freebsd-dev/usr.sbin/mrouted/mrinfo.c
Bill Fenner 7dae516d85 Release setuid immediately after initializing; this minimizes the
possibility of security holes allowing root penetration.

Inspired by:	Mark Handley <M.Handley@cs.ucl.ac.uk> and
		Theo de Raadt <deraadt@theos.com> independently
Submitted by:	Theo de Raadt <deraadt@theos.com>
1996-10-01 23:14:35 +00:00

636 lines
15 KiB
C

/*
* This tool requests configuration info from a multicast router
* and prints the reply (if any). Invoke it as:
*
* mrinfo router-name-or-address
*
* Written Wed Mar 24 1993 by Van Jacobson (adapted from the
* multicast mapper written by Pavel Curtis).
*
* The lawyers insist we include the following UC copyright notice.
* The mapper from which this is derived contained a Xerox copyright
* notice which follows the UC one. Try not to get depressed noting
* that the legal gibberish is larger than the program.
*
* Copyright (c) 1993 Regents of the University of California.
* 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. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the Computer Systems
* Engineering Group at Lawrence Berkeley Laboratory.
* 4. Neither the name of the University nor of the Laboratory 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.
* ---------------------------------
* Copyright (c) Xerox Corporation 1992. All rights reserved.
*
* License is granted to copy, to use, and to make and to use derivative works
* for research and evaluation purposes, provided that Xerox is acknowledged
* in all documentation pertaining to any such copy or derivative work. Xerox
* grants no other licenses expressed or implied. The Xerox trade name should
* not be used in any advertising without its written permission.
*
* XEROX CORPORATION MAKES NO REPRESENTATIONS CONCERNING EITHER THE
* MERCHANTABILITY OF THIS SOFTWARE OR THE SUITABILITY OF THIS SOFTWARE FOR
* ANY PARTICULAR PURPOSE. The software is provided "as is" without express
* or implied warranty of any kind.
*
* These notices must be retained in any copies of any part of this software.
*/
#ifndef lint
static char rcsid[] =
"@(#) $Id: mrinfo.c,v 1.8 1996/01/06 21:09:56 peter Exp $";
/* original rcsid:
"@(#) Header: mrinfo.c,v 1.6 93/04/08 15:14:16 van Exp (LBL)";
*/
#endif
#include <netdb.h>
#include <sys/time.h>
#include "defs.h"
#include <arpa/inet.h>
#ifdef __STDC__
#include <stdarg.h>
#else
#include <varargs.h>
#endif
#define DEFAULT_TIMEOUT 4 /* How long to wait before retrying requests */
#define DEFAULT_RETRIES 3 /* How many times to ask each router */
u_int32 our_addr, target_addr = 0; /* in NET order */
int debug = 0;
int nflag = 0;
int retries = DEFAULT_RETRIES;
int timeout = DEFAULT_TIMEOUT;
int target_level = 0;
vifi_t numvifs; /* to keep loader happy */
/* (see COPY_TABLES macro called in kern.c) */
char * inet_name __P((u_int32 addr));
void ask __P((u_int32 dst));
void ask2 __P((u_int32 dst));
int get_number __P((int *var, int deflt, char ***pargv,
int *pargc));
u_int32 host_addr __P((char *name));
void usage __P((void));
/* to shut up -Wstrict-prototypes */
int main __P((int argc, char *argv[]));
char *
inet_name(addr)
u_int32 addr;
{
struct hostent *e;
struct in_addr in;
if (addr == 0)
return "local";
if (nflag ||
(e = gethostbyaddr((char *)&addr, sizeof(addr), AF_INET)) == NULL) {
in.s_addr = addr;
return (inet_ntoa(in));
}
return (e->h_name);
}
/*
* Log errors and other messages to stderr, according to the severity of the
* message and the current debug level. For errors of severity LOG_ERR or
* worse, terminate the program.
*/
#ifdef __STDC__
void
log(int severity, int syserr, char *format, ...)
{
va_list ap;
char fmt[100];
va_start(ap, format);
#else
void
log(severity, syserr, format, va_alist)
int severity, syserr;
char *format;
va_dcl
{
va_list ap;
char fmt[100];
va_start(ap);
#endif
switch (debug) {
case 0:
if (severity > LOG_WARNING)
return;
case 1:
if (severity > LOG_NOTICE)
return;
case 2:
if (severity > LOG_INFO)
return;
default:
fmt[0] = '\0';
if (severity == LOG_WARNING)
strcat(fmt, "warning - ");
strncat(fmt, format, 80);
vfprintf(stderr, fmt, ap);
if (syserr == 0)
fprintf(stderr, "\n");
else if (syserr < sys_nerr)
fprintf(stderr, ": %s\n", sys_errlist[syserr]);
else
fprintf(stderr, ": errno %d\n", syserr);
}
if (severity <= LOG_ERR)
exit(-1);
}
/*
* Send a neighbors-list request.
*/
void
ask(dst)
u_int32 dst;
{
send_igmp(our_addr, dst, IGMP_DVMRP, DVMRP_ASK_NEIGHBORS,
htonl(MROUTED_LEVEL), 0);
}
void
ask2(dst)
u_int32 dst;
{
send_igmp(our_addr, dst, IGMP_DVMRP, DVMRP_ASK_NEIGHBORS2,
htonl(MROUTED_LEVEL), 0);
}
/*
* Process an incoming neighbor-list message.
*/
void
accept_neighbors(src, dst, p, datalen, level)
u_int32 src, dst, level;
u_char *p;
int datalen;
{
u_char *ep = p + datalen;
#define GET_ADDR(a) (a = ((u_int32)*p++ << 24), a += ((u_int32)*p++ << 16),\
a += ((u_int32)*p++ << 8), a += *p++)
printf("%s (%s):\n", inet_fmt(src, s1), inet_name(src));
while (p < ep) {
register u_int32 laddr;
register u_char metric;
register u_char thresh;
register int ncount;
GET_ADDR(laddr);
laddr = htonl(laddr);
metric = *p++;
thresh = *p++;
ncount = *p++;
while (--ncount >= 0) {
register u_int32 neighbor;
GET_ADDR(neighbor);
neighbor = htonl(neighbor);
printf(" %s -> ", inet_fmt(laddr, s1));
printf("%s (%s) [%d/%d]\n", inet_fmt(neighbor, s1),
inet_name(neighbor), metric, thresh);
}
}
}
void
accept_neighbors2(src, dst, p, datalen, level)
u_int32 src, dst, level;
u_char *p;
int datalen;
{
u_char *ep = p + datalen;
u_int broken_cisco = ((level & 0xffff) == 0x020a); /* 10.2 */
/* well, only possibly_broken_cisco, but that's too long to type. */
printf("%s (%s) [version %d.%d", inet_fmt(src, s1), inet_name(src),
level & 0xff, (level >> 8) & 0xff);
if ((level >> 16) & NF_LEAF) { printf (",leaf"); }
if ((level >> 16) & NF_PRUNE) { printf (",prune"); }
if ((level >> 16) & NF_GENID) { printf (",genid"); }
if ((level >> 16) & NF_MTRACE) { printf (",mtrace"); }
printf ("]:\n");
while (p < ep) {
register u_char metric;
register u_char thresh;
register u_char flags;
register int ncount;
register u_int32 laddr = *(u_int32*)p;
p += 4;
metric = *p++;
thresh = *p++;
flags = *p++;
ncount = *p++;
if (broken_cisco && ncount == 0) /* dumb Ciscos */
ncount = 1;
if (broken_cisco && ncount > 15) /* dumb Ciscos */
ncount = ncount & 0xf;
while (--ncount >= 0 && p < ep) {
register u_int32 neighbor = *(u_int32*)p;
p += 4;
printf(" %s -> ", inet_fmt(laddr, s1));
printf("%s (%s) [%d/%d", inet_fmt(neighbor, s1),
inet_name(neighbor), metric, thresh);
if (flags & DVMRP_NF_TUNNEL)
printf("/tunnel");
if (flags & DVMRP_NF_SRCRT)
printf("/srcrt");
if (flags & DVMRP_NF_PIM)
printf("/pim");
if (flags & DVMRP_NF_QUERIER)
printf("/querier");
if (flags & DVMRP_NF_DISABLED)
printf("/disabled");
if (flags & DVMRP_NF_DOWN)
printf("/down");
if (flags & DVMRP_NF_LEAF)
printf("/leaf");
printf("]\n");
}
}
}
int
get_number(var, deflt, pargv, pargc)
int *var, *pargc, deflt;
char ***pargv;
{
if ((*pargv)[0][2] == '\0') { /* Get the value from the next
* argument */
if (*pargc > 1 && isdigit((*pargv)[1][0])) {
(*pargv)++, (*pargc)--;
*var = atoi((*pargv)[0]);
return 1;
} else if (deflt >= 0) {
*var = deflt;
return 1;
} else
return 0;
} else { /* Get value from the rest of this argument */
if (isdigit((*pargv)[0][2])) {
*var = atoi((*pargv)[0] + 2);
return 1;
} else {
return 0;
}
}
}
void
usage()
{
fprintf(stderr,
"Usage: mrinfo [-n] [-t timeout] [-r retries] [router]\n");
exit(1);
}
int
main(argc, argv)
int argc;
char *argv[];
{
int tries;
int trynew;
struct timeval et;
struct hostent *hp;
struct hostent bogus;
char *host;
int curaddr;
if (geteuid() != 0) {
fprintf(stderr, "mrinfo: must be root\n");
exit(1);
}
init_igmp();
setuid(getuid());
setlinebuf(stderr);
argv++, argc--;
while (argc > 0 && argv[0][0] == '-') {
switch (argv[0][1]) {
case 'd':
if (!get_number(&debug, DEFAULT_DEBUG, &argv, &argc))
usage();
break;
case 'n':
++nflag;
break;
case 'r':
if (!get_number(&retries, -1, &argv, &argc))
usage();
break;
case 't':
if (!get_number(&timeout, -1, &argv, &argc))
usage();
break;
default:
usage();
}
argv++, argc--;
}
if (argc > 1)
usage();
if (argc == 1)
host = argv[0];
else
host = "127.0.0.1";
if ((target_addr = inet_addr(host)) != -1) {
hp = &bogus;
hp->h_length = sizeof(target_addr);
hp->h_addr_list = (char **)malloc(2 * sizeof(char *));
hp->h_addr_list[0] = malloc(hp->h_length);
memcpy(hp->h_addr_list[0], &target_addr, hp->h_length);
hp->h_addr_list[1] = 0;
} else
hp = gethostbyname(host);
if (hp == NULL) {
fprintf(stderr, "mrinfo: %s: no such host\n", argv[0]);
exit(1);
}
if (debug)
fprintf(stderr, "Debug level %u\n", debug);
/* Check all addresses; mrouters often have unreachable interfaces */
for (curaddr = 0; hp->h_addr_list[curaddr] != NULL; curaddr++) {
memcpy(&target_addr, hp->h_addr_list[curaddr], hp->h_length);
{ /* Find a good local address for us. */
int udp;
struct sockaddr_in addr;
int addrlen = sizeof(addr);
addr.sin_family = AF_INET;
#if (defined(BSD) && (BSD >= 199103))
addr.sin_len = sizeof addr;
#endif
addr.sin_addr.s_addr = target_addr;
addr.sin_port = htons(2000); /* any port over 1024 will
* do... */
if ((udp = socket(AF_INET, SOCK_DGRAM, 0)) < 0
|| connect(udp, (struct sockaddr *) & addr, sizeof(addr)) < 0
|| getsockname(udp, (struct sockaddr *) & addr, &addrlen) < 0) {
perror("Determining local address");
exit(-1);
}
close(udp);
our_addr = addr.sin_addr.s_addr;
}
tries = 0;
trynew = 1;
/*
* New strategy: send 'ask2' for two timeouts, then fall back
* to 'ask', since it's not very likely that we are going to
* find someone who only responds to 'ask' these days
*/
ask2(target_addr);
gettimeofday(&et, 0);
et.tv_sec += timeout;
/* Main receive loop */
for (;;) {
fd_set fds;
struct timeval tv, now;
int count, recvlen, dummy = 0;
register u_int32 src, dst, group;
struct ip *ip;
struct igmp *igmp;
int ipdatalen, iphdrlen, igmpdatalen;
FD_ZERO(&fds);
FD_SET(igmp_socket, &fds);
gettimeofday(&now, 0);
tv.tv_sec = et.tv_sec - now.tv_sec;
tv.tv_usec = et.tv_usec - now.tv_usec;
if (tv.tv_usec < 0) {
tv.tv_usec += 1000000L;
--tv.tv_sec;
}
if (tv.tv_sec < 0)
tv.tv_sec = tv.tv_usec = 0;
count = select(igmp_socket + 1, &fds, 0, 0, &tv);
if (count < 0) {
if (errno != EINTR)
perror("select");
continue;
} else if (count == 0) {
log(LOG_DEBUG, 0, "Timed out receiving neighbor lists");
if (++tries > retries)
break;
/* If we've tried ASK_NEIGHBORS2 twice with
* no response, fall back to ASK_NEIGHBORS
*/
if (tries == 2 && target_level == 0)
trynew = 0;
if (target_level == 0 && trynew == 0)
ask(target_addr);
else
ask2(target_addr);
gettimeofday(&et, 0);
et.tv_sec += timeout;
continue;
}
recvlen = recvfrom(igmp_socket, recv_buf, RECV_BUF_SIZE,
0, NULL, &dummy);
if (recvlen <= 0) {
if (recvlen && errno != EINTR)
perror("recvfrom");
continue;
}
if (recvlen < sizeof(struct ip)) {
log(LOG_WARNING, 0,
"packet too short (%u bytes) for IP header",
recvlen);
continue;
}
ip = (struct ip *) recv_buf;
if (ip->ip_p == 0)
continue; /* Request to install cache entry */
src = ip->ip_src.s_addr;
dst = ip->ip_dst.s_addr;
iphdrlen = ip->ip_hl << 2;
ipdatalen = ip->ip_len;
if (iphdrlen + ipdatalen != recvlen) {
log(LOG_WARNING, 0,
"packet shorter (%u bytes) than hdr+data length (%u+%u)",
recvlen, iphdrlen, ipdatalen);
continue;
}
igmp = (struct igmp *) (recv_buf + iphdrlen);
group = igmp->igmp_group.s_addr;
igmpdatalen = ipdatalen - IGMP_MINLEN;
if (igmpdatalen < 0) {
log(LOG_WARNING, 0,
"IP data field too short (%u bytes) for IGMP, from %s",
ipdatalen, inet_fmt(src, s1));
continue;
}
if (igmp->igmp_type != IGMP_DVMRP)
continue;
switch (igmp->igmp_code) {
case DVMRP_NEIGHBORS:
case DVMRP_NEIGHBORS2:
if (src != target_addr) {
fprintf(stderr, "mrinfo: got reply from %s",
inet_fmt(src, s1));
fprintf(stderr, " instead of %s\n",
inet_fmt(target_addr, s1));
/*continue;*/
}
break;
default:
continue; /* ignore all other DVMRP messages */
}
switch (igmp->igmp_code) {
case DVMRP_NEIGHBORS:
if (group) {
/* knows about DVMRP_NEIGHBORS2 msg */
if (target_level == 0) {
target_level = ntohl(group);
ask2(target_addr);
}
} else {
accept_neighbors(src, dst, (u_char *)(igmp + 1),
igmpdatalen, ntohl(group));
exit(0);
}
break;
case DVMRP_NEIGHBORS2:
accept_neighbors2(src, dst, (u_char *)(igmp + 1),
igmpdatalen, ntohl(group));
exit(0);
}
}
}
exit(1);
}
/* dummies */
void accept_probe(src, dst, p, datalen, level)
u_int32 src, dst, level;
char *p;
int datalen;
{
}
void accept_group_report(src, dst, group, r_type)
u_int32 src, dst, group;
int r_type;
{
}
void accept_neighbor_request2(src, dst)
u_int32 src, dst;
{
}
void accept_report(src, dst, p, datalen, level)
u_int32 src, dst, level;
char *p;
int datalen;
{
}
void accept_neighbor_request(src, dst)
u_int32 src, dst;
{
}
void accept_prune(src, dst, p, datalen)
u_int32 src, dst;
char *p;
int datalen;
{
}
void accept_graft(src, dst, p, datalen)
u_int32 src, dst;
char *p;
int datalen;
{
}
void accept_g_ack(src, dst, p, datalen)
u_int32 src, dst;
char *p;
int datalen;
{
}
void add_table_entry(origin, mcastgrp)
u_int32 origin, mcastgrp;
{
}
void check_vif_state()
{
}
void accept_leave_message(src, dst, group)
u_int32 src, dst, group;
{
}
void accept_mtrace(src, dst, group, data, no, datalen)
u_int32 src, dst, group;
char *data;
u_int no;
int datalen;
{
}
void accept_membership_query(src, dst, group, tmo)
u_int32 src, dst, group;
int tmo;
{
}
void accept_info_request(src, dst, p, datalen)
u_int32 src, dst;
u_char *p;
int datalen;
{
}
void accept_info_reply(src, dst, p, datalen)
u_int32 src, dst;
u_char *p;
int datalen;
{
}