Re: [PATCH 3/4] drivers/base: devres.c: use devm_kmemdup() from with devm_kstrdup()

[Rob Jones <rob.jones@codethink.co.uk>]

On 19/06/14 16:59, Joe Perches wrote:
> On Thu, 2014-06-19 at 16:46 +0100, Rob Jones wrote:
>> Avoid code duplication by using devm_kmemdup() to copy data instead
>> of having a separate loop within devm_kstrdup().
>>
>> Reviewed-by: Ian Molton <ian.molton@codethink.co.uk>
>> Signed-off-by: Rob Jones <rob.jones@codethink.co.uk>
> []
>> diff --git a/drivers/base/devres.c b/drivers/base/devres.c
> []
>> @@ -802,17 +802,10 @@ EXPORT_SYMBOL_GPL(devm_kmalloc);
>>    */
>>   char *devm_kstrdup(struct device *dev, const char *s, gfp_t gfp)
>>   {
> []
>> +	return devm_kmemdup(dev, s, (strlen(s) + 1), gfp);
>>   }
>>   EXPORT_SYMBOL_GPL(devm_kstrdup);
> Making this static inline in the header and dropping
> EXPORT_SYMBOL_GPL might be smaller code.
>
> static inline
> char *devm_kstrdup(struct device *dev, const char *s, gfp_t gfp)
> {
> 	return devm_kmemdup(dev, s, strlen(s) + 1, gfp);
> }
>

Is it wise not to check for (s == NULL)? Surely the behaviour of 
strlen(NULL) is undefined.

Consequently, I am, on balance, against making the function static 
inline. Given that inline is only a recommendation to the compiler, it 
may well not inline it anyway if the NULL test is present, in which case 
we would end up with a copy of the code in each module that uses it 
rather than a single, global, copy. If it didn't need the NULL test, I 
would agree that it would be best as a static inline but I think it does 
need it.

-- 
Rob Jones
Project Manager
Codethink Ltd
mailto:rob.jones@codethink.co.uk
tel:+44 161 236 5575