On Mon, Jan 23, 2017 at 9:04 AM, Jarno Rajahalme <[email protected]> wrote:
> Pravin,
>
> Sorry for the delay, I hope you find time to respond.
>
>> On Dec 28, 2016, at 1:44 AM, Pravin B Shelar <[email protected]> wrote:
>>
>> Today packet mark action is broaken for Tunnel ports with
>> tunnel monitoring.
>
> “broaken”->”broken”
>
ok.

> Could you be more specific here. Setting packet mark works, but apparently 
> not when sending the packet to a tunnel, or possibly when sending packet to a 
> tunnel when tunnel monitoring (you mean BFD?). What exactly is the problem?
>
>> User can write a flow to set pkt-mark for
>> any tunnel packet generated by vswitchd itself, like BFD packets.
>>
>
> So the problem is that if a flow sets the pkt_mark, it will be also used for 
> BFD packets? Or is it that it will not be used for BFD packets? What is the 
> desired behavior?
>
Problem is for tunnels with tunnel monitoring. Using OVS flow table we
can set pkt-mark for tunnel packets in datapath, but not for tunnel
monitoring packets which are originated from vswitchd.

>> Following patch introduces new option in OVSDB tunnel
>> configuration so that user can set skb-mark for given
>> tunnel endpoint. OVS would set the mark according to the
>> skb-mark option for all tunnel traffic including packets
>> generated by vSwitchd like tunnel monitoring BFD packet.
>>
>
> What happens to the tunnel packet mark when the new option is not specified? 
> Is that behavior changed?
>
Nope, it remains the same. As mentioned in the man page it is optional
parameter.

>> Signed-off-by: Pravin B Shelar <[email protected]>
>> ---
>> lib/netdev-vport.c       |  8 ++++++++
>> lib/netdev.h             |  1 +
>> ofproto/tunnel.c         |  5 +++++
>> tests/tunnel-push-pop.at | 16 ++++++++++++++++
>> vswitchd/vswitch.xml     |  4 ++++
>> 5 files changed, 34 insertions(+)
>>
>> diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c
>> index 02a246a..25dd0e8 100644
>> --- a/lib/netdev-vport.c
>> +++ b/lib/netdev-vport.c
>> @@ -505,6 +505,10 @@ set_tunnel_config(struct netdev *dev_, const struct 
>> smap *args)
>>             }
>>
>>             free(str);
>> +        } else if (!strcmp(node->key, "egress_pkt_mark")) {
>> +            char *ptr = NULL;
>> +
>> +            tnl_cfg.egress_pkt_mark = strtoul(node->value, &ptr, 10);
>
> Could skip the “ptr” and use NULL instead.
>
ok.

> Should document somewhere what happens when the option is not given, which 
> seems to be the same as setting the egress_pkt_mark to zero.
>
>>         } else {
>>             VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->key);
>>         }
>> @@ -623,6 +627,10 @@ get_tunnel_config(const struct netdev *dev, struct smap 
>> *args)
>>         smap_add(args, "df_default", "false");
>>     }
>>
>> +    if (tnl_cfg.egress_pkt_mark) {
>> +        smap_add_format(args, "egress_pkt_mark",
>> +                        "%"PRIu32, tnl_cfg.egress_pkt_mark);
>> +    }
>
>
> Does a zero mark mean that the mark is not used? If the zero value is still 
> used is leaving it out misleading?
>
This is good point. If user explicitly sets the mark then we need to
reset it, I will update the patch.

>>     return 0;
>> }
>>
>> diff --git a/lib/netdev.h b/lib/netdev.h
>> index a667fe3..5ec1274 100644
>> --- a/lib/netdev.h
>> +++ b/lib/netdev.h
>> @@ -89,6 +89,7 @@ struct netdev_tunnel_config {
>>     struct in6_addr ipv6_dst;
>>
>>     uint32_t exts;
>> +    uint32_t egress_pkt_mark;
>>
>>     uint8_t ttl;
>>     bool ttl_inherit;
>> diff --git a/ofproto/tunnel.c b/ofproto/tunnel.c
>> index ce727f4..6205948 100644
>> --- a/ofproto/tunnel.c
>> +++ b/ofproto/tunnel.c
>> @@ -461,6 +461,11 @@ tnl_port_send(const struct ofport_dpif *ofport, struct 
>> flow *flow,
>>         | (cfg->csum ? FLOW_TNL_F_CSUM : 0)
>>         | (cfg->out_key_present ? FLOW_TNL_F_KEY : 0);
>>
>> +    if (cfg->egress_pkt_mark) {
>> +        flow->pkt_mark = cfg->egress_pkt_mark;
>> +        wc->masks.pkt_mark = UINT32_MAX;
>
> Here the pkt_mark from the flow is not used or inherited, so why are we 
> setting the mask (which causes an exact match on the incoming pkt_mark value 
> (which is likely zero))?
>
The problem is in commit odp action. If I do not set the mask then the
odp action uses zero mask and key to generate skb-mark set action.
Result is pkt-mark is always zero in set pkt-mark action.


>> +    }
>> +
>
> If egress_pkt_mark is zero, do we further on in the tunnel code (maybe due to 
> later patches) use the pkt_mask value from the flow?
>
If egress_pkt_mark is set then it would overwrite the flow packet mark.

>>     if (pre_flow_str) {
>>         char *post_flow_str = flow_to_string(flow);
>>         char *tnl_str = tnl_port_fmt(tnl_port);
>> diff --git a/tests/tunnel-push-pop.at b/tests/tunnel-push-pop.at
>> index 700ef55..4aaa669 100644
>> --- a/tests/tunnel-push-pop.at
>> +++ b/tests/tunnel-push-pop.at
>> @@ -12,6 +12,8 @@ AT_CHECK([ovs-vsctl add-port int-br t2 -- set Interface t2 
>> type=vxlan \
>>                        options:remote_ip=1.1.2.93 options:out_key=flow 
>> options:csum=true ofport_request=4\
>>                     -- add-port int-br t4 -- set Interface t4 type=geneve \
>>                        options:remote_ip=flow options:key=123 
>> ofport_request=5\
>> +                    -- add-port int-br t5 -- set Interface t5 type=geneve \
>> +                       options:remote_ip=1.1.2.93 options:out_key=flow 
>> options:egress_pkt_mark=1234 ofport_request=6\
>>                        ], [0])
>>
>> AT_CHECK([ovs-appctl dpif/show], [0], [dnl
>> @@ -25,6 +27,7 @@ dummy@ovs-dummy: hit:0 missed:0
>>               t2 2/4789: (vxlan: key=123, remote_ip=1.1.2.92)
>>               t3 4/4789: (vxlan: csum=true, out_key=flow, remote_ip=1.1.2.93)
>>               t4 5/6081: (geneve: key=123, remote_ip=flow)
>> +             t5 6/6081: (geneve: egress_pkt_mark=1234, out_key=flow, 
>> remote_ip=1.1.2.93)
>> ])
>>
>> dnl First setup dummy interface IP address, then add the route
>> @@ -91,6 +94,12 @@ AT_CHECK([tail -1 stdout], [0],
>>   [Datapath actions: tnl_pop(6081)
>> ])
>>
>> +dnl Check Geneve tunnel (t6) pop
>> +AT_CHECK([ovs-appctl ofproto/trace ovs-dummy 
>> 'in_port(1),eth(src=f8:bc:12:44:34:b6,dst=aa:55:aa:55:00:00),eth_type(0x0800),ipv4(src=1.1.2.96,dst=1.1.2.88,proto=17,tos=0,ttl=64,frag=no),udp(src=51283,dst=6081)'],
>>  [0], [stdout])
>> +AT_CHECK([tail -1 stdout], [0],
>> +  [Datapath actions: tnl_pop(6081)
>> +])
>> +
>> dnl Check VXLAN tunnel push
>> AT_CHECK([ovs-ofctl add-flow int-br action=2])
>> AT_CHECK([ovs-appctl ofproto/trace ovs-dummy 
>> 'in_port(2),eth_type(0x0800),ipv4(src=1.1.3.88,dst=1.1.3.112,proto=47,tos=0,ttl=64,frag=no)'],
>>  [0], [stdout])
>> @@ -119,6 +128,13 @@ AT_CHECK([tail -1 stdout], [0],
>>   [Datapath actions: 
>> tnl_push(tnl_port(6081),header(size=50,type=5,eth(dst=f8:bc:12:44:34:b6,src=aa:55:aa:55:00:00,dl_type=0x0800),ipv4(src=1.1.2.88,dst=1.1.2.92,proto=17,tos=0,ttl=64,frag=0x4000),udp(src=0,dst=6081,csum=0x0),geneve(vni=0x7b)),out_port(100))
>> ])
>>
>> +dnl Check Geneve tunnel push with pkt-mark
>> +AT_CHECK([ovs-ofctl add-flow int-br "actions=set_tunnel:234,6"])
>> +AT_CHECK([ovs-appctl ofproto/trace ovs-dummy 
>> 'in_port(2),eth_type(0x0800),ipv4(src=1.1.3.88,dst=1.1.3.112,proto=47,tos=0,ttl=64,frag=no)'],
>>  [0], [stdout])
>> +AT_CHECK([tail -1 stdout], [0],
>> +  [Datapath actions: 
>> set(skb_mark(0x4d2)),tnl_push(tnl_port(6081),header(size=50,type=5,eth(dst=f8:bc:12:44:34:b7,src=aa:55:aa:55:00:00,dl_type=0x0800),ipv4(src=1.1.2.88,dst=1.1.2.93,proto=17,tos=0,ttl=64,frag=0x4000),udp(src=0,dst=6081,csum=0x0),geneve(vni=0xea)),out_port(100))
>> +])
>> +
>> dnl Check Geneve tunnel push with options
>> AT_CHECK([ovs-ofctl add-tlv-map int-br 
>> "{class=0xffff,type=0x80,len=4}->tun_metadata0"])
>> AT_CHECK([ovs-ofctl add-flow int-br 
>> "actions=set_field:1.1.2.92->tun_dst,set_field:0xa->tun_metadata0,5"])
>> diff --git a/vswitchd/vswitch.xml b/vswitchd/vswitch.xml
>> index b4af5a5..3ec8193 100644
>> --- a/vswitchd/vswitch.xml
>> +++ b/vswitchd/vswitch.xml
>> @@ -2226,6 +2226,10 @@
>>         to <code>false</code> to disable.
>>       </column>
>>
>> +      <column name="options" key="egress_pkt_mark">
>> +        Optional.  The pkt_mark to be set on the encapsulating packet.
>> +      </column>
>> +
>
> We should document the behavior in the absence of this option as well.
>
ok.

>>       <group title="Tunnel Options: vxlan only">
>>
>>         <column name="options" key="exts">
>> --
>> 2.9.3
>>
>> _______________________________________________
>> dev mailing list
>> [email protected]
>> https://mail.openvswitch.org/mailman/listinfo/ovs-dev
>
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to