mirror of
https://github.com/rsmarples/dhcpcd.git
synced 2024-11-28 12:33:49 +08:00
738 lines
18 KiB
C
738 lines
18 KiB
C
/*
|
|
* dhcpcd - DHCP client daemon
|
|
* Copyright (c) 2006-2013 Roy Marples <roy@marples.name>
|
|
* 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.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
|
|
*/
|
|
|
|
#include <asm/types.h> /* Needed for 2.4 kernels */
|
|
|
|
#include <sys/types.h>
|
|
#include <sys/socket.h>
|
|
#include <sys/ioctl.h>
|
|
#include <sys/param.h>
|
|
|
|
#include <linux/netlink.h>
|
|
#include <linux/rtnetlink.h>
|
|
|
|
/* Support older kernels */
|
|
#ifndef IFLA_WIRELESS
|
|
# define IFLA_WIRELESS (IFLA_MASTER + 1)
|
|
#endif
|
|
|
|
/* For some reason, glibc doesn't include newer flags from linux/if.h
|
|
* However, we cannot include linux/if.h directly as it conflicts
|
|
* with the glibc version. D'oh! */
|
|
#ifndef IFF_LOWER_UP
|
|
#define IFF_LOWER_UP 0x10000 /* driver signals L1 up */
|
|
#endif
|
|
|
|
#include <errno.h>
|
|
#include <ctype.h>
|
|
#include <stddef.h>
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <unistd.h>
|
|
|
|
#include "config.h"
|
|
#include "common.h"
|
|
#include "dhcp.h"
|
|
#include "ipv4.h"
|
|
#include "ipv6.h"
|
|
#include "net.h"
|
|
|
|
static int sock_fd;
|
|
static struct sockaddr_nl sock_nl;
|
|
|
|
int
|
|
if_init(struct interface *iface)
|
|
{
|
|
char path[PATH_MAX];
|
|
FILE *fp;
|
|
int n;
|
|
|
|
/* We enable promote_secondaries so that we can do this
|
|
* add 192.168.1.2/24
|
|
* add 192.168.1.3/24
|
|
* del 192.168.1.2/24
|
|
* and the subnet mask moves onto 192.168.1.3/24
|
|
* This matches the behaviour of BSD which makes coding dhcpcd
|
|
* a little easier as there's just one behaviour. */
|
|
snprintf(path, sizeof(path),
|
|
"/proc/sys/net/ipv4/conf/%s/promote_secondaries",
|
|
iface->name);
|
|
|
|
fp = fopen(path, "w");
|
|
if (fp == NULL)
|
|
return errno == ENOENT ? 0 : -1;
|
|
n = fprintf(fp, "1");
|
|
fclose(fp);
|
|
return n == -1 ? -1 : 0;
|
|
}
|
|
|
|
int
|
|
if_conf(struct interface *iface)
|
|
{
|
|
char path[PATH_MAX], buf[1];
|
|
FILE *fp;
|
|
|
|
/* Some qeth setups require the use of the broadcast flag. */
|
|
snprintf(path, sizeof(path),
|
|
"/sys/class/net/%s/device/layer2",
|
|
iface->name);
|
|
|
|
fp = fopen(path, "r");
|
|
if (fp == NULL)
|
|
return errno == ENOENT ? 0 : -1;
|
|
if (fgets(buf, sizeof(buf), fp) != NULL && buf[0] == '0')
|
|
iface->options->options |= DHCPCD_BROADCAST;
|
|
fclose(fp);
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
_open_link_socket(struct sockaddr_nl *nl)
|
|
{
|
|
int fd;
|
|
|
|
if ((fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE)) == -1)
|
|
return -1;
|
|
nl->nl_family = AF_NETLINK;
|
|
if (bind(fd, (struct sockaddr *)nl, sizeof(*nl)) == -1)
|
|
return -1;
|
|
set_cloexec(fd);
|
|
return fd;
|
|
}
|
|
|
|
int
|
|
open_sockets(void)
|
|
{
|
|
if ((socket_afnet = socket(AF_INET, SOCK_DGRAM, 0)) == -1)
|
|
return -1;
|
|
set_cloexec(socket_afnet);
|
|
sock_fd = _open_link_socket(&sock_nl);
|
|
set_cloexec(sock_fd);
|
|
return sock_fd;
|
|
}
|
|
|
|
int
|
|
open_link_socket(void)
|
|
{
|
|
struct sockaddr_nl snl;
|
|
|
|
memset(&snl, 0, sizeof(snl));
|
|
snl.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_ROUTE | RTMGRP_IPV4_IFADDR;
|
|
return _open_link_socket(&snl);
|
|
}
|
|
|
|
static int
|
|
get_netlink(int fd, int flags,
|
|
int (*callback)(struct nlmsghdr *))
|
|
{
|
|
char *buf = NULL, *nbuf;
|
|
ssize_t buflen = 0, bytes;
|
|
struct nlmsghdr *nlm;
|
|
struct sockaddr_nl nladdr;
|
|
socklen_t nladdr_len = sizeof(nladdr);
|
|
int r = -1;
|
|
|
|
for (;;) {
|
|
bytes = recv(fd, NULL, 0,
|
|
flags | MSG_PEEK | MSG_DONTWAIT | MSG_TRUNC);
|
|
if (bytes == -1) {
|
|
if (errno == EAGAIN) {
|
|
r = 0;
|
|
goto eexit;
|
|
}
|
|
if (errno == EINTR)
|
|
continue;
|
|
goto eexit;
|
|
} else if (bytes == buflen) {
|
|
/* Support kernels older than 2.6.22 */
|
|
if (bytes == 0)
|
|
bytes = 512;
|
|
else
|
|
bytes *= 2;
|
|
}
|
|
if (buflen < bytes) {
|
|
/* Alloc 1 more so we work with older kernels */
|
|
buflen = bytes + 1;
|
|
nbuf = realloc(buf, buflen);
|
|
if (nbuf == NULL)
|
|
goto eexit;
|
|
buf = nbuf;
|
|
}
|
|
bytes = recvfrom(fd, buf, buflen, flags,
|
|
(struct sockaddr *)&nladdr, &nladdr_len);
|
|
if (bytes == -1) {
|
|
if (errno == EAGAIN) {
|
|
r = 0;
|
|
goto eexit;
|
|
}
|
|
if (errno == EINTR)
|
|
continue;
|
|
goto eexit;
|
|
}
|
|
|
|
/* Check sender */
|
|
if (nladdr_len != sizeof(nladdr)) {
|
|
errno = EINVAL;
|
|
goto eexit;
|
|
}
|
|
/* Ignore message if it is not from kernel */
|
|
if (nladdr.nl_pid != 0)
|
|
continue;
|
|
|
|
for (nlm = (struct nlmsghdr *)(void *)buf;
|
|
NLMSG_OK(nlm, (size_t)bytes);
|
|
nlm = NLMSG_NEXT(nlm, bytes))
|
|
{
|
|
r = callback(nlm);
|
|
if (r != 0)
|
|
goto eexit;
|
|
}
|
|
}
|
|
|
|
eexit:
|
|
free(buf);
|
|
return r;
|
|
}
|
|
|
|
static int
|
|
err_netlink(struct nlmsghdr *nlm)
|
|
{
|
|
struct nlmsgerr *err;
|
|
int l;
|
|
|
|
if (nlm->nlmsg_type != NLMSG_ERROR)
|
|
return 0;
|
|
l = nlm->nlmsg_len - sizeof(*nlm);
|
|
if ((size_t)l < sizeof(*err)) {
|
|
errno = EBADMSG;
|
|
return -1;
|
|
}
|
|
err = (struct nlmsgerr *)NLMSG_DATA(nlm);
|
|
if (err->error == 0)
|
|
return l;
|
|
errno = -err->error;
|
|
return -1;
|
|
}
|
|
|
|
static int
|
|
link_route(struct nlmsghdr *nlm)
|
|
{
|
|
int len, idx, metric;
|
|
struct rtattr *rta;
|
|
struct rtmsg *rtm;
|
|
struct rt rt;
|
|
char ifn[IF_NAMESIZE + 1];
|
|
|
|
if (nlm->nlmsg_type != RTM_DELROUTE)
|
|
return 0;
|
|
|
|
len = nlm->nlmsg_len - sizeof(*nlm);
|
|
if ((size_t)len < sizeof(*rtm)) {
|
|
errno = EBADMSG;
|
|
return -1;
|
|
}
|
|
rtm = NLMSG_DATA(nlm);
|
|
if (rtm->rtm_type != RTN_UNICAST ||
|
|
rtm->rtm_table != RT_TABLE_MAIN ||
|
|
rtm->rtm_family != AF_INET ||
|
|
nlm->nlmsg_pid == (uint32_t)getpid())
|
|
return 1;
|
|
rta = (struct rtattr *)(void *)((char *)rtm +NLMSG_ALIGN(sizeof(*rtm)));
|
|
len = NLMSG_PAYLOAD(nlm, sizeof(*rtm));
|
|
memset(&rt, 0, sizeof(rt));
|
|
rt.dest.s_addr = INADDR_ANY;
|
|
rt.net.s_addr = INADDR_ANY;
|
|
rt.gate.s_addr = INADDR_ANY;
|
|
metric = 0;
|
|
while (RTA_OK(rta, len)) {
|
|
switch (rta->rta_type) {
|
|
case RTA_DST:
|
|
memcpy(&rt.dest.s_addr, RTA_DATA(rta),
|
|
sizeof(rt.dest.s_addr));
|
|
break;
|
|
case RTA_GATEWAY:
|
|
memcpy(&rt.gate.s_addr, RTA_DATA(rta),
|
|
sizeof(rt.gate.s_addr));
|
|
break;
|
|
case RTA_OIF:
|
|
idx = *(int *)RTA_DATA(rta);
|
|
if (if_indextoname(idx, ifn))
|
|
rt.iface = find_interface(ifn);
|
|
break;
|
|
case RTA_PRIORITY:
|
|
metric = *(int *)RTA_DATA(rta);
|
|
break;
|
|
}
|
|
rta = RTA_NEXT(rta, len);
|
|
}
|
|
if (rt.iface != NULL) {
|
|
if (metric == rt.iface->metric) {
|
|
#ifdef INET
|
|
inet_cidrtoaddr(rtm->rtm_dst_len, &rt.net);
|
|
ipv4_routedeleted(&rt);
|
|
#endif
|
|
}
|
|
}
|
|
return 1;
|
|
}
|
|
|
|
static int
|
|
link_addr(struct nlmsghdr *nlm)
|
|
{
|
|
#ifdef INET
|
|
int len;
|
|
struct rtattr *rta;
|
|
struct ifaddrmsg *ifa;
|
|
struct in_addr addr, net, dest;
|
|
char ifn[IF_NAMESIZE + 1];
|
|
struct interface *iface;
|
|
|
|
if (nlm->nlmsg_type != RTM_DELADDR && nlm->nlmsg_type != RTM_NEWADDR)
|
|
return 0;
|
|
|
|
len = nlm->nlmsg_len - sizeof(*nlm);
|
|
if ((size_t)len < sizeof(*ifa)) {
|
|
errno = EBADMSG;
|
|
return -1;
|
|
}
|
|
if (nlm->nlmsg_pid == (uint32_t)getpid())
|
|
return 1;
|
|
ifa = NLMSG_DATA(nlm);
|
|
if (if_indextoname(ifa->ifa_index, ifn) == NULL)
|
|
return -1;
|
|
iface = find_interface(ifn);
|
|
if (iface == NULL)
|
|
return 1;
|
|
rta = (struct rtattr *) IFA_RTA(ifa);
|
|
len = NLMSG_PAYLOAD(nlm, sizeof(*ifa));
|
|
addr.s_addr = dest.s_addr = INADDR_ANY;
|
|
dest.s_addr = INADDR_ANY;
|
|
inet_cidrtoaddr(ifa->ifa_prefixlen, &net);
|
|
while (RTA_OK(rta, len)) {
|
|
switch (rta->rta_type) {
|
|
case IFA_ADDRESS:
|
|
if (iface->flags & IFF_POINTOPOINT) {
|
|
memcpy(&dest.s_addr, RTA_DATA(rta),
|
|
sizeof(addr.s_addr));
|
|
}
|
|
break;
|
|
case IFA_LOCAL:
|
|
memcpy(&addr.s_addr, RTA_DATA(rta),
|
|
sizeof(addr.s_addr));
|
|
break;
|
|
}
|
|
rta = RTA_NEXT(rta, len);
|
|
}
|
|
ipv4_handleifa(nlm->nlmsg_type, ifn, &addr, &net, &dest);
|
|
#endif
|
|
return 1;
|
|
}
|
|
|
|
static int
|
|
link_netlink(struct nlmsghdr *nlm)
|
|
{
|
|
int len;
|
|
struct rtattr *rta;
|
|
struct ifinfomsg *ifi;
|
|
char ifn[IF_NAMESIZE + 1];
|
|
|
|
len = link_route(nlm);
|
|
if (len != 0)
|
|
return len;
|
|
len = link_addr(nlm);
|
|
if (len != 0)
|
|
return len;
|
|
|
|
if (nlm->nlmsg_type != RTM_NEWLINK && nlm->nlmsg_type != RTM_DELLINK)
|
|
return 0;
|
|
len = nlm->nlmsg_len - sizeof(*nlm);
|
|
if ((size_t)len < sizeof(*ifi)) {
|
|
errno = EBADMSG;
|
|
return -1;
|
|
}
|
|
ifi = NLMSG_DATA(nlm);
|
|
if (ifi->ifi_flags & IFF_LOOPBACK)
|
|
return 1;
|
|
rta = (struct rtattr *)(void *)((char *)ifi +NLMSG_ALIGN(sizeof(*ifi)));
|
|
len = NLMSG_PAYLOAD(nlm, sizeof(*ifi));
|
|
*ifn = '\0';
|
|
while (RTA_OK(rta, len)) {
|
|
switch (rta->rta_type) {
|
|
case IFLA_WIRELESS:
|
|
/* Ignore wireless messages */
|
|
if (nlm->nlmsg_type == RTM_NEWLINK &&
|
|
ifi->ifi_change == 0)
|
|
return 1;
|
|
break;
|
|
case IFLA_IFNAME:
|
|
strlcpy(ifn, RTA_DATA(rta), sizeof(ifn));
|
|
break;
|
|
}
|
|
rta = RTA_NEXT(rta, len);
|
|
}
|
|
|
|
if (nlm->nlmsg_type == RTM_DELLINK) {
|
|
handle_interface(-1, ifn);
|
|
return 1;
|
|
}
|
|
|
|
/* Bridge interfaces set IFF_LOWER_UP when they have a valid
|
|
* hardware address. To trigger a valid hardware address pickup
|
|
* we need to pretend that that don't exist until they have
|
|
* IFF_LOWER_UP set. */
|
|
if (ifi->ifi_flags & IFF_MASTER && !(ifi->ifi_flags & IFF_LOWER_UP)) {
|
|
handle_interface(-1, ifn);
|
|
return 1;
|
|
}
|
|
|
|
handle_carrier(ifi->ifi_flags & IFF_RUNNING ? 1 : -1,
|
|
ifi->ifi_flags, ifn);
|
|
return 1;
|
|
}
|
|
|
|
int
|
|
manage_link(int fd)
|
|
{
|
|
return get_netlink(fd, MSG_DONTWAIT, &link_netlink);
|
|
}
|
|
|
|
static int
|
|
send_netlink(struct nlmsghdr *hdr)
|
|
{
|
|
int r;
|
|
struct iovec iov;
|
|
struct msghdr msg;
|
|
static unsigned int seq;
|
|
|
|
memset(&iov, 0, sizeof(iov));
|
|
iov.iov_base = hdr;
|
|
iov.iov_len = hdr->nlmsg_len;
|
|
memset(&msg, 0, sizeof(msg));
|
|
msg.msg_name = &sock_nl;
|
|
msg.msg_namelen = sizeof(sock_nl);
|
|
msg.msg_iov = &iov;
|
|
msg.msg_iovlen = 1;
|
|
/* Request a reply */
|
|
hdr->nlmsg_flags |= NLM_F_ACK;
|
|
hdr->nlmsg_seq = ++seq;
|
|
|
|
if (sendmsg(sock_fd, &msg, 0) != -1)
|
|
r = get_netlink(sock_fd, 0, &err_netlink);
|
|
else
|
|
r = -1;
|
|
return r;
|
|
}
|
|
|
|
#define NLMSG_TAIL(nmsg) \
|
|
((struct rtattr *)(((ptrdiff_t)(nmsg))+NLMSG_ALIGN((nmsg)->nlmsg_len)))
|
|
|
|
static int
|
|
add_attr_l(struct nlmsghdr *n, unsigned int maxlen, int type,
|
|
const void *data, int alen)
|
|
{
|
|
int len = RTA_LENGTH(alen);
|
|
struct rtattr *rta;
|
|
|
|
if (NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len) > maxlen) {
|
|
errno = ENOBUFS;
|
|
return -1;
|
|
}
|
|
|
|
rta = NLMSG_TAIL(n);
|
|
rta->rta_type = type;
|
|
rta->rta_len = len;
|
|
memcpy(RTA_DATA(rta), data, alen);
|
|
n->nlmsg_len = NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
add_attr_32(struct nlmsghdr *n, unsigned int maxlen, int type, uint32_t data)
|
|
{
|
|
int len = RTA_LENGTH(sizeof(data));
|
|
struct rtattr *rta;
|
|
|
|
if (NLMSG_ALIGN(n->nlmsg_len) + len > maxlen) {
|
|
errno = ENOBUFS;
|
|
return -1;
|
|
}
|
|
|
|
rta = NLMSG_TAIL(n);
|
|
rta->rta_type = type;
|
|
rta->rta_len = len;
|
|
memcpy(RTA_DATA(rta), &data, sizeof(data));
|
|
n->nlmsg_len = NLMSG_ALIGN(n->nlmsg_len) + len;
|
|
|
|
return 0;
|
|
}
|
|
|
|
struct nlma
|
|
{
|
|
struct nlmsghdr hdr;
|
|
struct ifaddrmsg ifa;
|
|
char buffer[64];
|
|
};
|
|
|
|
struct nlmr
|
|
{
|
|
struct nlmsghdr hdr;
|
|
struct rtmsg rt;
|
|
char buffer[256];
|
|
};
|
|
|
|
#ifdef INET
|
|
int
|
|
if_address(const struct interface *iface,
|
|
const struct in_addr *address, const struct in_addr *netmask,
|
|
const struct in_addr *broadcast, int action)
|
|
{
|
|
struct nlma *nlm;
|
|
int retval = 0;
|
|
|
|
nlm = calloc(1, sizeof(*nlm));
|
|
if (nlm == NULL)
|
|
return -1;
|
|
nlm->hdr.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg));
|
|
nlm->hdr.nlmsg_flags = NLM_F_REQUEST;
|
|
if (action >= 0) {
|
|
nlm->hdr.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
|
|
nlm->hdr.nlmsg_type = RTM_NEWADDR;
|
|
} else
|
|
nlm->hdr.nlmsg_type = RTM_DELADDR;
|
|
nlm->ifa.ifa_index = iface->index;
|
|
nlm->ifa.ifa_family = AF_INET;
|
|
nlm->ifa.ifa_prefixlen = inet_ntocidr(*netmask);
|
|
/* This creates the aliased interface */
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), IFA_LABEL,
|
|
iface->name, strlen(iface->name) + 1);
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), IFA_LOCAL,
|
|
&address->s_addr, sizeof(address->s_addr));
|
|
if (action >= 0 && broadcast)
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), IFA_BROADCAST,
|
|
&broadcast->s_addr, sizeof(broadcast->s_addr));
|
|
|
|
if (send_netlink(&nlm->hdr) == -1)
|
|
retval = -1;
|
|
free(nlm);
|
|
return retval;
|
|
}
|
|
|
|
int
|
|
if_route(const struct rt *rt, int action)
|
|
{
|
|
struct nlmr *nlm;
|
|
int retval = 0;
|
|
struct dhcp_state *state;
|
|
|
|
nlm = calloc(1, sizeof(*nlm));
|
|
if (nlm == NULL)
|
|
return -1;
|
|
nlm->hdr.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
|
|
nlm->hdr.nlmsg_type = RTM_NEWROUTE;
|
|
if (action == 0)
|
|
nlm->hdr.nlmsg_flags = NLM_F_REPLACE;
|
|
else if (action == 1)
|
|
nlm->hdr.nlmsg_flags = NLM_F_CREATE | NLM_F_EXCL;
|
|
else
|
|
nlm->hdr.nlmsg_type = RTM_DELROUTE;
|
|
nlm->hdr.nlmsg_flags |= NLM_F_REQUEST;
|
|
nlm->rt.rtm_family = AF_INET;
|
|
nlm->rt.rtm_table = RT_TABLE_MAIN;
|
|
|
|
state = D_STATE(rt->iface);
|
|
if (action == -1 || action == -2)
|
|
nlm->rt.rtm_scope = RT_SCOPE_NOWHERE;
|
|
else {
|
|
nlm->hdr.nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
|
|
/* We only change route metrics for kernel routes */
|
|
if (rt->dest.s_addr ==
|
|
(state->addr.s_addr & state->net.s_addr) &&
|
|
rt->net.s_addr == state->net.s_addr)
|
|
nlm->rt.rtm_protocol = RTPROT_KERNEL;
|
|
else
|
|
nlm->rt.rtm_protocol = RTPROT_BOOT;
|
|
if (rt->gate.s_addr == INADDR_ANY ||
|
|
(rt->gate.s_addr == rt->dest.s_addr &&
|
|
rt->net.s_addr == INADDR_BROADCAST))
|
|
nlm->rt.rtm_scope = RT_SCOPE_LINK;
|
|
else
|
|
nlm->rt.rtm_scope = RT_SCOPE_UNIVERSE;
|
|
nlm->rt.rtm_type = RTN_UNICAST;
|
|
}
|
|
|
|
nlm->rt.rtm_dst_len = inet_ntocidr(rt->net);
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), RTA_DST,
|
|
&rt->dest.s_addr, sizeof(rt->dest.s_addr));
|
|
if (nlm->rt.rtm_protocol == RTPROT_KERNEL) {
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), RTA_PREFSRC,
|
|
&state->addr.s_addr, sizeof(state->addr.s_addr));
|
|
}
|
|
/* If destination == gateway then don't add the gateway */
|
|
if (rt->dest.s_addr != rt->gate.s_addr ||
|
|
rt->net.s_addr != INADDR_BROADCAST)
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), RTA_GATEWAY,
|
|
&rt->gate.s_addr, sizeof(rt->gate.s_addr));
|
|
|
|
add_attr_32(&nlm->hdr, sizeof(*nlm), RTA_OIF, rt->iface->index);
|
|
add_attr_32(&nlm->hdr, sizeof(*nlm), RTA_PRIORITY, rt->metric);
|
|
|
|
if (send_netlink(&nlm->hdr) == -1)
|
|
retval = -1;
|
|
free(nlm);
|
|
return retval;
|
|
}
|
|
#endif
|
|
|
|
#ifdef INET6
|
|
int
|
|
if_address6(const struct interface *ifp, const struct ipv6_addr *ap, int action)
|
|
{
|
|
struct nlma *nlm;
|
|
struct ifa_cacheinfo cinfo;
|
|
int retval = 0;
|
|
|
|
nlm = calloc(1, sizeof(*nlm));
|
|
if (nlm == NULL)
|
|
return -1;
|
|
nlm->hdr.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg));
|
|
nlm->hdr.nlmsg_flags = NLM_F_REQUEST;
|
|
if (action >= 0) {
|
|
nlm->hdr.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
|
|
nlm->hdr.nlmsg_type = RTM_NEWADDR;
|
|
} else
|
|
nlm->hdr.nlmsg_type = RTM_DELADDR;
|
|
nlm->ifa.ifa_index = ifp->index;
|
|
nlm->ifa.ifa_family = AF_INET6;
|
|
nlm->ifa.ifa_prefixlen = ap->prefix_len;
|
|
/* This creates the aliased interface */
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), IFA_LABEL,
|
|
ifp->name, strlen(ifp->name) + 1);
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), IFA_LOCAL,
|
|
&ap->addr.s6_addr, sizeof(ap->addr.s6_addr));
|
|
|
|
if (action >= 0) {
|
|
memset(&cinfo, 0, sizeof(cinfo));
|
|
cinfo.ifa_prefered = ap->prefix_pltime;
|
|
cinfo.ifa_valid = ap->prefix_vltime;
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), IFA_CACHEINFO,
|
|
&cinfo, sizeof(cinfo));
|
|
}
|
|
|
|
if (send_netlink(&nlm->hdr) == -1)
|
|
retval = -1;
|
|
free(nlm);
|
|
return retval;
|
|
}
|
|
|
|
static int
|
|
rta_add_attr_32(struct rtattr *rta, unsigned int maxlen,
|
|
int type, uint32_t data)
|
|
{
|
|
unsigned int len = RTA_LENGTH(sizeof(data));
|
|
struct rtattr *subrta;
|
|
|
|
if (RTA_ALIGN(rta->rta_len) + len > maxlen) {
|
|
errno = ENOBUFS;
|
|
return -1;
|
|
}
|
|
|
|
subrta = (struct rtattr*)(((char*)rta) + RTA_ALIGN(rta->rta_len));
|
|
subrta->rta_type = type;
|
|
subrta->rta_len = len;
|
|
memcpy(RTA_DATA(subrta), &data, sizeof(data));
|
|
rta->rta_len = NLMSG_ALIGN(rta->rta_len) + len;
|
|
return 0;
|
|
}
|
|
|
|
int
|
|
if_route6(const struct rt6 *rt, int action)
|
|
{
|
|
struct nlmr *nlm;
|
|
char metricsbuf[32];
|
|
struct rtattr *metrics = (void *)metricsbuf;
|
|
int retval = 0;
|
|
|
|
nlm = calloc(1, sizeof(*nlm));
|
|
if (nlm == NULL)
|
|
return -1;
|
|
nlm->hdr.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
|
|
nlm->hdr.nlmsg_type = RTM_NEWROUTE;
|
|
nlm->hdr.nlmsg_flags = NLM_F_REQUEST;
|
|
if (action == 0)
|
|
nlm->hdr.nlmsg_flags |= NLM_F_REPLACE;
|
|
else if (action == 1)
|
|
nlm->hdr.nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
|
|
else
|
|
nlm->hdr.nlmsg_type = RTM_DELROUTE;
|
|
nlm->rt.rtm_family = AF_INET6;
|
|
nlm->rt.rtm_table = RT_TABLE_MAIN;
|
|
|
|
if (action == -1 || action == -2)
|
|
nlm->rt.rtm_scope = RT_SCOPE_NOWHERE;
|
|
else {
|
|
nlm->hdr.nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
|
|
/* None interface subnet routes are static. */
|
|
if (IN6_IS_ADDR_UNSPECIFIED(&rt->gate)) {
|
|
nlm->rt.rtm_protocol = RTPROT_KERNEL;
|
|
nlm->rt.rtm_scope = RT_SCOPE_LINK;
|
|
} else
|
|
nlm->rt.rtm_protocol = RTPROT_BOOT;
|
|
nlm->rt.rtm_type = RTN_UNICAST;
|
|
}
|
|
|
|
nlm->rt.rtm_dst_len = ipv6_prefixlen(&rt->net);
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), RTA_DST,
|
|
&rt->dest.s6_addr, sizeof(rt->dest.s6_addr));
|
|
|
|
/* If destination == gateway then don't add the gateway */
|
|
if (!IN6_IS_ADDR_UNSPECIFIED(&rt->gate) &&
|
|
!IN6_ARE_ADDR_EQUAL(&rt->dest, &rt->gate))
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), RTA_GATEWAY,
|
|
&rt->gate.s6_addr, sizeof(rt->gate.s6_addr));
|
|
|
|
add_attr_32(&nlm->hdr, sizeof(*nlm), RTA_OIF, rt->iface->index);
|
|
add_attr_32(&nlm->hdr, sizeof(*nlm), RTA_PRIORITY, rt->metric);
|
|
|
|
if (rt->mtu) {
|
|
metrics->rta_type = RTA_METRICS;
|
|
metrics->rta_len = RTA_LENGTH(0);
|
|
rta_add_attr_32(metrics, sizeof(metricsbuf), RTAX_MTU, rt->mtu);
|
|
add_attr_l(&nlm->hdr, sizeof(*nlm), RTA_METRICS,
|
|
RTA_DATA(metrics), RTA_PAYLOAD(metrics));
|
|
}
|
|
|
|
if (send_netlink(&nlm->hdr) == -1)
|
|
retval = -1;
|
|
free(nlm);
|
|
return retval;
|
|
errno = ENOTSUP;
|
|
return -1;
|
|
}
|
|
#endif
|