[Xen-devel] HVM support to be removed from Debian Squeeze: call for volunteers

[Thomas Goirand <thomas@xxxxxxxxxx>]

[message cross posted to the pkg-xen and xen-devel list]

Dear everyone,

Bastian Blank - which is the person (among others, but mainly him) that
is packaging Xen in Debian -, has decided last summer that he doesn't
want to deal with the qemu-dm of Xen, thus removing Xen Qemu and support
for HVM in Debian. Here is what he wrote:

http://lists.alioth.debian.org/pipermail/pkg-xen-devel/2009-July/002366.html
http://lists.alioth.debian.org/pipermail/pkg-xen-devel/2009-July/002382.html

Many people have expressed their concerns about this regression.

Before anyone flames Bastian Blank, I'd like to highlight the fact that
maintaining a PV dom0 in a distribution needs quite some work, he does
it for free, and that he should be thanks for his work within Debian,
rather than being blame for his maintainers choice. If he is not
interested or doesn't have time to maintain the Qemu version of Xen, he
has right to do that, and in fact I think it's not a bad idea to have a
binary package separated from the hypervisor itself.

Now, if the qemu-dm binary have been removed from the hypervisor
packages, nothing prevents anyone to make a new Debian package that
would have only the HVM bits of Xen, depending on the hypervisor
package, and then having it uploaded to SID (so that it would go in
testing later). Later, we could ask the maintainers of the
xen-linux-system package to add yet another dependency to the new
"xen-hvm-support" package that we would create (are you ok with this new
package name?).

The issue here is that Squeeze (the current Debian testing) is said to
be going to be functionality frozen next march. That means that if we
want to have a xen-hvm-support package in the next Debian, and recover
from Bastian's choice, then we need to act FAST, otherwise after the
deadline, there's a big chance that the release managers will not accept
the package into testing.

I am quite comfortable with Debian packaging myself, so I believe I can
write Debian packages with a quality good enough to enter the
distribution without too much hassle. I manage already a dozen of
packages in Debian, and I intend to add even more. But I do not feel
comfortable enough with the Xen sources to do it alone. So I'm calling
here for volunteers to work together with me on this package. I also
don't want to be the only person responsible for the package, as I
wouldn't be able to address any security bug fixing issue (as I don't
know the internals of the Xen and Qemu code, and don't have the time to
dive into it). Someone that is willing to do this job of security
maintenance HAS to raise hand here, otherwise I will NOT shoot myself in
the foot by asking for an upload of a package that would have a bad
security maintenance team (this is the kind of behavior that people hate
in Debian, which I think is right: no package with poor maintenance
should enter the distribution).

Mainly, the work would be to have a script (or some Makefile entries
with DESTDIR= for example) that would install all needed files in
debian/xen-hvm-support (if we call that package xen-hvm-support). Also,
the source package should contain ONLY what is strictly concerning
Qemu-dm (not a big issue, but it would be cleaner this way and make the
source package smaller). As of now, I don't know who is maintaining Qemu
in Xen at all. Any volunteer here that would work with me does NOT need
to know anything about Debian or Debian packaging, but only about Qemu
and Xen itself.

Also, I'm not a registered Debian developer (yet), so I can't upload
directly in the distribution without going by a sponsor for my package,
which can seriously slow down everything. If a registered DD could raise
hand, that would help as well.

I think that the best way here, is to build a (small, 3 or 4 persons)
team to work on this issue so it can be addressed fast. I'd like to
highlight also, that my company (GPLHost) doesn't really need the HVM
part of Xen (we sell PV only VMs), so I'd be doing it only for the
community, and not for myself. So please volunteer, as this is the only
way it can work. Best could be to create a project in Alioth (which I
never did, maybe someone else who had this experience should do it for
the team), or anywhere else. GPLHost can host the git repository and/or
web space for that.

If you are interested in helping, please reply to this message.

Thomas

P.S. to Bastian: Are you willing to help the team with advices, even if
you don't take any responsibility in the package?

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel