Re: [PATCH 02/10] mm/hugetlb: make sure to get NULL when list is empty

[Mike Kravetz]

On 8/7/20 7:28 AM, Wei Yang wrote:
> On Fri, Aug 07, 2020 at 08:49:51PM +0800, Baoquan He wrote:
>> On 08/07/20 at 05:12pm, Wei Yang wrote:
>>> list_first_entry() may not return NULL even when the list is empty.
>>>
>>> Let's make sure the behavior by using list_first_entry_or_null(),
>>> otherwise it would corrupt the list.
>>>
>>> Signed-off-by: Wei Yang <richard.weiyang@xxxxxxxxxxxxxxxxx>
>>> ---
>>>  mm/hugetlb.c | 3 ++-
>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>> index 62ec74f6d03f..0a2f3851b828 100644
>>> --- a/mm/hugetlb.c
>>> +++ b/mm/hugetlb.c
>>> @@ -237,7 +237,8 @@ get_file_region_entry_from_cache(struct resv_map *resv, long from, long to)
>>>  	VM_BUG_ON(resv->region_cache_count <= 0);
>>
>>
>> We have had above line, is it possible to be NULL from list_first_entry?
>>
>>>  
>>>  	resv->region_cache_count--;
>>> -	nrg = list_first_entry(&resv->region_cache, struct file_region, link);
>>> +	nrg = list_first_entry_or_null(&resv->region_cache,
>>> +			struct file_region, link);
>>>  	VM_BUG_ON(!nrg);
> 
> Or we can remove this VM_BUG_ON()?
> 

I would prefer that we just remove the 'VM_BUG_ON(!nrg)'.  Code elsewhere
is responsible for making sure there is ALWAYS an entry in the cache.  That
is why the 'VM_BUG_ON(resv->region_cache_count <= 0)' is at the beginning
of the routine.

-- 
Mike Kravetz