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>
46 #include "rte_eth_pcap.h"
48 #define RTE_ETH_PCAP_SNAPSHOT_LEN 65535
49 #define RTE_ETH_PCAP_SNAPLEN 4096
50 #define RTE_ETH_PCAP_PROMISC 1
51 #define RTE_ETH_PCAP_TIMEOUT -1
52 #define ETH_PCAP_RX_PCAP_ARG "rx_pcap"
53 #define ETH_PCAP_TX_PCAP_ARG "tx_pcap"
54 #define ETH_PCAP_RX_IFACE_ARG "rx_iface"
55 #define ETH_PCAP_TX_IFACE_ARG "tx_iface"
56 #define ETH_PCAP_IFACE_ARG "iface"
58 static char errbuf[PCAP_ERRBUF_SIZE];
59 static struct timeval start_time;
60 static uint64_t start_cycles;
63 struct pcap_rx_queue {
65 struct rte_mempool *mb_pool;
66 volatile unsigned long rx_pkts;
67 volatile unsigned long err_pkts;
70 struct pcap_tx_queue {
71 pcap_dumper_t *dumper;
73 volatile unsigned long tx_pkts;
74 volatile unsigned long err_pkts;
79 pcap_t *pcaps[RTE_PMD_RING_MAX_RX_RINGS];
84 pcap_dumper_t *dumpers[RTE_PMD_RING_MAX_TX_RINGS];
85 pcap_t *pcaps[RTE_PMD_RING_MAX_RX_RINGS];
88 struct pmd_internals {
89 unsigned nb_rx_queues;
90 unsigned nb_tx_queues;
94 struct pcap_rx_queue rx_queue[RTE_PMD_RING_MAX_RX_RINGS];
95 struct pcap_tx_queue tx_queue[RTE_PMD_RING_MAX_TX_RINGS];
98 const char *valid_arguments[] = {
100 ETH_PCAP_TX_PCAP_ARG,
101 ETH_PCAP_RX_IFACE_ARG,
102 ETH_PCAP_TX_IFACE_ARG,
107 static struct ether_addr eth_addr = { .addr_bytes = { 0, 0, 0, 0x1, 0x2, 0x3 } };
108 static const char *drivername = "Pcap PMD";
109 static struct rte_eth_link pmd_link = {
111 .link_duplex = ETH_LINK_FULL_DUPLEX,
117 eth_pcap_rx(void *queue,
118 struct rte_mbuf **bufs,
122 struct pcap_pkthdr header;
123 const u_char *packet;
124 struct rte_mbuf *mbuf;
125 struct pcap_rx_queue *pcap_q = queue;
126 struct rte_pktmbuf_pool_private *mbp_priv;
130 if (unlikely(pcap_q->pcap == NULL || nb_pkts == 0))
133 /* Reads the given number of packets from the pcap file one by one
134 * and copies the packet data into a newly allocated mbuf to return.
136 for (i = 0; i < nb_pkts; i++) {
137 /* Get the next PCAP packet */
138 packet = pcap_next(pcap_q->pcap, &header);
139 if (unlikely(packet == NULL))
142 mbuf = rte_pktmbuf_alloc(pcap_q->mb_pool);
143 if (unlikely(mbuf == NULL))
146 /* Now get the space available for data in the mbuf */
147 mbp_priv = rte_mempool_get_priv(pcap_q->mb_pool);
148 buf_size = (uint16_t) (mbp_priv->mbuf_data_room_size -
149 RTE_PKTMBUF_HEADROOM);
151 if (header.len <= buf_size) {
152 /* pcap packet will fit in the mbuf, go ahead and copy */
153 rte_memcpy(mbuf->pkt.data, packet, header.len);
154 mbuf->pkt.data_len = (uint16_t)header.len;
155 mbuf->pkt.pkt_len = mbuf->pkt.data_len;
159 /* pcap packet will not fit in the mbuf, so drop packet */
161 "PCAP packet %d bytes will not fit in mbuf (%d bytes)\n",
162 header.len, buf_size);
163 rte_pktmbuf_free(mbuf);
166 pcap_q->rx_pkts += num_rx;
171 calculate_timestamp(struct timeval *ts) {
173 struct timeval cur_time;
175 cycles = rte_get_timer_cycles() - start_cycles;
176 cur_time.tv_sec = cycles / hz;
177 cur_time.tv_usec = (cycles % hz) * 10e6 / hz;
178 timeradd(&start_time, &cur_time, ts);
182 * Callback to handle writing packets to a pcap file.
185 eth_pcap_tx_dumper(void *queue,
186 struct rte_mbuf **bufs,
190 struct rte_mbuf *mbuf;
191 struct pcap_tx_queue *dumper_q = queue;
193 struct pcap_pkthdr header;
195 if (dumper_q->dumper == NULL || nb_pkts == 0)
198 /* writes the nb_pkts packets to the previously opened pcap file dumper */
199 for (i = 0; i < nb_pkts; i++) {
201 calculate_timestamp(&header.ts);
202 header.len = mbuf->pkt.data_len;
203 header.caplen = header.len;
204 pcap_dump((u_char*) dumper_q->dumper, &header, mbuf->pkt.data);
205 rte_pktmbuf_free(mbuf);
210 * Since there's no place to hook a callback when the forwarding
211 * process stops and to make sure the pcap file is actually written,
212 * we flush the pcap dumper within each burst.
214 pcap_dump_flush(dumper_q->dumper);
215 dumper_q->tx_pkts += num_tx;
216 dumper_q->err_pkts += nb_pkts - num_tx;
221 * Callback to handle sending packets through a real NIC.
224 eth_pcap_tx(void *queue,
225 struct rte_mbuf **bufs,
230 struct rte_mbuf *mbuf;
231 struct pcap_tx_queue *tx_queue = queue;
234 if (unlikely(nb_pkts == 0 || tx_queue->pcap == NULL))
237 for (i = 0; i < nb_pkts; i++) {
239 ret = pcap_sendpacket(tx_queue->pcap, (u_char*) mbuf->pkt.data,
243 rte_pktmbuf_free(mbuf);
246 tx_queue->tx_pkts += num_tx;
247 tx_queue->err_pkts += nb_pkts - num_tx;
252 eth_dev_start(struct rte_eth_dev *dev)
254 dev->data->dev_link.link_status = 1;
259 * This function gets called when the current port gets stopped.
260 * Is the only place for us to close all the tx streams dumpers.
261 * If not called the dumpers will be flushed within each tx burst.
264 eth_dev_stop(struct rte_eth_dev *dev)
267 pcap_dumper_t *dumper;
269 struct pmd_internals *internals = dev->data->dev_private;
271 for (i = 0; i < internals->nb_tx_queues; i++) {
272 dumper = internals->tx_queue[i].dumper;
274 pcap_dump_close(dumper);
275 pcap = internals->tx_queue[i].pcap;
280 dev->data->dev_link.link_status = 0;
284 eth_dev_configure(struct rte_eth_dev *dev __rte_unused)
290 eth_dev_info(struct rte_eth_dev *dev,
291 struct rte_eth_dev_info *dev_info)
293 struct pmd_internals *internals = dev->data->dev_private;
294 dev_info->driver_name = drivername;
295 dev_info->if_index = internals->if_index;
296 dev_info->max_mac_addrs = 1;
297 dev_info->max_rx_pktlen = (uint32_t) -1;
298 dev_info->max_rx_queues = (uint16_t)internals->nb_rx_queues;
299 dev_info->max_tx_queues = (uint16_t)internals->nb_tx_queues;
300 dev_info->min_rx_bufsize = 0;
301 dev_info->pci_dev = NULL;
305 eth_stats_get(struct rte_eth_dev *dev,
306 struct rte_eth_stats *igb_stats)
309 unsigned long rx_total = 0, tx_total = 0, tx_err_total = 0;
310 const struct pmd_internals *internal = dev->data->dev_private;
312 memset(igb_stats, 0, sizeof(*igb_stats));
313 for (i = 0; i < RTE_ETHDEV_QUEUE_STAT_CNTRS && i < internal->nb_rx_queues;
315 igb_stats->q_ipackets[i] = internal->rx_queue[i].rx_pkts;
316 rx_total += igb_stats->q_ipackets[i];
319 for (i = 0; i < RTE_ETHDEV_QUEUE_STAT_CNTRS && i < internal->nb_tx_queues;
321 igb_stats->q_opackets[i] = internal->tx_queue[i].tx_pkts;
322 igb_stats->q_errors[i] = internal->tx_queue[i].err_pkts;
323 tx_total += igb_stats->q_opackets[i];
324 tx_err_total += igb_stats->q_errors[i];
327 igb_stats->ipackets = rx_total;
328 igb_stats->opackets = tx_total;
329 igb_stats->oerrors = tx_err_total;
333 eth_stats_reset(struct rte_eth_dev *dev)
336 struct pmd_internals *internal = dev->data->dev_private;
337 for (i = 0; i < internal->nb_rx_queues; i++)
338 internal->rx_queue[i].rx_pkts = 0;
339 for (i = 0; i < internal->nb_tx_queues; i++) {
340 internal->tx_queue[i].tx_pkts = 0;
341 internal->tx_queue[i].err_pkts = 0;
346 eth_dev_close(struct rte_eth_dev *dev __rte_unused)
351 eth_queue_release(void *q __rte_unused)
356 eth_link_update(struct rte_eth_dev *dev __rte_unused,
357 int wait_to_complete __rte_unused)
363 eth_rx_queue_setup(struct rte_eth_dev *dev,
364 uint16_t rx_queue_id,
365 uint16_t nb_rx_desc __rte_unused,
366 unsigned int socket_id __rte_unused,
367 const struct rte_eth_rxconf *rx_conf __rte_unused,
368 struct rte_mempool *mb_pool)
370 struct pmd_internals *internals = dev->data->dev_private;
371 struct pcap_rx_queue *pcap_q = &internals->rx_queue[rx_queue_id];
372 pcap_q->mb_pool = mb_pool;
373 dev->data->rx_queues[rx_queue_id] = pcap_q;
378 eth_tx_queue_setup(struct rte_eth_dev *dev,
379 uint16_t tx_queue_id,
380 uint16_t nb_tx_desc __rte_unused,
381 unsigned int socket_id __rte_unused,
382 const struct rte_eth_txconf *tx_conf __rte_unused)
385 struct pmd_internals *internals = dev->data->dev_private;
386 dev->data->tx_queues[tx_queue_id] = &internals->tx_queue[tx_queue_id];
390 static struct eth_dev_ops ops = {
391 .dev_start = eth_dev_start,
392 .dev_stop = eth_dev_stop,
393 .dev_close = eth_dev_close,
394 .dev_configure = eth_dev_configure,
395 .dev_infos_get = eth_dev_info,
396 .rx_queue_setup = eth_rx_queue_setup,
397 .tx_queue_setup = eth_tx_queue_setup,
398 .rx_queue_release = eth_queue_release,
399 .tx_queue_release = eth_queue_release,
400 .link_update = eth_link_update,
401 .stats_get = eth_stats_get,
402 .stats_reset = eth_stats_reset,
406 * Function handler that opens the pcap file for reading a stores a
407 * reference of it for use it later on.
410 open_rx_pcap(const char *key __rte_unused, const char *value, void *extra_args)
413 const char *pcap_filename = value;
414 struct rx_pcaps *pcaps = extra_args;
417 for (i = 0; i < pcaps->num_of_rx; i++) {
418 if ((rx_pcap = pcap_open_offline(pcap_filename, errbuf)) == NULL) {
419 RTE_LOG(ERR, PMD, "Couldn't open %s: %s\n", pcap_filename, errbuf);
422 pcaps->pcaps[i] = rx_pcap;
429 * Opens a pcap file for writing and stores a reference to it
430 * for use it later on.
433 open_tx_pcap(const char *key __rte_unused, const char *value, void *extra_args)
436 const char *pcap_filename = value;
437 struct tx_pcaps *dumpers = extra_args;
439 pcap_dumper_t *dumper;
441 for (i = 0; i < dumpers->num_of_tx; i++) {
443 * We need to create a dummy empty pcap_t to use it
444 * with pcap_dump_open(). We create big enough an Ethernet
447 if ((tx_pcap = pcap_open_dead(DLT_EN10MB, RTE_ETH_PCAP_SNAPSHOT_LEN))
449 RTE_LOG(ERR, PMD, "Couldn't create dead pcap\n");
453 /* The dumper is created using the previous pcap_t reference */
454 if ((dumper = pcap_dump_open(tx_pcap, pcap_filename)) == NULL) {
455 RTE_LOG(ERR, PMD, "Couldn't open %s for writing.\n", pcap_filename);
458 dumpers->dumpers[i] = dumper;
465 * pcap_open_live wrapper function
468 open_iface_live(const char *iface, pcap_t **pcap) {
469 *pcap = pcap_open_live(iface, RTE_ETH_PCAP_SNAPLEN,
470 RTE_ETH_PCAP_PROMISC, RTE_ETH_PCAP_TIMEOUT, errbuf);
473 RTE_LOG(ERR, PMD, "Couldn't open %s: %s\n", iface, errbuf);
480 * Opens an interface for reading and writing
483 open_rx_tx_iface(const char *key __rte_unused, const char *value, void *extra_args)
485 const char *iface = value;
486 pcap_t **pcap = extra_args;
488 if(open_iface_live(iface, pcap) < 0)
494 * Opens a NIC for reading packets from it
497 open_rx_iface(const char *key __rte_unused, const char *value, void *extra_args)
500 const char *iface = value;
501 struct rx_pcaps *pcaps = extra_args;
504 for (i = 0; i < pcaps->num_of_rx; i++) {
505 if(open_iface_live(iface, &pcap) < 0)
507 pcaps->pcaps[i] = pcap;
514 * Opens a NIC for writing packets to it
517 open_tx_iface(const char *key __rte_unused, const char *value, void *extra_args)
520 const char *iface = value;
521 struct tx_pcaps *pcaps = extra_args;
524 for (i = 0; i < pcaps->num_of_tx; i++) {
525 if(open_iface_live(iface, &pcap) < 0)
527 pcaps->pcaps[i] = pcap;
535 rte_pmd_init_internals(const unsigned nb_rx_queues,
536 const unsigned nb_tx_queues,
537 const unsigned numa_node,
538 struct pmd_internals **internals,
539 struct rte_eth_dev **eth_dev,
540 struct rte_kvargs *kvlist)
542 struct rte_eth_dev_data *data = NULL;
543 struct rte_pci_device *pci_dev = NULL;
545 struct rte_kvargs_pair *pair = NULL;
547 for (k_idx = 0; k_idx < kvlist->count; k_idx++) {
548 pair = &kvlist->pairs[k_idx];
549 if (strstr(pair->key, ETH_PCAP_IFACE_ARG) != NULL)
554 "Creating pcap-backed ethdev on numa socket %u\n", numa_node);
556 /* now do all data allocation - for eth_dev structure, dummy pci driver
557 * and internal (private) data
559 data = rte_zmalloc_socket(NULL, sizeof(*data), 0, numa_node);
563 pci_dev = rte_zmalloc_socket(NULL, sizeof(*pci_dev), 0, numa_node);
567 *internals = rte_zmalloc_socket(NULL, sizeof(**internals), 0, numa_node);
568 if (*internals == NULL)
571 /* reserve an ethdev entry */
572 *eth_dev = rte_eth_dev_allocate();
573 if (*eth_dev == NULL)
576 /* now put it all together
577 * - store queue data in internals,
578 * - store numa_node info in pci_driver
579 * - point eth_dev_data to internals and pci_driver
580 * - and point eth_dev structure to new eth_dev_data structure
582 /* NOTE: we'll replace the data element, of originally allocated eth_dev
583 * so the rings are local per-process */
585 (*internals)->nb_rx_queues = nb_rx_queues;
586 (*internals)->nb_tx_queues = nb_tx_queues;
589 (*internals)->if_index = 0;
591 (*internals)->if_index = if_nametoindex(pair->value);
593 pci_dev->numa_node = numa_node;
595 data->dev_private = *internals;
596 data->port_id = (*eth_dev)->data->port_id;
597 data->nb_rx_queues = (uint16_t)nb_rx_queues;
598 data->nb_tx_queues = (uint16_t)nb_tx_queues;
599 data->dev_link = pmd_link;
600 data->mac_addrs = ð_addr;
602 (*eth_dev)->data = data;
603 (*eth_dev)->dev_ops = &ops;
604 (*eth_dev)->pci_dev = pci_dev;
613 rte_free(*internals);
618 rte_eth_from_pcaps_n_dumpers(pcap_t * const rx_queues[],
619 const unsigned nb_rx_queues,
620 pcap_dumper_t * const tx_queues[],
621 const unsigned nb_tx_queues,
622 const unsigned numa_node,
623 struct rte_kvargs *kvlist)
625 struct pmd_internals *internals = NULL;
626 struct rte_eth_dev *eth_dev = NULL;
629 /* do some parameter checking */
630 if (rx_queues == NULL && nb_rx_queues > 0)
632 if (tx_queues == NULL && nb_tx_queues > 0)
635 if (rte_pmd_init_internals(nb_rx_queues, nb_tx_queues, numa_node,
636 &internals, ð_dev, kvlist) < 0)
639 for (i = 0; i < nb_rx_queues; i++) {
640 internals->rx_queue->pcap = rx_queues[i];
642 for (i = 0; i < nb_tx_queues; i++) {
643 internals->tx_queue->dumper = tx_queues[i];
646 eth_dev->rx_pkt_burst = eth_pcap_rx;
647 eth_dev->tx_pkt_burst = eth_pcap_tx_dumper;
653 rte_eth_from_pcaps(pcap_t * const rx_queues[],
654 const unsigned nb_rx_queues,
655 pcap_t * const tx_queues[],
656 const unsigned nb_tx_queues,
657 const unsigned numa_node,
658 struct rte_kvargs *kvlist)
660 struct pmd_internals *internals = NULL;
661 struct rte_eth_dev *eth_dev = NULL;
664 /* do some parameter checking */
665 if (rx_queues == NULL && nb_rx_queues > 0)
667 if (tx_queues == NULL && nb_tx_queues > 0)
670 if (rte_pmd_init_internals(nb_rx_queues, nb_tx_queues, numa_node,
671 &internals, ð_dev, kvlist) < 0)
674 for (i = 0; i < nb_rx_queues; i++) {
675 internals->rx_queue->pcap = rx_queues[i];
677 for (i = 0; i < nb_tx_queues; i++) {
678 internals->tx_queue->pcap = tx_queues[i];
681 eth_dev->rx_pkt_burst = eth_pcap_rx;
682 eth_dev->tx_pkt_burst = eth_pcap_tx;
689 rte_pmd_pcap_init(const char *name, const char *params)
691 unsigned numa_node, using_dumpers = 0;
693 struct rte_kvargs *kvlist;
694 struct rx_pcaps pcaps;
695 struct tx_pcaps dumpers;
697 RTE_LOG(INFO, PMD, "Initializing pmd_pcap for %s\n", name);
699 numa_node = rte_socket_id();
701 gettimeofday(&start_time, NULL);
702 start_cycles = rte_get_timer_cycles();
703 hz = rte_get_timer_hz();
705 kvlist = rte_kvargs_parse(params, valid_arguments);
710 * If iface argument is passed we open the NICs and use them for
713 if (rte_kvargs_count(kvlist, ETH_PCAP_IFACE_ARG) == 1) {
715 ret = rte_kvargs_process(kvlist, ETH_PCAP_IFACE_ARG,
716 &open_rx_tx_iface, &pcaps.pcaps[0]);
720 return rte_eth_from_pcaps(pcaps.pcaps, 1, pcaps.pcaps, 1,
725 * We check whether we want to open a RX stream from a real NIC or a
728 if ((pcaps.num_of_rx = rte_kvargs_count(kvlist, ETH_PCAP_RX_PCAP_ARG))) {
729 ret = rte_kvargs_process(kvlist, ETH_PCAP_RX_PCAP_ARG,
730 &open_rx_pcap, &pcaps);
732 pcaps.num_of_rx = rte_kvargs_count(kvlist,
733 ETH_PCAP_RX_IFACE_ARG);
734 ret = rte_kvargs_process(kvlist, ETH_PCAP_RX_IFACE_ARG,
735 &open_rx_iface, &pcaps);
742 * We check whether we want to open a TX stream to a real NIC or a
745 if ((dumpers.num_of_tx = rte_kvargs_count(kvlist,
746 ETH_PCAP_TX_PCAP_ARG))) {
747 ret = rte_kvargs_process(kvlist, ETH_PCAP_TX_PCAP_ARG,
748 &open_tx_pcap, &dumpers);
751 dumpers.num_of_tx = rte_kvargs_count(kvlist,
752 ETH_PCAP_TX_IFACE_ARG);
753 ret = rte_kvargs_process(kvlist, ETH_PCAP_TX_IFACE_ARG,
754 &open_tx_iface, &dumpers);
761 return rte_eth_from_pcaps_n_dumpers(pcaps.pcaps, pcaps.num_of_rx,
762 dumpers.dumpers, dumpers.num_of_tx, numa_node, kvlist);
764 return rte_eth_from_pcaps(pcaps.pcaps, pcaps.num_of_rx, dumpers.pcaps,
765 dumpers.num_of_tx, numa_node, kvlist);