From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Wed, 31 Jan 2001 23:30:06 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Wed, 31 Jan 2001 23:29:57 -0500 Received: from bluebox.ne.mediaone.net ([24.91.117.20]:17414 "EHLO osiris.978.org") by vger.kernel.org with ESMTP id ; Wed, 31 Jan 2001 23:29:45 -0500 Date: Wed, 31 Jan 2001 23:29:44 -0500 From: Brian Ristuccia To: linux-kernel@vger.kernel.org Cc: user-mode-linux-devel@lists.sourceforge.net, Alex Pennace , Dennis Ristuccia Subject: usermode linux hoses 2.2.18-SMP host machine when run from regular user account Message-ID: <20010131232944.J16908@osiris.978.org> Mail-Followup-To: brianr@osiris.978.org, linux-kernel@vger.rutgers.edu, user-mode-linux-devel@lists.sourceforge.net Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.3.12i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org When run from a normal user account with its current working directory on a NFS filesystem, usermode linux causes the host machine's kernel to enter a hosed state. No processes (including UML) seem to respond, and the machine becomes unusable. I am running 2.2.18 built with SMP and hedrick's IDE patches on a dual Celeron 366@550 box with 384mB of ram. I have also reproduced the crash on a single Celeron 366@550 w/ 256mB of ram running 2.2.18-SMP w/ hedrick's IDE patches, and a dual PPro 166 w/ 256mB of RAM running 2.2.18pre23 unmodified from the version distributed by Alan Cox. The crash occurs with both the usermode linux 2.4.0 binary from their web site and the linux 2.4.1 binary there. I am using the root filesystem image file that comes in the tarball named deb-package-2.4.0.tar.bz2. The crash comes after the line "INIT: version 2.78 booting" during the virtual machine's boot sequence. The UML web site is at http://user-mode-linux.sourceforge.net/ The UML people report that UML doesn't work when being run from an NFS filesystem: UML doesn't work when it is being run from an NFS filesystem: This seems to be a similar situation with the resierfs problem above. Some versions of NFS seems not to handle mmap correctly, which UML depends on. The fix is to run UML in a non-NFS directory. The kernel binary and root filesystem be be mounted from NFS; UML itself just has to be running in a non-NFS directory. But I can't see how something done in userspace should hang the whole machine, especially when I set up low hard ulimits to prevent all system resources from being consumed. The statistics I'm able to gather from ctrl/shift-scrlock don't seem to indicate memory or process table overuse. Alt-Fx still switches consoles unless X is running. Ctrl/shift-scrlock still displays the usual debug stuff. The machine is still pingable and still accepts tcp/ip connections but you don't get any response from the listening daemon. Otherwise the box is hung. Ctrl-alt-del won't run the ctrlaltdel entry in inittab. Hardware reset is required to restore machine to normal operation. -- Brian Ristuccia brian@ristuccia.com bristucc@cs.uml.edu - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org Please read the FAQ at http://www.tux.org/lkml/