Re: [PATCH 03/11] rcu: Add mutex for rcu boost kthread spawning and affinity setting

[Paul E. McKenney]

On Thu, Dec 09, 2021 at 03:09:30PM +0000, David Woodhouse wrote:
> From: David Woodhouse <dwmw@xxxxxxxxxxxx>
> 
> As we handle parallel CPU bringup, we will need to take care to avoid
> spawning multiple boost threads, or race conditions when setting their
> affinity. Spotted by Paul McKenney.

And again, if testing goes well and you don't get it there first, I
expect to push this during the v5.18 merge window.  In case you
would like to push this with the rest of this series:

Acked-by: Paul E. McKenney <paulmck@xxxxxxxxxx>

> Signed-off-by: David Woodhouse <dwmw@xxxxxxxxxxxx>
> ---
>  kernel/rcu/tree.c        |  1 +
>  kernel/rcu/tree.h        |  3 +++
>  kernel/rcu/tree_plugin.h | 10 ++++++++--
>  3 files changed, 12 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index a1bb0b1229ed..809855474b39 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -4527,6 +4527,7 @@ static void __init rcu_init_one(void)
>  			init_waitqueue_head(&rnp->exp_wq[2]);
>  			init_waitqueue_head(&rnp->exp_wq[3]);
>  			spin_lock_init(&rnp->exp_lock);
> +			mutex_init(&rnp->boost_kthread_mutex);
>  		}
>  	}
>  
> diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
> index aff4cc9303fb..055e30b3e5e0 100644
> --- a/kernel/rcu/tree.h
> +++ b/kernel/rcu/tree.h
> @@ -108,6 +108,9 @@ struct rcu_node {
>  				/*  side effect, not as a lock. */
>  	unsigned long boost_time;
>  				/* When to start boosting (jiffies). */
> +	struct mutex boost_kthread_mutex;
> +				/* Exclusion for thread spawning and affinity */
> +				/*  manipulation. */
>  	struct task_struct *boost_kthread_task;
>  				/* kthread that takes care of priority */
>  				/*  boosting for this rcu_node structure. */
> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index 5199559fbbf0..3b4ee0933710 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -1162,15 +1162,16 @@ static void rcu_spawn_one_boost_kthread(struct rcu_node *rnp)
>  	struct sched_param sp;
>  	struct task_struct *t;
>  
> +	mutex_lock(&rnp->boost_kthread_mutex);
>  	if (rnp->boost_kthread_task || !rcu_scheduler_fully_active)
> -		return;
> +		goto out;
>  
>  	rcu_state.boost = 1;
>  
>  	t = kthread_create(rcu_boost_kthread, (void *)rnp,
>  			   "rcub/%d", rnp_index);
>  	if (WARN_ON_ONCE(IS_ERR(t)))
> -		return;
> +		goto out;
>  
>  	raw_spin_lock_irqsave_rcu_node(rnp, flags);
>  	rnp->boost_kthread_task = t;
> @@ -1178,6 +1179,9 @@ static void rcu_spawn_one_boost_kthread(struct rcu_node *rnp)
>  	sp.sched_priority = kthread_prio;
>  	sched_setscheduler_nocheck(t, SCHED_FIFO, &sp);
>  	wake_up_process(t); /* get to TASK_INTERRUPTIBLE quickly. */
> +
> + out:
> +	mutex_unlock(&rnp->boost_kthread_mutex);
>  }
>  
>  /*
> @@ -1200,6 +1204,7 @@ static void rcu_boost_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu)
>  		return;
>  	if (!zalloc_cpumask_var(&cm, GFP_KERNEL))
>  		return;
> +	mutex_lock(&rnp->boost_kthread_mutex);
>  	for_each_leaf_node_possible_cpu(rnp, cpu)
>  		if ((mask & leaf_node_cpu_bit(rnp, cpu)) &&
>  		    cpu != outgoingcpu)
> @@ -1207,6 +1212,7 @@ static void rcu_boost_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu)
>  	if (cpumask_weight(cm) == 0)
>  		cpumask_setall(cm);
>  	set_cpus_allowed_ptr(t, cm);
> +	mutex_unlock(&rnp->boost_kthread_mutex);
>  	free_cpumask_var(cm);
>  }
>  
> -- 
> 2.31.1
>