On Thu, Sep 11, 2014 at 09:11:40AM -0500, Michael Roth wrote: > Quoting Dietmar Maurer (2014-09-11 04:55:05) > > Seems that we get a crash when we live-migrate a VMs using virtio-net-pci > > (vhost=on). > > This worked in 2.1.0. Any ideas? > > Is this what you're seeing on the source side? > > qemu-system-x86_64: /home/mdroth/w/qemu3.git/hw/net/virtio-net.c:1348: > virtio_net_save: Assertion `!n->vhost_started' failed. > > If so I can reproduce on both stable and current master. A bisect of master > points to the following commit. > 269bd822e7f5ab80048b05fb7076236ed66ffbce is the first bad commit > commit 269bd822e7f5ab80048b05fb7076236ed66ffbce > Author: Michael S. Tsirkin <m...@redhat.com> > Date: Thu Sep 4 13:32:54 2014 +0300 > > virtio: don't call device on !vm_running > > On vm stop, virtio changes vm_running state > too soon, so callbacks can get envoked with > vm_running = false; > > Cc: qemu-sta...@nongnu.org > Cc: Jason Wang <jasow...@redhat.com> > Signed-off-by: Michael S. Tsirkin <m...@redhat.com> > Signed-off-by: Stefan Hajnoczi <stefa...@redhat.com> > > Any ideas Michael/Jason?
OK, I sent two patches to fix this. Will send upstream shortly, too. > This is with: > > qemu-system-x86_64 -enable-kvm -L build/build3-bios -M pc-i440fx-2.1 -m 512M > -kernel boot/vmlinuz-x86_64 -initrd boot/test-initramfs-x86_64.img.gz -vga > cirrus -append seed=1234 -drive file=disk1.img,if=virtio,cache=directsync > -drive file=disk2.img,if=virtio -device virtio-net-pci,netdev=net0 -netdev > tap,id=net0,vhost=on,script=/etc/qemu-ifup -incoming unix:/tmp/migrate.sock > -monitor unix:/tmp/vm-hmp-incoming.sock,server,nowait -qmp > unix:/tmp/vm-qmp-incoming.sock,server,nowait > > Since this is such a common use-case of live migration I'll likely be putting > out > a minor 2.1.2 update once we identify the fix. I unfortunately didn't have > vhost > enabled in the migration tests, which was a pretty big oversight on my part. Mee too :( > > > > > I am pleased to announce that the QEMU v2.1.1 stable release is now > > > available > > > at:
