int
netlink_link_down(vrrp_rt *vrrp)
{
int status = 1;
struct {
struct nlmsghdr n;
struct ifinfomsg ifi;
char buf[256];
} req;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_NEWLINK;
req.ifi.ifi_family = AF_UNSPEC;
req.ifi.ifi_index = IF_INDEX(vrrp->xmit_ifp);
req.ifi.ifi_change |= IFF_UP;
req.ifi.ifi_flags &= ~IFF_UP;
if (netlink_talk(&nl_cmd, &req.n) < 0)
status = -1;
return status;
}
int
netlink_link_del_vmac(vrrp_rt *vrrp)
{
int status = 1;
struct {
struct nlmsghdr n;
struct ifinfomsg ifi;
char buf[256];
} req;
if (!vrrp->ifp)
return -1;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_DELLINK;
req.ifi.ifi_family = AF_INET;
req.ifi.ifi_index = vrrp->vmac_ifindex;
if (netlink_talk(&nl_cmd, &req.n) < 0)
status = -1;
return status;
}
/* Link layer handling */
static int
netlink_link_setlladdr(vrrp_rt *vrrp)
{
int status = 1;
u_char ll_addr[ETH_ALEN] = {0x00, 0x00, 0x5e, 0x00, 0x01, vrrp->vrid};
struct {
struct nlmsghdr n;
struct ifinfomsg ifi;
char buf[256];
} req;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_NEWLINK;
req.ifi.ifi_family = AF_INET;
req.ifi.ifi_index = IF_INDEX(vrrp->xmit_ifp);
addattr_l(&req.n, sizeof(req), IFLA_ADDRESS, ll_addr, ETH_ALEN);
if (netlink_talk(&nl_cmd, &req.n) < 0)
status = -1;
else
memcpy(vrrp->xmit_ifp->hw_addr, ll_addr, ETH_ALEN);
return status;
}
int kernel_interface_set_master(struct interface *master,
struct interface *slave)
{
struct zebra_ns *zns = zebra_ns_lookup(NS_DEFAULT);
struct {
struct nlmsghdr n;
struct ifinfomsg ifa;
char buf[NL_PKT_BUF_SIZE];
} req;
memset(&req, 0, sizeof req);
req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_SETLINK;
req.n.nlmsg_pid = zns->netlink_cmd.snl.nl_pid;
req.ifa.ifi_index = slave->ifindex;
addattr_l(&req.n, sizeof req, IFLA_MASTER, &master->ifindex, 4);
addattr_l(&req.n, sizeof req, IFLA_LINK, &slave->ifindex, 4);
return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
0);
}
/* Interface address modification. */
static int netlink_address(int cmd, int family, struct interface *ifp,
struct connected *ifc)
{
int bytelen;
struct prefix *p;
struct {
struct nlmsghdr n;
struct ifaddrmsg ifa;
char buf[NL_PKT_BUF_SIZE];
} req;
struct zebra_ns *zns;
if (vrf_is_backend_netns())
zns = zebra_ns_lookup((ns_id_t)ifp->vrf_id);
else
zns = zebra_ns_lookup(NS_DEFAULT);
p = ifc->address;
memset(&req, 0, sizeof req - NL_PKT_BUF_SIZE);
bytelen = (family == AF_INET ? 4 : 16);
req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = cmd;
req.n.nlmsg_pid = zns->netlink_cmd.snl.nl_pid;
req.ifa.ifa_family = family;
req.ifa.ifa_index = ifp->ifindex;
addattr_l(&req.n, sizeof req, IFA_LOCAL, &p->u.prefix, bytelen);
if (family == AF_INET) {
if (CONNECTED_PEER(ifc)) {
p = ifc->destination;
addattr_l(&req.n, sizeof req, IFA_ADDRESS, &p->u.prefix,
bytelen);
} else if (cmd == RTM_NEWADDR && ifc->destination) {
p = ifc->destination;
addattr_l(&req.n, sizeof req, IFA_BROADCAST,
&p->u.prefix, bytelen);
}
}
/* p is now either ifc->address or ifc->destination */
req.ifa.ifa_prefixlen = p->prefixlen;
if (CHECK_FLAG(ifc->flags, ZEBRA_IFA_SECONDARY))
SET_FLAG(req.ifa.ifa_flags, IFA_F_SECONDARY);
if (ifc->label)
addattr_l(&req.n, sizeof req, IFA_LABEL, ifc->label,
strlen(ifc->label) + 1);
return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
0);
}
int
netlink_link_add_vmac(vrrp_rt *vrrp)
{
struct rtattr *linkinfo;
interface *ifp;
char ifname[IFNAMSIZ];
struct {
struct nlmsghdr n;
struct ifinfomsg ifi;
char buf[256];
} req;
if (!vrrp->ifp)
return -1;
memset(&req, 0, sizeof (req));
memset(ifname, 0, IFNAMSIZ);
strncpy(ifname, vrrp->vmac_ifname, IFNAMSIZ - 1);
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL;
req.n.nlmsg_type = RTM_NEWLINK;
req.ifi.ifi_family = AF_INET;
/* macvlan settings */
linkinfo = NLMSG_TAIL(&req.n);
addattr_l(&req.n, sizeof(req), IFLA_LINKINFO, NULL, 0);
addattr_l(&req.n, sizeof(req), IFLA_INFO_KIND, (void *)ll_kind, strlen(ll_kind));
linkinfo->rta_len = (void *)NLMSG_TAIL(&req.n) - (void *)linkinfo;
addattr_l(&req.n, sizeof(req), IFLA_LINK, &IF_INDEX(vrrp->ifp), sizeof(uint32_t));
addattr_l(&req.n, sizeof(req), IFLA_IFNAME, ifname, strlen(ifname));
if (netlink_talk(&nl_cmd, &req.n) < 0)
return -1;
/*
* Update interface queue and vrrp instance interface binding.
* bring it UP !
*/
netlink_interface_lookup();
ifp = if_get_by_ifname(ifname);
if (!ifp)
return -1;
vrrp->ifp = ifp;
vrrp->vmac |= 2;
netlink_link_setlladdr(vrrp);
netlink_link_up(vrrp);
/*
* By default MACVLAN interface are in VEPA mode which filters
* out received packets whose MAC source address matches that
* of the MACVLAN interface. Setting MACVLAN interface in private
* mode will not filter based on source MAC address.
*/
netlink_link_setmode(vrrp);
return 1;
}
/* Routing table change via netlink interface. */
int
netlink_route (int cmd, int family, void *dest, int length, void *gate,
int index, int zebra_flags, int table)
{
int ret;
int bytelen;
struct sockaddr_nl snl;
struct
{
struct nlmsghdr n;
struct rtmsg r;
char buf[1024];
} req;
memset (&req, 0, sizeof req);
bytelen = (family == AF_INET ? 4 : 16);
req.n.nlmsg_len = NLMSG_LENGTH (sizeof (struct rtmsg));
req.n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
req.n.nlmsg_type = cmd;
req.r.rtm_family = family;
req.r.rtm_table = table;
req.r.rtm_dst_len = length;
if (cmd == RTM_NEWROUTE)
{
req.r.rtm_protocol = RTPROT_ZEBRA;
req.r.rtm_scope = RT_SCOPE_UNIVERSE;
if (zebra_flags & ZEBRA_FLAG_BLACKHOLE)
req.r.rtm_type = RTN_BLACKHOLE;
else
req.r.rtm_type = RTN_UNICAST;
}
if (gate)
addattr_l (&req.n, sizeof req, RTA_GATEWAY, gate, bytelen);
if (dest)
addattr_l (&req.n, sizeof req, RTA_DST, dest, bytelen);
if (index > 0)
addattr32 (&req.n, sizeof req, RTA_OIF, index);
/* Destination netlink address. */
memset (&snl, 0, sizeof snl);
snl.nl_family = AF_NETLINK;
/* Talk to netlink socket. */
ret = netlink_talk (&req.n, &netlink);
if (ret < 0)
return -1;
return 0;
}
/* Interface address modification. */
static int
netlink_address (int cmd, int family, struct interface *ifp,
struct connected *ifc)
{
int bytelen;
struct prefix *p;
struct
{
struct nlmsghdr n;
struct ifaddrmsg ifa;
char buf[NL_PKT_BUF_SIZE];
} req;
p = ifc->address;
memset (&req, 0, sizeof req);
bytelen = (family == AF_INET ? 4 : 16);
req.n.nlmsg_len = NLMSG_LENGTH (sizeof (struct ifaddrmsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = cmd;
req.ifa.ifa_family = family;
req.ifa.ifa_index = ifp->ifindex;
req.ifa.ifa_prefixlen = p->prefixlen;
addattr_l (&req.n, sizeof req, IFA_LOCAL, &p->u.prefix, bytelen);
if (family == AF_INET && cmd == RTM_NEWADDR)
{
if (!CONNECTED_PEER(ifc) && ifc->destination)
{
p = ifc->destination;
addattr_l (&req.n, sizeof req, IFA_BROADCAST, &p->u.prefix,
bytelen);
}
}
if (CHECK_FLAG (ifc->flags, KROUTE_IFA_SECONDARY))
SET_FLAG (req.ifa.ifa_flags, IFA_F_SECONDARY);
if (ifc->label)
addattr_l (&req.n, sizeof req, IFA_LABEL, ifc->label,
strlen (ifc->label) + 1);
return netlink_talk (&req.n, &netlink_cmd);
}
static int
netlink_link_setmode(vrrp_rt *vrrp)
{
int status = 1;
struct {
struct nlmsghdr n;
struct ifinfomsg ifi;
char buf[256];
} req;
struct rtattr *linkinfo;
struct rtattr *data;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_NEWLINK;
req.ifi.ifi_family = AF_INET;
req.ifi.ifi_index = IF_INDEX(vrrp->xmit_ifp);
linkinfo = NLMSG_TAIL(&req.n);
addattr_l(&req.n, sizeof(req), IFLA_LINKINFO, NULL, 0);
addattr_l(&req.n, sizeof(req), IFLA_INFO_KIND, (void *)ll_kind,
strlen(ll_kind));
data = NLMSG_TAIL(&req.n);
addattr_l(&req.n, sizeof(req), IFLA_INFO_DATA, NULL, 0);
/*
* In private mode, macvlan will receive frames with same MAC addr
* as configured on the interface.
*/
#ifndef MACVLAN_MODE_VRRP
#define MACVLAN_MODE_VRRP 16
#endif
addattr32(&req.n, sizeof(req), IFLA_MACVLAN_MODE,
MACVLAN_MODE_VRRP);
data->rta_len = (void *)NLMSG_TAIL(&req.n) - (void *)data;
linkinfo->rta_len = (void *)NLMSG_TAIL(&req.n) - (void *)linkinfo;
if (netlink_talk(&nl_cmd, &req.n) < 0)
status = -1;
return status;
}
/* Add/Delete IP rule to/from a specific IP/network */
int
netlink_rule(ip_rule_t *iprule, int cmd)
{
int status = 1;
struct {
struct nlmsghdr n;
struct rtmsg r;
char buf[1024];
} req;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
req.n.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL;
req.n.nlmsg_type = cmd ? RTM_NEWRULE : RTM_DELRULE;
req.r.rtm_family = IP_FAMILY(iprule->addr);
if (iprule->table < 256)
req.r.rtm_table = iprule->table ? iprule->table : RT_TABLE_MAIN;
else {
req.r.rtm_table = RT_TABLE_UNSPEC;
addattr32(&req.n, sizeof(req), FRA_TABLE, iprule->table);
}
req.r.rtm_type = RTN_UNSPEC;
req.r.rtm_scope = RT_SCOPE_UNIVERSE;
req.r.rtm_flags = 0;
if (cmd) {
req.r.rtm_protocol = RTPROT_BOOT;
req.r.rtm_type = RTN_UNICAST;
}
/* Set rule entry */
if (iprule->dir == VRRP_RULE_FROM) {
req.r.rtm_src_len = iprule->mask;
add_addr2req(&req.n, sizeof(req), FRA_SRC, iprule->addr);
} else if (iprule->dir == VRRP_RULE_TO) {
req.r.rtm_dst_len = iprule->mask;
add_addr2req(&req.n, sizeof(req), FRA_DST, iprule->addr);
}
if (netlink_talk(&nl_cmd, &req.n) < 0)
status = -1;
return status;
}
int
netlink_link_del_vmac(vrrp_t *vrrp)
{
int status = 1;
interface_t *base_ifp ;
struct {
struct nlmsghdr n;
struct ifinfomsg ifi;
char buf[256];
} req;
if (!vrrp->ifp)
return -1;
/* Reset arp_ignore and arp_filter on the base interface if necessary */
if (vrrp->family == AF_INET) {
base_ifp = if_get_by_ifindex(vrrp->ifp->base_ifindex);
if (base_ifp)
reset_interface_parameters(base_ifp);
else
log_message(LOG_INFO, "Unable to find base interface for vrrp instance %s", vrrp->iname);
}
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_DELLINK;
req.ifi.ifi_family = AF_INET;
req.ifi.ifi_index = vrrp->vmac_ifindex;
if (netlink_talk(&nl_cmd, &req.n) < 0) {
log_message(LOG_INFO, "vmac: Error removing VMAC interface %s for vrrp_instance %s!!!"
, vrrp->vmac_ifname, vrrp->iname);
return -1;
}
log_message(LOG_INFO, "vmac: Success removing VMAC interface %s for vrrp_instance %s"
, vrrp->vmac_ifname, vrrp->iname);
return status;
}
static inline int
netlink_set_interface_flags(int ifindex, const sysctl_opts_t *sys_opts)
{
int status = 0;
struct {
struct nlmsghdr n;
struct ifinfomsg ifi;
char buf[64];
} req;
struct nlattr *start;
struct nlattr *inet_start;
struct nlattr *conf_start;
const sysctl_opts_t *so;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_NEWLINK;
req.ifi.ifi_family = AF_UNSPEC;
req.ifi.ifi_index = ifindex;
start = nest_start(&req.n, IFLA_AF_SPEC);
inet_start = nest_start(&req.n, AF_INET);
conf_start = nest_start(&req.n, IFLA_INET_CONF);
for (so = sys_opts; so->param; so++)
addattr32(&req.n, sizeof req, so->param, so->value);
nest_end(NLMSG_TAIL(&req.n), conf_start);
nest_end(NLMSG_TAIL(&req.n), inet_start);
nest_end(NLMSG_TAIL(&req.n), start);
if (netlink_talk(&nl_cmd, &req.n) < 0)
status = 1;
return status;
}
/* Routing table change via netlink interface. */
int
netlink_route_multipath (int cmd, struct prefix *p, struct rib *rib,
int family)
{
int bytelen;
struct sockaddr_nl snl;
struct nexthop *nexthop = NULL;
int nexthop_num = 0;
struct nlsock *nl;
struct
{
struct nlmsghdr n;
struct rtmsg r;
char buf[1024];
} req;
memset (&req, 0, sizeof req);
bytelen = (family == AF_INET ? 4 : 16);
req.n.nlmsg_len = NLMSG_LENGTH (sizeof (struct rtmsg));
req.n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
req.n.nlmsg_type = cmd;
req.r.rtm_family = family;
req.r.rtm_table = rib->table;
req.r.rtm_dst_len = p->prefixlen;
if (cmd == RTM_NEWROUTE)
{
req.r.rtm_protocol = RTPROT_ZEBRA;
req.r.rtm_scope = RT_SCOPE_UNIVERSE;
if (CHECK_FLAG (rib->flags, ZEBRA_FLAG_BLACKHOLE))
req.r.rtm_type = RTN_BLACKHOLE;
else
req.r.rtm_type = RTN_UNICAST;
}
addattr_l (&req.n, sizeof req, RTA_DST, &p->u.prefix, bytelen);
/* Metric. */
addattr32 (&req.n, sizeof req, RTA_PRIORITY, rib->metric);
/* Multipath case. */
if (rib->nexthop_active_num == 1 || MULTIPATH_NUM == 1)
{
for (nexthop = rib->nexthop; nexthop; nexthop = nexthop->next)
{
if ((cmd == RTM_NEWROUTE
&& CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_ACTIVE))
|| (cmd == RTM_DELROUTE
&& CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_FIB)))
{
if (CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
{
if (nexthop->rtype == NEXTHOP_TYPE_IPV4
|| nexthop->rtype == NEXTHOP_TYPE_IPV4_IFINDEX)
addattr_l (&req.n, sizeof req, RTA_GATEWAY,
&nexthop->rgate.ipv4, bytelen);
#ifdef HAVE_IPV6
if (nexthop->rtype == NEXTHOP_TYPE_IPV6
|| nexthop->rtype == NEXTHOP_TYPE_IPV6_IFINDEX
|| nexthop->rtype == NEXTHOP_TYPE_IPV6_IFNAME)
addattr_l (&req.n, sizeof req, RTA_GATEWAY,
&nexthop->rgate.ipv6, bytelen);
#endif /* HAVE_IPV6 */
if (nexthop->rtype == NEXTHOP_TYPE_IFINDEX
|| nexthop->rtype == NEXTHOP_TYPE_IFNAME
|| nexthop->rtype == NEXTHOP_TYPE_IPV4_IFINDEX
|| nexthop->rtype == NEXTHOP_TYPE_IPV6_IFINDEX
|| nexthop->rtype == NEXTHOP_TYPE_IPV6_IFNAME)
addattr32 (&req.n, sizeof req, RTA_OIF,
nexthop->rifindex);
}
else
{
if (nexthop->type == NEXTHOP_TYPE_IPV4
|| nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX)
addattr_l (&req.n, sizeof req, RTA_GATEWAY,
&nexthop->gate.ipv4, bytelen);
#ifdef HAVE_IPV6
if (nexthop->type == NEXTHOP_TYPE_IPV6
|| nexthop->type == NEXTHOP_TYPE_IPV6_IFNAME
|| nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)
addattr_l (&req.n, sizeof req, RTA_GATEWAY,
&nexthop->gate.ipv6, bytelen);
#endif /* HAVE_IPV6 */
if (nexthop->type == NEXTHOP_TYPE_IFINDEX
|| nexthop->type == NEXTHOP_TYPE_IFNAME
|| nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX
|| nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX
|| nexthop->type == NEXTHOP_TYPE_IPV6_IFNAME)
addattr32 (&req.n, sizeof req, RTA_OIF, nexthop->ifindex);
}
if (cmd == RTM_NEWROUTE)
SET_FLAG (nexthop->flags, NEXTHOP_FLAG_FIB);
nexthop_num++;
break;
}
}
}
else
{
char buf[1024];
struct rtattr *rta = (void *) buf;
struct rtnexthop *rtnh;
rta->rta_type = RTA_MULTIPATH;
rta->rta_len = RTA_LENGTH(0);
rtnh = RTA_DATA(rta);
nexthop_num = 0;
for (nexthop = rib->nexthop;
nexthop && (MULTIPATH_NUM == 0 || nexthop_num < MULTIPATH_NUM);
nexthop = nexthop->next)
{
if ((cmd == RTM_NEWROUTE
&& CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_ACTIVE))
|| (cmd == RTM_DELROUTE
&& CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_FIB)))
{
nexthop_num++;
rtnh->rtnh_len = sizeof (*rtnh);
rtnh->rtnh_flags = 0;
rtnh->rtnh_hops = 0;
rta->rta_len += rtnh->rtnh_len;
if (CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
{
if (nexthop->rtype == NEXTHOP_TYPE_IPV4
|| nexthop->rtype == NEXTHOP_TYPE_IPV4_IFINDEX)
{
rta_addattr_l (rta, 4096, RTA_GATEWAY,
&nexthop->rgate.ipv4, bytelen);
rtnh->rtnh_len += sizeof (struct rtattr) + 4;
}
#ifdef HAVE_IPV6
if (nexthop->rtype == NEXTHOP_TYPE_IPV6
|| nexthop->rtype == NEXTHOP_TYPE_IPV6_IFNAME
|| nexthop->rtype == NEXTHOP_TYPE_IPV6_IFINDEX)
rta_addattr_l (rta, 4096, RTA_GATEWAY,
&nexthop->rgate.ipv6, bytelen);
#endif /* HAVE_IPV6 */
/* ifindex */
if (nexthop->rtype == NEXTHOP_TYPE_IFINDEX
|| nexthop->rtype == NEXTHOP_TYPE_IFNAME
|| nexthop->rtype == NEXTHOP_TYPE_IPV4_IFINDEX
|| nexthop->rtype == NEXTHOP_TYPE_IPV6_IFINDEX
|| nexthop->rtype == NEXTHOP_TYPE_IPV6_IFNAME)
rtnh->rtnh_ifindex = nexthop->rifindex;
else
rtnh->rtnh_ifindex = 0;
}
else
{
if (nexthop->type == NEXTHOP_TYPE_IPV4
|| nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX)
{
rta_addattr_l (rta, 4096, RTA_GATEWAY,
&nexthop->gate.ipv4, bytelen);
rtnh->rtnh_len += sizeof (struct rtattr) + 4;
}
#ifdef HAVE_IPV6
if (nexthop->type == NEXTHOP_TYPE_IPV6
|| nexthop->type == NEXTHOP_TYPE_IPV6_IFNAME
|| nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)
rta_addattr_l (rta, 4096, RTA_GATEWAY,
&nexthop->gate.ipv6, bytelen);
#endif /* HAVE_IPV6 */
/* ifindex */
if (nexthop->type == NEXTHOP_TYPE_IFINDEX
|| nexthop->type == NEXTHOP_TYPE_IFNAME
|| nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX
|| nexthop->type == NEXTHOP_TYPE_IPV6_IFNAME
|| nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)
rtnh->rtnh_ifindex = nexthop->ifindex;
else
rtnh->rtnh_ifindex = 0;
}
rtnh = RTNH_NEXT(rtnh);
if (cmd == RTM_NEWROUTE)
SET_FLAG (nexthop->flags, NEXTHOP_FLAG_FIB);
}
}
if (rta->rta_len > RTA_LENGTH (0))
addattr_l (&req.n, 1024, RTA_MULTIPATH, RTA_DATA(rta),
RTA_PAYLOAD(rta));
}
/* If there is no useful nexthop then return. */
if (nexthop_num == 0)
{
if (IS_ZEBRA_DEBUG_KERNEL)
zlog_info ("netlink_route_multipath(): No useful nexthop.");
return 0;
}
/* Destination netlink address. */
memset (&snl, 0, sizeof snl);
snl.nl_family = AF_NETLINK;
if (family == AF_INET)
nl = &netlink_cmd;
else
nl = &netlink;
/* Talk to netlink socket. */
return netlink_talk (&req.n, nl);
}
int
kernel_route(int operation, const unsigned char *dest, unsigned short plen,
const unsigned char *gate, int ifindex, unsigned int metric,
const unsigned char *newgate, int newifindex,
unsigned int newmetric)
{
union { char raw[1024]; struct nlmsghdr nh; } buf;
struct rtmsg *rtm;
struct rtattr *rta;
int len = sizeof(buf.raw);
int rc, ipv4;
if(!nl_setup) {
fprintf(stderr,"kernel_route: netlink not initialized.\n");
errno = EIO;
return -1;
}
/* if the socket has been closed after an IO error, */
/* we try to re-open it. */
if(nl_command.sock < 0) {
rc = netlink_socket(&nl_command, 0);
if(rc < 0) {
int olderrno = errno;
perror("kernel_route: netlink_socket()");
errno = olderrno;
return -1;
}
}
/* Check that the protocol family is consistent. */
if(plen >= 96 && v4mapped(dest)) {
if(!v4mapped(gate)) {
errno = EINVAL;
return -1;
}
} else {
if(v4mapped(gate)) {
errno = EINVAL;
return -1;
}
}
ipv4 = v4mapped(gate);
if(operation == ROUTE_MODIFY) {
if(newmetric == metric && memcmp(newgate, gate, 16) == 0 &&
newifindex == ifindex)
return 0;
/* It would be better to add the new route before removing the
old one, to avoid losing packets. However, this causes
problems with non-multipath kernels, which sometimes
silently fail the request, causing "stuck" routes. Let's
stick with the naive approach, and hope that the window is
small enough to be negligible. */
kernel_route(ROUTE_FLUSH, dest, plen,
gate, ifindex, metric,
NULL, 0, 0);
rc = kernel_route(ROUTE_ADD, dest, plen,
newgate, newifindex, newmetric,
NULL, 0, 0);
if(rc < 0) {
if(errno == EEXIST)
rc = 1;
/* Should we try to re-install the flushed route on failure?
Error handling is hard. */
}
return rc;
}
kdebugf("kernel_route: %s %s/%d metric %d dev %d nexthop %s\n",
operation == ROUTE_ADD ? "add" :
operation == ROUTE_FLUSH ? "flush" : "???",
format_address(dest), plen, metric, ifindex,
format_address(gate));
/* Unreachable default routes cause all sort of weird interactions;
ignore them. */
if(metric >= KERNEL_INFINITY && (plen == 0 || (ipv4 && plen == 96)))
return 0;
memset(buf.raw, 0, sizeof(buf.raw));
if(operation == ROUTE_ADD) {
buf.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL;
buf.nh.nlmsg_type = RTM_NEWROUTE;
} else {
buf.nh.nlmsg_flags = NLM_F_REQUEST;
buf.nh.nlmsg_type = RTM_DELROUTE;
}
rtm = NLMSG_DATA(&buf.nh);
rtm->rtm_family = ipv4 ? AF_INET : AF_INET6;
rtm->rtm_dst_len = ipv4 ? plen - 96 : plen;
rtm->rtm_table = export_table;
rtm->rtm_scope = RT_SCOPE_UNIVERSE;
if(metric < KERNEL_INFINITY)
rtm->rtm_type = RTN_UNICAST;
else
rtm->rtm_type = RTN_UNREACHABLE;
rtm->rtm_protocol = RTPROT_BABEL;
rtm->rtm_flags |= RTNH_F_ONLINK;
rta = RTM_RTA(rtm);
if(ipv4) {
rta = RTA_NEXT(rta, len);
rta->rta_len = RTA_LENGTH(sizeof(struct in_addr));
rta->rta_type = RTA_DST;
memcpy(RTA_DATA(rta), dest + 12, sizeof(struct in_addr));
} else {
rta = RTA_NEXT(rta, len);
rta->rta_len = RTA_LENGTH(sizeof(struct in6_addr));
rta->rta_type = RTA_DST;
memcpy(RTA_DATA(rta), dest, sizeof(struct in6_addr));
}
rta = RTA_NEXT(rta, len);
rta->rta_len = RTA_LENGTH(sizeof(int));
rta->rta_type = RTA_PRIORITY;
if(metric < KERNEL_INFINITY) {
*(int*)RTA_DATA(rta) = metric;
rta = RTA_NEXT(rta, len);
rta->rta_len = RTA_LENGTH(sizeof(int));
rta->rta_type = RTA_OIF;
*(int*)RTA_DATA(rta) = ifindex;
if(ipv4) {
rta = RTA_NEXT(rta, len);
rta->rta_len = RTA_LENGTH(sizeof(struct in_addr));
rta->rta_type = RTA_GATEWAY;
memcpy(RTA_DATA(rta), gate + 12, sizeof(struct in_addr));
} else {
rta = RTA_NEXT(rta, len);
rta->rta_len = RTA_LENGTH(sizeof(struct in6_addr));
rta->rta_type = RTA_GATEWAY;
memcpy(RTA_DATA(rta), gate, sizeof(struct in6_addr));
}
} else {
*(int*)RTA_DATA(rta) = -1;
}
buf.nh.nlmsg_len = (char*)rta + rta->rta_len - buf.raw;
return netlink_talk(&buf.nh);
}
/* Install or uninstall specified rule for a specific interface.
* Form netlink message and ship it. Currently, notify status after
* waiting for netlink status.
*/
static int netlink_rule_update(int cmd, struct zebra_pbr_rule *rule)
{
int family;
int bytelen;
struct {
struct nlmsghdr n;
struct fib_rule_hdr frh;
char buf[NL_PKT_BUF_SIZE];
} req;
struct zebra_ns *zns = zebra_ns_lookup(NS_DEFAULT);
struct sockaddr_nl snl;
char buf1[PREFIX_STRLEN];
char buf2[PREFIX_STRLEN];
memset(&req, 0, sizeof(req) - NL_PKT_BUF_SIZE);
family = PREFIX_FAMILY(&rule->rule.filter.src_ip);
bytelen = (family == AF_INET ? 4 : 16);
req.n.nlmsg_type = cmd;
req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_pid = zns->netlink_cmd.snl.nl_pid;
req.frh.family = family;
req.frh.action = FR_ACT_TO_TBL;
/* rule's pref # */
addattr32(&req.n, sizeof(req), FRA_PRIORITY, rule->rule.priority);
/* interface on which applied */
if (rule->ifp)
addattr_l(&req.n, sizeof(req), FRA_IFNAME, rule->ifp->name,
strlen(rule->ifp->name) + 1);
/* source IP, if specified */
if (IS_RULE_FILTERING_ON_SRC_IP(rule)) {
req.frh.src_len = rule->rule.filter.src_ip.prefixlen;
addattr_l(&req.n, sizeof(req), FRA_SRC,
&rule->rule.filter.src_ip.u.prefix, bytelen);
}
/* destination IP, if specified */
if (IS_RULE_FILTERING_ON_DST_IP(rule)) {
req.frh.dst_len = rule->rule.filter.dst_ip.prefixlen;
addattr_l(&req.n, sizeof(req), FRA_DST,
&rule->rule.filter.dst_ip.u.prefix, bytelen);
}
/* fwmark, if specified */
if (IS_RULE_FILTERING_ON_FWMARK(rule)) {
addattr32(&req.n, sizeof(req), FRA_FWMARK,
rule->rule.filter.fwmark);
}
/* Route table to use to forward, if filter criteria matches. */
if (rule->rule.action.table < 256)
req.frh.table = rule->rule.action.table;
else {
req.frh.table = RT_TABLE_UNSPEC;
addattr32(&req.n, sizeof(req), FRA_TABLE,
rule->rule.action.table);
}
if (IS_ZEBRA_DEBUG_KERNEL)
zlog_debug(
"Tx %s family %s IF %s(%u) Pref %u Fwmark %u Src %s Dst %s Table %u",
nl_msg_type_to_str(cmd), nl_family_to_str(family),
rule->ifp ? rule->ifp->name : "Unknown",
rule->ifp ? rule->ifp->ifindex : 0, rule->rule.priority,
rule->rule.filter.fwmark,
prefix2str(&rule->rule.filter.src_ip, buf1,
sizeof(buf1)),
prefix2str(&rule->rule.filter.dst_ip, buf2,
sizeof(buf2)),
rule->rule.action.table);
/* Ship off the message.
* Note: Currently, netlink_talk() is a blocking call which returns
* back the status.
*/
memset(&snl, 0, sizeof(snl));
snl.nl_family = AF_NETLINK;
return netlink_talk(netlink_talk_filter, &req.n,
&zns->netlink_cmd, zns, 0);
}
int
netlink_link_add_vmac(vrrp_rt *vrrp)
{
struct rtattr *linkinfo;
interface *ifp;
char ifname[IFNAMSIZ];
struct {
struct nlmsghdr n;
struct ifinfomsg ifi;
char buf[256];
} req;
if (!vrrp->ifp)
return -1;
memset(&req, 0, sizeof (req));
memset(ifname, 0, IFNAMSIZ);
strncpy(ifname, vrrp->vmac_ifname, IFNAMSIZ - 1);
/*
* Check to see if this vmac interface was created
* by a previous instance.
*/
if (reload && (ifp = if_get_by_ifname(ifname))) {
vrrp->xmit_ifp = ifp;
/* Save ifindex for use on delete */
vrrp->vmac_ifindex = IF_INDEX(vrrp->xmit_ifp);
vrrp->vmac |= 2;
return 1;
}
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL;
req.n.nlmsg_type = RTM_NEWLINK;
req.ifi.ifi_family = AF_INET;
/* macvlan settings */
linkinfo = NLMSG_TAIL(&req.n);
addattr_l(&req.n, sizeof(req), IFLA_LINKINFO, NULL, 0);
addattr_l(&req.n, sizeof(req), IFLA_INFO_KIND, (void *)ll_kind, strlen(ll_kind));
linkinfo->rta_len = (void *)NLMSG_TAIL(&req.n) - (void *)linkinfo;
addattr_l(&req.n, sizeof(req), IFLA_LINK, &IF_INDEX(vrrp->ifp), sizeof(uint32_t));
addattr_l(&req.n, sizeof(req), IFLA_IFNAME, ifname, strlen(ifname));
if (netlink_talk(&nl_cmd, &req.n) < 0)
return -1;
/*
* Update interface queue and vrrp instance interface binding.
* bring it UP !
*/
netlink_interface_lookup();
ifp = if_get_by_ifname(ifname);
if (!ifp)
return -1;
vrrp->xmit_ifp = ifp;
vrrp->vmac_ifindex = IF_INDEX(vrrp->xmit_ifp); /* For use on delete */
vrrp->vmac |= 2;
netlink_link_setlladdr(vrrp);
vyatta_if_setup(ifname);
netlink_link_up(vrrp);
netlink_link_setmode(vrrp);
return 1;
}
/* Add/Delete IP route to/from a specific interface */
static int
netlink_route(ip_route_t *iproute, int cmd)
{
int status = 1;
struct {
struct nlmsghdr n;
struct rtmsg r;
char buf[1024];
} req;
char buf[1024];
struct rtattr *rta = (void*)buf;
struct rtnexthop *rtnh;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
req.n.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE;
req.n.nlmsg_type = cmd ? RTM_NEWROUTE : RTM_DELROUTE;
req.r.rtm_family = (iproute->dst) ? IP_FAMILY(iproute->dst) :
(iproute->src) ? IP_FAMILY(iproute->src) :
AF_INET;
if (iproute->table < 256)
req.r.rtm_table = iproute->table ? iproute->table : RT_TABLE_MAIN;
else {
req.r.rtm_table = RT_TABLE_UNSPEC;
addattr32(&req.n, sizeof(req), RTA_TABLE, iproute->table);
}
req.r.rtm_scope = RT_SCOPE_NOWHERE;
if (cmd == IPROUTE_ADD) {
req.r.rtm_protocol = RTPROT_BOOT;
req.r.rtm_scope = iproute->scope;
req.r.rtm_type = RTN_UNICAST;
}
if (iproute->blackhole)
req.r.rtm_type = RTN_BLACKHOLE;
/* Set routing entry */
if (iproute->dst) {
req.r.rtm_dst_len = iproute->dmask;
add_addr2req(&req.n, sizeof(req), RTA_DST, iproute->dst);
}
if ((!iproute->blackhole) && (!iproute->gw2))
add_addr2req(&req.n, sizeof(req), RTA_GATEWAY, iproute->gw);
if (iproute->gw && iproute->gw2) {
rta->rta_type = RTA_MULTIPATH;
rta->rta_len = RTA_LENGTH(0);
rtnh = RTA_DATA(rta);
#define MULTIPATH_ADD_GW(x) \
memset(rtnh, 0, sizeof(*rtnh)); \
rtnh->rtnh_len = sizeof(*rtnh); \
if (iproute->index) rtnh->rtnh_ifindex = iproute->index; \
rta->rta_len += rtnh->rtnh_len; \
add_addr2rta(rta, 1024, RTA_GATEWAY, x); \
rtnh->rtnh_len += sizeof(struct rtattr) + IP_SIZE(x); \
rtnh = RTNH_NEXT(rtnh);
MULTIPATH_ADD_GW(iproute->gw);
MULTIPATH_ADD_GW(iproute->gw2);
addattr_l(&req.n, sizeof(req), RTA_MULTIPATH, RTA_DATA(rta), RTA_PAYLOAD(rta));
}
if ((iproute->index) && (!iproute->gw2))
addattr32(&req.n, sizeof(req), RTA_OIF, iproute->index);
if (iproute->src)
add_addr2req(&req.n, sizeof(req), RTA_PREFSRC, iproute->src);
if (iproute->metric)
addattr32(&req.n, sizeof(req), RTA_PRIORITY, iproute->metric);
/* This returns ESRCH if the address of via address doesn't exist */
/* ENETDOWN if dev p33p1.40 for example is down */
if (netlink_talk(&nl_cmd, &req.n) < 0)
status = -1;
return status;
}
/* Add/Delete IP address to a specific interface_t */
int
netlink_ipaddress(ip_address_t *ipaddress, int cmd)
{
struct ifa_cacheinfo cinfo;
int status = 1;
struct {
struct nlmsghdr n;
struct ifaddrmsg ifa;
char buf[256];
} req;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifaddrmsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = (cmd == IPADDRESS_DEL) ? RTM_DELADDR : RTM_NEWADDR;
req.ifa = ipaddress->ifa;
if (IP_IS6(ipaddress)) {
if (cmd == IPADDRESS_ADD) {
/* Mark IPv6 address as deprecated (rfc3484) in order to prevent
* using VRRP VIP as source address in healthchecking use cases.
*/
if (ipaddress->ifa.ifa_prefixlen == 128) {
memset(&cinfo, 0, sizeof(cinfo));
cinfo.ifa_prefered = 0;
cinfo.ifa_valid = INFINITY_LIFE_TIME;
log_message(LOG_INFO, "%s has a prefix length of 128, setting "
"preferred_lft to 0", ipaddresstos(NULL, ipaddress));
addattr_l(&req.n, sizeof(req), IFA_CACHEINFO, &cinfo,
sizeof(cinfo));
}
/* Disable, per VIP, Duplicate Address Detection algorithm (DAD).
* Using the nodad flag has the following benefits:
*
* (1) The address becomes immediately usable after they're
* configured.
* (2) In the case of a temporary layer-2 / split-brain problem
* we can avoid that the active VIP transitions into the
* dadfailed phase and stays there forever - leaving us
* without service. HA/VRRP setups have their own "DAD"-like
* functionality, so it's not really needed from the IPv6 stack.
*/
#ifdef IFA_F_NODAD /* Since Linux 2.6.19 */
req.ifa.ifa_flags |= IFA_F_NODAD;
#endif
}
addattr_l(&req.n, sizeof(req), IFA_LOCAL,
&ipaddress->u.sin6_addr, sizeof(ipaddress->u.sin6_addr));
} else {
addattr_l(&req.n, sizeof(req), IFA_LOCAL,
&ipaddress->u.sin.sin_addr, sizeof(ipaddress->u.sin.sin_addr));
if (cmd == IPADDRESS_ADD) {
if (ipaddress->u.sin.sin_brd.s_addr)
addattr_l(&req.n, sizeof(req), IFA_BROADCAST,
&ipaddress->u.sin.sin_brd, sizeof(ipaddress->u.sin.sin_brd));
}
else {
/* IPADDRESS_DEL */
addattr_l(&req.n, sizeof(req), IFA_ADDRESS,
&ipaddress->u.sin.sin_addr, sizeof(ipaddress->u.sin.sin_addr));
}
}
if (cmd == IPADDRESS_ADD)
if (ipaddress->label)
addattr_l(&req.n, sizeof (req), IFA_LABEL,
ipaddress->label, strlen(ipaddress->label) + 1);
if (netlink_talk(&nl_cmd, &req.n) < 0)
status = -1;
return status;
}
static int netlink_receive(struct netlink_fd *fd, struct nlmsghdr *reply)
{
struct sockaddr_nl nladdr;
struct iovec iov;
struct msghdr msg = {
.msg_name = &nladdr,
.msg_namelen = sizeof(nladdr),
.msg_iov = &iov,
.msg_iovlen = 1,
};
int got_reply = FALSE, len;
char buf[16*1024];
iov.iov_base = buf;
while (!got_reply) {
int status;
struct nlmsghdr *h;
iov.iov_len = sizeof(buf);
status = recvmsg(fd->fd, &msg, MSG_DONTWAIT);
if (status < 0) {
if (errno == EINTR)
continue;
if (errno == EAGAIN)
return reply == NULL;
fprintf(stderr, "Netlink overrun\n");
continue;
}
if (status == 0) {
fprintf(stderr, "Netlink returned EOF\n");
return FALSE;
}
h = (struct nlmsghdr *) buf;
while (NLMSG_OK(h, status)) {
if (reply != NULL &&
h->nlmsg_seq == reply->nlmsg_seq) {
len = h->nlmsg_len;
if (len > reply->nlmsg_len) {
fprintf(stderr, "Netlink message "
"truncated\n");
len = reply->nlmsg_len;
}
memcpy(reply, h, len);
got_reply = TRUE;
} else if (h->nlmsg_type != NLMSG_DONE) {
fprintf(stderr,
"Unknown NLmsg: 0x%08x, len %d\n",
h->nlmsg_type, h->nlmsg_len);
}
h = NLMSG_NEXT(h, status);
}
}
return TRUE;
}
static int netlink_send(struct netlink_fd *fd, struct nlmsghdr *req)
{
struct sockaddr_nl nladdr;
struct iovec iov = {
.iov_base = (void*) req,
.iov_len = req->nlmsg_len
};
struct msghdr msg = {
.msg_name = &nladdr,
.msg_namelen = sizeof(nladdr),
.msg_iov = &iov,
.msg_iovlen = 1,
};
int status;
memset(&nladdr, 0, sizeof(nladdr));
nladdr.nl_family = AF_NETLINK;
req->nlmsg_seq = ++fd->seq;
status = sendmsg(fd->fd, &msg, 0);
if (status < 0) {
fprintf(stderr, "Cannot talk to rtnetlink\n");
return FALSE;
}
return TRUE;
}
static int netlink_talk(struct nlmsghdr *req, size_t replysize,
struct nlmsghdr *reply)
{
struct netlink_fd fd;
int ret = FALSE;
if (!netlink_open(&fd))
return FALSE;
if (reply == NULL)
req->nlmsg_flags |= NLM_F_ACK;
if (!netlink_send(&fd, req))
goto out;
if (reply != NULL) {
reply->nlmsg_len = replysize;
ret = netlink_receive(&fd, reply);
} else {
ret = TRUE;
}
out:
netlink_close(&fd);
return ret;
}
int netlink_route_get(struct sockaddr *dst, u_int16_t *mtu, char *ifname)
{
struct {
struct nlmsghdr n;
union {
struct rtmsg r;
struct ifinfomsg i;
};
char buf[1024];
} req;
struct rtmsg *r = NLMSG_DATA(&req.n);
struct rtattr *rta[RTA_MAX+1];
struct rtattr *rtax[RTAX_MAX+1];
struct rtattr *ifla[IFLA_MAX+1];
int index;
memset(&req, 0, sizeof(req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_GETROUTE;
req.r.rtm_family = dst->sa_family;
netlink_add_rtaddr_l(&req.n, sizeof(req), RTA_DST, dst);
req.r.rtm_dst_len = 32;
if (!netlink_talk(&req.n, sizeof(req), &req.n))
return FALSE;
netlink_parse_rtattr(rta, RTA_MAX, RTM_RTA(r),
RTM_PAYLOAD(&req.n));
if (mtu != NULL) {
if (rta[RTA_METRICS] == NULL)
return FALSE;
netlink_parse_rtattr(rtax, RTAX_MAX,
RTA_DATA(rta[RTA_METRICS]),
RTA_PAYLOAD(rta[RTA_METRICS]));
if (rtax[RTAX_MTU] == NULL)
return FALSE;
*mtu = *(int*) RTA_DATA(rtax[RTAX_MTU]);
}
if (ifname != NULL) {
if (rta[RTA_OIF] == NULL)
return FALSE;
index = *(int*) RTA_DATA(rta[RTA_OIF]);
memset(&req, 0, sizeof(req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_GETLINK;
req.i.ifi_index = index;
if (!netlink_talk(&req.n, sizeof(req), &req.n))
return FALSE;
netlink_parse_rtattr(ifla, IFLA_MAX, IFLA_RTA(r),
IFLA_PAYLOAD(&req.n));
if (ifla[IFLA_IFNAME] == NULL)
return FALSE;
memcpy(ifname, RTA_DATA(ifla[IFLA_IFNAME]),
RTA_PAYLOAD(ifla[IFLA_IFNAME]));
}
return TRUE;
}
int
netlink_link_add_vmac(vrrp_t *vrrp)
{
struct rtattr *linkinfo;
struct rtattr *data;
unsigned int base_ifindex;
interface_t *ifp;
interface_t *base_ifp;
char ifname[IFNAMSIZ];
struct {
struct nlmsghdr n;
struct ifinfomsg ifi;
char buf[256];
} req;
if (!vrrp->ifp || __test_bit(VRRP_VMAC_UP_BIT, &vrrp->vmac_flags) || !vrrp->vrid)
return -1;
if (vrrp->family == AF_INET6)
ll_addr[4] = 0x02;
else
ll_addr[4] = 0x01;
ll_addr[ETH_ALEN-1] = vrrp->vrid;
memset(&req, 0, sizeof (req));
memset(ifname, 0, IFNAMSIZ);
strncpy(ifname, vrrp->vmac_ifname, IFNAMSIZ - 1);
/*
* Check to see if this vmac interface was created
* by a previous instance.
*/
if ((ifp = if_get_by_ifname(ifname))) {
/* Check to see whether this interface has wrong mac ? */
if (memcmp((const void *) ifp->hw_addr,
(const void *) ll_addr, ETH_ALEN) != 0) {
/* We have found a VIF but the vmac do not match */
log_message(LOG_INFO, "vmac: Removing old VMAC interface %s due to conflicting "
"interface MAC for vrrp_instance %s!!!"
, vrrp->vmac_ifname, vrrp->iname);
/* Request that NETLINK remove the VIF interface first */
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_DELLINK;
req.ifi.ifi_family = AF_INET;
req.ifi.ifi_index = IF_INDEX(ifp);
if (netlink_talk(&nl_cmd, &req.n) < 0) {
log_message(LOG_INFO, "vmac: Error removing VMAC interface %s for "
"vrrp_instance %s!!!"
, vrrp->vmac_ifname, vrrp->iname);
return -1;
}
/* Interface successfully removed, now recreate */
}
}
/* Request that NETLINK create the VIF interface */
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL;
req.n.nlmsg_type = RTM_NEWLINK;
req.ifi.ifi_family = AF_INET;
/* macvlan settings */
linkinfo = NLMSG_TAIL(&req.n);
addattr_l(&req.n, sizeof(req), IFLA_LINKINFO, NULL, 0);
addattr_l(&req.n, sizeof(req), IFLA_INFO_KIND, (void *)macvlan_ll_kind, strlen(macvlan_ll_kind));
data = NLMSG_TAIL(&req.n);
addattr_l(&req.n, sizeof(req), IFLA_INFO_DATA, NULL, 0);
/*
* In private mode, macvlan will receive frames with same MAC addr
* as configured on the interface.
*/
addattr32(&req.n, sizeof(req), IFLA_MACVLAN_MODE,
MACVLAN_MODE_PRIVATE);
data->rta_len = (void *)NLMSG_TAIL(&req.n) - (void *)data;
linkinfo->rta_len = (void *)NLMSG_TAIL(&req.n) - (void *)linkinfo;
addattr_l(&req.n, sizeof(req), IFLA_LINK, &IF_INDEX(vrrp->ifp), sizeof(uint32_t));
addattr_l(&req.n, sizeof(req), IFLA_IFNAME, ifname, strlen(ifname));
addattr_l(&req.n, sizeof(req), IFLA_ADDRESS, ll_addr, ETH_ALEN);
if (netlink_talk(&nl_cmd, &req.n) < 0) {
log_message(LOG_INFO, "vmac: Error creating VMAC interface %s for vrrp_instance %s!!!"
, ifname, vrrp->iname);
return -1;
}
log_message(LOG_INFO, "vmac: Success creating VMAC interface %s for vrrp_instance %s"
, ifname, vrrp->iname);
/*
* Update interface queue and vrrp instance interface binding.
*/
netlink_interface_lookup();
ifp = if_get_by_ifname(ifname);
if (!ifp)
return -1;
base_ifp = vrrp->ifp;
base_ifindex = vrrp->ifp->ifindex;
ifp->flags = vrrp->ifp->flags; /* Copy base interface flags */
vrrp->ifp = ifp;
vrrp->ifp->base_ifindex = base_ifindex;
vrrp->ifp->vmac = 1;
vrrp->vmac_ifindex = IF_INDEX(vrrp->ifp); /* For use on delete */
if (vrrp->family == AF_INET) {
/* Set the necessary kernel parameters to make macvlans work for us */
set_interface_parameters(ifp, base_ifp);
/* We don't want IPv6 running on the interface unless we have some IPv6
* eVIPs, so disable it if not needed */
if (!vrrp->evip_add_ipv6)
link_disable_ipv6(ifp);
}
if (vrrp->family == AF_INET6 || vrrp->evip_add_ipv6) {
// We don't want a link-local address auto assigned - see RFC5798 paragraph 7.4.
// If we have a sufficiently recent kernel, we can stop a link local address
// based on the MAC address being automatically assigned. If not, then we have
// to delete the generated address after bringing the interface up (see below).
#ifdef IFLA_INET6_ADDR_GEN_MODE /* Since Linux 3.17 */
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof (struct ifinfomsg));
req.n.nlmsg_flags = NLM_F_REQUEST ;
req.n.nlmsg_type = RTM_NEWLINK;
req.ifi.ifi_family = AF_UNSPEC;
req.ifi.ifi_index = vrrp->vmac_ifindex;
u_char val = IN6_ADDR_GEN_MODE_NONE;
struct rtattr* spec;
spec = NLMSG_TAIL(&req.n);
addattr_l(&req.n, sizeof(req), IFLA_AF_SPEC, NULL,0);
data = NLMSG_TAIL(&req.n);
addattr_l(&req.n, sizeof(req), AF_INET6, NULL,0);
addattr_l(&req.n, sizeof(req), IFLA_INET6_ADDR_GEN_MODE, &val, sizeof(val));
data->rta_len = (void *)NLMSG_TAIL(&req.n) - (void *)data;
spec->rta_len = (void *)NLMSG_TAIL(&req.n) - (void *)spec;
if (netlink_talk(&nl_cmd, &req.n) < 0)
log_message(LOG_INFO, "vmac: Error setting ADDR_GEN_MODE to NONE");
#endif
if (vrrp->family == AF_INET6) {
/* Add link-local address. If a source address has been specified, use it,
* else use link-local address from underlying interface to vmac if there is one,
* otherwise construct a link-local address based on underlying interface's
* MAC address.
* This is so that VRRP advertisements will be sent from a non-VIP address, but
* using the VRRP MAC address */
ip_address_t ipaddress;
memset(&ipaddress, 0, sizeof(ipaddress));
ipaddress.ifp = ifp;
if (vrrp->saddr.ss_family == AF_INET6)
ipaddress.u.sin6_addr = ((struct sockaddr_in6*)&vrrp->saddr)->sin6_addr;
else if (base_ifp->sin6_addr.s6_addr32[0])
ipaddress.u.sin6_addr = base_ifp->sin6_addr;
else
make_link_local_address(&ipaddress.u.sin6_addr, base_ifp->hw_addr);
ipaddress.ifa.ifa_family = AF_INET6;
ipaddress.ifa.ifa_prefixlen = 64;
ipaddress.ifa.ifa_index = vrrp->vmac_ifindex;
if (netlink_ipaddress(&ipaddress, IPADDRESS_ADD) != 1)
log_message(LOG_INFO, "Adding link-local address to vmac failed");
/* Save the address as source for vrrp packets */
if (vrrp->saddr.ss_family == AF_UNSPEC)
inet_ip6tosockaddr(&ipaddress.u.sin6_addr, &vrrp->saddr);
inet_ip6scopeid(vrrp->vmac_ifindex, &vrrp->saddr);
}
}
/* bring it UP ! */
__set_bit(VRRP_VMAC_UP_BIT, &vrrp->vmac_flags);
netlink_link_up(vrrp);
#ifndef IFLA_INET6_ADDR_GEN_MODE
if (vrrp->family == AF_INET6 || vrrp->evip_add_ipv6) {
/* Delete the automatically created link-local address based on the
* MAC address if we weren't able to configure the interface not to
* create the address (see above).
* This isn't ideal, since the invalid address will exist momentarily,
* but is there any better way to do it? probably not otherwise
* ADDR_GEN_MODE wouldn't have been added to the kernel. */
ip_address_t ipaddress;
memset(&ipaddress, 0, sizeof(ipaddress));
ipaddress.u.sin6_addr = base_ifp->sin6_addr;
make_link_local_address(&ipaddress.u.sin6_addr, ll_addr);
ipaddress.ifa.ifa_family = AF_INET6;
ipaddress.ifa.ifa_prefixlen = 64;
ipaddress.ifa.ifa_index = vrrp->vmac_ifindex;
if (netlink_ipaddress(&ipaddress, IPADDRESS_DEL) != 1)
log_message(LOG_INFO, "Deleting auto link-local address from vmac failed");
}
#endif
return 1;
}
/* Add/Delete IP route to/from a specific interface */
static int
netlink_route(ip_route_t *iproute, int cmd)
{
int status = 1;
struct {
struct nlmsghdr n;
struct rtmsg r;
char buf[RTM_SIZE];
} req;
char buf[RTA_SIZE];
struct rtattr *rta = (void*)buf;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
if (cmd == IPROUTE_DEL) {
req.n.nlmsg_flags = NLM_F_REQUEST;
req.n.nlmsg_type = RTM_DELROUTE;
}
else {
req.n.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE;
if (cmd == IPROUTE_REPLACE)
req.n.nlmsg_flags |= NLM_F_REPLACE;
req.n.nlmsg_type = RTM_NEWROUTE;
}
rta->rta_type = RTA_METRICS;
rta->rta_len = RTA_LENGTH(0);
req.r.rtm_family = iproute->family;
if (iproute->table < 256)
req.r.rtm_table = iproute->table;
else {
req.r.rtm_table = RT_TABLE_UNSPEC;
addattr32(&req.n, sizeof(req), RTA_TABLE, iproute->table);
}
if (cmd == IPROUTE_DEL) {
req.r.rtm_scope = RT_SCOPE_NOWHERE;
if (iproute->mask & IPROUTE_BIT_TYPE)
req.r.rtm_type = iproute->type;
}
else {
req.r.rtm_protocol = RTPROT_BOOT;
req.r.rtm_scope = RT_SCOPE_UNIVERSE;
req.r.rtm_type = iproute->type;
}
if (iproute->mask & IPROUTE_BIT_PROTOCOL)
req.r.rtm_protocol = iproute->protocol;
if (iproute->mask & IPROUTE_BIT_SCOPE)
req.r.rtm_scope = iproute->scope;
if (iproute->dst) {
req.r.rtm_dst_len = iproute->dst->ifa.ifa_prefixlen;
add_addr2req(&req.n, sizeof(req), RTA_DST, iproute->dst);
}
if (iproute->src) {
req.r.rtm_src_len = iproute->src->ifa.ifa_prefixlen;
add_addr2req(&req.n, sizeof(req), RTA_SRC, iproute->src);
}
if (iproute->pref_src)
add_addr2req(&req.n, sizeof(req), RTA_PREFSRC, iproute->pref_src);
//#ifdef _HAVE_RTA_NEWDST_
// if (iproute->as_to)
// add_addr2req(&req.n, sizeof(req), RTA_NEWDST, iproute->as_to);
//#endif
if (iproute->via) {
if (iproute->via->ifa.ifa_family == iproute->family)
add_addr2req(&req.n, sizeof(req), RTA_GATEWAY, iproute->via);
#ifdef _HAVE_RTA_VIA_
else
add_addr_fam2req(&req.n, sizeof(req), RTA_VIA, iproute->via);
#endif
}
#ifdef _HAVE_RTA_ENCAP_
if (iproute->encap.type != LWTUNNEL_ENCAP_NONE) {
char encap_buf[ENCAP_RTA_SIZE];
struct rtattr *encap_rta = (void *)encap_buf;
encap_rta->rta_type = RTA_ENCAP;
encap_rta->rta_len = RTA_LENGTH(0);
add_encap(encap_rta, sizeof(encap_buf), &iproute->encap);
if (encap_rta->rta_len > RTA_LENGTH(0))
addraw_l(&req.n, sizeof(encap_buf), RTA_DATA(encap_rta), RTA_PAYLOAD(encap_rta));
}
#endif
if (iproute->mask & IPROUTE_BIT_DSFIELD)
req.r.rtm_tos = iproute->tos;
if (iproute->oif)
addattr32(&req.n, sizeof(req), RTA_OIF, iproute->oif->ifindex);
if (iproute->mask & IPROUTE_BIT_METRIC)
addattr32(&req.n, sizeof(req), RTA_PRIORITY, iproute->metric);
req.r.rtm_flags = iproute->flags;
if (iproute->realms)
addattr32(&req.n, sizeof(req), RTA_FLOW, iproute->realms);
#ifdef _HAVE_RTA_EXPIRES_
if (iproute->mask & IPROUTE_BIT_EXPIRES)
addattr32(&req.n, sizeof(req), RTA_EXPIRES, iproute->expires);
#endif
#ifdef RTAX_CC_ALGO
if (iproute->congctl)
rta_addattr_l(rta, sizeof(buf), RTAX_CC_ALGO, iproute->congctl, strlen(iproute->congctl));
#endif
if (iproute->mask & IPROUTE_BIT_RTT)
rta_addattr32(rta, sizeof(buf), RTAX_RTT, iproute->rtt);
if (iproute->mask & IPROUTE_BIT_RTTVAR)
rta_addattr32(rta, sizeof(buf), RTAX_RTTVAR, iproute->rttvar);
if (iproute->mask & IPROUTE_BIT_RTO_MIN)
rta_addattr32(rta, sizeof(buf), RTAX_RTO_MIN, iproute->rto_min);
#ifdef RTAX_FEATURES
if (iproute->features)
rta_addattr32(rta, sizeof(buf), RTAX_FEATURES, iproute->features);
#endif
if (iproute->mask & IPROUTE_BIT_MTU)
rta_addattr32(rta, sizeof(buf), RTAX_MTU, iproute->mtu);
if (iproute->mask & IPROUTE_BIT_WINDOW)
rta_addattr32(rta, sizeof(buf), RTAX_WINDOW, iproute->window);
if (iproute->mask & IPROUTE_BIT_SSTHRESH)
rta_addattr32(rta, sizeof(buf), RTAX_SSTHRESH, iproute->ssthresh);
if (iproute->mask & IPROUTE_BIT_CWND)
rta_addattr32(rta, sizeof(buf), RTAX_CWND, iproute->cwnd);
if (iproute->mask & IPROUTE_BIT_ADVMSS)
rta_addattr32(rta, sizeof(buf), RTAX_ADVMSS, iproute->advmss);
if (iproute->mask & IPROUTE_BIT_REORDERING)
rta_addattr32(rta, sizeof(buf), RTAX_REORDERING, iproute->reordering);
if (iproute->mask & IPROUTE_BIT_HOPLIMIT)
rta_addattr32(rta, sizeof(buf), RTAX_HOPLIMIT, iproute->hoplimit);
if (iproute->mask & IPROUTE_BIT_INITCWND)
rta_addattr32(rta, sizeof(buf), RTAX_INITCWND, iproute->initcwnd);
#ifdef RTAX_INITRWND
if (iproute->mask & IPROUTE_BIT_INITRWND)
rta_addattr32(rta, sizeof(buf), RTAX_INITRWND, iproute->initrwnd);
#endif
#ifdef RTAX_QUICKACK
if (iproute->mask & IPROUTE_BIT_QUICKACK)
rta_addattr32(rta, sizeof(buf), RTAX_QUICKACK, iproute->quickack);
#endif
#ifdef _HAVE_RTA_PREF_
if (iproute->mask & IPROUTE_BIT_PREF)
addattr8(&req.n, sizeof(req), RTA_PREF, iproute->pref);
#endif
if (rta->rta_len > RTA_LENGTH(0)) {
if (iproute->lock)
rta_addattr32(rta, sizeof(buf), RTAX_LOCK, iproute->lock);
addattr_l(&req.n, sizeof(req), RTA_METRICS, RTA_DATA(rta), RTA_PAYLOAD(rta));
}
if (!LIST_ISEMPTY(iproute->nhs))
add_nexthops(iproute, &req.n, &req.r);
#ifdef DEBUG_NETLINK_MSG
size_t i, j;
uint8_t *p;
char lbuf[3072];
char *op = lbuf;
log_message(LOG_INFO, "rtmsg buffer used %lu, rtattr buffer used %d", req.n.nlmsg_len - NLMSG_LENGTH(sizeof(struct rtmsg)), rta->rta_len);
op += snprintf(op, sizeof(lbuf) - (op - lbuf), "nlmsghdr %p(%u):", &req.n, req.n.nlmsg_len);
for (i = 0, p = (uint8_t*)&req.n; i < sizeof(struct nlmsghdr); i++)
op += snprintf(op, sizeof(lbuf) - (op - lbuf), " %2.2hhx", *(p++));
log_message(LOG_INFO, "%s\n", lbuf);
op = lbuf;
op += snprintf(op, sizeof(lbuf) - (op - lbuf), "rtmsg %p(%lu):", &req.r, req.n.nlmsg_len - sizeof(struct nlmsghdr));
for (i = 0, p = (uint8_t*)&req.r; i < + req.n.nlmsg_len - sizeof(struct nlmsghdr); i++)
op += snprintf(op, sizeof(lbuf) - (op - lbuf), " %2.2hhx", *(p++));
for (j = 0; lbuf + j < op; j+= MAX_LOG_MSG)
log_message(LOG_INFO, "%.*\n", MAX_LOG_MSG, lbuf+j);
#endif
/* This returns ESRCH if the address of via address doesn't exist */
/* ENETDOWN if dev p33p1.40 for example is down */
if (netlink_talk(&nl_cmd, &req.n) < 0) {
#ifdef _HAVE_RTA_EXPIRES_
/* If an expiry was set on the route, it may have disappeared already */
if (cmd != IPADDRESS_DEL || !(iproute->mask & IPROUTE_BIT_EXPIRES))
#endif
status = -1;
}
return status;
}
/* Add/Delete IP rule to/from a specific IP/network */
static int
netlink_rule(ip_rule_t *iprule, int cmd)
{
int status = 1;
struct {
struct nlmsghdr n;
struct fib_rule_hdr frh;
char buf[1024];
} req;
memset(&req, 0, sizeof (req));
req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
req.n.nlmsg_flags = NLM_F_REQUEST;
if (cmd != IPRULE_DEL) {
req.n.nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
req.n.nlmsg_type = RTM_NEWRULE;
req.frh.action = FR_ACT_UNSPEC;
}
else {
req.frh.action = FR_ACT_UNSPEC;
req.n.nlmsg_type = RTM_DELRULE;
}
req.frh.table = RT_TABLE_UNSPEC;
req.frh.flags = 0;
req.frh.tos = iprule->tos; // Hex value - 0xnn <= 255, or name from rt_dsfield
req.frh.family = iprule->family;
if (iprule->action == FR_ACT_TO_TBL
#if HAVE_DECL_FRA_L3MDEV
&& !iprule->l3mdev
#endif
) {
if (iprule->table < 256) // "Table" or "lookup"
req.frh.table = iprule->table ? iprule->table & 0xff : RT_TABLE_MAIN;
else {
req.frh.table = RT_TABLE_UNSPEC;
addattr32(&req.n, sizeof(req), FRA_TABLE, iprule->table);
}
}
if (iprule->invert)
req.frh.flags |= FIB_RULE_INVERT; // "not"
/* Set rule entry */
if (iprule->from_addr) { // can be "default"/"any"/"all" - and to addr => bytelen == bitlen == 0
add_addr2req(&req.n, sizeof(req), FRA_SRC, iprule->from_addr);
req.frh.src_len = iprule->from_addr->ifa.ifa_prefixlen;
}
if (iprule->to_addr) {
add_addr2req(&req.n, sizeof(req), FRA_DST, iprule->to_addr);
req.frh.dst_len = iprule->to_addr->ifa.ifa_prefixlen;
}
if (iprule->mask & IPRULE_BIT_PRIORITY) // "priority/order/preference"
addattr32(&req.n, sizeof(req), FRA_PRIORITY, iprule->priority);
if (iprule->mask & IPRULE_BIT_FWMARK) // "fwmark"
addattr32(&req.n, sizeof(req), FRA_FWMARK, iprule->fwmark);
if (iprule->mask & IPRULE_BIT_FWMASK) // "fwmark number followed by /nn"
addattr32(&req.n, sizeof(req), FRA_FWMASK, iprule->fwmask);
if (iprule->realms) // "realms u16[/u16] using rt_realms. after / is 16 msb (src), pre slash is 16 lsb (dest)"
addattr32(&req.n, sizeof(req), FRA_FLOW, iprule->realms);
#if HAVE_DECL_FRA_SUPPRESS_PREFIXLEN
if (iprule->suppress_prefix_len != -1) // "suppress_prefixlength" - only valid if table != 0
addattr32(&req.n, sizeof(req), FRA_SUPPRESS_PREFIXLEN, iprule->suppress_prefix_len);
#endif
#if HAVE_DECL_FRA_SUPPRESS_IFGROUP
if (iprule->mask & IPRULE_BIT_SUP_GROUP) // "suppress_ifgroup" or "sup_group" int32 - only valid if table !=0
addattr32(&req.n, sizeof(req), FRA_SUPPRESS_IFGROUP, iprule->suppress_group);
#endif
if (iprule->iif) // "dev/iif"
addattr_l(&req.n, sizeof(req), FRA_IFNAME, iprule->iif, strlen(iprule->iif->ifname)+1);
#if HAVE_DECL_FRA_OIFNAME
if (iprule->oif) // "oif"
addattr_l(&req.n, sizeof(req), FRA_OIFNAME, iprule->oif, strlen(iprule->oif->ifname)+1);
#endif
#if HAVE_DECL_FRA_TUN_ID
if (iprule->tunnel_id)
addattr64(&req.n, sizeof(req), FRA_TUN_ID, htobe64(iprule->tunnel_id));
#endif
#if HAVE_DECL_FRA_UID_RANGE
if (iprule->mask & IPRULE_BIT_UID_RANGE)
addattr_l(&req.n, sizeof(req), FRA_UID_RANGE, &iprule->uid_range, sizeof(iprule->uid_range));
#endif
#if HAVE_DECL_FRA_L3MDEV
if (iprule->l3mdev)
addattr8(&req.n, sizeof(req), FRA_L3MDEV, 1);
#endif
#if HAVE_DECL_FRA_PROTOCOL
if (iprule->mask & IPRULE_BIT_PROTOCOL)
addattr8(&req.n, sizeof(req), FRA_PROTOCOL, iprule->protocol);
#endif
#if HAVE_DECL_FRA_IP_PROTO
if (iprule->mask & IPRULE_BIT_IP_PROTO)
addattr8(&req.n, sizeof(req), FRA_IP_PROTO, iprule->ip_proto);
#endif
#if HAVE_DECL_FRA_SPORT_RANGE
if (iprule->mask & IPRULE_BIT_SPORT_RANGE)
addattr_l(&req.n, sizeof(req), FRA_SPORT_RANGE, &iprule->src_port, sizeof(iprule->src_port));
#endif
#if HAVE_DECL_FRA_DPORT_RANGE
if (iprule->mask & IPRULE_BIT_DPORT_RANGE)
addattr_l(&req.n, sizeof(req), FRA_DPORT_RANGE, &iprule->dst_port, sizeof(iprule->dst_port));
#endif
if (iprule->action == FR_ACT_GOTO) { // "goto"
addattr32(&req.n, sizeof(req), FRA_GOTO, iprule->goto_target);
req.frh.action = FR_ACT_GOTO;
}
req.frh.action = iprule->action;
if (netlink_talk(&nl_cmd, &req.n) < 0)
status = -1;
return status;
}
