[PATCH] slub: correct return errno on slab_sysfs_init failure

[PATCH] slub: correct return errno on slab_sysfs_init failure

[Jeff Liu]

From: Jie Liu <jeff.liu@oracle.com>

Return ENOMEM instead of ENOSYS if slab_sysfs_init() failed

Signed-off-by: Jie Liu <jeff.liu@oracle.com>
---
 mm/slub.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/slub.c b/mm/slub.c
index 2b1ce69..75ca109 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -5304,7 +5304,7 @@ static int __init slab_sysfs_init(void)
 	if (!slab_kset) {
 		mutex_unlock(&slab_mutex);
 		printk(KERN_ERR "Cannot register slab subsystem.\n");
-		return -ENOSYS;
+		return -ENOMEM;
 	}
 
 	slab_state = FULL;
-- 
1.8.3.2

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] slub: correct return errno on slab_sysfs_init failure

[Pekka Enberg]

On Thu, Jun 12, 2014 at 3:56 PM, Jeff Liu <jeff.liu@oracle.com> wrote:
> From: Jie Liu <jeff.liu@oracle.com>
>
> Return ENOMEM instead of ENOSYS if slab_sysfs_init() failed
>
> Signed-off-by: Jie Liu <jeff.liu@oracle.com>
> ---
>  mm/slub.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/slub.c b/mm/slub.c
> index 2b1ce69..75ca109 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -5304,7 +5304,7 @@ static int __init slab_sysfs_init(void)
>         if (!slab_kset) {
>                 mutex_unlock(&slab_mutex);
>                 printk(KERN_ERR "Cannot register slab subsystem.\n");
> -               return -ENOSYS;
> +               return -ENOMEM;

What is the motivation for this change? AFAICT, kset_create_and_add()
can fail for other reasons than ENOMEM, no?

>         }
>
>         slab_state = FULL;

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] slub: correct return errno on slab_sysfs_init failure

[Christoph Lameter]

On Thu, 12 Jun 2014, Jeff Liu wrote:

> From: Jie Liu <jeff.liu@oracle.com>
>
> Return ENOMEM instead of ENOSYS if slab_sysfs_init() failed

The reason that I used ENOSYS there is that the whole sysfs portion of the
slab allocator will be disabled. Could be due to a number of issues since
kset_create_and_add() returns NULL for any error.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] slub: correct return errno on slab_sysfs_init failure

[Jeff Liu]

On 06/13/2014 14:57 PM, Pekka Enberg wrote:
> On Thu, Jun 12, 2014 at 3:56 PM, Jeff Liu <jeff.liu@oracle.com> wrote:
>> From: Jie Liu <jeff.liu@oracle.com>
>>
>> Return ENOMEM instead of ENOSYS if slab_sysfs_init() failed
>>
>> Signed-off-by: Jie Liu <jeff.liu@oracle.com>
>> ---
>>  mm/slub.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/mm/slub.c b/mm/slub.c
>> index 2b1ce69..75ca109 100644
>> --- a/mm/slub.c
>> +++ b/mm/slub.c
>> @@ -5304,7 +5304,7 @@ static int __init slab_sysfs_init(void)
>>         if (!slab_kset) {
>>                 mutex_unlock(&slab_mutex);
>>                 printk(KERN_ERR "Cannot register slab subsystem.\n");
>> -               return -ENOSYS;
>> +               return -ENOMEM;
> 
> What is the motivation for this change? AFAICT, kset_create_and_add()
> can fail for other reasons than ENOMEM, no?

Originally I'd like to make it consistent to most other subsystems which are
return ENOMEM in case of sysfs init failed, but yes, kset_register() can fail
due to different reaons...


Cheers,
-Jeff

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] slub: correct return errno on slab_sysfs_init failure

[Jeff Liu]

On 06/13/2014 23:53 PM, Christoph Lameter wrote:
> 
> On Thu, 12 Jun 2014, Jeff Liu wrote:
> 
>> From: Jie Liu <jeff.liu@oracle.com>
>>
>> Return ENOMEM instead of ENOSYS if slab_sysfs_init() failed
> 
> The reason that I used ENOSYS there is that the whole sysfs portion of the
> slab allocator will be disabled. Could be due to a number of issues since
> kset_create_and_add() returns NULL for any error.

Thanks for your clarification and sorry for the noise.

Cheers,
-Jeff

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] slub: correct return errno on slab_sysfs_init failure

[Christoph Lameter]

On Sat, 14 Jun 2014, Jeff Liu wrote:

> Thanks for your clarification and sorry for the noise.

Dont be worried. I am not sure anymore that this was such a wise move.
Maybe get kset_create_and_add to return an error code instead and return
that instead of -ENOSYS?


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] slub: correct return errno on slab_sysfs_init failure

[Jeff Liu]

On 06/16/2014 22:00 PM, Christoph Lameter wrote:
> On Sat, 14 Jun 2014, Jeff Liu wrote:
> 
>> Thanks for your clarification and sorry for the noise.
> 
> Dont be worried. I am not sure anymore that this was such a wise move.
> Maybe get kset_create_and_add to return an error code instead and return
> that instead of -ENOSYS?

Personally, I prefer to get kset_create_and_add() to return an error which
can reflect the actual cause of the failure given that kset_register() can
failed due to different reasons.  If so, however, looks we have to make a
certain amount of change for the existing modules which are support sysfs
since they all return -ENOMEM if kset_create_and_add() return NULL, maybe
this is inherited from samples/kobject/kset-example.c...


Cheers,
-Jeff

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] slub: correct return errno on slab_sysfs_init failure

[Christoph Lameter]

On Mon, 16 Jun 2014, Jeff Liu wrote:

> >
> > Dont be worried. I am not sure anymore that this was such a wise move.
> > Maybe get kset_create_and_add to return an error code instead and return
> > that instead of -ENOSYS?
>
> Personally, I prefer to get kset_create_and_add() to return an error which
> can reflect the actual cause of the failure given that kset_register() can
> failed due to different reasons.  If so, however, looks we have to make a
> certain amount of change for the existing modules which are support sysfs
> since they all return -ENOMEM if kset_create_and_add() return NULL, maybe
> this is inherited from samples/kobject/kset-example.c...

Probably. Could you come up with patchset to clean this up? ERR_PTR() can
be used to return an error code in a pointer value.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] slub: correct return errno on slab_sysfs_init failure

[Jeff Liu]

On 06/16/2014 23:25 PM, Christoph Lameter wrote:
> On Mon, 16 Jun 2014, Jeff Liu wrote:
> 
>>>
>>> Dont be worried. I am not sure anymore that this was such a wise move.
>>> Maybe get kset_create_and_add to return an error code instead and return
>>> that instead of -ENOSYS?
>>
>> Personally, I prefer to get kset_create_and_add() to return an error which
>> can reflect the actual cause of the failure given that kset_register() can
>> failed due to different reasons.  If so, however, looks we have to make a
>> certain amount of change for the existing modules which are support sysfs
>> since they all return -ENOMEM if kset_create_and_add() return NULL, maybe
>> this is inherited from samples/kobject/kset-example.c...
> 
> Probably. Could you come up with patchset to clean this up? ERR_PTR() can
> be used to return an error code in a pointer value.

Sure, I'll work it out ASAP.


Cheers,
-Jeff

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>