Re: [PATCH 0 of 3 v5/leftover] Automatic NUMA placement for xl

From: Ian Campbell <ijc@hellion.org.uk>
To: Dario Faggioli <raistlin@linux.it>
Cc: Andre Przywara <andre.przywara@amd.com>,
	Stefano Stabellini <Stefano.Stabellini@eu.citrix.com>,
	George Dunlap <George.Dunlap@eu.citrix.com>,
	Juergen Gross <juergen.gross@ts.fujitsu.com>,
	Ian Jackson <Ian.Jackson@eu.citrix.com>,
	xen-devel <xen-devel@lists.xen.org>,
	David Vrabel <dvrabel@cantab.net>
Subject: Re: [PATCH 0 of 3 v5/leftover] Automatic NUMA placement for	xl
Date: Mon, 23 Jul 2012 11:42:07 +0100	[thread overview]
Message-ID: <1343040127.4091.11.camel@zakaz.uk.xensource.com> (raw)
In-Reply-To: <1343039906.4998.11.camel@Solace>

On Mon, 2012-07-23 at 11:38 +0100, Dario Faggioli wrote:
> On Fri, 2012-07-20 at 13:08 +0100, Ian Campbell wrote:
> > On Fri, 2012-07-20 at 13:00 +0100, Ian Campbell wrote:
> > > Even if we just blindly store the current result today as the expected
> > > one then when someone makes a tweak we can manually inspect the
> > > differences in the output and say "yes, that seems sane" or "no, that's
> > > mad". The need to do this will be encoded in the diff which makes the
> > > change (since you'd have to patch the test suite...)
> > 
> > BTW, this is mostly what we do for
> > tools/libxl/check-xl-{disk,vif}-parse.
> > 
> That is very nice, although our case here is a bit different. I can put
> together a similar script, and I can also think about some test cases
> and methods to check how things are going.
> 
> The tricky part is I _can't_ run xl just like we do in that file, as
> there's not a specific command for running just the placement, and I
> can't invoke `cl create', can I? :-P
> 
> So, it seems I should write a sort of C program from which to trigger
> the libxl domain creation, after taking care of intercepting and
> properly replacing quite a bit of functions to either nops or (e.g., in
> case of libxl_get_cpu_topology(), etc) proper reads of the test
> database. Also, this involves quite a bit of libxl internal function.
> Will I be able to intercept them too? Not sure, I need to check...
> 
> How does that sound? I'll think about and try to implement this in next
> days, any thoughts that may come to your mind, do feel free to
> share. :-D

I think it would be fine for a test program such as this to make use of
libxl internals and "look behind the curtain" as it were and call the
core of the placement algorithm directly with falsified inputs.

However I wouldn't prioritise this over any outstanding repost of the
series.

Ian.
-- 
Ian Campbell
Current Noise: Judas Priest - Painkiller

  After they got rid of capital punishment, they had to hang twice
  as many people as before.