[Qemu-devel] [Bug 584121] Re: migration always fails on 32bit qemu-kvm-0.12+ (sigsegv)

[Preston Connors]

I believe this issue to be affecting my systems too. I am willing to
provide more information as necessary.

Running two equivalent systems hardware wise - RAM type (not amount),
CPU, BIOS, Ubuntu version except qemu-kvm binary.

Both systems are running Ubuntu version 10.04.1 x86_64.

Running Ubuntu qemu-kvm 0.12.3+noroms-0ubuntu9.3 on both systems however
I have some running kvm processes that were running from qemu-kvm
version 0.12.3+noroms-0ubuntu9.2. The kvm processes running from qemu-
kvm version 0.12.3+noroms-0ubuntu9.2 are the kvm processes that are not
migrating.

Steps to reproduce:
1) Attempt to qemu+ssh migrate qemu-kvm 0.12.3+noroms-0ubuntu9.2 process to 
another server running either qemu-kvm 0.12.3+noroms-0ubuntu9.2 or qemu-kvm 
0.12.3+noroms-0ubuntu9.3 using command virsh migrate --live my_domain 
qemu+ssh://receiving.host/system
2) Wait for the migration to complete. It never completes and error 'qemu: 
warning: error while loading state for instance 0x0 of device 'ram'' is 
received on the receiving host.

During first migration attempt the guest will be created in the paused
state on the receiving host and the migration will not exit unless you
force quit (CTRL+C) the migration on the sending node. You can virsh
resume the domain on the receiving end and the guest will re-boot itself
and boot properly on the receiving end.

Further migration attempts with the same domain will fail with error
'error: Requested operation is not valid: domain is already active as
'my_domain''' on the sending host.

If you virsh destroy my_domain ; virsh undefine my_domain on the
receiving host (which will have this domain in a paused state)  you can
re-try the migration. After doing 'service libvirt-bin restart' on both
hosts you can re-try the migration multiple times. Around six tries of
restarting service libvirt-bin and also attempting the migration will
get the domain migrated over to the receiving host without any errors,
however during this migration that is supposedly successful the guest
domain will crash and re-boot itself on the receiving host.

After this crash/migration you can then migrate the domain back and
forth between hosts with no problems.

Log entries/errors generated during migration on receiving host:

/var/log/libvirt/qemu/$vm_name.log

LC_ALL=C PATH=/usr/local/sbin:/usr/local/bin:/usr/bin:/usr/sbin:/sbin:/bin 
QEMU_AUDIO_DRV=none /usr/bin/kvm -S -M pc-0.12 -enable-kvm -m 1024 -smp 1 -name 
27-1245 -uuid 47e4cbc4-bc91-12d4-3cd2-36b29de6f588 -chardev 
socket,id=monitor,path=/var/lib/libvirt/qemu/27-1245.monitor,server,nowait 
-monitor chardev:monitor -boot c -drive 
file=/dev/disk/by-id/scsi-3600144f0eae8810000004d0703550704,if=virtio,index=0,boot=on,format=raw,cache=none
 -net nic,macaddr=00:00:d1:d0:3e:b3,vlan=0,model=virtio,name=virtio.0 -net 
tap,fd=51,vlan=0,name=tap.0 -net 
nic,macaddr=00:00:0a:d0:3e:b3,vlan=1,model=virtio,name=virtio.1 -net 
tap,fd=52,vlan=1,name=tap.1 -chardev pty,id=serial0 -serial chardev:serial0 
-parallel none -usb -vnc 0.0.0.0:16,password -k en-us -vga cirrus -incoming 
tcp:0.0.0.0:49152 
char device redirected to /dev/pts/2
qemu: warning: error while loading state for instance 0x0 of device 'ram'
load of migration failed

Verbose libvirtd log output on receiving host:
http://pastebin.com/GPxLCjyS

Log entries/errors generated during migration on the sending host:

Verbose libvirtd log output on sending host:
http://pastebin.com/cMcU7SgS

If you would like any further information/output please let me know.

-- 
You received this bug notification because you are a member of qemu-
devel-ml, which is subscribed to QEMU.
https://bugs.launchpad.net/bugs/584121

Title:
  migration always fails on 32bit qemu-kvm-0.12+ (sigsegv)

Status in QEMU:
  Confirmed

Bug description:
  On a 32bit host (or when running 32bit userspace on 64bit host), migration 
always fails with a crash of qemu-kvm process.
See http://marc.info/?l=kvm&m=127351472231666 for more information.