On 12/03/14 08:34, Dario Faggioli wrote:
> On Thu, 2014-02-27 at 11:11 +0000, Andrew Cooper wrote:
>> These two new functions provide a substantially easier-to-use API, where libxc
>> itself takes care of all the appropriate bounce buffering.
>>
> So, because of the stalled discussion on patch 2 of this series, this
> patch was hardly considered, I guess, is that so?
>
> Personally, I think this is an improvement in its own right, and I'd
> like to have it, independently from the CPUID stuff. Are you up,
> perhaps, for submitting it as a separate patch?
>
> One thing I'm not sure is why you're introducing new *_bounced variants,
> rather than going ahead and modifying xc_topologyinfo() and
> xc_numainfo()  directly. libxc is not an API we committed to keep
> stable, is it?
>
> Thanks and Regards,
> Dario
>

I certainly can modify the original ones directly.  When developing, it in the 4.4 freeze, and I wanted my changes to easily co-exist with libxc.  (If you notice, these patches are in XenServer trunk for easy deployment against our entire set of weird & wacky hardware).

If there general agreement that making the old xc_{topology,numa}info() functions have the prototype and behaviour of my _bounced variants, then I will happy do that, and send some fixup to make libxl work against it.

~Andrew