I have a 9p mount inside a VM. I take a snapshot of the VM state with savevm <snapshot>, stop the VM, then restart it with that snapshot (with kvm -loadvm <snapshot>). When I try to interact with the mount by either umounting orlsing, the VM hangs in kernel-space.

The full command line with which I am launching the snapshot looks like this:

qemu-system-x86_64 -nographic -monitor telnet::6440,server,nowait \
  -m 1280M -balloon virtio -bios \
  external_sources/seabios/out/bios.bin \
  -drive file=testvm/deb.instance.integrate,if=virtio \
  -loadvm loaded \
  -virtfs local,path=/tmp/mymount,security_model=none,mount_tag=mymount

This happens with 64-bit Debian squeeze guests with kernel versions 2.6.32 and 2.6.38. The VERSION of qemu-kvm is 0.14.50, on an Ubuntu 10.04 amd64 host with kernel 2.6.32-30.

I am at a loss as to how to diagnose this problem further; my only option at this point is to replace 9p with a network filesystem of some sort.

2 Answers 2


According to the qemu-devel mailing list, virtio-9p doesn't currently support live migration at all, so savevm/loadvm can't work.

  1. this has nothing to do with kvm as such - qemu manages the snapshots.
  2. savevm/loadvm is basically migration into a file. Never tried it with 9p, as I'[m not too familiar with it, but if there's timing involved in the FS model, this might be the issue.
  3. ubuntu as a kvm host has proven itself (at least to me) to be suboptimal. dunno if it's the packaging or the OS itself, but the same versions of kvm and qemu have always been solid for me on fedora.
  4. I'd start by a) testing on a fedora or rhel box b) posting to the linux-kvm mailing list c)checking with whoever supports 9p if it can survive a live migration, hybernation and suspend.
  • 1
    I'm not using the Ubuntu-packaged qemu-kvm, but a source build. I've posted to qemu-devel, so I'll update here if I get an answer.
    – regularfry
    May 24, 2011 at 8:57

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .