Dual Loop Lookup and Adj-midchain

Change-Id: Id78b6dc637626c01b64bd5868d22425a30318348
Signed-off-by: Neale Ranns <neale.ranns@cisco.com>
diff --git a/vnet/vnet/adj/adj_midchain.c b/vnet/vnet/adj/adj_midchain.c
index 136f69b..c420155 100644
--- a/vnet/vnet/adj/adj_midchain.c
+++ b/vnet/vnet/adj/adj_midchain.c
@@ -64,10 +64,95 @@
 	/* set up to enqueue to our disposition with index = next_index */
 	vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
 
-	/*
-	 * FIXME DUAL LOOP
-	 */
 
+	while (n_left_from >= 4 && n_left_to_next > 2)
+	{
+	    u32 bi0, adj_index0, next0;
+	    const ip_adjacency_t * adj0;
+	    const dpo_id_t *dpo0;
+	    vlib_buffer_t * b0;
+	    u32 bi1, adj_index1, next1;
+	    const ip_adjacency_t * adj1;
+	    const dpo_id_t *dpo1;
+	    vlib_buffer_t * b1;
+
+	    /* Prefetch next iteration. */
+	    {
+		vlib_buffer_t * p2, * p3;
+
+		p2 = vlib_get_buffer (vm, from[2]);
+		p3 = vlib_get_buffer (vm, from[3]);
+
+		vlib_prefetch_buffer_header (p2, LOAD);
+		vlib_prefetch_buffer_header (p3, LOAD);
+
+		CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE);
+		CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE);
+	    }
+
+	    bi0 = from[0];
+	    to_next[0] = bi0;
+	    bi1 = from[1];
+	    to_next[1] = bi1;
+
+	    from += 2;
+	    to_next += 2;
+	    n_left_from -= 2;
+	    n_left_to_next -= 2;
+
+	    b0 = vlib_get_buffer(vm, bi0);
+	    b1 = vlib_get_buffer(vm, bi1);
+
+	    /* Follow the DPO on which the midchain is stacked */
+	    adj_index0 = vnet_buffer(b0)->ip.adj_index[VLIB_TX];
+	    adj_index1 = vnet_buffer(b1)->ip.adj_index[VLIB_TX];
+
+	    adj0 = adj_get(adj_index0);
+	    adj1 = adj_get(adj_index1);
+
+	    dpo0 = &adj0->sub_type.midchain.next_dpo;
+	    dpo1 = &adj1->sub_type.midchain.next_dpo;
+
+	    next0 = dpo0->dpoi_next_node;
+	    next1 = dpo1->dpoi_next_node;
+
+	    vnet_buffer(b1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index;
+	    vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index;
+
+	    if (interface_count)
+	    {
+		vlib_increment_combined_counter (im->combined_sw_if_counters
+						 + VNET_INTERFACE_COUNTER_TX,
+						 cpu_index,
+						 adj0->rewrite_header.sw_if_index,
+						 1,
+						 vlib_buffer_length_in_chain (vm, b0));
+		vlib_increment_combined_counter (im->combined_sw_if_counters
+						 + VNET_INTERFACE_COUNTER_TX,
+						 cpu_index,
+						 adj1->rewrite_header.sw_if_index,
+						 1,
+						 vlib_buffer_length_in_chain (vm, b1));
+	    }
+
+	    if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+		adj_midchain_tx_trace_t *tr = vlib_add_trace (vm, node,
+							      b0, sizeof (*tr));
+		tr->ai = adj_index0;
+	    }
+	    if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+		adj_midchain_tx_trace_t *tr = vlib_add_trace (vm, node,
+							      b1, sizeof (*tr));
+		tr->ai = adj_index1;
+	    }
+
+	    vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
+					     to_next, n_left_to_next,
+					     bi0, bi1,
+					     next0, next1);
+	}
 	while (n_left_from > 0 && n_left_to_next > 0)
 	{
 	    u32 bi0, adj_index0, next0;
diff --git a/vnet/vnet/dpo/lookup_dpo.c b/vnet/vnet/dpo/lookup_dpo.c
index ab3a7b9..c9e0fca 100644
--- a/vnet/vnet/dpo/lookup_dpo.c
+++ b/vnet/vnet/dpo/lookup_dpo.c
@@ -292,95 +292,246 @@
 
         vlib_get_next_frame(vm, node, next_index, to_next, n_left_to_next);
 
-        while (n_left_from > 0 && n_left_to_next > 0)
-        {
+	while (n_left_from >= 4 && n_left_to_next > 2)
+	{
 	    u32 bi0, lkdi0, lbi0, fib_index0, next0, hash_c0;
 	    flow_hash_config_t flow_hash_config0;
-            const ip4_address_t *input_addr;
-            const load_balance_t *lb0;
-            const lookup_dpo_t * lkd0;
-            const ip4_header_t * ip0;
-            const dpo_id_t *dpo0;
-            vlib_buffer_t * b0;
+	    const ip4_address_t *input_addr0;
+	    const load_balance_t *lb0;
+	    const lookup_dpo_t * lkd0;
+	    const ip4_header_t * ip0;
+	    const dpo_id_t *dpo0;
+	    vlib_buffer_t * b0;
+	    u32 bi1, lkdi1, lbi1, fib_index1, next1, hash_c1;
+	    flow_hash_config_t flow_hash_config1;
+	    const ip4_address_t *input_addr1;
+	    const load_balance_t *lb1;
+	    const lookup_dpo_t * lkd1;
+	    const ip4_header_t * ip1;
+	    const dpo_id_t *dpo1;
+	    vlib_buffer_t * b1;
 
-            bi0 = from[0];
-            to_next[0] = bi0;
-            from += 1;
-            to_next += 1;
-            n_left_from -= 1;
-            n_left_to_next -= 1;
+	    /* Prefetch next iteration. */
+	    {
+		vlib_buffer_t * p2, * p3;
 
-            b0 = vlib_get_buffer (vm, bi0);
-            ip0 = vlib_buffer_get_current (b0);
+		p2 = vlib_get_buffer (vm, from[2]);
+		p3 = vlib_get_buffer (vm, from[3]);
 
-            /* dst lookup was done by ip4 lookup */
-            lkdi0 = vnet_buffer(b0)->ip.adj_index[VLIB_TX];
-            lkd0 = lookup_dpo_get(lkdi0);
+		vlib_prefetch_buffer_header (p2, LOAD);
+		vlib_prefetch_buffer_header (p3, LOAD);
 
-            /*
-             * choose between a lookup using the fib index in the DPO
-             * or getting the FIB index from the interface.
-             */
-            if (table_from_interface)
-            {
-                fib_index0 = 
-                    ip4_fib_table_get_index_for_sw_if_index(
-                        vnet_buffer(b0)->sw_if_index[VLIB_RX]);
-            }
-            else
-            {
-                fib_index0 = lkd0->lkd_fib_index;
-            }
+		CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE);
+		CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE);
+	    }
 
-            /*
-             * choose between a source or destination address lookup in the table
-             */
-            if (input_src_addr)
-            {
-                input_addr = &ip0->src_address;
-            }
-            else
-            {
-                input_addr = &ip0->dst_address;
-            }
+	    bi0 = from[0];
+	    to_next[0] = bi0;
+	    bi1 = from[1];
+	    to_next[1] = bi1;
+	    from += 2;
+	    to_next += 2;
+	    n_left_from -= 2;
+	    n_left_to_next -= 2;
 
-            /* do lookup */
-            ip4_src_fib_lookup_one (fib_index0, input_addr, &lbi0);
-            lb0 = load_balance_get(lbi0);
+	    b0 = vlib_get_buffer (vm, bi0);
+	    ip0 = vlib_buffer_get_current (b0);
+	    b1 = vlib_get_buffer (vm, bi1);
+	    ip1 = vlib_buffer_get_current (b1);
+
+	    /* dst lookup was done by ip4 lookup */
+	    lkdi0 = vnet_buffer(b0)->ip.adj_index[VLIB_TX];
+	    lkdi1 = vnet_buffer(b1)->ip.adj_index[VLIB_TX];
+	    lkd0 = lookup_dpo_get(lkdi0);
+	    lkd1 = lookup_dpo_get(lkdi1);
+
+	    /*
+	     * choose between a lookup using the fib index in the DPO
+	     * or getting the FIB index from the interface.
+	     */
+	    if (table_from_interface)
+	    {
+		fib_index0 =
+		    ip4_fib_table_get_index_for_sw_if_index(
+			vnet_buffer(b0)->sw_if_index[VLIB_RX]);
+		fib_index1 =
+		    ip4_fib_table_get_index_for_sw_if_index(
+			vnet_buffer(b1)->sw_if_index[VLIB_RX]);
+	    }
+	    else
+	    {
+		fib_index0 = lkd0->lkd_fib_index;
+		fib_index1 = lkd1->lkd_fib_index;
+	    }
+
+	    /*
+	     * choose between a source or destination address lookup in the table
+	     */
+	    if (input_src_addr)
+	    {
+		input_addr0 = &ip0->src_address;
+		input_addr1 = &ip1->src_address;
+	    }
+	    else
+	    {
+		input_addr0 = &ip0->dst_address;
+		input_addr1 = &ip1->dst_address;
+	    }
+
+	    /* do lookup */
+	    ip4_src_fib_lookup_one (fib_index0, input_addr0, &lbi0);
+	    ip4_src_fib_lookup_one (fib_index1, input_addr1, &lbi1);
+	    lb0 = load_balance_get(lbi0);
+	    lb1 = load_balance_get(lbi1);
+
+	    /* Use flow hash to compute multipath adjacency. */
+	    hash_c0 = vnet_buffer (b0)->ip.flow_hash = 0;
+	    hash_c1 = vnet_buffer (b1)->ip.flow_hash = 0;
+
+	    if (PREDICT_FALSE (lb0->lb_n_buckets > 1))
+	    {
+		flow_hash_config0 = lb0->lb_hash_config;
+		hash_c0 = vnet_buffer (b0)->ip.flow_hash =
+		    ip4_compute_flow_hash (ip0, flow_hash_config0);
+	    }
+
+	    if (PREDICT_FALSE (lb1->lb_n_buckets > 1))
+	    {
+		flow_hash_config1 = lb1->lb_hash_config;
+		hash_c1 = vnet_buffer (b1)->ip.flow_hash =
+		    ip4_compute_flow_hash (ip1, flow_hash_config1);
+	    }
+
+	    dpo0 = load_balance_get_bucket_i(lb0,
+					     (hash_c0 &
+					      (lb0->lb_n_buckets_minus_1)));
+	    dpo1 = load_balance_get_bucket_i(lb1,
+					     (hash_c1 &
+					      (lb1->lb_n_buckets_minus_1)));
+
+	    next0 = dpo0->dpoi_next_node;
+	    next1 = dpo1->dpoi_next_node;
+	    vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index;
+	    vnet_buffer(b1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index;
+
+	    vlib_increment_combined_counter
+		(cm, cpu_index, lbi0, 1,
+		 vlib_buffer_length_in_chain (vm, b0));
+	    vlib_increment_combined_counter
+		(cm, cpu_index, lbi1, 1,
+		 vlib_buffer_length_in_chain (vm, b1));
+
+	    if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+		lookup_trace_t *tr = vlib_add_trace (vm, node,
+						     b0, sizeof (*tr));
+		tr->fib_index = fib_index0;
+		tr->lbi = lbi0;
+		tr->addr.ip4 = *input_addr0;
+	    }
+	    if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+		lookup_trace_t *tr = vlib_add_trace (vm, node,
+						     b1, sizeof (*tr));
+		tr->fib_index = fib_index1;
+		tr->lbi = lbi1;
+		tr->addr.ip4 = *input_addr1;
+	    }
+
+	    vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
+					     to_next, n_left_to_next,
+					     bi0, bi1, next0, next1);
+	}
+
+	while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	    u32 bi0, lkdi0, lbi0, fib_index0, next0, hash_c0;
+	    flow_hash_config_t flow_hash_config0;
+	    const ip4_address_t *input_addr;
+	    const load_balance_t *lb0;
+	    const lookup_dpo_t * lkd0;
+	    const ip4_header_t * ip0;
+	    const dpo_id_t *dpo0;
+	    vlib_buffer_t * b0;
+
+	    bi0 = from[0];
+	    to_next[0] = bi0;
+	    from += 1;
+	    to_next += 1;
+	    n_left_from -= 1;
+	    n_left_to_next -= 1;
+
+	    b0 = vlib_get_buffer (vm, bi0);
+	    ip0 = vlib_buffer_get_current (b0);
+
+	    /* dst lookup was done by ip4 lookup */
+	    lkdi0 = vnet_buffer(b0)->ip.adj_index[VLIB_TX];
+	    lkd0 = lookup_dpo_get(lkdi0);
+
+	    /*
+	     * choose between a lookup using the fib index in the DPO
+	     * or getting the FIB index from the interface.
+	     */
+	    if (table_from_interface)
+	    {
+		fib_index0 =
+		    ip4_fib_table_get_index_for_sw_if_index(
+			vnet_buffer(b0)->sw_if_index[VLIB_RX]);
+	    }
+	    else
+	    {
+		fib_index0 = lkd0->lkd_fib_index;
+	    }
+
+	    /*
+	     * choose between a source or destination address lookup in the table
+	     */
+	    if (input_src_addr)
+	    {
+		input_addr = &ip0->src_address;
+	    }
+	    else
+	    {
+		input_addr = &ip0->dst_address;
+	    }
+
+	    /* do lookup */
+	    ip4_src_fib_lookup_one (fib_index0, input_addr, &lbi0);
+	    lb0 = load_balance_get(lbi0);
 
 	    /* Use flow hash to compute multipath adjacency. */
 	    hash_c0 = vnet_buffer (b0)->ip.flow_hash = 0;
 
 	    if (PREDICT_FALSE (lb0->lb_n_buckets > 1))
-            {
+	    {
 		flow_hash_config0 = lb0->lb_hash_config;
 		hash_c0 = vnet_buffer (b0)->ip.flow_hash =
 		    ip4_compute_flow_hash (ip0, flow_hash_config0);
-            }
+	    }
 
 	    dpo0 = load_balance_get_bucket_i(lb0,
 					     (hash_c0 &
 					      (lb0->lb_n_buckets_minus_1)));
 
-            next0 = dpo0->dpoi_next_node;
-            vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index;
+	    next0 = dpo0->dpoi_next_node;
+	    vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index;
 
-            vlib_increment_combined_counter
-                (cm, cpu_index, lbi0, 1,
-                 vlib_buffer_length_in_chain (vm, b0));
+	    vlib_increment_combined_counter
+		(cm, cpu_index, lbi0, 1,
+		 vlib_buffer_length_in_chain (vm, b0));
 
-            if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) 
-            {
-                lookup_trace_t *tr = vlib_add_trace (vm, node, 
-                                                     b0, sizeof (*tr));
-                tr->fib_index = fib_index0;
-                tr->lbi = lbi0;
-                tr->addr.ip4 = *input_addr;
-            }
+	    if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+		lookup_trace_t *tr = vlib_add_trace (vm, node,
+						     b0, sizeof (*tr));
+		tr->fib_index = fib_index0;
+		tr->lbi = lbi0;
+		tr->addr.ip4 = *input_addr;
+	    }
 
-            vlib_validate_buffer_enqueue_x1(vm, node, next_index, to_next,
-                                            n_left_to_next, bi0, next0);
-        }
+	    vlib_validate_buffer_enqueue_x1(vm, node, next_index, to_next,
+					    n_left_to_next, bi0, next0);
+	}
         vlib_put_next_frame (vm, node, next_index, n_left_to_next);
     }
     return from_frame->n_vectors;
@@ -474,96 +625,249 @@
 
         vlib_get_next_frame(vm, node, next_index, to_next, n_left_to_next);
 
-        while (n_left_from > 0 && n_left_to_next > 0)
-        {
-            u32 bi0, lkdi0, lbi0, fib_index0, next0, hash_c0;
+	while (n_left_from >= 4 && n_left_to_next > 2)
+	{
+	    u32 bi0, lkdi0, lbi0, fib_index0, next0, hash_c0;
 	    flow_hash_config_t flow_hash_config0;
-            const ip6_address_t *input_addr0;
-            const load_balance_t *lb0;
-            const lookup_dpo_t * lkd0;
-            const ip6_header_t * ip0;
-            const dpo_id_t *dpo0;
-            vlib_buffer_t * b0;
+	    const ip6_address_t *input_addr0;
+	    const load_balance_t *lb0;
+	    const lookup_dpo_t * lkd0;
+	    const ip6_header_t * ip0;
+	    const dpo_id_t *dpo0;
+	    vlib_buffer_t * b0;
+	    u32 bi1, lkdi1, lbi1, fib_index1, next1, hash_c1;
+	    flow_hash_config_t flow_hash_config1;
+	    const ip6_address_t *input_addr1;
+	    const load_balance_t *lb1;
+	    const lookup_dpo_t * lkd1;
+	    const ip6_header_t * ip1;
+	    const dpo_id_t *dpo1;
+	    vlib_buffer_t * b1;
 
-            bi0 = from[0];
-            to_next[0] = bi0;
-            from += 1;
-            to_next += 1;
-            n_left_from -= 1;
-            n_left_to_next -= 1;
+	    /* Prefetch next iteration. */
+	    {
+		vlib_buffer_t * p2, * p3;
 
-            b0 = vlib_get_buffer (vm, bi0);
-            ip0 = vlib_buffer_get_current (b0);
+		p2 = vlib_get_buffer (vm, from[2]);
+		p3 = vlib_get_buffer (vm, from[3]);
 
-            /* dst lookup was done by ip6 lookup */
-            lkdi0 = vnet_buffer(b0)->ip.adj_index[VLIB_TX];
-            lkd0 = lookup_dpo_get(lkdi0);
+		vlib_prefetch_buffer_header (p2, LOAD);
+		vlib_prefetch_buffer_header (p3, LOAD);
 
-            /*
-             * choose between a lookup using the fib index in the DPO
-             * or getting the FIB index from the interface.
-             */
-            if (table_from_interface)
-            {
-                fib_index0 =
-                    ip4_fib_table_get_index_for_sw_if_index(
-                        vnet_buffer(b0)->sw_if_index[VLIB_RX]);
-            }
-            else
-            {
-                fib_index0 = lkd0->lkd_fib_index;
-            }
+		CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE);
+		CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE);
+	    }
 
-            /*
-             * choose between a source or destination address lookup in the table
-             */
-            if (input_src_addr)
-            {
-                input_addr0 = &ip0->src_address;
-            }
-            else
-            {
-                input_addr0 = &ip0->dst_address;
-            }
+	    bi0 = from[0];
+	    to_next[0] = bi0;
+	    bi1 = from[1];
+	    to_next[1] = bi1;
+	    from += 2;
+	    to_next += 2;
+	    n_left_from -= 2;
+	    n_left_to_next -= 2;
 
-            /* do src lookup */
-            lbi0 = ip6_fib_table_fwding_lookup(&ip6_main,
-                                               fib_index0,
-                                               input_addr0);
-            lb0 = load_balance_get(lbi0);
+	    b0 = vlib_get_buffer (vm, bi0);
+	    ip0 = vlib_buffer_get_current (b0);
+	    b1 = vlib_get_buffer (vm, bi1);
+	    ip1 = vlib_buffer_get_current (b1);
+
+	    /* dst lookup was done by ip6 lookup */
+	    lkdi0 = vnet_buffer(b0)->ip.adj_index[VLIB_TX];
+	    lkdi1 = vnet_buffer(b1)->ip.adj_index[VLIB_TX];
+	    lkd0 = lookup_dpo_get(lkdi0);
+	    lkd1 = lookup_dpo_get(lkdi1);
+
+	    /*
+	     * choose between a lookup using the fib index in the DPO
+	     * or getting the FIB index from the interface.
+	     */
+	    if (table_from_interface)
+	    {
+		fib_index0 =
+		    ip4_fib_table_get_index_for_sw_if_index(
+			vnet_buffer(b0)->sw_if_index[VLIB_RX]);
+		fib_index1 =
+		    ip4_fib_table_get_index_for_sw_if_index(
+			vnet_buffer(b1)->sw_if_index[VLIB_RX]);
+	    }
+	    else
+	    {
+		fib_index0 = lkd0->lkd_fib_index;
+		fib_index1 = lkd1->lkd_fib_index;
+	    }
+
+	    /*
+	     * choose between a source or destination address lookup in the table
+	     */
+	    if (input_src_addr)
+	    {
+		input_addr0 = &ip0->src_address;
+		input_addr1 = &ip1->src_address;
+	    }
+	    else
+	    {
+		input_addr0 = &ip0->dst_address;
+		input_addr1 = &ip1->dst_address;
+	    }
+
+	    /* do src lookup */
+	    lbi0 = ip6_fib_table_fwding_lookup(&ip6_main,
+					       fib_index0,
+					       input_addr0);
+	    lbi1 = ip6_fib_table_fwding_lookup(&ip6_main,
+					       fib_index1,
+					       input_addr1);
+	    lb0 = load_balance_get(lbi0);
+	    lb1 = load_balance_get(lbi1);
+
+	    /* Use flow hash to compute multipath adjacency. */
+	    hash_c0 = vnet_buffer (b0)->ip.flow_hash = 0;
+	    hash_c1 = vnet_buffer (b1)->ip.flow_hash = 0;
+
+	    if (PREDICT_FALSE (lb0->lb_n_buckets > 1))
+	    {
+		flow_hash_config0 = lb0->lb_hash_config;
+		hash_c0 = vnet_buffer (b0)->ip.flow_hash =
+		    ip6_compute_flow_hash (ip0, flow_hash_config0);
+	    }
+
+	    if (PREDICT_FALSE (lb1->lb_n_buckets > 1))
+	    {
+		flow_hash_config1 = lb1->lb_hash_config;
+		hash_c1 = vnet_buffer (b1)->ip.flow_hash =
+		    ip6_compute_flow_hash (ip1, flow_hash_config1);
+	    }
+
+	    dpo0 = load_balance_get_bucket_i(lb0,
+					     (hash_c0 &
+					      (lb0->lb_n_buckets_minus_1)));
+	    dpo1 = load_balance_get_bucket_i(lb1,
+					     (hash_c1 &
+					      (lb1->lb_n_buckets_minus_1)));
+
+	    next0 = dpo0->dpoi_next_node;
+	    next1 = dpo1->dpoi_next_node;
+	    vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index;
+	    vnet_buffer(b1)->ip.adj_index[VLIB_TX] = dpo1->dpoi_index;
+
+	    vlib_increment_combined_counter
+		(cm, cpu_index, lbi0, 1,
+		 vlib_buffer_length_in_chain (vm, b0));
+	    vlib_increment_combined_counter
+		(cm, cpu_index, lbi1, 1,
+		 vlib_buffer_length_in_chain (vm, b1));
+
+	    if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+		lookup_trace_t *tr = vlib_add_trace (vm, node,
+						     b0, sizeof (*tr));
+		tr->fib_index = fib_index0;
+		tr->lbi = lbi0;
+		tr->addr.ip6 = *input_addr0;
+	    }
+	    if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+		lookup_trace_t *tr = vlib_add_trace (vm, node,
+						     b1, sizeof (*tr));
+		tr->fib_index = fib_index1;
+		tr->lbi = lbi1;
+		tr->addr.ip6 = *input_addr1;
+	    }
+	    vlib_validate_buffer_enqueue_x2(vm, node, next_index, to_next,
+					    n_left_to_next, bi0, bi1,
+					    next0, next1);
+	}
+	while (n_left_from > 0 && n_left_to_next > 0)
+	{
+	    u32 bi0, lkdi0, lbi0, fib_index0, next0, hash_c0;
+	    flow_hash_config_t flow_hash_config0;
+	    const ip6_address_t *input_addr0;
+	    const load_balance_t *lb0;
+	    const lookup_dpo_t * lkd0;
+	    const ip6_header_t * ip0;
+	    const dpo_id_t *dpo0;
+	    vlib_buffer_t * b0;
+
+	    bi0 = from[0];
+	    to_next[0] = bi0;
+	    from += 1;
+	    to_next += 1;
+	    n_left_from -= 1;
+	    n_left_to_next -= 1;
+
+	    b0 = vlib_get_buffer (vm, bi0);
+	    ip0 = vlib_buffer_get_current (b0);
+
+	    /* dst lookup was done by ip6 lookup */
+	    lkdi0 = vnet_buffer(b0)->ip.adj_index[VLIB_TX];
+	    lkd0 = lookup_dpo_get(lkdi0);
+
+	    /*
+	     * choose between a lookup using the fib index in the DPO
+	     * or getting the FIB index from the interface.
+	     */
+	    if (table_from_interface)
+	    {
+		fib_index0 =
+		    ip4_fib_table_get_index_for_sw_if_index(
+			vnet_buffer(b0)->sw_if_index[VLIB_RX]);
+	    }
+	    else
+	    {
+		fib_index0 = lkd0->lkd_fib_index;
+	    }
+
+	    /*
+	     * choose between a source or destination address lookup in the table
+	     */
+	    if (input_src_addr)
+	    {
+		input_addr0 = &ip0->src_address;
+	    }
+	    else
+	    {
+		input_addr0 = &ip0->dst_address;
+	    }
+
+	    /* do src lookup */
+	    lbi0 = ip6_fib_table_fwding_lookup(&ip6_main,
+					       fib_index0,
+					       input_addr0);
+	    lb0 = load_balance_get(lbi0);
 
 	    /* Use flow hash to compute multipath adjacency. */
 	    hash_c0 = vnet_buffer (b0)->ip.flow_hash = 0;
 
 	    if (PREDICT_FALSE (lb0->lb_n_buckets > 1))
-            {
+	    {
 		flow_hash_config0 = lb0->lb_hash_config;
 		hash_c0 = vnet_buffer (b0)->ip.flow_hash =
 		    ip6_compute_flow_hash (ip0, flow_hash_config0);
-            }
+	    }
 
 	    dpo0 = load_balance_get_bucket_i(lb0,
 					     (hash_c0 &
 					      (lb0->lb_n_buckets_minus_1)));
 
-            next0 = dpo0->dpoi_next_node;
-            vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index;
+	    next0 = dpo0->dpoi_next_node;
+	    vnet_buffer(b0)->ip.adj_index[VLIB_TX] = dpo0->dpoi_index;
 
-            vlib_increment_combined_counter
-                (cm, cpu_index, lbi0, 1,
-                 vlib_buffer_length_in_chain (vm, b0));
+	    vlib_increment_combined_counter
+		(cm, cpu_index, lbi0, 1,
+		 vlib_buffer_length_in_chain (vm, b0));
 
-            if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) 
-            {
-                lookup_trace_t *tr = vlib_add_trace (vm, node, 
-                                                     b0, sizeof (*tr));
-                tr->fib_index = fib_index0;
-                tr->lbi = lbi0;
-                tr->addr.ip6 = *input_addr0;
-            }
-            vlib_validate_buffer_enqueue_x1(vm, node, next_index, to_next,
-                                            n_left_to_next, bi0, next0);
-        }
+	    if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
+	    {
+		lookup_trace_t *tr = vlib_add_trace (vm, node,
+						     b0, sizeof (*tr));
+		tr->fib_index = fib_index0;
+		tr->lbi = lbi0;
+		tr->addr.ip6 = *input_addr0;
+	    }
+	    vlib_validate_buffer_enqueue_x1(vm, node, next_index, to_next,
+					    n_left_to_next, bi0, next0);
+	}
         vlib_put_next_frame (vm, node, next_index, n_left_to_next);
     }
     return from_frame->n_vectors;