/* Note: on netlink systems, there should be a 1-to-1 mapping between interface
names and ifindex values. */
static void set_ifindex(struct interface *ifp, ifindex_t ifi_index,
struct zebra_ns *zns)
{
struct interface *oifp;
if (((oifp = if_lookup_by_index_per_ns(zns, ifi_index)) != NULL)
&& (oifp != ifp)) {
if (ifi_index == IFINDEX_INTERNAL)
flog_err(
LIB_ERR_INTERFACE,
"Netlink is setting interface %s ifindex to reserved internal value %u",
ifp->name, ifi_index);
else {
if (IS_ZEBRA_DEBUG_KERNEL)
zlog_debug(
"interface index %d was renamed from %s to %s",
ifi_index, oifp->name, ifp->name);
if (if_is_up(oifp))
flog_err(
LIB_ERR_INTERFACE,
"interface rename detected on up interface: index %d was renamed from %s to %s, results are uncertain!",
ifi_index, oifp->name, ifp->name);
if_delete_update(oifp);
}
}
if_set_index(ifp, ifi_index);
}
int bgp_set_socket_ttl(struct peer *peer, int bgp_sock)
{
char buf[INET_ADDRSTRLEN];
int ret = 0;
/* In case of peer is EBGP, we should set TTL for this connection. */
if (!peer->gtsm_hops && (peer_sort(peer) == BGP_PEER_EBGP)) {
ret = sockopt_ttl(peer->su.sa.sa_family, bgp_sock, peer->ttl);
if (ret) {
flog_err(
EC_LIB_SOCKET,
"%s: Can't set TxTTL on peer (rtrid %s) socket, err = %d",
__func__,
inet_ntop(AF_INET, &peer->remote_id, buf,
sizeof(buf)),
errno);
return ret;
}
} else if (peer->gtsm_hops) {
/* On Linux, setting minttl without setting ttl seems to mess
with the
outgoing ttl. Therefore setting both.
*/
ret = sockopt_ttl(peer->su.sa.sa_family, bgp_sock, MAXTTL);
if (ret) {
flog_err(
EC_LIB_SOCKET,
"%s: Can't set TxTTL on peer (rtrid %s) socket, err = %d",
__func__,
inet_ntop(AF_INET, &peer->remote_id, buf,
sizeof(buf)),
errno);
return ret;
}
ret = sockopt_minttl(peer->su.sa.sa_family, bgp_sock,
MAXTTL + 1 - peer->gtsm_hops);
if (ret) {
flog_err(
EC_LIB_SOCKET,
"%s: Can't set MinTTL on peer (rtrid %s) socket, err = %d",
__func__,
inet_ntop(AF_INET, &peer->remote_id, buf,
sizeof(buf)),
errno);
return ret;
}
}
return ret;
}
/* After TCP connection is established. Get local address and port. */
int bgp_getsockname(struct peer *peer)
{
if (peer->su_local) {
sockunion_free(peer->su_local);
peer->su_local = NULL;
}
if (peer->su_remote) {
sockunion_free(peer->su_remote);
peer->su_remote = NULL;
}
peer->su_local = sockunion_getsockname(peer->fd);
if (!peer->su_local)
return -1;
peer->su_remote = sockunion_getpeername(peer->fd);
if (!peer->su_remote)
return -1;
if (!bgp_zebra_nexthop_set(peer->su_local, peer->su_remote,
&peer->nexthop, peer)) {
flog_err(EC_BGP_NH_UPD,
"%s: nexthop_set failed, resetting connection - intf %p",
peer->host, peer->nexthop.ifp);
return -1;
}
return 0;
}
/* we need to be very cautious with this API as SA del too can trigger an
* upstream del and we will get stuck in a simple loop */
static void pim_msdp_sa_local_del_on_up_del(struct pim_instance *pim,
struct prefix_sg *sg)
{
struct pim_msdp_sa *sa;
sa = pim_msdp_sa_find(pim, sg);
if (sa) {
if (PIM_DEBUG_MSDP_INTERNAL) {
zlog_debug("MSDP local sa %s del on up del",
sa->sg_str);
}
/* if there is no local reference escape */
if (!(sa->flags & PIM_MSDP_SAF_LOCAL)) {
if (PIM_DEBUG_MSDP_INTERNAL) {
zlog_debug("MSDP local sa %s del; no local ref",
sa->sg_str);
}
return;
}
if (sa->flags & PIM_MSDP_SAF_UP_DEL_IN_PROG) {
/* MSDP is the one that triggered the upstream del. if
* this happens
* we most certainly have a bug in the PIM upstream
* state machine. We
* will not have a local reference unless the KAT is
* running. And if the
* KAT is running there MUST be an additional
* source-stream reference to
* the flow. Accounting for such cases requires lot of
* changes; perhaps
* address this in the next release? - XXX */
flog_err(
EC_LIB_DEVELOPMENT,
"MSDP sa %s SPT teardown is causing the local entry to be removed",
sa->sg_str);
return;
}
/* we are dropping the sa on upstream del we should not have an
* upstream reference */
if (sa->up) {
if (PIM_DEBUG_MSDP_INTERNAL) {
zlog_debug("MSDP local sa %s del; up non-NULL",
sa->sg_str);
}
sa->up = NULL;
}
pim_msdp_sa_deref(sa, PIM_MSDP_SAF_LOCAL);
}
}
void
install_route(struct babel_route *route)
{
int i, rc;
if(route->installed)
return;
if(!route_feasible(route))
flog_err(EC_BABEL_ROUTE, "WARNING: installing unfeasible route "
"(this shouldn't happen).");
i = find_route_slot(route->src->prefix, route->src->plen, NULL);
assert(i >= 0 && i < route_slots);
if(routes[i] != route && routes[i]->installed) {
flog_err(EC_BABEL_ROUTE,
"WARNING: attempting to install duplicate route "
"(this shouldn't happen).");
return;
}
rc = kernel_route(ROUTE_ADD, route->src->prefix, route->src->plen,
route->nexthop,
route->neigh->ifp->ifindex,
metric_to_kernel(route_metric(route)), NULL, 0, 0);
if(rc < 0) {
int save = errno;
flog_err(EC_BABEL_ROUTE, "kernel_route(ADD): %s",
safe_strerror(errno));
if(save != EEXIST)
return;
}
route->installed = 1;
move_installed_route(route, i);
}
void
uninstall_route(struct babel_route *route)
{
int rc;
if(!route->installed)
return;
rc = kernel_route(ROUTE_FLUSH, route->src->prefix, route->src->plen,
route->nexthop,
route->neigh->ifp->ifindex,
metric_to_kernel(route_metric(route)), NULL, 0, 0);
if(rc < 0)
flog_err(EC_BABEL_ROUTE, "kernel_route(FLUSH): %s",
safe_strerror(errno));
route->installed = 0;
}
struct source*
find_source(const unsigned char *id, const unsigned char *p, unsigned char plen,
int create, unsigned short seqno)
{
struct source *src;
for(src = srcs; src; src = src->next) {
/* This should really be a hash table. For now, check the
last byte first. */
if(src->id[7] != id[7])
continue;
if(memcmp(src->id, id, 8) != 0)
continue;
if(src->plen != plen)
continue;
if(memcmp(src->prefix, p, 16) == 0)
return src;
}
if(!create)
return NULL;
src = malloc(sizeof(struct source));
if(src == NULL) {
flog_err(BABEL_ERR_MEMORY, "malloc(source): %s", safe_strerror(errno));
return NULL;
}
memcpy(src->id, id, 8);
memcpy(src->prefix, p, 16);
src->plen = plen;
src->seqno = seqno;
src->metric = INFINITY;
src->time = babel_now.tv_sec;
src->route_count = 0;
src->next = srcs;
srcs = src;
return src;
}
/* passive peer socket accept */
static int pim_msdp_sock_accept(struct thread *thread)
{
union sockunion su;
struct pim_instance *pim = THREAD_ARG(thread);
int accept_sock;
int msdp_sock;
struct pim_msdp_peer *mp;
char buf[SU_ADDRSTRLEN];
sockunion_init(&su);
/* re-register accept thread */
accept_sock = THREAD_FD(thread);
if (accept_sock < 0) {
flog_err(LIB_ERR_DEVELOPMENT,
"accept_sock is negative value %d", accept_sock);
return -1;
}
pim->msdp.listener.thread = NULL;
thread_add_read(master, pim_msdp_sock_accept, pim, accept_sock,
&pim->msdp.listener.thread);
/* accept client connection. */
msdp_sock = sockunion_accept(accept_sock, &su);
if (msdp_sock < 0) {
flog_err_sys(LIB_ERR_SOCKET, "pim_msdp_sock_accept failed (%s)",
safe_strerror(errno));
return -1;
}
/* see if have peer config for this */
mp = pim_msdp_peer_find(pim, su.sin.sin_addr);
if (!mp || !PIM_MSDP_PEER_IS_LISTENER(mp)) {
++pim->msdp.rejected_accepts;
if (PIM_DEBUG_MSDP_EVENTS) {
flog_err(PIM_ERR_MSDP_PACKET,
"msdp peer connection refused from %s",
sockunion2str(&su, buf, SU_ADDRSTRLEN));
}
close(msdp_sock);
return -1;
}
if (PIM_DEBUG_MSDP_INTERNAL) {
zlog_debug("MSDP peer %s accept success%s", mp->key_str,
mp->fd >= 0 ? "(dup)" : "");
}
/* if we have an existing connection we need to kill that one
* with this one */
if (mp->fd >= 0) {
if (PIM_DEBUG_MSDP_EVENTS) {
zlog_notice(
"msdp peer new connection from %s stop old connection",
sockunion2str(&su, buf, SU_ADDRSTRLEN));
}
pim_msdp_peer_stop_tcp_conn(mp, true /* chg_state */);
}
mp->fd = msdp_sock;
set_nonblocking(mp->fd);
pim_msdp_update_sock_send_buffer_size(mp->fd);
pim_msdp_peer_established(mp);
return 0;
}
/* active peer socket setup */
int pim_msdp_sock_connect(struct pim_msdp_peer *mp)
{
int rc;
if (PIM_DEBUG_MSDP_INTERNAL) {
zlog_debug("MSDP peer %s attempt connect%s", mp->key_str,
mp->fd < 0 ? "" : "(dup)");
}
/* if we have an existing connection we need to kill that one
* with this one */
if (mp->fd >= 0) {
if (PIM_DEBUG_MSDP_EVENTS) {
zlog_notice(
"msdp duplicate connect to %s nuke old connection",
mp->key_str);
}
pim_msdp_peer_stop_tcp_conn(mp, false /* chg_state */);
}
/* Make socket for the peer. */
mp->fd = sockunion_socket(&mp->su_peer);
if (mp->fd < 0) {
flog_err_sys(LIB_ERR_SOCKET,
"pim_msdp_socket socket failure: %s",
safe_strerror(errno));
return -1;
}
if (mp->pim->vrf_id != VRF_DEFAULT) {
struct interface *ifp =
if_lookup_by_name(mp->pim->vrf->name, mp->pim->vrf_id);
if (!ifp) {
flog_err(LIB_ERR_INTERFACE,
"%s: Unable to lookup vrf interface: %s",
__PRETTY_FUNCTION__, mp->pim->vrf->name);
return -1;
}
if (pim_socket_bind(mp->fd, ifp)) {
flog_err_sys(LIB_ERR_SOCKET,
"%s: Unable to bind to socket: %s",
__PRETTY_FUNCTION__, safe_strerror(errno));
close(mp->fd);
mp->fd = -1;
return -1;
}
}
set_nonblocking(mp->fd);
/* Set socket send buffer size */
pim_msdp_update_sock_send_buffer_size(mp->fd);
sockopt_reuseaddr(mp->fd);
sockopt_reuseport(mp->fd);
/* source bind */
rc = sockunion_bind(mp->fd, &mp->su_local, 0, &mp->su_local);
if (rc < 0) {
flog_err_sys(LIB_ERR_SOCKET,
"pim_msdp_socket connect bind failure: %s",
safe_strerror(errno));
close(mp->fd);
mp->fd = -1;
return rc;
}
/* Connect to the remote mp. */
return (sockunion_connect(mp->fd, &mp->su_peer,
htons(PIM_MSDP_TCP_PORT), 0));
}
/* global listener for the MSDP well know TCP port */
int pim_msdp_sock_listen(struct pim_instance *pim)
{
int sock;
int socklen;
struct sockaddr_in sin;
int rc;
struct pim_msdp_listener *listener = &pim->msdp.listener;
if (pim->msdp.flags & PIM_MSDPF_LISTENER) {
/* listener already setup */
return 0;
}
sock = socket(AF_INET, SOCK_STREAM, 0);
if (sock < 0) {
flog_err_sys(LIB_ERR_SOCKET, "socket: %s",
safe_strerror(errno));
return sock;
}
memset(&sin, 0, sizeof(struct sockaddr_in));
sin.sin_family = AF_INET;
sin.sin_port = htons(PIM_MSDP_TCP_PORT);
socklen = sizeof(struct sockaddr_in);
#ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
sin.sin_len = socklen;
#endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
sockopt_reuseaddr(sock);
sockopt_reuseport(sock);
if (pim->vrf_id != VRF_DEFAULT) {
struct interface *ifp =
if_lookup_by_name(pim->vrf->name, pim->vrf_id);
if (!ifp) {
flog_err(LIB_ERR_INTERFACE,
"%s: Unable to lookup vrf interface: %s",
__PRETTY_FUNCTION__, pim->vrf->name);
close(sock);
return -1;
}
if (pim_socket_bind(sock, ifp)) {
flog_err_sys(LIB_ERR_SOCKET,
"%s: Unable to bind to socket: %s",
__PRETTY_FUNCTION__, safe_strerror(errno));
close(sock);
return -1;
}
}
frr_elevate_privs(&pimd_privs) {
/* bind to well known TCP port */
rc = bind(sock, (struct sockaddr *)&sin, socklen);
}
if (rc < 0) {
flog_err_sys(LIB_ERR_SOCKET,
"pim_msdp_socket bind to port %d: %s",
ntohs(sin.sin_port), safe_strerror(errno));
close(sock);
return rc;
}
rc = listen(sock, 3 /* backlog */);
if (rc < 0) {
flog_err_sys(LIB_ERR_SOCKET, "pim_msdp_socket listen: %s",
safe_strerror(errno));
close(sock);
return rc;
}
/* add accept thread */
listener->fd = sock;
memcpy(&listener->su, &sin, socklen);
listener->thread = NULL;
thread_add_read(pim->msdp.master, pim_msdp_sock_accept, pim, sock,
&listener->thread);
pim->msdp.flags |= PIM_MSDPF_LISTENER;
return 0;
}
struct isis_circuit *
isis_csm_state_change(int event, struct isis_circuit *circuit, void *arg)
{
int old_state;
old_state = circuit ? circuit->state : C_STATE_NA;
if (isis->debugs & DEBUG_EVENTS)
zlog_debug("CSM_EVENT: %s", EVENT2STR(event));
switch (old_state) {
case C_STATE_NA:
if (circuit)
zlog_warn("Non-null circuit while state C_STATE_NA");
assert(circuit == NULL);
switch (event) {
case ISIS_ENABLE:
circuit = isis_circuit_new();
isis_circuit_configure(circuit,
(struct isis_area *)arg);
circuit->state = C_STATE_CONF;
break;
case IF_UP_FROM_Z:
circuit = isis_circuit_new();
isis_circuit_if_add(circuit, (struct interface *)arg);
listnode_add(isis->init_circ_list, circuit);
circuit->state = C_STATE_INIT;
break;
case ISIS_DISABLE:
zlog_warn("circuit already disabled");
break;
case IF_DOWN_FROM_Z:
zlog_warn("circuit already disconnected");
break;
}
break;
case C_STATE_INIT:
assert(circuit);
switch (event) {
case ISIS_ENABLE:
isis_circuit_configure(circuit,
(struct isis_area *)arg);
if (isis_circuit_up(circuit) != ISIS_OK) {
isis_circuit_deconfigure(
circuit, (struct isis_area *)arg);
break;
}
circuit->state = C_STATE_UP;
isis_event_circuit_state_change(circuit, circuit->area,
1);
listnode_delete(isis->init_circ_list, circuit);
break;
case IF_UP_FROM_Z:
assert(circuit);
zlog_warn("circuit already connected");
break;
case ISIS_DISABLE:
zlog_warn("circuit already disabled");
break;
case IF_DOWN_FROM_Z:
isis_circuit_if_del(circuit, (struct interface *)arg);
listnode_delete(isis->init_circ_list, circuit);
isis_circuit_del(circuit);
circuit = NULL;
break;
}
break;
case C_STATE_CONF:
assert(circuit);
switch (event) {
case ISIS_ENABLE:
zlog_warn("circuit already enabled");
break;
case IF_UP_FROM_Z:
isis_circuit_if_add(circuit, (struct interface *)arg);
if (isis_circuit_up(circuit) != ISIS_OK) {
flog_err(
ISIS_ERR_CONFIG,
"Could not bring up %s because of invalid config.",
circuit->interface->name);
flog_err(
ISIS_ERR_CONFIG,
"Clearing config for %s. Please re-examine it.",
circuit->interface->name);
if (circuit->ip_router) {
circuit->ip_router = 0;
circuit->area->ip_circuits--;
}
if (circuit->ipv6_router) {
circuit->ipv6_router = 0;
circuit->area->ipv6_circuits--;
}
circuit_update_nlpids(circuit);
isis_circuit_deconfigure(circuit,
circuit->area);
listnode_add(isis->init_circ_list, circuit);
circuit->state = C_STATE_INIT;
break;
}
circuit->state = C_STATE_UP;
isis_event_circuit_state_change(circuit, circuit->area,
1);
break;
case ISIS_DISABLE:
isis_circuit_deconfigure(circuit,
(struct isis_area *)arg);
isis_circuit_del(circuit);
circuit = NULL;
break;
case IF_DOWN_FROM_Z:
zlog_warn("circuit already disconnected");
break;
}
break;
case C_STATE_UP:
assert(circuit);
switch (event) {
case ISIS_ENABLE:
zlog_warn("circuit already configured");
break;
case IF_UP_FROM_Z:
zlog_warn("circuit already connected");
break;
case ISIS_DISABLE:
isis_circuit_down(circuit);
isis_circuit_deconfigure(circuit,
(struct isis_area *)arg);
circuit->state = C_STATE_INIT;
isis_event_circuit_state_change(
circuit, (struct isis_area *)arg, 0);
listnode_add(isis->init_circ_list, circuit);
break;
case IF_DOWN_FROM_Z:
isis_circuit_down(circuit);
isis_circuit_if_del(circuit, (struct interface *)arg);
circuit->state = C_STATE_CONF;
isis_event_circuit_state_change(circuit, circuit->area,
0);
break;
}
break;
default:
zlog_warn("Invalid circuit state %d", old_state);
}
if (isis->debugs & DEBUG_EVENTS)
zlog_debug("CSM_STATE_CHANGE: %s -> %s ", STATE2STR(old_state),
circuit ? STATE2STR(circuit->state)
: STATE2STR(C_STATE_NA));
return circuit;
}
static int ssmpingd_socket(struct in_addr addr, int port, int mttl)
{
struct sockaddr_in sockaddr;
int fd;
fd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
if (fd < 0) {
flog_err_sys(LIB_ERR_SOCKET,
"%s: could not create socket: errno=%d: %s",
__PRETTY_FUNCTION__, errno, safe_strerror(errno));
return -1;
}
sockaddr.sin_family = AF_INET;
sockaddr.sin_addr = addr;
sockaddr.sin_port = htons(port);
if (bind(fd, (struct sockaddr *)&sockaddr, sizeof(sockaddr))) {
char addr_str[INET_ADDRSTRLEN];
pim_inet4_dump("<addr?>", addr, addr_str, sizeof(addr_str));
zlog_warn(
"%s: bind(fd=%d,addr=%s,port=%d,len=%zu) failure: errno=%d: %s",
__PRETTY_FUNCTION__, fd, addr_str, port,
sizeof(sockaddr), errno, safe_strerror(errno));
close(fd);
return -1;
}
/* Needed to obtain destination address from recvmsg() */
{
#if defined(HAVE_IP_PKTINFO)
/* Linux and Solaris IP_PKTINFO */
int opt = 1;
if (setsockopt(fd, IPPROTO_IP, IP_PKTINFO, &opt, sizeof(opt))) {
zlog_warn(
"%s: could not set IP_PKTINFO on socket fd=%d: errno=%d: %s",
__PRETTY_FUNCTION__, fd, errno,
safe_strerror(errno));
}
#elif defined(HAVE_IP_RECVDSTADDR)
/* BSD IP_RECVDSTADDR */
int opt = 1;
if (setsockopt(fd, IPPROTO_IP, IP_RECVDSTADDR, &opt,
sizeof(opt))) {
zlog_warn(
"%s: could not set IP_RECVDSTADDR on socket fd=%d: errno=%d: %s",
__PRETTY_FUNCTION__, fd, errno,
safe_strerror(errno));
}
#else
flog_err(
LIB_ERR_DEVELOPMENT,
"%s %s: missing IP_PKTINFO and IP_RECVDSTADDR: unable to get dst addr from recvmsg()",
__FILE__, __PRETTY_FUNCTION__);
close(fd);
return -1;
#endif
}
{
int reuse = 1;
if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (void *)&reuse,
sizeof(reuse))) {
zlog_warn(
"%s: could not set Reuse Address Option on socket fd=%d: errno=%d: %s",
__PRETTY_FUNCTION__, fd, errno,
safe_strerror(errno));
close(fd);
return -1;
}
}
if (setsockopt(fd, IPPROTO_IP, IP_MULTICAST_TTL, (void *)&mttl,
sizeof(mttl))) {
zlog_warn(
"%s: could not set multicast TTL=%d on socket fd=%d: errno=%d: %s",
__PRETTY_FUNCTION__, mttl, fd, errno,
safe_strerror(errno));
close(fd);
return -1;
}
if (setsockopt_ipv4_multicast_loop(fd, 0)) {
zlog_warn(
"%s: could not disable Multicast Loopback Option on socket fd=%d: errno=%d: %s",
__PRETTY_FUNCTION__, fd, errno, safe_strerror(errno));
close(fd);
return PIM_SOCK_ERR_LOOP;
}
if (setsockopt(fd, IPPROTO_IP, IP_MULTICAST_IF, (void *)&addr,
sizeof(addr))) {
zlog_warn(
"%s: could not set Outgoing Interface Option on socket fd=%d: errno=%d: %s",
__PRETTY_FUNCTION__, fd, errno, safe_strerror(errno));
close(fd);
return -1;
}
{
long flags;
flags = fcntl(fd, F_GETFL, 0);
if (flags < 0) {
zlog_warn(
"%s: could not get fcntl(F_GETFL,O_NONBLOCK) on socket fd=%d: errno=%d: %s",
__PRETTY_FUNCTION__, fd, errno,
safe_strerror(errno));
close(fd);
return -1;
}
if (fcntl(fd, F_SETFL, flags | O_NONBLOCK)) {
zlog_warn(
"%s: could not set fcntl(F_SETFL,O_NONBLOCK) on socket fd=%d: errno=%d: %s",
__PRETTY_FUNCTION__, fd, errno,
safe_strerror(errno));
close(fd);
return -1;
}
}
return fd;
}
/* IPv6 supported version of BGP server socket setup. */
int bgp_socket(struct bgp *bgp, unsigned short port, const char *address)
{
struct addrinfo *ainfo;
struct addrinfo *ainfo_save;
static const struct addrinfo req = {
.ai_family = AF_UNSPEC,
.ai_flags = AI_PASSIVE,
.ai_socktype = SOCK_STREAM,
};
int ret, count;
char port_str[BUFSIZ];
snprintf(port_str, sizeof(port_str), "%d", port);
port_str[sizeof(port_str) - 1] = '\0';
frr_elevate_privs(&bgpd_privs) {
ret = vrf_getaddrinfo(address, port_str, &req, &ainfo_save,
bgp->vrf_id);
}
if (ret != 0) {
flog_err_sys(EC_LIB_SOCKET, "getaddrinfo: %s",
gai_strerror(ret));
return -1;
}
if (bgp_option_check(BGP_OPT_NO_ZEBRA) &&
bgp->vrf_id != VRF_DEFAULT) {
freeaddrinfo(ainfo_save);
return -1;
}
count = 0;
for (ainfo = ainfo_save; ainfo; ainfo = ainfo->ai_next) {
int sock;
if (ainfo->ai_family != AF_INET && ainfo->ai_family != AF_INET6)
continue;
frr_elevate_privs(&bgpd_privs) {
sock = vrf_socket(ainfo->ai_family,
ainfo->ai_socktype,
ainfo->ai_protocol, bgp->vrf_id,
(bgp->inst_type
== BGP_INSTANCE_TYPE_VRF
? bgp->name : NULL));
}
if (sock < 0) {
flog_err_sys(EC_LIB_SOCKET, "socket: %s",
safe_strerror(errno));
continue;
}
/* if we intend to implement ttl-security, this socket needs
* ttl=255 */
sockopt_ttl(ainfo->ai_family, sock, MAXTTL);
ret = bgp_listener(sock, ainfo->ai_addr, ainfo->ai_addrlen,
bgp);
if (ret == 0)
++count;
else
close(sock);
}
freeaddrinfo(ainfo_save);
if (count == 0 && bgp->inst_type != BGP_INSTANCE_TYPE_VRF) {
flog_err(
EC_LIB_SOCKET,
"%s: no usable addresses please check other programs usage of specified port %d",
__func__, port);
flog_err_sys(EC_LIB_SOCKET, "%s: Program cannot continue",
__func__);
exit(-1);
}
return 0;
}
int netlink_interface_addr(struct nlmsghdr *h, ns_id_t ns_id, int startup)
{
int len;
struct ifaddrmsg *ifa;
struct rtattr *tb[IFA_MAX + 1];
struct interface *ifp;
void *addr;
void *broad;
uint8_t flags = 0;
char *label = NULL;
struct zebra_ns *zns;
zns = zebra_ns_lookup(ns_id);
ifa = NLMSG_DATA(h);
if (ifa->ifa_family != AF_INET && ifa->ifa_family != AF_INET6) {
zlog_warn(
"Invalid address family: %u received from kernel interface addr change: %u",
ifa->ifa_family, h->nlmsg_type);
return 0;
}
if (h->nlmsg_type != RTM_NEWADDR && h->nlmsg_type != RTM_DELADDR)
return 0;
len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct ifaddrmsg));
if (len < 0) {
zlog_err("%s: Message received from netlink is of a broken size: %d %zu",
__PRETTY_FUNCTION__,
h->nlmsg_len,
(size_t)NLMSG_LENGTH(sizeof(struct ifaddrmsg)));
return -1;
}
memset(tb, 0, sizeof tb);
netlink_parse_rtattr(tb, IFA_MAX, IFA_RTA(ifa), len);
ifp = if_lookup_by_index_per_ns(zns, ifa->ifa_index);
if (ifp == NULL) {
flog_err(
LIB_ERR_INTERFACE,
"netlink_interface_addr can't find interface by index %d",
ifa->ifa_index);
return -1;
}
if (IS_ZEBRA_DEBUG_KERNEL) /* remove this line to see initial ifcfg */
{
char buf[BUFSIZ];
zlog_debug("netlink_interface_addr %s %s flags 0x%x:",
nl_msg_type_to_str(h->nlmsg_type), ifp->name,
ifa->ifa_flags);
if (tb[IFA_LOCAL])
zlog_debug(" IFA_LOCAL %s/%d",
inet_ntop(ifa->ifa_family,
RTA_DATA(tb[IFA_LOCAL]), buf,
BUFSIZ),
ifa->ifa_prefixlen);
if (tb[IFA_ADDRESS])
zlog_debug(" IFA_ADDRESS %s/%d",
inet_ntop(ifa->ifa_family,
RTA_DATA(tb[IFA_ADDRESS]), buf,
BUFSIZ),
ifa->ifa_prefixlen);
if (tb[IFA_BROADCAST])
zlog_debug(" IFA_BROADCAST %s/%d",
inet_ntop(ifa->ifa_family,
RTA_DATA(tb[IFA_BROADCAST]), buf,
BUFSIZ),
ifa->ifa_prefixlen);
if (tb[IFA_LABEL] && strcmp(ifp->name, RTA_DATA(tb[IFA_LABEL])))
zlog_debug(" IFA_LABEL %s",
(char *)RTA_DATA(tb[IFA_LABEL]));
if (tb[IFA_CACHEINFO]) {
struct ifa_cacheinfo *ci = RTA_DATA(tb[IFA_CACHEINFO]);
zlog_debug(" IFA_CACHEINFO pref %d, valid %d",
ci->ifa_prefered, ci->ifa_valid);
}
}
/* logic copied from iproute2/ip/ipaddress.c:print_addrinfo() */
if (tb[IFA_LOCAL] == NULL)
tb[IFA_LOCAL] = tb[IFA_ADDRESS];
if (tb[IFA_ADDRESS] == NULL)
tb[IFA_ADDRESS] = tb[IFA_LOCAL];
/* local interface address */
addr = (tb[IFA_LOCAL] ? RTA_DATA(tb[IFA_LOCAL]) : NULL);
/* is there a peer address? */
if (tb[IFA_ADDRESS]
&& memcmp(RTA_DATA(tb[IFA_ADDRESS]), RTA_DATA(tb[IFA_LOCAL]),
RTA_PAYLOAD(tb[IFA_ADDRESS]))) {
broad = RTA_DATA(tb[IFA_ADDRESS]);
SET_FLAG(flags, ZEBRA_IFA_PEER);
} else
/* seeking a broadcast address */
broad = (tb[IFA_BROADCAST] ? RTA_DATA(tb[IFA_BROADCAST])
: NULL);
/* addr is primary key, SOL if we don't have one */
if (addr == NULL) {
zlog_debug("%s: NULL address", __func__);
return -1;
}
/* Flags. */
if (ifa->ifa_flags & IFA_F_SECONDARY)
SET_FLAG(flags, ZEBRA_IFA_SECONDARY);
/* Label */
if (tb[IFA_LABEL])
label = (char *)RTA_DATA(tb[IFA_LABEL]);
if (label && strcmp(ifp->name, label) == 0)
label = NULL;
/* Register interface address to the interface. */
if (ifa->ifa_family == AF_INET) {
if (ifa->ifa_prefixlen > IPV4_MAX_BITLEN) {
zlog_err(
"Invalid prefix length: %u received from kernel interface addr change: %u",
ifa->ifa_prefixlen, h->nlmsg_type);
return -1;
}
if (h->nlmsg_type == RTM_NEWADDR)
connected_add_ipv4(ifp, flags, (struct in_addr *)addr,
ifa->ifa_prefixlen,
(struct in_addr *)broad, label);
else
connected_delete_ipv4(
ifp, flags, (struct in_addr *)addr,
ifa->ifa_prefixlen, (struct in_addr *)broad);
}
if (ifa->ifa_family == AF_INET6) {
if (ifa->ifa_prefixlen > IPV6_MAX_BITLEN) {
zlog_err(
"Invalid prefix length: %u received from kernel interface addr change: %u",
ifa->ifa_prefixlen, h->nlmsg_type);
return -1;
}
if (h->nlmsg_type == RTM_NEWADDR) {
/* Only consider valid addresses; we'll not get a
* notification from
* the kernel till IPv6 DAD has completed, but at init
* time, Quagga
* does query for and will receive all addresses.
*/
if (!(ifa->ifa_flags
& (IFA_F_DADFAILED | IFA_F_TENTATIVE)))
connected_add_ipv6(ifp, flags,
(struct in6_addr *)addr,
(struct in6_addr *)broad,
ifa->ifa_prefixlen, label);
} else
connected_delete_ipv6(ifp, (struct in6_addr *)addr,
(struct in6_addr *)broad,
ifa->ifa_prefixlen);
}
return 0;
}
/*
* Obtain the BGP instance that the incoming connection should be processed
* against. This is important because more than one VRF could be using the
* same IP address space. The instance is got by obtaining the device to
* which the incoming connection is bound to. This could either be a VRF
* or it could be an interface, which in turn determines the VRF.
*/
static int bgp_get_instance_for_inc_conn(int sock, struct bgp **bgp_inst)
{
#ifndef SO_BINDTODEVICE
/* only Linux has SO_BINDTODEVICE, but we're in Linux-specific code here
* anyway since the assumption is that the interface name returned by
* getsockopt() is useful in identifying the VRF, particularly with
* Linux's
* VRF l3master device. The whole mechanism is specific to Linux, so...
* when other platforms add VRF support, this will need handling here as
* well. (or, some restructuring) */
*bgp_inst = bgp_get_default();
return !*bgp_inst;
#else
char name[VRF_NAMSIZ + 1];
socklen_t name_len = VRF_NAMSIZ;
struct bgp *bgp;
int rc;
struct listnode *node, *nnode;
*bgp_inst = NULL;
name[0] = '\0';
rc = getsockopt(sock, SOL_SOCKET, SO_BINDTODEVICE, name, &name_len);
if (rc != 0) {
#if defined(HAVE_CUMULUS)
flog_err(EC_LIB_SOCKET,
"[Error] BGP SO_BINDTODEVICE get failed (%s), sock %d",
safe_strerror(errno), sock);
return -1;
#endif
}
if (!strlen(name)) {
*bgp_inst = bgp_get_default();
return 0; /* default instance. */
}
/* First try match to instance; if that fails, check for interfaces. */
bgp = bgp_lookup_by_name(name);
if (bgp) {
if (!bgp->vrf_id) // unexpected
return -1;
*bgp_inst = bgp;
return 0;
}
/* TODO - This will be optimized once interfaces move into the NS */
for (ALL_LIST_ELEMENTS(bm->bgp, node, nnode, bgp)) {
struct interface *ifp;
if (bgp->inst_type == BGP_INSTANCE_TYPE_VIEW)
continue;
ifp = if_lookup_by_name(name, bgp->vrf_id);
if (ifp) {
*bgp_inst = bgp;
return 0;
}
}
/* We didn't match to either an instance or an interface. */
return -1;
#endif
}
static void netlink_vrf_change(struct nlmsghdr *h, struct rtattr *tb,
const char *name)
{
struct ifinfomsg *ifi;
struct rtattr *linkinfo[IFLA_INFO_MAX + 1];
struct rtattr *attr[IFLA_VRF_MAX + 1];
struct vrf *vrf;
struct zebra_vrf *zvrf;
uint32_t nl_table_id;
ifi = NLMSG_DATA(h);
memset(linkinfo, 0, sizeof linkinfo);
parse_rtattr_nested(linkinfo, IFLA_INFO_MAX, tb);
if (!linkinfo[IFLA_INFO_DATA]) {
if (IS_ZEBRA_DEBUG_KERNEL)
zlog_debug(
"%s: IFLA_INFO_DATA missing from VRF message: %s",
__func__, name);
return;
}
memset(attr, 0, sizeof attr);
parse_rtattr_nested(attr, IFLA_VRF_MAX, linkinfo[IFLA_INFO_DATA]);
if (!attr[IFLA_VRF_TABLE]) {
if (IS_ZEBRA_DEBUG_KERNEL)
zlog_debug(
"%s: IFLA_VRF_TABLE missing from VRF message: %s",
__func__, name);
return;
}
nl_table_id = *(uint32_t *)RTA_DATA(attr[IFLA_VRF_TABLE]);
if (h->nlmsg_type == RTM_NEWLINK) {
if (IS_ZEBRA_DEBUG_KERNEL)
zlog_debug("RTM_NEWLINK for VRF %s(%u) table %u", name,
ifi->ifi_index, nl_table_id);
/*
* vrf_get is implied creation if it does not exist
*/
vrf = vrf_get((vrf_id_t)ifi->ifi_index,
name); // It would create vrf
if (!vrf) {
flog_err(LIB_ERR_INTERFACE, "VRF %s id %u not created",
name, ifi->ifi_index);
return;
}
/*
* This is the only place that we get the actual kernel table_id
* being used. We need it to set the table_id of the routes
* we are passing to the kernel.... And to throw some totally
* awesome parties. that too.
*
* At this point we *must* have a zvrf because the vrf_create
* callback creates one. We *must* set the table id
* before the vrf_enable because of( at the very least )
* static routes being delayed for installation until
* during the vrf_enable callbacks.
*/
zvrf = (struct zebra_vrf *)vrf->info;
zvrf->table_id = nl_table_id;
/* Enable the created VRF. */
if (!vrf_enable(vrf)) {
flog_err(LIB_ERR_INTERFACE,
"Failed to enable VRF %s id %u", name,
ifi->ifi_index);
return;
}
} else // h->nlmsg_type == RTM_DELLINK
{
if (IS_ZEBRA_DEBUG_KERNEL)
zlog_debug("RTM_DELLINK for VRF %s(%u)", name,
ifi->ifi_index);
vrf = vrf_lookup_by_id((vrf_id_t)ifi->ifi_index);
if (!vrf) {
zlog_warn("%s: vrf not found", __func__);
return;
}
vrf_delete(vrf);
}
}
static int frrzmq_read_msg(struct thread *t)
{
struct frrzmq_cb **cbp = THREAD_ARG(t);
struct frrzmq_cb *cb;
zmq_msg_t msg;
unsigned partno;
unsigned char read = 0;
int ret, more;
size_t moresz;
if (!cbp)
return 1;
cb = (*cbp);
if (!cb || !cb->zmqsock)
return 1;
while (1) {
zmq_pollitem_t polli = {.socket = cb->zmqsock,
.events = ZMQ_POLLIN};
ret = zmq_poll(&polli, 1, 0);
if (ret < 0)
goto out_err;
if (!(polli.revents & ZMQ_POLLIN))
break;
if (cb->read.cb_msg) {
cb->read.cb_msg(cb->read.arg, cb->zmqsock);
read = 1;
if (cb->read.cancelled) {
frrzmq_check_events(cbp, &cb->write,
ZMQ_POLLOUT);
cb->read.thread = NULL;
if (cb->write.cancelled && !cb->write.thread)
XFREE(MTYPE_ZEROMQ_CB, cb);
return 0;
}
continue;
}
partno = 0;
if (zmq_msg_init(&msg))
goto out_err;
do {
ret = zmq_msg_recv(&msg, cb->zmqsock, ZMQ_NOBLOCK);
if (ret < 0) {
if (errno == EAGAIN)
break;
zmq_msg_close(&msg);
goto out_err;
}
read = 1;
cb->read.cb_part(cb->read.arg, cb->zmqsock, &msg,
partno);
if (cb->read.cancelled) {
zmq_msg_close(&msg);
frrzmq_check_events(cbp, &cb->write,
ZMQ_POLLOUT);
cb->read.thread = NULL;
if (cb->write.cancelled && !cb->write.thread)
XFREE(MTYPE_ZEROMQ_CB, cb);
return 0;
}
/* cb_part may have read additional parts of the
* message; don't use zmq_msg_more here */
moresz = sizeof(more);
more = 0;
ret = zmq_getsockopt(cb->zmqsock, ZMQ_RCVMORE, &more,
&moresz);
if (ret < 0) {
zmq_msg_close(&msg);
goto out_err;
}
partno++;
} while (more);
zmq_msg_close(&msg);
}
if (read)
frrzmq_check_events(cbp, &cb->write, ZMQ_POLLOUT);
funcname_thread_add_read_write(
THREAD_READ, t->master, frrzmq_read_msg, cbp, cb->fd,
&cb->read.thread, t->funcname, t->schedfrom, t->schedfrom_line);
return 0;
out_err:
flog_err(LIB_ERR_ZMQ, "ZeroMQ read error: %s(%d)", strerror(errno),
errno);
if (cb->read.cb_error)
cb->read.cb_error(cb->read.arg, cb->zmqsock);
return 1;
}
int funcname_frrzmq_thread_add_read(struct thread_master *master,
void (*msgfunc)(void *arg, void *zmqsock),
void (*partfunc)(void *arg, void *zmqsock,
zmq_msg_t *msg,
unsigned partnum),
void (*errfunc)(void *arg, void *zmqsock),
void *arg, void *zmqsock,
struct frrzmq_cb **cbp, debugargdef)
{
int fd, events;
size_t len;
struct frrzmq_cb *cb;
if (!cbp)
return -1;
if (!(msgfunc || partfunc) || (msgfunc && partfunc))
return -1;
len = sizeof(fd);
if (zmq_getsockopt(zmqsock, ZMQ_FD, &fd, &len))
return -1;
len = sizeof(events);
if (zmq_getsockopt(zmqsock, ZMQ_EVENTS, &events, &len))
return -1;
if (*cbp)
cb = *cbp;
else {
cb = XCALLOC(MTYPE_ZEROMQ_CB, sizeof(struct frrzmq_cb));
if (!cb)
return -1;
cb->write.cancelled = 1;
*cbp = cb;
}
cb->zmqsock = zmqsock;
cb->fd = fd;
cb->read.arg = arg;
cb->read.cb_msg = msgfunc;
cb->read.cb_part = partfunc;
cb->read.cb_error = errfunc;
cb->read.cancelled = 0;
if (events & ZMQ_POLLIN) {
if (cb->read.thread) {
thread_cancel(cb->read.thread);
cb->read.thread = NULL;
}
funcname_thread_add_event(master, frrzmq_read_msg, cbp, fd,
&cb->read.thread, funcname, schedfrom,
fromln);
} else
funcname_thread_add_read_write(
THREAD_READ, master, frrzmq_read_msg, cbp, fd,
&cb->read.thread, funcname, schedfrom, fromln);
return 0;
}
static int frrzmq_write_msg(struct thread *t)
{
struct frrzmq_cb **cbp = THREAD_ARG(t);
struct frrzmq_cb *cb;
unsigned char written = 0;
int ret;
if (!cbp)
return 1;
cb = (*cbp);
if (!cb || !cb->zmqsock)
return 1;
while (1) {
zmq_pollitem_t polli = {.socket = cb->zmqsock,
.events = ZMQ_POLLOUT};
ret = zmq_poll(&polli, 1, 0);
if (ret < 0)
goto out_err;
if (!(polli.revents & ZMQ_POLLOUT))
break;
if (cb->write.cb_msg) {
cb->write.cb_msg(cb->write.arg, cb->zmqsock);
written = 1;
if (cb->write.cancelled) {
frrzmq_check_events(cbp, &cb->read, ZMQ_POLLIN);
cb->write.thread = NULL;
if (cb->read.cancelled && !cb->read.thread)
XFREE(MTYPE_ZEROMQ_CB, cb);
return 0;
}
continue;
}
}
if (written)
frrzmq_check_events(cbp, &cb->read, ZMQ_POLLIN);
funcname_thread_add_read_write(THREAD_WRITE, t->master,
frrzmq_write_msg, cbp, cb->fd,
&cb->write.thread, t->funcname,
t->schedfrom, t->schedfrom_line);
return 0;
out_err:
flog_err(LIB_ERR_ZMQ, "ZeroMQ write error: %s(%d)", strerror(errno),
errno);
if (cb->write.cb_error)
cb->write.cb_error(cb->write.arg, cb->zmqsock);
return 1;
}
int funcname_frrzmq_thread_add_write(struct thread_master *master,
void (*msgfunc)(void *arg, void *zmqsock),
void (*errfunc)(void *arg, void *zmqsock),
void *arg, void *zmqsock,
struct frrzmq_cb **cbp, debugargdef)
{
int fd, events;
size_t len;
struct frrzmq_cb *cb;
if (!cbp)
return -1;
if (!msgfunc)
return -1;
len = sizeof(fd);
if (zmq_getsockopt(zmqsock, ZMQ_FD, &fd, &len))
return -1;
len = sizeof(events);
if (zmq_getsockopt(zmqsock, ZMQ_EVENTS, &events, &len))
return -1;
if (*cbp)
cb = *cbp;
else {
cb = XCALLOC(MTYPE_ZEROMQ_CB, sizeof(struct frrzmq_cb));
if (!cb)
return -1;
cb->read.cancelled = 1;
*cbp = cb;
}
cb->zmqsock = zmqsock;
cb->fd = fd;
cb->write.arg = arg;
cb->write.cb_msg = msgfunc;
cb->write.cb_part = NULL;
cb->write.cb_error = errfunc;
cb->write.cancelled = 0;
if (events & ZMQ_POLLOUT) {
if (cb->write.thread) {
thread_cancel(cb->write.thread);
cb->write.thread = NULL;
}
funcname_thread_add_event(master, frrzmq_write_msg, cbp, fd,
&cb->write.thread, funcname,
schedfrom, fromln);
} else
funcname_thread_add_read_write(
THREAD_WRITE, master, frrzmq_write_msg, cbp, fd,
&cb->write.thread, funcname, schedfrom, fromln);
return 0;
}
