On Tue, Jun 26, 2012 at 09:59:13PM -0700, Ben Pfaff wrote: > On Wed, Jun 27, 2012 at 01:26:55PM +0900, Isaku Yamahata wrote: > > This patch is preliminary to support set_field action. > > Although it is only compile tested, I'd like to hear if this is the right > > direction before going further. > > Thanks. I'm not sure. I considered going a similar direction myself, > but I think I decided that retaining discrete "set field" actions for > OF1.0 and OF1.1 make it easier to accurately reproduce the actions that > the controller sent (see my comments on the final patch for details). > > set-field.c is also larger than I would have guessed.
Thank you for quick reply. ofp_reg_load is super set of set_field, so I'll use ofpact_reg_load with compat = OFPAT12_SET_FIELD instead of introducing ofpact_set_field. I wasn't aware of the format issue. Although I think it can be address with using ofpact_reg_load with compat = OFPAT1[01]_SET_xxx, I'm not sure it's desirable. So I'll postpone the patch to eliminate existing OFPACT_SET_xxx. Is this direction OK for you? > > Once this is done, I'd like to add instruction support next. > > For that, introduce OFPACT_IT_xxx and teach struct action_xlate_ctx and > > do_xlate_actions() instructions. Is this the dirction to go? Or other way? > > Did you see the start of instruction support in my own "of1.1 v2" > series? So far, it only supports the "Apply-Actions" instruction, but I > think that adding support for other instructions should be fairly > straightforward. Do you mean "second version of "[of1.1 v2 00/12] of1.1 rollup" series"? As far as I understand, it only includes decoders. Or am I missing something? thanks, -- yamahata _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
