sched: Fix race between task_group and sched_task_group
authorKirill Tkhai <ktkhai@parallels.com>
Mon, 27 Oct 2014 10:18:25 +0000 (14:18 +0400)
committerIngo Molnar <mingo@kernel.org>
Tue, 28 Oct 2014 09:45:59 +0000 (10:45 +0100)
commiteeb61e53ea19be0c4015b00b2e8b3b2185436f2b
tree87ed33506d580130ac59cc231a6e112c4042b96d
parentcac7f2429872d3733dc3f9915857b1691da2eb2f
sched: Fix race between task_group and sched_task_group

The race may happen when somebody is changing task_group of a forking task.
Child's cgroup is the same as parent's after dup_task_struct() (there just
memory copying). Also, cfs_rq and rt_rq are the same as parent's.

But if parent changes its task_group before it's called cgroup_post_fork(),
we do not reflect this situation on child. Child's cfs_rq and rt_rq remain
the same, while child's task_group changes in cgroup_post_fork().

To fix this we introduce fork() method, which calls sched_move_task() directly.
This function changes sched_task_group on appropriate (also its logic has
no problem with freshly created tasks, so we shouldn't introduce something
special; we are able just to use it).

Possibly, this decides the Burke Libbey's problem: https://lkml.org/lkml/2014/10/24/456

Signed-off-by: Kirill Tkhai <ktkhai@parallels.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/1414405105.19914.169.camel@tkhai
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/sched/core.c