Re: Xen4.2-rc3 test result

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Thu, Sep 06, 2012 at 08:31:12AM +0100, Ian Campbell wrote:
> On Thu, 2012-09-06 at 06:59 +0100, Ren, Yongjie wrote:
> > > Have we ever supported HVM guest CPU remove? I thought not.
> > > 
> > > http://bugzilla.xen.org/bugzilla/show_bug.cgi?id=1822#c3 seems to
> > > describe the behaviour I would expect.
> > > 
> > If we don't want to support HVM guest CPU remove in near future, I want to close this bug.
> 
> Is this something Intel is considering working on? If so then someone
> should mention it to George in the 4.3 planning thread.
> 
> > > If this is supposed to be an existing feature then is this a regression
> > > with xl vs xm or from 4.1 to 4.2?
> > > 
> > No, it's not a regression from 4.1 to 4.2.
> > Neither Xen 4.1 nor 4.2 supports HVM guest CPU remove with xm or xl.

But I think the bug does not talk about 'remove' but 'offline'.

That functionality (from a Xen toolstack perspective) works - if you
do 'xl vcpu-set' it properly tells the guest (either PV or HVM) to decrease
the count.

The problem is with the Linux kernel - and with the generic code:

https://lkml.org/lkml/2012/4/30/198

(and no, I had not a chance to actually fix it. Looking for volunteers).

> 
> OK, then I can remove it from the TODO list for 4.2, since it certainly
> isn't happening for 4.2.0 at this stage.

Right. Its a Linux kernel issue.
> 
> Thanks for letting me know,
> Ian.
> 
>