On Thu, Sep 03, 2015 at 09:53:26AM +0100, Daniel P. Berrange wrote:
On Thu, Sep 03, 2015 at 10:48:58AM +0200, Andrea Bolognani wrote:On Wed, 2015-09-02 at 16:08 -0400, Laine Stump wrote: > > * change the default for all architectures (except arm) > > instead of just ppc64 > > Are we certain we want to do this even for x86 guests? I don't think > it's a good idea - it makes the default into something for which no > driver exists on the install media for *many* x86 guests, meaning it > is > highly likely that a "default" config would have non-functional > networking. AFAIR this is why we didn't make virtio the default > several > years ago when we began recording a default into the XML (and why the > last time changing the default was discussed, I believe it was > pointing > more towards something like e1000, i.e. something which 1) we are > certain has a driver on every guest OS installation media that might > be > found, 2) is better maintained in qemu than the rtl8139, and 3) > performs > better than rtl8139 (although obviously not as good as virtio). Good point. We could switch the default to virtio just for ppc64 then, leaving the capability check in place of course, and change the global default from rtl8139 to e1000 in a separate series.I think we should just do nothing. If people/apps care about optimal defaults, nothing libvirt changes will make them happy. Apps need to be just using libosinfo to get the sensible defaults on a per OS basis.
OK, so maybe at least using virtio instead of rtl8139 only in case rtl8139 is not available.
Regards, Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|
Attachment:
signature.asc
Description: PGP signature
-- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list