(backport) core: remove JoinControllers= configuration setting
authorLennart Poettering <lennart@poettering.net>
Thu, 15 Nov 2018 20:07:43 +0000 (21:07 +0100)
committerŁukasz Stelmach <l.stelmach@samsung.com>
Thu, 25 Jan 2024 12:31:35 +0000 (13:31 +0100)
commitaac02f5aa1ed4cbd371d37b35ce4a4a9080b90af
treee240f2db72d41ea8d59da55f0464750c74377edc
parent639b72a4c65389e7e6e0722e4c5947c5f2d2ec21
(backport) core: remove JoinControllers= configuration setting

This removes the ability to configure which cgroup controllers to mount
together. Instead, we'll now hardcode that "cpu" and "cpuacct" are
mounted together as well as "net_cls" and "net_prio".

The concept of mounting controllers together has no future as it does
not exist to cgroupsv2. Moreover, the current logic is systematically
broken, as revealed by the discussions in #10507. Also, we surveyed Red
Hat customers and couldn't find a single user of the concept (which
isn't particularly surprising, as it is broken...)

This reduced the (already way too complex) cgroup handling for us, since
we now know whenever we make a change to a cgroup for one controller to
which other controllers it applies.

Change-Id: I0eadd5f685980bd006134991310c0688ba7a86c5
man/systemd-system.conf.xml
src/core/main.c
src/core/mount-setup.c
src/core/mount-setup.h
src/core/system.conf.in
src/shared/conf-parser.c
src/shared/conf-parser.h
src/test/test-conf-parser.c
test/fuzz-corpus/unit-file/directives.service