X-Git-Url: http://git.sameswireless.fr/l2tpns.git/blobdiff_plain/bc5b25832a728dea4346782d84d63128035bcfaf..f2e59e98817d4959af549abccf4e798e6184f6d4:/bgp.c

diff --git a/bgp.c b/bgp.c
index cf8ab7e..f12efb1 100644
--- a/bgp.c
+++ b/bgp.c
@@ -10,7 +10,7 @@
  *   nor RFC2385 (which requires a kernel patch on 2.4 kernels).
  */
 
-char const *cvs_id_bgp = "$Id: bgp.c,v 1.8 2004/11/29 02:17:17 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>
@@ -34,7 +34,7 @@ static struct bgp_route_list *bgp_insert_route(struct bgp_route_list *head,
     struct bgp_route_list *new);
 
 static void bgp_free_routes(struct bgp_route_list *routes);
-static char const *bgp_msg_type_str(u8 type);
+static char const *bgp_msg_type_str(uint8_t type);
 static int bgp_connect(struct bgp_peer *peer);
 static int bgp_handle_connect(struct bgp_peer *peer);
 static int bgp_write(struct bgp_peer *peer);
@@ -43,9 +43,12 @@ static int bgp_handle_input(struct bgp_peer *peer);
 static int bgp_send_open(struct bgp_peer *peer);
 static int bgp_send_keepalive(struct bgp_peer *peer);
 static int bgp_send_update(struct bgp_peer *peer);
-static int bgp_send_notification(struct bgp_peer *peer, u8 code, u8 subcode);
+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 u16 our_as;
+static uint16_t our_as;
 static struct bgp_route_list *bgp_routes = 0;
 
 int bgp_configured = 0;
@@ -74,6 +77,10 @@ int bgp_setup(int as)
 
 	    return 0;
 	}
+
+	peer->edata.type = FD_TYPE_BGP;
+	peer->edata.index = i;
+	peer->events = 0;
     }
 
     if (as < 1)
@@ -89,7 +96,8 @@ int bgp_setup(int as)
 }
 
 /* start connection with a peer */
-int bgp_start(struct bgp_peer *peer, char *name, int as, int keepalive, int hold, int enable)
+int bgp_start(struct bgp_peer *peer, char *name, int as, int keepalive,
+    int hold, int enable)
 {
     struct hostent *h;
     int ibgp;
@@ -98,8 +106,8 @@ int bgp_start(struct bgp_peer *peer, char *name, int as, int keepalive, int hold
     char path_attrs[64];
     char *p = path_attrs;
     in_addr_t ip;
-    u32 metric = htonl(BGP_METRIC);
-    u32 no_export = htonl(BGP_COMMUNITY_NO_EXPORT);
+    uint32_t metric = htonl(BGP_METRIC);
+    uint32_t no_export = htonl(BGP_COMMUNITY_NO_EXPORT);
 
     if (!our_as)
 	return 0;
@@ -170,9 +178,9 @@ int bgp_start(struct bgp_peer *peer, char *name, int as, int keepalive, int hold
     {
 	/* just our AS */
 	struct {
-	    u8 type;
-	    u8 len;
-	    u16 value;
+	    uint8_t type;
+	    uint8_t len;
+	    uint16_t value;
 	} as_path = {
 	    BGP_PATH_ATTR_CODE_AS_PATH_AS_SEQUENCE,
 	    1,
@@ -204,7 +212,7 @@ int bgp_start(struct bgp_peer *peer, char *name, int as, int keepalive, int hold
 
     if (ibgp)
     {
-	u32 local_pref = htonl(BGP_LOCAL_PREF);
+	uint32_t local_pref = htonl(BGP_LOCAL_PREF);
 
 	/* LOCAL_PREF */
 	a.flags = BGP_PATH_ATTR_FLAG_TRANS;
@@ -268,6 +276,7 @@ static void bgp_clear(struct bgp_peer *peer)
     peer->inbuf->done = 0;
 
     peer->cli_flag = 0;
+    peer->events = 0;
 
     if (peer->state != peer->next_state)
     {
@@ -324,7 +333,7 @@ static void bgp_set_retry(struct bgp_peer *peer)
 static void bgp_cidr(in_addr_t ip, in_addr_t mask, struct bgp_ip_prefix *pfx)
 {
     int i;
-    u32 b;
+    uint32_t b;
 
     /* convert to prefix notation */
     pfx->len = 32;
@@ -485,130 +494,160 @@ void bgp_enable_routing(int enable)
     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->name && peer->cli_flag == BGP_CLI_RESTART)
+	{
+	    bgp_restart(peer);
+	    continue;
+	}
 
-    if (peer->state == Disabled)
-    	return 1;
+	if (peer->state == Disabled)
+	    continue;
 
-    if (peer->cli_flag)
-    {
-	switch (peer->cli_flag)
+	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;
@@ -641,7 +680,7 @@ char const *bgp_state_str(enum bgp_state state)
     return "?";
 }
 
-static char const *bgp_msg_type_str(u8 type)
+static char const *bgp_msg_type_str(uint8_t type)
 {
     switch (type)
     {
@@ -659,6 +698,7 @@ static int bgp_connect(struct bgp_peer *peer)
 {
     static int bgp_port = 0;
     struct sockaddr_in addr;
+    struct epoll_event ev;
 
     if (!bgp_port)
     {
@@ -681,6 +721,11 @@ static int bgp_connect(struct bgp_peer *peer)
 	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);
 
@@ -717,6 +762,8 @@ static int bgp_connect(struct bgp_peer *peer)
 
     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);
 }
 
@@ -724,7 +771,7 @@ static int bgp_connect(struct bgp_peer *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)
     {
@@ -740,6 +787,8 @@ static int bgp_handle_connect(struct bgp_peer *peer)
 
     LOG(4, 0, 0, "BGP peer %s: state Active\n", peer->name);
 
+    peer->handle_ipv6_routes = 0;
+
     return bgp_send_open(peer);
 }
 
@@ -859,6 +908,12 @@ static int bgp_handle_input(struct bgp_peer *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++)
 	    {
@@ -921,10 +976,91 @@ static int bgp_handle_input(struct bgp_peer *peer)
 	    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_PARAM_TYPE_CAPABILITY)
+		{
+		    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 *)&param->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 (ntohs(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;
@@ -953,10 +1089,21 @@ static int bgp_handle_input(struct bgp_peer *peer)
 	    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;
 	    }
 
+	    if (notification->error_code == BGP_ERR_OPEN
+		    && notification->error_subcode == BGP_ERR_OPN_UNSUP_CAP)
+	    {
+		/* the only capability we advertise is this one, so upon receiving
+		   an "unsupported capability" message, we disable IPv6 routes for
+		   this peer */
+		LOG(4, 0, 0, "BGP peer %s doesn't support IPv6 routes advertisement\n", peer->name);
+		peer->handle_ipv6_routes = 0;
+		break;
+	    }
+
 	    /* FIXME: should handle more notifications */
 	    LOG(4, 0, 0, "BGP peer %s sent unhandled NOTIFICATION %d\n",
 		peer->name, (int) notification->error_code);
@@ -987,7 +1134,10 @@ static int bgp_handle_input(struct bgp_peer *peer)
 static int bgp_send_open(struct bgp_peer *peer)
 {
     struct bgp_data_open data;
-    u16 len = sizeof(peer->outbuf->packet.header);
+    struct bgp_mp_cap_param mp_ipv6 = { htons(AF_INET6), 0, BGP_MP_SAFI_UNICAST };
+    struct bgp_capability cap_mp_ipv6;
+    struct bgp_opt_param param_cap_mp_ipv6;
+    uint16_t len = sizeof(peer->outbuf->packet.header);
 
     memset(peer->outbuf->packet.header.marker, 0xff,
 	sizeof(peer->outbuf->packet.header.marker));
@@ -998,7 +1148,18 @@ static int bgp_send_open(struct bgp_peer *peer)
     data.as = htons(our_as);
     data.hold_time = htons(peer->hold);
     data.identifier = my_address;
-    data.opt_len = 0;
+
+    /* construct the param and capability */
+    cap_mp_ipv6.code = BGP_CAP_CODE_MP;
+    cap_mp_ipv6.len = sizeof(mp_ipv6);
+    memcpy(&cap_mp_ipv6.value, &mp_ipv6, cap_mp_ipv6.len);
+
+    param_cap_mp_ipv6.type = BGP_PARAM_TYPE_CAPABILITY;
+    param_cap_mp_ipv6.len = 2 + sizeof(mp_ipv6);
+    memcpy(&param_cap_mp_ipv6.value, &cap_mp_ipv6, param_cap_mp_ipv6.len);
+
+    data.opt_len = 2 + param_cap_mp_ipv6.len;
+    memcpy(&data.opt_params, &param_cap_mp_ipv6, data.opt_len);
 
     memcpy(peer->outbuf->packet.data, &data, BGP_DATA_OPEN_SIZE);
     len += BGP_DATA_OPEN_SIZE;
@@ -1029,9 +1190,9 @@ static int bgp_send_keepalive(struct bgp_peer *peer)
 /* send/buffer UPDATE message */
 static int bgp_send_update(struct bgp_peer *peer)
 {
-    u16 unf_len = 0;
-    u16 attr_len;
-    u16 len = sizeof(peer->outbuf->packet.header);
+    uint16_t unf_len = 0;
+    uint16_t attr_len;
+    uint16_t len = sizeof(peer->outbuf->packet.header);
     struct bgp_route_list *have = peer->routes;
     struct bgp_route_list *want = peer->routing ? bgp_routes : 0;
     struct bgp_route_list *e = 0;
@@ -1169,10 +1330,17 @@ static int bgp_send_update(struct bgp_peer *peer)
 }
 
 /* send/buffer NOTIFICATION message */
-static int bgp_send_notification(struct bgp_peer *peer, u8 code, u8 subcode)
+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;
-    u16 len = 0;
+    uint16_t len = 0;
 
     data.error_code = code;
     len += sizeof(data.error_code);
@@ -1180,6 +1348,9 @@ static int bgp_send_notification(struct bgp_peer *peer, u8 code, u8 subcode)
     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));