summaryrefslogtreecommitdiffstats
path: root/usr.sbin
diff options
context:
space:
mode:
Diffstat (limited to 'usr.sbin')
-rw-r--r--usr.sbin/rarpd/Makefile5
-rw-r--r--usr.sbin/rarpd/arptab.c219
-rw-r--r--usr.sbin/rarpd/etherent.c224
-rw-r--r--usr.sbin/rarpd/rarpd.c809
4 files changed, 534 insertions, 723 deletions
diff --git a/usr.sbin/rarpd/Makefile b/usr.sbin/rarpd/Makefile
index c8fd9ce..09605f6 100644
--- a/usr.sbin/rarpd/Makefile
+++ b/usr.sbin/rarpd/Makefile
@@ -1,9 +1,10 @@
# from: @(#)Makefile 5.8 (Berkeley) 7/28/90
-# $Id: Makefile,v 1.2 1995/03/03 22:20:12 wpaul Exp $
+# $Id: Makefile,v 1.3 1995/04/02 01:35:53 wpaul Exp $
PROG= rarpd
MAN8= rarpd.8
-SRCS= rarpd.c arptab.c
+SRCS= rarpd.c
+
CFLAGS+= -DTFTP_DIR=\"/tftpboot\"
.include <bsd.prog.mk>
diff --git a/usr.sbin/rarpd/arptab.c b/usr.sbin/rarpd/arptab.c
deleted file mode 100644
index 7757814..0000000
--- a/usr.sbin/rarpd/arptab.c
+++ /dev/null
@@ -1,219 +0,0 @@
-/*
- * Copyright (c) 1984, 1993
- * The Regents of the University of California. All rights reserved.
- *
- * This code is derived from software contributed to Berkeley by
- * Sun Microsystems, Inc.
- *
- * 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) 1984, 1993\n\
- The Regents of the University of California. All rights reserved.\n";
-#endif /* not lint */
-
-#ifndef lint
-static char sccsid[] = "@(#)arp.c 8.2 (Berkeley) 1/2/94";
-#endif /* not lint */
-
-/*
- * set arp table entries
- */
-
-
-#include <sys/param.h>
-#include <sys/file.h>
-#include <sys/socket.h>
-#include <sys/sysctl.h>
-
-#include <net/if.h>
-#include <net/if_dl.h>
-#include <net/if_types.h>
-#include <net/route.h>
-
-#include <netinet/in.h>
-#include <netinet/if_ether.h>
-
-#include <arpa/inet.h>
-
-#include <netdb.h>
-#include <errno.h>
-#include <nlist.h>
-#include <stdio.h>
-#include <paths.h>
-#include <syslog.h>
-
-extern int errno;
-static int pid;
-static int s = -1;
-
-getsocket() {
- if (s < 0) {
- s = socket(PF_ROUTE, SOCK_RAW, 0);
- if (s < 0) {
- perror("arp: socket");
- exit(1);
- }
- }
-}
-
-struct sockaddr_in so_mask = {8, 0, 0, { 0xffffffff}};
-struct sockaddr_inarp blank_sin = {sizeof(blank_sin), AF_INET }, sin_m;
-struct sockaddr_dl blank_sdl = {sizeof(blank_sdl), AF_LINK }, sdl_m;
-int expire_time, flags, export_only, doing_proxy;
-struct {
- struct rt_msghdr m_rtm;
- char m_space[512];
-} m_rtmsg;
-
-/*
- * Set an individual arp entry
- */
-arptab_set(eaddr, host)
- u_char *eaddr;
- u_long host;
-{
- register struct sockaddr_inarp *sin = &sin_m;
- register struct sockaddr_dl *sdl;
- register struct rt_msghdr *rtm = &(m_rtmsg.m_rtm);
- struct timeval time;
-
- getsocket();
- pid = getpid();
- sdl_m = blank_sdl;
- sin_m = blank_sin;
- sin->sin_addr.s_addr = host;
- bcopy((char *)eaddr, (u_char *)LLADDR(&sdl_m), 6);
- sdl_m.sdl_alen = 6;
- doing_proxy = flags = export_only = expire_time = 0;
- gettimeofday(&time, 0);
- expire_time = time.tv_sec + 20 * 60;
-
-tryagain:
- if (rtmsg(RTM_GET) < 0) {
- syslog(LOG_ERR,"%s: %m", inet_ntoa(sin->sin_addr));
- return (1);
- }
- sin = (struct sockaddr_inarp *)(rtm + 1);
- sdl = (struct sockaddr_dl *)(sin->sin_len + (char *)sin);
- if (sin->sin_addr.s_addr == sin_m.sin_addr.s_addr) {
- if (sdl->sdl_family == AF_LINK &&
- (rtm->rtm_flags & RTF_LLINFO) &&
- !(rtm->rtm_flags & RTF_GATEWAY)) switch (sdl->sdl_type) {
- case IFT_ETHER: case IFT_FDDI: case IFT_ISO88023:
- case IFT_ISO88024: case IFT_ISO88025:
- goto overwrite;
- }
- if (doing_proxy == 0) {
- syslog(LOG_ERR, "arptab_set: can only proxy for %s\n", inet_ntoa(sin->sin_addr));
- return (1);
- }
- if (sin_m.sin_other & SIN_PROXY) {
- syslog(LOG_ERR,"arptab_set: proxy entry exists for non 802 device\n");
- return(1);
- }
- sin_m.sin_other = SIN_PROXY;
- export_only = 1;
- goto tryagain;
- }
-overwrite:
- if (sdl->sdl_family != AF_LINK) {
- syslog(LOG_ERR,"arptab_set: cannot intuit interface index and type for %s\n", inet_ntoa(sin->sin_addr));
- return (1);
- }
- sdl_m.sdl_type = sdl->sdl_type;
- sdl_m.sdl_index = sdl->sdl_index;
- return (rtmsg(RTM_ADD));
-}
-
-rtmsg(cmd)
-{
- static int seq;
- int rlen;
- register struct rt_msghdr *rtm = &m_rtmsg.m_rtm;
- register char *cp = m_rtmsg.m_space;
- register int l;
-
- errno = 0;
- if (cmd == RTM_DELETE)
- goto doit;
- bzero((char *)&m_rtmsg, sizeof(m_rtmsg));
- rtm->rtm_flags = flags;
- rtm->rtm_version = RTM_VERSION;
-
- switch (cmd) {
- default:
- syslog(LOG_ERR, "arptap_set: internal wrong cmd\n");
- exit(1);
- case RTM_ADD:
- rtm->rtm_addrs |= RTA_GATEWAY;
- rtm->rtm_rmx.rmx_expire = expire_time;
- rtm->rtm_inits = RTV_EXPIRE;
- rtm->rtm_flags |= (RTF_HOST | RTF_STATIC);
- sin_m.sin_other = 0;
- if (doing_proxy) {
- if (export_only)
- sin_m.sin_other = SIN_PROXY;
- else {
- rtm->rtm_addrs |= RTA_NETMASK;
- rtm->rtm_flags &= ~RTF_HOST;
- }
- }
- /* FALLTHROUGH */
- case RTM_GET:
- rtm->rtm_addrs |= RTA_DST;
- }
-#define NEXTADDR(w, s) \
- if (rtm->rtm_addrs & (w)) { \
- bcopy((char *)&s, cp, sizeof(s)); cp += sizeof(s);}
-
- NEXTADDR(RTA_DST, sin_m);
- NEXTADDR(RTA_GATEWAY, sdl_m);
- NEXTADDR(RTA_NETMASK, so_mask);
-
- rtm->rtm_msglen = cp - (char *)&m_rtmsg;
-doit:
- l = rtm->rtm_msglen;
- rtm->rtm_seq = ++seq;
- rtm->rtm_type = cmd;
- if ((rlen = write(s, (char *)&m_rtmsg, l)) < 0) {
- if (errno != ESRCH && errno != EEXIST) {
- syslog(LOG_ERR, "writing to routing socket: %m");
- return (-1);
- }
- }
- do {
- l = read(s, (char *)&m_rtmsg, sizeof(m_rtmsg));
- } while (l > 0 && (rtm->rtm_seq != seq || rtm->rtm_pid != pid));
- if (l < 0)
- syslog(LOG_ERR, "arptab_set: read from routing socket: %m");
- return (0);
-}
diff --git a/usr.sbin/rarpd/etherent.c b/usr.sbin/rarpd/etherent.c
deleted file mode 100644
index d682b63..0000000
--- a/usr.sbin/rarpd/etherent.c
+++ /dev/null
@@ -1,224 +0,0 @@
-/*
- * Copyright (c) 1990, 1993 The 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: (1) source code distributions
- * retain the above copyright notice and this paragraph in its entirety, (2)
- * distributions including binary code include the above copyright notice and
- * this paragraph in its entirety in the documentation or other materials
- * provided with the distribution, and (3) all advertising materials mentioning
- * features or use of this software display the following acknowledgement:
- * ``This product includes software developed by the University of California,
- * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of
- * the University nor the names of its contributors may be used to endorse
- * or promote products derived from this software without specific prior
- * written permission.
- * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED
- * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
- * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
- */
-#ifndef lint
-static char rcsid[] =
- "@(#) $Header: etherent.c,v 1.4 96/06/14 20:34:25 leres Exp $ (LBL)";
-#endif
-
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <sys/time.h>
-
-#include <ctype.h>
-#include <stdio.h>
-#include <string.h>
-
-#ifndef ETHERS_FILE
-#define ETHERS_FILE "/etc/ethers"
-#endif
-
-struct etherent {
- u_char addr[6];
- char name[122];
-};
-
-static FILE *ether_fp = NULL;
-
-
-/* Hex digit to integer. */
-static inline int
-xdtoi(c)
- int c;
-{
-
- if (isdigit(c))
- return c - '0';
- else if (islower(c))
- return c - 'a' + 10;
- else
- return c - 'A' + 10;
-}
-
-static inline int
-skip_space(f)
- FILE *f;
-{
- int c;
-
- do {
- c = getc(f);
- } while (isspace(c) && c != '\n');
-
- return c;
-}
-
-static inline int
-skip_line(f)
- FILE *f;
-{
- int c;
-
- do
- c = getc(f);
- while (c != '\n' && c != EOF);
-
- return c;
-}
-
-static struct etherent *
-next_etherent(fp)
- FILE *fp;
-{
- register int c, d, i;
- char *bp;
- static struct etherent e;
- static int nline = 1;
- top:
- while (nline) {
- /* Find addr */
- c = skip_space(fp);
- if (c == '\n')
- continue;
- /* If this is a comment, or first thing on line
- cannot be etehrnet address, skip the line. */
- else if (!isxdigit(c))
- c = skip_line(fp);
- else {
- /* must be the start of an address */
- for (i = 0; i < 6; i += 1) {
- d = xdtoi(c);
- c = getc(fp);
- if (c != ':') {
- d <<= 4;
- d |= xdtoi(c);
- c = getc(fp);
- }
- e.addr[i] = d;
- if (c != ':')
- break;
- c = getc(fp);
- }
- nline = 0;
- }
- if (c == EOF)
- return NULL;
- }
-
- /* If we started a new line, 'c' holds the char past the ether addr,
- which we assume is white space. If we are continuing a line,
- 'c' is garbage. In either case, we can throw it away. */
-
- c = skip_space(fp);
- if (c == '\n') {
- nline = 1;
- goto top;
- }
- else if (c == '#') {
- (void)skip_line(fp);
- nline = 1;
- goto top;
- }
- else if (c == EOF)
- return NULL;
-
- /* Must be a name. */
- bp = e.name;
- /* Use 'd' to prevent buffer overflow. */
- d = sizeof(e.name) - 1;
- do {
- *bp++ = c;
- c = getc(fp);
- } while (!isspace(c) && c != EOF && --d > 0);
- *bp = '\0';
- if (c == '\n')
- nline = 1;
-
- return &e;
-}
-
-/* Open/rewind the ethers files; returns 1 if file was reopened */
-int
-ether_rewind()
-{
- struct stat st;
- static long mtime = 0, ctime = 0;
-
- if (ether_fp != NULL) {
- if (fstat(fileno(ether_fp), &st) < 0 ||
- mtime != st.st_mtime || ctime != st.st_ctime ||
- fseek(ether_fp, 0L, SEEK_SET) < 0) {
- fclose(ether_fp);
- ether_fp = NULL;
- }
- }
- if (ether_fp == NULL) {
- ether_fp = fopen(ETHERS_FILE, "r");
- if (ether_fp == NULL)
- return (-1);
- if (fstat(fileno(ether_fp), &st) < 0) {
- fclose(ether_fp);
- ether_fp = NULL;
- return (-1);
- }
- mtime = st.st_mtime;
- ctime = st.st_ctime;
- return (1);
- }
- return (0);
-}
-
-/* Map an ethernet address to a name; returns 0 on success, else 1. */
-int
-ether_ntohost(name, ea)
- register char *name;
- register u_char *ea;
-{
- register struct etherent *ep;
-
- if (ether_rewind() < 0)
- return (1);
-
- while ((ep = next_etherent(ether_fp)) != NULL)
- if (bcmp(ep->addr, ea, 6) == 0) {
- strcpy(name, ep->name);
- return (0);
- }
- return (1);
-}
-
-/* Map an ethernet name to an address; returns 0 on success, else 1. */
-int
-ether_hostton(name, ea)
- register char *name;
- register u_char *ea;
-{
- register struct etherent *ep;
-
- if (ether_rewind() < 0)
- return (1);
-
- while ((ep = next_etherent(ether_fp)) != NULL)
- if (strcmp(ep->name, name) == 0) {
- bcopy(ep->addr, ea, 6);
- return (0);
- }
- return (1);
-}
diff --git a/usr.sbin/rarpd/rarpd.c b/usr.sbin/rarpd/rarpd.c
index 90ca240..f3a84f4 100644
--- a/usr.sbin/rarpd/rarpd.c
+++ b/usr.sbin/rarpd/rarpd.c
@@ -1,6 +1,6 @@
/*
- * Copyright (c) 1990 The Regents of the University of California.
- * All rights reserved.
+ * Copyright (c) 1990, 1991, 1992, 1993, 1996
+ * The 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: (1) source code distributions
@@ -20,60 +20,91 @@
*/
#ifndef lint
char copyright[] =
-"@(#) Copyright (c) 1990 The Regents of the University of California.\n\
- All rights reserved.\n";
+"@(#) Copyright (c) 1990, 1991, 1992, 1993, 1996\n\
+The Regents of the University of California. All rights reserved.\n";
#endif /* not lint */
#ifndef lint
-static char rcsid[] =
- "@(#) $Header: /home/ncvs/src/usr.sbin/rarpd/rarpd.c,v 1.6 1996/08/24 23:05:08 wpaul Exp $ (LBL)";
+static const char rcsid[] =
+ "@(#) $Header: rarpd.c,v 1.22 96/06/14 20:40:14 leres Exp $ (LBL)";
#endif
-
/*
* rarpd - Reverse ARP Daemon
*
- * Usage: rarpd -a [ -f ] [ hostname ]
- * rarpd [ -f ] interface [ hostname ]
+ * Usage: rarpd -a [ -fv ] [ hostname ]
+ * rarpd [ -fv ] interface [ hostname ]
*
* 'hostname' is optional solely for backwards compatibility with Sun's rarpd.
* Currently, the argument is ignored.
*/
-
-#include <stdio.h>
-#include <syslog.h>
-#include <string.h>
-#include <strings.h>
-#include <sys/types.h>
-/* SunOS 4.x defines this while 3.x does not. */
-#ifdef __sys_types_h
-#define SUNOS4
-#endif
+#include <sys/param.h>
+#include <sys/file.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
#include <sys/time.h>
+
#include <net/bpf.h>
-#include <sys/socket.h>
-#include <sys/ioctl.h>
#include <net/if.h>
+
+#if BSD >= 199100
+#include <net/if_types.h>
+#include <net/if_dl.h>
+#if BSD >= 199200
+#include <net/route.h>
+#endif
+#endif
+
#include <netinet/in.h>
#include <netinet/if_ether.h>
-#include <sys/errno.h>
-#include <sys/file.h>
+
+#include <arpa/inet.h>
+
+#include <errno.h>
#include <netdb.h>
+#include <stdio.h>
+#include <string.h>
+#include <syslog.h>
+#if BSD >= 199200
+#include <stdlib.h>
+#include <unistd.h>
+#else
-#if defined(SUNOS4) || defined(__FreeBSD__) /* XXX */
-#define HAVE_DIRENT_H
+extern char *optarg;
+extern int optind, opterr;
+
+extern int errno;
#endif
-#ifdef HAVE_DIRENT_H
+#ifdef SUNOS4
#include <dirent.h>
#else
#include <sys/dir.h>
#endif
+/* Cast a struct sockaddr to a structaddr_in */
+#define SATOSIN(sa) ((struct sockaddr_in *)(sa))
+
+#ifndef TFTP_DIR
+#define TFTP_DIR "/tftpboot"
+#endif
+
+#if BSD >= 199200
+#define ARPSECS (20 * 60) /* as per code in netinet/if_ether.c */
+#define REVARP_REQUEST ARPOP_REVREQUEST
+#define REVARP_REPLY ARPOP_REVREPLY
+#endif
+
+#ifndef ETHERTYPE_REVARP
+#define ETHERTYPE_REVARP 0x8035
+#define REVARP_REQUEST 3
+#define REVARP_REPLY 4
+#endif
+
/*
* Map field names in ether_arp struct. What a pain in the neck.
*/
-#if !defined(SUNOS4) && !defined(__FreeBSD__)
+#ifdef SUNOS3
#undef arp_sha
#undef arp_spa
#undef arp_tha
@@ -88,18 +119,16 @@ static char rcsid[] =
#define inline
#endif
-extern int errno;
-extern int ether_ntohost __P((char *, struct ether_addr *));
-
/*
* The structure for each interface.
*/
struct if_info {
- int ii_fd; /* BPF file descriptor */
- u_char ii_eaddr[6]; /* Ethernet address of this interface */
+ struct if_info *ii_next;
+ int ii_fd; /* BPF file descriptor */
u_long ii_ipaddr; /* IP address of this interface */
u_long ii_netmask; /* subnet or net mask */
- struct if_info *ii_next;
+ u_char ii_eaddr[6]; /* Ethernet address of this interface */
+ char ii_ifname[sizeof(((struct ifreq *)0)->ifr_name) + 1];
};
/*
@@ -108,31 +137,47 @@ struct if_info {
*/
struct if_info *iflist;
-extern char *malloc();
-extern void exit();
+int verbose; /* verbose messages */
+int s; /* inet datagram socket */
+char *tftp_dir = TFTP_DIR; /* tftp directory */
-u_long ipaddrtonetmask();
-void init_one();
-void init_all();
-void rarp_loop();
-void lookup_eaddr();
-void lookup_ipaddr();
+#ifndef __P
+#define __P(protos) ()
+#endif
+
+#if BSD < 199200
+extern char *malloc();
+extern void exit();
+#endif
+extern int ether_ntohost();
+
+void init __P((char *));
+void init_one __P((struct ifreq *, char *));
+char *intoa __P((u_long));
+u_long ipaddrtonetmask __P((u_long));
+char *eatoa __P((u_char *));
+int rarp_bootable __P((u_long));
+void rarp_loop __P((void));
+int rarp_open __P((char *));
+void rarp_process __P((struct if_info *, u_char *, u_int));
+void rarp_reply __P((struct if_info *, struct ether_header *, u_long, u_int));
+void update_arptab __P((u_char *, u_long));
+void usage __P((void));
+
+static u_char zero[6];
void
main(argc, argv)
int argc;
char **argv;
{
- int op, pid;
+ int op;
char *ifname, *hostname, *name;
int aflag = 0; /* listen on "all" interfaces */
int fflag = 0; /* don't fork */
- extern char *optarg;
- extern int optind, opterr;
-
- if (name = strrchr(argv[0], '/'))
+ if ((name = strrchr(argv[0], '/')) != NULL)
++name;
else
name = argv[0];
@@ -142,10 +187,10 @@ main(argc, argv)
/*
* All error reporting is done through syslogs.
*/
- openlog(name, LOG_PID, LOG_DAEMON);
+ openlog(name, LOG_PID | LOG_CONS, LOG_DAEMON);
opterr = 0;
- while ((op = getopt(argc, argv, "af")) != EOF) {
+ while ((op = getopt(argc, argv, "afv")) != EOF) {
switch (op) {
case 'a':
++aflag;
@@ -155,97 +200,209 @@ main(argc, argv)
++fflag;
break;
+ case 'v':
+ ++verbose;
+ break;
+
default:
usage();
/* NOTREACHED */
}
}
ifname = argv[optind++];
- hostname = ifname ? argv[optind] : 0;
- if ((aflag && ifname) || (!aflag && ifname == 0))
+ hostname = ifname ? argv[optind] : NULL;
+ if ((aflag && ifname) || (!aflag && ifname == NULL))
usage();
if (aflag)
- init_all();
+ init(NULL);
else
- init_one(ifname);
+ init(ifname);
- if (!fflag)
+ if (!fflag) {
if (daemon(0,0)) {
- perror("fork");
- exit(0);
+ syslog(LOG_ERR, "cannot fork");
+ exit(1);
}
+ }
rarp_loop();
}
/*
- * Add 'ifname' to the interface list. Lookup its IP address and network
- * mask and Ethernet address, and open a BPF file for it.
+ * Add to the interface list.
*/
void
-init_one(ifname)
- char *ifname;
+init_one(ifrp, target)
+ register struct ifreq *ifrp;
+ register char *target;
{
- struct if_info *p;
+ register struct if_info *ii;
+ register struct sockaddr_dl *ll;
+ int family;
+ struct ifreq ifr;
+ family = ifrp->ifr_addr.sa_family;
+ switch (family) {
- p = (struct if_info *)malloc(sizeof(*p));
- p->ii_next = iflist;
- iflist = p;
+ case AF_INET:
+#if BSD >= 199100
+ case AF_LINK:
+#endif
+ (void)strncpy(ifr.ifr_name, ifrp->ifr_name,
+ sizeof(ifrp->ifr_name));
+ if (ioctl(s, SIOCGIFFLAGS, (char *)&ifr) < 0) {
+ syslog(LOG_ERR,
+ "SIOCGIFFLAGS: %.*s: %m",
+ sizeof(ifrp->ifr_name), ifrp->ifr_name);
+ exit(1);
+ }
+ if ((ifr.ifr_flags & IFF_UP) == 0 ||
+ (ifr.ifr_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0)
+ return;
+ break;
- p->ii_fd = rarp_open(ifname);
- lookup_eaddr(p->ii_fd, p->ii_eaddr);
- lookup_ipaddr(ifname, &p->ii_ipaddr, &p->ii_netmask);
-}
+ default:
+ return;
+ }
+
+ /* Don't bother going any further if not the target interface */
+ if (target != NULL &&
+ strncmp(ifrp->ifr_name, target, sizeof(ifrp->ifr_name)) != 0)
+ return;
+
+ /* Look for interface in list */
+ for (ii = iflist; ii != NULL; ii = ii->ii_next)
+ if (strncmp(ifrp->ifr_name, ii->ii_ifname,
+ sizeof(ifrp->ifr_name)) == 0)
+ break;
+
+ /* Allocate a new one if not found */
+ if (ii == NULL) {
+ ii = (struct if_info *)malloc(sizeof(*ii));
+ if (ii == NULL) {
+ syslog(LOG_ERR, "malloc: %m");
+ exit(1);
+ }
+ bzero(ii, sizeof(*ii));
+ ii->ii_fd = -1;
+ (void)strncpy(ii->ii_ifname, ifrp->ifr_name,
+ sizeof(ifrp->ifr_name));
+ ii->ii_ifname[sizeof(ii->ii_ifname) - 1] = '\0';
+ ii->ii_next = iflist;
+ iflist = ii;
+ }
+
+ switch (family) {
+
+ case AF_INET:
+ if (ioctl(s, SIOCGIFADDR, (char *)&ifr) < 0) {
+ syslog(LOG_ERR, "ipaddr SIOCGIFADDR: %s: %m",
+ ii->ii_ifname);
+ exit(1);
+ }
+ ii->ii_ipaddr = SATOSIN(&ifr.ifr_addr)->sin_addr.s_addr;
+ if (ioctl(s, SIOCGIFNETMASK, (char *)&ifr) < 0) {
+ syslog(LOG_ERR, "SIOCGIFNETMASK: %m");
+ exit(1);
+ }
+ ii->ii_netmask = SATOSIN(&ifr.ifr_addr)->sin_addr.s_addr;
+ if (ii->ii_netmask == 0)
+ ii->ii_netmask = ipaddrtonetmask(ii->ii_ipaddr);
+ if (ii->ii_fd < 0) {
+ ii->ii_fd = rarp_open(ii->ii_ifname);
+#if BSD < 199100
+ /* Use BPF descriptor to get ethernet address. */
+ if (ioctl(ii->ii_fd, SIOCGIFADDR, (char *)&ifr) < 0) {
+ syslog(LOG_ERR, "eaddr SIOCGIFADDR: %s: %m",
+ ii->ii_ifname);
+ exit(1);
+ }
+ bcopy(&ifr.ifr_addr.sa_data[0], ii->ii_eaddr, 6);
+#endif
+ }
+ break;
+
+#if BSD >= 199100
+ case AF_LINK:
+ ll = (struct sockaddr_dl *)&ifrp->ifr_addr;
+ if (ll->sdl_type == IFT_ETHER)
+ bcopy(LLADDR(ll), ii->ii_eaddr, 6);
+ break;
+#endif
+ }
+}
/*
* Initialize all "candidate" interfaces that are in the system
* configuration list. A "candidate" is up, not loopback and not
* point to point.
*/
void
-init_all()
+init(target)
+ char *target;
{
- int fd;
- int ifflags;
- struct ifreq ibuf[8], tmp_ibuf, *ifptr, *n;
+ register int n;
+ register struct ifreq *ifrp, *ifend;
+ register struct if_info *ii, *nii, *lii;
struct ifconf ifc;
+ struct ifreq ibuf[16];
- if ((fd = socket(AF_INET, SOCK_DGRAM, 0)) < 0) {
+ if ((s = socket(AF_INET, SOCK_DGRAM, 0)) < 0) {
syslog(LOG_ERR, "socket: %m");
exit(1);
}
ifc.ifc_len = sizeof ibuf;
ifc.ifc_buf = (caddr_t)ibuf;
- if (ioctl(fd, SIOCGIFCONF, (char *)&ifc) < 0 ||
- ifc.ifc_len < sizeof(struct ifreq)) {
+ if (ioctl(s, SIOCGIFCONF, (char *)&ifc) < 0 ||
+ (u_int)ifc.ifc_len < sizeof(struct ifreq)) {
syslog(LOG_ERR, "SIOCGIFCONF: %m");
exit(1);
}
- ifptr = ifc.ifc_req;
- ifflags = ifptr->ifr_flags;
- n = (struct ifreq *) (ifc.ifc_buf + ifc.ifc_len);
- while (ifptr < n) {
- bcopy((char *)ifptr, (char *)&tmp_ibuf, sizeof(struct ifreq));
- if (ioctl(fd, SIOCGIFFLAGS, (char *)&tmp_ibuf) < 0) {
- syslog(LOG_ERR, "SIOCGIFFLAGS: %m");
- exit(1);
+ ifrp = ibuf;
+ ifend = (struct ifreq *)((char *)ibuf + ifc.ifc_len);
+ while (ifrp < ifend) {
+ init_one(ifrp, target);
+
+#if BSD >= 199100
+ n = ifrp->ifr_addr.sa_len + sizeof(ifrp->ifr_name);
+ if (n < sizeof(*ifrp))
+ n = sizeof(*ifrp);
+ ifrp = (struct ifreq *)((char *)ifrp + n);
+#else
+ ++ifrp;
+#endif
+ }
+
+ /* Throw away incomplete interfaces */
+ lii = NULL;
+ for (ii = iflist; ii != NULL; ii = nii) {
+ nii = ii->ii_next;
+ if (ii->ii_ipaddr == 0 ||
+ bcmp(ii->ii_eaddr, zero, 6) == 0) {
+ if (lii == NULL)
+ iflist = nii;
+ else
+ lii->ii_next = nii;
+ if (ii->ii_fd >= 0)
+ close(ii->ii_fd);
+ free(ii);
+ continue;
}
- if (ifptr->ifr_flags == ifflags && (tmp_ibuf.ifr_flags &
- (IFF_UP | IFF_LOOPBACK | IFF_POINTOPOINT)) == IFF_UP)
- init_one(ifptr->ifr_name);
- if(ifptr->ifr_addr.sa_len) /* Dohw! */
- ifptr = (struct ifreq *) ((caddr_t) ifptr +
- ifptr->ifr_addr.sa_len -
- sizeof(struct sockaddr));
- ifptr++;
+ lii = ii;
}
- (void)close(fd);
+
+ /* Verbose stuff */
+ if (verbose)
+ for (ii = iflist; ii != NULL; ii = ii->ii_next)
+ syslog(LOG_DEBUG, "%s %s 0x%x %s",
+ ii->ii_ifname, intoa(ii->ii_ipaddr),
+ ii->ii_netmask, eatoa(ii->ii_eaddr));
}
+void
usage()
{
- (void)fprintf(stderr, "usage: rarpd [ -af ] [ interface ]\n");
+ (void)fprintf(stderr, "usage: rarpd [ -afv ] [ interface ]\n");
exit(1);
}
@@ -266,7 +423,7 @@ bpf_open()
if (fd < 0) {
syslog(LOG_ERR, "%s: %m", device);
- exit(-1);
+ exit(1);
}
return fd;
}
@@ -281,22 +438,22 @@ rarp_open(device)
{
int fd;
struct ifreq ifr;
- int immediate, link_type;
+ u_int dlt;
+ int immediate;
static struct bpf_insn insns[] = {
- BPF_STMT(BPF_LD+BPF_H+BPF_ABS, 12),
- BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, ETHERTYPE_REVARP, 0, 3),
- BPF_STMT(BPF_LD+BPF_H+BPF_ABS, 20),
- BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, ARPOP_REVREQUEST, 0, 1),
- BPF_STMT(BPF_RET+BPF_K, sizeof(struct ether_arp) +
- sizeof(struct ether_header)),
- BPF_STMT(BPF_RET+BPF_K, 0),
- };
-
- static struct bpf_program filter = {
- sizeof insns / sizeof(insns[0]),
- (struct bpf_insn *)&insns
- };
+ BPF_STMT(BPF_LD|BPF_H|BPF_ABS, 12),
+ BPF_JUMP(BPF_JMP|BPF_JEQ|BPF_K, ETHERTYPE_REVARP, 0, 3),
+ BPF_STMT(BPF_LD|BPF_H|BPF_ABS, 20),
+ BPF_JUMP(BPF_JMP|BPF_JEQ|BPF_K, REVARP_REQUEST, 0, 1),
+ BPF_STMT(BPF_RET|BPF_K, sizeof(struct ether_arp) +
+ sizeof(struct ether_header)),
+ BPF_STMT(BPF_RET|BPF_K, 0),
+ };
+ static struct bpf_program filter = {
+ sizeof insns / sizeof(insns[0]),
+ insns
+ };
fd = bpf_open();
/*
@@ -316,12 +473,12 @@ rarp_open(device)
* Check that the data link layer is an Ethernet; this code won't
* work with anything else.
*/
- if (ioctl(fd, BIOCGDLT, &link_type) < 0) {
- syslog(LOG_ERR, "BIOCGDLP: %m");
+ if (ioctl(fd, BIOCGDLT, (caddr_t)&dlt) < 0) {
+ syslog(LOG_ERR, "BIOCGDLT: %m");
exit(1);
}
- if (link_type != DLT_EN10MB) {
- syslog(LOG_ERR, "%s not on ethernet", device);
+ if (dlt != DLT_EN10MB) {
+ syslog(LOG_ERR, "%s is not an ethernet", device);
exit(1);
}
/*
@@ -341,22 +498,23 @@ rarp_open(device)
static int
rarp_check(p, len)
u_char *p;
- int len;
+ u_int len;
{
struct ether_header *ep = (struct ether_header *)p;
struct ether_arp *ap = (struct ether_arp *)(p + sizeof(*ep));
if (len < sizeof(*ep) + sizeof(*ap)) {
syslog(LOG_ERR, "truncated request");
+ printf ("len: %d expected: %d\n", len, sizeof(*ep) + sizeof(*ap));
return 0;
}
/*
* XXX This test might be better off broken out...
*/
- if (ep->ether_type != htons(ETHERTYPE_REVARP) ||
- ap->arp_hrd != htons(ARPHRD_ETHER) ||
- ap->arp_op != htons(ARPOP_REVREQUEST) ||
- ap->arp_pro != htons(ETHERTYPE_IP) ||
+ if (ntohs(ep->ether_type) != ETHERTYPE_REVARP ||
+ ntohs(ap->arp_hrd) != ARPHRD_ETHER ||
+ ntohs(ap->arp_op) != REVARP_REQUEST ||
+ ntohs(ap->arp_pro) != ETHERTYPE_IP ||
ap->arp_hln != 6 || ap->arp_pln != 4) {
syslog(LOG_DEBUG, "request fails sanity check");
return 0;
@@ -385,14 +543,13 @@ rarp_check(p, len)
void
rarp_loop()
{
- struct bpf_hdr *bhp;
- u_char *pkt;
+ u_char *buf, *bp, *ep;
int cc, fd;
fd_set fds, listeners;
int bufsize, maxfd = 0;
struct if_info *ii;
- if (iflist == 0) {
+ if (iflist == NULL) {
syslog(LOG_ERR, "no interfaces");
exit(1);
}
@@ -400,102 +557,116 @@ rarp_loop()
syslog(LOG_ERR, "BIOCGBLEN: %m");
exit(1);
}
- bhp = (struct bpf_hdr *)malloc((unsigned)bufsize);
+ buf = (u_char *)malloc((unsigned)bufsize);
+ if (buf == NULL) {
+ syslog(LOG_ERR, "malloc: %m");
+ exit(1);
+ }
+ while (1) {
/*
* Find the highest numbered file descriptor for select().
* Initialize the set of descriptors to listen to.
*/
- FD_ZERO(&fds);
- for (ii = iflist; ii; ii = ii->ii_next) {
- FD_SET(ii->ii_fd, &fds);
- if (ii->ii_fd > maxfd)
- maxfd = ii->ii_fd;
- }
- while (1) {
+ FD_ZERO(&fds);
+ for (ii = iflist; ii != NULL; ii = ii->ii_next) {
+ FD_SET(ii->ii_fd, &fds);
+ if (ii->ii_fd > maxfd)
+ maxfd = ii->ii_fd;
+ }
listeners = fds;
- if (select(maxfd + 1, &listeners, (struct fd_set *)0,
- (struct fd_set *)0, (struct timeval *)0) < 0) {
+ if (select(maxfd + 1, &listeners, NULL, NULL, NULL) < 0) {
+ /* Don't choke when we get ptraced */
+ if (errno == EINTR)
+ continue;
syslog(LOG_ERR, "select: %m");
exit(1);
}
- for (ii = iflist; ii; ii = ii->ii_next) {
+ for (ii = iflist; ii != NULL; ii = ii->ii_next) {
fd = ii->ii_fd;
- if (FD_ISSET(fd, &listeners)) {
- again:
- cc = read(fd, (char *)bhp, bufsize);
- /*
- * Due to a SunOS bug, after 2^31 bytes, the
- * file offset overflows and read fails with
- * EINVAL. The lseek() to 0 will fix things.
- */
- if (cc < 0) {
- if (errno == EINVAL &&
- (long)(lseek(fd, 0L, SEEK_CUR) + bufsize) < 0) {
- (void)lseek(fd, 0, 0);
- goto again;
- }
- syslog(LOG_ERR, "read: %m");
- exit(1);
+ if (!FD_ISSET(fd, &listeners))
+ continue;
+ again:
+ cc = read(fd, (char *)buf, bufsize);
+ /* Don't choke when we get ptraced */
+ if (cc < 0 && errno == EINTR)
+ goto again;
+#if defined(SUNOS3) || defined(SUNOS4)
+ /*
+ * Due to a SunOS bug, after 2^31 bytes, the
+ * file offset overflows and read fails with
+ * EINVAL. The lseek() to 0 will fix things.
+ */
+ if (cc < 0) {
+ if (errno == EINVAL &&
+ (long)(tell(fd) + bufsize) < 0) {
+ (void)lseek(fd, 0, 0);
+ goto again;
}
- pkt = (u_char *)bhp + bhp->bh_hdrlen;
+ syslog(LOG_ERR, "read: %m");
+ exit(1);
+ }
+#endif
- if (rarp_check(pkt, (int)bhp->bh_datalen))
- rarp_process(ii, pkt);
+ /* Loop through the packet(s) */
+#define bhp ((struct bpf_hdr *)bp)
+ bp = buf;
+ ep = bp + cc;
+ while (bp < ep) {
+ register u_int caplen, hdrlen;
+
+ caplen = bhp->bh_caplen;
+ hdrlen = bhp->bh_hdrlen;
+ if (rarp_check(bp + hdrlen, caplen))
+ rarp_process(ii, bp + hdrlen, cc);
+ bp += BPF_WORDALIGN(hdrlen + caplen);
}
}
}
+#undef bhp
}
-#ifndef TFTP_DIR
-#define TFTP_DIR "/tftpboot"
-#endif
-
/*
* True if this server can boot the host whose IP address is 'addr'.
* This check is made by looking in the tftp directory for the
* configuration file.
*/
+int
rarp_bootable(addr)
u_long addr;
{
-
-#ifdef HAVE_DIRENT_H
+#ifdef SUNOS4
register struct dirent *dent;
#else
register struct direct *dent;
#endif
register DIR *d;
char ipname[9];
- static DIR *dd = 0;
+ static DIR *dd = NULL;
+
+ (void)sprintf(ipname, "%08X", (unsigned int )addr);
- /*
- * XXX Need to htonl() the IP address or it'll
- * come out backwards.
- */
- (void)sprintf(ipname, "%08X", htonl(addr));
/*
* If directory is already open, rewind it. Otherwise, open it.
*/
- if (d = dd)
+ if ((d = dd) != NULL)
rewinddir(d);
else {
- if (chdir(TFTP_DIR) == -1) {
- syslog(LOG_ERR, "chdir: %m");
+ if (chdir(tftp_dir) == -1) {
+ syslog(LOG_ERR, "chdir: %s: %m", tftp_dir);
exit(1);
}
d = opendir(".");
- if (d == 0) {
+ if (d == NULL) {
syslog(LOG_ERR, "opendir: %m");
exit(1);
}
dd = d;
}
- while (dent = readdir(d))
+ while ((dent = readdir(d)) != NULL)
if (strncmp(dent->d_name, ipname, 8) == 0)
return 1;
return 0;
-
}
/*
@@ -509,26 +680,21 @@ choose_ipaddr(alist, net, netmask)
u_long net;
u_long netmask;
{
- for (; *alist; ++alist) {
+ for (; *alist; ++alist)
if ((**alist & netmask) == net)
return **alist;
- }
return 0;
}
/*
- * A one entry ip/ethernet address cache.
- */
-static u_long cache_ipaddr;
-static u_char cache_eaddr[6];
-
-/*
* Answer the RARP request in 'pkt', on the interface 'ii'. 'pkt' has
* already been checked for validity. The reply is overlaid on the request.
*/
-rarp_process(ii, pkt)
+void
+rarp_process(ii, pkt, len)
struct if_info *ii;
u_char *pkt;
+ u_int len;
{
struct ether_header *ep;
struct hostent *hp;
@@ -536,107 +702,155 @@ rarp_process(ii, pkt)
char ename[256];
ep = (struct ether_header *)pkt;
+ if (ether_ntohost(ename, &ep->ether_shost) != 0)
+ return;
+
+ if((hp = gethostbyname(ename)) == NULL)
+ return;
/*
- * If the address in the one element cache, don't bother
- * looking up names.
+ * Choose correct address from list.
*/
- if (bcmp((char *)cache_eaddr, (char *)&ep->ether_shost, 6) == 0)
- target_ipaddr = cache_ipaddr;
- else {
- if (ether_ntohost(ename, (struct ether_addr *)&ep->ether_shost) != 0 ||
- (hp = gethostbyname(ename)) == 0)
- return;
- /*
- * Choose correct address from list.
- */
- if (hp->h_addrtype != AF_INET) {
- syslog(LOG_ERR, "cannot handle non IP addresses");
- exit(1);
- }
- target_ipaddr = choose_ipaddr((u_long **)hp->h_addr_list,
- ii->ii_ipaddr & ii->ii_netmask,
- ii->ii_netmask);
- if (target_ipaddr == 0) {
- syslog(LOG_ERR, "cannot find %s on %08x",
- ename, ii->ii_ipaddr & ii->ii_netmask);
- return;
- }
- bcopy((char *)&ep->ether_shost, (char *)cache_eaddr, 6);
- cache_ipaddr = target_ipaddr;
+ if (hp->h_addrtype != AF_INET) {
+ syslog(LOG_ERR, "cannot handle non IP addresses");
+ exit(1);
+ }
+ target_ipaddr = htonl(choose_ipaddr((u_long **)hp->h_addr_list,
+ ii->ii_ipaddr & ii->ii_netmask,
+ ii->ii_netmask));
+ if (target_ipaddr == 0) {
+ syslog(LOG_ERR, "cannot find %s on net %s",
+ ename, intoa(htonl(ii->ii_ipaddr & ii->ii_netmask)));
+ return;
}
if (rarp_bootable(target_ipaddr))
- rarp_reply(ii, ep, target_ipaddr);
+ rarp_reply(ii, ep, target_ipaddr, len);
+ else if (verbose > 1)
+ syslog(LOG_DEBUG, "%s %s at %s DENIED (not bootable)",
+ ii->ii_ifname,
+ eatoa(ii->ii_eaddr),
+ intoa(ii->ii_ipaddr));
}
/*
- * Lookup the ethernet address of the interface attached to the BPF
- * file descriptor 'fd'; return it in 'eaddr'.
+ * Poke the kernel arp tables with the ethernet/ip address combinataion
+ * given. When processing a reply, we must do this so that the booting
+ * host (i.e. the guy running rarpd), won't try to ARP for the hardware
+ * address of the guy being booted (he cannot answer the ARP).
*/
+#if BSD >= 199200
+static struct sockaddr_inarp sin_inarp = {
+ sizeof(struct sockaddr_inarp), AF_INET
+};
+static struct sockaddr_dl sin_dl = {
+ sizeof(struct sockaddr_dl), AF_LINK, 0, IFT_ETHER, 0, 6
+};
+static struct {
+ struct rt_msghdr rthdr;
+ char rtspace[512];
+} rtmsg;
+
void
-lookup_eaddr(fd, eaddr)
- int fd;
- u_char *eaddr;
+update_arptab(ep, ipaddr)
+ u_char *ep;
+ u_long ipaddr;
{
- struct ifreq ifr;
+ register int cc;
+ register struct sockaddr_inarp *ar, *ar2;
+ register struct sockaddr_dl *ll, *ll2;
+ register struct rt_msghdr *rt;
+ register int xtype, xindex;
+ static pid_t pid;
+ static int r, seq;
+ static init = 0;
+
+ if (!init) {
+ r = socket(PF_ROUTE, SOCK_RAW, 0);
+ if (r < 0) {
+ syslog(LOG_ERR, "raw route socket: %m");
+ exit(1);
+ }
+ pid = getpid();
+ ++init;
+ }
- /* Use BPF descriptor to get ethernet address. */
- if (ioctl(fd, SIOCGIFADDR, (char *)&ifr) < 0) {
- syslog(LOG_ERR, "SIOCGIFADDR: %m");
+ ar = &sin_inarp;
+ ar->sin_addr.s_addr = ipaddr;
+ ll = &sin_dl;
+ bcopy(ep, LLADDR(ll), 6);
+
+ /* Get the type and interface index */
+ rt = &rtmsg.rthdr;
+ bzero(rt, sizeof(rtmsg));
+ rt->rtm_version = RTM_VERSION;
+ rt->rtm_addrs = RTA_DST;
+ rt->rtm_type = RTM_GET;
+ rt->rtm_seq = ++seq;
+ ar2 = (struct sockaddr_inarp *)rtmsg.rtspace;
+ bcopy(ar, ar2, sizeof(*ar));
+ rt->rtm_msglen = sizeof(*rt) + sizeof(*ar);
+ errno = 0;
+ if (write(r, rt, rt->rtm_msglen) < 0 && errno != ESRCH) {
+ syslog(LOG_ERR, "rtmsg get write: %m");
exit(1);
}
- bcopy((char *)&ifr.ifr_addr.sa_data[0], (char *)eaddr, 6);
-}
-
-/*
- * Lookup the IP address and network mask of the interface named 'ifname'.
- */
-void
-lookup_ipaddr(ifname, addrp, netmaskp)
- char *ifname;
- u_long *addrp;
- u_long *netmaskp;
-{
- int fd;
- struct ifreq ifr;
-
- /* Use data gram socket to get IP address. */
- if ((fd = socket(AF_INET, SOCK_DGRAM, 0)) < 0) {
- syslog(LOG_ERR, "socket: %m");
+ do {
+ cc = read(r, rt, sizeof(rtmsg));
+ } while (cc > 0 && (rt->rtm_seq != seq || rt->rtm_pid != pid));
+ if (cc < 0) {
+ syslog(LOG_ERR, "rtmsg get read: %m");
exit(1);
}
- (void)strncpy(ifr.ifr_name, ifname, sizeof ifr.ifr_name);
- if (ioctl(fd, SIOCGIFADDR, (char *)&ifr) < 0) {
- syslog(LOG_ERR, "SIOCGIFADDR: %m");
+ ll2 = (struct sockaddr_dl *)((u_char *)ar2 + ar2->sin_len);
+ if (ll2->sdl_family != AF_LINK) {
+ /*
+ * XXX I think this means the ip address is not on a
+ * directly connected network (the family is AF_INET in
+ * this case).
+ */
+ syslog(LOG_ERR, "bogus link family (%d) wrong net?\n",
+ ll2->sdl_family);
+ return;
+ }
+ xtype = ll2->sdl_type;
+ xindex = ll2->sdl_index;
+
+ /* Set the new arp entry */
+ bzero(rt, sizeof(rtmsg));
+ rt->rtm_version = RTM_VERSION;
+ rt->rtm_addrs = RTA_DST | RTA_GATEWAY;
+ rt->rtm_inits = RTV_EXPIRE;
+ rt->rtm_rmx.rmx_expire = time(0) + ARPSECS;
+ rt->rtm_flags = RTF_HOST | RTF_STATIC;
+ rt->rtm_type = RTM_ADD;
+ rt->rtm_seq = ++seq;
+
+ bcopy(ar, ar2, sizeof(*ar));
+
+ ll2 = (struct sockaddr_dl *)((u_char *)ar2 + sizeof(*ar2));
+ bcopy(ll, ll2, sizeof(*ll));
+ ll2->sdl_type = xtype;
+ ll2->sdl_index = xindex;
+
+ rt->rtm_msglen = sizeof(*rt) + sizeof(*ar2) + sizeof(*ll2);
+ errno = 0;
+ if (write(r, rt, rt->rtm_msglen) < 0 && errno != EEXIST) {
+ syslog(LOG_ERR, "rtmsg add write: %m");
exit(1);
}
- *addrp = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr;
- if (ioctl(fd, SIOCGIFNETMASK, (char *)&ifr) < 0) {
- perror("SIOCGIFNETMASK");
+ do {
+ cc = read(r, rt, sizeof(rtmsg));
+ } while (cc > 0 && (rt->rtm_seq != seq || rt->rtm_pid != pid));
+ if (cc < 0) {
+ syslog(LOG_ERR, "rtmsg add read: %m");
exit(1);
}
- *netmaskp = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr;
- /*
- * If SIOCGIFNETMASK didn't work, figure out a mask from
- * the IP address class.
- */
- if (*netmaskp == 0)
- *netmaskp = ipaddrtonetmask(*addrp);
-
- (void)close(fd);
}
-
-/*
- * Poke the kernel arp tables with the ethernet/ip address combinataion
- * given. When processing a reply, we must do this so that the booting
- * host (i.e. the guy running rarpd), won't try to ARP for the hardware
- * address of the guy being booted (he cannot answer the ARP).
- */
+#else
+void
update_arptab(ep, ipaddr)
u_char *ep;
u_long ipaddr;
{
-#ifdef SIOCSARP
- int s;
struct arpreq request;
struct sockaddr_in *sin;
@@ -647,25 +861,20 @@ update_arptab(ep, ipaddr)
request.arp_ha.sa_family = AF_UNSPEC;
bcopy((char *)ep, (char *)request.arp_ha.sa_data, 6);
- s = socket(AF_INET, SOCK_DGRAM, 0);
if (ioctl(s, SIOCSARP, (caddr_t)&request) < 0)
syslog(LOG_ERR, "SIOCSARP: %m");
- (void)close(s);
-#else
- if (arptab_set(ep, ipaddr) > 0)
- syslog(LOG_ERR, "couldn't update arp table");
-#endif
}
+#endif
/*
* Build a reverse ARP packet and sent it out on the interface.
- * 'ep' points to a valid ARPOP_REVREQUEST. The ARPOP_REVREPLY is built
+ * 'ep' points to a valid REVARP_REQUEST. The REVARP_REPLY is built
* on top of the request, then written to the network.
*
* RFC 903 defines the ether_arp fields as follows. The following comments
* are taken (more or less) straight from this document.
*
- * ARPOP_REVREQUEST
+ * REVARP_REQUEST
*
* arp_sha is the hardware address of the sender of the packet.
* arp_spa is undefined.
@@ -675,7 +884,7 @@ update_arptab(ep, ipaddr)
* address of the sender.
* arp_tpa is undefined.
*
- * ARPOP_REVREPLY
+ * REVARP_REPLY
*
* arp_sha is the hardware address of the responder (the sender of the
* reply packet).
@@ -690,45 +899,45 @@ update_arptab(ep, ipaddr)
* address pair (arp_spa, arp_sha) may eliminate the need for a subsequent
* ARP request.
*/
-rarp_reply(ii, ep, ipaddr)
+void
+rarp_reply(ii, ep, ipaddr, len)
struct if_info *ii;
struct ether_header *ep;
u_long ipaddr;
+ u_int len;
{
int n;
struct ether_arp *ap = (struct ether_arp *)(ep + 1);
- int len, raw_sock;
update_arptab((u_char *)&ap->arp_sha, ipaddr);
/*
* Build the rarp reply by modifying the rarp request in place.
*/
- ap->arp_op = htons(ARPOP_REVREPLY);
-
- /*
- * XXX Using htons(ETHERTYPE_REVARP) doesn't work: you wind
- * up transmitting 0x3580 instead of the correct value of
- * 0x8035. What makes no sense is that the NetBSD people
- * do in fact use htons(ETHERTYPE_REVARP) in their rarpd.
- * (Thank god for tcpdump or I would never have figured this
- * out.)
- */
- ep->ether_type = htons(ETHERTYPE_REVARP);
+ ap->arp_op = htons(REVARP_REPLY);
+#ifdef BROKEN_BPF
+ ep->ether_type = ETHERTYPE_REVARP;
+#endif
bcopy((char *)&ap->arp_sha, (char *)&ep->ether_dhost, 6);
bcopy((char *)ii->ii_eaddr, (char *)&ep->ether_shost, 6);
bcopy((char *)ii->ii_eaddr, (char *)&ap->arp_sha, 6);
+ ipaddr = htonl(ipaddr);
bcopy((char *)&ipaddr, (char *)ap->arp_tpa, 4);
/* Target hardware is unchanged. */
bcopy((char *)&ii->ii_ipaddr, (char *)ap->arp_spa, 4);
- len = sizeof(*ep) + sizeof(*ap);
+ /* Zero possible garbage after packet. */
+ bzero((char *)ep + (sizeof(*ep) + sizeof(*ap)),
+ len - (sizeof(*ep) + sizeof(*ap)));
n = write(ii->ii_fd, (char *)ep, len);
- if (n != len) {
+ if (n != len)
syslog(LOG_ERR, "write: only %d of %d bytes written", n, len);
- }
+ if (verbose)
+ syslog(LOG_DEBUG, "%s %s at %s", ii->ii_ifname,
+ eatoa(ii->ii_eaddr),
+ intoa(ii->ii_ipaddr));
}
/*
@@ -749,3 +958,47 @@ ipaddrtonetmask(addr)
exit(1);
/* NOTREACHED */
}
+
+/*
+ * A faster replacement for inet_ntoa().
+ */
+char *
+intoa(addr)
+ u_long addr;
+{
+ register char *cp;
+ register u_int byte;
+ register int n;
+ static char buf[sizeof(".xxx.xxx.xxx.xxx")];
+
+ cp = &buf[sizeof buf];
+ *--cp = '\0';
+
+ n = 4;
+ do {
+ byte = addr & 0xff;
+ *--cp = byte % 10 + '0';
+ byte /= 10;
+ if (byte > 0) {
+ *--cp = byte % 10 + '0';
+ byte /= 10;
+ if (byte > 0)
+ *--cp = byte + '0';
+ }
+ *--cp = '.';
+ addr >>= 8;
+ } while (--n > 0);
+
+ return cp + 1;
+}
+
+char *
+eatoa(ea)
+ register u_char *ea;
+{
+ static char buf[sizeof("xx:xx:xx:xx:xx:xx")];
+
+ (void)sprintf(buf, "%x:%x:%x:%x:%x:%x",
+ ea[0], ea[1], ea[2], ea[3], ea[4], ea[5]);
+ return (buf);
+}
OpenPOWER on IntegriCloud