From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754906Ab2B0TOL (ORCPT ); Mon, 27 Feb 2012 14:14:11 -0500 Received: from mail-yx0-f174.google.com ([209.85.213.174]:49919 "EHLO mail-yx0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754281Ab2B0TOI convert rfc822-to-8bit (ORCPT ); Mon, 27 Feb 2012 14:14:08 -0500 Authentication-Results: mr.google.com; spf=pass (google.com: domain of keescook@google.com designates 10.60.28.10 as permitted sender) smtp.mail=keescook@google.com; dkim=pass header.i=keescook@google.com MIME-Version: 1.0 In-Reply-To: References: <1330140111-17201-1-git-send-email-wad@chromium.org> <1330140111-17201-7-git-send-email-wad@chromium.org> <20120227171132.GB10608@redhat.com> <20120227181434.GA13903@redhat.com> Date: Mon, 27 Feb 2012 11:14:04 -0800 X-Google-Sender-Auth: kUh_h40BUQREljCArS3U47TF0Co Message-ID: Subject: Re: [PATCH v11 07/12] seccomp: add SECCOMP_RET_ERRNO From: Kees Cook To: Andrew Lutomirski Cc: Oleg Nesterov , Will Drewry , linux-kernel@vger.kernel.org, linux-arch@vger.kernel.org, linux-doc@vger.kernel.org, kernel-hardening@lists.openwall.com, netdev@vger.kernel.org, x86@kernel.org, arnd@arndb.de, davem@davemloft.net, hpa@zytor.com, mingo@redhat.com, peterz@infradead.org, rdunlap@xenotime.net, mcgrathr@chromium.org, tglx@linutronix.de, eparis@redhat.com, serge.hallyn@canonical.com, djm@mindrot.org, scarybeasts@gmail.com, indan@nul.nu, pmoore@redhat.com, akpm@linux-foundation.org, corbet@lwn.net, eric.dumazet@gmail.com, markus@chromium.org, coreyb@linux.vnet.ibm.com Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8BIT X-System-Of-Record: true Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Feb 27, 2012 at 10:35 AM, Andrew Lutomirski wrote: > On Mon, Feb 27, 2012 at 10:14 AM, Oleg Nesterov wrote: >> On 02/27, Kees Cook wrote: >>> >>> On Mon, Feb 27, 2012 at 9:11 AM, Oleg Nesterov wrote: >>> > On 02/24, Will Drewry wrote: >>> >> >>> >>  static u32 seccomp_run_filters(int syscall) >>> >>  { >>> >>       struct seccomp_filter *f; >>> >> -     u32 ret = SECCOMP_RET_KILL; >>> >>       static const struct bpf_load_fn fns = { >>> >>               bpf_load, >>> >>               sizeof(struct seccomp_data), >>> >>       }; >>> >> +     u32 ret = SECCOMP_RET_ALLOW; >>> >>       const void *sc_ptr = (const void *)(uintptr_t)syscall; >>> >> >>> >> +     /* Ensure unexpected behavior doesn't result in failing open. */ >>> >> +     if (unlikely(current->seccomp.filter == NULL)) >>> >> +             ret = SECCOMP_RET_KILL; >>> > >>> > Is "seccomp.filter == NULL" really possible? >>> >>> It should not be, but I'm much more comfortable with this failing >>> closed. I think it's important to be as defensive as possible with >>> this code given its intended use. >> >> Can't resists... Sorry, I know I am troll but personally I think >> in this case the most defensive code is BUG_ON(->filter == NULL) >> or at least WARN_ON(). > > Linus will probably object because he objected (correctly) to a very > similar problem in my old vsyscall emulation series.  A userspace > security feature shouldn't have a failure mode in which it confuses > the kernel and results in an oops, unless the situation is really > unrecoverable.  So WARN_ON plus do_exit would be okay but BUG_ON would > not. Yeah, actually, add WARN_ON would be preferred here because it should be an impossible situation. It should still fail closed, though:     /* Ensure unexpected behavior doesn't result in failing open. */     if (WARN_ON(current->seccomp.filter == NULL))             return SECCOMP_RET_KILL; -Kees -- Kees Cook ChromeOS Security