diff options
Diffstat (limited to 'vnet/vnet/mpls')
-rw-r--r-- | vnet/vnet/mpls/interface.c | 992 | ||||
-rw-r--r-- | vnet/vnet/mpls/mpls.c | 451 | ||||
-rw-r--r-- | vnet/vnet/mpls/mpls.h | 78 | ||||
-rw-r--r-- | vnet/vnet/mpls/mpls_lookup.c | 27 | ||||
-rw-r--r-- | vnet/vnet/mpls/mpls_tunnel.c | 779 | ||||
-rw-r--r-- | vnet/vnet/mpls/mpls_tunnel.h | 98 | ||||
-rw-r--r-- | vnet/vnet/mpls/policy_encap.c | 181 |
7 files changed, 1033 insertions, 1573 deletions
diff --git a/vnet/vnet/mpls/interface.c b/vnet/vnet/mpls/interface.c index a0f6f2f2474..692a2d1eb62 100644 --- a/vnet/vnet/mpls/interface.c +++ b/vnet/vnet/mpls/interface.c @@ -22,243 +22,6 @@ #include <vnet/adj/adj_midchain.h> #include <vnet/dpo/classify_dpo.h> -/* manually added to the interface output node */ -#define MPLS_ETH_OUTPUT_NEXT_OUTPUT 1 - -static uword -mpls_eth_interface_tx (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - mpls_main_t * gm = &mpls_main; - vnet_main_t * vnm = gm->vnet_main; - u32 next_index; - u32 * from, * to_next, n_left_from, n_left_to_next; - - /* Vector of buffer / pkt indices we're supposed to process */ - from = vlib_frame_vector_args (frame); - - /* Number of buffers / pkts */ - n_left_from = frame->n_vectors; - - /* Speculatively send the first buffer to the last disposition we used */ - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - /* 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); - - /* - * As long as we have enough pkts left to process two pkts - * and prefetch two pkts... - */ - while (n_left_from >= 4 && n_left_to_next >= 2) - { - vlib_buffer_t * b0, * b1; - u32 bi0, next0, bi1, next1; - mpls_eth_tunnel_t * t0, * t1; - u32 sw_if_index0, sw_if_index1; - vnet_hw_interface_t * hi0, * hi1; - u8 * dst0, * dst1; - - /* Prefetch the 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); - - /* - * Prefetch packet data. We expect to overwrite - * the inbound L2 header with an ip header and a - * gre header. Might want to prefetch the last line - * of rewrite space as well; need profile data - */ - CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, STORE); - CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, STORE); - } - - /* Pick up the next two buffer indices */ - bi0 = from[0]; - bi1 = from[1]; - - /* Speculatively enqueue them where we sent the last buffer */ - to_next[0] = bi0; - to_next[1] = bi1; - from += 2; - to_next += 2; - n_left_to_next -= 2; - n_left_from -= 2; - - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - sw_if_index0 = vnet_buffer(b0)->sw_if_index [VLIB_TX]; - sw_if_index1 = vnet_buffer(b1)->sw_if_index [VLIB_TX]; - - /* get h/w intfcs */ - hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0); - hi1 = vnet_get_sup_hw_interface (vnm, sw_if_index1); - - /* hw_instance = tunnel pool index */ - t0 = pool_elt_at_index (gm->eth_tunnels, hi0->hw_instance); - t1 = pool_elt_at_index (gm->eth_tunnels, hi1->hw_instance); - - /* Apply rewrite - $$$$$ fixme don't use memcpy */ - vlib_buffer_advance (b0, -(word)vec_len(t0->rewrite_data)); - vlib_buffer_advance (b1, -(word)vec_len(t1->rewrite_data)); - - dst0 = vlib_buffer_get_current (b0); - dst1 = vlib_buffer_get_current (b1); - - clib_memcpy (dst0, t0->rewrite_data, vec_len(t0->rewrite_data)); - clib_memcpy (dst1, t1->rewrite_data, vec_len(t1->rewrite_data)); - - /* Fix TX fib indices */ - vnet_buffer(b0)->sw_if_index [VLIB_TX] = t0->tx_sw_if_index; - vnet_buffer(b1)->sw_if_index [VLIB_TX] = t1->tx_sw_if_index; - - /* mpls-post-rewrite takes it from here... */ - next0 = MPLS_ETH_OUTPUT_NEXT_OUTPUT; - next1 = MPLS_ETH_OUTPUT_NEXT_OUTPUT; - - if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) - { - mpls_eth_tx_trace_t *tr = vlib_add_trace (vm, node, - b0, sizeof (*tr)); - tr->lookup_miss = 0; - tr->tunnel_id = t0 - gm->eth_tunnels; - tr->tx_sw_if_index = t0->tx_sw_if_index; - tr->mpls_encap_index = t0->encap_index; - tr->length = b0->current_length; - hi0 = vnet_get_sup_hw_interface (vnm, t0->tx_sw_if_index); - clib_memcpy (tr->dst, hi0->hw_address, sizeof (tr->dst)); - } - if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED)) - { - mpls_eth_tx_trace_t *tr = vlib_add_trace (vm, node, - b1, sizeof (*tr)); - tr->lookup_miss = 0; - tr->tunnel_id = t1 - gm->eth_tunnels; - tr->tx_sw_if_index = t1->tx_sw_if_index; - tr->mpls_encap_index = t1->encap_index; - tr->length = b1->current_length; - hi1 = vnet_get_sup_hw_interface (vnm, t1->tx_sw_if_index); - clib_memcpy (tr->dst, hi1->hw_address, sizeof (tr->dst)); - } - - 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) - { - vlib_buffer_t * b0; - u32 bi0, next0; - mpls_eth_tunnel_t * t0; - u32 sw_if_index0; - vnet_hw_interface_t * hi0; - u8 * dst0; - - 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); - - sw_if_index0 = vnet_buffer(b0)->sw_if_index [VLIB_TX]; - - hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0); - - t0 = pool_elt_at_index (gm->eth_tunnels, hi0->hw_instance); - - /* Apply rewrite - $$$$$ fixme don't use memcpy */ - vlib_buffer_advance (b0, -(word)vec_len(t0->rewrite_data)); - - dst0 = vlib_buffer_get_current (b0); - - clib_memcpy (dst0, t0->rewrite_data, vec_len(t0->rewrite_data)); - - /* Fix the TX interface */ - vnet_buffer(b0)->sw_if_index [VLIB_TX] = t0->tx_sw_if_index; - - /* Send the packet */ - next0 = MPLS_ETH_OUTPUT_NEXT_OUTPUT; - - if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) - { - mpls_eth_tx_trace_t *tr = vlib_add_trace (vm, node, - b0, sizeof (*tr)); - tr->lookup_miss = 0; - tr->tunnel_id = t0 - gm->eth_tunnels; - tr->tx_sw_if_index = t0->tx_sw_if_index; - tr->mpls_encap_index = t0->encap_index; - tr->length = b0->current_length; - hi0 = vnet_get_sup_hw_interface (vnm, t0->tx_sw_if_index); - clib_memcpy (tr->dst, hi0->hw_address, sizeof (tr->dst)); - } - - 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); - } - - vlib_node_increment_counter (vm, mpls_input_node.index, - MPLS_ERROR_PKTS_ENCAP, frame->n_vectors); - - return frame->n_vectors; -} - -static u8 * format_mpls_eth_tunnel_name (u8 * s, va_list * args) -{ - u32 dev_instance = va_arg (*args, u32); - return format (s, "mpls-eth%d", dev_instance); -} - -static u8 * format_mpls_eth_device (u8 * s, va_list * args) -{ - u32 dev_instance = va_arg (*args, u32); - CLIB_UNUSED (int verbose) = va_arg (*args, int); - - s = format (s, "MPLS-ETH tunnel: id %d\n", dev_instance); - return s; -} - -VNET_DEVICE_CLASS (mpls_eth_device_class) = { - .name = "MPLS-ETH tunnel device", - .format_device_name = format_mpls_eth_tunnel_name, - .format_device = format_mpls_eth_device, - .format_tx_trace = format_mpls_eth_tx_trace, - .tx_function = mpls_eth_interface_tx, - .no_flatten_output_chains = 1, -#ifdef SOON - .clear counter = 0; - .admin_up_down_function = 0; -#endif -}; - -VLIB_DEVICE_TX_FUNCTION_MULTIARCH (mpls_eth_device_class, - mpls_eth_interface_tx) - -VNET_HW_INTERFACE_CLASS (mpls_eth_hw_interface_class) = { - .name = "MPLS-ETH", - .format_header = format_mpls_eth_header_with_length, -#if 0 - .unformat_header = unformat_mpls_eth_header, -#endif - .build_rewrite = default_build_rewrite, - .flags = VNET_HW_INTERFACE_CLASS_FLAG_P2P, -}; u8 mpls_sw_interface_is_enabled (u32 sw_if_index) @@ -308,753 +71,6 @@ mpls_sw_interface_enable_disable (mpls_main_t * mm, } -u8 * format_mpls_encap_index (u8 * s, va_list * args) -{ - mpls_main_t * mm = va_arg (*args, mpls_main_t *); - u32 entry_index = va_arg (*args, u32); - mpls_encap_t * e; - int i; - - e = pool_elt_at_index (mm->encaps, entry_index); - - for (i = 0; i < vec_len (e->labels); i++) - s = format - (s, "%d ", vnet_mpls_uc_get_label(clib_net_to_host_u32 - (e->labels[i].label_exp_s_ttl))); - - return s; -} - -u8 * format_mpls_ethernet_tunnel (u8 * s, va_list * args) -{ - mpls_eth_tunnel_t * t = va_arg (*args, mpls_eth_tunnel_t *); - mpls_main_t * mm = &mpls_main; - - s = format (s, "[%d]: dst %U, adj %U/%d, labels %U\n", - t - mm->eth_tunnels, - format_ethernet_address, &t->tunnel_dst, - format_ip4_address, &t->intfc_address, - t->mask_width, - format_mpls_encap_index, mm, t->encap_index); - - - s = format (s, " tx on %U, rx fib index %d", - format_vnet_sw_if_index_name, mm->vnet_main, t->tx_sw_if_index, - t->inner_fib_index); - - return s; -} - -static clib_error_t * -show_mpls_tunnel_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - mpls_main_t * mm = &mpls_main; - mpls_eth_tunnel_t * et; - - if (pool_elts (mm->eth_tunnels)) - { - vlib_cli_output (vm, "MPLS-Ethernet tunnels"); - pool_foreach (et, mm->eth_tunnels, - ({ - vlib_cli_output (vm, "%U", format_mpls_ethernet_tunnel, et); - })); - } - else - vlib_cli_output (vm, "No MPLS-Ethernet tunnels"); - - return 0; -} - -VLIB_CLI_COMMAND (show_mpls_tunnel_command, static) = { - .path = "show mpls tunnel", - .short_help = "show mpls tunnel", - .function = show_mpls_tunnel_command_fn, -}; - - -/* force inclusion from application's main.c */ -clib_error_t *mpls_interface_init (vlib_main_t *vm) -{ - clib_error_t * error; - - if ((error = vlib_call_init_function (vm, mpls_policy_encap_init))) - return error; - - return 0; -} -VLIB_INIT_FUNCTION(mpls_interface_init); - - -static u8 * mpls_ethernet_rewrite (mpls_main_t *mm, mpls_eth_tunnel_t * t) -{ - u8 * rewrite_data = 0; - mpls_encap_t * e; - mpls_unicast_header_t *lp0; - int i; - - /* look up the encap label stack using the RX FIB and adjacency address*/ - e = mpls_encap_by_fib_and_dest (mm, t->inner_fib_index, - t->intfc_address.as_u32); - - if (e == 0) - { - clib_warning ("no label for inner fib index %d, dst %U", - t->inner_fib_index, format_ip4_address, - &t->intfc_address); - return 0; - } - - vec_validate (rewrite_data, - sizeof (mpls_unicast_header_t) * vec_len(e->labels) -1); - - /* Copy the encap label stack */ - lp0 = (mpls_unicast_header_t *) rewrite_data; - - for (i = 0; i < vec_len(e->labels); i++) - lp0[i] = e->labels[i]; - - return (rewrite_data); -} - -int vnet_mpls_ethernet_add_del_tunnel (u8 *dst, - ip4_address_t *intfc, - u32 mask_width, - u32 inner_fib_id, - u32 tx_sw_if_index, - u32 * tunnel_sw_if_index, - u8 l2_only, - u8 is_add) -{ - ip4_main_t * im = &ip4_main; - ip_lookup_main_t * lm = &im->lookup_main; - mpls_main_t * mm = &mpls_main; - vnet_main_t * vnm = vnet_get_main(); - mpls_eth_tunnel_t *tp; - u32 inner_fib_index = 0; - ip_adjacency_t adj; - u32 adj_index; - u8 * rewrite_data; - int found_tunnel = 0; - mpls_encap_t * e = 0; - u32 hw_if_index = ~0; - vnet_hw_interface_t * hi; - u32 slot; - u32 dummy; - - if (tunnel_sw_if_index == 0) - tunnel_sw_if_index = &dummy; - - *tunnel_sw_if_index = ~0; - - if (inner_fib_id != (u32)~0) - { - uword * p; - - p = hash_get (im->fib_index_by_table_id, inner_fib_id); - if (! p) - return VNET_API_ERROR_NO_SUCH_FIB; - inner_fib_index = p[0]; - } - - /* suppress duplicate mpls interface generation. */ - pool_foreach (tp, mm->eth_tunnels, - ({ - /* - * If we have a tunnel which matches (src, dst, intfc/mask) - * AND the expected route is in the FIB, it's a dup - */ - if (!memcmp (&tp->tunnel_dst, dst, sizeof (*dst)) - && !memcmp (&tp->intfc_address, intfc, sizeof (*intfc)) - && tp->inner_fib_index == inner_fib_index - && FIB_NODE_INDEX_INVALID != tp->fei) - { - found_tunnel = 1; - - if (is_add) - { - if (l2_only) - return 1; - else - { - e = mpls_encap_by_fib_and_dest (mm, inner_fib_index, - intfc->as_u32); - if (e == 0) - return VNET_API_ERROR_NO_SUCH_LABEL; - - goto reinstall_it; - } - } - else - { - /* Delete */ - goto add_del_route; - } - - } - })); - - /* Delete, and we can't find the tunnel */ - if (is_add == 0 && found_tunnel == 0) - return VNET_API_ERROR_NO_SUCH_ENTRY; - - e = mpls_encap_by_fib_and_dest (mm, inner_fib_index, intfc->as_u32); - if (e == 0) - return VNET_API_ERROR_NO_SUCH_LABEL; - - pool_get(mm->eth_tunnels, tp); - memset (tp, 0, sizeof (*tp)); - - if (vec_len (mm->free_eth_sw_if_indices) > 0) - { - hw_if_index = - mm->free_eth_sw_if_indices[vec_len(mm->free_eth_sw_if_indices)-1]; - _vec_len (mm->free_eth_sw_if_indices) -= 1; - hi = vnet_get_hw_interface (vnm, hw_if_index); - hi->dev_instance = tp - mm->eth_tunnels; - hi->hw_instance = tp - mm->eth_tunnels; - } - else - { - hw_if_index = vnet_register_interface - (vnm, mpls_eth_device_class.index, tp - mm->eth_tunnels, - mpls_eth_hw_interface_class.index, - tp - mm->eth_tunnels); - hi = vnet_get_hw_interface (vnm, hw_if_index); - - /* ... to make the IP and L2 x-connect cases identical */ - slot = vlib_node_add_named_next_with_slot - (vnm->vlib_main, hi->tx_node_index, - "interface-output", MPLS_ETH_OUTPUT_NEXT_OUTPUT); - - ASSERT (slot == MPLS_ETH_OUTPUT_NEXT_OUTPUT); - } - - *tunnel_sw_if_index = hi->sw_if_index; - vnet_sw_interface_set_flags (vnm, hi->sw_if_index, - VNET_SW_INTERFACE_FLAG_ADMIN_UP); - - tp->hw_if_index = hw_if_index; - - reinstall_it: - clib_memcpy(tp->tunnel_dst, dst, sizeof (tp->tunnel_dst)); - tp->intfc_address.as_u32 = intfc->as_u32; - tp->mask_width = mask_width; - tp->inner_fib_index = inner_fib_index; - tp->encap_index = e - mm->encaps; - tp->tx_sw_if_index = tx_sw_if_index; - tp->l2_only = l2_only; - - /* Create the adjacency and add to v4 fib */ - memset(&adj, 0, sizeof (adj)); - adj.lookup_next_index = IP_LOOKUP_NEXT_REWRITE; - - rewrite_data = mpls_ethernet_rewrite (mm, tp); - if (rewrite_data == 0) - { - if (*tunnel_sw_if_index != ~0) - { - hi = vnet_get_hw_interface (vnm, tp->hw_if_index); - vnet_sw_interface_set_flags (vnm, hi->sw_if_index, - 0 /* admin down */); - vec_add1 (mm->free_eth_sw_if_indices, tp->hw_if_index); - } - - pool_put (mm->eth_tunnels, tp); - return VNET_API_ERROR_NO_SUCH_LABEL; - } - - vnet_rewrite_for_sw_interface - (vnm, - VNET_LINK_MPLS, - tx_sw_if_index, - ip4_rewrite_node.index, - tp->tunnel_dst, - &adj.rewrite_header, - sizeof (adj.rewrite_data)); - - /* - * Prepend the (0,1,2) VLAN tag ethernet header - * we just built to the mpls header stack - */ - vec_insert (rewrite_data, adj.rewrite_header.data_bytes, 0); - clib_memcpy(rewrite_data, - vnet_rewrite_get_data_internal(&adj.rewrite_header, - sizeof (adj.rewrite_data)), - adj.rewrite_header.data_bytes); - - vnet_rewrite_set_data_internal (&adj.rewrite_header, - sizeof(adj.rewrite_data), - rewrite_data, - vec_len(rewrite_data)); - - vec_free (tp->rewrite_data); - - tp->rewrite_data = rewrite_data; - - if (!l2_only) - ip_add_adjacency (lm, &adj, 1 /* one adj */, - &adj_index); - - add_del_route: - - if (!l2_only) - { - const fib_prefix_t pfx = { - .fp_addr = { - .ip4 = tp->intfc_address, - }, - .fp_len = tp->mask_width, - .fp_proto = FIB_PROTOCOL_IP4, - }; - if (is_add) - tp->fei = fib_table_entry_special_add(tp->inner_fib_index, - &pfx, - FIB_SOURCE_API, - FIB_ENTRY_FLAG_NONE, - adj_index); - else - { - fib_table_entry_delete(tp->inner_fib_index, &pfx, FIB_SOURCE_API); - tp->fei = FIB_NODE_INDEX_INVALID; - } - } - if (is_add == 0 && found_tunnel) - { - hi = vnet_get_hw_interface (vnm, tp->hw_if_index); - vnet_sw_interface_set_flags (vnm, hi->sw_if_index, - 0 /* admin down */); - vec_add1 (mm->free_eth_sw_if_indices, tp->hw_if_index); - vec_free (tp->rewrite_data); - pool_put (mm->eth_tunnels, tp); - } - - return 0; -} - -static clib_error_t * -create_mpls_ethernet_tunnel_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - unformat_input_t _line_input, * line_input = &_line_input; - vnet_main_t * vnm = vnet_get_main(); - ip4_address_t intfc; - int adj_set = 0; - u8 dst[6]; - int dst_set = 0, intfc_set = 0; - u32 mask_width; - u32 inner_fib_id = (u32)~0; - int rv; - u8 is_del = 0; - u8 l2_only = 0; - u32 tx_sw_if_index; - u32 sw_if_index = ~0; - - /* Get a line of input. */ - if (! unformat_user (input, unformat_line_input, line_input)) - return 0; - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "dst %U", - unformat_ethernet_address, &dst)) - dst_set = 1; - else if (unformat (line_input, "adj %U/%d", - unformat_ip4_address, &intfc, &mask_width)) - adj_set = 1; - else if (unformat (line_input, "tx-intfc %U", - unformat_vnet_sw_interface, vnm, &tx_sw_if_index)) - intfc_set = 1; - else if (unformat (line_input, "fib-id %d", &inner_fib_id)) - ; - else if (unformat (line_input, "l2-only")) - l2_only = 1; - else if (unformat (line_input, "del")) - is_del = 1; - else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, line_input); - } - - if (!intfc_set) - return clib_error_return (0, "missing tx-intfc"); - - if (!dst_set) - return clib_error_return (0, "missing: dst <ethernet-address>"); - - if (!adj_set) - return clib_error_return (0, "missing: intfc <ip-address>/<mask-width>"); - - - rv = vnet_mpls_ethernet_add_del_tunnel (dst, &intfc, mask_width, - inner_fib_id, tx_sw_if_index, - &sw_if_index, - l2_only, !is_del); - - switch (rv) - { - case 0: - if (!is_del) - vlib_cli_output(vm, "%U\n", format_vnet_sw_if_index_name, vnet_get_main(), sw_if_index); - break; - case VNET_API_ERROR_NO_SUCH_FIB: - return clib_error_return (0, "rx fib ID %d doesn't exist\n", - inner_fib_id); - - case VNET_API_ERROR_NO_SUCH_ENTRY: - return clib_error_return (0, "tunnel not found\n"); - - case VNET_API_ERROR_NO_SUCH_LABEL: - /* - * This happens when there's no MPLS label for the dst address - * no need for two error messages. - */ - return clib_error_return (0, "no label for %U in fib %d", - format_ip4_address, &intfc, inner_fib_id); - break; - - default: - return clib_error_return (0, "vnet_mpls_ethernet_add_del_tunnel returned %d", rv); - break; - } - return 0; -} - - -VLIB_CLI_COMMAND (create_mpls_ethernet_tunnel_command, static) = { - .path = "create mpls ethernet tunnel", - .short_help = - "create mpls ethernet tunnel [del] dst <mac-addr> intfc <addr>/<mw>", - .function = create_mpls_ethernet_tunnel_command_fn, -}; - - -int vnet_mpls_policy_tunnel_add_rewrite (mpls_main_t * mm, - mpls_encap_t * e, - u32 policy_tunnel_index) -{ - mpls_eth_tunnel_t * t; - ip_adjacency_t adj; - u8 * rewrite_data = 0; - u8 * label_start; - mpls_unicast_header_t *lp; - int i; - - if (pool_is_free_index (mm->eth_tunnels, policy_tunnel_index)) - return VNET_API_ERROR_NO_SUCH_ENTRY; - - t = pool_elt_at_index (mm->eth_tunnels, policy_tunnel_index); - - memset (&adj, 0, sizeof (adj)); - - /* Build L2 encap */ - vnet_rewrite_for_sw_interface - (mm->vnet_main, - VNET_LINK_MPLS, - t->tx_sw_if_index, - mpls_policy_encap_node.index, - t->tunnel_dst, - &adj.rewrite_header, - sizeof (adj.rewrite_data)); - - vec_validate (rewrite_data, adj.rewrite_header.data_bytes -1); - - clib_memcpy(rewrite_data, - vnet_rewrite_get_data_internal(&adj.rewrite_header, - sizeof (adj.rewrite_data)), - adj.rewrite_header.data_bytes); - - /* Append the label stack */ - - vec_add2 (rewrite_data, label_start, vec_len(e->labels) * sizeof (u32)); - - lp = (mpls_unicast_header_t *) label_start; - - for (i = 0; i < vec_len(e->labels); i++) - lp[i] = e->labels[i]; - - /* Remember the rewrite data */ - e->rewrite = rewrite_data; - e->output_next_index = adj.rewrite_header.next_index; - - return 0; -} - -int vnet_mpls_ethernet_add_del_policy_tunnel (u8 *dst, - ip4_address_t *intfc, - u32 mask_width, - u32 inner_fib_id, - u32 tx_sw_if_index, - u32 * tunnel_sw_if_index, - u32 classify_table_index, - u32 * new_tunnel_index, - u8 l2_only, - u8 is_add) -{ - ip4_main_t * im = &ip4_main; - mpls_main_t * mm = &mpls_main; - vnet_main_t * vnm = vnet_get_main(); - mpls_eth_tunnel_t *tp; - u32 inner_fib_index = 0; - int found_tunnel = 0; - mpls_encap_t * e = 0; - u32 hw_if_index = ~0; - vnet_hw_interface_t * hi; - u32 slot; - u32 dummy; - - if (tunnel_sw_if_index == 0) - tunnel_sw_if_index = &dummy; - - *tunnel_sw_if_index = ~0; - - if (inner_fib_id != (u32)~0) - { - uword * p; - - p = hash_get (im->fib_index_by_table_id, inner_fib_id); - if (! p) - return VNET_API_ERROR_NO_SUCH_FIB; - inner_fib_index = p[0]; - } - - /* suppress duplicate mpls interface generation. */ - pool_foreach (tp, mm->eth_tunnels, - ({ - /* - * If we have a tunnel which matches (src, dst, intfc/mask) - * AND the expected route is in the FIB, it's a dup - */ - if (!memcmp (&tp->tunnel_dst, dst, sizeof (*dst)) - && !memcmp (&tp->intfc_address, intfc, sizeof (*intfc)) - && tp->inner_fib_index == inner_fib_index - && FIB_NODE_INDEX_INVALID != tp->fei) - { - found_tunnel = 1; - - if (is_add) - { - if (l2_only) - return 1; - else - { - goto reinstall_it; - } - } - else - { - /* Delete */ - goto add_del_route; - } - - } - })); - - /* Delete, and we can't find the tunnel */ - if (is_add == 0 && found_tunnel == 0) - return VNET_API_ERROR_NO_SUCH_ENTRY; - - pool_get(mm->eth_tunnels, tp); - memset (tp, 0, sizeof (*tp)); - - if (vec_len (mm->free_eth_sw_if_indices) > 0) - { - hw_if_index = - mm->free_eth_sw_if_indices[vec_len(mm->free_eth_sw_if_indices)-1]; - _vec_len (mm->free_eth_sw_if_indices) -= 1; - hi = vnet_get_hw_interface (vnm, hw_if_index); - hi->dev_instance = tp - mm->eth_tunnels; - hi->hw_instance = tp - mm->eth_tunnels; - } - else - { - hw_if_index = vnet_register_interface - (vnm, mpls_eth_device_class.index, tp - mm->eth_tunnels, - mpls_eth_hw_interface_class.index, - tp - mm->eth_tunnels); - hi = vnet_get_hw_interface (vnm, hw_if_index); - - /* ... to make the IP and L2 x-connect cases identical */ - slot = vlib_node_add_named_next_with_slot - (vnm->vlib_main, hi->tx_node_index, - "interface-output", MPLS_ETH_OUTPUT_NEXT_OUTPUT); - - ASSERT (slot == MPLS_ETH_OUTPUT_NEXT_OUTPUT); - } - - *tunnel_sw_if_index = hi->sw_if_index; - vnet_sw_interface_set_flags (vnm, hi->sw_if_index, - VNET_SW_INTERFACE_FLAG_ADMIN_UP); - - tp->hw_if_index = hw_if_index; - - reinstall_it: - clib_memcpy(tp->tunnel_dst, dst, sizeof (tp->tunnel_dst)); - tp->intfc_address.as_u32 = intfc->as_u32; - tp->mask_width = mask_width; - tp->inner_fib_index = inner_fib_index; - tp->encap_index = e - mm->encaps; - tp->tx_sw_if_index = tx_sw_if_index; - tp->l2_only = l2_only; - tp->fei = FIB_NODE_INDEX_INVALID; - - if (new_tunnel_index) - *new_tunnel_index = tp - mm->eth_tunnels; - - add_del_route: - - if (!l2_only) - { - const fib_prefix_t pfx = { - .fp_addr = { - .ip4 = tp->intfc_address, - }, - .fp_len = tp->mask_width, - .fp_proto = FIB_PROTOCOL_IP4, - }; - dpo_id_t dpo = DPO_INVALID; - - if (is_add) - { - dpo_set(&dpo, - DPO_CLASSIFY, - DPO_PROTO_IP4, - classify_dpo_create(DPO_PROTO_IP4, - classify_table_index)); - - tp->fei = fib_table_entry_special_dpo_add(tp->inner_fib_index, - &pfx, - FIB_SOURCE_API, - FIB_ENTRY_FLAG_EXCLUSIVE, - &dpo); - dpo_reset(&dpo); - } - else - { - fib_table_entry_delete(tp->inner_fib_index, &pfx, FIB_SOURCE_API); - tp->fei = FIB_NODE_INDEX_INVALID; - } - } - if (is_add == 0 && found_tunnel) - { - hi = vnet_get_hw_interface (vnm, tp->hw_if_index); - vnet_sw_interface_set_flags (vnm, hi->sw_if_index, - 0 /* admin down */); - vec_add1 (mm->free_eth_sw_if_indices, tp->hw_if_index); - pool_put (mm->eth_tunnels, tp); - } - - return 0; -} - -static clib_error_t * -create_mpls_ethernet_policy_tunnel_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - unformat_input_t _line_input, * line_input = &_line_input; - vnet_main_t * vnm = vnet_get_main(); - ip4_address_t intfc; - int adj_set = 0; - u8 dst[6]; - int dst_set = 0, intfc_set = 0; - u32 mask_width; - u32 inner_fib_id = (u32)~0; - u32 classify_table_index = (u32)~0; - u32 new_tunnel_index; - int rv; - u8 is_del = 0; - u8 l2_only = 0; - u32 tx_sw_if_index; - - /* Get a line of input. */ - if (! unformat_user (input, unformat_line_input, line_input)) - return 0; - - while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (line_input, "dst %U", - unformat_ethernet_address, &dst)) - dst_set = 1; - else if (unformat (line_input, "adj %U/%d", - unformat_ip4_address, &intfc, &mask_width)) - adj_set = 1; - else if (unformat (line_input, "tx-intfc %U", - unformat_vnet_sw_interface, vnm, &tx_sw_if_index)) - intfc_set = 1; - else if (unformat (line_input, "classify-table-index %d", - &classify_table_index)) - ; - else if (unformat (line_input, "fib-id %d", &inner_fib_id)) - ; - else if (unformat (line_input, "l2-only")) - l2_only = 1; - else if (unformat (line_input, "del")) - is_del = 1; - else - return clib_error_return (0, "unknown input '%U'", - format_unformat_error, line_input); - } - - if (classify_table_index == ~0) - return clib_error_return (0, "missing classify_table_index"); - - if (!intfc_set) - return clib_error_return (0, "missing tx-intfc"); - - if (!dst_set) - return clib_error_return (0, "missing: dst <ethernet-address>"); - - if (!adj_set) - return clib_error_return (0, "missing: intfc <ip-address>/<mask-width>"); - - - rv = vnet_mpls_ethernet_add_del_policy_tunnel (dst, &intfc, mask_width, - inner_fib_id, tx_sw_if_index, - 0 /* tunnel sw_if_index */, - classify_table_index, - &new_tunnel_index, - l2_only, !is_del); - switch (rv) - { - case 0: - if (!is_del) - vlib_cli_output(vm, "%U\n", format_vnet_sw_if_index_name, vnet_get_main(), new_tunnel_index); - break; - case VNET_API_ERROR_NO_SUCH_FIB: - return clib_error_return (0, "rx fib ID %d doesn't exist\n", - inner_fib_id); - - case VNET_API_ERROR_NO_SUCH_ENTRY: - return clib_error_return (0, "tunnel not found\n"); - - case VNET_API_ERROR_NO_SUCH_LABEL: - /* - * This happens when there's no MPLS label for the dst address - * no need for two error messages. - */ - return clib_error_return (0, "no label for %U in fib %d", - format_ip4_address, &intfc, inner_fib_id); - break; - - default: - return clib_error_return (0, "vnet_mpls_ethernet_add_del_policy_tunnel returned %d", rv); - break; - } - - return 0; -} - -VLIB_CLI_COMMAND (create_mpls_ethernet_policy_tunnel_command, static) = { - .path = "create mpls ethernet policy tunnel", - .short_help = - "create mpls ethernet policy tunnel [del] dst <mac-addr> intfc <addr>/<mw>\n" - " classify-table-index <nn>", - .function = create_mpls_ethernet_policy_tunnel_command_fn, -}; - static clib_error_t * mpls_interface_enable_disable (vlib_main_t * vm, unformat_input_t * input, @@ -1090,6 +106,14 @@ mpls_interface_enable_disable (vlib_main_t * vm, return error; } +/*? + * This command enables an interface to accpet MPLS packets + * + * @cliexpar + * @cliexstart{set interface mpls} + * set interface mpls GigEthernet0/8/0 enable + * @cliexend + ?*/ VLIB_CLI_COMMAND (set_interface_ip_table_command, static) = { .path = "set interface mpls", .function = mpls_interface_enable_disable, diff --git a/vnet/vnet/mpls/mpls.c b/vnet/vnet/mpls/mpls.c index aa1d963dca1..e6ae4980067 100644 --- a/vnet/vnet/mpls/mpls.c +++ b/vnet/vnet/mpls/mpls.c @@ -121,45 +121,6 @@ unformat_mpls_header (unformat_input_t * input, va_list * args) return 1; } -u8 * format_mpls_eth_tx_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - mpls_eth_tx_trace_t * t = va_arg (*args, mpls_eth_tx_trace_t *); - mpls_main_t * mm = &mpls_main; - - if (t->lookup_miss) - s = format (s, "MPLS: lookup miss"); - else - { - s = format (s, "MPLS: tunnel %d labels %U len %d tx_sw_index %d dst %U", - t->tunnel_id, - format_mpls_encap_index, mm, t->mpls_encap_index, - clib_net_to_host_u16 (t->length), - t->tx_sw_if_index, - format_ethernet_address, t->dst); - } - return s; -} - -u8 * format_mpls_eth_header_with_length (u8 * s, va_list * args) -{ - ethernet_header_t * h = va_arg (*args, ethernet_header_t *); - mpls_unicast_header_t * m = (mpls_unicast_header_t *)(h+1); - u32 max_header_bytes = va_arg (*args, u32); - uword header_bytes; - - header_bytes = sizeof (h[0]); - if (max_header_bytes != 0 && header_bytes > max_header_bytes) - return format (s, "ethernet header truncated"); - - s = format - (s, "ETHERNET-MPLS label %d", - vnet_mpls_uc_get_label (clib_net_to_host_u32 (m->label_exp_s_ttl))); - - return s; -} - uword unformat_mpls_label_net_byte_order (unformat_input_t * input, va_list * args) @@ -176,157 +137,6 @@ unformat_mpls_label_net_byte_order (unformat_input_t * input, return 1; } -mpls_encap_t * -mpls_encap_by_fib_and_dest (mpls_main_t * mm, u32 rx_fib, u32 dst_address) -{ - uword * p; - mpls_encap_t * e; - u64 key; - - key = ((u64)rx_fib<<32) | ((u64) dst_address); - p = hash_get (mm->mpls_encap_by_fib_and_dest, key); - - if (!p) - return 0; - - e = pool_elt_at_index (mm->encaps, p[0]); - return e; -} - -int vnet_mpls_add_del_encap (ip4_address_t *dest, u32 fib_id, - u32 *labels_host_byte_order, - u32 policy_tunnel_index, - int no_dst_hash, u32 * indexp, int is_add) -{ - mpls_main_t * mm = &mpls_main; - ip4_main_t * im = &ip4_main; - mpls_encap_t * e; - u32 label_net_byte_order, label_host_byte_order; - u32 fib_index; - u64 key; - uword *p; - int i; - - p = hash_get (im->fib_index_by_table_id, fib_id); - if (! p) - return VNET_API_ERROR_NO_SUCH_FIB; - - fib_index = p[0]; - - key = ((u64)fib_index<<32) | ((u64) dest->as_u32); - - if (is_add) - { - pool_get (mm->encaps, e); - memset (e, 0, sizeof (*e)); - - for (i = 0; i < vec_len (labels_host_byte_order); i++) - { - mpls_unicast_header_t h; - label_host_byte_order = labels_host_byte_order[i]; - - /* Reformat label into mpls_unicast_header_t */ - label_host_byte_order <<= 12; - // FIXME NEOS AND EOS - //if (i == vec_len(labels_host_byte_order) - 1) - // label_host_byte_order |= 1<<8; /* S=1 */ - label_host_byte_order |= 0xff; /* TTL=FF */ - label_net_byte_order = clib_host_to_net_u32 (label_host_byte_order); - h.label_exp_s_ttl = label_net_byte_order; - vec_add1 (e->labels, h); - } - if (no_dst_hash == 0) - hash_set (mm->mpls_encap_by_fib_and_dest, key, e - mm->encaps); - if (indexp) - *indexp = e - mm->encaps; - if (policy_tunnel_index != ~0) - return vnet_mpls_policy_tunnel_add_rewrite (mm, e, policy_tunnel_index); - } - else - { - p = hash_get (mm->mpls_encap_by_fib_and_dest, key); - if (!p) - return VNET_API_ERROR_NO_SUCH_LABEL; - - e = pool_elt_at_index (mm->encaps, p[0]); - - vec_free (e->labels); - vec_free (e->rewrite); - pool_put(mm->encaps, e); - - if (no_dst_hash == 0) - hash_unset (mm->mpls_encap_by_fib_and_dest, key); - } - return 0; -} - -static clib_error_t * -mpls_add_encap_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - u32 fib_id; - u32 *labels = 0; - u32 this_label; - ip4_address_t dest; - u32 policy_tunnel_index = ~0; - int no_dst_hash = 0; - int rv; - int fib_set = 0; - int dest_set = 0; - - while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) - { - if (unformat (input, "fib %d", &fib_id)) - fib_set = 1; - else if (unformat (input, "dest %U", unformat_ip4_address, &dest)) - dest_set = 1; - else if (unformat (input, "no-dst-hash")) - no_dst_hash = 1; - else if (unformat (input, "label %d", &this_label)) - vec_add1 (labels, this_label); - else if (unformat (input, "policy-tunnel %d", &policy_tunnel_index)) - ; - else - break; - } - - if (fib_set == 0) - return clib_error_return (0, "fib-id missing"); - if (dest_set == 0) - return clib_error_return (0, "destination IP address missing"); - if (vec_len (labels) == 0) - return clib_error_return (0, "label stack missing"); - - rv = vnet_mpls_add_del_encap (&dest, fib_id, labels, - policy_tunnel_index, - no_dst_hash, 0 /* indexp */, - 1 /* is_add */); - vec_free (labels); - - switch (rv) - { - case 0: - break; - - case VNET_API_ERROR_NO_SUCH_FIB: - return clib_error_return (0, "fib id %d unknown", fib_id); - - default: - return clib_error_return (0, "vnet_mpls_add_del_encap returned %d", - rv); - } - - return 0; -} - -VLIB_CLI_COMMAND (mpls_add_encap_command, static) = { - .path = "mpls encap add", - .short_help = - "mpls encap add label <label> ... fib <id> dest <ip4-address>", - .function = mpls_add_encap_command_fn, -}; - u8 * format_mpls_unicast_header_host_byte_order (u8 * s, va_list * args) { mpls_unicast_header_t *h = va_arg(*args, mpls_unicast_header_t *); @@ -351,46 +161,6 @@ u8 * format_mpls_unicast_header_net_byte_order (u8 * s, va_list * args) &h_host); } -static clib_error_t * -mpls_del_encap_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - u32 fib_id; - ip4_address_t dest; - int rv; - - if (unformat (input, "fib %d dest %U", &fib_id, - unformat_ip4_address, &dest)) - { - rv = vnet_mpls_add_del_encap (&dest, fib_id, 0 /* labels */, - ~0 /* policy_tunnel_index */, - 0 /* no_dst_hash */, - 0 /* indexp */, - 0 /* is_add */); - switch (rv) - { - case VNET_API_ERROR_NO_SUCH_FIB: - return clib_error_return (0, "fib id %d unknown", fib_id); - case VNET_API_ERROR_NO_SUCH_ENTRY: - return clib_error_return (0, "dest %U not in fib %d", - format_ip4_address, &dest, fib_id); - default: - break; - } - return 0; - } - else - return clib_error_return (0, "unknown input `%U'", - format_unformat_error, input); -} - -VLIB_CLI_COMMAND (mpls_del_encap_command, static) = { - .path = "mpls encap delete", - .short_help = "mpls encap delete fib <id> dest <ip4-address>", - .function = mpls_del_encap_command_fn, -}; - int mpls_dest_cmp(void * a1, void * a2) { @@ -419,65 +189,22 @@ mpls_label_cmp(void * a1, void * a2) } static clib_error_t * -show_mpls_fib_command_fn (vlib_main_t * vm, - unformat_input_t * input, - vlib_cli_command_t * cmd) -{ - u64 key; - u32 value; - show_mpls_fib_t *records = 0; - show_mpls_fib_t *s; - mpls_main_t * mm = &mpls_main; - ip4_fib_t * rx_fib; - - hash_foreach (key, value, mm->mpls_encap_by_fib_and_dest, - ({ - vec_add2 (records, s, 1); - s->fib_index = (u32)(key>>32); - s->dest = (u32)(key & 0xFFFFFFFF); - s->entry_index = (u32) value; - })); - - if (!vec_len(records)) - { - vlib_cli_output (vm, "MPLS encap table empty"); - } - /* sort output by dst address within fib */ - vec_sort_with_function (records, mpls_dest_cmp); - vec_sort_with_function (records, mpls_fib_index_cmp); - vlib_cli_output (vm, "MPLS encap table"); - vlib_cli_output (vm, "%=6s%=16s%=16s", "Table", "Dest address", "Labels"); - vec_foreach (s, records) - { - rx_fib = ip4_fib_get (s->fib_index); - vlib_cli_output (vm, "%=6d%=16U%=16U", rx_fib->table_id, - format_ip4_address, &s->dest, - format_mpls_encap_index, mm, s->entry_index); - } - - vec_free(records); - return 0; -} - -VLIB_CLI_COMMAND (show_mpls_fib_command, static) = { - .path = "show mpls encap", - .short_help = "show mpls encap", - .function = show_mpls_fib_command_fn, -}; - -static clib_error_t * vnet_mpls_local_label (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cmd) { unformat_input_t _line_input, * line_input = &_line_input; fib_route_path_t *rpaths = NULL, rpath; - clib_error_t * error = 0; u32 table_id, is_del, is_ip; - fib_prefix_t pfx; mpls_label_t local_label; + mpls_label_t out_label; + clib_error_t * error; mpls_eos_bit_t eos; + vnet_main_t * vnm; + fib_prefix_t pfx; + vnm = vnet_get_main(); + error = NULL; is_ip = 0; table_id = 0; eos = MPLS_EOS; @@ -519,13 +246,99 @@ vnet_mpls_local_label (vlib_main_t * vm, pfx.fp_proto = FIB_PROTOCOL_IP6; is_ip = 1; } + else if (unformat (line_input, "via %U %U weight %u", + unformat_ip4_address, + &rpath.frp_addr.ip4, + unformat_vnet_sw_interface, vnm, + &rpath.frp_sw_if_index, + &rpath.frp_weight)) + { + rpath.frp_proto = FIB_PROTOCOL_IP4; + vec_add1(rpaths, rpath); + } + + else if (unformat (line_input, "via %U %U weight %u", + unformat_ip6_address, + &rpath.frp_addr.ip6, + unformat_vnet_sw_interface, vnm, + &rpath.frp_sw_if_index, + &rpath.frp_weight)) + { + rpath.frp_proto = FIB_PROTOCOL_IP6; + vec_add1(rpaths, rpath); + } + + else if (unformat (line_input, "via %U %U", + unformat_ip4_address, + &rpath.frp_addr.ip4, + unformat_vnet_sw_interface, vnm, + &rpath.frp_sw_if_index)) + { + rpath.frp_weight = 1; + rpath.frp_proto = FIB_PROTOCOL_IP4; + vec_add1(rpaths, rpath); + } + + else if (unformat (line_input, "via %U %U", + unformat_ip6_address, + &rpath.frp_addr.ip6, + unformat_vnet_sw_interface, vnm, + &rpath.frp_sw_if_index)) + { + rpath.frp_weight = 1; + rpath.frp_proto = FIB_PROTOCOL_IP6; + vec_add1(rpaths, rpath); + } + else if (unformat (line_input, "via %U next-hop-table %d", + unformat_ip4_address, + &rpath.frp_addr.ip4, + &rpath.frp_fib_index)) + { + rpath.frp_weight = 1; + rpath.frp_sw_if_index = ~0; + rpath.frp_proto = FIB_PROTOCOL_IP4; + vec_add1(rpaths, rpath); + } + else if (unformat (line_input, "via %U next-hop-table %d", + unformat_ip6_address, + &rpath.frp_addr.ip6, + &rpath.frp_fib_index)) + { + rpath.frp_weight = 1; + rpath.frp_sw_if_index = ~0; + rpath.frp_proto = FIB_PROTOCOL_IP6; + vec_add1(rpaths, rpath); + } + else if (unformat (line_input, "via %U", + unformat_ip4_address, + &rpath.frp_addr.ip4)) + { + /* + * the recursive next-hops are by default in the same table + * as the prefix + */ + rpath.frp_fib_index = table_id; + rpath.frp_weight = 1; + rpath.frp_sw_if_index = ~0; + rpath.frp_proto = FIB_PROTOCOL_IP4; + vec_add1(rpaths, rpath); + } + else if (unformat (line_input, "via %U", + unformat_ip6_address, + &rpath.frp_addr.ip6)) + { + rpath.frp_fib_index = table_id; + rpath.frp_weight = 1; + rpath.frp_sw_if_index = ~0; + rpath.frp_proto = FIB_PROTOCOL_IP6; + vec_add1(rpaths, rpath); + } else if (unformat (line_input, "%d", &local_label)) ; else if (unformat (line_input, "ip4-lookup-in-table %d", &rpath.frp_fib_index)) { - rpath.frp_label = MPLS_LABEL_INVALID; rpath.frp_proto = FIB_PROTOCOL_IP4; rpath.frp_sw_if_index = FIB_NODE_INDEX_INVALID; pfx.fp_payload_proto = DPO_PROTO_IP4; @@ -535,7 +348,6 @@ vnet_mpls_local_label (vlib_main_t * vm, "ip6-lookup-in-table %d", &rpath.frp_fib_index)) { - rpath.frp_label = MPLS_LABEL_INVALID; rpath.frp_proto = FIB_PROTOCOL_IP6; rpath.frp_sw_if_index = FIB_NODE_INDEX_INVALID; vec_add1(rpaths, rpath); @@ -545,16 +357,26 @@ vnet_mpls_local_label (vlib_main_t * vm, "mpls-lookup-in-table %d", &rpath.frp_fib_index)) { - rpath.frp_label = MPLS_LABEL_INVALID; rpath.frp_proto = FIB_PROTOCOL_MPLS; rpath.frp_sw_if_index = FIB_NODE_INDEX_INVALID; pfx.fp_payload_proto = DPO_PROTO_MPLS; vec_add1(rpaths, rpath); } + else if (unformat (line_input, "out-label %U", + unformat_mpls_unicast_label, + &out_label)) + { + if (vec_len(rpaths) == 0) + { + error = clib_error_return(0 , "Paths then labels"); + goto done; + } + vec_add1(rpaths[vec_len(rpaths)-1].frp_label_stack, out_label); + } else { error = clib_error_return (0, "unkown input: %U", - format_unformat_error, input); + format_unformat_error, line_input); goto done; } @@ -596,6 +418,7 @@ vnet_mpls_local_label (vlib_main_t * vm, pfx.fp_proto = FIB_PROTOCOL_MPLS; pfx.fp_len = 21; pfx.fp_label = local_label; + pfx.fp_payload_proto = DPO_PROTO_MPLS; if (NULL == rpaths) { @@ -606,17 +429,20 @@ vnet_mpls_local_label (vlib_main_t * vm, /* * the CLI parsing stored table Ids, swap to FIB indicies */ - fi = fib_table_id_find_fib_index(dpo_proto_to_fib(pfx.fp_payload_proto), - rpaths[0].frp_fib_index); - - if (~0 == fi) + if (FIB_NODE_INDEX_INVALID == rpath.frp_sw_if_index) { - error = clib_error_return(0 , "%U Via table %d does not exist", - format_fib_protocol, pfx.fp_payload_proto, - rpaths[0].frp_fib_index); - goto done; + fi = fib_table_id_find_fib_index(dpo_proto_to_fib(pfx.fp_payload_proto), + rpaths[0].frp_fib_index); + + if (~0 == fi) + { + error = clib_error_return(0 , "%U Via table %d does not exist", + format_fib_protocol, pfx.fp_payload_proto, + rpaths[0].frp_fib_index); + goto done; + } + rpaths[0].frp_fib_index = fi; } - rpaths[0].frp_fib_index = fi; fib_index = mpls_fib_index_from_table_id(table_id); @@ -653,44 +479,15 @@ VLIB_CLI_COMMAND (mpls_local_label_command, static) = { .short_help = "Create/Delete MPL local labels", }; -int mpls_fib_reset_labels (u32 fib_id) +int +mpls_fib_reset_labels (u32 fib_id) { - u64 key; - u32 value; - show_mpls_fib_t *records = 0; - show_mpls_fib_t *s; - mpls_main_t * mm = &mpls_main; - ip4_main_t * im = &ip4_main; - u32 fib_index; - uword *p; - - p = hash_get (im->fib_index_by_table_id, fib_id); - if (! p) - return VNET_API_ERROR_NO_SUCH_FIB; - - fib_index = p[0]; - - hash_foreach (key, value, mm->mpls_encap_by_fib_and_dest, - ({ - if (fib_index == (u32)(key>>32)) { - vec_add2 (records, s, 1); - s->dest = (u32)(key & 0xFFFFFFFF); - s->entry_index = (u32) value; - } - })); - - vec_foreach (s, records) - { - key = ((u64)fib_index<<32) | ((u64) s->dest); - hash_unset (mm->mpls_encap_by_fib_and_dest, key); - pool_put_index (mm->encaps, s->entry_index); - } - - vec_free(records); + // FIXME return 0; } -static clib_error_t * mpls_init (vlib_main_t * vm) +static clib_error_t * +mpls_init (vlib_main_t * vm) { mpls_main_t * mm = &mpls_main; clib_error_t * error; @@ -701,8 +498,6 @@ static clib_error_t * mpls_init (vlib_main_t * vm) if ((error = vlib_call_init_function (vm, ip_main_init))) return error; - mm->mpls_encap_by_fib_and_dest = hash_create (0, sizeof (uword)); - return vlib_call_init_function (vm, mpls_input_init); } diff --git a/vnet/vnet/mpls/mpls.h b/vnet/vnet/mpls/mpls.h index 59fc761ea2c..b6fdbce7d70 100644 --- a/vnet/vnet/mpls/mpls.h +++ b/vnet/vnet/mpls/mpls.h @@ -30,30 +30,6 @@ typedef enum { MPLS_N_ERROR, } mpls_error_t; -/* - * No protocol info, MPLS labels don't have a next-header field - * presumably the label field tells all... - */ -typedef struct { - u8 tunnel_dst[6]; - ip4_address_t intfc_address; - u32 tx_sw_if_index; - u32 inner_fib_index; - u32 mask_width; - u32 encap_index; - u32 hw_if_index; - u8 * rewrite_data; - u8 l2_only; - fib_node_index_t fei; -} mpls_eth_tunnel_t; - -typedef struct { - mpls_unicast_header_t *labels; - /* only for policy tunnels */ - u8 * rewrite; - u32 output_next_index; -} mpls_encap_t; - #define MPLS_FIB_DEFAULT_TABLE_ID 0 /** @@ -94,18 +70,6 @@ typedef struct { /** A hash table to lookup the mpls_fib by table ID */ uword *fib_index_by_table_id; - /* pool of ethernet tunnel instances */ - mpls_eth_tunnel_t *eth_tunnels; - u32 * free_eth_sw_if_indices; - - /* Encap side: map (fib, dst_address) to mpls label stack */ - mpls_encap_t * encaps; - uword * mpls_encap_by_fib_and_dest; - - /* mpls-o-e policy tunnel next index for ip4/ip6-classify */ - u32 ip4_classify_mpls_policy_encap_next_index; - u32 ip6_classify_mpls_policy_encap_next_index; - /* Feature arc indices */ u8 input_feature_arc_index; u8 output_feature_arc_index; @@ -123,7 +87,6 @@ extern mpls_main_t mpls_main; extern clib_error_t * mpls_feature_init(vlib_main_t * vm); format_function_t format_mpls_protocol; -format_function_t format_mpls_eth_header_with_length; format_function_t format_mpls_encap_index; format_function_t format_mpls_eos_bit; @@ -153,29 +116,8 @@ void mpls_sw_interface_enable_disable (mpls_main_t * mm, u8 mpls_sw_interface_is_enabled (u32 sw_if_index); -mpls_encap_t * -mpls_encap_by_fib_and_dest (mpls_main_t * mm, u32 rx_fib, u32 dst_address); - -int vnet_mpls_ethernet_add_del_tunnel (u8 *dst, - ip4_address_t *intfc, - u32 mask_width, - u32 inner_fib_id, - u32 tx_sw_if_index, - u32 * tunnel_sw_if_index, - u8 l2_only, - u8 is_add); - int mpls_fib_reset_labels (u32 fib_id); -int vnet_mpls_add_del_encap (ip4_address_t *dest, u32 fib_id, - u32 *labels_host_byte_order, - u32 policy_tunnel_index, - int no_dst_hash, u32 * indexp, int is_add); - -int vnet_mpls_policy_tunnel_add_rewrite (mpls_main_t * mm, - mpls_encap_t * e, - u32 policy_tunnel_index); - #define foreach_mpls_input_next \ _(DROP, "error-drop") \ _(LOOKUP, "mpls-lookup") @@ -211,26 +153,6 @@ typedef enum { } mpls_output_next_t; typedef struct { - u32 lookup_miss; - - /* Tunnel-id / index in tunnel vector */ - u32 tunnel_id; - - /* output interface */ - u32 tx_sw_if_index; - - /* mpls encap index */ - u32 mpls_encap_index; - - /* pkt length */ - u32 length; - - u8 dst[6]; -} mpls_eth_tx_trace_t; - -u8 * format_mpls_eth_tx_trace (u8 * s, va_list * args); - -typedef struct { u32 fib_index; u32 entry_index; u32 dest; diff --git a/vnet/vnet/mpls/mpls_lookup.c b/vnet/vnet/mpls/mpls_lookup.c index 34ba79e40fd..2d34cbde341 100644 --- a/vnet/vnet/mpls/mpls_lookup.c +++ b/vnet/vnet/mpls/mpls_lookup.c @@ -170,6 +170,19 @@ mpls_lookup (vlib_main_t * vm, vlib_buffer_length_in_chain (vm, b1)); /* + * before we pop the label copy th values we need to maintain. + * The label header is in network byte order. + * last byte is the TTL. + * bits 2 to 4 inclusive are the EXP bits + */ + vnet_buffer (b0)->mpls.ttl = ((char*)h0)[3]; + vnet_buffer (b0)->mpls.exp = (((char*)h0)[2] & 0xe) >> 1; + vnet_buffer (b0)->mpls.first = 1; + vnet_buffer (b1)->mpls.ttl = ((char*)h1)[3]; + vnet_buffer (b1)->mpls.exp = (((char*)h1)[2] & 0xe) >> 1; + vnet_buffer (b1)->mpls.first = 1; + + /* * pop the label that was just used in the lookup */ vlib_buffer_advance(b0, sizeof(*h0)); @@ -223,8 +236,8 @@ mpls_lookup (vlib_main_t * vm, lfib_index0 = vec_elt(mm->fib_index_by_sw_if_index, vnet_buffer(b0)->sw_if_index[VLIB_RX]); - lbi0 = mpls_fib_table_forwarding_lookup (lfib_index0, h0); - lb0 = load_balance_get(lbi0); + lbi0 = mpls_fib_table_forwarding_lookup(lfib_index0, h0); + lb0 = load_balance_get(lbi0); hash_c0 = vnet_buffer(b0)->ip.flow_hash = 0; if (PREDICT_FALSE(lb0->lb_n_buckets > 1)) @@ -248,6 +261,16 @@ mpls_lookup (vlib_main_t * vm, vlib_buffer_length_in_chain (vm, b0)); /* + * before we pop the label copy th values we need to maintain. + * The label header is in network byte order. + * last byte is the TTL. + * bits 2 to 4 inclusive are the EXP bits + */ + vnet_buffer (b0)->mpls.ttl = ((char*)h0)[3]; + vnet_buffer (b0)->mpls.exp = (((char*)h0)[2] & 0xe) >> 1; + vnet_buffer (b0)->mpls.first = 1; + + /* * pop the label that was just used in the lookup */ vlib_buffer_advance(b0, sizeof(*h0)); diff --git a/vnet/vnet/mpls/mpls_tunnel.c b/vnet/vnet/mpls/mpls_tunnel.c new file mode 100644 index 00000000000..656bf330b1a --- /dev/null +++ b/vnet/vnet/mpls/mpls_tunnel.c @@ -0,0 +1,779 @@ +/* + * mpls_tunnel.c: MPLS tunnel interfaces (i.e. for RSVP-TE) + * + * Copyright (c) 2012 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/vnet.h> +#include <vnet/pg/pg.h> +#include <vnet/mpls/mpls_tunnel.h> +#include <vnet/ip/ip.h> +#include <vnet/fib/fib_path_list.h> +#include <vnet/adj/adj_midchain.h> + +/** + * @brief pool of tunnel instances + */ +static mpls_tunnel_t *mpls_tunnel_pool; + +/** + * @brief Pool of free tunnel SW indices - i.e. recycled indices + */ +static u32 * mpls_tunnel_free_hw_if_indices; + +/** + * @brief DB of SW index to tunnel index + */ +static u32 *mpls_tunnel_db; + +/** + * @brief Get a tunnel object from a SW interface index + */ +static mpls_tunnel_t* +mpls_tunnel_get_from_sw_if_index (u32 sw_if_index) +{ + if ((vec_len(mpls_tunnel_db) < sw_if_index) || + (~0 == mpls_tunnel_db[sw_if_index])) + return (NULL); + + return (pool_elt_at_index(mpls_tunnel_pool, + mpls_tunnel_db[sw_if_index])); +} + +/** + * @brief Return true if the label stack is imp-null only + */ +static fib_forward_chain_type_t +mpls_tunnel_get_fwd_chain_type (const mpls_tunnel_t *mt) +{ + if ((1 == vec_len(mt->mt_label_stack)) && + (mt->mt_label_stack[0] == MPLS_IETF_IMPLICIT_NULL_LABEL)) + { + /* + * the only label in the label stack is implicit null + * we need to build an IP chain. + */ + if (FIB_PROTOCOL_IP4 == fib_path_list_get_proto(mt->mt_path_list)) + { + return (FIB_FORW_CHAIN_TYPE_UNICAST_IP4); + } + else + { + return (FIB_FORW_CHAIN_TYPE_UNICAST_IP6); + } + } + else + { + return (FIB_FORW_CHAIN_TYPE_MPLS_NON_EOS); + } +} + +/** + * @brief Build a rewrite string for the MPLS tunnel. + * + * We have choices here; + * 1 - have an Adjacency with a zero length string and stack it on + * MPLS label objects + * 2 - put the label header rewrites in the adjacency string. + * + * We choose 2 since it results in fewer graph nodes in the egress path + */ +static u8* +mpls_tunnel_build_rewrite (vnet_main_t * vnm, + u32 sw_if_index, + vnet_link_t link_type, + const void *dst_address) +{ + mpls_unicast_header_t *muh; + mpls_tunnel_t *mt; + u8 *rewrite; + u32 mti, ii; + + rewrite = NULL; + mti = mpls_tunnel_db[sw_if_index]; + mt = pool_elt_at_index(mpls_tunnel_pool, mti); + + vec_validate(rewrite, (sizeof(*muh) * vec_len(mt->mt_label_stack)) - 1); + muh = (mpls_unicast_header_t *)rewrite; + + /* + * The last (inner most) label in the stack may be EOS, all the rest Non-EOS + */ + for (ii = 0; ii < vec_len(mt->mt_label_stack)-1; ii++) + { + vnet_mpls_uc_set_label(&muh[ii].label_exp_s_ttl, mt->mt_label_stack[ii]); + vnet_mpls_uc_set_ttl(&muh[ii].label_exp_s_ttl, 255); + vnet_mpls_uc_set_exp(&muh[ii].label_exp_s_ttl, 0); + vnet_mpls_uc_set_s(&muh[ii].label_exp_s_ttl, MPLS_NON_EOS); + muh[ii].label_exp_s_ttl = clib_host_to_net_u32(muh[ii].label_exp_s_ttl); + } + + vnet_mpls_uc_set_label(&muh[ii].label_exp_s_ttl, mt->mt_label_stack[ii]); + vnet_mpls_uc_set_ttl(&muh[ii].label_exp_s_ttl, 255); + vnet_mpls_uc_set_exp(&muh[ii].label_exp_s_ttl, 0); + + if (VNET_LINK_MPLS == link_type && + mt->mt_label_stack[ii] != MPLS_IETF_IMPLICIT_NULL_LABEL) + vnet_mpls_uc_set_s(&muh[ii].label_exp_s_ttl, MPLS_NON_EOS); + else + vnet_mpls_uc_set_s(&muh[ii].label_exp_s_ttl, MPLS_EOS); + + muh[ii].label_exp_s_ttl = clib_host_to_net_u32(muh[ii].label_exp_s_ttl); + + return ((u8*)muh); +} + +/** + * mpls_tunnel_stack + * + * 'stack' (resolve the recursion for) the tunnel's midchain adjacency + */ +static void +mpls_tunnel_stack (adj_index_t ai) +{ + ip_adjacency_t *adj; + mpls_tunnel_t *mt; + u32 sw_if_index; + + adj = adj_get(ai); + sw_if_index = adj->rewrite_header.sw_if_index; + + mt = mpls_tunnel_get_from_sw_if_index(sw_if_index); + + if (NULL == mt) + return; + + /* + * find the adjacency that is contributed by the FIB path-list + * that this tunnel resovles via, and use it as the next adj + * in the midchain + */ + if (vnet_hw_interface_get_flags(vnet_get_main(), + mt->mt_hw_if_index) & + VNET_HW_INTERFACE_FLAG_LINK_UP) + { + dpo_id_t dpo = DPO_INVALID; + + fib_path_list_contribute_forwarding(mt->mt_path_list, + mpls_tunnel_get_fwd_chain_type(mt), + &dpo); + + if (DPO_LOAD_BALANCE == dpo.dpoi_type) + { + /* + * we don't support multiple paths, so no need to load-balance. + * pull the first and only choice and stack directly on that. + */ + load_balance_t *lb; + + lb = load_balance_get (dpo.dpoi_index); + + ASSERT(1 == lb->lb_n_buckets); + + dpo_copy(&dpo, load_balance_get_bucket_i (lb, 0)); + } + + adj_nbr_midchain_stack(ai, &dpo); + dpo_reset(&dpo); + } + else + { + adj_nbr_midchain_unstack(ai); + } +} + +/** + * @brief Call back when restacking all adjacencies on a MPLS interface + */ +static adj_walk_rc_t +mpls_adj_walk_cb (adj_index_t ai, + void *ctx) +{ + mpls_tunnel_stack(ai); + + return (ADJ_WALK_RC_CONTINUE); +} + +static void +mpls_tunnel_restack (mpls_tunnel_t *mt) +{ + fib_protocol_t proto; + + /* + * walk all the adjacencies on the MPLS interface and restack them + */ + FOR_EACH_FIB_PROTOCOL(proto) + { + adj_nbr_walk(mt->mt_sw_if_index, + proto, + mpls_adj_walk_cb, + NULL); + } +} + +static clib_error_t * +mpls_tunnel_admin_up_down (vnet_main_t * vnm, + u32 hw_if_index, + u32 flags) +{ + vnet_hw_interface_t * hi; + mpls_tunnel_t *mt; + + hi = vnet_get_hw_interface (vnm, hw_if_index); + + mt = mpls_tunnel_get_from_sw_if_index(hi->sw_if_index); + + if (NULL == mt) + return (NULL); + + if (flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) + vnet_hw_interface_set_flags (vnm, hw_if_index, + VNET_HW_INTERFACE_FLAG_LINK_UP); + else + vnet_hw_interface_set_flags (vnm, hw_if_index, 0 /* down */); + + mpls_tunnel_restack(mt); + + return (NULL); +} + +/** + * @brief Fixup the adj rewrite post encap. This is a no-op since the + * rewrite is a stack of labels. + */ +static void +mpls_tunnel_fixup (vlib_main_t *vm, + ip_adjacency_t *adj, + vlib_buffer_t *b0) +{ +} + +static void +mpls_tunnel_update_adj (vnet_main_t * vnm, + u32 sw_if_index, + adj_index_t ai) +{ + adj_nbr_midchain_update_rewrite( + ai, mpls_tunnel_fixup, + ADJ_MIDCHAIN_FLAG_NONE, + mpls_tunnel_build_rewrite(vnm, sw_if_index, + adj_get_link_type(ai), + NULL)); + + mpls_tunnel_stack(ai); +} + +static u8 * +format_mpls_tunnel_name (u8 * s, va_list * args) +{ + u32 dev_instance = va_arg (*args, u32); + return format (s, "mpls-tunnel%d", dev_instance); +} + +static u8 * +format_mpls_tunnel_device (u8 * s, va_list * args) +{ + u32 dev_instance = va_arg (*args, u32); + CLIB_UNUSED (int verbose) = va_arg (*args, int); + + return (format (s, "MPLS-tunnel: id %d\n", dev_instance)); +} + +/** + * @brief Packet trace structure + */ +typedef struct mpls_tunnel_trace_t_ +{ + /** + * Tunnel-id / index in tunnel vector + */ + u32 tunnel_id; +} mpls_tunnel_trace_t; + +static u8 * +format_mpls_tunnel_tx_trace (u8 * s, + va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + mpls_tunnel_trace_t * t = va_arg (*args, mpls_tunnel_trace_t *); + + s = format (s, "MPLS: tunnel %d", t->tunnel_id); + return s; +} + +/** + * @brief TX function. Only called L2. L3 traffic uses the adj-midchains + */ +static uword +mpls_tunnel_tx (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame) +{ + u32 next_index; + u32 * from, * to_next, n_left_from, n_left_to_next; + vnet_interface_output_runtime_t * rd = (void *) node->runtime_data; + const mpls_tunnel_t *mt; + + mt = pool_elt_at_index(mpls_tunnel_pool, rd->dev_instance); + + /* Vector of buffer / pkt indices we're supposed to process */ + from = vlib_frame_vector_args (frame); + + /* Number of buffers / pkts */ + n_left_from = frame->n_vectors; + + /* Speculatively send the first buffer to the last disposition we used */ + next_index = node->cached_next_index; + + while (n_left_from > 0) + { + /* 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 > 0 && n_left_to_next > 0) + { + vlib_buffer_t * b0; + u32 bi0; + + 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); + + vnet_buffer(b0)->ip.adj_index[VLIB_TX] = mt->mt_l2_adj; + + if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) + { + mpls_tunnel_trace_t *tr = vlib_add_trace (vm, node, + b0, sizeof (*tr)); + tr->tunnel_id = rd->dev_instance; + } + + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, + bi0, mt->mt_l2_tx_arc); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + return frame->n_vectors; +} + +VNET_DEVICE_CLASS (mpls_tunnel_class) = { + .name = "MPLS tunnel device", + .format_device_name = format_mpls_tunnel_name, + .format_device = format_mpls_tunnel_device, + .format_tx_trace = format_mpls_tunnel_tx_trace, + .tx_function = mpls_tunnel_tx, + .no_flatten_output_chains = 1, + .admin_up_down_function = mpls_tunnel_admin_up_down, +}; + +VNET_HW_INTERFACE_CLASS (mpls_tunnel_hw_interface_class) = { + .name = "MPLS-Tunnel", +// .format_header = format_mpls_eth_header_with_length, +// .unformat_header = unformat_mpls_eth_header, + .update_adjacency = mpls_tunnel_update_adj, + .build_rewrite = mpls_tunnel_build_rewrite, + .flags = VNET_HW_INTERFACE_CLASS_FLAG_P2P, +}; + +const mpls_tunnel_t * +mpls_tunnel_get (u32 mti) +{ + return (pool_elt_at_index(mpls_tunnel_pool, mti)); +} + +/** + * @brief Walk all the MPLS tunnels + */ +void +mpls_tunnel_walk (mpls_tunnel_walk_cb_t cb, + void *ctx) +{ + u32 mti; + + pool_foreach_index(mti, mpls_tunnel_pool, + ({ + cb(mti, ctx); + })); +} + +void +vnet_mpls_tunnel_del (u32 sw_if_index) +{ + mpls_tunnel_t *mt; + + mt = mpls_tunnel_get_from_sw_if_index(sw_if_index); + + if (NULL == mt) + return; + + fib_path_list_child_remove(mt->mt_path_list, + mt->mt_sibling_index); + if (ADJ_INDEX_INVALID != mt->mt_l2_adj) + adj_unlock(mt->mt_l2_adj); + + vec_free(mt->mt_label_stack); + + vec_add1 (mpls_tunnel_free_hw_if_indices, mt->mt_hw_if_index); + pool_put(mpls_tunnel_pool, mt); + mpls_tunnel_db[sw_if_index] = ~0; +} + +void +vnet_mpls_tunnel_add (fib_route_path_t *rpaths, + mpls_label_t *label_stack, + u8 l2_only, + u32 *sw_if_index) +{ + vnet_hw_interface_t * hi; + mpls_tunnel_t *mt; + vnet_main_t * vnm; + u32 mti; + + vnm = vnet_get_main(); + pool_get(mpls_tunnel_pool, mt); + memset (mt, 0, sizeof (*mt)); + mti = mt - mpls_tunnel_pool; + fib_node_init(&mt->mt_node, FIB_NODE_TYPE_MPLS_TUNNEL); + mt->mt_l2_adj = ADJ_INDEX_INVALID; + + /* + * Create a new, or re=use and old, tunnel HW interface + */ + if (vec_len (mpls_tunnel_free_hw_if_indices) > 0) + { + mt->mt_hw_if_index = + mpls_tunnel_free_hw_if_indices[vec_len(mpls_tunnel_free_hw_if_indices)-1]; + _vec_len (mpls_tunnel_free_hw_if_indices) -= 1; + hi = vnet_get_hw_interface (vnm, mt->mt_hw_if_index); + hi->hw_instance = mti; + hi->dev_instance = mti; + } + else + { + mt->mt_hw_if_index = vnet_register_interface( + vnm, + mpls_tunnel_class.index, + mti, + mpls_tunnel_hw_interface_class.index, + mti); + hi = vnet_get_hw_interface(vnm, mt->mt_hw_if_index); + } + + /* + * Add the new tunnel to the tunnel DB - key:SW if index + */ + mt->mt_sw_if_index = hi->sw_if_index; + vec_validate_init_empty(mpls_tunnel_db, mt->mt_sw_if_index, ~0); + mpls_tunnel_db[mt->mt_sw_if_index] = mti; + + /* + * construct a path-list from the path provided + */ + mt->mt_path_list = fib_path_list_create(FIB_PATH_LIST_FLAG_SHARED, rpaths); + mt->mt_sibling_index = fib_path_list_child_add(mt->mt_path_list, + FIB_NODE_TYPE_MPLS_TUNNEL, + mti); + + mt->mt_label_stack = vec_dup(label_stack); + + if (l2_only) + { + mt->mt_l2_adj = + adj_nbr_add_or_lock(fib_path_list_get_proto(mt->mt_path_list), + VNET_LINK_ETHERNET, + &zero_addr, + mt->mt_sw_if_index); + + mt->mt_l2_tx_arc = vlib_node_add_named_next(vlib_get_main(), + hi->tx_node_index, + "adj-l2-midchain"); + } + + *sw_if_index = mt->mt_sw_if_index; +} + +static clib_error_t * +vnet_create_mpls_tunnel_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + unformat_input_t _line_input, * line_input = &_line_input; + vnet_main_t * vnm = vnet_get_main(); + u8 is_del = 0; + u8 l2_only = 0; + fib_route_path_t rpath, *rpaths = NULL; + mpls_label_t out_label = MPLS_LABEL_INVALID, *labels = NULL; + u32 sw_if_index; + + memset(&rpath, 0, sizeof(rpath)); + + /* Get a line of input. */ + if (! unformat_user (input, unformat_line_input, line_input)) + return 0; + + while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (line_input, "del %U", + unformat_vnet_sw_interface, vnm, + &sw_if_index)) + is_del = 1; + else if (unformat (line_input, "add")) + is_del = 0; + else if (unformat (line_input, "out-label %U", + unformat_mpls_unicast_label, &out_label)) + { + vec_add1(labels, out_label); + } + else if (unformat (line_input, "via %U %U", + unformat_ip4_address, + &rpath.frp_addr.ip4, + unformat_vnet_sw_interface, vnm, + &rpath.frp_sw_if_index)) + { + rpath.frp_weight = 1; + rpath.frp_proto = FIB_PROTOCOL_IP4; + } + + else if (unformat (line_input, "via %U %U", + unformat_ip6_address, + &rpath.frp_addr.ip6, + unformat_vnet_sw_interface, vnm, + &rpath.frp_sw_if_index)) + { + rpath.frp_weight = 1; + rpath.frp_proto = FIB_PROTOCOL_IP6; + } + else if (unformat (line_input, "via %U", + unformat_ip6_address, + &rpath.frp_addr.ip6)) + { + rpath.frp_fib_index = 0; + rpath.frp_weight = 1; + rpath.frp_sw_if_index = ~0; + rpath.frp_proto = FIB_PROTOCOL_IP6; + } + else if (unformat (line_input, "via %U", + unformat_ip4_address, + &rpath.frp_addr.ip4)) + { + rpath.frp_fib_index = 0; + rpath.frp_weight = 1; + rpath.frp_sw_if_index = ~0; + rpath.frp_proto = FIB_PROTOCOL_IP4; + } + else if (unformat (line_input, "l2-only")) + l2_only = 1; + else + return clib_error_return (0, "unknown input '%U'", + format_unformat_error, line_input); + } + + if (is_del) + { + vnet_mpls_tunnel_del(sw_if_index); + } + else + { + if (0 == vec_len(labels)) + return clib_error_return (0, "No Output Labels '%U'", + format_unformat_error, line_input); + + vec_add1(rpaths, rpath); + vnet_mpls_tunnel_add(rpaths, labels, l2_only, &sw_if_index); + } + + vec_free(labels); + vec_free(rpaths); + + return (NULL); +} + +/*? + * This command create a uni-directional MPLS tunnel + * + * @cliexpar + * @cliexstart{create mpls tunnel} + * create mpls tunnel via 10.0.0.1 GigEthernet0/8/0 out-label 33 out-label 34 + * @cliexend + ?*/ +VLIB_CLI_COMMAND (create_mpls_tunnel_command, static) = { + .path = "mpls tunnel", + .short_help = + "mpls tunnel via [addr] [interface] [out-labels]", + .function = vnet_create_mpls_tunnel_command_fn, +}; + +static u8 * +format_mpls_tunnel (u8 * s, va_list * args) +{ + mpls_tunnel_t *mt = va_arg (*args, mpls_tunnel_t *); + int ii; + + s = format(s, "mpls_tunnel%d: sw_if_index:%d hw_if_index:%d", + mt - mpls_tunnel_pool, + mt->mt_sw_if_index, + mt->mt_hw_if_index); + s = format(s, "\n label-stack:\n "); + for (ii = 0; ii < vec_len(mt->mt_label_stack); ii++) + { + s = format(s, "%d, ", mt->mt_label_stack[ii]); + } + s = format(s, "\n via:\n"); + s = fib_path_list_format(mt->mt_path_list, s); + s = format(s, "\n"); + + return (s); +} + +static clib_error_t * +show_mpls_tunnel_command_fn (vlib_main_t * vm, + unformat_input_t * input, + vlib_cli_command_t * cmd) +{ + mpls_tunnel_t * mt; + u32 mti = ~0; + + if (pool_elts (mpls_tunnel_pool) == 0) + vlib_cli_output (vm, "No MPLS tunnels configured..."); + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "%d", &mti)) + ; + else + break; + } + + if (~0 == mti) + { + pool_foreach (mt, mpls_tunnel_pool, + ({ + vlib_cli_output (vm, "[@%d] %U", + mt - mpls_tunnel_pool, + format_mpls_tunnel, mt); + })); + } + else + { + if (pool_is_free_index(mpls_tunnel_pool, mti)) + return clib_error_return (0, "Not atunnel index %d", mti); + + mt = pool_elt_at_index(mpls_tunnel_pool, mti); + + vlib_cli_output (vm, "[@%d] %U", + mt - mpls_tunnel_pool, + format_mpls_tunnel, mt); + } + + return 0; +} + +/*? + * This command to show MPLS tunnels + * + * @cliexpar + * @cliexstart{sh mpls tunnel 2} + * [@2] mpls_tunnel2: sw_if_index:5 hw_if_index:5 + * label-stack: + * 3, + * via: + * index:26 locks:1 proto:ipv4 uPRF-list:26 len:1 itfs:[2, ] + * index:26 pl-index:26 ipv4 weight=1 attached-nexthop: oper-flags:resolved, + * 10.0.0.2 loop0 + * [@0]: ipv4 via 10.0.0.2 loop0: IP4: de:ad:00:00:00:00 -> 00:00:11:aa:bb:cc + * @cliexend + ?*/ +VLIB_CLI_COMMAND (show_mpls_tunnel_command, static) = { + .path = "show mpls tunnel", + .function = show_mpls_tunnel_command_fn, +}; + +static mpls_tunnel_t * +mpls_tunnel_from_fib_node (fib_node_t *node) +{ +#if (CLIB_DEBUG > 0) + ASSERT(FIB_NODE_TYPE_MPLS_TUNNEL == node->fn_type); +#endif + return ((mpls_tunnel_t*) (((char*)node) - + STRUCT_OFFSET_OF(mpls_tunnel_t, mt_node))); +} + +/** + * Function definition to backwalk a FIB node + */ +static fib_node_back_walk_rc_t +mpls_tunnel_back_walk (fib_node_t *node, + fib_node_back_walk_ctx_t *ctx) +{ + mpls_tunnel_restack(mpls_tunnel_from_fib_node(node)); + + return (FIB_NODE_BACK_WALK_CONTINUE); +} + +/** + * Function definition to get a FIB node from its index + */ +static fib_node_t* +mpls_tunnel_fib_node_get (fib_node_index_t index) +{ + mpls_tunnel_t * mt; + + mt = pool_elt_at_index(mpls_tunnel_pool, index); + + return (&mt->mt_node); +} + +/** + * Function definition to inform the FIB node that its last lock has gone. + */ +static void +mpls_tunnel_last_lock_gone (fib_node_t *node) +{ + /* + * The MPLS MPLS tunnel is a root of the graph. As such + * it never has children and thus is never locked. + */ + ASSERT(0); +} + +/* + * Virtual function table registered by MPLS MPLS tunnels + * for participation in the FIB object graph. + */ +const static fib_node_vft_t mpls_vft = { + .fnv_get = mpls_tunnel_fib_node_get, + .fnv_last_lock = mpls_tunnel_last_lock_gone, + .fnv_back_walk = mpls_tunnel_back_walk, +}; + +static clib_error_t * +mpls_tunnel_init (vlib_main_t *vm) +{ + fib_node_register_type(FIB_NODE_TYPE_MPLS_TUNNEL, &mpls_vft); + + return 0; +} +VLIB_INIT_FUNCTION(mpls_tunnel_init); diff --git a/vnet/vnet/mpls/mpls_tunnel.h b/vnet/vnet/mpls/mpls_tunnel.h new file mode 100644 index 00000000000..ee56c0fc8e3 --- /dev/null +++ b/vnet/vnet/mpls/mpls_tunnel.h @@ -0,0 +1,98 @@ +/* + * Copyright (c) 2015 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 __MPLS_TUNNEL_H__ +#define __MPLS_TUNNEL_H__ + +#include <vnet/mpls/mpls.h> + +/** + * @brief A uni-directional MPLS tunnel + */ +typedef struct mpls_tunnel_t_ +{ + /** + * @brief The tunnel hooks into the FIB control plane graph. + */ + fib_node_t mt_node; + + /** + * @brief If the tunnel is an L2 tunnel, this is the link type ETHERNET + * adjacency + */ + adj_index_t mt_l2_adj; + + /** + * @brief on a L2 tunnel this is the VLIB arc from the L2-tx to the l2-midchain + */ + u32 mt_l2_tx_arc; + + /** + * @brief The path-list over which the tunnel's destination is reachable + */ + fib_node_index_t mt_path_list; + + /** + * @brief sibling index on the path-list so notifications are received. + */ + u32 mt_sibling_index; + + /** + * @brief The Label stack to apply to egress packets + */ + mpls_label_t *mt_label_stack; + + /** + * @brief Flag to indicate the tunnel is only for L2 traffic, that is + * this tunnel belongs in a bridge domain. + */ + u8 mt_l2_only; + + /** + * @brief The HW interface index of the tunnel interfaces + */ + u32 mt_hw_if_index; + + /** + * @brief The SW interface index of the tunnel interfaces + */ + u32 mt_sw_if_index; + +} mpls_tunnel_t; + +/** + * @brief Create a new MPLS tunnel + */ +extern void vnet_mpls_tunnel_add (fib_route_path_t *rpath, + mpls_label_t *label_stack, + u8 l2_only, + u32 *sw_if_index); + +extern void vnet_mpls_tunnel_del (u32 sw_if_index); + +extern const mpls_tunnel_t *mpls_tunnel_get(u32 index); + +/** + * @brief Callback function invoked while walking MPLS tunnels + */ +typedef void (*mpls_tunnel_walk_cb_t)(u32 index, void *ctx); + +/** + * @brief Walk all the MPLS tunnels + */ +extern void mpls_tunnel_walk(mpls_tunnel_walk_cb_t cb, + void *ctx); + +#endif diff --git a/vnet/vnet/mpls/policy_encap.c b/vnet/vnet/mpls/policy_encap.c deleted file mode 100644 index d48a153b37d..00000000000 --- a/vnet/vnet/mpls/policy_encap.c +++ /dev/null @@ -1,181 +0,0 @@ -/* - * policy_encap.c: mpls-o-e policy encap - * - * Copyright (c) 2012-2014 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 <vlib/vlib.h> -#include <vnet/pg/pg.h> -#include <vnet/mpls/mpls.h> -#include <vnet/classify/vnet_classify.h> - -typedef struct { - u32 next_index; - u32 encap_index; -} mpls_policy_encap_trace_t; - -u8 * format_mpls_policy_encap_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - mpls_policy_encap_trace_t * t = va_arg (*args, mpls_policy_encap_trace_t *); - - s = format (s, "MPLS-POLICY-ENCAP: next-index %d encap-index %d", - t->next_index, t->encap_index); - - return s; -} - -vlib_node_registration_t mpls_policy_encap_node; - -#define foreach_mpls_policy_encap_next \ -_(DROP, "error-drop") - -typedef enum { -#define _(s,n) MPLS_POLICY_ENCAP_NEXT_##s, - foreach_mpls_policy_encap_next -#undef _ - MPLS_POLICY_ENCAP_N_NEXT, -} mpls_policy_encap_next_t; - -#define foreach_mpls_policy_error \ -_(PKTS_ENCAP, "mpls policy tunnel packets encapsulated") - -typedef enum { -#define _(n,s) MPLS_POLICY_ENCAP_ERROR_##n, - foreach_mpls_policy_error - MPLS_POLICY_ENCAP_N_ERROR, -#undef _ -} mpls_policy_encap_error_t; - -static char * mpls_policy_encap_error_strings[] = - { -#define _(n,s) s, - foreach_mpls_policy_error -#undef _ -}; - -static uword -mpls_policy_encap (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * from_frame) -{ - u32 n_left_from, next_index, * from, * to_next; - mpls_main_t * mm = &mpls_main; - - from = vlib_frame_vector_args (from_frame); - n_left_from = from_frame->n_vectors; - - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - 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; - vlib_buffer_t * b0; - u8 * h0; - u32 encap_index0; - u32 next0; - mpls_encap_t * e0; - - 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); - - encap_index0 = vnet_buffer(b0)->l2_classify.opaque_index; - - e0 = pool_elt_at_index (mm->encaps, encap_index0); - - vlib_buffer_advance (b0, -(word)vec_len(e0->rewrite)); - h0 = vlib_buffer_get_current (b0); - clib_memcpy (h0, e0->rewrite, vec_len(e0->rewrite)); - - next0 = e0->output_next_index; - - if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) - { - mpls_policy_encap_trace_t *tr = - vlib_add_trace (vm, node, b0, sizeof (*tr)); - tr->next_index = next0; - tr->encap_index = encap_index0; - } - 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); - } - vlib_node_increment_counter (vm, mpls_policy_encap_node.index, - MPLS_POLICY_ENCAP_ERROR_PKTS_ENCAP, - from_frame->n_vectors); - return from_frame->n_vectors; -} - -VLIB_REGISTER_NODE (mpls_policy_encap_node) = { - .function = mpls_policy_encap, - .name = "mpls-policy-encap", - /* Takes a vector of packets. */ - .vector_size = sizeof (u32), - - .runtime_data_bytes = 0, - - .n_errors = MPLS_POLICY_ENCAP_N_ERROR, - .error_strings = mpls_policy_encap_error_strings, - - .format_trace = format_mpls_policy_encap_trace, - - .n_next_nodes = MPLS_POLICY_ENCAP_N_NEXT, - .next_nodes = { -#define _(s,n) [MPLS_POLICY_ENCAP_NEXT_##s] = n, - foreach_mpls_policy_encap_next -#undef _ - }, -}; - -VLIB_NODE_FUNCTION_MULTIARCH (mpls_policy_encap_node, mpls_policy_encap) - -static clib_error_t * -mpls_policy_encap_init (vlib_main_t * vm) -{ - mpls_main_t * mm = &mpls_main; - clib_error_t * error; - - if ((error = vlib_call_init_function (vm, mpls_init))) - return error; - - mm->ip4_classify_mpls_policy_encap_next_index = - vlib_node_add_next (mm->vlib_main, - ip4_classify_node.index, - mpls_policy_encap_node.index); - - mm->ip6_classify_mpls_policy_encap_next_index = - vlib_node_add_next (mm->vlib_main, - ip6_classify_node.index, - mpls_policy_encap_node.index); - - return 0; -} - -VLIB_INIT_FUNCTION (mpls_policy_encap_init); |