-
Bug
-
Resolution: Done-Errata
-
Normal
-
rhel-9.4
-
qemu-kvm-8.2.0-3.el9
-
None
-
None
-
rhel-sst-virtualization
-
ssg_virtualization
-
19
-
23
-
None
-
False
-
-
None
-
None
-
Pass
-
None
-
If docs needed, set a value
-
-
Unspecified
-
None
Description of problem:
qemu crashed when migrate guest with blob resources enabled
Version-Release number of selected component (if applicable):
libvirt-9.5.0-6.el9.x86_64
qemu-kvm-8.0.0-13.el9.x86_64
How reproducible:
100%
Steps to Reproduce:
1. prepare a guest with the following xml snippet
- virsh dumpxml lizhu --xpath //video
<video>
<model type="virtio" heads="1" primary="yes" blob="on"/>
<alias name="video0"/>
<address type="pci" domain="0x0000" bus="0x00" slot="0x01" function="0x0"/>
</video>
2. migrate the guest to another host
- virsh migrate lizhu qemu+ssh://$target_host/system --verbose --live
Migration: [98.32 %]error: operation failed: domain is not running
3. check the qemu log on target host
#cat /var/log/libvirt/qemu/lizhu.log
...
2023-08-30 14:35:30.653+0000: Domain id=1 is tainted: host-cpu
char device redirected to /dev/pts/2 (label charserial0)
2023-08-30T14:35:37.352236Z qemu-kvm: Failed to load virtio-gpu:virtio-gpu
2023-08-30T14:35:37.352266Z qemu-kvm: error while loading state for
instance 0x0 of device '0000:00:01.0/virtio-gpu'
2023-08-30T14:35:37.352442Z qemu-kvm: load of migration failed: Invalid argument
2023-08-30 14:35:37.754+0000: shutting down, reason=crashed
4. check the qemu log on source host
#cat /var/log/libvirt/qemu/lizhu.log
...
2023-08-30 14:35:30.974+0000: initiating migration
2023-08-30 14:35:37.789+0000: shutting down, reason=crashed
5. check the guest states on source host
- virsh domstate lizhu --reason
shut off (unknown)
Actual results:
Migration failed, qemu crashed
Expected results:
Qemu should not crash.
- depends on
-
RHEL-14111 Rebase qemu-kvm to QEMU 8.2.0
-
- Closed
-
- external trackers
- links to
-
RHBA-2023:121202
qemu-kvm bug fix and enhancement update
- mentioned on
[RHEL-7565] qemu crashed when migrate guest with blob resources enabled
Simply verify the latest behavior - virtio-vga/gpu is not migratble:.
step1:
Boot a VM with blob enabled:
... <memoryBacking> <source type='memfd'/> </memoryBacking> ... <video> <model type='virtio' heads='1' primary='yes' blob="on"/> </video> ...
step2:
Save the VM's state to a file:
# virsh save rhel current.stat
After step2, it will prompt:
error: Failed to save domain 'rhel' to current.stat
error: Requested operation is not valid: cannot migrate domain: virtio-gpu blob VMs are currently not migratable.
Fix included in qemu-kvm-8.2.0-3.el9
Fixed by merge request 'virtio-gpu: block migration of VMs with blob=true' ( https://gitlab.com/redhat/centos-stream/src/qemu-kvm/-/merge_requests/217 )
gitlab-bot
added a comment - Miroslav Rezanina mentioned this issue in a merge request of Red Hat / centos-stream / rpms / qemu-kvm on branch next:
Update to qemu-kvm-8.2.0-3.el9
gitlab-bot
added a comment - Miroslav Rezanina mentioned this issue in a merge request of Red Hat / centos-stream / rpms / qemu-kvm on branch next : Update to qemu-kvm-8.2.0-3.el9 
Marc-Andre Lureau
added a comment - upstream patches for 9.0 "[PATCH 0/2] virtio-gpu: fix blob scanout post-load". Most likely not good candidates for backport, given they change the migration stream version.
Marc-Andre Lureau
added a comment - upstream patches for 9.0 " [PATCH 0/2] virtio-gpu: fix blob scanout post-load". Most likely not good candidates for backport, given they change the migration stream version. Marc-Andre Lureau
added a comment - rhn-support-zhguo as long as support for blob migration isn't a customer requirement, I think this is the best thing to do. I am going to send a patch for upstream, but it will bump the migration stream version, and hence not a suitable change for a stable version.
Marc-Andre Lureau
added a comment - rhn-support-zhguo as long as support for blob migration isn't a customer requirement, I think this is the best thing to do. I am going to send a patch for upstream, but it will bump the migration stream version, and hence not a suitable change for a stable version. Sigh, this is going to be a bit more difficult to handle. Maybe we should just revert the support for migration at this point in RHEL (revert commit 10b9ddbc83b94986cbdf989e26fb7269fb2e9f72).
I do agree no crash happen is the best option for rhel 9.4. If you choose this path, I can modify the title of this Jira issue and open a new issue for tracking the live migration capability. WDYT?
Marc-Andre Lureau
added a comment - I didn't realize blob resources could be used for scanouts, and didn't reach that code path during testing.
Sigh, this is going to be a bit more difficult to handle. Maybe we should just revert the support for migration at this point in RHEL (revert commit 10b9ddbc83b94986cbdf989e26fb7269fb2e9f72).
Marc-Andre Lureau
added a comment - I didn't realize blob resources could be used for scanouts, and didn't reach that code path during testing.
Sigh, this is going to be a bit more difficult to handle. Maybe we should just revert the support for migration at this point in RHEL (revert commit 10b9ddbc83b94986cbdf989e26fb7269fb2e9f72). 
Since the problem described in this issue should be resolved in a recent advisory, it has been closed.
For information on the advisory (Moderate: qemu-kvm security update), and where to find the updated files, follow the link below.
If the solution does not work for you, open a new bug report.
https://access.redhat.com/errata/RHSA-2024:2135