From: Ze Zuo zuoze1@huawei.com
The NUMA fault behavior didn't split THP if migration is failed due to lack of memory on the target node. But the THP migration does split THP, so keep the old behavior for misplaced NUMA page migration.
Signed-off-by: Ze Zuo zuoze1@huawei.com --- mm/migrate.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/mm/migrate.c b/mm/migrate.c index a3f9cbbfbfe5..4c2a1d99db53 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -1487,6 +1487,7 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page, struct page *page2; int swapwrite = current->flags & PF_SWAPWRITE; int rc, nr_subpages; + bool nosplit = (reason == MR_NUMA_MISPLACED);
if (!swapwrite) current->flags |= PF_SWAPWRITE; @@ -1527,8 +1528,9 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page, * pages are added to the tail of the list so * we encounter them after the rest of the list * is processed. + * THP NUMA acces migrate doesn't split THP to retry. */ - if (is_thp) { + if (is_thp && !nosplit) { lock_page(page); rc = split_huge_page_to_list(page, from); unlock_page(page);