diff --git a/README.md b/README.md
index 2146d2b415d3..2bbe50221082 100644
--- a/README.md
+++ b/README.md
@@ -133,6 +133,7 @@ pair of .c and .py files, and some are directories of files.
 - tools/[tcplife](tools/tcplife.py): Trace TCP sessions and summarize lifespan. [Examples](tools/tcplife_example.txt).
 - tools/[tcpretrans](tools/tcpretrans.py): Trace TCP retransmits and TLPs. [Examples](tools/tcpretrans_example.txt).
 - tools/[tcptop](tools/tcptop.py): Summarize TCP send/recv throughput by host. Top for TCP. [Examples](tools/tcptop_example.txt).
+- tools/[tcptracer](tools/tcptracer.py): Trace TCP established connections (connect(), accept(), close()). [Examples](tools/tcptracer_example.txt).
 - tools/[tplist](tools/tplist.py): Display kernel tracepoints or USDT probes and their formats. [Examples](tools/tplist_example.txt).
 - tools/[trace](tools/trace.py): Trace arbitrary functions, with filters. [Examples](tools/trace_example.txt).
 - tools/[ttysnoop](tools/ttysnoop.py): Watch live output from a tty or pts device. [Examples](tools/ttysnoop_example.txt).
diff --git a/man/man8/tcptracer.8 b/man/man8/tcptracer.8
new file mode 100644
index 000000000000..b5b3061712ad
--- /dev/null
+++ b/man/man8/tcptracer.8
@@ -0,0 +1,98 @@
+.TH tcptracer 8  "2017-03-27" "USER COMMANDS"
+.SH NAME
+tcptracer \- Trace TCP established connections. Uses Linux eBPF/bcc.
+.SH SYNOPSIS
+.B tcptracer [\-h] [\-v] [\-p PID] [\-N NETNS]
+.SH DESCRIPTION
+This tool traces established TCP connections that open and close while tracing,
+and prints a line of output per connect, accept and close events. This includes
+the type of event, PID, IP addresses and ports.
+
+This tool works by using kernel dynamic tracing, and will need to be updated if
+the kernel implementation changes. Only established TCP connections are traced,
+so it is expected that the overhead of this tool is rather low.
+
+Since this uses BPF, only the root user can use this tool.
+.SH REQUIREMENTS
+CONFIG_BPF and bcc.
+.SH OPTIONS
+.TP
+\-h
+Print usage message.
+.TP
+\-v
+Print full lines, with long event type names and network namespace numbers.
+.TP
+\-p PID
+Trace this process ID only (filtered in-kernel).
+.TP
+\-N NETNS
+Trace this network namespace only (filtered in-kernel).
+.TP
+.SH EXAMPLES
+.TP
+Trace all TCP established connections:
+#
+.B tcptracer
+.TP
+Trace all TCP established connections with verbose lines:
+#
+.B tcptracer \-v
+.TP
+Trace PID 181 only:
+#
+.B tcptracer \-p 181
+.TP
+Trace connections in network namespace 4026531969 only:
+#
+.B tcptracer \-N 4026531969
+.SH FIELDS
+.TP
+TYPE
+Type of event. In non-verbose mode: C for connect, A for accept, X for close.
+.TP
+PID
+Process ID
+.TP
+COMM
+Process name
+.TP
+IP
+IP address family (4 or 6)
+.TP
+SADDR
+Source IP address.
+.TP
+DADDR
+Destination IP address.
+.TP
+SPORT
+Source port.
+.TP
+DPORT
+Destination port.
+.TP
+NETNS
+Network namespace where the event originated.
+.SH OVERHEAD
+This traces the kernel inet accept function, and the TCP connect, close,
+and set state functions. However, it only prints information for connections
+that are established, so it shouldn't have a huge overhead.
+
+As always, test and understand this tools overhead for your types of workloads
+before production use.
+.SH SOURCE
+This is from bcc.
+.IP
+https://github.com/iovisor/bcc
+.PP
+Also look in the bcc distribution for a companion _examples.txt file containing
+example usage, output, and commentary for this tool.
+.SH OS
+Linux
+.SH STABILITY
+Unstable - in development.
+.SH AUTHOR
+Iago López Galeiras
+.SH SEE ALSO
+tcpaccept(8), tcpconnect(8), tcptop(8), tcplife(8)
diff --git a/tools/tcptracer.py b/tools/tcptracer.py
new file mode 100755
index 000000000000..de8c0f52c907
--- /dev/null
+++ b/tools/tcptracer.py
@@ -0,0 +1,654 @@
+#!/usr/bin/python
+#
+# tcpv4tracer   Trace TCP connections.
+#               For Linux, uses BCC, eBPF. Embedded C.
+#
+# USAGE: tcpv4tracer [-h] [-v] [-p PID] [-N NETNS]
+#
+# You should generally try to avoid writing long scripts that measure multiple
+# functions and walk multiple kernel structures, as they will be a burden to
+# maintain as the kernel changes.
+# The following code should be replaced, and simplified, when static TCP probes
+# exist.
+#
+# Copyright 2017 Kinvolk GmbH
+#
+# Licensed under the Apache License, Version 2.0 (the "License")
+from __future__ import print_function
+from bcc import BPF
+
+import argparse as ap
+import ctypes
+from socket import inet_ntop, AF_INET, AF_INET6
+from struct import pack
+
+parser = ap.ArgumentParser(description="Trace TCP connections",
+                           formatter_class=ap.RawDescriptionHelpFormatter)
+parser.add_argument("-p", "--pid", default=0, type=int,
+                    help="trace this PID only")
+parser.add_argument("-N", "--netns", default=0, type=int,
+                    help="trace this Network Namespace only")
+parser.add_argument("-v", "--verbose", action="store_true",
+                    help="include Network Namespace in the output")
+args = parser.parse_args()
+
+bpf_text = """
+#include <uapi/linux/ptrace.h>
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wtautological-compare"
+#include <net/sock.h>
+#pragma clang diagnostic pop
+#include <net/inet_sock.h>
+#include <net/net_namespace.h>
+#include <bcc/proto.h>
+
+#define TCP_EVENT_TYPE_CONNECT 1
+#define TCP_EVENT_TYPE_ACCEPT  2
+#define TCP_EVENT_TYPE_CLOSE   3
+
+struct tcp_ipv4_event_t {
+    u32 type;
+    u32 pid;
+    char comm[TASK_COMM_LEN];
+    u8 ip;
+    u32 saddr;
+    u32 daddr;
+    u16 sport;
+    u16 dport;
+    u32 netns;
+};
+BPF_PERF_OUTPUT(tcp_ipv4_event);
+
+struct tcp_ipv6_event_t {
+    u32 type;
+    u32 pid;
+    char comm[TASK_COMM_LEN];
+    u8 ip;
+    unsigned __int128 saddr;
+    unsigned __int128 daddr;
+    u16 sport;
+    u16 dport;
+    u32 netns;
+};
+BPF_PERF_OUTPUT(tcp_ipv6_event);
+
+// tcp_set_state doesn't run in the context of the process that initiated the
+// connection so we need to store a map TUPLE -> PID to send the right PID on
+// the event
+struct ipv4_tuple_t {
+    u32 saddr;
+    u32 daddr;
+    u16 sport;
+    u16 dport;
+    u32 netns;
+};
+
+struct ipv6_tuple_t {
+    unsigned __int128 saddr;
+    unsigned __int128 daddr;
+    u16 sport;
+    u16 dport;
+    u32 netns;
+};
+
+struct pid_comm_t {
+    u64 pid;
+    char comm[TASK_COMM_LEN];
+};
+
+BPF_HASH(tuplepid_ipv4, struct ipv4_tuple_t, struct pid_comm_t);
+BPF_HASH(tuplepid_ipv6, struct ipv6_tuple_t, struct pid_comm_t);
+
+BPF_HASH(connectsock, u64, struct sock *);
+
+static int read_ipv4_tuple(struct ipv4_tuple_t *tuple, struct sock *skp)
+{
+  u32 saddr = 0, daddr = 0, net_ns_inum = 0;
+  u16 sport = 0, dport = 0;
+  possible_net_t skc_net;
+
+  bpf_probe_read(&saddr, sizeof(saddr), &skp->__sk_common.skc_rcv_saddr);
+  bpf_probe_read(&daddr, sizeof(daddr), &skp->__sk_common.skc_daddr);
+  bpf_probe_read(&sport, sizeof(sport),
+                 &((struct inet_sock *)skp)->inet_sport);
+  bpf_probe_read(&dport, sizeof(dport), &skp->__sk_common.skc_dport);
+#ifdef CONFIG_NET_NS
+  bpf_probe_read(&skc_net, sizeof(skc_net), &skp->__sk_common.skc_net);
+  bpf_probe_read(&net_ns_inum, sizeof(net_ns_inum), &skc_net.net->ns.inum);
+#else
+  net_ns_inum = 0;
+#endif
+
+  ##FILTER_NETNS##
+
+  tuple->saddr = saddr;
+  tuple->daddr = daddr;
+  tuple->sport = sport;
+  tuple->dport = dport;
+  tuple->netns = net_ns_inum;
+
+  // if addresses or ports are 0, ignore
+  if (saddr == 0 || daddr == 0 || sport == 0 || dport == 0) {
+      return 0;
+  }
+
+  return 1;
+}
+
+static int read_ipv6_tuple(struct ipv6_tuple_t *tuple, struct sock *skp)
+{
+  u32 net_ns_inum = 0;
+  u16 sport = 0, dport = 0;
+  unsigned __int128 saddr = 0, daddr = 0;
+  possible_net_t skc_net;
+
+  bpf_probe_read(&sport, sizeof(sport),
+                 &((struct inet_sock *)skp)->inet_sport);
+  bpf_probe_read(&dport, sizeof(dport), &skp->__sk_common.skc_dport);
+  bpf_probe_read(&saddr, sizeof(saddr),
+                 &skp->__sk_common.skc_v6_rcv_saddr.in6_u.u6_addr32);
+  bpf_probe_read(&daddr, sizeof(daddr),
+                 &skp->__sk_common.skc_v6_daddr.in6_u.u6_addr32);
+#ifdef CONFIG_NET_NS
+  bpf_probe_read(&skc_net, sizeof(skc_net), &skp->__sk_common.skc_net);
+  bpf_probe_read(&net_ns_inum, sizeof(net_ns_inum), &skc_net.net->ns.inum);
+#else
+  net_ns_inum = 0;
+#endif
+
+  ##FILTER_NETNS##
+
+  tuple->saddr = saddr;
+  tuple->daddr = daddr;
+  tuple->sport = sport;
+  tuple->dport = dport;
+  tuple->netns = net_ns_inum;
+
+  // if addresses or ports are 0, ignore
+  if (saddr == 0 || daddr == 0 || sport == 0 || dport == 0) {
+      return 0;
+  }
+
+  return 1;
+}
+
+static bool check_family(struct sock *sk, u16 expected_family) {
+  u64 zero = 0;
+  u16 family = 0;
+
+  bpf_probe_read(&family, sizeof(family), &sk->__sk_common.skc_family);
+
+  return family == expected_family;
+}
+
+int trace_connect_v4_entry(struct pt_regs *ctx, struct sock *sk)
+{
+  u64 pid = bpf_get_current_pid_tgid();
+
+  ##FILTER_PID##
+
+  // stash the sock ptr for lookup on return
+  connectsock.update(&pid, &sk);
+
+  return 0;
+}
+
+int trace_connect_v4_return(struct pt_regs *ctx)
+{
+  int ret = PT_REGS_RC(ctx);
+  u64 pid = bpf_get_current_pid_tgid();
+
+  struct sock **skpp;
+  skpp = connectsock.lookup(&pid);
+  if (skpp == 0) {
+      return 0;       // missed entry
+  }
+
+  connectsock.delete(&pid);
+
+  if (ret != 0) {
+      // failed to send SYNC packet, may not have populated
+      // socket __sk_common.{skc_rcv_saddr, ...}
+      return 0;
+  }
+
+  // pull in details
+  struct sock *skp = *skpp;
+  struct ipv4_tuple_t t = { };
+  if (!read_ipv4_tuple(&t, skp)) {
+      return 0;
+  }
+
+  struct pid_comm_t p = { };
+  p.pid = pid;
+  bpf_get_current_comm(&p.comm, sizeof(p.comm));
+
+  tuplepid_ipv4.update(&t, &p);
+
+  return 0;
+}
+
+int trace_connect_v6_entry(struct pt_regs *ctx, struct sock *sk)
+{
+  u64 pid = bpf_get_current_pid_tgid();
+
+  ##FILTER_PID##
+
+  // stash the sock ptr for lookup on return
+  connectsock.update(&pid, &sk);
+
+  return 0;
+}
+
+int trace_connect_v6_return(struct pt_regs *ctx)
+{
+  int ret = PT_REGS_RC(ctx);
+  u64 pid = bpf_get_current_pid_tgid();
+
+  struct sock **skpp;
+  skpp = connectsock.lookup(&pid);
+  if (skpp == 0) {
+      return 0;       // missed entry
+  }
+
+  connectsock.delete(&pid);
+
+  if (ret != 0) {
+      // failed to send SYNC packet, may not have populated
+      // socket __sk_common.{skc_rcv_saddr, ...}
+      return 0;
+  }
+
+  // pull in details
+  struct sock *skp = *skpp;
+  struct ipv6_tuple_t t = { };
+  if (!read_ipv6_tuple(&t, skp)) {
+      return 0;
+  }
+
+  struct pid_comm_t p = { };
+  p.pid = pid;
+  bpf_get_current_comm(&p.comm, sizeof(p.comm));
+
+  tuplepid_ipv6.update(&t, &p);
+
+  return 0;
+}
+
+int trace_tcp_set_state_entry(struct pt_regs *ctx, struct sock *sk, int state)
+{
+  if (state != TCP_ESTABLISHED && state != TCP_CLOSE) {
+      return 0;
+  }
+
+  struct sock *skp;
+  bpf_probe_read(&skp, sizeof(struct sock *), &sk);
+
+  u8 ipver = 0;
+  if (check_family(skp, AF_INET)) {
+      ipver = 4;
+      struct ipv4_tuple_t t = { };
+      if (!read_ipv4_tuple(&t, skp)) {
+          return 0;
+      }
+
+      if (state == TCP_CLOSE) {
+          tuplepid_ipv4.delete(&t);
+          return 0;
+      }
+
+      struct pid_comm_t *p;
+      p = tuplepid_ipv4.lookup(&t);
+      if (p == 0) {
+          return 0;       // missed entry
+      }
+
+      struct tcp_ipv4_event_t evt4 = { };
+      evt4.type = TCP_EVENT_TYPE_CONNECT;
+      evt4.pid = p->pid >> 32;
+      evt4.ip = ipver;
+      evt4.saddr = t.saddr;
+      evt4.daddr = t.daddr;
+      evt4.sport = ntohs(t.sport);
+      evt4.dport = ntohs(t.dport);
+      evt4.netns = t.netns;
+
+      int i;
+      for (i = 0; i < TASK_COMM_LEN; i++) {
+          evt4.comm[i] = p->comm[i];
+      }
+
+      tcp_ipv4_event.perf_submit(ctx, &evt4, sizeof(evt4));
+      tuplepid_ipv4.delete(&t);
+  } else if (check_family(skp, AF_INET6)) {
+      ipver = 6;
+      struct ipv6_tuple_t t = { };
+      if (!read_ipv6_tuple(&t, skp)) {
+          return 0;
+      }
+
+      if (state == TCP_CLOSE) {
+          tuplepid_ipv6.delete(&t);
+          return 0;
+      }
+
+      struct pid_comm_t *p;
+      p = tuplepid_ipv6.lookup(&t);
+      if (p == 0) {
+          return 0;       // missed entry
+      }
+
+      struct tcp_ipv6_event_t evt6 = { };
+      evt6.type = TCP_EVENT_TYPE_CONNECT;
+      evt6.pid = p->pid >> 32;
+      evt6.ip = ipver;
+      evt6.saddr = t.saddr;
+      evt6.daddr = t.daddr;
+      evt6.sport = ntohs(t.sport);
+      evt6.dport = ntohs(t.dport);
+      evt6.netns = t.netns;
+
+      int i;
+      for (i = 0; i < TASK_COMM_LEN; i++) {
+          evt6.comm[i] = p->comm[i];
+      }
+
+      tcp_ipv6_event.perf_submit(ctx, &evt6, sizeof(evt6));
+      tuplepid_ipv6.delete(&t);
+  }
+  // else drop
+
+  return 0;
+}
+
+int trace_close_entry(struct pt_regs *ctx, struct sock *sk)
+{
+  u64 pid = bpf_get_current_pid_tgid();
+
+  ##FILTER_PID##
+
+  // pull in details
+  struct sock *skp;
+  bpf_probe_read(&skp, sizeof(struct sock *), &sk);
+
+  u8 oldstate = 0;
+  bpf_probe_read(&oldstate, sizeof(oldstate), (u8 *)&skp->sk_state);
+  // Don't generate close events for connections that were never
+  // established in the first place.
+  if (oldstate == TCP_SYN_SENT ||
+      oldstate == TCP_SYN_RECV ||
+      oldstate == TCP_NEW_SYN_RECV)
+      return 0;
+
+  u8 ipver = 0;
+  if (check_family(skp, AF_INET)) {
+      ipver = 4;
+      struct ipv4_tuple_t t = { };
+      if (!read_ipv4_tuple(&t, skp)) {
+          return 0;
+      }
+
+      struct tcp_ipv4_event_t evt4 = { };
+      evt4.type = TCP_EVENT_TYPE_CLOSE;
+      evt4.pid = pid >> 32;
+      evt4.ip = ipver;
+      evt4.saddr = t.saddr;
+      evt4.daddr = t.daddr;
+      evt4.sport = ntohs(t.sport);
+      evt4.dport = ntohs(t.dport);
+      evt4.netns = t.netns;
+      bpf_get_current_comm(&evt4.comm, sizeof(evt4.comm));
+
+      tcp_ipv4_event.perf_submit(ctx, &evt4, sizeof(evt4));
+  } else if (check_family(skp, AF_INET6)) {
+      ipver = 6;
+      struct ipv6_tuple_t t = { };
+      if (!read_ipv6_tuple(&t, skp)) {
+          return 0;
+      }
+
+      struct tcp_ipv6_event_t evt6 = { };
+      evt6.type = TCP_EVENT_TYPE_CLOSE;
+      evt6.pid = pid >> 32;
+      evt6.ip = ipver;
+      evt6.saddr = t.saddr;
+      evt6.daddr = t.daddr;
+      evt6.sport = ntohs(t.sport);
+      evt6.dport = ntohs(t.dport);
+      evt6.netns = t.netns;
+      bpf_get_current_comm(&evt6.comm, sizeof(evt6.comm));
+
+      tcp_ipv6_event.perf_submit(ctx, &evt6, sizeof(evt6));
+  }
+  // else drop
+
+  return 0;
+};
+
+int trace_accept_return(struct pt_regs *ctx)
+{
+  struct sock *newsk = (struct sock *)PT_REGS_RC(ctx);
+  u64 pid = bpf_get_current_pid_tgid();
+
+  ##FILTER_PID##
+
+  if (newsk == NULL) {
+      return 0;
+  }
+
+  // pull in details
+  u16 lport = 0, dport = 0;
+  u32 net_ns_inum = 0;
+  u8 ipver = 0;
+
+  bpf_probe_read(&dport, sizeof(dport), &newsk->__sk_common.skc_dport);
+  bpf_probe_read(&lport, sizeof(lport), &newsk->__sk_common.skc_num);
+
+  // Get network namespace id, if kernel supports it
+#ifdef CONFIG_NET_NS
+  possible_net_t skc_net;
+  bpf_probe_read(&skc_net, sizeof(skc_net), &newsk->__sk_common.skc_net);
+  bpf_probe_read(&net_ns_inum, sizeof(net_ns_inum), &skc_net.net->ns.inum);
+#else
+  net_ns_inum = 0;
+#endif
+
+  ##FILTER_NETNS##
+
+  if (check_family(newsk, AF_INET)) {
+      ipver = 4;
+
+      struct tcp_ipv4_event_t evt4 = { 0 };
+
+      evt4.type = TCP_EVENT_TYPE_ACCEPT;
+      evt4.netns = net_ns_inum;
+      evt4.pid = pid >> 32;
+      evt4.ip = ipver;
+
+      bpf_probe_read(&evt4.saddr, sizeof(evt4.saddr),
+                     &newsk->__sk_common.skc_rcv_saddr);
+      bpf_probe_read(&evt4.daddr, sizeof(evt4.daddr),
+                     &newsk->__sk_common.skc_daddr);
+
+      evt4.sport = lport;
+      evt4.dport = ntohs(dport);
+      bpf_get_current_comm(&evt4.comm, sizeof(evt4.comm));
+
+      // do not send event if IP address is 0.0.0.0 or port is 0
+      if (evt4.saddr != 0 && evt4.daddr != 0 &&
+          evt4.sport != 0 && evt4.dport != 0) {
+          tcp_ipv4_event.perf_submit(ctx, &evt4, sizeof(evt4));
+      }
+  } else if (check_family(newsk, AF_INET6)) {
+      ipver = 6;
+
+      struct tcp_ipv6_event_t evt6 = { 0 };
+
+      evt6.type = TCP_EVENT_TYPE_ACCEPT;
+      evt6.netns = net_ns_inum;
+      evt6.pid = pid >> 32;
+      evt6.ip = ipver;
+
+      bpf_probe_read(&evt6.saddr, sizeof(evt6.saddr),
+                     &newsk->__sk_common.skc_v6_rcv_saddr.in6_u.u6_addr32);
+      bpf_probe_read(&evt6.daddr, sizeof(evt6.daddr),
+                     &newsk->__sk_common.skc_v6_daddr.in6_u.u6_addr32);
+
+      evt6.sport = lport;
+      evt6.dport = ntohs(dport);
+      bpf_get_current_comm(&evt6.comm, sizeof(evt6.comm));
+
+      // do not send event if IP address is 0.0.0.0 or port is 0
+      if (evt6.saddr != 0 && evt6.daddr != 0 &&
+          evt6.sport != 0 && evt6.dport != 0) {
+          tcp_ipv6_event.perf_submit(ctx, &evt6, sizeof(evt6));
+      }
+  }
+  // else drop
+
+  return 0;
+}
+"""
+
+TASK_COMM_LEN = 16   # linux/sched.h
+
+
+class TCPIPV4Evt(ctypes.Structure):
+    _fields_ = [
+            ("type", ctypes.c_uint),
+            ("pid", ctypes.c_uint),
+            ("comm", ctypes.c_char * TASK_COMM_LEN),
+            ("ip", ctypes.c_ubyte),
+            ("saddr", ctypes.c_uint),
+            ("daddr", ctypes.c_uint),
+            ("sport", ctypes.c_ushort),
+            ("dport", ctypes.c_ushort),
+            ("netns", ctypes.c_uint)
+    ]
+
+
+class TCPIPV6Evt(ctypes.Structure):
+    _fields_ = [
+            ("type", ctypes.c_uint),
+            ("pid", ctypes.c_uint),
+            ("comm", ctypes.c_char * TASK_COMM_LEN),
+            ("ip", ctypes.c_ubyte),
+            ("saddr", (ctypes.c_ulong * 2)),
+            ("daddr", (ctypes.c_ulong * 2)),
+            ("sport", ctypes.c_ushort),
+            ("dport", ctypes.c_ushort),
+            ("netns", ctypes.c_uint)
+    ]
+
+
+verbose_types = {"C": "connect", "A": "accept",
+                 "X": "close", "U": "unknown"}
+
+
+def print_ipv4_event(cpu, data, size):
+    event = ctypes.cast(data, ctypes.POINTER(TCPIPV4Evt)).contents
+    if event.type == 1:
+        type_str = "C"
+    elif event.type == 2:
+        type_str = "A"
+    elif event.type == 3:
+        type_str = "X"
+    else:
+        type_str = "U"
+
+    if args.verbose:
+        print("%-12s " % (verbose_types[type_str]), end="")
+    else:
+        print("%-2s " % (type_str), end="")
+
+    print("%-6d %-16s %-2d %-16s %-16s %-6d %-6d" %
+          (event.pid, event.comm.decode('utf-8'),
+           event.ip,
+           inet_ntop(AF_INET, pack("I", event.saddr)),
+           inet_ntop(AF_INET, pack("I", event.daddr)),
+           event.sport,
+           event.dport), end="")
+    if args.verbose and not args.netns:
+        print(" %-8d" % event.netns)
+    else:
+        print()
+
+
+def print_ipv6_event(cpu, data, size):
+    event = ctypes.cast(data, ctypes.POINTER(TCPIPV6Evt)).contents
+    if event.type == 1:
+        type_str = "C"
+    elif event.type == 2:
+        type_str = "A"
+    elif event.type == 3:
+        type_str = "X"
+    else:
+        type_str = "U"
+
+    if args.verbose:
+        print("%-12s " % (verbose_types[type_str]), end="")
+    else:
+        print("%-2s " % (type_str), end="")
+
+    print("%-6d %-16s %-2d %-16s %-16s %-6d %-6d" %
+          (event.pid, event.comm.decode('utf-8'),
+           event.ip,
+           "["+inet_ntop(AF_INET6, event.saddr)+"]",
+           "["+inet_ntop(AF_INET6, event.daddr)+"]",
+           event.sport,
+           event.dport), end="")
+    if args.verbose and not args.netns:
+        print(" %-8d" % event.netns)
+    else:
+        print()
+
+
+pid_filter = ""
+netns_filter = ""
+
+if args.pid:
+    pid_filter = 'if (pid >> 32 != %d) { return 0; }' % args.pid
+if args.netns:
+    netns_filter = 'if (net_ns_inum != %d) { return 0; }' % args.netns
+
+bpf_text = bpf_text.replace('##FILTER_PID##', pid_filter)
+bpf_text = bpf_text.replace('##FILTER_NETNS##', netns_filter)
+
+# initialize BPF
+b = BPF(text=bpf_text)
+b.attach_kprobe(event="tcp_v4_connect", fn_name="trace_connect_v4_entry")
+b.attach_kretprobe(event="tcp_v4_connect", fn_name="trace_connect_v4_return")
+b.attach_kprobe(event="tcp_v6_connect", fn_name="trace_connect_v6_entry")
+b.attach_kretprobe(event="tcp_v6_connect", fn_name="trace_connect_v6_return")
+b.attach_kprobe(event="tcp_set_state", fn_name="trace_tcp_set_state_entry")
+b.attach_kprobe(event="tcp_close", fn_name="trace_close_entry")
+b.attach_kretprobe(event="inet_csk_accept", fn_name="trace_accept_return")
+
+print("Tracing TCP established connections. Ctrl-C to end.")
+
+# header
+if args.verbose:
+    print("%-12s %-6s %-16s %-2s %-16s %-16s %-6s %-7s" % ("TYPE",
+          "PID", "COMM", "IP", "SADDR", "DADDR", "SPORT", "DPORT"), end="")
+    if not args.netns:
+        print("%-8s" % "NETNS", end="")
+    print()
+else:
+    print("%-2s %-6s %-16s %-2s %-16s %-16s %-6s %-6s" %
+          ("T", "PID", "COMM", "IP", "SADDR", "DADDR", "SPORT", "DPORT"))
+
+
+def inet_ntoa(addr):
+    dq = ''
+    for i in range(0, 4):
+        dq = dq + str(addr & 0xff)
+        if (i != 3):
+            dq = dq + '.'
+        addr = addr >> 8
+    return dq
+
+
+b["tcp_ipv4_event"].open_perf_buffer(print_ipv4_event)
+b["tcp_ipv6_event"].open_perf_buffer(print_ipv6_event)
+while True:
+    b.kprobe_poll()
diff --git a/tools/tcptracer_example.txt b/tools/tcptracer_example.txt
new file mode 100644
index 000000000000..7798610aa06a
--- /dev/null
+++ b/tools/tcptracer_example.txt
@@ -0,0 +1,26 @@
+Demonstrations of tcptracer, the Linux eBPF/bcc version.
+
+
+This tool traces the kernel function performing TCP connections (eg, via a
+connect() or accept() syscalls) and closing them (explicitly or if the process
+dies). Some example output (IP addresses are fake):
+
+```
+# ./tcptracer
+Tracing TCP established connections. Ctrl-C to end.
+T  PID    COMM             IP SADDR            DADDR            SPORT  DPORT
+C  28943  telnet           4  192.168.1.2      192.168.1.1      59306  23
+C  28818  curl             6  [::1]            [::1]            55758  80
+X  28943  telnet           4  192.168.1.2      192.168.1.1      59306  23
+A  28817  nc               6  [::1]            [::1]            80     55758
+X  28818  curl             6  [::1]            [::1]            55758  80
+X  28817  nc               6  [::1]            [::1]            80     55758
+A  28978  nc               4  10.202.210.1     10.202.109.12    8080   59160
+X  28978  nc               4  10.202.210.1     10.202.109.12    8080   59160
+```
+
+This output shows three conections, one outgoing from a "telnet" process, one
+outgoing from "curl" to a local netcat, and one incoming received by the "nc"
+process. The output details show the kind of event (C for connection, X for
+close and A for accept), PID, IP version, source address, destination address,
+source port and destination port.