Re: [Qemu-devel] [ANNOUNCE] QEMU 1.5.0-rc2 is now available

[mdroth <mdroth@linux.vnet.ibm.com>]

On Wed, May 15, 2013 at 06:53:47PM -0500, Anthony Liguori wrote:
> Hi,
> 
> On behalf of the QEMU Team, I'd like to announce the availability of the
> third release candidate for the QEMU 1.5 release.  This release is meant
> for testing purposes and should not be used in a production environment.
> 
> http://wiki.qemu.org/download/qemu-1.5.0-rc2.tar.bz2
> 
> You can help improve the quality of the QEMU 1.5 release by testing this
> release and reporting bugs on Launchpad:
> 

Sorry to chime in on this so late in the cycle, but I just noticed what
seems to be a pretty serious problem with migration to/from 1.4. This is
the failure for 1.4 -> 1.5-rc2

(qemu) migrate unix:/tmp/migrate.sock
Unknown savevm section or instance '0000:00:03.0/virtio-net' 0

Configuration:

source: v14/x86_64-softmmu/qemu-system-x86_64 -enable-kvm -L v14-bios -M
pc-i440fx-1.4 -m 512M -kernel boot/vmlinuz-x86_64 -initrd
boot/test-initramfs-x86_64.img.gz -vga std -append seed=1234 -drive
file=disk1.img,if=virtio -drive file=disk2.img,if=virtio -net
nic,model=virtio -net user -monitor unix:/tmp/vm-hmp.sock,server,nowait
-qmp unix:/tmp/vm-qmp.sock,server,nowait -vnc :100

target: v15rc2/x86_64-softmmu/qemu-system-x86_64 -enable-kvm -L temp-bios
-M pc-i440fx-1.4 -m 512M -kernel boot/vmlinuz-x86_64 -initrd
boot/test-initramfs-x86_64.img.gz -vga std -append seed=1234 -drive
file=disk1.img,if=virtio -drive file=disk2.img,if=virtio -net
nic,model=virtio -net user -incoming unix:/tmp/migrate.sock -monitor
unix:/tmp/vm-hmp-incoming.sock,server,nowait -qmp
unix:/tmp/vm-qmp-incoming.sock,server,nowait -vnc :101
QEMU 1.4.0 monitor - type 'help' for more information

This seems to have been introduced with the virtio refactoring:

commit e37da3945fa2fde161e1b217f937fc318c4b7639
Author: KONRAD Frederic <fred.konrad@greensocs.com>
Date:   Thu Apr 11 16:29:58 2013 +0200

    virtio-net-pci: switch to the new API.
    
    Here the virtio-net-pci is modified for the new API. The device
    virtio-net-pci extends virtio-pci. It creates and connects a
    virtio-net-device during the init. The properties are not changed.
    
    Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
    Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
    Message-id: 1365690602-22729-4-git-send-email-fred.konrad@greensocs.com
    Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

And if we roll that back, we have similar failures for virtio-blk, and most
likely the other virtio devices touched by the refactoring.

The issue seems to be a change the way section id strings are generated in
vmstate_register(). In v1.4.x we had:

se->instance_id: 0, se->idstr: 0000:00:03.0/virtio-net

In v1.5.0-rc2 we have:

se->instance_id: 0, se->idstr: virtio-net

This seems to be due to the fact that these devices now sit on a
TYPE_VIRTIO_BUS that has no implementation of TYPE_BUS's get_dev_path()
interface, which is what savevm uses to calculate the id prefix for
se->idstr.

Prior to the refactoring, the device sat on a TYPE_PCI_BUS which used
pcibus_get_dev_path() to calculate this.

I'm not sure what the best fix is for this. I looking at implementing
get_dev_path() for TYPE_VIRTIO_BUS, but to maintain migration
compatibility we'd end up baking in PCI-specific stuff which from what
I gather is exactly what we were trying to avoid there.

I think adding a compat string property to TYPE_VIRTIO_DEVICE and having
that get set somewhere like virtio_bus_plug_device() is a better
approach, but vmstate_register() gets call during TYPE_VIRTIO_DEVICE
init which I think happens before then.

Still looking at it but if someone more familiar with this code has
some ideas or wants to whip up a patch please jump right in.

> 
> Regards,
> 
> Anthony Liguori
> 
>