[ovs-dev] [PATCH v4] ofproto: per-table statistics

[Simon Horman]

On Thu, May 01, 2014 at 08:02:32AM -0700, Ben Pfaff wrote:
> On Thu, May 01, 2014 at 05:58:00PM +0900, Simon Horman wrote:
> > On Wed, Apr 30, 2014 at 09:54:30AM +0900, YAMAMOTO Takashi wrote:
> > > > On Tue, Apr 29, 2014 at 06:34:22PM +0900, Simon Horman wrote:
> > > >> Add per-table counters. This resolves some short-comings
> > > >> in the data provided in a table stats reply message.
> > > >> 
> > > >> * Lookups and matches are calculated based on table
> > > >>   accesses rather than datapath flow hits and misses.
> > > >> 
> > > >> * Lookups and matches are credited to the table where they
> > > >>   occurred rather than all being credited to table 0.
> > > >> 
> > > >> These problems were observed when running make check-ryu
> > > >> and this patch allows many of its tester.py match checks
> > > >> to pass.
> > > > 
> > > > I'm not sure that atomic_uint64 is portable to all platforms (notably I
> > > > think some 32-bit platforms don't have atomic 64-bit ints).  I'd feel
> > > > more comfortable using atomic_ulong.
> > > 
> > > this is not the first use of them.  cf. udpif::n_flows
> > > 
> > > if you want to avoid the use of them, how about removing them
> > > from ovs-atomic.h?
> > 
> > Hi Ben,
> > 
> > I'm happy to make a patch to update the existing user of atomic_uint64_t
> > and, if you like, remove it from ovs-atomic.h.
> > 
> > Please let me know if that is how you would like to handle things.
> 
> Let's do that.
> 
> My original rationale behind including these at all was that they could
> be useful in code that only runs on 64-bit systems.  But I'm not sure
> that we actually have any code like that, and anyway now that the ATOMIC
> macro exists we can construct a 64-bit atomic int with ATOMIC(uint64_t).

Thanks.

I am on vacation this week.
I should be able to prepare some patches after I return next week.