From a9374df5f351d25e968f5f90a827796203cbafdd Mon Sep 17 00:00:00 2001 From: Neale Ranns Date: Thu, 2 Feb 2017 02:18:18 -0800 Subject: Fix SR multicast post mfib commit 1 - use the SR policy to construct the replicate DPO. Each bucket therein is a SR tunnel. 2 - install a special mfib entry that links via this replicate 3 - forwarding is now mfib-lookup -> replicate -> sr_rewrite (per-tunnel) no need for a separate sr_replicate node. 4 - Stack the sr tunnel on the forwarding DPO of the first-hop FIB entry. no need for a second lookup post SR encap. 5 - fix some path-list lock leaks in the MFIB entry. Change-Id: I20de96ea4c4be4fae252625bde159d9c435c8315 Signed-off-by: Neale Ranns --- src/vnet/sr/sr.c | 327 +++++++++++++++++++----------- src/vnet/sr/sr.h | 30 ++- src/vnet/sr/sr_api.c | 5 - src/vnet/sr/sr_replicate.c | 491 --------------------------------------------- 4 files changed, 235 insertions(+), 618 deletions(-) delete mode 100644 src/vnet/sr/sr_replicate.c (limited to 'src/vnet/sr') diff --git a/src/vnet/sr/sr.c b/src/vnet/sr/sr.c index f30c0da940e..012d45428ec 100644 --- a/src/vnet/sr/sr.c +++ b/src/vnet/sr/sr.c @@ -23,7 +23,9 @@ #include #include #include +#include #include +#include #include @@ -35,6 +37,11 @@ static vlib_node_registration_t sr_local_node; */ static dpo_type_t sr_dpo_type; +/** + * @brief Dynamically added SR FIB Node type + */ +static fib_node_type_t sr_fib_node_type; + /** * @brief Use passed HMAC key in ip6_sr_header_t in OpenSSL HMAC routines * @@ -258,20 +265,10 @@ format_ip6_sr_header_with_length (u8 * s, va_list * args) /** * @brief Defined valid next nodes - * @note Cannot call replicate yet without DPDK */ -#if DPDK > 0 -#define foreach_sr_rewrite_next \ -_(ERROR, "error-drop") \ -_(IP6_LOOKUP, "ip6-lookup") \ -_(SR_LOCAL, "sr-local") \ -_(SR_REPLICATE,"sr-replicate") -#else #define foreach_sr_rewrite_next \ _(ERROR, "error-drop") \ -_(IP6_LOOKUP, "ip6-lookup") \ _(SR_LOCAL, "sr-local") -#endif /* DPDK */ /** * @brief Struct for defined valid next nodes @@ -384,8 +381,8 @@ sr_rewrite (vlib_main_t * vm, ip6_header_t *ip0, *ip1; ip6_sr_header_t *sr0, *sr1; ip6_sr_tunnel_t *t0, *t1; - u32 next0 = SR_REWRITE_NEXT_IP6_LOOKUP; - u32 next1 = SR_REWRITE_NEXT_IP6_LOOKUP; + u32 next0; + u32 next1; u16 new_l0 = 0; u16 new_l1 = 0; @@ -433,16 +430,6 @@ sr_rewrite (vlib_main_t * vm, ip0 = vlib_buffer_get_current (b0); ip1 = vlib_buffer_get_current (b1); -#if DPDK > 0 /* Cannot call replication node yet without DPDK */ - /* add a replication node */ - if (PREDICT_FALSE (t0->policy_index != ~0)) - { - vnet_buffer (b0)->ip.save_protocol = t0->policy_index; - next0 = SR_REWRITE_NEXT_SR_REPLICATE; - sr0 = (ip6_sr_header_t *) (t0->rewrite); - goto processnext; - } -#endif /* DPDK */ /* * SR-unaware service chaining case: pkt coming back from @@ -506,8 +493,11 @@ sr_rewrite (vlib_main_t * vm, sr_fix_hmac (sm, ip0, sr0); - next0 = sr_local_cb ? sr_local_cb (vm, node, b0, ip0, sr0) : - next0; + vnet_buffer (b0)->ip.adj_index[VLIB_TX] = + t0->first_hop_dpo.dpoi_index; + next0 = t0->first_hop_dpo.dpoi_next_node; + next0 = (sr_local_cb ? + sr_local_cb (vm, node, b0, ip0, sr0) : next0); /* * Ignore "do not rewrite" shtik in this path @@ -519,17 +509,7 @@ sr_rewrite (vlib_main_t * vm, b0->error = node->errors[SR_REWRITE_ERROR_APP_CALLBACK]; } } -#if DPDK > 0 /* Cannot call replication node yet without DPDK */ - processnext: - /* add a replication node */ - if (PREDICT_FALSE (t1->policy_index != ~0)) - { - vnet_buffer (b1)->ip.save_protocol = t1->policy_index; - next1 = SR_REWRITE_NEXT_SR_REPLICATE; - sr1 = (ip6_sr_header_t *) (t1->rewrite); - goto trace00; - } -#endif /* DPDK */ + if (PREDICT_FALSE (ip1->protocol == IPPROTO_IPV6_ROUTE)) { vlib_buffer_advance (b1, sizeof (ip1)); @@ -584,8 +564,11 @@ sr_rewrite (vlib_main_t * vm, sr_fix_hmac (sm, ip1, sr1); - next1 = sr_local_cb ? sr_local_cb (vm, node, b1, ip1, sr1) : - next1; + vnet_buffer (b1)->ip.adj_index[VLIB_TX] = + t1->first_hop_dpo.dpoi_index; + next1 = t1->first_hop_dpo.dpoi_next_node; + next1 = (sr_local_cb ? + sr_local_cb (vm, node, b1, ip1, sr1) : next1); /* * Ignore "do not rewrite" shtik in this path @@ -597,9 +580,6 @@ sr_rewrite (vlib_main_t * vm, b1->error = node->errors[SR_REWRITE_ERROR_APP_CALLBACK]; } } -#if DPDK > 0 /* Cannot run replicate without DPDK and only replicate uses this label */ - trace00: -#endif /* DPDK */ if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { @@ -641,7 +621,7 @@ sr_rewrite (vlib_main_t * vm, ip6_header_t *ip0 = 0; ip6_sr_header_t *sr0 = 0; ip6_sr_tunnel_t *t0; - u32 next0 = SR_REWRITE_NEXT_IP6_LOOKUP; + u32 next0; u16 new_l0 = 0; bi0 = from[0]; @@ -661,16 +641,6 @@ sr_rewrite (vlib_main_t * vm, t0 = pool_elt_at_index (sm->tunnels, vnet_buffer (b0)->ip.adj_index[VLIB_TX]); -#if DPDK > 0 /* Cannot call replication node yet without DPDK */ - /* add a replication node */ - if (PREDICT_FALSE (t0->policy_index != ~0)) - { - vnet_buffer (b0)->ip.save_protocol = t0->policy_index; - next0 = SR_REWRITE_NEXT_SR_REPLICATE; - sr0 = (ip6_sr_header_t *) (t0->rewrite); - goto trace0; - } -#endif /* DPDK */ ASSERT (VLIB_BUFFER_PRE_DATA_SIZE >= ((word) vec_len (t0->rewrite)) + b0->current_data); @@ -740,8 +710,11 @@ sr_rewrite (vlib_main_t * vm, sr_fix_hmac (sm, ip0, sr0); - next0 = sr_local_cb ? sr_local_cb (vm, node, b0, ip0, sr0) : - next0; + vnet_buffer (b0)->ip.adj_index[VLIB_TX] = + t0->first_hop_dpo.dpoi_index; + next0 = t0->first_hop_dpo.dpoi_next_node; + next0 = (sr_local_cb ? + sr_local_cb (vm, node, b0, ip0, sr0) : next0); /* * Ignore "do not rewrite" shtik in this path @@ -753,9 +726,6 @@ sr_rewrite (vlib_main_t * vm, b0->error = node->errors[SR_REWRITE_ERROR_APP_CALLBACK]; } } -#if DPDK > 0 /* Cannot run replicate without DPDK and only replicate uses this label */ - trace0: -#endif /* DPDK */ if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { @@ -809,20 +779,69 @@ VLIB_NODE_FUNCTION_MULTIARCH (sr_rewrite_node, sr_rewrite) /* *INDENT-ON* */ static int -ip6_delete_route_no_next_hop (ip6_address_t * dst_address_arg, - u32 dst_address_length, u32 rx_table_id) +ip6_routes_add_del (ip6_sr_tunnel_t * t, int is_del) { + ip6_sr_main_t *sm = &sr_main; + + /* + * the prefix for the tunnel's destination + */ + /* *INDENT-OFF* */ fib_prefix_t pfx = { - .fp_len = dst_address_length, .fp_proto = FIB_PROTOCOL_IP6, + .fp_len = t->dst_mask_width, .fp_addr = { - .ip6 = *dst_address_arg, - } + .ip6 = t->key.dst, + } }; + /* *INDENT-ON* */ - fib_table_entry_delete (fib_table_id_find_fib_index (FIB_PROTOCOL_IP6, - rx_table_id), - &pfx, FIB_SOURCE_SR); + if (is_del) + { + fib_table_entry_delete (t->rx_fib_index, &pfx, FIB_SOURCE_SR); + } + else + { + dpo_id_t dpo = DPO_INVALID; + + dpo_set (&dpo, sr_dpo_type, DPO_PROTO_IP6, t - sm->tunnels); + fib_table_entry_special_dpo_add (t->rx_fib_index, + &pfx, + FIB_SOURCE_SR, + FIB_ENTRY_FLAG_EXCLUSIVE, &dpo); + dpo_reset (&dpo); + } + + /* + * Track the first hop address so we don't need to perform an extra + * lookup in the data-path + */ + /* *INDENT-OFF* */ + const fib_prefix_t first_hop_pfx = { + .fp_len = 128, + .fp_proto = FIB_PROTOCOL_IP6, + .fp_addr = { + .ip6 = t->first_hop, + } + }; + /* *INDENT-ON* */ + + if (is_del) + { + fib_entry_child_remove (t->fib_entry_index, t->sibling_index); + fib_table_entry_delete_index (t->fib_entry_index, FIB_SOURCE_RR); + } + else + { + t->fib_entry_index = + fib_table_entry_special_add (t->rx_fib_index, + &first_hop_pfx, + FIB_SOURCE_RR, + FIB_ENTRY_FLAG_NONE, ADJ_INDEX_INVALID); + t->sibling_index = + fib_entry_child_add (t->fib_entry_index, + sr_fib_node_type, t - sm->tunnels); + } return 0; } @@ -885,6 +904,18 @@ find_or_add_shared_secret (ip6_sr_main_t * sm, u8 * secret, u32 * indexp) return (key); } +/** + * @brief Stack a tunnel on the forwarding chain of the first-hop + */ +static void +sr_tunnel_stack (ip6_sr_tunnel_t * st) +{ + dpo_stack (sr_dpo_type, + DPO_PROTO_IP6, + &st->first_hop_dpo, + fib_entry_contribute_ip_forwarding (st->fib_entry_index)); +} + /** * @brief Add or Delete a Segment Routing tunnel. * @@ -909,7 +940,6 @@ ip6_sr_add_del_tunnel (ip6_sr_add_del_tunnel_args_t * a) u8 hmac_key_index = 0; ip6_sr_policy_t *pt; int i; - dpo_id_t dpo = DPO_INVALID; /* Make sure that the rx FIB exists */ p = hash_get (im->fib_index_by_table_id, a->rx_table_id); @@ -981,8 +1011,8 @@ ip6_sr_add_del_tunnel (ip6_sr_add_del_tunnel_args_t * a) /* Delete existing tunnel */ t = pool_elt_at_index (sm->tunnels, p[0]); - ip6_delete_route_no_next_hop (&t->key.dst, t->dst_mask_width, - a->rx_table_id); + ip6_routes_add_del (t, 1); + vec_free (t->rewrite); /* Remove tunnel from any policy if associated */ if (t->policy_index != ~0) @@ -1014,6 +1044,7 @@ ip6_sr_add_del_tunnel (ip6_sr_add_del_tunnel_args_t * a) hash_unset_mem (sm->tunnel_index_by_name, t->name); vec_free (t->name); } + dpo_reset (&t->first_hop_dpo); pool_put (sm->tunnels, t); hp = hash_get_pair (sm->tunnel_index_by_key, &key); key_copy = (void *) (hp->key); @@ -1026,6 +1057,7 @@ ip6_sr_add_del_tunnel (ip6_sr_add_del_tunnel_args_t * a) pool_get (sm->tunnels, t); memset (t, 0, sizeof (*t)); t->policy_index = ~0; + fib_node_init (&t->node, sr_fib_node_type); clib_memcpy (&t->key, &key, sizeof (t->key)); t->dst_mask_width = a->dst_mask_width; @@ -1124,20 +1156,13 @@ ip6_sr_add_del_tunnel (ip6_sr_add_del_tunnel_args_t * a) * We don't handle ugly RFC-related cases yet, but I'm sure PL will complain * at some point... */ - dpo_set (&dpo, sr_dpo_type, DPO_PROTO_IP6, t - sm->tunnels); - fib_prefix_t pfx = { - .fp_proto = FIB_PROTOCOL_IP6, - .fp_len = a->dst_mask_width, - .fp_addr = { - .ip6 = *a->dst_address, - } - }; - fib_table_entry_special_dpo_add (rx_fib_index, - &pfx, - FIB_SOURCE_SR, - FIB_ENTRY_FLAG_EXCLUSIVE, &dpo); - dpo_reset (&dpo); + /* + * Add the routes for the tunnel destination and first-hop, then stack + * the tunnel on the appropriate forwarding DPOs. + */ + ip6_routes_add_del (t, 0); + sr_tunnel_stack (t); if (a->policy_name) { @@ -1197,7 +1222,7 @@ format_sr_dpo (u8 * s, va_list * args) return (format (s, "SR: tunnel:[%d]", index)); } -const static dpo_vft_t sr_vft = { +const static dpo_vft_t sr_dpo_vft = { .dv_lock = sr_dpo_lock, .dv_unlock = sr_dpo_unlock, .dv_format = format_sr_dpo, @@ -1212,6 +1237,65 @@ const static char *const *const sr_nodes[DPO_PROTO_NUM] = { [DPO_PROTO_IP6] = sr_ip6_nodes, }; +static ip6_sr_tunnel_t * +sr_tunnel_from_fib_node (fib_node_t * node) +{ +#if (CLIB_DEBUG > 0) + ASSERT (sr_fib_node_type == node->fn_type); +#endif + return ((ip6_sr_tunnel_t *) (((char *) node) - + STRUCT_OFFSET_OF (ip6_sr_tunnel_t, node))); +} + +/** + * Function definition to backwalk a FIB node + */ +static fib_node_back_walk_rc_t +sr_tunnel_back_walk (fib_node_t * node, fib_node_back_walk_ctx_t * ctx) +{ + sr_tunnel_stack (sr_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 * +sr_tunnel_fib_node_get (fib_node_index_t index) +{ + ip6_sr_tunnel_t *st; + ip6_sr_main_t *sm; + + sm = &sr_main; + st = pool_elt_at_index (sm->tunnels, index); + + return (&st->node); +} + +/** + * Function definition to inform the FIB node that its last lock has gone. + */ +static void +sr_tunnel_last_lock_gone (fib_node_t * node) +{ + /* + * The SR 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 SR tunnels + * for participation in the FIB object graph. + */ +const static fib_node_vft_t sr_fib_vft = { + .fnv_get = sr_tunnel_fib_node_get, + .fnv_last_lock = sr_tunnel_last_lock_gone, + .fnv_back_walk = sr_tunnel_back_walk, +}; + /** * @brief CLI parser for Add or Delete a Segment Routing tunnel. * @@ -1764,6 +1848,8 @@ ip6_sr_add_del_multicastmap (ip6_sr_add_del_multicastmap_args_t * a) ip6_sr_tunnel_t *t; ip6_sr_main_t *sm = &sr_main; ip6_sr_policy_t *pt; + index_t rep; + u32 ii; if (a->is_del) { @@ -1803,23 +1889,49 @@ ip6_sr_add_del_multicastmap (ip6_sr_add_del_multicastmap_args_t * a) * We don't handle ugly RFC-related cases yet, but I'm sure PL will complain * at some point... */ - dpo_id_t dpo = DPO_INVALID; - - dpo_set (&dpo, sr_dpo_type, DPO_PROTO_IP6, t - sm->tunnels); - /* Construct a FIB entry for multicast using the rx/tx fib from the first tunnel */ - fib_prefix_t pfx = { + /* + * Construct an mFIB entry for the multicast address, + * using the rx/tx fib from the first tunnel. + * There is no RPF information for this address (I need to discuss this with + * Pablo), so for now accept from anywhere... + */ + /* *INDENT-OFF* */ + mfib_prefix_t pfx = { .fp_proto = FIB_PROTOCOL_IP6, .fp_len = 128, - .fp_addr = { - .ip6 = *a->multicast_address, - } + .fp_grp_addr = { + .ip6 = *a->multicast_address, + } }; - fib_table_entry_special_dpo_add (t->rx_fib_index, - &pfx, - FIB_SOURCE_SR, - FIB_ENTRY_FLAG_EXCLUSIVE, &dpo); - dpo_reset (&dpo); + /* *INDENT-ON* */ + + if (a->is_del) + mfib_table_entry_delete (t->rx_fib_index, &pfx, MFIB_SOURCE_SRv6); + else + { + /* + * Construct a replicate DPO that will replicate received packets over + * each tunnel in the policy + */ + dpo_id_t dpo = DPO_INVALID; + + rep = replicate_create (vec_len (pt->tunnel_indices), DPO_PROTO_IP6); + + vec_foreach_index (ii, pt->tunnel_indices) + { + dpo_set (&dpo, sr_dpo_type, DPO_PROTO_IP6, pt->tunnel_indices[ii]); + + replicate_set_bucket (rep, ii, &dpo); + } + + mfib_table_entry_special_add (t->rx_fib_index, + &pfx, + MFIB_SOURCE_SRv6, + MFIB_ENTRY_FLAG_ACCEPT_ALL_ITF, rep); + + dpo_reset (&dpo); + } u8 *mcast_copy = 0; mcast_copy = vec_new (ip6_address_t, 1); @@ -1829,13 +1941,12 @@ ip6_sr_add_del_multicastmap (ip6_sr_add_del_multicastmap_args_t * a) { hash_unset_mem (sm->policy_index_by_multicast_address, mcast_copy); vec_free (mcast_copy); - return 0; } - /* else */ - - hash_set_mem (sm->policy_index_by_multicast_address, mcast_copy, - pt - sm->policies); - + else + { + hash_set_mem (sm->policy_index_by_multicast_address, mcast_copy, + pt - sm->policies); + } return 0; } @@ -1888,12 +1999,7 @@ sr_add_del_multicast_map_command_fn (vlib_main_t * vm, a->multicast_address = &multicast_address; a->policy_name = policy_name; -#if DPDK > 0 /*Cannot call replicate or configure multicast map yet without DPDK */ rv = ip6_sr_add_del_multicastmap (a); -#else - return clib_error_return (0, - "cannot use multicast replicate spray case without DPDK installed"); -#endif /* DPDK */ switch (rv) { @@ -2295,12 +2401,6 @@ sr_init (vlib_main_t * vm) ip6_rewrite_node = vlib_get_node_by_name (vm, (u8 *) "ip6-rewrite"); ASSERT (ip6_rewrite_node); -#if DPDK > 0 /* Cannot run replicate without DPDK */ - /* Add a disposition to sr_replicate for the sr multicast replicate node */ - sm->ip6_lookup_sr_replicate_index = - vlib_node_add_next (vm, ip6_lookup_node->index, sr_replicate_node.index); -#endif /* DPDK */ - /* Add a disposition to ip6_rewrite for the sr dst address hack node */ sm->ip6_rewrite_sr_next_index = vlib_node_add_next (vm, ip6_rewrite_node->index, @@ -2311,7 +2411,8 @@ sr_init (vlib_main_t * vm) sm->md = (void *) EVP_get_digestbyname ("sha1"); sm->hmac_ctx = clib_mem_alloc (sizeof (HMAC_CTX)); - sr_dpo_type = dpo_register_new_type (&sr_vft, sr_nodes); + sr_dpo_type = dpo_register_new_type (&sr_dpo_vft, sr_nodes); + sr_fib_node_type = fib_node_register_new_type (&sr_fib_vft); return error; } @@ -3087,7 +3188,7 @@ set_ip6_sr_rewrite_fn (vlib_main_t * vm, hi = vnet_get_sup_hw_interface (vnm, sw_if_index); adj->rewrite_header.node_index = sr_fix_dst_addr_node.index; - /* $$$$$ hack... steal the mcast group index */ + /* $$$$$ hack... steal the interface address index */ adj->if_address_index = vlib_node_add_next (vm, sr_fix_dst_addr_node.index, hi->output_node_index); diff --git a/src/vnet/sr/sr.h b/src/vnet/sr/sr.h index 610b36996f3..3c50b7358a1 100644 --- a/src/vnet/sr/sr.h +++ b/src/vnet/sr/sr.h @@ -15,8 +15,6 @@ /** * @file * @brief Segment Routing header - * - * @note sr_replicate only works using DPDK today */ #ifndef included_vnet_sr_h #define included_vnet_sr_h @@ -71,6 +69,27 @@ typedef struct /** Indicates that this tunnel is part of a policy comprising of multiple tunnels. If == ~0 tunnel is not part of a policy */ u32 policy_index; + + /** + * The FIB node graph linkage + */ + fib_node_t node; + + /** + * The FIB entry index for the first hop. We track this so we + * don't need an extra lookup for it in the data plane + */ + fib_node_index_t fib_entry_index; + + /** + * This tunnel's sibling index in the children of the FIB entry + */ + u32 sibling_index; + + /** + * The DPO contributed by the first-hop FIB entry. + */ + dpo_id_t first_hop_dpo; } ip6_sr_tunnel_t; /** @@ -205,9 +224,6 @@ typedef struct /** ip6-rewrite next index for reinstalling the original dst address */ u32 ip6_rewrite_sr_next_index; - /** ip6-replicate next index for multicast tunnel */ - u32 ip6_lookup_sr_replicate_index; - /** application API callback */ void *sr_local_cb; @@ -238,10 +254,6 @@ format_function_t format_ip6_sr_header_with_length; vlib_node_registration_t ip6_sr_input_node; -#if DPDK > 0 -extern vlib_node_registration_t sr_replicate_node; -#endif /* DPDK */ - int ip6_sr_add_del_tunnel (ip6_sr_add_del_tunnel_args_t * a); int ip6_sr_add_del_policy (ip6_sr_add_del_policy_args_t * a); int ip6_sr_add_del_multicastmap (ip6_sr_add_del_multicastmap_args_t * a); diff --git a/src/vnet/sr/sr_api.c b/src/vnet/sr/sr_api.c index 6c6eb9b6449..bab0fc84aba 100644 --- a/src/vnet/sr/sr_api.c +++ b/src/vnet/sr/sr_api.c @@ -190,12 +190,7 @@ static void vl_api_sr_multicast_map_add_del_t_handler goto out; } -#if DPDK > 0 /* Cannot call replicate without DPDK */ rv = ip6_sr_add_del_multicastmap (a); -#else - clib_warning ("multicast replication without DPDK not implemented"); - rv = VNET_API_ERROR_UNIMPLEMENTED; -#endif /* DPDK */ out: diff --git a/src/vnet/sr/sr_replicate.c b/src/vnet/sr/sr_replicate.c deleted file mode 100644 index fa5a68c31c2..00000000000 --- a/src/vnet/sr/sr_replicate.c +++ /dev/null @@ -1,491 +0,0 @@ -/* - * sr_replicate.c: ipv6 segment routing replicator for multicast - * - * Copyright (c) 2016 Cisco and/or its affiliates. - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at: - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ -/** - * @file - * @brief Functions for replicating packets across SR tunnels. - * - * Leverages rte_pktmbuf_clone() so there is no memcpy for - * invariant parts of the packet. - * - * @note Currently requires DPDK -*/ - -#if DPDK > 0 /* Cannot run replicate without DPDK */ -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -/** - * @brief sr_replicate state. - * -*/ -typedef struct -{ - /* convenience */ - vlib_main_t *vlib_main; - vnet_main_t *vnet_main; -} sr_replicate_main_t; - -sr_replicate_main_t sr_replicate_main; - -/** - * @brief Information to display in packet trace. - * -*/ -typedef struct -{ - ip6_address_t src, dst; - u16 length; - u32 next_index; - u32 tunnel_index; - u8 sr[256]; -} sr_replicate_trace_t; - -/** - * @brief packet trace format function. - * - * @param *s u8 used for string output - * @param *args va_list structured input to va_arg to output @ref sr_replicate_trace_t - * @return *s u8 - formatted trace output -*/ -static u8 * -format_sr_replicate_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 *); - sr_replicate_trace_t *t = va_arg (*args, sr_replicate_trace_t *); - ip6_sr_main_t *sm = &sr_main; - ip6_sr_tunnel_t *tun = pool_elt_at_index (sm->tunnels, t->tunnel_index); - ip6_fib_t *rx_fib, *tx_fib; - - rx_fib = ip6_fib_get (tun->rx_fib_index); - tx_fib = ip6_fib_get (tun->tx_fib_index); - - s = format - (s, "SR-REPLICATE: next %s ip6 src %U dst %U len %u\n" - " rx-fib-id %d tx-fib-id %d\n%U", - "ip6-lookup", - format_ip6_address, &t->src, - format_ip6_address, &t->dst, t->length, - rx_fib->table_id, tx_fib->table_id, - format_ip6_sr_header, t->sr, 0 /* print_hmac */ ); - return s; - -} - -#define foreach_sr_replicate_error \ -_(REPLICATED, "sr packets replicated") \ -_(NO_BUFFERS, "error allocating buffers for replicas") \ -_(NO_REPLICAS, "no replicas were needed") \ -_(NO_BUFFER_DROPS, "sr no buffer drops") - -/** - * @brief Struct for SR replicate errors - */ -typedef enum -{ -#define _(sym,str) SR_REPLICATE_ERROR_##sym, - foreach_sr_replicate_error -#undef _ - SR_REPLICATE_N_ERROR, -} sr_replicate_error_t; - -/** - * @brief Error strings for SR replicate - */ -static char *sr_replicate_error_strings[] = { -#define _(sym,string) string, - foreach_sr_replicate_error -#undef _ -}; - -/** - * @brief Defines next-nodes for packet processing. - * -*/ -typedef enum -{ - SR_REPLICATE_NEXT_IP6_LOOKUP, - SR_REPLICATE_N_NEXT, -} sr_replicate_next_t; - -/** - * @brief Single loop packet replicator. - * - * @node sr-replicate - * @param vm vlib_main_t - * @return frame->n_vectors uword -*/ -static uword -sr_replicate_node_fn (vlib_main_t * vm, - vlib_node_runtime_t * node, vlib_frame_t * frame) -{ - dpdk_main_t *dm = &dpdk_main; - u32 n_left_from, *from, *to_next; - sr_replicate_next_t next_index; - int pkts_replicated = 0; - ip6_sr_main_t *sm = &sr_main; - int no_buffer_drops = 0; - vlib_buffer_free_list_t *fl; - unsigned socket_id = rte_socket_id (); - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - fl = vlib_buffer_get_free_list (vm, VLIB_BUFFER_DEFAULT_FREE_LIST_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, hdr_bi0; - vlib_buffer_t *b0, *orig_b0; - struct rte_mbuf *orig_mb0 = 0, *hdr_mb0 = 0, *clone0 = 0; - struct rte_mbuf **hdr_vec = 0, **rte_mbuf_vec = 0; - ip6_sr_policy_t *pol0 = 0; - ip6_sr_tunnel_t *t0 = 0; - ip6_sr_header_t *hdr_sr0 = 0; - ip6_header_t *ip0 = 0, *hdr_ip0 = 0; - int num_replicas = 0; - int i; - u32 len_bytes = sizeof (ip6_header_t); - u8 next_hdr, ip_next_hdr = IPPROTO_IPV6_ROUTE; - - bi0 = from[0]; - - b0 = vlib_get_buffer (vm, bi0); - orig_b0 = b0; - - pol0 = pool_elt_at_index (sm->policies, - vnet_buffer (b0)->ip.save_protocol); - - ip0 = vlib_buffer_get_current (b0); - /* Skip forward to the punch-in point */ - vlib_buffer_advance (b0, sizeof (*ip0)); - next_hdr = ip0->protocol; - - /* HBH must immediately follow ipv6 header */ - if (PREDICT_FALSE - (ip0->protocol == IP_PROTOCOL_IP6_HOP_BY_HOP_OPTIONS)) - { - ip6_hop_by_hop_ext_t *ext_hdr = - (ip6_hop_by_hop_ext_t *) ip6_next_header (ip0); - u32 ext_hdr_len = 0; - ext_hdr_len = ip6_ext_header_len ((ip6_ext_header_t *) ext_hdr); - len_bytes += ext_hdr_len; - next_hdr = ext_hdr->next_hdr; - ext_hdr->next_hdr = IPPROTO_IPV6_ROUTE; - ip_next_hdr = IP_PROTOCOL_IP6_HOP_BY_HOP_OPTIONS; - /* Skip forward to the punch-in point */ - vlib_buffer_advance (b0, ext_hdr_len); - - } - - orig_mb0 = rte_mbuf_from_vlib_buffer (b0); - - i16 delta0 = vlib_buffer_length_in_chain (vm, orig_b0) - - (i16) orig_mb0->pkt_len; - - u16 new_data_len0 = (u16) ((i16) orig_mb0->data_len + delta0); - u16 new_pkt_len0 = (u16) ((i16) orig_mb0->pkt_len + delta0); - - orig_mb0->data_len = new_data_len0; - orig_mb0->pkt_len = new_pkt_len0; - orig_mb0->data_off += (u16) (b0->current_data); - - /* - Before entering loop determine if we can allocate: - - all the new HEADER RTE_MBUFs and assign them to a vector - - all the clones - - if successful, then iterate over vectors of resources - - */ - num_replicas = vec_len (pol0->tunnel_indices); - - if (PREDICT_FALSE (num_replicas == 0)) - { - b0->error = node->errors[SR_REPLICATE_ERROR_NO_REPLICAS]; - goto do_trace0; - } - - vec_reset_length (hdr_vec); - vec_reset_length (rte_mbuf_vec); - - for (i = 0; i < num_replicas; i++) - { - uint8_t nb_seg; - struct rte_mbuf *clone0i; - vlib_buffer_t *clone0_c, *clone_b0; - - t0 = vec_elt_at_index (sm->tunnels, pol0->tunnel_indices[i]); - hdr_mb0 = rte_pktmbuf_alloc (dm->pktmbuf_pools[socket_id]); - - if (i < (num_replicas - 1)) - { - /* Not the last tunnel to process */ - clone0 = rte_pktmbuf_clone - (orig_mb0, dm->pktmbuf_pools[socket_id]); - if (clone0 == 0) - goto clone_fail; - nb_seg = 0; - clone0i = clone0; - clone0_c = NULL; - while ((clone0->nb_segs >= 1) && (nb_seg < clone0->nb_segs)) - { - - clone_b0 = vlib_buffer_from_rte_mbuf (clone0i); - vlib_buffer_init_for_free_list (clone_b0, fl); - - ASSERT ((clone_b0->flags & VLIB_BUFFER_NEXT_PRESENT) == - 0); - ASSERT (clone_b0->current_data == 0); - - clone_b0->current_data = - (clone0i->buf_addr + clone0i->data_off) - - (void *) clone_b0->data; - - clone_b0->current_length = clone0i->data_len; - if (PREDICT_FALSE (clone0_c != NULL)) - { - clone0_c->flags |= VLIB_BUFFER_NEXT_PRESENT; - clone0_c->next_buffer = - vlib_get_buffer_index (vm, clone_b0); - } - clone0_c = clone_b0; - clone0i = clone0i->next; - nb_seg++; - } - } - else - /* First tunnel to process, use original MB */ - clone0 = orig_mb0; - - - if (PREDICT_FALSE (!clone0 || !hdr_mb0)) - { - clone_fail: - b0->error = node->errors[SR_REPLICATE_ERROR_NO_BUFFERS]; - - vec_foreach_index (i, rte_mbuf_vec) - { - rte_pktmbuf_free (rte_mbuf_vec[i]); - } - vec_free (rte_mbuf_vec); - - vec_foreach_index (i, hdr_vec) - { - rte_pktmbuf_free (hdr_vec[i]); - } - vec_free (hdr_vec); - - goto do_trace0; - } - - vec_add1 (hdr_vec, hdr_mb0); - vec_add1 (rte_mbuf_vec, clone0); - - } - - for (i = 0; i < num_replicas; i++) - { - vlib_buffer_t *hdr_b0; - u16 new_l0 = 0; - - t0 = vec_elt_at_index (sm->tunnels, pol0->tunnel_indices[i]); - /* Our replicas */ - hdr_mb0 = hdr_vec[i]; - clone0 = rte_mbuf_vec[i]; - - hdr_mb0->data_len = len_bytes + vec_len (t0->rewrite); - hdr_mb0->pkt_len = hdr_mb0->data_len + - vlib_buffer_length_in_chain (vm, orig_b0); - - hdr_b0 = vlib_buffer_from_rte_mbuf (hdr_mb0); - - vlib_buffer_init_for_free_list (hdr_b0, fl); - - memcpy (hdr_b0->data, ip0, len_bytes); - memcpy (hdr_b0->data + len_bytes, t0->rewrite, - vec_len (t0->rewrite)); - - hdr_b0->current_data = 0; - hdr_b0->current_length = len_bytes + vec_len (t0->rewrite); - hdr_b0->flags = orig_b0->flags | VLIB_BUFFER_NEXT_PRESENT; - hdr_b0->trace_index = orig_b0->trace_index; - vnet_buffer (hdr_b0)->l2_classify.opaque_index = 0; - - hdr_b0->total_length_not_including_first_buffer = - hdr_mb0->pkt_len - hdr_b0->current_length; - vnet_buffer (hdr_b0)->sw_if_index[VLIB_TX] = t0->tx_fib_index; - - hdr_ip0 = (ip6_header_t *) hdr_b0->data; - new_l0 = clib_net_to_host_u16 (ip0->payload_length) + - vec_len (t0->rewrite); - hdr_ip0->payload_length = clib_host_to_net_u16 (new_l0); - hdr_sr0 = (ip6_sr_header_t *) ((u8 *) hdr_ip0 + len_bytes); - /* $$$ tune */ - clib_memcpy (hdr_sr0, t0->rewrite, vec_len (t0->rewrite)); - hdr_sr0->protocol = next_hdr; - hdr_ip0->protocol = ip_next_hdr; - - /* Copy dst address into the DA slot in the segment list */ - clib_memcpy (hdr_sr0->segments, ip0->dst_address.as_u64, - sizeof (ip6_address_t)); - - /* Rewrite the ip6 dst address */ - hdr_ip0->dst_address.as_u64[0] = t0->first_hop.as_u64[0]; - hdr_ip0->dst_address.as_u64[1] = t0->first_hop.as_u64[1]; - - sr_fix_hmac (sm, hdr_ip0, hdr_sr0); - - /* prepend new header to invariant piece */ - hdr_mb0->next = clone0; - hdr_b0->next_buffer = - vlib_get_buffer_index (vm, - vlib_buffer_from_rte_mbuf (clone0)); - - /* update header's fields */ - hdr_mb0->pkt_len = - (uint16_t) (hdr_mb0->data_len + clone0->pkt_len); - hdr_mb0->nb_segs = (uint8_t) (clone0->nb_segs + 1); - - /* copy metadata from source packet */ - hdr_mb0->port = clone0->port; - hdr_mb0->vlan_tci = clone0->vlan_tci; - hdr_mb0->vlan_tci_outer = clone0->vlan_tci_outer; - hdr_mb0->tx_offload = clone0->tx_offload; - hdr_mb0->hash = clone0->hash; - - hdr_mb0->ol_flags = clone0->ol_flags & ~(IND_ATTACHED_MBUF); - - __rte_mbuf_sanity_check (hdr_mb0, 1); - - hdr_bi0 = vlib_get_buffer_index (vm, hdr_b0); - - to_next[0] = hdr_bi0; - to_next += 1; - n_left_to_next -= 1; - - if (n_left_to_next == 0) - { - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - vlib_get_next_frame (vm, node, next_index, - to_next, n_left_to_next); - - } - pkts_replicated++; - } - - from += 1; - n_left_from -= 1; - - do_trace0: - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) - { - sr_replicate_trace_t *tr = vlib_add_trace (vm, node, - b0, sizeof (*tr)); - tr->tunnel_index = t0 - sm->tunnels; - tr->length = 0; - if (hdr_ip0) - { - memcpy (tr->src.as_u8, hdr_ip0->src_address.as_u8, - sizeof (tr->src.as_u8)); - memcpy (tr->dst.as_u8, hdr_ip0->dst_address.as_u8, - sizeof (tr->dst.as_u8)); - if (hdr_ip0->payload_length) - tr->length = clib_net_to_host_u16 - (hdr_ip0->payload_length); - } - tr->next_index = next_index; - if (hdr_sr0) - memcpy (tr->sr, hdr_sr0, sizeof (tr->sr)); - } - - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - vlib_node_increment_counter (vm, sr_replicate_node.index, - SR_REPLICATE_ERROR_REPLICATED, - pkts_replicated); - - vlib_node_increment_counter (vm, sr_replicate_node.index, - SR_REPLICATE_ERROR_NO_BUFFER_DROPS, - no_buffer_drops); - - return frame->n_vectors; -} - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (sr_replicate_node) = { - .function = sr_replicate_node_fn, - .name = "sr-replicate", - .vector_size = sizeof (u32), - .format_trace = format_sr_replicate_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - - .n_errors = ARRAY_LEN(sr_replicate_error_strings), - .error_strings = sr_replicate_error_strings, - - .n_next_nodes = SR_REPLICATE_N_NEXT, - - .next_nodes = { - [SR_REPLICATE_NEXT_IP6_LOOKUP] = "ip6-lookup", - }, -}; - -VLIB_NODE_FUNCTION_MULTIARCH (sr_replicate_node, sr_replicate_node_fn) -/* *INDENT-ON* */ - -clib_error_t * -sr_replicate_init (vlib_main_t * vm) -{ - sr_replicate_main_t *msm = &sr_replicate_main; - - msm->vlib_main = vm; - msm->vnet_main = vnet_get_main (); - - return 0; -} - -VLIB_INIT_FUNCTION (sr_replicate_init); - -#endif /* DPDK */ - -/* - * fd.io coding-style-patch-verification: ON - * - * Local Variables: - * eval: (c-set-style "gnu") - * End: - */ -- cgit 1.2.3-korg