-
Bug
-
Resolution: Unresolved
-
Normal
-
rhel-9.1.0
-
None
-
Moderate
-
rhel-sst-virtualization-storage
-
ssg_virtualization
-
5
-
QE ack, Dev ack
-
False
-
-
None
-
None
-
Fail
-
None
-
If docs needed, set a value
-
-
Unspecified
-
None
Up until now, qemu-kvm has had a dependency on the qemu-block-curl plugin, which allows qemu guests to use http/https network locations as a disk source. However, in rhel 9, this dependency is being removed from qemu-kvm, which results in issues such as bug #2014229.
The long-term solution is to switch to using the more featureful and maintained nbdkit curl plugin (https://libguestfs.org/nbdkit-curl-plugin.1.html) within libvirt.
From an email from Rich:
It would need to instead run an nbdkit subprocess first, eg:
nbdkit -U /tmp/randomly-named-sock --exit-with-parent -f -r \
--pidfile /tmp/randomly-named-file \
curl http://foo protocols=http,https,ftp,ftps &
- wait for pidfile to get created
- pass nbd:unix:/tmp/randomly-named-sock as qemu parameter
- or:
- <disk type='network' device='disk'>
- <driver name='qemu'/>
- <source protocol='nbd'>
- <host transport='unix' socket='/tmp/randomly-named-sock'/>
- </source>
- <target dev='vda' bus='virtio'/>
- </disk>
- wait for qemu then:
kill `cat /tmp/randomly-named-file`
rm /tmp/randomly-named-sock
(or nbdkit will clean itself up because of --exit-with-parent)
- is blocked by
-
RHEL-56029 SELinux prevents the rpc-virtqemud from starting a guest which uses nbdkit
- Planning
-
RHEL-5173 libvirt is unable to start nbdkit process when SELinux is enforcing
- Closed
-
RHEL-5174 Create a selinux policy for nbdkit
- Closed
- external trackers
- links to
- mentioned on