From mboxrd@z Thu Jan 1 00:00:00 1970 From: Oliver Paukstadt Subject: Re: [RFC] kvm-s390: userspace snapshot Date: Tue, 10 Jun 2008 07:55:58 +0200 Message-ID: <1213077358.3120.33.camel@nerve> References: <1212767659.8861.11.camel@cotte.boeblingen.de.ibm.com> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: kvm@vger.kernel.org, Avi Kivity , aliguori@us.ibm.com, jblunck@suse.de, ihno@suse.de, rvdheij@gmail.com, rusty@rustcorp.com.au, Christian Borntraeger To: Carsten Otte Return-path: Received: from outpost.paukstadt.de ([213.239.197.141]:39186 "EHLO mx01.paukstadt.de" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1750728AbYFJGCQ (ORCPT ); Tue, 10 Jun 2008 02:02:16 -0400 In-Reply-To: <1212767659.8861.11.camel@cotte.boeblingen.de.ibm.com> Sender: kvm-owner@vger.kernel.org List-ID: On Fri, 2008-06-06 at 17:54 +0200, Carsten Otte wrote: > This patch is a full snapshot of "kuli", our current userspace for > kvm. > It is <3000 lines of code, and it contains a bootloader as well as > virtio backeds for console, network, and block. It's command line > syntax > is same as the common kvm userspace, but not all options are > supported. > See --help for details. > Our next step will be to get rid of the guest phys == user virt > address > mapping, so that we can have the guest memory anywhere on a megabyte > boundary in userland. After that, we'd like to integrate this into the > common kvm userspace, it is not intended for customer production use > but > is a good starting point to explore what kvm can look alike on s390. > > Have fun reading I started playing around with the stuff. I was able to get a running host kernel based on yesterdays kvm.git + Christian's VIRTIO_BLK_F_BLK_SIZE patch + Frank's latest qeth fixes. I decided to use the same kernel/initrd for my first guest system and expected the kernel to run at least through initial setup code, but there was an unexpected intercept: init_logging: KULI logging initialized do_ipl: perparing for IPL get_pages: map guest memory at 0x10000000 (pfn: 65536) get_pages: map guest memory at 0x10001000 (pfn: 65537) get_pages: map guest memory at 0x10003000 (pfn: 65539) get_pages: map guest memory at 0x10005000 (pfn: 65541) load_ipl: Booting from kernel image file "image-2.6.26-rc5kvm-20080609-01433-gdf4245d-dirty" load_from_files: loading ramdisk file initrd-2.6.26-rc5kvm-20080609-01433-gdf4245d-dirty at 800000 load_from_files: loading kernel parameter file parmfile do_ipl: IPL from address 10000 launch_cpu_ipl: starting guest (ipl) run_cpu: cpu 0: activated, running work... handle_should_not_happen: PANIC:cpu 0 caught unexpected intercept. Magic number: 0x20 Did I miss anything, like setting up virtual consoles, addressing mode or is there another kernel option for guests I did not see? Documentation is a little short, any hint appreciated ;-) I did this test running the host system as guest in z/VM 5.3 on z900, is this "supported"? Regards, Oliver