On Mon, Apr 18, 2016 at 09:06:35AM +0800, kernel test robot wrote: > FYI, we noticed the below changes on > > https://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs.git testing.lookups > commit 96352435afd792d713ec91e2d5693bdcf40a6b03 ("parallel lookups machinery, > part 2")
Very odd. All stores in that commit are to ->d_parent->d_inode->i_dir_seq of in-lookup dentries. And if ->d_parent in those gets buggered, we were really deep in trouble. Could you try to reproduce that without __d_move() or __d_add() parts and see which one ends up triggering that crap? Note that we don't *use* ->i_dir_seq yet...