[PATCH 2/2] f2fs: fix to wait all node page writeback

From: Chao Yu
Date: Fri Feb 14 2020 - 04:46:11 EST


There is a race condition that we may miss to wait for all node pages
writeback, fix it.

- fsync() - shrink
- f2fs_do_sync_file
- __write_node_page
- set_page_writeback(page#0)
: remove DIRTY/TOWRITE flag
- f2fs_fsync_node_pages
: won't find page #0 as TOWRITE flag was removeD
- f2fs_wait_on_node_pages_writeback
: wont' wait page #0 writeback as it was not in fsync_node_list list.
- f2fs_add_fsync_node_entry

Fixes: 50fa53eccf9f ("f2fs: fix to avoid broken of dnode block list")
Signed-off-by: Chao Yu <yuchao0@xxxxxxxxxx>
---
fs/f2fs/node.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index 332cebc2033c..9824f055a5bc 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -1561,15 +1561,16 @@ static int __write_node_page(struct page *page, bool atomic, bool *submitted,
if (atomic && !test_opt(sbi, NOBARRIER))
fio.op_flags |= REQ_PREFLUSH | REQ_FUA;

- set_page_writeback(page);
- ClearPageError(page);
-
+ /* should add to global list before clearing PAGECACHE status */
if (f2fs_in_warm_node_list(sbi, page)) {
seq = f2fs_add_fsync_node_entry(sbi, page);
if (seq_id)
*seq_id = seq;
}

+ set_page_writeback(page);
+ ClearPageError(page);
+
fio.old_blkaddr = ni.blk_addr;
f2fs_do_write_node_page(nid, &fio);
set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page));
--
2.18.0.rc1