Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190529113157.227380-1-jannh@google.com> <20190530123452.GF22536@redhat.com>
In-Reply-To: <20190530123452.GF22536@redhat.com>
From:   Jann Horn <jannh@google.com>
Date:   Fri, 31 May 2019 13:56:47 +0200
Message-ID: <CAG48ez0ivQ+gfwKMife-3ZwBuqAuc1BhDGW3dtYTHMq0sByuNw@mail.gmail.com>
Subject: Re: [PATCH] ptrace: restore smp_rmb() in __ptrace_may_access()
To:     Oleg Nesterov <oleg@redhat.com>
Cc:     "Eric W . Biederman" <ebiederm@xmission.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Kees Cook <keescook@chromium.org>,
        David Howells <dhowells@redhat.com>,
        kernel list <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, May 30, 2019 at 2:35 PM Oleg Nesterov <oleg@redhat.com> wrote:
> On 05/29, Jann Horn wrote:
> > --- a/kernel/cred.c
> > +++ b/kernel/cred.c
> > @@ -450,6 +450,15 @@ int commit_creds(struct cred *new)
> >               if (task->mm)
> >                       set_dumpable(task->mm, suid_dumpable);
> >               task->pdeath_signal = 0;
> > +             /*
> > +              * If a task drops privileges and becomes nondumpable,
> > +              * the dumpability change must become visible before
> > +              * the credential change; otherwise, a __ptrace_may_access()
> > +              * racing with this change may be able to attach to a task it
> > +              * shouldn't be able to attach to (as if the task had dropped
> > +              * privileges without becoming nondumpable).
> > +              * Pairs with a read barrier in __ptrace_may_access().
> > +              */
> >               smp_wmb();
>
> Hmm. Now that I tried to actually read this patch I do not understand this wmb().
>
> commit_creds() does rcu_assign_pointer(real_cred) which implies smp_store_release(),
> the dumpability change must be visible before ->real_cred is updated without any
> additional barriers?

Oh, yes, I think you're right.

So I guess I should make a v2 that still adds the smp_rmb() in
__ptrace_may_access(), but gets rid of the smp_wmb() in
commit_creds()? (With a comment above the rcu_assign_pointer() that
explains the ordering?)