Re: [PATCH 1/8] mm/vmscan: remove redundant folio_test_swapbacked check when folio is file lru

["Huang, Ying" <ying.huang@intel.com>]

Muchun Song <songmuchun@bytedance.com> writes:

> On Tue, Mar 29, 2022 at 9:26 PM Miaohe Lin <linmiaohe@huawei.com> wrote:
>>
>> When folio is file lru, folio_test_swapbacked is guaranteed to be true. So
>> it's unnecessary to check it here again. No functional change intended.
>>
>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>> ---
>>  mm/vmscan.c | 3 +--
>>  1 file changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/mm/vmscan.c b/mm/vmscan.c
>> index 1678802e03e7..7c1a9713bfc9 100644
>> --- a/mm/vmscan.c
>> +++ b/mm/vmscan.c
>> @@ -1434,8 +1434,7 @@ static void folio_check_dirty_writeback(struct folio *folio,
>>          * Anonymous pages are not handled by flushers and must be written
>>          * from reclaim context. Do not stall reclaim based on them
>>          */
>> -       if (!folio_is_file_lru(folio) ||
>> -           (folio_test_anon(folio) && !folio_test_swapbacked(folio))) {
>> +       if (!folio_is_file_lru(folio) || folio_test_anon(folio)) {
>
> At least your login is no problem since folio_is_file_lru() is equal to
> !folio_test_swapbacked().  But the new code is not clear to me.
> The old code is easy to understand, e.g. folio_test_anon(folio) &&
> !folio_test_swapbacked(folio) tells us that the anon pages which
> do not need to be swapped should be skipped.

That is for MADV_FREE pages.  The code is introduced in commit
802a3a92ad7a ("mm: reclaim MADV_FREE pages").

So I think the original code is better.  It's an implementation detail
that folio_is_file_lru() equals !folio_test_swapbacked().  It may be
better to add some comments here for MADV_FREE pages.

> So I'm neutral on the patch.

Best Regards,
Huang, Ying