f2fs: don't give partially written atomic data from process crash

This patch resolves the below scenario.

== Process 1 ==     == Process 2 ==
open(w)             open(rw)
begin
write(new_#1)
process_crash
  f_op->flush
  locks_remove_posix
  f_op>release
                    read (new_#1)

In order to avoid corrupted database caused by new_#1, we must do roll-back
at process_crash time. In order to check that, this patch keeps task which
triggers transaction begin, and does roll-back in f_op->flush before removing
file locks.

Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

2 files changed, 19 insertions, 0 deletions

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index a02645780fd7..8ec1afac4897 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -603,6 +603,7 @@ struct f2fs_inode_info {
 	struct list_head dirty_list;	/* dirty list for dirs and files */
 	struct list_head gdirty_list;	/* linked in global dirty list */
 	struct list_head inmem_pages;	/* inmemory pages managed by f2fs */
+	struct task_struct *inmem_task;	/* store inmemory task */
 	struct mutex inmem_lock;	/* lock for inmemory pages */
 	struct extent_tree *extent_tree;	/* cached extent_tree entry */
 	struct rw_semaphore dio_rwsem[2];/* avoid racing between dio and gc */
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 435927e6c6f3..368aa332c833 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -1480,6 +1480,22 @@ static int f2fs_release_file(struct inode *inode, struct file *filp)
 	return 0;
 }
 
+static int f2fs_file_flush(struct file *file, fl_owner_t id)
+{
+	struct inode *inode = file_inode(file);
+
+	/*
+	 * If the process doing a transaction is crashed, we should do
+	 * roll-back. Otherwise, other reader/write can see corrupted database
+	 * until all the writers close its file. Since this should be done
+	 * before dropping file lock, it needs to do in ->flush.
+	 */
+	if (f2fs_is_atomic_file(inode) &&
+			F2FS_I(inode)->inmem_task == current)
+		drop_inmem_pages(inode);
+	return 0;
+}
+
 #define F2FS_REG_FLMASK		(~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
 #define F2FS_OTHER_FLMASK	(FS_NODUMP_FL | FS_NOATIME_FL)
 
@@ -1599,6 +1615,7 @@ static int f2fs_ioc_start_atomic_write(struct file *filp)
 	}
 
 inc_stat:
+	F2FS_I(inode)->inmem_task = current;
 	stat_inc_atomic_write(inode);
 	stat_update_max_atomic_write(inode);
 out:
@@ -2496,6 +2513,7 @@ const struct file_operations f2fs_file_operations = {
 	.open		= f2fs_file_open,
 	.release	= f2fs_release_file,
 	.mmap		= f2fs_file_mmap,
+	.flush		= f2fs_file_flush,
 	.fsync		= f2fs_sync_file,
 	.fallocate	= f2fs_fallocate,
 	.unlocked_ioctl	= f2fs_ioctl,