[PATCH v2 12/17] mm: ensure that we set mapping error if writeout() fails

From: Jeff Layton
Date: Wed Apr 12 2017 - 08:06:49 EST


If writepage fails during a page migration, then we need to ensure that
fsync will see it by flagging the mapping.

Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
---
mm/migrate.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/mm/migrate.c b/mm/migrate.c
index ed97c2c14fa8..a77b9e769d0b 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -792,7 +792,11 @@ static int writeout(struct address_space *mapping, struct page *page)
/* unlocked. Relock */
lock_page(page);

- return (rc < 0) ? -EIO : -EAGAIN;
+ if (rc < 0) {
+ mapping_set_error(mapping, rc);
+ return -EIO;
+ }
+ return -EAGAIN;
}

/*
--
2.9.3