[PATCH] sched: do not call __put_task_struct() on rt if pi_blocked_on is set

[PATCH] sched: do not call __put_task_struct() on rt if pi_blocked_on is set

[Luis Claudio R. Goncalves]

With PREEMPT_RT enabled, some of the calls to put_task_struct() coming
from rt_mutex_adjust_prio_chain() could happen in preemptible context and
with a mutex enqueued. That could lead to this sequence:

	rt_mutex_adjust_prio_chain()
	  put_task_struct()
	    __put_task_struct()
	      sched_ext_free()
	        spin_lock_irqsave()
	          rtlock_lock() --->  TRIGGERS
	                              lockdep_assert(!current->pi_blocked_on);

Adjust the check in put_task_struct() to also consider pi_blocked_on before
calling __put_task_struct(), resorting to the deferred call in case it is
set.

Suggested-by: Crystal Wood <crwood@redhat.com>
Signed-off-by: Luis Claudio R. Goncalves <lgoncalv@redhat.com>
---
 include/linux/sched/task.h |   16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/include/linux/sched/task.h b/include/linux/sched/task.h
index 0f2aeb37bbb04..638114f66a4d7 100644
--- a/include/linux/sched/task.h
+++ b/include/linux/sched/task.h
@@ -130,14 +130,22 @@ extern void __put_task_struct_rcu_cb(struct rcu_head *rhp);
 
 static inline void put_task_struct(struct task_struct *t)
 {
+	bool defer = false;
+
 	if (!refcount_dec_and_test(&t->usage))
 		return;
 
 	/*
 	 * In !RT, it is always safe to call __put_task_struct().
-	 * Under RT, we can only call it in preemptible context.
+	 * Under RT, we can only call it in preemptible context,
+	 * when not blocked on a PI chain.
 	 */
-	if (!IS_ENABLED(CONFIG_PREEMPT_RT) || preemptible()) {
+#ifdef CONFIG_PREEMPT_RT
+	if (!preemptible() || current->pi_blocked_on)
+		defer = true;
+#endif
+
+	if (!defer) {
 		static DEFINE_WAIT_OVERRIDE_MAP(put_task_map, LD_WAIT_SLEEP);
 
 		lock_map_acquire_try(&put_task_map);
@@ -149,7 +156,9 @@ static inline void put_task_struct(struct task_struct *t)
 	/*
 	 * under PREEMPT_RT, we can't call put_task_struct
 	 * in atomic context because it will indirectly
-	 * acquire sleeping locks.
+	 * acquire sleeping locks. The same is true if the
+	 * current process has a mutex enqueued (blocked on
+	 * a PI chain).
 	 *
 	 * call_rcu() will schedule delayed_put_task_struct_rcu()
 	 * to be called in process context.
-- 
2.49.0

Re: [PATCH] sched: do not call __put_task_struct() on rt if pi_blocked_on is set

[Steven Rostedt]

On Tue, 8 Apr 2025 21:27:37 -0300
"Luis Claudio R. Goncalves" <lgoncalv@redhat.com> wrote:

>  {
> +	bool defer = false;
> +
>  	if (!refcount_dec_and_test(&t->usage))
>  		return;
>  
>  	/*
>  	 * In !RT, it is always safe to call __put_task_struct().
> -	 * Under RT, we can only call it in preemptible context.
> +	 * Under RT, we can only call it in preemptible context,
> +	 * when not blocked on a PI chain.
>  	 */
> -	if (!IS_ENABLED(CONFIG_PREEMPT_RT) || preemptible()) {
> +#ifdef CONFIG_PREEMPT_RT
> +	if (!preemptible() || current->pi_blocked_on)
> +		defer = true;
> +#endif

Why add the ugly #ifdef back?

	if (!IS_ENABLED(CONFIG_PREEMPT_RT) ||
	    (preemptible() && !current->pi_blocked_on)) {

-- Steve

> +
> +	if (!defer) {
>  		static DEFINE_WAIT_OVERRIDE_MAP(put_task_map, LD_WAIT_SLEEP);
>  
>  

Re: [PATCH] sched: do not call __put_task_struct() on rt if pi_blocked_on is set

[Luis Claudio R. Goncalves]

On Wed, Apr 09, 2025 at 11:43:30AM -0400, Steven Rostedt wrote:
> On Tue, 8 Apr 2025 21:27:37 -0300
> "Luis Claudio R. Goncalves" <lgoncalv@redhat.com> wrote:
> 
> >  {
> > +	bool defer = false;
> > +
> >  	if (!refcount_dec_and_test(&t->usage))
> >  		return;
> >  
> >  	/*
> >  	 * In !RT, it is always safe to call __put_task_struct().
> > -	 * Under RT, we can only call it in preemptible context.
> > +	 * Under RT, we can only call it in preemptible context,
> > +	 * when not blocked on a PI chain.
> >  	 */
> > -	if (!IS_ENABLED(CONFIG_PREEMPT_RT) || preemptible()) {
> > +#ifdef CONFIG_PREEMPT_RT
> > +	if (!preemptible() || current->pi_blocked_on)
> > +		defer = true;
> > +#endif
> 
> Why add the ugly #ifdef back?
> 
> 	if (!IS_ENABLED(CONFIG_PREEMPT_RT) ||
> 	    (preemptible() && !current->pi_blocked_on)) {

I had to add the #ifdef to avoid the build failing if CONFIG_RT_MUTEXES is
not set. I do know SMP, FUTEX, I2C, PSTORE and a few more things depend on
CONFIG_RT_MUTEXES being enabled, but I opted for being thorough.

I would be more than glad getting rid of the #ifdef and simplifying the patch
if that possible build failure is not a case to worry about, if RT_MUTEXES are
always enabled nowadays.

Luis

> -- Steve
> 
> > +
> > +	if (!defer) {
> >  		static DEFINE_WAIT_OVERRIDE_MAP(put_task_map, LD_WAIT_SLEEP);
> >  
> >  
> 
---end quoted text---

Re: [PATCH] sched: do not call __put_task_struct() on rt if pi_blocked_on is set

[Steven Rostedt]

On Wed, 9 Apr 2025 13:27:12 -0300
"Luis Claudio R. Goncalves" <lgoncalv@redhat.com> wrote:

> > > -	if (!IS_ENABLED(CONFIG_PREEMPT_RT) || preemptible()) {
> > > +#ifdef CONFIG_PREEMPT_RT
> > > +	if (!preemptible() || current->pi_blocked_on)
> > > +		defer = true;
> > > +#endif  
> > 
> > Why add the ugly #ifdef back?
> > 
> > 	if (!IS_ENABLED(CONFIG_PREEMPT_RT) ||
> > 	    (preemptible() && !current->pi_blocked_on)) {  
> 
> I had to add the #ifdef to avoid the build failing if CONFIG_RT_MUTEXES is
> not set. I do know SMP, FUTEX, I2C, PSTORE and a few more things depend on
> CONFIG_RT_MUTEXES being enabled, but I opted for being thorough.
> 
> I would be more than glad getting rid of the #ifdef and simplifying the patch
> if that possible build failure is not a case to worry about, if RT_MUTEXES are
> always enabled nowadays.
> 

Because pi_blocked_on is only defined when CONFIG_RT_MUTEXES is enabled?

OK, then perhaps we should add in sched.h:

#ifdef CONFIG_RT_MUTEXES
static inline bool tsk_is_pi_blocked_on(struct task_struct *tsk)
{
	return tsk->pi_blocked_on != NULL;
}
#else
static inline bool tsk_is_pi_blocked_on(strut task_struct *tsk)
{
	return false;
}
#endif

??

-- Steve

Re: [PATCH] sched: do not call __put_task_struct() on rt if pi_blocked_on is set

[Luis Claudio R. Goncalves]

On Wed, Apr 09, 2025 at 12:39:44PM -0400, Steven Rostedt wrote:
> On Wed, 9 Apr 2025 13:27:12 -0300
> "Luis Claudio R. Goncalves" <lgoncalv@redhat.com> wrote:
> 
> > > > -	if (!IS_ENABLED(CONFIG_PREEMPT_RT) || preemptible()) {
> > > > +#ifdef CONFIG_PREEMPT_RT
> > > > +	if (!preemptible() || current->pi_blocked_on)
> > > > +		defer = true;
> > > > +#endif  
> > > 
> > > Why add the ugly #ifdef back?
> > > 
> > > 	if (!IS_ENABLED(CONFIG_PREEMPT_RT) ||
> > > 	    (preemptible() && !current->pi_blocked_on)) {  
> > 
> > I had to add the #ifdef to avoid the build failing if CONFIG_RT_MUTEXES is
> > not set. I do know SMP, FUTEX, I2C, PSTORE and a few more things depend on
> > CONFIG_RT_MUTEXES being enabled, but I opted for being thorough.
> > 
> > I would be more than glad getting rid of the #ifdef and simplifying the patch
> > if that possible build failure is not a case to worry about, if RT_MUTEXES are
> > always enabled nowadays.
> > 
> 
> Because pi_blocked_on is only defined when CONFIG_RT_MUTEXES is enabled?
> 
> OK, then perhaps we should add in sched.h:
> 
> #ifdef CONFIG_RT_MUTEXES
> static inline bool tsk_is_pi_blocked_on(struct task_struct *tsk)
> {
> 	return tsk->pi_blocked_on != NULL;
> }
> #else
> static inline bool tsk_is_pi_blocked_on(strut task_struct *tsk)
> {
> 	return false;
> }
> #endif
> 
> ??

Thank you for the suggestion! I will rework the patch with that and send v2
after basic testing.

Luis
 
> -- Steve
> 
---end quoted text---