4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
5 * Copyright(c) 2014 6WIND S.A.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * * Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * * Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * * Neither the name of Intel Corporation nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
37 #include <rte_ethdev.h>
38 #include <rte_malloc.h>
39 #include <rte_memcpy.h>
40 #include <rte_string_fns.h>
41 #include <rte_cycles.h>
42 #include <rte_kvargs.h>
44 #include "rte_eth_pcap.h"
46 #define RTE_ETH_PCAP_SNAPSHOT_LEN 65535
47 #define RTE_ETH_PCAP_SNAPLEN 4096
48 #define RTE_ETH_PCAP_PROMISC 1
49 #define RTE_ETH_PCAP_TIMEOUT -1
50 #define ETH_PCAP_RX_PCAP_ARG "rx_pcap"
51 #define ETH_PCAP_TX_PCAP_ARG "tx_pcap"
52 #define ETH_PCAP_RX_IFACE_ARG "rx_iface"
53 #define ETH_PCAP_TX_IFACE_ARG "tx_iface"
54 #define ETH_PCAP_IFACE_ARG "iface"
56 static char errbuf[PCAP_ERRBUF_SIZE];
57 static struct timeval start_time;
58 static uint64_t start_cycles;
61 struct pcap_rx_queue {
63 struct rte_mempool *mb_pool;
64 volatile unsigned long rx_pkts;
65 volatile unsigned long err_pkts;
68 struct pcap_tx_queue {
69 pcap_dumper_t *dumper;
71 volatile unsigned long tx_pkts;
72 volatile unsigned long err_pkts;
77 pcap_t *pcaps[RTE_PMD_RING_MAX_RX_RINGS];
82 pcap_dumper_t *dumpers[RTE_PMD_RING_MAX_TX_RINGS];
83 pcap_t *pcaps[RTE_PMD_RING_MAX_RX_RINGS];
86 struct pmd_internals {
87 unsigned nb_rx_queues;
88 unsigned nb_tx_queues;
90 struct pcap_rx_queue rx_queue[RTE_PMD_RING_MAX_RX_RINGS];
91 struct pcap_tx_queue tx_queue[RTE_PMD_RING_MAX_TX_RINGS];
94 const char *valid_arguments[] = {
97 ETH_PCAP_RX_IFACE_ARG,
98 ETH_PCAP_TX_IFACE_ARG,
103 static struct ether_addr eth_addr = { .addr_bytes = { 0, 0, 0, 0x1, 0x2, 0x3 } };
104 static const char *drivername = "Pcap PMD";
105 static struct rte_eth_link pmd_link = {
107 .link_duplex = ETH_LINK_FULL_DUPLEX,
113 eth_pcap_rx(void *queue,
114 struct rte_mbuf **bufs,
118 struct pcap_pkthdr header;
119 const u_char *packet;
120 struct rte_mbuf *mbuf;
121 struct pcap_rx_queue *pcap_q = queue;
122 struct rte_pktmbuf_pool_private *mbp_priv;
126 if (unlikely(pcap_q->pcap == NULL || nb_pkts == 0))
129 /* Reads the given number of packets from the pcap file one by one
130 * and copies the packet data into a newly allocated mbuf to return.
132 for (i = 0; i < nb_pkts; i++) {
133 /* Get the next PCAP packet */
134 packet = pcap_next(pcap_q->pcap, &header);
135 if (unlikely(packet == NULL))
138 mbuf = rte_pktmbuf_alloc(pcap_q->mb_pool);
139 if (unlikely(mbuf == NULL))
142 /* Now get the space available for data in the mbuf */
143 mbp_priv = rte_mempool_get_priv(pcap_q->mb_pool);
144 buf_size = (uint16_t) (mbp_priv->mbuf_data_room_size -
145 RTE_PKTMBUF_HEADROOM);
147 if (header.len <= buf_size) {
148 /* pcap packet will fit in the mbuf, go ahead and copy */
149 rte_memcpy(mbuf->pkt.data, packet, header.len);
150 mbuf->pkt.data_len = (uint16_t)header.len;
151 mbuf->pkt.pkt_len = mbuf->pkt.data_len;
155 /* pcap packet will not fit in the mbuf, so drop packet */
157 "PCAP packet %d bytes will not fit in mbuf (%d bytes)\n",
158 header.len, buf_size);
159 rte_pktmbuf_free(mbuf);
162 pcap_q->rx_pkts += num_rx;
167 calculate_timestamp(struct timeval *ts) {
169 struct timeval cur_time;
171 cycles = rte_get_timer_cycles() - start_cycles;
172 cur_time.tv_sec = cycles / hz;
173 cur_time.tv_usec = (cycles % hz) * 10e6 / hz;
174 timeradd(&start_time, &cur_time, ts);
178 * Callback to handle writing packets to a pcap file.
181 eth_pcap_tx_dumper(void *queue,
182 struct rte_mbuf **bufs,
186 struct rte_mbuf *mbuf;
187 struct pcap_tx_queue *dumper_q = queue;
189 struct pcap_pkthdr header;
191 if (dumper_q->dumper == NULL || nb_pkts == 0)
194 /* writes the nb_pkts packets to the previously opened pcap file dumper */
195 for (i = 0; i < nb_pkts; i++) {
197 calculate_timestamp(&header.ts);
198 header.len = mbuf->pkt.data_len;
199 header.caplen = header.len;
200 pcap_dump((u_char*) dumper_q->dumper, &header, mbuf->pkt.data);
201 rte_pktmbuf_free(mbuf);
206 * Since there's no place to hook a callback when the forwarding
207 * process stops and to make sure the pcap file is actually written,
208 * we flush the pcap dumper within each burst.
210 pcap_dump_flush(dumper_q->dumper);
211 dumper_q->tx_pkts += num_tx;
212 dumper_q->err_pkts += nb_pkts - num_tx;
218 * Callback to handle sending packets through a real NIC.
221 eth_pcap_tx(void *queue,
222 struct rte_mbuf **bufs,
227 struct rte_mbuf *mbuf;
228 struct pcap_tx_queue *tx_queue = queue;
231 if (unlikely(nb_pkts == 0 || tx_queue->pcap == NULL))
234 for (i = 0; i < nb_pkts; i++) {
236 ret = pcap_sendpacket(tx_queue->pcap, (u_char*) mbuf->pkt.data,
240 rte_pktmbuf_free(mbuf);
243 tx_queue->tx_pkts += num_tx;
244 tx_queue->err_pkts += nb_pkts - num_tx;
249 eth_pcap_tx(__rte_unused void *queue,
250 __rte_unused struct rte_mbuf **bufs,
251 __rte_unused uint16_t nb_pkts)
253 RTE_LOG(ERR, PMD, "pcap library cannot send packets, please rebuild "
254 "with a more up to date libpcap\n");
260 eth_dev_start(struct rte_eth_dev *dev)
262 dev->data->dev_link.link_status = 1;
267 * This function gets called when the current port gets stopped.
268 * Is the only place for us to close all the tx streams dumpers.
269 * If not called the dumpers will be flushed within each tx burst.
272 eth_dev_stop(struct rte_eth_dev *dev)
275 pcap_dumper_t *dumper;
277 struct pmd_internals *internals = dev->data->dev_private;
279 for (i = 0; i < internals->nb_tx_queues; i++) {
280 dumper = internals->tx_queue[i].dumper;
282 pcap_dump_close(dumper);
283 pcap = internals->tx_queue[i].pcap;
288 dev->data->dev_link.link_status = 0;
292 eth_dev_configure(struct rte_eth_dev *dev __rte_unused)
298 eth_dev_info(struct rte_eth_dev *dev,
299 struct rte_eth_dev_info *dev_info)
301 struct pmd_internals *internals = dev->data->dev_private;
302 dev_info->driver_name = drivername;
303 dev_info->max_mac_addrs = 1;
304 dev_info->max_rx_pktlen = (uint32_t) -1;
305 dev_info->max_rx_queues = (uint16_t)internals->nb_rx_queues;
306 dev_info->max_tx_queues = (uint16_t)internals->nb_tx_queues;
307 dev_info->min_rx_bufsize = 0;
308 dev_info->pci_dev = NULL;
312 eth_stats_get(struct rte_eth_dev *dev,
313 struct rte_eth_stats *igb_stats)
316 unsigned long rx_total = 0, tx_total = 0, tx_err_total = 0;
317 const struct pmd_internals *internal = dev->data->dev_private;
319 memset(igb_stats, 0, sizeof(*igb_stats));
320 for (i = 0; i < RTE_ETHDEV_QUEUE_STAT_CNTRS && i < internal->nb_rx_queues;
322 igb_stats->q_ipackets[i] = internal->rx_queue[i].rx_pkts;
323 rx_total += igb_stats->q_ipackets[i];
326 for (i = 0; i < RTE_ETHDEV_QUEUE_STAT_CNTRS && i < internal->nb_tx_queues;
328 igb_stats->q_opackets[i] = internal->tx_queue[i].tx_pkts;
329 igb_stats->q_errors[i] = internal->tx_queue[i].err_pkts;
330 tx_total += igb_stats->q_opackets[i];
331 tx_err_total += igb_stats->q_errors[i];
334 igb_stats->ipackets = rx_total;
335 igb_stats->opackets = tx_total;
336 igb_stats->oerrors = tx_err_total;
340 eth_stats_reset(struct rte_eth_dev *dev)
343 struct pmd_internals *internal = dev->data->dev_private;
344 for (i = 0; i < internal->nb_rx_queues; i++)
345 internal->rx_queue[i].rx_pkts = 0;
346 for (i = 0; i < internal->nb_tx_queues; i++) {
347 internal->tx_queue[i].tx_pkts = 0;
348 internal->tx_queue[i].err_pkts = 0;
353 eth_dev_close(struct rte_eth_dev *dev __rte_unused)
358 eth_queue_release(void *q __rte_unused)
363 eth_link_update(struct rte_eth_dev *dev __rte_unused,
364 int wait_to_complete __rte_unused)
370 eth_rx_queue_setup(struct rte_eth_dev *dev,
371 uint16_t rx_queue_id,
372 uint16_t nb_rx_desc __rte_unused,
373 unsigned int socket_id __rte_unused,
374 const struct rte_eth_rxconf *rx_conf __rte_unused,
375 struct rte_mempool *mb_pool)
377 struct pmd_internals *internals = dev->data->dev_private;
378 struct pcap_rx_queue *pcap_q = &internals->rx_queue[rx_queue_id];
379 pcap_q->mb_pool = mb_pool;
380 dev->data->rx_queues[rx_queue_id] = pcap_q;
385 eth_tx_queue_setup(struct rte_eth_dev *dev,
386 uint16_t tx_queue_id,
387 uint16_t nb_tx_desc __rte_unused,
388 unsigned int socket_id __rte_unused,
389 const struct rte_eth_txconf *tx_conf __rte_unused)
392 struct pmd_internals *internals = dev->data->dev_private;
393 dev->data->tx_queues[tx_queue_id] = &internals->tx_queue[tx_queue_id];
397 static struct eth_dev_ops ops = {
398 .dev_start = eth_dev_start,
399 .dev_stop = eth_dev_stop,
400 .dev_close = eth_dev_close,
401 .dev_configure = eth_dev_configure,
402 .dev_infos_get = eth_dev_info,
403 .rx_queue_setup = eth_rx_queue_setup,
404 .tx_queue_setup = eth_tx_queue_setup,
405 .rx_queue_release = eth_queue_release,
406 .tx_queue_release = eth_queue_release,
407 .link_update = eth_link_update,
408 .stats_get = eth_stats_get,
409 .stats_reset = eth_stats_reset,
413 * Function handler that opens the pcap file for reading a stores a
414 * reference of it for use it later on.
417 open_rx_pcap(const char *key __rte_unused, const char *value, void *extra_args)
420 const char *pcap_filename = value;
421 struct rx_pcaps *pcaps = extra_args;
424 for (i = 0; i < pcaps->num_of_rx; i++) {
425 if ((rx_pcap = pcap_open_offline(pcap_filename, errbuf)) == NULL) {
426 RTE_LOG(ERR, PMD, "Couldn't open %s: %s\n", pcap_filename, errbuf);
429 pcaps->pcaps[i] = rx_pcap;
436 * Opens a pcap file for writing and stores a reference to it
437 * for use it later on.
440 open_tx_pcap(const char *key __rte_unused, const char *value, void *extra_args)
443 const char *pcap_filename = value;
444 struct tx_pcaps *dumpers = extra_args;
446 pcap_dumper_t *dumper;
448 for (i = 0; i < dumpers->num_of_tx; i++) {
450 * We need to create a dummy empty pcap_t to use it
451 * with pcap_dump_open(). We create big enough an Ethernet
454 if ((tx_pcap = pcap_open_dead(DLT_EN10MB, RTE_ETH_PCAP_SNAPSHOT_LEN))
456 RTE_LOG(ERR, PMD, "Couldn't create dead pcap\n");
460 /* The dumper is created using the previous pcap_t reference */
461 if ((dumper = pcap_dump_open(tx_pcap, pcap_filename)) == NULL) {
462 RTE_LOG(ERR, PMD, "Couldn't open %s for writing.\n", pcap_filename);
465 dumpers->dumpers[i] = dumper;
472 * pcap_open_live wrapper function
475 open_iface_live(const char *iface, pcap_t **pcap) {
476 *pcap = pcap_open_live(iface, RTE_ETH_PCAP_SNAPLEN,
477 RTE_ETH_PCAP_PROMISC, RTE_ETH_PCAP_TIMEOUT, errbuf);
480 RTE_LOG(ERR, PMD, "Couldn't open %s: %s\n", iface, errbuf);
487 * Opens an interface for reading and writing
490 open_rx_tx_iface(const char *key __rte_unused, const char *value, void *extra_args)
492 const char *iface = value;
493 pcap_t **pcap = extra_args;
495 if(open_iface_live(iface, pcap) < 0)
501 * Opens a NIC for reading packets from it
504 open_rx_iface(const char *key __rte_unused, const char *value, void *extra_args)
507 const char *iface = value;
508 struct rx_pcaps *pcaps = extra_args;
511 for (i = 0; i < pcaps->num_of_rx; i++) {
512 if(open_iface_live(iface, &pcap) < 0)
514 pcaps->pcaps[i] = pcap;
521 * Opens a NIC for writing packets to it
524 open_tx_iface(const char *key __rte_unused, const char *value, void *extra_args)
527 const char *iface = value;
528 struct tx_pcaps *pcaps = extra_args;
531 for (i = 0; i < pcaps->num_of_tx; i++) {
532 if(open_iface_live(iface, &pcap) < 0)
534 pcaps->pcaps[i] = pcap;
542 rte_pmd_init_internals(const unsigned nb_rx_queues,
543 const unsigned nb_tx_queues,
544 const unsigned numa_node,
545 struct pmd_internals **internals,
546 struct rte_eth_dev **eth_dev)
548 struct rte_eth_dev_data *data = NULL;
549 struct rte_pci_device *pci_dev = NULL;
552 "Creating pcap-backed ethdev on numa socket %u\n", numa_node);
554 /* now do all data allocation - for eth_dev structure, dummy pci driver
555 * and internal (private) data
557 data = rte_zmalloc_socket(NULL, sizeof(*data), 0, numa_node);
561 pci_dev = rte_zmalloc_socket(NULL, sizeof(*pci_dev), 0, numa_node);
565 *internals = rte_zmalloc_socket(NULL, sizeof(**internals), 0, numa_node);
566 if (*internals == NULL)
569 /* reserve an ethdev entry */
570 *eth_dev = rte_eth_dev_allocate();
571 if (*eth_dev == NULL)
574 /* now put it all together
575 * - store queue data in internals,
576 * - store numa_node info in pci_driver
577 * - point eth_dev_data to internals and pci_driver
578 * - and point eth_dev structure to new eth_dev_data structure
580 /* NOTE: we'll replace the data element, of originally allocated eth_dev
581 * so the rings are local per-process */
583 (*internals)->nb_rx_queues = nb_rx_queues;
584 (*internals)->nb_tx_queues = nb_tx_queues;
586 pci_dev->numa_node = numa_node;
588 data->dev_private = *internals;
589 data->port_id = (*eth_dev)->data->port_id;
590 data->nb_rx_queues = (uint16_t)nb_rx_queues;
591 data->nb_tx_queues = (uint16_t)nb_tx_queues;
592 data->dev_link = pmd_link;
593 data->mac_addrs = ð_addr;
595 (*eth_dev)->data = data;
596 (*eth_dev)->dev_ops = &ops;
597 (*eth_dev)->pci_dev = pci_dev;
606 rte_free(*internals);
611 rte_eth_from_pcaps_n_dumpers(pcap_t * const rx_queues[],
612 const unsigned nb_rx_queues,
613 pcap_dumper_t * const tx_queues[],
614 const unsigned nb_tx_queues,
615 const unsigned numa_node)
617 struct pmd_internals *internals = NULL;
618 struct rte_eth_dev *eth_dev = NULL;
621 /* do some parameter checking */
622 if (rx_queues == NULL && nb_rx_queues > 0)
624 if (tx_queues == NULL && nb_tx_queues > 0)
627 if (rte_pmd_init_internals(nb_rx_queues, nb_tx_queues, numa_node,
628 &internals, ð_dev) < 0)
631 for (i = 0; i < nb_rx_queues; i++) {
632 internals->rx_queue->pcap = rx_queues[i];
634 for (i = 0; i < nb_tx_queues; i++) {
635 internals->tx_queue->dumper = tx_queues[i];
638 eth_dev->rx_pkt_burst = eth_pcap_rx;
639 eth_dev->tx_pkt_burst = eth_pcap_tx_dumper;
645 rte_eth_from_pcaps(pcap_t * const rx_queues[],
646 const unsigned nb_rx_queues,
647 pcap_t * const tx_queues[],
648 const unsigned nb_tx_queues,
649 const unsigned numa_node)
651 struct pmd_internals *internals = NULL;
652 struct rte_eth_dev *eth_dev = NULL;
655 /* do some parameter checking */
656 if (rx_queues == NULL && nb_rx_queues > 0)
658 if (tx_queues == NULL && nb_tx_queues > 0)
661 if (rte_pmd_init_internals(nb_rx_queues, nb_tx_queues, numa_node,
662 &internals, ð_dev) < 0)
665 for (i = 0; i < nb_rx_queues; i++) {
666 internals->rx_queue->pcap = rx_queues[i];
668 for (i = 0; i < nb_tx_queues; i++) {
669 internals->tx_queue->pcap = tx_queues[i];
672 eth_dev->rx_pkt_burst = eth_pcap_rx;
673 eth_dev->tx_pkt_burst = eth_pcap_tx;
680 rte_pmd_pcap_init(const char *name, const char *params)
682 unsigned numa_node, using_dumpers = 0;
684 struct rte_kvargs *kvlist;
685 struct rx_pcaps pcaps;
686 struct tx_pcaps dumpers;
688 RTE_LOG(INFO, PMD, "Initializing pmd_pcap for %s\n", name);
690 numa_node = rte_socket_id();
692 gettimeofday(&start_time, NULL);
693 start_cycles = rte_get_timer_cycles();
694 hz = rte_get_timer_hz();
696 kvlist = rte_kvargs_parse(params, valid_arguments);
701 * If iface argument is passed we open the NICs and use them for
704 if (rte_kvargs_count(kvlist, ETH_PCAP_IFACE_ARG) == 1) {
706 ret = rte_kvargs_process(kvlist, ETH_PCAP_IFACE_ARG,
707 &open_rx_tx_iface, &pcaps.pcaps[0]);
711 return rte_eth_from_pcaps(pcaps.pcaps, 1, pcaps.pcaps, 1, numa_node);
715 * We check whether we want to open a RX stream from a real NIC or a
718 if ((pcaps.num_of_rx = rte_kvargs_count(kvlist, ETH_PCAP_RX_PCAP_ARG))) {
719 ret = rte_kvargs_process(kvlist, ETH_PCAP_RX_PCAP_ARG,
720 &open_rx_pcap, &pcaps);
722 pcaps.num_of_rx = rte_kvargs_count(kvlist,
723 ETH_PCAP_RX_IFACE_ARG);
724 ret = rte_kvargs_process(kvlist, ETH_PCAP_RX_IFACE_ARG,
725 &open_rx_iface, &pcaps);
732 * We check whether we want to open a TX stream to a real NIC or a
735 if ((dumpers.num_of_tx = rte_kvargs_count(kvlist,
736 ETH_PCAP_TX_PCAP_ARG))) {
737 ret = rte_kvargs_process(kvlist, ETH_PCAP_TX_PCAP_ARG,
738 &open_tx_pcap, &dumpers);
741 dumpers.num_of_tx = rte_kvargs_count(kvlist,
742 ETH_PCAP_TX_IFACE_ARG);
743 ret = rte_kvargs_process(kvlist, ETH_PCAP_TX_IFACE_ARG,
744 &open_tx_iface, &dumpers);
751 return rte_eth_from_pcaps_n_dumpers(pcaps.pcaps, pcaps.num_of_rx,
752 dumpers.dumpers, dumpers.num_of_tx, numa_node);
754 return rte_eth_from_pcaps(pcaps.pcaps, pcaps.num_of_rx, dumpers.pcaps,
755 dumpers.num_of_tx, numa_node);