tick/nohz: Fix cpu_is_hotpluggable() by checking with nohz subsystem

  For CONFIG_NO_HZ_FULL systems, the tick_do_timer_cpu cannot be offlined.
However, cpu_is_hotpluggable() still returns true for those CPUs. This causes
torture tests that do offlining to end up trying to offline this CPU causing
test failures. Such failure happens on all architectures.

Fix it by asking the opinion of the nohz subsystem on whether the CPU can
be hotplugged.

[ Apply Frederic Weisbecker feedback on refactoring tick_nohz_cpu_down(). ]

Cc: Frederic Weisbecker <frederic@kernel.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Zhouyi Zhou <zhouzhouyi@gmail.com>
Cc: Will Deacon <will@kernel.org>
Cc: Marc Zyngier <maz@kernel.org>
Cc: rcu <rcu@vger.kernel.org>
Fixes: 2987557f52b9 ("driver-core/cpu: Expose hotpluggability to the rest of the kernel")
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
---
 drivers/base/cpu.c       |  3 ++-
 include/linux/tick.h     |  2 ++
 kernel/time/tick-sched.c | 12 +++++++++++-
 3 files changed, 15 insertions(+), 2 deletions(-)
  

On Fri, Jan 20, 2023 at 4:45 AM Joel Fernandes (Google)
<joel@joelfernandes.org> wrote:
>
> For CONFIG_NO_HZ_FULL systems, the tick_do_timer_cpu cannot be offlined.
> However, cpu_is_hotpluggable() still returns true for those CPUs. This causes
> torture tests that do offlining to end up trying to offline this CPU causing
> test failures. Such failure happens on all architectures.
>
> Fix it by asking the opinion of the nohz subsystem on whether the CPU can
> be hotplugged.
>
> [ Apply Frederic Weisbecker feedback on refactoring tick_nohz_cpu_down(). ]
Thanks for your fantastic work
I applied this fix to linux-5.15.y, and perform new round of rcu
torture test on PPC VM of Open Source Lab of Oregon State University.
Could you please wait for the test to finish?

The test results of linux-5.15.y before your patch can be viewed at [1]
The patched source code of linux-5.15.y can be viewed at [2]
The ongoing test of patched linux-5.15.y can be viewed at [3]

[1] http://140.211.169.189/linux-stable-rc/tools/testing/selftests/rcutorture/res/2023.01.18-13.22.39-torture/
[2] http://140.211.169.189/linux-stable-rc/
[3] http://140.211.169.189/linux-stable-rc/tools/testing/selftests/rcutorture/res/2023.01.19-23.40.55-torture/

Hope to continue to benefit the community.

Thank you all
Zhouyi
>
> Cc: Frederic Weisbecker <frederic@kernel.org>
> Cc: "Paul E. McKenney" <paulmck@kernel.org>
> Cc: Zhouyi Zhou <zhouzhouyi@gmail.com>
> Cc: Will Deacon <will@kernel.org>
> Cc: Marc Zyngier <maz@kernel.org>
> Cc: rcu <rcu@vger.kernel.org>
> Fixes: 2987557f52b9 ("driver-core/cpu: Expose hotpluggability to the rest of the kernel")
> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> ---
>  drivers/base/cpu.c       |  3 ++-
>  include/linux/tick.h     |  2 ++
>  kernel/time/tick-sched.c | 12 +++++++++++-
>  3 files changed, 15 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/base/cpu.c b/drivers/base/cpu.c
> index 55405ebf23ab..450dca235a2f 100644
> --- a/drivers/base/cpu.c
> +++ b/drivers/base/cpu.c
> @@ -487,7 +487,8 @@ static const struct attribute_group *cpu_root_attr_groups[] = {
>  bool cpu_is_hotpluggable(unsigned int cpu)
>  {
>         struct device *dev = get_cpu_device(cpu);
> -       return dev && container_of(dev, struct cpu, dev)->hotpluggable;
> +       return dev && container_of(dev, struct cpu, dev)->hotpluggable
> +               && tick_nohz_cpu_hotpluggable(cpu);
>  }
>  EXPORT_SYMBOL_GPL(cpu_is_hotpluggable);
>
> diff --git a/include/linux/tick.h b/include/linux/tick.h
> index bfd571f18cfd..9459fef5b857 100644
> --- a/include/linux/tick.h
> +++ b/include/linux/tick.h
> @@ -216,6 +216,7 @@ extern void tick_nohz_dep_set_signal(struct task_struct *tsk,
>                                      enum tick_dep_bits bit);
>  extern void tick_nohz_dep_clear_signal(struct signal_struct *signal,
>                                        enum tick_dep_bits bit);
> +extern bool tick_nohz_cpu_hotpluggable(unsigned int cpu);
>
>  /*
>   * The below are tick_nohz_[set,clear]_dep() wrappers that optimize off-cases
> @@ -280,6 +281,7 @@ static inline void tick_nohz_full_add_cpus_to(struct cpumask *mask) { }
>
>  static inline void tick_nohz_dep_set_cpu(int cpu, enum tick_dep_bits bit) { }
>  static inline void tick_nohz_dep_clear_cpu(int cpu, enum tick_dep_bits bit) { }
> +static inline bool tick_nohz_cpu_hotpluggable(unsigned int cpu) { return true; }
>
>  static inline void tick_dep_set(enum tick_dep_bits bit) { }
>  static inline void tick_dep_clear(enum tick_dep_bits bit) { }
> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index 9c6f661fb436..383a060f30c5 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -510,7 +510,7 @@ void __init tick_nohz_full_setup(cpumask_var_t cpumask)
>         tick_nohz_full_running = true;
>  }
>
> -static int tick_nohz_cpu_down(unsigned int cpu)
> +static int tick_nohz_cpu_hotplug_ret(unsigned int cpu)
>  {
>         /*
>          * The tick_do_timer_cpu CPU handles housekeeping duty (unbound
> @@ -522,6 +522,16 @@ static int tick_nohz_cpu_down(unsigned int cpu)
>         return 0;
>  }
>
> +static int tick_nohz_cpu_down(unsigned int cpu)
> +{
> +       return tick_nohz_cpu_hotplug_ret(cpu);
> +}
> +
> +bool tick_nohz_cpu_hotpluggable(unsigned int cpu)
> +{
> +       return tick_nohz_cpu_hotplug_ret(cpu) == 0;
> +}
> +
>  void __init tick_nohz_init(void)
>  {
>         int cpu, ret;
> --
> 2.39.0.246.g2a6d74b583-goog
>
  

On Thu, Jan 19, 2023 at 08:44:35PM +0000, Joel Fernandes (Google) wrote:
> For CONFIG_NO_HZ_FULL systems, the tick_do_timer_cpu cannot be offlined.
> However, cpu_is_hotpluggable() still returns true for those CPUs. This causes
> torture tests that do offlining to end up trying to offline this CPU causing
> test failures. Such failure happens on all architectures.
> 
> Fix it by asking the opinion of the nohz subsystem on whether the CPU can
> be hotplugged.
> 
> [ Apply Frederic Weisbecker feedback on refactoring tick_nohz_cpu_down(). ]
> 
> Cc: Frederic Weisbecker <frederic@kernel.org>
> Cc: "Paul E. McKenney" <paulmck@kernel.org>
> Cc: Zhouyi Zhou <zhouzhouyi@gmail.com>
> Cc: Will Deacon <will@kernel.org>
> Cc: Marc Zyngier <maz@kernel.org>
> Cc: rcu <rcu@vger.kernel.org>
> Fixes: 2987557f52b9 ("driver-core/cpu: Expose hotpluggability to the rest of the kernel")
> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>

Also want to cc: stable on the patch?

Anyway, for the driver core portion:

Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  

> On Jan 20, 2023, at 2:05 AM, Greg Kroah-Hartman <gregkh@linuxfoundation.org> wrote:
> 
> On Thu, Jan 19, 2023 at 08:44:35PM +0000, Joel Fernandes (Google) wrote:
>> For CONFIG_NO_HZ_FULL systems, the tick_do_timer_cpu cannot be offlined.
>> However, cpu_is_hotpluggable() still returns true for those CPUs. This causes
>> torture tests that do offlining to end up trying to offline this CPU causing
>> test failures. Such failure happens on all architectures.
>> 
>> Fix it by asking the opinion of the nohz subsystem on whether the CPU can
>> be hotplugged.
>> 
>> [ Apply Frederic Weisbecker feedback on refactoring tick_nohz_cpu_down(). ]
>> 
>> Cc: Frederic Weisbecker <frederic@kernel.org>
>> Cc: "Paul E. McKenney" <paulmck@kernel.org>
>> Cc: Zhouyi Zhou <zhouzhouyi@gmail.com>
>> Cc: Will Deacon <will@kernel.org>
>> Cc: Marc Zyngier <maz@kernel.org>
>> Cc: rcu <rcu@vger.kernel.org>
>> Fixes: 2987557f52b9 ("driver-core/cpu: Expose hotpluggability to the rest of the kernel")
>> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> 
> Also want to cc: stable on the patch?

Oh sure, sorry. For some reason I thought Sasha and your AI scripts were looking at the Linux-kernel list as well. Or are they, and a Cc to stable is just to be doubly sure?

> Anyway, for the driver core portion:

Thanks!

 - Joel

> 
> Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  

On Fri, Jan 20, 2023 at 08:32:30AM -0500, Joel Fernandes wrote:
> 
> 
> > On Jan 20, 2023, at 2:05 AM, Greg Kroah-Hartman <gregkh@linuxfoundation.org> wrote:
> > 
> > On Thu, Jan 19, 2023 at 08:44:35PM +0000, Joel Fernandes (Google) wrote:
> >> For CONFIG_NO_HZ_FULL systems, the tick_do_timer_cpu cannot be offlined.
> >> However, cpu_is_hotpluggable() still returns true for those CPUs. This causes
> >> torture tests that do offlining to end up trying to offline this CPU causing
> >> test failures. Such failure happens on all architectures.
> >> 
> >> Fix it by asking the opinion of the nohz subsystem on whether the CPU can
> >> be hotplugged.
> >> 
> >> [ Apply Frederic Weisbecker feedback on refactoring tick_nohz_cpu_down(). ]
> >> 
> >> Cc: Frederic Weisbecker <frederic@kernel.org>
> >> Cc: "Paul E. McKenney" <paulmck@kernel.org>
> >> Cc: Zhouyi Zhou <zhouzhouyi@gmail.com>
> >> Cc: Will Deacon <will@kernel.org>
> >> Cc: Marc Zyngier <maz@kernel.org>
> >> Cc: rcu <rcu@vger.kernel.org>
> >> Fixes: 2987557f52b9 ("driver-core/cpu: Expose hotpluggability to the rest of the kernel")
> >> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > 
> > Also want to cc: stable on the patch?
> 
> Oh sure, sorry. For some reason I thought Sasha and your AI scripts
> were looking at the Linux-kernel list as well. Or are they, and a Cc
> to stable is just to be doubly sure?

As per the rules we have had for the last 15+ years, always add a cc:
stable to be sure that the patch will be considered for stable releases.
If not, you are on you own and sometimes we might notice it, others not.

See:
    https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
for the details.

thanks,

greg k-h
  

On Thu, Jan 19, 2023 at 08:44:35PM +0000, Joel Fernandes (Google) wrote:
> -static int tick_nohz_cpu_down(unsigned int cpu)
> +static int tick_nohz_cpu_hotplug_ret(unsigned int cpu)
>  {
>  	/*
>  	 * The tick_do_timer_cpu CPU handles housekeeping duty (unbound
> @@ -522,6 +522,16 @@ static int tick_nohz_cpu_down(unsigned int cpu)
>  	return 0;
>  }
>  
> +static int tick_nohz_cpu_down(unsigned int cpu)
> +{
> +	return tick_nohz_cpu_hotplug_ret(cpu);
> +}
> +
> +bool tick_nohz_cpu_hotpluggable(unsigned int cpu)
> +{
> +	return tick_nohz_cpu_hotplug_ret(cpu) == 0;

This is still calling the hotplug function for the CPU in order to know if the
CPU is hotpluggable...

Why not:

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index ba2ac1469d47..a46506f7ec6d 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -532,7 +532,7 @@ void __init tick_nohz_full_setup(cpumask_var_t cpumask)
 	tick_nohz_full_running = true;
 }
 
-static int tick_nohz_cpu_down(unsigned int cpu)
+bool tick_nohz_cpu_hotpluggable(unsigned int cpu)
 {
 	/*
 	 * The tick_do_timer_cpu CPU handles housekeeping duty (unbound
@@ -540,8 +540,13 @@ static int tick_nohz_cpu_down(unsigned int cpu)
 	 * CPUs. It must remain online when nohz full is enabled.
 	 */
 	if (tick_nohz_full_running && tick_do_timer_cpu == cpu)
-		return -EBUSY;
-	return 0;
+		return false;
+	return true;
+}
+
+static int tick_nohz_cpu_down(unsigned int cpu)
+{
+	return tick_nohz_cpu_hotpluggable(cpu) ? 0 : -EBUSY;
 }
 
 void __init tick_nohz_init(void)




>  void __init tick_nohz_init(void)
>  {
>  	int cpu, ret;
> -- 
> 2.39.0.246.g2a6d74b583-goog
>
  

On Fri, Jan 20, 2023 at 7:53 AM Zhouyi Zhou <zhouzhouyi@gmail.com> wrote:
>
> On Fri, Jan 20, 2023 at 4:45 AM Joel Fernandes (Google)
> <joel@joelfernandes.org> wrote:
> >
> > For CONFIG_NO_HZ_FULL systems, the tick_do_timer_cpu cannot be offlined.
> > However, cpu_is_hotpluggable() still returns true for those CPUs. This causes
> > torture tests that do offlining to end up trying to offline this CPU causing
> > test failures. Such failure happens on all architectures.
> >
> > Fix it by asking the opinion of the nohz subsystem on whether the CPU can
> > be hotplugged.
> >
> > [ Apply Frederic Weisbecker feedback on refactoring tick_nohz_cpu_down(). ]
> Thanks for your fantastic work
> I applied this fix to linux-5.15.y, and perform new round of rcu
> torture test on PPC VM of Open Source Lab of Oregon State University.
> Could you please wait for the test to finish?
>
> The test results of linux-5.15.y before your patch can be viewed at [1]
> The patched source code of linux-5.15.y can be viewed at [2]
> The ongoing test of patched linux-5.15.y can be viewed at [3]
>
> [1] http://140.211.169.189/linux-stable-rc/tools/testing/selftests/rcutorture/res/2023.01.18-13.22.39-torture/
> [2] http://140.211.169.189/linux-stable-rc/
> [3] http://140.211.169.189/linux-stable-rc/tools/testing/selftests/rcutorture/res/2023.01.19-23.40.55-torture/
The rcutorture test on [3] has finished, by comparing the logs of [1]
and [3], we can see Joel's patch fix the HOTPLUG problem without bring
any new issues.

Tested-by: Zhouyi Zhou <zhouzhouyi@gmail.com>

Thanks
Zhouyi
>
> Hope to continue to benefit the community.
>
> Thank you all
> Zhouyi
> >
> > Cc: Frederic Weisbecker <frederic@kernel.org>
> > Cc: "Paul E. McKenney" <paulmck@kernel.org>
> > Cc: Zhouyi Zhou <zhouzhouyi@gmail.com>
> > Cc: Will Deacon <will@kernel.org>
> > Cc: Marc Zyngier <maz@kernel.org>
> > Cc: rcu <rcu@vger.kernel.org>
> > Fixes: 2987557f52b9 ("driver-core/cpu: Expose hotpluggability to the rest of the kernel")
> > Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > ---
> >  drivers/base/cpu.c       |  3 ++-
> >  include/linux/tick.h     |  2 ++
> >  kernel/time/tick-sched.c | 12 +++++++++++-
> >  3 files changed, 15 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/base/cpu.c b/drivers/base/cpu.c
> > index 55405ebf23ab..450dca235a2f 100644
> > --- a/drivers/base/cpu.c
> > +++ b/drivers/base/cpu.c
> > @@ -487,7 +487,8 @@ static const struct attribute_group *cpu_root_attr_groups[] = {
> >  bool cpu_is_hotpluggable(unsigned int cpu)
> >  {
> >         struct device *dev = get_cpu_device(cpu);
> > -       return dev && container_of(dev, struct cpu, dev)->hotpluggable;
> > +       return dev && container_of(dev, struct cpu, dev)->hotpluggable
> > +               && tick_nohz_cpu_hotpluggable(cpu);
> >  }
> >  EXPORT_SYMBOL_GPL(cpu_is_hotpluggable);
> >
> > diff --git a/include/linux/tick.h b/include/linux/tick.h
> > index bfd571f18cfd..9459fef5b857 100644
> > --- a/include/linux/tick.h
> > +++ b/include/linux/tick.h
> > @@ -216,6 +216,7 @@ extern void tick_nohz_dep_set_signal(struct task_struct *tsk,
> >                                      enum tick_dep_bits bit);
> >  extern void tick_nohz_dep_clear_signal(struct signal_struct *signal,
> >                                        enum tick_dep_bits bit);
> > +extern bool tick_nohz_cpu_hotpluggable(unsigned int cpu);
> >
> >  /*
> >   * The below are tick_nohz_[set,clear]_dep() wrappers that optimize off-cases
> > @@ -280,6 +281,7 @@ static inline void tick_nohz_full_add_cpus_to(struct cpumask *mask) { }
> >
> >  static inline void tick_nohz_dep_set_cpu(int cpu, enum tick_dep_bits bit) { }
> >  static inline void tick_nohz_dep_clear_cpu(int cpu, enum tick_dep_bits bit) { }
> > +static inline bool tick_nohz_cpu_hotpluggable(unsigned int cpu) { return true; }
> >
> >  static inline void tick_dep_set(enum tick_dep_bits bit) { }
> >  static inline void tick_dep_clear(enum tick_dep_bits bit) { }
> > diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> > index 9c6f661fb436..383a060f30c5 100644
> > --- a/kernel/time/tick-sched.c
> > +++ b/kernel/time/tick-sched.c
> > @@ -510,7 +510,7 @@ void __init tick_nohz_full_setup(cpumask_var_t cpumask)
> >         tick_nohz_full_running = true;
> >  }
> >
> > -static int tick_nohz_cpu_down(unsigned int cpu)
> > +static int tick_nohz_cpu_hotplug_ret(unsigned int cpu)
> >  {
> >         /*
> >          * The tick_do_timer_cpu CPU handles housekeeping duty (unbound
> > @@ -522,6 +522,16 @@ static int tick_nohz_cpu_down(unsigned int cpu)
> >         return 0;
> >  }
> >
> > +static int tick_nohz_cpu_down(unsigned int cpu)
> > +{
> > +       return tick_nohz_cpu_hotplug_ret(cpu);
> > +}
> > +
> > +bool tick_nohz_cpu_hotpluggable(unsigned int cpu)
> > +{
> > +       return tick_nohz_cpu_hotplug_ret(cpu) == 0;
> > +}
> > +
> >  void __init tick_nohz_init(void)
> >  {
> >         int cpu, ret;
> > --
> > 2.39.0.246.g2a6d74b583-goog
> >
  

On Fri, Jan 20, 2023 at 5:25 PM Frederic Weisbecker <frederic@kernel.org> wrote:
>
> On Thu, Jan 19, 2023 at 08:44:35PM +0000, Joel Fernandes (Google) wrote:
> > -static int tick_nohz_cpu_down(unsigned int cpu)
> > +static int tick_nohz_cpu_hotplug_ret(unsigned int cpu)
> >  {
> >       /*
> >        * The tick_do_timer_cpu CPU handles housekeeping duty (unbound
> > @@ -522,6 +522,16 @@ static int tick_nohz_cpu_down(unsigned int cpu)
> >       return 0;
> >  }
> >
> > +static int tick_nohz_cpu_down(unsigned int cpu)
> > +{
> > +     return tick_nohz_cpu_hotplug_ret(cpu);
> > +}
> > +
> > +bool tick_nohz_cpu_hotpluggable(unsigned int cpu)
> > +{
> > +     return tick_nohz_cpu_hotplug_ret(cpu) == 0;
>
> This is still calling the hotplug function for the CPU in order to know if the
> CPU is hotpluggable...
>
> Why not:
>
> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index ba2ac1469d47..a46506f7ec6d 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -532,7 +532,7 @@ void __init tick_nohz_full_setup(cpumask_var_t cpumask)
>         tick_nohz_full_running = true;
>  }
>
> -static int tick_nohz_cpu_down(unsigned int cpu)
> +bool tick_nohz_cpu_hotpluggable(unsigned int cpu)
>  {
>         /*
>          * The tick_do_timer_cpu CPU handles housekeeping duty (unbound
> @@ -540,8 +540,13 @@ static int tick_nohz_cpu_down(unsigned int cpu)
>          * CPUs. It must remain online when nohz full is enabled.
>          */
>         if (tick_nohz_full_running && tick_do_timer_cpu == cpu)
> -               return -EBUSY;
> -       return 0;
> +               return false;
> +       return true;
> +}
> +
> +static int tick_nohz_cpu_down(unsigned int cpu)
> +{
> +       return tick_nohz_cpu_hotpluggable(cpu) ? 0 : -EBUSY;
>  }
>

Yes, this looks better. I will do it this way. Thanks!

 - Joel