Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1682200 > unrolled thread

[PATCH] cpu_pm: replace raw_notifier to atomic_notifier

Back to article view | Back to linux.kernel

Contents

  [PATCH] cpu_pm: replace raw_notifier to atomic_notifier Alex Shi <alex.shi@linaro.org> - 2017-07-06 10:50 +0200
    Re: [PATCH] cpu_pm: replace raw_notifier to atomic_notifier Sebastian Andrzej Siewior <bigeasy@linutronix.de> - 2017-07-07 12:30 +0200
      Re: [PATCH] cpu_pm: replace raw_notifier to atomic_notifier Alex Shi <alex.shi@linaro.org> - 2017-07-08 09:30 +0200
    Re: [PATCH] cpu_pm: replace raw_notifier to atomic_notifier Alex Shi <alex.shi@linaro.org> - 2017-07-11 16:50 +0200
      Re: [PATCH] cpu_pm: replace raw_notifier to atomic_notifier Sebastian Andrzej Siewior <bigeasy@linutronix.de> - 2017-07-11 17:10 +0200
        Re: [PATCH] cpu_pm: replace raw_notifier to atomic_notifier "Rafael J. Wysocki" <rjw@rjwysocki.net> - 2017-07-11 17:10 +0200
          Re: [PATCH] cpu_pm: replace raw_notifier to atomic_notifier Sebastian Andrzej Siewior <bigeasy@linutronix.de> - 2017-07-11 17:20 +0200
            Re: [PATCH] cpu_pm: replace raw_notifier to atomic_notifier Alex Shi <alex.shi@linaro.org> - 2017-07-12 03:00 +0200

#1682200 — [PATCH] cpu_pm: replace raw_notifier to atomic_notifier

This patch replace a rwlock and raw notifier by atomic notifier which
protected by spin_lock and rcu.

The first to reason to have this replace is due to a 'scheduling while
 atomic' bug of RT kernel on arm/arm64 platform. On arm/arm64, rwlock
cpu_pm_notifier_lock in cpu_pm cause a potential schedule after irq
disable in idle call chain:

cpu_startup_entry
  cpu_idle_loop
    local_irq_disable()
    cpuidle_idle_call
      call_cpuidle
        cpuidle_enter
          cpuidle_enter_state
            ->enter :arm_enter_idle_state
              cpu_pm_enter/exit
                CPU_PM_CPU_IDLE_ENTER
                  read_lock(&cpu_pm_notifier_lock); <-- sleep in idle
                     __rt_spin_lock();
                        schedule();

The kernel panic is here:
[    4.609601] BUG: scheduling while atomic: swapper/1/0/0x00000002
[    4.609608] [<ffff0000086fae70>] arm_enter_idle_state+0x18/0x70
[    4.609614] Modules linked in:
[    4.609615] [<ffff0000086f9298>] cpuidle_enter_state+0xf0/0x218
[    4.609620] [<ffff0000086f93f8>] cpuidle_enter+0x18/0x20
[    4.609626] Preemption disabled at:
[    4.609627] [<ffff0000080fa234>] call_cpuidle+0x24/0x40
[    4.609635] [<ffff000008882fa4>] schedule_preempt_disabled+0x1c/0x28
[    4.609639] [<ffff0000080fa49c>] cpu_startup_entry+0x154/0x1f8
[    4.609645] [<ffff00000808e004>] secondary_start_kernel+0x15c/0x1a0

Daniel Lezcano said this notification is needed on arm/arm64 platforms.
Sebastian suggested using atomic_notifier instead of rwlock, which is not
only removing the sleeping in idle, but also getting better latency
improvement.

This patch passed Fengguang's 0day testing.

Signed-off-by: Alex Shi <alex.shi@linaro.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Anders Roxell <anders.roxell@linaro.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: linux-rt-users <linux-rt-users@vger.kernel.org>
---
 kernel/cpu_pm.c | 43 ++++++-------------------------------------
 1 file changed, 6 insertions(+), 37 deletions(-)

diff --git a/kernel/cpu_pm.c b/kernel/cpu_pm.c
index 009cc9a..10f4640 100644
--- a/kernel/cpu_pm.c
+++ b/kernel/cpu_pm.c
@@ -22,14 +22,13 @@
 #include <linux/spinlock.h>
 #include <linux/syscore_ops.h>
 
-static DEFINE_RWLOCK(cpu_pm_notifier_lock);
-static RAW_NOTIFIER_HEAD(cpu_pm_notifier_chain);
+static ATOMIC_NOTIFIER_HEAD(cpu_pm_notifier_chain);
 
 static int cpu_pm_notify(enum cpu_pm_event event, int nr_to_call, int *nr_calls)
 {
 	int ret;
 
-	ret = __raw_notifier_call_chain(&cpu_pm_notifier_chain, event, NULL,
+	ret = __atomic_notifier_call_chain(&cpu_pm_notifier_chain, event, NULL,
 		nr_to_call, nr_calls);
 
 	return notifier_to_errno(ret);
@@ -47,14 +46,7 @@ static int cpu_pm_notify(enum cpu_pm_event event, int nr_to_call, int *nr_calls)
  */
 int cpu_pm_register_notifier(struct notifier_block *nb)
 {
-	unsigned long flags;
-	int ret;
-
-	write_lock_irqsave(&cpu_pm_notifier_lock, flags);
-	ret = raw_notifier_chain_register(&cpu_pm_notifier_chain, nb);
-	write_unlock_irqrestore(&cpu_pm_notifier_lock, flags);
-
-	return ret;
+	return atomic_notifier_chain_register(&cpu_pm_notifier_chain, nb);
 }
 EXPORT_SYMBOL_GPL(cpu_pm_register_notifier);
 
@@ -69,14 +61,7 @@ EXPORT_SYMBOL_GPL(cpu_pm_register_notifier);
  */
 int cpu_pm_unregister_notifier(struct notifier_block *nb)
 {
-	unsigned long flags;
-	int ret;
-
-	write_lock_irqsave(&cpu_pm_notifier_lock, flags);
-	ret = raw_notifier_chain_unregister(&cpu_pm_notifier_chain, nb);
-	write_unlock_irqrestore(&cpu_pm_notifier_lock, flags);
-
-	return ret;
+	return atomic_notifier_chain_unregister(&cpu_pm_notifier_chain, nb);
 }
 EXPORT_SYMBOL_GPL(cpu_pm_unregister_notifier);
 
@@ -100,7 +85,6 @@ int cpu_pm_enter(void)
 	int nr_calls;
 	int ret = 0;
 
-	read_lock(&cpu_pm_notifier_lock);
 	ret = cpu_pm_notify(CPU_PM_ENTER, -1, &nr_calls);
 	if (ret)
 		/*
@@ -108,7 +92,6 @@ int cpu_pm_enter(void)
 		 * PM entry who are notified earlier to prepare for it.
 		 */
 		cpu_pm_notify(CPU_PM_ENTER_FAILED, nr_calls - 1, NULL);
-	read_unlock(&cpu_pm_notifier_lock);
 
 	return ret;
 }
@@ -128,13 +111,7 @@ EXPORT_SYMBOL_GPL(cpu_pm_enter);
  */
 int cpu_pm_exit(void)
 {
-	int ret;
-
-	read_lock(&cpu_pm_notifier_lock);
-	ret = cpu_pm_notify(CPU_PM_EXIT, -1, NULL);
-	read_unlock(&cpu_pm_notifier_lock);
-
-	return ret;
+	return cpu_pm_notify(CPU_PM_EXIT, -1, NULL);
 }
 EXPORT_SYMBOL_GPL(cpu_pm_exit);
 
@@ -159,7 +136,6 @@ int cpu_cluster_pm_enter(void)
 	int nr_calls;
 	int ret = 0;
 
-	read_lock(&cpu_pm_notifier_lock);
 	ret = cpu_pm_notify(CPU_CLUSTER_PM_ENTER, -1, &nr_calls);
 	if (ret)
 		/*
@@ -167,7 +143,6 @@ int cpu_cluster_pm_enter(void)
 		 * PM entry who are notified earlier to prepare for it.
 		 */
 		cpu_pm_notify(CPU_CLUSTER_PM_ENTER_FAILED, nr_calls - 1, NULL);
-	read_unlock(&cpu_pm_notifier_lock);
 
 	return ret;
 }
@@ -190,13 +165,7 @@ EXPORT_SYMBOL_GPL(cpu_cluster_pm_enter);
  */
 int cpu_cluster_pm_exit(void)
 {
-	int ret;
-
-	read_lock(&cpu_pm_notifier_lock);
-	ret = cpu_pm_notify(CPU_CLUSTER_PM_EXIT, -1, NULL);
-	read_unlock(&cpu_pm_notifier_lock);
-
-	return ret;
+	return cpu_pm_notify(CPU_CLUSTER_PM_EXIT, -1, NULL);
 }
 EXPORT_SYMBOL_GPL(cpu_cluster_pm_exit);
 
-- 
2.7.4

[toc] | [next] | [standalone]

#1683112

On 2017-07-06 16:47:46 [+0800], Alex Shi wrote:
> This patch replace a rwlock and raw notifier by atomic notifier which
+ is 
> protected by spin_lock and rcu.
…
> Sebastian suggested using atomic_notifier instead of rwlock, which is not
> only removing the sleeping in idle, but also getting better latency
> improvement.

Did you measure this / have numbers or did you write this because this
is what RCU does in general?

Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>

Sebastian

[toc] | [prev] | [next] | [standalone]

#1683551

On 07/07/2017 06:25 PM, Sebastian Andrzej Siewior wrote:
> On 2017-07-06 16:47:46 [+0800], Alex Shi wrote:
>> This patch replace a rwlock and raw notifier by atomic notifier which
> + is 
>> protected by spin_lock and rcu.
> …
>> Sebastian suggested using atomic_notifier instead of rwlock, which is not
>> only removing the sleeping in idle, but also getting better latency
>> improvement.
> 
> Did you measure this / have numbers or did you write this because this
> is what RCU does in general?

Uh, I have no number on the performance. The benefit is in the theory.
https://lwn.net/Articles/263130/
But above article give some comparison data of rwlock and RCU.

> 
> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> 

Thanks a lot!

Regards
Alex

[toc] | [prev] | [next] | [standalone]

#1685096

It is a serious bug: add a waiting lock in idle and cause boot failure
in arm/arm64 RT.

Any more comments for this change?

On 07/06/2017 04:47 PM, Alex Shi wrote:
> This patch replace a rwlock and raw notifier by atomic notifier which
> protected by spin_lock and rcu.
> 
> The first to reason to have this replace is due to a 'scheduling while
>  atomic' bug of RT kernel on arm/arm64 platform. On arm/arm64, rwlock
> cpu_pm_notifier_lock in cpu_pm cause a potential schedule after irq
> disable in idle call chain:
> 
> cpu_startup_entry
>   cpu_idle_loop
>     local_irq_disable()
>     cpuidle_idle_call
>       call_cpuidle
>         cpuidle_enter
>           cpuidle_enter_state
>             ->enter :arm_enter_idle_state
>               cpu_pm_enter/exit
>                 CPU_PM_CPU_IDLE_ENTER
>                   read_lock(&cpu_pm_notifier_lock); <-- sleep in idle
>                      __rt_spin_lock();
>                         schedule();
> 
> The kernel panic is here:
> [    4.609601] BUG: scheduling while atomic: swapper/1/0/0x00000002
> [    4.609608] [<ffff0000086fae70>] arm_enter_idle_state+0x18/0x70
> [    4.609614] Modules linked in:
> [    4.609615] [<ffff0000086f9298>] cpuidle_enter_state+0xf0/0x218
> [    4.609620] [<ffff0000086f93f8>] cpuidle_enter+0x18/0x20
> [    4.609626] Preemption disabled at:
> [    4.609627] [<ffff0000080fa234>] call_cpuidle+0x24/0x40
> [    4.609635] [<ffff000008882fa4>] schedule_preempt_disabled+0x1c/0x28
> [    4.609639] [<ffff0000080fa49c>] cpu_startup_entry+0x154/0x1f8
> [    4.609645] [<ffff00000808e004>] secondary_start_kernel+0x15c/0x1a0
> 
> Daniel Lezcano said this notification is needed on arm/arm64 platforms.
> Sebastian suggested using atomic_notifier instead of rwlock, which is not
> only removing the sleeping in idle, but also getting better latency
> improvement.
> 
> This patch passed Fengguang's 0day testing.
> 
> Signed-off-by: Alex Shi <alex.shi@linaro.org>
> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Anders Roxell <anders.roxell@linaro.org>
> Cc: Rik van Riel <riel@redhat.com>
> Cc: Steven Rostedt <rostedt@goodmis.org>
> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
> Cc: linux-rt-users <linux-rt-users@vger.kernel.org>
> ---
>  kernel/cpu_pm.c | 43 ++++++-------------------------------------
>  1 file changed, 6 insertions(+), 37 deletions(-)
> 
> diff --git a/kernel/cpu_pm.c b/kernel/cpu_pm.c
> index 009cc9a..10f4640 100644
> --- a/kernel/cpu_pm.c
> +++ b/kernel/cpu_pm.c
> @@ -22,14 +22,13 @@
>  #include <linux/spinlock.h>
>  #include <linux/syscore_ops.h>
>  
> -static DEFINE_RWLOCK(cpu_pm_notifier_lock);
> -static RAW_NOTIFIER_HEAD(cpu_pm_notifier_chain);
> +static ATOMIC_NOTIFIER_HEAD(cpu_pm_notifier_chain);
>  
>  static int cpu_pm_notify(enum cpu_pm_event event, int nr_to_call, int *nr_calls)
>  {
>  	int ret;
>  
> -	ret = __raw_notifier_call_chain(&cpu_pm_notifier_chain, event, NULL,
> +	ret = __atomic_notifier_call_chain(&cpu_pm_notifier_chain, event, NULL,
>  		nr_to_call, nr_calls);
>  
>  	return notifier_to_errno(ret);
> @@ -47,14 +46,7 @@ static int cpu_pm_notify(enum cpu_pm_event event, int nr_to_call, int *nr_calls)
>   */
>  int cpu_pm_register_notifier(struct notifier_block *nb)
>  {
> -	unsigned long flags;
> -	int ret;
> -
> -	write_lock_irqsave(&cpu_pm_notifier_lock, flags);
> -	ret = raw_notifier_chain_register(&cpu_pm_notifier_chain, nb);
> -	write_unlock_irqrestore(&cpu_pm_notifier_lock, flags);
> -
> -	return ret;
> +	return atomic_notifier_chain_register(&cpu_pm_notifier_chain, nb);
>  }
>  EXPORT_SYMBOL_GPL(cpu_pm_register_notifier);
>  
> @@ -69,14 +61,7 @@ EXPORT_SYMBOL_GPL(cpu_pm_register_notifier);
>   */
>  int cpu_pm_unregister_notifier(struct notifier_block *nb)
>  {
> -	unsigned long flags;
> -	int ret;
> -
> -	write_lock_irqsave(&cpu_pm_notifier_lock, flags);
> -	ret = raw_notifier_chain_unregister(&cpu_pm_notifier_chain, nb);
> -	write_unlock_irqrestore(&cpu_pm_notifier_lock, flags);
> -
> -	return ret;
> +	return atomic_notifier_chain_unregister(&cpu_pm_notifier_chain, nb);
>  }
>  EXPORT_SYMBOL_GPL(cpu_pm_unregister_notifier);
>  
> @@ -100,7 +85,6 @@ int cpu_pm_enter(void)
>  	int nr_calls;
>  	int ret = 0;
>  
> -	read_lock(&cpu_pm_notifier_lock);
>  	ret = cpu_pm_notify(CPU_PM_ENTER, -1, &nr_calls);
>  	if (ret)
>  		/*
> @@ -108,7 +92,6 @@ int cpu_pm_enter(void)
>  		 * PM entry who are notified earlier to prepare for it.
>  		 */
>  		cpu_pm_notify(CPU_PM_ENTER_FAILED, nr_calls - 1, NULL);
> -	read_unlock(&cpu_pm_notifier_lock);
>  
>  	return ret;
>  }
> @@ -128,13 +111,7 @@ EXPORT_SYMBOL_GPL(cpu_pm_enter);
>   */
>  int cpu_pm_exit(void)
>  {
> -	int ret;
> -
> -	read_lock(&cpu_pm_notifier_lock);
> -	ret = cpu_pm_notify(CPU_PM_EXIT, -1, NULL);
> -	read_unlock(&cpu_pm_notifier_lock);
> -
> -	return ret;
> +	return cpu_pm_notify(CPU_PM_EXIT, -1, NULL);
>  }
>  EXPORT_SYMBOL_GPL(cpu_pm_exit);
>  
> @@ -159,7 +136,6 @@ int cpu_cluster_pm_enter(void)
>  	int nr_calls;
>  	int ret = 0;
>  
> -	read_lock(&cpu_pm_notifier_lock);
>  	ret = cpu_pm_notify(CPU_CLUSTER_PM_ENTER, -1, &nr_calls);
>  	if (ret)
>  		/*
> @@ -167,7 +143,6 @@ int cpu_cluster_pm_enter(void)
>  		 * PM entry who are notified earlier to prepare for it.
>  		 */
>  		cpu_pm_notify(CPU_CLUSTER_PM_ENTER_FAILED, nr_calls - 1, NULL);
> -	read_unlock(&cpu_pm_notifier_lock);
>  
>  	return ret;
>  }
> @@ -190,13 +165,7 @@ EXPORT_SYMBOL_GPL(cpu_cluster_pm_enter);
>   */
>  int cpu_cluster_pm_exit(void)
>  {
> -	int ret;
> -
> -	read_lock(&cpu_pm_notifier_lock);
> -	ret = cpu_pm_notify(CPU_CLUSTER_PM_EXIT, -1, NULL);
> -	read_unlock(&cpu_pm_notifier_lock);
> -
> -	return ret;
> +	return cpu_pm_notify(CPU_CLUSTER_PM_EXIT, -1, NULL);
>  }
>  EXPORT_SYMBOL_GPL(cpu_cluster_pm_exit);
>  
> 

[toc] | [prev] | [next] | [standalone]

#1685116

On 2017-07-11 22:42:04 [+0800], Alex Shi wrote:
> It is a serious bug: add a waiting lock in idle and cause boot failure
> in arm/arm64 RT.
> 
> Any more comments for this change?

As far as RT is concerned, I am taking this for the next v4.11 release.
I would appreciate if upstream would apply this as well.
Rafael do you feel responsible for this?

Sebastian

[toc] | [prev] | [next] | [standalone]

#1685117

On Tuesday, July 11, 2017 05:06:35 PM Sebastian Andrzej Siewior wrote:
> On 2017-07-11 22:42:04 [+0800], Alex Shi wrote:
> > It is a serious bug: add a waiting lock in idle and cause boot failure
> > in arm/arm64 RT.
> > 
> > Any more comments for this change?
> 
> As far as RT is concerned, I am taking this for the next v4.11 release.
> I would appreciate if upstream would apply this as well.
> Rafael do you feel responsible for this?

I can apply this if no one else wants to. :-)

However, I'd appreciate a resend with a CC to linux-pm@vger.kernel.org for
easier processing.

Thanks,
Rafael

[toc] | [prev] | [next] | [standalone]

#1685129

On 2017-07-11 17:01:09 [+0200], Rafael J. Wysocki wrote:
> > As far as RT is concerned, I am taking this for the next v4.11 release.
> > I would appreciate if upstream would apply this as well.
> > Rafael do you feel responsible for this?
> 
> I can apply this if no one else wants to. :-)

prosze.

> However, I'd appreciate a resend with a CC to linux-pm@vger.kernel.org for
> easier processing.

Alex: please repost.

> Thanks,
> Rafael

Sebastian

[toc] | [prev] | [next] | [standalone]

#1685425

On 07/11/2017 11:16 PM, Sebastian Andrzej Siewior wrote:
> On 2017-07-11 17:01:09 [+0200], Rafael J. Wysocki wrote:
>>> As far as RT is concerned, I am taking this for the next v4.11 release.
>>> I would appreciate if upstream would apply this as well.
>>> Rafael do you feel responsible for this?
>>
>> I can apply this if no one else wants to. :-)
> 
> prosze.
> 
>> However, I'd appreciate a resend with a CC to linux-pm@vger.kernel.org for
>> easier processing.
> 
> Alex: please repost.

Reposted. Thanks for you all! :)

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web