#include "queue.h"
#include "hash.h"
#include "filter.h"
+#include "ns.h"
+#include "lib_errors.h"
#include "bgpd/bgpd.h"
#include "bgpd/bgp_open.h"
#include "bgpd/bgp_fsm.h"
#include "bgpd/bgp_attr.h"
#include "bgpd/bgp_debug.h"
+#include "bgpd/bgp_errors.h"
#include "bgpd/bgp_network.h"
extern struct zebra_privs_t bgpd_privs;
-static int bgp_bind(struct peer *);
+static char *bgp_get_bound_name(struct peer *peer);
/* BGP listening socket. */
struct bgp_listener {
int fd;
union sockunion su;
struct thread *thread;
+ struct bgp *bgp;
};
/*
int ret = -1;
#if HAVE_DECL_TCP_MD5SIG
- if (bgpd_privs.change(ZPRIVS_RAISE)) {
- zlog_err("%s: could not raise privs", __func__);
- return ret;
+ frr_elevate_privs(&bgpd_privs) {
+ ret = bgp_md5_set_socket(socket, su, password);
}
-
- ret = bgp_md5_set_socket(socket, su, password);
-
- if (bgpd_privs.change(ZPRIVS_LOWER))
- zlog_err("%s: could not lower privs", __func__);
#endif /* HAVE_TCP_MD5SIG */
return ret;
int ret = 0;
struct bgp_listener *listener;
- if (bgpd_privs.change(ZPRIVS_RAISE)) {
- zlog_err("%s: could not raise privs", __func__);
- return -1;
- }
-
+ frr_elevate_privs(&bgpd_privs) {
/* Set or unset the password on the listen socket(s). Outbound
- * connections
- * are taken care of in bgp_connect() below.
+ * connections are taken care of in bgp_connect() below.
*/
- for (ALL_LIST_ELEMENTS_RO(bm->listen_sockets, node, listener))
- if (listener->su.sa.sa_family == peer->su.sa.sa_family) {
- ret = bgp_md5_set_socket(listener->fd, &peer->su,
- password);
- break;
- }
-
- if (bgpd_privs.change(ZPRIVS_LOWER))
- zlog_err("%s: could not lower privs", __func__);
-
+ for (ALL_LIST_ELEMENTS_RO(bm->listen_sockets, node, listener))
+ if (listener->su.sa.sa_family
+ == peer->su.sa.sa_family) {
+ ret = bgp_md5_set_socket(listener->fd,
+ &peer->su, password);
+ break;
+ }
+ }
return ret;
}
if (!peer->gtsm_hops && (peer_sort(peer) == BGP_PEER_EBGP)) {
ret = sockopt_ttl(peer->su.sa.sa_family, bgp_sock, peer->ttl);
if (ret) {
- zlog_err(
+ flog_err(
+ LIB_ERR_SOCKET,
"%s: Can't set TxTTL on peer (rtrid %s) socket, err = %d",
- __func__, inet_ntop(AF_INET, &peer->remote_id,
- buf, sizeof(buf)),
+ __func__,
+ inet_ntop(AF_INET, &peer->remote_id, buf,
+ sizeof(buf)),
errno);
return ret;
}
*/
ret = sockopt_ttl(peer->su.sa.sa_family, bgp_sock, MAXTTL);
if (ret) {
- zlog_err(
+ flog_err(
+ LIB_ERR_SOCKET,
"%s: Can't set TxTTL on peer (rtrid %s) socket, err = %d",
- __func__, inet_ntop(AF_INET, &peer->remote_id,
- buf, sizeof(buf)),
+ __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) {
- zlog_err(
+ flog_err(
+ LIB_ERR_SOCKET,
"%s: Can't set MinTTL on peer (rtrid %s) socket, err = %d",
- __func__, inet_ntop(AF_INET, &peer->remote_id,
- buf, sizeof(buf)),
+ __func__,
+ inet_ntop(AF_INET, &peer->remote_id, buf,
+ sizeof(buf)),
errno);
return ret;
}
rc = getsockopt(sock, SOL_SOCKET, SO_BINDTODEVICE, name, &name_len);
if (rc != 0) {
#if defined(HAVE_CUMULUS)
- zlog_err("[Error] BGP SO_BINDTODEVICE get failed (%s), sock %d",
- safe_strerror(errno), sock);
+ flog_err(
+ LIB_ERR_SOCKET,
+ "[Error] BGP SO_BINDTODEVICE get failed (%s), sock %d",
+ safe_strerror(errno), sock);
return -1;
#endif
}
/* Register accept thread. */
accept_sock = THREAD_FD(thread);
if (accept_sock < 0) {
- zlog_err("accept_sock is nevative value %d", accept_sock);
+ flog_err_sys(LIB_ERR_SOCKET, "accept_sock is nevative value %d",
+ accept_sock);
return -1;
}
listener->thread = NULL;
+
thread_add_read(bm->master, bgp_accept, listener, accept_sock,
&listener->thread);
/* Accept client connection. */
bgp_sock = sockunion_accept(accept_sock, &su);
if (bgp_sock < 0) {
- zlog_err("[Error] BGP socket accept failed (%s)",
- safe_strerror(errno));
+ flog_err_sys(LIB_ERR_SOCKET,
+ "[Error] BGP socket accept failed (%s)",
+ safe_strerror(errno));
return -1;
}
set_nonblocking(bgp_sock);
- /* Obtain BGP instance this connection is meant for. */
- if (bgp_get_instance_for_inc_conn(bgp_sock, &bgp)) {
+ /* Obtain BGP instance this connection is meant for.
+ * - if it is a VRF netns sock, then BGP is in listener structure
+ * - otherwise, the bgp instance need to be demultiplexed
+ */
+ if (listener->bgp)
+ bgp = listener->bgp;
+ else if (bgp_get_instance_for_inc_conn(bgp_sock, &bgp)) {
if (bgp_debug_neighbor_events(NULL))
zlog_debug(
"[Event] Could not get instance for incoming conn from %s",
peer->doppelganger = peer1;
peer1->doppelganger = peer;
peer->fd = bgp_sock;
- bgp_bind(peer);
+ vrf_bind(peer->bgp->vrf_id, bgp_sock, bgp_get_bound_name(peer));
bgp_fsm_change_status(peer, Active);
BGP_TIMER_OFF(peer->t_start); /* created in peer_create() */
}
/* BGP socket bind. */
-static int bgp_bind(struct peer *peer)
+static char *bgp_get_bound_name(struct peer *peer)
{
-#ifdef SO_BINDTODEVICE
- int ret;
- int myerrno;
char *name = NULL;
- /* If not bound to an interface or part of a VRF, we don't care. */
- if (!peer->bgp->vrf_id && !peer->ifname && !peer->conf_if)
- return 0;
+ if (!peer)
+ return NULL;
+
+ if ((peer->bgp->vrf_id == VRF_DEFAULT) && !peer->ifname
+ && !peer->conf_if)
+ return NULL;
if (peer->su.sa.sa_family != AF_INET
&& peer->su.sa.sa_family != AF_INET6)
- return 0; // unexpected
+ return NULL; // unexpected
/* For IPv6 peering, interface (unnumbered or link-local with interface)
* takes precedence over VRF. For IPv4 peering, explicit interface or
else
name = peer->ifname ? peer->ifname : peer->bgp->name;
- if (!name)
- return 0;
-
- if (bgp_debug_neighbor_events(peer))
- zlog_debug("%s Binding to interface %s", peer->host, name);
-
- if (bgpd_privs.change(ZPRIVS_RAISE))
- zlog_err("bgp_bind: could not raise privs");
-
- ret = setsockopt(peer->fd, SOL_SOCKET, SO_BINDTODEVICE, name,
- strlen(name));
- myerrno = errno;
-
- if (bgpd_privs.change(ZPRIVS_LOWER))
- zlog_err("bgp_bind: could not lower privs");
-
- if (ret < 0) {
- if (bgp_debug_neighbor_events(peer))
- zlog_debug("bind to interface %s failed, errno=%d",
- name, myerrno);
- return ret;
- }
-#endif /* SO_BINDTODEVICE */
- return 0;
+ return name;
}
static int bgp_update_address(struct interface *ifp, const union sockunion *dst,
zlog_debug("Peer address not learnt: Returning from connect");
return 0;
}
+ frr_elevate_privs(&bgpd_privs) {
/* Make socket for the peer. */
- peer->fd = sockunion_socket(&peer->su);
+ peer->fd = vrf_sockunion_socket(&peer->su, peer->bgp->vrf_id,
+ bgp_get_bound_name(peer));
+ }
if (peer->fd < 0)
return -1;
peer->host, safe_strerror(errno));
#ifdef IPTOS_PREC_INTERNETCONTROL
- if (bgpd_privs.change(ZPRIVS_RAISE))
- zlog_err("%s: could not raise privs", __func__);
- if (sockunion_family(&peer->su) == AF_INET)
- setsockopt_ipv4_tos(peer->fd, IPTOS_PREC_INTERNETCONTROL);
- else if (sockunion_family(&peer->su) == AF_INET6)
- setsockopt_ipv6_tclass(peer->fd, IPTOS_PREC_INTERNETCONTROL);
- if (bgpd_privs.change(ZPRIVS_LOWER))
- zlog_err("%s: could not lower privs", __func__);
+ frr_elevate_privs(&bgpd_privs) {
+ if (sockunion_family(&peer->su) == AF_INET)
+ setsockopt_ipv4_tos(peer->fd,
+ IPTOS_PREC_INTERNETCONTROL);
+ else if (sockunion_family(&peer->su) == AF_INET6)
+ setsockopt_ipv6_tclass(peer->fd,
+ IPTOS_PREC_INTERNETCONTROL);
+ }
#endif
if (peer->password)
bgp_md5_set_connect(peer->fd, &peer->su, peer->password);
- /* Bind socket. */
- bgp_bind(peer);
-
/* Update source bind. */
if (bgp_update_source(peer) < 0) {
return connect_error;
if (bgp_nexthop_set(peer->su_local, peer->su_remote, &peer->nexthop,
peer)) {
#if defined(HAVE_CUMULUS)
- zlog_err(
+ flog_err(
+ BGP_ERR_NH_UPD,
"%s: nexthop_set failed, resetting connection - intf %p",
peer->host, peer->nexthop.ifp);
return -1;
#endif
}
-
return 0;
}
-static int bgp_listener(int sock, struct sockaddr *sa, socklen_t salen)
+static int bgp_listener(int sock, struct sockaddr *sa, socklen_t salen,
+ struct bgp *bgp)
{
struct bgp_listener *listener;
int ret, en;
sockopt_reuseaddr(sock);
sockopt_reuseport(sock);
- if (bgpd_privs.change(ZPRIVS_RAISE))
- zlog_err("%s: could not raise privs", __func__);
+ frr_elevate_privs(&bgpd_privs) {
#ifdef IPTOS_PREC_INTERNETCONTROL
- if (sa->sa_family == AF_INET)
- setsockopt_ipv4_tos(sock, IPTOS_PREC_INTERNETCONTROL);
- else if (sa->sa_family == AF_INET6)
- setsockopt_ipv6_tclass(sock, IPTOS_PREC_INTERNETCONTROL);
+ if (sa->sa_family == AF_INET)
+ setsockopt_ipv4_tos(sock, IPTOS_PREC_INTERNETCONTROL);
+ else if (sa->sa_family == AF_INET6)
+ setsockopt_ipv6_tclass(sock,
+ IPTOS_PREC_INTERNETCONTROL);
#endif
- sockopt_v6only(sa->sa_family, sock);
+ sockopt_v6only(sa->sa_family, sock);
- ret = bind(sock, sa, salen);
- en = errno;
- if (bgpd_privs.change(ZPRIVS_LOWER))
- zlog_err("%s: could not lower privs", __func__);
+ ret = bind(sock, sa, salen);
+ en = errno;
+ }
if (ret < 0) {
- zlog_err("bind: %s", safe_strerror(en));
+ flog_err_sys(LIB_ERR_SOCKET, "bind: %s", safe_strerror(en));
return ret;
}
ret = listen(sock, SOMAXCONN);
if (ret < 0) {
- zlog_err("listen: %s", safe_strerror(errno));
+ flog_err_sys(LIB_ERR_SOCKET, "listen: %s",
+ safe_strerror(errno));
return ret;
}
- listener = XMALLOC(MTYPE_BGP_LISTENER, sizeof(*listener));
+ listener = XCALLOC(MTYPE_BGP_LISTENER, sizeof(*listener));
listener->fd = sock;
+
+ /* this socket needs a change of ns. record bgp back pointer */
+ if (bgp->vrf_id != VRF_DEFAULT && vrf_is_mapped_on_netns(
+ vrf_lookup_by_id(bgp->vrf_id)))
+ listener->bgp = bgp;
+
memcpy(&listener->su, sa, salen);
listener->thread = NULL;
thread_add_read(bm->master, bgp_accept, listener, sock,
}
/* IPv6 supported version of BGP server socket setup. */
-int bgp_socket(unsigned short port, const char *address)
+int bgp_socket(struct bgp *bgp, unsigned short port, const char *address)
{
struct addrinfo *ainfo;
struct addrinfo *ainfo_save;
snprintf(port_str, sizeof(port_str), "%d", port);
port_str[sizeof(port_str) - 1] = '\0';
- ret = getaddrinfo(address, port_str, &req, &ainfo_save);
+ frr_elevate_privs(&bgpd_privs) {
+ ret = vrf_getaddrinfo(address, port_str, &req, &ainfo_save,
+ bgp->vrf_id);
+ }
if (ret != 0) {
- zlog_err("getaddrinfo: %s", gai_strerror(ret));
+ flog_err_sys(LIB_ERR_SOCKET, "getaddrinfo: %s",
+ gai_strerror(ret));
return -1;
}
if (ainfo->ai_family != AF_INET && ainfo->ai_family != AF_INET6)
continue;
- sock = socket(ainfo->ai_family, ainfo->ai_socktype,
- ainfo->ai_protocol);
+ 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) {
- zlog_err("socket: %s", safe_strerror(errno));
+ flog_err_sys(LIB_ERR_SOCKET, "socket: %s",
+ safe_strerror(errno));
continue;
}
* ttl=255 */
sockopt_ttl(ainfo->ai_family, sock, MAXTTL);
- ret = bgp_listener(sock, ainfo->ai_addr, ainfo->ai_addrlen);
+ ret = bgp_listener(sock, ainfo->ai_addr, ainfo->ai_addrlen,
+ bgp);
if (ret == 0)
++count;
else
close(sock);
}
freeaddrinfo(ainfo_save);
- if (count == 0) {
- zlog_err("%s: no usable addresses please check other programs usage of specified port %d",
- __func__, port);
- zlog_err("%s: Program cannot continue", __func__);
+ if (count == 0 && bgp->inst_type != BGP_INSTANCE_TYPE_VRF) {
+ flog_err(
+ LIB_ERR_SOCKET,
+ "%s: no usable addresses please check other programs usage of specified port %d",
+ __func__, port);
+ flog_err_sys(LIB_ERR_SOCKET, "%s: Program cannot continue",
+ __func__);
exit(-1);
}
return 0;
}
+/* this function closes vrf socket
+ * this should be called only for vrf socket with netns backend
+ */
+void bgp_close_vrf_socket(struct bgp *bgp)
+{
+ struct listnode *node, *next;
+ struct bgp_listener *listener;
+
+ if (!bgp)
+ return;
+
+ if (bm->listen_sockets == NULL)
+ return;
+
+ for (ALL_LIST_ELEMENTS(bm->listen_sockets, node, next, listener)) {
+ if (listener->bgp == bgp) {
+ thread_cancel(listener->thread);
+ close(listener->fd);
+ listnode_delete(bm->listen_sockets, listener);
+ XFREE(MTYPE_BGP_LISTENER, listener);
+ }
+ }
+}
+
+/* this function closes main socket
+ */
void bgp_close(void)
{
struct listnode *node, *next;
return;
for (ALL_LIST_ELEMENTS(bm->listen_sockets, node, next, listener)) {
+ if (listener->bgp)
+ continue;
thread_cancel(listener->thread);
close(listener->fd);
listnode_delete(bm->listen_sockets, listener);