Re: [PATCH 2/3] srcu: queue work without holding the lock

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

On Fri, Sep 22, 2017 at 05:28:05PM +0200, Sebastian Andrzej Siewior wrote:
> On RT we can't invoke queue_delayed_work() within an atomic section
> (which is provided by raw_spin_lock_irqsave()).
> srcu_reschedule() invokes queue_delayed_work() outside of the
> raw_spin_lock_irq_rcu_node() section so this should be fine here, too.
> If the remaining callers of call_srcu() aren't atomic
> (spin_lock_irqsave() is fine) then this should work on RT, too.

Just to make sure I understand...   The problem is not the _irqsave,
but rather the raw_?

							Thanx, Paul

> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> ---
>  kernel/rcu/srcutree.c | 11 ++++++++---
>  1 file changed, 8 insertions(+), 3 deletions(-)
> 
> diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> index d190af0e56f8..3ee4ef40f23e 100644
> --- a/kernel/rcu/srcutree.c
> +++ b/kernel/rcu/srcutree.c
> @@ -648,12 +648,17 @@ static void srcu_funnel_gp_start(struct srcu_struct *sp, struct srcu_data *sdp,
>  	/* If grace period not already done and none in progress, start it. */
>  	if (!rcu_seq_done(&sp->srcu_gp_seq, s) &&
>  	    rcu_seq_state(sp->srcu_gp_seq) == SRCU_STATE_IDLE) {
> +		unsigned long delay;
> +
>  		WARN_ON_ONCE(ULONG_CMP_GE(sp->srcu_gp_seq, sp->srcu_gp_seq_needed));
>  		srcu_gp_start(sp);
> +		delay = srcu_get_delay(sp);
> +		raw_spin_unlock_irqrestore_rcu_node(sp, flags);
> +
>  		queue_delayed_work(system_power_efficient_wq, &sp->work,
> -				   srcu_get_delay(sp));
> -	}
> -	raw_spin_unlock_irqrestore_rcu_node(sp, flags);
> +				   delay);
> +	} else
> +		raw_spin_unlock_irqrestore_rcu_node(sp, flags);
>  }
>  
>  /*
> -- 
> 2.14.1
>