From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264359AbUBEFS5 (ORCPT ); Thu, 5 Feb 2004 00:18:57 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264472AbUBEFS5 (ORCPT ); Thu, 5 Feb 2004 00:18:57 -0500 Received: from sitemail3.everyone.net ([216.200.145.37]:56993 "EHLO omta06.mta.everyone.net") by vger.kernel.org with ESMTP id S264359AbUBEFSz (ORCPT ); Thu, 5 Feb 2004 00:18:55 -0500 Content-Type: text/plain Content-Disposition: inline Content-Transfer-Encoding: 7bit Mime-Version: 1.0 X-Mailer: MIME-tools 5.41 (Entity 5.404) Date: Wed, 4 Feb 2004 21:17:36 -0800 (PST) From: john moser To: linux-kernel@vger.kernel.org Subject: US/KS performance question Reply-To: bluefoxicy@linux.net X-Originating-Ip: [68.33.187.247] Message-Id: <20040205051736.89307CF33@sitemail.everyone.net> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org CC me all replies. I'm doing a little work on ACLs. Instead of a LSM, I'm writing up a little bit of code and some hooks to allow a userspace daemon to do the decision making. The plan is as follows: - acld connects to kernel, sleeps. - Kernel needs to ask something of acld, ques request in US - Kernel wakes acld up - Kernel puts the requesting thread to sleep - acld processes requests - Each request processed is followed by a syscall to the kernel with the result - Kernel takes the result, finds the thread that wanted it, wakes thread up - When there's no more requests, acld sleeps Now, I have two major bottlenecks: - Userspace acld MAC daemon has to make a syscall and pass data from US to KS - Kernel has to sometimes wake up acld and always has to wake up the requesting thread after getting the result back from acld Any ideas on how much overhead this could add? If it takes only a few more microseconds to wake up acld, sleep the process, process the syscall, and finally wake the process; as opposed to calling a kernel function to make the decision, then it should be okay. If it's going to be some 10 or 20 miliseconds per request, I might have problems. If it's going to take more than 100 mS for this, I need a redesign. _____________________________________________________________ Linux.Net -->Open Source to everyone Powered by Linare Corporation http://www.linare.com/