[PATCH] lib/flex_proportions.c: Remove local_irq_ops in fprop_new_period()

wuchi posted 1 patch 3 years, 10 months ago
lib/flex_proportions.c | 10 ++--------
1 file changed, 2 insertions(+), 8 deletions(-)
[PATCH] lib/flex_proportions.c: Remove local_irq_ops in fprop_new_period()
Posted by wuchi 3 years, 10 months ago
The commit <e78d4833c03e28> (lib: Fix possible deadlock in flexible
proportion code) adds the local_irq_ops because percpu_counter_{sum
|add} ops'lock can cause deadlock by interrupts. Now percpu_counter
_{sum|add} ops use raw_spin_(un)lock_irq*, so revert the commit and
resolve the conflict.

Signed-off-by: wuchi <wuchi.zero@gmail.com>
---
 lib/flex_proportions.c | 10 ++--------
 1 file changed, 2 insertions(+), 8 deletions(-)

diff --git a/lib/flex_proportions.c b/lib/flex_proportions.c
index 53e7eb1dd76c..05cccbcf1661 100644
--- a/lib/flex_proportions.c
+++ b/lib/flex_proportions.c
@@ -63,18 +63,13 @@ void fprop_global_destroy(struct fprop_global *p)
  */
 bool fprop_new_period(struct fprop_global *p, int periods)
 {
-	s64 events;
-	unsigned long flags;
+	s64 events = percpu_counter_sum(&p->events);
 
-	local_irq_save(flags);
-	events = percpu_counter_sum(&p->events);
 	/*
 	 * Don't do anything if there are no events.
 	 */
-	if (events <= 1) {
-		local_irq_restore(flags);
+	if (events <= 1)
 		return false;
-	}
 	write_seqcount_begin(&p->sequence);
 	if (periods < 64)
 		events -= events >> periods;
@@ -82,7 +77,6 @@ bool fprop_new_period(struct fprop_global *p, int periods)
 	percpu_counter_add(&p->events, -events);
 	p->period += periods;
 	write_seqcount_end(&p->sequence);
-	local_irq_restore(flags);
 
 	return true;
 }
-- 
2.20.1
Re: [PATCH] lib/flex_proportions.c: Remove local_irq_ops in fprop_new_period()
Posted by Jan Kara 3 years, 10 months ago
On Sat 04-06-22 21:15:02, wuchi wrote:
> The commit <e78d4833c03e28> (lib: Fix possible deadlock in flexible
> proportion code) adds the local_irq_ops because percpu_counter_{sum
> |add} ops'lock can cause deadlock by interrupts. Now percpu_counter
> _{sum|add} ops use raw_spin_(un)lock_irq*, so revert the commit and
> resolve the conflict.
> 
> Signed-off-by: wuchi <wuchi.zero@gmail.com>

Yes, good point. Thanks for cleaning this up. Feel free to add:

Reviewed-by: Jan Kara <jack@suse.cz>

I guess Andrew will pickup this patch, or should I do it Andrew?

								Honza

> ---
>  lib/flex_proportions.c | 10 ++--------
>  1 file changed, 2 insertions(+), 8 deletions(-)
> 
> diff --git a/lib/flex_proportions.c b/lib/flex_proportions.c
> index 53e7eb1dd76c..05cccbcf1661 100644
> --- a/lib/flex_proportions.c
> +++ b/lib/flex_proportions.c
> @@ -63,18 +63,13 @@ void fprop_global_destroy(struct fprop_global *p)
>   */
>  bool fprop_new_period(struct fprop_global *p, int periods)
>  {
> -	s64 events;
> -	unsigned long flags;
> +	s64 events = percpu_counter_sum(&p->events);
>  
> -	local_irq_save(flags);
> -	events = percpu_counter_sum(&p->events);
>  	/*
>  	 * Don't do anything if there are no events.
>  	 */
> -	if (events <= 1) {
> -		local_irq_restore(flags);
> +	if (events <= 1)
>  		return false;
> -	}
>  	write_seqcount_begin(&p->sequence);
>  	if (periods < 64)
>  		events -= events >> periods;
> @@ -82,7 +77,6 @@ bool fprop_new_period(struct fprop_global *p, int periods)
>  	percpu_counter_add(&p->events, -events);
>  	p->period += periods;
>  	write_seqcount_end(&p->sequence);
> -	local_irq_restore(flags);
>  
>  	return true;
>  }
> -- 
> 2.20.1
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR
Re: [PATCH] lib/flex_proportions.c: Remove local_irq_ops in fprop_new_period()
Posted by Jan Kara 3 years, 10 months ago
On Mon 06-06-22 09:44:44, Jan Kara wrote:
> On Sat 04-06-22 21:15:02, wuchi wrote:
> > The commit <e78d4833c03e28> (lib: Fix possible deadlock in flexible
> > proportion code) adds the local_irq_ops because percpu_counter_{sum
> > |add} ops'lock can cause deadlock by interrupts. Now percpu_counter
> > _{sum|add} ops use raw_spin_(un)lock_irq*, so revert the commit and
> > resolve the conflict.
> > 
> > Signed-off-by: wuchi <wuchi.zero@gmail.com>
> 
> Yes, good point. Thanks for cleaning this up. Feel free to add:
> 
> Reviewed-by: Jan Kara <jack@suse.cz>
> 
> I guess Andrew will pickup this patch, or should I do it Andrew?

Oh, I can see Andrew already did pickup the patch. So everything is fine
:).
								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR