Re: [PATCH] sched: remove rcu_read_lock from wake_affine

From: Paul E. McKenney
Date: Tue Jun 07 2011 - 14:11:41 EST


On Tue, Jun 07, 2011 at 07:29:23PM +0200, Peter Zijlstra wrote:
> On Tue, 2011-06-07 at 10:26 -0700, Paul E. McKenney wrote:
> >
> > Nikunj, one such approach is is "WARN_ON_ONCE(!rcu_read_lock_held())".
> >
> > This will complain if this function is called without an rcu_read_lock()
> > in effect, but only if CONFIG_PROVE_RCU=y.
>
> rcu_lockdep_assert(rcu_read_lock_held()) would be nicer,

Good point!

> however, since
> the below:
>
> > > > static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync)
> > > > {
> > > > s64 this_load, load;
> > > > @@ -1481,7 +1482,6 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync)
> > > > * effect of the currently running task from the load
> > > > * of the current CPU:
> > > > */
> > > > - rcu_read_lock();
> > > > if (sync) {
> > > > tg = task_group(current);
> > > > weight = current->se.load.weight;
>
> task_group() has an rcu_dereference_check() in, its really not needed,
> the thing will yell if we get this wrong.

Fair enough! The main reason for adding it at this level as well is
to prevent people from "fixing" splats by adding rcu_read_lock() and
rcu_read_unlock() at this level. But you would see any such patch, so
such a "fix" would not get far. ;-)

Thanx, Paul
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/