[ovs-dev] [PATCH/RFC v2 3/8] Support decoding of NMX selection method

Simon Horman simon.horman at netronome.com
Wed Feb 18 14:01:31 UTC 2015


On Tue, Feb 17, 2015 at 11:43:21AM -0800, Ben Pfaff wrote:
> On Sun, Feb 15, 2015 at 02:13:14PM -0500, Simon Horman wrote:
> > On Sun, Feb 15, 2015 at 08:51:31AM -0800, Ben Pfaff wrote:
> > > On Fri, Feb 13, 2015 at 06:14:28PM -0500, Simon Horman wrote:
> > > > On Fri, Feb 13, 2015 at 02:55:58PM -0800, Ben Pfaff wrote:
> > > > > On Fri, Jan 30, 2015 at 11:41:51AM +0900, Simon Horman wrote:
> > > > > > This is in preparation for supporting group mod and desc reply
> > > > > > messages with an NMX selection method group experimenter property.
> > > > > > 
> > > > > > NMX selection method
> > > > > > Signed-off-by: Simon Horman <simon.horman at netronome.com>
> > > > > > 
> > > > > > ---
> > > > > > v2 Use list of struct field_array of TLVs rather than OF1.1 match
> > > > > >    for fields field of NMX selection method property
> > > > > 
> > > > > It might be simpler to use an actual array for struct field_array.  I
> > > > > guess that we can be sure there will be no more than MFF_N_IDS
> > > > > elements?
> > > > 
> > > > As duplicates are rejected I believe that we can make that assumption.
> > > > 
> > > > I think that should simplify the code somewhat at the
> > > > expense of some memory cost if the property is present.
> > > > I'm assuming that you are happy about the latter and
> > > > I'll see about making it so.
> > > 
> > > I'm not sure yet as I haven't read the rest of the series yet.  If
> > > you're unsure then you might want to wait for the rest of the reviews.
> > 
> > Thanks. I'll wait.
> > 
> > After writing my previous email I recalled the original motivation
> > for the empty string indicating that the existing behaviour should occur.
> > 
> > My original proposal extended the group mod message and as such the fields
> > to describe the selection method were always present. In that scenario
> > using a empty field to denote that the feature was turned off made sense
> > (at least to me).  Later, on your advice IIRC, I reworked things to use a
> > group mod property which may be present or absent. And at that time I
> > didn't modify the empty field behaviour which, as you pointed out, seems
> > unnecessary now.
> 
> I understand now.  I'd prefer to only allow such a property if the
> string names a supported method.  (This might require squashing some
> patches; that's fine.)

Thanks, I'll see about making it so.

> Would you mind rebasing the series now and reposting it?  I'd like to
> do build testing as I go and I'm having trouble finding a base that I
> can apply everything to cleanly.  Alternatively it would be fine to
> push the series as-is to a branch somewhere.

I have pushed the series to:

https://github.com/horms/openvswitch.git devel/ext-350+selection_method

The base is somewhat old.



More information about the dev mailing list