Re: [PATCH RT v2 2/3] sched: migrate_enable: Use sleeping_lock to indicate involuntary sleep
From: Paul E. McKenney
Date: Wed Aug 28 2019 - 08:54:42 EST
On Wed, Aug 28, 2019 at 11:27:39AM +0200, Sebastian Andrzej Siewior wrote:
> On 2019-08-27 08:53:06 [-0700], Paul E. McKenney wrote:
> > > > On the other hand, within a PREEMPT=n kernel, the call to schedule()
> > > > would split even an rcu_read_lock() critical section. Which is why I
> > > > asked earlier if sleeping_lock_inc() and sleeping_lock_dec() are no-ops
> > > > in !PREEMPT_RT_BASE kernels. We would after all want the usual lockdep
> > > > complaints in that case.
> > >
> > > sleeping_lock_inc() +dec() is only RT specific. It is part of RT's
> > > spin_lock() implementation and used by RCU (rcu_note_context_switch())
> > > to not complain if invoked within a critical section.
> > Then this is being called when we have something like this, correct?
> > DEFINE_SPINLOCK(mylock); // As opposed to DEFINE_RAW_SPINLOCK().
> > ...
> > rcu_read_lock();
> > do_something();
> > spin_lock(&mylock); // Can block in -rt, thus needs sleeping_lock_inc()
> > ...
> > rcu_read_unlock();
> > Without sleeping_lock_inc(), lockdep would complain about a voluntary
> > schedule within an RCU read-side critical section. But in -rt, voluntary
> > schedules due to sleeping on a "spinlock" are OK.
> > Am I understanding this correctly?
> Everything perfect except that it is not lockdep complaining but the
> WARN_ON_ONCE() in rcu_note_context_switch().
This one, right?
WARN_ON_ONCE(!preempt && t->rcu_read_lock_nesting > 0);
Another approach would be to change that WARN_ON_ONCE(). This fix might
be too extreme, as it would suppress other issues:
WARN_ON_ONCE(IS_ENABLED(CONFIG_PREEMPT_RT_BASE) && !preempt && t->rcu_read_lock_nesting > 0);
But maybe what is happening under the covers is that preempt is being
set when sleeping on a spinlock. Is that the case?