Re: [PATCH] mm/migrate: fix deadlock in migrate_pages_batch() on large folios

[Gao Xiang <hsiangkao@linux.alibaba.com>]

On 2024/8/16 13:12, Gao Xiang wrote:
> 
> Hi Andrew,
> 
> On 2024/8/16 13:02, Andrew Morton wrote:
>> On Mon, 29 Jul 2024 09:58:02 +0800 Gao Xiang <hsiangkao@linux.alibaba.com> wrote:
>>
>>>> For the fix, I think that we should still respect migrate_mode because
>>>> users may prefer migration success over blocking.
>>>>
>>>> @@ -1492,11 +1492,17 @@ static int unmap_and_move_huge_page(new_folio_t get_new_folio,
>>>>        return rc;
>>>>    }
>>>> -static inline int try_split_folio(struct folio *folio, struct list_head *split_folios)
>>>> +static inline int try_split_folio(struct folio *folio, struct list_head *split_folios,
>>>> +                  enum migrate_mode mode)
>>>>    {
>>>>        int rc;
>>>> -    folio_lock(folio);
>>>> +    if (mode == MIGRATE_ASYNC) {
>>>> +        if (!folio_trylock(folio))
>>>> +            return -EAGAIN;
>>>> +    } else {
>>>> +        folio_lock(folio);
>>>> +    }
>>>>        rc = split_folio_to_list(folio, split_folios);
>>>>        folio_unlock(folio);
>>>>        if (!rc)
>>>
>>> Okay, yeah it looks better since it seems I missed the fallback
>>> part in migrate_pages_sync().
>>>
>>> Let me send the next version to follow your advice, thanks.
>>
>> The author seems to have disappeared.  Should we merge this as-is or
>> does someone want to take a look at developing a v2?
> 
> I've replied your email last week, I'm not sure why it has not
> been addressed?
> 
> https://lore.kernel.org/linux-mm/20240729021306.398286-1-hsiangkao@linux.alibaba.com/
> 
> The patch in your queue is already v2? No?

Really confused about this, since the comment above was about v1.
and v2 is already sent (in July 29) and in -next for two weeks
with

Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: David Hildenbrand <david@redhat.com>

What else I need to do to resolve this already resolved comment
so that I could enable large folios without deadlocks anymore?

Thanks,
Gao Xiang

> 
> Thanks,
> Gao Xiang