Re: [PATCH] securityfs: Add missing d_delete() call on removal

[Al Viro]

On Wed, May 06, 2020 at 08:34:29AM -0700, Kees Cook wrote:

> Just posted the whole series:
> https://lore.kernel.org/lkml/20200506152114.50375-1-keescook@xxxxxxxxxxxx/
> 
> But the specific question was driven by this patch:
> https://lore.kernel.org/lkml/20200506152114.50375-11-keescook@xxxxxxxxxxxx/

Yecchh...  First of all, you are leaving a dangling pointer in your
struct pstore_private ->dentry.  What's more, in your case d_delete()
is definitely wrong - either there are other references to dentry
(in which case d_delete() is the same as d_drop()), or dput() right
after it will drive ->d_count to zero and since you end up using
simple_dentry_operations, dentry will be freed immediately after
that.

I have not looked at the locking in that series yet, so no comments
on the races, but in any case - that d_delete() is a misspelled d_drop().