From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ian Campbell Subject: Re: [PATCH v7 15/33] net: xen-netback - set name assign type Date: Thu, 10 Jul 2014 13:13:10 +0100 Message-ID: <1404994390.8558.4.camel@kazak.uk.xensource.com> References: <1404980258-30853-1-git-send-email-teg@jklm.no> <1404980258-30853-16-git-send-email-teg@jklm.no> <1404982886.30660.4.camel@hastur.hellion.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: netdev , LKML , David Miller , David Herrmann , Kay Sievers , Wei Liu , xen-devel To: Tom Gundersen Return-path: In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Thu, 2014-07-10 at 12:46 +0200, Tom Gundersen wrote: > On Thu, Jul 10, 2014 at 11:01 AM, Ian Campbell wrote: > > On Thu, 2014-07-10 at 10:17 +0200, Tom Gundersen wrote: > >> The name contains then xen handle, which is not guaranteed to be > >> stable between restarts, so label this NET_NAME_ENUM. > > > > FWIW the N'th interface for domain with domid D will always be named > > vifD.N. > > > > If you reboot domain D then it's domid will change to E and the new > > device will be vifE.N. > > > > I don't know if that counts as predictable as far as the interface you > > are introducing is concerned. > > That was my understanding, thanks for confirming. > > The idea is that the same interface on the same machine must get the > same name between reboots in order to be considered predictable. So > under the assumption that a rebooted domain should be considered "the > same machine" (i.e., it still has the same local configuration, > possibly referencing the interface names), then we should consider the > names not predictable, so NET_NAME_ENUM. OK, thanks. I'll defer to you understanding of what this interface is supposed to mean: Acked-by: Ian Campbell Ian.