On Fri, 15 Oct 2010 17:35:18 +0200 Hans de Goede <hdegoede@xxxxxxxxxx> wrote: > Hi All, > > It took me a while to figure the below out, I hope it gives some clues > into the problem I'm seeing. I'm running qemu compiled from git > from the following repo: > > http://cgit.freedesktop.org/spice/qemu/ > > From the spice.v20 branch. > > So this is basically qemu HEAD (not qemu-kvm but plain qemu), with spice > patches added. I can reproduce this without the use of the spice vga device > however! > > When I start Fedora 14 i386 inside a qemu vm with the following cmdline: > qemu-system-x86_64 -enable-kvm -cpu host \ > -m 1024 -name F14 -smp 2 \ > -drive file=/mnt/rhel6_x86_64/images/f14-i386.qcow2,if=virtio,media=disk \ > -net nic,macaddr=52:54:00:7a:b4:7d,vlan=0,model=virtio,name=virtio.0 -net tap,vlan=0 \ > -monitor stdio > > And then: > -wait till it has booted into gdm > -switch to tty2 using sendkey ctrl+alt+f2 from the monitor > -login as root > -run the following: > while true; fortune; sleep 1; done; > -wait (10 minutes or so at a maximum at my machine) > -note qemu cpu load goes to 100% all of a sudden, monitor is dead, > ctrl+alt+1 to go to serial console is dead, guest is dead > -attach a debugger, see it is executing guest instructions > (doing kvm_ioctl), also if you let it hang long enough you > will get BUG: soft lockup in dmesg (once you unstuck it) > -detach, hang is gone everything works again, entered > monitor commands during the hang are executed, etc. > > Alternatively sending SIGCHLD to the qemu process also unstucks qemu + > the guest. I've seen similar things on uniprocessor PowerPC, caused by a dyntick timer not getting reprogrammed after expiring. It went away when I switched to using the I/O thread. -Scott -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
