/* * Copyright (c) 2016 Cisco and/or its affiliates. * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at: * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ #include #include #include #include //GC runs at most once every so many seconds #define LB_GARBAGE_RUN 60 //After so many seconds. It is assumed that inter-core race condition will not occur. #define LB_CONCURRENCY_TIMEOUT 10 lb_main_t lb_main; #define lb_get_writer_lock() do {} while(__sync_lock_test_and_set (lb_main.writer_lock, 1)) #define lb_put_writer_lock() lb_main.writer_lock[0] = 0 static void lb_as_stack (lb_as_t *as); const static char * const lb_dpo_gre4_ip4[] = { "lb4-gre4" , NULL }; const static char * const lb_dpo_gre4_ip6[] = { "lb6-gre4" , NULL }; const static char* const * const lb_dpo_gre4_nodes[DPO_PROTO_NUM] = { [DPO_PROTO_IP4] = lb_dpo_gre4_ip4, [DPO_PROTO_IP6] = lb_dpo_gre4_ip6, }; const static char * const lb_dpo_gre6_ip4[] = { "lb4-gre6" , NULL }; const static char * const lb_dpo_gre6_ip6[] = { "lb6-gre6" , NULL }; const static char* const * const lb_dpo_gre6_nodes[DPO_PROTO_NUM] = { [DPO_PROTO_IP4] = lb_dpo_gre6_ip4, [DPO_PROTO_IP6] = lb_dpo_gre6_ip6, }; u32 lb_hash_time_now(vlib_main_t * vm) { return (u32) (vlib_time_now(vm) + 10000); } u8 *format_lb_main (u8 * s, va_list * args) { vlib_thread_main_t *tm = vlib_get_thread_main(); lb_main_t *lbm = &lb_main; s = format(s, "lb_main"); s = format(s, " ip4-src-address: %U \n", format_ip4_address, &lbm->ip4_src_address); s = format(s, " ip6-src-address: %U \n", format_ip6_address, &lbm->ip6_src_address); s = format(s, " #vips: %u\n", pool_elts(lbm->vips)); s = format(s, " #ass: %u\n", pool_elts(lbm->ass) - 1); u32 thread_index; for(thread_index = 0; thread_index < tm->n_vlib_mains; thread_index++ ) { lb_hash_t *h = lbm->per_cpu[thread_index].sticky_ht; if (h) { s = format(s, "core %d\n", thread_index); s = format(s, " timeout: %ds\n", h->timeout); s = format(s, " usage: %d / %d\n", lb_hash_elts(h, lb_hash_time_now(vlib_get_main())), lb_hash_size(h)); } } return s; } static char *lb_vip_type_strings[] = { [LB_VIP_TYPE_IP6_GRE6] = "ip6-gre6", [LB_VIP_TYPE_IP6_GRE4] = "ip6-gre4", [LB_VIP_TYPE_IP4_GRE6] = "ip4-gre6", [LB_VIP_TYPE_IP4_GRE4] = "ip4-gre4", }; u8 *format_lb_vip_type (u8 * s, va_list * args) { lb_vip_type_t vipt = va_arg (*args, lb_vip_type_t); u32 i; for (i=0; itype, format_ip46_prefix, &vip->prefix, vip->plen, IP46_TYPE_ANY, vip->new_flow_table_mask + 1, pool_elts(vip->as_indexes), (vip->flags & LB_VIP_FLAGS_USED)?"":" removed"); } u8 *format_lb_as (u8 * s, va_list * args) { lb_as_t *as = va_arg (*args, lb_as_t *); return format(s, "%U %s", format_ip46_address, &as->address, IP46_TYPE_ANY, (as->flags & LB_AS_FLAGS_USED)?"used":"removed"); } u8 *format_lb_vip_detailed (u8 * s, va_list * args) { lb_main_t *lbm = &lb_main; lb_vip_t *vip = va_arg (*args, lb_vip_t *); u32 indent = format_get_indent (s); s = format(s, "%U %U [%lu] %U%s\n" "%U new_size:%u\n", format_white_space, indent, format_lb_vip_type, vip->type, vip - lbm->vips, format_ip46_prefix, &vip->prefix, (u32) vip->plen, IP46_TYPE_ANY, (vip->flags & LB_VIP_FLAGS_USED)?"":" removed", format_white_space, indent, vip->new_flow_table_mask + 1); //Print counters s = format(s, "%U counters:\n", format_white_space, indent); u32 i; for (i=0; ivip_counters[i].name, vlib_get_simple_counter(&lbm->vip_counters[i], vip - lbm->vips)); s = format(s, "%U #as:%u\n", format_white_space, indent, 
---
name: Link Layer Discovery Protocol
maintainer: Klement Sekera <ksekera@cisco.com>
features:
  - link layer discovery protocol implementation
description: "Link Layer Discovery Protocol (LLDP) implementation"
state: production
properties: [API, CLI, STATS, MULTITHREAD]
_t *as; u32 *asi; pool_foreach(asi, vip->as_indexes, { as = &lbm->ass[*asi]; if (as->vip_index == (vip - lbm->vips) && as->address.as_u64[0] == address->as_u64[0] && as->address.as_u64[1] == address->as_u64[1]) { *as_index = as - lbm->ass; return 0; } }); return -1; } int lb_vip_add_ass(u32 vip_index, ip46_address_t *addresses, u32 n) { lb_main_t *lbm = &lb_main; lb_get_writer_lock(); lb_vip_t *vip; if (!(vip = lb_vip_get_by_index(vip_index))) { lb_put_writer_lock(); return VNET_API_ERROR_NO_SUCH_ENTRY; } ip46_type_t type = lb_vip_is_gre4(vip)?IP46_TYPE_IP4:IP46_TYPE_IP6; u32 *to_be_added = 0; u32 *to_be_updated = 0; u32 i; u32 *ip; //Sanity check while (n--) { if (!lb_as_find_index_vip(vip, &addresses[n], &i)) { if (lbm->ass[i].flags & LB_AS_FLAGS_USED) { vec_free(to_be_added); vec_free(to_be_updated); lb_put_writer_lock(); return VNET_API_ERROR_VALUE_EXIST; } vec_add1(to_be_updated, i); goto next; } if (ip46_address_type(&addresses[n]) != type) { vec_free(to_be_added); vec_free(to_be_updated); lb_put_writer_lock(); return VNET_API_ERROR_INVALID_ADDRESS_FAMILY; } if (n) { u32 n2 = n; while(n2--) //Check for duplicates if (addresses[n2].as_u64[0] == addresses[n].as_u64[0] && addresses[n2].as_u64[1] == addresses[n].as_u64[1]) goto next; } vec_add1(to_be_added, n); next: continue; } //Update reused ASs vec_foreach(ip, to_be_updated) { lbm->ass[*ip].flags = LB_AS_FLAGS_USED; } vec_free(to_be_updated); //Create those who have to be created vec_foreach(ip, to_be_added) { lb_as_t *as; u32 *as_index; pool_get(lbm->ass, as); as->address = addresses[*ip]; as->flags = LB_AS_FLAGS_USED; as->vip_index = vip_index; pool_get(vip->as_indexes, as_index); *as_index = as - lbm->ass; /* * become a child of the FIB entry * so we are informed when its forwarding changes */ fib_prefix_t nh = {}; if (lb_vip_is_gre4(vip)) { nh.fp_addr.ip4 = as->address.ip4; nh.fp_len = 32; nh.fp_proto = FIB_PROTOCOL_IP4; } else { nh.fp_addr.ip6 = as->address.ip6; nh.fp_len = 128; nh.fp_proto = FIB_PROTOCOL_IP6; } as->next_hop_fib_entry_index = fib_table_entry_special_add(0, &nh, FIB_SOURCE_RR, FIB_ENTRY_FLAG_NONE); as->next_hop_child_index = fib_entry_child_add(as->next_hop_fib_entry_index, lbm->fib_node_type, as - lbm->ass); lb_as_stack(as); } vec_free(to_be_added); //Recompute flows lb_vip_update_new_flow_table(vip); //Garbage collection maybe lb_vip_garbage_collection(vip); lb_put_writer_lock(); return 0; } int lb_vip_del_ass_withlock(u32 vip_index, ip46_address_t *addresses, u32 n) { lb_main_t *lbm = &lb_main; u32 now = (u32) vlib_time_now(vlib_get_main()); u32 *ip = 0; lb_vip_t *vip; if (!(vip = lb_vip_get_by_index(vip_index))) { return VNET_API_ERROR_NO_SUCH_ENTRY; } u32 *indexes = NULL; while (n--) { u32 i; if (lb_as_find_index_vip(vip, &addresses[n], &i)) { vec_free(indexes); return VNET_API_ERROR_NO_SUCH_ENTRY; } if (n) { //Check for duplicates u32 n2 = n - 1; while(n2--) { if (addresses[n2].as_u64[0] == addresses[n].as_u64[0] && addresses[n2].as_u64[1] == addresses[n].as_u64[1]) goto next; } } vec_add1(indexes, i); next: continue; } //Garbage collection maybe lb_vip_garbage_collection(vip); if (indexes != NULL) { vec_foreach(ip, indexes) { lbm->ass[*ip].flags &= ~LB_AS_FLAGS_USED; lbm->ass[*ip].last_used = now; } //Recompute flows lb_vip_update_new_flow_table(vip); } vec_free(indexes); return 0; } int lb_vip_del_ass(u32 vip_index, ip46_address_t *addresses, u32 n) { lb_get_writer_lock(); int ret = lb_vip_del_ass_withlock(vip_index, addresses, n); lb_put_writer_lock(); return ret; } /** * Add the VIP adjacency to the ip4 or ip6 fib */ static void lb_vip_add_adjacency(lb_main_t *lbm, lb_vip_t *vip) { dpo_proto_t proto = 0; dpo_id_t dpo = DPO_INVALID; fib_prefix_t pfx = {}; if (lb_vip_is_ip4(vip)) { pfx.fp_addr.ip4 = vip->prefix.ip4; pfx.fp_len = vip->plen - 96; pfx.fp_proto = FIB_PROTOCOL_IP4; proto = DPO_PROTO_IP4; } else { pfx.fp_addr.ip6 = vip->prefix.ip6; pfx.fp_len = vip->plen; pfx.fp_proto = FIB_PROTOCOL_IP6; proto = DPO_PROTO_IP6; } dpo_set(&dpo, lb_vip_is_gre4(vip)?lbm->dpo_gre4_type:lbm->dpo_gre6_type, proto, vip - lbm->vips); fib_table_entry_special_dpo_add(0, &pfx, FIB_SOURCE_PLUGIN_HI, FIB_ENTRY_FLAG_EXCLUSIVE, &dpo); dpo_reset(&dpo); } /** * Deletes the adjacency associated with the VIP */ static void lb_vip_del_adjacency(lb_main_t *lbm, lb_vip_t *vip) { fib_prefix_t pfx = {}; if (lb_vip_is_ip4(vip)) { pfx.fp_addr.ip4 = vip->prefix.ip4; pfx.fp_len = vip->plen - 96; pfx.fp_proto = FIB_PROTOCOL_IP4; } else { pfx.fp_addr.ip6 = vip->prefix.ip6; pfx.fp_len = vip->plen; pfx.fp_proto = FIB_PROTOCOL_IP6; } fib_table_entry_special_remove(0, &pfx, FIB_SOURCE_PLUGIN_HI); } int lb_vip_add(ip46_address_t *prefix, u8 plen, lb_vip_type_t type, u32 new_length, u32 *vip_index) { lb_main_t *lbm = &lb_main; lb_vip_t *vip; lb_get_writer_lock(); ip46_prefix_normalize(prefix, plen); if (!lb_vip_find_index_with_lock(prefix, plen, vip_index)) { lb_put_writer_lock(); return VNET_API_ERROR_VALUE_EXIST; } if (!is_pow2(new_length)) { lb_put_writer_lock(); return VNET_API_ERROR_INVALID_MEMORY_SIZE; } if (ip46_prefix_is_ip4(prefix, plen) && (type != LB_VIP_TYPE_IP4_GRE4) && (type != LB_VIP_TYPE_IP4_GRE6)) return VNET_API_ERROR_INVALID_ADDRESS_FAMILY; //Allocate pool_get(lbm->vips, vip); //Init vip->prefix = *prefix; vip->plen = plen; vip->last_garbage_collection = (u32) vlib_time_now(vlib_get_main()); vip->type = type; vip->flags = LB_VIP_FLAGS_USED; vip->as_indexes = 0; //Validate counters u32 i; for (i = 0; i < LB_N_VIP_COUNTERS; i++) { vlib_validate_simple_counter(&lbm->vip_counters[i], vip - lbm->vips); vlib_zero_simple_counter(&lbm->vip_counters[i], vip - lbm->vips); } //Configure new flow table vip->new_flow_table_mask = new_length - 1; vip->new_flow_table = 0; //Create a new flow hash table full of the default entry lb_vip_update_new_flow_table(vip); //Create adjacency to direct traffic lb_vip_add_adjacency(lbm, vip); //Return result *vip_index = vip - lbm->vips; lb_put_writer_lock(); return 0; } int lb_vip_del(u32 vip_index) { lb_main_t *lbm = &lb_main; lb_vip_t *vip; lb_get_writer_lock(); if (!(vip = lb_vip_get_by_index(vip_index))) { lb_put_writer_lock(); return VNET_API_ERROR_NO_SUCH_ENTRY; } //FIXME: This operation is actually not working //We will need to remove state before performing this. { //Remove all ASs ip46_address_t *ass = 0; lb_as_t *as; u32 *as_index; pool_foreach(as_index, vip->as_indexes, { as = &lbm->ass[*as_index]; vec_add1(ass, as->address); }); if (vec_len(ass)) lb_vip_del_ass_withlock(vip_index, ass, vec_len(ass)); vec_free(ass); } //Delete adjacency lb_vip_del_adjacency(lbm, vip); //Set the VIP as unused vip->flags &= ~LB_VIP_FLAGS_USED; lb_put_writer_lock(); return 0; } /* *INDENT-OFF* */ VLIB_PLUGIN_REGISTER () = { .version = VPP_BUILD_VER, .description = "Load Balancer", }; /* *INDENT-ON* */ u8 *format_lb_dpo (u8 * s, va_list * va) { index_t index = va_arg (*va, index_t); CLIB_UNUSED(u32 indent) = va_arg (*va, u32); lb_main_t *lbm = &lb_main; lb_vip_t *vip = pool_elt_at_index (lbm->vips, index); return format (s, "%U", format_lb_vip, vip); } static void lb_dpo_lock (dpo_id_t *dpo) {} static void lb_dpo_unlock (dpo_id_t *dpo) {} static fib_node_t * lb_fib_node_get_node (fib_node_index_t index) { lb_main_t *lbm = &lb_main; lb_as_t *as = pool_elt_at_index (lbm->ass, index); return (&as->fib_node); } static void lb_fib_node_last_lock_gone (fib_node_t *node) { } static lb_as_t * lb_as_from_fib_node (fib_node_t *node) { return ((lb_as_t*)(((char*)node) - STRUCT_OFFSET_OF(lb_as_t, fib_node))); } static void lb_as_stack (lb_as_t *as) { lb_main_t *lbm = &lb_main; lb_vip_t *vip = &lbm->vips[as->vip_index]; dpo_stack(lb_vip_is_gre4(vip)?lbm->dpo_gre4_type:lbm->dpo_gre6_type, lb_vip_is_ip4(vip)?DPO_PROTO_IP4:DPO_PROTO_IP6, &as->dpo, fib_entry_contribute_ip_forwarding( as->next_hop_fib_entry_index)); } static fib_node_back_walk_rc_t lb_fib_node_back_walk_notify (fib_node_t *node, fib_node_back_walk_ctx_t *ctx) { lb_as_stack(lb_as_from_fib_node(node)); return (FIB_NODE_BACK_WALK_CONTINUE); } clib_error_t * lb_init (vlib_main_t * vm) { vlib_thread_main_t *tm = vlib_get_thread_main (); lb_main_t *lbm = &lb_main; lb_as_t *default_as; fib_node_vft_t lb_fib_node_vft = { .fnv_get = lb_fib_node_get_node, .fnv_last_lock = lb_fib_node_last_lock_gone, .fnv_back_walk = lb_fib_node_back_walk_notify, }; dpo_vft_t lb_vft = { .dv_lock = lb_dpo_lock, .dv_unlock = lb_dpo_unlock, .dv_format = format_lb_dpo, }; lbm->vips = 0; lbm->per_cpu = 0; vec_validate(lbm->per_cpu, tm->n_vlib_mains - 1); lbm->writer_lock = clib_mem_alloc_aligned (CLIB_CACHE_LINE_BYTES, CLIB_CACHE_LINE_BYTES); lbm->writer_lock[0] = 0; lbm->per_cpu_sticky_buckets = LB_DEFAULT_PER_CPU_STICKY_BUCKETS; lbm->flow_timeout = LB_DEFAULT_FLOW_TIMEOUT; lbm->ip4_src_address.as_u32 = 0xffffffff; lbm->ip6_src_address.as_u64[0] = 0xffffffffffffffffL; lbm->ip6_src_address.as_u64[1] = 0xffffffffffffffffL; lbm->dpo_gre4_type = dpo_register_new_type(&lb_vft, lb_dpo_gre4_nodes); lbm->dpo_gre6_type = dpo_register_new_type(&lb_vft, lb_dpo_gre6_nodes); lbm->fib_node_type = fib_node_register_new_type(&lb_fib_node_vft); //Init AS reference counters vlib_refcount_init(&lbm->as_refcount); //Allocate and init default AS. lbm->ass = 0; pool_get(lbm->ass, default_as); default_as->flags = 0; default_as->dpo.dpoi_next_node = LB_NEXT_DROP; default_as->vip_index = ~0; default_as->address.ip6.as_u64[0] = 0xffffffffffffffffL; default_as->address.ip6.as_u64[1] = 0xffffffffffffffffL; #define _(a,b,c) lbm->vip_counters[c].name = b; lb_foreach_vip_counter #undef _ return NULL; } VLIB_INIT_FUNCTION (lb_init);