I am trying to work out whether it is practical to propose Dom0 xen support as a feature for Fedora 15.
The current situation is that patches to allow a kernel to boot as a xen Domain 0 are in 2.6.37 but they aren't enough to run a reasonable xen guest system from it. The xen developers are aiming to get enough drivers accepted in 2.6.38 so that you can run at least basic guest machines.
Given the roughly 3 month kernel release cycle, I would expect Fedora 15 to ship with a 2.6.37.x kernel, with 2.6.38 coming out at around the time of the Fedora 15 release but too late for it to be included, and that Fedora 15 will move to 2.6.38 at some later point. Thus Fedora 15 should be usable as a xen Domain 0 at some point in its life cycle.
However for xen dom0 support to be a Fedora 15 feature we would really need to have the appropriate drivers in the kernel when Fedora 15 ships, so my main question is how likely is it that backported xen drivers would be accepted into the Fedora kernel? I would also be interested in any other comments on this proposal.
Michael Young
On Tue, Nov 9, 2010 at 7:12 PM, M A Young m.a.young@durham.ac.uk wrote:
I am trying to work out whether it is practical to propose Dom0 xen support as a feature for Fedora 15.
The current situation is that patches to allow a kernel to boot as a xen Domain 0 are in 2.6.37 but they aren't enough to run a reasonable xen guest system from it. The xen developers are aiming to get enough drivers accepted in 2.6.38 so that you can run at least basic guest machines.
Given the roughly 3 month kernel release cycle, I would expect Fedora 15 to ship with a 2.6.37.x kernel, with 2.6.38 coming out at around the time of the Fedora 15 release but too late for it to be included, and that Fedora 15 will move to 2.6.38 at some later point. Thus Fedora 15 should be usable as a xen Domain 0 at some point in its life cycle.
However for xen dom0 support to be a Fedora 15 feature we would really need to have the appropriate drivers in the kernel when Fedora 15 ships, so my main question is how likely is it that backported xen drivers would be accepted into the Fedora kernel? I would also be interested in any other comments on this proposal.
I don't think backported drivers would be the right approach. There are several times where large backports of function have caused issues in the upgrade path or otherwise been painful.
I won't comment on Xen overall, but it certainly doesn't seem like a suitable feature for F15. Best case I would recommend F16.
josh
On Wed, Nov 10, 2010 at 12:12:15AM +0000, M A Young wrote:
However for xen dom0 support to be a Fedora 15 feature we would really need to have the appropriate drivers in the kernel when Fedora 15 ships, so my main question is how likely is it that backported xen drivers would be accepted into the Fedora kernel?
Not particularly likely, I'm really sick of carrying backports of crap for aeons on a "promise" that things will get upstream. If things are already /merged/ in 2.6.37-git by then, then maybe, but see below.
I would also be interested in any other comments on this proposal.
What's the overhead? Why would we want to support it? What's the upside?
regards, Kyle
Hi,
2010/11/10 Kyle McMartin kyle@mcmartin.ca:
What's the overhead? Why would we want to support it?
I've got two fairly new machines - one Atom 330 and one laptop with T5270. Both CPU's doesn't have VMX support.
Is in this case Xen not the only reasonable solution in terms of speed?
What's the upside?
regards, Kyle _______________________________________________ kernel mailing list kernel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/kernel
Kind regards, Michal
On 11/10/10 01:59, Kyle McMartin wrote:
On Wed, Nov 10, 2010 at 12:12:15AM +0000, M A Young wrote:
However for xen dom0 support to be a Fedora 15 feature we would really need to have the appropriate drivers in the kernel when Fedora 15 ships, so my main question is how likely is it that backported xen drivers would be accepted into the Fedora kernel?
Not particularly likely, I'm really sick of carrying backports of crap for aeons on a "promise" that things will get upstream.
Well, unlike 20 kernel releases back the xen guys are actually getting their stuff upstream now. Chances are pretty good that the drivers make it into the 2.6.38 merge window.
If things are already /merged/ in 2.6.37-git by then, then maybe, but see below.
Could being in linux-next convince you too?
I would also be interested in any other comments on this proposal.
What's the overhead?
Just some xen specific drivers below drivers/xen/
Why would we want to support it? What's the upside?
Fedora will be able to work as Xen host?
cheers, Gerd
On Wed, Nov 10, 2010 at 5:39 AM, Gerd Hoffmann kraxel@redhat.com wrote:
On 11/10/10 01:59, Kyle McMartin wrote:
On Wed, Nov 10, 2010 at 12:12:15AM +0000, M A Young wrote:
However for xen dom0 support to be a Fedora 15 feature we would really need to have the appropriate drivers in the kernel when Fedora 15 ships, so my main question is how likely is it that backported xen drivers would be accepted into the Fedora kernel?
Not particularly likely, I'm really sick of carrying backports of crap for aeons on a "promise" that things will get upstream.
Well, unlike 20 kernel releases back the xen guys are actually getting their stuff upstream now. Chances are pretty good that the drivers make it into the 2.6.38 merge window.
Good, sounds like 2.6.38 will be a suitable base for this work for those that want it.
If things are already /merged/ in 2.6.37-git by then, then maybe, but see below.
Could being in linux-next convince you too?
In my opinion, no. Things have been in linux-next but not merged during the actual merge window. I personally don't believe there is any substitution for "it's merged and in a released kernel".
josh
On Wed, Nov 10, 2010 at 11:39:46AM +0100, Gerd Hoffmann wrote:
Could being in linux-next convince you too?
It depends really, I don't want to suck back giant patches just to add features that will be added in the next release anyway.
The problem with linux-next is it hasn't met the Linus merge criteria yet.
What's the overhead?
Just some xen specific drivers below drivers/xen/
For dom0? Do we need to build yet another flavour of kernel?
regards, Kyle
Hi,
What's the overhead?
Just some xen specific drivers below drivers/xen/
For dom0?
Yes. The host backend drivers for guest devices. It is just the last piece to have a useful xen host, everything else has been merged upstream already. 2.6.37 already can run xen guests without network and storage devices.
Do we need to build yet another flavour of kernel?
No. The standard kernel which can run as domU (aka xen guest) for a long time will be able to run as dom0 (xen host) too. Just some config options need to be enabled.
cheers, Gerd
kernel@lists.fedoraproject.org
-
Gerd Hoffmann -
Josh Boyer -
Kyle McMartin -
M A Young -
Michał Piotrowski