The tun
interface is a software loopback mechanism that
can be loosely described as the network interface analog of the
pty(4), that
is, tun
does for network interfaces what the
pty(4)
driver does for terminals.
The tun
driver, like the
pty(4)
driver, provides two interfaces: an interface like the usual facility it is
simulating (a network interface in the case of tun
,
or a terminal for
pty(4)),
and a character-special device “control” interface. A client
program transfers IP (by default) packets to or from the
tun
“control” interface. The
tap(4)
interface provides similar functionality at the Ethernet layer: a client
will transfer Ethernet frames to or from a
tap(4)
“control” interface.
The network interfaces are named
“tun0
”,
“tun1
”, etc., one for each control
device that has been opened. These network interfaces persist until the
if_tuntap.ko module is unloaded, or until removed
with the
ifconfig(8)
command.
tun
devices are created using interface
cloning. This is done using the “ifconfig tunN
create” command. This is the preferred method
of creating tun
devices. The same method allows
removal of interfaces. For this, use the “ifconfig
tunN destroy” command.
If the
sysctl(8)
variable net.link.tun.devfs_cloning is non-zero, the
tun
interface permits opens on the special control
device /dev/tun. When this device is opened,
tun
will return a handle for the lowest unused
tun
device (use
devname(3)
to determine which).
Disabling the legacy devfs cloning functionality may break
existing applications which use
tun
, such as
ppp(8) and
ssh(1). It
therefore defaults to being enabled until further notice.
Control devices (once successfully opened) persist until
if_tuntap.ko is unloaded in the same way that
network interfaces persist (see above).
Each interface supports the usual network-interface
ioctl(2)s,
such as SIOCAIFADDR
and thus can be used with
ifconfig(8)
like any other interface. At boot time, they are
POINTOPOINT
interfaces, but this can be changed; see
the description of the control device, below. When the system chooses to
transmit a packet on the network interface, the packet can be read from the
control device (it appears as “input” there); writing a packet
to the control device generates an input packet on the network interface, as
if the (non-existent) hardware had just received it.
The tunnel device
(/dev/tunN) is exclusive-open
(it cannot be opened if it is already open). A
read(2)
call will return an error (EHOSTDOWN
) if the
interface is not “ready” (which means that the control device
is open and the interface's address has been set).
Once the interface is ready,
read(2)
will return a packet if one is available; if not, it will either block until
one is or return EWOULDBLOCK
, depending on whether
non-blocking I/O has been enabled. If the packet is longer than is allowed
for in the buffer passed to
read(2),
the extra data will be silently dropped.
If the TUNSLMODE
ioctl has been set,
packets read from the control device will be prepended with the destination
address as presented to the network interface output routine,
tunoutput
(). The destination address is in
struct sockaddr format. The actual length of the
prepended address is in the member sa_len. If the
TUNSIFHEAD
ioctl has been set, packets will be
prepended with a four byte address family in network byte order.
TUNSLMODE
and TUNSIFHEAD
are
mutually exclusive. In any case, the packet data follows immediately.
A
write(2)
call passes a packet in to be “received” on the
pseudo-interface. If the TUNSIFHEAD
ioctl has been
set, the address family must be prepended, otherwise the packet is assumed
to be of type AF_INET
. Each
write(2)
call supplies exactly one packet; the packet length is taken from the amount
of data provided to
write(2)
(minus any supplied address family). Writes will not block; if the packet
cannot be accepted for a transient reason (e.g., no buffer space available),
it is silently dropped; if the reason is not transient (e.g., packet too
large), an error is returned.
The following
ioctl(2)
calls are supported (defined in
<net/if_tun.h>
):
TUNSDEBUG
- The argument should be a pointer to an int; this
sets the internal debugging variable to that value. What, if anything,
this variable controls is not documented here; see the source code.
TUNGDEBUG
- The argument should be a pointer to an int; this
stores the internal debugging variable's value into it.
TUNSIFINFO
- The argument should be a pointer to an struct
tuninfo and allows setting the MTU and the baudrate of the tunnel
device. The type must be the same as returned by
TUNGIFINFO
or set to
IFT_PPP
else the
ioctl(2)
call will fail. The struct tuninfo is declared in
<net/if_tun.h>
.
The use of this ioctl is restricted to the super-user.
TUNGIFINFO
- The argument should be a pointer to an struct
tuninfo, where the current MTU, type, and baudrate will be
stored.
TUNSIFMODE
- The argument should be a pointer to an int; its
value must be either
IFF_POINTOPOINT
or
IFF_BROADCAST
and should have
IFF_MULTICAST
OR'd into the value if multicast
support is required. The type of the corresponding
“tun
N”
interface is set to the supplied type. If the value is outside the above
range, an EINVAL
error is returned. The interface
must be down at the time; if it is up, an EBUSY
error is returned.
TUNSLMODE
- The argument should be a pointer to an int; a
non-zero value turns off “multi-af” mode and turns on
“link-layer” mode, causing packets read from the tunnel
device to be prepended with the network destination address (see
above).
TUNSIFPID
- Will set the pid owning the tunnel device to the current process's
pid.
TUNSIFHEAD
- The argument should be a pointer to an int; a
non-zero value turns off “link-layer” mode, and enables
“multi-af” mode, where every packet is preceded with a four
byte address family.
TUNGIFHEAD
- The argument should be a pointer to an int; the
ioctl sets the value to one if the device is in “multi-af”
mode, and zero otherwise.
FIONBIO
- Turn non-blocking I/O for reads off or on, according as the argument
int's value is or is not zero. (Writes are always
non-blocking.)
FIOASYNC
- Turn asynchronous I/O for reads (i.e., generation of
SIGIO
when data is available to be read) off or
on, according as the argument int's value is or is
not zero.
FIONREAD
- If any packets are queued to be read, store the size of the first one into
the argument int; otherwise, store zero.
TIOCSPGRP
- Set the process group to receive
SIGIO
signals,
when asynchronous I/O is enabled, to the argument
int value.
TIOCGPGRP
- Retrieve the process group value for
SIGIO
signals
into the argument int value.
The control device also supports
select(2)
for read; selecting for write is pointless, and always succeeds, since
writes are always non-blocking.
On the last close of the data device, by default, the interface is
brought down (as if with ifconfig
tunN down
). All queued packets
are thrown away. If the interface is up when the data device is not open
output packets are always thrown away rather than letting them pile up.