[CentOS-virt] Building Xen 4.4 rpms for centos7

Tue Jan 20 15:06:00 UTC 2015
George Dunlap <dunlapg at umich.edu>

On Tue, Jan 20, 2015 at 2:35 PM, Johnny Hughes <johnny at centos.org> wrote:
> On 01/20/2015 08:10 AM, Pasi Kärkkäinen wrote:
>> Hello,
>> Now that the Xen 4.4.1 rpms + libvirt 1.2.10 rpms for CentOS 6 seem to be in good shape
>> it would be a good time to start experimenting with el7 builds aswell.
>> I think it's best to start with the same set of packages we have for el6,
>> and just do the minimum required changes to get them working on centos7.
>> That probably means fixing some dependencies in the spec file, and some changes for systemd support.
>> Note that all those changes are already done in Fedora 21 Xen rpms, for example:
>> ftp://ftp.funet.fi/pub/mirrors/fedora.redhat.com/pub/fedora/linux/updates/21/SRPMS/x/xen-4.4.1-12.fc21.src.rpm
>> A few months ago I also saw a Xen 4.4 el7 rpms port, so we could find those src.rpms aswell.
> Agreed on this too .. let's use as much the same as we can, and we can
> use %if statements in the SPEC to differentiate el6 and el7 things, if
> necessary.  systemd versus init and maybe some version number changes
> for buildrequires should be the changes we need to be concerned about.
>> And then there's the dom0 kernel.. Stock rhel7/centos7 is based on Linux 3.10 kernel,
>> so we probably can use the exact same 3.10.63 kernel we have today for CentOS 6 Xen.
> I agree that we can likely use the current centos-6 kernel.  We do need
> to decide when we want to shift the kernel to 3.14 .. the support for
> 3.10 as a LTS kernel ends in September 2015 (so 9 months).

I intend to do this as soon as I have time; but ATM I've got probably
less than one day a week to devote to it, so it may be slow.

> https://www.kernel.org/category/releases.html
> We can try to take the RH 3.10 el7 kernel, mod it for xen, and use it ..
> or we can shift to 3.14.x and that buys us at least one more year .. or
> wait until they name the next LTS kernel and go for that.  Likely the
> next LTS kernel will be the easiest option (the RH modified kernel will
> not support xen and rolling in stuff externally will be hard because of
> the backports RH does to the kernel (things that go into a standard
> kernel will not apply cleanly to the RH kernel).

Sticking with the stable upstream kernels would allow us to easily get
fixes and updates to Xen-specific issues, rather than having to
manually port them over.