Merge branch 'master' into add-path

Conflicts:

	filter/filter.c
	nest/proto.c
	nest/rt-table.c
	proto/bgp/bgp.h
	proto/bgp/config.Y

8 files changed, 2564 insertions, 0 deletions

diff --git a/proto/bfd/Doc b/proto/bfd/Doc
new file mode 100644
index 00000000..7ee5d3ef
--- /dev/null
+++ b/proto/bfd/Doc
@@ -0,0 +1 @@
+S bfd.c
diff --git a/proto/bfd/Makefile b/proto/bfd/Makefile
new file mode 100644
index 00000000..c28cedec
--- /dev/null
+++ b/proto/bfd/Makefile
@@ -0,0 +1,5 @@
+source=bfd.c packets.c io.c
+root-rel=../../
+dir-name=proto/bfd
+
+include ../../Rules
diff --git a/proto/bfd/bfd.c b/proto/bfd/bfd.c
new file mode 100644
index 00000000..5ebfadc1
--- /dev/null
+++ b/proto/bfd/bfd.c
@@ -0,0 +1,1114 @@
+/*
+ *	BIRD -- Bidirectional Forwarding Detection (BFD)
+ *
+ *	Can be freely distributed and used under the terms of the GNU GPL.
+ */
+
+/**
+ * DOC: Bidirectional Forwarding Detection
+ *
+ * The BFD protocol is implemented in three files: |bfd.c| containing the
+ * protocol logic and the protocol glue with BIRD core, |packets.c| handling BFD
+ * packet processing, RX, TX and protocol sockets. |io.c| then contains generic
+ * code for the event loop, threads and event sources (sockets, microsecond
+ * timers). This generic code will be merged to the main BIRD I/O code in the
+ * future.
+ *
+ * The BFD implementation uses a separate thread with an internal event loop for
+ * handling the protocol logic, which requires high-res and low-latency timing,
+ * so it is not affected by the rest of BIRD, which has several low-granularity
+ * hooks in the main loop, uses second-based timers and cannot offer good
+ * latency. The core of BFD protocol (the code related to BFD sessions,
+ * interfaces and packets) runs in the BFD thread, while the rest (the code
+ * related to BFD requests, BFD neighbors and the protocol glue) runs in the
+ * main thread.
+ *
+ * BFD sessions are represented by structure &bfd_session that contains a state
+ * related to the session and two timers (TX timer for periodic packets and hold
+ * timer for session timeout). These sessions are allocated from @session_slab
+ * and are accessible by two hash tables, @session_hash_id (by session ID) and
+ * @session_hash_ip (by IP addresses of neighbors). Slab and both hashes are in
+ * the main protocol structure &bfd_proto. The protocol logic related to BFD
+ * sessions is implemented in internal functions bfd_session_*(), which are
+ * expected to be called from the context of BFD thread, and external functions
+ * bfd_add_session(), bfd_remove_session() and bfd_reconfigure_session(), which
+ * form an interface to the BFD core for the rest and are expected to be called
+ * from the context of main thread.
+ *
+ * Each BFD session has an associated BFD interface, represented by structure
+ * &bfd_iface. A BFD interface contains a socket used for TX (the one for RX is
+ * shared in &bfd_proto), an interface configuration and reference counter.
+ * Compared to interface structures of other protocols, these structures are not
+ * created and removed based on interface notification events, but according to
+ * the needs of BFD sessions. When a new session is created, it requests a
+ * proper BFD interface by function bfd_get_iface(), which either finds an
+ * existing one in &iface_list (from &bfd_proto) or allocates a new one. When a
+ * session is removed, an associated iface is dicharged by bfd_free_iface().
+ *
+ * BFD requests are the external API for the other protocols. When a protocol
+ * wants a BFD session, it calls bfd_request_session(), which creates a
+ * structure &bfd_request containing approprite information and an notify hook.
+ * This structure is a resource associated with the caller's resource pool. When
+ * a BFD protocol is available, a BFD request is submitted to the protocol, an
+ * appropriate BFD session is found or created and the request is attached to
+ * the session. When a session changes state, all attached requests (and related
+ * protocols) are notified. Note that BFD requests do not depend on BFD protocol
+ * running. When the BFD protocol is stopped or removed (or not available from
+ * beginning), related BFD requests are stored in @bfd_wait_list, where waits
+ * for a new protocol.
+ *
+ * BFD neighbors are just a way to statically configure BFD sessions without
+ * requests from other protocol. Structures &bfd_neighbor are part of BFD
+ * configuration (like static routes in the static protocol). BFD neighbors are
+ * handled by BFD protocol like it is a BFD client -- when a BFD neighbor is
+ * ready, the protocol just creates a BFD request like any other protocol.
+ * 
+ * The protocol uses a new generic event loop (structure &birdloop) from |io.c|,
+ * which supports sockets, timers and events like the main loop. Timers
+ * (structure &timer2) are new microsecond based timers, while sockets and
+ * events are the same. A birdloop is associated with a thread (field @thread)
+ * in which event hooks are executed. Most functions for setting event sources
+ * (like sk_start() or tm2_start()) must be called from the context of that
+ * thread. Birdloop allows to temporarily acquire the context of that thread for
+ * the main thread by calling birdloop_enter() and then birdloop_leave(), which
+ * also ensures mutual exclusion with all event hooks. Note that resources
+ * associated with a birdloop (like timers) should be attached to the
+ * independent resource pool, detached from the main resource tree.
+ *
+ * There are two kinds of interaction between the BFD core (running in the BFD
+ * thread) and the rest of BFD (running in the main thread). The first kind are
+ * configuration calls from main thread to the BFD thread (like bfd_add_session()).
+ * These calls are synchronous and use birdloop_enter() mechanism for mutual
+ * exclusion. The second kind is a notification about session changes from the
+ * BFD thread to the main thread. This is done in an asynchronous way, sesions
+ * with pending notifications are linked (in the BFD thread) to @notify_list in
+ * &bfd_proto, and then bfd_notify_hook() in the main thread is activated using
+ * bfd_notify_kick() and a pipe. The hook then processes scheduled sessions and
+ * calls hooks from associated BFD requests. This @notify_list (and state fields
+ * in structure &bfd_session) is protected by a spinlock in &bfd_proto and
+ * functions bfd_lock_sessions() / bfd_unlock_sessions().
+ *
+ * There are few data races (accessing @p->p.debug from TRACE() from the BFD
+ * thread and accessing some some private fields of %bfd_session from
+ * bfd_show_sessions() from the main thread, but these are harmless (i hope).
+ *
+ * TODO: document functions and access restrictions for fields in BFD structures.
+ *
+ * Supported standards:
+ * - RFC 5880 - main BFD standard
+ * - RFC 5881 - BFD for IP links
+ * - RFC 5882 - generic application of BFD
+ * - RFC 5883 - BFD for multihop paths
+ */
+
+#include "bfd.h"
+
+
+#define HASH_ID_KEY(n)		n->loc_id
+#define HASH_ID_NEXT(n)		n->next_id
+#define HASH_ID_EQ(a,b)		(a == b)
+#define HASH_ID_FN(k)		(k)
+
+#define HASH_IP_KEY(n)		n->addr
+#define HASH_IP_NEXT(n)		n->next_ip
+#define HASH_IP_EQ(a,b)		ipa_equal(a,b)
+#define HASH_IP_FN(k)		ipa_hash(k)
+
+static list bfd_proto_list;
+static list bfd_wait_list;
+
+const char *bfd_state_names[] = { "AdminDown", "Down", "Init", "Up" };
+
+static void bfd_session_set_min_tx(struct bfd_session *s, u32 val);
+static struct bfd_iface *bfd_get_iface(struct bfd_proto *p, ip_addr local, struct iface *iface);
+static void bfd_free_iface(struct bfd_iface *ifa);
+static inline void bfd_notify_kick(struct bfd_proto *p);
+
+
+/*
+ *	BFD sessions
+ */
+
+static void 
+bfd_session_update_state(struct bfd_session *s, uint state, uint diag)
+{
+  struct bfd_proto *p = s->ifa->bfd;
+  uint old_state = s->loc_state; 
+  int notify;
+
+  if (state == old_state)
+    return;
+
+  TRACE(D_EVENTS, "Session to %I changed state from %s to %s",
+	s->addr, bfd_state_names[old_state], bfd_state_names[state]);
+
+  bfd_lock_sessions(p);
+  s->loc_state = state;
+  s->loc_diag = diag;
+
+  notify = !NODE_VALID(&s->n);
+  if (notify)
+    add_tail(&p->notify_list, &s->n);
+  bfd_unlock_sessions(p);
+
+  if (state == BFD_STATE_UP)
+    bfd_session_set_min_tx(s, s->ifa->cf->min_tx_int);
+
+  if (old_state == BFD_STATE_UP)
+    bfd_session_set_min_tx(s, s->ifa->cf->idle_tx_int);
+
+  if (notify)
+    bfd_notify_kick(p);
+}
+
+static void
+bfd_session_update_tx_interval(struct bfd_session *s)
+{
+  u32 tx_int = MAX(s->des_min_tx_int, s->rem_min_rx_int);
+  u32 tx_int_l = tx_int - (tx_int / 4);	 // 75 %
+  u32 tx_int_h = tx_int - (tx_int / 10); // 90 %
+
+  s->tx_timer->recurrent = tx_int_l;
+  s->tx_timer->randomize = tx_int_h - tx_int_l;
+
+  /* Do not set timer if no previous event */
+  if (!s->last_tx)
+    return;
+
+  /* Set timer relative to last tx_timer event */
+  tm2_set(s->tx_timer, s->last_tx + tx_int_l);
+}
+
+static void
+bfd_session_update_detection_time(struct bfd_session *s, int kick)
+{
+  btime timeout = (btime) MAX(s->req_min_rx_int, s->rem_min_tx_int) * s->rem_detect_mult;
+
+  if (kick)
+    s->last_rx = current_time();
+
+  if (!s->last_rx)
+    return;
+
+  tm2_set(s->hold_timer, s->last_rx + timeout);
+}
+
+static void
+bfd_session_control_tx_timer(struct bfd_session *s, int reset)
+{
+  // if (!s->opened) goto stop;
+
+  if (s->passive && (s->rem_id == 0))
+    goto stop;
+
+  if (s->rem_demand_mode && 
+      !s->poll_active && 
+      (s->loc_state == BFD_STATE_UP) &&
+      (s->rem_state == BFD_STATE_UP))
+    goto stop;
+
+  if (s->rem_min_rx_int == 0)
+    goto stop;
+
+  /* So TX timer should run */
+  if (reset || !tm2_active(s->tx_timer))
+  {
+    s->last_tx = 0;
+    tm2_start(s->tx_timer, 0);
+  }
+
+  return;
+
+ stop:
+  tm2_stop(s->tx_timer);
+  s->last_tx = 0;
+}
+
+static void
+bfd_session_request_poll(struct bfd_session *s, u8 request)
+{
+  /* Not sure about this, but doing poll in this case does not make sense */
+  if (s->rem_id == 0)
+    return;
+
+  s->poll_scheduled |= request;
+
+  if (s->poll_active)
+    return;
+
+  s->poll_active = s->poll_scheduled;
+  s->poll_scheduled = 0;
+
+  bfd_session_control_tx_timer(s, 1);
+}
+
+static void
+bfd_session_terminate_poll(struct bfd_session *s)
+{
+  u8 poll_done = s->poll_active & ~s->poll_scheduled;
+
+  if (poll_done & BFD_POLL_TX)
+    s->des_min_tx_int = s->des_min_tx_new;
+
+  if (poll_done & BFD_POLL_RX)
+    s->req_min_rx_int = s->req_min_rx_new;
+
+  s->poll_active = s->poll_scheduled;
+  s->poll_scheduled = 0;
+
+  /* Timers are updated by caller - bfd_session_process_ctl() */
+}
+
+void
+bfd_session_process_ctl(struct bfd_session *s, u8 flags, u32 old_tx_int, u32 old_rx_int)
+{
+  if (s->poll_active && (flags & BFD_FLAG_FINAL))
+    bfd_session_terminate_poll(s);
+
+  if ((s->des_min_tx_int != old_tx_int) || (s->rem_min_rx_int != old_rx_int))
+    bfd_session_update_tx_interval(s);
+
+  bfd_session_update_detection_time(s, 1);
+
+  /* Update session state */
+  int next_state = 0;
+  int diag = BFD_DIAG_NOTHING;
+
+  switch (s->loc_state)
+  {
+  case BFD_STATE_ADMIN_DOWN:
+    return;
+
+  case BFD_STATE_DOWN:
+    if (s->rem_state == BFD_STATE_DOWN)		next_state = BFD_STATE_INIT;
+    else if (s->rem_state == BFD_STATE_INIT)	next_state = BFD_STATE_UP;
+    break;
+
+  case BFD_STATE_INIT:
+    if (s->rem_state == BFD_STATE_ADMIN_DOWN)	next_state = BFD_STATE_DOWN, diag = BFD_DIAG_NEIGHBOR_DOWN;
+    else if (s->rem_state >= BFD_STATE_INIT)	next_state = BFD_STATE_UP;
+    break;
+
+  case BFD_STATE_UP:
+    if (s->rem_state <= BFD_STATE_DOWN)		next_state = BFD_STATE_DOWN, diag = BFD_DIAG_NEIGHBOR_DOWN;
+    break;
+  }
+
+  if (next_state)
+    bfd_session_update_state(s, next_state, diag);
+
+  bfd_session_control_tx_timer(s, 0);
+
+  if (flags & BFD_FLAG_POLL)
+    bfd_send_ctl(s->ifa->bfd, s, 1);
+}
+
+static void 
+bfd_session_timeout(struct bfd_session *s)
+{
+  struct bfd_proto *p = s->ifa->bfd;
+
+  TRACE(D_EVENTS, "Session to %I expired", s->addr);
+
+  s->rem_state = BFD_STATE_DOWN;
+  s->rem_id = 0;
+  s->rem_min_tx_int = 0;
+  s->rem_min_rx_int = 1;
+  s->rem_demand_mode = 0;
+  s->rem_detect_mult = 0;
+
+  s->poll_active = 0;
+  s->poll_scheduled = 0;
+
+  bfd_session_update_state(s, BFD_STATE_DOWN, BFD_DIAG_TIMEOUT);
+
+  bfd_session_control_tx_timer(s, 1);
+}
+
+static void
+bfd_session_set_min_tx(struct bfd_session *s, u32 val)
+{
+  /* Note that des_min_tx_int <= des_min_tx_new */
+
+  if (val == s->des_min_tx_new)
+    return;
+
+  s->des_min_tx_new = val;
+
+  /* Postpone timer update if des_min_tx_int increases and the session is up */
+  if ((s->loc_state != BFD_STATE_UP) || (val < s->des_min_tx_int))
+  {
+    s->des_min_tx_int = val;
+    bfd_session_update_tx_interval(s);
+  }
+
+  bfd_session_request_poll(s, BFD_POLL_TX);
+}
+
+static void
+bfd_session_set_min_rx(struct bfd_session *s, u32 val)
+{
+  /* Note that req_min_rx_int >= req_min_rx_new */
+
+  if (val == s->req_min_rx_new)
+    return;
+
+  s->req_min_rx_new = val; 
+
+  /* Postpone timer update if req_min_rx_int decreases and the session is up */
+  if ((s->loc_state != BFD_STATE_UP) || (val > s->req_min_rx_int))
+  {
+    s->req_min_rx_int = val;
+    bfd_session_update_detection_time(s, 0);
+  }
+
+  bfd_session_request_poll(s, BFD_POLL_RX);
+}
+
+struct bfd_session *
+bfd_find_session_by_id(struct bfd_proto *p, u32 id)
+{
+  return HASH_FIND(p->session_hash_id, HASH_ID, id);
+}
+
+struct bfd_session *
+bfd_find_session_by_addr(struct bfd_proto *p, ip_addr addr)
+{
+  return HASH_FIND(p->session_hash_ip, HASH_IP, addr);
+}
+
+static void
+bfd_tx_timer_hook(timer2 *t)
+{
+  struct bfd_session *s = t->data;
+
+  s->last_tx = current_time();
+  bfd_send_ctl(s->ifa->bfd, s, 0);
+}
+
+static void
+bfd_hold_timer_hook(timer2 *t)
+{
+  bfd_session_timeout(t->data);
+}
+
+static u32
+bfd_get_free_id(struct bfd_proto *p)
+{
+  u32 id;
+  for (id = random_u32(); 1; id++)
+    if (id && !bfd_find_session_by_id(p, id))
+      break;
+
+  return id;
+}
+
+static struct bfd_session *
+bfd_add_session(struct bfd_proto *p, ip_addr addr, ip_addr local, struct iface *iface)
+{
+  birdloop_enter(p->loop);
+
+  struct bfd_iface *ifa = bfd_get_iface(p, local, iface);
+
+  struct bfd_session *s = sl_alloc(p->session_slab);
+  bzero(s, sizeof(struct bfd_session));
+
+  s->addr = addr;
+  s->ifa = ifa;
+  s->loc_id = bfd_get_free_id(p);
+
+  HASH_INSERT(p->session_hash_id, HASH_ID, s);
+  HASH_INSERT(p->session_hash_ip, HASH_IP, s);
+
+
+  /* Initialization of state variables - see RFC 5880 6.8.1 */
+  s->loc_state = BFD_STATE_DOWN;
+  s->rem_state = BFD_STATE_DOWN;
+  s->des_min_tx_int = s->des_min_tx_new = ifa->cf->idle_tx_int;
+  s->req_min_rx_int = s->req_min_rx_new = ifa->cf->min_rx_int;
+  s->rem_min_rx_int = 1;
+  s->detect_mult = ifa->cf->multiplier;
+  s->passive = ifa->cf->passive;
+
+  s->tx_timer = tm2_new_init(p->tpool, bfd_tx_timer_hook, s, 0, 0);
+  s->hold_timer = tm2_new_init(p->tpool, bfd_hold_timer_hook, s, 0, 0);
+  bfd_session_update_tx_interval(s);
+  bfd_session_control_tx_timer(s, 1);
+
+  init_list(&s->request_list);
+  s->last_state_change = now;
+
+  TRACE(D_EVENTS, "Session to %I added", s->addr);
+
+  birdloop_leave(p->loop);
+
+  return s;
+}
+
+/*
+static void
+bfd_open_session(struct bfd_proto *p, struct bfd_session *s, ip_addr local, struct iface *ifa)
+{
+  birdloop_enter(p->loop);
+
+  s->opened = 1;
+
+  bfd_session_control_tx_timer(s);
+
+  birdloop_leave(p->loop);
+}
+
+static void
+bfd_close_session(struct bfd_proto *p, struct bfd_session *s)
+{
+  birdloop_enter(p->loop);
+
+  s->opened = 0;
+
+  bfd_session_update_state(s, BFD_STATE_DOWN, BFD_DIAG_PATH_DOWN);
+  bfd_session_control_tx_timer(s);
+
+  birdloop_leave(p->loop);
+}
+*/
+
+static void
+bfd_remove_session(struct bfd_proto *p, struct bfd_session *s)
+{
+  ip_addr ip = s->addr;
+
+  birdloop_enter(p->loop);
+
+  bfd_free_iface(s->ifa);
+
+  rfree(s->tx_timer);
+  rfree(s->hold_timer);
+
+  HASH_REMOVE(p->session_hash_id, HASH_ID, s);
+  HASH_REMOVE(p->session_hash_ip, HASH_IP, s);
+
+  sl_free(p->session_slab, s);
+
+  TRACE(D_EVENTS, "Session to %I removed", ip);
+
+  birdloop_leave(p->loop);
+}
+
+static void
+bfd_reconfigure_session(struct bfd_proto *p, struct bfd_session *s)
+{
+  birdloop_enter(p->loop);
+
+  struct bfd_iface_config *cf = s->ifa->cf;
+
+  u32 tx = (s->loc_state == BFD_STATE_UP) ? cf->min_tx_int : cf->idle_tx_int;
+  bfd_session_set_min_tx(s, tx);
+  bfd_session_set_min_rx(s, cf->min_rx_int);
+  s->detect_mult = cf->multiplier;
+  s->passive = cf->passive;
+
+  bfd_session_control_tx_timer(s, 0);
+
+  birdloop_leave(p->loop);
+
+  TRACE(D_EVENTS, "Session to %I reconfigured", s->addr);
+}
+
+
+/*
+ *	BFD interfaces
+ */
+
+static struct bfd_iface_config bfd_default_iface = {
+  .min_rx_int = BFD_DEFAULT_MIN_RX_INT,
+  .min_tx_int = BFD_DEFAULT_MIN_TX_INT,
+  .idle_tx_int = BFD_DEFAULT_IDLE_TX_INT,
+  .multiplier = BFD_DEFAULT_MULTIPLIER
+};
+
+static inline struct bfd_iface_config *
+bfd_find_iface_config(struct bfd_config *cf, struct iface *iface)
+{
+  struct bfd_iface_config *ic;
+
+  ic = iface ? (void *) iface_patt_find(&cf->patt_list, iface, NULL) : cf->multihop;
+
+  return ic ? ic : &bfd_default_iface;
+}
+
+static struct bfd_iface *
+bfd_get_iface(struct bfd_proto *p, ip_addr local, struct iface *iface)
+{
+  struct bfd_iface *ifa;
+
+  WALK_LIST(ifa, p->iface_list)
+    if (ipa_equal(ifa->local, local) && (ifa->iface == iface))
+      return ifa->uc++, ifa;
+
+  struct bfd_config *cf = (struct bfd_config *) (p->p.cf);
+  struct bfd_iface_config *ic = bfd_find_iface_config(cf, iface);
+
+  ifa = mb_allocz(p->tpool, sizeof(struct bfd_iface));
+  ifa->local = local;
+  ifa->iface = iface;
+  ifa->cf = ic;
+  ifa->bfd = p;
+
+  ifa->sk = bfd_open_tx_sk(p, local, iface);
+  ifa->uc = 1;
+
+  add_tail(&p->iface_list, &ifa->n);
+
+  return ifa;
+}
+
+static void
+bfd_free_iface(struct bfd_iface *ifa)
+{
+  if (!ifa || --ifa->uc)
+    return;
+
+  rem_node(&ifa->n);
+  sk_stop(ifa->sk);
+  rfree(ifa->sk);
+  mb_free(ifa);
+}
+
+static void
+bfd_reconfigure_iface(struct bfd_proto *p, struct bfd_iface *ifa, struct bfd_config *nc)
+{
+  struct bfd_iface_config *nic = bfd_find_iface_config(nc, ifa->iface);
+  ifa->changed = !!memcmp(nic, ifa->cf, sizeof(struct bfd_iface_config));
+
+  /* This should be probably changed to not access ifa->cf from the BFD thread */
+  birdloop_enter(p->loop);
+  ifa->cf = nic;
+  birdloop_leave(p->loop);
+}
+
+
+/*
+ *	BFD requests
+ */
+
+static void
+bfd_request_notify(struct bfd_request *req, u8 state, u8 diag)
+{
+  u8 old_state = req->state;
+
+  if (state == old_state)
+    return;
+
+  req->state = state;
+  req->diag = diag;
+  req->old_state = old_state;
+  req->down = (old_state == BFD_STATE_UP) && (state == BFD_STATE_DOWN);
+
+  if (req->hook)
+    req->hook(req);
+}
+
+static int
+bfd_add_request(struct bfd_proto *p, struct bfd_request *req)
+{
+  struct bfd_session *s = bfd_find_session_by_addr(p, req->addr);
+  u8 state, diag;
+
+  if (!s)
+    s = bfd_add_session(p, req->addr, req->local, req->iface);
+
+  rem_node(&req->n);
+  add_tail(&s->request_list, &req->n);
+  req->session = s;
+
+  bfd_lock_sessions(p);
+  state = s->loc_state;
+  diag = s->loc_diag;
+  bfd_unlock_sessions(p);
+
+  bfd_request_notify(req, state, diag);
+
+  return 1;
+}
+
+static void
+bfd_submit_request(struct bfd_request *req)
+{
+  node *n;
+
+  WALK_LIST(n, bfd_proto_list)
+    if (bfd_add_request(SKIP_BACK(struct bfd_proto, bfd_node, n), req))
+      return;
+
+  rem_node(&req->n);
+  add_tail(&bfd_wait_list, &req->n);
+  req->session = NULL;
+  bfd_request_notify(req, BFD_STATE_ADMIN_DOWN, 0);
+}
+
+static void
+bfd_take_requests(struct bfd_proto *p)
+{
+  node *n, *nn;
+
+  WALK_LIST_DELSAFE(n, nn, bfd_wait_list)
+    bfd_add_request(p, SKIP_BACK(struct bfd_request, n, n));
+}
+
+static void
+bfd_drop_requests(struct bfd_proto *p)
+{
+  node *n;
+
+  HASH_WALK(p->session_hash_id, next_id, s)
+  {
+    /* We assume that p is not in bfd_proto_list */
+    WALK_LIST_FIRST(n, s->request_list)
+      bfd_submit_request(SKIP_BACK(struct bfd_request, n, n));
+  }
+  HASH_WALK_END;
+}
+
+static struct resclass bfd_request_class;
+
+struct bfd_request *
+bfd_request_session(pool *p, ip_addr addr, ip_addr local, struct iface *iface,
+		    void (*hook)(struct bfd_request *), void *data)
+{
+  struct bfd_request *req = ralloc(p, &bfd_request_class);
+
+  /* Hack: self-link req->n, we will call rem_node() on it */
+  req->n.prev = req->n.next = &req->n;
+
+  req->addr = addr;
+  req->local = local;
+  req->iface = iface;
+
+  bfd_submit_request(req);
+
+  req->hook = hook;
+  req->data = data;
+
+  return req;
+}
+
+static void
+bfd_request_free(resource *r)
+{
+  struct bfd_request *req = (struct bfd_request *) r;
+  struct bfd_session *s = req->session;
+
+  rem_node(&req->n);
+
+  /* Remove the session if there is no request for it. Skip that if
+     inside notify hooks, will be handled by bfd_notify_hook() itself */
+
+  if (s && EMPTY_LIST(s->request_list) && !s->notify_running)
+    bfd_remove_session(s->ifa->bfd, s);
+}
+
+static void
+bfd_request_dump(resource *r)
+{
+  struct bfd_request *req = (struct bfd_request *) r;
+
+  debug("(code %p, data %p)\n", req->hook, req->data);
+}
+
+static struct resclass bfd_request_class = {
+  "BFD request",
+  sizeof(struct bfd_request),
+  bfd_request_free,
+  bfd_request_dump,
+  NULL,
+  NULL
+};
+
+
+/*
+ *	BFD neighbors
+ */
+
+static void
+bfd_neigh_notify(struct neighbor *nb)
+{
+  struct bfd_proto *p = (struct bfd_proto *) nb->proto;
+  struct bfd_neighbor *n = nb->data;
+
+  if (!n)
+    return;
+
+  if ((nb->scope > 0) && !n->req)
+  {
+    ip_addr local = ipa_nonzero(n->local) ? n->local : nb->iface->addr->ip;
+    n->req = bfd_request_session(p->p.pool, n->addr, local, nb->iface, NULL, NULL);
+  }
+
+  if ((nb->scope <= 0) && n->req)
+  {
+    rfree(n->req);
+    n->req = NULL;
+  }
+}
+
+static void
+bfd_start_neighbor(struct bfd_proto *p, struct bfd_neighbor *n)
+{
+  n->active = 1;
+
+  if (n->multihop)
+  {
+    n->req = bfd_request_session(p->p.pool, n->addr, n->local, NULL, NULL, NULL);
+    return;
+  }
+
+  struct neighbor *nb = neigh_find2(&p->p, &n->addr, n->iface, NEF_STICKY);
+  if (!nb)
+  {
+    log(L_ERR "%s: Invalid remote address %I%J", p->p.name, n->addr, n->iface);
+    return;
+  }
+
+  if (nb->data)
+  {
+    log(L_ERR "%s: Duplicate neighbor %I", p->p.name, n->addr);
+    return;
+  }
+
+  n->neigh = nb;
+  nb->data = n;
+
+  if (nb->scope > 0)
+    bfd_neigh_notify(nb);
+  else
+    TRACE(D_EVENTS, "Waiting for %I%J to become my neighbor", n->addr, n->iface);
+}
+
+static void
+bfd_stop_neighbor(struct bfd_proto *p, struct bfd_neighbor *n)
+{
+  if (n->neigh)
+    n->neigh->data = NULL;
+  n->neigh = NULL;
+
+  rfree(n->req);
+  n->req = NULL;
+}
+
+static inline int
+bfd_same_neighbor(struct bfd_neighbor *x, struct bfd_neighbor *y)
+{
+  return ipa_equal(x->addr, y->addr) && ipa_equal(x->local, y->local) &&
+    (x->iface == y->iface) && (x->multihop == y->multihop);
+}
+
+static void
+bfd_reconfigure_neighbors(struct bfd_proto *p, struct bfd_config *new)
+{
+  struct bfd_config *old = (struct bfd_config *) (p->p.cf);
+  struct bfd_neighbor *on, *nn;
+
+  WALK_LIST(on, old->neigh_list)
+  {
+    WALK_LIST(nn, new->neigh_list)
+      if (bfd_same_neighbor(nn, on))
+      {
+	nn->neigh = on->neigh;
+	if (nn->neigh)
+	  nn->neigh->data = nn;
+
+	nn->req = on->req;
+	nn->active = 1;
+	return;
+      }
+
+    bfd_stop_neighbor(p, on);
+  }
+
+  WALK_LIST(nn, new->neigh_list)
+    if (!nn->active)
+      bfd_start_neighbor(p, nn);
+}
+
+
+/*
+ *	BFD notify socket
+ */
+
+/* This core notify code should be replaced after main loop transition to birdloop */
+
+int pipe(int pipefd[2]);
+void pipe_drain(int fd);
+void pipe_kick(int fd);
+
+static int
+bfd_notify_hook(sock *sk, int len)
+{
+  struct bfd_proto *p = sk->data;
+  struct bfd_session *s;
+  list tmp_list;
+  u8 state, diag;
+  node *n, *nn;
+
+  pipe_drain(sk->fd);
+
+  bfd_lock_sessions(p);
+  init_list(&tmp_list);
+  add_tail_list(&tmp_list, &p->notify_list);
+  init_list(&p->notify_list);
+  bfd_unlock_sessions(p);
+
+  WALK_LIST_FIRST(s, tmp_list)
+  {
+    bfd_lock_sessions(p);
+    rem2_node(&s->n);
+    state = s->loc_state;
+    diag = s->loc_diag;
+    bfd_unlock_sessions(p);
+
+    /* FIXME: convert to btime and move to bfd_session_update_state() */ 
+    s->last_state_change = now;
+
+    s->notify_running = 1;
+    WALK_LIST_DELSAFE(n, nn, s->request_list)
+      bfd_request_notify(SKIP_BACK(struct bfd_request, n, n), state, diag);
+    s->notify_running = 0;
+
+    /* Remove the session if all requests were removed in notify hooks */
+    if (EMPTY_LIST(s->request_list))
+      bfd_remove_session(p, s);
+  }
+
+  return 0;
+}
+
+static inline void
+bfd_notify_kick(struct bfd_proto *p)
+{
+  pipe_kick(p->notify_ws->fd);
+}
+
+static void
+bfd_noterr_hook(sock *sk, int err)
+{
+  struct bfd_proto *p = sk->data;
+  log(L_ERR "%s: Notify socket error: %m", p->p.name, err);
+}
+
+static void
+bfd_notify_init(struct bfd_proto *p)
+{
+  int pfds[2];
+  sock *sk;
+
+  int rv = pipe(pfds);
+  if (rv < 0)
+    die("pipe: %m");
+
+  sk = sk_new(p->p.pool);
+  sk->type = SK_MAGIC;
+  sk->rx_hook = bfd_notify_hook;
+  sk->err_hook = bfd_noterr_hook;
+  sk->fd = pfds[0];
+  sk->data = p;
+  if (sk_open(sk) < 0)
+    die("bfd: sk_open failed");
+  p->notify_rs = sk;
+
+  /* The write sock is not added to any event loop */
+  sk = sk_new(p->p.pool);
+  sk->type = SK_MAGIC;
+  sk->fd = pfds[1];
+  sk->data = p;
+  sk->flags = SKF_THREAD;
+  if (sk_open(sk) < 0)
+    die("bfd: sk_open failed");
+  p->notify_ws = sk;
+}
+
+
+/*
+ *	BFD protocol glue
+ */
+
+void
+bfd_init_all(void)
+{
+  init_list(&bfd_proto_list);
+  init_list(&bfd_wait_list);
+}
+
+static struct proto *
+bfd_init(struct proto_config *c)
+{
+  struct proto *p = proto_new(c, sizeof(struct bfd_proto));
+
+  p->neigh_notify = bfd_neigh_notify;
+
+  return p;
+}
+
+static int
+bfd_start(struct proto *P)
+{
+  struct bfd_proto *p = (struct bfd_proto *) P;
+  struct bfd_config *cf = (struct bfd_config *) (P->cf);
+
+  p->loop = birdloop_new();
+  p->tpool = rp_new(NULL, "BFD thread root");
+  pthread_spin_init(&p->lock, PTHREAD_PROCESS_PRIVATE);
+
+  p->session_slab = sl_new(P->pool, sizeof(struct bfd_session));
+  HASH_INIT(p->session_hash_id, P->pool, 8);
+  HASH_INIT(p->session_hash_ip, P->pool, 8);
+
+  init_list(&p->iface_list);
+
+  init_list(&p->notify_list);
+  bfd_notify_init(p);
+
+  add_tail(&bfd_proto_list, &p->bfd_node);
+
+  birdloop_enter(p->loop);
+  p->rx_1 = bfd_open_rx_sk(p, 0);
+  p->rx_m = bfd_open_rx_sk(p, 1);
+  birdloop_leave(p->loop);
+
+  bfd_take_requests(p);
+
+  struct bfd_neighbor *n;
+  WALK_LIST(n, cf->neigh_list)
+    bfd_start_neighbor(p, n);
+
+  birdloop_start(p->loop);
+
+  return PS_UP;
+}
+
+
+static int
+bfd_shutdown(struct proto *P)
+{
+  struct bfd_proto *p = (struct bfd_proto *) P;
+  struct bfd_config *cf = (struct bfd_config *) (P->cf);
+
+  rem_node(&p->bfd_node);
+
+  birdloop_stop(p->loop);
+
+  struct bfd_neighbor *n;
+  WALK_LIST(n, cf->neigh_list)
+    bfd_stop_neighbor(p, n);
+
+  bfd_drop_requests(p);
+
+  /* FIXME: This is hack */
+  birdloop_enter(p->loop);
+  rfree(p->tpool);
+  birdloop_leave(p->loop);
+
+  birdloop_free(p->loop);
+
+  return PS_DOWN;
+}
+
+static int
+bfd_reconfigure(struct proto *P, struct proto_config *c)
+{
+  struct bfd_proto *p = (struct bfd_proto *) P;
+  // struct bfd_config *old = (struct bfd_config *) (P->cf);
+  struct bfd_config *new = (struct bfd_config *) c;
+  struct bfd_iface *ifa;
+
+  birdloop_mask_wakeups(p->loop);
+
+  WALK_LIST(ifa, p->iface_list)
+    bfd_reconfigure_iface(p, ifa, new);
+
+  HASH_WALK(p->session_hash_id, next_id, s)
+  {
+    if (s->ifa->changed)
+      bfd_reconfigure_session(p, s);
+  }
+  HASH_WALK_END;
+
+  bfd_reconfigure_neighbors(p, new);
+
+  birdloop_unmask_wakeups(p->loop);
+
+  return 1;
+}
+
+/* Ensure one instance */
+struct bfd_config *bfd_cf;
+
+static void
+bfd_preconfig(struct protocol *P UNUSED, struct config *c UNUSED)
+{
+  bfd_cf = NULL;
+}
+
+static void
+bfd_copy_config(struct proto_config *dest, struct proto_config *src)
+{
+  struct bfd_config *d = (struct bfd_config *) dest;
+  // struct bfd_config *s = (struct bfd_config *) src;
+
+  /* We clean up patt_list and neigh_list, neighbors and ifaces are non-sharable */
+  init_list(&d->patt_list);  
+  init_list(&d->neigh_list);
+}
+
+void
+bfd_show_sessions(struct proto *P)
+{
+  byte tbuf[TM_DATETIME_BUFFER_SIZE];
+  struct bfd_proto *p = (struct bfd_proto *) P;
+  uint state, diag;
+  u32 tx_int, timeout;
+  const char *ifname;
+
+  if (p->p.proto_state != PS_UP)
+  {
+    cli_msg(-1013, "%s: is not up", p->p.name);
+    cli_msg(0, "");
+    return;
+  }
+
+  cli_msg(-1013, "%s:", p->p.name);
+  cli_msg(-1013, "%-25s %-10s %-10s %-10s  %8s %8s",
+	  "IP address", "Interface", "State", "Since", "Interval", "Timeout");
+
+
+  HASH_WALK(p->session_hash_id, next_id, s)
+  {
+    /* FIXME: this is thread-unsafe, but perhaps harmless */
+    state = s->loc_state;
+    diag = s->loc_diag;
+    ifname = (s->ifa && s->ifa->sk->iface) ? s->ifa->sk->iface->name : "---";
+    tx_int = s->last_tx ? (MAX(s->des_min_tx_int, s->rem_min_rx_int) TO_MS) : 0;
+    timeout = (MAX(s->req_min_rx_int, s->rem_min_tx_int) TO_MS) * s->rem_detect_mult;
+
+    state = (state < 4) ? state : 0;
+    tm_format_datetime(tbuf, &config->tf_proto, s->last_state_change);
+
+    cli_msg(-1013, "%-25I %-10s %-10s %-10s  %3u.%03u  %3u.%03u",
+	    s->addr, ifname, bfd_state_names[state], tbuf,
+	    tx_int / 1000, tx_int % 1000, timeout / 1000, timeout % 1000);
+  }
+  HASH_WALK_END;
+
+  cli_msg(0, "");
+}
+
+
+struct protocol proto_bfd = {
+  .name =		"BFD",
+  .template =		"bfd%d",
+  .init =		bfd_init,
+  .start =		bfd_start,
+  .shutdown =		bfd_shutdown,
+  .reconfigure =	bfd_reconfigure,
+  .preconfig = 		bfd_preconfig,
+  .copy_config =	bfd_copy_config,
+};
diff --git a/proto/bfd/bfd.h b/proto/bfd/bfd.h
new file mode 100644
index 00000000..9b61be64
--- /dev/null
+++ b/proto/bfd/bfd.h
@@ -0,0 +1,191 @@
+/*
+ *	BIRD -- Bidirectional Forwarding Detection (BFD)
+ *
+ *	Can be freely distributed and used under the terms of the GNU GPL.
+ */
+
+#ifndef _BIRD_BFD_H_
+#define _BIRD_BFD_H_
+
+#include <pthread.h>
+
+#include "nest/bird.h"
+#include "nest/cli.h"
+#include "nest/iface.h"
+#include "nest/protocol.h"
+#include "nest/route.h"
+#include "conf/conf.h"
+#include "lib/hash.h"
+#include "lib/resource.h"
+#include "lib/socket.h"
+#include "lib/string.h"
+
+#include "nest/bfd.h"
+#include "io.h"
+
+
+#define BFD_CONTROL_PORT	3784
+#define BFD_ECHO_PORT		3785
+#define BFD_MULTI_CTL_PORT	4784
+
+#define BFD_DEFAULT_MIN_RX_INT	(10 MS_)
+#define BFD_DEFAULT_MIN_TX_INT	(100 MS_)
+#define BFD_DEFAULT_IDLE_TX_INT	(1 S_)
+#define BFD_DEFAULT_MULTIPLIER	5
+
+
+struct bfd_iface_config;
+
+struct bfd_config
+{
+  struct proto_config c;
+  list patt_list;		/* List of iface configs (struct bfd_iface_config) */
+  list neigh_list;		/* List of configured neighbors (struct bfd_neighbor) */
+  struct bfd_iface_config *multihop; /* Multihop pseudoiface config */
+};
+
+struct bfd_iface_config
+{
+  struct iface_patt i;
+  u32 min_rx_int;
+  u32 min_tx_int;
+  u32 idle_tx_int;
+  u8 multiplier;
+  u8 passive;
+};
+
+struct bfd_neighbor
+{
+  node n;
+  ip_addr addr;
+  ip_addr local;
+  struct iface *iface;
+
+  struct neighbor *neigh;
+  struct bfd_request *req;
+
+  u8 multihop;
+  u8 active;
+};
+
+struct bfd_proto
+{
+  struct proto p;
+  struct birdloop *loop;
+  pool *tpool;
+  pthread_spinlock_t lock;
+  node bfd_node;
+
+  slab *session_slab;
+  HASH(struct bfd_session) session_hash_id;
+  HASH(struct bfd_session) session_hash_ip;
+
+  sock *notify_rs;
+  sock *notify_ws;
+  list notify_list;
+
+  sock *rx_1;
+  sock *rx_m;
+  list iface_list;
+};
+
+struct bfd_iface
+{
+  node n;
+  ip_addr local;
+  struct iface *iface;
+  struct bfd_iface_config *cf;
+  struct bfd_proto *bfd;
+
+  sock *sk;
+  u32 uc;
+  u8 changed;
+};
+
+struct bfd_session
+{
+  node n;
+  ip_addr addr;				/* Address of session */
+  struct bfd_iface *ifa;		/* Iface associated with session */
+  struct bfd_session *next_id;		/* Next in bfd.session_hash_id */
+  struct bfd_session *next_ip;		/* Next in bfd.session_hash_ip */
+
+  u8 opened_unused;
+  u8 passive;
+  u8 poll_active;
+  u8 poll_scheduled;
+  
+  u8 loc_state;
+  u8 rem_state;
+  u8 loc_diag;
+  u8 rem_diag;
+  u32 loc_id;				/* Local session ID (local discriminator) */
+  u32 rem_id;				/* Remote session ID (remote discriminator) */
+  u32 des_min_tx_int;			/* Desired min rx interval, local option */
+  u32 des_min_tx_new;			/* Used for des_min_tx_int change */
+  u32 req_min_rx_int;			/* Required min tx interval, local option */
+  u32 req_min_rx_new;			/* Used for req_min_rx_int change */
+  u32 rem_min_tx_int;			/* Last received des_min_tx_int */
+  u32 rem_min_rx_int;			/* Last received req_min_rx_int */
+  u8 demand_mode;			/* Currently unused */
+  u8 rem_demand_mode;
+  u8 detect_mult;			/* Announced detect_mult, local option */
+  u8 rem_detect_mult;			/* Last received detect_mult */
+
+  btime last_tx;			/* Time of last sent periodic control packet */
+  btime last_rx;			/* Time of last received valid control packet */
+
+  timer2 *tx_timer;			/* Periodic control packet timer */
+  timer2 *hold_timer;			/* Timer for session down detection time */
+
+  list request_list;			/* List of client requests (struct bfd_request) */
+  bird_clock_t last_state_change;	/* Time of last state change */
+  u8 notify_running;			/* 1 if notify hooks are running */
+};
+
+
+extern const char *bfd_state_names[];
+
+#define BFD_STATE_ADMIN_DOWN	0
+#define BFD_STATE_DOWN		1
+#define BFD_STATE_INIT		2
+#define BFD_STATE_UP		3
+
+#define BFD_DIAG_NOTHING	0
+#define BFD_DIAG_TIMEOUT	1
+#define BFD_DIAG_ECHO_FAILED	2
+#define BFD_DIAG_NEIGHBOR_DOWN	3
+#define BFD_DIAG_FWD_RESET	4
+#define BFD_DIAG_PATH_DOWN	5
+#define BFD_DIAG_C_PATH_DOWN	6
+#define BFD_DIAG_ADMIN_DOWN	7
+#define BFD_DIAG_RC_PATH_DOWN	8
+
+#define BFD_POLL_TX		1
+#define BFD_POLL_RX		2
+
+#define BFD_FLAGS		0x3f
+#define BFD_FLAG_POLL		(1 << 5)
+#define BFD_FLAG_FINAL		(1 << 4)
+#define BFD_FLAG_CPI		(1 << 3)
+#define BFD_FLAG_AP		(1 << 2)
+#define BFD_FLAG_DEMAND		(1 << 1)
+#define BFD_FLAG_MULTIPOINT	(1 << 0)
+
+
+static inline void bfd_lock_sessions(struct bfd_proto *p) { pthread_spin_lock(&p->lock); }
+static inline void bfd_unlock_sessions(struct bfd_proto *p) { pthread_spin_unlock(&p->lock); }
+
+/* bfd.c */
+struct bfd_session * bfd_find_session_by_id(struct bfd_proto *p, u32 id);
+struct bfd_session * bfd_find_session_by_addr(struct bfd_proto *p, ip_addr addr);
+void bfd_session_process_ctl(struct bfd_session *s, u8 flags, u32 old_tx_int, u32 old_rx_int);
+void bfd_show_sessions(struct proto *P);
+
+/* packets.c */
+void bfd_send_ctl(struct bfd_proto *p, struct bfd_session *s, int final);
+sock * bfd_open_rx_sk(struct bfd_proto *p, int multihop);
+sock * bfd_open_tx_sk(struct bfd_proto *p, ip_addr local, struct iface *ifa);
+
+
+#endif /* _BIRD_BFD_H_ */
diff --git a/proto/bfd/config.Y b/proto/bfd/config.Y
new file mode 100644
index 00000000..1bf8764f
--- /dev/null
+++ b/proto/bfd/config.Y
@@ -0,0 +1,138 @@
+/*
+ *	BIRD -- Router Advertisement Configuration
+ *
+ *
+ *	Can be freely distributed and used under the terms of the GNU GPL.
+ */
+
+CF_HDR
+
+#include "proto/bfd/bfd.h"
+
+CF_DEFINES
+
+#define BFD_CFG ((struct bfd_config *) this_proto)
+#define BFD_IFACE ((struct bfd_iface_config *) this_ipatt)
+#define BFD_NEIGHBOR this_bfd_neighbor
+
+static struct bfd_neighbor *this_bfd_neighbor;
+
+extern struct bfd_config *bfd_cf;
+
+CF_DECLS
+
+CF_KEYWORDS(BFD, MIN, IDLE, RX, TX, INTERVAL, MULTIPLIER, PASSIVE,
+	INTERFACE, MULTIHOP, NEIGHBOR, DEV, LOCAL)
+
+%type <iface> bfd_neigh_iface
+%type <a> bfd_neigh_local
+%type <i> bfd_neigh_multihop
+
+CF_GRAMMAR
+
+CF_ADDTO(proto, bfd_proto)
+
+bfd_proto_start: proto_start BFD
+{
+  this_proto = proto_config_new(&proto_bfd, sizeof(struct bfd_config), $1);
+  init_list(&BFD_CFG->patt_list);
+  init_list(&BFD_CFG->neigh_list);
+
+  if (bfd_cf)
+    cf_error("Only one BFD instance allowed");
+  bfd_cf = BFD_CFG;
+};
+
+bfd_proto_item:
+   proto_item
+ | INTERFACE bfd_iface
+ | MULTIHOP bfd_multihop
+ | NEIGHBOR bfd_neighbor
+ ;
+
+bfd_proto_opts:
+   /* empty */
+ | bfd_proto_opts bfd_proto_item ';'
+ ;
+
+bfd_proto:
+   bfd_proto_start proto_name '{' bfd_proto_opts '}';
+
+
+bfd_iface_start:
+{
+  this_ipatt = cfg_allocz(sizeof(struct bfd_iface_config));
+  init_list(&this_ipatt->ipn_list);
+
+  BFD_IFACE->min_rx_int = BFD_DEFAULT_MIN_RX_INT;
+  BFD_IFACE->min_tx_int = BFD_DEFAULT_MIN_TX_INT;
+  BFD_IFACE->idle_tx_int = BFD_DEFAULT_IDLE_TX_INT;
+  BFD_IFACE->multiplier = BFD_DEFAULT_MULTIPLIER;
+};
+
+bfd_iface_item:
+   INTERVAL expr_us { BFD_IFACE->min_rx_int = BFD_IFACE->min_tx_int = $2; }
+ | MIN RX INTERVAL expr_us { BFD_IFACE->min_rx_int = $4; }
+ | MIN TX INTERVAL expr_us { BFD_IFACE->min_tx_int = $4; }
+ | IDLE TX INTERVAL expr_us { BFD_IFACE->idle_tx_int = $4; }
+ | MULTIPLIER expr { BFD_IFACE->multiplier = $2; }
+ | PASSIVE bool { BFD_IFACE->passive = $2; }
+ ;
+
+bfd_iface_opts:
+   /* empty */
+ | bfd_iface_opts bfd_iface_item ';'
+ ;
+
+bfd_iface_opt_list:
+   /* empty */
+ | '{' bfd_iface_opts '}'
+ ;
+
+bfd_iface: bfd_iface_start iface_patt_list bfd_iface_opt_list
+{ add_tail(&BFD_CFG->patt_list, NODE this_ipatt); };
+
+bfd_multihop: bfd_iface_start bfd_iface_opt_list
+{ BFD_CFG->multihop = BFD_IFACE; };
+
+
+bfd_neigh_iface:
+   /* empty */ { $$ = NULL; }
+ | '%' SYM { $$ = if_get_by_name($2->name); }
+ | DEV TEXT { $$ = if_get_by_name($2); }
+ ;
+
+bfd_neigh_local:
+   /* empty */ { $$ = IPA_NONE; }
+ | LOCAL ipa { $$ = $2; }
+ ;
+
+bfd_neigh_multihop:
+   /* empty */ { $$ = 0; }
+ | MULTIHOP bool { $$ = $2; }
+ ;
+
+bfd_neighbor: ipa bfd_neigh_iface bfd_neigh_local bfd_neigh_multihop
+{
+  this_bfd_neighbor = cfg_allocz(sizeof(struct bfd_neighbor));
+  add_tail(&BFD_CFG->neigh_list, NODE this_bfd_neighbor);
+
+  BFD_NEIGHBOR->addr = $1;
+  BFD_NEIGHBOR->local = $3;
+  BFD_NEIGHBOR->iface = $2;
+  BFD_NEIGHBOR->multihop = $4;
+
+  if ($4 && $2)
+    cf_error("Neighbor cannot set both interface and multihop");
+
+  if ($4 && ipa_zero($3))
+    cf_error("Multihop neighbor requires specified local address");
+};
+
+
+CF_CLI(SHOW BFD SESSIONS, optsym, [<name>], [[Show information about BFD sessions]])
+{ bfd_show_sessions(proto_get_named($4, &proto_bfd)); };
+
+CF_CODE
+
+CF_END
diff --git a/proto/bfd/io.c b/proto/bfd/io.c
new file mode 100644
index 00000000..fb150040
--- /dev/null
+++ b/proto/bfd/io.c
@@ -0,0 +1,768 @@
+/*
+ *	BIRD -- I/O and event loop
+ *
+ *	Can be freely distributed and used under the terms of the GNU GPL.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <poll.h>
+#include <pthread.h>
+#include <time.h>
+#include <sys/time.h>
+
+#include "nest/bird.h"
+#include "proto/bfd/io.h"
+
+#include "lib/buffer.h"
+#include "lib/heap.h"
+#include "lib/lists.h"
+#include "lib/resource.h"
+#include "lib/event.h"
+#include "lib/socket.h"
+
+
+struct birdloop
+{
+  pool *pool;
+  pthread_t thread;
+  pthread_mutex_t mutex;
+
+  btime last_time;
+  btime real_time;
+  u8 use_monotonic_clock;
+
+  u8 stop_called;
+  u8 poll_active;
+  u8 wakeup_masked;
+  int wakeup_fds[2];
+
+  BUFFER(timer2 *) timers;
+  list event_list;
+  list sock_list;
+  uint sock_num;
+
+  BUFFER(sock *) poll_sk;
+  BUFFER(struct pollfd) poll_fd;
+  u8 poll_changed;
+  u8 close_scheduled;
+};
+
+
+/*
+ *	Current thread context
+ */
+
+static pthread_key_t current_loop_key;
+
+static inline struct birdloop *
+birdloop_current(void)
+{
+  return pthread_getspecific(current_loop_key);
+}
+
+static inline void
+birdloop_set_current(struct birdloop *loop)
+{
+  pthread_setspecific(current_loop_key, loop);
+}
+
+static inline void
+birdloop_init_current(void)
+{
+  pthread_key_create(&current_loop_key, NULL);
+}
+
+
+/*
+ *	Time clock
+ */
+
+static void times_update_alt(struct birdloop *loop);
+
+static void
+times_init(struct birdloop *loop)
+{
+  struct timespec ts;
+  int rv;
+
+  rv = clock_gettime(CLOCK_MONOTONIC, &ts);
+  if (rv < 0)
+  {
+    log(L_WARN "Monotonic clock is missing");
+
+    loop->use_monotonic_clock = 0;
+    loop->last_time = 0;
+    loop->real_time = 0;
+    times_update_alt(loop);
+    return;
+  }
+
+  if ((ts.tv_sec < 0) || (((s64) ts.tv_sec) > ((s64) 1 << 40)))
+    log(L_WARN "Monotonic clock is crazy");
+
+  loop->use_monotonic_clock = 1;
+  loop->last_time = ((s64) ts.tv_sec S) + (ts.tv_nsec / 1000);
+  loop->real_time = 0;
+}
+
+static void
+times_update_pri(struct birdloop *loop)
+{
+  struct timespec ts;
+  int rv;
+
+  rv = clock_gettime(CLOCK_MONOTONIC, &ts);
+  if (rv < 0)
+    die("clock_gettime: %m");
+
+  btime new_time = ((s64) ts.tv_sec S) + (ts.tv_nsec / 1000);
+
+  if (new_time < loop->last_time)
+    log(L_ERR "Monotonic clock is broken");
+
+  loop->last_time = new_time;
+  loop->real_time = 0;
+}
+
+static void
+times_update_alt(struct birdloop *loop)
+{
+  struct timeval tv;
+  int rv;
+
+  rv = gettimeofday(&tv, NULL);
+  if (rv < 0)
+    die("gettimeofday: %m");
+
+  btime new_time = ((s64) tv.tv_sec S) + tv.tv_usec;
+  btime delta = new_time - loop->real_time;
+
+  if ((delta < 0) || (delta > (60 S)))
+  {
+    if (loop->real_time)
+      log(L_WARN "Time jump, delta %d us", (int) delta);
+
+    delta = 100 MS;
+  }
+
+  loop->last_time += delta;
+  loop->real_time = new_time;
+}
+
+static void
+times_update(struct birdloop *loop)
+{
+  if (loop->use_monotonic_clock)
+    times_update_pri(loop);
+  else
+    times_update_alt(loop);
+}
+
+btime
+current_time(void)
+{
+  return birdloop_current()->last_time;
+}
+
+
+/*
+ *	Wakeup code for birdloop
+ */
+
+static void
+pipe_new(int *pfds)
+{
+  int rv = pipe(pfds);
+  if (rv < 0)
+    die("pipe: %m");
+
+  if (fcntl(pfds[0], F_SETFL, O_NONBLOCK) < 0)
+    die("fcntl(O_NONBLOCK): %m");
+
+  if (fcntl(pfds[1], F_SETFL, O_NONBLOCK) < 0)
+    die("fcntl(O_NONBLOCK): %m");
+}
+
+void
+pipe_drain(int fd)
+{
+  char buf[64];
+  int rv;
+  
+ try:
+  rv = read(fd, buf, 64);
+  if (rv < 0)
+  {
+    if (errno == EINTR)
+      goto try;
+    if (errno == EAGAIN)
+      return;
+    die("wakeup read: %m");
+  }
+  if (rv == 64)
+    goto try;
+}
+
+void
+pipe_kick(int fd)
+{
+  u64 v = 1;
+  int rv;
+
+ try:
+  rv = write(fd, &v, sizeof(u64));
+  if (rv < 0)
+  {
+    if (errno == EINTR)
+      goto try;
+    if (errno == EAGAIN)
+      return;
+    die("wakeup write: %m");
+  }
+}
+
+static inline void
+wakeup_init(struct birdloop *loop)
+{
+  pipe_new(loop->wakeup_fds);
+}
+
+static inline void
+wakeup_drain(struct birdloop *loop)
+{
+  pipe_drain(loop->wakeup_fds[0]);
+}
+
+static inline void
+wakeup_do_kick(struct birdloop *loop) 
+{
+  pipe_kick(loop->wakeup_fds[1]);
+}
+
+static inline void
+wakeup_kick(struct birdloop *loop)
+{
+  if (!loop->wakeup_masked)
+    wakeup_do_kick(loop);
+  else
+    loop->wakeup_masked = 2;
+}
+
+
+/*
+ *	Events
+ */
+
+static inline uint
+events_waiting(struct birdloop *loop)
+{
+  return !EMPTY_LIST(loop->event_list);
+}
+
+static inline void
+events_init(struct birdloop *loop)
+{
+  init_list(&loop->event_list);
+}
+
+static void
+events_fire(struct birdloop *loop)
+{
+  times_update(loop);
+  ev_run_list(&loop->event_list);
+}
+
+void
+ev2_schedule(event *e)
+{
+  struct birdloop *loop = birdloop_current();
+
+  if (loop->poll_active && EMPTY_LIST(loop->event_list))
+    wakeup_kick(loop);
+
+  if (e->n.next)
+    rem_node(&e->n);
+
+  add_tail(&loop->event_list, &e->n);
+}
+
+
+/*
+ *	Timers
+ */
+
+#define TIMER_LESS(a,b)		((a)->expires < (b)->expires)
+#define TIMER_SWAP(heap,a,b,t)	(t = heap[a], heap[a] = heap[b], heap[b] = t, \
+				   heap[a]->index = (a), heap[b]->index = (b))
+
+static inline uint timers_count(struct birdloop *loop)
+{ return loop->timers.used - 1; }
+
+static inline timer2 *timers_first(struct birdloop *loop)
+{ return (loop->timers.used > 1) ? loop->timers.data[1] : NULL; }
+
+
+static void
+tm2_free(resource *r)
+{
+  timer2 *t = (timer2 *) r;
+
+  tm2_stop(t);
+}
+
+static void
+tm2_dump(resource *r)
+{
+  timer2 *t = (timer2 *) r;
+
+  debug("(code %p, data %p, ", t->hook, t->data);
+  if (t->randomize)
+    debug("rand %d, ", t->randomize);
+  if (t->recurrent)
+    debug("recur %d, ", t->recurrent);
+  if (t->expires)
+    debug("expires in %d ms)\n", (t->expires - current_time()) TO_MS);
+  else
+    debug("inactive)\n");
+}
+
+
+static struct resclass tm2_class = {
+  "Timer",
+  sizeof(timer2),
+  tm2_free,
+  tm2_dump,
+  NULL,
+  NULL
+};
+
+timer2 *
+tm2_new(pool *p)
+{
+  timer2 *t = ralloc(p, &tm2_class);
+  t->index = -1;
+  return t;
+}
+
+void
+tm2_set(timer2 *t, btime when)
+{
+  struct birdloop *loop = birdloop_current();
+  uint tc = timers_count(loop);
+
+  if (!t->expires)
+  {
+    t->index = ++tc;
+    t->expires = when;
+    BUFFER_PUSH(loop->timers) = t;
+    HEAP_INSERT(loop->timers.data, tc, timer2 *, TIMER_LESS, TIMER_SWAP);
+  }
+  else if (t->expires < when)
+  {
+    t->expires = when;
+    HEAP_INCREASE(loop->timers.data, tc, timer2 *, TIMER_LESS, TIMER_SWAP, t->index);
+  }
+  else if (t->expires > when)
+  {
+    t->expires = when;
+    HEAP_DECREASE(loop->timers.data, tc, timer2 *, TIMER_LESS, TIMER_SWAP, t->index);
+  }
+
+  if (loop->poll_active && (t->index == 1))
+    wakeup_kick(loop);
+}
+
+void
+tm2_start(timer2 *t, btime after)
+{
+  tm2_set(t, current_time() + MAX(after, 0));
+}
+
+void
+tm2_stop(timer2 *t)
+{
+  if (!t->expires)
+    return;
+
+  struct birdloop *loop = birdloop_current();
+  uint tc = timers_count(loop);
+
+  HEAP_DELETE(loop->timers.data, tc, timer2 *, TIMER_LESS, TIMER_SWAP, t->index);
+  BUFFER_POP(loop->timers);
+
+  t->index = -1;
+  t->expires = 0;
+}
+
+static void
+timers_init(struct birdloop *loop)
+{
+  BUFFER_INIT(loop->timers, loop->pool, 4);
+  BUFFER_PUSH(loop->timers) = NULL;
+}
+
+static void
+timers_fire(struct birdloop *loop)
+{
+  btime base_time;
+  timer2 *t;
+
+  times_update(loop);
+  base_time = loop->last_time;
+
+  while (t = timers_first(loop))
+  {
+    if (t->expires > base_time)
+      return;
+
+    if (t->recurrent)
+    {
+      btime when = t->expires + t->recurrent;
+      
+      if (when <= loop->last_time)
+	when = loop->last_time + t->recurrent;
+
+      if (t->randomize)
+	when += random() % (t->randomize + 1);
+
+      tm2_set(t, when);
+    }
+    else
+      tm2_stop(t);
+
+    t->hook(t);
+  }
+}
+
+
+/*
+ *	Sockets
+ */
+
+static void
+sockets_init(struct birdloop *loop)
+{
+  init_list(&loop->sock_list);
+  loop->sock_num = 0;
+
+  BUFFER_INIT(loop->poll_sk, loop->pool, 4);
+  BUFFER_INIT(loop->poll_fd, loop->pool, 4);
+  loop->poll_changed = 1;	/* add wakeup fd */
+}
+
+static void
+sockets_add(struct birdloop *loop, sock *s)
+{
+  add_tail(&loop->sock_list, &s->n);
+  loop->sock_num++;
+
+  s->index = -1;
+  loop->poll_changed = 1;
+
+  if (loop->poll_active)
+    wakeup_kick(loop);
+}
+
+void
+sk_start(sock *s)
+{
+  struct birdloop *loop = birdloop_current();
+
+  sockets_add(loop, s);
+}
+
+static void
+sockets_remove(struct birdloop *loop, sock *s)
+{
+  rem_node(&s->n);
+  loop->sock_num--;
+
+  if (s->index >= 0)
+    loop->poll_sk.data[s->index] = NULL;
+
+  s->index = -1;
+  loop->poll_changed = 1;
+
+  /* Wakeup moved to sk_stop() */
+}
+
+void
+sk_stop(sock *s)
+{
+  struct birdloop *loop = birdloop_current();
+
+  sockets_remove(loop, s);
+
+  if (loop->poll_active)
+  {
+    loop->close_scheduled = 1;
+    wakeup_kick(loop);
+  }
+  else
+    close(s->fd);
+
+  s->fd = -1;
+}
+
+static inline uint sk_want_events(sock *s)
+{ return (s->rx_hook ? POLLIN : 0) | ((s->ttx != s->tpos) ? POLLOUT : 0); }
+
+/*
+FIXME: this should be called from sock code
+
+static void
+sockets_update(struct birdloop *loop, sock *s)
+{
+  if (s->index >= 0)
+    loop->poll_fd.data[s->index].events = sk_want_events(s);
+}
+*/
+
+static void
+sockets_prepare(struct birdloop *loop)
+{
+  BUFFER_SET(loop->poll_sk, loop->sock_num + 1);
+  BUFFER_SET(loop->poll_fd, loop->sock_num + 1);
+
+  struct pollfd *pfd = loop->poll_fd.data;
+  sock **psk = loop->poll_sk.data;
+  int i = 0;
+  node *n;
+
+  WALK_LIST(n, loop->sock_list)
+  {
+    sock *s = SKIP_BACK(sock, n, n);
+
+    ASSERT(i < loop->sock_num);
+
+    s->index = i;
+    *psk = s;
+    pfd->fd = s->fd;
+    pfd->events = sk_want_events(s);
+    pfd->revents = 0;
+
+    pfd++;
+    psk++;
+    i++;
+  }
+
+  ASSERT(i == loop->sock_num);
+
+  /* Add internal wakeup fd */
+  *psk = NULL;
+  pfd->fd = loop->wakeup_fds[0];
+  pfd->events = POLLIN;
+  pfd->revents = 0;
+
+  loop->poll_changed = 0;
+}
+
+static void
+sockets_close_fds(struct birdloop *loop)
+{
+  struct pollfd *pfd = loop->poll_fd.data;
+  sock **psk = loop->poll_sk.data;
+  int poll_num = loop->poll_fd.used - 1;
+
+  int i;
+  for (i = 0; i < poll_num; i++)
+    if (psk[i] == NULL)
+      close(pfd[i].fd);
+
+  loop->close_scheduled = 0;
+}
+
+int sk_read(sock *s);
+int sk_write(sock *s);
+
+static void
+sockets_fire(struct birdloop *loop)
+{
+  struct pollfd *pfd = loop->poll_fd.data;
+  sock **psk = loop->poll_sk.data;
+  int poll_num = loop->poll_fd.used - 1;
+
+  times_update(loop);
+
+  /* Last fd is internal wakeup fd */
+  if (pfd[loop->sock_num].revents & POLLIN)
+    wakeup_drain(loop);
+
+  int i;
+  for (i = 0; i < poll_num; pfd++, psk++, i++)
+  {
+    int e = 1;
+
+    if (! pfd->revents)
+      continue;
+
+    if (pfd->revents & POLLNVAL)
+      die("poll: invalid fd %d", pfd->fd);
+
+    if (pfd->revents & POLLIN)
+      while (e && *psk && (*psk)->rx_hook)
+	e = sk_read(*psk);
+
+    e = 1;
+    if (pfd->revents & POLLOUT)
+      while (e && *psk)
+	e = sk_write(*psk);
+  }
+}
+
+
+/*
+ *	Birdloop
+ */
+
+static void * birdloop_main(void *arg);
+
+struct birdloop *
+birdloop_new(void)
+{
+  /* FIXME: this init should be elsewhere and thread-safe */
+  static int init = 0;
+  if (!init)
+    { birdloop_init_current(); init = 1; }
+
+  pool *p = rp_new(NULL, "Birdloop root");
+  struct birdloop *loop = mb_allocz(p, sizeof(struct birdloop));
+  loop->pool = p;
+  pthread_mutex_init(&loop->mutex, NULL);
+
+  times_init(loop);
+  wakeup_init(loop);
+
+  events_init(loop);
+  timers_init(loop);
+  sockets_init(loop);
+
+  return loop;
+}
+
+void
+birdloop_start(struct birdloop *loop)
+{
+  int rv = pthread_create(&loop->thread, NULL, birdloop_main, loop);
+  if (rv)
+    die("pthread_create(): %M", rv);
+}
+
+void
+birdloop_stop(struct birdloop *loop)
+{
+  pthread_mutex_lock(&loop->mutex);
+  loop->stop_called = 1;
+  wakeup_do_kick(loop);
+  pthread_mutex_unlock(&loop->mutex);
+
+  int rv = pthread_join(loop->thread, NULL);
+  if (rv)
+    die("pthread_join(): %M", rv);
+}
+
+void
+birdloop_free(struct birdloop *loop)
+{
+  rfree(loop->pool);
+}
+
+
+void
+birdloop_enter(struct birdloop *loop)
+{
+  /* TODO: these functions could save and restore old context */
+  pthread_mutex_lock(&loop->mutex);
+  birdloop_set_current(loop);
+}
+
+void
+birdloop_leave(struct birdloop *loop)
+{
+  /* TODO: these functions could save and restore old context */
+  birdloop_set_current(NULL);
+  pthread_mutex_unlock(&loop->mutex);
+}
+
+void
+birdloop_mask_wakeups(struct birdloop *loop)
+{
+  pthread_mutex_lock(&loop->mutex);
+  loop->wakeup_masked = 1;
+  pthread_mutex_unlock(&loop->mutex);
+}
+
+void
+birdloop_unmask_wakeups(struct birdloop *loop)
+{
+  pthread_mutex_lock(&loop->mutex);
+  if (loop->wakeup_masked == 2)
+    wakeup_do_kick(loop);
+  loop->wakeup_masked = 0;
+  pthread_mutex_unlock(&loop->mutex);
+}
+
+static void *
+birdloop_main(void *arg)
+{
+  struct birdloop *loop = arg;
+  timer2 *t;
+  int rv, timeout;
+
+  birdloop_set_current(loop);
+
+  pthread_mutex_lock(&loop->mutex);
+  while (1)
+  {
+    events_fire(loop);
+    timers_fire(loop);
+
+    times_update(loop);
+    if (events_waiting(loop))
+      timeout = 0;
+    else if (t = timers_first(loop))
+      timeout = (tm2_remains(t) TO_MS) + 1;
+    else
+      timeout = -1;
+
+    if (loop->poll_changed)
+      sockets_prepare(loop);
+
+    loop->poll_active = 1;
+    pthread_mutex_unlock(&loop->mutex);
+
+  try:
+    rv = poll(loop->poll_fd.data, loop->poll_fd.used, timeout);
+    if (rv < 0)
+    {
+      if (errno == EINTR || errno == EAGAIN)
+	goto try;
+      die("poll: %m");
+    }
+
+    pthread_mutex_lock(&loop->mutex);
+    loop->poll_active = 0;
+
+    if (loop->close_scheduled)
+      sockets_close_fds(loop);
+
+    if (loop->stop_called)
+      break;
+
+    if (rv)
+      sockets_fire(loop);
+
+    timers_fire(loop);
+  }
+
+  loop->stop_called = 0;
+  pthread_mutex_unlock(&loop->mutex);
+
+  return NULL;
+}
+
+
diff --git a/proto/bfd/io.h b/proto/bfd/io.h
new file mode 100644
index 00000000..641ee054
--- /dev/null
+++ b/proto/bfd/io.h
@@ -0,0 +1,99 @@
+/*
+ *	BIRD -- I/O and event loop
+ *
+ *	Can be freely distributed and used under the terms of the GNU GPL.
+ */
+
+#ifndef _BIRD_BFD_IO_H_
+#define _BIRD_BFD_IO_H_
+
+#include "nest/bird.h"
+#include "lib/lists.h"
+#include "lib/resource.h"
+#include "lib/event.h"
+#include "lib/socket.h"
+// #include "lib/timer.h"
+
+
+typedef struct timer2
+{
+  resource r;
+  void (*hook)(struct timer2 *);
+  void *data;
+
+  btime expires;			/* 0=inactive */
+  uint randomize;			/* Amount of randomization */
+  uint recurrent;			/* Timer recurrence */
+
+  int index;
+} timer2;
+
+
+btime current_time(void);
+
+void ev2_schedule(event *e);
+
+
+timer2 *tm2_new(pool *p);
+void tm2_set(timer2 *t, btime when);
+void tm2_start(timer2 *t, btime after);
+void tm2_stop(timer2 *t);
+
+static inline int
+tm2_active(timer2 *t)
+{
+  return t->expires != 0;
+}
+
+static inline btime 
+tm2_remains(timer2 *t)
+{
+  btime now = current_time();
+  return (t->expires > now) ? (t->expires - now) : 0;
+}
+
+static inline timer2 *
+tm2_new_init(pool *p, void (*hook)(struct timer2 *), void *data, uint rec, uint rand)
+{
+  timer2 *t = tm2_new(p);
+  t->hook = hook;
+  t->data = data;
+  t->recurrent = rec;
+  t->randomize = rand;
+  return t;
+}
+
+static inline void
+tm2_set_max(timer2 *t, btime when)
+{
+  if (when > t->expires)
+    tm2_set(t, when);
+}
+
+/*
+static inline void
+tm2_start_max(timer2 *t, btime after)
+{
+  btime rem = tm2_remains(t);
+  tm2_start(t, MAX_(rem, after));
+}
+*/
+
+
+void sk_start(sock *s);
+void sk_stop(sock *s);
+
+
+
+struct birdloop *birdloop_new(void);
+void birdloop_start(struct birdloop *loop);
+void birdloop_stop(struct birdloop *loop);
+void birdloop_free(struct birdloop *loop);
+
+void birdloop_enter(struct birdloop *loop);
+void birdloop_leave(struct birdloop *loop);
+void birdloop_mask_wakeups(struct birdloop *loop);
+void birdloop_unmask_wakeups(struct birdloop *loop);
+
+
+#endif /* _BIRD_BFD_IO_H_ */
diff --git a/proto/bfd/packets.c b/proto/bfd/packets.c
new file mode 100644
index 00000000..fc2616ca
--- /dev/null
+++ b/proto/bfd/packets.c
@@ -0,0 +1,248 @@
+/*
+ *	BIRD -- Bidirectional Forwarding Detection (BFD)
+ *
+ *	Can be freely distributed and used under the terms of the GNU GPL.
+ */
+
+#include "bfd.h"
+
+
+struct bfd_ctl_packet
+{
+  u8 vdiag;			/* version and diagnostic */
+  u8 flags;			/* state and flags */
+  u8 detect_mult;
+  u8 length;
+  u32 snd_id;			/* sender ID, aka 'my discriminator' */
+  u32 rcv_id;			/* receiver ID, aka 'your discriminator' */
+  u32 des_min_tx_int;
+  u32 req_min_rx_int;
+  u32 req_min_echo_rx_int;
+};
+
+#define BFD_BASE_LEN	sizeof(struct bfd_ctl_packet)
+#define BFD_MAX_LEN	64
+
+static inline u8 bfd_pack_vdiag(u8 version, u8 diag)
+{ return (version << 5) | diag; }
+
+static inline u8 bfd_pack_flags(u8 state, u8 flags)
+{ return (state << 6) | flags; }
+
+static inline u8 bfd_pkt_get_version(struct bfd_ctl_packet *pkt)
+{ return pkt->vdiag >> 5; }
+
+static inline u8 bfd_pkt_get_diag(struct bfd_ctl_packet *pkt)
+{ return pkt->vdiag && 0x1f; }
+
+
+static inline u8 bfd_pkt_get_state(struct bfd_ctl_packet *pkt)
+{ return pkt->flags >> 6; }
+
+static inline void bfd_pkt_set_state(struct bfd_ctl_packet *pkt, u8 val)
+{ pkt->flags = val << 6; }
+
+
+char *
+bfd_format_flags(u8 flags, char *buf)
+{
+  char *bp = buf;
+  if (flags & BFD_FLAGS)	*bp++ = ' ';
+  if (flags & BFD_FLAG_POLL)	*bp++ = 'P';
+  if (flags & BFD_FLAG_FINAL)	*bp++ = 'F';
+  if (flags & BFD_FLAG_CPI)	*bp++ = 'C';
+  if (flags & BFD_FLAG_AP)	*bp++ = 'A';
+  if (flags & BFD_FLAG_DEMAND)	*bp++ = 'D';
+  if (flags & BFD_FLAG_MULTIPOINT) *bp++ = 'M';
+  *bp = 0;
+
+  return buf;
+}
+
+void
+bfd_send_ctl(struct bfd_proto *p, struct bfd_session *s, int final)
+{
+  sock *sk = s->ifa->sk;
+  struct bfd_ctl_packet *pkt = (struct bfd_ctl_packet *) sk->tbuf;
+  char fb[8];
+
+  pkt->vdiag = bfd_pack_vdiag(1, s->loc_diag);
+  pkt->flags = bfd_pack_flags(s->loc_state, 0);
+  pkt->detect_mult = s->detect_mult;
+  pkt->length = BFD_BASE_LEN;
+  pkt->snd_id = htonl(s->loc_id);
+  pkt->rcv_id = htonl(s->rem_id);
+  pkt->des_min_tx_int = htonl(s->des_min_tx_new);
+  pkt->req_min_rx_int = htonl(s->req_min_rx_new);
+  pkt->req_min_echo_rx_int = 0;
+
+  if (final)
+    pkt->flags |= BFD_FLAG_FINAL;
+  else if (s->poll_active)
+    pkt->flags |= BFD_FLAG_POLL;
+
+  if (sk->tbuf != sk->tpos)
+    log(L_WARN "%s: Old packet overwritten in TX buffer", p->p.name);
+
+  TRACE(D_PACKETS, "Sending CTL to %I [%s%s]", s->addr,
+	bfd_state_names[s->loc_state], bfd_format_flags(pkt->flags, fb));
+
+  sk_send_to(sk, pkt->length, s->addr, sk->dport);
+}
+
+#define DROP(DSC,VAL) do { err_dsc = DSC; err_val = VAL; goto drop; } while(0)
+
+static int
+bfd_rx_hook(sock *sk, int len)
+{
+  struct bfd_proto *p =  sk->data;
+  struct bfd_ctl_packet *pkt = (struct bfd_ctl_packet *) sk->rbuf;
+  const char *err_dsc = NULL;
+  uint err_val = 0;
+  char fb[8];
+
+  if ((sk->sport == BFD_CONTROL_PORT) && (sk->ttl < 255))
+    DROP("wrong TTL", sk->ttl);
+
+  if (len < BFD_BASE_LEN)
+    DROP("too short", len);
+
+  u8 version = bfd_pkt_get_version(pkt);
+  if (version != 1)
+    DROP("version mismatch", version);
+
+  if ((pkt->length < BFD_BASE_LEN) || (pkt->length > len))
+    DROP("length mismatch", pkt->length);
+
+  if (pkt->detect_mult == 0)
+    DROP("invalid detect mult", 0);
+
+  if ((pkt->flags & BFD_FLAG_MULTIPOINT) ||
+      ((pkt->flags & BFD_FLAG_POLL) && (pkt->flags & BFD_FLAG_FINAL)))
+    DROP("invalid flags", pkt->flags);
+
+  if (pkt->snd_id == 0)
+    DROP("invalid my discriminator", 0);
+
+  struct bfd_session *s;
+  u32 id = ntohl(pkt->rcv_id);
+
+  if (id)
+  {
+    s = bfd_find_session_by_id(p, id);
+
+    if (!s)
+      DROP("unknown session id", id);
+  }
+  else
+  {
+    u8 ps = bfd_pkt_get_state(pkt);
+    if (ps > BFD_STATE_DOWN)
+      DROP("invalid init state", ps);
+      
+    s = bfd_find_session_by_addr(p, sk->faddr);
+
+    /* FIXME: better session matching and message */
+    if (!s)
+      return 1;
+  }
+
+  /* FIXME: better authentication handling and message */
+  if (pkt->flags & BFD_FLAG_AP)
+    DROP("authentication not supported", 0);
+
+
+  u32 old_tx_int = s->des_min_tx_int;
+  u32 old_rx_int = s->rem_min_rx_int;
+
+  s->rem_id= ntohl(pkt->snd_id);
+  s->rem_state = bfd_pkt_get_state(pkt);
+  s->rem_diag = bfd_pkt_get_diag(pkt);
+  s->rem_demand_mode = pkt->flags & BFD_FLAG_DEMAND;
+  s->rem_min_tx_int = ntohl(pkt->des_min_tx_int);
+  s->rem_min_rx_int = ntohl(pkt->req_min_rx_int);
+  s->rem_detect_mult = pkt->detect_mult;
+
+  TRACE(D_PACKETS, "CTL received from %I [%s%s]", sk->faddr,
+	bfd_state_names[s->rem_state], bfd_format_flags(pkt->flags, fb));
+
+  bfd_session_process_ctl(s, pkt->flags, old_tx_int, old_rx_int);
+  return 1;
+
+ drop:
+  log(L_REMOTE "%s: Bad packet from %I - %s (%u)", p->p.name, sk->faddr, err_dsc, err_val);
+  return 1;
+}
+
+static void
+bfd_err_hook(sock *sk, int err)
+{
+  struct bfd_proto *p = sk->data;
+  log(L_ERR "%s: Socket error: %m", p->p.name, err);
+}
+
+sock *
+bfd_open_rx_sk(struct bfd_proto *p, int multihop)
+{
+  sock *sk = sk_new(p->tpool);
+  sk->type = SK_UDP;
+  sk->sport = !multihop ? BFD_CONTROL_PORT : BFD_MULTI_CTL_PORT;
+  sk->data = p;
+
+  sk->rbsize = BFD_MAX_LEN;
+  sk->rx_hook = bfd_rx_hook;
+  sk->err_hook = bfd_err_hook;
+
+  /* TODO: configurable ToS and priority */
+  sk->tos = IP_PREC_INTERNET_CONTROL;
+  sk->priority = sk_priority_control;
+  sk->flags = SKF_THREAD | SKF_LADDR_RX | (!multihop ? SKF_TTL_RX : 0);
+
+#ifdef IPV6
+  sk->flags |= SKF_V6ONLY;
+#endif
+
+  if (sk_open(sk) < 0)
+    goto err;
+
+  sk_start(sk);
+  return sk;
+
+ err:
+  rfree(sk);
+  return NULL;
+}
+
+sock *
+bfd_open_tx_sk(struct bfd_proto *p, ip_addr local, struct iface *ifa)
+{
+  sock *sk = sk_new(p->tpool);
+  sk->type = SK_UDP;
+  sk->saddr = local;
+  sk->dport = ifa ? BFD_CONTROL_PORT : BFD_MULTI_CTL_PORT;
+  sk->iface = ifa;
+  sk->data = p;
+
+  sk->tbsize = BFD_MAX_LEN;
+  sk->err_hook = bfd_err_hook;
+
+  /* TODO: configurable ToS, priority and TTL security */
+  sk->tos = IP_PREC_INTERNET_CONTROL;
+  sk->priority = sk_priority_control;
+  sk->ttl = ifa ? 255 : -1;
+  sk->flags = SKF_THREAD;
+
+#ifdef IPV6
+  sk->flags |= SKF_V6ONLY;
+#endif
+
+  if (sk_open(sk) < 0)
+    goto err;
+
+  sk_start(sk);
+  return sk;
+
+ err:
+  rfree(sk);
+  return NULL;
+}