Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753360AbaBCWJG (ORCPT ); Mon, 3 Feb 2014 17:09:06 -0500 Received: from mail-vb0-f52.google.com ([209.85.212.52]:48681 "EHLO mail-vb0-f52.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751195AbaBCWJE (ORCPT ); Mon, 3 Feb 2014 17:09:04 -0500 MIME-Version: 1.0 In-Reply-To: <1400745.hJLGxDlRLu@x2> References: <1400745.hJLGxDlRLu@x2> From: Andy Lutomirski Date: Mon, 3 Feb 2014 14:08:43 -0800 Message-ID: Subject: Re: [PATCH] audit: Only use the syscall slowpath when syscall audit rules exist To: Steve Grubb Cc: linux-audit@redhat.com, "linux-kernel@vger.kernel.org" , Andi Kleen , Oleg Nesterov , Eric Paris Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Feb 3, 2014 at 12:23 PM, Steve Grubb wrote: > On Monday, February 03, 2014 09:53:23 AM Andy Lutomirski wrote: >> This toggles TIF_SYSCALL_AUDIT as needed when rules change instead of >> leaving it set whenever rules might be set in the future. This reduces >> syscall latency from >60ns to closer to 40ns on my laptop. > > Does this mean that we have processes that don't have the TIF_SYSCALL_AUDIT > flag set? When rules get loaded, how do we get the flag put back into all > processes? By looping over all processes and setting the flag, which is what my patch does. > > The theory of ops is supposed to be that for anyone not needing audit, there > is only the cost of "if (tif & TIF_SYSCALL_AUDIT)". On current kernels *all* processes have TIF_SYSCALL_AUDIT, even if they don't need auditing because there's nothing to audit. So everything pays the full cost. > That should be it. If you > have audit enabled or had it enabled (which means it might be loaded with new > rules), we want to inspect the syscall. > My point is that there's nothing to inspect -- there are no rules. Unless the audit code needs to do something just in case a non-syscall audit event gets written, in which case the audit code should IMO be fixed. (This is what Eric is talking about, I think.) --Andy -- 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/