f2fs: introduce FI_ATOMIC_COMMIT
This patch introduces a new flag to indicate inode status of doing atomic
write committing, so that, we can keep atomic write status for inode
during atomic committing, then we can skip GCing pages of atomic write inode,
that avoids random GCed datas being mixed with current transaction, so
isolation of transaction can be kept.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 6c33518..e455221 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -1569,14 +1569,15 @@ static int f2fs_ioc_commit_atomic_write(struct file *filp)
goto err_out;
if (f2fs_is_atomic_file(inode)) {
- clear_inode_flag(inode, FI_ATOMIC_FILE);
ret = commit_inmem_pages(inode);
- if (ret) {
- set_inode_flag(inode, FI_ATOMIC_FILE);
+ if (ret)
goto err_out;
- }
+
ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
- stat_dec_atomic_write(inode);
+ if (!ret) {
+ clear_inode_flag(inode, FI_ATOMIC_FILE);
+ stat_dec_atomic_write(inode);
+ }
} else {
ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
}