On Thu, Jun 02, 2016 at 10:22:05PM -0700, Guru Shetty wrote: > On 2 June 2016 at 14:19, Ben Pfaff <b...@ovn.org> wrote: > > > On Thu, May 19, 2016 at 01:02:33PM -0700, Gurucharan Shetty wrote: > > > OVS NAT currently cannot do snat and dnat in the same zone. > > > So we need two zones per gateway router. > > > > > > Signed-off-by: Gurucharan Shetty <g...@ovn.org> > > > > We're running out of registers quickly, but we're also using a full > > 32-bit register when we only need 16 bits, so there's considerable room > > to economize later if necessary. > > > I agree. I will work on atleast combining the 2 registers being used here > into one.
It might require some new infrastructure, since I think there is some code around oriented around using a whole MFF_* field instead of a subfield. If so, I don't think it's essential for this patch. > > There are a couple of instances of > > + char *dnat = xasprintf(UUID_FMT"_%s", > > + > > UUID_ARGS(&binding->datapath->header_.uuid), > > + "dnat"); > > + char *snat = xasprintf(UUID_FMT"_%s", > > + > > UUID_ARGS(&binding->datapath->header_.uuid), > > + "snat"); > > or similar. Do you think it would be worth having a helper function (or > > two) so that they are harder to get out-of-sync? > > > > How about something like the following? Looks good to me, thanks! _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
