[PATCH 2/2] sched: use slab in cpupri_init()

[PATCH 2/2] sched: use slab in cpupri_init()

[Pekka J Enberg]

From: Pekka Enberg <penberg@cs.helsinki.fi>

Lets not use the bootmem allocator in cpupri_init() as slab is already up when
it is run.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
---
 kernel/sched.c        |    2 +-
 kernel/sched_cpupri.c |    8 +++++---
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index 12cc09c..dcf2dc2 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -7795,7 +7795,7 @@ static int __init_refok init_rootdomain(struct root_domain *rd, bool bootmem)
 	if (!alloc_cpumask_var(&rd->rto_mask, gfp))
 		goto free_online;
 
-	if (cpupri_init(&rd->cpupri, false) != 0)
+	if (cpupri_init(&rd->cpupri, bootmem) != 0)
 		goto free_rto_mask;
 	return 0;
 
diff --git a/kernel/sched_cpupri.c b/kernel/sched_cpupri.c
index 344712a..7deffc9 100644
--- a/kernel/sched_cpupri.c
+++ b/kernel/sched_cpupri.c
@@ -154,8 +154,12 @@ void cpupri_set(struct cpupri *cp, int cpu, int newpri)
  */
 int __init_refok cpupri_init(struct cpupri *cp, bool bootmem)
 {
+	gfp_t gfp = GFP_KERNEL;
 	int i;
 
+	if (bootmem)
+		gfp = GFP_NOWAIT;
+
 	memset(cp, 0, sizeof(*cp));
 
 	for (i = 0; i < CPUPRI_NR_PRIORITIES; i++) {
@@ -163,9 +167,7 @@ int __init_refok cpupri_init(struct cpupri *cp, bool bootmem)
 
 		spin_lock_init(&vec->lock);
 		vec->count = 0;
-		if (bootmem)
-			alloc_bootmem_cpumask_var(&vec->mask);
-		else if (!zalloc_cpumask_var(&vec->mask, GFP_KERNEL))
+		if (!zalloc_cpumask_var(&vec->mask, gfp))
 			goto cleanup;
 	}
 
-- 
1.6.0.4

Re: [PATCH 2/2] sched: use slab in cpupri_init()

[Li Zefan]

> @@ -163,9 +167,7 @@ int __init_refok cpupri_init(struct cpupri *cp, bool bootmem)

so the __init_refok tag can be removed, no?

>  
>  		spin_lock_init(&vec->lock);
>  		vec->count = 0;
> -		if (bootmem)
> -			alloc_bootmem_cpumask_var(&vec->mask);
> -		else if (!zalloc_cpumask_var(&vec->mask, GFP_KERNEL))
> +		if (!zalloc_cpumask_var(&vec->mask, gfp))
>  			goto cleanup;
>  	}
>  

Re: [PATCH 2/2] sched: use slab in cpupri_init()

[Pekka Enberg]

Hi Li!

On Thu, 2009-06-11 at 14:10 +0800, Li Zefan wrote:
> > @@ -163,9 +167,7 @@ int __init_refok cpupri_init(struct cpupri *cp, bool bootmem)
> 
> so the __init_refok tag can be removed, no?

I am not sure how this is related to my patch but as long as
init_rootdomain() has it, cpupri_init() should too AFAICT.

			Pekka

> >  
> >  		spin_lock_init(&vec->lock);
> >  		vec->count = 0;
> > -		if (bootmem)
> > -			alloc_bootmem_cpumask_var(&vec->mask);
> > -		else if (!zalloc_cpumask_var(&vec->mask, GFP_KERNEL))
> > +		if (!zalloc_cpumask_var(&vec->mask, gfp))
> >  			goto cleanup;
> >  	}
> >  

Re: [PATCH 2/2] sched: use slab in cpupri_init()

[Li Zefan]

Pekka Enberg wrote:
> Hi Li!
> 
> On Thu, 2009-06-11 at 14:10 +0800, Li Zefan wrote:
>>> @@ -163,9 +167,7 @@ int __init_refok cpupri_init(struct cpupri *cp, bool bootmem)
>> so the __init_refok tag can be removed, no?
> 
> I am not sure how this is related to my patch but as long as
> init_rootdomain() has it, cpupri_init() should too AFAICT.
> 

cpupri_init() is marked as __init_refok, because it calls
alloc_bootmem_cpumask_var() not because init_rootdomain()
has the tag. Now this patch removes this function call,
so the tag is no long needed.

> 
>>>  
>>>  		spin_lock_init(&vec->lock);
>>>  		vec->count = 0;
>>> -		if (bootmem)
>>> -			alloc_bootmem_cpumask_var(&vec->mask);
>>> -		else if (!zalloc_cpumask_var(&vec->mask, GFP_KERNEL))
>>> +		if (!zalloc_cpumask_var(&vec->mask, gfp))
>>>  			goto cleanup;
>>>  	}
>>>  
> 

Re: [PATCH 2/2] sched: use slab in cpupri_init()

[Pekka Enberg]

On Thu, 2009-06-11 at 15:37 +0800, Li Zefan wrote:
> cpupri_init() is marked as __init_refok, because it calls
> alloc_bootmem_cpumask_var() not because init_rootdomain()
> has the tag. Now this patch removes this function call,
> so the tag is no long needed.

Aah, OK, makes sense. Care to send a patch to Ingo once we have this
thing merged?

			Pekka

Re: [PATCH 2/2] sched: use slab in cpupri_init()

[Li Zefan]

Pekka Enberg wrote:
> On Thu, 2009-06-11 at 15:37 +0800, Li Zefan wrote:
>> cpupri_init() is marked as __init_refok, because it calls
>> alloc_bootmem_cpumask_var() not because init_rootdomain()
>> has the tag. Now this patch removes this function call,
>> so the tag is no long needed.
> 
> Aah, OK, makes sense. Care to send a patch to Ingo once we have this
> thing merged?
> 

OK. Actually it's me that added it. ;)
(see db2f59c8c9b315f2b88b1dac159b988c6009034d)