[ovs-discuss] [ovs-dev] ovn: Issue in ACL stage when the dhcp reply packet is resumed

Tue May 10 07:00:10 UTC 2016

Thanks Jarno, sure I will do that.

On Tue, May 10, 2016 at 8:08 AM, Jarno Rajahalme <jarno at ovn.org> wrote:

> Numan,
>
> This seems to be the right fix. In general, recirculations due to
> NXT_RESUME will fail if the packet metadata is not restored prior to the
> packet execution.
>
> Would you send a patch formatted by git format-patch with your
> Signed-off-by?
>
>   Jarno
>
> > On May 9, 2016, at 1:55 AM, Numan Siddique <nusiddiq at redhat.com> wrote:
> >
> > The below fix is resolving the issue. But I am not sure if that's the
> right
> > solution.
> > May be the packet metadata should also be saved in "struct
> > ofputil_packet_in" in "emit_continuation()".
> >
> > ---
> > ofproto/ofproto-dpif.c | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c
> > index a1e0407..00369b6 100644
> > --- a/ofproto/ofproto-dpif.c
> > +++ b/ofproto/ofproto-dpif.c
> > @@ -4399,6 +4399,7 @@ nxt_resume(struct ofproto *ofproto_,
> >     struct dp_packet packet;
> >     dp_packet_init(&packet, pin->public.packet_len);
> >     dp_packet_put(&packet, pin->public.packet, pin->public.packet_len);
> > +    pkt_metadata_from_flow(&packet.md,
> &pin->public.flow_metadata.flow);
> >
> >     struct flow headers;
> >     flow_extract(&packet, &headers);
> > --
> >
> > Thanks
> > Numan
> >
> >
> > On Fri, May 6, 2016 at 6:35 PM, Numan Siddique <nusiddiq at redhat.com>
> wrote:
> >
> >>
> >> Thanks Babu for your comments.
> >>
> >> I am seeing the problem even if NXM_OF_IN_PORT is not set to zero after
> >> the packet is resumed.
> >>
> >> The issue is that a resumed packet is getting dropped when it executes
> the
> >> action 'OFPACT_CT'.
> >> 'xlate_lookup()' is failing in the 'upcall_receive()'
> >>
> >> Ben, if you want to reproduce the issue you can run the below commands
> >>
> >> Thanks
> >> Numan
> >>
> >> -----------------------------------------------------------------------
> >> BRIDGE=ovs-test
> >> ovs-vsctl add-br $BRIDGE
> >>
> >> ovs-ofctl add-flow $BRIDGE
> >> "table=0,priority=1000,actions=controller(pause), resubmit(,1)"
> >> ovs-ofctl add-flow $BRIDGE "table=1,priority=1000,ip,
> >> actions=ct(table=2,zone=1)"
> >> ovs-ofctl add-flow $BRIDGE "table=1,priority=1000,arp, actions=FLOOD"
> >> ovs-ofctl add-flow $BRIDGE
> >> "table=2,priority=1000,actions=ct(commit,zone=1),resubmit(,3)"
> >> ovs-ofctl add-flow $BRIDGE "table=3,priority=1000,actions=NORMAL"
> >>
> >> ip netns add ns1
> >> ip netns add ns2
> >>
> >> ovs-vsctl add-port $BRIDGE tap1 -- set Interface tap1 type=internal
> >> ip link set tap1 netns ns1
> >> ip netns exec ns1 ip link set dev tap1 up
> >> ip netns exec ns1 ip addr add 10.0.0.2/24 dev tap1
> >>
> >> ovs-vsctl add-port $BRIDGE tap2 -- set Interface tap2 type=internal
> >> ip link set tap2 netns ns2
> >> ip netns exec ns2 ip link set dev tap2 up
> >> ip netns exec ns2 ip addr add 10.0.0.3/24 dev tap2
> >>
> >> # start ovs-ofctl monitor on $BRIDGE in another shell
> >> ovs-ofctl monitor $BRIDGE resume
> >>
> >> ip netns exec ns1 ping -c 1 10.0.0.3
> >> -------------------------------------------------
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >> On Thu, May 5, 2016 at 5:36 PM, Babu Shanmugam <bschanmu at redhat.com>
> >> wrote:
> >>
> >>> Numan,
> >>> It happens because flow chain is paused at the DHCP flow and while
> >>> resuming, the inport is set 0. When the packet traverse down the
> pipeline,
> >>> while processing an upcall for conntrack in the 49th table, a lookup
> for DP
> >>> inport 0 is performed and could not be found. Hence, it was not able to
> >>> proceed further.
> >>>
> >>> I think using packet-out with necessary actions in the userdata for
> >>> packet traversal would be better. We can use logical flow syntax such
> as
> >>> "output(inport)" to imply that the response is a packet-out to the
> inport
> >>> from which packet came. Suggestion are welcome.
> >>>
> >>> Thank you,
> >>> Babu
> >>>
> >>>
> >>> On Wednesday 13 April 2016 09:23 PM, Numan Siddique wrote:
> >>>
> >>>> Hi,
> >>>>
> >>>> I am seeing an issue while testing the ovn native dhcp implementation.
> >>>>
> >>>> When the dhcp reply packet from ovn-controller is resumed by
> >>>> ovs-vswitchd,
> >>>> the packet is getting dropped in the egress pipeline - ACL stage.
> >>>>
> >>>> I see the below messages in the vswitchd.log
> >>>>
> >>>> ------------------
> >>>>
> >>>>
> 2016-04-13T14:14:03.813Z|00002|ofproto_dpif_upcall(handler1)|INFO|received
> >>>> packet on unassociated datapath port 0
> >>>>
> >>>>
> 2016-04-13T14:14:03.860Z|00001|ofproto_dpif_upcall(revalidator2)|WARN|Failed
> >>>> to acquire udpif_key corresponding to unexpected flow (Invalid
> argument):
> >>>> ufid:a33e557f-369f-404e-aefb-2e3da678b09a
> >>>> ------------------------
> >>>>
> >>>>
> >>>> In the function upcall_receive(), I could see that xlate_lookup() is
> >>>> returning error.
> >>>>
> >>>> Also I could see that in the function odp_flow_key_to_flow__()
> >>>> L5191,  nl_attr_get_odp_port(attrs[OVS_KEY_ATTR_IN_PORT]) is
> returning 0.
> >>>>
> >>>>
> >>>> Please let me know if you need further information.
> >>>>
> >>>> Thanks
> >>>> Numan
> >>>> _______________________________________________
> >>>> dev mailing list
> >>>> dev at openvswitch.org
> >>>> http://openvswitch.org/mailman/listinfo/dev
> >>>>
> >>>
> >>>
> >>
> > _______________________________________________
> > dev mailing list
> > dev at openvswitch.org
> > http://openvswitch.org/mailman/listinfo/dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20160510/b028cfa2/attachment-0002.html>