From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Thu, 31 May 2001 03:13:30 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Thu, 31 May 2001 03:13:21 -0400 Received: from mailhost.idcomm.com ([207.40.196.14]:7812 "EHLO mailhost.idcomm.com") by vger.kernel.org with ESMTP id ; Thu, 31 May 2001 03:13:08 -0400 Message-ID: <3B15EF16.89B18D@idcomm.com> Date: Thu, 31 May 2001 01:13:26 -0600 From: "D. Stimits" Reply-To: stimits@idcomm.com X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.4.5-ac5-1 i686) X-Accept-Language: en MIME-Version: 1.0 To: kernel-list Subject: missing sysrq 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 compiled the magic sysrq as enabled in most kernels I've used for quite a while. Most recently 2.4.5-ac5, on a RH 7.1 SMP machine with APIC disabled. The /etc/sysctl.conf contains this line: kernel.sysrq = 1 However, if I go to /proc/sys/kernel/sysrq does not exist. Nor is it possible for root to echo to that file name. Attempts to use the magic sysrq keys, such as for sync, prove that it truly is not enabled, or is otherwise missing. Has something changed in the enabling of magic sysrq? Or is this one of those strange "it should be there" things? A second observation, maybe related (maybe not), is the /var/log/messages line: sysctl: error: permission denied on key 'vm.freepages' This particular line has occurred since install of RH 7.1, even with its original kernel, but continues into 2.4.5-ac5. The relevant line in /etc/sysctl.conf: vm.freepages = 383 766 1149 This line is how the original RH 7.1 install set it up. This particular machine has 256 MB of ram, and somewhat over a 1 GB of swap. Is the vm.freepages not intended to be set in /etc/sysctl.conf? Or maybe the specs are off for this particular hardware? I know a lot of vm changes are going on in the kernel, and wondering if this could be something that used to be supported but no longer is. D. Stimits, stimits@idcomm.com