freebsd-dev/sbin/ping6/ping6.c
Hajimu UMEMOTO 4adb3cea54 Show standard deviation.
PR:		bin/35433
Submitted by:	Morten Rodal <morten@rodal.no>,
		Maxim Konovalov <maxim@macomnet.ru>
MFC after:	1 week
2002-03-01 09:49:48 +00:00

2711 lines
65 KiB
C

/* $KAME: ping6.c,v 1.126 2001/05/17 03:39:08 itojun Exp $ */
/*
* Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the project 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 PROJECT 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 PROJECT 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.
*/
/* BSDI ping.c,v 2.3 1996/01/21 17:56:50 jch Exp */
/*
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Mike Muuss.
*
* 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. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. 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.
*/
#ifndef lint
static char copyright[] =
"@(#) Copyright (c) 1989, 1993\n\
The Regents of the University of California. All rights reserved.\n";
#endif /* not lint */
#ifndef lint
#if 0
static char sccsid[] = "@(#)ping.c 8.1 (Berkeley) 6/5/93";
#endif
static const char rcsid[] =
"$FreeBSD$";
#endif /* not lint */
/*
* Using the InterNet Control Message Protocol (ICMP) "ECHO" facility,
* measure round-trip-delays and packet loss across network paths.
*
* Author -
* Mike Muuss
* U. S. Army Ballistic Research Laboratory
* December, 1983
*
* Status -
* Public Domain. Distribution Unlimited.
* Bugs -
* More statistics could always be gathered.
* This program has to run SUID to ROOT to access the ICMP socket.
*/
/*
* NOTE:
* USE_SIN6_SCOPE_ID assumes that sin6_scope_id has the same semantics
* as IPV6_PKTINFO. Some people object it (sin6_scope_id specifies *link*
* while IPV6_PKTINFO specifies *interface*. Link is defined as collection of
* network attached to 1 or more interfaces)
*/
#include <sys/param.h>
#include <sys/uio.h>
#include <sys/socket.h>
#include <sys/time.h>
#include <net/if.h>
#include <net/route.h>
#include <netinet/in.h>
#include <netinet/ip6.h>
#include <netinet/icmp6.h>
#include <arpa/inet.h>
#include <arpa/nameser.h>
#include <netdb.h>
#include <ctype.h>
#include <err.h>
#include <errno.h>
#include <fcntl.h>
#if defined(__OpenBSD__) || defined(__NetBSD__) || defined(__FreeBSD__)
#include <math.h>
#endif
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#ifdef IPSEC
#include <netinet6/ah.h>
#include <netinet6/ipsec.h>
#endif
#if defined(__NetBSD__) || defined(__FreeBSD__) || defined(__OpenBSD__)
#include <md5.h>
#else
#include "md5.h"
#endif
#define MAXPACKETLEN 131072
#define IP6LEN 40
#define ICMP6ECHOLEN 8 /* icmp echo header length excluding time */
#define ICMP6ECHOTMLEN sizeof(struct timeval)
#define ICMP6_NIQLEN (ICMP6ECHOLEN + 8)
/* FQDN case, 64 bits of nonce + 32 bits ttl */
#define ICMP6_NIRLEN (ICMP6ECHOLEN + 12)
#define EXTRA 256 /* for AH and various other headers. weird. */
#define DEFDATALEN ICMP6ECHOTMLEN
#define MAXDATALEN MAXPACKETLEN - IP6LEN - ICMP6ECHOLEN
#define NROUTES 9 /* number of record route slots */
#define A(bit) rcvd_tbl[(bit)>>3] /* identify byte in array */
#define B(bit) (1 << ((bit) & 0x07)) /* identify bit in byte */
#define SET(bit) (A(bit) |= B(bit))
#define CLR(bit) (A(bit) &= (~B(bit)))
#define TST(bit) (A(bit) & B(bit))
#define F_FLOOD 0x0001
#define F_INTERVAL 0x0002
#define F_PINGFILLED 0x0008
#define F_QUIET 0x0010
#define F_RROUTE 0x0020
#define F_SO_DEBUG 0x0040
#define F_VERBOSE 0x0100
#ifdef IPSEC
#ifdef IPSEC_POLICY_IPSEC
#define F_POLICY 0x0400
#else
#define F_AUTHHDR 0x0200
#define F_ENCRYPT 0x0400
#endif /*IPSEC_POLICY_IPSEC*/
#endif /*IPSEC*/
#define F_NODEADDR 0x0800
#define F_FQDN 0x1000
#define F_INTERFACE 0x2000
#define F_SRCADDR 0x4000
#ifdef IPV6_REACHCONF
#define F_REACHCONF 0x8000
#endif
#define F_HOSTNAME 0x10000
#define F_FQDNOLD 0x20000
#define F_NIGROUP 0x40000
#define F_SUPTYPES 0x80000
#define F_NOMINMTU 0x100000
#define F_NOUSERDATA (F_NODEADDR | F_FQDN | F_FQDNOLD | F_SUPTYPES)
u_int options;
#define IN6LEN sizeof(struct in6_addr)
#define SA6LEN sizeof(struct sockaddr_in6)
#define DUMMY_PORT 10101
#define SIN6(s) ((struct sockaddr_in6 *)(s))
/*
* MAX_DUP_CHK is the number of bits in received table, i.e. the maximum
* number of received sequence numbers we can keep track of. Change 128
* to 8192 for complete accuracy...
*/
#define MAX_DUP_CHK (8 * 8192)
int mx_dup_ck = MAX_DUP_CHK;
char rcvd_tbl[MAX_DUP_CHK / 8];
struct addrinfo *res;
struct sockaddr_in6 dst; /* who to ping6 */
struct sockaddr_in6 src; /* src addr of this packet */
int datalen = DEFDATALEN;
int s; /* socket file descriptor */
u_char outpack[MAXPACKETLEN];
char BSPACE = '\b'; /* characters written for flood */
char DOT = '.';
char *hostname;
int ident; /* process id to identify our packets */
u_int8_t nonce[8]; /* nonce field for node information */
struct in6_addr srcaddr;
int hoplimit = -1; /* hoplimit */
int pathmtu = 0; /* path MTU for the destination. 0 = unspec. */
/* counters */
long npackets; /* max packets to transmit */
long nreceived; /* # of packets we got back */
long nrepeats; /* number of duplicates */
long ntransmitted; /* sequence # for outbound packets = #sent */
struct timeval interval = {1, 0}; /* interval between packets */
/* timing */
int timing; /* flag to do timing */
double tmin = 999999999.0; /* minimum round trip time */
double tmax = 0.0; /* maximum round trip time */
double tsum = 0.0; /* sum of all times, for doing average */
#if defined(__OpenBSD__) || defined(__NetBSD__) || defined(__FreeBSD__)
double tsumsq = 0.0; /* sum of all times squared, for std. dev. */
#endif
/* for node addresses */
u_short naflags;
/* for ancillary data(advanced API) */
struct msghdr smsghdr;
struct iovec smsgiov;
char *scmsg = 0;
volatile int signo;
volatile sig_atomic_t seenalrm;
volatile sig_atomic_t seenint;
#ifdef SIGINFO
volatile sig_atomic_t seeninfo;
#endif
int main __P((int, char *[]));
void fill __P((char *, char *));
int get_hoplim __P((struct msghdr *));
int get_pathmtu __P((struct msghdr *));
void set_pathmtu __P((int));
struct in6_pktinfo *get_rcvpktinfo __P((struct msghdr *));
void onsignal __P((int));
void retransmit __P((void));
void onint __P((int));
size_t pingerlen __P((void));
int pinger __P((void));
const char *pr_addr __P((struct sockaddr *, int));
void pr_icmph __P((struct icmp6_hdr *, u_char *));
void pr_iph __P((struct ip6_hdr *));
void pr_suptypes __P((struct icmp6_nodeinfo *, size_t));
void pr_nodeaddr __P((struct icmp6_nodeinfo *, int));
int myechoreply __P((const struct icmp6_hdr *));
int mynireply __P((const struct icmp6_nodeinfo *));
char *dnsdecode __P((const u_char **, const u_char *, const u_char *,
u_char *, size_t));
void pr_pack __P((u_char *, int, struct msghdr *));
void pr_exthdrs __P((struct msghdr *));
void pr_ip6opt __P((void *));
void pr_rthdr __P((void *));
int pr_bitrange __P((u_int32_t, int, int));
void pr_retip __P((struct ip6_hdr *, u_char *));
void summary __P((void));
void tvsub __P((struct timeval *, struct timeval *));
int setpolicy __P((int, char *));
char *nigroup __P((char *));
void usage __P((void));
int
main(argc, argv)
int argc;
char *argv[];
{
struct itimerval itimer;
struct sockaddr_in6 from;
struct timeval timeout, *tv;
struct addrinfo hints;
fd_set *fdmaskp;
int fdmasks;
register int cc, i;
int ch, fromlen, hold, packlen, preload, optval, ret_ga;
u_char *datap, *packet;
char *e, *target, *ifname = NULL;
int ip6optlen = 0;
struct cmsghdr *scmsgp = NULL;
int sockbufsize = 0;
int usepktinfo = 0;
struct in6_pktinfo *pktinfo = NULL;
#ifdef USE_RFC2292BIS
struct ip6_rthdr *rthdr = NULL;
#endif
#ifdef IPSEC_POLICY_IPSEC
char *policy_in = NULL;
char *policy_out = NULL;
#endif
double intval;
size_t rthlen;
/* just to be sure */
memset(&smsghdr, 0, sizeof(&smsghdr));
memset(&smsgiov, 0, sizeof(&smsgiov));
preload = 0;
datap = &outpack[ICMP6ECHOLEN + ICMP6ECHOTMLEN];
#ifndef IPSEC
#define ADDOPTS
#else
#ifdef IPSEC_POLICY_IPSEC
#define ADDOPTS "P:"
#else
#define ADDOPTS "AE"
#endif /*IPSEC_POLICY_IPSEC*/
#endif
while ((ch = getopt(argc, argv,
"a:b:c:dfHh:I:i:l:mnNp:qRS:s:tvwW" ADDOPTS)) != -1) {
#undef ADDOPTS
switch (ch) {
case 'a':
{
char *cp;
options &= ~F_NOUSERDATA;
options |= F_NODEADDR;
for (cp = optarg; *cp != '\0'; cp++) {
switch (*cp) {
case 'a':
naflags |= NI_NODEADDR_FLAG_ALL;
break;
case 'c':
case 'C':
naflags |= NI_NODEADDR_FLAG_COMPAT;
break;
case 'l':
case 'L':
naflags |= NI_NODEADDR_FLAG_LINKLOCAL;
break;
case 's':
case 'S':
naflags |= NI_NODEADDR_FLAG_SITELOCAL;
break;
case 'g':
case 'G':
naflags |= NI_NODEADDR_FLAG_GLOBAL;
break;
case 'A': /* experimental. not in the spec */
#ifdef NI_NODEADDR_FLAG_ANYCAST
naflags |= NI_NODEADDR_FLAG_ANYCAST;
break;
#else
errx(1,
"-a A is not supported on the platform");
/*NOTREACHED*/
#endif
default:
usage();
/*NOTREACHED*/
}
}
break;
}
case 'b':
#if defined(SO_SNDBUF) && defined(SO_RCVBUF)
sockbufsize = atoi(optarg);
#else
errx(1,
"-b option ignored: SO_SNDBUF/SO_RCVBUF socket options not supported");
#endif
break;
case 'c':
npackets = strtol(optarg, &e, 10);
if (npackets <= 0 || *optarg == '\0' || *e != '\0')
errx(1,
"illegal number of packets -- %s", optarg);
break;
case 'd':
options |= F_SO_DEBUG;
break;
case 'f':
if (getuid())
errx(1, "must be superuser to flood ping");
options |= F_FLOOD;
setbuf(stdout, (char *)NULL);
break;
case 'H':
options |= F_HOSTNAME;
break;
case 'h': /* hoplimit */
hoplimit = strtol(optarg, &e, 10);
if (255 < hoplimit || hoplimit < -1)
errx(1,
"illegal hoplimit -- %s", optarg);
break;
case 'I':
ifname = optarg;
options |= F_INTERFACE;
#ifndef USE_SIN6_SCOPE_ID
usepktinfo++;
#endif
break;
case 'i': /* wait between sending packets */
intval = strtod(optarg, &e);
if (*optarg == '\0' || *e != '\0')
errx(1, "illegal timing interval %s", optarg);
if (intval < 1 && getuid()) {
errx(1, "%s: only root may use interval < 1s",
strerror(EPERM));
}
interval.tv_sec = (long)intval;
interval.tv_usec =
(long)((intval - interval.tv_sec) * 1000000);
if (interval.tv_sec < 0)
errx(1, "illegal timing interval %s", optarg);
/* less than 1/hz does not make sense */
if (interval.tv_sec == 0 && interval.tv_usec < 10000) {
warnx("too small interval, raised to 0.01");
interval.tv_usec = 10000;
}
options |= F_INTERVAL;
break;
case 'l':
if (getuid())
errx(1, "must be superuser to preload");
preload = strtol(optarg, &e, 10);
if (preload < 0 || *optarg == '\0' || *e != '\0')
errx(1, "illegal preload value -- %s", optarg);
break;
case 'm':
#ifdef IPV6_USE_MIN_MTU
options |= F_NOMINMTU;
break;
#else
errx(1, "-%c is not supported on this platform", ch);
/*NOTREACHED*/
#endif
case 'n':
options &= ~F_HOSTNAME;
break;
case 'N':
options |= F_NIGROUP;
break;
case 'p': /* fill buffer with user pattern */
options |= F_PINGFILLED;
fill((char *)datap, optarg);
break;
case 'q':
options |= F_QUIET;
break;
case 'R':
#ifdef IPV6_REACHCONF
options |= F_REACHCONF;
break;
#else
errx(1, "-R is not supported in this configuration");
#endif
case 'S':
/* XXX: use getaddrinfo? */
if (inet_pton(AF_INET6, optarg, (void *)&srcaddr) != 1)
errx(1, "invalid IPv6 address: %s", optarg);
options |= F_SRCADDR;
usepktinfo++;
break;
case 's': /* size of packet to send */
datalen = strtol(optarg, &e, 10);
if (datalen <= 0 || *optarg == '\0' || *e != '\0')
errx(1, "illegal datalen value -- %s", optarg);
if (datalen > MAXDATALEN) {
errx(1,
"datalen value too large, maximum is %d",
MAXDATALEN);
}
break;
case 't':
options &= ~F_NOUSERDATA;
options |= F_SUPTYPES;
break;
case 'v':
options |= F_VERBOSE;
break;
case 'w':
options &= ~F_NOUSERDATA;
options |= F_FQDN;
break;
case 'W':
options &= ~F_NOUSERDATA;
options |= F_FQDNOLD;
break;
#ifdef IPSEC
#ifdef IPSEC_POLICY_IPSEC
case 'P':
options |= F_POLICY;
if (!strncmp("in", optarg, 2)) {
if ((policy_in = strdup(optarg)) == NULL)
errx(1, "strdup");
} else if (!strncmp("out", optarg, 3)) {
if ((policy_out = strdup(optarg)) == NULL)
errx(1, "strdup");
} else
errx(1, "invalid security policy");
break;
#else
case 'A':
options |= F_AUTHHDR;
break;
case 'E':
options |= F_ENCRYPT;
break;
#endif /*IPSEC_POLICY_IPSEC*/
#endif /*IPSEC*/
default:
usage();
/*NOTREACHED*/
}
}
argc -= optind;
argv += optind;
if (argc < 1) {
usage();
/*NOTREACHED*/
}
if (argc > 1) {
#ifdef IPV6_RECVRTHDR /* 2292bis */
rthlen = CMSG_SPACE(inet6_rth_space(IPV6_RTHDR_TYPE_0,
argc - 1));
#else /* RFC2292 */
rthlen = inet6_rthdr_space(IPV6_RTHDR_TYPE_0, argc - 1);
#endif
if (rthlen == 0) {
errx(1, "too many intermediate hops");
/*NOTREACHED*/
}
ip6optlen += rthlen;
}
if (options & F_NIGROUP) {
target = nigroup(argv[argc - 1]);
if (target == NULL) {
usage();
/*NOTREACHED*/
}
} else
target = argv[argc - 1];
/* getaddrinfo */
bzero(&hints, sizeof(struct addrinfo));
hints.ai_flags = AI_CANONNAME;
hints.ai_family = AF_INET6;
hints.ai_socktype = SOCK_RAW;
hints.ai_protocol = IPPROTO_ICMPV6;
ret_ga = getaddrinfo(target, NULL, &hints, &res);
if (ret_ga)
errx(1, "%s", gai_strerror(ret_ga));
if (res->ai_canonname)
hostname = res->ai_canonname;
else
hostname = target;
if (!res->ai_addr)
errx(1, "getaddrinfo failed");
(void)memcpy(&dst, res->ai_addr, res->ai_addrlen);
if ((s = socket(res->ai_family, res->ai_socktype,
res->ai_protocol)) < 0)
err(1, "socket");
/*
* let the kerel pass extension headers of incoming packets,
* for privileged socket options
*/
if ((options & F_VERBOSE) != 0) {
int opton = 1;
#ifdef IPV6_RECVHOPOPTS
if (setsockopt(s, IPPROTO_IPV6, IPV6_RECVHOPOPTS, &opton,
sizeof(opton)))
err(1, "setsockopt(IPV6_RECVHOPOPTS)");
#else /* old adv. API */
if (setsockopt(s, IPPROTO_IPV6, IPV6_HOPOPTS, &opton,
sizeof(opton)))
err(1, "setsockopt(IPV6_HOPOPTS)");
#endif
#ifdef IPV6_RECVDSTOPTS
if (setsockopt(s, IPPROTO_IPV6, IPV6_RECVDSTOPTS, &opton,
sizeof(opton)))
err(1, "setsockopt(IPV6_RECVDSTOPTS)");
#else /* old adv. API */
if (setsockopt(s, IPPROTO_IPV6, IPV6_DSTOPTS, &opton,
sizeof(opton)))
err(1, "setsockopt(IPV6_DSTOPTS)");
#endif
#ifdef IPV6_RECVRTHDRDSTOPTS
if (setsockopt(s, IPPROTO_IPV6, IPV6_RECVRTHDRDSTOPTS, &opton,
sizeof(opton)))
err(1, "setsockopt(IPV6_RECVRTHDRDSTOPTS)");
#endif
}
/* revoke root privilege */
seteuid(getuid());
setuid(getuid());
if (options & F_FLOOD && options & F_INTERVAL)
errx(1, "-f and -i incompatible options");
if ((options & F_NOUSERDATA) == 0) {
if (datalen >= sizeof(struct timeval)) {
/* we can time transfer */
timing = 1;
} else
timing = 0;
/* in F_VERBOSE case, we may get non-echoreply packets*/
if (options & F_VERBOSE)
packlen = 2048 + IP6LEN + ICMP6ECHOLEN + EXTRA;
else
packlen = datalen + IP6LEN + ICMP6ECHOLEN + EXTRA;
} else {
/* suppress timing for node information query */
timing = 0;
datalen = 2048;
packlen = 2048 + IP6LEN + ICMP6ECHOLEN + EXTRA;
}
if (!(packet = (u_char *)malloc((u_int)packlen)))
errx(1, "unable to allocate packet");
if (!(options & F_PINGFILLED))
for (i = ICMP6ECHOLEN; i < packlen; ++i)
*datap++ = i;
ident = getpid() & 0xFFFF;
#ifndef __OpenBSD__
gettimeofday(&timeout, NULL);
srand((unsigned int)(timeout.tv_sec ^ timeout.tv_usec ^ (long)ident));
memset(nonce, 0, sizeof(nonce));
for (i = 0; i < sizeof(nonce); i += sizeof(int))
*((int *)&nonce[i]) = rand();
#else
memset(nonce, 0, sizeof(nonce));
for (i = 0; i < sizeof(nonce); i += sizeof(u_int32_t))
*((u_int32_t *)&nonce[i]) = arc4random();
#endif
hold = 1;
if (options & F_SO_DEBUG)
(void)setsockopt(s, SOL_SOCKET, SO_DEBUG, (char *)&hold,
sizeof(hold));
optval = IPV6_DEFHLIM;
if (IN6_IS_ADDR_MULTICAST(&dst.sin6_addr))
if (setsockopt(s, IPPROTO_IPV6, IPV6_MULTICAST_HOPS,
&optval, sizeof(optval)) == -1)
err(1, "IPV6_MULTICAST_HOPS");
#ifdef IPV6_USE_MIN_MTU
if ((options & F_NOMINMTU) == 0) {
optval = 1;
if (setsockopt(s, IPPROTO_IPV6, IPV6_USE_MIN_MTU,
&optval, sizeof(optval)) == -1)
err(1, "setsockopt(IPV6_USE_MIN_MTU)");
}
#ifdef IPV6_RECVPATHMTU
else {
optval = 1;
if (setsockopt(s, IPPROTO_IPV6, IPV6_RECVPATHMTU,
&optval, sizeof(optval)) == -1)
err(1, "setsockopt(IPV6_RECVPATHMTU)");
}
#endif /* IPV6_RECVPATHMTU */
#endif /* IPV6_USE_MIN_MTU */
#ifdef IPSEC
#ifdef IPSEC_POLICY_IPSEC
if (options & F_POLICY) {
if (setpolicy(s, policy_in) < 0)
errx(1, "%s", ipsec_strerror());
if (setpolicy(s, policy_out) < 0)
errx(1, "%s", ipsec_strerror());
}
#else
if (options & F_AUTHHDR) {
optval = IPSEC_LEVEL_REQUIRE;
#ifdef IPV6_AUTH_TRANS_LEVEL
if (setsockopt(s, IPPROTO_IPV6, IPV6_AUTH_TRANS_LEVEL,
&optval, sizeof(optval)) == -1)
err(1, "setsockopt(IPV6_AUTH_TRANS_LEVEL)");
#else /* old def */
if (setsockopt(s, IPPROTO_IPV6, IPV6_AUTH_LEVEL,
&optval, sizeof(optval)) == -1)
err(1, "setsockopt(IPV6_AUTH_LEVEL)");
#endif
}
if (options & F_ENCRYPT) {
optval = IPSEC_LEVEL_REQUIRE;
if (setsockopt(s, IPPROTO_IPV6, IPV6_ESP_TRANS_LEVEL,
&optval, sizeof(optval)) == -1)
err(1, "setsockopt(IPV6_ESP_TRANS_LEVEL)");
}
#endif /*IPSEC_POLICY_IPSEC*/
#endif
#ifdef ICMP6_FILTER
{
struct icmp6_filter filt;
if (!(options & F_VERBOSE)) {
ICMP6_FILTER_SETBLOCKALL(&filt);
if ((options & F_FQDN) || (options & F_FQDNOLD) ||
(options & F_NODEADDR) || (options & F_SUPTYPES))
ICMP6_FILTER_SETPASS(ICMP6_NI_REPLY, &filt);
else
ICMP6_FILTER_SETPASS(ICMP6_ECHO_REPLY, &filt);
} else {
ICMP6_FILTER_SETPASSALL(&filt);
}
if (setsockopt(s, IPPROTO_ICMPV6, ICMP6_FILTER, &filt,
sizeof(filt)) < 0)
err(1, "setsockopt(ICMP6_FILTER)");
}
#endif /*ICMP6_FILTER*/
/* let the kerel pass extension headers of incoming packets */
if ((options & F_VERBOSE) != 0) {
int opton = 1;
#ifdef IPV6_RECVRTHDR
if (setsockopt(s, IPPROTO_IPV6, IPV6_RECVRTHDR, &opton,
sizeof(opton)))
err(1, "setsockopt(IPV6_RECVRTHDR)");
#else /* old adv. API */
if (setsockopt(s, IPPROTO_IPV6, IPV6_RTHDR, &opton,
sizeof(opton)))
err(1, "setsockopt(IPV6_RTHDR)");
#endif
}
/*
optval = 1;
if (IN6_IS_ADDR_MULTICAST(&dst.sin6_addr))
if (setsockopt(s, IPPROTO_IPV6, IPV6_MULTICAST_LOOP,
&optval, sizeof(optval)) == -1)
err(1, "IPV6_MULTICAST_LOOP");
*/
/* Specify the outgoing interface and/or the source address */
if (usepktinfo)
ip6optlen += CMSG_SPACE(sizeof(struct in6_pktinfo));
if (hoplimit != -1)
ip6optlen += CMSG_SPACE(sizeof(int));
#ifdef IPV6_REACHCONF
if (options & F_REACHCONF)
ip6optlen += CMSG_SPACE(0);
#endif
/* set IP6 packet options */
if (ip6optlen) {
if ((scmsg = (char *)malloc(ip6optlen)) == 0)
errx(1, "can't allocate enough memory");
smsghdr.msg_control = (caddr_t)scmsg;
smsghdr.msg_controllen = ip6optlen;
scmsgp = (struct cmsghdr *)scmsg;
}
if (usepktinfo) {
pktinfo = (struct in6_pktinfo *)(CMSG_DATA(scmsgp));
memset(pktinfo, 0, sizeof(*pktinfo));
scmsgp->cmsg_len = CMSG_LEN(sizeof(struct in6_pktinfo));
scmsgp->cmsg_level = IPPROTO_IPV6;
scmsgp->cmsg_type = IPV6_PKTINFO;
scmsgp = CMSG_NXTHDR(&smsghdr, scmsgp);
}
/* set the outgoing interface */
if (ifname) {
#ifndef USE_SIN6_SCOPE_ID
/* pktinfo must have already been allocated */
if ((pktinfo->ipi6_ifindex = if_nametoindex(ifname)) == 0)
errx(1, "%s: invalid interface name", ifname);
#else
if ((dst.sin6_scope_id = if_nametoindex(ifname)) == 0)
errx(1, "%s: invalid interface name", ifname);
#endif
}
/* set the source address */
if (options & F_SRCADDR)/* pktinfo must be valid */
pktinfo->ipi6_addr = srcaddr;
if (hoplimit != -1) {
scmsgp->cmsg_len = CMSG_LEN(sizeof(int));
scmsgp->cmsg_level = IPPROTO_IPV6;
scmsgp->cmsg_type = IPV6_HOPLIMIT;
*(int *)(CMSG_DATA(scmsgp)) = hoplimit;
scmsgp = CMSG_NXTHDR(&smsghdr, scmsgp);
}
#ifdef IPV6_REACHCONF
if (options & F_REACHCONF) {
scmsgp->cmsg_len = CMSG_LEN(0);
scmsgp->cmsg_level = IPPROTO_IPV6;
scmsgp->cmsg_type = IPV6_REACHCONF;
scmsgp = CMSG_NXTHDR(&smsghdr, scmsgp);
}
#endif
if (argc > 1) { /* some intermediate addrs are specified */
int hops, error;
#ifdef USE_RFC2292BIS
int rthdrlen;
#endif
#ifdef USE_RFC2292BIS
rthdrlen = inet6_rth_space(IPV6_RTHDR_TYPE_0, argc - 1);
scmsgp->cmsg_len = CMSG_LEN(rthdrlen);
scmsgp->cmsg_level = IPPROTO_IPV6;
scmsgp->cmsg_type = IPV6_RTHDR;
rthdr = (struct ip6_rthdr *)CMSG_DATA(scmsgp);
rthdr = inet6_rth_init((void *)rthdr, rthdrlen,
IPV6_RTHDR_TYPE_0, argc - 1);
if (rthdr == NULL)
errx(1, "can't initialize rthdr");
#else /* old advanced API */
if ((scmsgp = (struct cmsghdr *)inet6_rthdr_init(scmsgp,
IPV6_RTHDR_TYPE_0)) == 0)
errx(1, "can't initialize rthdr");
#endif /* USE_RFC2292BIS */
for (hops = 0; hops < argc - 1; hops++) {
struct addrinfo *iaip;
if ((error = getaddrinfo(argv[hops], NULL, &hints,
&iaip)))
errx(1, "%s", gai_strerror(error));
if (SIN6(iaip->ai_addr)->sin6_family != AF_INET6)
errx(1,
"bad addr family of an intermediate addr");
#ifdef USE_RFC2292BIS
if (inet6_rth_add(rthdr,
&(SIN6(iaip->ai_addr))->sin6_addr))
errx(1, "can't add an intermediate node");
#else /* old advanced API */
if (inet6_rthdr_add(scmsgp,
&(SIN6(iaip->ai_addr))->sin6_addr,
IPV6_RTHDR_LOOSE))
errx(1, "can't add an intermediate node");
#endif /* USE_RFC2292BIS */
freeaddrinfo(iaip);
}
#ifndef USE_RFC2292BIS
if (inet6_rthdr_lasthop(scmsgp, IPV6_RTHDR_LOOSE))
errx(1, "can't set the last flag");
#endif
scmsgp = CMSG_NXTHDR(&smsghdr, scmsgp);
}
{
/*
* source selection
*/
int dummy, len = sizeof(src);
if ((dummy = socket(AF_INET6, SOCK_DGRAM, 0)) < 0)
err(1, "UDP socket");
src.sin6_family = AF_INET6;
src.sin6_addr = dst.sin6_addr;
src.sin6_port = ntohs(DUMMY_PORT);
src.sin6_scope_id = dst.sin6_scope_id;
#ifdef USE_RFC2292BIS
if (pktinfo &&
setsockopt(dummy, IPPROTO_IPV6, IPV6_PKTINFO,
(void *)pktinfo, sizeof(*pktinfo)))
err(1, "UDP setsockopt(IPV6_PKTINFO)");
if (hoplimit != -1 &&
setsockopt(dummy, IPPROTO_IPV6, IPV6_HOPLIMIT,
(void *)&hoplimit, sizeof(hoplimit)))
err(1, "UDP setsockopt(IPV6_HOPLIMIT)");
if (rthdr &&
setsockopt(dummy, IPPROTO_IPV6, IPV6_RTHDR,
(void *)rthdr, (rthdr->ip6r_len + 1) << 3))
err(1, "UDP setsockopt(IPV6_RTHDR)");
#else /* old advanced API */
if (smsghdr.msg_control &&
setsockopt(dummy, IPPROTO_IPV6, IPV6_PKTOPTIONS,
(void *)smsghdr.msg_control, smsghdr.msg_controllen))
err(1, "UDP setsockopt(IPV6_PKTOPTIONS)");
#endif
if (connect(dummy, (struct sockaddr *)&src, len) < 0)
err(1, "UDP connect");
if (getsockname(dummy, (struct sockaddr *)&src, &len) < 0)
err(1, "getsockname");
close(dummy);
}
#if defined(SO_SNDBUF) && defined(SO_RCVBUF)
if (sockbufsize) {
if (datalen > sockbufsize)
warnx("you need -b to increase socket buffer size");
if (setsockopt(s, SOL_SOCKET, SO_SNDBUF, &sockbufsize,
sizeof(sockbufsize)) < 0)
err(1, "setsockopt(SO_SNDBUF)");
if (setsockopt(s, SOL_SOCKET, SO_RCVBUF, &sockbufsize,
sizeof(sockbufsize)) < 0)
err(1, "setsockopt(SO_RCVBUF)");
}
else {
if (datalen > 8 * 1024) /*XXX*/
warnx("you need -b to increase socket buffer size");
/*
* When pinging the broadcast address, you can get a lot of
* answers. Doing something so evil is useful if you are trying
* to stress the ethernet, or just want to fill the arp cache
* to get some stuff for /etc/ethers.
*/
hold = 48 * 1024;
setsockopt(s, SOL_SOCKET, SO_RCVBUF, (char *)&hold,
sizeof(hold));
}
#endif
optval = 1;
#ifndef USE_SIN6_SCOPE_ID
#ifdef IPV6_RECVPKTINFO
if (setsockopt(s, IPPROTO_IPV6, IPV6_RECVPKTINFO, &optval,
sizeof(optval)) < 0)
warn("setsockopt(IPV6_RECVPKTINFO)"); /* XXX err? */
#else /* old adv. API */
if (setsockopt(s, IPPROTO_IPV6, IPV6_PKTINFO, &optval,
sizeof(optval)) < 0)
warn("setsockopt(IPV6_PKTINFO)"); /* XXX err? */
#endif
#endif /* USE_SIN6_SCOPE_ID */
#ifdef IPV6_RECVHOPLIMIT
if (setsockopt(s, IPPROTO_IPV6, IPV6_RECVHOPLIMIT, &optval,
sizeof(optval)) < 0)
warn("setsockopt(IPV6_RECVHOPLIMIT)"); /* XXX err? */
#else /* old adv. API */
if (setsockopt(s, IPPROTO_IPV6, IPV6_HOPLIMIT, &optval,
sizeof(optval)) < 0)
warn("setsockopt(IPV6_HOPLIMIT)"); /* XXX err? */
#endif
printf("PING6(%lu=40+8+%lu bytes) ", (unsigned long)(40 + pingerlen()),
(unsigned long)(pingerlen() - 8));
printf("%s --> ", pr_addr((struct sockaddr *)&src, sizeof(src)));
printf("%s\n", pr_addr((struct sockaddr *)&dst, sizeof(dst)));
while (preload--) /* Fire off them quickies. */
(void)pinger();
(void)signal(SIGINT, onsignal);
#ifdef SIGINFO
(void)signal(SIGINFO, onsignal);
#endif
if ((options & F_FLOOD) == 0) {
(void)signal(SIGALRM, onsignal);
itimer.it_interval = interval;
itimer.it_value = interval;
(void)setitimer(ITIMER_REAL, &itimer, NULL);
if (ntransmitted == 0)
retransmit();
}
fdmasks = howmany(s + 1, NFDBITS) * sizeof(fd_mask);
if ((fdmaskp = malloc(fdmasks)) == NULL)
err(1, "malloc");
signo = seenalrm = seenint = 0;
#ifdef SIGINFO
seeninfo = 0;
#endif
for (;;) {
struct msghdr m;
struct cmsghdr *cm;
u_char buf[1024];
struct iovec iov[2];
/* signal handling */
if (seenalrm) {
retransmit();
seenalrm = 0;
continue;
}
if (seenint) {
onint(SIGINT);
seenint = 0;
continue;
}
#ifdef SIGINFO
if (seeninfo) {
summary();
seeninfo = 0;
continue;
}
#endif
if (options & F_FLOOD) {
(void)pinger();
timeout.tv_sec = 0;
timeout.tv_usec = 10000;
tv = &timeout;
} else
tv = NULL;
memset(fdmaskp, 0, fdmasks);
FD_SET(s, fdmaskp);
cc = select(s + 1, fdmaskp, NULL, NULL, tv);
if (cc < 0) {
if (errno != EINTR) {
warn("select");
sleep(1);
}
continue;
} else if (cc == 0)
continue;
fromlen = sizeof(from);
m.msg_name = (caddr_t)&from;
m.msg_namelen = sizeof(from);
memset(&iov, 0, sizeof(iov));
iov[0].iov_base = (caddr_t)packet;
iov[0].iov_len = packlen;
m.msg_iov = iov;
m.msg_iovlen = 1;
cm = (struct cmsghdr *)buf;
m.msg_control = (caddr_t)buf;
m.msg_controllen = sizeof(buf);
cc = recvmsg(s, &m, 0);
if (cc < 0) {
if (errno != EINTR) {
warn("recvmsg");
sleep(1);
}
continue;
} else if (cc == 0) {
int mtu;
/*
* receive control messages only. Process the
* exceptions (currently the only possiblity is
* a path MTU notification.)
*/
if ((mtu = get_pathmtu(&m)) > 0) {
if ((options & F_VERBOSE) != 0) {
printf("new path MTU (%d) is "
"notified\n", mtu);
}
set_pathmtu(mtu);
}
continue;
} else {
/*
* an ICMPv6 message (probably an echoreply) arrived.
*/
pr_pack(packet, cc, &m);
}
if (npackets && nreceived >= npackets)
break;
}
summary();
exit(nreceived == 0);
}
void
onsignal(sig)
int sig;
{
signo = sig;
switch (sig) {
case SIGALRM:
seenalrm++;
break;
case SIGINT:
seenint++;
break;
#ifdef SIGINFO
case SIGINFO:
seeninfo++;
break;
#endif
}
}
/*
* retransmit --
* This routine transmits another ping6.
*/
void
retransmit()
{
struct itimerval itimer;
if (pinger() == 0)
return;
/*
* If we're not transmitting any more packets, change the timer
* to wait two round-trip times if we've received any packets or
* ten seconds if we haven't.
*/
#define MAXWAIT 10
if (nreceived) {
itimer.it_value.tv_sec = 2 * tmax / 1000;
if (itimer.it_value.tv_sec == 0)
itimer.it_value.tv_sec = 1;
} else
itimer.it_value.tv_sec = MAXWAIT;
itimer.it_interval.tv_sec = 0;
itimer.it_interval.tv_usec = 0;
itimer.it_value.tv_usec = 0;
(void)signal(SIGALRM, onint);
(void)setitimer(ITIMER_REAL, &itimer, NULL);
}
/*
* pinger --
* Compose and transmit an ICMP ECHO REQUEST packet. The IP packet
* will be added on by the kernel. The ID field is our UNIX process ID,
* and the sequence number is an ascending integer. The first 8 bytes
* of the data portion are used to hold a UNIX "timeval" struct in VAX
* byte-order, to compute the round-trip time.
*/
size_t
pingerlen()
{
size_t l;
if (options & F_FQDN)
l = ICMP6_NIQLEN + sizeof(dst.sin6_addr);
else if (options & F_FQDNOLD)
l = ICMP6_NIQLEN;
else if (options & F_NODEADDR)
l = ICMP6_NIQLEN + sizeof(dst.sin6_addr);
else if (options & F_SUPTYPES)
l = ICMP6_NIQLEN;
else
l = ICMP6ECHOLEN + datalen;
return l;
}
int
pinger()
{
struct icmp6_hdr *icp;
struct iovec iov[2];
int i, cc;
struct icmp6_nodeinfo *nip;
int seq;
if (npackets && ntransmitted >= npackets)
return(-1); /* no more transmission */
icp = (struct icmp6_hdr *)outpack;
nip = (struct icmp6_nodeinfo *)outpack;
memset(icp, 0, sizeof(*icp));
icp->icmp6_cksum = 0;
seq = ntransmitted++;
CLR(seq % mx_dup_ck);
if (options & F_FQDN) {
icp->icmp6_type = ICMP6_NI_QUERY;
icp->icmp6_code = ICMP6_NI_SUBJ_IPV6;
nip->ni_qtype = htons(NI_QTYPE_FQDN);
nip->ni_flags = htons(0);
memcpy(nip->icmp6_ni_nonce, nonce,
sizeof(nip->icmp6_ni_nonce));
*(u_int16_t *)nip->icmp6_ni_nonce = ntohs(seq);
memcpy(&outpack[ICMP6_NIQLEN], &dst.sin6_addr,
sizeof(dst.sin6_addr));
cc = ICMP6_NIQLEN + sizeof(dst.sin6_addr);
datalen = 0;
} else if (options & F_FQDNOLD) {
/* packet format in 03 draft - no Subject data on queries */
icp->icmp6_type = ICMP6_NI_QUERY;
icp->icmp6_code = 0; /* code field is always 0 */
nip->ni_qtype = htons(NI_QTYPE_FQDN);
nip->ni_flags = htons(0);
memcpy(nip->icmp6_ni_nonce, nonce,
sizeof(nip->icmp6_ni_nonce));
*(u_int16_t *)nip->icmp6_ni_nonce = ntohs(seq);
cc = ICMP6_NIQLEN;
datalen = 0;
} else if (options & F_NODEADDR) {
icp->icmp6_type = ICMP6_NI_QUERY;
icp->icmp6_code = ICMP6_NI_SUBJ_IPV6;
nip->ni_qtype = htons(NI_QTYPE_NODEADDR);
nip->ni_flags = naflags;
memcpy(nip->icmp6_ni_nonce, nonce,
sizeof(nip->icmp6_ni_nonce));
*(u_int16_t *)nip->icmp6_ni_nonce = ntohs(seq);
memcpy(&outpack[ICMP6_NIQLEN], &dst.sin6_addr,
sizeof(dst.sin6_addr));
cc = ICMP6_NIQLEN + sizeof(dst.sin6_addr);
datalen = 0;
} else if (options & F_SUPTYPES) {
icp->icmp6_type = ICMP6_NI_QUERY;
icp->icmp6_code = ICMP6_NI_SUBJ_FQDN; /*empty*/
nip->ni_qtype = htons(NI_QTYPE_SUPTYPES);
/* we support compressed bitmap */
nip->ni_flags = NI_SUPTYPE_FLAG_COMPRESS;
memcpy(nip->icmp6_ni_nonce, nonce,
sizeof(nip->icmp6_ni_nonce));
*(u_int16_t *)nip->icmp6_ni_nonce = ntohs(seq);
cc = ICMP6_NIQLEN;
datalen = 0;
} else {
icp->icmp6_type = ICMP6_ECHO_REQUEST;
icp->icmp6_code = 0;
icp->icmp6_id = htons(ident);
icp->icmp6_seq = ntohs(seq);
if (timing)
(void)gettimeofday((struct timeval *)
&outpack[ICMP6ECHOLEN], NULL);
cc = ICMP6ECHOLEN + datalen;
}
#ifdef DIAGNOSTIC
if (pingerlen() != cc)
errx(1, "internal error; length mismatch");
#endif
smsghdr.msg_name = (caddr_t)&dst;
smsghdr.msg_namelen = sizeof(dst);
memset(&iov, 0, sizeof(iov));
iov[0].iov_base = (caddr_t)outpack;
iov[0].iov_len = cc;
smsghdr.msg_iov = iov;
smsghdr.msg_iovlen = 1;
i = sendmsg(s, &smsghdr, 0);
if (i < 0 || i != cc) {
if (i < 0)
warn("sendmsg");
(void)printf("ping6: wrote %s %d chars, ret=%d\n",
hostname, cc, i);
}
if (!(options & F_QUIET) && options & F_FLOOD)
(void)write(STDOUT_FILENO, &DOT, 1);
return(0);
}
int
myechoreply(icp)
const struct icmp6_hdr *icp;
{
if (ntohs(icp->icmp6_id) == ident)
return 1;
else
return 0;
}
int
mynireply(nip)
const struct icmp6_nodeinfo *nip;
{
if (memcmp(nip->icmp6_ni_nonce + sizeof(u_int16_t),
nonce + sizeof(u_int16_t),
sizeof(nonce) - sizeof(u_int16_t)) == 0)
return 1;
else
return 0;
}
char *
dnsdecode(sp, ep, base, buf, bufsiz)
const u_char **sp;
const u_char *ep;
const u_char *base; /*base for compressed name*/
u_char *buf;
size_t bufsiz;
{
int i;
const u_char *cp;
char cresult[MAXDNAME + 1];
const u_char *comp;
int l;
cp = *sp;
*buf = '\0';
if (cp >= ep)
return NULL;
while (cp < ep) {
i = *cp;
if (i == 0 || cp != *sp) {
if (strlcat(buf, ".", bufsiz) >= bufsiz)
return NULL; /*result overrun*/
}
if (i == 0)
break;
cp++;
if ((i & 0xc0) == 0xc0 && cp - base > (i & 0x3f)) {
/* DNS compression */
if (!base)
return NULL;
comp = base + (i & 0x3f);
if (dnsdecode(&comp, cp, base, cresult,
sizeof(cresult)) == NULL)
return NULL;
if (strlcat(buf, cresult, bufsiz) >= bufsiz)
return NULL; /*result overrun*/
break;
} else if ((i & 0x3f) == i) {
if (i > ep - cp)
return NULL; /*source overrun*/
while (i-- > 0 && cp < ep) {
l = snprintf(cresult, sizeof(cresult),
isprint(*cp) ? "%c" : "\\%03o", *cp & 0xff);
if (l < 0 || l >= sizeof(cresult))
return NULL;
if (strlcat(buf, cresult, bufsiz) >= bufsiz)
return NULL; /*result overrun*/
cp++;
}
} else
return NULL; /*invalid label*/
}
if (i != 0)
return NULL; /*not terminated*/
cp++;
*sp = cp;
return buf;
}
/*
* pr_pack --
* Print out the packet, if it came from us. This logic is necessary
* because ALL readers of the ICMP socket get a copy of ALL ICMP packets
* which arrive ('tis only fair). This permits multiple copies of this
* program to be run without having intermingled output (or statistics!).
*/
void
pr_pack(buf, cc, mhdr)
u_char *buf;
int cc;
struct msghdr *mhdr;
{
#define safeputc(c) printf((isprint((c)) ? "%c" : "\\%03o"), c)
struct icmp6_hdr *icp;
struct icmp6_nodeinfo *ni;
int i;
int hoplim;
struct sockaddr *from;
int fromlen;
u_char *cp = NULL, *dp, *end = buf + cc;
struct in6_pktinfo *pktinfo = NULL;
struct timeval tv, *tp;
double triptime = 0;
int dupflag;
size_t off;
int oldfqdn;
u_int16_t seq;
char dnsname[MAXDNAME + 1];
(void)gettimeofday(&tv, NULL);
if (!mhdr || !mhdr->msg_name ||
mhdr->msg_namelen != sizeof(struct sockaddr_in6) ||
((struct sockaddr *)mhdr->msg_name)->sa_family != AF_INET6) {
if (options & F_VERBOSE)
warnx("invalid peername\n");
return;
}
from = (struct sockaddr *)mhdr->msg_name;
fromlen = mhdr->msg_namelen;
if (cc < sizeof(struct icmp6_hdr)) {
if (options & F_VERBOSE)
warnx("packet too short (%d bytes) from %s", cc,
pr_addr(from, fromlen));
return;
}
icp = (struct icmp6_hdr *)buf;
ni = (struct icmp6_nodeinfo *)buf;
off = 0;
if ((hoplim = get_hoplim(mhdr)) == -1) {
warnx("failed to get receiving hop limit");
return;
}
if ((pktinfo = get_rcvpktinfo(mhdr)) == NULL) {
warnx("failed to get receiving packet information");
return;
}
if (icp->icmp6_type == ICMP6_ECHO_REPLY && myechoreply(icp)) {
seq = ntohs(icp->icmp6_seq);
++nreceived;
if (timing) {
tp = (struct timeval *)(icp + 1);
tvsub(&tv, tp);
triptime = ((double)tv.tv_sec) * 1000.0 +
((double)tv.tv_usec) / 1000.0;
tsum += triptime;
#if defined(__OpenBSD__) || defined(__NetBSD__) || defined(__FreeBSD__)
tsumsq += triptime * triptime;
#endif
if (triptime < tmin)
tmin = triptime;
if (triptime > tmax)
tmax = triptime;
}
if (TST(seq % mx_dup_ck)) {
++nrepeats;
--nreceived;
dupflag = 1;
} else {
SET(seq % mx_dup_ck);
dupflag = 0;
}
if (options & F_QUIET)
return;
if (options & F_FLOOD)
(void)write(STDOUT_FILENO, &BSPACE, 1);
else {
(void)printf("%d bytes from %s, icmp_seq=%u", cc,
pr_addr(from, fromlen), seq);
(void)printf(" hlim=%d", hoplim);
if ((options & F_VERBOSE) != 0) {
struct sockaddr_in6 dstsa;
memset(&dstsa, 0, sizeof(dstsa));
dstsa.sin6_family = AF_INET6;
#ifdef SIN6_LEN
dstsa.sin6_len = sizeof(dstsa);
#endif
dstsa.sin6_scope_id = pktinfo->ipi6_ifindex;
dstsa.sin6_addr = pktinfo->ipi6_addr;
(void)printf(" dst=%s",
pr_addr((struct sockaddr *)&dstsa,
sizeof(dstsa)));
}
if (timing)
(void)printf(" time=%g ms", triptime);
if (dupflag)
(void)printf("(DUP!)");
/* check the data */
cp = buf + off + ICMP6ECHOLEN + ICMP6ECHOTMLEN;
dp = outpack + ICMP6ECHOLEN + ICMP6ECHOTMLEN;
for (i = 8; cp < end; ++i, ++cp, ++dp) {
if (*cp != *dp) {
(void)printf("\nwrong data byte #%d should be 0x%x but was 0x%x", i, *dp, *cp);
break;
}
}
}
} else if (icp->icmp6_type == ICMP6_NI_REPLY && mynireply(ni)) {
seq = ntohs(*(u_int16_t *)ni->icmp6_ni_nonce);
++nreceived;
if (TST(seq % mx_dup_ck)) {
++nrepeats;
--nreceived;
dupflag = 1;
} else {
SET(seq % mx_dup_ck);
dupflag = 0;
}
if (options & F_QUIET)
return;
(void)printf("%d bytes from %s: ", cc, pr_addr(from, fromlen));
switch (ntohs(ni->ni_code)) {
case ICMP6_NI_SUCCESS:
break;
case ICMP6_NI_REFUSED:
printf("refused, type 0x%x", ntohs(ni->ni_type));
goto fqdnend;
case ICMP6_NI_UNKNOWN:
printf("unknown, type 0x%x", ntohs(ni->ni_type));
goto fqdnend;
default:
printf("unknown code 0x%x, type 0x%x",
ntohs(ni->ni_code), ntohs(ni->ni_type));
goto fqdnend;
}
switch (ntohs(ni->ni_qtype)) {
case NI_QTYPE_NOOP:
printf("NodeInfo NOOP");
break;
case NI_QTYPE_SUPTYPES:
pr_suptypes(ni, end - (u_char *)ni);
break;
case NI_QTYPE_NODEADDR:
pr_nodeaddr(ni, end - (u_char *)ni);
break;
case NI_QTYPE_FQDN:
default: /* XXX: for backward compatibility */
cp = (u_char *)ni + ICMP6_NIRLEN;
if (buf[off + ICMP6_NIRLEN] ==
cc - off - ICMP6_NIRLEN - 1)
oldfqdn = 1;
else
oldfqdn = 0;
if (oldfqdn) {
cp++; /* skip length */
while (cp < end) {
safeputc(*cp & 0xff);
cp++;
}
} else {
i = 0;
while (cp < end) {
if (dnsdecode((const u_char **)&cp, end,
(const u_char *)(ni + 1), dnsname,
sizeof(dnsname)) == NULL) {
printf("???");
break;
}
/*
* name-lookup special handling for
* truncated name
*/
if (cp + 1 <= end && !*cp &&
strlen(dnsname) > 0) {
dnsname[strlen(dnsname) - 1] = '\0';
cp++;
}
printf("%s%s", i > 0 ? "," : "",
dnsname);
}
}
if (options & F_VERBOSE) {
int32_t ttl;
int comma = 0;
(void)printf(" ("); /*)*/
switch (ni->ni_code) {
case ICMP6_NI_REFUSED:
(void)printf("refused");
comma++;
break;
case ICMP6_NI_UNKNOWN:
(void)printf("unknown qtype");
comma++;
break;
}
if ((end - (u_char *)ni) < ICMP6_NIRLEN) {
/* case of refusion, unknown */
/*(*/
putchar(')');
goto fqdnend;
}
ttl = (int32_t)ntohl(*(u_long *)&buf[off+ICMP6ECHOLEN+8]);
if (comma)
printf(",");
if (!(ni->ni_flags & NI_FQDN_FLAG_VALIDTTL)) {
(void)printf("TTL=%d:meaningless",
(int)ttl);
} else {
if (ttl < 0) {
(void)printf("TTL=%d:invalid",
ttl);
} else
(void)printf("TTL=%d", ttl);
}
comma++;
if (oldfqdn) {
if (comma)
printf(",");
printf("03 draft");
comma++;
} else {
cp = (u_char *)ni + ICMP6_NIRLEN;
if (cp == end) {
if (comma)
printf(",");
printf("no name");
comma++;
}
}
if (buf[off + ICMP6_NIRLEN] !=
cc - off - ICMP6_NIRLEN - 1 && oldfqdn) {
if (comma)
printf(",");
(void)printf("invalid namelen:%d/%lu",
buf[off + ICMP6_NIRLEN],
(u_long)cc - off - ICMP6_NIRLEN - 1);
comma++;
}
/*(*/
putchar(')');
}
fqdnend:
;
}
} else {
/* We've got something other than an ECHOREPLY */
if (!(options & F_VERBOSE))
return;
(void)printf("%d bytes from %s: ", cc, pr_addr(from, fromlen));
pr_icmph(icp, end);
}
if (!(options & F_FLOOD)) {
(void)putchar('\n');
if (options & F_VERBOSE)
pr_exthdrs(mhdr);
(void)fflush(stdout);
}
#undef safeputc
}
void
pr_exthdrs(mhdr)
struct msghdr *mhdr;
{
struct cmsghdr *cm;
for (cm = (struct cmsghdr *)CMSG_FIRSTHDR(mhdr); cm;
cm = (struct cmsghdr *)CMSG_NXTHDR(mhdr, cm)) {
if (cm->cmsg_level != IPPROTO_IPV6)
continue;
switch (cm->cmsg_type) {
case IPV6_HOPOPTS:
printf(" HbH Options: ");
pr_ip6opt(CMSG_DATA(cm));
break;
case IPV6_DSTOPTS:
#ifdef IPV6_RTHDRDSTOPTS
case IPV6_RTHDRDSTOPTS:
#endif
printf(" Dst Options: ");
pr_ip6opt(CMSG_DATA(cm));
break;
case IPV6_RTHDR:
printf(" Routing: ");
pr_rthdr(CMSG_DATA(cm));
break;
}
}
}
#ifdef USE_RFC2292BIS
void
pr_ip6opt(void *extbuf)
{
struct ip6_hbh *ext;
int currentlen;
u_int8_t type;
size_t extlen, len;
void *databuf;
size_t offset;
u_int16_t value2;
u_int32_t value4;
ext = (struct ip6_hbh *)extbuf;
extlen = (ext->ip6h_len + 1) * 8;
printf("nxt %u, len %u (%lu bytes)\n", ext->ip6h_nxt,
(unsigned int)ext->ip6h_len, (unsigned long)extlen);
currentlen = 0;
while (1) {
currentlen = inet6_opt_next(extbuf, extlen, currentlen,
&type, &len, &databuf);
if (currentlen == -1)
break;
switch (type) {
/*
* Note that inet6_opt_next automatically skips any padding
* optins.
*/
case IP6OPT_JUMBO:
offset = 0;
offset = inet6_opt_get_val(databuf, offset,
&value4, sizeof(value4));
printf(" Jumbo Payload Opt: Length %u\n",
(u_int32_t)ntohl(value4));
break;
case IP6OPT_ROUTER_ALERT:
offset = 0;
offset = inet6_opt_get_val(databuf, offset,
&value2, sizeof(value2));
printf(" Router Alert Opt: Type %u\n",
ntohs(value2));
break;
default:
printf(" Received Opt %u len %lu\n",
type, (unsigned long)len);
break;
}
}
return;
}
#else /* !USE_RFC2292BIS */
/* ARGSUSED */
void
pr_ip6opt(void *extbuf)
{
putchar('\n');
return;
}
#endif /* USE_RFC2292BIS */
#ifdef USE_RFC2292BIS
void
pr_rthdr(void *extbuf)
{
struct in6_addr *in6;
char ntopbuf[INET6_ADDRSTRLEN];
struct ip6_rthdr *rh = (struct ip6_rthdr *)extbuf;
int i, segments;
/* print fixed part of the header */
printf("nxt %u, len %u (%d bytes), type %u, ", rh->ip6r_nxt,
rh->ip6r_len, (rh->ip6r_len + 1) << 3, rh->ip6r_type);
if ((segments = inet6_rth_segments(extbuf)) >= 0)
printf("%d segments, ", segments);
else
printf("segments unknown, ");
printf("%d left\n", rh->ip6r_segleft);
for (i = 0; i < segments; i++) {
in6 = inet6_rth_getaddr(extbuf, i);
if (in6 == NULL)
printf(" [%d]<NULL>\n", i);
else {
if (!inet_ntop(AF_INET6, in6, ntopbuf,
sizeof(ntopbuf)))
strncpy(ntopbuf, "?", sizeof(ntopbuf));
printf(" [%d]%s\n", i, ntopbuf);
}
}
return;
}
#else /* !USE_RFC2292BIS */
/* ARGSUSED */
void
pr_rthdr(void *extbuf)
{
putchar('\n');
return;
}
#endif /* USE_RFC2292BIS */
int
pr_bitrange(v, s, ii)
u_int32_t v;
int s;
int ii;
{
int off;
int i;
off = 0;
while (off < 32) {
/* shift till we have 0x01 */
if ((v & 0x01) == 0) {
if (ii > 1)
printf("-%u", s + off - 1);
ii = 0;
switch (v & 0x0f) {
case 0x00:
v >>= 4;
off += 4;
continue;
case 0x08:
v >>= 3;
off += 3;
continue;
case 0x04: case 0x0c:
v >>= 2;
off += 2;
continue;
default:
v >>= 1;
off += 1;
continue;
}
}
/* we have 0x01 with us */
for (i = 0; i < 32 - off; i++) {
if ((v & (0x01 << i)) == 0)
break;
}
if (!ii)
printf(" %u", s + off);
ii += i;
v >>= i; off += i;
}
return ii;
}
void
pr_suptypes(ni, nilen)
struct icmp6_nodeinfo *ni; /* ni->qtype must be SUPTYPES */
size_t nilen;
{
size_t clen;
u_int32_t v;
const u_char *cp, *end;
u_int16_t cur;
struct cbit {
u_int16_t words; /*32bit count*/
u_int16_t skip;
} cbit;
#define MAXQTYPES (1 << 16)
size_t off;
int b;
cp = (u_char *)(ni + 1);
end = ((u_char *)ni) + nilen;
cur = 0;
b = 0;
printf("NodeInfo Supported Qtypes");
if (options & F_VERBOSE) {
if (ni->ni_flags & NI_SUPTYPE_FLAG_COMPRESS)
printf(", compressed bitmap");
else
printf(", raw bitmap");
}
while (cp < end) {
clen = (size_t)(end - cp);
if ((ni->ni_flags & NI_SUPTYPE_FLAG_COMPRESS) == 0) {
if (clen == 0 || clen > MAXQTYPES / 8 ||
clen % sizeof(v)) {
printf("???");
return;
}
} else {
if (clen < sizeof(cbit) || clen % sizeof(v))
return;
memcpy(&cbit, cp, sizeof(cbit));
if (sizeof(cbit) + ntohs(cbit.words) * sizeof(v) >
clen)
return;
cp += sizeof(cbit);
clen = ntohs(cbit.words) * sizeof(v);
if (cur + clen * 8 + (u_long)ntohs(cbit.skip) * 32 >
MAXQTYPES)
return;
}
for (off = 0; off < clen; off += sizeof(v)) {
memcpy(&v, cp + off, sizeof(v));
v = (u_int32_t)ntohl(v);
b = pr_bitrange(v, (int)(cur + off * 8), b);
}
/* flush the remaining bits */
b = pr_bitrange(0, (int)(cur + off * 8), b);
cp += clen;
cur += clen * 8;
if ((ni->ni_flags & NI_SUPTYPE_FLAG_COMPRESS) != 0)
cur += ntohs(cbit.skip) * 32;
}
}
void
pr_nodeaddr(ni, nilen)
struct icmp6_nodeinfo *ni; /* ni->qtype must be NODEADDR */
int nilen;
{
u_char *cp = (u_char *)(ni + 1);
char ntop_buf[INET6_ADDRSTRLEN];
int withttl = 0;
nilen -= sizeof(struct icmp6_nodeinfo);
if (options & F_VERBOSE) {
switch (ni->ni_code) {
case ICMP6_NI_REFUSED:
(void)printf("refused");
break;
case ICMP6_NI_UNKNOWN:
(void)printf("unknown qtype");
break;
}
if (ni->ni_flags & NI_NODEADDR_FLAG_TRUNCATE)
(void)printf(" truncated");
}
putchar('\n');
if (nilen <= 0)
printf(" no address\n");
/*
* In icmp-name-lookups 05 and later, TTL of each returned address
* is contained in the resposne. We try to detect the version
* by the length of the data, but note that the detection algorithm
* is incomplete. We assume the latest draft by default.
*/
if (nilen % (sizeof(u_int32_t) + sizeof(struct in6_addr)) == 0)
withttl = 1;
while (nilen > 0) {
u_int32_t ttl;
if (withttl) {
/* XXX: alignment? */
ttl = (u_int32_t)ntohl(*(u_int32_t *)cp);
cp += sizeof(u_int32_t);
nilen -= sizeof(u_int32_t);
}
if (inet_ntop(AF_INET6, cp, ntop_buf, sizeof(ntop_buf)) ==
NULL)
strncpy(ntop_buf, "?", sizeof(ntop_buf));
printf(" %s", ntop_buf);
if (withttl) {
if (ttl == 0xffffffff) {
/*
* XXX: can this convention be applied to all
* type of TTL (i.e. non-ND TTL)?
*/
printf("(TTL=infty)");
}
else
printf("(TTL=%u)", ttl);
}
putchar('\n');
nilen -= sizeof(struct in6_addr);
cp += sizeof(struct in6_addr);
}
}
int
get_hoplim(mhdr)
struct msghdr *mhdr;
{
struct cmsghdr *cm;
for (cm = (struct cmsghdr *)CMSG_FIRSTHDR(mhdr); cm;
cm = (struct cmsghdr *)CMSG_NXTHDR(mhdr, cm)) {
if (cm->cmsg_len == 0)
return(-1);
if (cm->cmsg_level == IPPROTO_IPV6 &&
cm->cmsg_type == IPV6_HOPLIMIT &&
cm->cmsg_len == CMSG_LEN(sizeof(int)))
return(*(int *)CMSG_DATA(cm));
}
return(-1);
}
struct in6_pktinfo *
get_rcvpktinfo(mhdr)
struct msghdr *mhdr;
{
struct cmsghdr *cm;
for (cm = (struct cmsghdr *)CMSG_FIRSTHDR(mhdr); cm;
cm = (struct cmsghdr *)CMSG_NXTHDR(mhdr, cm)) {
if (cm->cmsg_len == 0)
return(NULL);
if (cm->cmsg_level == IPPROTO_IPV6 &&
cm->cmsg_type == IPV6_PKTINFO &&
cm->cmsg_len == CMSG_LEN(sizeof(struct in6_pktinfo)))
return((struct in6_pktinfo *)CMSG_DATA(cm));
}
return(NULL);
}
int
get_pathmtu(mhdr)
struct msghdr *mhdr;
{
#ifdef IPV6_RECVPATHMTU
struct cmsghdr *cm;
struct ip6_mtuinfo *mtuctl = NULL;
for (cm = (struct cmsghdr *)CMSG_FIRSTHDR(mhdr); cm;
cm = (struct cmsghdr *)CMSG_NXTHDR(mhdr, cm)) {
if (cm->cmsg_len == 0)
return(0);
if (cm->cmsg_level == IPPROTO_IPV6 &&
cm->cmsg_type == IPV6_PATHMTU &&
cm->cmsg_len == CMSG_LEN(sizeof(struct ip6_mtuinfo))) {
mtuctl = (struct ip6_mtuinfo *)CMSG_DATA(cm);
/*
* If the notified destination is different from
* the one we are pinging, just ignore the info.
* We check the scope ID only when both notified value
* and our own value have non-0 values, because we may
* have used the default scope zone ID for sending,
* in which case the scope ID value is 0.
*/
if (!IN6_ARE_ADDR_EQUAL(&mtuctl->ip6m_addr.sin6_addr,
&dst.sin6_addr) ||
(mtuctl->ip6m_addr.sin6_scope_id &&
dst.sin6_scope_id &&
mtuctl->ip6m_addr.sin6_scope_id !=
dst.sin6_scope_id)) {
if ((options & F_VERBOSE) != 0) {
printf("path MTU for %s is notified. "
"(ignored)\n",
pr_addr((struct sockaddr *)&mtuctl->ip6m_addr,
sizeof(mtuctl->ip6m_addr)));
}
return(0);
}
/*
* Ignore an invalid MTU. XXX: can we just believe
* the kernel check?
*/
if (mtuctl->ip6m_mtu < IPV6_MMTU)
return(0);
/* notification for our destination. return the MTU. */
return((int)mtuctl->ip6m_mtu);
}
}
#endif
return(0);
}
void
set_pathmtu(mtu)
int mtu;
{
#ifdef IPV6_USE_MTU
static int firsttime = 1;
struct cmsghdr *cm;
if (firsttime) {
int oldlen = smsghdr.msg_controllen;
char *oldbuf = smsghdr.msg_control;
/* XXX: We need to enlarge control message buffer */
firsttime = 0; /* prevent further enlargement */
smsghdr.msg_controllen = oldlen + CMSG_SPACE(sizeof(int));
if ((smsghdr.msg_control =
(char *)malloc(smsghdr.msg_controllen)) == NULL)
err(1, "set_pathmtu: malloc");
cm = (struct cmsghdr *)CMSG_FIRSTHDR(&smsghdr);
cm->cmsg_len = CMSG_LEN(sizeof(int));
cm->cmsg_level = IPPROTO_IPV6;
cm->cmsg_type = IPV6_USE_MTU;
cm = (struct cmsghdr *)CMSG_NXTHDR(&smsghdr, cm);
if (oldlen)
memcpy((void *)cm, (void *)oldbuf, oldlen);
free(oldbuf);
}
/*
* look for a cmsgptr that points MTU structure.
* XXX: this procedure seems redundant at this moment, but we'd better
* keep the code generic enough for future extensions.
*/
for (cm = CMSG_FIRSTHDR(&smsghdr); cm;
cm = (struct cmsghdr *)CMSG_NXTHDR(&smsghdr, cm)) {
if (cm->cmsg_len == 0) /* XXX: paranoid check */
errx(1, "set_pathmtu: internal error");
if (cm->cmsg_level == IPPROTO_IPV6 &&
cm->cmsg_type == IPV6_USE_MTU &&
cm->cmsg_len == CMSG_LEN(sizeof(int)))
break;
}
if (cm == NULL)
errx(1, "set_pathmtu: internal error: no space for path MTU");
*(int *)CMSG_DATA(cm) = mtu;
#endif
}
/*
* tvsub --
* Subtract 2 timeval structs: out = out - in. Out is assumed to
* be >= in.
*/
void
tvsub(out, in)
register struct timeval *out, *in;
{
if ((out->tv_usec -= in->tv_usec) < 0) {
--out->tv_sec;
out->tv_usec += 1000000;
}
out->tv_sec -= in->tv_sec;
}
/*
* onint --
* SIGINT handler.
*/
/* ARGSUSED */
void
onint(notused)
int notused;
{
summary();
(void)signal(SIGINT, SIG_DFL);
(void)kill(getpid(), SIGINT);
/* NOTREACHED */
exit(1);
}
/*
* summary --
* Print out statistics.
*/
void
summary()
{
(void)printf("\n--- %s ping6 statistics ---\n", hostname);
(void)printf("%ld packets transmitted, ", ntransmitted);
(void)printf("%ld packets received, ", nreceived);
if (nrepeats)
(void)printf("+%ld duplicates, ", nrepeats);
if (ntransmitted) {
if (nreceived > ntransmitted)
(void)printf("-- somebody's printing up packets!");
else
(void)printf("%d%% packet loss",
(int) (((ntransmitted - nreceived) * 100) /
ntransmitted));
}
(void)putchar('\n');
if (nreceived && timing) {
/* Only display average to microseconds */
double num = nreceived + nrepeats;
double avg = tsum / num;
#if defined(__OpenBSD__) || defined(__NetBSD__) || defined(__FreeBSD__)
double dev = sqrt(tsumsq / num - avg * avg);
(void)printf(
"round-trip min/avg/max/std-dev = %.3f/%.3f/%.3f/%.3f ms\n",
tmin, avg, tmax, dev);
#else
(void)printf(
"round-trip min/avg/max = %.3f/%.3f/%.3f ms\n",
tmin, avg, tmax);
#endif
(void)fflush(stdout);
}
(void)fflush(stdout);
}
/*subject type*/
static char *niqcode[] = {
"IPv6 address",
"DNS label", /*or empty*/
"IPv4 address",
};
/*result code*/
static char *nircode[] = {
"Success", "Refused", "Unknown",
};
/*
* pr_icmph --
* Print a descriptive string about an ICMP header.
*/
void
pr_icmph(icp, end)
struct icmp6_hdr *icp;
u_char *end;
{
char ntop_buf[INET6_ADDRSTRLEN];
struct nd_redirect *red;
struct icmp6_nodeinfo *ni;
char dnsname[MAXDNAME + 1];
const u_char *cp;
size_t l;
switch (icp->icmp6_type) {
case ICMP6_DST_UNREACH:
switch (icp->icmp6_code) {
case ICMP6_DST_UNREACH_NOROUTE:
(void)printf("No Route to Destination\n");
break;
case ICMP6_DST_UNREACH_ADMIN:
(void)printf("Destination Administratively "
"Unreachable\n");
break;
case ICMP6_DST_UNREACH_BEYONDSCOPE:
(void)printf("Destination Unreachable Beyond Scope\n");
break;
case ICMP6_DST_UNREACH_ADDR:
(void)printf("Destination Host Unreachable\n");
break;
case ICMP6_DST_UNREACH_NOPORT:
(void)printf("Destination Port Unreachable\n");
break;
default:
(void)printf("Destination Unreachable, Bad Code: %d\n",
icp->icmp6_code);
break;
}
/* Print returned IP header information */
pr_retip((struct ip6_hdr *)(icp + 1), end);
break;
case ICMP6_PACKET_TOO_BIG:
(void)printf("Packet too big mtu = %d\n",
(int)ntohl(icp->icmp6_mtu));
pr_retip((struct ip6_hdr *)(icp + 1), end);
break;
case ICMP6_TIME_EXCEEDED:
switch (icp->icmp6_code) {
case ICMP6_TIME_EXCEED_TRANSIT:
(void)printf("Time to live exceeded\n");
break;
case ICMP6_TIME_EXCEED_REASSEMBLY:
(void)printf("Frag reassembly time exceeded\n");
break;
default:
(void)printf("Time exceeded, Bad Code: %d\n",
icp->icmp6_code);
break;
}
pr_retip((struct ip6_hdr *)(icp + 1), end);
break;
case ICMP6_PARAM_PROB:
(void)printf("Parameter problem: ");
switch (icp->icmp6_code) {
case ICMP6_PARAMPROB_HEADER:
(void)printf("Erroneous Header ");
break;
case ICMP6_PARAMPROB_NEXTHEADER:
(void)printf("Unknown Nextheader ");
break;
case ICMP6_PARAMPROB_OPTION:
(void)printf("Unrecognized Option ");
break;
default:
(void)printf("Bad code(%d) ", icp->icmp6_code);
break;
}
(void)printf("pointer = 0x%02x\n",
(u_int32_t)ntohl(icp->icmp6_pptr));
pr_retip((struct ip6_hdr *)(icp + 1), end);
break;
case ICMP6_ECHO_REQUEST:
(void)printf("Echo Request");
/* XXX ID + Seq + Data */
break;
case ICMP6_ECHO_REPLY:
(void)printf("Echo Reply");
/* XXX ID + Seq + Data */
break;
case ICMP6_MEMBERSHIP_QUERY:
(void)printf("Listener Query");
break;
case ICMP6_MEMBERSHIP_REPORT:
(void)printf("Listener Report");
break;
case ICMP6_MEMBERSHIP_REDUCTION:
(void)printf("Listener Done");
break;
case ND_ROUTER_SOLICIT:
(void)printf("Router Solicitation");
break;
case ND_ROUTER_ADVERT:
(void)printf("Router Advertisement");
break;
case ND_NEIGHBOR_SOLICIT:
(void)printf("Neighbor Solicitation");
break;
case ND_NEIGHBOR_ADVERT:
(void)printf("Neighbor Advertisement");
break;
case ND_REDIRECT:
red = (struct nd_redirect *)icp;
(void)printf("Redirect\n");
if (!inet_ntop(AF_INET6, &red->nd_rd_dst, ntop_buf,
sizeof(ntop_buf)))
strncpy(ntop_buf, "?", sizeof(ntop_buf));
(void)printf("Destination: %s", ntop_buf);
if (!inet_ntop(AF_INET6, &red->nd_rd_target, ntop_buf,
sizeof(ntop_buf)))
strncpy(ntop_buf, "?", sizeof(ntop_buf));
(void)printf(" New Target: %s", ntop_buf);
break;
case ICMP6_NI_QUERY:
(void)printf("Node Information Query");
/* XXX ID + Seq + Data */
ni = (struct icmp6_nodeinfo *)icp;
l = end - (u_char *)(ni + 1);
printf(", ");
switch (ntohs(ni->ni_qtype)) {
case NI_QTYPE_NOOP:
(void)printf("NOOP");
break;
case NI_QTYPE_SUPTYPES:
(void)printf("Supported qtypes");
break;
case NI_QTYPE_FQDN:
(void)printf("DNS name");
break;
case NI_QTYPE_NODEADDR:
(void)printf("nodeaddr");
break;
case NI_QTYPE_IPV4ADDR:
(void)printf("IPv4 nodeaddr");
break;
default:
(void)printf("unknown qtype");
break;
}
if (options & F_VERBOSE) {
switch (ni->ni_code) {
case ICMP6_NI_SUBJ_IPV6:
if (l == sizeof(struct in6_addr) &&
inet_ntop(AF_INET6, ni + 1, ntop_buf,
sizeof(ntop_buf)) != NULL) {
(void)printf(", subject=%s(%s)",
niqcode[ni->ni_code], ntop_buf);
} else {
#if 1
/* backward compat to -W */
(void)printf(", oldfqdn");
#else
(void)printf(", invalid");
#endif
}
break;
case ICMP6_NI_SUBJ_FQDN:
if (end == (u_char *)(ni + 1)) {
(void)printf(", no subject");
break;
}
printf(", subject=%s", niqcode[ni->ni_code]);
cp = (const u_char *)(ni + 1);
if (dnsdecode(&cp, end, NULL, dnsname,
sizeof(dnsname)) != NULL)
printf("(%s)", dnsname);
else
printf("(invalid)");
break;
case ICMP6_NI_SUBJ_IPV4:
if (l == sizeof(struct in_addr) &&
inet_ntop(AF_INET, ni + 1, ntop_buf,
sizeof(ntop_buf)) != NULL) {
(void)printf(", subject=%s(%s)",
niqcode[ni->ni_code], ntop_buf);
} else
(void)printf(", invalid");
break;
default:
(void)printf(", invalid");
break;
}
}
break;
case ICMP6_NI_REPLY:
(void)printf("Node Information Reply");
/* XXX ID + Seq + Data */
ni = (struct icmp6_nodeinfo *)icp;
printf(", ");
switch (ntohs(ni->ni_qtype)) {
case NI_QTYPE_NOOP:
(void)printf("NOOP");
break;
case NI_QTYPE_SUPTYPES:
(void)printf("Supported qtypes");
break;
case NI_QTYPE_FQDN:
(void)printf("DNS name");
break;
case NI_QTYPE_NODEADDR:
(void)printf("nodeaddr");
break;
case NI_QTYPE_IPV4ADDR:
(void)printf("IPv4 nodeaddr");
break;
default:
(void)printf("unknown qtype");
break;
}
if (options & F_VERBOSE) {
if (ni->ni_code > sizeof(nircode) / sizeof(nircode[0]))
printf(", invalid");
else
printf(", %s", nircode[ni->ni_code]);
}
break;
default:
(void)printf("Bad ICMP type: %d", icp->icmp6_type);
}
}
/*
* pr_iph --
* Print an IP6 header.
*/
void
pr_iph(ip6)
struct ip6_hdr *ip6;
{
u_int32_t flow = ip6->ip6_flow & IPV6_FLOWLABEL_MASK;
u_int8_t tc;
char ntop_buf[INET6_ADDRSTRLEN];
tc = *(&ip6->ip6_vfc + 1); /* XXX */
tc = (tc >> 4) & 0x0f;
tc |= (ip6->ip6_vfc << 4);
printf("Vr TC Flow Plen Nxt Hlim\n");
printf(" %1x %02x %05x %04x %02x %02x\n",
(ip6->ip6_vfc & IPV6_VERSION_MASK) >> 4, tc, (u_int32_t)ntohl(flow),
ntohs(ip6->ip6_plen), ip6->ip6_nxt, ip6->ip6_hlim);
if (!inet_ntop(AF_INET6, &ip6->ip6_src, ntop_buf, sizeof(ntop_buf)))
strncpy(ntop_buf, "?", sizeof(ntop_buf));
printf("%s->", ntop_buf);
if (!inet_ntop(AF_INET6, &ip6->ip6_dst, ntop_buf, sizeof(ntop_buf)))
strncpy(ntop_buf, "?", sizeof(ntop_buf));
printf("%s\n", ntop_buf);
}
/*
* pr_addr --
* Return an ascii host address as a dotted quad and optionally with
* a hostname.
*/
const char *
pr_addr(addr, addrlen)
struct sockaddr *addr;
int addrlen;
{
static char buf[NI_MAXHOST];
int flag;
#ifdef NI_WITHSCOPEID
flag = NI_WITHSCOPEID;
#else
flag = 0;
#endif
if ((options & F_HOSTNAME) == 0)
flag |= NI_NUMERICHOST;
if (getnameinfo(addr, addrlen, buf, sizeof(buf), NULL, 0, flag) == 0)
return (buf);
else
return "?";
}
/*
* pr_retip --
* Dump some info on a returned (via ICMPv6) IPv6 packet.
*/
void
pr_retip(ip6, end)
struct ip6_hdr *ip6;
u_char *end;
{
u_char *cp = (u_char *)ip6, nh;
int hlen;
if (end - (u_char *)ip6 < sizeof(*ip6)) {
printf("IP6");
goto trunc;
}
pr_iph(ip6);
hlen = sizeof(*ip6);
nh = ip6->ip6_nxt;
cp += hlen;
while (end - cp >= 8) {
switch (nh) {
case IPPROTO_HOPOPTS:
printf("HBH ");
hlen = (((struct ip6_hbh *)cp)->ip6h_len+1) << 3;
nh = ((struct ip6_hbh *)cp)->ip6h_nxt;
break;
case IPPROTO_DSTOPTS:
printf("DSTOPT ");
hlen = (((struct ip6_dest *)cp)->ip6d_len+1) << 3;
nh = ((struct ip6_dest *)cp)->ip6d_nxt;
break;
case IPPROTO_FRAGMENT:
printf("FRAG ");
hlen = sizeof(struct ip6_frag);
nh = ((struct ip6_frag *)cp)->ip6f_nxt;
break;
case IPPROTO_ROUTING:
printf("RTHDR ");
hlen = (((struct ip6_rthdr *)cp)->ip6r_len+1) << 3;
nh = ((struct ip6_rthdr *)cp)->ip6r_nxt;
break;
#ifdef IPSEC
case IPPROTO_AH:
printf("AH ");
hlen = (((struct ah *)cp)->ah_len+2) << 2;
nh = ((struct ah *)cp)->ah_nxt;
break;
#endif
case IPPROTO_ICMPV6:
printf("ICMP6: type = %d, code = %d\n",
*cp, *(cp + 1));
return;
case IPPROTO_ESP:
printf("ESP\n");
return;
case IPPROTO_TCP:
printf("TCP: from port %u, to port %u (decimal)\n",
(*cp * 256 + *(cp + 1)),
(*(cp + 2) * 256 + *(cp + 3)));
return;
case IPPROTO_UDP:
printf("UDP: from port %u, to port %u (decimal)\n",
(*cp * 256 + *(cp + 1)),
(*(cp + 2) * 256 + *(cp + 3)));
return;
default:
printf("Unknown Header(%d)\n", nh);
return;
}
if ((cp += hlen) >= end)
goto trunc;
}
if (end - cp < 8)
goto trunc;
putchar('\n');
return;
trunc:
printf("...\n");
return;
}
void
fill(bp, patp)
char *bp, *patp;
{
register int ii, jj, kk;
int pat[16];
char *cp;
for (cp = patp; *cp; cp++)
if (!isxdigit(*cp))
errx(1, "patterns must be specified as hex digits");
ii = sscanf(patp,
"%2x%2x%2x%2x%2x%2x%2x%2x%2x%2x%2x%2x%2x%2x%2x%2x",
&pat[0], &pat[1], &pat[2], &pat[3], &pat[4], &pat[5], &pat[6],
&pat[7], &pat[8], &pat[9], &pat[10], &pat[11], &pat[12],
&pat[13], &pat[14], &pat[15]);
/* xxx */
if (ii > 0)
for (kk = 0;
kk <= MAXDATALEN - (8 + sizeof(struct timeval) + ii);
kk += ii)
for (jj = 0; jj < ii; ++jj)
bp[jj + kk] = pat[jj];
if (!(options & F_QUIET)) {
(void)printf("PATTERN: 0x");
for (jj = 0; jj < ii; ++jj)
(void)printf("%02x", bp[jj] & 0xFF);
(void)printf("\n");
}
}
#ifdef IPSEC
#ifdef IPSEC_POLICY_IPSEC
int
setpolicy(so, policy)
int so;
char *policy;
{
char *buf;
if (policy == NULL)
return 0; /* ignore */
buf = ipsec_set_policy(policy, strlen(policy));
if (buf == NULL)
errx(1, "%s", ipsec_strerror());
if (setsockopt(s, IPPROTO_IPV6, IPV6_IPSEC_POLICY, buf,
ipsec_get_policylen(buf)) < 0)
warnx("unable to set IPSec policy");
free(buf);
return 0;
}
#endif
#endif
char *
nigroup(name)
char *name;
{
char *p;
unsigned char *q;
MD5_CTX ctxt;
u_int8_t digest[16];
u_int8_t c;
size_t l;
char hbuf[NI_MAXHOST];
struct in6_addr in6;
p = strchr(name, '.');
if (!p)
p = name + strlen(name);
l = p - name;
if (l > 63 || l > sizeof(hbuf) - 1)
return NULL; /*label too long*/
strncpy(hbuf, name, l);
hbuf[(int)l] = '\0';
for (q = name; *q; q++) {
if (isupper(*q))
*q = tolower(*q);
}
/* generate 8 bytes of pseudo-random value. */
bzero(&ctxt, sizeof(ctxt));
MD5Init(&ctxt);
c = l & 0xff;
MD5Update(&ctxt, &c, sizeof(c));
MD5Update(&ctxt, name, l);
MD5Final(digest, &ctxt);
if (inet_pton(AF_INET6, "ff02::2:0000:0000", &in6) != 1)
return NULL; /*XXX*/
bcopy(digest, &in6.s6_addr[12], 4);
if (inet_ntop(AF_INET6, &in6, hbuf, sizeof(hbuf)) == NULL)
return NULL;
return strdup(hbuf);
}
void
usage()
{
(void)fprintf(stderr,
"usage: ping6 [-dfH"
#ifdef IPV6_USE_MIN_MTU
"m"
#endif
"nNqtvwW"
#ifdef IPV6_REACHCONF
"R"
#endif
#ifdef IPSEC
#ifdef IPSEC_POLICY_IPSEC
"] [-P policy"
#else
"AE"
#endif
#endif
"] [-a [aAclsg]] [-b sockbufsiz] [-c count]\n"
"\t[-I interface] [-i wait] [-l preload] [-p pattern] "
"[-S sourceaddr]\n"
"\t[-s packetsize] [-h hoplimit] [hops...] host\n");
exit(1);
}