Re: [PATCH] mm/hugetlb: hide nr_nodes in the internal of for_each_node_mask_to_[alloc|free]

[Wei Yang <richard.weiyang@linux.alibaba.com>]

On Tue, Jul 14, 2020 at 02:12:03PM -0700, Mike Kravetz wrote:
>On 7/14/20 3:02 AM, Vlastimil Babka wrote:
>> On 7/14/20 11:57 AM, Wei Yang wrote:
>>> On Tue, Jul 14, 2020 at 11:22:03AM +0200, Vlastimil Babka wrote:
>>>> On 7/14/20 11:13 AM, Vlastimil Babka wrote:
>>>>> On 7/14/20 9:34 AM, Wei Yang wrote:
>>>>>> The second parameter of for_each_node_mask_to_[alloc|free] is a loop
>>>>>> variant, which is not used outside of loop iteration.
>>>>>>
>>>>>> Let's hide this.
>>>>>>
>>>>>> Signed-off-by: Wei Yang <richard.weiyang@linux.alibaba.com>
>>>>>> ---
>>>>>>  mm/hugetlb.c | 38 ++++++++++++++++++++------------------
>>>>>>  1 file changed, 20 insertions(+), 18 deletions(-)
>>>>>>
>>>>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>>>>> index 57ece74e3aae..9c3d15fb317e 100644
>>>>>> --- a/mm/hugetlb.c
>>>>>> +++ b/mm/hugetlb.c
>>>>>> @@ -1196,17 +1196,19 @@ static int hstate_next_node_to_free(struct hstate *h, nodemask_t *nodes_allowed)
>>>>>>  	return nid;
>>>>>>  }
>>>>>>  
>>>>>> -#define for_each_node_mask_to_alloc(hs, nr_nodes, node, mask)		\
>>>>>> -	for (nr_nodes = nodes_weight(*mask);				\
>>>>>> -		nr_nodes > 0 &&						\
>>>>>> +#define for_each_node_mask_to_alloc(hs, node, mask)			\
>>>>>> +	int __nr_nodes;							\
>>>>>> +	for (__nr_nodes = nodes_weight(*mask);				\
>>>>>
>>>>> The problem with this is that if I use the macro twice in the same block, this
>>>>> will redefine __nr_nodes and fail to compile, no?
>>>>> In that case it's better to avoid setting up this trap, IMHO.
>>>>
>>>> Ah, and it will also generate the following warning, if the use of for_each*
>>>> macro is not the first thing after variable declarations, but there's another
>>>> statement before:
>>>>
>>>> warning: ISO C90 forbids mixed declarations and code [-Wdeclaration-after-statement]
>>>>
>>>> Instead we should switch to C99 and declare it as "for (int __nr_nodes" :P
>>>
>>> Hmm... I tried what you suggested, but compiler complains.
>>>
>>> 'for' loop initial declarations are only allowed in C99 or C11 mode
>> 
>> Yes, by "we should switch to C99" I meant that the kernel kbuild system would
>> need to switch. Not a trivial change...
>> Without that, I don't see how your patch is possible to do safely.
>
>Vlastimil, thanks for pointing out future potential issues with this patch.
>I likely would have missed that.
>
>Wei, thanks for taking the time to put together the patch.  However, I tend
>to agree with Vlastimil's assesment.  The cleanup is not worth the risk of
>running into issues if someone uses multiple instances of the macro.

Yep, thanks all for your feedback.

>-- 
>Mike Kravetz

-- 
Wei Yang
Help you, Help me