Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757626AbaFYREH (ORCPT ); Wed, 25 Jun 2014 13:04:07 -0400 Received: from mail-la0-f43.google.com ([209.85.215.43]:33519 "EHLO mail-la0-f43.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757421AbaFYREE (ORCPT ); Wed, 25 Jun 2014 13:04:04 -0400 MIME-Version: 1.0 In-Reply-To: References: <1403642893-23107-1-git-send-email-keescook@chromium.org> <1403642893-23107-6-git-send-email-keescook@chromium.org> <20140625135121.GB7892@redhat.com> From: Andy Lutomirski Date: Wed, 25 Jun 2014 10:03:42 -0700 Message-ID: Subject: Re: [PATCH v8 5/9] seccomp: split mode set routines To: Kees Cook Cc: Oleg Nesterov , LKML , "Michael Kerrisk (man-pages)" , Alexei Starovoitov , Andrew Morton , Daniel Borkmann , Will Drewry , Julien Tinnes , David Drysdale , Linux API , "x86@kernel.org" , "linux-arm-kernel@lists.infradead.org" , linux-mips@linux-mips.org, linux-arch , linux-security-module Content-Type: text/plain; charset=UTF-8 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Jun 25, 2014 at 9:54 AM, Kees Cook wrote: > On Wed, Jun 25, 2014 at 9:10 AM, Andy Lutomirski wrote: >> On Wed, Jun 25, 2014 at 7:51 AM, Kees Cook wrote: >>> On Wed, Jun 25, 2014 at 6:51 AM, Oleg Nesterov wrote: >>>> On 06/24, Kees Cook wrote: >>>>> >>>>> +static inline void seccomp_assign_mode(struct task_struct *task, >>>>> + unsigned long seccomp_mode) >>>>> +{ >>>>> + BUG_ON(!spin_is_locked(&task->sighand->siglock)); >>>>> + >>>>> + task->seccomp.mode = seccomp_mode; >>>>> + set_tsk_thread_flag(task, TIF_SECCOMP); >>>>> +} >>>> >>>> OK, but unless task == current this can race with secure_computing(). >>>> I think this needs smp_mb__before_atomic() and secure_computing() needs >>>> rmb() after test_bit(TIF_SECCOMP). >>>> >>>> Otherwise, can't __secure_computing() hit BUG() if it sees the old >>>> mode == SECCOMP_MODE_DISABLED ? >>>> >>>> Or seccomp_run_filters() can see ->filters == NULL and WARN(), >>>> smp_load_acquire() only serializes that LOAD with the subsequent memory >>>> operations. >>> >>> Hm, actually, now I'm worried about smp_load_acquire() being too slow >>> in run_filters(). >>> >>> The ordering must be: >>> - task->seccomp.filter must be valid before >>> - task->seccomp.mode is set, which must be valid before >>> - TIF_SECCOMP is set >>> >>> But I don't want to impact secure_computing(). What's the best way to >>> make sure this ordering is respected? >> >> Remove the ordering requirement, perhaps? >> >> What if you moved mode into seccomp.filter? Then there would be >> little reason to check TIF_SECCOMP from secure_computing; instead, you >> could smp_load_acquire (or read_barrier_depends, maybe) seccomp.filter >> from secure_computing and pass the result as a parameter to >> __secure_computing. Or you could even remove the distinction between >> secure_computing and __secure_computing -- it's essentially useless >> anyway to split entry hook approaches like my x86 fastpath prototype. > > The TIF_SECCOMP is needed for the syscall entry path. The check in > secure_computing() is just because the "I am being traced" trigger > includes a call to secure_computing, which filters out tracing > reasons. Right. I'm suggesting just checking a single indication that seccomp is on from the process in the seccomp code so that the order doesn't matter. IOW, TIF_SECCOMP causes __secure_computing to be invoked, but the race only seems to matter because of the warning and the BUG. I think that both can be fixed if you merge mode into filter so that __secure_computing atomically checks one condition. > > Your fast path work would clean a lot of that up, as you say. But it > still doesn't change the ordering check here. TIF_SECCOMP indicates > seccomp.mode must be checked, so that ordering will remain, and if > mode == FILTER, seccomp.filter must be valid. > > Isn't there a way we can force the assignment ordering in seccomp_assign_mode()? Write the filter, then smp_mb (or maybe a weaker barrier is okay), then set the bit. --Andy > > -Kees > > -- > Kees Cook > Chrome OS Security -- Andy Lutomirski AMA Capital Management, LLC -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/