bgp_packet.c

大名鼎鼎的路由器源码。程序分ZEBRA、OSPFRIP等3个包。程序框架采用一个路由协议一个进程的方式

/* BGP packet management routine.
   Copyright (C) 1999 Kunihiro Ishiguro

This file is part of GNU Zebra.

GNU Zebra is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2, or (at your option) any
later version.

GNU Zebra is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
General Public License for more details.

You should have received a copy of the GNU General Public License
along with GNU Zebra; see the file COPYING.  If not, write to the Free
Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
02111-1307, USA.  */

#include <zebra.h>

#include "thread.h"
#include "stream.h"
#include "network.h"
#include "prefix.h"
#include "command.h"
#include "log.h"
#include "memory.h"
#include "sockunion.h"		/* for inet_ntop () */
#include "linklist.h"
#include "plist.h"

#include "bgpd/bgpd.h"
#include "bgpd/bgp_table.h"
#include "bgpd/bgp_dump.h"
#include "bgpd/bgp_attr.h"
#include "bgpd/bgp_debug.h"
#include "bgpd/bgp_fsm.h"
#include "bgpd/bgp_route.h"
#include "bgpd/bgp_packet.h"
#include "bgpd/bgp_open.h"
#include "bgpd/bgp_aspath.h"
#include "bgpd/bgp_community.h"
#include "bgpd/bgp_ecommunity.h"
#include "bgpd/bgp_network.h"
#include "bgpd/bgp_mplsvpn.h"
#include "bgpd/bgp_advertise.h"

int stream_put_prefix (struct stream *, struct prefix *);

/* Set up BGP packet marker and packet type. */
static int
bgp_packet_set_marker (struct stream *s, u_char type)
{
  int i;

  /* Fill in marker. */
  for (i = 0; i < BGP_MARKER_SIZE; i++)
    stream_putc (s, 0xff);

  /* Dummy total length. This field is should be filled in later on. */
  stream_putw (s, 0);

  /* BGP packet type. */
  stream_putc (s, type);

  /* Return current stream size. */
  return stream_get_putp (s);
}

/* Set BGP packet header size entry.  If size is zero then use current
   stream size. */
static int
bgp_packet_set_size (struct stream *s)
{
  int cp;

  /* Preserve current pointer. */
  cp = stream_get_putp (s);
  stream_set_putp (s, BGP_MARKER_SIZE);
  stream_putw (s, cp);

  /* Write back current pointer. */
  stream_set_putp (s, cp);

  return cp;
}

/* Add new packet to the peer. */
void
bgp_packet_add (struct peer *peer, struct stream *s)
{
  /* Add packet to the end of list. */
  stream_fifo_push (peer->obuf, s);
}

/* Free first packet. */
void
bgp_packet_delete (struct peer *peer)
{
  stream_free (stream_fifo_pop (peer->obuf));
}

/* Duplicate packet. */
struct stream *
bgp_packet_dup (struct stream *s)
{
  struct stream *new;

  new = stream_new (stream_get_endp (s));

  new->endp = s->endp;
  new->putp = s->putp;
  new->getp = s->getp;

  memcpy (new->data, s->data, stream_get_endp (s));

  return new;
}

/* Check file descriptor whether connect is established. */
static void
bgp_connect_check (struct peer *peer)
{
  int status;
  int slen;
  int ret;

  /* Anyway I have to reset read and write thread. */
  BGP_READ_OFF (peer->t_read);
  BGP_WRITE_OFF (peer->t_write);

  /* Check file descriptor. */
  slen = sizeof (status);
  ret = getsockopt(peer->fd, SOL_SOCKET, SO_ERROR, (void *) &status, &slen);

  /* If getsockopt is fail, this is fatal error. */
  if (ret < 0)
    {
      zlog (peer->log, LOG_INFO, "can't get sockopt for nonblocking connect");
      BGP_EVENT_ADD (peer, TCP_fatal_error);
      return;
    }      

  /* When status is 0 then TCP connection is established. */
  if (status == 0)
    {
      BGP_EVENT_ADD (peer, TCP_connection_open);
    }
  else
    {
      if (BGP_DEBUG (events, EVENTS))
	  plog_info (peer->log, "%s [Event] Connect failed (%s)",
		     peer->host, strerror (errno));
      BGP_EVENT_ADD (peer, TCP_connection_open_failed);
    }
}

/* Make BGP update packet.  */
struct stream *
bgp_update_packet (struct peer *peer, afi_t afi, safi_t safi)
{
  struct stream *s;
  struct bgp_adj_out *adj;
  struct bgp_advertise *adv;
  struct stream *packet;
  struct bgp_node *rn = NULL;
  struct bgp_info *binfo = NULL;
  bgp_size_t total_attr_len = 0;
  unsigned long pos;
  char buf[BUFSIZ];
  struct prefix_rd *prd = NULL;
  char *tag = NULL;

  s = peer->work;
  stream_reset (s);

  adv = FIFO_HEAD (&peer->sync[afi][safi]->update);

  while (adv)
    {
      if (adv->rn)
        rn = adv->rn;
      adj = adv->adj;
      if (adv->binfo)
        binfo = adv->binfo;
#ifdef MPLS_VPN
      if (rn)
        prd = (struct prefix_rd *) &rn->prn->p;
      if (binfo)
        tag = binfo->tag;
#endif /* MPLS_VPN */

      /* When remaining space can't include NLRI and it's length.  */
      if (rn && STREAM_REMAIN (s) <= BGP_NLRI_LENGTH + PSIZE (rn->p.prefixlen))
	break;

      /* If packet is empty, set attribute. */
      if (stream_empty (s))
	{
	  bgp_packet_set_marker (s, BGP_MSG_UPDATE);
	  stream_putw (s, 0);		
	  pos = stream_get_putp (s);
	  stream_putw (s, 0);
	  total_attr_len = bgp_packet_attribute (NULL, peer, s,
	                			 adv->baa->attr,
						 &rn->p, afi, safi,
						 binfo->peer, prd, tag);
	  stream_putw_at (s, pos, total_attr_len);
	}

      if (afi == AFI_IP && safi == SAFI_UNICAST)
	stream_put_prefix (s, &rn->p);
      
      if (BGP_DEBUG (update, UPDATE_OUT))
	zlog (peer->log, LOG_INFO, "%s send UPDATE %s/%d",
	      peer->host,
	      inet_ntop (rn->p.family, &(rn->p.u.prefix), buf, BUFSIZ),
	      rn->p.prefixlen);

      /* Synchnorize attribute.  */
      if (adj->attr)
	bgp_attr_unintern (adj->attr);
      else
	peer->scount[afi][safi]++;

      adj->attr = bgp_attr_intern (adv->baa->attr);

      adv = bgp_advertise_clean (peer, adj, afi, safi);

      if (! (afi == AFI_IP && safi == SAFI_UNICAST))
	break;
    }
	 
  if (! stream_empty (s))
    {
      bgp_packet_set_size (s);
      packet = bgp_packet_dup (s);
      bgp_packet_add (peer, packet);
      stream_reset (s);
      return packet;
    }
  return NULL;

}

/* Make BGP withdraw packet.  */
struct stream *
bgp_withdraw_packet (struct peer *peer, afi_t afi, safi_t safi)
{
  struct stream *s;
  struct stream *packet;
  struct bgp_adj_out *adj;
  struct bgp_advertise *adv;
  struct bgp_node *rn;
  unsigned long pos;
  bgp_size_t unfeasible_len;
  bgp_size_t total_attr_len;
  char buf[BUFSIZ];
  struct prefix_rd *prd = NULL;

  s = peer->work;
  stream_reset (s);

  while ((adv = FIFO_HEAD (&peer->sync[afi][safi]->withdraw)) != NULL)
    {
      adj = adv->adj;
      rn = adv->rn;
#ifdef MPLS_VPN
      prd = (struct prefix_rd *) &rn->prn->p;
#endif /* MPLS_VPN */

      if (STREAM_REMAIN (s) 
	  <= (BGP_NLRI_LENGTH + BGP_TOTAL_ATTR_LEN + PSIZE (rn->p.prefixlen)))
	break;

      if (stream_empty (s))
	{
	  bgp_packet_set_marker (s, BGP_MSG_UPDATE);
	  stream_putw (s, 0);
	}

      if (afi == AFI_IP && safi == SAFI_UNICAST)
	stream_put_prefix (s, &rn->p);
      else
	{
	  pos = stream_get_putp (s);
	  stream_putw (s, 0);
	  total_attr_len
	    = bgp_packet_withdraw (peer, s, &rn->p, afi, safi, prd, NULL);
      
	  /* Set total path attribute length. */
	  stream_putw_at (s, pos, total_attr_len);
	}

      if (BGP_DEBUG (update, UPDATE_OUT))
	zlog (peer->log, LOG_INFO, "%s send UPDATE %s/%d -- unreachable",
	      peer->host,
	      inet_ntop (rn->p.family, &(rn->p.u.prefix), buf, BUFSIZ),
	      rn->p.prefixlen);

      peer->scount[afi][safi]--;

      bgp_adj_out_remove (rn, adj, peer, afi, safi);
      bgp_unlock_node (rn);

      if (! (afi == AFI_IP && safi == SAFI_UNICAST))
	break;
    }

  if (! stream_empty (s))
    {
      if (afi == AFI_IP && safi == SAFI_UNICAST)
	{
	  unfeasible_len 
	    = stream_get_putp (s) - BGP_HEADER_SIZE - BGP_UNFEASIBLE_LEN;
	  stream_putw_at (s, BGP_HEADER_SIZE, unfeasible_len);
	  stream_putw (s, 0);
	}
      bgp_packet_set_size (s);
      packet = bgp_packet_dup (s);
      bgp_packet_add (peer, packet);
      stream_reset (s);
      return packet;
    }

  return NULL;
}

void
bgp_default_update_send (struct peer *peer, struct attr *attr,
			 afi_t afi, safi_t safi, struct peer *from)
{
  struct stream *s;
  struct stream *packet;
  struct prefix p;
  unsigned long pos;
  bgp_size_t total_attr_len;
  char attrstr[BUFSIZ];
  char buf[BUFSIZ];

#ifdef DISABLE_BGP_ANNOUNCE
  return;
#endif /* DISABLE_BGP_ANNOUNCE */

  if (afi == AFI_IP)
    str2prefix ("0.0.0.0/0", &p);
#ifdef HAVE_IPV6
  else 
    str2prefix ("::/0", &p);
#endif /* HAVE_IPV6 */

  /* Logging the attribute. */
  if (BGP_DEBUG (update, UPDATE_OUT))
    {
      bgp_dump_attr (peer, attr, attrstr, BUFSIZ);
      zlog (peer->log, LOG_INFO, "%s send UPDATE %s/%d %s",
	    peer->host, inet_ntop(p.family, &(p.u.prefix), buf, BUFSIZ),
	    p.prefixlen, attrstr);
    }

  s = stream_new (BGP_MAX_PACKET_SIZE);

  /* Make BGP update packet. */
  bgp_packet_set_marker (s, BGP_MSG_UPDATE);

  /* Unfeasible Routes Length. */
  stream_putw (s, 0);

  /* Make place for total attribute length.  */
  pos = stream_get_putp (s);
  stream_putw (s, 0);
  total_attr_len = bgp_packet_attribute (NULL, peer, s, attr, &p, afi, safi, from, NULL, NULL);

  /* Set Total Path Attribute Length. */
  stream_putw_at (s, pos, total_attr_len);

  /* NLRI set. */
  if (p.family == AF_INET && safi == SAFI_UNICAST)
    stream_put_prefix (s, &p);

  /* Set size. */
  bgp_packet_set_size (s);

  packet = bgp_packet_dup (s);
  stream_free (s);

  /* Dump packet if debug option is set. */
#ifdef DEBUG
  bgp_packet_dump (packet);
#endif /* DEBUG */

  /* Add packet to the peer. */
  bgp_packet_add (peer, packet);

  BGP_WRITE_ON (peer->t_write, bgp_write, peer->fd);
}

void
bgp_default_withdraw_send (struct peer *peer, afi_t afi, safi_t safi)
{
  struct stream *s;
  struct stream *packet;
  struct prefix p;
  unsigned long pos;
  unsigned long cp;
  bgp_size_t unfeasible_len;
  bgp_size_t total_attr_len;
  char buf[BUFSIZ];

#ifdef DISABLE_BGP_ANNOUNCE
  return;
#endif /* DISABLE_BGP_ANNOUNCE */

  if (afi == AFI_IP)
    str2prefix ("0.0.0.0/0", &p);
#ifdef HAVE_IPV6
  else 
    str2prefix ("::/0", &p);
#endif /* HAVE_IPV6 */

  total_attr_len = 0;
  pos = 0;

  if (BGP_DEBUG (update, UPDATE_OUT))
    zlog (peer->log, LOG_INFO, "%s send UPDATE %s/%d -- unreachable",
          peer->host, inet_ntop(p.family, &(p.u.prefix), buf, BUFSIZ),
          p.prefixlen);

  s = stream_new (BGP_MAX_PACKET_SIZE);

  /* Make BGP update packet. */
  bgp_packet_set_marker (s, BGP_MSG_UPDATE);

  /* Unfeasible Routes Length. */;
  cp = stream_get_putp (s);
  stream_putw (s, 0);

  /* Withdrawn Routes. */
  if (p.family == AF_INET && safi == SAFI_UNICAST)
    {
      stream_put_prefix (s, &p);

      unfeasible_len = stream_get_putp (s) - cp - 2;

      /* Set unfeasible len.  */
      stream_putw_at (s, cp, unfeasible_len);

      /* Set total path attribute length. */
      stream_putw (s, 0);
    }
  else
    {
      pos = stream_get_putp (s);
      stream_putw (s, 0);
      total_attr_len = bgp_packet_withdraw (peer, s, &p, afi, safi, NULL, NULL);

      /* Set total path attribute length. */
      stream_putw_at (s, pos, total_attr_len);
    }

  bgp_packet_set_size (s);

  packet = bgp_packet_dup (s);
  stream_free (s);

  /* Add packet to the peer. */
  bgp_packet_add (peer, packet);

  BGP_WRITE_ON (peer->t_write, bgp_write, peer->fd);
}

/* Get next packet to be written.  */
struct stream *
bgp_write_packet (struct peer *peer)
{
  afi_t afi;
  safi_t safi;
  struct stream *s = NULL;
  struct bgp_advertise *adv;

  s = stream_fifo_head (peer->obuf);
  if (s)
    return s;

  for (afi = AFI_IP; afi < AFI_MAX; afi++)
    for (safi = SAFI_UNICAST; safi < SAFI_MAX; safi++)
      {
	adv = FIFO_HEAD (&peer->sync[afi][safi]->withdraw);
	if (adv)
	  {
	    s = bgp_withdraw_packet (peer, afi, safi);
	    if (s)
	      return s;
	  }
      }
    
  for (afi = AFI_IP; afi < AFI_MAX; afi++)
    for (safi = SAFI_UNICAST; safi < SAFI_MAX; safi++)
      {
	adv = FIFO_HEAD (&peer->sync[afi][safi]->update);
	if (adv)
	  {
            if (adv->binfo && adv->binfo->uptime < peer->synctime)
              s = bgp_update_packet (peer, afi, safi);

	    if (s)
	      return s;
	  }
      }

  return NULL;
}

/* Is there partially written packet or updates we can send right
   now.  */
int
bgp_write_proceed (struct peer *peer)
{
  afi_t afi;
  safi_t safi;
  struct bgp_advertise *adv;

  if (stream_fifo_head (peer->obuf))
    return 1;

  for (afi = AFI_IP; afi < AFI_MAX; afi++)
    for (safi = SAFI_UNICAST; safi < SAFI_MAX; safi++)
      if (FIFO_HEAD (&peer->sync[afi][safi]->withdraw))
	return 1;

  for (afi = AFI_IP; afi < AFI_MAX; afi++)
    for (safi = SAFI_UNICAST; safi < SAFI_MAX; safi++)
      if ((adv = FIFO_HEAD (&peer->sync[afi][safi]->update)) != NULL)
	if (adv->binfo->uptime < peer->synctime)
	  return 1;

  return 0;
}

/* Write packet to the peer. */
int
bgp_write (struct thread *thread)
{
  struct peer *peer;
  u_char type;
  struct stream *s; 
  int num;
  int count = 0;
  int write_errno;

  /* Yes first of all get peer pointer. */
  peer = THREAD_ARG (thread);
  peer->t_write = NULL;

  /* For non-blocking IO check. */
  if (peer->status == Connect)
    {
      bgp_connect_check (peer);
      return 0;
    }

    /* Nonblocking write until TCP output buffer is full.  */