[PATCH] kobject: Fix kobject_set_name_vargs()

[PATCH] kobject: Fix kobject_set_name_vargs()

[Eric Dumazet]

In case of kvasprintf() failure, we can leak old kobject name.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
---
diff --git a/lib/kobject.c b/lib/kobject.c
index 8115eb1..1247c57 100644
--- a/lib/kobject.c
+++ b/lib/kobject.c
@@ -222,8 +222,10 @@ int kobject_set_name_vargs(struct kobject *kobj, const char *fmt,
 		return 0;
 
 	kobj->name = kvasprintf(GFP_KERNEL, fmt, vargs);
-	if (!kobj->name)
+	if (!kobj->name) {
+		kobj->name = old_name;
 		return -ENOMEM;
+	}
 
 	/* ewww... some of these buggers have '/' in the name ... */
 	while ((s = strchr(kobj->name, '/')))

Re: [PATCH] kobject: Fix kobject_set_name_vargs()

[Greg KH]

On Thu, Mar 11, 2010 at 09:50:52PM +0100, Eric Dumazet wrote:
> In case of kvasprintf() failure, we can leak old kobject name.
> 
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> ---
> diff --git a/lib/kobject.c b/lib/kobject.c
> index 8115eb1..1247c57 100644
> --- a/lib/kobject.c
> +++ b/lib/kobject.c
> @@ -222,8 +222,10 @@ int kobject_set_name_vargs(struct kobject *kobj, const char *fmt,
>  		return 0;
>  
>  	kobj->name = kvasprintf(GFP_KERNEL, fmt, vargs);
> -	if (!kobj->name)
> +	if (!kobj->name) {
> +		kobj->name = old_name;
>  		return -ENOMEM;
> +	}

Are you sure?  I think we've been over this very thing many times in the
past...

Kay, I can't recall the issue here, can you?

thanks,

greg k-h

Re: [PATCH] kobject: Fix kobject_set_name_vargs()

[Kay Sievers]

On Fri, Mar 12, 2010 at 05:15, Greg KH <greg@kroah.com> wrote:
> On Thu, Mar 11, 2010 at 09:50:52PM +0100, Eric Dumazet wrote:
>> In case of kvasprintf() failure, we can leak old kobject name.
>>
>> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
>> ---
>> diff --git a/lib/kobject.c b/lib/kobject.c
>> index 8115eb1..1247c57 100644
>> --- a/lib/kobject.c
>> +++ b/lib/kobject.c
>> @@ -222,8 +222,10 @@ int kobject_set_name_vargs(struct kobject *kobj, const char *fmt,
>>               return 0;
>>
>>       kobj->name = kvasprintf(GFP_KERNEL, fmt, vargs);
>> -     if (!kobj->name)
>> +     if (!kobj->name) {
>> +             kobj->name = old_name;
>>               return -ENOMEM;
>> +     }
>
> Are you sure?  I think we've been over this very thing many times in the
> past...
>
> Kay, I can't recall the issue here, can you?

I think we just lost attention to it, and never made a final decision.
There have been some issues, I also don't really remember.

I myself have an unsubmitted patch for this problem here since 9 months:
  http://git.kernel.org/?p=linux/kernel/git/kay/patches.git;a=blob;f=kobj-leak.patch;hb=HEAD

Seems, we should start thinking about the problem again. And if we
decide to do nothing at least add a comment why we don't. :)

Thanks,
Kay

Re: [PATCH] kobject: Fix kobject_set_name_vargs()

[Kay Sievers]

On Fri, Mar 12, 2010 at 05:39, Kay Sievers <kay.sievers@vrfy.org> wrote:
> On Fri, Mar 12, 2010 at 05:15, Greg KH <greg@kroah.com> wrote:
>> On Thu, Mar 11, 2010 at 09:50:52PM +0100, Eric Dumazet wrote:
>>> In case of kvasprintf() failure, we can leak old kobject name.
>>>
>>> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
>>> ---
>>> diff --git a/lib/kobject.c b/lib/kobject.c
>>> index 8115eb1..1247c57 100644
>>> --- a/lib/kobject.c
>>> +++ b/lib/kobject.c
>>> @@ -222,8 +222,10 @@ int kobject_set_name_vargs(struct kobject *kobj, const char *fmt,
>>>               return 0;
>>>
>>>       kobj->name = kvasprintf(GFP_KERNEL, fmt, vargs);
>>> -     if (!kobj->name)
>>> +     if (!kobj->name) {
>>> +             kobj->name = old_name;
>>>               return -ENOMEM;
>>> +     }
>>
>> Are you sure?  I think we've been over this very thing many times in the
>> past...
>>
>> Kay, I can't recall the issue here, can you?
>
> I think we just lost attention to it, and never made a final decision.
> There have been some issues, I also don't really remember.
>
> I myself have an unsubmitted patch for this problem here since 9 months:
>  http://git.kernel.org/?p=linux/kernel/git/kay/patches.git;a=blob;f=kobj-leak.patch;hb=HEAD
>
> Seems, we should start thinking about the problem again. And if we
> decide to do nothing at least add a comment why we don't. :)

Found the old discussion, which ended in no action:
  http://lkml.org/lkml/2009/6/27/160

It was about simplifying the logic and not to allow to set a kobject
name several times in a row.

Sounds good to me to try that, and raise a BUG if the name is set a
second time. If we don't support that, all can be simplified, and the
original issue, that has come up again now, will be gone with that.

Thanks,
Kay