[FFE] Merge 0.9.3-5 from debian unstable

[Serge Hallyn]

We are currently on 0.9.2-4ubuntu11. Please sync 0.9.3-5

[Jamie Strandboge (jdstrand)]

libvirt has a significant Ubuntu delta, not least of which is adjusting it for AppArmor, apport, libvirt-migrate-qemu-disks (needed until 12.04), different groups, openiscsi, etc, etc. Glancing at the Debian changelog, these have not been upstreamed. AFAICT, this should be an FFe for a merge, not sync.

Serge, can you elaborate?

[Serge Hallyn (serge-hallyn)]

Sorry, I assume that's what I meant.

Note I intended to do the merge myself. But it seems this should get FFE approval (or else not be done), as version-to-version libvirt changes could be fragile.

Also, if we decide this is a bad idea, then we should probably also not merge qemu-0.15.0 (which also has an open FFE request which I was going to tackle).

I'm hoping to get candidate packages for both by tomorrow.

[Serge Hallyn (serge-hallyn)]

The debdiff is rather large (due to libvirt tree changes), but is at http://people.canonical.com/~serge/libvirt.debdiff

[Jamie Strandboge (jdstrand)]

Does this pass QRT?

[Serge Hallyn (serge-hallyn)]

@Jamie,

no, here is a list of errors.

At this point I personally feel it is too late to risk this merge. Unless you or someone else says otherwise, I think we should mark this wontfix and merge 0.9.5 for P when it opens.

[Jamie Strandboge (jdstrand)]

This is an interesting error:
error: Failed to connect socket to '@/home/tucxnXJm/.libvirt/libvirt-sock': Connection refused
error: failed to connect to the hypervisor

This is a new location of the socket. Do you have any apparmor denials in kern.log?

[Serge Hallyn (serge-hallyn)]

Quoting Jamie Strandboge (<email address hidden>):
> This is an interesting error:
> error: Failed to connect socket to '@/home/tucxnXJm/.libvirt/libvirt-sock': Connection refused
> error: failed to connect to the hypervisor
>
> This is a new location of the socket. Do you have any apparmor denials
> in kern.log?

No, the only thing I saw was several entries like:

Sep 2 13:39:24 sergelap kernel: [19175.248710] type=1400 audit(1314988764.860:26): apparmor="STATUS" operation="profile_remove" info="profile does not exist" error=-2 name="/usr/sbin/libvirtd" pid=6460 comm="apparmor_parser"

[Jamie Strandboge (jdstrand)]

This points to the failure happening before the dynamic profile is generated.

[Serge Hallyn (serge-hallyn)]

@Jamie,

does that indicate a problem in the apparmor security driver code?

Should I wait to mark this bug 'wontfix'?

[Jamie Strandboge (jdstrand)]

I don't think it is the apparmor security driver code based on where the error is occurring and the test output:
 Test initialization ... (default:default, qemu:///session, apparmor=False) ... FAIL

The apparmor driver is disabled and it is still failing (granted, in the session code).

[Jamie Strandboge (jdstrand)]

I should also note I am not actively investigating this issue at present.

[Serge Hallyn (serge-hallyn)]

Ok thanks, Jamie. I will look into it then when I do a libvirt merge at the start of the P cycle.