RE: [PATCH 1/2] viridian: remove implicit limit of 64 VPs per partition

[Paul Durrant <xadimgnik@xxxxxxxxx>]

> -----Original Message-----
> From: Jan Beulich <jbeulich@xxxxxxxx>
> Sent: 11 January 2021 13:38
> To: Igor Druzhinin <igor.druzhinin@xxxxxxxxxx>; paul@xxxxxxx
> Cc: wl@xxxxxxx; iwj@xxxxxxxxxxxxxx; anthony.perard@xxxxxxxxxx; 
> andrew.cooper3@xxxxxxxxxx;
> george.dunlap@xxxxxxxxxx; julien@xxxxxxx; sstabellini@xxxxxxxxxx; 
> roger.pau@xxxxxxxxxx; xen-
> devel@xxxxxxxxxxxxxxxxxxxx
> Subject: Re: [PATCH 1/2] viridian: remove implicit limit of 64 VPs per 
> partition
> 
> On 11.01.2021 14:34, Igor Druzhinin wrote:
> > On 11/01/2021 09:16, Jan Beulich wrote:
> >> On 11.01.2021 10:12, Paul Durrant wrote:
> >>>> From: Paul Durrant <xadimgnik@xxxxxxxxx>
> >>>> Sent: 11 January 2021 09:10
> >>>>
> >>>>> From: Jan Beulich <jbeulich@xxxxxxxx>
> >>>>> Sent: 11 January 2021 09:00
> >>>>>
> >>>>> On 11.01.2021 09:45, Paul Durrant wrote:
> >>>>>> You can add my R-b to the patch.
> >>>>>
> >>>>> That's the unchanged patch then, including the libxl change that
> >>>>> I had asked about and that I have to admit I don't fully follow
> >>>>> Igor's responses? I'm hesitant to give an ack for that aspect of
> >>>>> the change, yet I suppose the libxl maintainers will defer to
> >>>>> x86 ones there. Alternatively Andrew or Roger could of course
> >>>>> ack this ...
> >>>>>
> >>>>
> >>>> I don't think we really need specific control in xl.cfg as this is a fix 
> >>>> for some poorly
> documented
> >>>> semantics in the spec. The flag simply prevents the leaf magically 
> >>>> appearing on migrate and I
> think
> >>>> that's enough.
> >>>
> >>> ... although adding an option in xl/libxl isn't that much work, I suppose.
> >>>
> >>> Igor, would you be ok plumbing it through?
> >>
> >> This back and forth leaves unclear to me what I should do. I
> >> would have asked on irc, but you're not there as it seems.
> >
> > I don't see a scenario where somebody would want to opt out of unlimited
> > VPs per domain given the leaf with -1 is supported on all Windows versions.
> 
> So Paul - commit patch as is then?
> 
> > I can make it configurable in the future if reports re-surface it causes
> > troubles somewhere.
> 
> This is the slight concern I have: Having to make it configurable
> once someone has reported trouble would look a little late to me.
> Otoh I agree it may end up being dead code if no problems get
> ever encountered.
> 

I think I'm persuaded by your caution. Since it's not a massive amount of code, 
let's have flags for both wired through to xl and default them to on, so I 
withdraw my R-b for the libxl_x86.c hunk.

  Paul

> Jan