Re: [PATCH v6 3/7] kernfs: use VFS negative dentry caching

[Al Viro]

On Sat, Jun 12, 2021 at 09:08:05AM +0800, Ian Kent wrote:

> But if I change to take the read lock to ensure there's no operation
> in progress for the revision check I would need the dget_parent(), yes?

WTF for?  ->d_parent can change *ONLY* when ->d_lock is held on all
dentries involved (including old and new parents).

And it very definitely does *not* change for negative dentries.  I mean,
look at the very beginning of __d_move().