Re: [v3 PATCH 2/2] mm: vmscan: correct some vmscan counters for THP swapout

2019-05-21 Thread Yang Shi
On 5/22/19 9:23 AM, Huang, Ying wrote: Yang Shi writes: Since commit bd4c82c22c36 ("mm, THP, swap: delay splitting THP after swapped out"), THP can be swapped out in a whole. But, nr_reclaimed and some other vm counters still get inc'ed by one even though a whole THP (512 pages) gets swapp

Re: [v3 PATCH 2/2] mm: vmscan: correct some vmscan counters for THP swapout

2019-05-21 Thread Yang Shi
On 5/22/19 12:00 AM, Johannes Weiner wrote: On Tue, May 21, 2019 at 05:40:42PM +0800, Yang Shi wrote: Since commit bd4c82c22c36 ("mm, THP, swap: delay splitting THP after swapped out"), THP can be swapped out in a whole. But, nr_reclaimed and some other vm counters still get inc'ed by one ev

Re: [v3 PATCH 2/2] mm: vmscan: correct some vmscan counters for THP swapout

2019-05-21 Thread Huang, Ying
Yang Shi writes: > Since commit bd4c82c22c36 ("mm, THP, swap: delay splitting THP after > swapped out"), THP can be swapped out in a whole. But, nr_reclaimed > and some other vm counters still get inc'ed by one even though a whole > THP (512 pages) gets swapped out. > > This doesn't make too muc

Re: [v3 PATCH 2/2] mm: vmscan: correct some vmscan counters for THP swapout

2019-05-21 Thread Johannes Weiner
On Tue, May 21, 2019 at 05:40:42PM +0800, Yang Shi wrote: > Since commit bd4c82c22c36 ("mm, THP, swap: delay splitting THP after > swapped out"), THP can be swapped out in a whole. But, nr_reclaimed > and some other vm counters still get inc'ed by one even though a whole > THP (512 pages) gets swa

[v3 PATCH 2/2] mm: vmscan: correct some vmscan counters for THP swapout

2019-05-21 Thread Yang Shi
Since commit bd4c82c22c36 ("mm, THP, swap: delay splitting THP after swapped out"), THP can be swapped out in a whole. But, nr_reclaimed and some other vm counters still get inc'ed by one even though a whole THP (512 pages) gets swapped out. This doesn't make too much sense to memory reclaim. Fo