[PATCH 1/2] f2fs: fix to avoid race in between aio and gc

[PATCH 1/2] f2fs: fix to avoid race in between aio and gc

[Chao Yu]

We won't wait DIO synchronously when doing AIO, so there will be potential
IO reorder in between AIO and GC, which will cause data corruption.

This patch adds inode_dio_wait to serialize aio and data GC to avoid this
issue.

Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/gc.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index 57bea2182f30..cd147e7c71e8 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -875,6 +875,9 @@ static void gc_data_segment(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
 					continue;
 				}
 				locked = true;
+
+				/* wait for all inflight aio data */
+				inode_dio_wait(inode);
 			}
 
 			start_bidx = start_bidx_of_node(nofs, inode)
-- 
2.13.1.388.g69e6b9b4f4a9

[PATCH 2/2] f2fs: trigger fdatasync for non-atomic_write file

[Chao Yu]

Sqlite only cares about synchronization of file data instead of other data
unrelated attribute of inode, so in commit flow, call fdatasync is enough.

Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/file.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 5b180177570c..b9c4930fc74b 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -1700,7 +1700,7 @@ static int f2fs_ioc_commit_atomic_write(struct file *filp)
 			stat_dec_atomic_write(inode);
 		}
 	} else {
-		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, false);
+		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
 	}
 err_out:
 	inode_unlock(inode);
-- 
2.13.1.388.g69e6b9b4f4a9