Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <CAHC9VhSSafrkW4VbTVoAUJmjFQdCwPTGDqTP8yBnLBqc7rW7iQ@mail.gmail.com>
 <CAHk-=wj_6Df1NAMs14S0OOqX1Z=460j-mfwn_qm-7EK1eK76qw@mail.gmail.com> <20221109143834.GB24561@mail.hallyn.com>
In-Reply-To: <20221109143834.GB24561@mail.hallyn.com>
From:   Paul Moore <paul@paul-moore.com>
Date:   Wed, 9 Nov 2022 15:13:40 -0500
Message-ID: <CAHC9VhQgEej_yKXRJFtLHUQkL8hGrBYqRPye5SwBz_SU+2eK2A@mail.gmail.com>
Subject: Re: [GIT PULL] LSM fixes for v6.1 (#1)
To:     "Serge E. Hallyn" <serge@hallyn.com>
Cc:     Linus Torvalds <torvalds@linux-foundation.org>,
        linux-security-module@vger.kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Nov 9, 2022 at 9:38 AM Serge E. Hallyn <serge@hallyn.com> wrote:
> On Mon, Oct 31, 2022 at 12:22:29PM -0700, Linus Torvalds wrote:
> > On Mon, Oct 31, 2022 at 4:07 AM Paul Moore <paul@paul-moore.com> wrote:
> > >
> > > A single patch to the capabilities code to fix a potential memory leak
> > > in the xattr allocation error handling.  Please apply for v6.1-rcX.
> >
> > Pulled.
> >
> > However, I react to the strange test condition. Sure, it's
> > pre-existing, but does it really make sense?
> >
> > It does
> >
> > +       if (ret < 0 || !tmpbuf) {
> > +               size = ret;
> > +               goto out_free;
> > +       }
> >
> > and how the heck can 'tmpbuf' be NULL if vfs_getxattr_alloc() succeeded?
>
> I had to go through the history a bit - the !tmpbuf check was added
>
> https://www.spinics.net/lists/stable/msg463010.html
>
> because of a gcc warning.  Perhaps there's a better way to tell gcc
> that it can't remain NULL if ret was < 0 ?

Ooof, that's ugly, but thanks for digging it up.  As it turns out I
happen to be working on a patch for vfs_getxattr_alloc() to fix the
return value type right now, but it looks like I'll leave that gcc
hack in place ... although I might leave a comment about it so the
next person doesn't have to wonder.

> > I think that's not only impossible in the first place, but if it *was*
> > possible, then that
> >
> >                 size = ret;
> >                goto out_free;
> >
> > would be wrong, because this function would return success even if it
> > wasn't successful.
> >
> > That whole "cast to int, and then cast back to size_t" also smells of
> > some serious confusion in the return value handling. It looks to me
> > like vfs_getxattr_alloc() fundamentally returns an 'int', not a
> > 'ssize_t', just by looking at the ->get function. But it just all
> > looks weird.
> >
> > So this code has all kinds of oddities.
> >
> >                Linus

-- 
paul-moore.com