* nor RFC2385 (which requires a kernel patch on 2.4 kernels).
*/
-char const *cvs_id_bgp = "$Id: bgp.c,v 1.9 2004/12/16 08:49:52 bodea Exp $";
+char const *cvs_id_bgp = "$Id: bgp.c,v 1.12 2005/09/02 23:39:36 bodea Exp $";
#include <stdlib.h>
#include <unistd.h>
static int bgp_send_update(struct bgp_peer *peer);
static int bgp_send_notification(struct bgp_peer *peer, uint8_t code,
uint8_t subcode);
+static int bgp_send_notification_full(struct bgp_peer *peer, uint8_t code,
+ uint8_t subcode, char *notification_data, uint16_t data_len);
static uint16_t our_as;
static struct bgp_route_list *bgp_routes = 0;
return 0;
}
+
+ peer->edata.type = FD_TYPE_BGP;
+ peer->edata.index = i;
+ peer->events = 0;
}
if (as < 1)
peer->inbuf->done = 0;
peer->cli_flag = 0;
+ peer->events = 0;
if (peer->state != peer->next_state)
{
LOG(4, 0, 0, "%s BGP routing\n", enable ? "Enabled" : "Suspended");
}
-/* return a bitmask indicating if the socket should be added to the
- read set (1) and or write set (2) for select */
-int bgp_select_state(struct bgp_peer *peer)
+#ifdef HAVE_EPOLL
+# include <sys/epoll.h>
+#else
+# include "fake_epoll.h"
+#endif
+
+/* return a bitmask of the events required to poll this peer's fd */
+int bgp_set_poll()
{
- int flags = 0;
+ int i;
if (!bgp_configured)
return 0;
- if (peer->state == Disabled || peer->state == Idle)
- return 0;
+ for (i = 0; i < BGP_NUM_PEERS; i++)
+ {
+ struct bgp_peer *peer = &bgp_peers[i];
+ int events = 0;
+
+ if (peer->state == Disabled || peer->state == Idle)
+ continue;
+
+ if (peer->inbuf->done < BGP_MAX_PACKET_SIZE)
+ events |= EPOLLIN;
- if (peer->inbuf->done < BGP_MAX_PACKET_SIZE)
- flags |= 1;
+ if (peer->state == Connect || /* connection in progress */
+ peer->update_routes || /* routing updates */
+ peer->outbuf->packet.header.len) /* pending output */
+ events |= EPOLLOUT;
- if (peer->state == Connect || /* connection in progress */
- peer->update_routes || /* routing updates */
- peer->outbuf->packet.header.len) /* pending output */
- flags |= 2;
+ if (peer->events != events)
+ {
+ struct epoll_event ev;
- return flags;
+ ev.events = peer->events = events;
+ ev.data.ptr = &peer->edata;
+ epoll_ctl(epollfd, EPOLL_CTL_MOD, peer->sock, &ev);
+ }
+ }
+
+ return 1;
}
-/* process bgp peer */
-int bgp_process(struct bgp_peer *peer, int readable, int writable)
+/* process bgp events/timers */
+int bgp_process(uint32_t events[])
{
+ int i;
+
if (!bgp_configured)
return 0;
- if (*peer->name && peer->cli_flag == BGP_CLI_RESTART)
- return bgp_restart(peer);
+ for (i = 0; i < BGP_NUM_PEERS; i++)
+ {
+ struct bgp_peer *peer = &bgp_peers[i];
- if (peer->state == Disabled)
- return 1;
+ if (*peer->name && peer->cli_flag == BGP_CLI_RESTART)
+ {
+ bgp_restart(peer);
+ continue;
+ }
- if (peer->cli_flag)
- {
- switch (peer->cli_flag)
+ if (peer->state == Disabled)
+ continue;
+
+ if (peer->cli_flag)
{
- case BGP_CLI_SUSPEND:
- if (peer->routing)
+ switch (peer->cli_flag)
{
- peer->routing = 0;
- if (peer->state == Established)
- peer->update_routes = 1;
- }
+ case BGP_CLI_SUSPEND:
+ if (peer->routing)
+ {
+ peer->routing = 0;
+ if (peer->state == Established)
+ peer->update_routes = 1;
+ }
- break;
+ break;
- case BGP_CLI_ENABLE:
- if (!peer->routing)
- {
- peer->routing = 1;
- if (peer->state == Established)
- peer->update_routes = 1;
+ case BGP_CLI_ENABLE:
+ if (!peer->routing)
+ {
+ peer->routing = 1;
+ if (peer->state == Established)
+ peer->update_routes = 1;
+ }
+
+ break;
}
- break;
+ peer->cli_flag = 0;
}
- peer->cli_flag = 0;
- }
-
- /* handle empty/fill of buffers */
- if (writable)
- {
- int r = 1;
- if (peer->state == Connect)
- r = bgp_handle_connect(peer);
- else if (peer->outbuf->packet.header.len)
- r = bgp_write(peer);
+ /* handle empty/fill of buffers */
+ if (events[i] & EPOLLOUT)
+ {
+ int r = 1;
+ if (peer->state == Connect)
+ r = bgp_handle_connect(peer);
+ else if (peer->outbuf->packet.header.len)
+ r = bgp_write(peer);
- if (!r)
- return 0;
- }
+ if (!r)
+ continue;
+ }
- if (readable)
- {
- if (!bgp_read(peer))
- return 0;
- }
+ if (events[i] & (EPOLLIN|EPOLLHUP))
+ {
+ if (!bgp_read(peer))
+ continue;
+ }
- /* process input buffer contents */
- while (peer->inbuf->done >= sizeof(peer->inbuf->packet.header)
- && !peer->outbuf->packet.header.len) /* may need to queue a response */
- {
- if (bgp_handle_input(peer) < 0)
- return 0;
- }
+ /* process input buffer contents */
+ while (peer->inbuf->done >= sizeof(peer->inbuf->packet.header)
+ && !peer->outbuf->packet.header.len) /* may need to queue a response */
+ {
+ if (bgp_handle_input(peer) < 0)
+ continue;
+ }
- /* process pending updates */
- if (peer->update_routes
- && !peer->outbuf->packet.header.len) /* ditto */
- {
- if (!bgp_send_update(peer))
- return 0;
- }
+ /* process pending updates */
+ if (peer->update_routes
+ && !peer->outbuf->packet.header.len) /* ditto */
+ {
+ if (!bgp_send_update(peer))
+ continue;
+ }
- /* process timers */
- if (peer->state == Established)
- {
- if (time_now > peer->expire_time)
+ /* process timers */
+ if (peer->state == Established)
{
- LOG(1, 0, 0, "No message from BGP peer %s in %ds\n",
- peer->name, peer->hold);
+ if (time_now > peer->expire_time)
+ {
+ LOG(1, 0, 0, "No message from BGP peer %s in %ds\n",
+ peer->name, peer->hold);
- bgp_send_notification(peer, BGP_ERR_HOLD_TIMER_EXP, 0);
- return 0;
- }
+ bgp_send_notification(peer, BGP_ERR_HOLD_TIMER_EXP, 0);
+ continue;
+ }
- if (time_now > peer->keepalive_time && !peer->outbuf->packet.header.len)
- bgp_send_keepalive(peer);
- }
- else if (peer->state == Idle)
- {
- if (time_now > peer->retry_time)
- return bgp_connect(peer);
- }
- else if (time_now > peer->state_time + BGP_STATE_TIME)
- {
- LOG(1, 0, 0, "%s timer expired for BGP peer %s\n",
- bgp_state_str(peer->state), peer->name);
+ if (time_now > peer->keepalive_time && !peer->outbuf->packet.header.len)
+ bgp_send_keepalive(peer);
+ }
+ else if (peer->state == Idle)
+ {
+ if (time_now > peer->retry_time)
+ bgp_connect(peer);
+ }
+ else if (time_now > peer->state_time + BGP_STATE_TIME)
+ {
+ LOG(1, 0, 0, "%s timer expired for BGP peer %s\n",
+ bgp_state_str(peer->state), peer->name);
- return bgp_restart(peer);
+ bgp_restart(peer);
+ }
}
return 1;
{
static int bgp_port = 0;
struct sockaddr_in addr;
+ struct epoll_event ev;
if (!bgp_port)
{
return 0;
}
+ /* add to poll set */
+ ev.events = peer->events = EPOLLOUT;
+ ev.data.ptr = &peer->edata;
+ epoll_ctl(epollfd, EPOLL_CTL_ADD, peer->sock, &ev);
+
/* set to non-blocking */
fcntl(peer->sock, F_SETFL, fcntl(peer->sock, F_GETFL, 0) | O_NONBLOCK);
LOG(4, 0, 0, "BGP peer %s: state Active\n", inet_ntoa(addr.sin_addr));
+ peer->handle_ipv6_routes = 0;
+
return bgp_send_open(peer);
}
static int bgp_handle_connect(struct bgp_peer *peer)
{
int err = 0;
- int len = sizeof(int);
+ socklen_t len = sizeof(int);
getsockopt(peer->sock, SOL_SOCKET, SO_ERROR, &err, &len);
if (err)
{
LOG(4, 0, 0, "BGP peer %s: state Active\n", peer->name);
+ peer->handle_ipv6_routes = 0;
+
return bgp_send_open(peer);
}
struct bgp_data_open data;
int hold;
int i;
+ off_t param_offset, capability_offset;
+ struct bgp_opt_param *param;
+ uint8_t capabilities_len;
+ char *capabilities = NULL;
+ struct bgp_capability *capability;
+ struct bgp_mp_cap_param *mp_cap;
for (i = 0; i < sizeof(p->header.marker); i++)
{
if (peer->keepalive * 3 > peer->hold)
peer->keepalive = peer->hold / 3;
+ /* check for optional parameters */
+ /* 2 is for the size of type + len (both uint8_t) */
+ for (param_offset = 0;
+ param_offset < data.opt_len;
+ param_offset += 2 + param->len)
+ {
+ param = (struct bgp_opt_param *)(&data.opt_params + param_offset);
+
+ /* sensible check */
+ if (data.opt_len - param_offset < 2
+ || param->len > data.opt_len - param_offset - 2)
+ {
+ LOG(1, 0, 0, "Malformed Optional Parameter list from BGP peer %s\n",
+ peer->name);
+
+ bgp_send_notification(peer, BGP_ERR_OPEN, BGP_ERR_UNSPEC);
+ return 0;
+ }
+
+ /* we know only one parameter type */
+ if (param->type != BGP_CAPABILITY_PARAM_TYPE)
+ {
+ LOG(1, 0, 0, "Unsupported Optional Parameter type %d from BGP peer %s\n",
+ param->type, peer->name);
+
+ bgp_send_notification(peer, BGP_ERR_OPEN, BGP_ERR_OPN_UNSUP_PARAM);
+ return 0;
+ }
+
+ capabilities_len = param->len;
+ capabilities = (char *)¶m->value;
+ }
+
+ /* look for BGP multiprotocol capability */
+ if (capabilities)
+ {
+ for (capability_offset = 0;
+ capability_offset < capabilities_len;
+ capability_offset += 2 + capability->len)
+ {
+ capability = (struct bgp_capability *)(capabilities + capability_offset);
+
+ /* sensible check */
+ if (capabilities_len - capability_offset < 2
+ || capability->len > capabilities_len - capability_offset - 2)
+ {
+ LOG(1, 0, 0, "Malformed Capabilities list from BGP peer %s\n",
+ peer->name);
+
+ bgp_send_notification(peer, BGP_ERR_OPEN, BGP_ERR_UNSPEC);
+ return 0;
+ }
+
+ /* we only know one capability code */
+ if (capability->code != BGP_CAP_CODE_MP
+ && capability->len != sizeof(struct bgp_mp_cap_param))
+ {
+ LOG(4, 0, 0, "Unsupported Capability code %d from BGP peer %s\n",
+ capability->code, peer->name);
+
+ bgp_send_notification_full(peer, BGP_ERR_OPEN, BGP_ERR_OPN_UNSUP_CAP,
+ (char *)capability, 2 + capability->len);
+ /* we don't terminate, still; we just jump to the next one */
+ continue;
+ }
+
+ mp_cap = (struct bgp_mp_cap_param *)&capability->value;
+ /* the only <AFI, SAFI> tuple we support */
+ if (mp_cap->afi != AF_INET6 && mp_cap->safi != BGP_MP_SAFI_UNICAST)
+ {
+ LOG(4, 0, 0, "Unsupported multiprotocol AFI %d and SAFI %d from BGP peer %s\n",
+ mp_cap->afi, mp_cap->safi, peer->name);
+
+ bgp_send_notification_full(peer, BGP_ERR_OPEN, BGP_ERR_OPN_UNSUP_CAP,
+ (char *)capability, 2 + capability->len);
+ /* we don't terminate, still; we just jump to the next one */
+ continue;
+ }
+
+ peer->handle_ipv6_routes = 1;
+ }
+ }
+
/* next transition requires an exchange of keepalives */
bgp_send_keepalive(peer);
-
- /* FIXME: may need to check for optional params */
}
break;
if (notification->error_code == BGP_ERR_CEASE)
{
LOG(4, 0, 0, "BGP peer %s sent CEASE\n", peer->name);
- bgp_halt(peer);
+ bgp_restart(peer);
return 0;
}
/* send/buffer NOTIFICATION message */
static int bgp_send_notification(struct bgp_peer *peer, uint8_t code,
uint8_t subcode)
+{
+ return bgp_send_notification_full(peer, code, subcode, NULL, 0);
+}
+
+static int bgp_send_notification_full(struct bgp_peer *peer, uint8_t code,
+ uint8_t subcode, char *notification_data, uint16_t data_len)
{
struct bgp_data_notification data;
uint16_t len = 0;
data.error_subcode = subcode;
len += sizeof(data.error_code);
+ memcpy(data.data, notification_data, data_len);
+ len += data_len;
+
memset(peer->outbuf->packet.header.marker, 0xff,
sizeof(peer->outbuf->packet.header.marker));