Re: [PATCH v2] credit: generalize __vcpu_has_soft_affinity()

[Dario Faggioli <dario.faggioli@citrix.com>]

[-- Attachment #1.1: Type: text/plain, Size: 1779 bytes --]

On Fri, 2015-03-06 at 09:53 +0000, George Dunlap wrote:
> On 03/06/2015 07:36 AM, Jan Beulich wrote:

> It looks like the comment above this line could use changing too;
> 
It probably does.

> ---
> Hard affinity balancing is always necessary and must never be skipped.
> But soft affinity need only be considered when it has a functionally
> different effect than other constraints (such as hard affinity, cpus
> online, or cpupools).
> 
> Soft affinity only needs to be considered if:
> * The cpus in the cpupool are not a subset of soft affinity
> * The hard affinity is not a subset of soft affinity
> * There is an overlap between the soft affinity and the mask which is
> currently being considered.
> ---
The original comment was more aimed at trying to make as evident as
possible to the reader why the occurrence of a particular combination of
hard and soft affinity makes using the latter pointless (perhaps it was
not that effective at that, but that's another story! :-D).

So, if changing it, I'd go for something more 'high level'; something
that helps the reader grasp more quickly what is the actual meaning of
those _subsect() and _intersect() operations. Perhaps:

---
Hard affinity balancing is always necessary and must never be skipped.
Soft affinity balancing is only useful if it, potentially, makes a
difference.

In more details, there is no point checking soft affinity if:
* there is no online pCPU in the domain's cpupool that is not in the
  soft affinity mask
* there is no pCPU in the hard affinity mask that is not in the soft
  affinity mask
* there is no overlapping between the soft affinity mask and the mask
  being considered (which, in most cases, is the hard affinity mask)
---

Regards,
Dario

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

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