linux-kernel - Re: [PATCH] seccomp: fix the usage of get/put_seccomp_filter() in seccomp_get

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20170921113150.GA1416@redhat.com>
Date:   Thu, 21 Sep 2017 13:31:50 +0200
From:   Oleg Nesterov <oleg@...hat.com>
To:     Kees Cook <keescook@...omium.org>
Cc:     Chris Salls <chrissalls5@...il.com>,
        Andy Lutomirski <luto@...capital.net>,
        Will Drewry <wad@...omium.org>,
        "security@...nel.org" <security@...nel.org>,
        LKML <linux-kernel@...r.kernel.org>,
        Tycho Andersen <tycho@...ker.com>
Subject: Re: [PATCH] seccomp: fix the usage of get/put_seccomp_filter() in
 seccomp_get_filter()

On 09/20, Kees Cook wrote:
>
> I like doing these sanity checks -- this isn't fast-path at all.

Yes, but see another "introduce get_nth_filter()" cleanup I sent, it is
similar but more suitable for Tycho's "retrieving seccomp flags" patch.

> > +       for (filter = orig; count > 1; filter = filter->prev)
                                ^^^^^^^^^
I just noticed that I forgot to replace this check with "count != 1".
Correctness wise this doesn't matter, but looks more clean.

> >                 count--;
> > -       }
> > -
> > -       if (WARN_ON(count != 1 || !filter)) {
> > -               /* The filter tree shouldn't shrink while we're using it. */
> > -               ret = -ENOENT;
> > -               goto out;
> > -       }
>
> Similarly, there's no reason to remove this check either.

Well, I disagree, but this is subjective so I won't insist.

Why do we want this WARN_ON() ? The sanity check can only fail if we have
a bug in 10 lines above. Lets look at the code after this cleanup,

	count = 0;
	for (filter = orig; filter; filter = filter->prev)
		count++;

	if (filter_off >= count)
		goto out;

	count -= filter_off;
	for (filter = orig; count != 1; filter = filter->prev)
		count--;


Do we want to check "count == 1" after the 2nd loop? I don't think so.
filter != NULL ? IMO makes no sense. Again, it can only be NULL if the
quoted code above is wrong, and in this case the next line

	refcount_inc(&filter->usage);

will crash.

Oleg.