Re: [PULL] qemu-xen stable update to 1.6.2

[George Dunlap <george.dunlap@eu.citrix.com>]

On 01/24/2014 07:34 PM, Stefano Stabellini wrote:
> On Fri, 17 Jan 2014, Stefano Stabellini wrote:
>> On Fri, 17 Jan 2014, Anthony PERARD wrote:
>>> On Fri, Jan 17, 2014 at 01:17:55PM +0000, Stefano Stabellini wrote:
>>>> On Fri, 17 Jan 2014, Anthony PERARD wrote:
>>>>> On Thu, Jan 16, 2014 at 03:51:42PM +0000, Stefano Stabellini wrote:
>>>>>> On Thu, 16 Jan 2014, Anthony PERARD wrote:
>>>>>>> On Thu, Jan 16, 2014 at 03:42:17PM +0000, Stefano Stabellini wrote:
>>>>>>>> On Wed, 15 Jan 2014, Ian Campbell wrote:
>>>>>>>>> On Wed, 2014-01-15 at 14:35 +0000, Anthony PERARD wrote:
>>>>>>>>>> On Wed, Jan 15, 2014 at 09:35:10AM +0000, Ian Campbell wrote:
>>>>>>>>>>> On Tue, 2014-01-14 at 18:52 +0000, Anthony PERARD wrote:
>>>>>>>>>>>> There is an update of QEMU 1.6, I have done a merge and put that in a tree:
>>>>>>>>>>>> git://xenbits.xen.org/people/aperard/qemu-dm.git  merge-1.6.2
>>>>>>>>>>> Based on the above I have no idea whether a freeze exception should be
>>>>>>>>>>> granted for this, so my default answer is no. I'm not sure what else you
>>>>>>>>>>> could have expected.
>>>>>>>>>>>
>>>>>>>>>>> If you think there are changes here which should be in 4.4.0 then please
>>>>>>>>>>> enumerate all changes included in this merge which have any relation to
>>>>>>>>>>> Xen and their potential impact on the release.
>>>>>>>>>> I have a list the change here that have a potential impact on Xen, with
>>>>>>>>>> the ones that I think are quite important at the beginning. Either the
>>>>>>>>>> commit title speak for itself or I added a small description on what is
>>>>>>>>>> affected.
>>>>>>>>> Thanks but there's not a lot here for me to go on WRT making a decision
>>>>>>>>> on a freeze exception. Did you refer to
>>>>>>>>> http://wiki.xen.org/wiki/Xen_Roadmap/4.4#Exception_guidelines_for_after_the_code_freeze
>>>>>>>>> like I said? A freeze exception needs an analysis of benefits and risks,
>>>>>>>>> not the very briefest words you can possibly manage.
>>>>>>>>>
>>>>>>>>> Anyway it appears this is a grab bag of things we might want and misc
>>>>>>>>> fixes which are perhaps nice to have, I'm nowhere near comfortable
>>>>>>>>> giving it a blanket exemption based on what you've presented here, or
>>>>>>>>> even of cherry picking what might be the important ones. If you think
>>>>>>>>> any or all of it is actually important for 4.4 please make a proper case
>>>>>>>>> for inclusion, either of the aggregate or of the individual changes.
>>>>>>>> Anthony, did you simply update the tree by pulling from the upstream 1.6
>>>>>>>> stable tree?
>>>>>>> Yes, a simple merge.
>>>>>>>
>>>>>>>> I also assume that you tested at the very least the basic
>>>>>>>> PV and HVM configurations?
>>>>>>> :(, no, I haven't try PV. But I did try HVM.
>>>>>>>
>>>>>>> There is one thing that I may want to try, it's migration from the
>>>>>>> previous version of Xen. There is one patch that change (fix?) that.
>>>>>> Please do and let me know if it works as expected.
>>>>> I have tryied a pv guest, it does work fine.
>>>>>
>>>>> I also try a migration (xl save/restore) from Xen 4.3.1 to Xen 4.4 with
>>>>> both the current qemu-xen tree and with the merge of 1.6.2, but the
>>>>> migration fail in both cases because of the same error reported by qemu.
>>>>> (Unknown savevm section or instance '0000:02.0/cirrus_vga' 0). I did not
>>>>> investigate in that. Might just be a issue with my compile script ...
>>>>> (using the wrong qemu-xen tree).
>>>> It is important that we identify what is the cause of the problem.
>>>> Especially if you think that it could be a "compile script" issue,
>>>> because I imagine that if it is, it might invalidate your previous
>>>> positive tests too.
>>> I was compiling with always the master branch of qemu-xen. So I had a
>>> Xen 4.3 with QEMU 1.6 instead of 1.3. So it only invalidate the
>>> migration test.
>> OK, good. Can you double check how migration from  Xen 4.3 with QEMU 1.3
>> works?
> I tested migration myself: although the update fixes a couple of
> problems with migration from 1.3, it also introduces a new one, that
> fortunately is fixed in QEMU 1.7.0.
>
> I have a branch based on 1.6.2 plus the backported fix from 1.7.0 that
> works well, I recommend getting a release exception for it and pushing
> it as soon as possible.

Sorry, didn't catch the "I recommend getting a release exception". The 
qemu issues (including the migration one) are definitely the main 
blockers to the release at the moment, so:

Release-acked-by: George Dunlap <george.dunlap@eu.citrix.com>