From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S268979AbUIXRw0 (ORCPT ); Fri, 24 Sep 2004 13:52:26 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S268971AbUIXRw0 (ORCPT ); Fri, 24 Sep 2004 13:52:26 -0400 Received: from rproxy.gmail.com ([64.233.170.207]:50058 "EHLO mproxy.gmail.com") by vger.kernel.org with ESMTP id S268980AbUIXRt6 (ORCPT ); Fri, 24 Sep 2004 13:49:58 -0400 Message-ID: <4ef5fec604092410491fbae33b@mail.gmail.com> Date: Fri, 24 Sep 2004 10:49:29 -0700 From: Martin Peck Reply-To: Martin Peck To: linux-kernel@vger.kernel.org Subject: Longhaul cpu frequency scaling leads to segmentation violation in userspace Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org I have encountered a strange scenario using 2.6.8.1 on a VIA C5P mini-itx system when Longhaul CPU frequency scaling is enabled. When filling a buffer with entropy via xstore in kernel space a tight loop can be performed without any problems. If the xstore is invoked in the same iterative manner in user space for a sufficient number of iterations (~ 120 or more) the application is terminated due to a segmentation violation. Adding a short delay between iterations or turning off CPU frequency scaling resolves the problem. Is there something associated with CPU frequency scaling and processor intensive operations in user space that leads to unexpected timeslice overrun (or any other side effect) causing a segmentation violation? This behavior is very strange; any insight would be appreciated. Regards,