]> git.karo-electronics.de Git - karo-tx-linux.git/commit
cgroup: cgroup_subsys->fork() should be called after the task is added to css_set
authorTejun Heo <tj@kernel.org>
Tue, 16 Oct 2012 22:03:14 +0000 (15:03 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 11 Jan 2013 17:18:49 +0000 (09:18 -0800)
commit838a4f8bb1a039ee4ea6de544624d8fd620bf455
treef53162dab9f09f14f4ed3fbce4da096875dcd8d6
parent27f0a83c49a639749f9e26c5af4b70164922f304
cgroup: cgroup_subsys->fork() should be called after the task is added to css_set

commit 5edee61edeaaebafe584f8fb7074c1ef4658596b upstream.

cgroup core has a bug which violates a basic rule about event
notifications - when a new entity needs to be added, you add that to
the notification list first and then make the new entity conform to
the current state.  If done in the reverse order, an event happening
inbetween will be lost.

cgroup_subsys->fork() is invoked way before the new task is added to
the css_set.  Currently, cgroup_freezer is the only user of ->fork()
and uses it to make new tasks conform to the current state of the
freezer.  If FROZEN state is requested while fork is in progress
between cgroup_fork_callbacks() and cgroup_post_fork(), the child
could escape freezing - the cgroup isn't frozen when ->fork() is
called and the freezer couldn't see the new task on the css_set.

This patch moves cgroup_subsys->fork() invocation to
cgroup_post_fork() after the new task is added to the css_set.
cgroup_fork_callbacks() is removed.

Because now a task may be migrated during cgroup_subsys->fork(),
freezer_fork() is updated so that it adheres to the usual RCU locking
and the rather pointless comment on why locking can be different there
is removed (if it doesn't make anything simpler, why even bother?).

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
include/linux/cgroup.h
kernel/cgroup.c
kernel/cgroup_freezer.c
kernel/fork.c