Re: [RFC][PATCH] vmscan: balance local_irq_disable() and local_irq_enable()

Re: [RFC][PATCH] vmscan: balance local_irq_disable() and local_irq_enable()

[Steven Rostedt]

t On Wed, 2010-02-03 at 20:53 +0100, John Kacur wrote:
> Balance local_irq_disable() and local_irq_enable() as well as
> spin_lock_irq() and spin_lock_unlock_irq
> 
> Signed-off-by: John Kacur <jkacur@redhat.com>
> ---
>  mm/vmscan.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index c26986c..b895025 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1200,8 +1200,9 @@ static unsigned long shrink_inactive_list(unsigned long max_scan,
>  		if (current_is_kswapd())
>  			__count_vm_events(KSWAPD_STEAL, nr_freed);
>  		__count_zone_vm_events(PGSTEAL, zone, nr_freed);
> +		local_irq_enable();
>  
> -		spin_lock(&zone->lru_lock);
> +		spin_lock_irq(&zone->lru_lock);
>  		/*
>  		 * Put back any unfreeable pages.
>  		 */


The above looks wrong. I don't know the code, but just by looking at
where the locking and interrupts are, I can take a guess.

Lets add a little more of the code:

                local_irq_disable();
                if (current_is_kswapd())
                        __count_vm_events(KSWAPD_STEAL, nr_freed);
                __count_zone_vm_events(PGSTEAL, zone, nr_freed);

                spin_lock(&zone->lru_lock);
                /*

I'm guessing the __count_zone_vm_events and friends need interrupts
disabled here, probably due to per cpu stuff. But if you enable
interrupts before the spin_lock() you may let an interrupt come in and
invalidate what was done above it.

So no, I do not think enabling interrupts here is a good thing.

-- Steve

Re: [RFC][PATCH] vmscan: balance local_irq_disable() and local_irq_enable()

[John Kacur]

On Wed, Feb 3, 2010 at 9:09 PM, Steven Rostedt <rostedt@goodmis.org> wrote:
> t On Wed, 2010-02-03 at 20:53 +0100, John Kacur wrote:
>> Balance local_irq_disable() and local_irq_enable() as well as
>> spin_lock_irq() and spin_lock_unlock_irq
>>
>> Signed-off-by: John Kacur <jkacur@redhat.com>
>> ---
>>  mm/vmscan.c |    3 ++-
>>  1 files changed, 2 insertions(+), 1 deletions(-)
>>
>> diff --git a/mm/vmscan.c b/mm/vmscan.c
>> index c26986c..b895025 100644
>> --- a/mm/vmscan.c
>> +++ b/mm/vmscan.c
>> @@ -1200,8 +1200,9 @@ static unsigned long shrink_inactive_list(unsigned long max_scan,
>>               if (current_is_kswapd())
>>                       __count_vm_events(KSWAPD_STEAL, nr_freed);
>>               __count_zone_vm_events(PGSTEAL, zone, nr_freed);
>> +             local_irq_enable();
>>
>> -             spin_lock(&zone->lru_lock);
>> +             spin_lock_irq(&zone->lru_lock);
>>               /*
>>                * Put back any unfreeable pages.
>>                */
>
>
> The above looks wrong. I don't know the code, but just by looking at
> where the locking and interrupts are, I can take a guess.
>
> Lets add a little more of the code:
>
>                local_irq_disable();
>                if (current_is_kswapd())
>                        __count_vm_events(KSWAPD_STEAL, nr_freed);
>                __count_zone_vm_events(PGSTEAL, zone, nr_freed);
>
>                spin_lock(&zone->lru_lock);
>                /*
>
> I'm guessing the __count_zone_vm_events and friends need interrupts
> disabled here, probably due to per cpu stuff. But if you enable
> interrupts before the spin_lock() you may let an interrupt come in and
> invalidate what was done above it.
>
> So no, I do not think enabling interrupts here is a good thing.
>

okay, and since we have already done local_irq_disable(), then that is
why we only need the spin_lock() and not the spin_lock_irq() flavour?

Re: [RFC][PATCH] vmscan: balance local_irq_disable() and local_irq_enable()

[KOSAKI Motohiro]

> On Wed, Feb 3, 2010 at 9:09 PM, Steven Rostedt <rostedt@goodmis.org> wrote:
> > t On Wed, 2010-02-03 at 20:53 +0100, John Kacur wrote:
> >> Balance local_irq_disable() and local_irq_enable() as well as
> >> spin_lock_irq() and spin_lock_unlock_irq
> >>
> >> Signed-off-by: John Kacur <jkacur@redhat.com>
> >> ---
> >>  mm/vmscan.c |    3 ++-
> >>  1 files changed, 2 insertions(+), 1 deletions(-)
> >>
> >> diff --git a/mm/vmscan.c b/mm/vmscan.c
> >> index c26986c..b895025 100644
> >> --- a/mm/vmscan.c
> >> +++ b/mm/vmscan.c
> >> @@ -1200,8 +1200,9 @@ static unsigned long shrink_inactive_list(unsigned long max_scan,
> >>               if (current_is_kswapd())
> >>                       __count_vm_events(KSWAPD_STEAL, nr_freed);
> >>               __count_zone_vm_events(PGSTEAL, zone, nr_freed);
> >> +             local_irq_enable();
> >>
> >> -             spin_lock(&zone->lru_lock);
> >> +             spin_lock_irq(&zone->lru_lock);
> >>               /*
> >>                * Put back any unfreeable pages.
> >>                */
> >
> >
> > The above looks wrong. I don't know the code, but just by looking at
> > where the locking and interrupts are, I can take a guess.
> >
> > Lets add a little more of the code:
> >
> >                local_irq_disable();
> >                if (current_is_kswapd())
> >                        __count_vm_events(KSWAPD_STEAL, nr_freed);
> >                __count_zone_vm_events(PGSTEAL, zone, nr_freed);
> >
> >                spin_lock(&zone->lru_lock);
> >                /*
> >
> > I'm guessing the __count_zone_vm_events and friends need interrupts
> > disabled here, probably due to per cpu stuff. But if you enable
> > interrupts before the spin_lock() you may let an interrupt come in and
> > invalidate what was done above it.
> >
> > So no, I do not think enabling interrupts here is a good thing.
> >
> 
> okay, and since we have already done local_irq_disable(), then that is
> why we only need the spin_lock() and not the spin_lock_irq() flavour?

Yes, spin_lock_irq() is equivalent to spin_lock() + irq_disable().
Now, we already disabled irq. then, we only need spin_lock().

So, I don't think shrink_inactive_list need any fix.

Re: [RFC][PATCH] vmscan: balance local_irq_disable() and local_irq_enable()

[John Kacur]

On Thu, Feb 4, 2010 at 1:22 AM, KOSAKI Motohiro
<kosaki.motohiro@jp.fujitsu.com> wrote:
>> On Wed, Feb 3, 2010 at 9:09 PM, Steven Rostedt <rostedt@goodmis.org> wrote:
>> > t On Wed, 2010-02-03 at 20:53 +0100, John Kacur wrote:
>> >> Balance local_irq_disable() and local_irq_enable() as well as
>> >> spin_lock_irq() and spin_lock_unlock_irq
>> >>
>> >> Signed-off-by: John Kacur <jkacur@redhat.com>
>> >> ---
>> >>  mm/vmscan.c |    3 ++-
>> >>  1 files changed, 2 insertions(+), 1 deletions(-)
>> >>
>> >> diff --git a/mm/vmscan.c b/mm/vmscan.c
>> >> index c26986c..b895025 100644
>> >> --- a/mm/vmscan.c
>> >> +++ b/mm/vmscan.c
>> >> @@ -1200,8 +1200,9 @@ static unsigned long shrink_inactive_list(unsigned long max_scan,
>> >>               if (current_is_kswapd())
>> >>                       __count_vm_events(KSWAPD_STEAL, nr_freed);
>> >>               __count_zone_vm_events(PGSTEAL, zone, nr_freed);
>> >> +             local_irq_enable();
>> >>
>> >> -             spin_lock(&zone->lru_lock);
>> >> +             spin_lock_irq(&zone->lru_lock);
>> >>               /*
>> >>                * Put back any unfreeable pages.
>> >>                */
>> >
>> >
>> > The above looks wrong. I don't know the code, but just by looking at
>> > where the locking and interrupts are, I can take a guess.
>> >
>> > Lets add a little more of the code:
>> >
>> >                local_irq_disable();
>> >                if (current_is_kswapd())
>> >                        __count_vm_events(KSWAPD_STEAL, nr_freed);
>> >                __count_zone_vm_events(PGSTEAL, zone, nr_freed);
>> >
>> >                spin_lock(&zone->lru_lock);
>> >                /*
>> >
>> > I'm guessing the __count_zone_vm_events and friends need interrupts
>> > disabled here, probably due to per cpu stuff. But if you enable
>> > interrupts before the spin_lock() you may let an interrupt come in and
>> > invalidate what was done above it.
>> >
>> > So no, I do not think enabling interrupts here is a good thing.
>> >
>>
>> okay, and since we have already done local_irq_disable(), then that is
>> why we only need the spin_lock() and not the spin_lock_irq() flavour?
>
> Yes, spin_lock_irq() is equivalent to spin_lock() + irq_disable().
> Now, we already disabled irq. then, we only need spin_lock().
>
> So, I don't think shrink_inactive_list need any fix.
>

Thanks for the explanation!