Re: [PATCH] perf: perf_event_exit_task_context: s/rcu_dereference/rcu_dereference_raw/

[Oleg Nesterov <oleg@redhat.com>]

On 01/21, Paul E. McKenney wrote:
>
> On Fri, Jan 21, 2011 at 06:53:45PM +0100, Oleg Nesterov wrote:
> >
> > But rcu_dereference_raw() looks a bit confusing, and it is not
> > very convenient to use read_barrier_depends() directly.
> >
> > Paul, may be it makes sense to add the new trivial helper which
> > can be used instead?
> >
> > Yes, this is only cosmetic issue, I know ;)
>
> Cosmetic issues can be pretty important to the poor guy trying to read
> the code.  ;-)

Agreed!

> What keeps the structure that rcu_dereference_raw() returns a pointer
> to from going away?

It can't go away, current owns its ->perf_event_ctxp[] pointers. But
the pointer can be installed at any time by sys_perf_event_open().

Currently the code does

	ctx = current->perf_event_ctxp[ctxn];
	if (ctx)
		do_something(ctx);

and in theory we need smp_read_barrier_depends() in between.

> Best would be if a lockdep condition could be
> constructed from the answer to this question and added to the appropriate
> rcu_dereference() primitive.

In this case the condition is "true", so we can use rcu_dereference_raw().
The only problem, it looks confusing. Especially because you actually
need rcu_read_lock() if you look at not_current_task->perf_event_ctxp[].

Oleg.