libxl backports for 4.4 and 4.5

libxl backports for 4.4 and 4.5

[Anthony PERARD]

Hi Jan and Ian,

Could we have the following commit been backport to 4.4?
9369988 libxl: event handling: Break out ao_work_outstanding
f1335f0 libxl: event handling: ao_inprogress does waits while reports outstanding
4783c99 libxl: In domain death search, start search at first domid we want
188e9c5 libxl: Domain destroy: fork

And the following commit to 4.5?
9369988 libxl: event handling: Break out ao_work_outstanding
f1335f0 libxl: event handling: ao_inprogress does waits while reports outstanding

There are usefull to have libvirt working well and not hang/freeze.

Thanks,

-- 
Anthony PERARD

Re: libxl backports for 4.4 and 4.5

[Jan Beulich]

>>> On 11.06.15 at 17:29, <anthony.perard@citrix.com> wrote:
> Could we have the following commit been backport to 4.4?
> 9369988 libxl: event handling: Break out ao_work_outstanding
> f1335f0 libxl: event handling: ao_inprogress does waits while reports 
> outstanding
> 4783c99 libxl: In domain death search, start search at first domid we want
> 188e9c5 libxl: Domain destroy: fork
> 
> And the following commit to 4.5?
> 9369988 libxl: event handling: Break out ao_work_outstanding
> f1335f0 libxl: event handling: ao_inprogress does waits while reports 
> outstanding
> 
> There are usefull to have libvirt working well and not hang/freeze.

Right, but unless really urgent/important I think this is too late for
4.5.1 now and needs to be put on the list of backports for 4.5.2.

Jan

Re: libxl backports for 4.4 and 4.5

[Ian Jackson]

Anthony PERARD writes ("libxl backports for 4.4 and 4.5"):
> Hi Jan and Ian,
> 
> Could we have the following commit been backport to 4.4?
> 9369988 libxl: event handling: Break out ao_work_outstanding
> f1335f0 libxl: event handling: ao_inprogress does waits while reports outstanding
> 4783c99 libxl: In domain death search, start search at first domid we want
> 188e9c5 libxl: Domain destroy: fork
> 
> And the following commit to 4.5?
> 9369988 libxl: event handling: Break out ao_work_outstanding
> f1335f0 libxl: event handling: ao_inprogress does waits while reports outstanding
> 
> There are usefull to have libvirt working well and not hang/freeze.

Yes, good idea.

Thanks,
Ian.

Re: libxl backports for 4.4 and 4.5

[Lars Kurth]

Ian, Jan,

you seem to be disagreeing, or more likely Ian missed Jan's reply, so I wanted to double check.

> Right, but unless really urgent/important I think this is too late for
> 4.5.1 now and needs to be put on the list of backports for 4.5.2.

Jan, just to clarify, are you saying back porting for 4.4.3 is OK, but 4.5.1 is too tight?  Looking at the 4 months cadence, 4.4.3 would be sometimes in July: so I can live with this as a compromise.

Even though having a released version of Xen which works out of the box with OpenStack is pretty important.

@Stefano, @Anthony: please keep on top of necessary back ports for OpenStack in future. I thought these have been requested and only found out when I talked to Anthony yesterday that this wasn't the case.

Regards
Lars

Re: libxl backports for 4.4 and 4.5

[Jan Beulich]

>>> On 12.06.15 at 11:32, <lars.kurth.xen@gmail.com> wrote:
> Ian, Jan,
> 
> you seem to be disagreeing, or more likely Ian missed Jan's reply, so I 
> wanted to double check.
> 
>> Right, but unless really urgent/important I think this is too late for
>> 4.5.1 now and needs to be put on the list of backports for 4.5.2.
> 
> Jan, just to clarify, are you saying back porting for 4.4.3 is OK, but 4.5.1 
> is too tight?  Looking at the 4 months cadence, 4.4.3 would be sometimes in 
> July: so I can live with this as a compromise.
> 
> Even though having a released version of Xen which works out of the box with 
> OpenStack is pretty important.

The release being stuck anyway, perhaps we can put them in with
the understanding that this time round we'll need at least another
RC. Ian - you know how risky the changes are, please use your
judgment.

(Btw., as to 4.4.3, I don't think I want to kick off that process until
4.5.1 gets out of its stuck state, i.e. as all stable trees are affected
by the underlying problem, I don't want both trees to be hold up half
way through the stable release process.)

Jan

Re: libxl backports for 4.4 and 4.5

[Lars Kurth]

> On 12 Jun 2015, at 10:46, Jan Beulich <jbeulich@suse.com> wrote:
> 
>>>> On 12.06.15 at 11:32, <lars.kurth.xen@gmail.com> wrote:
>> 
>> Even though having a released version of Xen which works out of the box with 
>> OpenStack is pretty important.
> 
> The release being stuck anyway, perhaps we can put them in with
> the understanding that this time round we'll need at least another
> RC. Ian - you know how risky the changes are, please use your
> judgment.

That works. Awaiting Ian's reply. 
@Ian: we have been using these patches in the OpenStack CI loop for months now, but only on top of 4.4.1-0ubuntu0.14.04.4

> 
> (Btw., as to 4.4.3, I don't think I want to kick off that process until
> 4.5.1 gets out of its stuck state, i.e. as all stable trees are affected
> by the underlying problem, I don't want both trees to be hold up half
> way through the stable release process.)

Makes sense

Lars

Re: libxl backports for 4.4 and 4.5

[Ian Jackson]

Jan Beulich writes ("Re: [Xen-devel] libxl backports for 4.4 and 4.5"):
> On 12.06.15 at 11:32, <lars.kurth.xen@gmail.com> wrote:
> > you seem to be disagreeing, or more likely Ian missed Jan's reply, so I 
> > wanted to double check.

Thanks for pointing this out, Lars.  I had just spotted it myself.

> The release being stuck anyway, perhaps we can put them in with
> the understanding that this time round we'll need at least another
> RC. Ian - you know how risky the changes are, please use your
> judgment.

These changes are old, well-tested, and they are important bugfixes.
I think they should go in.

> (Btw., as to 4.4.3, I don't think I want to kick off that process until
> 4.5.1 gets out of its stuck state, i.e. as all stable trees are affected
> by the underlying problem, I don't want both trees to be hold up half
> way through the stable release process.)

Right.  So I take you to mean that I should apply backports to 4.4
freely.


Jan, things that I have queued for 4.5 are therefore:

 d83bf9d2 libxl: In libxl_set_vcpuonline check for
                 maximum number of VCPUs against the cpumap

 12e817e2 Fix build of 32bit toolstacks on CentOS 5.x following XSA-125

 9369988  libxl: event handling: Break out ao_work_outstanding

 f1335f0  libxl: event handling: ao_inprogress does waits while reports
                 outstanding

Of these, all but 12e817e2 have been in master for a long time.
12e817e2 was pushed this morning.

I propose to push all of these today, unless you object.  I will
double-check with you on irc.

Ian.

Re: libxl backports for 4.4 and 4.5

[Ian Jackson]

Ian Jackson writes ("Re: [Xen-devel] libxl backports for 4.4 and 4.5"):
> I propose to push all of these today, unless you object.  I will
> double-check with you on irc.

Now done.

Ian.

Re: libxl backports for 4.4 and 4.5

[Lars Kurth]

> On 12 Jun 2015, at 13:01, Ian Jackson <ian.jackson@eu.citrix.com> wrote:
> 
> Ian Jackson writes ("Re: [Xen-devel] libxl backports for 4.4 and 4.5"):
>> I propose to push all of these today, unless you object.  I will
>> double-check with you on irc.
> 
> Now done.

Thanks
Lars