rt_netlink.c

zebra测试源代码用于 SOCKET 通信

  struct iovec iov = { (void*) n, n->nlmsg_len };
  struct msghdr msg = {(void*) &snl, sizeof snl, &iov, 1, NULL, 0, 0};
  int flags = 0;
  
  memset (&snl, 0, sizeof snl);
  snl.nl_family = AF_NETLINK;
  
  n->nlmsg_seq = ++netlink_cmd.seq;

  /* Request an acknowledgement by setting NLM_F_ACK */
  n->nlmsg_flags |= NLM_F_ACK;
  
  if (IS_ZEBRA_DEBUG_KERNEL) 
    zlog_info ("netlink_talk: %s type %s(%u), seq=%u", netlink_cmd.name,
	      lookup (nlmsg_str, n->nlmsg_type), n->nlmsg_type,
	      n->nlmsg_seq);

  /* Send message to netlink interface. */
  status = sendmsg (nl->sock, &msg, 0);
  if (status < 0)
    {
      zlog (NULL, LOG_ERR, "netlink_talk sendmsg() error: %s",
	    strerror (errno));
      return -1;
    }
  
  /* 
   * Change socket flags for blocking I/O. 
   * This ensures we wait for a reply in netlink_parse_info().
   */
  if((flags = fcntl(nl->sock, F_GETFL, 0)) < 0) 
    {
      zlog (NULL, LOG_ERR, "%s:%i F_GETFL error: %s", 
              __FUNCTION__, __LINE__, strerror (errno));
    }
  flags &= ~O_NONBLOCK;
  if(fcntl(nl->sock, F_SETFL, flags) < 0) 
    {
      zlog (NULL, LOG_ERR, "%s:%i F_SETFL error: %s", 
              __FUNCTION__, __LINE__, strerror (errno));
    }

  /* 
   * Get reply from netlink socket. 
   * The reply should either be an acknowlegement or an error.
   */
  status = netlink_parse_info (netlink_talk_filter, nl);
  
  /* Restore socket flags for nonblocking I/O */
  flags |= O_NONBLOCK;
  if(fcntl(nl->sock, F_SETFL, flags) < 0) 
    {
      zlog (NULL, LOG_ERR, "%s:%i F_SETFL error: %s", 
              __FUNCTION__, __LINE__, strerror (errno));
    }
  
  return status;
}

/* Routing table change via netlink interface. */
int
netlink_route (int cmd, int family, void *dest, int length, void *gate,
	       int index, int zebra_flags, int table)
{
  int ret;
  int bytelen;
  struct sockaddr_nl snl;
  int discard;

  struct 
  {
    struct nlmsghdr n;
    struct rtmsg r;
    char buf[1024];
  } req;

  memset (&req, 0, sizeof req);

  bytelen = (family == AF_INET ? 4 : 16);

  req.n.nlmsg_len = NLMSG_LENGTH (sizeof (struct rtmsg));
  req.n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
  req.n.nlmsg_type = cmd;
  req.r.rtm_family = family;
  req.r.rtm_table = table;
  req.r.rtm_dst_len = length;

  if (zebra_flags & ZEBRA_FLAG_BLACKHOLE)
    discard = 1;
  else
    discard = 0;

  if (cmd == RTM_NEWROUTE) 
    {
#if 0
      req.r.rtm_protocol = RTPROT_ZEBRA;
#endif
      req.r.rtm_scope = RT_SCOPE_UNIVERSE;

      if (discard)
	req.r.rtm_type = RTN_BLACKHOLE;
      else
	req.r.rtm_type = RTN_UNICAST;
    }

  if (dest)
    addattr_l (&req.n, sizeof req, RTA_DST, dest, bytelen);

  if (! discard)
    {
      if (gate)
	addattr_l (&req.n, sizeof req, RTA_GATEWAY, gate, bytelen);
      if (index > 0)
	addattr32 (&req.n, sizeof req, RTA_OIF, index);
    }

  /* Destination netlink address. */
  memset (&snl, 0, sizeof snl);
  snl.nl_family = AF_NETLINK;

  /* Talk to netlink socket. */
  ret = netlink_talk (&req.n, &netlink);
  if (ret < 0)
    return -1;

  return 0;
}

/* Routing table change via netlink interface. */
int
netlink_route_multipath (int cmd, struct prefix *p, struct rib *rib,
			 int family)
{
  int bytelen;
  struct sockaddr_nl snl;
  struct nexthop *nexthop = NULL;
  int nexthop_num = 0;
  struct nlsock *nl;
  int discard;

  struct 
  {
    struct nlmsghdr n;
    struct rtmsg r;
    char buf[1024];
  } req;

  memset (&req, 0, sizeof req);

  bytelen = (family == AF_INET ? 4 : 16);

  req.n.nlmsg_len = NLMSG_LENGTH (sizeof (struct rtmsg));
  req.n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
  req.n.nlmsg_type = cmd;
  req.r.rtm_family = family;
  req.r.rtm_table = rib->table;
  req.r.rtm_dst_len = p->prefixlen;

#ifdef RTM_F_EQUALIZE
  req.r.rtm_flags |= RTM_F_EQUALIZE;
#endif /* RTM_F_EQUALIZE */

  if (rib->flags & ZEBRA_FLAG_BLACKHOLE)
    discard = 1;
  else
    discard = 0;

  if (cmd == RTM_NEWROUTE) 
    {
      req.r.rtm_protocol = RTPROT_ZEBRA;
      req.r.rtm_scope = RT_SCOPE_UNIVERSE;

      if (discard)
	req.r.rtm_type = RTN_BLACKHOLE;
      else
	req.r.rtm_type = RTN_UNICAST;
    }

  addattr_l (&req.n, sizeof req, RTA_DST, &p->u.prefix, bytelen);

  /* Metric. */
  addattr32 (&req.n, sizeof req, RTA_PRIORITY, rib->metric);

  if (discard)
    {
      if (cmd == RTM_NEWROUTE)
	for (nexthop = rib->nexthop; nexthop; nexthop = nexthop->next)
	  SET_FLAG (nexthop->flags, NEXTHOP_FLAG_FIB);
      goto skip;
    }

  /* Multipath case. */
  if (rib->nexthop_active_num == 1 || MULTIPATH_NUM == 1)
    {
      for (nexthop = rib->nexthop; nexthop; nexthop = nexthop->next)
	{
	  if ((cmd == RTM_NEWROUTE 
	       && CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_ACTIVE))
	      || (cmd == RTM_DELROUTE
		  && CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_FIB)))
	    {
	      if (CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
		{
		  if (nexthop->rtype == NEXTHOP_TYPE_IPV4 
		      || nexthop->rtype == NEXTHOP_TYPE_IPV4_IFINDEX)
		    addattr_l (&req.n, sizeof req, RTA_GATEWAY,
			       &nexthop->rgate.ipv4, bytelen);
#ifdef HAVE_IPV6
		  if (nexthop->rtype == NEXTHOP_TYPE_IPV6 
		      || nexthop->rtype == NEXTHOP_TYPE_IPV6_IFINDEX 
		      || nexthop->rtype == NEXTHOP_TYPE_IPV6_IFNAME)
		    addattr_l (&req.n, sizeof req, RTA_GATEWAY,
			       &nexthop->rgate.ipv6, bytelen);
#endif /* HAVE_IPV6 */
		  if (nexthop->rtype == NEXTHOP_TYPE_IFINDEX
		      || nexthop->rtype == NEXTHOP_TYPE_IFNAME
		      || nexthop->rtype == NEXTHOP_TYPE_IPV4_IFINDEX
		      || nexthop->rtype == NEXTHOP_TYPE_IPV6_IFINDEX
		      || nexthop->rtype == NEXTHOP_TYPE_IPV6_IFNAME)
		    addattr32 (&req.n, sizeof req, RTA_OIF,
			       nexthop->rifindex);
		}
	      else
		{
		  if (nexthop->type == NEXTHOP_TYPE_IPV4 
		      || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX)
		    addattr_l (&req.n, sizeof req, RTA_GATEWAY,
			       &nexthop->gate.ipv4, bytelen);
#ifdef HAVE_IPV6
		  if (nexthop->type == NEXTHOP_TYPE_IPV6 
		      || nexthop->type == NEXTHOP_TYPE_IPV6_IFNAME
		      || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)
		    addattr_l (&req.n, sizeof req, RTA_GATEWAY,
			       &nexthop->gate.ipv6, bytelen);
#endif /* HAVE_IPV6 */
		  if (nexthop->type == NEXTHOP_TYPE_IFINDEX
		      || nexthop->type == NEXTHOP_TYPE_IFNAME
		      || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX
		      || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX
		      || nexthop->type == NEXTHOP_TYPE_IPV6_IFNAME)
		    addattr32 (&req.n, sizeof req, RTA_OIF, nexthop->ifindex);
		}

	      if (cmd == RTM_NEWROUTE)
		SET_FLAG (nexthop->flags, NEXTHOP_FLAG_FIB);

	      nexthop_num++;
	      break;
	    }
	}
    }
  else
    {
      char buf[1024];
      struct rtattr *rta = (void *) buf;
      struct rtnexthop *rtnh;

      rta->rta_type = RTA_MULTIPATH;
      rta->rta_len = RTA_LENGTH(0);
      rtnh = RTA_DATA(rta);

      nexthop_num = 0;
      for (nexthop = rib->nexthop;
	   nexthop && (MULTIPATH_NUM == 0 || nexthop_num < MULTIPATH_NUM);
	   nexthop = nexthop->next)
	{
	  if ((cmd == RTM_NEWROUTE 
	       && CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_ACTIVE))
	      || (cmd == RTM_DELROUTE
		  && CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_FIB)))
	    {
	      nexthop_num++;

	      rtnh->rtnh_len = sizeof (*rtnh);
	      rtnh->rtnh_flags = 0;
	      rtnh->rtnh_hops = 0;
	      rta->rta_len += rtnh->rtnh_len;

	      if (CHECK_FLAG (nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
		{
		  if (nexthop->rtype == NEXTHOP_TYPE_IPV4
		      || nexthop->rtype == NEXTHOP_TYPE_IPV4_IFINDEX)
		    {
		      rta_addattr_l (rta, 4096, RTA_GATEWAY,
				     &nexthop->rgate.ipv4, bytelen);
		      rtnh->rtnh_len += sizeof (struct rtattr) + 4;
		    }
#ifdef HAVE_IPV6
		  if (nexthop->rtype == NEXTHOP_TYPE_IPV6
		      || nexthop->rtype == NEXTHOP_TYPE_IPV6_IFNAME
		      || nexthop->rtype == NEXTHOP_TYPE_IPV6_IFINDEX)
		    rta_addattr_l (rta, 4096, RTA_GATEWAY,
				   &nexthop->rgate.ipv6, bytelen);
#endif /* HAVE_IPV6 */
		  /* ifindex */
		  if (nexthop->rtype == NEXTHOP_TYPE_IFINDEX
		      || nexthop->rtype == NEXTHOP_TYPE_IFNAME
		      || nexthop->rtype == NEXTHOP_TYPE_IPV4_IFINDEX
		      || nexthop->rtype == NEXTHOP_TYPE_IPV6_IFINDEX
		      || nexthop->rtype == NEXTHOP_TYPE_IPV6_IFNAME)
		    rtnh->rtnh_ifindex = nexthop->rifindex;
		  else
		    rtnh->rtnh_ifindex = 0;
		}
	      else
		{
		  if (nexthop->type == NEXTHOP_TYPE_IPV4
		      || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX)
		    {
		      rta_addattr_l (rta, 4096, RTA_GATEWAY,
				     &nexthop->gate.ipv4, bytelen);
		      rtnh->rtnh_len += sizeof (struct rtattr) + 4;
		    }
#ifdef HAVE_IPV6
		  if (nexthop->type == NEXTHOP_TYPE_IPV6
		      || nexthop->type == NEXTHOP_TYPE_IPV6_IFNAME
		      || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)
		    rta_addattr_l (rta, 4096, RTA_GATEWAY,
				   &nexthop->gate.ipv6, bytelen);
#endif /* HAVE_IPV6 */
		  /* ifindex */
		  if (nexthop->type == NEXTHOP_TYPE_IFINDEX
		      || nexthop->type == NEXTHOP_TYPE_IFNAME
		      || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX
		      || nexthop->type == NEXTHOP_TYPE_IPV6_IFNAME
		      || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)
		    rtnh->rtnh_ifindex = nexthop->ifindex;
		  else
		    rtnh->rtnh_ifindex = 0;
		}
	      rtnh = RTNH_NEXT(rtnh);

	      if (cmd == RTM_NEWROUTE)
		SET_FLAG (nexthop->flags, NEXTHOP_FLAG_FIB);
	    }
	}

      if (rta->rta_len > RTA_LENGTH (0))
	addattr_l (&req.n, 1024, RTA_MULTIPATH, RTA_DATA(rta),
		   RTA_PAYLOAD(rta));
    }

  /* If there is no useful nexthop then return. */
  if (nexthop_num == 0)
    {
      if (IS_ZEBRA_DEBUG_KERNEL)
	zlog_info ("netlink_route_multipath(): No useful nexthop.");
      return 0;
    }

 skip:

  /* Destination netlink address. */
  memset (&snl, 0, sizeof snl);
  snl.nl_family = AF_NETLINK;

  if (family == AF_INET)
    nl = &netlink_cmd;
  else
    nl = &netlink;

  /* Talk to netlink socket. */
  return netlink_talk (&req.n, nl);
}

int
kernel_add_ipv4 (struct prefix *p, struct rib *rib)
{
  return netlink_route_multipath (RTM_NEWROUTE, p, rib, AF_INET);
}

int
kernel_delete_ipv4 (struct prefix *p, struct rib *rib)
{
  return netlink_route_multipath (RTM_DELROUTE, p, rib, AF_INET);
}

#ifdef HAVE_IPV6
int
kernel_add_ipv6 (struct prefix *p, struct rib *rib)
{
  return netlink_route_multipath (RTM_NEWROUTE, p, rib, AF_INET6);
}

int
kernel_delete_ipv6 (struct prefix *p, struct rib *rib)
{
  return netlink_route_multipath (RTM_DELROUTE, p, rib, AF_INET6);
}

/* Delete IPv6 route from the kernel. */
int
kernel_delete_ipv6_old (struct prefix_ipv6 *dest, struct in6_addr *gate,
			int index, int flags, int table)
{
  return netlink_route (RTM_DELROUTE, AF_INET6, &dest->prefix, dest->prefixlen,
			gate, index, flags, table);
}
#endif /* HAVE_IPV6 */

/* Interface address modification. */
int
netlink_address (int cmd, int family, struct interface *ifp,
		 struct connected *ifc)
{
  int bytelen;
  struct prefix *p;

  struct 
  {
    struct nlmsghdr n;
    struct ifaddrmsg ifa;
    char buf[1024];
  } req;

  p = ifc->address;
  memset (&req, 0, sizeof req);

  bytelen = (family == AF_INET ? 4 : 16);

  req.n.nlmsg_len = NLMSG_LENGTH (sizeof(struct ifaddrmsg));
  req.n.nlmsg_flags = NLM_F_REQUEST;
  req.n.nlmsg_type = cmd;
  req.ifa.ifa_family = family;

  req.ifa.ifa_index = ifp->ifindex;
  req.ifa.ifa_prefixlen = p->prefixlen;

  addattr_l (&req.n, sizeof req, IFA_LOCAL, &p->u.prefix, bytelen);

  if (family == AF_INET && cmd == RTM_NEWADDR)
    {
      if (if_is_broadcast (ifp) && ifc->destination)
	{
	  p = ifc->destination;
	  addattr_l(&req.n, sizeof req, IFA_BROADCAST, &p->u.prefix, bytelen);
	}
    }

  if (CHECK_FLAG (ifc->flags, ZEBRA_IFA_SECONDARY))
    SET_FLAG (req.ifa.ifa_flags, IFA_F_SECONDARY);
    
  if (ifc->label)
    addattr_l (&req.n, sizeof req, IFA_LABEL, ifc->label,
	       strlen (ifc->label) + 1);

  return netlink_talk (&req.n, &netlink_cmd);
}

int
kernel_address_add_ipv4 (struct interface *ifp, struct connected *ifc)
{
  return netlink_address (RTM_NEWADDR, AF_INET, ifp, ifc);
}

int
kernel_address_delete_ipv4 (struct interface *ifp, struct connected *ifc)
{
  return netlink_address (RTM_DELADDR, AF_INET, ifp, ifc);
}

#include "thread.h"

extern struct thread_master *master;

/* Kernel route reflection. */
int
kernel_read (struct thread *thread)
{
  int ret;
  int sock;

  sock = THREAD_FD (thread);
  ret = netlink_parse_info (netlink_information_fetch, &netlink);
  thread_add_read (master, kernel_read, NULL, netlink.sock);

  return 0;
}

/* Exported interface function.  This function simply calls
   netlink_socket (). */
void
kernel_init ()
{
  unsigned long groups;

  groups = RTMGRP_LINK|RTMGRP_IPV4_ROUTE|RTMGRP_IPV4_IFADDR;
#ifdef HAVE_IPV6
  groups |= RTMGRP_IPV6_ROUTE|RTMGRP_IPV6_IFADDR;
#endif /* HAVE_IPV6 */
  netlink_socket (&netlink, groups);
  netlink_socket (&netlink_cmd, 0);

  /* Register kernel socket. */
  if (netlink.sock > 0)
    thread_add_read (master, kernel_read, NULL, netlink.sock);
}