X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fbfd.c;h=2d53bd27d6faff51c4df3f157ca7666f4766294c;hb=HEAD;hp=a06611a8f9309d8cb4276fb64bbb45c18f143503;hpb=fab52e16f7d8ccdbad53c5037cd04d5863a2cb4a;p=sliver-openvswitch.git

diff --git a/lib/bfd.c b/lib/bfd.c
index a06611a8f..2d53bd27d 100644
--- a/lib/bfd.c
+++ b/lib/bfd.c
@@ -1,4 +1,4 @@
-/* Copyright (c) 2013 Nicira, Inc.
+/* Copyright (c) 2013, 2014 Nicira, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -15,8 +15,13 @@
 #include <config.h>
 #include "bfd.h"
 
+#include <sys/types.h>
 #include <arpa/inet.h>
+#include <netinet/in_systm.h>
+#include <netinet/ip.h>
 
+#include "byte-order.h"
+#include "connectivity.h"
 #include "csum.h"
 #include "dpif.h"
 #include "dynamic-string.h"
@@ -24,15 +29,18 @@
 #include "hash.h"
 #include "hmap.h"
 #include "list.h"
-#include "netlink.h"
+#include "netdev.h"
 #include "odp-util.h"
 #include "ofpbuf.h"
+#include "ovs-thread.h"
 #include "openvswitch/types.h"
 #include "packets.h"
 #include "poll-loop.h"
 #include "random.h"
+#include "seq.h"
 #include "smap.h"
 #include "timeval.h"
+#include "unaligned.h"
 #include "unixctl.h"
 #include "util.h"
 #include "vlog.h"
@@ -54,9 +62,7 @@ VLOG_DEFINE_THIS_MODULE(bfd);
  *
  * - Unit tests.
  *
- * - BFD show into ovs-bugtool.
- *
- * - Set TOS/PCP on inner BFD frame, and outer tunnel header when encapped.
+ * - Set TOS/PCP on the outer tunnel header when encapped.
  *
  * - Sending BFD messages should be in its own thread/process.
  *
@@ -148,6 +154,9 @@ struct bfd {
     bool cpath_down;              /* Concatenated Path Down. */
     uint8_t mult;                 /* bfd.DetectMult. */
 
+    struct netdev *netdev;
+    uint64_t rx_packets;          /* Packets received by 'netdev'. */
+
     enum state state;             /* bfd.SessionState. */
     enum state rmt_state;         /* bfd.RemoteSessionState. */
 
@@ -159,6 +168,12 @@ struct bfd {
 
     uint32_t rmt_disc;            /* bfd.RemoteDiscr. */
 
+    uint8_t eth_dst[ETH_ADDR_LEN];/* Ethernet destination address. */
+    bool eth_dst_set;             /* 'eth_dst' set through database. */
+
+    ovs_be32 ip_src;              /* IPv4 source address. */
+    ovs_be32 ip_dst;              /* IPv4 destination address. */
+
     uint16_t udp_src;             /* UDP source port. */
 
     /* All timers in milliseconds. */
@@ -176,61 +191,138 @@ struct bfd {
     long long int next_tx;        /* Next TX time. */
     long long int detect_time;    /* RFC 5880 6.8.4 Detection time. */
 
-    int ref_cnt;
+    bool last_forwarding;         /* Last calculation of forwarding flag. */
     int forwarding_override;      /* Manual override of 'forwarding' status. */
-    bool check_tnl_key;           /* Verify tunnel key of inbound packets? */
+
+    atomic_bool check_tnl_key;    /* Verify tunnel key of inbound packets? */
+    struct ovs_refcount ref_cnt;
+
+    /* When forward_if_rx is true, bfd_forwarding() will return
+     * true as long as there are incoming packets received.
+     * Note, forwarding_override still has higher priority. */
+    bool forwarding_if_rx;
+    long long int forwarding_if_rx_detect_time;
+
+    /* When 'bfd->forwarding_if_rx' is set, at least one bfd control packet
+     * is required to be received every 100 * bfd->cfg_min_rx.  If bfd
+     * control packet is not received within this interval, even if data
+     * packets are received, the bfd->forwarding will still be false. */
+    long long int demand_rx_bfd_time;
+
+    /* BFD decay related variables. */
+    bool in_decay;                /* True when bfd is in decay. */
+    int decay_min_rx;             /* min_rx is set to decay_min_rx when */
+                                  /* in decay. */
+    int decay_rx_ctl;             /* Count bfd packets received within decay */
+                                  /* detect interval. */
+    uint64_t decay_rx_packets;    /* Packets received by 'netdev'. */
+    long long int decay_detect_time; /* Decay detection time. */
+
+    uint64_t flap_count;          /* Counts bfd forwarding flaps. */
+
+    /* True when the variables returned by bfd_get_status() are changed
+     * since last check. */
+    bool status_changed;
 };
 
-static bool bfd_in_poll(const struct bfd *);
-static void bfd_poll(struct bfd *bfd);
-static const char *bfd_diag_str(enum diag);
-static const char *bfd_state_str(enum state);
-static long long int bfd_min_tx(const struct bfd *);
-static long long int bfd_tx_interval(const struct bfd *);
-static long long int bfd_rx_interval(const struct bfd *);
-static void bfd_set_next_tx(struct bfd *);
-static void bfd_set_state(struct bfd *, enum state, enum diag);
-static uint32_t generate_discriminator(void);
-static void bfd_put_details(struct ds *, const struct bfd *);
+static struct ovs_mutex mutex = OVS_MUTEX_INITIALIZER;
+static struct hmap all_bfds__ = HMAP_INITIALIZER(&all_bfds__);
+static struct hmap *const all_bfds OVS_GUARDED_BY(mutex) = &all_bfds__;
+
+static bool bfd_lookup_ip(const char *host_name, struct in_addr *)
+    OVS_REQUIRES(mutex);
+static bool bfd_forwarding__(struct bfd *) OVS_REQUIRES(mutex);
+static bool bfd_in_poll(const struct bfd *) OVS_REQUIRES(mutex);
+static void bfd_poll(struct bfd *bfd) OVS_REQUIRES(mutex);
+static const char *bfd_diag_str(enum diag) OVS_REQUIRES(mutex);
+static const char *bfd_state_str(enum state) OVS_REQUIRES(mutex);
+static long long int bfd_min_tx(const struct bfd *) OVS_REQUIRES(mutex);
+static long long int bfd_tx_interval(const struct bfd *)
+    OVS_REQUIRES(mutex);
+static long long int bfd_rx_interval(const struct bfd *)
+    OVS_REQUIRES(mutex);
+static void bfd_set_next_tx(struct bfd *) OVS_REQUIRES(mutex);
+static void bfd_set_state(struct bfd *, enum state, enum diag)
+    OVS_REQUIRES(mutex);
+static uint32_t generate_discriminator(void) OVS_REQUIRES(mutex);
+static void bfd_put_details(struct ds *, const struct bfd *)
+    OVS_REQUIRES(mutex);
+static uint64_t bfd_rx_packets(const struct bfd *) OVS_REQUIRES(mutex);
+static void bfd_try_decay(struct bfd *) OVS_REQUIRES(mutex);
+static void bfd_decay_update(struct bfd *) OVS_REQUIRES(mutex);
+static void bfd_status_changed(struct bfd *) OVS_REQUIRES(mutex);
+
+static void bfd_forwarding_if_rx_update(struct bfd *) OVS_REQUIRES(mutex);
 static void bfd_unixctl_show(struct unixctl_conn *, int argc,
                              const char *argv[], void *aux OVS_UNUSED);
 static void bfd_unixctl_set_forwarding_override(struct unixctl_conn *,
                                                 int argc, const char *argv[],
                                                 void *aux OVS_UNUSED);
 static void log_msg(enum vlog_level, const struct msg *, const char *message,
-                    const struct bfd *);
+                    const struct bfd *) OVS_REQUIRES(mutex);
 
 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(20, 20);
-static struct hmap all_bfds = HMAP_INITIALIZER(&all_bfds);
 
 /* Returns true if the interface on which 'bfd' is running may be used to
  * forward traffic according to the BFD session state. */
 bool
-bfd_forwarding(const struct bfd *bfd)
+bfd_forwarding(struct bfd *bfd) OVS_EXCLUDED(mutex)
 {
-    if (bfd->forwarding_override != -1) {
-        return bfd->forwarding_override == 1;
+    bool ret;
+
+    ovs_mutex_lock(&mutex);
+    ret = bfd_forwarding__(bfd);
+    ovs_mutex_unlock(&mutex);
+    return ret;
+}
+
+/* When forwarding_if_rx is enabled, if there are packets received,
+ * updates forwarding_if_rx_detect_time. */
+void
+bfd_account_rx(struct bfd *bfd, const struct dpif_flow_stats *stats)
+{
+    if (stats->n_packets && bfd->forwarding_if_rx) {
+        ovs_mutex_lock(&mutex);
+        bfd_forwarding__(bfd);
+        bfd_forwarding_if_rx_update(bfd);
+        bfd_forwarding__(bfd);
+        ovs_mutex_unlock(&mutex);
     }
+}
+
+/* Returns and resets the 'bfd->status_changed'. */
+bool
+bfd_check_status_change(struct bfd *bfd) OVS_EXCLUDED(mutex)
+{
+    bool ret;
+
+    ovs_mutex_lock(&mutex);
+    ret = bfd->status_changed;
+    bfd->status_changed = false;
+    ovs_mutex_unlock(&mutex);
 
-    return bfd->state == STATE_UP
-        && bfd->rmt_diag != DIAG_PATH_DOWN
-        && bfd->rmt_diag != DIAG_CPATH_DOWN
-        && bfd->rmt_diag != DIAG_RCPATH_DOWN;
+    return ret;
 }
 
 /* Returns a 'smap' of key value pairs representing the status of 'bfd'
  * intended for the OVS database. */
 void
 bfd_get_status(const struct bfd *bfd, struct smap *smap)
+    OVS_EXCLUDED(mutex)
 {
-    smap_add(smap, "forwarding", bfd_forwarding(bfd) ? "true" : "false");
+    ovs_mutex_lock(&mutex);
+    smap_add(smap, "forwarding",
+             bfd_forwarding__(CONST_CAST(struct bfd *, bfd))
+             ? "true" : "false");
     smap_add(smap, "state", bfd_state_str(bfd->state));
     smap_add(smap, "diagnostic", bfd_diag_str(bfd->diag));
+    smap_add_format(smap, "flap_count", "%"PRIu64, bfd->flap_count);
 
     if (bfd->state != STATE_DOWN) {
         smap_add(smap, "remote_state", bfd_state_str(bfd->rmt_state));
         smap_add(smap, "remote_diagnostic", bfd_diag_str(bfd->rmt_diag));
     }
+    ovs_mutex_unlock(&mutex);
 }
 
 /* Initializes, destroys, or reconfigures the BFD session 'bfd' (named 'name'),
@@ -239,22 +331,28 @@ bfd_get_status(const struct bfd *bfd, struct smap *smap)
  * handle for the session, or NULL if BFD is not enabled according to 'cfg'.
  * Also returns NULL if cfg is NULL. */
 struct bfd *
-bfd_configure(struct bfd *bfd, const char *name,
-              const struct smap *cfg)
+bfd_configure(struct bfd *bfd, const char *name, const struct smap *cfg,
+              struct netdev *netdev) OVS_EXCLUDED(mutex)
 {
-    static uint16_t udp_src = 0;
-    static bool init = false;
+    static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
+    static atomic_uint16_t udp_src = ATOMIC_VAR_INIT(0);
 
+    int decay_min_rx;
     long long int min_tx, min_rx;
-    bool cpath_down;
-
-    if (!init) {
+    bool need_poll = false;
+    bool cfg_min_rx_changed = false;
+    bool cpath_down, forwarding_if_rx;
+    const char *hwaddr, *ip_src, *ip_dst;
+    struct in_addr in_addr;
+    uint8_t ea[ETH_ADDR_LEN];
+
+    if (ovsthread_once_start(&once)) {
         unixctl_command_register("bfd/show", "[interface]", 0, 1,
                                  bfd_unixctl_show, NULL);
         unixctl_command_register("bfd/set-forwarding",
                                  "[interface] normal|false|true", 1, 2,
                                  bfd_unixctl_set_forwarding_override, NULL);
-        init = true;
+        ovsthread_once_done(&once);
     }
 
     if (!cfg || !smap_get_bool(cfg, "enable", false)) {
@@ -262,29 +360,38 @@ bfd_configure(struct bfd *bfd, const char *name,
         return NULL;
     }
 
+    ovs_mutex_lock(&mutex);
     if (!bfd) {
         bfd = xzalloc(sizeof *bfd);
         bfd->name = xstrdup(name);
         bfd->forwarding_override = -1;
         bfd->disc = generate_discriminator();
-        hmap_insert(&all_bfds, &bfd->node, bfd->disc);
+        hmap_insert(all_bfds, &bfd->node, bfd->disc);
 
         bfd->diag = DIAG_NONE;
         bfd->min_tx = 1000;
         bfd->mult = 3;
-        bfd->ref_cnt = 1;
+        ovs_refcount_init(&bfd->ref_cnt);
+        bfd->netdev = netdev_ref(netdev);
+        bfd->rx_packets = bfd_rx_packets(bfd);
+        bfd->in_decay = false;
+        bfd->flap_count = 0;
 
         /* RFC 5881 section 4
          * The source port MUST be in the range 49152 through 65535.  The same
          * UDP source port number MUST be used for all BFD Control packets
          * associated with a particular session.  The source port number SHOULD
          * be unique among all BFD sessions on the system. */
-        bfd->udp_src = (udp_src++ % 16384) + 49152;
+        atomic_add(&udp_src, 1, &bfd->udp_src);
+        bfd->udp_src = (bfd->udp_src % 16384) + 49152;
 
         bfd_set_state(bfd, STATE_DOWN, DIAG_NONE);
+
+        memcpy(bfd->eth_dst, eth_addr_bfd, ETH_ADDR_LEN);
     }
 
-    bfd->check_tnl_key = smap_get_bool(cfg, "check_tnl_key", false);
+    atomic_store(&bfd->check_tnl_key,
+                 smap_get_bool(cfg, "check_tnl_key", false));
     min_tx = smap_get_int(cfg, "min_tx", 100);
     min_tx = MAX(min_tx, 100);
     if (bfd->cfg_min_tx != min_tx) {
@@ -293,7 +400,7 @@ bfd_configure(struct bfd *bfd, const char *name,
             || (!bfd_in_poll(bfd) && bfd->cfg_min_tx < bfd->min_tx)) {
             bfd->min_tx = bfd->cfg_min_tx;
         }
-        bfd_poll(bfd);
+        need_poll = true;
     }
 
     min_rx = smap_get_int(cfg, "min_rx", 1000);
@@ -304,17 +411,69 @@ bfd_configure(struct bfd *bfd, const char *name,
             || (!bfd_in_poll(bfd) && bfd->cfg_min_rx > bfd->min_rx)) {
             bfd->min_rx = bfd->cfg_min_rx;
         }
-        bfd_poll(bfd);
+        cfg_min_rx_changed = true;
+        need_poll = true;
+    }
+
+    decay_min_rx = smap_get_int(cfg, "decay_min_rx", 0);
+    if (bfd->decay_min_rx != decay_min_rx || cfg_min_rx_changed) {
+        if (decay_min_rx > 0 && decay_min_rx < bfd->cfg_min_rx) {
+            VLOG_WARN("%s: decay_min_rx cannot be less than %lld ms",
+                      bfd->name, bfd->cfg_min_rx);
+            bfd->decay_min_rx = 0;
+        } else {
+            bfd->decay_min_rx = decay_min_rx;
+        }
+        /* Resets decay. */
+        bfd->in_decay = false;
+        bfd_decay_update(bfd);
+        need_poll = true;
     }
 
     cpath_down = smap_get_bool(cfg, "cpath_down", false);
     if (bfd->cpath_down != cpath_down) {
         bfd->cpath_down = cpath_down;
-        if (bfd->diag == DIAG_NONE || bfd->diag == DIAG_CPATH_DOWN) {
-            bfd_set_state(bfd, bfd->state, DIAG_NONE);
+        bfd_set_state(bfd, bfd->state, DIAG_NONE);
+        need_poll = true;
+    }
+
+    hwaddr = smap_get(cfg, "bfd_dst_mac");
+    if (hwaddr && eth_addr_from_string(hwaddr, ea) && !eth_addr_is_zero(ea)) {
+        memcpy(bfd->eth_dst, ea, ETH_ADDR_LEN);
+        bfd->eth_dst_set = true;
+    } else if (bfd->eth_dst_set) {
+        memcpy(bfd->eth_dst, eth_addr_bfd, ETH_ADDR_LEN);
+        bfd->eth_dst_set = false;
+    }
+
+    ip_src = smap_get(cfg, "bfd_src_ip");
+    if (ip_src && bfd_lookup_ip(ip_src, &in_addr)) {
+        memcpy(&bfd->ip_src, &in_addr, sizeof in_addr);
+    } else {
+        bfd->ip_src = htonl(0xA9FE0100); /* 169.254.1.0. */
+    }
+
+    ip_dst = smap_get(cfg, "bfd_dst_ip");
+    if (ip_dst && bfd_lookup_ip(ip_dst, &in_addr)) {
+        memcpy(&bfd->ip_dst, &in_addr, sizeof in_addr);
+    } else {
+        bfd->ip_dst = htonl(0xA9FE0101); /* 169.254.1.1. */
+    }
+
+    forwarding_if_rx = smap_get_bool(cfg, "forwarding_if_rx", false);
+    if (bfd->forwarding_if_rx != forwarding_if_rx) {
+        bfd->forwarding_if_rx = forwarding_if_rx;
+        if (bfd->state == STATE_UP && bfd->forwarding_if_rx) {
+            bfd_forwarding_if_rx_update(bfd);
+        } else {
+            bfd->forwarding_if_rx_detect_time = 0;
         }
+    }
+
+    if (need_poll) {
         bfd_poll(bfd);
     }
+    ovs_mutex_unlock(&mutex);
     return bfd;
 }
 
@@ -323,59 +482,96 @@ bfd_ref(const struct bfd *bfd_)
 {
     struct bfd *bfd = CONST_CAST(struct bfd *, bfd_);
     if (bfd) {
-        ovs_assert(bfd->ref_cnt > 0);
-        bfd->ref_cnt++;
+        ovs_refcount_ref(&bfd->ref_cnt);
     }
     return bfd;
 }
 
 void
-bfd_unref(struct bfd *bfd)
+bfd_unref(struct bfd *bfd) OVS_EXCLUDED(mutex)
 {
-    if (bfd) {
-        ovs_assert(bfd->ref_cnt > 0);
-        if (!--bfd->ref_cnt) {
-            hmap_remove(&all_bfds, &bfd->node);
-            free(bfd->name);
-            free(bfd);
-        }
+    if (bfd && ovs_refcount_unref(&bfd->ref_cnt) == 1) {
+        ovs_mutex_lock(&mutex);
+        hmap_remove(all_bfds, &bfd->node);
+        netdev_close(bfd->netdev);
+        free(bfd->name);
+        free(bfd);
+        ovs_mutex_unlock(&mutex);
     }
 }
 
 void
-bfd_wait(const struct bfd *bfd)
+bfd_wait(const struct bfd *bfd) OVS_EXCLUDED(mutex)
 {
-    if (bfd->flags & FLAG_FINAL) {
-        poll_immediate_wake();
+    poll_timer_wait_until(bfd_wake_time(bfd));
+}
+
+/* Returns the next wake up time. */
+long long int
+bfd_wake_time(const struct bfd *bfd) OVS_EXCLUDED(mutex)
+{
+    long long int retval;
+
+    if (!bfd) {
+        return LLONG_MAX;
     }
 
-    poll_timer_wait_until(bfd->next_tx);
-    if (bfd->state > STATE_DOWN) {
-        poll_timer_wait_until(bfd->detect_time);
+    ovs_mutex_lock(&mutex);
+    if (bfd->flags & FLAG_FINAL) {
+        retval = 0;
+    } else {
+        retval = bfd->next_tx;
+        if (bfd->state > STATE_DOWN) {
+            retval = MIN(bfd->detect_time, retval);
+        }
     }
+    ovs_mutex_unlock(&mutex);
+    return retval;
 }
 
 void
-bfd_run(struct bfd *bfd)
+bfd_run(struct bfd *bfd) OVS_EXCLUDED(mutex)
 {
-    if (bfd->state > STATE_DOWN && time_msec() >= bfd->detect_time) {
+    long long int now;
+    bool old_in_decay;
+
+    ovs_mutex_lock(&mutex);
+    now = time_msec();
+    old_in_decay = bfd->in_decay;
+
+    if (bfd->state > STATE_DOWN && now >= bfd->detect_time) {
         bfd_set_state(bfd, STATE_DOWN, DIAG_EXPIRED);
     }
+    bfd_forwarding__(bfd);
 
-    if (bfd->min_tx != bfd->cfg_min_tx || bfd->min_rx != bfd->cfg_min_rx) {
+    /* Decay may only happen when state is STATE_UP, bfd->decay_min_rx is
+     * configured, and decay_detect_time is reached. */
+    if (bfd->state == STATE_UP && bfd->decay_min_rx > 0
+        && now >= bfd->decay_detect_time) {
+        bfd_try_decay(bfd);
+    }
+
+    if (bfd->min_tx != bfd->cfg_min_tx
+        || (bfd->min_rx != bfd->cfg_min_rx && bfd->min_rx != bfd->decay_min_rx)
+        || bfd->in_decay != old_in_decay) {
         bfd_poll(bfd);
     }
+    ovs_mutex_unlock(&mutex);
 }
 
 bool
-bfd_should_send_packet(const struct bfd *bfd)
+bfd_should_send_packet(const struct bfd *bfd) OVS_EXCLUDED(mutex)
 {
-    return bfd->flags & FLAG_FINAL || time_msec() >= bfd->next_tx;
+    bool ret;
+    ovs_mutex_lock(&mutex);
+    ret = bfd->flags & FLAG_FINAL || time_msec() >= bfd->next_tx;
+    ovs_mutex_unlock(&mutex);
+    return ret;
 }
 
 void
 bfd_put_packet(struct bfd *bfd, struct ofpbuf *p,
-               uint8_t eth_src[ETH_ADDR_LEN])
+               uint8_t eth_src[ETH_ADDR_LEN]) OVS_EXCLUDED(mutex)
 {
     long long int min_tx, min_rx;
     struct udp_header *udp;
@@ -383,11 +579,12 @@ bfd_put_packet(struct bfd *bfd, struct ofpbuf *p,
     struct ip_header *ip;
     struct msg *msg;
 
+    ovs_mutex_lock(&mutex);
     if (bfd->next_tx) {
         long long int delay = time_msec() - bfd->next_tx;
         long long int interval = bfd_tx_interval(bfd);
         if (delay > interval * 3 / 2) {
-            VLOG_WARN("%s: long delay of %lldms (expected %lldms) sending BFD"
+            VLOG_INFO("%s: long delay of %lldms (expected %lldms) sending BFD"
                       " control message", bfd->name, delay, interval);
         }
     }
@@ -399,17 +596,18 @@ bfd_put_packet(struct bfd *bfd, struct ofpbuf *p,
 
     ofpbuf_reserve(p, 2); /* Properly align after the ethernet header. */
     eth = ofpbuf_put_uninit(p, sizeof *eth);
-    memcpy(eth->eth_dst, eth_addr_broadcast, ETH_ADDR_LEN);
     memcpy(eth->eth_src, eth_src, ETH_ADDR_LEN);
+    memcpy(eth->eth_dst, bfd->eth_dst, ETH_ADDR_LEN);
     eth->eth_type = htons(ETH_TYPE_IP);
 
     ip = ofpbuf_put_zeros(p, sizeof *ip);
     ip->ip_ihl_ver = IP_IHL_VER(5, 4);
     ip->ip_tot_len = htons(sizeof *ip + sizeof *udp + sizeof *msg);
-    ip->ip_ttl = 255;
+    ip->ip_ttl = MAXTTL;
+    ip->ip_tos = IPTOS_LOWDELAY | IPTOS_THROUGHPUT;
     ip->ip_proto = IPPROTO_UDP;
-    ip->ip_src = htonl(0xA9FE0100); /* 169.254.1.0 Link Local. */
-    ip->ip_dst = htonl(0xA9FE0101); /* 169.254.1.1 Link Local. */
+    put_16aligned_be32(&ip->ip_src, bfd->ip_src);
+    put_16aligned_be32(&ip->ip_dst, bfd->ip_dst);
     ip->ip_csum = csum(ip, sizeof *ip);
 
     udp = ofpbuf_put_zeros(p, sizeof *udp);
@@ -444,45 +642,69 @@ bfd_put_packet(struct bfd *bfd, struct ofpbuf *p,
 
     bfd->last_tx = time_msec();
     bfd_set_next_tx(bfd);
+    ovs_mutex_unlock(&mutex);
 }
 
 bool
-bfd_should_process_flow(const struct bfd *bfd, const struct flow *flow,
+bfd_should_process_flow(const struct bfd *bfd_, const struct flow *flow,
                         struct flow_wildcards *wc)
 {
+    struct bfd *bfd = CONST_CAST(struct bfd *, bfd_);
+    bool check_tnl_key;
+
+    memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
+    if (bfd->eth_dst_set && memcmp(bfd->eth_dst, flow->dl_dst, ETH_ADDR_LEN)) {
+        return false;
+    }
+
     memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
     memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
-    if (bfd->check_tnl_key) {
+
+    atomic_read(&bfd->check_tnl_key, &check_tnl_key);
+    if (check_tnl_key) {
         memset(&wc->masks.tunnel.tun_id, 0xff, sizeof wc->masks.tunnel.tun_id);
     }
     return (flow->dl_type == htons(ETH_TYPE_IP)
             && flow->nw_proto == IPPROTO_UDP
-            && flow->tp_dst == htons(3784)
-            && (!bfd->check_tnl_key || flow->tunnel.tun_id == htonl(0)));
+            && flow->tp_dst == htons(BFD_DEST_PORT)
+            && (!check_tnl_key || flow->tunnel.tun_id == htonll(0)));
 }
 
 void
 bfd_process_packet(struct bfd *bfd, const struct flow *flow,
-                   const struct ofpbuf *p)
+                   const struct ofpbuf *p) OVS_EXCLUDED(mutex)
 {
     uint32_t rmt_min_rx, pkt_your_disc;
     enum state rmt_state;
     enum flags flags;
     uint8_t version;
     struct msg *msg;
+    const uint8_t *l7 = ofpbuf_get_udp_payload(p);
+
+    if (!l7) {
+        return; /* No UDP payload. */
+    }
 
     /* This function is designed to follow section RFC 5880 6.8.6 closely. */
 
+    ovs_mutex_lock(&mutex);
+    /* Increments the decay rx counter. */
+    bfd->decay_rx_ctl++;
+
+    bfd_forwarding__(bfd);
+
     if (flow->nw_ttl != 255) {
         /* XXX Should drop in the kernel to prevent DOS. */
-        return;
+        goto out;
     }
 
-    msg = ofpbuf_at(p, (uint8_t *)p->l7 - (uint8_t *)p->data, BFD_PACKET_LEN);
+    msg = ofpbuf_at(p, l7 - (uint8_t *)ofpbuf_data(p), BFD_PACKET_LEN);
     if (!msg) {
-        VLOG_INFO_RL(&rl, "%s: Received unparseable BFD control message.",
-                     bfd->name);
-        return;
+        VLOG_INFO_RL(&rl, "%s: Received too-short BFD control message (only "
+                     "%"PRIdPTR" bytes long, at least %d required).",
+                     bfd->name, (uint8_t *) ofpbuf_tail(p) - l7,
+                     BFD_PACKET_LEN);
+        goto out;
     }
 
     /* RFC 5880 Section 6.8.6
@@ -502,7 +724,7 @@ bfd_process_packet(struct bfd *bfd, const struct flow *flow,
 
     if (version != BFD_VERSION) {
         log_msg(VLL_WARN, msg, "Incorrect version", bfd);
-        return;
+        goto out;
     }
 
     /* Technically this should happen after the length check. We don't support
@@ -510,29 +732,29 @@ bfd_process_packet(struct bfd *bfd, const struct flow *flow,
     if (flags & FLAG_AUTH) {
         log_msg(VLL_WARN, msg, "Authenticated control message with"
                    " authentication disabled", bfd);
-        return;
+        goto out;
     }
 
     if (msg->length != BFD_PACKET_LEN) {
         log_msg(VLL_WARN, msg, "Unexpected length", bfd);
         if (msg->length < BFD_PACKET_LEN) {
-            return;
+            goto out;
         }
     }
 
     if (!msg->mult) {
         log_msg(VLL_WARN, msg, "Zero multiplier", bfd);
-        return;
+        goto out;
     }
 
     if (flags & FLAG_MULTIPOINT) {
         log_msg(VLL_WARN, msg, "Unsupported multipoint flag", bfd);
-        return;
+        goto out;
     }
 
     if (!msg->my_disc) {
         log_msg(VLL_WARN, msg, "NULL my_disc", bfd);
-        return;
+        goto out;
     }
 
     pkt_your_disc = ntohl(msg->your_disc);
@@ -544,11 +766,15 @@ bfd_process_packet(struct bfd *bfd, const struct flow *flow,
          * well, so in this respect, we are not compliant. */
        if (pkt_your_disc != bfd->disc) {
            log_msg(VLL_WARN, msg, "Incorrect your_disc", bfd);
-           return;
+           goto out;
        }
     } else if (rmt_state > STATE_DOWN) {
         log_msg(VLL_WARN, msg, "Null your_disc", bfd);
-        return;
+        goto out;
+    }
+
+    if (bfd->rmt_state != rmt_state) {
+        bfd_status_changed(bfd);
     }
 
     bfd->rmt_disc = ntohl(msg->my_disc);
@@ -576,7 +802,9 @@ bfd_process_packet(struct bfd *bfd, const struct flow *flow,
     rmt_min_rx = MAX(ntohl(msg->min_rx) / 1000, 1);
     if (bfd->rmt_min_rx != rmt_min_rx) {
         bfd->rmt_min_rx = rmt_min_rx;
-        bfd_set_next_tx(bfd);
+        if (bfd->next_tx) {
+            bfd_set_next_tx(bfd);
+        }
         log_msg(VLL_INFO, msg, "New remote min_rx", bfd);
     }
 
@@ -585,7 +813,7 @@ bfd_process_packet(struct bfd *bfd, const struct flow *flow,
 
     if (bfd->state == STATE_ADMIN_DOWN) {
         VLOG_DBG_RL(&rl, "Administratively down, dropping control message.");
-        return;
+        goto out;
     }
 
     if (rmt_state == STATE_ADMIN_DOWN) {
@@ -614,26 +842,97 @@ bfd_process_packet(struct bfd *bfd, const struct flow *flow,
             break;
         case STATE_ADMIN_DOWN:
         default:
-            NOT_REACHED();
+            OVS_NOT_REACHED();
         }
     }
     /* XXX: RFC 5880 Section 6.8.6 Demand mode related calculations here. */
+
+    if (bfd->forwarding_if_rx) {
+        bfd->demand_rx_bfd_time = time_msec() + 100 * bfd->cfg_min_rx;
+    }
+
+out:
+    bfd_forwarding__(bfd);
+    ovs_mutex_unlock(&mutex);
+}
+
+/* Must be called when the netdev owned by 'bfd' should change. */
+void
+bfd_set_netdev(struct bfd *bfd, const struct netdev *netdev)
+    OVS_EXCLUDED(mutex)
+{
+    ovs_mutex_lock(&mutex);
+    if (bfd->netdev != netdev) {
+        netdev_close(bfd->netdev);
+        bfd->netdev = netdev_ref(netdev);
+        if (bfd->decay_min_rx && bfd->state == STATE_UP) {
+            bfd_decay_update(bfd);
+        }
+        if (bfd->forwarding_if_rx && bfd->state == STATE_UP) {
+            bfd_forwarding_if_rx_update(bfd);
+        }
+        bfd->rx_packets = bfd_rx_packets(bfd);
+    }
+    ovs_mutex_unlock(&mutex);
 }
+
 
+/* Updates the forwarding flag.  If override is not configured and
+ * the forwarding flag value changes, increments the flap count.
+ *
+ * Note this function may be called multiple times in a function
+ * (e.g. bfd_account_rx) before and after the bfd state or status
+ * change.  This is to capture any forwarding flag flap. */
+static bool
+bfd_forwarding__(struct bfd *bfd) OVS_REQUIRES(mutex)
+{
+    long long int now = time_msec();
+    bool forwarding_if_rx;
+    bool last_forwarding = bfd->last_forwarding;
+
+    if (bfd->forwarding_override != -1) {
+        return bfd->forwarding_override == 1;
+    }
+
+    forwarding_if_rx = bfd->forwarding_if_rx
+                       && bfd->forwarding_if_rx_detect_time > now
+                       && bfd->demand_rx_bfd_time > now;
+
+    bfd->last_forwarding = (bfd->state == STATE_UP || forwarding_if_rx)
+                           && bfd->rmt_diag != DIAG_PATH_DOWN
+                           && bfd->rmt_diag != DIAG_CPATH_DOWN
+                           && bfd->rmt_diag != DIAG_RCPATH_DOWN;
+    if (bfd->last_forwarding != last_forwarding) {
+        bfd->flap_count++;
+        bfd_status_changed(bfd);
+    }
+    return bfd->last_forwarding;
+}
+
 /* Helpers. */
 static bool
-bfd_in_poll(const struct bfd *bfd)
+bfd_lookup_ip(const char *host_name, struct in_addr *addr)
+{
+    if (!inet_pton(AF_INET, host_name, addr)) {
+        VLOG_ERR_RL(&rl, "\"%s\" is not a valid IP address", host_name);
+        return false;
+    }
+    return true;
+}
+
+static bool
+bfd_in_poll(const struct bfd *bfd) OVS_REQUIRES(mutex)
 {
     return (bfd->flags & FLAG_POLL) != 0;
 }
 
 static void
-bfd_poll(struct bfd *bfd)
+bfd_poll(struct bfd *bfd) OVS_REQUIRES(mutex)
 {
     if (bfd->state > STATE_DOWN && !bfd_in_poll(bfd)
         && !(bfd->flags & FLAG_FINAL)) {
         bfd->poll_min_tx = bfd->cfg_min_tx;
-        bfd->poll_min_rx = bfd->cfg_min_rx;
+        bfd->poll_min_rx = bfd->in_decay ? bfd->decay_min_rx : bfd->cfg_min_rx;
         bfd->flags |= FLAG_POLL;
         bfd->next_tx = 0;
         VLOG_INFO_RL(&rl, "%s: Initiating poll sequence", bfd->name);
@@ -641,7 +940,7 @@ bfd_poll(struct bfd *bfd)
 }
 
 static long long int
-bfd_min_tx(const struct bfd *bfd)
+bfd_min_tx(const struct bfd *bfd) OVS_REQUIRES(mutex)
 {
     /* RFC 5880 Section 6.8.3
      * When bfd.SessionState is not Up, the system MUST set
@@ -653,20 +952,20 @@ bfd_min_tx(const struct bfd *bfd)
 }
 
 static long long int
-bfd_tx_interval(const struct bfd *bfd)
+bfd_tx_interval(const struct bfd *bfd) OVS_REQUIRES(mutex)
 {
     long long int interval = bfd_min_tx(bfd);
     return MAX(interval, bfd->rmt_min_rx);
 }
 
 static long long int
-bfd_rx_interval(const struct bfd *bfd)
+bfd_rx_interval(const struct bfd *bfd) OVS_REQUIRES(mutex)
 {
     return MAX(bfd->min_rx, bfd->rmt_min_tx);
 }
 
 static void
-bfd_set_next_tx(struct bfd *bfd)
+bfd_set_next_tx(struct bfd *bfd) OVS_REQUIRES(mutex)
 {
     long long int interval = bfd_tx_interval(bfd);
     interval -= interval * random_range(26) / 100;
@@ -707,6 +1006,8 @@ bfd_flag_str(enum flags flags)
         ds_put_cstr(&ds, "poll ");
     }
 
+    /* Do not copy the trailing whitespace. */
+    ds_chomp(&ds, ' ');
     ovs_strlcpy(flag_str, ds_cstr(&ds), sizeof flag_str);
     ds_destroy(&ds);
     return flag_str;
@@ -742,7 +1043,7 @@ bfd_diag_str(enum diag diag) {
 
 static void
 log_msg(enum vlog_level level, const struct msg *p, const char *message,
-        const struct bfd *bfd)
+        const struct bfd *bfd) OVS_REQUIRES(mutex)
 {
     struct ds ds = DS_EMPTY_INITIALIZER;
 
@@ -774,8 +1075,9 @@ log_msg(enum vlog_level level, const struct msg *p, const char *message,
 
 static void
 bfd_set_state(struct bfd *bfd, enum state state, enum diag diag)
+    OVS_REQUIRES(mutex)
 {
-    if (diag == DIAG_NONE && bfd->cpath_down) {
+    if (bfd->cpath_down) {
         diag = DIAG_CPATH_DOWN;
     }
 
@@ -803,10 +1105,80 @@ bfd_set_state(struct bfd *bfd, enum state state, enum diag diag)
             bfd->rmt_flags = 0;
             bfd->rmt_disc = 0;
             bfd->rmt_min_tx = 0;
+            /* Resets the min_rx if in_decay. */
+            if (bfd->in_decay) {
+                bfd->min_rx = bfd->cfg_min_rx;
+                bfd->in_decay = false;
+            }
+        }
+        /* Resets the decay when state changes to STATE_UP
+         * and decay_min_rx is configured. */
+        if (bfd->state == STATE_UP && bfd->decay_min_rx) {
+            bfd_decay_update(bfd);
         }
+
+        bfd_status_changed(bfd);
+    }
+}
+
+static uint64_t
+bfd_rx_packets(const struct bfd *bfd) OVS_REQUIRES(mutex)
+{
+    struct netdev_stats stats;
+
+    if (!netdev_get_stats(bfd->netdev, &stats)) {
+        return stats.rx_packets;
+    } else {
+        return 0;
     }
 }
 
+/* Decays the bfd->min_rx to bfd->decay_min_rx when 'diff' is less than
+ * the 'expect' value. */
+static void
+bfd_try_decay(struct bfd *bfd) OVS_REQUIRES(mutex)
+{
+    int64_t diff, expect;
+
+    /* The 'diff' is the difference between current interface rx_packets
+     * stats and last-time check.  The 'expect' is the recorded number of
+     * bfd control packets received within an approximately decay_min_rx
+     * (2000 ms if decay_min_rx is less than 2000 ms) interval.
+     *
+     * Since the update of rx_packets stats at interface happens
+     * asynchronously to the bfd_rx_packets() function, the 'diff' value
+     * can be jittered.  Thusly, we double the decay_rx_ctl to provide
+     * more wiggle room. */
+    diff = bfd_rx_packets(bfd) - bfd->decay_rx_packets;
+    expect = 2 * MAX(bfd->decay_rx_ctl, 1);
+    bfd->in_decay = diff <= expect ? true : false;
+    bfd_decay_update(bfd);
+}
+
+/* Updates the rx_packets, decay_rx_ctl and decay_detect_time. */
+static void
+bfd_decay_update(struct bfd * bfd) OVS_REQUIRES(mutex)
+{
+    bfd->decay_rx_packets = bfd_rx_packets(bfd);
+    bfd->decay_rx_ctl = 0;
+    bfd->decay_detect_time = MAX(bfd->decay_min_rx, 2000) + time_msec();
+}
+
+/* Records the status change and changes the global connectivity seq. */
+static void
+bfd_status_changed(struct bfd *bfd) OVS_REQUIRES(mutex)
+{
+    seq_change(connectivity_seq_get());
+    bfd->status_changed = true;
+}
+
+static void
+bfd_forwarding_if_rx_update(struct bfd *bfd) OVS_REQUIRES(mutex)
+{
+    int64_t incr = bfd_rx_interval(bfd) * bfd->mult;
+    bfd->forwarding_if_rx_detect_time = MAX(incr, 2000) + time_msec();
+}
+
 static uint32_t
 generate_discriminator(void)
 {
@@ -823,7 +1195,7 @@ generate_discriminator(void)
         /* 'disc' is by definition random, so there's no reason to waste time
          * hashing it. */
         disc = random_uint32();
-        HMAP_FOR_EACH_IN_BUCKET (bfd, node, disc, &all_bfds) {
+        HMAP_FOR_EACH_IN_BUCKET (bfd, node, disc, all_bfds) {
             if (bfd->disc == disc) {
                 disc = 0;
                 break;
@@ -835,11 +1207,11 @@ generate_discriminator(void)
 }
 
 static struct bfd *
-bfd_find_by_name(const char *name)
+bfd_find_by_name(const char *name) OVS_REQUIRES(mutex)
 {
     struct bfd *bfd;
 
-    HMAP_FOR_EACH (bfd, node, &all_bfds) {
+    HMAP_FOR_EACH (bfd, node, all_bfds) {
         if (!strcmp(bfd->name, name)) {
             return bfd;
         }
@@ -848,10 +1220,11 @@ bfd_find_by_name(const char *name)
 }
 
 static void
-bfd_put_details(struct ds *ds, const struct bfd *bfd)
+bfd_put_details(struct ds *ds, const struct bfd *bfd) OVS_REQUIRES(mutex)
 {
     ds_put_format(ds, "\tForwarding: %s\n",
-                  bfd_forwarding(bfd) ? "true" : "false");
+                  bfd_forwarding__(CONST_CAST(struct bfd *, bfd))
+                  ? "true" : "false");
     ds_put_format(ds, "\tDetect Multiplier: %d\n", bfd->mult);
     ds_put_format(ds, "\tConcatenated Path Down: %s\n",
                   bfd->cpath_down ? "true" : "false");
@@ -891,37 +1264,43 @@ bfd_put_details(struct ds *ds, const struct bfd *bfd)
 
 static void
 bfd_unixctl_show(struct unixctl_conn *conn, int argc, const char *argv[],
-                 void *aux OVS_UNUSED)
+                 void *aux OVS_UNUSED) OVS_EXCLUDED(mutex)
 {
     struct ds ds = DS_EMPTY_INITIALIZER;
     struct bfd *bfd;
 
+    ovs_mutex_lock(&mutex);
     if (argc > 1) {
         bfd = bfd_find_by_name(argv[1]);
         if (!bfd) {
             unixctl_command_reply_error(conn, "no such bfd object");
-            return;
+            goto out;
         }
         bfd_put_details(&ds, bfd);
     } else {
-        HMAP_FOR_EACH (bfd, node, &all_bfds) {
+        HMAP_FOR_EACH (bfd, node, all_bfds) {
             ds_put_format(&ds, "---- %s ----\n", bfd->name);
             bfd_put_details(&ds, bfd);
         }
     }
     unixctl_command_reply(conn, ds_cstr(&ds));
     ds_destroy(&ds);
+
+out:
+    ovs_mutex_unlock(&mutex);
 }
 
 
 static void
 bfd_unixctl_set_forwarding_override(struct unixctl_conn *conn, int argc,
                                     const char *argv[], void *aux OVS_UNUSED)
+    OVS_EXCLUDED(mutex)
 {
     const char *forward_str = argv[argc - 1];
     int forwarding_override;
     struct bfd *bfd;
 
+    ovs_mutex_lock(&mutex);
     if (!strcasecmp("true", forward_str)) {
         forwarding_override = 1;
     } else if (!strcasecmp("false", forward_str)) {
@@ -930,21 +1309,26 @@ bfd_unixctl_set_forwarding_override(struct unixctl_conn *conn, int argc,
         forwarding_override = -1;
     } else {
         unixctl_command_reply_error(conn, "unknown fault string");
-        return;
+        goto out;
     }
 
     if (argc > 2) {
         bfd = bfd_find_by_name(argv[1]);
         if (!bfd) {
             unixctl_command_reply_error(conn, "no such BFD object");
-            return;
+            goto out;
         }
         bfd->forwarding_override = forwarding_override;
+        bfd_status_changed(bfd);
     } else {
-        HMAP_FOR_EACH (bfd, node, &all_bfds) {
+        HMAP_FOR_EACH (bfd, node, all_bfds) {
             bfd->forwarding_override = forwarding_override;
+            bfd_status_changed(bfd);
         }
     }
 
     unixctl_command_reply(conn, "OK");
+
+out:
+    ovs_mutex_unlock(&mutex);
 }