Linux – Difference Between pkill and killall

killlinuxprocessprocess-management

I know that pkill has more filtering rules than killall. My question is, what is the difference between:

pkill [signal] name

and 

killall [signal] name

I've read that killall is more effective and kill all processes and subprocesses (and recursively) that match with name program. pkill doesn't do this too?

Best Answer

The pgrep and pkill utilities were introduced in Sun's Solaris 7 and, as g33klord noted, they take a pattern as argument which is matched against the names of running processes. While pgrep merely prints a list of matching processes, pkill will send the specified signal (or SIGTERM by default) to the processes. The common options and semantics between pgrep and pkill comes in handy when you want to be careful and first review the list matching processes with pgrep, then proceed to kill them with pkill. pgrep and pkill are provided by the the procps package, which also provides other /proc file system utilities, such as ps, top, free, uptime among others.

The killall command is provided by the psmisc package, and differs from pkill in that, by default, it matches the argument name exactly (up to the first 15 characters) when determining the processes signals will be sent to. The -e, --exact option can be specified to also require exact matches for names longer than 15 characters. This makes killall somewhat safer to use compared to pkill. If the specified argument contains slash (/) characters, the argument is interpreted as a file name and processes running that particular file will be selected as signal recipients. killall also supports regular expression matching of process names, via the -r, --regexp option.

There are other differences as well. The killall command for instance has options for matching processes by age (-o, --older-than and -y, --younger-than), while pkill can be told to only kill processes on a specific terminal (via the -t option). Clearly then, the two commands have specific niches.

Note that the killall command on systems descendant from Unix System V (notably Sun's Solaris, IBM's AIX and HP's HP-UX) kills all processes killable by a particular user, effectively shutting down the system if run by root.

The Linux psmisc utilities have been ported to BSD (and in extension Mac OS X), hence killall there follows the "kill processes by name" semantics.

Related Question