IPv4/IPv6 integrated socket code.

3 files changed, 886 insertions, 704 deletions

diff --git a/sysdep/unix/io.c b/sysdep/unix/io.c
index 428f24cc..5a0c07e5 100644
--- a/sysdep/unix/io.c
+++ b/sysdep/unix/io.c
@@ -22,6 +22,7 @@
 #include <unistd.h>
 #include <fcntl.h>
 #include <errno.h>
+#include <net/if.h>
 #include <netinet/in.h>
 #include <netinet/tcp.h>
 #include <netinet/udp.h>
@@ -470,6 +471,7 @@ tm_format_datetime(char *x, struct timeformat *fmt_spec, bird_clock_t t)
     strcpy(x, "<too-long>");
 }
 
+
 /**
  * DOC: Sockets
  *
@@ -496,234 +498,147 @@ tm_format_datetime(char *x, struct timeformat *fmt_spec, bird_clock_t t)
 #endif
 
 
-static list sock_list;
-static struct birdsock *current_sock;
-static struct birdsock *stored_sock;
-static int sock_recalc_fdsets_p;
+/*
+ *	Sockaddr helper functions
+ */
 
-static inline sock *
-sk_next(sock *s)
+static inline int sockaddr_length(int af)
+{ return (af == AF_INET) ? sizeof(struct sockaddr_in) : sizeof(struct sockaddr_in6); }
+
+static inline void
+sockaddr_fill4(struct sockaddr_in *sa, ip_addr a, struct iface *ifa, uint port)
 {
-  if (!s->n.next->next)
-    return NULL;
-  else
-    return SKIP_BACK(sock, n, s->n.next);
+  memset(sa, 0, sizeof(struct sockaddr_in));
+#ifdef HAVE_SIN_LEN
+  sa->sin_len = sizeof(struct sockaddr_in);
+#endif
+  sa->sin_family = AF_INET;
+  sa->sin_port = htons(port);
+  sa->sin_addr = ipa_to_in4(a);
 }
 
-static void
-sk_alloc_bufs(sock *s)
+static inline void
+sockaddr_fill6(struct sockaddr_in6 *sa, ip_addr a, struct iface *ifa, uint port)
 {
-  if (!s->rbuf && s->rbsize)
-    s->rbuf = s->rbuf_alloc = xmalloc(s->rbsize);
-  s->rpos = s->rbuf;
-  if (!s->tbuf && s->tbsize)
-    s->tbuf = s->tbuf_alloc = xmalloc(s->tbsize);
-  s->tpos = s->ttx = s->tbuf;
+  memset(sa, 0, sizeof(struct sockaddr_in6));
+#ifdef SIN6_LEN
+  sa->sin6_len = sizeof(struct sockaddr_in6);
+#endif
+  sa->sin6_family = AF_INET6;
+  sa->sin6_port = htons(port);
+  sa->sin6_flowinfo = 0;
+  sa->sin6_addr = ipa_to_in6(a);
+
+  if (ifa && ipa_is_link_local(a))
+    sa->sin6_scope_id = ifa->index;
 }
 
-static void
-sk_free_bufs(sock *s)
+void
+sockaddr_fill(sockaddr *sa, int af, ip_addr a, struct iface *ifa, uint port)
 {
-  if (s->rbuf_alloc)
-    {
-      xfree(s->rbuf_alloc);
-      s->rbuf = s->rbuf_alloc = NULL;
-    }
-  if (s->tbuf_alloc)
-    {
-      xfree(s->tbuf_alloc);
-      s->tbuf = s->tbuf_alloc = NULL;
-    }
+  if (af == AF_INET)
+    sockaddr_fill4((struct sockaddr_in *) sa, a, ifa, port);
+  else if (af == AF_INET6)
+    sockaddr_fill6((struct sockaddr_in6 *) sa, a, ifa, port);
+  else
+    bug("Unknown AF");
 }
 
-static void
-sk_free(resource *r)
+static inline void
+sockaddr_read4(struct sockaddr_in *sa, ip_addr *a, struct iface **ifa, uint *port)
 {
-  sock *s = (sock *) r;
-
-  sk_free_bufs(s);
-  if (s->fd >= 0)
-    {
-      close(s->fd);
-
-      /* FIXME: we should call sk_stop() for SKF_THREAD sockets */
-      if (s->flags & SKF_THREAD)
-	return;
-
-      if (s == current_sock)
-	current_sock = sk_next(s);
-      if (s == stored_sock)
-	stored_sock = sk_next(s);
-      rem_node(&s->n);
-      sock_recalc_fdsets_p = 1;
-    }
+  *port = ntohs(sa->sin_port);
+  *a = ipa_from_in4(sa->sin_addr);
 }
 
-void
-sk_set_rbsize(sock *s, uint val)
+static inline void
+sockaddr_read6(struct sockaddr_in6 *sa, ip_addr *a, struct iface **ifa, uint *port)
 {
-  ASSERT(s->rbuf_alloc == s->rbuf);
-
-  if (s->rbsize == val)
-    return;
+  *port = ntohs(sa->sin6_port);
+  *a = ipa_from_in6(sa->sin6_addr);
 
-  s->rbsize = val;
-  xfree(s->rbuf_alloc);
-  s->rbuf_alloc = xmalloc(val);
-  s->rpos = s->rbuf = s->rbuf_alloc;
+  if (ifa && ipa_is_link_local(*a))
+    *ifa = if_find_by_index(sa->sin6_scope_id);
 }
 
-void
-sk_set_tbsize(sock *s, uint val)
+int
+sockaddr_read(sockaddr *sa, int af, ip_addr *a, struct iface **ifa, uint *port)
 {
-  ASSERT(s->tbuf_alloc == s->tbuf);
+  if (sa->sa.sa_family != af)
+    goto fail;
 
-  if (s->tbsize == val)
-    return;
-
-  byte *old_tbuf = s->tbuf;
+  if (af == AF_INET)
+    sockaddr_read4((struct sockaddr_in *) sa, a, ifa, port);
+  else if (af == AF_INET6)
+    sockaddr_read6((struct sockaddr_in6 *) sa, a, ifa, port);
+  else
+    goto fail;
 
-  s->tbsize = val;
-  s->tbuf = s->tbuf_alloc = xrealloc(s->tbuf_alloc, val);
-  s->tpos = s->tbuf + (s->tpos - old_tbuf);
-  s->ttx  = s->tbuf + (s->ttx  - old_tbuf);
-}
+  return 0;
 
-void
-sk_set_tbuf(sock *s, void *tbuf)
-{
-  s->tbuf = tbuf ?: s->tbuf_alloc;
-  s->ttx = s->tpos = s->tbuf;
+ fail:
+  *a = IPA_NONE;
+  *port = 0;
+  return -1;
 }
 
-void
-sk_reallocate(sock *s)
-{
-  sk_free_bufs(s);
-  sk_alloc_bufs(s);
-}
 
-static void
-sk_dump(resource *r)
-{
-  sock *s = (sock *) r;
-  static char *sk_type_names[] = { "TCP<", "TCP>", "TCP", "UDP", "UDP/MC", "IP", "IP/MC", "MAGIC", "UNIX<", "UNIX", "DEL!" };
+/*
+ *	IPv6 multicast syscalls
+ */
 
-  debug("(%s, ud=%p, sa=%08x, sp=%d, da=%08x, dp=%d, tos=%d, ttl=%d, if=%s)\n",
-	sk_type_names[s->type],
-	s->data,
-	s->saddr,
-	s->sport,
-	s->daddr,
-	s->dport,
-	s->tos,
-	s->ttl,
-	s->iface ? s->iface->name : "none");
-}
+/* Fortunately standardized in RFC 3493 */
 
-static struct resclass sk_class = {
-  "Socket",
-  sizeof(sock),
-  sk_free,
-  sk_dump,
-  NULL,
-  NULL
-};
+#define INIT_MREQ6(maddr,ifa) \
+  { .ipv6mr_multiaddr = ipa_to_in6(maddr), .ipv6mr_interface = ifa->index }
 
-/**
- * sk_new - create a socket
- * @p: pool
- *
- * This function creates a new socket resource. If you want to use it,
- * you need to fill in all the required fields of the structure and
- * call sk_open() to do the actual opening of the socket.
- *
- * The real function name is sock_new(), sk_new() is a macro wrapper
- * to avoid collision with OpenSSL.
- */
-sock *
-sock_new(pool *p)
+static inline int
+sk_setup_multicast6(sock *s)
 {
-  sock *s = ralloc(p, &sk_class);
-  s->pool = p;
-  // s->saddr = s->daddr = IPA_NONE;
-  s->tos = s->priority = s->ttl = -1;
-  s->fd = -1;
-  return s;
-}
+  int index = s->iface->index;
+  int ttl = s->ttl;
+  int n = 0;
 
-static void
-sk_insert(sock *s)
-{
-  add_tail(&sock_list, &s->n);
-  sock_recalc_fdsets_p = 1;
-}
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_IF, &index, sizeof(index)) < 0)
+    ERR("IPV6_MULTICAST_IF");
 
-#ifdef IPV6
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_HOPS, &ttl, sizeof(ttl)) < 0)
+    ERR("IPV6_MULTICAST_HOPS");
 
-void
-fill_in_sockaddr(struct sockaddr_in6 *sa, ip_addr a, struct iface *ifa, unsigned port)
-{
-  memset(sa, 0, sizeof (struct sockaddr_in6));
-  sa->sin6_family = AF_INET6;
-  sa->sin6_port = htons(port);
-  sa->sin6_flowinfo = 0;
-#ifdef HAVE_SIN_LEN
-  sa->sin6_len = sizeof(struct sockaddr_in6);
-#endif
-  set_inaddr(&sa->sin6_addr, a);
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_LOOP, &n, sizeof(n)) < 0)
+    ERR("IPV6_MULTICAST_LOOP");
 
-  if (ifa && ipa_has_link_scope(a))
-    sa->sin6_scope_id = ifa->index;
+  return 0;
 }
 
-void
-get_sockaddr(struct sockaddr_in6 *sa, ip_addr *a, struct iface **ifa, unsigned *port, int check)
+static inline int
+sk_join_group6(sock *s, ip_addr maddr)
 {
-  if (check && sa->sin6_family != AF_INET6)
-    bug("get_sockaddr called for wrong address family (%d)", sa->sin6_family);
-  if (port)
-    *port = ntohs(sa->sin6_port);
-  memcpy(a, &sa->sin6_addr, sizeof(*a));
-  ipa_ntoh(*a);
+  struct ipv6_mreq mr = INIT_MREQ6(maddr, s->iface);
 
-  if (ifa && ipa_has_link_scope(*a))
-    *ifa = if_find_by_index(sa->sin6_scope_id);
-}
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_JOIN_GROUP, &mr, sizeof(mr)) < 0)
+    ERR("IPV6_JOIN_GROUP");
 
-#else
-
-void
-fill_in_sockaddr(struct sockaddr_in *sa, ip_addr a, struct iface *ifa, unsigned port)
-{
-  memset (sa, 0, sizeof (struct sockaddr_in));
-  sa->sin_family = AF_INET;
-  sa->sin_port = htons(port);
-#ifdef HAVE_SIN_LEN
-  sa->sin_len = sizeof(struct sockaddr_in);
-#endif
-  set_inaddr(&sa->sin_addr, a);
+  return 0;
 }
 
-void
-get_sockaddr(struct sockaddr_in *sa, ip_addr *a, struct iface **ifa, unsigned *port, int check)
+static inline int
+sk_leave_group6(sock *s, ip_addr maddr)
 {
-  if (check && sa->sin_family != AF_INET)
-    bug("get_sockaddr called for wrong address family (%d)", sa->sin_family);
-  if (port)
-    *port = ntohs(sa->sin_port);
-  memcpy(a, &sa->sin_addr.s_addr, sizeof(*a));
-  ipa_ntoh(*a);
-}
+  struct ipv6_mreq mr = INIT_MREQ6(maddr, s->iface);
 
-#endif
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_LEAVE_GROUP, &mr, sizeof(mr)) < 0)
+    ERR("IPV6_LEAVE_GROUP");
+
+  return 0;
+}
 
 
-#ifdef IPV6
+/*
+ *	IPv6 packet control messages
+ */
 
-/* PKTINFO handling is also standardized in IPv6 */
-#define CMSG_RX_SPACE (CMSG_SPACE(sizeof(struct in6_pktinfo)) + CMSG_SPACE(sizeof(int)))
-#define CMSG_TX_SPACE CMSG_SPACE(sizeof(struct in6_pktinfo))
+/* Also standardized, in RFC 3542 */
 
 /*
  * RFC 2292 uses IPV6_PKTINFO for both the socket option and the cmsg
@@ -741,60 +656,52 @@ get_sockaddr(struct sockaddr_in *sa, ip_addr *a, struct iface **ifa, unsigned *p
 #define IPV6_RECVHOPLIMIT IPV6_HOPLIMIT
 #endif
 
-static char *
-sysio_register_cmsgs(sock *s)
-{
-  int ok = 1;
 
-  if ((s->flags & SKF_LADDR_RX) &&
-      (setsockopt(s->fd, SOL_IPV6, IPV6_RECVPKTINFO, &ok, sizeof(ok)) < 0))
-    return "IPV6_RECVPKTINFO";
+#define CMSG6_SPACE_PKTINFO CMSG_SPACE(sizeof(struct in6_pktinfo))
+#define CMSG6_SPACE_TTL CMSG_SPACE(sizeof(int))
+
+static inline int
+sk_request_cmsg6_pktinfo(sock *s)
+{
+  int y = 1;
 
-  if ((s->flags & SKF_TTL_RX) &&
-      (setsockopt(s->fd, SOL_IPV6, IPV6_RECVHOPLIMIT, &ok, sizeof(ok)) < 0))
-    return "IPV6_RECVHOPLIMIT";
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_RECVPKTINFO, &y, sizeof(y)) < 0)
+    ERR("IPV6_RECVPKTINFO");
 
-  return NULL;
+  return 0;
 }
 
-static void
-sysio_process_rx_cmsgs(sock *s, struct msghdr *msg)
+static inline int
+sk_request_cmsg6_ttl(sock *s)
 {
-  struct cmsghdr *cm;
-  struct in6_pktinfo *pi = NULL;
-  int *hlim = NULL;
+  int y = 1;
 
-  for (cm = CMSG_FIRSTHDR(msg); cm != NULL; cm = CMSG_NXTHDR(msg, cm))
-  {
-    if (cm->cmsg_level == SOL_IPV6 && cm->cmsg_type == IPV6_PKTINFO)
-      pi = (struct in6_pktinfo *) CMSG_DATA(cm);
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_RECVHOPLIMIT, &y, sizeof(y)) < 0)
+    ERR("IPV6_RECVHOPLIMIT");
 
-    if (cm->cmsg_level == SOL_IPV6 && cm->cmsg_type == IPV6_HOPLIMIT)
-      hlim = (int *) CMSG_DATA(cm);
-  }
+  return 0;
+}
 
-  if (s->flags & SKF_LADDR_RX)
+static inline void
+sk_process_cmsg6_pktinfo(sock *s, struct cmsghdr *cm)
+{
+  if (cm->cmsg_type == IPV6_PKTINFO)
   {
-    if (pi)
-    {
-      get_inaddr(&s->laddr, &pi->ipi6_addr);
-      s->lifindex = pi->ipi6_ifindex;
-    }
-    else
-    {
-      s->laddr = IPA_NONE;
-      s->lifindex = 0;
-    }
+    struct in6_pktinfo *pi = (struct in6_pktinfo *) CMSG_DATA(cm);
+    s->laddr = ipa_from_in6(pi->ipi6_addr);
+    s->lifindex = pi->ipi6_ifindex;
   }
+}
 
-  if (s->flags & SKF_TTL_RX)
-    s->ttl = hlim ? *hlim : -1;
-
-  return;
+static inline void
+sk_process_cmsg6_ttl(sock *s, struct cmsghdr *cm)
+{
+  if (cm->cmsg_type == IPV6_HOPLIMIT)
+    s->rcv_ttl = * (int *) CMSG_DATA(cm);
 }
 
-static void
-sysio_prepare_tx_cmsgs(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen)
+static inline void
+sk_prepare_cmsgs6(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen)
 {
   struct cmsghdr *cm;
   struct in6_pktinfo *pi;
@@ -809,104 +716,147 @@ sysio_prepare_tx_cmsgs(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen)
 
   pi = (struct in6_pktinfo *) CMSG_DATA(cm);
   pi->ipi6_ifindex = s->iface ? s->iface->index : 0;
-  set_inaddr(&pi->ipi6_addr, s->saddr);
+  pi->ipi6_addr = ipa_to_in6(s->saddr);
 
   msg->msg_controllen = cm->cmsg_len;
 }
 
-#endif
 
-static char *
-sk_set_ttl_int(sock *s)
+/*
+ *	Miscellaneous socket syscalls
+ */
+
+static inline int
+sk_set_ttl4(sock *s, int ttl)
 {
-#ifdef IPV6
-  if (setsockopt(s->fd, SOL_IPV6, IPV6_UNICAST_HOPS, &s->ttl, sizeof(s->ttl)) < 0)
-    return "IPV6_UNICAST_HOPS";
-#else
-  if (setsockopt(s->fd, SOL_IP, IP_TTL, &s->ttl, sizeof(s->ttl)) < 0)
-    return "IP_TTL";
-#endif
-  return NULL;
+  if (setsockopt(s->fd, SOL_IP, IP_TTL, &ttl, sizeof(ttl)) < 0)
+    ERR("IP_TTL");
+
+  return 0;
 }
 
-#define ERR(x) do { err = x; goto bad; } while(0)
-#define WARN(x) log(L_WARN "sk_setup: %s: %m", x)
+static inline int
+sk_set_ttl6(sock *s, int ttl)
+{
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_UNICAST_HOPS, &ttl, sizeof(ttl)) < 0)
+    ERR("IPV6_UNICAST_HOPS");
 
-static char *
-sk_setup(sock *s)
+  return 0;
+}
+
+static inline int
+sk_set_tos4(sock *s, int tos)
 {
-  int one = 1;
-  int fd = s->fd;
-  char *err = NULL;
+  if (setsockopt(s->fd, SOL_IP, IP_TOS, &tos, sizeof(tos)) < 0)
+    ERR("IP_TOS");
 
-  if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0)
-    ERR("fcntl(O_NONBLOCK)");
-  if (s->type == SK_UNIX)
-    return NULL;
+  return 0;
+}
 
-  if (ipa_nonzero(s->saddr) && !(s->flags & SKF_BIND))
-    s->flags |= SKF_PKTINFO;
+static inline int
+sk_set_tos6(sock *s, int tos)
+{
+  if (setsockopt(s->fd, SOL_IPV6, IPV6_TCLASS, &tos, sizeof(tos)) < 0)
+    ERR("IPV6_TCLASS");
 
-#ifdef CONFIG_USE_HDRINCL
-  if ((s->type == SK_IP) && (s->flags & SKF_PKTINFO))
-  {
-    s->flags &= ~SKF_PKTINFO;
-    s->flags |= SKF_HDRINCL;
-    if (setsockopt(fd, SOL_IP, IP_HDRINCL, &one, sizeof(one)) < 0)
-      ERR("IP_HDRINCL");
-  }
-#endif
+  return 0;
+}
 
-  if (s->iface)
-  {
-#ifdef SO_BINDTODEVICE
-    struct ifreq ifr;
-    strcpy(ifr.ifr_name, s->iface->name);
-    if (setsockopt(s->fd, SOL_SOCKET, SO_BINDTODEVICE, &ifr, sizeof(ifr)) < 0)
-      ERR("SO_BINDTODEVICE");
-#endif
 
-#ifdef CONFIG_UNIX_DONTROUTE
-    if (setsockopt(s->fd, SOL_SOCKET, SO_DONTROUTE, &one, sizeof(one)) < 0)
-      ERR("SO_DONTROUTE");
-#endif
-  }
+/*
+ *	Public socket functions
+ */
 
-  if ((s->ttl >= 0) && (err = sk_set_ttl_int(s)))
-    goto bad;
+/**
+ * sk_setup_multicast - enable multicast for given socket
+ * @s: socket
+ *
+ * Prepare transmission of multicast packets for given datagram socket.
+ * The socket must have defined @iface.
+ *
+ * Result: 0 for success, -1 for an error.
+ */
 
-  if (err = sysio_register_cmsgs(s))
-    goto bad;
+int
+sk_setup_multicast(sock *s)
+{
+  ASSERT(s->iface);
 
+  if (sk_is_ipv4(s))
+    return sk_setup_multicast4(s);
+  else
+    return sk_setup_multicast6(s);
+}
 
-#ifdef IPV6
-  if ((s->tos >= 0) && setsockopt(fd, SOL_IPV6, IPV6_TCLASS, &s->tos, sizeof(s->tos)) < 0)
-    WARN("IPV6_TCLASS");
-#else
-  if ((s->tos >= 0) && setsockopt(fd, SOL_IP, IP_TOS, &s->tos, sizeof(s->tos)) < 0)
-    WARN("IP_TOS");
-#endif
+/**
+ * sk_join_group - join multicast group for given socket
+ * @s: socket
+ * @maddr: multicast address
+ *
+ * Join multicast group for given datagram socket and associated interface.
+ * The socket must have defined @iface.
+ *
+ * Result: 0 for success, -1 for an error.
+ */
 
-  if (s->priority >= 0)
-    sk_set_priority(s, s->priority);
+int
+sk_join_group(sock *s, ip_addr maddr)
+{
+  if (sk_is_ipv4(s))
+    return sk_join_group4(s, maddr);
+  else
+    return sk_join_group6(s, maddr);
+}
 
-#ifdef IPV6
-  if ((s->flags & SKF_V6ONLY) && setsockopt(fd, SOL_IPV6, IPV6_V6ONLY, &one, sizeof(one)) < 0)
-    WARN("IPV6_V6ONLY");
-#endif
+/**
+ * sk_leave_group - leave multicast group for given socket
+ * @s: socket
+ * @maddr: multicast address
+ *
+ * Leave multicast group for given datagram socket and associated interface.
+ * The socket must have defined @iface.
+ *
+ * Result: 0 for success, -1 for an error.
+ */
+
+int
+sk_leave_group(sock *s, ip_addr maddr)
+{
+  if (sk_is_ipv4(s))
+    return sk_leave_group4(s, maddr);
+  else
+    return sk_leave_group6(s, maddr);
+}
+
+/**
+ * sk_setup_broadcast - enable broadcast for given socket
+ * @s: socket
+ *
+ * Allow reception and transmission of broadcast packets for given datagram
+ * socket. The socket must have defined @iface. For transmission, packets should
+ * be send to @brd address of @iface.
+ *
+ * Result: 0 for success, -1 for an error.
+ */
+
+int
+sk_setup_broadcast(sock *s)
+{
+  int y = 1;
+
+  if (setsockopt(s->fd, SOL_SOCKET, SO_BROADCAST, &y, sizeof(y)) < 0)
+    ERR("SO_BROADCAST");
 
-bad:
-  return err;
+  return 0;
 }
 
 /**
- * sk_set_ttl - set transmit TTL for given socket.
+ * sk_set_ttl - set transmit TTL for given socket
  * @s: socket
  * @ttl: TTL value
  *
- * Set TTL for already opened connections when TTL was not set before.
- * Useful for accepted connections when different ones should have 
- * different TTL.
+ * Set TTL for already opened connections when TTL was not set before. Useful
+ * for accepted connections when different ones should have different TTL.
  *
  * Result: 0 for success, -1 for an error.
  */
@@ -914,21 +864,21 @@ bad:
 int
 sk_set_ttl(sock *s, int ttl)
 {
-  char *err;
-
   s->ttl = ttl;
-  if (err = sk_set_ttl_int(s))
-    log(L_ERR "sk_set_ttl: %s: %m", err);
 
-  return (err ? -1 : 0);
+  if (sk_is_ipv4(s))
+    return sk_set_ttl4(s, ttl);
+  else
+    return sk_set_ttl6(s, ttl);
 }
 
 /**
- * sk_set_min_ttl - set minimal accepted TTL for given socket.
+ * sk_set_min_ttl - set minimal accepted TTL for given socket
  * @s: socket
  * @ttl: TTL value
  *
- * Can be used in TTL security implementation
+ * Set minimal accepted TTL for given socket. Can be used for TTL security.
+ * implementations.
  *
  * Result: 0 for success, -1 for an error.
  */
@@ -936,28 +886,24 @@ sk_set_ttl(sock *s, int ttl)
 int
 sk_set_min_ttl(sock *s, int ttl)
 {
-  int err;
-#ifdef IPV6
-  err = sk_set_min_ttl6(s, ttl);
-#else
-  err = sk_set_min_ttl4(s, ttl);
-#endif
-
-  return err;
+  if (sk_is_ipv4(s))
+    return sk_set_min_ttl4(s, ttl);
+  else
+    return sk_set_min_ttl6(s, ttl);
 }
 
+#if 0
 /**
- * sk_set_md5_auth - add / remove MD5 security association for given socket.
+ * sk_set_md5_auth - add / remove MD5 security association for given socket
  * @s: socket
  * @a: IP address of the other side
  * @ifa: Interface for link-local IP address
  * @passwd: password used for MD5 authentication
  *
- * In TCP MD5 handling code in kernel, there is a set of pairs
- * (address, password) used to choose password according to
- * address of the other side. This function is useful for
- * listening socket, for active sockets it is enough to set
- * s->password field.
+ * In TCP MD5 handling code in kernel, there is a set of pairs (address,
+ * password) used to choose password according to address of the other side.
+ * This function is useful for listening socket, for active sockets it is enough
+ * to set s->password field.
  *
  * When called with passwd != NULL, the new pair is added,
  * When called with passwd == NULL, the existing pair is removed.
@@ -967,41 +913,33 @@ sk_set_min_ttl(sock *s, int ttl)
 
 int
 sk_set_md5_auth(sock *s, ip_addr a, struct iface *ifa, char *passwd)
-{
-  sockaddr sa;
-  fill_in_sockaddr(&sa, a, ifa, 0);
-  return sk_set_md5_auth_int(s, &sa, passwd);
-}
-
-int
-sk_set_broadcast(sock *s, int enable)
-{
-  if (setsockopt(s->fd, SOL_SOCKET, SO_BROADCAST, &enable, sizeof(enable)) < 0)
-    {
-      log(L_ERR "sk_set_broadcast: SO_BROADCAST: %m");
-      return -1;
-    }
-
-  return 0;
-}
-
+{ DUMMY; }
+#endif
 
-#ifdef IPV6
+/**
+ * sk_set_ipv6_checksum - specify IPv6 checksum offset for given socket
+ * @s: socket
+ * @offset: offset
+ *
+ * Specify IPv6 checksum field offset for given raw IPv6 socket. After that, the
+ * kernel will automatically fill it for outgoing packets and check it for
+ * incoming packets. Should not be used on ICMPv6 sockets, where the position is
+ * known to the kernel.
+ *
+ * Result: 0 for success, -1 for an error.
+ */
 
 int
 sk_set_ipv6_checksum(sock *s, int offset)
 {
   if (setsockopt(s->fd, SOL_IPV6, IPV6_CHECKSUM, &offset, sizeof(offset)) < 0)
-    {
-      log(L_ERR "sk_set_ipv6_checksum: IPV6_CHECKSUM: %m");
-      return -1;
-    }
+    ERR("IPV6_CHECKSUM");
 
   return 0;
 }
 
 int
-sk_set_icmp_filter(sock *s, int p1, int p2)
+sk_set_icmp6_filter(sock *s, int p1, int p2)
 {
   /* a bit of lame interface, but it is here only for Radv */
   struct icmp6_filter f;
@@ -1011,132 +949,291 @@ sk_set_icmp_filter(sock *s, int p1, int p2)
   ICMP6_FILTER_SETPASS(p2, &f);
 
   if (setsockopt(s->fd, SOL_ICMPV6, ICMP6_FILTER, &f, sizeof(f)) < 0)
-    {
-      log(L_ERR "sk_setup_icmp_filter: ICMP6_FILTER: %m");
-      return -1;
-    }
+    ERR("ICMP6_FILTER");
 
   return 0;
 }
 
-int
-sk_setup_multicast(sock *s)
+void
+sk_log_error(sock *s, const char *p)
 {
-  char *err;
-  int zero = 0;
-  int index;
+  log(L_ERR "%s: Socket error: %s%#m", p, s->err);
+}
 
-  ASSERT(s->iface);
 
-  index = s->iface->index;
-  if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_HOPS, &s->ttl, sizeof(s->ttl)) < 0)
-    ERR("IPV6_MULTICAST_HOPS");
-  if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_LOOP, &zero, sizeof(zero)) < 0)
-    ERR("IPV6_MULTICAST_LOOP");
-  if (setsockopt(s->fd, SOL_IPV6, IPV6_MULTICAST_IF, &index, sizeof(index)) < 0)
-    ERR("IPV6_MULTICAST_IF");
+/*
+ *	Actual struct birdsock code
+ */
 
-  return 0;
+static list sock_list;
+static struct birdsock *current_sock;
+static struct birdsock *stored_sock;
+static int sock_recalc_fdsets_p;
 
-bad:
-  log(L_ERR "sk_setup_multicast: %s: %m", err);
-  return -1;
+static inline sock *
+sk_next(sock *s)
+{
+  if (!s->n.next->next)
+    return NULL;
+  else
+    return SKIP_BACK(sock, n, s->n.next);
 }
 
-#ifdef CONFIG_IPV6_GLIBC_20
-#define ipv6mr_interface ipv6mr_ifindex
-#endif
-
-int
-sk_join_group(sock *s, ip_addr maddr)
+static void
+sk_alloc_bufs(sock *s)
 {
-  struct ipv6_mreq mreq;
+  if (!s->rbuf && s->rbsize)
+    s->rbuf = s->rbuf_alloc = xmalloc(s->rbsize);
+  s->rpos = s->rbuf;
+  if (!s->tbuf && s->tbsize)
+    s->tbuf = s->tbuf_alloc = xmalloc(s->tbsize);
+  s->tpos = s->ttx = s->tbuf;
+}
 
-  set_inaddr(&mreq.ipv6mr_multiaddr, maddr);
-  mreq.ipv6mr_interface = s->iface->index;
+static void
+sk_free_bufs(sock *s)
+{
+  if (s->rbuf_alloc)
+  {
+    xfree(s->rbuf_alloc);
+    s->rbuf = s->rbuf_alloc = NULL;
+  }
+  if (s->tbuf_alloc)
+  {
+    xfree(s->tbuf_alloc);
+    s->tbuf = s->tbuf_alloc = NULL;
+  }
+}
 
-  if (setsockopt(s->fd, SOL_IPV6, IPV6_JOIN_GROUP, &mreq, sizeof(mreq)) < 0)
-    {
-      log(L_ERR "sk_join_group: IPV6_JOIN_GROUP: %m");
-      return -1;
-    }
+static void
+sk_free(resource *r)
+{
+  sock *s = (sock *) r;
 
-  return 0;
+  sk_free_bufs(s);
+  if (s->fd >= 0)
+  {
+    close(s->fd);
+
+    /* FIXME: we should call sk_stop() for SKF_THREAD sockets */
+    if (s->flags & SKF_THREAD)
+      return;
+
+    if (s == current_sock)
+      current_sock = sk_next(s);
+    if (s == stored_sock)
+      stored_sock = sk_next(s);
+    rem_node(&s->n);
+    sock_recalc_fdsets_p = 1;
+  }
 }
 
-int
-sk_leave_group(sock *s, ip_addr maddr)
+void
+sk_set_rbsize(sock *s, uint val)
 {
-  struct ipv6_mreq mreq;
-	
-  set_inaddr(&mreq.ipv6mr_multiaddr, maddr);
-  mreq.ipv6mr_interface = s->iface->index;
+  ASSERT(s->rbuf_alloc == s->rbuf);
 
-  if (setsockopt(s->fd, SOL_IPV6, IPV6_LEAVE_GROUP, &mreq, sizeof(mreq)) < 0)
-    {
-      log(L_ERR "sk_leave_group: IPV6_LEAVE_GROUP: %m");
-      return -1;
-    }
+  if (s->rbsize == val)
+    return;
 
-  return 0;
+  s->rbsize = val;
+  xfree(s->rbuf_alloc);
+  s->rbuf_alloc = xmalloc(val);
+  s->rpos = s->rbuf = s->rbuf_alloc;
 }
 
-#else /* IPV4 */
-
-int
-sk_setup_multicast(sock *s)
+void
+sk_set_tbsize(sock *s, uint val)
 {
-  char *err;
+  ASSERT(s->tbuf_alloc == s->tbuf);
 
-  ASSERT(s->iface);
+  if (s->tbsize == val)
+    return;
 
-  if (err = sysio_setup_multicast(s))
-    {
-      log(L_ERR "sk_setup_multicast: %s: %m", err);
-      return -1;
-    }
+  byte *old_tbuf = s->tbuf;
 
-  return 0;
+  s->tbsize = val;
+  s->tbuf = s->tbuf_alloc = xrealloc(s->tbuf_alloc, val);
+  s->tpos = s->tbuf + (s->tpos - old_tbuf);
+  s->ttx  = s->tbuf + (s->ttx  - old_tbuf);
 }
 
-int
-sk_join_group(sock *s, ip_addr maddr)
+void
+sk_set_tbuf(sock *s, void *tbuf)
 {
- char *err;
+  s->tbuf = tbuf ?: s->tbuf_alloc;
+  s->ttx = s->tpos = s->tbuf;
+}
 
- if (err = sysio_join_group(s, maddr))
-    {
-      log(L_ERR "sk_join_group: %s: %m", err);
-      return -1;
-    }
+void
+sk_reallocate(sock *s)
+{
+  sk_free_bufs(s);
+  sk_alloc_bufs(s);
+}
 
-  return 0;
+static void
+sk_dump(resource *r)
+{
+  sock *s = (sock *) r;
+  static char *sk_type_names[] = { "TCP<", "TCP>", "TCP", "UDP", NULL, "IP", NULL, "MAGIC", "UNIX<", "UNIX", "DEL!" };
+
+  debug("(%s, ud=%p, sa=%08x, sp=%d, da=%08x, dp=%d, tos=%d, ttl=%d, if=%s)\n",
+	sk_type_names[s->type],
+	s->data,
+	s->saddr,
+	s->sport,
+	s->daddr,
+	s->dport,
+	s->tos,
+	s->ttl,
+	s->iface ? s->iface->name : "none");
 }
 
-int
-sk_leave_group(sock *s, ip_addr maddr)
+static struct resclass sk_class = {
+  "Socket",
+  sizeof(sock),
+  sk_free,
+  sk_dump,
+  NULL,
+  NULL
+};
+
+/**
+ * sk_new - create a socket
+ * @p: pool
+ *
+ * This function creates a new socket resource. If you want to use it,
+ * you need to fill in all the required fields of the structure and
+ * call sk_open() to do the actual opening of the socket.
+ *
+ * The real function name is sock_new(), sk_new() is a macro wrapper
+ * to avoid collision with OpenSSL.
+ */
+sock *
+sock_new(pool *p)
+{
+  sock *s = ralloc(p, &sk_class);
+  s->pool = p;
+  // s->saddr = s->daddr = IPA_NONE;
+  s->tos = s->priority = s->ttl = -1;
+  s->fd = -1;
+  return s;
+}
+
+static int
+sk_setup(sock *s)
 {
- char *err;
+  int y = 1;
+  int fd = s->fd;
 
- if (err = sysio_leave_group(s, maddr))
-    {
-      log(L_ERR "sk_leave_group: %s: %m", err);
+  if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0)
+    ERR("O_NONBLOCK");
+
+  if (!s->af)
+    return 0;
+
+  if (ipa_nonzero(s->saddr) && !(s->flags & SKF_BIND))
+    s->flags |= SKF_PKTINFO;
+
+#ifdef CONFIG_USE_HDRINCL
+  if (sk_is_ipv4(s) && (s->type == SK_IP) && (s->flags & SKF_PKTINFO))
+  {
+    s->flags &= ~SKF_PKTINFO;
+    s->flags |= SKF_HDRINCL;
+    if (setsockopt(fd, SOL_IP, IP_HDRINCL, &y, sizeof(y)) < 0)
+      ERR("IP_HDRINCL");
+  }
+#endif
+
+  if (s->iface)
+  {
+#ifdef SO_BINDTODEVICE
+    struct ifreq ifr;
+    strcpy(ifr.ifr_name, s->iface->name);
+    if (setsockopt(s->fd, SOL_SOCKET, SO_BINDTODEVICE, &ifr, sizeof(ifr)) < 0)
+      ERR("SO_BINDTODEVICE");
+#endif
+
+#ifdef CONFIG_UNIX_DONTROUTE
+    if (setsockopt(s->fd, SOL_SOCKET, SO_DONTROUTE, &y, sizeof(y)) < 0)
+      ERR("SO_DONTROUTE");
+#endif
+  }
+
+  if (s->priority >= 0)
+    if (sk_set_priority(s, s->priority) < 0)
       return -1;
-    }
+
+  if (sk_is_ipv4(s))
+  {
+    if (s->flags & SKF_LADDR_RX)
+      if (sk_request_cmsg4_pktinfo(s) < 0)
+	return -1;
+
+    if (s->flags & SKF_TTL_RX)
+      if (sk_request_cmsg4_ttl(s) < 0)
+	return -1;
+
+    if ((s->type == SK_UDP) || (s->type == SK_IP))
+      if (sk_disable_mtu_disc4(s) < 0)
+	return -1;
+
+    if (s->ttl >= 0)
+      if (sk_set_ttl4(s, s->ttl) < 0)
+	return -1;
+
+    if (s->tos >= 0)
+      if (sk_set_tos4(s, s->tos) < 0)
+	return -1;
+  }
+
+  if (sk_is_ipv6(s))
+  {
+    if (s->flags & SKF_V6ONLY)
+      if (setsockopt(fd, SOL_IPV6, IPV6_V6ONLY, &y, sizeof(y)) < 0)
+	ERR("IPV6_V6ONLY");
+
+    if (s->flags & SKF_LADDR_RX)
+      if (sk_request_cmsg6_pktinfo(s) < 0)
+	return -1;
+
+    if (s->flags & SKF_TTL_RX)
+      if (sk_request_cmsg6_ttl(s) < 0)
+	return -1;
+
+    if ((s->type == SK_UDP) || (s->type == SK_IP))
+      if (sk_disable_mtu_disc6(s) < 0)
+	return -1;
+
+    if (s->ttl >= 0)
+      if (sk_set_ttl6(s, s->ttl) < 0)
+	return -1;
+
+    if (s->tos >= 0)
+      if (sk_set_tos6(s, s->tos) < 0)
+	return -1;
+  }
 
   return 0;
 }
 
-#endif 
-
+static void
+sk_insert(sock *s)
+{
+  add_tail(&sock_list, &s->n);
+  sock_recalc_fdsets_p = 1;
+}
 
 static void
 sk_tcp_connected(sock *s)
 {
-  sockaddr lsa;
-  int lsa_len = sizeof(lsa);
-  if (getsockname(s->fd, (struct sockaddr *) &lsa, &lsa_len) == 0)
-    get_sockaddr(&lsa, &s->saddr, &s->iface, &s->sport, 1);
+  sockaddr sa;
+  int sa_len = sizeof(sa);
+
+  if ((getsockname(s->fd, &sa.sa, &sa_len) < 0) ||
+      (sockaddr_read(&sa, s->af, &s->saddr, &s->iface, &s->sport) < 0))
+    log(L_WARN "SOCK: Cannot get local IP address for TCP>");
 
   s->type = SK_TCP;
   sk_alloc_bufs(s);
@@ -1144,44 +1241,55 @@ sk_tcp_connected(sock *s)
 }
 
 static int
-sk_passive_connected(sock *s, struct sockaddr *sa, int al, int type)
+sk_passive_connected(sock *s, int type)
 {
-  int fd = accept(s->fd, sa, &al);
-  if (fd >= 0)
-    {
-      sock *t = sk_new(s->pool);
-      char *err;
-      t->type = type;
-      t->fd = fd;
-      t->ttl = s->ttl;
-      t->tos = s->tos;
-      t->rbsize = s->rbsize;
-      t->tbsize = s->tbsize;
-      if (type == SK_TCP)
-	{
-	  sockaddr lsa;
-	  int lsa_len = sizeof(lsa);
-	  if (getsockname(fd, (struct sockaddr *) &lsa, &lsa_len) == 0)
-	    get_sockaddr(&lsa, &t->saddr, &t->iface, &t->sport, 1);
+  sockaddr loc_sa, rem_sa;
+  int loc_sa_len = sizeof(loc_sa);
+  int rem_sa_len = sizeof(rem_sa);
 
-	  get_sockaddr((sockaddr *) sa, &t->daddr, &t->iface, &t->dport, 1);
-	}
-      sk_insert(t);
-      if (err = sk_setup(t))
-	{
-	  log(L_ERR "Incoming connection: %s: %m", err);
-	  rfree(t);
-	  return 1;
-	}
-      sk_alloc_bufs(t);
-      s->rx_hook(t, 0);
-      return 1;
-    }
-  else if (errno != EINTR && errno != EAGAIN)
-    {
+  int fd = accept(s->fd, ((type == SK_TCP) ? &rem_sa.sa : NULL), &rem_sa_len);
+  if (fd < 0)
+  {
+    if ((errno != EINTR) && (errno != EAGAIN))
       s->err_hook(s, errno);
-    }
-  return 0;
+    return 0;
+  }
+
+  sock *t = sk_new(s->pool);
+  t->type = type;
+  t->fd = fd;
+  t->af = s->af;
+  t->ttl = s->ttl;
+  t->tos = s->tos;
+  t->rbsize = s->rbsize;
+  t->tbsize = s->tbsize;
+
+  if (type == SK_TCP)
+  {
+    if ((getsockname(fd, &loc_sa.sa, &loc_sa_len) < 0) ||
+	(sockaddr_read(&loc_sa, s->af, &t->saddr, &t->iface, &t->sport) < 0))
+      log(L_WARN "SOCK: Cannot get local IP address for TCP<");
+
+    if (sockaddr_read(&rem_sa, s->af, &t->daddr, &t->iface, &t->dport) < 0)
+      log(L_WARN "SOCK: Cannot get remote IP address for TCP<");
+  }
+
+  if (sk_setup(t) < 0)
+  {
+    /* FIXME: Call err_hook instead ? */
+    log(L_ERR "SOCK: Incoming connection: %s%#m", t->err);
+
+    /* FIXME: handle it better in rfree() */
+    close(t->fd);	
+    t->fd = -1;
+    rfree(t);
+    return 1;
+  }
+
+  sk_insert(t);
+  sk_alloc_bufs(t);
+  s->rx_hook(t, 0);
+  return 1;
 }
 
 /**
@@ -1197,161 +1305,185 @@ sk_passive_connected(sock *s, struct sockaddr *sa, int al, int type)
 int
 sk_open(sock *s)
 {
-  int fd;
-  int one = 1;
+  int af = BIRD_AF;
+  int fd = -1;
   int do_bind = 0;
   int bind_port = 0;
   ip_addr bind_addr = IPA_NONE;
   sockaddr sa;
-  char *err;
 
   switch (s->type)
-    {
-    case SK_TCP_ACTIVE:
-      s->ttx = "";			/* Force s->ttx != s->tpos */
-      /* Fall thru */
-    case SK_TCP_PASSIVE:
-      fd = socket(BIRD_PF, SOCK_STREAM, IPPROTO_TCP);
-      bind_port = s->sport;
-      bind_addr = s->saddr;
-      do_bind = bind_port || ipa_nonzero(bind_addr);
-      break;
-
-    case SK_UDP:
-      fd = socket(BIRD_PF, SOCK_DGRAM, IPPROTO_UDP);
-      bind_port = s->sport;
-      bind_addr = (s->flags & SKF_BIND) ? s->saddr : IPA_NONE;
-      do_bind = 1;
-      break;
-
-    case SK_IP:
-      fd = socket(BIRD_PF, SOCK_RAW, s->dport);
-      bind_port = 0;
-      bind_addr = (s->flags & SKF_BIND) ? s->saddr : IPA_NONE;
-      do_bind = ipa_nonzero(bind_addr);
-      break;
-
-    case SK_MAGIC:
-      fd = s->fd;
-      break;
-
-    default:
-      bug("sk_open() called for invalid sock type %d", s->type);
-    }
+  {
+  case SK_TCP_ACTIVE:
+    s->ttx = "";			/* Force s->ttx != s->tpos */
+    /* Fall thru */
+  case SK_TCP_PASSIVE:
+    fd = socket(af, SOCK_STREAM, IPPROTO_TCP);
+    bind_port = s->sport;
+    bind_addr = s->saddr;
+    do_bind = bind_port || ipa_nonzero(bind_addr);
+    break;
+  
+  case SK_UDP:
+    fd = socket(af, SOCK_DGRAM, IPPROTO_UDP);
+    bind_port = s->sport;
+    bind_addr = (s->flags & SKF_BIND) ? s->saddr : IPA_NONE;
+    do_bind = 1;
+    break;
+
+  case SK_IP:
+    fd = socket(af, SOCK_RAW, s->dport);
+    bind_port = 0;
+    bind_addr = (s->flags & SKF_BIND) ? s->saddr : IPA_NONE;
+    do_bind = ipa_nonzero(bind_addr);
+    break;
+
+  case SK_MAGIC:
+    af = 0;
+    fd = s->fd;
+    break;
+
+  default:
+    bug("sk_open() called for invalid sock type %d", s->type);
+  }
+
   if (fd < 0)
-    die("sk_open: socket: %m");
+    ERR("socket");
+
+  s->af = af;
   s->fd = fd;
 
-  if (err = sk_setup(s))
-    goto bad;
+  if (sk_setup(s) < 0)
+    goto err;
 
   if (do_bind)
+  {
+    if (bind_port)
     {
-      if (bind_port)
-	{
-	  if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &one, sizeof(one)) < 0)
-	    ERR("SO_REUSEADDR");
+      int y = 1;
+
+      if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &y, sizeof(y)) < 0)
+	ERR2("SO_REUSEADDR");
 
 #ifdef CONFIG_NO_IFACE_BIND
-	  /* Workaround missing ability to bind to an iface */
-	  if ((s->type == SK_UDP) && s->iface && ipa_zero(bind_addr))
-	  {
-	    if (setsockopt(fd, SOL_SOCKET, SO_REUSEPORT, &one, sizeof(one)) < 0)
-	      ERR("SO_REUSEPORT");
-	  }
+      /* Workaround missing ability to bind to an iface */
+      if ((s->type == SK_UDP) && s->iface && ipa_zero(bind_addr))
+      {
+	if (setsockopt(fd, SOL_SOCKET, SO_REUSEPORT, &y, sizeof(y)) < 0)
+	  ERR2("SO_REUSEPORT");
+      }
 #endif
-	}
-
-      fill_in_sockaddr(&sa, bind_addr, s->iface, bind_port);
-      if (bind(fd, (struct sockaddr *) &sa, sizeof(sa)) < 0)
-	ERR("bind");
     }
 
-  fill_in_sockaddr(&sa, s->daddr, s->iface, s->dport);
+    sockaddr_fill(&sa, af, bind_addr, s->iface, bind_port);
+    if (bind(fd, &sa.sa, SA_LEN(sa)) < 0)
+      ERR2("bind");
+  }
 
   if (s->password)
-    {
-      int rv = sk_set_md5_auth_int(s, &sa, s->password);
-      if (rv < 0)
-	goto bad_no_log;
-    }
+    if (sk_set_md5_auth(s, s->daddr, s->iface, s->password) < 0)
+      goto err;
 
   switch (s->type)
-    {
-    case SK_TCP_ACTIVE:
-      if (connect(fd, (struct sockaddr *) &sa, sizeof(sa)) >= 0)
-	sk_tcp_connected(s);
-      else if (errno != EINTR && errno != EAGAIN && errno != EINPROGRESS &&
-	       errno != ECONNREFUSED && errno != EHOSTUNREACH && errno != ENETUNREACH)
-	ERR("connect");
-      break;
-    case SK_TCP_PASSIVE:
-      if (listen(fd, 8))
-	ERR("listen");
-      break;
-    case SK_MAGIC:
-      break;
-    default:
-      sk_alloc_bufs(s);
-#ifdef IPV6
-#ifdef IPV6_MTU_DISCOVER
-      {
-	int dont = IPV6_PMTUDISC_DONT;
-	if (setsockopt(fd, SOL_IPV6, IPV6_MTU_DISCOVER, &dont, sizeof(dont)) < 0)
-	  ERR("IPV6_MTU_DISCOVER");
-      }
-#endif
-#else
-#ifdef IP_PMTUDISC
-      {
-	int dont = IP_PMTUDISC_DONT;
-	if (setsockopt(fd, SOL_IP, IP_PMTUDISC, &dont, sizeof(dont)) < 0)
-	  ERR("IP_PMTUDISC");
-      }
-#endif
-#endif
-    }
+  {
+  case SK_TCP_ACTIVE:
+    sockaddr_fill(&sa, af, s->daddr, s->iface, s->dport);
+    if (connect(fd, &sa.sa, SA_LEN(sa)) >= 0)
+      sk_tcp_connected(s);
+    else if (errno != EINTR && errno != EAGAIN && errno != EINPROGRESS &&
+	     errno != ECONNREFUSED && errno != EHOSTUNREACH && errno != ENETUNREACH)
+      ERR2("connect");
+    break;
+
+  case SK_TCP_PASSIVE:
+    if (listen(fd, 8) < 0)
+      ERR2("listen");
+    break;
+
+  case SK_MAGIC:
+    break;
+
+  default:
+    sk_alloc_bufs(s);
+  }
 
   if (!(s->flags & SKF_THREAD))
     sk_insert(s);
   return 0;
 
-bad:
-  log(L_ERR "sk_open: %s: %m", err);
-bad_no_log:
+err:
   close(fd);
   s->fd = -1;
   return -1;
 }
 
-void
+int
 sk_open_unix(sock *s, char *name)
 {
-  int fd;
   struct sockaddr_un sa;
-  char *err;
+  int fd;
+
+  /* We are sloppy during error (leak fd and not set s->err), but we die anyway */
 
   fd = socket(AF_UNIX, SOCK_STREAM, 0);
   if (fd < 0)
-    ERR("socket");
-  s->fd = fd;
-  if (err = sk_setup(s))
-    goto bad;
-  unlink(name);
+    return -1;
+
+  if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0)
+    return -1;
 
   /* Path length checked in test_old_bird() */
   sa.sun_family = AF_UNIX;
   strcpy(sa.sun_path, name);
+
   if (bind(fd, (struct sockaddr *) &sa, SUN_LEN(&sa)) < 0)
-    ERR("bind");
-  if (listen(fd, 8))
-    ERR("listen");
+    return -1;
+
+  if (listen(fd, 8) < 0)
+    return -1;
+
+  s->fd = fd;
   sk_insert(s);
-  return;
+  return 0;
+}
+
+
+#define CMSG_RX_SPACE MAX(CMSG4_SPACE_PKTINFO+CMSG4_SPACE_TTL, \
+			  CMSG6_SPACE_PKTINFO+CMSG6_SPACE_TTL)
+#define CMSG_TX_SPACE MAX(CMSG4_SPACE_PKTINFO,CMSG6_SPACE_PKTINFO)
+
+static void
+sk_prepare_cmsgs(sock *s, struct msghdr *msg, void *cbuf, size_t cbuflen)
+{
+  if (sk_is_ipv4(s))
+    sk_prepare_cmsgs4(s, msg, cbuf, cbuflen);
+  else
+    sk_prepare_cmsgs6(s, msg, cbuf, cbuflen);
+}
+
+static void
+sk_process_cmsgs(sock *s, struct msghdr *msg)
+{
+  struct cmsghdr *cm;
+
+  s->laddr = IPA_NONE;
+  s->lifindex = 0;
+  s->rcv_ttl = -1;
 
- bad:
-  log(L_ERR "sk_open_unix: %s: %m", err);
-  die("Unable to create control socket %s", name);
+  for (cm = CMSG_FIRSTHDR(msg); cm != NULL; cm = CMSG_NXTHDR(msg, cm))
+  {
+    if ((cm->cmsg_level == SOL_IP) && sk_is_ipv4(s))
+    {
+      sk_process_cmsg4_pktinfo(s, cm);
+      sk_process_cmsg4_ttl(s, cm);
+    }
+
+    if ((cm->cmsg_level == SOL_IPV6) && sk_is_ipv6(s))
+    {
+      sk_process_cmsg6_pktinfo(s, cm);
+      sk_process_cmsg6_ttl(s, cm);
+    }
+  }
 }
 
 
@@ -1362,11 +1494,11 @@ sk_sendmsg(sock *s)
   byte cmsg_buf[CMSG_TX_SPACE];
   sockaddr dst;
 
-  fill_in_sockaddr(&dst, s->daddr, s->iface, s->dport);
+  sockaddr_fill(&dst, s->af, s->daddr, s->iface, s->dport);
 
   struct msghdr msg = {
-    .msg_name = &dst,
-    .msg_namelen = sizeof(dst),
+    .msg_name = &dst.sa,
+    .msg_namelen = SA_LEN(dst),
     .msg_iov = &iov,
     .msg_iovlen = 1
   };
@@ -1377,14 +1509,14 @@ sk_sendmsg(sock *s)
 
   if (s->flags & SKF_HDRINCL)
   {
-    fill_ip_header(s, hdr, iov.iov_len);
+    sk_prepare_ip_header(s, hdr, iov.iov_len);
     msg.msg_iov = iov2;
     msg.msg_iovlen = 2;
   }
 #endif
 
   if (s->flags & SKF_PKTINFO)
-    sysio_prepare_tx_cmsgs(s, &msg, cmsg_buf, sizeof(cmsg_buf));
+    sk_prepare_cmsgs(s, &msg, cmsg_buf, sizeof(cmsg_buf));
 
   return sendmsg(s->fd, &msg, 0);
 }
@@ -1397,8 +1529,8 @@ sk_recvmsg(sock *s)
   sockaddr src;
 
   struct msghdr msg = {
-    .msg_name = &src,
-    .msg_namelen = sizeof(src),
+    .msg_name = &src.sa,
+    .msg_namelen = sizeof(src), // XXXX ??
     .msg_iov = &iov,
     .msg_iovlen = 1,
     .msg_control = cmsg_buf,
@@ -1415,8 +1547,8 @@ sk_recvmsg(sock *s)
   //    rv = ipv4_skip_header(pbuf, rv);
   //endif
 
-  get_sockaddr(&src, &s->faddr, NULL, &s->fport, 1);
-  sysio_process_rx_cmsgs(s, &msg);
+  sockaddr_read(&src, s->af, &s->faddr, NULL, &s->fport);
+  sk_process_cmsgs(s, &msg);
 
   if (msg.msg_flags & MSG_TRUNC)
     s->flags |= SKF_TRUNCATED;
@@ -1435,55 +1567,57 @@ sk_maybe_write(sock *s)
   int e;
 
   switch (s->type)
+  {
+  case SK_TCP:
+  case SK_MAGIC:
+  case SK_UNIX:
+    while (s->ttx != s->tpos)
     {
-    case SK_TCP:
-    case SK_MAGIC:
-    case SK_UNIX:
-      while (s->ttx != s->tpos)
+      e = write(s->fd, s->ttx, s->tpos - s->ttx);
+
+      if (e < 0)
+      {
+	if (errno != EINTR && errno != EAGAIN)
 	{
-	  e = write(s->fd, s->ttx, s->tpos - s->ttx);
-	  if (e < 0)
-	    {
-	      if (errno != EINTR && errno != EAGAIN)
-		{
-		  reset_tx_buffer(s);
-		  /* EPIPE is just a connection close notification during TX */
-		  s->err_hook(s, (errno != EPIPE) ? errno : 0);
-		  return -1;
-		}
-	      return 0;
-	    }
-	  s->ttx += e;
+	  reset_tx_buffer(s);
+	  /* EPIPE is just a connection close notification during TX */
+	  s->err_hook(s, (errno != EPIPE) ? errno : 0);
+	  return -1;
 	}
-      reset_tx_buffer(s);
-      return 1;
-    case SK_UDP:
-    case SK_IP:
-      {
-	if (s->tbuf == s->tpos)
-	  return 1;
-
-	e = sk_sendmsg(s);
-
-	if (e < 0)
-	  {
-	    if (errno != EINTR && errno != EAGAIN)
-	      {
-		reset_tx_buffer(s);
-		s->err_hook(s, errno);
-		return -1;
-	      }
-
-	    if (!s->tx_hook)
-	      reset_tx_buffer(s);
-	    return 0;
-	  }
-	reset_tx_buffer(s);
+	return 0;
+      }
+      s->ttx += e;
+    }
+    reset_tx_buffer(s);
+    return 1;
+
+  case SK_UDP:
+  case SK_IP:
+    {
+      if (s->tbuf == s->tpos)
 	return 1;
+
+      e = sk_sendmsg(s);
+
+      if (e < 0)
+      {
+	if (errno != EINTR && errno != EAGAIN)
+	{
+	  reset_tx_buffer(s);
+	  s->err_hook(s, errno);
+	  return -1;
+	}
+
+	if (!s->tx_hook)
+	  reset_tx_buffer(s);
+	return 0;
       }
-    default:
-      bug("sk_maybe_write: unknown socket type %d", s->type);
+      reset_tx_buffer(s);
+      return 1;
     }
+  default:
+    bug("sk_maybe_write: unknown socket type %d", s->type);
+  }
 }
 
 int
@@ -1573,88 +1707,86 @@ int
 sk_read(sock *s)
 {
   switch (s->type)
+  {
+  case SK_TCP_PASSIVE:
+    return sk_passive_connected(s, SK_TCP);
+
+  case SK_UNIX_PASSIVE:
+    return sk_passive_connected(s, SK_UNIX);
+
+  case SK_TCP:
+  case SK_UNIX:
     {
-    case SK_TCP_PASSIVE:
-      {
-	sockaddr sa;
-	return sk_passive_connected(s, (struct sockaddr *) &sa, sizeof(sa), SK_TCP);
-      }
-    case SK_UNIX_PASSIVE:
+      int c = read(s->fd, s->rpos, s->rbuf + s->rbsize - s->rpos);
+
+      if (c < 0)
       {
-	struct sockaddr_un sa;
-	return sk_passive_connected(s, (struct sockaddr *) &sa, sizeof(sa), SK_UNIX);
+	if (errno != EINTR && errno != EAGAIN)
+	  s->err_hook(s, errno);
       }
-    case SK_TCP:
-    case SK_UNIX:
+      else if (!c)
+	s->err_hook(s, 0);
+      else
       {
-	int c = read(s->fd, s->rpos, s->rbuf + s->rbsize - s->rpos);
-
-	if (c < 0)
-	  {
-	    if (errno != EINTR && errno != EAGAIN)
-	      s->err_hook(s, errno);
-	  }
-	else if (!c)
-	  s->err_hook(s, 0);
-	else
-	  {
-	    s->rpos += c;
-	    if (s->rx_hook(s, s->rpos - s->rbuf))
-	      {
-		/* We need to be careful since the socket could have been deleted by the hook */
-		if (current_sock == s)
-		  s->rpos = s->rbuf;
-	      }
-	    return 1;
-	  }
-	return 0;
+	s->rpos += c;
+	if (s->rx_hook(s, s->rpos - s->rbuf))
+	{
+	  /* We need to be careful since the socket could have been deleted by the hook */
+	  if (current_sock == s)
+	    s->rpos = s->rbuf;
+	}
+	return 1;
       }
-    case SK_MAGIC:
-      return s->rx_hook(s, 0);
-    default:
-      {
-	int e;
+      return 0;
+    }
 
-	e = sk_recvmsg(s);
+  case SK_MAGIC:
+    return s->rx_hook(s, 0);
 
-	if (e < 0)
-	  {
-	    if (errno != EINTR && errno != EAGAIN)
-	      s->err_hook(s, errno);
-	    return 0;
-	  }
+  default:
+    {
+      int e = sk_recvmsg(s);
 
-	s->rpos = s->rbuf + e;
-	s->rx_hook(s, e);
-	return 1;
+      if (e < 0)
+      {
+	if (errno != EINTR && errno != EAGAIN)
+	  s->err_hook(s, errno);
+	return 0;
       }
+
+      s->rpos = s->rbuf + e;
+      s->rx_hook(s, e);
+      return 1;
     }
+  }
 }
 
 int
 sk_write(sock *s)
 {
   switch (s->type)
+  {
+  case SK_TCP_ACTIVE:
     {
-    case SK_TCP_ACTIVE:
-      {
-	sockaddr sa;
-	fill_in_sockaddr(&sa, s->daddr, s->iface, s->dport);
-	if (connect(s->fd, (struct sockaddr *) &sa, sizeof(sa)) >= 0 || errno == EISCONN)
-	  sk_tcp_connected(s);
-	else if (errno != EINTR && errno != EAGAIN && errno != EINPROGRESS)
-	  s->err_hook(s, errno);
-	return 0;
-      }
-    default:
-      if (s->ttx != s->tpos && sk_maybe_write(s) > 0)
-	{
-	  if (s->tx_hook)
-	    s->tx_hook(s);
-	  return 1;
-	}
+      sockaddr sa;
+      sockaddr_fill(&sa, s->af, s->daddr, s->iface, s->dport);
+
+      if (connect(s->fd, &sa.sa, SA_LEN(sa)) >= 0 || errno == EISCONN)
+	sk_tcp_connected(s);
+      else if (errno != EINTR && errno != EAGAIN && errno != EINPROGRESS)
+	s->err_hook(s, errno);
       return 0;
     }
+
+  default:
+    if (s->ttx != s->tpos && sk_maybe_write(s) > 0)
+    {
+      if (s->tx_hook)
+	s->tx_hook(s);
+      return 1;
+    }
+    return 0;
+  }
 }
 
 void
@@ -1665,16 +1797,14 @@ sk_dump_all(void)
 
   debug("Open sockets:\n");
   WALK_LIST(n, sock_list)
-    {
-      s = SKIP_BACK(sock, n, n);
-      debug("%p ", s);
-      sk_dump(&s->r);
-    }
+  {
+    s = SKIP_BACK(sock, n, n);
+    debug("%p ", s);
+    sk_dump(&s->r);
+  }
   debug("\n");
 }
 
-#undef ERR
-#undef WARN
 
 /*
  *	Main I/O Loop
diff --git a/sysdep/unix/main.c b/sysdep/unix/main.c
index 31094c52..61b306dc 100644
--- a/sysdep/unix/main.c
+++ b/sysdep/unix/main.c
@@ -463,7 +463,12 @@ cli_init_unix(uid_t use_uid, gid_t use_gid)
   s->type = SK_UNIX_PASSIVE;
   s->rx_hook = cli_connect;
   s->rbsize = 1024;
-  sk_open_unix(s, path_control_socket);
+
+  /* Return value intentionally ignored */
+  unlink(path_control_socket);
+
+  if (sk_open_unix(s, path_control_socket) < 0)
+    die("Cannot create control socket %s: %m", path_control_socket);
 
   if (use_uid || use_gid)
     if (chown(path_control_socket, use_uid, use_gid) < 0)
diff --git a/sysdep/unix/unix.h b/sysdep/unix/unix.h
index 346adcf2..518713bc 100644
--- a/sysdep/unix/unix.h
+++ b/sysdep/unix/unix.h
@@ -12,6 +12,8 @@
 #include <sys/socket.h>
 
 struct pool;
+struct iface;
+struct birdsock;
 
 /* main.c */
 
@@ -27,36 +29,81 @@ void cmd_shutdown(void);
 
 #define UNIX_DEFAULT_CONFIGURE_TIMEOUT 300
 
+
 /* io.c */
 
-volatile int async_config_flag;
-volatile int async_dump_flag;
-volatile int async_shutdown_flag;
+#define ERR(c) do { s->err = c; return -1; } while (0)
+#define ERR2(c) do { s->err = c; goto err; } while (0)
+#define ERR_MSG(c) do { errno = 0; s->err = c; return -1; } while (0)
+
+
+#define SOCKADDR_SIZE 32
+
+typedef struct sockaddr_bird {
+  struct sockaddr sa;
+  char padding[SOCKADDR_SIZE - sizeof(struct sockaddr)];
+} sockaddr;
+
 
 #ifdef IPV6
-#define BIRD_PF PF_INET6
 #define BIRD_AF AF_INET6
-typedef struct sockaddr_in6 sockaddr;
-static inline int sa_family_check(sockaddr *sa) { return sa->sin6_family == AF_INET6; }
+#define _MI6(x1,x2,x3,x4) _MI(x1, x2, x3, x4)
+#define ipa_is_link_local(x) ipa_has_link_scope(x)
+#define ipa_from_sa(x) ipa_from_sa6(x)
+#define ipa_from_u32(x) _MI6(0,0,0xffff,x)
+#define ipa_to_u32(x) _I3(x)
 #else
-#define BIRD_PF PF_INET
 #define BIRD_AF AF_INET
-typedef struct sockaddr_in sockaddr;
-static inline int sa_family_check(sockaddr *sa) { return sa->sin_family == AF_INET; }
+#define _I0(X) 0
+#define _I1(X) 0
+#define _I2(X) 0
+#define _I3(X) 0
+#define _MI6(x1,x2,x3,x4) IPA_NONE
+#define ipa_is_link_local(x) 0
+#define ipa_from_sa(x) ipa_from_sa4(x)
 #endif
 
+
+/* This is sloppy hack, it should be detected by configure script */
+/* Linux systems have it defined so this is definition for BSD systems */
+#ifndef s6_addr32
+#define s6_addr32 __u6_addr.__u6_addr32
+#endif
+
+
+static inline ip_addr ipa_from_in4(struct in_addr a)
+{ return ipa_from_u32(ntohl(a.s_addr)); }
+
+static inline ip_addr ipa_from_in6(struct in6_addr a)
+{ return _MI6(ntohl(a.s6_addr32[0]), ntohl(a.s6_addr32[1]), ntohl(a.s6_addr32[2]), ntohl(a.s6_addr32[3])); }
+
+static inline ip_addr ipa_from_sa4(sockaddr *sa)
+{ return ipa_from_in4(((struct sockaddr_in *) sa)->sin_addr); }
+
+static inline ip_addr ipa_from_sa6(sockaddr *sa)
+{ return ipa_from_in6(((struct sockaddr_in6 *) sa)->sin6_addr); }
+
+static inline struct in_addr ipa_to_in4(ip_addr a)
+{ return (struct in_addr) { htonl(ipa_to_u32(a)) }; }
+
+static inline struct in6_addr ipa_to_in6(ip_addr a)
+{ return (struct in6_addr) { .s6_addr32 = { htonl(_I0(a)), htonl(_I1(a)), htonl(_I2(a)), htonl(_I3(a)) } }; }
+
+void sockaddr_fill(sockaddr *sa, int af, ip_addr a, struct iface *ifa, uint port);
+int sockaddr_read(sockaddr *sa, int af, ip_addr *a, struct iface **ifa, uint *port);
+
+
 #ifndef SUN_LEN
 #define SUN_LEN(ptr) ((size_t) (((struct sockaddr_un *) 0)->sun_path) + strlen ((ptr)->sun_path))
 #endif
 
-struct birdsock;
-struct iface;
+volatile int async_config_flag;
+volatile int async_dump_flag;
+volatile int async_shutdown_flag;
 
 void io_init(void);
 void io_loop(void);
-void fill_in_sockaddr(sockaddr *sa, ip_addr a, struct iface *ifa, unsigned port);
-void get_sockaddr(sockaddr *sa, ip_addr *a, struct iface **ifa, unsigned *port, int check);
-void sk_open_unix(struct birdsock *s, char *name);
+int sk_open_unix(struct birdsock *s, char *name);
 void *tracked_fopen(struct pool *, char *name, char *mode);
 void test_old_bird(char *path);