From:   Chen Wandun <chenwandun@huawei.com>
To:     <akpm@linux-foundation.org>, <linux-mm@kvack.org>,
        <linux-kernel@vger.kernel.org>, <ying.huang@intel.com>
CC:     <chenwandun@huawei.com>
Subject: [PATCH] mm/migrate: Continue to migrate for small pages
Date:   Sun, 29 Jan 2023 10:54:04 +0800
Message-ID: <20230129025404.1262745-1-chenwandun@huawei.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 7BIT
Content-Type:   text/plain; charset=US-ASCII
Precedence: bulk

migrate_hugetlbs returns -ENOMEM when no enough huge page,
however maybe there are still free small pages, so continue
to migrate for small pages.

Signed-off-by: Chen Wandun <chenwandun@huawei.com>
---
 mm/migrate.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/mm/migrate.c b/mm/migrate.c
index b971edbf32fc..c9d0d2058036 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -1939,7 +1939,12 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
 
 	rc_gather = migrate_hugetlbs(from, get_new_page, put_new_page, private,
 				     mode, reason, &stats, &ret_folios);
-	if (rc_gather < 0)
+	/* Maybe there are free small pages, continue to migrate.
+	 * Nowdays the only negtive return value of migrate_hugetlbs
+	 * is -ENOMEM, keep compatible for coming negtive return
+	 * value instead of ignore return value.
+	 */
+	if (rc_gather < 0 && rc_gather != -ENOMEM)
 		goto out;
 again:
 	nr_pages = 0;
-- 
2.25.1