Currently the kernel is not correctly updating the numa stats for NR_FILE_PAGES and NR_SHMEM on THP migration. Fix that. For NR_FILE_DIRTY and NR_ZONE_WRITE_PENDING, although at the moment there is no need to handle THP migration as kernel still does not have write support for file THP but to be more future proof, this patch adds the THP support for those stats as well.

Fixes: e71769ae52609 ("mm: enable thp migration for shmem thp") Signed-off-by: Shakeel Butt shakeelb@google.com Cc: stable@vger.kernel.org --- mm/migrate.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-)

diff --git a/mm/migrate.c b/mm/migrate.c index 613794f6a433..ade163c6ecdf 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -402,6 +402,7 @@ int migrate_page_move_mapping(struct address_space *mapping, struct zone *oldzone, *newzone; int dirty; int expected_count = expected_page_refs(mapping, page) + extra_count; + int nr = thp_nr_pages(page);

if (!mapping) { /* Anonymous page without mapping */ @@ -437,7 +438,7 @@ int migrate_page_move_mapping(struct address_space *mapping, */ newpage->index = page->index; newpage->mapping = page->mapping; - page_ref_add(newpage, thp_nr_pages(page)); /* add cache reference */ + page_ref_add(newpage, nr); /* add cache reference */ if (PageSwapBacked(page)) { __SetPageSwapBacked(newpage); if (PageSwapCache(page)) { @@ -459,7 +460,7 @@ int migrate_page_move_mapping(struct address_space *mapping, if (PageTransHuge(page)) { int i;

- for (i = 1; i < HPAGE_PMD_NR; i++) { + for (i = 1; i < nr; i++) { xas_next(&xas); xas_store(&xas, newpage); } @@ -470,7 +471,7 @@ int migrate_page_move_mapping(struct address_space *mapping, * to one less reference. * We know this isn't the last reference. */ - page_ref_unfreeze(page, expected_count - thp_nr_pages(page)); + page_ref_unfreeze(page, expected_count - nr);

xas_unlock(&xas); /* Leave irq disabled to prevent preemption while updating stats */ @@ -493,17 +494,17 @@ int migrate_page_move_mapping(struct address_space *mapping, old_lruvec = mem_cgroup_lruvec(memcg, oldzone->zone_pgdat); new_lruvec = mem_cgroup_lruvec(memcg, newzone->zone_pgdat);

- __dec_lruvec_state(old_lruvec, NR_FILE_PAGES); - __inc_lruvec_state(new_lruvec, NR_FILE_PAGES); + __mod_lruvec_state(old_lruvec, NR_FILE_PAGES, -nr); + __mod_lruvec_state(new_lruvec, NR_FILE_PAGES, nr); if (PageSwapBacked(page) && !PageSwapCache(page)) { - __dec_lruvec_state(old_lruvec, NR_SHMEM); - __inc_lruvec_state(new_lruvec, NR_SHMEM); + __mod_lruvec_state(old_lruvec, NR_SHMEM, -nr); + __mod_lruvec_state(new_lruvec, NR_SHMEM, nr); } if (dirty && mapping_can_writeback(mapping)) { - __dec_lruvec_state(old_lruvec, NR_FILE_DIRTY); - __dec_zone_state(oldzone, NR_ZONE_WRITE_PENDING); - __inc_lruvec_state(new_lruvec, NR_FILE_DIRTY); - __inc_zone_state(newzone, NR_ZONE_WRITE_PENDING); + __mod_lruvec_state(old_lruvec, NR_FILE_DIRTY, -nr); + __mod_zone_page_tate(oldzone, NR_ZONE_WRITE_PENDING, -nr); + __mod_lruvec_state(new_lruvec, NR_FILE_DIRTY, nr); + __mod_zone_page_state(newzone, NR_ZONE_WRITE_PENDING, nr); } } local_irq_enable();

On Mon, Dec 28, 2020 at 2:13 AM Shakeel Butt shakeelb@google.com wrote:

The kernel updates the per-node NR_FILE_DIRTY stats on page migration but not the memcg numa stats. That was not an issue until recently the commit 5f9a4f4a7096 ("mm: memcontrol: add the missing numa_stat interface for cgroup v2") exposed numa stats for the memcg. So fixing the file_dirty per-memcg numa stat.

Fixes: 5f9a4f4a7096 ("mm: memcontrol: add the missing numa_stat interface for cgroup v2")

Thanks for catching this problem.

Reviewed-by: Muchun Song songmuchun@bytedance.com

Signed-off-by: Shakeel Butt shakeelb@google.com Cc: stable@vger.kernel.org

---

mm/migrate.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mm/migrate.c b/mm/migrate.c index ee5e612b4cd8..613794f6a433 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -500,9 +500,9 @@ int migrate_page_move_mapping(struct address_space *mapping, __inc_lruvec_state(new_lruvec, NR_SHMEM); } if (dirty && mapping_can_writeback(mapping)) {

•                   __dec_node_state(oldzone->zone_pgdat, NR_FILE_DIRTY);
  

•                   __dec_lruvec_state(old_lruvec, NR_FILE_DIRTY);
                    __dec_zone_state(oldzone, NR_ZONE_WRITE_PENDING);
  

•                   __inc_node_state(newzone->zone_pgdat, NR_FILE_DIRTY);
  

•                   __inc_lruvec_state(new_lruvec, NR_FILE_DIRTY);
                    __inc_zone_state(newzone, NR_ZONE_WRITE_PENDING);
            }
    }
  

-- 2.29.2.729.g45daf8777d-goog

-- Yours, Muchun

On Sun, Dec 27, 2020 at 10:13:09AM -0800, Shakeel Butt wrote:

The kernel updates the per-node NR_FILE_DIRTY stats on page migration but not the memcg numa stats. That was not an issue until recently the commit 5f9a4f4a7096 ("mm: memcontrol: add the missing numa_stat interface for cgroup v2") exposed numa stats for the memcg. So fixing the file_dirty per-memcg numa stat.

Fixes: 5f9a4f4a7096 ("mm: memcontrol: add the missing numa_stat interface for cgroup v2") Signed-off-by: Shakeel Butt shakeelb@google.com Cc: stable@vger.kernel.org

Reviewed-by: Roman Gushchin guro@fb.com

Thanks!

---

mm/migrate.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mm/migrate.c b/mm/migrate.c index ee5e612b4cd8..613794f6a433 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -500,9 +500,9 @@ int migrate_page_move_mapping(struct address_space *mapping, __inc_lruvec_state(new_lruvec, NR_SHMEM); } if (dirty && mapping_can_writeback(mapping)) {

• 	__dec_node_state(oldzone->zone_pgdat, NR_FILE_DIRTY);
  

• 	__dec_lruvec_state(old_lruvec, NR_FILE_DIRTY);
  __dec_zone_state(oldzone, NR_ZONE_WRITE_PENDING);
  

• 	__inc_node_state(newzone->zone_pgdat, NR_FILE_DIRTY);
  

• 	__inc_lruvec_state(new_lruvec, NR_FILE_DIRTY);
  __inc_zone_state(newzone, NR_ZONE_WRITE_PENDING);
  

  } }

-- 2.29.2.729.g45daf8777d-goog