On Thu, 2015-09-03 at 10:17 +0100, Daniel P. Berrange wrote:
This happens because we assume rtl8139 will be available, but that's not the case eg. on RHEL ppc64.
Right, but changing upstream to default to virtio, just because RHEL downstream disabled other devices is not right.
Just assuming rtl8139 will be available, as we're doing right now, is not very nice either :)
Either upstream will need to dynamically change if it finds the device not available, or downstream RHEL libvirt should have a downstream only patch to workaround the RHEL QEMU PPC64 changes.
Yeah, Martin suggested doing something similar to the first option as well. Let's just probe for a bunch of network devices and use the first one that's available, okay? 1. rtl8139 2. e1000 3. virtio-net Any other we should try? Any drawbacks to this approach? Cheers. -- Andrea Bolognani Software Engineer - Virtualization Team