Re: [libvirt] interface "peer address" patches are broken
On 04/26/2016 12:55 PM, Vasiliy Tolstov wrote:
2016-04-26 18:42 GMT+03:00 Laine Stump <laine(a)laine.org>:
> The simplest fix would be to just swap peer and address in the set-address
> call for QEMU, but the fact that the problem didn't leap out at anyone
> indicates that the current name "peer" may not be the best choice (even
> though that's what it really is in the netlink message that sets the
> addresses; of course that *won't* be the case after it's fixed -
"address"
> will be set as "peer", and "peer" as "address"). So, as
I suggested in my
> earlier message, maybe we can come up with a better name for what is now
> called "peer" ("address" must stay the same, since it has already
been in
> several releases). Does "hostAddress" sound okay?
>
> Either of these patches (the one to fix the missing attribute in the XML or
> the one to rename the attribute) could be pushed even after DV begins the
> freeze period for the release - they qualify as bug fixes.
>
> (I actually already have a patch to fix the missing peer attribute in the
> XML, and am testing it, so the big question is whether to change the name
> from "peer" to something more accurate, and if so, what to change it to).
Thanks! May be combine Host and Peer to something like HostPeer ? =)
Hmm. That makes it sound like it's the address for the peer of the host,
but it's actually the address on the host itself. I think the original
"peer" is better than that, as long as it's understood that it is "the
peer address for the guest interface" (I still am hoping we can find
something better though, since "peer" is ambiguous unless it is
documented and the user reads the documentation).