} else {
1388 		rcvsock6 = rsock;
1389 		sndsock6 = ssock;
1390 	}
1391 	/* Revert to non-privileged user after configuring sockets */
1392 	(void) __priv_bracket(PRIV_OFF);
1393 }
1394 
1395 /*
1396  * If we are "probing all", this function calls traceroute() for each IP address
1397  * of the target, otherwise calls only once. Returns _B_FALSE if traceroute()
1398  * fails.
1399  */
1400 static void
1401 trace_it(struct addrinfo *ai_dst)
1402 {
1403 	struct msghdr msg6;
1404 	int num_dst_IPaddrs;
1405 	struct addrinfo *aip;
1406 	int i;
1407 
1408 	if (!probe_all)
1409 		num_dst_IPaddrs = 1;
1410 	else
1411 		num_dst_IPaddrs = num_v4 + num_v6;
1412 
1413 	/*
1414 	 * Initialize the msg6 structure using the hoplimit for the first
1415 	 * probe packet, gateway addresses and the outgoing interface index.
1416 	 */
1417 	if (ai_dst->ai_family == AF_INET6 || (probe_all && num_v6)) {
1418 		msg6.msg_control = NULL;
1419 		msg6.msg_controllen = 0;
1420 		set_ancillary_data(&msg6, first_ttl, pr6->gwIPlist, gw_count,
1421 		    if_index);
1422 	}
1423 
1424 	/* run traceroute for all the IP addresses of the multihomed dest */
1425 	for (aip = ai_dst, i = 0; i < num_dst_IPaddrs && aip != NULL; i++) {
1426 		union any_in_addr *addrp;
1427 		if (aip->ai_family == AF_INET) {
1428 			addrp = (union any_in_addr *)
1429 			    /* LINTED E_BAD_PTR_CAST_ALIGN */
1430 			    &((struct sockaddr_in *)
1431 			    aip->ai_addr)->sin_addr;
1432 			set_sin((struct sockaddr *)pr4->to, addrp,
1433 			    aip->ai_family);
1434 			traceroute(addrp, &msg6, pr4, num_ifs4, al4);
1435 		} else {
1436 			addrp = (union any_in_addr *)
1437 			    /* LINTED E_BAD_PTR_CAST_ALIGN */
1438 			    &((struct sockaddr_in6 *)
1439 			    aip->ai_addr)->sin6_addr;
1440 			set_sin((struct sockaddr *)pr6->to, addrp,
1441 			    aip->ai_family);
1442 			traceroute(addrp, &msg6, pr6, num_ifs6, al6);
1443 		}
1444 		aip = aip->ai_next;
1445 		if (i < (num_dst_IPaddrs - 1))
1446 			(void) putchar('\n');
1447 	}
1448 }
1449 
1450 /*
1451  * set the IP address in a sockaddr struct
1452  */
1453 static void
1454 set_sin(struct sockaddr *sock, union any_in_addr *addr, int family)
1455 {
1456 	sock->sa_family = family;
1457 
1458 	if (family == AF_INET)
1459 		/* LINTED E_BAD_PTR_CAST_ALIGN */
1460 		((struct sockaddr_in *)sock)->sin_addr = addr->addr;
1461 	else
1462 		/* LINTED E_BAD_PTR_CAST_ALIGN */
1463 		((struct sockaddr_in6 *)sock)->sin6_addr = addr->addr6;
1464 }
1465 
1466 /*
1467  * returns the IF name on which the given IP address is configured
1468  */
1469 static char *
1470 device_name(struct ifaddrlist *al, int len, union any_in_addr *ip_addr,
1471     struct pr_set *pr)
1472 {
1473 	int i;
1474 	struct ifaddrlist *tmp_al;
1475 
1476 	tmp_al = al;
1477 
1478 	for (i = 0; i < len; i++, tmp_al++) {
1479 		if (memcmp(&tmp_al->addr, ip_addr, pr->addr_len) == 0) {
1480 			return (tmp_al->device);
1481 		}
1482 	}
1483 
1484 	return (NULL);
1485 }
1486 
1487 /*
1488  * Trace the route to the host with given IP address.
1489  */
1490 static void
1491 traceroute(union any_in_addr *ip_addr, struct msghdr *msg6, struct pr_set *pr,
1492     int num_ifs, struct ifaddrlist *al)
1493 {
1494 	int ttl;
1495 	int probe;
1496 	uchar_t type;				/* icmp type */
1497 	uchar_t code;				/* icmp code */
1498 	int reply;
1499 	int seq = 0;
1500 	char abuf[INET6_ADDRSTRLEN];		/* use for inet_ntop() */
1501 	int longjmp_return;			/* return value from longjump */
1502 	struct ip *ip = (struct ip *)packet;
1503 	boolean_t got_there = _B_FALSE;		/* we hit the destination */
1504 	static boolean_t first_pkt = _B_TRUE;
1505 	int hoplimit;				/* hoplimit for IPv6 packets */
1506 	struct in6_addr addr6;
1507 	int num_src_ifs;			/* excludes down and loopback */
1508 	struct msghdr in_msg;
1509 	struct iovec iov;
1510 	int *intp;
1511 	int sndsock;
1512 	int rcvsock;
1513 
1514 	msg6->msg_name = pr->to;
1515 	msg6->msg_namelen = sizeof (struct sockaddr_in6);
1516 	sndsock =  (pr->family == AF_INET) ? sndsock4 : sndsock6;
1517 	rcvsock =  (pr->family == AF_INET) ? rcvsock4 : rcvsock6;
1518 
1519 	/* carry out the source address selection */
1520 	if (pick_src) {
1521 		union any_in_addr src_addr;
1522 		char *dev_name;
1523 		int i;
1524 
1525 		/*
1526 		 * If there's a gateway, a routing header as a consequence, our
1527 		 * kernel picks the source address based on the first hop
1528 		 * address, rather than final destination address.
1529 		 */
1530 		if (gw_count > 0) {
1531 			(void) select_src_addr(pr->gwIPlist, &src_addr,
1532 			    pr->family);
1533 		} else {
1534 			(void) select_src_addr(ip_addr, &src_addr, pr->family);
1535 		}
1536 		set_sin(pr->from, &src_addr, pr->family);
1537 
1538 		/* filter out down and loopback interfaces */
1539 		num_src_ifs = 0;
1540 		for (i = 0; i < num_ifs; i++) {
1541 			if (!(al[i].flags & IFF_LOOPBACK) &&
1542 			    (al[i].flags & IFF_UP))
1543 				num_src_ifs++;
1544 		}
1545 
1546 		if (num_src_ifs > 1) {
1547 			dev_name = device_name(al, num_ifs, &src_addr, pr);
1548 			if (dev_name == NULL)
1549 				dev_name = "?";
1550 
1551 			(void) inet_ntop(pr->family, pr->from_sin_addr, abuf,
1552 			    sizeof (abuf));
1553 			Fprintf(stderr,
1554 			    "%s: Warning: Multiple interfaces found;"
1555 			    " using %s @ %s\n", prog, abuf, dev_name);
1556 		}
1557 	}
1558 
1559 	if (pr->family == AF_INET) {
1560 		outip4->ip_src = *(struct in_addr *)pr->from_sin_addr;
1561 		outip4->ip_dst = ip_addr->addr;
1562 	}
1563 
1564 	/*
1565 	 * If the hostname is an IPv6 literal address, let's not print it twice.
1566 	 */
1567 	if (pr->family == AF_INET6 &&
1568 	    inet_pton(AF_INET6, hostname, &addr6) > 0) {
1569 		Fprintf(stderr, "%s to %s", prog, hostname);
1570 	} else {
1571 		Fprintf(stderr, "%s to %s (%s)", prog, hostname,
1572 		    inet_ntop(pr->family, ip_addr, abuf, sizeof (abuf)));
1573 	}
1574 
1575 	if (source)
1576 		Fprintf(stderr, " from %s", source);
1577 	Fprintf(stderr, ", %d hops max, %d byte packets\n", max_ttl,
1578 	    pr->packlen);
1579 	(void) fflush(stderr);
1580 
1581 	/*
1582 	 * Setup the source routing for IPv4. For IPv6, we did the required
1583 	 * setup in the caller function, trace_it(), because it's independent
1584 	 * from the IP address of target.
1585 	 */
1586 	if (pr->family == AF_INET && gw_count > 0)
1587 		set_IPv4opt_sourcerouting(sndsock, ip_addr, pr->gwIPlist);
1588 
1589 	if (probe_all) {
1590 		/* interrupt handler sig_handler() jumps back to here */
1591 		if ((longjmp_return = setjmp(env)) != 0) {
1592 			switch (longjmp_return) {
1593 			case SIGINT:
1594 				Printf("(skipping)\n");
1595 				return;
1596 			case SIGQUIT:
1597 				Printf("(exiting)\n");
1598 				exit(EXIT_SUCCESS);
1599 			default:	/* should never happen */
1600 				exit(EXIT_FAILURE);
1601 			}
1602 		}
1603 		(void) signal(SIGINT, sig_handler);
1604 	}
1605 
1606 	for (ttl = first_ttl; ttl <= max_ttl; ++ttl) {
1607 		union any_in_addr lastaddr;
1608 		int timeouts = 0;
1609 		double rtt;		/* for statistics */
1610 		int nreceived = 0;
1611 		double rttmin, rttmax;
1612 		double rttsum, rttssq;
1613 		int unreachable;
1614 
1615 		got_there = _B_FALSE;
1616 		unreachable = 0;
1617 
1618 		/*
1619 		 * The following line clears both IPv4 and IPv6 address stored
1620 		 * in the union.
1621 		 */
1622 		lastaddr.addr6 = in6addr_any;
1623 
1624 		if ((ttl == (first_ttl + 1)) && (options & SO_DONTROUTE)) {
1625 			Fprintf(stderr,
1626 			    "%s: host %s is not on a directly-attached"
1627 			    " network\n", prog, hostname);
1628 			break;
1629 		}
1630 
1631 		Printf("%2d ", ttl);
1632 		(void) fflush(stdout);
1633 
1634 		for (probe = 0; (probe < nprobes) && (timeouts < max_timeout);
1635 		    ++probe) {
1636 			int cc;
1637 			struct timeval t1, t2;
1638 
1639 			/*
1640 			 * Put a delay before sending this probe packet. Don't
1641 			 * delay it if it's the very first packet.
1642 			 */
1643 			if (!first_pkt) {
1644 				if (delay.tv_sec > 0)
1645 					(void) sleep((uint_t)delay.tv_sec);
1646 				if (delay.tv_usec > 0)
1647 					(void) usleep(delay.tv_usec);
1648 			} else {
1649 				first_pkt = _B_FALSE;
1650 			}
1651 
1652 			(void) gettimeofday(&t1, NULL);
1653 
1654 			if (pr->family == AF_INET) {
1655 				send_probe(sndsock, pr->to, outip4, seq, ttl,
1656 				    &t1, pr->packlen);
1657 			} else {
1658 				send_probe6(sndsock, msg6, outip6, seq, ttl,
1659 				    &t1, pr->packlen);
1660 			}
1661 
1662 			/* prepare msghdr for recvmsg() */
1663 			in_msg.msg_name = pr->from;
1664 			in_msg.msg_namelen = pr->sock_size;
1665 
1666 			iov.iov_base = (char *)packet;
1667 			iov.iov_len = sizeof (packet);
1668 
1669 			in_msg.msg_iov = &iov;
1670 			in_msg.msg_iovlen = 1;
1671 
1672 			in_msg.msg_control = ancillary_data;
1673 			in_msg.msg_controllen = sizeof (ancillary_data);
1674 
1675 			while ((cc = wait_for_reply(rcvsock, &in_msg,
1676 			    &t1)) != 0) {
1677 				(void) gettimeofday(&t2, NULL);
1678 
1679 				reply = (*pr->check_reply_fn) (&in_msg, cc, seq,
1680 				    &type, &code);
1681 
1682 				in_msg.msg_controllen =
1683 				    sizeof (ancillary_data);
1684 				/* Skip short packet */
1685 				if (reply == REPLY_SHORT_PKT) {
1686 					continue;
1687 				}
1688 
1689 				timeouts = 0;
1690 
1691 				/*
1692 				 * if reply comes from a different host, print
1693 				 * the hostname
1694 				 */
1695 				if (memcmp(pr->from_sin_addr, &lastaddr,
1696 				    pr->addr_len) != 0) {
1697 					(*pr->print_addr_fn) ((uchar_t *)packet,
1698 					    cc, pr->from);
1699 					/* store the address response */
1700 					(void) memcpy(&lastaddr,
1701 					    pr->from_sin_addr, pr->addr_len);
1702 				}
1703 
1704 				rtt = deltaT(&t1, &t2);
1705 				if (collect_stat) {
1706 					record_stats(rtt, &nreceived, &rttmin,
1707 					    &rttmax, &rttsum, &rttssq);
1708 				} else {
1709 					Printf("  %.3f ms", rtt);
1710 				}
1711 
1712 				if (pr->family == AF_INET6) {
1713 					intp = find_ancillary_data(&in_msg,
1714 					    IPPROTO_IPV6, IPV6_HOPLIMIT);
1715 					if (intp == NULL) {
1716 						Fprintf(stderr,
1717 						    "%s: can't find "
1718 						    "IPV6_HOPLIMIT ancillary "
1719 						    "data\n", prog);
1720 						exit(EXIT_FAILURE);
1721 					}
1722 					hoplimit = *intp;
1723 				}
1724 
1725 				if (reply == REPLY_GOT_TARGET) {
1726 					got_there = _B_TRUE;
1727 
1728 					if (((pr->family == AF_INET) &&
1729 					    (ip->ip_ttl <= 1)) ||
1730 					    ((pr->family == AF_INET6) &&
1731 					    (hoplimit <= 1)))
1732 						Printf(" !");
1733 				}
1734 
1735 				if (!collect_stat && showttl) {
1736 					if (pr->family == AF_INET) {
1737 						Printf(" (ttl=%d)",
1738 						    (int)ip->ip_ttl);
1739 					} else if (hoplimit != -1) {
1740 						Printf(" (hop limit=%d)",
1741 						    hoplimit);
1742 					}
1743 				}
1744 
1745 				if (reply == REPLY_GOT_OTHER) {
1746 					if ((*pr->print_icmp_other_fn)
1747 					    (type, code)) {
1748 						unreachable++;
1749 					}
1750 				}
1751 
1752 				/* special case */
1753 				if (pr->family == AF_INET &&
1754 				    type == ICMP_UNREACH &&
1755 				    code == ICMP_UNREACH_PROTOCOL)
1756 					got_there = _B_TRUE;
1757 
1758 				break;
1759 			}
1760 
1761 			seq = (seq + 1) % (MAX_SEQ + 1);
1762 
1763 			if (cc == 0) {
1764 				Printf(" *");
1765 				timeouts++;
1766 			}
1767 
1768 			(void) fflush(stdout);
1769 		}
1770 
1771 		if (collect_stat) {
1772 			print_stats(probe, nreceived, rttmin, rttmax, rttsum,
1773 			    rttssq);
1774 		}
1775 
1776 		(void) putchar('\n');
1777 
1778 		/* either we hit the target or received too many unreachables */
1779 		if (got_there ||
1780 		    (unreachable > 0 && unreachable >= nprobes - 1))
1781 			break;
1782 	}
1783 
1784 	/* Ignore the SIGINT between traceroute() runs */
1785 	if (probe_all)
1786 		(void) signal(SIGINT, SIG_IGN);
1787 }
1788 
1789 /*
1790  * for a given destination address and address family, it finds out what
1791  * source address kernel is going to pick
1792  */
1793 static void
1794 select_src_addr(union any_in_addr *dst_addr, union any_in_addr *src_addr,
1795     int family)
1796 {
1797 	int tmp_fd;
1798 	struct sockaddr *sock;
1799 	struct sockaddr_in *sin;
1800 	struct sockaddr_in6 *sin6;
1801 	size_t sock_len;
1802 
1803 	sock = (struct sockaddr *)malloc(sizeof (struct sockaddr_in6));
1804 	if (sock == NULL) {
1805 		Fprintf(stderr, "%s: malloc %s\n", prog, strerror(errno));
1806 		exit(EXIT_FAILURE);
1807 	}
1808 	(void) bzero(sock, sizeof (struct sockaddr_in6));
1809 
1810 	if (family == AF_INET) {
1811 		/* LINTED E_BAD_PTR_CAST_ALIGN */
1812 		sin = (struct sockaddr_in *)sock;
1813 		sin->sin_family = AF_INET;
1814 		sin->sin_addr = dst_addr->addr;
1815 		sin->sin_port = IPPORT_ECHO;	/* port shouldn't be 0 */
1816 		sock_len = sizeof (struct sockaddr_in);
1817 	} else {
1818 		/* LINTED E_BAD_PTR_CAST_ALIGN */
1819 		sin6 = (struct sockaddr_in6 *)sock;
1820 		sin6->sin6_family = AF_INET6;
1821 		sin6->sin6_addr = dst_addr->addr6;
1822 		sin6->sin6_port = IPPORT_ECHO;	/* port shouldn't be 0 */
1823 		sock_len = sizeof (struct sockaddr_in6);
1824 	}
1825 
1826 	/* open a UDP socket */
1827 	if ((tmp_fd = socket(family, SOCK_DGRAM, 0)) < 0) {
1828 		Fprintf(stderr, "%s: udp socket: %s\n", prog,
1829 		    strerror(errno));
1830 		exit(EXIT_FAILURE);
1831 	}
1832 
1833 	/* connect it */
1834 	if (connect(tmp_fd, sock, sock_len) < 0) {
1835 		/*
1836 		 * If there's no route to the destination, this connect() call
1837 		 * fails. We just return all-zero (wildcard) as the source
1838 		 * address, so that user can get to see "no route to dest"
1839 		 * message, as it'll try to send the probe packet out and will
1840 		 * receive ICMP unreachable.
1841 		 */
1842 		if (family == AF_INET)
1843 			src_addr->addr.s_addr = INADDR_ANY;
1844 		else
1845 			src_addr->addr6 = in6addr_any;
1846 		free(sock);
1847 		return;
1848 	}
1849 
1850 	/* get the local sock info */
1851 	if (getsockname(tmp_fd, sock, &sock_len) < 0) {
1852 		Fprintf(stderr, "%s: getsockname: %s\n", prog,
1853 		    strerror(errno));
1854 		exit(EXIT_FAILURE);
1855 	}
1856 
1857 	if (family == AF_INET) {
1858 		/* LINTED E_BAD_PTR_CAST_ALIGN */
1859 		sin = (struct sockaddr_in *)sock;
1860 		src_addr->addr = sin->sin_addr;
1861 	} else {
1862 		/* LINTED E_BAD_PTR_CAST_ALIGN */
1863 		sin6 = (struct sockaddr_in6 *)sock;
1864 		src_addr->addr6 = sin6->sin6_addr;
1865 	}
1866 
1867 	free(sock);
1868 	(void) close(tmp_fd);
1869 }
1870 
1871 /*
1872  * Checksum routine for Internet Protocol family headers (C Version)
1873  */
1874 ushort_t
1875 in_cksum(ushort_t *addr, int len)
1876 {
1877 	int nleft = len;
1878 	ushort_t *w = addr;
1879 	ushort_t answer;
1880 	int sum = 0;
1881 
1882 	/*
1883 	 *  Our algorithm is simple, using a 32 bit accumulator (sum),
1884 	 *  we add sequential 16 bit words to it, and at the end, fold
1885 	 *  back all the carry bits from the top 16 bits into the lower
1886 	 *  16 bits.
1887 	 */
1888 	while (nleft > 1)  {
1889 		sum += *w++;
1890 		nleft -= 2;
1891 	}
1892 
1893 	/* mop up an odd byte, if necessary */
1894 	if (nleft == 1)
1895 		sum += *(uchar_t *)w;
1896 
1897 	/* add back carry outs from top 16 bits to low 16 bits */
1898 	sum = (sum >> 16) + (sum & 0xffff);	/* add hi 16 to low 16 */
1899 	sum += (sum >> 16);			/* add carry */
1900 	answer = ~sum;				/* truncate to 16 bits */
1901 	return (answer);
1902 }
1903 
1904 /*
1905  * Wait until a reply arrives or timeout occurs. If packet arrived, read it
1906  * return the size of the packet read.
1907  */
1908 static int
1909 wait_for_reply(int sock, struct msghdr *msg, struct timeval *tp)
1910 {
1911 	fd_set fds;
1912 	struct timeval now, wait;
1913 	int cc = 0;
1914 	int result;
1915 
1916 	(void) FD_ZERO(&fds);
1917 	FD_SET(sock, &fds);
1918 
1919 	wait.tv_sec = tp->tv_sec + waittime;
1920 	wait.tv_usec = tp->tv_usec;
1921 	(void) gettimeofday(&now, NULL);
1922 	tv_sub(&wait, &now);
1923 
1924 	if (wait.tv_sec < 0 || wait.tv_usec < 0)
1925 		return (0);
1926 
1927 	result = select(sock + 1, &fds, (fd_set *)NULL, (fd_set *)NULL, &wait);
1928 
1929 	if (result == -1) {
1930 		if (errno != EINTR) {
1931 			Fprintf(stderr, "%s: select: %s\n", prog,
1932 			    strerror(errno));
1933 		}
1934 	} else if (result > 0)
1935 		cc = recvmsg(sock, msg, 0);
1936 
1937 	return (cc);
1938 }
1939 
1940 /*
1941  * Construct an Internet address representation. If the nflag has been supplied,
1942  * give numeric value, otherwise try for symbolic name.
1943  */
1944 char *
1945 inet_name(union any_in_addr *in, int family)
1946 {
1947 	char *cp;
1948 	static boolean_t first = _B_TRUE;
1949 	static char domain[NI_MAXHOST + 1];
1950 	static char line[NI_MAXHOST + 1];	/* assuming		*/
1951 				/* (NI_MAXHOST + 1) >= INET6_ADDRSTRLEN */
1952 	char hbuf[NI_MAXHOST];
1953 	socklen_t slen;
1954 	struct sockaddr_in sin;
1955 	struct sockaddr_in6 sin6;
1956 	struct sockaddr *sa;
1957 	int flags;
1958 
1959 	switch (family) {
1960 	case AF_INET:
1961 		slen = sizeof (struct sockaddr_in);
1962 		sin.sin_addr = in->addr;
1963 		sin.sin_port = 0;
1964 		sa = (struct sockaddr *)&sin;
1965 		break;
1966 	case AF_INET6:
1967 		slen = sizeof (struct sockaddr_in6);
1968 		sin6.sin6_addr = in->addr6;
1969 		sin6.sin6_port = 0;
1970 		sin6.sin6_scope_id = 0;
1971 		sa = (struct sockaddr *)&sin6;
1972 		break;
1973 	default:
1974 		(void) snprintf(line, sizeof (line),
1975 		    "<invalid address family>");
1976 		return (line);
1977 	}
1978 	sa->sa_family = family;
1979 
1980 	if (first && !nflag) {
1981 		/* find out the domain name */
1982 		first = _B_FALSE;
1983 		if (gethostname(domain, MAXHOSTNAMELEN) == 0 &&
1984 		    (cp = strchr(domain, '.')) != NULL) {
1985 			(void) strncpy(domain, cp + 1, sizeof (domain) - 1);
1986 			domain[sizeof (domain) - 1] = '\0';
1987 		} else {
1988 			domain[0] = '\0';
1989 		}
1990 	}
1991 
1992 	flags = (nflag) ? NI_NUMERICHOST : NI_NAMEREQD;
1993 	if (getnameinfo(sa, slen, hbuf, sizeof (hbuf), NULL, 0, flags) != 0) {
1994 		if (inet_ntop(family, (const void *)&in->addr6,
1995 		    hbuf, sizeof (hbuf)) == NULL)
1996 			hbuf[0] = 0;
1997 	} else if (!nflag && (cp = strchr(hbuf, '.')) != NULL &&
1998 	    strcmp(cp + 1, domain) == 0) {
1999 		*cp = '\0';
2000 	}
2001 	(void) strlcpy(line, hbuf, sizeof (line));
2002 
2003 	return (line);
2004 }
2005 
2006 /*
2007  * return the difference (in msec) between two time values
2008  */
2009 static double
2010 deltaT(struct timeval *t1p, struct timeval *t2p)
2011 {
2012 	double dt;
2013 
2014 	dt = (double)(t2p->tv_sec - t1p->tv_sec) * 1000.0 +
2015 	    (double)(t2p->tv_usec - t1p->tv_usec) / 1000.0;
2016 	return (dt);
2017 }
2018 
2019 /*
2020  * Subtract 2 timeval structs:  out = out - in.
2021  * Out is assumed to be >= in.
2022  */
2023 static void
2024 tv_sub(struct timeval *out, struct timeval *in)
2025 {
2026 	if ((out->tv_usec -= in->tv_usec) < 0)   {
2027 		--out->tv_sec;
2028 		out->tv_usec += 1000000;
2029 	}
2030 	out->tv_sec -= in->tv_sec;
2031 }
2032 
2033 /*
2034  * record statistics
2035  */
2036 static void
2037 record_stats(double rtt, int *nreceived, double *rttmin, double *rttmax,
2038     double *rttsum, double *rttssq)
2039 {
2040 	if (*nreceived == 0) {
2041 		*rttmin = rtt;
2042 		*rttmax = rtt;
2043 		*rttsum = rtt;
2044 		*rttssq = rtt * rtt;
2045 	} else {
2046 		if (rtt < *rttmin)
2047 			*rttmin = rtt;
2048 
2049 		if (rtt > *rttmax)
2050 			*rttmax = rtt;
2051 
2052 		*rttsum += rtt;
2053 		*rttssq += rtt * rtt;
2054 	}
2055 
2056 	(*nreceived)++;
2057 }
2058 
2059 /*
2060  * display statistics
2061  */
2062 static void
2063 print_stats(int ntransmitted, int nreceived, double rttmin, double rttmax,
2064     double rttsum, double rttssq)
2065 {
2066 	double rttavg;			/* average round-trip time */
2067 	double rttstd;			/* rtt standard deviation */
2068 
2069 	if (ntransmitted > 0 && ntransmitted >= nreceived) {
2070 		int missed = ntransmitted - nreceived;
2071 		double loss = 100 * (double)missed / (double)ntransmitted;
2072 
2073 		if (nreceived > 0) {
2074 			rttavg = rttsum / nreceived;
2075 			rttstd = rttssq - (rttavg * rttsum);
2076 			rttstd = xsqrt(rttstd / nreceived);
2077 
2078 			Printf("  %.3f", rttmin);
2079 			Printf("/%.3f", rttavg);
2080 			Printf("/%.3f", rttmax);
2081 
2082 			Printf(" (%.3f) ms ", rttstd);
2083 		}
2084 
2085 		Printf(" %d/%d pkts", nreceived, ntransmitted);
2086 
2087 		if (nreceived == 0)
2088 			Printf(" (100%% loss)");
2089 		else
2090 			Printf(" (%.2g%% loss)", loss);
2091 	}
2092 }
2093 
2094 /*
2095  * square root function
2096  */
2097 double
2098 xsqrt(double y)
2099 {
2100 	double t, x;
2101 
2102 	if (y <= 0) {
2103 		return (0.0);
2104 	}
2105 
2106 	x = (y < 1.0) ? 1.0 : y;
2107 	do {
2108 		t = x;
2109 		x = (t + (y/t))/2.0;
2110 	} while (0 < x && x < t);
2111 
2112 	return (x);
2113 }
2114 
2115 /*
2116  * String to double with optional min and max.
2117  */
2118 static double
2119 str2dbl(const char *str, const char *what, double mi, double ma)
2120 {
2121 	double val;
2122 	char *ep;
2123 
2124 	errno = 0;
2125 
2126 	val = strtod(str, &ep);
2127 	if (errno != 0 || *ep != '\0') {
2128 		Fprintf(stderr, "%s: \"%s\" bad value for %s \n",
2129 		    prog, str, what);
2130 		exit(EXIT_FAILURE);
2131 	}
2132 	if (val < mi && mi >= 0) {
2133 		Fprintf(stderr, "%s: %s must be >= %f\n", prog, what, mi);
2134 		exit(EXIT_FAILURE);
2135 	}
2136 	if (val > ma && ma >= 0) {
2137 		Fprintf(stderr, "%s: %s must be <= %f\n", prog, what, ma);
2138 		exit(EXIT_FAILURE);
2139 	}
2140 	return (val);
2141 }
2142 
2143 /*
2144  * String to int with optional min and max. Handles decimal and hex.
2145  */
2146 static int
2147 str2int(const char *str, const char *what, int mi, int ma)
2148 {
2149 	const char *cp;
2150 	int val;
2151 	char *ep;
2152 
2153 	errno = 0;
2154 
2155 	if (str[0] == '0' && (str[1] == 'x' || str[1] == 'X')) {
2156 		cp = str + 2;
2157 		val = (int)strtol(cp, &ep, 16);
2158 	} else {
2159 		val = (int)strtol(str, &ep, 10);
2160 	}
2161 	if (errno != 0 || *ep != '\0') {
2162 		Fprintf(stderr, "%s: \"%s\" bad value for %s \n",
2163 		    prog, str, what);
2164 		exit(EXIT_FAILURE);
2165 	}
2166 	if (val < mi && mi >= 0) {
2167 		if (mi == 0) {
2168 			Fprintf(stderr, "%s: %s must be >= %d\n",
2169 			    prog, what, mi);
2170 		} else {
2171 			Fprintf(stderr, "%s: %s must be > %d\n",
2172 			    prog, what, mi - 1);
2173 		}
2174 		exit(EXIT_FAILURE);
2175 	}
2176 	if (val > ma && ma >= 0) {
2177 		Fprintf(stderr, "%s: %s must be <= %d\n", prog, what, ma);
2178 		exit(EXIT_FAILURE);
2179 	}
2180 	return (val);
2181 }
2182 
2183 /*
2184  * This is the interrupt handler for SIGINT and SIGQUIT. It's completely handled
2185  * where it jumps to.
2186  */
2187 static void
2188 sig_handler(int sig)
2189 {
2190 	longjmp(env, sig);
2191 }
2192 
2193 /*
2194  * display the usage of traceroute
2195  */
2196 static void
2197 usage(void)
2198 {
2199 	Fprintf(stderr, "Usage: %s [-adFIlnSvx] [-A address_family] "
2200 	    "[-c traffic_class]\n"
2201 	    "\t[-f first_hop] [-g gateway [-g gateway ...]| -r] [-i iface]\n"
2202 	    "\t[-L flow_label] [-m max_hop] [-P pause_sec] [-p port] "
2203 	    "[-Q max_timeout]\n"
2204 	    "\t[-q nqueries] [-s src_addr] [-t tos] [-w wait_time] host "
2205 	    "[packetlen]\n", prog);
2206 	exit(EXIT_FAILURE);
2207 }
2208