.\" Copyright (c) 2006 Vadim Goncharov <vadimnuclight@tpu.ru>
.\" All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\"    notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\"    notice, this list of conditions and the following disclaimer in the
.\"    documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.Dd June 10, 2006
.Dt NG_TAG 4
.Os
.Sh NAME
.Nm ng_tag
.Nd "mbuf tags manipulating netgraph node type"
.Sh SYNOPSIS
.In netgraph/ng_tag.h
.Sh DESCRIPTION
The
.Nm tag
node type allows mbuf packet tags (see
.Xr mbuf_tags 9 )
to be examined, stripped or applied to data travelling through a
Netgraph network.
Mbuf tags are used in many parts of the
.Fx
kernel network subsystem,
including the storage of VLAN tags as described in
.Xr vlan 4 ,
Mandatory Access Control (MAC) labels as described in
.Xr mac 9 ,
IPsec policy information as described in
.Xr ipsec 4 ,
and packet filter tags used by
.Xr pf 4 .
One should also consider useful setting or checking
.Xr ipfw 8
tags, which are implemented as mbuf tags, too.
.Pp
Each node allows an arbitrary number of connections to arbitrarily
named hooks.
With each hook is associated a tag which will be searched in the list
of all tags attached to a packet incoming to this hook, a destination hook
for matching packets, a destination hook for non-matching packets,
a tag which will be appended to data leaving node through this hook,
and various statistics counters.
.Pp
The list of incoming packet's tags is traversed to find a tag with
specified
.Va type
and
.Va cookie
values.
Upon match, if specified
.Va tag_len
is non-zero,
.Va tag_data
of tag is checked to be identical to that specified in the hook structure.
Packets with matched tags are forwarded to
.Dq match
destination hook, or forwarded to
.Dq non-match
hook otherwise.
Either or both destination hooks can be an empty string, or may
not exist, in which case the packet is dropped.
.Pp
Tag list of packets leaving the node is extended with a new tag
specified in outgoing hook structure (it is possible to avoid appending
a new tag to pass packet completely unchanged by specifying zero
.Va type
and
.Va cookie
values in the structure of the corresponding outgoing hook).
Additionally,
a tag can be stripped from incoming packet after match if
.Va strip
flag is set.
This can be used for simple tag removal or tag replacement, if combined
with tag addition on outgoing matching hook.
Note that new tag is appended unconditionally, without checking if
such a tag is already present in the list (it is up to user to check
if this is a concern).
.Pp
New hooks are initially configured to drop all incoming packets
(as all hook names are empty strings; zero values can be specified
to forward all packets to non-matching hook),
and to forward all outgoing packets without any tag appending.
.Pp
Data payload of packets passing through the node is completely
unchanged, all operations can affect tag list only.
.Sh HOOKS
This node type supports any number of hooks having arbitrary names.
In order to allow internal optimizations, user should never try to
configure a hook with a structure pointing to hooks which do not exist yet.
The safe way is to create all hooks first, then begin to configure them.
.Sh CONTROL MESSAGES
This node type supports the generic control messages, plus the following:
.Bl -tag -width foo
.It Dv NGM_TAG_SET_HOOKIN Pq Ic sethookin
This command sets tag values which will be searched in the tag list of
incoming packets on a hook.
The following structure must be supplied as an argument:
.Bd -literal -offset 4n
struct ng_tag_hookin {
  char		  thisHook[NG_HOOKSIZ];     /* name of hook */
  char		  ifMatch[NG_HOOKSIZ];	    /* match dest hook */
  char		  ifNotMatch[NG_HOOKSIZ];   /* !match dest hook */
  uint8_t	  strip;		    /* strip tag if found */
  uint32_t	  tag_cookie;		    /* ABI/Module ID */
  uint16_t	  tag_id;		    /* tag ID */
  uint16_t	  tag_len;		    /* length of data */
  uint8_t	  tag_data[0];		    /* tag data */
};
.Ed
.Pp
The hook to be updated is specified in
.Va thisHook .
Data bytes of tag corresponding to specified
.Va tag_id
(type) and
.Va tag_cookie
are placed in the
.Va tag_data
array; there must be
.Va tag_len
of them.
Matching and non-matching incoming packets are delivered out the hooks named
.Va ifMatch
and
.Va ifNotMatch ,
respectively.
If
.Va strip
flag is non-zero, then found tag is deleted from list of packet tags.
.It Dv NGM_TAG_GET_HOOKIN Pq Ic gethookin
This command takes an ASCII string argument, the hook name, and returns the
corresponding
.Vt "struct ng_tag_hookin"
as shown above.
.It Dv NGM_TAG_SET_HOOKOUT Pq Ic sethookout
This command sets tags values which will be applied to outgoing
packets.
The following structure must be supplied as an argument:
.Bd -literal -offset 4n
struct ng_tag_hookout {
  char		  thisHook[NG_HOOKSIZ];     /* name of hook */
  uint32_t	  tag_cookie;		    /* ABI/Module ID */
  uint16_t	  tag_id;		    /* tag ID */
  uint16_t	  tag_len;		    /* length of data */
  uint8_t	  tag_data[0];		    /* tag data */
};
.Ed
.Pp
The hook to be updated is specified in
.Va thisHook .
Other variables mean basically the same as in
.Vt "struct ng_tag_hookin"
shown above, except used for setting values in a new tag.
.It Dv NGM_TAG_GET_HOOKOUT Pq Ic gethookout
This command takes an ASCII string argument, the hook name, and returns the
corresponding
.Vt "struct ng_tag_hookout"
as shown above.
.It Dv NGM_TAG_GET_STATS Pq Ic getstats
This command takes an ASCII string argument, the hook name, and returns the
statistics associated with the hook as a
.Vt "struct ng_tag_hookstat" .
.It Dv NGM_TAG_CLR_STATS Pq Ic clrstats
This command takes an ASCII string argument, the hook name, and clears the
statistics associated with the hook.
.It Dv NGM_TAG_GETCLR_STATS Pq Ic getclrstats
This command is identical to
.Dv NGM_TAG_GET_STATS ,
except that the statistics are also atomically cleared.
.El
.Pp
.Em Note :
statistics counters as well as three statistics messages above work
only if code was compiled with the
.Dv NG_TAG_DEBUG
option.
The reason for this is that statistics is rarely used in practice,
but still consumes CPU cycles for every packet.
Moreover, it is even not accurate on SMP systems due to lack of
synchronization between threads, as this is very expensive.
.Sh SHUTDOWN
This node shuts down upon receipt of a
.Dv NGM_SHUTDOWN
control message, or when all hooks have been disconnected.
.Sh EXAMPLES
It is possible to do a simple L7 filtering by using
.Xr ipfw 8
tags in conjunction with
.Xr ng_bpf 4
traffic analyzer.
Example below explains how to filter DirectConnect P2P network data traffic,
which cannot be done by usual means as it uses random ports.
It is known that such data connection always contains a TCP packet with
6-byte payload string "$Send|".
So ipfw's
.Cm netgraph
action will be used to divert all TCP packets to an
.Xr ng_bpf 4
node which will check for the specified string and return non-matching
packets to
.Xr ipfw 8 .
Matching packets are passed to
.Nm
node, which will set a tag and pass them back to
.Xr ng_bpf 4
node on a hook programmed to accept all packets and pass them back to
.Xr ipfw 8 .
A script provided in
.Xr ng_bpf 4
manual page will be used for programming node.
Note that packets diverted from
.Xr ipfw 8
to Netgraph have no link-level header, so offsets in
.Xr tcpdump 1
expressions must be altered accordingly.
Thus, there will be expression
.Dq Li "ether[40:2]=0x244c && ether[42:4]=0x6f636b20"
on incoming hook and empty expression to match all packets from
.Nm .
.Pp
So, this is
.Xr ngctl 8
script for nodes creating and naming for easier access:
.Bd -literal -offset 4n
/usr/sbin/ngctl -f- <<-SEQ
	mkpeer ipfw: bpf 41 ipfw
	name ipfw:41 dcbpf
	mkpeer dcbpf: tag matched th1
	name dcbpf:matched ngdc
SEQ
.Ed
.Pp
Now
.Dq Li ngdc
node (which is of type
.Nm )
must be programmed to echo all packets received on the
.Dq Li th1
hook back, with the
.Xr ipfw 8
tag 412 attached.
.Dv MTAG_IPFW
value for
.Va tag_cookie
was taken from file
.In netinet/ip_fw.h
and value for
.Va tag_id
is tag number (412), with zero tag length:
.Bd -literal -offset 4n
ngctl msg ngdc: sethookin { thisHook=\e"th1\e" ifNotMatch=\e"th1\e" }
ngctl msg ngdc: sethookout { thisHook=\e"th1\e" \e
  tag_cookie=1148380143 \e
  tag_id=412 }
.Ed
.Pp
Do not forget to program
.Xr ng_bpf 4
.Dq Li ipfw
hook with the above expression (see
.Xr ng_bpf 4
for script doing this) and
.Dq Li matched
hook with an empty expression:
.Bd -literal -offset 4n
ngctl msg dcbpf: setprogram { thisHook=\e"matched\e" ifMatch=\e"ipfw\e" \e
  bpf_prog_len=1 bpf_prog=[ { code=6 k=8192 } ] }
.Ed
.Pp
After finishing with
.Xr netgraph 4
nodes,
.Xr ipfw 8
rules must be added to enable packet flow:
.Bd -literal -offset 4n
ipfw add 100 netgraph 41 tcp from any to any iplen 46
ipfw add 110 reset tcp from any to any tagged 412
.Ed
.Pp
Note: one should ensure that packets are returned to ipfw after processing
inside
.Xr netgraph 4 ,
by setting appropriate
.Xr sysctl 8
variable:
.Bd -literal -offset 4n
sysctl net.inet.ip.fw.one_pass=0
.Ed
.Sh SEE ALSO
.Xr netgraph 4 ,
.Xr ng_bpf 4 ,
.Xr ng_ipfw 4 ,
.Xr ipfw 8 ,
.Xr ngctl 8 ,
.Xr mbuf_tags 9
.Sh HISTORY
The
.Nm
node type was implemented in
.Fx 6.2 .
.Sh AUTHORS
.An Vadim Goncharov Aq Mt vadimnuclight@tpu.ru
.Sh BUGS
For manipulating any tags with data payload (that is, all tags with non-zero
.Va tag_len )
one should care about non-portable machine-dependent representation of
tags on the low level as byte stream.
Perhaps this should be done by another program rather than manually.