When performing O_SYNC+AIO+DIO writes to block devices, use the DIO_SYNC_WRITES
flag so that flushes are issued /after/ the write completes, not before.

Note, however, that for block devices, the DIO setup code ensures that a flush
wq is attached to the superblock of the bdevfs filesystem, not the filesystem
that the device node happens to reside in.  This means that unlike regular
files, iocb->ki_filp->f_mapping->host->i_sb != inode->i_sb.  Therefore, adjust
Jeff's earlier patch to keep the pointer use consistent and avoid a NULL deref.

Signed-off-by: Darrick J. Wong <redacted>
---
 fs/block_dev.c |    5 +++--
 fs/direct-io.c |    3 ++-
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/fs/block_dev.c b/fs/block_dev.c
index 1a1e5e3..05ff33a 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c

@@ -235,7 +235,8 @@ blkdev_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
 	struct inode *inode = file->f_mapping->host;
 
 	return __blockdev_direct_IO(rw, iocb, inode, I_BDEV(inode), iov, offset,
-				    nr_segs, blkdev_get_blocks, NULL, NULL, 0);
+				    nr_segs, blkdev_get_blocks, NULL, NULL,
+				    DIO_SYNC_WRITES);
 }
 
 int __sync_blockdev(struct block_device *bdev, int wait)

@@ -1631,7 +1632,7 @@ ssize_t blkdev_aio_write(struct kiocb *iocb, const struct iovec *iov,
 	percpu_down_read(&bdev->bd_block_size_semaphore);
 
 	ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos);
-	if (ret > 0 || ret == -EIOCBQUEUED) {
+	if (ret > 0) {
 		ssize_t err;
 
 		err = generic_write_sync(file, pos, ret);

diff --git a/fs/direct-io.c b/fs/direct-io.c
index b7391d4..c626c43 100644
--- a/fs/direct-io.c
+++ b/fs/direct-io.c

@@ -258,7 +258,8 @@ void generic_dio_end_io(struct kiocb *iocb, loff_t offset, ssize_t bytes,
 		work->ret = ret;
 		work->offset = offset;
 		work->len = bytes;
-		queue_work(inode->i_sb->s_dio_flush_wq, &work->work);
+		queue_work(iocb->ki_filp->f_mapping->host->i_sb->s_dio_flush_wq,
+			   &work->work);