The kill command sends a signal to a process. This can terminate a process (the default action), interrupt it, suspend it, crash it, and so on. You must own the process, or be the superuser, to affect it. To terminate process 13243, for example, run:

$ kill 13243

If this does not work—some programs catch this signal without terminating—add the -KILL or (equivalently) -9 option:

$ kill -KILL 13243

which is virtually guaranteed to work. However, this is not a clean exit for the program, which may leave resources allocated (or cause other inconsistencies) upon its death.

If you don’t know the PID of a process, run ps and examine the output:

$ ps -uax | grep emacs

or even better, try the pidof command, which looks up and prints the PID of a process by its name:

$ pidof emacs
8374

Now you can kill a process knowing only its program name in a single line, using shell backquotes to execute pidof:

$ kill `pidof emacs`

In addition to the kill program in the filesystem (usually /bin/kill), most shells have built-in kill commands, but their syntax and behavior differ. However, they all support the following usage:

$ kill -N PID
$ kill -NAME PID

where N is a signal number, and NAME is a signal name without its leading “SIG” (e.g., use -HUP to send the SIGHUP signal). To see a complete list of signals transmitted by kill, run kill -l, though its output differs depending on which kill you’re running. For descriptions of the signals, run man 7 signal.