Re: [PATCH 04/14] mm,migration: Allow the migration of PageSwapCache pages

[Minchan Kim]

On Thu, Apr 22, 2010 at 8:59 AM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:
> On Wed, 21 Apr 2010 09:30:20 -0500 (CDT)
> Christoph Lameter <cl@xxxxxxxxxxxxxxxxxxxx> wrote:
>
>> On Tue, 20 Apr 2010, Mel Gorman wrote:
>>
>> > @@ -520,10 +521,12 @@ static int move_to_new_page(struct page *newpage, struct page *page)
>> > Â Â else
>> > Â Â Â Â Â Â rc = fallback_migrate_page(mapping, newpage, page);
>> >
>> > - Â if (!rc)
>> > - Â Â Â Â Â remove_migration_ptes(page, newpage);
>> > - Â else
>> > + Â if (rc) {
>> > Â Â Â Â Â Â newpage->mapping = NULL;
>> > + Â } else {
>> > + Â Â Â Â Â if (remap_swapcache)
>> > + Â Â Â Â Â Â Â Â Â remove_migration_ptes(page, newpage);
>> > + Â }
>>
>> You are going to keep the migration ptes after the page has been unlocked?
>> Or is remap_swapcache true if its not a swapcache page?
>>
>> Maybe you meant
>>
>> if (!remap_swapcache)
>>
>
> Ah....Can I confirm my understanding ?
>
> remap_swapcache == true only when
> ÂThe old page was ANON && it is not mapped. && it is SwapCache.
>
> We do above check under lock_page(). So, this SwapCache is never mapped until
> we release lock_page() on the old page. So, we don't use migration_pte in
> this case because try_to_unmap() do nothing and don't need to call
> remove_migration_pte().

Yes. so I thought what kinds of race happened.
Firstly I doubt fork and migration. but It isn't.
I can't understand how this bug happens.
Apparently, We have been missed something.
I will look into this further.


-- 
Kind regards,
Minchan Kim
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/