Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 1 Sep 2021 10:12:00 +0100
From:   Christoph Hellwig <hch@infradead.org>
To:     Dmitry Kadashev <dkadashev@gmail.com>
Cc:     Stephen Brennan <stephen.s.brennan@oracle.com>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Jens Axboe <axboe@kernel.dk>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] namei: Fix use after free in kern_path_locked
Message-ID: <YS9D4AlEsaCxLFV0@infradead.org>
References: <20210901001341.79887-1-stephen.s.brennan@oracle.com>
 <CAOKbgA49wFL3+-QAQ+DEnNVzCjYcN0qmnVHGo1x=eXeyzNxvsw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAOKbgA49wFL3+-QAQ+DEnNVzCjYcN0qmnVHGo1x=eXeyzNxvsw@mail.gmail.com>
Precedence: bulk

On Wed, Sep 01, 2021 at 02:35:08PM +0700, Dmitry Kadashev wrote:
> On Wed, Sep 1, 2021 at 7:13 AM Stephen Brennan
> <stephen.s.brennan@oracle.com> wrote:
> >
> > In 0ee50b47532a ("namei: change filename_parentat() calling
> > conventions"), filename_parentat() was made to always put the struct
> > filename before returning, and kern_path_locked() was migrated to this
> > calling convention. However, kern_path_locked() uses the "last"
> > parameter to lookup and potentially create a new dentry. The last
> > parameter contains the last component of the path and points within the
> > filename, which was recently freed at the end of filename_parentat().
> > Thus, when kern_path_locked() calls __lookup_hash(), it is using the
> > filename after it has already been freed.
> >
> > To avoid this, switch back to __filename_parentat() and place a putname
> > at the end of the function, once all uses are completed.
> 
> Ouch. Thanks for taking care of this, Stephen. I guess
> filename_parentat() should be killed, since kern_path_locked() was the
> only place it's used in and it always results in danging "last",
> provoking bugs just like this one. I can send a patch on top of this if
> you prefer.

Yes.  And then rename __filename_parentat to filename_parentat, please.