mm: numa: take THP into account when migrating pages for NUMA balancing

Wanpeng Li pointed out that numamigrate_isolate_page() assumes that only
one base page is being migrated when in fact it can also be checking
THP.

The consequences are that a migration will be attempted when a target
node is nearly full and fail later.  It's unlikely to be user-visible
but it should be fixed.  While we are there, migrate_balanced_pgdat()
should treat nr_migrate_pages as an unsigned long as it is treated as a
watermark.

Signed-off-by: Mel Gorman <[email protected]>
Suggested-by: Wanpeng Li <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Andrea Arcangeli <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Simon Jeons <[email protected]>
Cc: Hugh Dickins <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm/migrate.c

@@ -1461,7 +1461,7 @@ int migrate_vmas(struct mm_struct *mm, const nodemask_t *to,
  * pages. Currently it only checks the watermarks which crude
  */
 static bool migrate_balanced_pgdat(struct pglist_data *pgdat,
-				   int nr_migrate_pages)
+				   unsigned long nr_migrate_pages)
 {
 	int z;
 	for (z = pgdat->nr_zones - 1; z >= 0; z--) {
@@ -1559,8 +1559,10 @@ int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 {
 	int ret = 0;
 
+	VM_BUG_ON(compound_order(page) && !PageTransHuge(page));
+
 	/* Avoid migrating to a node that is nearly full */
-	if (migrate_balanced_pgdat(pgdat, 1)) {
+	if (migrate_balanced_pgdat(pgdat, 1UL << compound_order(page))) {
 		int page_lru;
 
 		if (isolate_lru_page(page)) {