Keyboard input

From ArchWiki

Prerequisite for modifying the key mapping is knowing how a key press results in a symbol:

  1. The keyboard sends a scancode to the computer.
  2. The Linux kernel maps the scancode to a keycode; see Map scancodes to keycodes.
  3. The keyboard layout maps the keycode to a symbol or keysym, depending on what modifier keys are pressed.

Most of your keys should already have a keycode, or at least a scancode. Keys without a scancode are not recognized by the kernel; these can include additional keys from "gaming" keyboards, etc.

In Xorg, some keysyms (e.g. XF86AudioPlay, XF86AudioRaiseVolume etc.) can be mapped to actions (i.e. launching an external application). See Keyboard shortcuts#Xorg for details.

In Linux console, some keysyms (e.g. F1 to F246) can be mapped to certain actions (e.g. switch to other console or print some sequence of characters). See Console keyboard configuration#Creating a custom keymap for details.

Identifying scancodes

Using evtest

The most reliable way to obtain a scancode is to reference the MSC_SCAN evdev event produced when the key is pressed [1]. There are multiple evdev API testers, but the most straightforward is evtest(1) from the evtest package:

# evtest /dev/input/event12
...
Event: time 1434666536.001123, type 4 (EV_MSC), code 4 (MSC_SCAN), value 70053
Event: time 1434666536.001123, type 1 (EV_KEY), code 69 (KEY_NUMLOCK), value 0
Event: time 1434666536.001123, -------------- EV_SYN ------------
Tip: If you do not know which event number has device of your interest, you can run evtest without parameters and it will show you list of devices with their event numbers, then you can enter needed number.

Use the "value" field of MSC_SCAN. This example shows that NumLock has scancode 70053 and keycode 69.

Using showkey

Note: The manual page remarks that the raw scancode mode "is not very raw at all" on modern kernels. It appears that this method produces incorrect results for USB devices in particular.

The traditional way to get a scancode is to use the showkey(1) utility. showkey waits for a key to be pressed, or exits if no keys are pressed within 10 seconds. For showkey to work you need to be in a virtual console, not in a graphical environment or logged in via a network connection. Run the following command:

# showkey --scancodes

and try to push keyboard keys; you should see scancodes being printed to the output.

Using dmesg

Note: This method only identifies unknown keys.

You can get the scancode of a key by pressing the desired key and looking at the output of dmesg. For example, if you get:

Unknown key pressed (translated set 2, code 0xa0 on isa0060/serio0

then the scancode you need is 0xa0.

Identifying keycodes

The Linux keycodes are defined in /usr/include/linux/input-event-codes.h (see the KEY_ variables).

Identifying keycodes in console

The keycodes for virtual console are reported by the showkey(1) utility. showkey waits for a key to be pressed and if none are, in a span of 10 seconds, it quits. To execute showkey, you need to be in a virtual console, not in a graphical environment. Run the following command:

# showkey --keycodes

and try to push keyboard keys; you should see keycodes being printed to the output.

Identifying keycodes in Xorg

This article or section needs expansion.

Reason: xev also reports keysyms. Mention that you need to focus the "Event Tester" window. (Discuss in Talk:Keyboard input)
Note: The Xorg keycodes are incremented by 8 compared to the Linux keycodes.[2]

The keycodes used by Xorg are reported by a utility called xev(1), which is provided by the xorg-xev package. Of course to execute xev, you need to be in a graphical environment, not in the console.

With the following command you can start xev and show only the relevant parts:

$ xev | awk -F'[ )]+' '/^KeyPress/ { a[NR+2] } NR in a { printf "%-3s %s\n", $5, $8 }'
38  a
55  v
54  c
50  Shift_L
133 Super_L
135 Menu

Xbindkeys is another wrapper to xev that reports keycodes.

If you press a key and nothing appears in the terminal, it means that either the key does not have a scancode, the scancode is not mapped to a keycode, or some other process is capturing the keypress. If you suspect that a process listening to X server is capturing the keypress, you can try running xev from a clean X session:

$ xinit /usr/bin/xterm -- :1

Identifying keycodes in Wayland

Although xev works through xwayland, you can also use wev to access keycodes under pure Wayland.

For example, this command lets you retrieve only key names and their UTF-8 equivalent:

$ wev | grep 'sym'

Tips and tricks

Configuration of VIA compatible keyboards

VIA is a program to remap keys directly into compatible keyboards. In case you have one of those, in order for the keyboard to be picked up by the browser and configure it online, you need to add a custom udev rule changing the permissions of devices accessed through the hidraw driver.

Note: This method might pose a security risk.

Create the following udev rule:

/etc/udev/rules.d/99-via.rules
KERNEL=="hidraw*", SUBSYSTEM=="hidraw", MODE="0666", TAG+="uaccess", TAG+="udev-acl"

In order for this to take effect you need to reload udev with:

# udevadm control --reload

See also

  • kbd-project - official website of the showkeys utility
  • interception-tools - a set of utilities to control and customize the behavior of keyboard input mappings
  • kmonad - an advanced key rebinding and remapping daemon
  • Hawck - another key rebinding daemon
  • keyd - simplistic key rebinding daemon
  • Vial - VIA standalone program