TAP(4)

okay so bestie, welcome to TAP(4) — that's the page from the Kernel Interfaces Manual — and honestly? she's giving networking royalty 👑. We're talking tap, we're talking vmnet, we're talking the whole Ethernet tunnel software network interface situation. Buckle up, grab a strawberry 🍓, and let's get into it.

🌸 NAME 🌸

tap, vmnet — Ethernet tunnel software network interface

---

✨ SYNOPSIS ✨

device tuntap

---

🎀 DESCRIPTION 🎀

The tap interface is a software loopback mechanism, and the loosest possible way to describe it is: it's the network-interface version of pty(4). Like, tap does for network interfaces exactly what the pty(4) driver does for terminals. Periodt. 💅

The tap driver, same as the pty(4) driver, hands you two interfaces: one that looks like the usual facility it's simulating (that's an Ethernet network interface when we're talking tap, or a terminal when we're talking pty(4)), plus a character-special device "control" interface. A client program moves Ethernet frames to and from the tap "control" interface. Meanwhile the tun(4) interface does basically the same trick but at the network layer — a client transfers IP (by default) packets to or from a tun(4) "control" interface. Same vibe, different layer 🦋.

The network interfaces get named "tap0", "tap1", etc., one for every control device that's been opened. These Ethernet network interfaces stick around until if_tuntap.ko module is unloaded, or until they get removed with "ifconfig destroy" (more on that below).

tap devices are created using interface cloning, and you do that with the "ifconfig tapN create" command. This is the preferred method of creating tap devices, no cap. The very same method lets you remove interfaces too — for that you use the "ifconfig tapN destroy" command. ✨

If the sysctl(8) variable net.link.tap.devfs_cloning is non-zero, the tap interface permits opens on the special control device /dev/tap. When this device gets opened, tap will return a handle for the lowest unused tap device (use devname(3) to figure out which one that was) 👀.

Disabling the legacy devfs cloning functionality may break existing applications which use tap, such as VMware and ssh(1). It therefore defaults to being enabled until further notice. So like… don't touch it unless you know what you're doing, bestie 😭

Control devices, once successfully opened, persist until if_tuntap.ko is unloaded or the interface is destroyed.

Each interface supports the usual Ethernet network interface ioctl(2)s, which means you can use it with ifconfig(8) like any other Ethernet interface. When the system chooses to transmit an Ethernet frame on the network interface, that frame can be read from the control device (it shows up as "input" over there); and writing an Ethernet frame to the control device generates an input frame on the network interface, as if the (non-existent) hardware had just received it. Iconic behaviour honestly 🫶.

The Ethernet tunnel device, normally /dev/tapN, is exclusive-open (meaning it cannot be opened if it's already open) and is restricted to the super-user — unless the sysctl(8) variable net.link.tap.user_open is non-zero. If the sysctl(8) variable net.link.tap.up_on_open is non-zero, the tunnel device will be marked "up" when the control device is opened. A read() call will return an error (EHOSTDOWN) if the interface is not "ready". Once the interface is ready, read() will return an Ethernet frame if one is available; if not, it will either block until one shows up or return EWOULDBLOCK, depending on whether non-blocking I/O has been enabled. And heads up: if the frame is longer than is allowed for in the buffer passed to read(), the extra data will be silently dropped 💀.

A write(2) call passes an Ethernet frame in to be "received" on the pseudo-interface. Each write() call supplies exactly one frame; the frame length is taken from the amount of data provided to write(). Writes will not block; if the frame 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., frame too large), an error is returned.

🎮 the ioctl(2) calls 🎮

The following ioctl(2) calls are supported (defined in <net/if_tap.h>):

• TAPSIFINFO — Set network interface information (line speed and MTU). The type must be the same as returned by TAPGIFINFO or set to IFT_ETHER, else the ioctl(2) call will fail. The argument should be a pointer to a struct tapinfo.
• TAPGIFINFO — Retrieve network interface information (line speed, MTU and type). The argument should be a pointer to a struct tapinfo.
• TAPSDEBUG — 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. (iykyk 🔍)
• TAPGDEBUG — The argument should be a pointer to an int; this stores the internal debugging variable's value into it.
• TAPGIFNAME — Retrieve network interface name. The argument should be a pointer to a struct ifreq. The interface name will be returned in the ifr_name field.
• 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 nonblocking).
• 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 frames 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.
• SIOCGIFADDR — Retrieve the Media Access Control (MAC) address of the "remote" side. This command is used by the VMware port and expected to be executed on descriptor, associated with control device (usually /dev/vmnetN or /dev/tapN). The buffer, which is passed as the argument, is expected to have enough space to store the MAC address. At the open time both "local" and "remote" MAC addresses are the same, so this command could be used to retrieve the "local" MAC address.
• SIOCSIFADDR — Set the Media Access Control (MAC) address of the "remote" side. This command is used by VMware port and expected to be executed on a descriptor, associated with control device (usually /dev/vmnetN).

---

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, the interface is brought down (as if with "ifconfig tapN down") and has all of its configured addresses deleted — unless the device is a VMnet device, or has IFF_LINK0 flag set. All queued frames are thrown away. And if the interface is up when the data device is not open, output frames are thrown away rather than letting them pile up. Not me being dramatic, that's just what it does 😭

The tap device can also be used with the VMware port as a replacement for the old VMnet device driver. VMnet devices do not ifconfig(8) themselves down when the control device is closed. Everything else is the same.

In addition to the above mentioned ioctl(2) calls, there is an additional one for the VMware port.

• VMIO_SIOCSIFFLAGS — VMware SIOCSIFFLAGS.

---

🌈 SEE ALSO 🌈

inet(4), intro(4), tun(4)

---

and that's the whole page, bestie — you ate that 💅✨

this lil masterpiece came to you from the good people at Debian, dated January 13, 2020 — go forth and tunnel those frames, you absolute main character 🫶👑