implement SO_BINDTODEVICE for inet protocols

bind to device is needed to properly support VRF-Lite under
Linux (see [1] for details).

[1]: https://www.kernel.org/doc/Documentation/networking/vrf.txt

4 files changed, 99 insertions, 6 deletions

diff --git a/lib/kernel/doc/src/inet.xml b/lib/kernel/doc/src/inet.xml
index 076e50cd10..947e4d4560 100644
--- a/lib/kernel/doc/src/inet.xml
+++ b/lib/kernel/doc/src/inet.xml
@@ -897,6 +897,32 @@ setcap cap_sys_admin,cap_sys_ptrace,cap_dac_read_search+epi beam.smp</code>
 	      <seealso marker="file#native_name_encoding/0"><c>file:native_name_encoding/0</c></seealso>.</p></item>
 	    </list>
           </item>
+	  <tag><c>{bind_to_device, Ifname :: binary()}</c></tag>
+	  <item>
+	    <p>Binds a socket to a specific network interface. This option
+	      must be used in a function call that creates a socket, that is,
+	      <seealso marker="gen_tcp#connect/3"><c>gen_tcp:connect/3,4</c></seealso>,
+	      <seealso marker="gen_tcp#listen/2"><c>gen_tcp:listen/2</c></seealso>,
+	      <seealso marker="gen_udp#open/1"><c>gen_udp:open/1,2</c></seealso>, or
+	      <seealso marker="gen_sctp#open/0"><c>gen_sctp:open/0,1,2</c></seealso>.</p>
+	    <p>Unlike <seealso marker="#getifaddrs/0"><c>getifaddrs/0</c></seealso>, Ifname
+	      is encoded a binary. In the unlikely case that a system is using
+	      non-7-bit-ASCII characters in network device names, special care
+	      has to be taken when encoding this argument.</p>
+	    <p>This option uses the Linux-specific socket option
+	      <c>SO_BINDTODEVICE</c>, such as in Linux kernel 2.0.30 or later,
+	      and therefore only exists when the runtime system
+	      is compiled for such an operating system.</p>
+	    <p>Before Linux 3.8, this socket option could be set, but could not retrieved
+	      with <seealso marker="#getopts/2"><c>getopts/2</c></seealso>. Since Linux 3.8,
+	      it is readable.</p>
+	    <p>The virtual machine also needs elevated privileges, either
+	      running as superuser or (for Linux) having capability
+	    <c>CAP_NET_RAW</c>.</p>
+	    <p>The primary use case for this option is to bind sockets into
+	      <url href="http://www.kernel.org/doc/Documentation/networking/vrf.txt">Linux VRF instances</url>.
+	    </p>
+	  </item>
           <tag><c>list</c></tag>
           <item>
             <p>Received <c>Packet</c> is delivered as a list.</p>
diff --git a/lib/kernel/src/inet.erl b/lib/kernel/src/inet.erl
index f5c13ecdd7..5be790b7d9 100644
--- a/lib/kernel/src/inet.erl
+++ b/lib/kernel/src/inet.erl
@@ -702,7 +702,7 @@ connect_options() ->
      header, active, packet, packet_size, buffer, mode, deliver, line_delimiter,
      exit_on_close, high_watermark, low_watermark, high_msgq_watermark,
      low_msgq_watermark, send_timeout, send_timeout_close, delay_send, raw,
-     show_econnreset].
+     show_econnreset, bind_to_device].
     
 connect_options(Opts, Mod) ->
     BaseOpts = 
@@ -770,7 +770,7 @@ listen_options() ->
      header, active, packet, buffer, mode, deliver, backlog, ipv6_v6only,
      exit_on_close, high_watermark, low_watermark, high_msgq_watermark,
      low_msgq_watermark, send_timeout, send_timeout_close, delay_send,
-     packet_size, raw, show_econnreset].
+     packet_size, raw, show_econnreset, bind_to_device].
 
 listen_options(Opts, Mod) ->
     BaseOpts = 
@@ -850,7 +850,7 @@ udp_options() ->
      deliver, ipv6_v6only,
      broadcast, dontroute, multicast_if, multicast_ttl, multicast_loop,
      add_membership, drop_membership, read_packets,raw,
-     high_msgq_watermark, low_msgq_watermark].
+     high_msgq_watermark, low_msgq_watermark, bind_to_device].
 
 
 udp_options(Opts, Mod) ->
@@ -919,6 +919,7 @@ sctp_options() ->
 [   % The following are generic inet options supported for SCTP sockets:
     mode, active, buffer, tos, tclass, priority, dontroute, reuseaddr, linger, sndbuf,
     recbuf, ipv6_v6only, high_msgq_watermark, low_msgq_watermark,
+    bind_to_device,
 
     % Other options are SCTP-specific (though they may be similar to their
     % TCP and UDP counter-parts):
@@ -1055,7 +1056,6 @@ binary2filename(Bin) ->
 	    Bin
     end.
 
-
 translate_ip(any,      inet) -> {0,0,0,0};
 translate_ip(loopback, inet) -> {127,0,0,1};
 translate_ip(any,      inet6) -> {0,0,0,0,0,0,0,0};
diff --git a/lib/kernel/src/inet_int.hrl b/lib/kernel/src/inet_int.hrl
index 4e8f59a3b9..e6cd48935a 100644
--- a/lib/kernel/src/inet_int.hrl
+++ b/lib/kernel/src/inet_int.hrl
@@ -154,6 +154,7 @@
 -define(INET_LOPT_TCP_SHOW_ECONNRESET, 39).
 -define(INET_LOPT_LINE_DELIM,     40).
 -define(INET_OPT_TCLASS,          41).
+-define(INET_OPT_BIND_TO_DEVICE,  42).
 % Specific SCTP options: separate range:
 -define(SCTP_OPT_RTOINFO,	 	100).
 -define(SCTP_OPT_ASSOCINFO,	 	101).
diff --git a/lib/kernel/test/inet_SUITE.erl b/lib/kernel/test/inet_SUITE.erl
index f60c13d2e3..86f6b95fb9 100644
--- a/lib/kernel/test/inet_SUITE.erl
+++ b/lib/kernel/test/inet_SUITE.erl
@@ -40,7 +40,8 @@
 	 lookup_bad_search_option/1,
 	 getif/1,
 	 getif_ifr_name_overflow/1,getservbyname_overflow/1, getifaddrs/1,
-	 parse_strict_address/1, simple_netns/1, simple_netns_open/1]).
+	 parse_strict_address/1, simple_netns/1, simple_netns_open/1,
+         simple_bind_to_device/1, simple_bind_to_device_open/1]).
 
 -export([get_hosts/1, get_ipv6_hosts/1, parse_hosts/1, parse_address/1,
 	 kill_gethost/0, parallell_gethost/0, test_netns/0]).
@@ -58,7 +59,8 @@ all() ->
      gethostnative_debug_level, gethostnative_soft_restart,
      lookup_bad_search_option,
      getif, getif_ifr_name_overflow, getservbyname_overflow,
-     getifaddrs, parse_strict_address, simple_netns, simple_netns_open].
+     getifaddrs, parse_strict_address, simple_netns, simple_netns_open,
+     simple_bind_to_device, simple_bind_to_device_open].
 
 groups() -> 
     [{parse, [], [parse_hosts, parse_address]}].
@@ -1247,3 +1249,67 @@ cmd(CmdString) ->
     io:put_chars(["# ",CmdString,io_lib:nl()]),
     io:put_chars([os:cmd(CmdString++" ; echo '  =>' $?")]),
     ok.
+
+-define(CAP_NET_RAW, 13).        %% from /usr/include/linux/capability.h
+
+can_bind_to_device({unix, linux}, {Major, _, _})
+  when Major > 2 ->
+    Status = os:cmd("cat /proc/self/status | grep CapEff"),
+    [_, CapEffStr] = string:tokens(Status, [$\n, $\t]),
+    CapEff = list_to_integer(CapEffStr, 16),
+    if CapEff band (1 bsl ?CAP_NET_RAW) =/= 0 ->
+            ok;
+       true ->
+            {skip,"insufficient capabilities, CAP_NET_RAW not granted"}
+    end;
+can_bind_to_device(_OS, _Version) ->
+    {skip,"socket option bind_to_device not supported on this OS or version"}.
+
+simple_bind_to_device(Config) when is_list(Config) ->
+    case can_bind_to_device(os:type(), os:version()) of
+        ok ->
+            {ok,U} = gen_udp:open(0),
+            jog_bind_to_device_opt(U),
+            ok = gen_udp:close(U),
+            %%
+            {ok,L} = gen_tcp:listen(0, []),
+            jog_bind_to_device_opt(L),
+            ok = gen_tcp:close(L),
+            %%
+            case gen_sctp:open() of
+                {ok,S} ->
+                    jog_bind_to_device_opt(S),
+                    ok = gen_sctp:close(S);
+                {error,eprotonosupport} ->
+                    ok
+            end;
+        Other ->
+            Other
+    end.
+
+%% Smoke test bind_to_device support.
+simple_bind_to_device_open(Config) when is_list(Config) ->
+    case can_bind_to_device(os:type(), os:version()) of
+        ok ->
+            {ok,U} = gen_udp:open(0, [binary,{bind_to_device,<<"lo">>},inet]),
+            ok = gen_udp:close(U),
+            {ok,T} = gen_tcp:listen(0, [binary,{bind_to_device,<<"lo">>},inet]),
+            ok = gen_tcp:close(T),
+
+            case gen_sctp:open(0, [binary,{bind_to_device,<<"lo">>},inet]) of
+                {ok,S} ->
+                    ok = gen_sctp:close(S);
+                {error,eprotonosupport} ->
+                    ok
+            end;
+        Other ->
+            Other
+    end.
+
+jog_bind_to_device_opt(S) ->
+    %% This is just jogging the option mechanics
+    ok = inet:setopts(S, [{bind_to_device,<<>>}]),
+    {ok,[{bind_to_device,<<>>}]} = inet:getopts(S, [bind_to_device]),
+    ok = inet:setopts(S, [{bind_to_device,<<"lo">>}]),
+    {ok,[{bind_to_device,<<"lo">>}]} = inet:getopts(S, [bind_to_device]),
+    ok.