On Sat, 2006-07-22 at 09:31 -0400, Aron Griffis wrote:
Hi guys,
There are four ways you could be a big help to the xen-ia64 effort.
I have mentioned these in the past but never pulled them together into
a single request.
Sorry for getting so long in answering you, but A PAE bug and getting
linux-2.6-xen working on 2.6.18-rc2 have required all my attention :(
1. Publish Juan's tree which is the result of his merge of
linux-2.6.tip-xen, linux-2.6 and linux-2.6-xen. How about
http://people.redhat.com/quintela/linux-2.6.tip-xen-fedora.hg?
As things stand right now, it's very difficult for ia64 devs to
contribute to Juan's patch, for 2 reasons: (1) we never see it
until after a new kernel rpm is published, (2) all we get is the
final result, lacking the extremely helpful changeset history.
It is based on 2.6.18-rc2. It is on:
http://hg.et.redhat.com/kernel/linux-2.6-xen-fedora
It is public only since Tuesday (I was waiting for a place where to
publish it). Tree will not work on ia64, I didn't forward ported the
ia64 changes, noly x86 & x86_64 (and it was painfull enough, time
source, smp-alternatives x86_64, irqtrace, vDSO on x86 & friends meaned
that I had to go back & re-apply series of patches one at a time to find
a coulpe of bugs).
2. Use matched xenlinux/hypervisor pairs. At the OLS Xen
mini-summit
there was some discussion of compatibility. The statements were:
Believe me that we _try_, and very hard.
- old domU should run on new hypervisor
Agreed. I normally test plain fc5 domU on all my new kernels.
- new domU NOT guaranteed to run on old hypervisor
We have found that lately this "normally" works, versus bugs.
- dom0 and hypervisor should be MATCHED
Guess why HV on fedora is on the same package that the kernel, and they
have indeed the same version number?
Presently (kernel-2.6.17-1.2432.fc6.src.rpm) it appears that the
hypervisor and xen patch are both dated 20060719. That's a good
sign! :-) It would be great if you could provide some indication
of your intent to continue using matched pairs. Are they actually
based on the same xen-unstable changeset? We, the users, can't
tell because the tarball is dated instead of cset-stamped, and
Juan's tree isn't available. See #1 ;-)
Normally I add cset numbers in the changelog, but will try to put them
into the HV version number (it maks as much sense as the date,
actually).
About the source tree, it is already public. It hasn't been published
sooner due to lack of somobdy setting up a server. Now it is done.
3. Apply my kernel.spec and config changes at the end of this email.
This would make it much easier for the ia64 developers to do test
builds and track down failures. Presently it's a pain for me to
maintain this out of tree.
Will do today. Will send one email once this is done.
4. Apply the xen.spec and libvirt.spec changes.
https://www.redhat.com/archives/fedora-xen/2006-July/msg00022.html
https://www.redhat.com/archives/fedora-xen/2006-July/msg00021.html
Regarding libvirt.spec, I've talked with Daniel and he's just
waiting for xen.spec to be ia64-enabled...
Will talk with daniel when he is back from OLS.
Objections/comments? I'll attempt to update any relevant BZs
today.
Will comment on the patches in a follow-up.
Later, Juan.