Re: [PATCH v5 12/13] tools: add tools support for Intel CAT

[Chao Peng <chao.p.peng@linux.intel.com>]

On Tue, Apr 21, 2015 at 04:39:05PM +0200, Dario Faggioli wrote:
> On Tue, 2015-04-21 at 15:01 +0100, Ian Campbell wrote:
> > On Tue, 2015-04-21 at 17:49 +0800, Chao Peng wrote:
> 
> > > I do like this suggestion and I have ever considered it actually. The
> > > only thing prevents me is that we need an extra _get_socket_count in xl
> > > for TARGET_ALL case. So libxl__count_physical_sockets is needed to be
> > > public. If Ian/Wei have no concerns for this, then I'm glad to do this.
> > 
> > I don't think you need libxl__count_physical_sockets for this,
> >
> I also don't think that is necessary.
> 
> >  the
> > existing xl code for similar things just uses libxl_bitmap_set_any to
> > handle the all case.
> > 
> That or, if you want to do something similar to what we have in *libxl*
> now you can look at the implementation of libxl_cpu_bitmap_alloc() (or
> libxl_node_bitmap_alloc() ) and define yours
> libxl_socket_bitmap_alloc().
> 
> In there, libxl_get_max_cpus() ( *_nodes()) is used, internally, to
> determine the size of the bitmap, if the user did not provide it
> explicitly. In your case, you can use libxl__count_physical_sockets() to
> the same effect, without the need for it to be public.
> 
> But then again, that is probably even not necessary.
> 
I tend to add libxl_socket_bitmap_alloc(), otherwise a prediefined
MAX_SOCKETS is needed.

Chao