diff mbox series

[ovs-dev,RFC,ovn] ovn: introduce IP_SRC_POLICY stage in ingress router pipeline

Message ID a99d46b275a10a6d8278434f7cefa0466bb78af5.1589557561.git.lorenzo.bianconi@redhat.com
State New
Headers show
Series [ovs-dev,RFC,ovn] ovn: introduce IP_SRC_POLICY stage in ingress router pipeline | expand

Commit Message

Lorenzo Bianconi May 15, 2020, 3:49 p.m. UTC
In order to fix the issues introduced by commit
c0bf32d72f8b ("Manage ARP process locally in a DVR scenario "), restore
previous configuration of table 9 in ingress router pipeline and
introduce a new stage called 'ip_src_policy' used to set the src address
info in order to not distribute FIP traffic if DVR is enabled

Fixes: c0bf32d72f8b ("Manage ARP process locally in a DVR scenario ")
Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
---
 northd/ovn-northd.8.xml | 65 ++++++++++++++++++++---------------------
 northd/ovn-northd.c     | 38 ++++++++++--------------
 tests/ovn.at            | 28 +++++-------------
 3 files changed, 54 insertions(+), 77 deletions(-)

Comments

Dumitru Ceara May 15, 2020, 5:29 p.m. UTC | #1
On 5/15/20 5:49 PM, Lorenzo Bianconi wrote:
> In order to fix the issues introduced by commit
> c0bf32d72f8b ("Manage ARP process locally in a DVR scenario "), restore
> previous configuration of table 9 in ingress router pipeline and
> introduce a new stage called 'ip_src_policy' used to set the src address
> info in order to not distribute FIP traffic if DVR is enabled
> 
> Fixes: c0bf32d72f8b ("Manage ARP process locally in a DVR scenario ")
> Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>

Hi Lorenzo,

Thanks for working on fixing this issue!

As discussed offline, I'm a bit concerned about adding a new stage to
the OVN pipeline and I'd prefer if we could find a different way to
address this.

This being said, if there's no other easy way of fixing this, I think
your change looks ok and will handle all the bugs reported, including [0].

Thanks,
Dumitru

[0] https://bugzilla.redhat.com/show_bug.cgi?id=1834433

> ---
>  northd/ovn-northd.8.xml | 65 ++++++++++++++++++++---------------------
>  northd/ovn-northd.c     | 38 ++++++++++--------------
>  tests/ovn.at            | 28 +++++-------------
>  3 files changed, 54 insertions(+), 77 deletions(-)
> 
> diff --git a/northd/ovn-northd.8.xml b/northd/ovn-northd.8.xml
> index 8f224b07f..09dbb52b4 100644
> --- a/northd/ovn-northd.8.xml
> +++ b/northd/ovn-northd.8.xml
> @@ -2484,37 +2484,6 @@ output;
>          </p>
>        </li>
>  
> -      <li>
> -        <p>
> -          For distributed logical routers where one of the logical router ports
> -          specifies a <code>redirect-chassis</code>, a priority-400 logical
> -          flow for each <code>dnat_and_snat</code> NAT rules configured.
> -          These flows will allow to properly forward traffic to the external
> -          connections if available and avoid sending it through the tunnel.
> -          Assuming the following NAT rule has been configured:
> -        </p>
> -
> -        <pre>
> -external_ip = <var>A</var>;
> -external_mac = <var>B</var>;
> -logical_ip = <var>C</var>;
> -        </pre>
> -
> -        <p>
> -          the following action will be applied:
> -        </p>
> -
> -        <pre>
> -ip.ttl--;
> -reg0 = <var>ip.dst</var>;
> -reg1 = <var>A</var>;
> -eth.src = <var>B</var>;
> -outport = <var>router-port</var>;
> -next;
> -        </pre>
> -
> -      </li>
> -
>        <li>
>          <p>
>            IPv4 routing table.  For each route to IPv4 network <var>N</var> with
> @@ -2660,7 +2629,35 @@ outport = <var>P</var>;
>        </li>
>      </ul>
>  
> -    <h3>Ingress Table 12: ARP/ND Resolution</h3>
> +    <h3>Ingress Table 12: IP Source Policy</h3>
> +
> +    <p>
> +      This table contains for distributed logical routers where one of
> +      the logical router ports specifies a <code>redirect-chassis</code>,
> +      a priority-100 logical flow for each <code>dnat_and_snat</code>
> +      NAT rules configured.
> +      These flows will allow to properly forward traffic to the external
> +      connections if available and avoid sending it through the tunnel.
> +      Assuming the following NAT rule has been configured:
> +    </p>
> +
> +    <pre>
> +external_ip = <var>A</var>;
> +external_mac = <var>B</var>;
> +logical_ip = <var>C</var>;
> +    </pre>
> +
> +    <p>
> +      the following action will be applied:
> +    </p>
> +
> +    <pre>
> +reg1 = <var>A</var>;
> +eth.src = <var>B</var>;
> +next;
> +    </pre>
> +
> +    <h3>Ingress Table 13: ARP/ND Resolution</h3>
>  
>      <p>
>        Any packet that reaches this table is an IP packet whose next-hop
> @@ -2819,7 +2816,7 @@ outport = <var>P</var>;
>  
>      </ul>
>  
> -    <h3>Ingress Table 13: Check packet length</h3>
> +    <h3>Ingress Table 14: Check packet length</h3>
>  
>      <p>
>        For distributed logical routers with distributed gateway port configured
> @@ -2849,7 +2846,7 @@ REGBIT_PKT_LARGER = check_pkt_larger(<var>L</var>); next;
>        and advances to the next table.
>      </p>
>  
> -    <h3>Ingress Table 14: Handle larger packets</h3>
> +    <h3>Ingress Table 15: Handle larger packets</h3>
>  
>      <p>
>        For distributed logical routers with distributed gateway port configured
> diff --git a/northd/ovn-northd.c b/northd/ovn-northd.c
> index 3c0070ea7..d5f3997a9 100644
> --- a/northd/ovn-northd.c
> +++ b/northd/ovn-northd.c
> @@ -175,11 +175,12 @@ enum ovn_stage {
>      PIPELINE_STAGE(ROUTER, IN,  IP_ROUTING,      9, "lr_in_ip_routing")   \
>      PIPELINE_STAGE(ROUTER, IN,  IP_ROUTING_ECMP, 10, "lr_in_ip_routing_ecmp") \
>      PIPELINE_STAGE(ROUTER, IN,  POLICY,          11, "lr_in_policy")       \
> -    PIPELINE_STAGE(ROUTER, IN,  ARP_RESOLVE,     12, "lr_in_arp_resolve")  \
> -    PIPELINE_STAGE(ROUTER, IN,  CHK_PKT_LEN   ,  13, "lr_in_chk_pkt_len")   \
> -    PIPELINE_STAGE(ROUTER, IN,  LARGER_PKTS,     14,"lr_in_larger_pkts")   \
> -    PIPELINE_STAGE(ROUTER, IN,  GW_REDIRECT,     15, "lr_in_gw_redirect")  \
> -    PIPELINE_STAGE(ROUTER, IN,  ARP_REQUEST,     16, "lr_in_arp_request")  \
> +    PIPELINE_STAGE(ROUTER, IN,  IP_SRC_POLICY,   12, "lr_in_ip_src_policy") \
> +    PIPELINE_STAGE(ROUTER, IN,  ARP_RESOLVE,     13, "lr_in_arp_resolve")  \
> +    PIPELINE_STAGE(ROUTER, IN,  CHK_PKT_LEN   ,  14, "lr_in_chk_pkt_len")   \
> +    PIPELINE_STAGE(ROUTER, IN,  LARGER_PKTS,     15,"lr_in_larger_pkts")   \
> +    PIPELINE_STAGE(ROUTER, IN,  GW_REDIRECT,     16, "lr_in_gw_redirect")  \
> +    PIPELINE_STAGE(ROUTER, IN,  ARP_REQUEST,     17, "lr_in_arp_request")  \
>                                                                        \
>      /* Logical router egress stages. */                               \
>      PIPELINE_STAGE(ROUTER, OUT, UNDNAT,    0, "lr_out_undnat")        \
> @@ -7103,8 +7104,6 @@ build_routing_policy_flow(struct hmap *lflows, struct ovn_datapath *od,
>      ds_destroy(&actions);
>  }
>  
> -/* default logical flow prioriry for distributed routes */
> -#define DROUTE_PRIO 400
>  struct parsed_route {
>      struct ovs_list list_node;
>      struct v46_ip prefix;
> @@ -7493,7 +7492,7 @@ build_ecmp_route_flow(struct hmap *lflows, struct ovn_datapath *od,
>  }
>  
>  static void
> -add_distributed_routes(struct hmap *lflows, struct ovn_datapath *od)
> +add_ip_src_policy_flows(struct hmap *lflows, struct ovn_datapath *od)
>  {
>      struct ds actions = DS_EMPTY_INITIALIZER;
>      struct ds match = DS_EMPTY_INITIALIZER;
> @@ -7511,12 +7510,9 @@ add_distributed_routes(struct hmap *lflows, struct ovn_datapath *od)
>                        is_ipv4 ? "4" : "6", nat->logical_ip,
>                        nat->logical_port);
>          char *prefix = is_ipv4 ? "" : "xx";
> -        ds_put_format(&actions, "outport = %s; eth.src = %s; "
> -                      "%sreg0 = ip%s.dst; %sreg1 = %s; next;",
> -                      od->l3dgw_port->json_key, nat->external_mac,
> -                      prefix, is_ipv4 ? "4" : "6",
> -                      prefix, nat->external_ip);
> -        ovn_lflow_add(lflows, od, S_ROUTER_IN_IP_ROUTING, DROUTE_PRIO,
> +        ds_put_format(&actions, "eth.src = %s; %sreg1 = %s; next;",
> +                      nat->external_mac, prefix, nat->external_ip);
> +        ovn_lflow_add(lflows, od, S_ROUTER_IN_IP_SRC_POLICY, 100,
>                        ds_cstr(&match), ds_cstr(&actions));
>          ds_clear(&match);
>          ds_clear(&actions);
> @@ -7547,12 +7543,6 @@ add_route(struct hmap *lflows, const struct ovn_port *op,
>      }
>      build_route_match(op_inport, network_s, plen, is_src_route, is_ipv4,
>                        &match, &priority);
> -    /* traffic for internal IPs of logical switch ports must be sent to
> -     * the gw controller through the overlay tunnels
> -     */
> -    if (op->nbrp && !op->nbrp->n_gateway_chassis) {
> -        priority += DROUTE_PRIO;
> -    }
>  
>      struct ds actions = DS_EMPTY_INITIALIZER;
>      ds_put_format(&actions, "ip.ttl--; "REG_ECMP_GROUP_ID" = 0; %sreg0 = ",
> @@ -9519,9 +9509,13 @@ build_lrouter_flows(struct hmap *datapaths, struct hmap *ports,
>       * logical router
>       */
>      HMAP_FOR_EACH (od, key_node, datapaths) {
> -        if (od->nbr && od->l3dgw_port) {
> -            add_distributed_routes(lflows, od);
> +        if (!od->nbr) {
> +            continue;
> +        }
> +        if (od->l3dgw_port) {
> +            add_ip_src_policy_flows(lflows, od);
>          }
> +        ovn_lflow_add(lflows, od, S_ROUTER_IN_IP_SRC_POLICY, 0, "1", "next;");
>      }
>  
>      /* Logical router ingress table IP_ROUTING & IP_ROUTING_ECMP: IP Routing.
> diff --git a/tests/ovn.at b/tests/ovn.at
> index f39fda2e4..fcc34fd5d 100644
> --- a/tests/ovn.at
> +++ b/tests/ovn.at
> @@ -9637,20 +9637,6 @@ AT_CHECK([as hv3 ovs-vsctl set Open_vSwitch . external-ids:ovn-bridge-mappings=p
>  OVS_WAIT_UNTIL([test 1 = `as hv3 ovs-vsctl show | \
>  grep "Port patch-br-int-to-ln_port" | wc -l`])
>  
> -AT_CHECK([test 1 = `ovn-sbctl dump-flows lr0 | grep lr_in_ip_routing | \
> -grep "ip4.src == 10.0.0.3 && is_chassis_resident(\"foo1\")" -c`])
> -AT_CHECK([test 1 = `ovn-sbctl dump-flows lr0 | grep lr_in_ip_routing | \
> -grep "ip4.src == 10.0.0.4 && is_chassis_resident(\"foo2\")" -c`])
> -
> -key=`ovn-sbctl --bare --columns tunnel_key list datapath_Binding lr0`
> -# Check that the OVS flows appear for the dnat_and_snat entries in
> -# lr_in_ip_routing table.
> -OVS_WAIT_UNTIL([test 1 = `as hv3 ovs-ofctl dump-flows br-int table=17 | \
> -grep "priority=400,ip,metadata=0x$key,nw_src=10.0.0.3" -c`])
> -
> -OVS_WAIT_UNTIL([test 1 = `as hv3 ovs-ofctl dump-flows br-int table=17 | \
> -grep "priority=400,ip,metadata=0x$key,nw_src=10.0.0.4" -c`])
> -
>  # Re-add nat-addresses option
>  ovn-nbctl lsp-set-options lrp0-rp router-port=lrp0 nat-addresses="router"
>  
> @@ -15141,7 +15127,7 @@ ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
>  # Since the sw0-vir is not claimed by any chassis, eth.dst should be set to
>  # zero if the ip4.dst is the virtual ip in the router pipeline.
>  AT_CHECK([cat lflows.txt], [0], [dnl
> -  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 00:00:00:00:00:00; next;)
> +  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 00:00:00:00:00:00; next;)
>  ])
>  
>  ip_to_hex() {
> @@ -15192,7 +15178,7 @@ ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
>  # There should be an arp resolve flow to resolve the virtual_ip with the
>  # sw0-p1's MAC.
>  AT_CHECK([cat lflows.txt], [0], [dnl
> -  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:03; next;)
> +  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:03; next;)
>  ])
>  
>  # Forcibly clear virtual_parent. ovn-controller should release the binding
> @@ -15233,7 +15219,7 @@ ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
>  # There should be an arp resolve flow to resolve the virtual_ip with the
>  # sw0-p2's MAC.
>  AT_CHECK([cat lflows.txt], [0], [dnl
> -  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:05; next;)
> +  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:05; next;)
>  ])
>  
>  # send the garp from sw0-p2 (in hv2). hv2 should claim sw0-vir
> @@ -15256,7 +15242,7 @@ ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
>  # There should be an arp resolve flow to resolve the virtual_ip with the
>  # sw0-p3's MAC.
>  AT_CHECK([cat lflows.txt], [0], [dnl
> -  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:04; next;)
> +  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:04; next;)
>  ])
>  
>  # Now send arp reply from sw0-p1. hv1 should claim sw0-vir
> @@ -15277,7 +15263,7 @@ ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
>  > lflows.txt
>  
>  AT_CHECK([cat lflows.txt], [0], [dnl
> -  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:03; next;)
> +  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:03; next;)
>  ])
>  
>  # Delete hv1-vif1 port. hv1 should release sw0-vir
> @@ -15295,7 +15281,7 @@ ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
>  > lflows.txt
>  
>  AT_CHECK([cat lflows.txt], [0], [dnl
> -  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 00:00:00:00:00:00; next;)
> +  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 00:00:00:00:00:00; next;)
>  ])
>  
>  # Now send arp reply from sw0-p2. hv2 should claim sw0-vir
> @@ -15316,7 +15302,7 @@ ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
>  > lflows.txt
>  
>  AT_CHECK([cat lflows.txt], [0], [dnl
> -  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:04; next;)
> +  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:04; next;)
>  ])
>  
>  # Delete sw0-p2 logical port
>
diff mbox series

Patch

diff --git a/northd/ovn-northd.8.xml b/northd/ovn-northd.8.xml
index 8f224b07f..09dbb52b4 100644
--- a/northd/ovn-northd.8.xml
+++ b/northd/ovn-northd.8.xml
@@ -2484,37 +2484,6 @@  output;
         </p>
       </li>
 
-      <li>
-        <p>
-          For distributed logical routers where one of the logical router ports
-          specifies a <code>redirect-chassis</code>, a priority-400 logical
-          flow for each <code>dnat_and_snat</code> NAT rules configured.
-          These flows will allow to properly forward traffic to the external
-          connections if available and avoid sending it through the tunnel.
-          Assuming the following NAT rule has been configured:
-        </p>
-
-        <pre>
-external_ip = <var>A</var>;
-external_mac = <var>B</var>;
-logical_ip = <var>C</var>;
-        </pre>
-
-        <p>
-          the following action will be applied:
-        </p>
-
-        <pre>
-ip.ttl--;
-reg0 = <var>ip.dst</var>;
-reg1 = <var>A</var>;
-eth.src = <var>B</var>;
-outport = <var>router-port</var>;
-next;
-        </pre>
-
-      </li>
-
       <li>
         <p>
           IPv4 routing table.  For each route to IPv4 network <var>N</var> with
@@ -2660,7 +2629,35 @@  outport = <var>P</var>;
       </li>
     </ul>
 
-    <h3>Ingress Table 12: ARP/ND Resolution</h3>
+    <h3>Ingress Table 12: IP Source Policy</h3>
+
+    <p>
+      This table contains for distributed logical routers where one of
+      the logical router ports specifies a <code>redirect-chassis</code>,
+      a priority-100 logical flow for each <code>dnat_and_snat</code>
+      NAT rules configured.
+      These flows will allow to properly forward traffic to the external
+      connections if available and avoid sending it through the tunnel.
+      Assuming the following NAT rule has been configured:
+    </p>
+
+    <pre>
+external_ip = <var>A</var>;
+external_mac = <var>B</var>;
+logical_ip = <var>C</var>;
+    </pre>
+
+    <p>
+      the following action will be applied:
+    </p>
+
+    <pre>
+reg1 = <var>A</var>;
+eth.src = <var>B</var>;
+next;
+    </pre>
+
+    <h3>Ingress Table 13: ARP/ND Resolution</h3>
 
     <p>
       Any packet that reaches this table is an IP packet whose next-hop
@@ -2819,7 +2816,7 @@  outport = <var>P</var>;
 
     </ul>
 
-    <h3>Ingress Table 13: Check packet length</h3>
+    <h3>Ingress Table 14: Check packet length</h3>
 
     <p>
       For distributed logical routers with distributed gateway port configured
@@ -2849,7 +2846,7 @@  REGBIT_PKT_LARGER = check_pkt_larger(<var>L</var>); next;
       and advances to the next table.
     </p>
 
-    <h3>Ingress Table 14: Handle larger packets</h3>
+    <h3>Ingress Table 15: Handle larger packets</h3>
 
     <p>
       For distributed logical routers with distributed gateway port configured
diff --git a/northd/ovn-northd.c b/northd/ovn-northd.c
index 3c0070ea7..d5f3997a9 100644
--- a/northd/ovn-northd.c
+++ b/northd/ovn-northd.c
@@ -175,11 +175,12 @@  enum ovn_stage {
     PIPELINE_STAGE(ROUTER, IN,  IP_ROUTING,      9, "lr_in_ip_routing")   \
     PIPELINE_STAGE(ROUTER, IN,  IP_ROUTING_ECMP, 10, "lr_in_ip_routing_ecmp") \
     PIPELINE_STAGE(ROUTER, IN,  POLICY,          11, "lr_in_policy")       \
-    PIPELINE_STAGE(ROUTER, IN,  ARP_RESOLVE,     12, "lr_in_arp_resolve")  \
-    PIPELINE_STAGE(ROUTER, IN,  CHK_PKT_LEN   ,  13, "lr_in_chk_pkt_len")   \
-    PIPELINE_STAGE(ROUTER, IN,  LARGER_PKTS,     14,"lr_in_larger_pkts")   \
-    PIPELINE_STAGE(ROUTER, IN,  GW_REDIRECT,     15, "lr_in_gw_redirect")  \
-    PIPELINE_STAGE(ROUTER, IN,  ARP_REQUEST,     16, "lr_in_arp_request")  \
+    PIPELINE_STAGE(ROUTER, IN,  IP_SRC_POLICY,   12, "lr_in_ip_src_policy") \
+    PIPELINE_STAGE(ROUTER, IN,  ARP_RESOLVE,     13, "lr_in_arp_resolve")  \
+    PIPELINE_STAGE(ROUTER, IN,  CHK_PKT_LEN   ,  14, "lr_in_chk_pkt_len")   \
+    PIPELINE_STAGE(ROUTER, IN,  LARGER_PKTS,     15,"lr_in_larger_pkts")   \
+    PIPELINE_STAGE(ROUTER, IN,  GW_REDIRECT,     16, "lr_in_gw_redirect")  \
+    PIPELINE_STAGE(ROUTER, IN,  ARP_REQUEST,     17, "lr_in_arp_request")  \
                                                                       \
     /* Logical router egress stages. */                               \
     PIPELINE_STAGE(ROUTER, OUT, UNDNAT,    0, "lr_out_undnat")        \
@@ -7103,8 +7104,6 @@  build_routing_policy_flow(struct hmap *lflows, struct ovn_datapath *od,
     ds_destroy(&actions);
 }
 
-/* default logical flow prioriry for distributed routes */
-#define DROUTE_PRIO 400
 struct parsed_route {
     struct ovs_list list_node;
     struct v46_ip prefix;
@@ -7493,7 +7492,7 @@  build_ecmp_route_flow(struct hmap *lflows, struct ovn_datapath *od,
 }
 
 static void
-add_distributed_routes(struct hmap *lflows, struct ovn_datapath *od)
+add_ip_src_policy_flows(struct hmap *lflows, struct ovn_datapath *od)
 {
     struct ds actions = DS_EMPTY_INITIALIZER;
     struct ds match = DS_EMPTY_INITIALIZER;
@@ -7511,12 +7510,9 @@  add_distributed_routes(struct hmap *lflows, struct ovn_datapath *od)
                       is_ipv4 ? "4" : "6", nat->logical_ip,
                       nat->logical_port);
         char *prefix = is_ipv4 ? "" : "xx";
-        ds_put_format(&actions, "outport = %s; eth.src = %s; "
-                      "%sreg0 = ip%s.dst; %sreg1 = %s; next;",
-                      od->l3dgw_port->json_key, nat->external_mac,
-                      prefix, is_ipv4 ? "4" : "6",
-                      prefix, nat->external_ip);
-        ovn_lflow_add(lflows, od, S_ROUTER_IN_IP_ROUTING, DROUTE_PRIO,
+        ds_put_format(&actions, "eth.src = %s; %sreg1 = %s; next;",
+                      nat->external_mac, prefix, nat->external_ip);
+        ovn_lflow_add(lflows, od, S_ROUTER_IN_IP_SRC_POLICY, 100,
                       ds_cstr(&match), ds_cstr(&actions));
         ds_clear(&match);
         ds_clear(&actions);
@@ -7547,12 +7543,6 @@  add_route(struct hmap *lflows, const struct ovn_port *op,
     }
     build_route_match(op_inport, network_s, plen, is_src_route, is_ipv4,
                       &match, &priority);
-    /* traffic for internal IPs of logical switch ports must be sent to
-     * the gw controller through the overlay tunnels
-     */
-    if (op->nbrp && !op->nbrp->n_gateway_chassis) {
-        priority += DROUTE_PRIO;
-    }
 
     struct ds actions = DS_EMPTY_INITIALIZER;
     ds_put_format(&actions, "ip.ttl--; "REG_ECMP_GROUP_ID" = 0; %sreg0 = ",
@@ -9519,9 +9509,13 @@  build_lrouter_flows(struct hmap *datapaths, struct hmap *ports,
      * logical router
      */
     HMAP_FOR_EACH (od, key_node, datapaths) {
-        if (od->nbr && od->l3dgw_port) {
-            add_distributed_routes(lflows, od);
+        if (!od->nbr) {
+            continue;
+        }
+        if (od->l3dgw_port) {
+            add_ip_src_policy_flows(lflows, od);
         }
+        ovn_lflow_add(lflows, od, S_ROUTER_IN_IP_SRC_POLICY, 0, "1", "next;");
     }
 
     /* Logical router ingress table IP_ROUTING & IP_ROUTING_ECMP: IP Routing.
diff --git a/tests/ovn.at b/tests/ovn.at
index f39fda2e4..fcc34fd5d 100644
--- a/tests/ovn.at
+++ b/tests/ovn.at
@@ -9637,20 +9637,6 @@  AT_CHECK([as hv3 ovs-vsctl set Open_vSwitch . external-ids:ovn-bridge-mappings=p
 OVS_WAIT_UNTIL([test 1 = `as hv3 ovs-vsctl show | \
 grep "Port patch-br-int-to-ln_port" | wc -l`])
 
-AT_CHECK([test 1 = `ovn-sbctl dump-flows lr0 | grep lr_in_ip_routing | \
-grep "ip4.src == 10.0.0.3 && is_chassis_resident(\"foo1\")" -c`])
-AT_CHECK([test 1 = `ovn-sbctl dump-flows lr0 | grep lr_in_ip_routing | \
-grep "ip4.src == 10.0.0.4 && is_chassis_resident(\"foo2\")" -c`])
-
-key=`ovn-sbctl --bare --columns tunnel_key list datapath_Binding lr0`
-# Check that the OVS flows appear for the dnat_and_snat entries in
-# lr_in_ip_routing table.
-OVS_WAIT_UNTIL([test 1 = `as hv3 ovs-ofctl dump-flows br-int table=17 | \
-grep "priority=400,ip,metadata=0x$key,nw_src=10.0.0.3" -c`])
-
-OVS_WAIT_UNTIL([test 1 = `as hv3 ovs-ofctl dump-flows br-int table=17 | \
-grep "priority=400,ip,metadata=0x$key,nw_src=10.0.0.4" -c`])
-
 # Re-add nat-addresses option
 ovn-nbctl lsp-set-options lrp0-rp router-port=lrp0 nat-addresses="router"
 
@@ -15141,7 +15127,7 @@  ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
 # Since the sw0-vir is not claimed by any chassis, eth.dst should be set to
 # zero if the ip4.dst is the virtual ip in the router pipeline.
 AT_CHECK([cat lflows.txt], [0], [dnl
-  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 00:00:00:00:00:00; next;)
+  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 00:00:00:00:00:00; next;)
 ])
 
 ip_to_hex() {
@@ -15192,7 +15178,7 @@  ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
 # There should be an arp resolve flow to resolve the virtual_ip with the
 # sw0-p1's MAC.
 AT_CHECK([cat lflows.txt], [0], [dnl
-  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:03; next;)
+  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:03; next;)
 ])
 
 # Forcibly clear virtual_parent. ovn-controller should release the binding
@@ -15233,7 +15219,7 @@  ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
 # There should be an arp resolve flow to resolve the virtual_ip with the
 # sw0-p2's MAC.
 AT_CHECK([cat lflows.txt], [0], [dnl
-  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:05; next;)
+  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:05; next;)
 ])
 
 # send the garp from sw0-p2 (in hv2). hv2 should claim sw0-vir
@@ -15256,7 +15242,7 @@  ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
 # There should be an arp resolve flow to resolve the virtual_ip with the
 # sw0-p3's MAC.
 AT_CHECK([cat lflows.txt], [0], [dnl
-  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:04; next;)
+  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:04; next;)
 ])
 
 # Now send arp reply from sw0-p1. hv1 should claim sw0-vir
@@ -15277,7 +15263,7 @@  ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
 > lflows.txt
 
 AT_CHECK([cat lflows.txt], [0], [dnl
-  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:03; next;)
+  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:03; next;)
 ])
 
 # Delete hv1-vif1 port. hv1 should release sw0-vir
@@ -15295,7 +15281,7 @@  ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
 > lflows.txt
 
 AT_CHECK([cat lflows.txt], [0], [dnl
-  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 00:00:00:00:00:00; next;)
+  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 00:00:00:00:00:00; next;)
 ])
 
 # Now send arp reply from sw0-p2. hv2 should claim sw0-vir
@@ -15316,7 +15302,7 @@  ovn-sbctl dump-flows lr0 | grep lr_in_arp_resolve | grep "reg0 == 10.0.0.10" \
 > lflows.txt
 
 AT_CHECK([cat lflows.txt], [0], [dnl
-  table=12(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:04; next;)
+  table=13(lr_in_arp_resolve  ), priority=100  , match=(outport == "lr0-sw0" && reg0 == 10.0.0.10), action=(eth.dst = 50:54:00:00:00:04; next;)
 ])
 
 # Delete sw0-p2 logical port