Re: processes hung after sys_renameat, and 'missing' processes

[Al Viro <viro@ZenIV.linux.org.uk>]

On Wed, Jun 06, 2012 at 08:29:14PM -0400, Dave Jones wrote:
>  void d_move(struct dentry *dentry, struct dentry *target)
>  {
>         write_seqlock(&rename_lock);
> +
> +       BUG_ON(!spin_is_locked(&target->d_parent->d_lock));
> +
> +       if (dentry->d_parent != NULL)
> +               BUG_ON(!spin_is_locked(&dentry->d_parent->d_lock));
> +
>         __d_move(dentry, target);
>         write_sequnlock(&rename_lock);
>  }
> 
> 
> To be clear, do you want me to try that with or without the reverts ?

No.  Not ->d_lock; ->d_inode->i_mutex.  _That_ needs to be held for
d_move() to be safe (and on cross-directory move you need ->s_vfs_rename_mutex
as well).  I'd do it with WARN_ON, BTW, and without the reverts - that
way we'll have a clear indication if that code is stepped on.