NFS: Fix a race between mmap() and O_DIRECT

When locking the file in order to do O_DIRECT on it, we must unmap
any mmapped ranges on the pagecache so that we can flush out the
dirty data.

Fixes: a5864c999de67 ("NFS: Do not serialise O_DIRECT reads and writes")
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Cc: stable@vger.kernel.org # v4.8+

diff --git a/fs/nfs/io.c b/fs/nfs/io.c
index 20fef85d2bb15782c4ede8bdbaa083867e522d82..9034b4926909a91a86e7cbf95cd227178f953270 100644 (file)
--- a/fs/nfs/io.c
+++ b/fs/nfs/io.c
@@ -99,7 +99,7 @@ static void nfs_block_buffered(struct nfs_inode *nfsi, struct inode *inode)
 {
        if (!test_bit(NFS_INO_ODIRECT, &nfsi->flags)) {
                set_bit(NFS_INO_ODIRECT, &nfsi->flags);
-               nfs_wb_all(inode);
+               nfs_sync_mapping(inode->i_mapping);
        }
 }