busybox/networking/route.c
Ron Yorston 5c69ad0ecd build system: drop PLATFORM_LINUX
PLATFORM_LINUX is a hidden configuration option which is disabled by
default and enabled at over a hundred locations for features that are
deemed to be Linux specific.

The only effect of PLATFORM_LINUX is to control compilation of
libbb/match_fstype.c.  This file is only needed by mount and umount.

Remove all references to PLATFORM_LINUX and compile match_fstype.c
if mount or umount is enabled.

Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2020-08-13 17:12:56 +02:00

727 lines
18 KiB
C

/* vi: set sw=4 ts=4: */
/*
* route
*
* Similar to the standard Unix route, but with only the necessary
* parts for AF_INET and AF_INET6
*
* Bjorn Wesen, Axis Communications AB
*
* Author of the original route:
* Fred N. van Kempen, <waltje@uwalt.nl.mugnet.org>
* (derived from FvK's 'route.c 1.70 01/04/94')
*
* Licensed under GPLv2 or later, see file LICENSE in this source tree.
*
*
* displayroute() code added by Vladimir N. Oleynik <dzo@simtreas.ru>
* adjustments by Larry Doolittle <LRDoolittle@lbl.gov>
*
* IPV6 support added by Bart Visscher <magick@linux-fan.com>
*/
/* 2004/03/09 Manuel Novoa III <mjn3@codepoet.org>
*
* Rewritten to fix several bugs, add additional error checking, and
* remove ridiculous amounts of bloat.
*/
//config:config ROUTE
//config: bool "route (8.7 kb)"
//config: default y
//config: help
//config: Route displays or manipulates the kernel's IP routing tables.
//applet:IF_ROUTE(APPLET(route, BB_DIR_SBIN, BB_SUID_DROP))
//kbuild:lib-$(CONFIG_ROUTE) += route.o
//usage:#define route_trivial_usage
//usage: "[{add|del|delete}]"
//usage:#define route_full_usage "\n\n"
//usage: "Edit kernel routing tables\n"
//usage: "\n -n Don't resolve names"
//usage: "\n -e Display other/more information"
//usage: "\n -A inet" IF_FEATURE_IPV6("{6}") " Select address family"
#include <net/route.h>
#include <net/if.h>
#include "libbb.h"
#include "inet_common.h"
#ifndef RTF_UP
/* Keep this in sync with /usr/src/linux/include/linux/route.h */
#define RTF_UP 0x0001 /* route usable */
#define RTF_GATEWAY 0x0002 /* destination is a gateway */
#define RTF_HOST 0x0004 /* host entry (net otherwise) */
#define RTF_REINSTATE 0x0008 /* reinstate route after tmout */
#define RTF_DYNAMIC 0x0010 /* created dyn. (by redirect) */
#define RTF_MODIFIED 0x0020 /* modified dyn. (by redirect) */
#define RTF_MTU 0x0040 /* specific MTU for this route */
#ifndef RTF_MSS
#define RTF_MSS RTF_MTU /* Compatibility :-( */
#endif
#define RTF_WINDOW 0x0080 /* per route window clamping */
#define RTF_IRTT 0x0100 /* Initial round trip time */
#define RTF_REJECT 0x0200 /* Reject route */
#define RTF_NONEXTHOP 0x00200000 /* route with no nexthop */
#endif
#if defined(SIOCADDRTOLD) || defined(RTF_IRTT) /* route */
#define HAVE_NEW_ADDRT 1
#endif
#if HAVE_NEW_ADDRT
#define mask_in_addr(x) (((struct sockaddr_in *)&((x).rt_genmask))->sin_addr.s_addr)
#define full_mask(x) (x)
#else
#define mask_in_addr(x) ((x).rt_genmask)
#define full_mask(x) (((struct sockaddr_in *)&(x))->sin_addr.s_addr)
#endif
/* The RTACTION entries must agree with tbl_verb[] below! */
#define RTACTION_ADD 1
#define RTACTION_DEL 2
/* For the various tbl_*[] arrays, the 1st byte is the offset to
* the next entry and the 2nd byte is return value. */
#define NET_FLAG 1
#define HOST_FLAG 2
/* We remap '-' to '#' to avoid problems with getopt. */
static const char tbl_hash_net_host[] ALIGN1 =
"\007\001#net\0"
/* "\010\002#host\0" */
"\007\002#host" /* Since last, we can save a byte. */
;
#define KW_TAKES_ARG 020
#define KW_SETS_FLAG 040
#define KW_IPVx_METRIC 020
#define KW_IPVx_NETMASK 021
#define KW_IPVx_GATEWAY 022
#define KW_IPVx_MSS 023
#define KW_IPVx_WINDOW 024
#define KW_IPVx_IRTT 025
#define KW_IPVx_DEVICE 026
#define KW_IPVx_FLAG_ONLY 040
#define KW_IPVx_REJECT 040
#define KW_IPVx_MOD 041
#define KW_IPVx_DYN 042
#define KW_IPVx_REINSTATE 043
static const char tbl_ipvx[] ALIGN1 =
/* 020 is the "takes an arg" bit */
#if HAVE_NEW_ADDRT
"\011\020metric\0"
#endif
"\012\021netmask\0"
"\005\022gw\0"
"\012\022gateway\0"
"\006\023mss\0"
"\011\024window\0"
#ifdef RTF_IRTT
"\007\025irtt\0"
#endif
"\006\026dev\0"
"\011\026device\0"
/* 040 is the "sets a flag" bit - MUST match flags_ipvx[] values below. */
#ifdef RTF_REJECT
"\011\040reject\0"
#endif
"\006\041mod\0"
"\006\042dyn\0"
/* "\014\043reinstate\0" */
"\013\043reinstate" /* Since last, we can save a byte. */
;
static const uint16_t flags_ipvx[] = { /* MUST match tbl_ipvx[] values above. */
#ifdef RTF_REJECT
RTF_REJECT,
#endif
RTF_MODIFIED,
RTF_DYNAMIC,
RTF_REINSTATE
};
static int kw_lookup(const char *kwtbl, char ***pargs)
{
if (**pargs) {
do {
if (strcmp(kwtbl+2, **pargs) == 0) { /* Found a match. */
*pargs += 1;
if (kwtbl[1] & KW_TAKES_ARG) {
if (!**pargs) { /* No more args! */
bb_show_usage();
}
*pargs += 1; /* Calling routine will use args[-1]. */
}
return kwtbl[1];
}
kwtbl += *kwtbl;
} while (*kwtbl);
}
return 0;
}
/* Add or delete a route, depending on action. */
static NOINLINE void INET_setroute(int action, char **args)
{
/* char buffer instead of bona-fide struct avoids aliasing warning */
char rt_buf[sizeof(struct rtentry)];
struct rtentry *const rt = (void *)rt_buf;
const char *netmask = NULL;
int skfd, isnet, xflag;
/* Grab the -net or -host options. Remember they were transformed. */
xflag = kw_lookup(tbl_hash_net_host, &args);
/* If we did grab -net or -host, make sure we still have an arg left. */
if (*args == NULL) {
bb_show_usage();
}
/* Clean out the RTREQ structure. */
memset(rt, 0, sizeof(*rt));
{
const char *target = *args++;
char *prefix;
/* recognize x.x.x.x/mask format. */
prefix = strchr(target, '/');
if (prefix) {
int prefix_len;
prefix_len = xatoul_range(prefix+1, 0, 32);
mask_in_addr(*rt) = htonl( ~(0xffffffffUL >> prefix_len));
*prefix = '\0';
#if HAVE_NEW_ADDRT
rt->rt_genmask.sa_family = AF_INET;
#endif
} else {
/* Default netmask. */
netmask = "default";
}
/* Prefer hostname lookup is -host flag (xflag==1) was given. */
isnet = INET_resolve(target, (struct sockaddr_in *) &rt->rt_dst,
(xflag & HOST_FLAG));
if (isnet < 0) {
bb_error_msg_and_die("resolving %s", target);
}
if (prefix) {
/* do not destroy prefix for process args */
*prefix = '/';
}
}
if (xflag) { /* Reinit isnet if -net or -host was specified. */
isnet = (xflag & NET_FLAG);
}
/* Fill in the other fields. */
rt->rt_flags = ((isnet) ? RTF_UP : (RTF_UP | RTF_HOST));
while (*args) {
int k = kw_lookup(tbl_ipvx, &args);
const char *args_m1 = args[-1];
if (k & KW_IPVx_FLAG_ONLY) {
rt->rt_flags |= flags_ipvx[k & 3];
continue;
}
#if HAVE_NEW_ADDRT
if (k == KW_IPVx_METRIC) {
rt->rt_metric = xatoul(args_m1) + 1;
continue;
}
#endif
if (k == KW_IPVx_NETMASK) {
struct sockaddr mask;
if (mask_in_addr(*rt)) {
bb_show_usage();
}
netmask = args_m1;
isnet = INET_resolve(netmask, (struct sockaddr_in *) &mask, 0);
if (isnet < 0) {
bb_error_msg_and_die("resolving %s", netmask);
}
rt->rt_genmask = full_mask(mask);
continue;
}
if (k == KW_IPVx_GATEWAY) {
if (rt->rt_flags & RTF_GATEWAY) {
bb_show_usage();
}
isnet = INET_resolve(args_m1,
(struct sockaddr_in *) &rt->rt_gateway, 1);
rt->rt_flags |= RTF_GATEWAY;
if (isnet) {
if (isnet < 0) {
bb_error_msg_and_die("resolving %s", args_m1);
}
bb_error_msg_and_die("gateway %s is a NETWORK", args_m1);
}
continue;
}
if (k == KW_IPVx_MSS) { /* Check valid MSS bounds. */
rt->rt_flags |= RTF_MSS;
rt->rt_mss = xatoul_range(args_m1, 64, 32768);
continue;
}
if (k == KW_IPVx_WINDOW) { /* Check valid window bounds. */
rt->rt_flags |= RTF_WINDOW;
rt->rt_window = xatoul_range(args_m1, 128, INT_MAX);
continue;
}
#ifdef RTF_IRTT
if (k == KW_IPVx_IRTT) {
rt->rt_flags |= RTF_IRTT;
rt->rt_irtt = xatoul(args_m1);
rt->rt_irtt *= (bb_clk_tck() / 100); /* FIXME */
#if 0 /* FIXME: do we need to check anything of this? */
if (rt->rt_irtt < 1 || rt->rt_irtt > (120 * HZ)) {
bb_error_msg_and_die("bad irtt");
}
#endif
continue;
}
#endif
/* Device is special in that it can be the last arg specified
* and doesn't require the dev/device keyword in that case. */
if (!rt->rt_dev && ((k == KW_IPVx_DEVICE) || (!k && !*++args))) {
/* Don't use args_m1 here since args may have changed! */
rt->rt_dev = args[-1];
continue;
}
/* Nothing matched. */
bb_show_usage();
}
#ifdef RTF_REJECT
if ((rt->rt_flags & RTF_REJECT) && !rt->rt_dev) {
rt->rt_dev = (char*)"lo";
}
#endif
/* sanity checks.. */
if (mask_in_addr(*rt)) {
uint32_t mask = mask_in_addr(*rt);
mask = ~ntohl(mask);
if ((rt->rt_flags & RTF_HOST) && mask != 0xffffffff) {
bb_error_msg_and_die("netmask %.8x and host route conflict",
(unsigned int) mask);
}
if (mask & (mask + 1)) {
bb_error_msg_and_die("bogus netmask %s", netmask);
}
mask = ((struct sockaddr_in *) &rt->rt_dst)->sin_addr.s_addr;
if (mask & ~(uint32_t)mask_in_addr(*rt)) {
bb_simple_error_msg_and_die("netmask and route address conflict");
}
}
/* Fill out netmask if still unset */
if ((action == RTACTION_ADD) && (rt->rt_flags & RTF_HOST)) {
mask_in_addr(*rt) = 0xffffffff;
}
/* Create a socket to the INET kernel. */
skfd = xsocket(AF_INET, SOCK_DGRAM, 0);
if (action == RTACTION_ADD)
xioctl(skfd, SIOCADDRT, rt);
else
xioctl(skfd, SIOCDELRT, rt);
if (ENABLE_FEATURE_CLEAN_UP) close(skfd);
}
#if ENABLE_FEATURE_IPV6
static NOINLINE void INET6_setroute(int action, char **args)
{
struct sockaddr_in6 sa6;
struct in6_rtmsg rt;
int prefix_len, skfd;
const char *devname;
/* We know args isn't NULL from the check in route_main. */
const char *target = *args++;
if (strcmp(target, "default") == 0) {
prefix_len = 0;
memset(&sa6, 0, sizeof(sa6));
} else {
char *cp;
cp = strchr(target, '/'); /* Yes... const to non is ok. */
if (cp) {
*cp = '\0';
prefix_len = xatoul_range(cp + 1, 0, 128);
} else {
prefix_len = 128;
}
if (INET6_resolve(target, (struct sockaddr_in6 *) &sa6) < 0) {
bb_error_msg_and_die("resolving %s", target);
}
}
/* Clean out the RTREQ structure. */
memset(&rt, 0, sizeof(rt));
memcpy(&rt.rtmsg_dst, sa6.sin6_addr.s6_addr, sizeof(struct in6_addr));
/* Fill in the other fields. */
rt.rtmsg_dst_len = prefix_len;
rt.rtmsg_flags = ((prefix_len == 128) ? (RTF_UP|RTF_HOST) : RTF_UP);
rt.rtmsg_metric = 1;
devname = NULL;
while (*args) {
int k = kw_lookup(tbl_ipvx, &args);
const char *args_m1 = args[-1];
if ((k == KW_IPVx_MOD) || (k == KW_IPVx_DYN)) {
rt.rtmsg_flags |= flags_ipvx[k & 3];
continue;
}
if (k == KW_IPVx_METRIC) {
rt.rtmsg_metric = xatoul(args_m1);
continue;
}
if (k == KW_IPVx_GATEWAY) {
if (rt.rtmsg_flags & RTF_GATEWAY) {
bb_show_usage();
}
if (INET6_resolve(args_m1, (struct sockaddr_in6 *) &sa6) < 0) {
bb_error_msg_and_die("resolving %s", args_m1);
}
memcpy(&rt.rtmsg_gateway, sa6.sin6_addr.s6_addr,
sizeof(struct in6_addr));
rt.rtmsg_flags |= RTF_GATEWAY;
continue;
}
/* Device is special in that it can be the last arg specified
* and doesn't require the dev/device keyword in that case. */
if (!devname && ((k == KW_IPVx_DEVICE) || (!k && !*++args))) {
/* Don't use args_m1 here since args may have changed! */
devname = args[-1];
continue;
}
/* Nothing matched. */
bb_show_usage();
}
/* Create a socket to the INET6 kernel. */
skfd = xsocket(AF_INET6, SOCK_DGRAM, 0);
rt.rtmsg_ifindex = 0;
if (devname) {
struct ifreq ifr;
/*memset(&ifr, 0, sizeof(ifr)); - SIOCGIFINDEX does not need to clear all */
strncpy_IFNAMSIZ(ifr.ifr_name, devname);
xioctl(skfd, SIOCGIFINDEX, &ifr);
rt.rtmsg_ifindex = ifr.ifr_ifindex;
}
/* Tell the kernel to accept this route. */
if (action == RTACTION_ADD)
xioctl(skfd, SIOCADDRT, &rt);
else
xioctl(skfd, SIOCDELRT, &rt);
if (ENABLE_FEATURE_CLEAN_UP) close(skfd);
}
#endif
static const
IF_NOT_FEATURE_IPV6(uint16_t)
IF_FEATURE_IPV6(unsigned)
flagvals[] = { /* Must agree with flagchars[]. */
RTF_UP,
RTF_GATEWAY,
RTF_HOST,
RTF_REINSTATE,
RTF_DYNAMIC,
RTF_MODIFIED,
#if ENABLE_FEATURE_IPV6
RTF_DEFAULT,
RTF_ADDRCONF,
RTF_CACHE,
RTF_REJECT,
RTF_NONEXTHOP, /* this one doesn't fit into 16 bits */
#endif
};
/* Must agree with flagvals[]. */
static const char flagchars[] ALIGN1 =
"UGHRDM"
#if ENABLE_FEATURE_IPV6
"DAC!n"
#endif
;
#define IPV4_MASK (RTF_UP|RTF_GATEWAY|RTF_HOST|RTF_REINSTATE|RTF_DYNAMIC|RTF_MODIFIED)
#define IPV6_MASK (RTF_UP|RTF_GATEWAY|RTF_HOST|RTF_DEFAULT|RTF_ADDRCONF|RTF_CACHE|RTF_REJECT|RTF_NONEXTHOP)
static void set_flags(char *flagstr, int flags)
{
int i;
for (i = 0; (*flagstr = flagchars[i]) != 0; i++) {
if (flags & flagvals[i]) {
++flagstr;
}
}
}
/* also used in netstat */
void FAST_FUNC bb_displayroutes(int noresolve, int netstatfmt)
{
char devname[64], flags[16], *sdest, *sgw;
unsigned long d, g, m;
int r;
int flgs, ref, use, metric, mtu, win, ir;
struct sockaddr_in s_addr;
struct in_addr mask;
FILE *fp = xfopen_for_read("/proc/net/route");
printf("Kernel IP routing table\n"
"Destination Gateway Genmask Flags %s Iface\n",
netstatfmt ? " MSS Window irtt" : "Metric Ref Use");
/* Skip the first line. */
r = fscanf(fp, "%*[^\n]\n");
if (r < 0) {
/* Empty line, read error, or EOF. Yes, if routing table
* is completely empty, /proc/net/route has no header.
*/
goto ERROR;
}
while (1) {
r = fscanf(fp, "%63s%lx%lx%X%d%d%d%lx%d%d%d\n",
devname, &d, &g, &flgs, &ref, &use, &metric, &m,
&mtu, &win, &ir);
if (r != 11) {
ERROR:
if ((r < 0) && feof(fp)) { /* EOF with no (nonspace) chars read. */
break;
}
bb_simple_perror_msg_and_die(bb_msg_read_error);
}
if (!(flgs & RTF_UP)) { /* Skip interfaces that are down. */
continue;
}
set_flags(flags, (flgs & IPV4_MASK));
#ifdef RTF_REJECT
if (flgs & RTF_REJECT) {
flags[0] = '!';
}
#endif
memset(&s_addr, 0, sizeof(struct sockaddr_in));
s_addr.sin_family = AF_INET;
s_addr.sin_addr.s_addr = d;
sdest = INET_rresolve(&s_addr, (noresolve | 0x8000), m); /* 'default' instead of '*' */
s_addr.sin_addr.s_addr = g;
sgw = INET_rresolve(&s_addr, (noresolve | 0x4000), m); /* Host instead of net */
mask.s_addr = m;
/* "%15.15s" truncates hostnames, do we really want that? */
printf("%-15.15s %-15.15s %-16s%-6s", sdest, sgw, inet_ntoa(mask), flags);
free(sdest);
free(sgw);
if (netstatfmt) {
printf("%5d %-5d %6d %s\n", mtu, win, ir, devname);
} else {
printf("%-6d %-2d %7d %s\n", metric, ref, use, devname);
}
}
fclose(fp);
}
#if ENABLE_FEATURE_IPV6
static void INET6_displayroutes(void)
{
char addr6[128], *naddr6;
/* In addr6x, we store both 40-byte ':'-delimited ipv6 addresses.
* We read the non-delimited strings into the tail of the buffer
* using fscanf and then modify the buffer by shifting forward
* while inserting ':'s and the nul terminator for the first string.
* Hence the strings are at addr6x and addr6x+40. This generates
* _much_ less code than the previous (upstream) approach. */
char addr6x[80];
char iface[16], flags[16];
int iflags, metric, refcnt, use, prefix_len, slen;
struct sockaddr_in6 snaddr6;
FILE *fp = xfopen_for_read("/proc/net/ipv6_route");
printf("Kernel IPv6 routing table\n%-44s%-40s"
"Flags Metric Ref Use Iface\n",
"Destination", "Next Hop");
while (1) {
int r;
r = fscanf(fp, "%32s%x%*s%x%32s%x%x%x%x%s\n",
addr6x+14, &prefix_len, &slen, addr6x+40+7,
&metric, &refcnt, &use, &iflags, iface);
if (r != 9) {
if ((r < 0) && feof(fp)) { /* EOF with no (nonspace) chars read. */
break;
}
ERROR:
bb_simple_perror_msg_and_die(bb_msg_read_error);
}
/* Do the addr6x shift-and-insert changes to ':'-delimit addresses.
* For now, always do this to validate the proc route format, even
* if the interface is down. */
{
int i = 0;
char *p = addr6x+14;
do {
if (!*p) {
if (i == 40) { /* nul terminator for 1st address? */
addr6x[39] = 0; /* Fixup... need 0 instead of ':'. */
++p; /* Skip and continue. */
continue;
}
goto ERROR;
}
addr6x[i++] = *p++;
if (!((i+1) % 5)) {
addr6x[i++] = ':';
}
} while (i < 40+28+7);
}
set_flags(flags, (iflags & IPV6_MASK));
r = 0;
while (1) {
memset(&snaddr6, 0, sizeof(snaddr6));
inet_pton(AF_INET6, addr6x + r,
(struct sockaddr *) &snaddr6.sin6_addr);
snaddr6.sin6_family = AF_INET6;
naddr6 = INET6_rresolve((struct sockaddr_in6 *) &snaddr6,
0x0fff /* Apparently, upstream never resolves. */
);
if (!r) { /* 1st pass */
snprintf(addr6, sizeof(addr6), "%s/%d", naddr6, prefix_len);
r += 40;
free(naddr6);
} else { /* 2nd pass */
/* Print the info. */
printf("%-43s %-39s %-5s %-6d %-2d %7d %-8s\n",
addr6, naddr6, flags, metric, refcnt, use, iface);
free(naddr6);
break;
}
}
}
fclose(fp);
}
#endif
#define ROUTE_OPT_A 0x01
#define ROUTE_OPT_n 0x02
#define ROUTE_OPT_e 0x04
#define ROUTE_OPT_INET6 0x08 /* Not an actual option. See below. */
/* 1st byte is offset to next entry offset. 2nd byte is return value. */
/* 2nd byte matches RTACTION_* code */
static const char tbl_verb[] ALIGN1 =
"\006\001add\0"
"\006\002del\0"
/* "\011\002delete\0" */
"\010\002delete" /* Since it's last, we can save a byte. */
;
int route_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
int route_main(int argc UNUSED_PARAM, char **argv)
{
unsigned opt;
int what;
char *family;
char **p;
/* First, remap '-net' and '-host' to avoid getopt problems. */
p = argv;
while (*++p) {
if (strcmp(*p, "-net") == 0 || strcmp(*p, "-host") == 0) {
p[0][0] = '#';
}
}
opt = getopt32(argv, "A:ne", &family);
if ((opt & ROUTE_OPT_A) && strcmp(family, "inet") != 0) {
#if ENABLE_FEATURE_IPV6
if (strcmp(family, "inet6") == 0) {
opt |= ROUTE_OPT_INET6; /* Set flag for ipv6. */
} else
#endif
bb_show_usage();
}
argv += optind;
/* No more args means display the routing table. */
if (!*argv) {
int noresolve = (opt & ROUTE_OPT_n) ? 0x0fff : 0;
#if ENABLE_FEATURE_IPV6
if (opt & ROUTE_OPT_INET6)
INET6_displayroutes();
else
#endif
bb_displayroutes(noresolve, opt & ROUTE_OPT_e);
fflush_stdout_and_exit(EXIT_SUCCESS);
}
/* Check verb. At the moment, must be add, del, or delete. */
what = kw_lookup(tbl_verb, &argv);
if (!what || !*argv) { /* Unknown verb or no more args. */
bb_show_usage();
}
#if ENABLE_FEATURE_IPV6
if (opt & ROUTE_OPT_INET6)
INET6_setroute(what, argv);
else
#endif
INET_setroute(what, argv);
return EXIT_SUCCESS;
}