On Fri, 2010-04-23 at 09:05 +0800, Li Zefan wrote:
>
>
> You are right in that taking task_lock() is sufficient (I forgot
> this lock rule), but it's not true that whatever locks are held
> in the ->attach method can pin a task's cgroup.
Ah, can you be more specific about the ->attach() case?
The way I read it, cgroup_attach_task():
for_each_subsys(root, ss) {
if (ss->attach)
ss->attach(ss, cgrp, oldcgrp, tsk, false);
}
set_bit(CGRP_RELEASABLE, &oldcgrp->flags);
synchronize_rcu();
put_css_set(cg);
So if you hold a lock that any of those ->attach() methods will use, it
will in fact delay the put_css_set().
Ah, indeed I see your point, it doesn't indeed pin the task to the
cgroup, but does avoid the cgroup from being freed.
Hrmm,.. so anything wanting to really pin a task to its cgroup will have
to use task_lock()? I'll have to see if that works for
sched_setscheduler().
_______________________________________________
Containers mailing list
Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx
https://lists.linux-foundation.org/mailman/listinfo/containers
