[PATCH] genirq: update irq stats from NMI handlers

[PATCH] genirq: update irq stats from NMI handlers

[Shijith Thotton]

The NMI handlers handle_percpu_devid_fasteoi_nmi() and
handle_fasteoi_nmi() added by commit 2dcf1fbcad35 ("genirq: Provide NMI
handlers") do not update the interrupt counts. Due to that the NMI
interrupt count does not show up correctly in /proc/interrupts.

Update the functions to fix this. With this change, we can see stats of
the perf NMI interrupts on arm64.

Fixes: 2dcf1fbcad35 ("genirq: Provide NMI handlers")

Signed-off-by: Shijith Thotton <sthotton@marvell.com>
---
 kernel/irq/chip.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
index 29d6c7d070b4..88d1e054c6ea 100644
--- a/kernel/irq/chip.c
+++ b/kernel/irq/chip.c
@@ -748,6 +748,8 @@ void handle_fasteoi_nmi(struct irq_desc *desc)
 	unsigned int irq = irq_desc_get_irq(desc);
 	irqreturn_t res;
 
+	kstat_incr_irqs_this_cpu(desc);
+
 	trace_irq_handler_entry(irq, action);
 	/*
 	 * NMIs cannot be shared, there is only one action.
@@ -962,6 +964,8 @@ void handle_percpu_devid_fasteoi_nmi(struct irq_desc *desc)
 	unsigned int irq = irq_desc_get_irq(desc);
 	irqreturn_t res;
 
+	__kstat_incr_irqs_this_cpu(desc);
+
 	trace_irq_handler_entry(irq, action);
 	res = action->handler(irq, raw_cpu_ptr(action->percpu_dev_id));
 	trace_irq_handler_exit(irq, action, res);
-- 
2.17.0

Re: [PATCH] genirq: update irq stats from NMI handlers

[Julien Thierry]

Hi Shijith,

On 04/07/2019 05:22, Shijith Thotton wrote:
> The NMI handlers handle_percpu_devid_fasteoi_nmi() and
> handle_fasteoi_nmi() added by commit 2dcf1fbcad35 ("genirq: Provide NMI
> handlers") do not update the interrupt counts. Due to that the NMI
> interrupt count does not show up correctly in /proc/interrupts.
> 
> Update the functions to fix this. With this change, we can see stats of
> the perf NMI interrupts on arm64.
> 
> Fixes: 2dcf1fbcad35 ("genirq: Provide NMI handlers")
> 
> Signed-off-by: Shijith Thotton <sthotton@marvell.com>
> ---
>  kernel/irq/chip.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
> index 29d6c7d070b4..88d1e054c6ea 100644
> --- a/kernel/irq/chip.c
> +++ b/kernel/irq/chip.c
> @@ -748,6 +748,8 @@ void handle_fasteoi_nmi(struct irq_desc *desc)
>  	unsigned int irq = irq_desc_get_irq(desc);
>  	irqreturn_t res;
>  
> +	kstat_incr_irqs_this_cpu(desc);
> +

This needs to be called with the desc lock taken, otherwise we're likely
to corrupt desc->tot_count.
But taking the desc lock is something we can't do in NMI context (
*spin_lock_irq*() won't prevent an NMI from happening).

>  	trace_irq_handler_entry(irq, action);
>  	/*
>  	 * NMIs cannot be shared, there is only one action.
> @@ -962,6 +964,8 @@ void handle_percpu_devid_fasteoi_nmi(struct irq_desc *desc)
>  	unsigned int irq = irq_desc_get_irq(desc);
>  	irqreturn_t res;
>  
> +	__kstat_incr_irqs_this_cpu(desc);
> +

Looking at handle_percpu_irq(), I think this might be acceptable. But
does it make sense to only have kstats for percpu NMIs?

Cheers,

-- 
Julien Thierry

Re: [PATCH] genirq: update irq stats from NMI handlers

[Shijith Thotton]

Hi Julien,

On 7/4/19 12:13 AM, Julien Thierry wrote:
> On 04/07/2019 05:22, Shijith Thotton wrote:
>> The NMI handlers handle_percpu_devid_fasteoi_nmi() and
>> handle_fasteoi_nmi() added by commit 2dcf1fbcad35 ("genirq: Provide NMI
>> handlers") do not update the interrupt counts. Due to that the NMI
>> interrupt count does not show up correctly in /proc/interrupts.
>>
>> Update the functions to fix this. With this change, we can see stats of
>> the perf NMI interrupts on arm64.
>>
>> Fixes: 2dcf1fbcad35 ("genirq: Provide NMI handlers")
>>
>> Signed-off-by: Shijith Thotton <sthotton@marvell.com>
>> ---
>>   kernel/irq/chip.c | 4 ++++
>>   1 file changed, 4 insertions(+)
>>
>> diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
>> index 29d6c7d070b4..88d1e054c6ea 100644
>> --- a/kernel/irq/chip.c
>> +++ b/kernel/irq/chip.c
>> @@ -748,6 +748,8 @@ void handle_fasteoi_nmi(struct irq_desc *desc)
>>   	unsigned int irq = irq_desc_get_irq(desc);
>>   	irqreturn_t res;
>>   
>> +	kstat_incr_irqs_this_cpu(desc);
>> +
> 
> This needs to be called with the desc lock taken, otherwise we're likely
> to corrupt desc->tot_count.
> But taking the desc lock is something we can't do in NMI context (
> *spin_lock_irq*() won't prevent an NMI from happening).
> 
>>   	trace_irq_handler_entry(irq, action);
>>   	/*
>>   	 * NMIs cannot be shared, there is only one action.
>> @@ -962,6 +964,8 @@ void handle_percpu_devid_fasteoi_nmi(struct irq_desc *desc)
>>   	unsigned int irq = irq_desc_get_irq(desc);
>>   	irqreturn_t res;
>>   
>> +	__kstat_incr_irqs_this_cpu(desc);
>> +
> 
> Looking at handle_percpu_irq(), I think this might be acceptable. But
> does it make sense to only have kstats for percpu NMIs?
> 

It would be better to have stats for both.

handle_fasteoi_nmi() can use __kstat_incr_irqs_this_cpu() if below 
change can be added to kstat_irqs_cpu().

diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c
index a92b33593b8d..9484e88dabc2 100644
--- a/kernel/irq/irqdesc.c
+++ b/kernel/irq/irqdesc.c
@@ -950,6 +950,11 @@ unsigned int kstat_irqs_cpu(unsigned int irq, int cpu)
                         *per_cpu_ptr(desc->kstat_irqs, cpu) : 0;
  }

+static bool irq_is_nmi(struct irq_desc *desc)
+{
+       return desc->istate & IRQS_NMI;
+}
+
  /**
   * kstat_irqs - Get the statistics for an interrupt
   * @irq:       The interrupt number
@@ -967,7 +972,8 @@ unsigned int kstat_irqs(unsigned int irq)
         if (!desc || !desc->kstat_irqs)
                 return 0;
         if (!irq_settings_is_per_cpu_devid(desc) &&
-           !irq_settings_is_per_cpu(desc))
+           !irq_settings_is_per_cpu(desc) &&
+           !irq_is_nmi(desc))
             return desc->tot_count;

         for_each_possible_cpu(cpu)


Thomas,
Please suggest a better way if any.

Thanks,
Shijith

Re: [PATCH] genirq: update irq stats from NMI handlers

[Thomas Gleixner]

On Thu, 4 Jul 2019, Shijith Thotton wrote:
> On 7/4/19 12:13 AM, Julien Thierry wrote:
> > Looking at handle_percpu_irq(), I think this might be acceptable. But
> > does it make sense to only have kstats for percpu NMIs?
> > 
> 
> It would be better to have stats for both.
> 
> handle_fasteoi_nmi() can use __kstat_incr_irqs_this_cpu() if below 
> change can be added to kstat_irqs_cpu().
> 
> diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c
> index a92b33593b8d..9484e88dabc2 100644
> --- a/kernel/irq/irqdesc.c
> +++ b/kernel/irq/irqdesc.c
> @@ -950,6 +950,11 @@ unsigned int kstat_irqs_cpu(unsigned int irq, int cpu)
>                          *per_cpu_ptr(desc->kstat_irqs, cpu) : 0;
>   }
> 
> +static bool irq_is_nmi(struct irq_desc *desc)
> +{
> +       return desc->istate & IRQS_NMI;
> +}
> +
>   /**
>    * kstat_irqs - Get the statistics for an interrupt
>    * @irq:       The interrupt number
> @@ -967,7 +972,8 @@ unsigned int kstat_irqs(unsigned int irq)
>          if (!desc || !desc->kstat_irqs)
>                  return 0;
>          if (!irq_settings_is_per_cpu_devid(desc) &&
> -           !irq_settings_is_per_cpu(desc))
> +           !irq_settings_is_per_cpu(desc) &&
> +           !irq_is_nmi(desc))
>              return desc->tot_count;
> 
>          for_each_possible_cpu(cpu)
> 
> 
> Thomas,
> Please suggest a better way if any.

Looks good.

Thanks,

	tglx

Re: [PATCH] genirq: update irq stats from NMI handlers

[Shijith Thotton]

On 7/4/19 9:19 AM, Thomas Gleixner wrote:
> On Thu, 4 Jul 2019, Shijith Thotton wrote:
>> On 7/4/19 12:13 AM, Julien Thierry wrote:
>>> Looking at handle_percpu_irq(), I think this might be acceptable. But
>>> does it make sense to only have kstats for percpu NMIs?
>>>
>>
>> It would be better to have stats for both.
>>
>> handle_fasteoi_nmi() can use __kstat_incr_irqs_this_cpu() if below
>> change can be added to kstat_irqs_cpu().
>>
>> diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c
>> index a92b33593b8d..9484e88dabc2 100644
>> --- a/kernel/irq/irqdesc.c
>> +++ b/kernel/irq/irqdesc.c
>> @@ -950,6 +950,11 @@ unsigned int kstat_irqs_cpu(unsigned int irq, int cpu)
>>                           *per_cpu_ptr(desc->kstat_irqs, cpu) : 0;
>>    }
>>
>> +static bool irq_is_nmi(struct irq_desc *desc)
>> +{
>> +       return desc->istate & IRQS_NMI;
>> +}
>> +
>>    /**
>>     * kstat_irqs - Get the statistics for an interrupt
>>     * @irq:       The interrupt number
>> @@ -967,7 +972,8 @@ unsigned int kstat_irqs(unsigned int irq)
>>           if (!desc || !desc->kstat_irqs)
>>                   return 0;
>>           if (!irq_settings_is_per_cpu_devid(desc) &&
>> -           !irq_settings_is_per_cpu(desc))
>> +           !irq_settings_is_per_cpu(desc) &&
>> +           !irq_is_nmi(desc))
>>               return desc->tot_count;
>>
>>           for_each_possible_cpu(cpu)
>>
>>
>> Thomas,
>> Please suggest a better way if any.
> 
> Looks good.
> 

Thanks Thomas. Will share v2 with the changes.

Thanks,
Shijith

[PATCH v2] genirq: update irq stats from NMI handlers

[Shijith Thotton]

The NMI handlers handle_percpu_devid_fasteoi_nmi() and
handle_fasteoi_nmi() added by commit 2dcf1fbcad35 ("genirq: Provide NMI
handlers") do not update the interrupt counts. Due to that the NMI
interrupt count does not show up correctly in /proc/interrupts.

Update the functions to fix this. With this change, we can see stats of
the perf NMI interrupts on arm64.

As NMI handlers can't update tot_count in irq descriptor, kstat_irqs()
has been updated not to return tot_count for NMI.

Fixes: 2dcf1fbcad35 ("genirq: Provide NMI handlers")

Signed-off-by: Shijith Thotton <sthotton@marvell.com>
---
Changes since v1:
- Don't touch tot_count from NMI handler.
- Update kstat_irqs() to not return tot_count for NMI.

 kernel/irq/chip.c    | 4 ++++
 kernel/irq/irqdesc.c | 8 +++++++-
 2 files changed, 11 insertions(+), 1 deletion(-)

diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
index 29d6c7d070b4..04c850fb70cb 100644
--- a/kernel/irq/chip.c
+++ b/kernel/irq/chip.c
@@ -748,6 +748,8 @@ void handle_fasteoi_nmi(struct irq_desc *desc)
 	unsigned int irq = irq_desc_get_irq(desc);
 	irqreturn_t res;
 
+	__kstat_incr_irqs_this_cpu(desc);
+
 	trace_irq_handler_entry(irq, action);
 	/*
 	 * NMIs cannot be shared, there is only one action.
@@ -962,6 +964,8 @@ void handle_percpu_devid_fasteoi_nmi(struct irq_desc *desc)
 	unsigned int irq = irq_desc_get_irq(desc);
 	irqreturn_t res;
 
+	__kstat_incr_irqs_this_cpu(desc);
+
 	trace_irq_handler_entry(irq, action);
 	res = action->handler(irq, raw_cpu_ptr(action->percpu_dev_id));
 	trace_irq_handler_exit(irq, action, res);
diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c
index c52b737ab8e3..9149dde5a7b0 100644
--- a/kernel/irq/irqdesc.c
+++ b/kernel/irq/irqdesc.c
@@ -946,6 +946,11 @@ unsigned int kstat_irqs_cpu(unsigned int irq, int cpu)
 			*per_cpu_ptr(desc->kstat_irqs, cpu) : 0;
 }
 
+static bool irq_is_nmi(struct irq_desc *desc)
+{
+	return desc->istate & IRQS_NMI;
+}
+
 /**
  * kstat_irqs - Get the statistics for an interrupt
  * @irq:	The interrupt number
@@ -963,7 +968,8 @@ unsigned int kstat_irqs(unsigned int irq)
 	if (!desc || !desc->kstat_irqs)
 		return 0;
 	if (!irq_settings_is_per_cpu_devid(desc) &&
-	    !irq_settings_is_per_cpu(desc))
+	    !irq_settings_is_per_cpu(desc) &&
+	    !irq_is_nmi(desc))
 	    return desc->tot_count;
 
 	for_each_possible_cpu(cpu)
-- 
2.17.0

[tip:irq/core] genirq: Update irq stats from NMI handlers

[tip-bot for Shijith Thotton]

Commit-ID:  c09cb1293523dd786ae54a12fd88001542cba2f6
Gitweb:     https://git.kernel.org/tip/c09cb1293523dd786ae54a12fd88001542cba2f6
Author:     Shijith Thotton <sthotton@marvell.com>
AuthorDate: Fri, 5 Jul 2019 07:56:20 +0000
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Sat, 6 Jul 2019 10:40:19 +0200

genirq: Update irq stats from NMI handlers

The NMI handlers handle_percpu_devid_fasteoi_nmi() and handle_fasteoi_nmi()
do not update the interrupt counts. Due to that the NMI interrupt count
does not show up correctly in /proc/interrupts.

Add the statistics and treat the NMI handlers in the same way as per cpu
interrupts and prevent them from updating irq_desc::tot_count as this might
be corrupted due to concurrency.

[ tglx: Massaged changelog ]

Fixes: 2dcf1fbcad35 ("genirq: Provide NMI handlers")
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/1562313336-11888-1-git-send-email-sthotton@marvell.com
---
 kernel/irq/chip.c    | 4 ++++
 kernel/irq/irqdesc.c | 8 +++++++-
 2 files changed, 11 insertions(+), 1 deletion(-)

diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
index 29d6c7d070b4..04c850fb70cb 100644
--- a/kernel/irq/chip.c
+++ b/kernel/irq/chip.c
@@ -748,6 +748,8 @@ void handle_fasteoi_nmi(struct irq_desc *desc)
 	unsigned int irq = irq_desc_get_irq(desc);
 	irqreturn_t res;
 
+	__kstat_incr_irqs_this_cpu(desc);
+
 	trace_irq_handler_entry(irq, action);
 	/*
 	 * NMIs cannot be shared, there is only one action.
@@ -962,6 +964,8 @@ void handle_percpu_devid_fasteoi_nmi(struct irq_desc *desc)
 	unsigned int irq = irq_desc_get_irq(desc);
 	irqreturn_t res;
 
+	__kstat_incr_irqs_this_cpu(desc);
+
 	trace_irq_handler_entry(irq, action);
 	res = action->handler(irq, raw_cpu_ptr(action->percpu_dev_id));
 	trace_irq_handler_exit(irq, action, res);
diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c
index c52b737ab8e3..9149dde5a7b0 100644
--- a/kernel/irq/irqdesc.c
+++ b/kernel/irq/irqdesc.c
@@ -946,6 +946,11 @@ unsigned int kstat_irqs_cpu(unsigned int irq, int cpu)
 			*per_cpu_ptr(desc->kstat_irqs, cpu) : 0;
 }
 
+static bool irq_is_nmi(struct irq_desc *desc)
+{
+	return desc->istate & IRQS_NMI;
+}
+
 /**
  * kstat_irqs - Get the statistics for an interrupt
  * @irq:	The interrupt number
@@ -963,7 +968,8 @@ unsigned int kstat_irqs(unsigned int irq)
 	if (!desc || !desc->kstat_irqs)
 		return 0;
 	if (!irq_settings_is_per_cpu_devid(desc) &&
-	    !irq_settings_is_per_cpu(desc))
+	    !irq_settings_is_per_cpu(desc) &&
+	    !irq_is_nmi(desc))
 	    return desc->tot_count;
 
 	for_each_possible_cpu(cpu)