[dpdk-dev] [RFC 17.05 v1 0/3] Merge l3fwd-acl and l3fwd

[Ravi Kerur]

Hi Konstantin,

Thanks for the inputs.

Initialization starts with ETH_RSS_IP and gets updated with L4 if ACL is
selected in parse_args function. This change is to fix rss_hf.

I have submitted 'v4' patch, please let me know if additional changes are
needed.

Thanks.

On Sun, Mar 5, 2017 at 5:15 AM, Ananyev, Konstantin <
konstantin.ananyev at intel.com> wrote:

>
> Hi Ravi,
>
> >
> > Hi Konstantin,
> >
> > Sorry for this one, I had to resend patch series as 'v3' as additional
> checkpatch warnings were seen after the submission which didn't show
> > up in my run.
> >
> > 'v3' patch should have all fixed except the ones I mentioned in my
> earlier email on which I need inputs from you.
> >
> > Thanks.
> >
> > On Sat, Mar 4, 2017 at 11:49 AM, Ravi Kerur <rkerur at gmail.com> wrote:
> > Hi Konstantin,
> >
> > I have sent 'v2' patchset. I need clarifications on following things, if
> they should be fixed I will send out 'v3' so please let me know.
> >
> > Following code changes were done by me manually, not merged.
> > +++ b/examples/l3fwd/main.c
> > @@ -161,7 +163,9 @@  static struct rte_eth_conf port_conf = {
> >         .rx_adv_conf = {
> >                 .rss_conf = {
> >                         .rss_key = NULL,
> > -                       .rss_hf = ETH_RSS_IP,
> > +                       .rss_hf = ETH_RSS_IP | ETH_RSS_UDP |
> > +                               ETH_RSS_TCP | ETH_RSS_SCTP,
> > +
> >                 },
> >
> > The reason I did it is because
> >
> > LPM/EM has .rss_hf = ETH_RSS_IP
> > ACL has .rss_hf = ETH_RSS_IP | ETH_RSS_UDP | ETH_RSS_TCP | ETH_RSS_SCTP,
> >
> > ACL looks like a superset of LPM/EM and functional testing didn't reveal
> any issues hence I kept ACL version.
>
> But at least for LPM, we probably don't want L4 ports affect packet
> distribution?
> Probably the safest way would be to have a separate port_conf for each
> case (LPM/EM/ACL).
> That way will preserve the original behavior.
>
> >
> > 2. Checkpatch errors are all fixed. Some warnings are not fixed and they
> are
> >
> > 2.a, string length greater than 80 characters
> > 2.b GET_CB_FIELD macro. I could have changed GET_CB_FIELD to inline
> function, however, function names cannot be in capital letters. I
> > could have changed it to 'get_cb_field' inline function, but didn't do
> it as I thought it may not be worth the change.
>
> It is ok by me to leave as it is by now.
> Thanks
> Konstantin
>
>
>
>