[PATCH] mm/slub: clean up create_unique_id()

[PATCH] mm/slub: clean up create_unique_id()

[Chao Yu]

From: Chao Yu <chao.yu@oppo.com>

As Christophe JAILLET suggested:

In create_unique_id(),

"looks that ID_STR_LENGTH could even be reduced to 32 or 16.

The 2nd BUG_ON at the end of the function could certainly be just
removed as well or remplaced by a:
        if (p > name + ID_STR_LENGTH - 1) {
                kfree(name);
                return -E<something>;
        }
"

According to above suggestion, let's do below cleanups:
1. reduce ID_STR_LENGTH to 32, as the buffer size should be enough;
2. remove BUG_ON() and return error if check condition is true.

Link: https://lore.kernel.org/linux-mm/2025305d-16db-abdf-6cd3-1fb93371c2b4@wanadoo.fr/
Suggested-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Fixes: 81819f0fc828 ("SLUB core")
Signed-off-by: Chao Yu <chao.yu@oppo.com>
---
 mm/slub.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/mm/slub.c b/mm/slub.c
index 5ba6db62a5ab..a045c1ca8772 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -5883,7 +5883,7 @@ static inline struct kset *cache_kset(struct kmem_cache *s)
 	return slab_kset;
 }
 
-#define ID_STR_LENGTH 64
+#define ID_STR_LENGTH 32
 
 /* Create a unique string id for a slab cache:
  *
@@ -5919,7 +5919,10 @@ static char *create_unique_id(struct kmem_cache *s)
 		*p++ = '-';
 	p += sprintf(p, "%07u", s->size);
 
-	BUG_ON(p > name + ID_STR_LENGTH - 1);
+	if (p > name + ID_STR_LENGTH - 1) {
+		kfree(name);
+		return ERR_PTR(-EINVAL);
+	}
 	return name;
 }
 
-- 
2.25.1

Re: [PATCH] mm/slub: clean up create_unique_id()

[Hyeonggon Yoo]

On Sun, Sep 18, 2022 at 05:21:46PM +0800, Chao Yu wrote:
> From: Chao Yu <chao.yu@oppo.com>
> 
> As Christophe JAILLET suggested:
> 
> In create_unique_id(),
> 
> "looks that ID_STR_LENGTH could even be reduced to 32 or 16.
> 
> The 2nd BUG_ON at the end of the function could certainly be just
> removed as well or remplaced by a:
>         if (p > name + ID_STR_LENGTH - 1) {
>                 kfree(name);
>                 return -E<something>;
>         }
> "
> 
> According to above suggestion, let's do below cleanups:
> 1. reduce ID_STR_LENGTH to 32, as the buffer size should be enough;
> 2. remove BUG_ON() and return error if check condition is true.
> 
> Link: https://lore.kernel.org/linux-mm/2025305d-16db-abdf-6cd3-1fb93371c2b4@wanadoo.fr/
> Suggested-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> Fixes: 81819f0fc828 ("SLUB core")
> Signed-off-by: Chao Yu <chao.yu@oppo.com>
> ---
>  mm/slub.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/slub.c b/mm/slub.c
> index 5ba6db62a5ab..a045c1ca8772 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -5883,7 +5883,7 @@ static inline struct kset *cache_kset(struct kmem_cache *s)
>  	return slab_kset;
>  }
>  
> -#define ID_STR_LENGTH 64
> +#define ID_STR_LENGTH 32
>  
>  /* Create a unique string id for a slab cache:
>   *
> @@ -5919,7 +5919,10 @@ static char *create_unique_id(struct kmem_cache *s)
>  		*p++ = '-';
>  	p += sprintf(p, "%07u", s->size);
>  
> -	BUG_ON(p > name + ID_STR_LENGTH - 1);
> +	if (p > name + ID_STR_LENGTH - 1) {
> +		kfree(name);
> +		return ERR_PTR(-EINVAL);
> +	}
>  	return name;

IIUC 16 bytes will be still safe ;)

Anyway, for either size:

Looks good to me,
Reviewed-by: Hyeonggon Yoo <42.hyeyoo@gmail.com>

Thanks!

>  }
>  
> -- 
> 2.25.1
> 

-- 
Thanks,
Hyeonggon

Re: [PATCH] mm/slub: clean up create_unique_id()

[Vlastimil Babka]

On 9/18/22 11:21, Chao Yu wrote:
> From: Chao Yu <chao.yu@oppo.com>
> 
> As Christophe JAILLET suggested:
> 
> In create_unique_id(),
> 
> "looks that ID_STR_LENGTH could even be reduced to 32 or 16.
> 
> The 2nd BUG_ON at the end of the function could certainly be just
> removed as well or remplaced by a:
>         if (p > name + ID_STR_LENGTH - 1) {
>                 kfree(name);
>                 return -E<something>;
>         }
> "
> 
> According to above suggestion, let's do below cleanups:
> 1. reduce ID_STR_LENGTH to 32, as the buffer size should be enough;
> 2. remove BUG_ON() and return error if check condition is true.

I'd leave a WARN_ON there as we really don't expect this to happen, so if it
does, we should be loud about it and not silently fail.

> Link: https://lore.kernel.org/linux-mm/2025305d-16db-abdf-6cd3-1fb93371c2b4@wanadoo.fr/
> Suggested-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> Fixes: 81819f0fc828 ("SLUB core")
> Signed-off-by: Chao Yu <chao.yu@oppo.com>
> ---
>  mm/slub.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/slub.c b/mm/slub.c
> index 5ba6db62a5ab..a045c1ca8772 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -5883,7 +5883,7 @@ static inline struct kset *cache_kset(struct kmem_cache *s)
>  	return slab_kset;
>  }
>  
> -#define ID_STR_LENGTH 64
> +#define ID_STR_LENGTH 32
>  
>  /* Create a unique string id for a slab cache:
>   *
> @@ -5919,7 +5919,10 @@ static char *create_unique_id(struct kmem_cache *s)
>  		*p++ = '-';
>  	p += sprintf(p, "%07u", s->size);

Hm but sprintf() will happily overflow, so if we only detect that
afterwards, it's kinda too late to gracefully fail.
Should use snprintf() then?

> -	BUG_ON(p > name + ID_STR_LENGTH - 1);
> +	if (p > name + ID_STR_LENGTH - 1) {
> +		kfree(name);
> +		return ERR_PTR(-EINVAL);
> +	}
>  	return name;
>  }
>  

Re: [PATCH] mm/slub: clean up create_unique_id()

[Chao Yu]

On 2022/9/23 4:47, Vlastimil Babka wrote:
> On 9/18/22 11:21, Chao Yu wrote:
>> From: Chao Yu <chao.yu@oppo.com>
>>
>> As Christophe JAILLET suggested:
>>
>> In create_unique_id(),
>>
>> "looks that ID_STR_LENGTH could even be reduced to 32 or 16.
>>
>> The 2nd BUG_ON at the end of the function could certainly be just
>> removed as well or remplaced by a:
>>          if (p > name + ID_STR_LENGTH - 1) {
>>                  kfree(name);
>>                  return -E<something>;
>>          }
>> "
>>
>> According to above suggestion, let's do below cleanups:
>> 1. reduce ID_STR_LENGTH to 32, as the buffer size should be enough;
>> 2. remove BUG_ON() and return error if check condition is true.
> 
> I'd leave a WARN_ON there as we really don't expect this to happen, so if it
> does, we should be loud about it and not silently fail.

Agreed.

> 
>> Link: https://lore.kernel.org/linux-mm/2025305d-16db-abdf-6cd3-1fb93371c2b4@wanadoo.fr/
>> Suggested-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
>> Fixes: 81819f0fc828 ("SLUB core")
>> Signed-off-by: Chao Yu <chao.yu@oppo.com>
>> ---
>>   mm/slub.c | 7 +++++--
>>   1 file changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/mm/slub.c b/mm/slub.c
>> index 5ba6db62a5ab..a045c1ca8772 100644
>> --- a/mm/slub.c
>> +++ b/mm/slub.c
>> @@ -5883,7 +5883,7 @@ static inline struct kset *cache_kset(struct kmem_cache *s)
>>   	return slab_kset;
>>   }
>>   
>> -#define ID_STR_LENGTH 64
>> +#define ID_STR_LENGTH 32
>>   
>>   /* Create a unique string id for a slab cache:
>>    *
>> @@ -5919,7 +5919,10 @@ static char *create_unique_id(struct kmem_cache *s)
>>   		*p++ = '-';
>>   	p += sprintf(p, "%07u", s->size);
> 
> Hm but sprintf() will happily overflow, so if we only detect that
> afterwards, it's kinda too late to gracefully fail.
> Should use snprintf() then?

Correct.

Updated in v2.

Thanks,

> 
>> -	BUG_ON(p > name + ID_STR_LENGTH - 1);
>> +	if (p > name + ID_STR_LENGTH - 1) {
>> +		kfree(name);
>> +		return ERR_PTR(-EINVAL);
>> +	}
>>   	return name;
>>   }
>>   
>