Date:   Wed, 7 Dec 2022 14:14:23 -0800
From:   "Paul E. McKenney" <paulmck@...nel.org>
To:     "Joel Fernandes (Google)" <joel@...lfernandes.org>
Cc:     linux-kernel@...r.kernel.org, Ben Segall <bsegall@...gle.com>,
        Daniel Bristot de Oliveira <bristot@...hat.com>,
        Davidlohr Bueso <dave@...olabs.net>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        Ingo Molnar <mingo@...hat.com>,
        Josh Triplett <josh@...htriplett.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Mel Gorman <mgorman@...e.de>,
        Peter Zijlstra <peterz@...radead.org>,
        Steven Rostedt <rostedt@...dmis.org>,
        Valentin Schneider <vschneid@...hat.com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        kernel-team@...roid.com, John Stultz <jstultz@...gle.com>,
        Joel Fernandes <joelaf@...gle.com>,
        Qais Yousef <qais.yousef@....com>,
        Will Deacon <will@...nel.org>,
        Waiman Long <longman@...hat.com>,
        Boqun Feng <boqun.feng@...il.com>,
        Connor O'Brien <connoro@...gle.com>
Subject: Re: [PATCH RFC 2/3] locktorture: Allow non-rtmutex lock types to be
 boosted

On Wed, Nov 23, 2022 at 01:21:03AM +0000, Joel Fernandes (Google) wrote:
> Currently RT boosting is only done for rtmutex_lock, however with proxy
> execution, we also have the mutex_lock participating in priorities. To
> exercise the testing better, add RT boosting to other lock testing types
> as well, using a new knob (rt_boost).
> 
> Tested with boot parameters:
> locktorture.torture_type=mutex_lock
> locktorture.onoff_interval=1
> locktorture.nwriters_stress=8
> locktorture.stutter=0
> locktorture.rt_boost=1
> locktorture.rt_boost_factor=1
> locktorture.nlocks=3
> 
> For the rtmutex test, rt_boost is always enabled even if disabling is
> requested.
> 
> Signed-off-by: Joel Fernandes (Google) <joel@...lfernandes.org>
> ---
>  kernel/locking/locktorture.c | 91 +++++++++++++++++++-----------------
>  1 file changed, 48 insertions(+), 43 deletions(-)
> 
> diff --git a/kernel/locking/locktorture.c b/kernel/locking/locktorture.c
> index bc3557677eed..5a388ac96a9b 100644
> --- a/kernel/locking/locktorture.c
> +++ b/kernel/locking/locktorture.c
> @@ -46,6 +46,7 @@ torture_param(int, shutdown_secs, 0, "Shutdown time (j), <= zero to disable.");
>  torture_param(int, stat_interval, 60,
>  	     "Number of seconds between stats printk()s");
>  torture_param(int, stutter, 5, "Number of jiffies to run/halt test, 0=disable");
> +torture_param(int, rt_boost, 0, "Perform an rt-boost from the writer, always 1 for rtmutex_lock");
>  torture_param(int, verbose, 1,
>  	     "Enable verbose debugging printk()s");
>  torture_param(int, nlocks, 1,
> @@ -129,15 +130,44 @@ static void torture_lock_busted_write_unlock(int tid __maybe_unused)
>  	  /* BUGGY, do not use in real life!!! */
>  }
>  
> -static void torture_boost_dummy(struct torture_random_state *trsp)

We no longer have torture_boot_dummy().  Is the point that the
"spinlocks" to priority boosting in PREEMPT_RT kernels?  If so,
would it make sense to do something like this for spinlock?

	.task_boost     = IS_ENABLED(CONFIG_PREEMPT_RT) ? torture_rt_boost : torture_boost_dummy,

Or maybe using a similar approach for the default value of the rt_boost
module parameter?

Or is there some benefit of priority boosting for spinlocks even in
non-PREEMPT_RT kernels that I am missing?

> +static void torture_rt_boost(struct torture_random_state *trsp)
>  {
> -	/* Only rtmutexes care about priority */
> +	const unsigned int factor = 50000; /* yes, quite arbitrary */

OK, this one looks like code movement combined with 50000 being named
"factor".  Whoever originally wrote these comments needs to have done
a better job.  ;-)

> +
> +	if (!rt_boost)
> +		return;
> +
> +	if (!rt_task(current)) {
> +		/*
> +		 * Boost priority once every ~50k operations. When the
> +		 * task tries to take the lock, the rtmutex it will account
> +		 * for the new priority, and do any corresponding pi-dance.
> +		 */
> +		if (trsp && !(torture_random(trsp) %
> +			      (cxt.nrealwriters_stress * factor))) {
> +			sched_set_fifo(current);
> +		} else /* common case, do nothing */
> +			return;
> +	} else {
> +		/*
> +		 * The task will remain boosted for another ~500k operations,
> +		 * then restored back to its original prio, and so forth.
> +		 *
> +		 * When @trsp is nil, we want to force-reset the task for
> +		 * stopping the kthread.
> +		 */
> +		if (!trsp || !(torture_random(trsp) %
> +			       (cxt.nrealwriters_stress * factor * 2))) {
> +			sched_set_normal(current, 0);
> +		} else /* common case, do nothing */
> +			return;
> +	}
>  }
>  
>  static struct lock_torture_ops lock_busted_ops = {
>  	.writelock	= torture_lock_busted_write_lock,
>  	.write_delay	= torture_lock_busted_write_delay,
> -	.task_boost     = torture_boost_dummy,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_lock_busted_write_unlock,
>  	.readlock       = NULL,
>  	.read_delay     = NULL,
> @@ -181,7 +211,7 @@ __releases(torture_spinlock)
>  static struct lock_torture_ops spin_lock_ops = {
>  	.writelock	= torture_spin_lock_write_lock,
>  	.write_delay	= torture_spin_lock_write_delay,
> -	.task_boost     = torture_boost_dummy,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_spin_lock_write_unlock,
>  	.readlock       = NULL,
>  	.read_delay     = NULL,
> @@ -208,7 +238,7 @@ __releases(torture_spinlock)
>  static struct lock_torture_ops spin_lock_irq_ops = {
>  	.writelock	= torture_spin_lock_write_lock_irq,
>  	.write_delay	= torture_spin_lock_write_delay,
> -	.task_boost     = torture_boost_dummy,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_lock_spin_write_unlock_irq,
>  	.readlock       = NULL,
>  	.read_delay     = NULL,
> @@ -277,7 +307,7 @@ __releases(torture_rwlock)
>  static struct lock_torture_ops rw_lock_ops = {
>  	.writelock	= torture_rwlock_write_lock,
>  	.write_delay	= torture_rwlock_write_delay,
> -	.task_boost     = torture_boost_dummy,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_rwlock_write_unlock,
>  	.readlock       = torture_rwlock_read_lock,
>  	.read_delay     = torture_rwlock_read_delay,
> @@ -320,7 +350,7 @@ __releases(torture_rwlock)
>  static struct lock_torture_ops rw_lock_irq_ops = {
>  	.writelock	= torture_rwlock_write_lock_irq,
>  	.write_delay	= torture_rwlock_write_delay,
> -	.task_boost     = torture_boost_dummy,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_rwlock_write_unlock_irq,
>  	.readlock       = torture_rwlock_read_lock_irq,
>  	.read_delay     = torture_rwlock_read_delay,
> @@ -362,7 +392,7 @@ __releases(torture_mutex)
>  static struct lock_torture_ops mutex_lock_ops = {
>  	.writelock	= torture_mutex_lock,
>  	.write_delay	= torture_mutex_delay,
> -	.task_boost     = torture_boost_dummy,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_mutex_unlock,
>  	.readlock       = NULL,
>  	.read_delay     = NULL,
> @@ -460,7 +490,7 @@ static struct lock_torture_ops ww_mutex_lock_ops = {
>  	.exit		= torture_ww_mutex_exit,
>  	.writelock	= torture_ww_mutex_lock,
>  	.write_delay	= torture_mutex_delay,
> -	.task_boost     = torture_boost_dummy,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_ww_mutex_unlock,
>  	.readlock       = NULL,
>  	.read_delay     = NULL,
> @@ -471,6 +501,11 @@ static struct lock_torture_ops ww_mutex_lock_ops = {
>  #ifdef CONFIG_RT_MUTEXES
>  static DEFINE_RT_MUTEX(torture_rtmutex);
>  
> +static void torture_rtmutex_init(void)
> +{
> +	rt_boost = 1;
> +}
> +
>  static int torture_rtmutex_lock(int tid __maybe_unused)
>  __acquires(torture_rtmutex)
>  {
> @@ -478,37 +513,6 @@ __acquires(torture_rtmutex)
>  	return 0;
>  }
>  
> -static void torture_rtmutex_boost(struct torture_random_state *trsp)
> -{
> -	const unsigned int factor = 50000; /* yes, quite arbitrary */
> -
> -	if (!rt_task(current)) {
> -		/*
> -		 * Boost priority once every ~50k operations. When the
> -		 * task tries to take the lock, the rtmutex it will account
> -		 * for the new priority, and do any corresponding pi-dance.
> -		 */
> -		if (trsp && !(torture_random(trsp) %
> -			      (cxt.nrealwriters_stress * factor))) {
> -			sched_set_fifo(current);
> -		} else /* common case, do nothing */
> -			return;
> -	} else {
> -		/*
> -		 * The task will remain boosted for another ~500k operations,
> -		 * then restored back to its original prio, and so forth.
> -		 *
> -		 * When @trsp is nil, we want to force-reset the task for
> -		 * stopping the kthread.
> -		 */
> -		if (!trsp || !(torture_random(trsp) %
> -			       (cxt.nrealwriters_stress * factor * 2))) {
> -			sched_set_normal(current, 0);
> -		} else /* common case, do nothing */
> -			return;
> -	}
> -}
> -
>  static void torture_rtmutex_delay(struct torture_random_state *trsp)
>  {
>  	const unsigned long shortdelay_us = 2;
> @@ -535,9 +539,10 @@ __releases(torture_rtmutex)
>  }
>  
>  static struct lock_torture_ops rtmutex_lock_ops = {
> +	.init		= torture_rtmutex_init,

OK, so rt_boost defaults on for rtmutex.  In fact, it cannot be disabled,
which might make things more difficult for debugging.

Another approach would to do something similar to the test_boost module
parameter for RCU.  This defaults to "1", which means "Boost if it
makes sense in this situation".  It can be set to "0", which means
"Never boost", and also to "2", which means "Boost even if it makes no
sense to do so.  This last helps verify rcutorture's ability to detect
boost failures.  There is a can_boost field in the rcu_torture_ops
structure that defines when it makes sense to boost, and this field
is initialized based on CONFIG_RCU_BOOST.

In this case, it makes sense to boost rt_mutex always, and it makes
sense to boost exclusive spinlocks in PREEMPT_RT kernels.  It might make
sense to boost reader-writer spinlock situations involving only writers,
but that would likely require additional changes.

Or is there some reason why this approach would not work well?

							Thanx, Paul

>  	.writelock	= torture_rtmutex_lock,
>  	.write_delay	= torture_rtmutex_delay,
> -	.task_boost     = torture_rtmutex_boost,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_rtmutex_unlock,
>  	.readlock       = NULL,
>  	.read_delay     = NULL,
> @@ -604,7 +609,7 @@ __releases(torture_rwsem)
>  static struct lock_torture_ops rwsem_lock_ops = {
>  	.writelock	= torture_rwsem_down_write,
>  	.write_delay	= torture_rwsem_write_delay,
> -	.task_boost     = torture_boost_dummy,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_rwsem_up_write,
>  	.readlock       = torture_rwsem_down_read,
>  	.read_delay     = torture_rwsem_read_delay,
> @@ -656,7 +661,7 @@ static struct lock_torture_ops percpu_rwsem_lock_ops = {
>  	.exit		= torture_percpu_rwsem_exit,
>  	.writelock	= torture_percpu_rwsem_down_write,
>  	.write_delay	= torture_rwsem_write_delay,
> -	.task_boost     = torture_boost_dummy,
> +	.task_boost     = torture_rt_boost,
>  	.writeunlock	= torture_percpu_rwsem_up_write,
>  	.readlock       = torture_percpu_rwsem_down_read,
>  	.read_delay     = torture_rwsem_read_delay,
> -- 
> 2.38.1.584.g0f3c55d4c2-goog
> 

Powered by blists - more mailing lists