Re: [PATCH v2] CMA: page_isolation: check buddy before access it

[Vlastimil Babka <vbabka@suse.cz>]

On 6.5.2015 8:28, Joonsoo Kim wrote:
> On Tue, May 05, 2015 at 11:22:59AM +0800, Hui Zhu wrote:
>>
>> diff --git a/mm/page_isolation.c b/mm/page_isolation.c
>> index 755a42c..eb22d1f 100644
>> --- a/mm/page_isolation.c
>> +++ b/mm/page_isolation.c
>> @@ -101,7 +101,8 @@ void unset_migratetype_isolate(struct page *page, unsigned migratetype)
>>  			buddy_idx = __find_buddy_index(page_idx, order);
>>  			buddy = page + (buddy_idx - page_idx);
>>  
>> -			if (!is_migrate_isolate_page(buddy)) {
>> +			if (!pfn_valid_within(page_to_pfn(buddy))
>> +			    || !is_migrate_isolate_page(buddy)) {
>>  				__isolate_free_page(page, order);
>>  				kernel_map_pages(page, (1 << order), 1);
>>  				set_page_refcounted(page);
> 
> Hello,
> 
> This isolation is for merging buddy pages. If buddy is not valid, we
> don't need to isolate page, because we can't merge them.
> I think that correct code would be:
> 
> pfn_valid_within(page_to_pfn(buddy)) &&
>         !is_migrate_isolate_page(buddy)
> 
> But, isolation and free here is safe operation so your code will work
> fine.

Ah damnit, you're right. But now you got me thinking about it more, and
paranoid... I thought I saw more bugs since the buddy might be in different zone
and we are not locking that zone, but then again it's probably fine, just very
tricky. Then I thought it could be simplified but then not again. Guess I'll
just run away fast :)

> Thanks.
> 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>