netlink.c

BIRD的路由器守护程序源代码

/*
 *	BIRD -- Linux Netlink Interface
 *
 *	(c) 1999--2000 Martin Mares <[e-mail]>
 *
 *	Can be freely distributed and used under the terms of the GNU GPL.
 */

#include <stdio.h>
#include <fcntl.h>
#include <net/if.h>
#include <sys/socket.h>
#include <sys/uio.h>
#include <errno.h>

#undef LOCAL_DEBUG

#include "nest/bird.h"
#include "nest/route.h"
#include "nest/protocol.h"
#include "nest/iface.h"
#include "lib/timer.h"
#include "lib/unix.h"
#include "lib/krt.h"
#include "lib/socket.h"
#include "lib/string.h"
#include "conf/conf.h"

#include <asm/types.h>
#include <linux/netlink.h>
#include <linux/rtnetlink.h>

#ifndef MSG_TRUNC			/* Hack: Several versions of glibc miss this one :( */
#define MSG_TRUNC 0x20
#endif

/*
 *	Synchronous Netlink interface
 */

static int nl_sync_fd = -1;		/* Unix socket for synchronous netlink actions */
static u32 nl_sync_seq;			/* Sequence number of last request sent */

static byte *nl_rx_buffer;		/* Receive buffer */
#define NL_RX_SIZE 8192

static struct nlmsghdr *nl_last_hdr;	/* Recently received packet */
static unsigned int nl_last_size;

static void
nl_open(void)
{
  if (nl_sync_fd < 0)
    {
      nl_sync_fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
      if (nl_sync_fd < 0)
	die("Unable to open rtnetlink socket: %m");
      nl_sync_seq = now;
      nl_rx_buffer = xmalloc(NL_RX_SIZE);
    }
}

static void
nl_send(struct nlmsghdr *nh)
{
  struct sockaddr_nl sa;

  memset(&sa, 0, sizeof(sa));
  sa.nl_family = AF_NETLINK;
  nh->nlmsg_pid = 0;
  nh->nlmsg_seq = ++nl_sync_seq;
  if (sendto(nl_sync_fd, nh, nh->nlmsg_len, 0, (struct sockaddr *)&sa, sizeof(sa)) < 0)
    die("rtnetlink sendto: %m");
  nl_last_hdr = NULL;
}

static void
nl_request_dump(int cmd)
{
  struct {
    struct nlmsghdr nh;
    struct rtgenmsg g;
  } req;
  req.nh.nlmsg_type = cmd;
  req.nh.nlmsg_len = sizeof(req);
  req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP;
  req.g.rtgen_family = BIRD_PF;
  nl_send(&req.nh);
}

static struct nlmsghdr *
nl_get_reply(void)
{
  for(;;)
    {
      if (!nl_last_hdr)
	{
	  struct iovec iov = { nl_rx_buffer, NL_RX_SIZE };
	  struct sockaddr_nl sa;
	  struct msghdr m = { (struct sockaddr *) &sa, sizeof(sa), &iov, 1, NULL, 0, 0 };
	  int x = recvmsg(nl_sync_fd, &m, 0);
	  if (x < 0)
	    die("nl_get_reply: %m");
	  if (sa.nl_pid)		/* It isn't from the kernel */
	    {
	      DBG("Non-kernel packet\n");
	      continue;
	    }
	  nl_last_size = x;
	  nl_last_hdr = (void *) nl_rx_buffer;
	  if (m.msg_flags & MSG_TRUNC)
	    bug("nl_get_reply: got truncated reply which should be impossible");
	}
      if (NLMSG_OK(nl_last_hdr, nl_last_size))
	{
	  struct nlmsghdr *h = nl_last_hdr;
	  nl_last_hdr = NLMSG_NEXT(h, nl_last_size);
	  if (h->nlmsg_seq != nl_sync_seq)
	    {
	      log(L_WARN "nl_get_reply: Ignoring out of sequence netlink packet (%x != %x)",
		  h->nlmsg_seq, nl_sync_seq);
	      continue;
	    }
	  return h;
	}
      if (nl_last_size)
	log(L_WARN "nl_get_reply: Found packet remnant of size %d", nl_last_size);
      nl_last_hdr = NULL;
    }
}

static struct rate_limit rl_netlink_err;

static int
nl_error(struct nlmsghdr *h)
{
  struct nlmsgerr *e;
  int ec;

  if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr)))
    {
      log(L_WARN "Netlink: Truncated error message received");
      return ENOBUFS;
    }
  e = (struct nlmsgerr *) NLMSG_DATA(h);
  ec = -e->error;
  if (ec)
    log_rl(&rl_netlink_err, L_WARN "Netlink: %s", strerror(ec));
  return ec;
}

static struct nlmsghdr *
nl_get_scan(void)
{
  struct nlmsghdr *h = nl_get_reply();

  if (h->nlmsg_type == NLMSG_DONE)
    return NULL;
  if (h->nlmsg_type == NLMSG_ERROR)
    {
      nl_error(h);
      return NULL;
    }
  return h;
}

static int
nl_exchange(struct nlmsghdr *pkt)
{
  struct nlmsghdr *h;

  nl_send(pkt);
  for(;;)
    {
      h = nl_get_reply();
      if (h->nlmsg_type == NLMSG_ERROR)
	break;
      log(L_WARN "nl_exchange: Unexpected reply received");
    }
  return nl_error(h);
}

/*
 *	Netlink attributes
 */

static int nl_attr_len;

static void *
nl_checkin(struct nlmsghdr *h, int lsize)
{
  nl_attr_len = h->nlmsg_len - NLMSG_LENGTH(lsize);
  if (nl_attr_len < 0)
    {
      log(L_ERR "nl_checkin: underrun by %d bytes", -nl_attr_len);
      return NULL;
    }
  return NLMSG_DATA(h);
}

static int
nl_parse_attrs(struct rtattr *a, struct rtattr **k, int ksize)
{
  int max = ksize / sizeof(struct rtattr *);
  bzero(k, ksize);
  while (RTA_OK(a, nl_attr_len))
    {
      if (a->rta_type < max)
	k[a->rta_type] = a;
      a = RTA_NEXT(a, nl_attr_len);
    }
  if (nl_attr_len)
    {
      log(L_ERR "nl_parse_attrs: remnant of size %d", nl_attr_len);
      return 0;
    }
  else
    return 1;
}

static void
nl_add_attr_u32(struct nlmsghdr *h, unsigned maxsize, int code, u32 data)
{
  unsigned len = RTA_LENGTH(4);
  struct rtattr *a;

  if (NLMSG_ALIGN(h->nlmsg_len) + len > maxsize)
    bug("nl_add_attr32: packet buffer overflow");
  a = (struct rtattr *)((char *)h + NLMSG_ALIGN(h->nlmsg_len));
  a->rta_type = code;
  a->rta_len = len;
  memcpy(RTA_DATA(a), &data, 4);
  h->nlmsg_len = NLMSG_ALIGN(h->nlmsg_len) + len;
}

static void
nl_add_attr_ipa(struct nlmsghdr *h, unsigned maxsize, int code, ip_addr ipa)
{
  unsigned len = RTA_LENGTH(sizeof(ipa));
  struct rtattr *a;

  if (NLMSG_ALIGN(h->nlmsg_len) + len > maxsize)
    bug("nl_add_attr_ipa: packet buffer overflow");
  a = (struct rtattr *)((char *)h + NLMSG_ALIGN(h->nlmsg_len));
  a->rta_type = code;
  a->rta_len = len;
  ipa_hton(ipa);
  memcpy(RTA_DATA(a), &ipa, sizeof(ipa));
  h->nlmsg_len = NLMSG_ALIGN(h->nlmsg_len) + len;
}

/*
 *	Scanning of interfaces
 */

static void
nl_parse_link(struct nlmsghdr *h, int scan)
{
  struct ifinfomsg *i;
  struct rtattr *a[IFLA_WIRELESS+1];
  int new = h->nlmsg_type == RTM_NEWLINK;
  struct iface f;
  struct iface *ifi;
  char *name;
  u32 mtu;
  unsigned int fl;

  if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFLA_RTA(i), a, sizeof(a)))
    return;
  if (!a[IFLA_IFNAME] || RTA_PAYLOAD(a[IFLA_IFNAME]) < 2 ||
      !a[IFLA_MTU] || RTA_PAYLOAD(a[IFLA_MTU]) != 4)
    {
      if (scan || !a[IFLA_WIRELESS])
        log(L_ERR "nl_parse_link: Malformed message received");
      return;
    }
  name = RTA_DATA(a[IFLA_IFNAME]);
  memcpy(&mtu, RTA_DATA(a[IFLA_MTU]), sizeof(u32));

  ifi = if_find_by_index(i->ifi_index);
  if (!new)
    {
      DBG("KIF: IF%d(%s) goes down\n", i->ifi_index, name);
      if (ifi && !scan)
	{
	  memcpy(&f, ifi, sizeof(struct iface));
	  f.flags |= IF_ADMIN_DOWN;
	  if_update(&f);
	}
    }
  else
    {
      DBG("KIF: IF%d(%s) goes up (mtu=%d,flg=%x)\n", i->ifi_index, name, mtu, i->ifi_flags);
      if (ifi)
	memcpy(&f, ifi, sizeof(f));
      else
	{
	  bzero(&f, sizeof(f));
	  f.index = i->ifi_index;
	}
      strncpy(f.name, RTA_DATA(a[IFLA_IFNAME]), sizeof(f.name)-1);
      f.mtu = mtu;
      f.flags = 0;
      fl = i->ifi_flags;
      if (fl & IFF_UP)
	f.flags |= IF_LINK_UP;
      if (fl & IFF_LOOPBACK)		/* Loopback */
	f.flags |= IF_MULTIACCESS | IF_LOOPBACK | IF_IGNORE;
      else if (fl & IFF_POINTOPOINT)	/* PtP */
	f.flags |= IF_MULTICAST;
      else if (fl & IFF_BROADCAST)	/* Broadcast */
	f.flags |= IF_MULTIACCESS | IF_BROADCAST | IF_MULTICAST;
      else
	f.flags |= IF_MULTIACCESS;	/* NBMA */
      if_update(&f);
    }
}

static void
nl_parse_addr(struct nlmsghdr *h)
{
  struct ifaddrmsg *i;
  struct rtattr *a[IFA_ANYCAST+1];
  int new = h->nlmsg_type == RTM_NEWADDR;
  struct ifa ifa;
  struct iface *ifi;
  int scope;

  if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFA_RTA(i), a, sizeof(a)))
    return;
  if (i->ifa_family != BIRD_AF)
    return;
  if (!a[IFA_ADDRESS] || RTA_PAYLOAD(a[IFA_ADDRESS]) != sizeof(ip_addr)
#ifdef IPV6
      || a[IFA_LOCAL] && RTA_PAYLOAD(a[IFA_LOCAL]) != sizeof(ip_addr)
#else
      || !a[IFA_LOCAL] || RTA_PAYLOAD(a[IFA_LOCAL]) != sizeof(ip_addr)
      || (a[IFA_BROADCAST] && RTA_PAYLOAD(a[IFA_BROADCAST]) != sizeof(ip_addr))
#endif
      )
    {
      log(L_ERR "nl_parse_addr: Malformed message received");
      return;
    }

  ifi = if_find_by_index(i->ifa_index);
  if (!ifi)
    {
      log(L_ERR "KIF: Received address message for unknown interface %d", i->ifa_index);
      return;
    }

  bzero(&ifa, sizeof(ifa));
  ifa.iface = ifi;
  if (i->ifa_flags & IFA_F_SECONDARY)
    ifa.flags |= IA_SECONDARY;

  /* IFA_LOCAL can be unset for IPv6 interfaces */
  memcpy(&ifa.ip, RTA_DATA(a[IFA_LOCAL] ? : a[IFA_ADDRESS]), sizeof(ifa.ip));
  ipa_ntoh(ifa.ip);
  ifa.pxlen = i->ifa_prefixlen;
  if (i->ifa_prefixlen > BITS_PER_IP_ADDRESS ||
      i->ifa_prefixlen == BITS_PER_IP_ADDRESS - 1)
    {
      log(L_ERR "KIF: Invalid prefix length for interface %s: %d", ifi->name, i->ifa_prefixlen);
      new = 0;
    }
  if (i->ifa_prefixlen == BITS_PER_IP_ADDRESS)
    {
      ifa.flags |= IA_UNNUMBERED;
      memcpy(&ifa.opposite, RTA_DATA(a[IFA_ADDRESS]), sizeof(ifa.opposite));
      ipa_ntoh(ifa.opposite);
      ifa.prefix = ifa.brd = ifa.opposite;
    }
  else
    {
      ip_addr netmask = ipa_mkmask(ifa.pxlen);
      ip_addr xbrd;
      ifa.prefix = ipa_and(ifa.ip, netmask);
      ifa.brd = ipa_or(ifa.ip, ipa_not(netmask));
#ifndef IPV6
      if (i->ifa_prefixlen == BITS_PER_IP_ADDRESS - 2)
	ifa.opposite = ipa_opposite(ifa.ip, i->ifa_prefixlen);
      if ((ifi->flags & IF_BROADCAST) && a[IFA_BROADCAST])
	{
	  memcpy(&xbrd, RTA_DATA(a[IFA_BROADCAST]), sizeof(xbrd));
	  ipa_ntoh(xbrd);
	  if (ipa_equal(xbrd, ifa.prefix) || ipa_equal(xbrd, ifa.brd))
	    ifa.brd = xbrd;
	  else if (ifi->flags & IF_TMP_DOWN) /* Complain only during the first scan */
	    log(L_ERR "KIF: Invalid broadcast address %I for %s", xbrd, ifi->name);
	}
#endif
    }

  scope = ipa_classify(ifa.ip);
  if (scope < 0)
    {
      log(L_ERR "KIF: Invalid interface address %I for %s", ifa.ip, ifi->name);
      return;
    }
  ifa.scope = scope & IADDR_SCOPE_MASK;

  DBG("KIF: IF%d(%s): %s IPA %I, flg %x, net %I/%d, brd %I, opp %I\n",
      ifi->index, ifi->name,
      new ? "added" : "removed",
      ifa.ip, ifa.flags, ifa.prefix, ifa.pxlen, ifa.brd, ifa.opposite);
  if (new)
    ifa_update(&ifa);
  else
    ifa_delete(&ifa);
}

void
krt_if_scan(struct kif_proto *p UNUSED)
{
  struct nlmsghdr *h;

  if_start_update();

  nl_request_dump(RTM_GETLINK);
  while (h = nl_get_scan())
    if (h->nlmsg_type == RTM_NEWLINK || h->nlmsg_type == RTM_DELLINK)
      nl_parse_link(h, 1);
    else
      log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type);

  nl_request_dump(RTM_GETADDR);
  while (h = nl_get_scan())
    if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR)
      nl_parse_addr(h);
    else
      log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type);

  if_end_update();
}

/*
 *	Routes
 */

static struct krt_proto *nl_table_map[NL_NUM_TABLES];

int
krt_capable(rte *e)
{
  rta *a = e->attrs;

  if (a->cast != RTC_UNICAST
#if 0
      && a->cast != RTC_ANYCAST
#endif
      )
    return 0;
  if (a->source == RTS_DEVICE)	/* Kernel takes care of device routes itself */
    return 0;
  switch (a->dest)
    {
    case RTD_ROUTER:
    case RTD_DEVICE:
    case RTD_BLACKHOLE:
    case RTD_UNREACHABLE: