On Wed, Apr 2, 2014 at 7:05 PM, Jesse Gross <je...@nicira.com> wrote: > On Wed, Apr 2, 2014 at 9:56 AM, Zoltan Kiss <zoltan.k...@citrix.com> wrote: >> On 01/04/14 20:26, Jesse Gross wrote: >>> >>> On Tue, Apr 1, 2014 at 11:41 AM, Kyle Mestery <mest...@noironetworks.com> >>> wrote: >>>> >>>> On Tue, Apr 1, 2014 at 1:27 PM, Zoltan Kiss <zoltan.k...@citrix.com> >>>> wrote: >>>>> >>>>> Hi, >>>>> >>>>> I have a recent patch on net-next which affects OVS as well: >>>>> >>>>> >>>>> http://git.kernel.org/cgit/linux/kernel/git/davem/net-next.git/commit/?id=36d5fe6a000790f56039afe26834265db0a3ad4c >>>>> >>>>> I guess the changes for datapatch.c will turn up automagically sometime >>>>> during a merge, is that correct? >>>>> But there is a change in the core skbuff API which need to be included >>>>> in >>>>> datapath/linux/compat/skbuff-openvswitch.c, otherwise upstream OVS won't >>>>> build on older kernels. How does this process works? Is there any >>>>> guarantee >>>>> that even if it's forgotten during the merge, a build system somewhere >>>>> will >>>>> try to compile upstream OVS with older kernel, and sends a message when >>>>> it >>>>> fails? >>>> >>>> >>>> Hi Zoltan: >>>> >>>> I just proposed a patch to support Linux 3.13 and 3.14, and this >>>> includes your patch >>>> folded in. The mailing lists server appears to be down now, otherwise >>>> I would have >>>> included a link. But I've taken care of the skbuff API change as well. >>>> Feedback on >>>> this patch is appreciated! >>> >>> >>> To answer the more general question, there's unfortunately nothing >>> automatic about it. In many cases, changes will cause compilation to >>> fail on older kernels if an upstream patch is missed when adding >>> support for a newer kernel but that's not always the case (including >>> this one I believe). The best thing is if authors submit their changes >>> to both repositories (with backports if necessary), otherwise it falls >>> to somebody to do it like Kyle has done here. >> >> >> My understanding is that things from the main OVS repo's datapath directory >> merge to git://git.kernel.org/pub/scm/linux/kernel/git/jesse/openvswitch >> after a while, and then to net(-next), is that correct? >> I think the merge from OVS repo to Jesse's kernel.org repo are made >> manually, shouldn't you add a step to compare net(-next) with your repo as >> well? I mean the history for the net/openvswitch directory. That way you can >> see if something changed OVS in the mainline kernel tree and take action to >> merge it backwards to the main OVS repo. > > Right, this is essentially what Kyle has done with his patch.
+1 _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev