Re: [v2 PATCH 2/2] mm: mempolicy: handle vma with unmovable pages mapped correctly in mbind

[Yang Shi <yang.shi@linux.alibaba.com>]

On 7/16/19 5:07 AM, Vlastimil Babka wrote:
> On 6/22/19 2:20 AM, Yang Shi wrote:
>> @@ -969,10 +975,21 @@ static long do_get_mempolicy(int *policy, nodemask_t *nmask,
>>   /*
>>    * page migration, thp tail pages can be passed.
>>    */
>> -static void migrate_page_add(struct page *page, struct list_head *pagelist,
>> +static int migrate_page_add(struct page *page, struct list_head *pagelist,
>>   				unsigned long flags)
>>   {
>>   	struct page *head = compound_head(page);
>> +
>> +	/*
>> +	 * Non-movable page may reach here.  And, there may be
>> +	 * temporaty off LRU pages or non-LRU movable pages.
>> +	 * Treat them as unmovable pages since they can't be
>> +	 * isolated, so they can't be moved at the moment.  It
>> +	 * should return -EIO for this case too.
>> +	 */
>> +	if (!PageLRU(head) && (flags & MPOL_MF_STRICT))
>> +		return -EIO;
>> +
> Hm but !PageLRU() is not the only way why queueing for migration can
> fail, as can be seen from the rest of the function. Shouldn't all cases
> be reported?

Do you mean the shared pages and isolation failed pages? I'm not sure 
whether we should consider these cases break the semantics or not, so I 
leave them as they are. But, strictly speaking they should be reported 
too, at least for the isolation failed page.

Thanks,
Yang

>
>>   	/*
>>   	 * Avoid migrating a page that is shared with others.
>>   	 */
>> @@ -984,6 +1001,8 @@ static void migrate_page_add(struct page *page, struct list_head *pagelist,
>>   				hpage_nr_pages(head));
>>   		}
>>   	}
>> +
>> +	return 0;
>>   }
>>   
>>   /* page allocation callback for NUMA node migration */
>> @@ -1186,9 +1205,10 @@ static struct page *new_page(struct page *page, unsigned long start)
>>   }
>>   #else
>>   
>> -static void migrate_page_add(struct page *page, struct list_head *pagelist,
>> +static int migrate_page_add(struct page *page, struct list_head *pagelist,
>>   				unsigned long flags)
>>   {
>> +	return -EIO;
>>   }
>>   
>>   int do_migrate_pages(struct mm_struct *mm, const nodemask_t *from,
>>