2014-10-22 17:07:46

by Boris Ostrovsky

[permalink] [raw]
Subject: [PATCH v3] xen/smp: Use die_complete completion when taking CPU down

Commit 2ed53c0d6cc9 ("x86/smpboot: Speed up suspend/resume by avoiding
100ms sleep for CPU offline during S3") introduced completions to CPU
offlining process. These completions are not initialized on Xen kernels
causing a panic in play_dead_common().

Move handling of die_complete into common routines to make them
available to Xen guests.

(While at it, move die_complete definition under #ifdef CONFIG_HOTPLUG_CPU)

Signed-off-by: Boris Ostrovsky <[email protected]>
Reviewed-by: David Vrabel <[email protected]>
---

v3: Don't #ifdef function prototypes in smp.h
v2: Keep die_complete private to smpboot.c


arch/x86/include/asm/smp.h | 1 +
arch/x86/kernel/smpboot.c | 13 +++++++++++--
arch/x86/xen/smp.c | 3 +++
3 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
index 8cd27e0..8cd1cc3 100644
--- a/arch/x86/include/asm/smp.h
+++ b/arch/x86/include/asm/smp.h
@@ -150,6 +150,7 @@ static inline void arch_send_call_function_ipi_mask(const struct cpumask *mask)
}

void cpu_disable_common(void);
+void cpu_die_common(unsigned int cpu);
void native_smp_prepare_boot_cpu(void);
void native_smp_prepare_cpus(unsigned int max_cpus);
void native_smp_cpus_done(unsigned int max_cpus);
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 2d5200e..6fe0fc1 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -102,7 +102,9 @@ DEFINE_PER_CPU_READ_MOSTLY(cpumask_var_t, cpu_llc_shared_map);
DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
EXPORT_PER_CPU_SYMBOL(cpu_info);

+#ifdef CONFIG_HOTPLUG_CPU
static DEFINE_PER_CPU(struct completion, die_complete);
+#endif

atomic_t init_deasserted;

@@ -1309,6 +1311,8 @@ void cpu_disable_common(void)
{
int cpu = smp_processor_id();

+ init_completion(&per_cpu(die_complete, smp_processor_id()));
+
remove_siblinginfo(cpu);

/* It's now safe to remove this processor from the online map */
@@ -1327,16 +1331,21 @@ int native_cpu_disable(void)
return ret;

clear_local_APIC();
- init_completion(&per_cpu(die_complete, smp_processor_id()));
cpu_disable_common();

return 0;
}

+void cpu_die_common(unsigned int cpu)
+{
+ wait_for_completion_timeout(&per_cpu(die_complete, cpu), HZ);
+}
+
void native_cpu_die(unsigned int cpu)
{
/* We don't do anything here: idle task is faking death itself. */
- wait_for_completion_timeout(&per_cpu(die_complete, cpu), HZ);
+
+ cpu_die_common(cpu);

/* They ack this in play_dead() by setting CPU_DEAD */
if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
index 8650cdb..4c071ae 100644
--- a/arch/x86/xen/smp.c
+++ b/arch/x86/xen/smp.c
@@ -510,6 +510,9 @@ static void xen_cpu_die(unsigned int cpu)
current->state = TASK_UNINTERRUPTIBLE;
schedule_timeout(HZ/10);
}
+
+ cpu_die_common(cpu);
+
xen_smp_intr_free(cpu);
xen_uninit_lock_cpu(cpu);
xen_teardown_timer(cpu);
--
1.8.4.2


2014-10-29 19:36:15

by Boris Ostrovsky

[permalink] [raw]
Subject: Re: [PATCH v3] xen/smp: Use die_complete completion when taking CPU down

On 10/22/2014 01:09 PM, Boris Ostrovsky wrote:
> Commit 2ed53c0d6cc9 ("x86/smpboot: Speed up suspend/resume by avoiding
> 100ms sleep for CPU offline during S3") introduced completions to CPU
> offlining process. These completions are not initialized on Xen kernels
> causing a panic in play_dead_common().
>
> Move handling of die_complete into common routines to make them
> available to Xen guests.
>
> (While at it, move die_complete definition under #ifdef CONFIG_HOTPLUG_CPU)
>
> Signed-off-by: Boris Ostrovsky <[email protected]>
> Reviewed-by: David Vrabel <[email protected]>
> ---
>
> v3: Don't #ifdef function prototypes in smp.h
> v2: Keep die_complete private to smpboot.c
>
>
> arch/x86/include/asm/smp.h | 1 +
> arch/x86/kernel/smpboot.c | 13 +++++++++++--
> arch/x86/xen/smp.c | 3 +++
> 3 files changed, 15 insertions(+), 2 deletions(-)


Does anything else need to be done in this patch? We have CPU hotplug
broken currently in Xen and it would be nice to have it fixed in rc3.

-boris


>
> diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
> index 8cd27e0..8cd1cc3 100644
> --- a/arch/x86/include/asm/smp.h
> +++ b/arch/x86/include/asm/smp.h
> @@ -150,6 +150,7 @@ static inline void arch_send_call_function_ipi_mask(const struct cpumask *mask)
> }
>
> void cpu_disable_common(void);
> +void cpu_die_common(unsigned int cpu);
> void native_smp_prepare_boot_cpu(void);
> void native_smp_prepare_cpus(unsigned int max_cpus);
> void native_smp_cpus_done(unsigned int max_cpus);
> diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
> index 2d5200e..6fe0fc1 100644
> --- a/arch/x86/kernel/smpboot.c
> +++ b/arch/x86/kernel/smpboot.c
> @@ -102,7 +102,9 @@ DEFINE_PER_CPU_READ_MOSTLY(cpumask_var_t, cpu_llc_shared_map);
> DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
> EXPORT_PER_CPU_SYMBOL(cpu_info);
>
> +#ifdef CONFIG_HOTPLUG_CPU
> static DEFINE_PER_CPU(struct completion, die_complete);
> +#endif
>
> atomic_t init_deasserted;
>
> @@ -1309,6 +1311,8 @@ void cpu_disable_common(void)
> {
> int cpu = smp_processor_id();
>
> + init_completion(&per_cpu(die_complete, smp_processor_id()));
> +
> remove_siblinginfo(cpu);
>
> /* It's now safe to remove this processor from the online map */
> @@ -1327,16 +1331,21 @@ int native_cpu_disable(void)
> return ret;
>
> clear_local_APIC();
> - init_completion(&per_cpu(die_complete, smp_processor_id()));
> cpu_disable_common();
>
> return 0;
> }
>
> +void cpu_die_common(unsigned int cpu)
> +{
> + wait_for_completion_timeout(&per_cpu(die_complete, cpu), HZ);
> +}
> +
> void native_cpu_die(unsigned int cpu)
> {
> /* We don't do anything here: idle task is faking death itself. */
> - wait_for_completion_timeout(&per_cpu(die_complete, cpu), HZ);
> +
> + cpu_die_common(cpu);
>
> /* They ack this in play_dead() by setting CPU_DEAD */
> if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
> diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
> index 8650cdb..4c071ae 100644
> --- a/arch/x86/xen/smp.c
> +++ b/arch/x86/xen/smp.c
> @@ -510,6 +510,9 @@ static void xen_cpu_die(unsigned int cpu)
> current->state = TASK_UNINTERRUPTIBLE;
> schedule_timeout(HZ/10);
> }
> +
> + cpu_die_common(cpu);
> +
> xen_smp_intr_free(cpu);
> xen_uninit_lock_cpu(cpu);
> xen_teardown_timer(cpu);

2014-10-31 10:13:43

by Ingo Molnar

[permalink] [raw]
Subject: Re: [Xen-devel] [PATCH v3] xen/smp: Use die_complete completion when taking CPU down


* Boris Ostrovsky <[email protected]> wrote:

> On 10/22/2014 01:09 PM, Boris Ostrovsky wrote:
> >Commit 2ed53c0d6cc9 ("x86/smpboot: Speed up suspend/resume by avoiding
> >100ms sleep for CPU offline during S3") introduced completions to CPU
> >offlining process. These completions are not initialized on Xen kernels
> >causing a panic in play_dead_common().
> >
> >Move handling of die_complete into common routines to make them
> >available to Xen guests.
> >
> >(While at it, move die_complete definition under #ifdef CONFIG_HOTPLUG_CPU)
> >
> >Signed-off-by: Boris Ostrovsky <[email protected]>
> >Reviewed-by: David Vrabel <[email protected]>
> >---
> >
> >v3: Don't #ifdef function prototypes in smp.h
> >v2: Keep die_complete private to smpboot.c
> >
> >
> > arch/x86/include/asm/smp.h | 1 +
> > arch/x86/kernel/smpboot.c | 13 +++++++++++--
> > arch/x86/xen/smp.c | 3 +++
> > 3 files changed, 15 insertions(+), 2 deletions(-)
>
>
> Does anything else need to be done in this patch? We have CPU
> hotplug broken currently in Xen and it would be nice to have it
> fixed in rc3.

There's a build warning fix that this change conflicts with:

db6a00b4bed3 x86/smpboot: Move data structure to its primary usage scope

it would be nice if you could rebase it to tip:master or
tip:x86/urgent so I can push it upstream ASAP.

Thanks,

Ingo

2014-10-31 13:34:42

by Boris Ostrovsky

[permalink] [raw]
Subject: Re: [Xen-devel] [PATCH v3] xen/smp: Use die_complete completion when taking CPU down

On 10/31/2014 06:13 AM, Ingo Molnar wrote:
> * Boris Ostrovsky <[email protected]> wrote:
>
>> On 10/22/2014 01:09 PM, Boris Ostrovsky wrote:
>>> Commit 2ed53c0d6cc9 ("x86/smpboot: Speed up suspend/resume by avoiding
>>> 100ms sleep for CPU offline during S3") introduced completions to CPU
>>> offlining process. These completions are not initialized on Xen kernels
>>> causing a panic in play_dead_common().
>>>
>>> Move handling of die_complete into common routines to make them
>>> available to Xen guests.
>>>
>>> (While at it, move die_complete definition under #ifdef CONFIG_HOTPLUG_CPU)
>>>
>>> Signed-off-by: Boris Ostrovsky <[email protected]>
>>> Reviewed-by: David Vrabel <[email protected]>
>>> ---
>>>
>>> v3: Don't #ifdef function prototypes in smp.h
>>> v2: Keep die_complete private to smpboot.c
>>>
>>>
>>> arch/x86/include/asm/smp.h | 1 +
>>> arch/x86/kernel/smpboot.c | 13 +++++++++++--
>>> arch/x86/xen/smp.c | 3 +++
>>> 3 files changed, 15 insertions(+), 2 deletions(-)
>>
>> Does anything else need to be done in this patch? We have CPU
>> hotplug broken currently in Xen and it would be nice to have it
>> fixed in rc3.
> There's a build warning fix that this change conflicts with:
>
> db6a00b4bed3 x86/smpboot: Move data structure to its primary usage scope
>
> it would be nice if you could rebase it to tip:master or
> tip:x86/urgent so I can push it upstream ASAP.


Thanks Ingo. Will do.

-boris