]> git.ipfire.org Git - thirdparty/systemd.git/commit
cgroup: remove logic for maintaining /control subcgroup for the service unit type
authorLennart Poettering <lennart@poettering.net>
Fri, 24 Nov 2017 19:11:27 +0000 (20:11 +0100)
committerLennart Poettering <lennart@poettering.net>
Sat, 25 Nov 2017 16:08:21 +0000 (17:08 +0100)
commite9a4f67609cfabfeccbda8565bc4daf5b9e45c54
tree550c944062afe58e0aa9ffb3e911a3f8778bbd0c
parent60c728adf7e716ca23e2d300d3955c60265a28c6
cgroup: remove logic for maintaining /control subcgroup for the service unit type

Previously, in the service unit type we ran all control processes in a
special subcgroup /control of the unit's main cgroup. Remove that, and
run the control program in the main cgroup instead.

The concept conflicts with cgroupv2's logic of "no processes in inner
nodes": if a unit has a main daemon process running in the main cgroup,
and a reload control process would be started in the /control subcgroup,
then this would necessarily fail, as the main daemon process would
become an inner node process that way.

We could in theory continue to support this in cgroupv1, but in the
interest in keeping behaviour similar in both hierarchies, let's drop
this altogether.

Philosophically maybe it wasn't the greatest idea anyway to just go
berserk and SIGKILL all those processes — loud warning logging might
have sufficed, too.
src/core/service.c