ceph/src/dpdk/app/test-pmd/txonly.c

   1 /*-
   2  *   BSD LICENSE
   3  *
   4  *   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
   5  *   All rights reserved.
   6  *
   7  *   Redistribution and use in source and binary forms, with or without
   8  *   modification, are permitted provided that the following conditions
   9  *   are met:
  10  *
  11  *     * Redistributions of source code must retain the above copyright
  12  *       notice, this list of conditions and the following disclaimer.
  13  *     * Redistributions in binary form must reproduce the above copyright
  14  *       notice, this list of conditions and the following disclaimer in
  15  *       the documentation and/or other materials provided with the
  16  *       distribution.
  17  *     * Neither the name of Intel Corporation nor the names of its
  18  *       contributors may be used to endorse or promote products derived
  19  *       from this software without specific prior written permission.
  20  *
  21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  32  */
  33
  34 #include <stdarg.h>
  35 #include <string.h>
  36 #include <stdio.h>
  37 #include <errno.h>
  38 #include <stdint.h>
  39 #include <unistd.h>
  40 #include <inttypes.h>
  41
  42 #include <sys/queue.h>
  43 #include <sys/stat.h>
  44
  45 #include <rte_common.h>
  46 #include <rte_byteorder.h>
  47 #include <rte_log.h>
  48 #include <rte_debug.h>
  49 #include <rte_cycles.h>
  50 #include <rte_memory.h>
  51 #include <rte_memcpy.h>
  52 #include <rte_memzone.h>
  53 #include <rte_launch.h>
  54 #include <rte_eal.h>
  55 #include <rte_per_lcore.h>
  56 #include <rte_lcore.h>
  57 #include <rte_atomic.h>
  58 #include <rte_branch_prediction.h>
  59 #include <rte_memory.h>
  60 #include <rte_mempool.h>
  61 #include <rte_mbuf.h>
  62 #include <rte_memcpy.h>
  63 #include <rte_interrupts.h>
  64 #include <rte_pci.h>
  65 #include <rte_ether.h>
  66 #include <rte_ethdev.h>
  67 #include <rte_ip.h>
  68 #include <rte_tcp.h>
  69 #include <rte_udp.h>
  70 #include <rte_string_fns.h>
  71
  72 #include "testpmd.h"
  73
  74 #define UDP_SRC_PORT 1024
  75 #define UDP_DST_PORT 1024
  76
  77 #define IP_SRC_ADDR ((192U << 24) | (168 << 16) | (0 << 8) | 1)
  78 #define IP_DST_ADDR ((192U << 24) | (168 << 16) | (0 << 8) | 2)
  79
  80 #define IP_DEFTTL  64   /* from RFC 1340. */
  81 #define IP_VERSION 0x40
  82 #define IP_HDRLEN  0x05 /* default IP header length == five 32-bits words. */
  83 #define IP_VHL_DEF (IP_VERSION | IP_HDRLEN)
  84
  85 static struct ipv4_hdr  pkt_ip_hdr;  /**< IP header of transmitted packets. */
  86 static struct udp_hdr pkt_udp_hdr; /**< UDP header of transmitted packets. */
  87
  88 static void
  89 copy_buf_to_pkt_segs(void* buf, unsigned len, struct rte_mbuf *pkt,
  90                      unsigned offset)
  91 {
  92         struct rte_mbuf *seg;
  93         void *seg_buf;
  94         unsigned copy_len;
  95
  96         seg = pkt;
  97         while (offset >= seg->data_len) {
  98                 offset -= seg->data_len;
  99                 seg = seg->next;
 100         }
 101         copy_len = seg->data_len - offset;
 102         seg_buf = rte_pktmbuf_mtod_offset(seg, char *, offset);
 103         while (len > copy_len) {
 104                 rte_memcpy(seg_buf, buf, (size_t) copy_len);
 105                 len -= copy_len;
 106                 buf = ((char*) buf + copy_len);
 107                 seg = seg->next;
 108                 seg_buf = rte_pktmbuf_mtod(seg, char *);
 109         }
 110         rte_memcpy(seg_buf, buf, (size_t) len);
 111 }
 112
 113 static inline void
 114 copy_buf_to_pkt(void* buf, unsigned len, struct rte_mbuf *pkt, unsigned offset)
 115 {
 116         if (offset + len <= pkt->data_len) {
 117                 rte_memcpy(rte_pktmbuf_mtod_offset(pkt, char *, offset),
 118                         buf, (size_t) len);
 119                 return;
 120         }
 121         copy_buf_to_pkt_segs(buf, len, pkt, offset);
 122 }
 123
 124 static void
 125 setup_pkt_udp_ip_headers(struct ipv4_hdr *ip_hdr,
 126                          struct udp_hdr *udp_hdr,
 127                          uint16_t pkt_data_len)
 128 {
 129         uint16_t *ptr16;
 130         uint32_t ip_cksum;
 131         uint16_t pkt_len;
 132
 133         /*
 134          * Initialize UDP header.
 135          */
 136         pkt_len = (uint16_t) (pkt_data_len + sizeof(struct udp_hdr));
 137         udp_hdr->src_port = rte_cpu_to_be_16(UDP_SRC_PORT);
 138         udp_hdr->dst_port = rte_cpu_to_be_16(UDP_DST_PORT);
 139         udp_hdr->dgram_len      = RTE_CPU_TO_BE_16(pkt_len);
 140         udp_hdr->dgram_cksum    = 0; /* No UDP checksum. */
 141
 142         /*
 143          * Initialize IP header.
 144          */
 145         pkt_len = (uint16_t) (pkt_len + sizeof(struct ipv4_hdr));
 146         ip_hdr->version_ihl   = IP_VHL_DEF;
 147         ip_hdr->type_of_service   = 0;
 148         ip_hdr->fragment_offset = 0;
 149         ip_hdr->time_to_live   = IP_DEFTTL;
 150         ip_hdr->next_proto_id = IPPROTO_UDP;
 151         ip_hdr->packet_id = 0;
 152         ip_hdr->total_length   = RTE_CPU_TO_BE_16(pkt_len);
 153         ip_hdr->src_addr = rte_cpu_to_be_32(IP_SRC_ADDR);
 154         ip_hdr->dst_addr = rte_cpu_to_be_32(IP_DST_ADDR);
 155
 156         /*
 157          * Compute IP header checksum.
 158          */
 159         ptr16 = (unaligned_uint16_t*) ip_hdr;
 160         ip_cksum = 0;
 161         ip_cksum += ptr16[0]; ip_cksum += ptr16[1];
 162         ip_cksum += ptr16[2]; ip_cksum += ptr16[3];
 163         ip_cksum += ptr16[4];
 164         ip_cksum += ptr16[6]; ip_cksum += ptr16[7];
 165         ip_cksum += ptr16[8]; ip_cksum += ptr16[9];
 166
 167         /*
 168          * Reduce 32 bit checksum to 16 bits and complement it.
 169          */
 170         ip_cksum = ((ip_cksum & 0xFFFF0000) >> 16) +
 171                 (ip_cksum & 0x0000FFFF);
 172         if (ip_cksum > 65535)
 173                 ip_cksum -= 65535;
 174         ip_cksum = (~ip_cksum) & 0x0000FFFF;
 175         if (ip_cksum == 0)
 176                 ip_cksum = 0xFFFF;
 177         ip_hdr->hdr_checksum = (uint16_t) ip_cksum;
 178 }
 179
 180 /*
 181  * Transmit a burst of multi-segments packets.
 182  */
 183 static void
 184 pkt_burst_transmit(struct fwd_stream *fs)
 185 {
 186         struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
 187         struct rte_port *txp;
 188         struct rte_mbuf *pkt;
 189         struct rte_mbuf *pkt_seg;
 190         struct rte_mempool *mbp;
 191         struct ether_hdr eth_hdr;
 192         uint16_t nb_tx;
 193         uint16_t nb_pkt;
 194         uint16_t vlan_tci, vlan_tci_outer;
 195         uint32_t retry;
 196         uint64_t ol_flags = 0;
 197         uint8_t  i;
 198 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
 199         uint64_t start_tsc;
 200         uint64_t end_tsc;
 201         uint64_t core_cycles;
 202 #endif
 203         uint32_t nb_segs, pkt_len;
 204
 205 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
 206         start_tsc = rte_rdtsc();
 207 #endif
 208
 209         mbp = current_fwd_lcore()->mbp;
 210         txp = &ports[fs->tx_port];
 211         vlan_tci = txp->tx_vlan_id;
 212         vlan_tci_outer = txp->tx_vlan_id_outer;
 213         if (txp->tx_ol_flags & TESTPMD_TX_OFFLOAD_INSERT_VLAN)
 214                 ol_flags = PKT_TX_VLAN_PKT;
 215         if (txp->tx_ol_flags & TESTPMD_TX_OFFLOAD_INSERT_QINQ)
 216                 ol_flags |= PKT_TX_QINQ_PKT;
 217         for (nb_pkt = 0; nb_pkt < nb_pkt_per_burst; nb_pkt++) {
 218                 pkt = rte_mbuf_raw_alloc(mbp);
 219                 if (pkt == NULL) {
 220                 nomore_mbuf:
 221                         if (nb_pkt == 0)
 222                                 return;
 223                         break;
 224                 }
 225
 226                 /*
 227                  * Using raw alloc is good to improve performance,
 228                  * but some consumers may use the headroom and so
 229                  * decrement data_off. We need to make sure it is
 230                  * reset to default value.
 231                  */
 232                 rte_pktmbuf_reset_headroom(pkt);
 233                 pkt->data_len = tx_pkt_seg_lengths[0];
 234                 pkt_seg = pkt;
 235                 if (tx_pkt_split == TX_PKT_SPLIT_RND)
 236                         nb_segs = random() % tx_pkt_nb_segs + 1;
 237                 else
 238                         nb_segs = tx_pkt_nb_segs;
 239                 pkt_len = pkt->data_len;
 240                 for (i = 1; i < nb_segs; i++) {
 241                         pkt_seg->next = rte_mbuf_raw_alloc(mbp);
 242                         if (pkt_seg->next == NULL) {
 243                                 pkt->nb_segs = i;
 244                                 rte_pktmbuf_free(pkt);
 245                                 goto nomore_mbuf;
 246                         }
 247                         pkt_seg = pkt_seg->next;
 248                         pkt_seg->data_len = tx_pkt_seg_lengths[i];
 249                         pkt_len += pkt_seg->data_len;
 250                 }
 251                 pkt_seg->next = NULL; /* Last segment of packet. */
 252
 253                 /*
 254                  * Initialize Ethernet header.
 255                  */
 256                 ether_addr_copy(&peer_eth_addrs[fs->peer_addr],&eth_hdr.d_addr);
 257                 ether_addr_copy(&ports[fs->tx_port].eth_addr, &eth_hdr.s_addr);
 258                 eth_hdr.ether_type = rte_cpu_to_be_16(ETHER_TYPE_IPv4);
 259
 260                 /*
 261                  * Copy headers in first packet segment(s).
 262                  */
 263                 copy_buf_to_pkt(&eth_hdr, sizeof(eth_hdr), pkt, 0);
 264                 copy_buf_to_pkt(&pkt_ip_hdr, sizeof(pkt_ip_hdr), pkt,
 265                                 sizeof(struct ether_hdr));
 266                 copy_buf_to_pkt(&pkt_udp_hdr, sizeof(pkt_udp_hdr), pkt,
 267                                 sizeof(struct ether_hdr) +
 268                                 sizeof(struct ipv4_hdr));
 269
 270                 /*
 271                  * Complete first mbuf of packet and append it to the
 272                  * burst of packets to be transmitted.
 273                  */
 274                 pkt->nb_segs = nb_segs;
 275                 pkt->pkt_len = pkt_len;
 276                 pkt->ol_flags = ol_flags;
 277                 pkt->vlan_tci = vlan_tci;
 278                 pkt->vlan_tci_outer = vlan_tci_outer;
 279                 pkt->l2_len = sizeof(struct ether_hdr);
 280                 pkt->l3_len = sizeof(struct ipv4_hdr);
 281                 pkts_burst[nb_pkt] = pkt;
 282         }
 283         nb_tx = rte_eth_tx_burst(fs->tx_port, fs->tx_queue, pkts_burst, nb_pkt);
 284         /*
 285          * Retry if necessary
 286          */
 287         if (unlikely(nb_tx < nb_pkt) && fs->retry_enabled) {
 288                 retry = 0;
 289                 while (nb_tx < nb_pkt && retry++ < burst_tx_retry_num) {
 290                         rte_delay_us(burst_tx_delay_time);
 291                         nb_tx += rte_eth_tx_burst(fs->tx_port, fs->tx_queue,
 292                                         &pkts_burst[nb_tx], nb_pkt - nb_tx);
 293                 }
 294         }
 295         fs->tx_packets += nb_tx;
 296
 297 #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
 298         fs->tx_burst_stats.pkt_burst_spread[nb_tx]++;
 299 #endif
 300         if (unlikely(nb_tx < nb_pkt)) {
 301                 if (verbose_level > 0 && fs->fwd_dropped == 0)
 302                         printf("port %d tx_queue %d - drop "
 303                                "(nb_pkt:%u - nb_tx:%u)=%u packets\n",
 304                                fs->tx_port, fs->tx_queue,
 305                                (unsigned) nb_pkt, (unsigned) nb_tx,
 306                                (unsigned) (nb_pkt - nb_tx));
 307                 fs->fwd_dropped += (nb_pkt - nb_tx);
 308                 do {
 309                         rte_pktmbuf_free(pkts_burst[nb_tx]);
 310                 } while (++nb_tx < nb_pkt);
 311         }
 312
 313 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
 314         end_tsc = rte_rdtsc();
 315         core_cycles = (end_tsc - start_tsc);
 316         fs->core_cycles = (uint64_t) (fs->core_cycles + core_cycles);
 317 #endif
 318 }
 319
 320 static void
 321 tx_only_begin(__attribute__((unused)) portid_t pi)
 322 {
 323         uint16_t pkt_data_len;
 324
 325         pkt_data_len = (uint16_t) (tx_pkt_length - (sizeof(struct ether_hdr) +
 326                                                     sizeof(struct ipv4_hdr) +
 327                                                     sizeof(struct udp_hdr)));
 328         setup_pkt_udp_ip_headers(&pkt_ip_hdr, &pkt_udp_hdr, pkt_data_len);
 329 }
 330
 331 struct fwd_engine tx_only_engine = {
 332         .fwd_mode_name  = "txonly",
 333         .port_fwd_begin = tx_only_begin,
 334         .port_fwd_end   = NULL,
 335         .packet_fwd     = pkt_burst_transmit,
 336 };