On 22 March 2016 at 08:33, Russell Bryant <russ...@ovn.org> wrote: > On Tue, Mar 22, 2016 at 11:22 AM, Guru Shetty <g...@ovn.org> wrote: > > > > > > > On 21 March 2016 at 16:40, Russell Bryant <russ...@ovn.org> wrote: > > > >> On Mon, Mar 21, 2016 at 11:23 AM, Russell Bryant <russ...@ovn.org> > wrote: > >> > >>> > >>> > >>> On Mon, Mar 21, 2016 at 11:16 AM, Guru Shetty <g...@ovn.org> wrote: > >>> > >>>> > >>>> > >>>> On 21 March 2016 at 07:54, Russell Bryant <russ...@ovn.org> wrote: > >>>> > >>>>> Update the "ct_commit;" logical flow action to optionally take > >>>>> one or two parameters, setting the value of "ct_mark" or "ct_label". > >>>>> Supported ct_commit syntax now includes: > >>>>> > >>>>> ct_commit; > >>>>> ct_commit(); > >>>>> ct_commit(ct_mark=1); > >>>>> ct_commit(ct_label=1); > >>>>> ct_commit(ct_mark=1, ct_label=1); > >>>>> > >>>>> Setting ct_mark via this type of logical flow results in an OpenFlow > >>>>> flow that looks like: > >>>>> > >>>>> > >>>>> > actions=ct(commit,zone=NXM_NX_REG5[0..15],exec(set_field:0x1->ct_mark)) > >>>>> > >>>>> Similarly, setting ct_label results in: > >>>>> > >>>>> > >>>>> > actions=ct(commit,zone=NXM_NX_REG5[0..15],exec(set_field:0x1->ct_label)) > >>>>> > >>>> > >>>> I think this feature makes it tricky to share zones with other > stateful > >>>> additions. If you want to commit only once for all stateful services, > then > >>>> set-field for ct_mark and ct_label will need to be loaded to > registers in > >>>> advance, which I guess would mean that you loose 2 registers for this > >>>> purpose. > >>>> > >>> > >>> Do you have any thoughts on how I could change this to cause you less > >>> pain on the LB series? The next patch shows how this is used. In > >>> practice, it only uses ct_label. > >>> > >> > >> I seem to recall that you had used 2 registers, but then Ben pointed out > >> that you could just use 2 bits of a single register. The next patch > only > >> uses a single bit of ct_label, so it could just use another bit out of > an > >> "OVN bit flags" register to signal the value of ct_label. > >> > > > > Right. So I will use 2 bits of a single register. I had a quick look at > > the 3rd patch (and its different versions). So looks like ct_mark will > not > > be used? And only one bit in ct_label is used. I wonder whether the > action > > setting here in the logical flows should be a macro instead of INT. The > > macro will signify a single bit. > > > > That's right. It's currently only using a single bit of ct_label and not > using ct_mark at all. I was just trying add more generally useful > Logical_Flow support for any future needs. > > I'm not sure what you mean by using a macro here. Can you expand on the > idea? > Macro was probably wrong use of word. I mean to say, something like (very crude): ct_commit(ct_label=MARK_FOR_DELETION)
And you are only allowed to set certain values to ct_label and those values only set certain bits. This will likely mean that we can share the ct_mark and ct_label better across different features. > > -- > Russell Bryant > _______________________________________________ > dev mailing list > dev@openvswitch.org > http://openvswitch.org/mailman/listinfo/dev > _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
