Re: [RFC] rcu: Warn that rcu ktheads cannot be spawned

[Paul E. McKenney]

On Tue, Jun 25, 2019 at 11:41:00AM +0900, Byungchul Park wrote:
> On Mon, Jun 24, 2019 at 10:25:51AM -0700, Paul E. McKenney wrote:
> > On Mon, Jun 24, 2019 at 12:46:24PM -0400, Joel Fernandes wrote:
> > > On Mon, Jun 24, 2019 at 05:27:32PM +0900, Byungchul Park wrote:
> > > > Hello rcu folks,
> > > > 
> > > > I thought it'd better to announce it if those spawnings fail because of
> > > > !rcu_scheduler_fully_active.
> > > > 
> > > > Of course, with the current code, it never happens though.
> > > > 
> > > > Thoughts?
> > > 
> > > It seems in the right spirit, but with your patch a warning always fires.
> > > rcu_prepare_cpu() is called multiple times, once from rcu_init() and then
> > > from hotplug paths.
> > > 
> > > Warning splat stack looks like:
> > > 
> > > [    0.398767] Call Trace:
> > > [    0.398775]  rcu_init+0x6aa/0x724
> > > [    0.398779]  start_kernel+0x220/0x4a2
> > > [    0.398780]  ? copy_bootdata+0x12/0xac
> > > [    0.398782]  secondary_startup_64+0xa4/0xb0
> > 
> > Thank you both, and I will remove this from my testing queue.
> > 
> > As Joel says, this is called at various points in the boot sequence, not
> > all of which are far enough along to support spawning kthreads.
> > 
> > The real question here is "What types of bugs are we trying to defend
> > against?"  But keeping in mind existing diagnostics.  For example, are
> > there any kthreads for which a persistent failure to spawn would not
> > emit any error message.  My belief is that any such persistent failure
> > would result in either an in-kernel diagnostic or an rcutorture failure,
> > but I might well be missing something.
> > 
> > Thoughts?  Or, more to the point, tests demonstrating silence in face
> > of such a persistent failure?
> 
> You are right. There wouldn't be a persistent failure because the path
> turning cpus on always tries to spawn them, *even* in case that the
> booting sequence is wrong. The current code anyway goes right though.
> 
> I thought a hole can be there if the code changes so that those kthreads
> cannot be spawned until the cpu being up, which is the case I was
> interested in. Again, it's gonna never happen with the current code
> because it spawns them after setting rcu_scheduler_fully_active to 1 in
> rcu_spawn_gp_kthead().
> 
> And I wrongly thought you placed the rcu_scheduler_fully_active check on
> spawning just in case. But it seems to be not the case.
> 
> So I'd better stop working on the warning patch. :) Instead, please
> check the following trivial fix.
> 
> Thanks,
> Byungchul
> 
> ---8<---
> >From 1293d19bb7abf7553d656c81182118eff54e7dc9 Mon Sep 17 00:00:00 2001
> From: Byungchul Park <byungchul.park@xxxxxxx>
> Date: Mon, 24 Jun 2019 16:22:11 +0900
> Subject: [PATCH] rcu: Make rcu_spawn_one_boost_kthread() return void
> 
> The return value of rcu_spawn_one_boost_kthread() is not used any
> longer. Change the return type from int to void.
> 
> Signed-off-by: Byungchul Park <byungchul.park@xxxxxxx>

Looks pretty good, just one comment below.  Plus could you please tell
me what you are developing this against?

							Thanx, Paul

> ---
>  kernel/rcu/tree_plugin.h | 15 +++++++--------
>  1 file changed, 7 insertions(+), 8 deletions(-)
> 
> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index 1102765..4e11aa4 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -1131,7 +1131,7 @@ static void rcu_preempt_boost_start_gp(struct rcu_node *rnp)
>   * already exist.  We only create this kthread for preemptible RCU.
>   * Returns zero if all is well, a negated errno otherwise.
>   */
> -static int rcu_spawn_one_boost_kthread(struct rcu_node *rnp)
> +static void rcu_spawn_one_boost_kthread(struct rcu_node *rnp)
>  {
>  	int rnp_index = rnp - rcu_get_root();
>  	unsigned long flags;
> @@ -1139,25 +1139,24 @@ static int rcu_spawn_one_boost_kthread(struct rcu_node *rnp)
>  	struct task_struct *t;
>  
>  	if (!IS_ENABLED(CONFIG_PREEMPT_RCU))
> -		return 0;
> +		return;
>  
>  	if (!rcu_scheduler_fully_active || rcu_rnp_online_cpus(rnp) == 0)
> -		return 0;
> +		return;
>  
>  	rcu_state.boost = 1;
>  	if (rnp->boost_kthread_task != NULL)
> -		return 0;
> +		return;
>  	t = kthread_create(rcu_boost_kthread, (void *)rnp,
>  			   "rcub/%d", rnp_index);
>  	if (IS_ERR(t))

This would be a change in behavior, but it might be good to have a
WARN_ON_ONCE() above.  Assuming that it doesn't splat on every boot.  ;-)

> -		return PTR_ERR(t);
> +		return;
>  	raw_spin_lock_irqsave_rcu_node(rnp, flags);
>  	rnp->boost_kthread_task = t;
>  	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
>  	sp.sched_priority = kthread_prio;
>  	sched_setscheduler_nocheck(t, SCHED_FIFO, &sp);
>  	wake_up_process(t); /* get to TASK_INTERRUPTIBLE quickly. */
> -	return 0;
>  }
>  
>  static void rcu_cpu_kthread_setup(unsigned int cpu)
> @@ -1265,7 +1264,7 @@ static void __init rcu_spawn_boost_kthreads(void)
>  	if (WARN_ONCE(smpboot_register_percpu_thread(&rcu_cpu_thread_spec), "%s: Could not start rcub kthread, OOM is now expected behavior\n", __func__))
>  		return;
>  	rcu_for_each_leaf_node(rnp)
> -		(void)rcu_spawn_one_boost_kthread(rnp);
> +		rcu_spawn_one_boost_kthread(rnp);
>  }
>  
>  static void rcu_prepare_kthreads(int cpu)
> @@ -1275,7 +1274,7 @@ static void rcu_prepare_kthreads(int cpu)
>  
>  	/* Fire up the incoming CPU's kthread and leaf rcu_node kthread. */
>  	if (rcu_scheduler_fully_active)
> -		(void)rcu_spawn_one_boost_kthread(rnp);
> +		rcu_spawn_one_boost_kthread(rnp);
>  }
>  
>  #else /* #ifdef CONFIG_RCU_BOOST */
> -- 
> 1.9.1
>