Re: [PATCH 02/11] sched, rcu: Make RCU use resched_cpu()

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Mon, Sep 23, 2013 at 09:55:40AM -0700, Paul E. McKenney wrote:
> On Tue, Sep 17, 2013 at 04:40:12PM +0200, Peter Zijlstra wrote:
> > Thomas Meyer reported a UP build fail, should be fixed.
> > 
> > ---
> > Subject: sched, rcu: Make RCU use resched_cpu()
> > From: Peter Zijlstra <peterz@infradead.org>
> > Date: Tue Sep 17 09:30:55 CEST 2013
> > 
> > We're going to deprecate and remove set_need_resched() for it will do
> > the wrong thing. Make an exception for RCU and allow it to use
> > resched_cpu() which will do the right thing.
> > 
> > Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
> > Signed-off-by: Peter Zijlstra <peterz@infradead.org>
> 
> Queued for 3.13, thank you!

Passes light testing, but I am confused about why the change to
resched_task() when only resched_cpu() is invoked elsewhere in the patch.
Enlightenment?

							Thanx, Paul

> > ---
> >  kernel/rcutree.c    |   15 ++++++++++++++-
> >  kernel/sched/core.c |   10 ++--------
> >  2 files changed, 16 insertions(+), 9 deletions(-)
> > 
> > --- a/kernel/rcutree.c
> > +++ b/kernel/rcutree.c
> > @@ -898,6 +898,12 @@ static void print_other_cpu_stall(struct
> >  	force_quiescent_state(rsp);  /* Kick them all. */
> >  }
> > 
> > +/*
> > + * This function really isn't for public consumption, but RCU is special in
> > + * that context switches can allow the state machine to make progress.
> > + */
> > +extern void resched_cpu(int cpu);
> > +
> >  static void print_cpu_stall(struct rcu_state *rsp)
> >  {
> >  	int cpu;
> > @@ -927,7 +933,14 @@ static void print_cpu_stall(struct rcu_s
> >  				     3 * rcu_jiffies_till_stall_check() + 3;
> >  	raw_spin_unlock_irqrestore(&rnp->lock, flags);
> > 
> > -	set_need_resched();  /* kick ourselves to get things going. */
> > +	/*
> > +	 * Attempt to revive the RCU machinery by forcing a context switch.
> > +	 *
> > +	 * A context switch would normally allow the RCU state machine to make
> > +	 * progress and it could be we're stuck in kernel space without context
> > +	 * switches for an entirely unreasonable amount of time.
> > +	 */
> > +	resched_cpu(smp_processor_id());
> >  }
> > 
> >  static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -513,12 +513,11 @@ static inline void init_hrtick(void)
> >   * might also involve a cross-CPU call to trigger the scheduler on
> >   * the target CPU.
> >   */
> > -#ifdef CONFIG_SMP
> >  void resched_task(struct task_struct *p)
> >  {
> >  	int cpu;
> > 
> > -	assert_raw_spin_locked(&task_rq(p)->lock);
> > +	lockdep_assert_held(&task_rq(p)->lock);
> > 
> >  	if (test_tsk_need_resched(p))
> >  		return;
> > @@ -546,6 +545,7 @@ void resched_cpu(int cpu)
> >  	raw_spin_unlock_irqrestore(&rq->lock, flags);
> >  }
> > 
> > +#ifdef CONFIG_SMP
> >  #ifdef CONFIG_NO_HZ_COMMON
> >  /*
> >   * In the semi idle case, use the nearest busy cpu for migrating timers
> > @@ -693,12 +693,6 @@ void sched_avg_update(struct rq *rq)
> >  	}
> >  }
> > 
> > -#else /* !CONFIG_SMP */
> > -void resched_task(struct task_struct *p)
> > -{
> > -	assert_raw_spin_locked(&task_rq(p)->lock);
> > -	set_tsk_need_resched(p);
> > -}
> >  #endif /* CONFIG_SMP */
> > 
> >  #if defined(CONFIG_RT_GROUP_SCHED) || (defined(CONFIG_FAIR_GROUP_SCHED) && \
> > 
> >