[Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests?
I have a remote system that boots fine when I use: qemu -vnc :4 -kernel-kqemu -smp 2 -no-reboot -monitor unix:/vservers/qemu/cms1/cms1.sock,server,nowait -boot c -hda hda1.img -hdb swap.img -net nic,vlan=0,macaddr=00:16:3e:00:0f:00 -net tap,vlan=0,ifname=br0-cms1,script=/etc/qemu-ifup-br0 -m 500 But this is not using kqemu and very slow. When I boot the following, the guest never gets past the init process. According to top, the guest qemu system uses 101% of the hosts cpu amd x2 3800. The guest is ubuntu 7.4 running (postgesql, Apache, and ssh) qemu -vnc :4 -kernel-kqemu -no-reboot -monitor unix:/vservers/qemu/cms1/cms1.sock,server,nowait -boot c -hda hda1.img -hdb swap.img -net nic,vlan=0,macaddr=00:16:3e:00:0f:00 -net tap,vlan=0,ifname=br0-cms1,script=/etc/qemu-ifup-br0 -m 500 Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests? Is there a way in the proc file system on the guest to see kqemu if is working? Is there a plan to create a stand alone qemu-monitor that can be attached to a running qemu process? Thank you for your time and a great project, Nicholas A. Schembri State College, PA USA PS. qemu -smp 2 has 90 day's uptime on the production hardware.
Re: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests?
virtualBox support page talks about a race condition. Is this a race in the guest kernel or user land? It looks like I boot the kernel just fine. the cpu spikes around init time. - Original Message From: n schembr [EMAIL PROTECTED] To: qemu-devel@nongnu.org Sent: Wednesday, July 11, 2007 11:18:14 AM Subject: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests? I have a remote system that boots fine when I use: qemu -vnc :4 -kernel-kqemu -smp 2 -no-reboot -monitor unix:/vservers/qemu/cms1/cms1.sock,server,nowait -boot c -hda hda1.img -hdb swap.img -net nic,vlan=0,macaddr=00:16:3e:00:0f:00 -net tap,vlan=0,ifname=br0-cms1,script=/etc/qemu-ifup-br0 -m 500 But this is not using kqemu and very slow. When I boot the following, the guest never gets past the init process. According to top, the guest qemu system uses 101% of the hosts cpu amd x2 3800. The guest is ubuntu 7.4 running (postgesql, Apache, and ssh) qemu -vnc :4 -kernel-kqemu -no-reboot -monitor unix:/vservers/qemu/cms1/cms1.sock,server,nowait -boot c -hda hda1.img -hdb swap.img -net nic,vlan=0,macaddr=00:16:3e:00:0f:00 -net tap,vlan=0,ifname=br0-cms1,script=/etc/qemu-ifup-br0 -m 500 Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests? Is there a way in the proc file system on the guest to see kqemu if is working? Is there a plan to create a stand alone qemu-monitor that can be attached to a running qemu process? Thank you for your time and a great project, Nicholas A. Schembri State College, PA USA PS. qemu -smp 2 has 90 day's uptime on the production hardware.
Re: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests?
ok, This error is in kqemu. The guest boots with -no-kqemu option. The guest is booting ubuntu 7.04 linux 2.6.20-16-server. Note: the host is running 7.4 and I compiled with ./configure --prefix=/usr/local --cc=gcc-3.4 --host-cc=gcc-3.4 --kernel-path=/usr/src/linux-headers-2.6.20-16-server/ The host has a second guest running ubuntu dapper linux 2.6.15-28-686 qemu -vnc :1 -kernel-kqemu -daemonize -no-reboot -monitor unix:/vservers/qemu/mail1/mail1.sock,server,nowait -boot c -hda /dev/sdc -net nic,vlan=0,macaddr=00:16:3e:00:0f:10 -net tap,vlan=0,ifname=br0-mail1,script=/etc/qemu-ifup-br0 -m 512 I also found that qemu is not exiting after guest is halted with init 0. - Original Message From: n schembr [EMAIL PROTECTED] To: qemu-devel@nongnu.org Sent: Wednesday, July 11, 2007 12:27:16 PM Subject: Re: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests? virtualBox support page talks about a race condition. Is this a race in the guest kernel or user land? It looks like I boot the kernel just fine. the cpu spikes around init time. - Original Message From: n schembr [EMAIL PROTECTED] To: qemu-devel@nongnu.org Sent: Wednesday, July 11, 2007 11:18:14 AM Subject: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests? I have a remote system that boots fine when I use: qemu -vnc :4 -kernel-kqemu -smp 2 -no-reboot -monitor unix:/vservers/qemu/cms1/cms1.sock,server,nowait -boot c -hda hda1.img -hdb swap.img -net nic,vlan=0,macaddr=00:16:3e:00:0f:00 -net tap,vlan=0,ifname=br0-cms1,script=/etc/qemu-ifup-br0 -m 500 But this is not using kqemu and very slow. When I boot the following, the guest never gets past the init process. According to top, the guest qemu system uses 101% of the hosts cpu amd x2 3800. The guest is ubuntu 7.4 running (postgesql, Apache, and ssh) qemu -vnc :4 -kernel-kqemu -no-reboot -monitor unix:/vservers/qemu/cms1/cms1.sock,server,nowait -boot c -hda hda1.img -hdb swap.img -net nic,vlan=0,macaddr=00:16:3e:00:0f:00 -net tap,vlan=0,ifname=br0-cms1,script=/etc/qemu-ifup-br0 -m 500 Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests? Is there a way in the proc file system on the guest to see kqemu if is working? Is there a plan to create a stand alone qemu-monitor that can be attached to a running qemu process? Thank you for your time and a great project, Nicholas A. Schembri State College, PA USA PS. qemu -smp 2 has 90 day's uptime on the production hardware.
Re: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests?
Hi, On 11/07/07, n schembr [EMAIL PROTECTED] wrote: ok, This error is in kqemu. The guest boots with -no-kqemu option. The guest is booting ubuntu 7.04 linux 2.6.20-16-server. Note: the host is running 7.4 and I compiled with ./configure --prefix=/usr/local --cc=gcc-3.4 --host-cc=gcc-3.4 --kernel-path=/usr/src/linux-headers-2.6.20-16-server/ The host has a second guest running ubuntu dapper linux 2.6.15-28-686 qemu -vnc :1 -kernel-kqemu -daemonize -no-reboot -monitor unix:/vservers/qemu/mail1/mail1.sock,server,nowait -boot c -hda /dev/sdc -net nic,vlan=0,macaddr=00:16:3e:00:0f:10 -net tap,vlan=0,ifname=br0-mail1,script=/etc/qemu-ifup-br0 -m 512 Just an idea, if your host is SMP too: try booting the guest kernel with notsc param or setting qemu's affinity to a single CPU using taskset. It may help or not :) Cheers, Andrzej
Re: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests?
- Original Message From: andrzej zaborowski [EMAIL PROTECTED] To: qemu-devel@nongnu.org Sent: Wednesday, July 11, 2007 2:38:28 PM Subject: Re: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests? Hi, On 11/07/07, n schembr [EMAIL PROTECTED] wrote: ok, This error is in kqemu. The guest boots with -no-kqemu option. The guest is booting ubuntu 7.04 linux 2.6.20-16-server. Note: the host is running 7.4 and I compiled with ./configure --prefix=/usr/local --cc=gcc-3.4 --host-cc=gcc-3.4 --kernel-path=/usr/src/linux-headers-2.6.20-16-server/ The host has a second guest running ubuntu dapper linux 2.6.15-28-686 qemu -vnc :1 -kernel-kqemu -daemonize -no-reboot -monitor unix:/vservers/qemu/mail1/mail1.sock,server,nowait -boot c -hda /dev/sdc -net nic,vlan=0,macaddr=00:16:3e:00:0f:10 -net tap,vlan=0,ifname=br0-mail1,script=/etc/qemu-ifup-br0 -m 512 Just an idea, if your host is SMP too: try booting the guest kernel with notsc param or setting qemu's affinity to a single CPU using taskset. It may help or not :) Cheers, Andrzej ./taskset -c 1 qemu ...etc worked great. What will notsc do to the guest kernel? You just saved my project, thank you. Nicholas A. Schembri State College PA USA For google qemu race condition kqemu ubuntu 7.4 feisty lockup hang
Re: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests?
On 11/07/07, n schembr [EMAIL PROTECTED] wrote: ./taskset -c 1 qemu ...etc worked great. What will notsc do to the guest kernel? Ok, then it is the rdtsc issue with kqemu. See http://lists.gnu.org/archive/html/qemu-devel/2007-03/msg00652.html for explanation and proposed fix. Cheers, Andrzej
Re: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests?
- Original Message From: andrzej zaborowski [EMAIL PROTECTED] To: qemu-devel@nongnu.org Sent: Wednesday, July 11, 2007 4:19:23 PM Subject: Re: [Qemu-devel] Is there a patch for kqemu 1.3.0pre11 and qemu 0.9 to allow accelerated smp guests? On 11/07/07, n schembr [EMAIL PROTECTED] wrote: ./taskset -c 1 qemu ...etc worked great. What will notsc do to the guest kernel? Ok, then it is the rdtsc issue with kqemu. See http://lists.gnu.org/archive/html/qemu-devel/2007-03/msg00652.html for explanation and proposed fix. Cheers, Andrzej It it better to boot the guest with notsc and have processor migration or set qemu to one processor? It seems cleaner to set notsc in grub and let it fly. In my case, I'm booting a real hard drive inside qemu. If the host system's hard drive fails, the reboot will load the host off of the guests hard drive. If I add notsc to grub will it still boot on the real hardware? I'm 11 hours from the host machine. :)