Re: [Xen-devel] [PATCH RESEND v7 1/9] xen: vnuma topology and subop hypercalls

[Elena Ufimtseva <ufimtseva@xxxxxxxxx>]

On Fri, Aug 22, 2014 at 1:38 PM, Elena Ufimtseva <ufimtseva@xxxxxxxxx> wrote:
> On Fri, Aug 22, 2014 at 9:54 AM, Dario Faggioli
> <dario.faggioli@xxxxxxxxxx> wrote:
>> On ven, 2014-08-22 at 14:17 +0100, Jan Beulich wrote:
>>
>>> > +    if ( ret )
>>> > +        return ret;
>>> > +
>>> > +    ret = -EFAULT;
>>> > +
>>> > +    if ( copy_from_guest(v->vdistance, u_vnuma->vdistance,
>>> > +                         nr_vnodes * nr_vnodes) )
>>> > +        goto vnuma_fail;
>>> > +
>>> > +    if ( copy_from_guest(v->vmemrange, u_vnuma->vmemrange, nr_vnodes) )
>>> > +        goto vnuma_fail;
>>>
>>> Isn't a single memory range per vnode rather limiting? Physical
>>> machines frequently have at least one node with two ranges to
>>> accommodate the hole below 4Gb. And since the interface is for
>>> all guest kinds I'm afraid this will harm setting up guests rather
>>> sooner than later. (I'm sorry for thinking of this only now.)
>>>
>> This actually was one concern of mine too, during the early stage of
>> Elena's work. I clearly remember wondering, and even asking, considering
>> that Linux uses more ranges per node, why using only one was ok for us.
>>
>> Unfortunately, I OTOH forgot how that discussion ended and how we got to
>> this point, without me or anyone else (continuing to) complaining... So,
>> sorry from me too. :-(
>>
>> Let's see what Elena and others think and, after that, how big of a
>> piece of work is to introduce support for more than one ranges (or
>> re-introduce, as I also think I remember this to be some kind of list,
>> during early versions of the series).
>>
>> Regards,
>> Dario
>>
>> --
>> <<This happens because I choose it to happen!>> (Raistlin Majere)
>> -----------------------------------------------------------------
>> Dario Faggioli, Ph.D, http://about.me/dario.faggioli
>> Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)
>>
>
>
> Jan, Dario
>
> Thank you for review. I briefly read them and I will give it a more
> through read after work.
>
> As for many-regions nodes, I will give it a thought as there maybe a
> simple extension to memory structure.
>
> I will try to evaluate asap and reply.
>
> Elena
> --
> Elena


Dario

Can you please take a look at libxl placement part and check if you
have any suggestions.
That probably will change in other versions, I just wanted to run it
with you for the case I have now.
You are going for vacation next week right?

I will try to work over the weekend and see how I can extend interface
with minimum invasion to
accommodate multi-region nodes, but placement part is unlikely will change.

Thank you!

-- 
Elena

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel