2 // Copyright (c) 2010-2017 Intel Corporation
4 // Licensed under the Apache License, Version 2.0 (the "License");
5 // you may not use this file except in compliance with the License.
6 // You may obtain a copy of the License at
8 // http://www.apache.org/licenses/LICENSE-2.0
10 // Unless required by applicable law or agreed to in writing, software
11 // distributed under the License is distributed on an "AS IS" BASIS,
12 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 // See the License for the specific language governing permissions and
14 // limitations under the License.
18 #include <rte_cycles.h>
20 #include <rte_version.h>
22 #include <rte_byteorder.h>
25 #include "prox_lua_types.h"
29 #include "handle_routing.h"
33 #include "prox_port_cfg.h"
36 #include "hash_entry_types.h"
41 #include "prox_shared.h"
42 #include "prox_cksum.h"
43 #include "mbuf_utils.h"
46 struct task_base base;
47 uint8_t runtime_flags;
48 struct lcore_cfg *lconf;
49 struct rte_lpm *ipv4_lpm;
50 struct next_hop *next_hops;
52 uint32_t number_free_rules;
55 uint64_t src_mac[PROX_MAX_PORTS];
58 static void routing_update(struct task_base *tbase, void **data, uint16_t n_msgs)
60 struct task_routing *task = (struct task_routing *)tbase;
61 struct route_msg *msg;
63 for (uint16_t i = 0; i < n_msgs; ++i) {
64 msg = (struct route_msg *)data[i];
66 if (task->number_free_rules == 0) {
67 plog_warn("Failed adding route: %u.%u.%u.%u/%u: lpm table full\n",
68 msg->ip_bytes[0], msg->ip_bytes[1], msg->ip_bytes[2],
69 msg->ip_bytes[3], msg->prefix);
71 if (rte_lpm_add(task->ipv4_lpm, rte_bswap32(msg->ip), msg->prefix, msg->nh)) {
72 plog_warn("Failed adding route: %u.%u.%u.%u/%u\n",
73 msg->ip_bytes[0], msg->ip_bytes[1], msg->ip_bytes[2],
74 msg->ip_bytes[3], msg->prefix);
76 task->number_free_rules--;
82 static void init_task_routing(struct task_base *tbase, struct task_args *targ)
84 struct task_routing *task = (struct task_routing *)tbase;
85 const int socket_id = rte_lcore_to_socket_id(targ->lconf->id);
88 task->lconf = targ->lconf;
89 task->qinq_tag = targ->qinq_tag;
90 task->runtime_flags = targ->runtime_flags;
92 PROX_PANIC(!strcmp(targ->route_table, ""), "route table not specified\n");
93 if (targ->flags & TASK_ARG_LOCAL_LPM) {
94 int ret = lua_to_lpm4(prox_lua(), GLOBAL, targ->route_table, socket_id, &lpm);
95 PROX_PANIC(ret, "Failed to load IPv4 LPM:\n%s\n", get_lua_to_errors());
96 prox_sh_add_socket(socket_id, targ->route_table, lpm);
98 task->number_free_rules = lpm->n_free_rules;
101 lpm = prox_sh_find_socket(socket_id, targ->route_table);
103 int ret = lua_to_lpm4(prox_lua(), GLOBAL, targ->route_table, socket_id, &lpm);
104 PROX_PANIC(ret, "Failed to load IPv4 LPM:\n%s\n", get_lua_to_errors());
105 prox_sh_add_socket(socket_id, targ->route_table, lpm);
108 task->ipv4_lpm = lpm->rte_lpm;
109 task->next_hops = lpm->next_hops;
110 task->number_free_rules = lpm->n_free_rules;
112 for (uint32_t i = 0; i < MAX_HOP_INDEX; i++) {
113 int tx_port = task->next_hops[i].mac_port.out_idx;
114 if ((tx_port > targ->nb_txports - 1) && (tx_port > targ->nb_txrings - 1)) {
115 PROX_PANIC(1, "Routing Table contains port %d but only %d tx port/ %d ring:\n", tx_port, targ->nb_txports, targ->nb_txrings);
119 if (targ->nb_txrings) {
120 struct task_args *dtarg;
122 for (uint32_t i = 0; i < targ->nb_txrings; ++i) {
123 ct = targ->core_task_set[0].core_task[i];
124 dtarg = core_targ_get(ct.core, ct.task);
125 dtarg = find_reachable_task_sending_to_port(dtarg);
126 if (task->runtime_flags & TASK_MPLS_TAGGING) {
127 task->src_mac[i] = (0x0000ffffffffffff & ((*(uint64_t*)&prox_port_cfg[dtarg->tx_port_queue[0].port].eth_addr))) | ((uint64_t)ETYPE_MPLSU << (64 - 16));
129 task->src_mac[i] = (0x0000ffffffffffff & ((*(uint64_t*)&prox_port_cfg[dtarg->tx_port_queue[0].port].eth_addr))) | ((uint64_t)ETYPE_IPv4 << (64 - 16));
133 for (uint32_t i = 0; i < targ->nb_txports; ++i) {
134 if (task->runtime_flags & TASK_MPLS_TAGGING) {
135 task->src_mac[i] = (0x0000ffffffffffff & ((*(uint64_t*)&prox_port_cfg[targ->tx_port_queue[i].port].eth_addr))) | ((uint64_t)ETYPE_MPLSU << (64 - 16));
137 task->src_mac[i] = (0x0000ffffffffffff & ((*(uint64_t*)&prox_port_cfg[targ->tx_port_queue[i].port].eth_addr))) | ((uint64_t)ETYPE_IPv4 << (64 - 16));
142 for (uint32_t i = 0; i < 4; ++i) {
143 task->marking[i] = rte_bswap32(targ->marking[i] << 9);
146 struct prox_port_cfg *port = find_reachable_port(targ);
148 task->offload_crc = port->capabilities.tx_offload_cksum;
151 targ->lconf->ctrl_func_m[targ->task] = routing_update;
152 targ->lconf->ctrl_timeout = freq_to_tsc(20);
155 static inline uint8_t handle_routing(struct task_routing *task, struct rte_mbuf *mbuf);
157 static int handle_routing_bulk(struct task_base *tbase, struct rte_mbuf **mbufs, uint16_t n_pkts)
159 struct task_routing *task = (struct task_routing *)tbase;
160 uint8_t out[MAX_PKT_BURST];
163 prefetch_first(mbufs, n_pkts);
165 for (j = 0; j + PREFETCH_OFFSET < n_pkts; ++j) {
166 #ifdef PROX_PREFETCH_OFFSET
167 PREFETCH0(mbufs[j + PREFETCH_OFFSET]);
168 PREFETCH0(rte_pktmbuf_mtod(mbufs[j + PREFETCH_OFFSET - 1], void *));
170 out[j] = handle_routing(task, mbufs[j]);
172 #ifdef PROX_PREFETCH_OFFSET
173 PREFETCH0(rte_pktmbuf_mtod(mbufs[n_pkts - 1], void *));
174 for (; j < n_pkts; ++j) {
175 out[j] = handle_routing(task, mbufs[j]);
179 return task->base.tx_pkt(&task->base, mbufs, n_pkts, out);
182 static void set_l2(struct task_routing *task, struct rte_mbuf *mbuf, uint8_t nh_idx)
184 struct ether_hdr *peth = rte_pktmbuf_mtod(mbuf, struct ether_hdr *);
185 *((uint64_t *)(&peth->d_addr)) = task->next_hops[nh_idx].mac_port_8bytes;
186 *((uint64_t *)(&peth->s_addr)) = task->src_mac[task->next_hops[nh_idx].mac_port.out_idx];
189 static void set_l2_mpls(struct task_routing *task, struct rte_mbuf *mbuf, uint8_t nh_idx, uint16_t l2_len)
191 struct ether_hdr *peth = (struct ether_hdr *)rte_pktmbuf_prepend(mbuf, sizeof(struct mpls_hdr));
192 l2_len += sizeof(struct mpls_hdr);
193 prox_ip_cksum(mbuf, (struct ipv4_hdr *)((uint8_t *)peth + l2_len), l2_len, sizeof(struct ipv4_hdr), task->offload_crc);
195 *((uint64_t *)(&peth->d_addr)) = task->next_hops[nh_idx].mac_port_8bytes;
196 *((uint64_t *)(&peth->s_addr)) = task->src_mac[task->next_hops[nh_idx].mac_port.out_idx];
197 /* MPLSU ether_type written as high word of 64bit src_mac prepared by init_task_routing */
198 struct mpls_hdr *mpls = (struct mpls_hdr *)(peth + 1);
200 if (task->runtime_flags & TASK_MARK) {
201 enum rte_meter_color color = rte_sched_port_pkt_read_color(mbuf);
203 *(uint32_t *)mpls = task->next_hops[nh_idx].mpls | task->marking[color] | 0x00010000; // Set BoS to 1
206 *(uint32_t *)mpls = task->next_hops[nh_idx].mpls | 0x00010000; // Set BoS to 1
210 static uint8_t route_ipv4(struct task_routing *task, uint8_t *beg, uint32_t ip_offset, struct rte_mbuf *mbuf)
212 struct ipv4_hdr *ip = (struct ipv4_hdr*)(beg + ip_offset);
213 struct ether_hdr *peth_out;
217 if (unlikely(ip->version_ihl >> 4 != 4)) {
218 plog_warn("Offset: %d\n", ip_offset);
219 plog_warn("Expected to receive IPv4 packet but IP version was %d\n",
220 ip->version_ihl >> 4);
224 switch(ip->next_proto_id) {
226 struct gre_hdr *pgre = (struct gre_hdr *)(ip + 1);
227 dst_ip = ((struct ipv4_hdr *)(pgre + 1))->dst_addr;
232 dst_ip = ip->dst_addr;
235 /* Routing for other protocols is not implemented */
239 #if RTE_VERSION >= RTE_VERSION_NUM(16,4,0,1)
240 uint32_t next_hop_index;
242 uint8_t next_hop_index;
244 if (unlikely(rte_lpm_lookup(task->ipv4_lpm, rte_bswap32(dst_ip), &next_hop_index) != 0)) {
245 uint8_t* dst_ipp = (uint8_t*)&dst_ip;
246 plog_warn("lpm_lookup failed for ip %d.%d.%d.%d: rc = %d\n",
247 dst_ipp[0], dst_ipp[1], dst_ipp[2], dst_ipp[3], -ENOENT);
251 tx_port = task->next_hops[next_hop_index].mac_port.out_idx;
252 if (task->runtime_flags & TASK_MPLS_TAGGING) {
253 uint16_t padlen = rte_pktmbuf_pkt_len(mbuf) - rte_be_to_cpu_16(ip->total_length) - ip_offset;
255 rte_pktmbuf_trim(mbuf, padlen);
258 set_l2_mpls(task, mbuf, next_hop_index, ip_offset);
261 set_l2(task, mbuf, next_hop_index);
266 static inline uint8_t handle_routing(struct task_routing *task, struct rte_mbuf *mbuf)
268 struct qinq_hdr *qinq;
269 struct ether_hdr *peth = rte_pktmbuf_mtod(mbuf, struct ether_hdr *);
271 switch (peth->ether_type) {
273 struct qinq_hdr *qinq = (struct qinq_hdr *)peth;
274 if ((qinq->cvlan.eth_proto != ETYPE_VLAN)) {
275 plog_warn("Unexpected proto in QinQ = %#04x\n", qinq->cvlan.eth_proto);
279 return route_ipv4(task, (uint8_t*)qinq, sizeof(*qinq), mbuf);
282 return route_ipv4(task, (uint8_t*)peth, sizeof(*peth), mbuf);
284 /* skip MPLS headers if any for routing */
285 struct mpls_hdr *mpls = (struct mpls_hdr *)(peth + 1);
286 uint32_t count = sizeof(struct ether_hdr);
287 while (!(mpls->bytes & 0x00010000)) {
289 count += sizeof(struct mpls_hdr);
291 count += sizeof(struct mpls_hdr);
293 return route_ipv4(task, (uint8_t*)peth, count, mbuf);
296 if (peth->ether_type == task->qinq_tag) {
297 struct qinq_hdr *qinq = (struct qinq_hdr *)peth;
298 if ((qinq->cvlan.eth_proto != ETYPE_VLAN)) {
299 plog_warn("Unexpected proto in QinQ = %#04x\n", qinq->cvlan.eth_proto);
303 return route_ipv4(task, (uint8_t*)qinq, sizeof(*qinq), mbuf);
305 plog_warn("Failed routing packet: ether_type %#06x is unknown\n", peth->ether_type);
310 static struct task_init task_init_routing = {
311 .mode_str = "routing",
312 .init = init_task_routing,
313 .handle = handle_routing_bulk,
314 .flag_features = TASK_FEATURE_ROUTING,
315 .size = sizeof(struct task_routing)
318 __attribute__((constructor)) static void reg_task_routing(void)
320 reg_task(&task_init_routing);