Re: [PATCH] knav_qmss_queue: fix a missing-check bug in knav_pool_create()

[santosh.shilimkar@oracle.com]

On 6/11/19 3:08 AM, Gen Zhang wrote:
> On Tue, Jun 11, 2019 at 10:54:15AM +0100, Marc Zyngier wrote:
>> Hi Gen,
>>
>> No idea why I'm being cc'd on this but hey... ;-)
> I copied email address ftom thid commit:-)
> https://github.com/torvalds/linux/commit/832ad0e3da4510fd17f98804abe512ea9a747035#diff-f2a24befc247191f4b81af93e9336785
>>
>> On 11/06/2019 10:37, Gen Zhang wrote:
>>> On Thu, May 30, 2019 at 11:39:49AM +0800, Gen Zhang wrote:
>>>> In knav_pool_create(), 'pool->name' is allocated by kstrndup(). It
>>>> returns NULL when fails. So 'pool->name' should be checked. And free
>>>> 'pool' when error.
>>>>
>>>> Signed-off-by: Gen Zhang <blackgod016574@gmail.com>
>>>> ---
>>>> diff --git a/drivers/soc/ti/knav_qmss_queue.c b/drivers/soc/ti/knav_qmss_queue.c
>>>> index 8b41837..0f8cb28 100644
>>>> --- a/drivers/soc/ti/knav_qmss_queue.c
>>>> +++ b/drivers/soc/ti/knav_qmss_queue.c
>>>> @@ -814,6 +814,12 @@ void *knav_pool_create(const char *name,
>>>>   	}
>>>>   
>>>>   	pool->name = kstrndup(name, KNAV_NAME_SIZE - 1, GFP_KERNEL);
>>>> +	if (!pool->name) {
>>>> +		dev_err(kdev->dev, "failed to duplicate for pool(%s)\n",
>>>> +			name);
>>
>> There is no need to output anything, the kernel will be loud enough if
>> you run out of memory.
> Thanks for your comments.
>>
>>>> +		ret = -ENOMEM;
>>>> +		goto err_name;
>>>> +	}
>>>>   	pool->kdev = kdev;
>>>>   	pool->dev = kdev->dev;
>>>>   
>>>> @@ -864,6 +870,7 @@ void *knav_pool_create(const char *name,
>>>>   	mutex_unlock(&knav_dev_lock);
>>>>   err:
>>>>   	kfree(pool->name);
>>>> +err_name:
>>
>> kfree(NULL) is perfectly valid, there is no need to create a second
>> label. Just branch to the existing error label.
> Sure, better not to add redundant codes.
>>
>>>>   	devm_kfree(kdev->dev, pool);
>>>>   	return ERR_PTR(ret);
>>>>   }
>>> Can anyone look into this patch?
>>>
>>> Thanks
>>> Gen
>>>
>>
>> The real question is whether this is actually an error at all.
>> pool->name doesn't seem to be used for anything but debug information,
>> and the printing code can perfectly accommodate a NULL pointer.
> That sounds reasonable. This patch just fixes a *theoretical* bug.
> 
Not even theoretical bug.