MFIB: recurse resolution through an MFIB entry

Change-Id: I8dc261e40b8398c5c8ab6bb69ecebbd0176055d9
Signed-off-by: Neale Ranns <nranns@cisco.com>

24 files changed, 2097 insertions, 199 deletions

diff --git a/src/plugins/unittest/mfib_test.c b/src/plugins/unittest/mfib_test.c
index bbe6709f4d7..0707a6f57c8 100644
--- a/src/plugins/unittest/mfib_test.c
+++ b/src/plugins/unittest/mfib_test.c
@@ -34,9 +34,6 @@
         fformat(stderr, "FAIL:%d: " _comment "\n",		\
                 __LINE__, ##_args);				\
         res = 1;                                                \
-    } else {							\
-        fformat(stderr, "PASS:%d: " _comment "\n",		\
-                __LINE__, ##_args);				\
     }								\
     res;							\
 })
@@ -223,16 +220,16 @@ mfib_test_entry (fib_node_index_t fei,
                  int n_buckets,
                  ...)
 {
+    const mfib_prefix_t *pfx;
     const mfib_entry_t *mfe;
     const replicate_t *rep;
-    mfib_prefix_t pfx;
     va_list ap;
     int res;
 
 
     res = 0;
     mfe = mfib_entry_get(fei);
-    mfib_entry_get_prefix(fei, &pfx);
+    pfx = mfib_entry_get_prefix(fei);
 
     MFIB_TEST_REP((eflags == mfe->mfe_flags),
                   "%U has %U expect %U",
@@ -253,7 +250,8 @@ mfib_test_entry (fib_node_index_t fei,
 
         mfib_entry_contribute_forwarding(
             fei,
-            fib_forw_chain_type_from_fib_proto(pfx.fp_proto),
+            mfib_forw_chain_type_from_fib_proto(pfx->fp_proto),
+            MFIB_ENTRY_FWD_FLAG_NONE,
             &tmp);
         rep = replicate_get(tmp.dpoi_index);
 
@@ -278,15 +276,15 @@ mfib_test_entry_itf (fib_node_index_t fei,
                      u32 sw_if_index,
                      mfib_itf_flags_t flags)
 {
+    const mfib_prefix_t *pfx;
     const mfib_entry_t *mfe;
     const mfib_itf_t *mfi;
-    mfib_prefix_t pfx;
     int res;
 
     res = 0;
     mfe = mfib_entry_get(fei);
     mfi = mfib_entry_get_itf(mfe, sw_if_index);
-    mfib_entry_get_prefix(fei, &pfx);
+    pfx = mfib_entry_get_prefix(fei);
 
     MFIB_TEST_REP((NULL != mfi),
                   "%U has interface %d",
@@ -305,19 +303,19 @@ static int
 mfib_test_entry_no_itf (fib_node_index_t fei,
                         u32 sw_if_index)
 {
+    const mfib_prefix_t *pfx;
     const mfib_entry_t *mfe;
     const mfib_itf_t *mfi;
-    mfib_prefix_t pfx;
     int res;
 
     res = 0;
     mfe = mfib_entry_get(fei);
     mfi = mfib_entry_get_itf(mfe, sw_if_index);
-    mfib_entry_get_prefix(fei, &pfx);
+    pfx = mfib_entry_get_prefix(fei);
 
     MFIB_TEST_REP((NULL == mfi),
                   "%U has no interface %d",
-                  format_mfib_prefix, &pfx, sw_if_index);
+                  format_mfib_prefix, pfx, sw_if_index);
 
     return (res);
 }
@@ -1143,11 +1141,12 @@ mfib_test_i (fib_protocol_t PROTO,
     dpo_set(&td, DPO_ADJACENCY_MCAST, fib_proto_to_dpo(PROTO), ai_1);
     replicate_set_bucket(repi2, 0, &td);
 
-    mfei = mfib_table_entry_special_add(fib_index,
-                                        pfx_star_g_3,
-                                        MFIB_SOURCE_SRv6,
-                                        MFIB_ENTRY_FLAG_ACCEPT_ALL_ITF,
-                                        repi2);
+    mfib_entry_update(mfei,
+                      MFIB_SOURCE_SRv6,
+                      (MFIB_ENTRY_FLAG_ACCEPT_ALL_ITF |
+                       MFIB_ENTRY_FLAG_EXCLUSIVE),
+                      MFIB_RPF_ID_NONE,
+                      repi2);
     MFIB_TEST(!mfib_test_entry(mfei,
                                (MFIB_ENTRY_FLAG_ACCEPT_ALL_ITF |
                                 MFIB_ENTRY_FLAG_EXCLUSIVE),
@@ -1278,6 +1277,11 @@ mfib_test_i (fib_protocol_t PROTO,
     /*
      * Unlock the table - it's the last lock so should be gone thereafter
      */
+    MFIB_TEST(((PROTO == FIB_PROTOCOL_IP4 ? 1 : 5) ==
+               mfib_table_get_n_routes(fib_index, PROTO)),
+              "1 = %d route left in the FIB",
+              mfib_table_get_n_routes(fib_index, PROTO));
+
     mfib_table_unlock(fib_index, PROTO, MFIB_SOURCE_API);
 
     MFIB_TEST((FIB_NODE_INDEX_INVALID ==
@@ -1494,6 +1498,435 @@ mfib_test_v6 (void)
                         &nbr2));
 }
 
+static int
+mfib_test_rr_i (fib_protocol_t FPROTO,
+                dpo_proto_t DPROTO,
+                vnet_link_t LINKT,
+                const mfib_prefix_t *pfx_cover,
+                const mfib_prefix_t *pfx_host1,
+                const mfib_prefix_t *pfx_host2)
+{
+    fib_node_index_t mfei_cover, mfei_host1, mfei_host2, ai_1, ai_2;
+    u32 fib_index, n_entries, n_itfs, n_reps, n_pls;
+    test_main_t *tm;
+    int res;
+
+    res = 0;
+    n_entries = pool_elts(mfib_entry_pool);
+    n_itfs = pool_elts(mfib_itf_pool);
+    n_reps = pool_elts(replicate_pool);
+    n_pls = fib_path_list_pool_size();
+    tm = &test_main;
+
+    fib_index = 0;
+    ai_1 = adj_mcast_add_or_lock(FPROTO,
+                                 LINKT,
+                                 tm->hw[1]->sw_if_index);
+    ai_2 = adj_mcast_add_or_lock(FPROTO,
+                                 LINKT,
+                                 tm->hw[2]->sw_if_index);
+
+    fib_route_path_t path_via_if0 = {
+        .frp_proto = DPROTO,
+        .frp_addr = zero_addr,
+        .frp_sw_if_index = tm->hw[0]->sw_if_index,
+        .frp_fib_index = ~0,
+        .frp_weight = 0,
+        .frp_flags = 0,
+    };
+    fib_route_path_t path_via_if1 = {
+        .frp_proto = DPROTO,
+        .frp_addr = zero_addr,
+        .frp_sw_if_index = tm->hw[1]->sw_if_index,
+        .frp_fib_index = ~0,
+        .frp_weight = 0,
+        .frp_flags = 0,
+    };
+    fib_route_path_t path_via_if2 = {
+        .frp_proto = DPROTO,
+        .frp_addr = zero_addr,
+        .frp_sw_if_index = tm->hw[2]->sw_if_index,
+        .frp_fib_index = ~0,
+        .frp_weight = 0,
+        .frp_flags = 0,
+    };
+    fib_route_path_t path_for_us = {
+        .frp_proto = DPROTO,
+        .frp_addr = zero_addr,
+        .frp_sw_if_index = 0xffffffff,
+        .frp_fib_index = ~0,
+        .frp_weight = 0,
+        .frp_flags = FIB_ROUTE_PATH_LOCAL,
+    };
+
+    /*
+     * with only the default in place, recusre thru the /32
+     */
+    mfei_host1 = mfib_table_entry_special_add(fib_index, pfx_host1,
+                                              MFIB_SOURCE_RR,
+                                              MFIB_ENTRY_FLAG_NONE,
+                                              INDEX_INVALID);
+    /*
+     * expect its forwarding to match the cover's
+     */
+    MFIB_TEST(!mfib_test_entry(mfei_host1,
+                               MFIB_ENTRY_FLAG_DROP,
+                               0),
+              "%U no replications OK",
+              format_mfib_prefix, pfx_host1);
+
+    /*
+     * Insert the less specific /28
+     */
+    mfib_table_entry_path_update(fib_index,
+                                 pfx_cover,
+                                 MFIB_SOURCE_API,
+                                 &path_via_if1,
+                                 MFIB_ITF_FLAG_FORWARD);
+
+    mfei_cover = mfib_table_lookup_exact_match(fib_index, pfx_cover);
+
+    MFIB_TEST(!mfib_test_entry(mfei_cover,
+                               MFIB_ENTRY_FLAG_NONE,
+                               1,
+                               DPO_ADJACENCY_MCAST, ai_1),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_cover);
+
+    /*
+     * expect the /32 forwarding to match the new cover's
+     */
+    MFIB_TEST(!mfib_test_entry(mfei_host1,
+                               MFIB_ENTRY_FLAG_NONE,
+                               1,
+                               DPO_ADJACENCY_MCAST, ai_1),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_host1);
+
+    /*
+     * add another path to the cover
+     */
+    mfib_table_entry_path_update(fib_index,
+                                 pfx_cover,
+                                 MFIB_SOURCE_API,
+                                 &path_via_if2,
+                                 MFIB_ITF_FLAG_FORWARD);
+
+    /*
+     * expect the /32 and /28 to be via both boths
+     */
+    MFIB_TEST(!mfib_test_entry(mfei_cover,
+                               MFIB_ENTRY_FLAG_NONE,
+                               2,
+                               DPO_ADJACENCY_MCAST, ai_1,
+                               DPO_ADJACENCY_MCAST, ai_2),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_cover);
+    MFIB_TEST(!mfib_test_entry(mfei_host1,
+                               MFIB_ENTRY_FLAG_NONE,
+                               2,
+                               DPO_ADJACENCY_MCAST, ai_1,
+                               DPO_ADJACENCY_MCAST, ai_2),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_host1);
+
+    /*
+     * and the other host whilst all is ready
+     */
+    mfei_host2 = mfib_table_entry_special_add(fib_index, pfx_host2,
+                                              MFIB_SOURCE_RR,
+                                              MFIB_ENTRY_FLAG_NONE,
+                                              INDEX_INVALID);
+    MFIB_TEST(!mfib_test_entry(mfei_host2,
+                               MFIB_ENTRY_FLAG_NONE,
+                               2,
+                               DPO_ADJACENCY_MCAST, ai_1,
+                               DPO_ADJACENCY_MCAST, ai_2),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_host2);
+
+    /*
+     * repaet multiple time to simulate multiple recursve children
+     */
+    mfei_host2 = mfib_table_entry_special_add(fib_index, pfx_host2,
+                                              MFIB_SOURCE_RR,
+                                              MFIB_ENTRY_FLAG_NONE,
+                                              INDEX_INVALID);
+    mfei_host2 = mfib_table_entry_special_add(fib_index, pfx_host2,
+                                              MFIB_SOURCE_RR,
+                                              MFIB_ENTRY_FLAG_NONE,
+                                              INDEX_INVALID);
+    mfei_host2 = mfib_table_entry_special_add(fib_index, pfx_host2,
+                                              MFIB_SOURCE_RR,
+                                              MFIB_ENTRY_FLAG_NONE,
+                                              INDEX_INVALID);
+
+    /*
+     * add an accepting path to the cover
+     */
+    mfib_table_entry_path_update(fib_index,
+                                 pfx_cover,
+                                 MFIB_SOURCE_API,
+                                 &path_via_if0,
+                                 MFIB_ITF_FLAG_ACCEPT);
+
+    /*
+     * expect the /32 and /28 to be via both boths
+     */
+    MFIB_TEST(!mfib_test_entry(mfei_cover,
+                               MFIB_ENTRY_FLAG_NONE,
+                               2,
+                               DPO_ADJACENCY_MCAST, ai_1,
+                               DPO_ADJACENCY_MCAST, ai_2),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_cover);
+    MFIB_TEST(!mfib_test_entry(mfei_host1,
+                               MFIB_ENTRY_FLAG_NONE,
+                               2,
+                               DPO_ADJACENCY_MCAST, ai_1,
+                               DPO_ADJACENCY_MCAST, ai_2),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_cover);
+    MFIB_TEST_NS(!mfib_test_entry_itf(mfei_host1, tm->hw[0]->sw_if_index,
+                                      MFIB_ITF_FLAG_ACCEPT));
+    MFIB_TEST_NS(!mfib_test_entry_itf(mfei_cover, tm->hw[0]->sw_if_index,
+                                      MFIB_ITF_FLAG_ACCEPT));
+    MFIB_TEST_NS(!mfib_test_entry_itf(mfei_host1, tm->hw[1]->sw_if_index,
+                                      MFIB_ITF_FLAG_FORWARD));
+    MFIB_TEST_NS(!mfib_test_entry_itf(mfei_cover, tm->hw[1]->sw_if_index,
+                                      MFIB_ITF_FLAG_FORWARD));
+    MFIB_TEST_NS(!mfib_test_entry_itf(mfei_host1, tm->hw[2]->sw_if_index,
+                                      MFIB_ITF_FLAG_FORWARD));
+    MFIB_TEST_NS(!mfib_test_entry_itf(mfei_cover, tm->hw[2]->sw_if_index,
+                                      MFIB_ITF_FLAG_FORWARD));
+
+    /*
+     * add a for-us path to the cover
+     */
+    mfib_table_entry_path_update(fib_index,
+                                 pfx_cover,
+                                 MFIB_SOURCE_API,
+                                 &path_for_us,
+                                 MFIB_ITF_FLAG_FORWARD);
+
+    /*
+     * expect the /32 and /28 to be via all three paths
+     */
+    MFIB_TEST(!mfib_test_entry(mfei_cover,
+                               MFIB_ENTRY_FLAG_NONE,
+                               3,
+                               DPO_ADJACENCY_MCAST, ai_1,
+                               DPO_ADJACENCY_MCAST, ai_2,
+                               DPO_RECEIVE, 0),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_cover);
+    MFIB_TEST(!mfib_test_entry(mfei_host1,
+                               MFIB_ENTRY_FLAG_NONE,
+                               3,
+                               DPO_ADJACENCY_MCAST, ai_1,
+                               DPO_ADJACENCY_MCAST, ai_2,
+                               DPO_RECEIVE, 0),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_cover);
+
+    /*
+     * get the forwarding chain from the RR prefix
+     */
+    replicate_t *rep;
+    dpo_id_t dpo = DPO_INVALID;
+
+    mfib_entry_contribute_forwarding(
+        mfei_host1,
+        mfib_forw_chain_type_from_dpo_proto(DPROTO),
+        MFIB_ENTRY_FWD_FLAG_NONE,
+        &dpo);
+
+    rep = replicate_get(dpo.dpoi_index);
+    MFIB_TEST((3 == rep->rep_n_buckets),
+              "%U replicate 3 buckets",
+              format_mfib_prefix, pfx_host1);
+
+    /*
+     * get the forwarding chain from the RR prefix without local paths
+     */
+    mfib_entry_contribute_forwarding(
+        mfei_host1,
+        mfib_forw_chain_type_from_dpo_proto(DPROTO),
+        MFIB_ENTRY_FWD_FLAG_NO_LOCAL,
+        &dpo);
+
+    rep = replicate_get(dpo.dpoi_index);
+    MFIB_TEST((2 == rep->rep_n_buckets),
+              "%U no-local replicate 2 buckets",
+              format_mfib_prefix, pfx_host1);
+
+    dpo_reset(&dpo);
+
+    /*
+     * delete the cover, expect the /32 to be via the default
+     */
+    mfib_table_entry_delete(fib_index, pfx_cover, MFIB_SOURCE_API);
+    MFIB_TEST(!mfib_test_entry(mfei_host1,
+                               MFIB_ENTRY_FLAG_DROP,
+                               0),
+              "%U no replications OK",
+              format_mfib_prefix, pfx_host1);
+
+    /*
+     * source the /32 with its own path
+     */
+    mfib_table_entry_path_update(fib_index,
+                                 pfx_host1,
+                                 MFIB_SOURCE_API,
+                                 &path_via_if2,
+                                 MFIB_ITF_FLAG_FORWARD);
+    MFIB_TEST(!mfib_test_entry(mfei_host1,
+                               MFIB_ENTRY_FLAG_NONE,
+                               1,
+                               DPO_ADJACENCY_MCAST, ai_2),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_host1);
+
+    /*
+     * remove host2 - as many times as it was added
+     */
+    mfib_table_entry_delete(fib_index, pfx_host2,
+                            MFIB_SOURCE_RR);
+    mfib_table_entry_delete(fib_index, pfx_host2,
+                            MFIB_SOURCE_RR);
+    mfib_table_entry_delete(fib_index, pfx_host2,
+                            MFIB_SOURCE_RR);
+    mfib_table_entry_delete(fib_index, pfx_host2,
+                            MFIB_SOURCE_RR);
+
+
+    /*
+     * remove the RR source with paths present
+     */
+    mfib_table_entry_delete(fib_index, pfx_host1,
+                            MFIB_SOURCE_RR);
+
+    /*
+     * add the RR back then remove the path and RR
+     */
+    mfib_table_entry_path_update(fib_index,
+                                 pfx_host1,
+                                 MFIB_SOURCE_API,
+                                 &path_via_if2,
+                                 MFIB_ITF_FLAG_FORWARD);
+    MFIB_TEST(!mfib_test_entry(mfei_host1,
+                               MFIB_ENTRY_FLAG_NONE,
+                               1,
+                               DPO_ADJACENCY_MCAST, ai_2),
+              "%U replicate OK",
+              format_mfib_prefix, pfx_cover);
+
+    mfib_table_entry_delete(fib_index, pfx_host1,
+                            MFIB_SOURCE_API);
+    mfib_table_entry_delete(fib_index, pfx_host1,
+                            MFIB_SOURCE_RR);
+
+    /*
+     * test we've leaked no resources
+     */
+    adj_unlock(ai_1);
+    adj_unlock(ai_2);
+    MFIB_TEST(0 == adj_mcast_db_size(), "%d MCAST adjs", adj_mcast_db_size());
+    MFIB_TEST(n_pls == fib_path_list_pool_size(), "%d=%d path-lists",
+              n_pls, fib_path_list_pool_size());
+    MFIB_TEST(n_reps == pool_elts(replicate_pool), "%d=%d replicates",
+              n_reps, pool_elts(replicate_pool));
+    MFIB_TEST(n_entries == pool_elts(mfib_entry_pool),
+              " No more entries %d!=%d",
+              n_entries, pool_elts(mfib_entry_pool));
+    MFIB_TEST(n_itfs == pool_elts(mfib_itf_pool),
+              " No more Interfaces %d!=%d",
+              n_itfs, pool_elts(mfib_itf_pool));
+    return (res);
+}
+
+static int
+mfib_test_rr_v4 (void)
+{
+    /*
+     * 2 length of prefix to play with
+     */
+    const mfib_prefix_t pfx_host1 = {
+        .fp_len = 32,
+        .fp_proto = FIB_PROTOCOL_IP4,
+        .fp_grp_addr = {
+            .ip4.as_u32 = clib_host_to_net_u32(0xe0001011),
+        },
+    };
+    const mfib_prefix_t pfx_host2 = {
+        .fp_len = 64,
+        .fp_proto = FIB_PROTOCOL_IP4,
+        .fp_grp_addr = {
+            .ip4.as_u32 = clib_host_to_net_u32(0xe0001011),
+        },
+        .fp_src_addr = {
+            .ip4.as_u32 = clib_host_to_net_u32(0x10101010),
+        },
+    };
+    const mfib_prefix_t pfx_cover = {
+        .fp_len = 28,
+        .fp_proto = FIB_PROTOCOL_IP4,
+        .fp_grp_addr = {
+            .ip4.as_u32 = clib_host_to_net_u32(0xe0001010),
+        },
+    };
+
+    return (mfib_test_rr_i(FIB_PROTOCOL_IP4,
+                           DPO_PROTO_IP4,
+                           VNET_LINK_IP4,
+                           &pfx_cover,
+                           &pfx_host1,
+                           &pfx_host2));
+}
+
+static int
+mfib_test_rr_v6 (void)
+{
+    /*
+     * 2 length of prefix to play with
+     */
+    const mfib_prefix_t pfx_host1 = {
+        .fp_len = 128,
+        .fp_proto = FIB_PROTOCOL_IP6,
+        .fp_grp_addr = {
+            .ip6.as_u64[0] = clib_host_to_net_u64(0xff03000000000000),
+            .ip6.as_u64[1] = clib_host_to_net_u64(0x0000000000000001),
+        },
+    };
+    const mfib_prefix_t pfx_host2 = {
+        .fp_len = 256,
+        .fp_proto = FIB_PROTOCOL_IP6,
+        .fp_grp_addr = {
+            .ip6.as_u64[0] = clib_host_to_net_u64(0xff03000000000000),
+            .ip6.as_u64[1] = clib_host_to_net_u64(0x0000000000000001),
+        },
+        .fp_src_addr = {
+            .ip6.as_u64[0] = clib_host_to_net_u64(0x2001000000000000),
+            .ip6.as_u64[1] = clib_host_to_net_u64(0x0000000000000001),
+        },
+    };
+    const mfib_prefix_t pfx_cover = {
+        .fp_len = 64,
+        .fp_proto = FIB_PROTOCOL_IP6,
+        .fp_grp_addr = {
+            .ip6.as_u64[0] = clib_host_to_net_u64(0xff03000000000000),
+            .ip6.as_u64[1] = clib_host_to_net_u64(0x0000000000000000),
+        },
+    };
+
+    return (mfib_test_rr_i(FIB_PROTOCOL_IP6,
+                           DPO_PROTO_IP6,
+                           VNET_LINK_IP6,
+                           &pfx_cover,
+                           &pfx_host1,
+                           &pfx_host2));
+}
+
 static clib_error_t *
 mfib_test (vlib_main_t * vm,
            unformat_input_t * input,
@@ -1502,23 +1935,35 @@ mfib_test (vlib_main_t * vm,
     int res = 0;
 
     res += mfib_test_mk_intf(4);
-    res += mfib_test_v4();
+    res += mfib_test_rr_v4();
 
     if (res)
     {
-        return clib_error_return(0, "MFIB Unit Test Failed");
+        return clib_error_return(0, "MFIB RR V4 Unit Test Failed");
     }
 
-    res += mfib_test_v6();
+    res += mfib_test_rr_v6();
 
     if (res)
     {
-        return clib_error_return(0, "MFIB Unit Test Failed");
+        return clib_error_return(0, "MFIB RR V6 Unit Test Failed");
     }
-    else
+
+    res += mfib_test_v4();
+
+    if (res)
     {
-        return (NULL);
+        return clib_error_return(0, "MFIB V4 Unit Test Failed");
     }
+
+    res += mfib_test_v6();
+
+    if (res)
+    {
+        return clib_error_return(0, "MFIB V6 Unit Test Failed");
+    }
+
+    return (NULL);
 }
 
 VLIB_CLI_COMMAND (test_fib_command, static) = {
diff --git a/src/vnet/CMakeLists.txt b/src/vnet/CMakeLists.txt
index 62bd57582b4..353bd3f7a6d 100644
--- a/src/vnet/CMakeLists.txt
+++ b/src/vnet/CMakeLists.txt
@@ -1279,6 +1279,11 @@ list(APPEND VNET_SOURCES
   mfib/mfib_signal.c
   mfib/mfib_itf.c
   mfib/mfib_entry.c
+  mfib/mfib_entry.c
+  mfib/mfib_entry_cover.c
+  mfib/mfib_entry_delegate.c
+  mfib/mfib_entry_src.c
+  mfib/mfib_entry_src_rr.c
   mfib/mfib_table.c
 )
 
diff --git a/src/vnet/dpo/receive_dpo.c b/src/vnet/dpo/receive_dpo.c
index 30ddceeb23e..949dbfa6587 100644
--- a/src/vnet/dpo/receive_dpo.c
+++ b/src/vnet/dpo/receive_dpo.c
@@ -25,6 +25,12 @@
  */
 receive_dpo_t *receive_dpo_pool;
 
+int
+dpo_is_receive (const dpo_id_t *dpo)
+{
+    return (dpo->dpoi_type == DPO_RECEIVE);
+}
+
 static receive_dpo_t *
 receive_dpo_alloc (void)
 {
diff --git a/src/vnet/dpo/receive_dpo.h b/src/vnet/dpo/receive_dpo.h
index 9ab86a824c6..ee3c7836aff 100644
--- a/src/vnet/dpo/receive_dpo.h
+++ b/src/vnet/dpo/receive_dpo.h
@@ -47,6 +47,8 @@ typedef struct receive_dpo_t_
     u16 rd_locks;
 } receive_dpo_t;
 
+extern int dpo_is_receive(const dpo_id_t *dpo);
+
 extern void receive_dpo_add_or_lock (dpo_proto_t proto,
                                      u32 sw_if_index,
                                      const ip46_address_t *nh_addr,
diff --git a/src/vnet/dpo/replicate_dpo.c b/src/vnet/dpo/replicate_dpo.c
index 39b17f103c9..69e367aa368 100644
--- a/src/vnet/dpo/replicate_dpo.c
+++ b/src/vnet/dpo/replicate_dpo.c
@@ -16,6 +16,7 @@
 #include <vnet/ip/lookup.h>
 #include <vnet/dpo/replicate_dpo.h>
 #include <vnet/dpo/drop_dpo.h>
+#include <vnet/dpo/receive_dpo.h>
 #include <vnet/adj/adj.h>
 #include <vnet/mpls/mpls_types.h>
 
@@ -101,6 +102,23 @@ replicate_alloc_i (void)
 }
 
 static u8*
+format_replicate_flags (u8 *s, va_list *args)
+{
+    int flags = va_arg (*args, int);
+
+    if (flags == REPLICATE_FLAGS_NONE)
+    {
+        s = format (s, "none");
+    }
+    else if (flags & REPLICATE_FLAGS_HAS_LOCAL)
+    {
+        s = format (s, "has-local ");
+    }
+
+    return (s);
+}
+
+static u8*
 replicate_format (index_t repi,
                   replicate_format_flags_t flags,
                   u32 indent,
@@ -118,6 +136,7 @@ replicate_format (index_t repi,
 
     s = format(s, "%U: ", format_dpo_type, DPO_REPLICATE);
     s = format(s, "[index:%d buckets:%d ", repi, rep->rep_n_buckets);
+    s = format(s, "flags:[%U] ", format_replicate_flags, rep->rep_flags);
     s = format(s, "to:[%Ld:%Ld]]", to.packets, to.bytes);
 
     for (i = 0; i < rep->rep_n_buckets; i++)
@@ -182,6 +201,14 @@ replicate_set_bucket_i (replicate_t *rep,
                         dpo_id_t *buckets,
                         const dpo_id_t *next)
 {
+    if (dpo_is_receive(&buckets[bucket]))
+    {
+        rep->rep_flags &= ~REPLICATE_FLAGS_HAS_LOCAL;
+    }
+    if (dpo_is_receive(next))
+    {
+        rep->rep_flags |= REPLICATE_FLAGS_HAS_LOCAL;
+    }
     dpo_stack(DPO_REPLICATE, rep->rep_proto, &buckets[bucket], next);
 }
 
@@ -492,6 +519,67 @@ replicate_lock (dpo_id_t *dpo)
     rep->rep_locks++;
 }
 
+index_t
+replicate_dup (replicate_flags_t flags,
+               index_t repi)
+{
+    replicate_t *rep, *copy;
+
+    rep = replicate_get(repi);
+
+    if (rep->rep_flags == flags ||
+        flags & REPLICATE_FLAGS_HAS_LOCAL)
+    {
+        /*
+         * we can include all the buckets from the original in the copy
+         */
+        return (repi);
+    }
+    else
+    {
+        /*
+         * caller doesn't want the local paths that the original has
+         */
+        if (rep->rep_n_buckets == 1)
+        {
+            /*
+             * original has only one bucket that is the local, so create
+             * a new one with only the drop
+             */
+            copy = replicate_create_i (1, rep->rep_proto);
+
+            replicate_set_bucket_i(copy, 0,
+                                   replicate_get_buckets(copy),
+                                   drop_dpo_get(rep->rep_proto));
+        }
+        else
+        {
+            dpo_id_t *old_buckets, *copy_buckets;
+            u16 bucket, pos;
+
+            copy = replicate_create_i(rep->rep_n_buckets - 1,
+                                      rep->rep_proto);
+
+            rep = replicate_get(repi);
+            old_buckets = replicate_get_buckets(rep);
+            copy_buckets = replicate_get_buckets(copy);
+            pos = 0;
+
+            for (bucket = 0; bucket < rep->rep_n_buckets; bucket++)
+            {
+                if (!dpo_is_receive(&old_buckets[bucket]))
+                {
+                    replicate_set_bucket_i(copy, pos, copy_buckets,
+                                           (&old_buckets[bucket]));
+                    pos++;
+                }
+            }
+        }
+    }
+
+    return (replicate_get_index(copy));
+}
+
 static void
 replicate_destroy (replicate_t *rep)
 {
diff --git a/src/vnet/dpo/replicate_dpo.h b/src/vnet/dpo/replicate_dpo.h
index 7b075394bc3..70faef3e28f 100644
--- a/src/vnet/dpo/replicate_dpo.h
+++ b/src/vnet/dpo/replicate_dpo.h
@@ -47,6 +47,15 @@ extern replicate_main_t replicate_main;
 #define REP_NUM_INLINE_BUCKETS 4
 
 /**
+ * Flags on the replicate DPO
+ */
+typedef enum replicate_flags_t_
+{
+    REPLICATE_FLAGS_NONE,
+    REPLICATE_FLAGS_HAS_LOCAL,
+} __clib_packed replicate_flags_t;
+
+/**
  * The FIB DPO provieds;
  *  - load-balancing over the next DPOs in the chain/graph
  *  - per-route counters
@@ -71,6 +80,11 @@ typedef struct replicate_t_ {
     dpo_proto_t rep_proto;
 
     /**
+     * Flags specifying the replicate properties/behaviour
+     */
+    replicate_flags_t rep_flags;
+
+    /**
      * The number of locks, which is approximately the number of users,
      * of this load-balance.
      * Load-balance objects of via-entries are heavily shared by recursives,
@@ -121,6 +135,9 @@ extern int replicate_is_drop(const dpo_id_t *dpo);
 
 extern u16 replicate_n_buckets(index_t repi);
 
+extern index_t replicate_dup(replicate_flags_t flags,
+                             index_t repi);
+
 /**
  * The encapsulation breakages are for fast DP access
  */
diff --git a/src/vnet/ip/ip_api.c b/src/vnet/ip/ip_api.c
index b7ee17cfca8..5a2aa60b057 100644
--- a/src/vnet/ip/ip_api.c
+++ b/src/vnet/ip/ip_api.c
@@ -408,13 +408,13 @@ send_ip_mfib_details (vl_api_registration_t * reg,
 {
   fib_route_path_encode_t *api_rpath, *api_rpaths = NULL;
   vl_api_ip_mfib_details_t *mp;
+  const mfib_prefix_t *pfx;
   mfib_entry_t *mfib_entry;
   vl_api_fib_path_t *fp;
-  mfib_prefix_t pfx;
   int path_count;
 
   mfib_entry = mfib_entry_get (mfei);
-  mfib_entry_get_prefix (mfei, &pfx);
+  pfx = mfib_entry_get_prefix (mfei);
   mfib_entry_encode (mfei, &api_rpaths);
 
   path_count = vec_len (api_rpaths);
@@ -428,11 +428,11 @@ send_ip_mfib_details (vl_api_registration_t * reg,
   mp->rpf_id = mfib_entry->mfe_rpf_id;
   mp->entry_flags = mfib_entry->mfe_flags;
   mp->table_id = htonl (table_id);
-  mp->address_length = pfx.fp_len;
-  memcpy (mp->grp_address, &pfx.fp_grp_addr.ip4,
-	  sizeof (pfx.fp_grp_addr.ip4));
-  memcpy (mp->src_address, &pfx.fp_src_addr.ip4,
-	  sizeof (pfx.fp_src_addr.ip4));
+  mp->address_length = pfx->fp_len;
+  memcpy (mp->grp_address, &pfx->fp_grp_addr.ip4,
+	  sizeof (pfx->fp_grp_addr.ip4));
+  memcpy (mp->src_address, &pfx->fp_src_addr.ip4,
+	  sizeof (pfx->fp_src_addr.ip4));
 
   mp->count = htonl (path_count);
   fp = mp->path;
@@ -503,7 +503,7 @@ static void
 send_ip6_mfib_details (vpe_api_main_t * am,
 		       vl_api_registration_t * reg,
 		       u32 table_id,
-		       mfib_prefix_t * pfx,
+		       const mfib_prefix_t * pfx,
 		       fib_route_path_encode_t * api_rpaths, u32 context)
 {
   vl_api_ip6_mfib_details_t *mp;
@@ -559,8 +559,8 @@ vl_api_ip6_mfib_dump_t_handler (vl_api_ip6_mfib_dump_t * mp)
   vl_api_registration_t *reg;
   ip6_main_t *im = &ip6_main;
   mfib_table_t *mfib_table;
+  const mfib_prefix_t *pfx;
   fib_node_index_t *mfeip;
-  mfib_prefix_t pfx;
   fib_route_path_encode_t *api_rpaths = NULL;
   vl_api_ip6_mfib_dump_ctc_t ctx = {
     .entries = NULL,
@@ -582,11 +582,11 @@ vl_api_ip6_mfib_dump_t_handler (vl_api_ip6_mfib_dump_t * mp)
 
     vec_foreach(mfeip, ctx.entries)
     {
-      mfib_entry_get_prefix (*mfeip, &pfx);
+      pfx = mfib_entry_get_prefix (*mfeip);
       mfib_entry_encode (*mfeip, &api_rpaths);
       send_ip6_mfib_details (am, reg,
                              mfib_table->mft_table_id,
-                             &pfx, api_rpaths,
+                             pfx, api_rpaths,
                              mp->context);
     }
     vec_reset_length (api_rpaths);
@@ -1833,7 +1833,7 @@ vl_mfib_signal_send_one (vl_api_registration_t * reg,
 			 u32 context, const mfib_signal_t * mfs)
 {
   vl_api_mfib_signal_details_t *mp;
-  mfib_prefix_t prefix;
+  const mfib_prefix_t *prefix;
   mfib_table_t *mfib;
   mfib_itf_t *mfi;
 
@@ -1844,25 +1844,25 @@ vl_mfib_signal_send_one (vl_api_registration_t * reg,
   mp->context = context;
 
   mfi = mfib_itf_get (mfs->mfs_itf);
-  mfib_entry_get_prefix (mfs->mfs_entry, &prefix);
+  prefix = mfib_entry_get_prefix (mfs->mfs_entry);
   mfib = mfib_table_get (mfib_entry_get_fib_index (mfs->mfs_entry),
-			 prefix.fp_proto);
+			 prefix->fp_proto);
   mp->table_id = ntohl (mfib->mft_table_id);
   mp->sw_if_index = ntohl (mfi->mfi_sw_if_index);
 
-  if (FIB_PROTOCOL_IP4 == prefix.fp_proto)
+  if (FIB_PROTOCOL_IP4 == prefix->fp_proto)
     {
-      mp->grp_address_len = ntohs (prefix.fp_len);
+      mp->grp_address_len = ntohs (prefix->fp_len);
 
-      memcpy (mp->grp_address, &prefix.fp_grp_addr.ip4, 4);
-      if (prefix.fp_len > 32)
+      memcpy (mp->grp_address, &prefix->fp_grp_addr.ip4, 4);
+      if (prefix->fp_len > 32)
 	{
-	  memcpy (mp->src_address, &prefix.fp_src_addr.ip4, 4);
+	  memcpy (mp->src_address, &prefix->fp_src_addr.ip4, 4);
 	}
     }
   else
     {
-      mp->grp_address_len = ntohs (prefix.fp_len);
+      mp->grp_address_len = ntohs (prefix->fp_len);
 
       ASSERT (0);
     }
diff --git a/src/vnet/mfib/ip4_mfib.c b/src/vnet/mfib/ip4_mfib.c
index 9d70f0b664f..d20d1f6c2b9 100644
--- a/src/vnet/mfib/ip4_mfib.c
+++ b/src/vnet/mfib/ip4_mfib.c
@@ -224,6 +224,38 @@ ip4_mfib_table_lookup (const ip4_mfib_t *mfib,
     return (FIB_NODE_INDEX_INVALID);
 }
 
+fib_node_index_t
+ip4_mfib_table_get_less_specific (const ip4_mfib_t *mfib,
+                                  const ip4_address_t *src,
+                                  const ip4_address_t *grp,
+                                  u32 len)
+{
+    u32 mask_len;
+
+    /*
+     * in the absence of a tree structure for the table that allows for an O(1)
+     * parent get, a cheeky way to find the cover is to LPM for the prefix with
+     * mask-1.
+     * there should always be a cover, though it may be the default route. the
+     * default route's cover is the default route.
+     */
+    if (len == 64)
+    {
+        /* go from (S,G) to (*,G*) */
+        mask_len = 32;
+    }
+    else if (len != 0)
+    {
+	mask_len = len - 1;
+    }
+    else
+    {
+        mask_len = len;
+    }
+
+    return (ip4_mfib_table_lookup(mfib, src, grp, mask_len));
+}
+
 void
 ip4_mfib_table_entry_insert (ip4_mfib_t *mfib,
                              const ip4_address_t *grp,
diff --git a/src/vnet/mfib/ip4_mfib.h b/src/vnet/mfib/ip4_mfib.h
index 7767d795cad..5b225596469 100644
--- a/src/vnet/mfib/ip4_mfib.h
+++ b/src/vnet/mfib/ip4_mfib.h
@@ -38,6 +38,10 @@ extern fib_node_index_t ip4_mfib_table_lookup_exact_match(const ip4_mfib_t *fib,
                                                           const ip4_address_t *grp,
                                                           const ip4_address_t *src,
                                                           u32 len);
+extern fib_node_index_t ip4_mfib_table_get_less_specific (const ip4_mfib_t *mfib,
+                                                          const ip4_address_t *src,
+                                                          const ip4_address_t *grp,
+                                                          u32 len);
 
 extern void ip4_mfib_table_entry_remove(ip4_mfib_t *fib,
                                         const ip4_address_t *grp,
diff --git a/src/vnet/mfib/ip6_mfib.c b/src/vnet/mfib/ip6_mfib.c
index 554a932844f..d47887be699 100644
--- a/src/vnet/mfib/ip6_mfib.c
+++ b/src/vnet/mfib/ip6_mfib.c
@@ -365,7 +365,6 @@ ip6_mfib_table_fwd_lookup (const ip6_mfib_t *mfib,
 
 	ASSERT(len >= 0 && len <= 256);
         IP6_MFIB_MK_KEY(mfib, grp, src, len, key);
-
 	rv = clib_bihash_search_inline_2_40_8(&table->ip6_mhash, &key, &value);
 	if (rv == 0)
 	    return value.value;
@@ -374,6 +373,39 @@ ip6_mfib_table_fwd_lookup (const ip6_mfib_t *mfib,
     return (FIB_NODE_INDEX_INVALID);
 }
 
+
+fib_node_index_t
+ip6_mfib_table_get_less_specific (const ip6_mfib_t *mfib,
+                                  const ip6_address_t *src,
+                                  const ip6_address_t *grp,
+                                  u32 len)
+{
+    u32 mask_len;
+
+    /*
+     * in the absence of a tree structure for the table that allows for an O(1)
+     * parent get, a cheeky way to find the cover is to LPM for the prefix with
+     * mask-1.
+     * there should always be a cover, though it may be the default route. the
+     * default route's cover is the default route.
+     */
+    if (len == 256)
+    {
+        /* go from (S,G) to (*,G*) */
+        mask_len = 128;
+    }
+    else if (len != 0)
+    {
+	mask_len = len - 1;
+    }
+    else
+    {
+        mask_len = len;
+    }
+
+    return (ip6_mfib_table_lookup(mfib, src, grp, mask_len));
+}
+
 /*
  * ip6_fib_table_lookup
  *
@@ -501,12 +533,23 @@ ip6_mfib_table_show_one (ip6_mfib_t *mfib,
                          vlib_main_t * vm,
                          ip6_address_t *src,
                          ip6_address_t *grp,
-                         u32 mask_len)
+                         u32 mask_len,
+                         u32 cover)
 {
-    vlib_cli_output(vm, "%U",
-                    format_mfib_entry,
-                    ip6_mfib_table_lookup(mfib, src, grp, mask_len),
-                    MFIB_ENTRY_FORMAT_DETAIL);
+    if (cover)
+    {
+        vlib_cli_output(vm, "%U",
+                        format_mfib_entry,
+                        ip6_mfib_table_get_less_specific(mfib, src, grp, mask_len),
+                        MFIB_ENTRY_FORMAT_DETAIL);
+    }
+    else
+    {
+        vlib_cli_output(vm, "%U",
+                        format_mfib_entry,
+                        ip6_mfib_table_lookup(mfib, src, grp, mask_len),
+                        MFIB_ENTRY_FORMAT_DETAIL);
+    }
 }
 
 typedef struct ip6_mfib_show_ctx_t_ {
@@ -600,11 +643,12 @@ ip6_show_mfib (vlib_main_t * vm,
     mfib_table_t *mfib_table;
     int verbose, matching;
     ip6_address_t grp, src = {{0}};
-    u32 mask = 32;
+    u32 mask = 128, cover;
     int table_id = -1, fib_index = ~0;
 
     verbose = 1;
     matching = 0;
+    cover = 0;
 
     while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
     {
@@ -634,6 +678,8 @@ ip6_show_mfib (vlib_main_t * vm,
             ;
         else if (unformat (input, "index %d", &fib_index))
             ;
+        else if (unformat (input, "cover"))
+            cover = 1;
         else
             break;
     }
@@ -671,7 +717,7 @@ ip6_show_mfib (vlib_main_t * vm,
         }
         else
         {
-            ip6_mfib_table_show_one(mfib, vm, &src, &grp, mask);
+            ip6_mfib_table_show_one(mfib, vm, &src, &grp, mask, cover);
         }
     }));
 
diff --git a/src/vnet/mfib/ip6_mfib.h b/src/vnet/mfib/ip6_mfib.h
index 5ed330b30af..f197b9d9fe8 100644
--- a/src/vnet/mfib/ip6_mfib.h
+++ b/src/vnet/mfib/ip6_mfib.h
@@ -41,6 +41,10 @@ extern fib_node_index_t ip6_mfib_table_lookup_exact_match(const ip6_mfib_t *fib,
                                                           const ip6_address_t *grp,
                                                           const ip6_address_t *src,
                                                           u32 len);
+extern fib_node_index_t ip6_mfib_table_get_less_specific (const ip6_mfib_t *mfib,
+                                                          const ip6_address_t *src,
+                                                          const ip6_address_t *grp,
+                                                          u32 len);
 
 extern void ip6_mfib_table_entry_remove(ip6_mfib_t *fib,
                                         const ip6_address_t *grp,
diff --git a/src/vnet/mfib/mfib_entry.c b/src/vnet/mfib/mfib_entry.c
index 90b223e3c25..51d775dd295 100644
--- a/src/vnet/mfib/mfib_entry.c
+++ b/src/vnet/mfib/mfib_entry.c
@@ -16,7 +16,10 @@
 #include <vlib/vlib.h>
 
 #include <vnet/mfib/mfib_entry.h>
+#include <vnet/mfib/mfib_entry_src.h>
+#include <vnet/mfib/mfib_entry_cover.h>
 #include <vnet/fib/fib_path_list.h>
+#include <vnet/fib/fib_walk.h>
 
 #include <vnet/dpo/drop_dpo.h>
 #include <vnet/dpo/replicate_dpo.h>
@@ -27,66 +30,6 @@
 vlib_log_class_t mfib_entry_logger;
 
 /**
- * Debug macro
- */
-#define MFIB_ENTRY_DBG(_e, _fmt, _args...)		\
-{                                                       \
-    vlib_log_debug(mfib_entry_logger,                   \
-                   "e:[%d:%U]: " _fmt,                  \
-                   mfib_entry_get_index(_e),		\
-                   format_mfib_prefix,			\
-                   &_e->mfe_prefix,                     \
-                   ##_args);                            \
-}
-
-/**
- * MFIB extensions to each path
- */
-typedef struct mfib_path_ext_t_
-{
-    mfib_itf_flags_t mfpe_flags;
-    fib_node_index_t mfpe_path;
-} mfib_path_ext_t;
-
-/**
- * The source of an MFIB entry
- */
-typedef struct mfib_entry_src_t_
-{
-    /**
-     * Which source this is
-     */
-    mfib_source_t mfes_src;
-
-    /**
-     * Route flags
-     */
-    mfib_entry_flags_t mfes_flags;
-
-    /**
-     * The path-list of forwarding interfaces
-     */
-    fib_node_index_t mfes_pl;
-
-    /**
-     * RPF-ID
-     */
-    fib_rpf_id_t mfes_rpf_id;
-
-    /**
-     * Hash table of path extensions
-     */
-    mfib_path_ext_t *mfes_exts;
-
-    /**
-     * The hash table of all interfaces.
-     *  This is forwarding time information derived from the paths
-     *  and their extensions.
-     */
-    mfib_itf_t *mfes_itfs;
-} mfib_entry_src_t;
-
-/**
  * Pool of path extensions
  */
 static mfib_path_ext_t *mfib_path_ext_pool;
@@ -184,8 +127,14 @@ format_mfib_entry (u8 * s, va_list * args)
         s = format (s, " locks:%d\n", mfib_entry->mfe_node.fn_locks);
         vec_foreach(msrc, mfib_entry->mfe_srcs)
         {
-            s = format (s, "  src:%s", mfib_source_names[msrc->mfes_src]);
-            s = format (s, ": %U\n", format_mfib_entry_flags, msrc->mfes_flags);
+            s = format (s, "  src:%s locks:%d:",
+                        mfib_source_names[msrc->mfes_src],
+                        msrc->mfes_ref_count);
+            if (msrc->mfes_cover != FIB_NODE_INDEX_INVALID)
+            {
+                s = format (s, " cover:%d", msrc->mfes_cover);
+            }
+            s = format (s, " %U\n", format_mfib_entry_flags, msrc->mfes_flags);
             if (FIB_NODE_INDEX_INVALID != msrc->mfes_pl)
             {
                 s = fib_path_list_format(msrc->mfes_pl, s);
@@ -254,6 +203,9 @@ mfib_entry_src_init (mfib_entry_t *mfib_entry,
         .mfes_pl = FIB_NODE_INDEX_INVALID,
         .mfes_flags = MFIB_ENTRY_FLAG_NONE,
         .mfes_src = source,
+        .mfes_cover = FIB_NODE_INDEX_INVALID,
+        .mfes_sibling = FIB_NODE_INDEX_INVALID,
+        .mfes_ref_count = 1,
     };
 
     vec_add1(mfib_entry->mfe_srcs, esrc);
@@ -294,19 +246,51 @@ static mfib_entry_src_t *
 mfib_entry_src_find_or_create (mfib_entry_t *mfib_entry,
                                mfib_source_t source)
 {
-    mfib_entry_src_t *esrc;
+    mfib_entry_src_t *msrc;
 
-    esrc = mfib_entry_src_find(mfib_entry, source, NULL);
+    msrc = mfib_entry_src_find(mfib_entry, source, NULL);
 
-    if (NULL == esrc)
+    if (NULL == msrc)
     {
         mfib_entry_src_init(mfib_entry, source);
+        msrc = mfib_entry_src_find(mfib_entry, source, NULL);
     }
 
-    return (mfib_entry_src_find(mfib_entry, source, NULL));
+    return (msrc);
+}
+
+static mfib_entry_src_t *
+mfib_entry_src_update (mfib_entry_t *mfib_entry,
+                       mfib_source_t source,
+                       fib_rpf_id_t rpf_id,
+                       mfib_entry_flags_t entry_flags)
+{
+    mfib_entry_src_t *msrc;
+
+    msrc = mfib_entry_src_find_or_create(mfib_entry, source);
+
+    msrc->mfes_flags = entry_flags;
+    msrc->mfes_rpf_id = rpf_id;
+
+    return (msrc);
 }
 
-static mfib_entry_src_t*
+static mfib_entry_src_t *
+mfib_entry_src_update_and_lock (mfib_entry_t *mfib_entry,
+                                mfib_source_t source,
+                                fib_rpf_id_t rpf_id,
+                                mfib_entry_flags_t entry_flags)
+{
+    mfib_entry_src_t *msrc;
+
+    msrc = mfib_entry_src_update(mfib_entry, source, rpf_id, entry_flags);
+
+    msrc->mfes_ref_count++;
+
+    return (msrc);
+}
+
+mfib_entry_src_t*
 mfib_entry_get_best_src (const mfib_entry_t *mfib_entry)
 {
     mfib_entry_src_t *bsrc;
@@ -326,6 +310,16 @@ mfib_entry_get_best_src (const mfib_entry_t *mfib_entry)
     return (bsrc);
 }
 
+static mfib_source_t
+mfib_entry_get_best_source (const mfib_entry_t *mfib_entry)
+{
+    mfib_entry_src_t *bsrc;
+
+    bsrc = mfib_entry_get_best_src(mfib_entry);
+
+    return (bsrc->mfes_src);
+}
+
 int
 mfib_entry_is_sourced (fib_node_index_t mfib_entry_index,
                        mfib_source_t source)
@@ -337,6 +331,13 @@ mfib_entry_is_sourced (fib_node_index_t mfib_entry_index,
     return (NULL != mfib_entry_src_find(mfib_entry, source, NULL));
 }
 
+int
+mfib_entry_is_host (fib_node_index_t mfib_entry_index)
+{
+    return (mfib_prefix_is_host(mfib_entry_get_prefix(mfib_entry_index)));
+}
+
+
 static void
 mfib_entry_src_flush (mfib_entry_src_t *msrc)
 {
@@ -364,8 +365,19 @@ mfib_entry_src_remove (mfib_entry_t *mfib_entry,
 
     if (NULL != msrc)
     {
-        mfib_entry_src_flush(msrc);
-        vec_del1(mfib_entry->mfe_srcs, index);
+        ASSERT(0 != msrc->mfes_ref_count);
+        msrc->mfes_ref_count--;
+
+        if (0 == msrc->mfes_ref_count)
+        {
+            mfib_entry_src_deactivate(mfib_entry, msrc);
+            mfib_entry_src_flush(msrc);
+
+            vec_del1(mfib_entry->mfe_srcs, index);
+            if (vec_len (mfib_entry->mfe_srcs) > 1)
+                vec_sort_with_function(mfib_entry->mfe_srcs,
+                                       mfib_entry_src_cmp_for_sort);
+        }
     }
 }
 
@@ -552,8 +564,7 @@ mfib_entry_stack (mfib_entry_t *mfib_entry,
                                    mfib_entry->mfe_sibling);
     }
 
-    if (NULL != msrc &&
-        FIB_NODE_INDEX_INVALID != msrc->mfes_pl)
+    if (NULL != msrc)
     {
         mfib_entry_collect_forwarding_ctx_t ctx = {
             .next_hops = NULL,
@@ -561,10 +572,27 @@ mfib_entry_stack (mfib_entry_t *mfib_entry,
             .msrc = msrc,
         };
 
-        fib_path_list_walk(msrc->mfes_pl,
-                           mfib_entry_src_collect_forwarding,
-                           &ctx);
+        /*
+         * link the entry to the path-list.
+         * The entry needs to be a child so that we receive the back-walk
+         * updates to recalculate forwarding.
+         */
+        mfib_entry->mfe_pl = msrc->mfes_pl;
+        mfib_entry->mfe_flags = msrc->mfes_flags;
+        mfib_entry->mfe_itfs = msrc->mfes_itfs;
+        mfib_entry->mfe_rpf_id = msrc->mfes_rpf_id;
 
+        if (FIB_NODE_INDEX_INVALID != mfib_entry->mfe_pl)
+        {
+            mfib_entry->mfe_sibling =
+                fib_path_list_child_add(mfib_entry->mfe_pl,
+                                        FIB_NODE_TYPE_MFIB_ENTRY,
+                                        mfib_entry_get_index(mfib_entry));
+
+            fib_path_list_walk(mfib_entry->mfe_pl,
+                               mfib_entry_src_collect_forwarding,
+                               &ctx);
+        }
         if (!(MFIB_ENTRY_FLAG_EXCLUSIVE & mfib_entry->mfe_flags))
         {
             if (NULL == ctx.next_hops)
@@ -616,17 +644,6 @@ mfib_entry_stack (mfib_entry_t *mfib_entry,
             dpo_reset(&ctx.next_hops[0].path_dpo);
             vec_free(ctx.next_hops);
         }
-
-        /*
-         * link the entry to the path-list.
-         * The entry needs to be a child so that we receive the back-walk
-         * updates to recalculate forwarding.
-         */
-        mfib_entry->mfe_pl = msrc->mfes_pl;
-        mfib_entry->mfe_sibling =
-            fib_path_list_child_add(mfib_entry->mfe_pl,
-                                    FIB_NODE_TYPE_MFIB_ENTRY,
-                                    mfib_entry_get_index(mfib_entry));
     }
     else
     {
@@ -634,6 +651,17 @@ mfib_entry_stack (mfib_entry_t *mfib_entry,
                   &mfib_entry->mfe_rep,
                   drop_dpo_get(dp));
     }
+
+    /*
+     * time for walkies fido.
+     */
+    fib_node_back_walk_ctx_t bw_ctx = {
+        .fnbw_reason = FIB_NODE_BW_REASON_FLAG_EVALUATE,
+    };
+
+    fib_walk_sync(FIB_NODE_TYPE_MFIB_ENTRY,
+                  mfib_entry_get_index(mfib_entry),
+                  &bw_ctx);
 }
 
 static fib_node_index_t
@@ -689,23 +717,36 @@ mfib_entry_src_path_remove (mfib_entry_src_t *msrc,
 }
 
 static void
-mfib_entry_recalculate_forwarding (mfib_entry_t *mfib_entry)
+mfib_entry_recalculate_forwarding (mfib_entry_t *mfib_entry,
+                                   mfib_source_t old_best)
 {
-    mfib_entry_src_t *bsrc;
+    mfib_entry_src_t *bsrc, *osrc;
 
     /*
      * copy the forwarding data from the bast source
      */
     bsrc = mfib_entry_get_best_src(mfib_entry);
+    osrc = mfib_entry_src_find(mfib_entry, old_best, NULL);
 
     if (NULL != bsrc)
     {
-        mfib_entry->mfe_flags = bsrc->mfes_flags;
-        mfib_entry->mfe_itfs = bsrc->mfes_itfs;
-        mfib_entry->mfe_rpf_id = bsrc->mfes_rpf_id;
+        if (bsrc->mfes_src != old_best)
+        {
+            /*
+             * we are changing from one source to another
+             * deactivate the old, and activate the new
+             */
+            mfib_entry_src_deactivate(mfib_entry, osrc);
+            mfib_entry_src_activate(mfib_entry, bsrc);
+        }
+    }
+    else
+    {
+        mfib_entry_src_deactivate(mfib_entry, osrc);
     }
 
     mfib_entry_stack(mfib_entry, bsrc);
+    mfib_entry_cover_update_notify(mfib_entry);
 }
 
 
@@ -714,7 +755,8 @@ mfib_entry_create (u32 fib_index,
                    mfib_source_t source,
                    const mfib_prefix_t *prefix,
                    fib_rpf_id_t rpf_id,
-                   mfib_entry_flags_t entry_flags)
+                   mfib_entry_flags_t entry_flags,
+                   index_t repi)
 {
     fib_node_index_t mfib_entry_index;
     mfib_entry_t *mfib_entry;
@@ -722,11 +764,37 @@ mfib_entry_create (u32 fib_index,
 
     mfib_entry = mfib_entry_alloc(fib_index, prefix,
                                   &mfib_entry_index);
-    msrc = mfib_entry_src_find_or_create(mfib_entry, source);
-    msrc->mfes_flags = entry_flags;
-    msrc->mfes_rpf_id = rpf_id;
+    msrc = mfib_entry_src_update(mfib_entry, source,
+                                 rpf_id, entry_flags);
 
-    mfib_entry_recalculate_forwarding(mfib_entry);
+    if (INDEX_INVALID != repi)
+    {
+        /*
+         * The source is providing its own replicate DPO.
+         * Create a sepcial path-list to manage it, that way
+         * this entry and the source are equivalent to a normal
+         * entry
+         */
+        fib_node_index_t old_pl_index;
+        dpo_proto_t dp;
+        dpo_id_t dpo = DPO_INVALID;
+
+        dp = fib_proto_to_dpo(mfib_entry_get_proto(mfib_entry));
+        old_pl_index = msrc->mfes_pl;
+
+        dpo_set(&dpo, DPO_REPLICATE, dp, repi);
+
+        msrc->mfes_pl =
+            fib_path_list_create_special(dp,
+                                         FIB_PATH_LIST_FLAG_EXCLUSIVE,
+                                         &dpo);
+
+        dpo_reset(&dpo);
+        fib_path_list_lock(msrc->mfes_pl);
+        fib_path_list_unlock(old_pl_index);
+    }
+
+    mfib_entry_recalculate_forwarding(mfib_entry, MFIB_SOURCE_NONE);
 
     return (mfib_entry_index);
 }
@@ -740,25 +808,18 @@ mfib_entry_ok_for_delete (mfib_entry_t *mfib_entry)
 static int
 mfib_entry_src_ok_for_delete (const mfib_entry_src_t *msrc)
 {
-    return ((MFIB_ENTRY_FLAG_NONE == msrc->mfes_flags &&
+    return ((INDEX_INVALID == msrc->mfes_cover &&
+             MFIB_ENTRY_FLAG_NONE == msrc->mfes_flags &&
              0 == fib_path_list_get_n_paths(msrc->mfes_pl)));
 }
 
-int
-mfib_entry_update (fib_node_index_t mfib_entry_index,
-                   mfib_source_t source,
-                   mfib_entry_flags_t entry_flags,
-                   fib_rpf_id_t rpf_id,
-                   index_t repi)
-{
-    mfib_entry_t *mfib_entry;
-    mfib_entry_src_t *msrc;
-
-    mfib_entry = mfib_entry_get(mfib_entry_index);
-    msrc = mfib_entry_src_find_or_create(mfib_entry, source);
-    msrc->mfes_flags = entry_flags;
-    msrc->mfes_rpf_id = rpf_id;
 
+static void
+mfib_entry_update_i (mfib_entry_t *mfib_entry,
+                     mfib_entry_src_t *msrc,
+                     mfib_source_t current_best,
+                     index_t repi)
+{
     if (INDEX_INVALID != repi)
     {
         /*
@@ -792,10 +853,50 @@ mfib_entry_update (fib_node_index_t mfib_entry_index,
          * this source has no interfaces and no flags.
          * it has nothing left to give - remove it
          */
-        mfib_entry_src_remove(mfib_entry, source);
+        mfib_entry_src_remove(mfib_entry, msrc->mfes_src);
     }
 
-    mfib_entry_recalculate_forwarding(mfib_entry);
+    mfib_entry_recalculate_forwarding(mfib_entry, current_best);
+}
+
+int
+mfib_entry_special_add (fib_node_index_t mfib_entry_index,
+                        mfib_source_t source,
+                        mfib_entry_flags_t entry_flags,
+                        fib_rpf_id_t rpf_id,
+                        index_t repi)
+{
+    mfib_source_t current_best;
+    mfib_entry_t *mfib_entry;
+    mfib_entry_src_t *msrc;
+
+    mfib_entry = mfib_entry_get(mfib_entry_index);
+    current_best = mfib_entry_get_best_source(mfib_entry);
+
+    msrc = mfib_entry_src_update_and_lock(mfib_entry, source, rpf_id,
+                                          entry_flags);
+
+    mfib_entry_update_i(mfib_entry, msrc, current_best, repi);
+
+    return (mfib_entry_ok_for_delete(mfib_entry));
+}
+
+int
+mfib_entry_update (fib_node_index_t mfib_entry_index,
+                   mfib_source_t source,
+                   mfib_entry_flags_t entry_flags,
+                   fib_rpf_id_t rpf_id,
+                   index_t repi)
+{
+    mfib_source_t current_best;
+    mfib_entry_t *mfib_entry;
+    mfib_entry_src_t *msrc;
+
+    mfib_entry = mfib_entry_get(mfib_entry_index);
+    current_best = mfib_entry_get_best_source(mfib_entry);
+    msrc = mfib_entry_src_update(mfib_entry, source, rpf_id, entry_flags);
+
+    mfib_entry_update_i(mfib_entry, msrc, current_best, repi);
 
     return (mfib_entry_ok_for_delete(mfib_entry));
 }
@@ -828,6 +929,7 @@ mfib_entry_path_update (fib_node_index_t mfib_entry_index,
                         mfib_itf_flags_t itf_flags)
 {
     fib_node_index_t path_index;
+    mfib_source_t current_best;
     mfib_path_ext_t *path_ext;
     mfib_entry_t *mfib_entry;
     mfib_entry_src_t *msrc;
@@ -835,6 +937,7 @@ mfib_entry_path_update (fib_node_index_t mfib_entry_index,
 
     mfib_entry = mfib_entry_get(mfib_entry_index);
     ASSERT(NULL != mfib_entry);
+    current_best = mfib_entry_get_best_source(mfib_entry);
     msrc = mfib_entry_src_find_or_create(mfib_entry, source);
 
     /*
@@ -897,7 +1000,7 @@ mfib_entry_path_update (fib_node_index_t mfib_entry_index,
         }
     }
 
-    mfib_entry_recalculate_forwarding(mfib_entry);
+    mfib_entry_recalculate_forwarding(mfib_entry, current_best);
 }
 
 /*
@@ -912,11 +1015,13 @@ mfib_entry_path_remove (fib_node_index_t mfib_entry_index,
                         const fib_route_path_t *rpath)
 {
     fib_node_index_t path_index;
+    mfib_source_t current_best;
     mfib_entry_t *mfib_entry;
     mfib_entry_src_t *msrc;
 
     mfib_entry = mfib_entry_get(mfib_entry_index);
     ASSERT(NULL != mfib_entry);
+    current_best = mfib_entry_get_best_source(mfib_entry);
     msrc = mfib_entry_src_find(mfib_entry, source, NULL);
 
     if (NULL == msrc)
@@ -968,7 +1073,7 @@ mfib_entry_path_remove (fib_node_index_t mfib_entry_index,
         mfib_entry_src_remove(mfib_entry, source);
     }
 
-    mfib_entry_recalculate_forwarding(mfib_entry);
+    mfib_entry_recalculate_forwarding(mfib_entry, current_best);
 
     return (mfib_entry_ok_for_delete(mfib_entry));
 }
@@ -982,12 +1087,14 @@ int
 mfib_entry_delete (fib_node_index_t mfib_entry_index,
                    mfib_source_t source)
 {
+    mfib_source_t current_best;
     mfib_entry_t *mfib_entry;
 
     mfib_entry = mfib_entry_get(mfib_entry_index);
+    current_best = mfib_entry_get_best_source(mfib_entry);
     mfib_entry_src_remove(mfib_entry, source);
 
-    mfib_entry_recalculate_forwarding(mfib_entry);
+    mfib_entry_recalculate_forwarding(mfib_entry, current_best);
 
     return (mfib_entry_ok_for_delete(mfib_entry));
 }
@@ -1116,7 +1223,11 @@ static fib_node_back_walk_rc_t
 mfib_entry_back_walk_notify (fib_node_t *node,
                             fib_node_back_walk_ctx_t *ctx)
 {
-    mfib_entry_recalculate_forwarding(mfib_entry_from_fib_node(node));
+    mfib_entry_t *mfib_entry;
+
+    mfib_entry = mfib_entry_from_fib_node(node);
+    mfib_entry_recalculate_forwarding(mfib_entry,
+                                      mfib_entry_get_best_source(mfib_entry));
 
     return (FIB_NODE_BACK_WALK_CONTINUE);
 }
@@ -1219,15 +1330,14 @@ mfib_entry_encode (fib_node_index_t mfib_entry_index,
     }
 }
 
-
-void
-mfib_entry_get_prefix (fib_node_index_t mfib_entry_index,
-                      mfib_prefix_t *pfx)
+const mfib_prefix_t *
+mfib_entry_get_prefix (fib_node_index_t mfib_entry_index)
 {
     mfib_entry_t *mfib_entry;
 
     mfib_entry = mfib_entry_get(mfib_entry_index);
-    *pfx = mfib_entry->mfe_prefix;
+
+    return (&mfib_entry->mfe_prefix);
 }
 
 u32
@@ -1253,6 +1363,7 @@ mfib_entry_contribute_ip_forwarding (fib_node_index_t mfib_entry_index)
 void
 mfib_entry_contribute_forwarding (fib_node_index_t mfib_entry_index,
                                   fib_forward_chain_type_t type,
+                                  mfib_entry_fwd_flags_t flags,
                                   dpo_id_t *dpo)
 {
     /*
@@ -1267,9 +1378,26 @@ mfib_entry_contribute_forwarding (fib_node_index_t mfib_entry_index,
 
     dp = fib_proto_to_dpo(mfib_entry->mfe_prefix.fp_proto);
 
-    if (type == fib_forw_chain_type_from_dpo_proto(dp))
+    if (type == mfib_forw_chain_type_from_dpo_proto(dp))
     {
-        dpo_copy(dpo, &mfib_entry->mfe_rep);
+        replicate_t * rep;
+
+        rep = replicate_get(mfib_entry->mfe_rep.dpoi_index);
+
+        if ((rep->rep_flags & REPLICATE_FLAGS_HAS_LOCAL) &&
+            (flags & MFIB_ENTRY_FWD_FLAG_NO_LOCAL))
+        {
+            /*
+             * caller does not want the local paths that the entry has
+             */
+            dpo_set(dpo, DPO_REPLICATE, rep->rep_proto,
+                    replicate_dup(REPLICATE_FLAGS_NONE,
+                                  mfib_entry->mfe_rep.dpoi_index));
+        }
+        else
+        {
+            dpo_copy(dpo, &mfib_entry->mfe_rep);
+        }
     }
     else
     {
@@ -1277,6 +1405,55 @@ mfib_entry_contribute_forwarding (fib_node_index_t mfib_entry_index,
     }
 }
 
+/*
+ * fib_entry_cover_changed
+ *
+ * this entry is tracking its cover and that cover has changed.
+ */
+void
+mfib_entry_cover_changed (fib_node_index_t mfib_entry_index)
+{
+    mfib_entry_t *mfib_entry;
+    mfib_entry_src_t *msrc;
+    mfib_src_res_t res;
+
+    mfib_entry = mfib_entry_get(mfib_entry_index);
+    msrc = mfib_entry_get_best_src(mfib_entry);
+
+    res = mfib_entry_src_cover_change(mfib_entry, msrc);
+
+    if (MFIB_SRC_REEVALUATE == res)
+    {
+        mfib_entry_recalculate_forwarding(mfib_entry, msrc->mfes_src);
+    }
+    MFIB_ENTRY_DBG(mfib_entry, "cover-changed");
+}
+
+/*
+ * mfib_entry_cover_updated
+ *
+ * this entry is tracking its cover and that cover has been updated
+ * (i.e. its forwarding information has changed).
+ */
+void
+mfib_entry_cover_updated (fib_node_index_t mfib_entry_index)
+{
+    mfib_entry_t *mfib_entry;
+    mfib_entry_src_t *msrc;
+    mfib_src_res_t res;
+
+    mfib_entry = mfib_entry_get(mfib_entry_index);
+    msrc = mfib_entry_get_best_src(mfib_entry);
+
+    res = mfib_entry_src_cover_update(mfib_entry, msrc);
+
+    if (MFIB_SRC_REEVALUATE == res)
+    {
+        mfib_entry_recalculate_forwarding(mfib_entry, msrc->mfes_src);
+    }
+    MFIB_ENTRY_DBG(mfib_entry, "cover-updated");
+}
+
 u32
 mfib_entry_pool_size (void)
 {
diff --git a/src/vnet/mfib/mfib_entry.h b/src/vnet/mfib/mfib_entry.h
index 880c1e2933c..8ab7cee5ea7 100644
--- a/src/vnet/mfib/mfib_entry.h
+++ b/src/vnet/mfib/mfib_entry.h
@@ -19,6 +19,7 @@
 #include <vnet/fib/fib_node.h>
 #include <vnet/mfib/mfib_types.h>
 #include <vnet/mfib/mfib_itf.h>
+#include <vnet/mfib/mfib_entry_delegate.h>
 #include <vnet/ip/ip.h>
 #include <vnet/dpo/dpo.h>
 
@@ -34,10 +35,12 @@ typedef struct mfib_entry_t_ {
      * Base class. The entry's node representation in the graph.
      */
     fib_node_t mfe_node;
+
     /**
      * The prefix of the route
      */
     mfib_prefix_t mfe_prefix;
+
     /**
      * The index of the FIB table this entry is in
      */
@@ -82,8 +85,28 @@ typedef struct mfib_entry_t_ {
      * A hash table of interfaces
      */
     mfib_itf_t *mfe_itfs;
+
+    /**
+     * A vector of delegates.
+     */
+    mfib_entry_delegate_t *fe_delegates;
 } mfib_entry_t;
 
+/**
+ * Debug macro
+ */
+extern vlib_log_class_t mfib_entry_logger;
+
+#define MFIB_ENTRY_DBG(_e, _fmt, _args...)		\
+{                                                       \
+    vlib_log_debug(mfib_entry_logger,                   \
+                   "e:[%d:%U]: " _fmt,                  \
+                   mfib_entry_get_index(_e),		\
+                   format_mfib_prefix,			\
+                   &_e->mfe_prefix,                     \
+                   ##_args);                            \
+}
+
 #define MFIB_ENTRY_FORMAT_BRIEF   (0x0)
 #define MFIB_ENTRY_FORMAT_DETAIL  (0x1)
 #define MFIB_ENTRY_FORMAT_DETAIL2 (0x2)
@@ -95,7 +118,8 @@ extern fib_node_index_t mfib_entry_create(u32 fib_index,
                                           mfib_source_t source,
                                           const mfib_prefix_t *prefix,
                                           fib_rpf_id_t rpf_id,
-                                          mfib_entry_flags_t entry_flags);
+                                          mfib_entry_flags_t entry_flags,
+                                          index_t repi);
 
 extern int mfib_entry_update(fib_node_index_t fib_entry_index,
                              mfib_source_t source,
@@ -103,6 +127,12 @@ extern int mfib_entry_update(fib_node_index_t fib_entry_index,
                              fib_rpf_id_t rpf_id,
                              index_t rep_dpo);
 
+extern int mfib_entry_special_add(fib_node_index_t fib_entry_index,
+                                  mfib_source_t source,
+                                  mfib_entry_flags_t entry_flags,
+                                  fib_rpf_id_t rpf_id,
+                                  index_t rep_dpo);
+
 extern void mfib_entry_path_update(fib_node_index_t fib_entry_index,
                                    mfib_source_t source,
                                    const fib_route_path_t *rpath,
@@ -127,19 +157,35 @@ extern void mfib_entry_child_remove(fib_node_index_t mfib_entry_index,
 extern void mfib_entry_lock(fib_node_index_t fib_entry_index);
 extern void mfib_entry_unlock(fib_node_index_t fib_entry_index);
 
-extern void mfib_entry_get_prefix(fib_node_index_t fib_entry_index,
-                                  mfib_prefix_t *pfx);
+extern const mfib_prefix_t *mfib_entry_get_prefix(fib_node_index_t fib_entry_index);
 extern u32 mfib_entry_get_fib_index(fib_node_index_t fib_entry_index);
 extern int mfib_entry_is_sourced(fib_node_index_t fib_entry_index,
                                  mfib_source_t source);
+extern int mfib_entry_is_host(fib_node_index_t fib_entry_index);
 extern u32 mfib_entry_get_stats_index(fib_node_index_t fib_entry_index);
+extern void mfib_entry_cover_changed(fib_node_index_t fib_entry_index);
+extern void mfib_entry_cover_updated(fib_node_index_t fib_entry_index);
 
 extern const dpo_id_t*mfib_entry_contribute_ip_forwarding(
     fib_node_index_t mfib_entry_index);
 
+/**
+ * Flags to control what is present in the replicate DPO returned when
+ * the entry contributes forwarding
+ */
+typedef enum mfib_entry_fwd_flags_t_
+{
+    MFIB_ENTRY_FWD_FLAG_NONE,
+    /**
+     * Do not reutrn any local replications in the set
+     */
+    MFIB_ENTRY_FWD_FLAG_NO_LOCAL,
+} mfib_entry_fwd_flags_t;
+
 extern void mfib_entry_contribute_forwarding(
     fib_node_index_t mfib_entry_index,
     fib_forward_chain_type_t type,
+    mfib_entry_fwd_flags_t flags,
     dpo_id_t *dpo);
 
 extern void mfib_entry_encode(fib_node_index_t fib_entry_index,
diff --git a/src/vnet/mfib/mfib_entry_cover.c b/src/vnet/mfib/mfib_entry_cover.c
new file mode 100644
index 00000000000..6caeb1b6928
--- /dev/null
+++ b/src/vnet/mfib/mfib_entry_cover.c
@@ -0,0 +1,180 @@
+/*
+ * Copyright (c) 2016 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <vnet/mfib/mfib_entry_cover.h>
+#include <vnet/mfib/mfib_entry_src.h>
+#include <vnet/fib/fib_node_list.h>
+
+u32
+mfib_entry_cover_track (mfib_entry_t* cover,
+		       fib_node_index_t covered)
+{
+    mfib_entry_delegate_t *mfed;
+
+    MFIB_ENTRY_DBG(cover, "cover-track %d", covered);
+
+    ASSERT(mfib_entry_get_index(cover) != covered);
+
+    mfed = mfib_entry_delegate_get(cover, MFIB_ENTRY_DELEGATE_COVERED);
+
+    if (NULL == mfed)
+    {
+        mfed = mfib_entry_delegate_find_or_add(cover, MFIB_ENTRY_DELEGATE_COVERED);
+        mfed->mfd_list = fib_node_list_create();
+    }
+
+    return (fib_node_list_push_front(mfed->mfd_list,
+                                     0, FIB_NODE_TYPE_MFIB_ENTRY,
+                                     covered));
+}
+
+void
+mfib_entry_cover_untrack (mfib_entry_t* cover,
+			 u32 tracked_index)
+{
+    mfib_entry_delegate_t *mfed;
+
+    MFIB_ENTRY_DBG(cover, "cover-untrack @ %d", tracked_index);
+
+    mfed = mfib_entry_delegate_get(cover, MFIB_ENTRY_DELEGATE_COVERED);
+
+    if (NULL == mfed)
+        return;
+
+    fib_node_list_remove(mfed->mfd_list, tracked_index);
+
+    if (0 == fib_node_list_get_size(mfed->mfd_list))
+    {
+        fib_node_list_destroy(&mfed->mfd_list);
+        mfib_entry_delegate_remove(cover, MFIB_ENTRY_DELEGATE_COVERED);        
+    }
+}
+
+/**
+ * Internal struct to hold user supplied paraneters for the cover walk
+ */
+typedef struct mfib_enty_cover_walk_ctx_t_ {
+    mfib_entry_t *cover;
+    mfib_entry_covered_walk_t walk;
+    void *ctx;
+} mfib_enty_cover_walk_ctx_t;
+
+static int
+mfib_entry_cover_walk_node_ptr (fib_node_ptr_t *depend,
+                                void *args)
+{
+    mfib_enty_cover_walk_ctx_t *ctx = args;
+
+    ctx->walk(ctx->cover, depend->fnp_index, ctx->ctx);
+
+    /* continue */
+    return (1);
+}
+
+void
+mfib_entry_cover_walk (mfib_entry_t *cover,
+		      mfib_entry_covered_walk_t walk,
+		      void *args)
+{
+    mfib_entry_delegate_t *mfed;
+
+    mfed = mfib_entry_delegate_get(cover, MFIB_ENTRY_DELEGATE_COVERED);
+
+    if (NULL == mfed)
+        return;
+
+    mfib_enty_cover_walk_ctx_t ctx = {
+        .cover = cover,
+        .walk = walk,
+        .ctx = args,
+    };
+
+    fib_node_list_walk(mfed->mfd_list,
+                       mfib_entry_cover_walk_node_ptr,
+                       &ctx);
+}
+
+static int
+mfib_entry_cover_change_one (mfib_entry_t *cover,
+			    fib_node_index_t covered,
+			    void *args)
+{
+    fib_node_index_t new_cover;
+
+    /*
+     * The 3 entries involved here are:
+     *   cover - the least specific. It will cover both the others
+     *  new_cover - the enty just inserted below the cover
+     *  covered - the entry that was tracking the cover.
+     *
+     * The checks below are to determine if new_cover is a cover for covered.
+     */
+    new_cover = pointer_to_uword(args);
+
+    if (FIB_NODE_INDEX_INVALID == new_cover)
+    {
+	/*
+	 * nothing has been inserted, which implies the cover was removed.
+	 * 'cover' is thus the new cover.
+	 */
+	mfib_entry_cover_changed(covered);
+    }
+    else if (new_cover != covered)
+    {
+	const mfib_prefix_t *pfx_covered, *pfx_new_cover;
+
+	pfx_covered = mfib_entry_get_prefix(covered);
+	pfx_new_cover = mfib_entry_get_prefix(new_cover);
+
+	if (mfib_prefix_is_cover(pfx_new_cover, pfx_covered))
+	{
+	    mfib_entry_cover_changed(covered);
+	}
+    }
+    /* continue */
+    return (1);
+}
+
+void
+mfib_entry_cover_change_notify (fib_node_index_t cover_index,
+                                fib_node_index_t covered)
+{
+    mfib_entry_t *cover;
+
+    cover = mfib_entry_get(cover_index);
+
+    mfib_entry_cover_walk(cover, 
+                          mfib_entry_cover_change_one,
+                          uword_to_pointer(covered, void*));
+}
+
+static int
+mfib_entry_cover_update_one (mfib_entry_t *cover,
+			    fib_node_index_t covered,
+			    void *args)
+{
+    mfib_entry_cover_updated(covered);
+
+    /* continue */
+    return (1);
+}
+
+void
+mfib_entry_cover_update_notify (mfib_entry_t *mfib_entry)
+{
+    mfib_entry_cover_walk(mfib_entry, 
+			 mfib_entry_cover_update_one,
+			 NULL);
+}
diff --git a/src/vnet/mfib/mfib_entry_cover.h b/src/vnet/mfib/mfib_entry_cover.h
new file mode 100644
index 00000000000..8f165c29dfd
--- /dev/null
+++ b/src/vnet/mfib/mfib_entry_cover.h
@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2018 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __MFIB_ENTRY_COVER_H__
+#define __MFIB_ENTRY_COVER_H__
+
+#include <vnet/mfib/mfib_entry.h>
+
+/**
+ * callback function used when walking the covered entries
+ */
+typedef int (*mfib_entry_covered_walk_t)(mfib_entry_t *cover,
+                                         fib_node_index_t covered,
+                                         void *ctx);
+
+extern u32 mfib_entry_cover_track(mfib_entry_t *cover,
+                                  fib_node_index_t covered);
+
+extern void mfib_entry_cover_untrack(mfib_entry_t *cover,
+                                     u32 tracked_index);
+
+extern void mfib_entry_cover_walk(mfib_entry_t *cover,
+                                  mfib_entry_covered_walk_t walk,
+                                  void *ctx);
+
+extern void mfib_entry_cover_change_notify(fib_node_index_t cover_index,
+                                           fib_node_index_t covered_index);
+extern void mfib_entry_cover_update_notify(mfib_entry_t *cover);
+
+#endif
diff --git a/src/vnet/mfib/mfib_entry_delegate.c b/src/vnet/mfib/mfib_entry_delegate.c
new file mode 100644
index 00000000000..95915cb3807
--- /dev/null
+++ b/src/vnet/mfib/mfib_entry_delegate.c
@@ -0,0 +1,143 @@
+/*
+ * Copyright (c) 2018 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <vnet/mfib/mfib_entry_delegate.h>
+#include <vnet/mfib/mfib_entry.h>
+
+static mfib_entry_delegate_t *
+mfib_entry_delegate_find_i (const mfib_entry_t *mfib_entry,
+                            mfib_entry_delegate_type_t type,
+                            u32 *index)
+{
+    mfib_entry_delegate_t *delegate;
+    int ii;
+
+    ii = 0;
+    vec_foreach(delegate, mfib_entry->fe_delegates)
+    {
+        if (delegate->mfd_type == type)
+        {
+            if (NULL != index)
+                *index = ii;
+
+            return (delegate);
+        }
+        else
+        {
+            ii++;
+        }
+    }
+
+    return (NULL);
+}
+
+mfib_entry_delegate_t *
+mfib_entry_delegate_get (const mfib_entry_t *mfib_entry,
+                         mfib_entry_delegate_type_t type)
+{
+    return (mfib_entry_delegate_find_i(mfib_entry, type, NULL));
+}
+
+void
+mfib_entry_delegate_remove (mfib_entry_t *mfib_entry,
+                            mfib_entry_delegate_type_t type)
+{
+    mfib_entry_delegate_t *fed;
+    u32 index = ~0;
+
+    fed = mfib_entry_delegate_find_i(mfib_entry, type, &index);
+
+    ASSERT(NULL != fed);
+
+    vec_del1(mfib_entry->fe_delegates, index);
+}
+
+static int
+mfib_entry_delegate_cmp_for_sort (void * v1,
+                                  void * v2)
+{
+    mfib_entry_delegate_t *delegate1 = v1, *delegate2 = v2;
+
+    return (delegate1->mfd_type - delegate2->mfd_type);
+}
+
+static void
+mfib_entry_delegate_init (mfib_entry_t *mfib_entry,
+                          mfib_entry_delegate_type_t type)
+
+{
+    mfib_entry_delegate_t delegate = {
+        .mfd_entry_index = mfib_entry_get_index(mfib_entry),
+        .mfd_type = type,
+    };
+
+    vec_add1(mfib_entry->fe_delegates, delegate);
+    vec_sort_with_function(mfib_entry->fe_delegates,
+                           mfib_entry_delegate_cmp_for_sort);
+}
+
+mfib_entry_delegate_t *
+mfib_entry_delegate_find_or_add (mfib_entry_t *mfib_entry,
+                                 mfib_entry_delegate_type_t fdt)
+{
+    mfib_entry_delegate_t *delegate;
+
+    delegate = mfib_entry_delegate_get(mfib_entry, fdt);
+
+    if (NULL == delegate)
+    {
+        mfib_entry_delegate_init(mfib_entry, fdt);
+    }
+
+    return (mfib_entry_delegate_get(mfib_entry, fdt));
+}
+
+/**
+ * typedef for printing a delegate
+ */
+typedef u8 * (*mfib_entry_delegate_format_t)(const mfib_entry_delegate_t *fed,
+                                             u8 *s);
+
+/**
+ * Print a delegate that represents cover tracking
+ */
+static u8 *
+mfib_entry_delegate_fmt_covered (const mfib_entry_delegate_t *fed,
+                                 u8 *s)
+{
+    s = format(s, "covered:[");
+    s = fib_node_children_format(fed->mfd_list, s);
+    s = format(s, "]");
+
+    return (s);
+}
+
+/**
+ * A delegate type to formatter map
+ */
+static mfib_entry_delegate_format_t fed_formatters[] =
+{
+    [MFIB_ENTRY_DELEGATE_COVERED] = mfib_entry_delegate_fmt_covered,
+};
+
+u8 *
+format_mfib_entry_deletegate (u8 * s, va_list * args)
+{
+    mfib_entry_delegate_t *fed;
+
+    fed = va_arg (*args, mfib_entry_delegate_t *);
+
+    return (fed_formatters[fed->mfd_type](fed, s));
+}
diff --git a/src/vnet/mfib/mfib_entry_delegate.h b/src/vnet/mfib/mfib_entry_delegate.h
new file mode 100644
index 00000000000..0454a0f157a
--- /dev/null
+++ b/src/vnet/mfib/mfib_entry_delegate.h
@@ -0,0 +1,90 @@
+/*
+ * Copyright (c) 2018 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __MFIB_ENTRY_DELEGATE_T__
+#define __MFIB_ENTRY_DELEGATE_T__
+
+#include <vnet/fib/fib_node.h>
+
+/**
+ * Delegate types
+ */
+typedef enum mfib_entry_delegate_type_t_ {
+    /**
+     * Dependency list of covered entries.
+     * these are more specific entries that are interested in changes
+     * to their respective cover
+     */
+    MFIB_ENTRY_DELEGATE_COVERED,
+} mfib_entry_delegate_type_t;
+
+#define FOR_EACH_MFIB_DELEGATE(_entry, _fdt, _fed, _body)      \
+{                                                              \
+    for (_fdt = MFIB_ENTRY_DELEGATE_CHAIN_UNICAST_IP4;         \
+         _fdt <= MFIB_ENTRY_DELEGATE_ATTACHED_EXPORT;          \
+         _fdt++)                                               \
+    {                                                          \
+        _fed = mfib_entry_delegate_get(_entry, _fdt);          \
+        if (NULL != _fed) {                                    \
+            _body;                                             \
+        }                                                      \
+    }                                                          \
+}
+
+/**
+ * A Delagate is a means to implmenet the Delagation design pattern; the extension of an
+ * objects functionality through the composition of, and delgation to, other objects.
+ * These 'other' objects are delegates. Delagates are thus attached to other MFIB objects
+ * to extend their functionality.
+ */
+typedef struct mfib_entry_delegate_t_
+{
+    /**
+     * The MFIB entry object to which the delagate is attached
+     */
+    fib_node_index_t mfd_entry_index;
+
+    /**
+     * The delagate type
+     */
+    mfib_entry_delegate_type_t mfd_type;
+
+    /**
+     * A union of data for the different delegate types
+     * These delegates are stored in a sparse vector on the entry, so they
+     * must all be of the same size.
+     */
+    union
+    {
+        /**
+         * For the cover tracking. The node list;
+         */
+        fib_node_list_t mfd_list;
+    };
+} mfib_entry_delegate_t;
+
+struct mfib_entry_t_;
+
+extern void mfib_entry_delegate_remove(struct mfib_entry_t_ *mfib_entry,
+                                      mfib_entry_delegate_type_t type);
+
+extern mfib_entry_delegate_t *mfib_entry_delegate_find_or_add(struct mfib_entry_t_ *mfib_entry,
+                                                            mfib_entry_delegate_type_t fdt);
+extern mfib_entry_delegate_t *mfib_entry_delegate_get(const struct mfib_entry_t_ *mfib_entry,
+                                                    mfib_entry_delegate_type_t type);
+
+extern u8 *format_mfib_entry_deletegate(u8 * s, va_list * args);
+
+#endif
diff --git a/src/vnet/mfib/mfib_entry_src.c b/src/vnet/mfib/mfib_entry_src.c
new file mode 100644
index 00000000000..ca5c788a546
--- /dev/null
+++ b/src/vnet/mfib/mfib_entry_src.c
@@ -0,0 +1,101 @@
+/*
+ * Copyright (c) 2018 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+
+#include <vnet/mfib/mfib_entry_src.h>
+
+static mfib_entry_src_vft mfib_entry_src_vfts[MFIB_N_SOURCES];
+
+static void
+mfib_entry_src_default_deactiviate (mfib_entry_t *mfib_entry,
+                                    mfib_entry_src_t *msrc)
+{
+}
+
+static void
+mfib_entry_src_default_activiate (mfib_entry_t *mfib_entry,
+                                  mfib_entry_src_t *msrc)
+{
+}
+
+static mfib_src_res_t
+mfib_entry_src_default_cover_change (mfib_entry_t *mfib_entry,
+                                     mfib_entry_src_t *msrc)
+{
+    return (MFIB_SRC_OK);
+}
+
+static mfib_src_res_t
+mfib_entry_src_default_cover_update (mfib_entry_t *mfib_entry,
+                                     mfib_entry_src_t *msrc)
+{
+    return (MFIB_SRC_OK);
+}
+
+void
+mfib_entry_src_register (mfib_source_t source,
+                         const mfib_entry_src_vft *mvft)
+{
+    mfib_entry_src_vfts[source] = *mvft;
+}
+
+void
+mfib_entry_src_deactivate (mfib_entry_t *mfib_entry,
+                           mfib_entry_src_t *msrc)
+{
+    if (NULL != msrc)
+        mfib_entry_src_vfts[msrc->mfes_src].mev_deactivate(mfib_entry, msrc);
+}
+
+void
+mfib_entry_src_activate (mfib_entry_t *mfib_entry,
+                         mfib_entry_src_t *msrc)
+{
+    if (NULL != msrc)
+        mfib_entry_src_vfts[msrc->mfes_src].mev_activate(mfib_entry, msrc);
+}
+
+mfib_src_res_t
+mfib_entry_src_cover_change (mfib_entry_t *mfib_entry,
+                             mfib_entry_src_t *msrc)
+{
+    return (mfib_entry_src_vfts[msrc->mfes_src].mev_cover_change(mfib_entry, msrc));
+}
+
+mfib_src_res_t
+mfib_entry_src_cover_update (mfib_entry_t *mfib_entry,
+                             mfib_entry_src_t *msrc)
+{
+    return (mfib_entry_src_vfts[msrc->mfes_src].mev_cover_update(mfib_entry, msrc));
+}
+
+void
+mfib_entry_src_module_init (void)
+{
+    mfib_entry_src_vft mvft = {
+        .mev_activate = mfib_entry_src_default_activiate,
+        .mev_deactivate = mfib_entry_src_default_deactiviate,
+        .mev_cover_change = mfib_entry_src_default_cover_change,
+        .mev_cover_update = mfib_entry_src_default_cover_update,
+    };
+    mfib_source_t source;
+
+    FOREACH_MFIB_SOURCE(source)
+    {
+        mfib_entry_src_register(source, &mvft);
+    }
+
+    mfib_entry_src_rr_module_init();
+}
diff --git a/src/vnet/mfib/mfib_entry_src.h b/src/vnet/mfib/mfib_entry_src.h
new file mode 100644
index 00000000000..86752e071ec
--- /dev/null
+++ b/src/vnet/mfib/mfib_entry_src.h
@@ -0,0 +1,142 @@
+/*
+ * Copyright (c) 2018 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef __MFIB_ENTRY_SRC_H__
+#define __MFIB_ENTRY_SRC_H__
+
+#include <vnet/mfib/mfib_entry.h>
+
+/**
+ * MFIB extensions to each path
+ */
+typedef struct mfib_path_ext_t_
+{
+    mfib_itf_flags_t mfpe_flags;
+    fib_node_index_t mfpe_path;
+} mfib_path_ext_t;
+
+/**
+ * The source of an MFIB entry
+ */
+typedef struct mfib_entry_src_t_
+{
+    /**
+     * Which source this is
+     */
+    mfib_source_t mfes_src;
+
+    /**
+     * Route flags
+     */
+    mfib_entry_flags_t mfes_flags;
+
+    /**
+     * The reference count on the entry. this is a u32
+     * since there is no path-list sharing in mfib, so the number
+     * os children could be high.
+     */
+    u32 mfes_ref_count;
+
+    /**
+     * The path-list of forwarding interfaces
+     */
+    fib_node_index_t mfes_pl;
+
+    /**
+     * RPF-ID
+     */
+    fib_rpf_id_t mfes_rpf_id;
+
+    /**
+     * Hash table of path extensions
+     */
+    mfib_path_ext_t *mfes_exts;
+
+    /**
+     * Covering entry (if needed)
+     */
+    struct {
+        fib_node_index_t mfes_cover;
+        u32 mfes_sibling;
+    };
+
+    /**
+     * The hash table of all interfaces.
+     *  This is forwarding time information derived from the paths
+     *  and their extensions.
+     */
+    mfib_itf_t *mfes_itfs;
+} mfib_entry_src_t;
+
+/**
+ * signals from the sources to the caller
+ */
+typedef enum mfib_src_res_t_
+{
+    MFIB_SRC_OK,
+    MFIB_SRC_REEVALUATE,
+} mfib_src_res_t;
+
+/**
+ * A function provided by each source to be invoked when it is activated
+ */
+typedef void (*mfib_entry_src_activiate_t) (mfib_entry_t*, mfib_entry_src_t*);
+
+/**
+ * A function provided by each source to be invoked when it is deactivated
+ */
+typedef void (*mfib_entry_src_deactiviate_t) (mfib_entry_t*, mfib_entry_src_t*);
+
+/**
+ * A function provided by each source to be invoked when the cover changes
+ */
+typedef mfib_src_res_t (*mfib_entry_src_cover_change_t) (mfib_entry_t*, mfib_entry_src_t*);
+
+/**
+ * A function provided by each source to be invoked when the cover is updated
+ */
+typedef mfib_src_res_t (*mfib_entry_src_cover_update_t) (mfib_entry_t*, mfib_entry_src_t*);
+
+/**
+ * Virtual function table provided by each_source
+ */
+typedef struct mfib_entry_src_vft_t_
+{
+    mfib_entry_src_activiate_t mev_activate;
+    mfib_entry_src_deactiviate_t mev_deactivate;
+    mfib_entry_src_cover_change_t mev_cover_change;
+    mfib_entry_src_cover_update_t mev_cover_update;
+} mfib_entry_src_vft;
+
+extern void mfib_entry_src_register(mfib_source_t, const mfib_entry_src_vft*);
+
+extern void mfib_entry_src_deactivate(mfib_entry_t *mfib_entry,
+                                      mfib_entry_src_t *bsrc);
+
+extern void mfib_entry_src_activate(mfib_entry_t *mfib_entry,
+                                    mfib_entry_src_t *bsrc);
+
+extern mfib_src_res_t mfib_entry_src_cover_change(mfib_entry_t *mfib_entry,
+                                                  mfib_entry_src_t *bsrc);
+
+extern mfib_src_res_t mfib_entry_src_cover_update(mfib_entry_t *mfib_entry,
+                                                  mfib_entry_src_t *bsrc);
+
+extern mfib_entry_src_t* mfib_entry_get_best_src(const mfib_entry_t *mfib_entry);
+
+extern void mfib_entry_src_module_init(void);
+extern void mfib_entry_src_rr_module_init(void);
+
+#endif
diff --git a/src/vnet/mfib/mfib_entry_src_rr.c b/src/vnet/mfib/mfib_entry_src_rr.c
new file mode 100644
index 00000000000..4512f74a9f5
--- /dev/null
+++ b/src/vnet/mfib/mfib_entry_src_rr.c
@@ -0,0 +1,112 @@
+/*
+ * Copyright (c) 2018 Cisco and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+
+#include <vnet/mfib/mfib_entry_src.h>
+#include <vnet/mfib/mfib_entry_cover.h>
+#include <vnet/mfib/mfib_table.h>
+#include <vnet/fib/fib_path_list.h>
+
+static void
+mfib_entry_src_rr_deactiviate (mfib_entry_t *mfib_entry,
+                               mfib_entry_src_t *msrc)
+{
+    mfib_entry_t *cover;
+
+    /*
+     * remove the depednecy on the covering entry
+     */
+    if (FIB_NODE_INDEX_INVALID != msrc->mfes_cover)
+    {
+	cover = mfib_entry_get(msrc->mfes_cover);
+	mfib_entry_cover_untrack(cover, msrc->mfes_sibling);
+	msrc->mfes_cover = FIB_NODE_INDEX_INVALID;
+    }
+
+    fib_path_list_unlock(msrc->mfes_pl);
+    msrc->mfes_pl = FIB_NODE_INDEX_INVALID;
+    msrc->mfes_itfs = NULL;
+    msrc->mfes_exts = NULL;
+}
+
+static void
+mfib_entry_src_rr_activiate (mfib_entry_t *mfib_entry,
+                             mfib_entry_src_t *msrc)
+{
+    mfib_entry_src_t *csrc;
+    mfib_entry_t *cover;
+
+    msrc->mfes_cover = mfib_table_get_less_specific(mfib_entry->mfe_fib_index,
+                                                    &mfib_entry->mfe_prefix);
+
+    ASSERT(FIB_NODE_INDEX_INVALID != msrc->mfes_cover);
+
+    cover = mfib_entry_get(msrc->mfes_cover);
+
+    msrc->mfes_sibling =
+	mfib_entry_cover_track(cover, mfib_entry_get_index(mfib_entry));
+
+    csrc = mfib_entry_get_best_src(cover);
+
+    msrc->mfes_pl = csrc->mfes_pl;
+    fib_path_list_lock(msrc->mfes_pl);
+    msrc->mfes_flags = csrc->mfes_flags;
+    msrc->mfes_itfs = csrc->mfes_itfs;
+    msrc->mfes_exts = csrc->mfes_exts;
+    msrc->mfes_rpf_id = csrc->mfes_rpf_id;
+}
+
+static mfib_src_res_t
+mfib_entry_src_rr_cover_change (mfib_entry_t *mfib_entry,
+                                mfib_entry_src_t *msrc)
+{
+    mfib_entry_src_rr_deactiviate(mfib_entry, msrc);
+    mfib_entry_src_rr_activiate(mfib_entry, msrc);
+
+    return (MFIB_SRC_REEVALUATE);
+}
+
+static mfib_src_res_t
+mfib_entry_src_rr_cover_update (mfib_entry_t *mfib_entry,
+                                mfib_entry_src_t *msrc)
+{
+    /*
+     * path lists are updated (i.e. not shared) in the mfib world,
+     * so there's no need to check for a new one. but we do need to
+     * copy down any new flags and input interfaces
+     */
+    mfib_entry_t *cover;
+
+    cover = mfib_entry_get(msrc->mfes_cover);
+
+    msrc->mfes_flags = cover->mfe_flags;
+    msrc->mfes_itfs = cover->mfe_itfs;
+    msrc->mfes_rpf_id = cover->mfe_rpf_id;
+
+    return (MFIB_SRC_REEVALUATE);
+}
+
+void
+mfib_entry_src_rr_module_init (void)
+{
+    mfib_entry_src_vft mvft = {
+        .mev_activate = mfib_entry_src_rr_activiate,
+        .mev_deactivate = mfib_entry_src_rr_deactiviate,
+        .mev_cover_change = mfib_entry_src_rr_cover_change,
+        .mev_cover_update = mfib_entry_src_rr_cover_update,
+    };
+
+    mfib_entry_src_register(MFIB_SOURCE_RR, &mvft);
+}
diff --git a/src/vnet/mfib/mfib_table.c b/src/vnet/mfib/mfib_table.c
index 82431e370ee..8ae4a784684 100644
--- a/src/vnet/mfib/mfib_table.c
+++ b/src/vnet/mfib/mfib_table.c
@@ -20,6 +20,8 @@
 #include <vnet/mfib/ip4_mfib.h>
 #include <vnet/mfib/ip6_mfib.h>
 #include <vnet/mfib/mfib_entry.h>
+#include <vnet/mfib/mfib_entry_src.h>
+#include <vnet/mfib/mfib_entry_cover.h>
 #include <vnet/mfib/mfib_signal.h>
 
 mfib_table_t *
@@ -99,10 +101,41 @@ mfib_table_lookup_exact_match (u32 fib_index,
                                             prefix));
 }
 
+static fib_node_index_t
+mfib_table_get_less_specific_i (const mfib_table_t *mfib_table,
+                                const mfib_prefix_t *prefix)
+{
+    switch (prefix->fp_proto)
+    {
+    case FIB_PROTOCOL_IP4:
+        return (ip4_mfib_table_get_less_specific(&mfib_table->v4,
+                                                 &prefix->fp_src_addr.ip4,
+                                                 &prefix->fp_grp_addr.ip4,
+                                                 prefix->fp_len));
+    case FIB_PROTOCOL_IP6:
+        return (ip6_mfib_table_get_less_specific(&mfib_table->v6,
+                                                 &prefix->fp_src_addr.ip6,
+                                                 &prefix->fp_grp_addr.ip6,
+                                                 prefix->fp_len));
+    case FIB_PROTOCOL_MPLS:
+        break;
+    }
+    return (FIB_NODE_INDEX_INVALID);
+}
+
+fib_node_index_t
+mfib_table_get_less_specific (u32 fib_index,
+                              const mfib_prefix_t *prefix)
+{
+    return (mfib_table_get_less_specific_i(mfib_table_get(fib_index,
+                                                          prefix->fp_proto),
+                                           prefix));
+}
+
 static void
 mfib_table_entry_remove (mfib_table_t *mfib_table,
                          const mfib_prefix_t *prefix,
-                         fib_node_index_t fib_entry_index)
+                         fib_node_index_t mfib_entry_index)
 {
     vlib_smp_unsafe_warning();
 
@@ -127,10 +160,45 @@ mfib_table_entry_remove (mfib_table_t *mfib_table,
         break;
     }
 
-    mfib_entry_unlock(fib_entry_index);
+    mfib_entry_cover_change_notify(mfib_entry_index,
+                                   FIB_NODE_INDEX_INVALID);
+    mfib_entry_unlock(mfib_entry_index);
 }
 
 static void
+mfib_table_post_insert_actions (mfib_table_t *mfib_table,
+                                const mfib_prefix_t *prefix,
+                                fib_node_index_t mfib_entry_index)
+{
+    fib_node_index_t mfib_entry_cover_index;
+
+    /*
+     * find  the covering entry
+     */
+    mfib_entry_cover_index = mfib_table_get_less_specific_i(mfib_table,
+                                                            prefix);
+    /*
+     * the indicies are the same when the default route is first added
+     */
+    if (mfib_entry_cover_index != mfib_entry_index)
+    {
+        /*
+         * inform the covering entry that a new more specific
+         * has been inserted beneath it.
+         * If the prefix that has been inserted is a host route
+         * then it is not possible that it will be the cover for any
+         * other entry, so we can elide the walk.
+         */
+        if (!mfib_entry_is_host(mfib_entry_index))
+        {
+            mfib_entry_cover_change_notify(mfib_entry_cover_index,
+                                           mfib_entry_index);
+        }
+    }
+}
+
+
+static void
 mfib_table_entry_insert (mfib_table_t *mfib_table,
                          const mfib_prefix_t *prefix,
                          fib_node_index_t mfib_entry_index)
@@ -159,6 +227,8 @@ mfib_table_entry_insert (mfib_table_t *mfib_table,
     case FIB_PROTOCOL_MPLS:
         break;
     }
+
+    mfib_table_post_insert_actions(mfib_table, prefix, mfib_entry_index);
 }
 
 fib_node_index_t
@@ -183,7 +253,8 @@ mfib_table_entry_update (u32 fib_index,
              */
             mfib_entry_index = mfib_entry_create(fib_index, source,
                                                  prefix, rpf_id,
-                                                 entry_flags);
+                                                 entry_flags,
+                                                 INDEX_INVALID);
 
             mfib_table_entry_insert(mfib_table, prefix, mfib_entry_index);
         }
@@ -234,16 +305,23 @@ mfib_table_entry_path_update (u32 fib_index,
                                              source,
                                              prefix,
                                              MFIB_RPF_ID_NONE,
-                                             MFIB_ENTRY_FLAG_NONE);
+                                             MFIB_ENTRY_FLAG_NONE,
+                                             INDEX_INVALID);
+
+        mfib_entry_path_update(mfib_entry_index,
+                               source,
+                               rpath,
+                               itf_flags);
 
         mfib_table_entry_insert(mfib_table, prefix, mfib_entry_index);
     }
-
-    mfib_entry_path_update(mfib_entry_index,
-                           source,
-                           rpath,
-                           itf_flags);
-
+    else
+    {
+        mfib_entry_path_update(mfib_entry_index,
+                               source,
+                               rpath,
+                               itf_flags);
+    }
     return (mfib_entry_index);
 }
 
@@ -295,7 +373,7 @@ mfib_table_entry_special_add (u32 fib_index,
                               const mfib_prefix_t *prefix,
                               mfib_source_t source,
                               mfib_entry_flags_t entry_flags,
-                              index_t rep_dpo)
+                              index_t repi)
 {
     fib_node_index_t mfib_entry_index;
     mfib_table_t *mfib_table;
@@ -303,21 +381,27 @@ mfib_table_entry_special_add (u32 fib_index,
     mfib_table = mfib_table_get(fib_index, prefix->fp_proto);
     mfib_entry_index = mfib_table_lookup_exact_match_i(mfib_table, prefix);
 
+    if (INDEX_INVALID != repi)
+    {
+        entry_flags |= MFIB_ENTRY_FLAG_EXCLUSIVE;
+    }
+
     if (FIB_NODE_INDEX_INVALID == mfib_entry_index)
     {
         mfib_entry_index = mfib_entry_create(fib_index,
                                              source,
                                              prefix,
                                              MFIB_RPF_ID_NONE,
-                                             MFIB_ENTRY_FLAG_NONE);
+                                             entry_flags,
+                                             repi);
 
         mfib_table_entry_insert(mfib_table, prefix, mfib_entry_index);
     }
-
-    mfib_entry_update(mfib_entry_index, source,
-                      (MFIB_ENTRY_FLAG_EXCLUSIVE | entry_flags),
-                      MFIB_RPF_ID_NONE,
-                      rep_dpo);
+    else
+    {
+        mfib_entry_special_add(mfib_entry_index, source, entry_flags,
+                               MFIB_RPF_ID_NONE, repi);
+    }
 
     return (mfib_entry_index);
 }
@@ -380,12 +464,12 @@ void
 mfib_table_entry_delete_index (fib_node_index_t mfib_entry_index,
                                mfib_source_t source)
 {
-    mfib_prefix_t prefix;
+    const mfib_prefix_t *prefix;
 
-    mfib_entry_get_prefix(mfib_entry_index, &prefix);
+    prefix = mfib_entry_get_prefix(mfib_entry_index);
 
     mfib_table_entry_delete_i(mfib_entry_get_fib_index(mfib_entry_index),
-                              mfib_entry_index, &prefix, source);
+                              mfib_entry_index, prefix, source);
 }
 
 u32
@@ -595,6 +679,17 @@ mfib_table_lock (u32 fib_index,
     mfib_table->mft_locks[MFIB_TABLE_TOTAL_LOCKS]++;
 }
 
+u32
+mfib_table_get_n_routes (fib_node_index_t fib_index,
+                         fib_protocol_t proto)
+{
+    mfib_table_t *mfib_table;
+
+    mfib_table = mfib_table_get(fib_index, proto);
+
+    return (mfib_table->mft_total_route_counts);
+}
+
 void
 mfib_table_walk (u32 fib_index,
                  fib_protocol_t proto,
@@ -642,6 +737,7 @@ mfib_module_init (vlib_main_t * vm)
 {
     clib_error_t * error;
 
+    mfib_entry_src_module_init();
     mfib_entry_module_init();
     mfib_signal_module_init();
 
diff --git a/src/vnet/mfib/mfib_table.h b/src/vnet/mfib/mfib_table.h
index b8ade8b5cd9..89f194f254e 100644
--- a/src/vnet/mfib/mfib_table.h
+++ b/src/vnet/mfib/mfib_table.h
@@ -405,6 +405,22 @@ extern u32 mfib_table_get_num_entries(u32 fib_index,
 
 /**
  * @brief
+ *  Get the less specific (covering) prefix
+ *
+ * @param fib_index
+ *  The index of the FIB
+ *
+ * @param prefix
+ *  The prefix to lookup
+ *
+ * @return
+ *  The index of the less specific fib_entry_t.
+ */
+extern fib_node_index_t mfib_table_get_less_specific(u32 fib_index,
+						    const mfib_prefix_t *prefix);
+
+/**
+ * @brief
  * Get a pointer to a FIB table
  */
 extern mfib_table_t *mfib_table_get(fib_node_index_t index,
@@ -430,4 +446,11 @@ extern void mfib_table_walk(u32 fib_index,
  */
 extern u8 * format_mfib_table_memory(u8 * s, va_list * args);
 
+/**
+ * To assit UT
+ */
+extern u32 mfib_table_get_n_routes(fib_node_index_t index,
+                                   fib_protocol_t proto);
+
+
 #endif
diff --git a/src/vnet/mfib/mfib_types.c b/src/vnet/mfib/mfib_types.c
index a3ed46a4b8d..19583ea18f4 100644
--- a/src/vnet/mfib/mfib_types.c
+++ b/src/vnet/mfib/mfib_types.c
@@ -26,6 +26,82 @@ static const char *mfib_flag_names_long[] = MFIB_ENTRY_NAMES_LONG;
 static const char *mfib_itf_flag_long_names[] = MFIB_ITF_NAMES_LONG;
 static const char *mfib_itf_flag_names[] = MFIB_ITF_NAMES_SHORT;
 
+int
+mfib_prefix_is_cover (const mfib_prefix_t *p1,
+                      const mfib_prefix_t *p2)
+{
+    if (!ip46_address_is_equal(&p1->fp_src_addr, &p2->fp_src_addr))
+        return (0);
+
+    switch (p1->fp_proto)
+    {
+    case FIB_PROTOCOL_IP4:
+	return (ip4_destination_matches_route(&ip4_main,
+					      &p1->fp_grp_addr.ip4,
+					      &p2->fp_grp_addr.ip4,
+					      p1->fp_len));
+    case FIB_PROTOCOL_IP6:
+	return (ip6_destination_matches_route(&ip6_main,
+					      &p1->fp_grp_addr.ip6,
+					      &p2->fp_grp_addr.ip6,
+					      p1->fp_len));
+    case FIB_PROTOCOL_MPLS:
+	break;
+    }
+    return (0);
+}
+
+int
+mfib_prefix_is_host (const mfib_prefix_t *pfx)
+{
+    switch (pfx->fp_proto)
+    {
+    case FIB_PROTOCOL_IP4:
+	return (64 == pfx->fp_len);
+    case FIB_PROTOCOL_IP6:
+	return (256 == pfx->fp_len);
+    case FIB_PROTOCOL_MPLS:
+        ASSERT(0);
+	break;
+    }
+    return (0);
+}
+
+fib_forward_chain_type_t
+mfib_forw_chain_type_from_dpo_proto (dpo_proto_t proto)
+{
+    switch (proto)
+    {
+    case DPO_PROTO_IP4:
+	return (FIB_FORW_CHAIN_TYPE_MCAST_IP4);
+    case DPO_PROTO_IP6:
+	return (FIB_FORW_CHAIN_TYPE_MCAST_IP6);
+    case DPO_PROTO_MPLS:
+    case DPO_PROTO_ETHERNET:
+    case DPO_PROTO_NSH:
+    case DPO_PROTO_BIER:
+        break;
+    }
+    ASSERT(0);
+    return (FIB_FORW_CHAIN_TYPE_MCAST_IP4);
+}
+
+fib_forward_chain_type_t
+mfib_forw_chain_type_from_fib_proto (fib_protocol_t proto)
+{
+    switch (proto)
+    {
+    case FIB_PROTOCOL_IP4:
+	return (FIB_FORW_CHAIN_TYPE_MCAST_IP4);
+    case FIB_PROTOCOL_IP6:
+	return (FIB_FORW_CHAIN_TYPE_MCAST_IP6);
+    case FIB_PROTOCOL_MPLS:
+        break;
+    }
+    ASSERT(0);
+    return (FIB_FORW_CHAIN_TYPE_MCAST_IP4);
+}
+
 u8 *
 format_mfib_prefix (u8 * s, va_list * args)
 {
diff --git a/src/vnet/mfib/mfib_types.h b/src/vnet/mfib/mfib_types.h
index 58b7c987ab7..aea4a768a0b 100644
--- a/src/vnet/mfib/mfib_types.h
+++ b/src/vnet/mfib/mfib_types.h
@@ -169,10 +169,10 @@ typedef enum mfib_source_t_
     MFIB_SOURCE_SRv6,
     MFIB_SOURCE_GTPU,
     MFIB_SOURCE_VXLAN_GPE,
-    MFIB_SOURCE_RR,
     MFIB_SOURCE_GENEVE,
     MFIB_SOURCE_IGMP,
     MFIB_SOURCE_VXLAN_GBP,
+    MFIB_SOURCE_RR,
     MFIB_SOURCE_DEFAULT_ROUTE,
 } mfib_source_t;
 
@@ -186,14 +186,20 @@ typedef enum mfib_source_t_
     [MFIB_SOURCE_SRv6] = "SRv6",                   \
     [MFIB_SOURCE_GTPU] = "GTPU",                   \
     [MFIB_SOURCE_VXLAN_GPE] = "VXLAN-GPE",         \
-    [MFIB_SOURCE_RR] = "Recursive-resolution",     \
     [MFIB_SOURCE_GENEVE] = "Geneve",               \
     [MFIB_SOURCE_IGMP] = "IGMP",                   \
     [MFIB_SOURCE_VXLAN_GBP] = "VXLAN-GBP",         \
+    [MFIB_SOURCE_RR] = "Recursive-resolution",     \
     [MFIB_SOURCE_DEFAULT_ROUTE] = "Default Route", \
 }
 
-#define MFIB_N_SOURCES (MFIB_SOURCE_DEFAULT_ROUTE)
+#define FOREACH_MFIB_SOURCE(_ms)                \
+    for (_ms = MFIB_SOURCE_SPECIAL;             \
+         _ms <= MFIB_SOURCE_DEFAULT_ROUTE;      \
+         _ms++)
+
+#define MFIB_N_SOURCES (MFIB_SOURCE_DEFAULT_ROUTE + 1)
+#define MFIB_SOURCE_NONE (MFIB_SOURCE_DEFAULT_ROUTE + 1)
 
 /**
  * \brief Compare two prefixes for equality
@@ -209,5 +215,20 @@ extern uword unformat_mfib_itf_flags(unformat_input_t * input,
                                      va_list * args);
 extern uword unformat_mfib_entry_flags(unformat_input_t * input,
                                        va_list * args);
+/**
+ * \brief Compare two prefixes for covering relationship
+ *
+ * \return non-zero if the first prefix is a cover for the second
+ */
+extern int mfib_prefix_is_cover(const mfib_prefix_t *p1,
+                                const mfib_prefix_t *p2);
+
+/**
+ * \brief Return true is the prefix is a host prefix
+ */
+extern int mfib_prefix_is_host(const mfib_prefix_t *p);
+
+extern fib_forward_chain_type_t mfib_forw_chain_type_from_dpo_proto(dpo_proto_t proto);
+extern fib_forward_chain_type_t mfib_forw_chain_type_from_fib_proto(fib_protocol_t proto);
 
 #endif