Handle IPv6 and ECMP routes to be programmed to ASIC (sonic-net#1711)

*In case of standby mux, associated routes may come to orch with ifname of tun0. Handle the case when nexthop is non-zero
*For ECMP, multiple nexthop IPs can have the same sai nexthop id (tunnel NH). Existing data structure is unable to handle such case. Added a secondary map if nexthops are shared

Author: prsunny

cfgmgr/tunnelmgr.cpp

@@ -64,21 +64,43 @@ static int cmdIpTunnelRouteAdd(const std::string& pfx, std::string & res)
     // ip route add/replace {{ip prefix}} dev {{tunnel intf}}
     // Replace route if route already exists
     ostringstream cmd;
-    cmd << IP_CMD " route replace "
-        << shellquote(pfx)
-        << " dev "
-        << TUNIF;
+    if (IpPrefix(pfx).isV4())
+    {
+        cmd << IP_CMD " route replace "
+            << shellquote(pfx)
+            << " dev "
+            << TUNIF;
+    }
+    else
+    {
+        cmd << IP_CMD " -6 route replace "
+            << shellquote(pfx)
+            << " dev "
+            << TUNIF;
+    }
+
     return swss::exec(cmd.str(), res);
 }
 
 static int cmdIpTunnelRouteDel(const std::string& pfx, std::string & res)
 {
     // ip route del {{ip prefix}} dev {{tunnel intf}}
     ostringstream cmd;
-    cmd << IP_CMD " route del "
-        << shellquote(pfx)
-        << " dev "
-        << TUNIF;
+    if (IpPrefix(pfx).isV4())
+    {
+        cmd << IP_CMD " route del "
+            << shellquote(pfx)
+            << " dev "
+            << TUNIF;
+    }
+    else
+    {
+        cmd << IP_CMD " -6 route del "
+            << shellquote(pfx)
+            << " dev "
+            << TUNIF;
+    }
+
     return swss::exec(cmd.str(), res);
 }
 

orchagent/muxorch.cpp

@@ -1278,12 +1278,6 @@ void MuxCableOrch::updateMuxState(string portName, string muxState)
 
 void MuxCableOrch::addTunnelRoute(const NextHopKey &nhKey)
 {
-    if (!nhKey.ip_address.isV4())
-    {
-        SWSS_LOG_INFO("IPv6 tunnel route add '%s' - (Not Implemented)", nhKey.ip_address.to_string().c_str());
-        return;
-    }
-
     vector<FieldValueTuple> data;
     string key, alias = nhKey.alias;
 
@@ -1299,12 +1293,6 @@ void MuxCableOrch::addTunnelRoute(const NextHopKey &nhKey)
 
 void MuxCableOrch::removeTunnelRoute(const NextHopKey &nhKey)
 {
-    if (!nhKey.ip_address.isV4())
-    {
-        SWSS_LOG_INFO("IPv6 tunnel route remove '%s' - (Not Implemented)", nhKey.ip_address.to_string().c_str());
-        return;
-    }
-
     string key, alias = nhKey.alias;
 
     IpPrefix pfx = nhKey.ip_address.to_string();

orchagent/neighorch.cpp

@@ -313,7 +313,10 @@ bool NeighOrch::isNextHopFlagSet(const NextHopKey &nexthop, const uint32_t nh_fl
 
     auto nhop = m_syncdNextHops.find(nexthop);
 
-    assert(nhop != m_syncdNextHops.end());
+    if (nhop == m_syncdNextHops.end())
+    {
+        return false;
+    }
 
     if (nhop->second.nh_flags & nh_flag)
     {

orchagent/routeorch.cpp

@@ -574,7 +574,7 @@ void RouteOrch::doTask(Consumer& consumer)
                      * way is to create loopback interface and then create
                      * route pointing to it, so that we can traps packets to
                      * CPU */
-                    if (alias == "eth0" || alias == "docker0" || alias == "tun0" ||
+                    if (alias == "eth0" || alias == "docker0" ||
                         alias == "lo" || !alias.compare(0, strlen(LOOPBACK_PREFIX), LOOPBACK_PREFIX))
                     {
                         excp_intfs_flag = true;
@@ -599,10 +599,18 @@ void RouteOrch::doTask(Consumer& consumer)
 
                 if (overlay_nh == false)
                 {
+                    if (alsv[0] == "tun0" && !(IpAddress(ipv[0]).isZero()))
+                    {
+                        alsv[0] = gIntfsOrch->getRouterIntfsAlias(ipv[0]);
+                    }
                     nhg_str = ipv[0] + NH_DELIMITER + alsv[0];
 
                     for (uint32_t i = 1; i < ipv.size(); i++)
                     {
+                        if (alsv[i] == "tun0" && !(IpAddress(ipv[i]).isZero()))
+                        {
+                            alsv[i] = gIntfsOrch->getRouterIntfsAlias(ipv[i]);
+                        }
                         nhg_str += NHG_DELIMITER + ipv[i] + NH_DELIMITER + alsv[i];
                     }
 
@@ -628,6 +636,11 @@ void RouteOrch::doTask(Consumer& consumer)
                         /* add addBlackholeRoute or addRoute support empty nhg */
                         it = consumer.m_toSync.erase(it);
                     }
+                    /* skip direct routes to tun0 */
+                    else if (alsv[0] == "tun0")
+                    {
+                        it = consumer.m_toSync.erase(it);
+                    }
                     /* directly connected route to VRF interface which come from kernel */
                     else if (!alsv[0].compare(0, strlen(VRF_PREFIX), VRF_PREFIX))
                     {
@@ -1009,6 +1022,7 @@ bool RouteOrch::addNextHopGroup(const NextHopGroupKey &nexthops)
     vector<sai_object_id_t> next_hop_ids;
     set<NextHopKey> next_hop_set = nexthops.getNextHops();
     std::map<sai_object_id_t, NextHopKey> nhopgroup_members_set;
+    std::map<sai_object_id_t, set<NextHopKey>> nhopgroup_shared_set;
 
     /* Assert each IP address exists in m_syncdNextHops table,
      * and add the corresponding next_hop_id to next_hop_ids. */
@@ -1029,7 +1043,14 @@ bool RouteOrch::addNextHopGroup(const NextHopGroupKey &nexthops)
 
         sai_object_id_t next_hop_id = m_neighOrch->getNextHopId(it);
         next_hop_ids.push_back(next_hop_id);
-        nhopgroup_members_set[next_hop_id] = it;
+        if (nhopgroup_members_set.find(next_hop_id) == nhopgroup_members_set.end())
+        {
+            nhopgroup_members_set[next_hop_id] = it;
+        }
+        else
+        {
+            nhopgroup_shared_set[next_hop_id].insert(it);
+        }
     }
 
     sai_attribute_t nhg_attr;
@@ -1103,8 +1124,20 @@ bool RouteOrch::addNextHopGroup(const NextHopGroupKey &nexthops)
         gCrmOrch->incCrmResUsedCounter(CrmResourceType::CRM_NEXTHOP_GROUP_MEMBER);
 
         // Save the membership into next hop structure
-        next_hop_group_entry.nhopgroup_members[nhopgroup_members_set.find(nhid)->second] =
-                                                                nhgm_id;
+        if (nhopgroup_shared_set.find(nhid) != nhopgroup_shared_set.end())
+        {
+            auto it = nhopgroup_shared_set[nhid].begin();
+            next_hop_group_entry.nhopgroup_members[*it] = nhgm_id;
+            nhopgroup_shared_set[nhid].erase(it);
+            if (nhopgroup_shared_set[nhid].empty())
+            {
+                nhopgroup_shared_set.erase(nhid);
+            }
+        }
+        else
+        {
+            next_hop_group_entry.nhopgroup_members[nhopgroup_members_set.find(nhid)->second] = nhgm_id;
+        }
     }
 
     /* Increment the ref_count for the next hops used by the next hop group. */
@@ -1118,7 +1151,6 @@ bool RouteOrch::addNextHopGroup(const NextHopGroupKey &nexthops)
     next_hop_group_entry.ref_count = 0;
     m_syncdNextHopGroups[nexthops] = next_hop_group_entry;
 
-
     return true;
 }
 

tests/test_mux.py

@@ -59,6 +59,7 @@ def create_vlan_interface(self, confdb, asicdb, dvs):
         fvs = {"NULL": "NULL"}
         confdb.create_entry("VLAN_INTERFACE", "Vlan1000", fvs)
         confdb.create_entry("VLAN_INTERFACE", "Vlan1000|192.168.0.1/24", fvs)
+        confdb.create_entry("VLAN_INTERFACE", "Vlan1000|fc02:1000::1/64", fvs)
 
         dvs.runcmd("config interface startup Ethernet0")
         dvs.runcmd("config interface startup Ethernet4")
@@ -334,6 +335,44 @@ def create_and_test_route(self, appdb, asicdb, dvs, dvs_route):
         self.set_mux_state(appdb, "Ethernet4", "active")
         self.check_nexthop_group_in_asic_db(asicdb, rtkeys[0])
 
+        ps._del(rtprefix)
+
+        # Test IPv6 ECMP routes and start with standby config
+        self.set_mux_state(appdb, "Ethernet0", "standby")
+        self.set_mux_state(appdb, "Ethernet4", "standby")
+
+        rtprefix = "2020::/64"
+
+        dvs_route.check_asicdb_deleted_route_entries([rtprefix])
+
+        ps = swsscommon.ProducerStateTable(pdb.db_connection, "ROUTE_TABLE")
+
+        fvs = swsscommon.FieldValuePairs([("nexthop", self.SERV1_IPV6 + "," + self.SERV2_IPV6), ("ifname", "tun0,tun0")])
+
+        ps.set(rtprefix, fvs)
+
+        # Check if route was propagated to ASIC DB
+        rtkeys = dvs_route.check_asicdb_route_entries([rtprefix])
+
+        # Check for nexthop group and validate nexthop group member in asic db
+        self.check_nexthop_group_in_asic_db(asicdb, rtkeys[0], 2)
+
+        # Step: 1 - Change one NH to active and verify ecmp route
+        self.set_mux_state(appdb, "Ethernet0", "active")
+        self.check_nexthop_group_in_asic_db(asicdb, rtkeys[0], 1)
+
+        # Step: 2 - Change the other NH to active and verify ecmp route
+        self.set_mux_state(appdb, "Ethernet4", "active")
+        self.check_nexthop_group_in_asic_db(asicdb, rtkeys[0])
+
+        # Step: 3 - Change one NH to back to standby and verify ecmp route
+        self.set_mux_state(appdb, "Ethernet0", "standby")
+        self.check_nexthop_group_in_asic_db(asicdb, rtkeys[0], 1)
+
+        # Step: 4 - Change the other NH to standby and verify ecmp route
+        self.set_mux_state(appdb, "Ethernet4", "standby")
+        self.check_nexthop_group_in_asic_db(asicdb, rtkeys[0], 2)
+
 
     def get_expected_sai_qualifiers(self, portlist, dvs_acl):
         expected_sai_qualifiers = {