4 * Copyright(c) 2016 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
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.
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.
34 #include <rte_atomic.h>
35 #include <rte_common.h>
37 #include <rte_ethdev.h>
38 #include <rte_malloc.h>
40 #include <rte_kvargs.h>
43 #include <sys/types.h>
45 #include <sys/socket.h>
46 #include <sys/ioctl.h>
52 #include <arpa/inet.h>
54 #include <linux/if_tun.h>
55 #include <linux/if_ether.h>
58 /* Linux based path to the TUN device */
59 #define TUN_TAP_DEV_PATH "/dev/net/tun"
60 #define DEFAULT_TAP_NAME "dtap"
62 #define ETH_TAP_IFACE_ARG "iface"
63 #define ETH_TAP_SPEED_ARG "speed"
65 #ifdef IFF_MULTI_QUEUE
66 #define RTE_PMD_TAP_MAX_QUEUES 16
68 #define RTE_PMD_TAP_MAX_QUEUES 1
71 static struct rte_vdev_driver pmd_tap_drv;
73 static const char *valid_arguments[] = {
81 static volatile uint32_t tap_trigger; /* Rx trigger */
83 static struct rte_eth_link pmd_link = {
84 .link_speed = ETH_SPEED_NUM_10G,
85 .link_duplex = ETH_LINK_FULL_DUPLEX,
86 .link_status = ETH_LINK_DOWN,
87 .link_autoneg = ETH_LINK_SPEED_AUTONEG
91 uint64_t opackets; /* Number of output packets */
92 uint64_t ipackets; /* Number of input packets */
93 uint64_t obytes; /* Number of bytes on output */
94 uint64_t ibytes; /* Number of bytes on input */
95 uint64_t errs; /* Number of error packets */
99 struct rte_mempool *mp; /* Mempool for RX packets */
100 uint32_t trigger_seen; /* Last seen Rx trigger value */
101 uint16_t in_port; /* Port ID */
104 struct pkt_stats stats; /* Stats for this RX queue */
109 struct pkt_stats stats; /* Stats for this TX queue */
112 struct pmd_internals {
113 char name[RTE_ETH_NAME_MAX_LEN]; /* Internal Tap device name */
114 uint16_t nb_queues; /* Number of queues supported */
115 struct ether_addr eth_addr; /* Mac address of the device port */
117 int if_index; /* IF_INDEX for the port */
118 int ioctl_sock; /* socket for ioctl calls */
120 struct rx_queue rxq[RTE_PMD_TAP_MAX_QUEUES]; /* List of RX queues */
121 struct tx_queue txq[RTE_PMD_TAP_MAX_QUEUES]; /* List of TX queues */
125 tap_trigger_cb(int sig __rte_unused)
127 /* Valid trigger values are nonzero */
128 tap_trigger = (tap_trigger + 1) | 0x80000000;
132 tap_ioctl(struct pmd_internals *pmd, unsigned long request,
133 struct ifreq *ifr, int set);
135 /* Tun/Tap allocation routine
137 * name is the number of the interface to use, unless NULL to take the host
141 tun_alloc(struct pmd_internals *pmd, uint16_t qid)
144 #ifdef IFF_MULTI_QUEUE
145 unsigned int features;
149 memset(&ifr, 0, sizeof(struct ifreq));
151 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
152 snprintf(ifr.ifr_name, IFNAMSIZ, "%s", pmd->name);
154 RTE_LOG(DEBUG, PMD, "ifr_name '%s'\n", ifr.ifr_name);
156 fd = open(TUN_TAP_DEV_PATH, O_RDWR);
158 RTE_LOG(ERR, PMD, "Unable to create TAP interface");
162 #ifdef IFF_MULTI_QUEUE
163 /* Grab the TUN features to verify we can work multi-queue */
164 if (ioctl(fd, TUNGETFEATURES, &features) < 0) {
165 RTE_LOG(ERR, PMD, "TAP unable to get TUN/TAP features\n");
168 RTE_LOG(DEBUG, PMD, " TAP Features %08x\n", features);
170 if (features & IFF_MULTI_QUEUE) {
171 RTE_LOG(DEBUG, PMD, " Multi-queue support for %d queues\n",
172 RTE_PMD_TAP_MAX_QUEUES);
173 ifr.ifr_flags |= IFF_MULTI_QUEUE;
177 ifr.ifr_flags |= IFF_ONE_QUEUE;
178 RTE_LOG(DEBUG, PMD, " Single queue only support\n");
181 /* Set the TUN/TAP configuration and set the name if needed */
182 if (ioctl(fd, TUNSETIFF, (void *)&ifr) < 0) {
183 RTE_LOG(WARNING, PMD,
184 "Unable to set TUNSETIFF for %s\n",
190 /* Always set the file descriptor to non-blocking */
191 if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0) {
192 RTE_LOG(WARNING, PMD,
193 "Unable to set %s to nonblocking\n",
195 perror("F_SETFL, NONBLOCK");
199 /* Set up trigger to optimize empty Rx bursts */
203 int flags = fcntl(fd, F_GETFL);
205 if (flags == -1 || sigaction(SIGIO, NULL, &sa) == -1)
207 if (sa.sa_handler != tap_trigger_cb) {
209 * Make sure SIGIO is not already taken. This is done
210 * as late as possible to leave the application a
211 * chance to set up its own signal handler first.
213 if (sa.sa_handler != SIG_IGN &&
214 sa.sa_handler != SIG_DFL) {
218 sa = (struct sigaction){
219 .sa_flags = SA_RESTART,
220 .sa_handler = tap_trigger_cb,
222 if (sigaction(SIGIO, &sa, NULL) == -1)
225 /* Enable SIGIO on file descriptor */
226 fcntl(fd, F_SETFL, flags | O_ASYNC);
227 fcntl(fd, F_SETOWN, getpid());
230 /* Disable trigger globally in case of error */
232 RTE_LOG(WARNING, PMD, "Rx trigger disabled: %s\n",
239 if (tap_ioctl(pmd, SIOCGIFHWADDR, &ifr, 0) < 0)
241 rte_memcpy(&pmd->eth_addr, ifr.ifr_hwaddr.sa_data,
253 /* Callback to handle the rx burst of packets to the correct interface and
254 * file descriptor(s) in a multi-queue setup.
257 pmd_rx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
260 struct rte_mbuf *mbuf;
261 struct rx_queue *rxq = queue;
263 unsigned long num_rx_bytes = 0;
264 uint32_t trigger = tap_trigger;
266 if (trigger == rxq->trigger_seen)
269 rxq->trigger_seen = trigger;
270 rte_compiler_barrier();
271 for (num_rx = 0; num_rx < nb_pkts; ) {
272 /* allocate the next mbuf */
273 mbuf = rte_pktmbuf_alloc(rxq->mp);
274 if (unlikely(!mbuf)) {
275 RTE_LOG(WARNING, PMD, "TAP unable to allocate mbuf\n");
279 len = read(rxq->fd, rte_pktmbuf_mtod(mbuf, char *),
280 rte_pktmbuf_tailroom(mbuf));
282 rte_pktmbuf_free(mbuf);
286 mbuf->data_len = len;
288 mbuf->port = rxq->in_port;
289 mbuf->packet_type = rte_net_get_ptype(mbuf, NULL,
292 /* account for the receive frame */
293 bufs[num_rx++] = mbuf;
294 num_rx_bytes += mbuf->pkt_len;
296 rxq->stats.ipackets += num_rx;
297 rxq->stats.ibytes += num_rx_bytes;
302 /* Callback to handle sending packets from the tap interface
305 pmd_tx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
307 struct rte_mbuf *mbuf;
308 struct tx_queue *txq = queue;
310 unsigned long num_tx_bytes = 0;
313 if (unlikely(nb_pkts == 0))
316 for (i = 0; i < nb_pkts; i++) {
317 /* copy the tx frame data */
320 rte_pktmbuf_mtod(mbuf, void *),
321 rte_pktmbuf_pkt_len(mbuf));
326 num_tx_bytes += mbuf->pkt_len;
327 rte_pktmbuf_free(mbuf);
330 txq->stats.opackets += num_tx;
331 txq->stats.errs += nb_pkts - num_tx;
332 txq->stats.obytes += num_tx_bytes;
338 tap_ioctl(struct pmd_internals *pmd, unsigned long request,
339 struct ifreq *ifr, int set)
341 short req_flags = ifr->ifr_flags;
343 snprintf(ifr->ifr_name, IFNAMSIZ, "%s", pmd->name);
346 /* fetch current flags to leave other flags untouched */
347 if (ioctl(pmd->ioctl_sock, SIOCGIFFLAGS, ifr) < 0)
350 ifr->ifr_flags |= req_flags;
352 ifr->ifr_flags &= ~req_flags;
359 RTE_LOG(WARNING, PMD, "%s: ioctl() called with wrong arg\n",
363 if (ioctl(pmd->ioctl_sock, request, ifr) < 0)
368 RTE_LOG(ERR, PMD, "%s: ioctl(%lu) failed with error: %s\n",
369 ifr->ifr_name, request, strerror(errno));
374 tap_link_set_down(struct rte_eth_dev *dev)
376 struct pmd_internals *pmd = dev->data->dev_private;
377 struct ifreq ifr = { .ifr_flags = IFF_UP };
379 dev->data->dev_link.link_status = ETH_LINK_DOWN;
380 return tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 0);
384 tap_link_set_up(struct rte_eth_dev *dev)
386 struct pmd_internals *pmd = dev->data->dev_private;
387 struct ifreq ifr = { .ifr_flags = IFF_UP };
389 dev->data->dev_link.link_status = ETH_LINK_UP;
390 return tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 1);
394 tap_dev_start(struct rte_eth_dev *dev)
396 return tap_link_set_up(dev);
399 /* This function gets called when the current port gets stopped.
402 tap_dev_stop(struct rte_eth_dev *dev)
404 tap_link_set_down(dev);
408 tap_dev_configure(struct rte_eth_dev *dev __rte_unused)
414 tap_dev_speed_capa(void)
416 uint32_t speed = pmd_link.link_speed;
419 if (speed >= ETH_SPEED_NUM_10M)
420 capa |= ETH_LINK_SPEED_10M;
421 if (speed >= ETH_SPEED_NUM_100M)
422 capa |= ETH_LINK_SPEED_100M;
423 if (speed >= ETH_SPEED_NUM_1G)
424 capa |= ETH_LINK_SPEED_1G;
425 if (speed >= ETH_SPEED_NUM_5G)
426 capa |= ETH_LINK_SPEED_2_5G;
427 if (speed >= ETH_SPEED_NUM_5G)
428 capa |= ETH_LINK_SPEED_5G;
429 if (speed >= ETH_SPEED_NUM_10G)
430 capa |= ETH_LINK_SPEED_10G;
431 if (speed >= ETH_SPEED_NUM_20G)
432 capa |= ETH_LINK_SPEED_20G;
433 if (speed >= ETH_SPEED_NUM_25G)
434 capa |= ETH_LINK_SPEED_25G;
435 if (speed >= ETH_SPEED_NUM_40G)
436 capa |= ETH_LINK_SPEED_40G;
437 if (speed >= ETH_SPEED_NUM_50G)
438 capa |= ETH_LINK_SPEED_50G;
439 if (speed >= ETH_SPEED_NUM_56G)
440 capa |= ETH_LINK_SPEED_56G;
441 if (speed >= ETH_SPEED_NUM_100G)
442 capa |= ETH_LINK_SPEED_100G;
448 tap_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
450 struct pmd_internals *internals = dev->data->dev_private;
452 dev_info->if_index = internals->if_index;
453 dev_info->max_mac_addrs = 1;
454 dev_info->max_rx_pktlen = (uint32_t)ETHER_MAX_VLAN_FRAME_LEN;
455 dev_info->max_rx_queues = internals->nb_queues;
456 dev_info->max_tx_queues = internals->nb_queues;
457 dev_info->min_rx_bufsize = 0;
458 dev_info->pci_dev = NULL;
459 dev_info->speed_capa = tap_dev_speed_capa();
463 tap_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *tap_stats)
465 unsigned int i, imax;
466 unsigned long rx_total = 0, tx_total = 0, tx_err_total = 0;
467 unsigned long rx_bytes_total = 0, tx_bytes_total = 0;
468 const struct pmd_internals *pmd = dev->data->dev_private;
470 imax = (pmd->nb_queues < RTE_ETHDEV_QUEUE_STAT_CNTRS) ?
471 pmd->nb_queues : RTE_ETHDEV_QUEUE_STAT_CNTRS;
473 for (i = 0; i < imax; i++) {
474 tap_stats->q_ipackets[i] = pmd->rxq[i].stats.ipackets;
475 tap_stats->q_ibytes[i] = pmd->rxq[i].stats.ibytes;
476 rx_total += tap_stats->q_ipackets[i];
477 rx_bytes_total += tap_stats->q_ibytes[i];
479 tap_stats->q_opackets[i] = pmd->txq[i].stats.opackets;
480 tap_stats->q_errors[i] = pmd->txq[i].stats.errs;
481 tap_stats->q_obytes[i] = pmd->txq[i].stats.obytes;
482 tx_total += tap_stats->q_opackets[i];
483 tx_err_total += tap_stats->q_errors[i];
484 tx_bytes_total += tap_stats->q_obytes[i];
487 tap_stats->ipackets = rx_total;
488 tap_stats->ibytes = rx_bytes_total;
489 tap_stats->opackets = tx_total;
490 tap_stats->oerrors = tx_err_total;
491 tap_stats->obytes = tx_bytes_total;
495 tap_stats_reset(struct rte_eth_dev *dev)
498 struct pmd_internals *pmd = dev->data->dev_private;
500 for (i = 0; i < pmd->nb_queues; i++) {
501 pmd->rxq[i].stats.ipackets = 0;
502 pmd->rxq[i].stats.ibytes = 0;
504 pmd->txq[i].stats.opackets = 0;
505 pmd->txq[i].stats.errs = 0;
506 pmd->txq[i].stats.obytes = 0;
511 tap_dev_close(struct rte_eth_dev *dev __rte_unused)
514 struct pmd_internals *internals = dev->data->dev_private;
516 tap_link_set_down(dev);
518 for (i = 0; i < internals->nb_queues; i++) {
519 if (internals->rxq[i].fd != -1)
520 close(internals->rxq[i].fd);
521 internals->rxq[i].fd = -1;
522 internals->txq[i].fd = -1;
527 tap_rx_queue_release(void *queue)
529 struct rx_queue *rxq = queue;
531 if (rxq && (rxq->fd > 0)) {
538 tap_tx_queue_release(void *queue)
540 struct tx_queue *txq = queue;
542 if (txq && (txq->fd > 0)) {
549 tap_link_update(struct rte_eth_dev *dev __rte_unused,
550 int wait_to_complete __rte_unused)
556 tap_promisc_enable(struct rte_eth_dev *dev)
558 struct pmd_internals *pmd = dev->data->dev_private;
559 struct ifreq ifr = { .ifr_flags = IFF_PROMISC };
561 dev->data->promiscuous = 1;
562 tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 1);
566 tap_promisc_disable(struct rte_eth_dev *dev)
568 struct pmd_internals *pmd = dev->data->dev_private;
569 struct ifreq ifr = { .ifr_flags = IFF_PROMISC };
571 dev->data->promiscuous = 0;
572 tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 0);
576 tap_allmulti_enable(struct rte_eth_dev *dev)
578 struct pmd_internals *pmd = dev->data->dev_private;
579 struct ifreq ifr = { .ifr_flags = IFF_ALLMULTI };
581 dev->data->all_multicast = 1;
582 tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 1);
586 tap_allmulti_disable(struct rte_eth_dev *dev)
588 struct pmd_internals *pmd = dev->data->dev_private;
589 struct ifreq ifr = { .ifr_flags = IFF_ALLMULTI };
591 dev->data->all_multicast = 0;
592 tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 0);
597 tap_mac_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
599 struct pmd_internals *pmd = dev->data->dev_private;
602 if (is_zero_ether_addr(mac_addr)) {
603 RTE_LOG(ERR, PMD, "%s: can't set an empty MAC address\n",
608 ifr.ifr_hwaddr.sa_family = AF_LOCAL;
609 rte_memcpy(ifr.ifr_hwaddr.sa_data, mac_addr, ETHER_ADDR_LEN);
610 if (tap_ioctl(pmd, SIOCSIFHWADDR, &ifr, 1) < 0)
612 rte_memcpy(&pmd->eth_addr, mac_addr, ETHER_ADDR_LEN);
616 tap_setup_queue(struct rte_eth_dev *dev,
617 struct pmd_internals *internals,
620 struct pmd_internals *pmd = dev->data->dev_private;
621 struct rx_queue *rx = &internals->rxq[qid];
622 struct tx_queue *tx = &internals->txq[qid];
629 RTE_LOG(INFO, PMD, "Add queue to TAP %s for qid %d\n",
631 fd = tun_alloc(pmd, qid);
633 RTE_LOG(ERR, PMD, "tun_alloc(%s, %d) failed\n",
640 ifr.ifr_mtu = dev->data->mtu;
641 if (tap_ioctl(pmd, SIOCSIFMTU, &ifr, 1) < 0) {
656 rx_setup_queue(struct rte_eth_dev *dev,
657 struct pmd_internals *internals,
660 dev->data->rx_queues[qid] = &internals->rxq[qid];
662 return tap_setup_queue(dev, internals, qid);
666 tx_setup_queue(struct rte_eth_dev *dev,
667 struct pmd_internals *internals,
670 dev->data->tx_queues[qid] = &internals->txq[qid];
672 return tap_setup_queue(dev, internals, qid);
676 tap_rx_queue_setup(struct rte_eth_dev *dev,
677 uint16_t rx_queue_id,
678 uint16_t nb_rx_desc __rte_unused,
679 unsigned int socket_id __rte_unused,
680 const struct rte_eth_rxconf *rx_conf __rte_unused,
681 struct rte_mempool *mp)
683 struct pmd_internals *internals = dev->data->dev_private;
687 if ((rx_queue_id >= internals->nb_queues) || !mp) {
688 RTE_LOG(WARNING, PMD,
689 "nb_queues %d too small or mempool NULL\n",
690 internals->nb_queues);
694 internals->rxq[rx_queue_id].mp = mp;
695 internals->rxq[rx_queue_id].trigger_seen = 1; /* force initial burst */
696 internals->rxq[rx_queue_id].in_port = dev->data->port_id;
698 /* Now get the space available for data in the mbuf */
699 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(mp) -
700 RTE_PKTMBUF_HEADROOM);
702 if (buf_size < ETH_FRAME_LEN) {
703 RTE_LOG(WARNING, PMD,
704 "%s: %d bytes will not fit in mbuf (%d bytes)\n",
705 dev->data->name, ETH_FRAME_LEN, buf_size);
709 fd = rx_setup_queue(dev, internals, rx_queue_id);
713 RTE_LOG(DEBUG, PMD, " RX TAP device name %s, qid %d on fd %d\n",
714 internals->name, rx_queue_id, internals->rxq[rx_queue_id].fd);
720 tap_tx_queue_setup(struct rte_eth_dev *dev,
721 uint16_t tx_queue_id,
722 uint16_t nb_tx_desc __rte_unused,
723 unsigned int socket_id __rte_unused,
724 const struct rte_eth_txconf *tx_conf __rte_unused)
726 struct pmd_internals *internals = dev->data->dev_private;
729 if (tx_queue_id >= internals->nb_queues)
732 ret = tx_setup_queue(dev, internals, tx_queue_id);
736 RTE_LOG(DEBUG, PMD, " TX TAP device name %s, qid %d on fd %d\n",
737 internals->name, tx_queue_id, internals->txq[tx_queue_id].fd);
743 tap_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
745 struct pmd_internals *pmd = dev->data->dev_private;
746 struct ifreq ifr = { .ifr_mtu = mtu };
749 err = tap_ioctl(pmd, SIOCSIFMTU, &ifr, 1);
751 dev->data->mtu = mtu;
757 tap_set_mc_addr_list(struct rte_eth_dev *dev __rte_unused,
758 struct ether_addr *mc_addr_set __rte_unused,
759 uint32_t nb_mc_addr __rte_unused)
762 * Nothing to do actually: the tap has no filtering whatsoever, every
763 * packet is received.
768 static const uint32_t*
769 tap_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
771 static const uint32_t ptypes[] = {
772 RTE_PTYPE_INNER_L2_ETHER,
773 RTE_PTYPE_INNER_L2_ETHER_VLAN,
774 RTE_PTYPE_INNER_L2_ETHER_QINQ,
775 RTE_PTYPE_INNER_L3_IPV4,
776 RTE_PTYPE_INNER_L3_IPV4_EXT,
777 RTE_PTYPE_INNER_L3_IPV6,
778 RTE_PTYPE_INNER_L3_IPV6_EXT,
779 RTE_PTYPE_INNER_L4_FRAG,
780 RTE_PTYPE_INNER_L4_UDP,
781 RTE_PTYPE_INNER_L4_TCP,
782 RTE_PTYPE_INNER_L4_SCTP,
784 RTE_PTYPE_L2_ETHER_VLAN,
785 RTE_PTYPE_L2_ETHER_QINQ,
787 RTE_PTYPE_L3_IPV4_EXT,
788 RTE_PTYPE_L3_IPV6_EXT,
799 static const struct eth_dev_ops ops = {
800 .dev_start = tap_dev_start,
801 .dev_stop = tap_dev_stop,
802 .dev_close = tap_dev_close,
803 .dev_configure = tap_dev_configure,
804 .dev_infos_get = tap_dev_info,
805 .rx_queue_setup = tap_rx_queue_setup,
806 .tx_queue_setup = tap_tx_queue_setup,
807 .rx_queue_release = tap_rx_queue_release,
808 .tx_queue_release = tap_tx_queue_release,
809 .link_update = tap_link_update,
810 .dev_set_link_up = tap_link_set_up,
811 .dev_set_link_down = tap_link_set_down,
812 .promiscuous_enable = tap_promisc_enable,
813 .promiscuous_disable = tap_promisc_disable,
814 .allmulticast_enable = tap_allmulti_enable,
815 .allmulticast_disable = tap_allmulti_disable,
816 .mac_addr_set = tap_mac_set,
817 .mtu_set = tap_mtu_set,
818 .set_mc_addr_list = tap_set_mc_addr_list,
819 .stats_get = tap_stats_get,
820 .stats_reset = tap_stats_reset,
821 .dev_supported_ptypes_get = tap_dev_supported_ptypes_get,
825 eth_dev_tap_create(const char *name, char *tap_name)
827 int numa_node = rte_socket_id();
828 struct rte_eth_dev *dev = NULL;
829 struct pmd_internals *pmd = NULL;
830 struct rte_eth_dev_data *data = NULL;
833 RTE_LOG(DEBUG, PMD, " TAP device on numa %u\n", rte_socket_id());
835 data = rte_zmalloc_socket(tap_name, sizeof(*data), 0, numa_node);
837 RTE_LOG(ERR, PMD, "TAP Failed to allocate data\n");
841 pmd = rte_zmalloc_socket(tap_name, sizeof(*pmd), 0, numa_node);
843 RTE_LOG(ERR, PMD, "TAP Unable to allocate internal struct\n");
847 /* name in allocation and data->name must be consistent */
848 snprintf(data->name, sizeof(data->name), "%s", name);
849 dev = rte_eth_dev_allocate(name);
851 RTE_LOG(ERR, PMD, "TAP Unable to allocate device struct\n");
855 snprintf(pmd->name, sizeof(pmd->name), "%s", tap_name);
857 pmd->nb_queues = RTE_PMD_TAP_MAX_QUEUES;
859 pmd->ioctl_sock = socket(AF_INET, SOCK_DGRAM, 0);
860 if (pmd->ioctl_sock == -1) {
862 "TAP Unable to get a socket for management: %s\n",
867 /* Setup some default values */
868 data->dev_private = pmd;
869 data->port_id = dev->data->port_id;
870 data->mtu = dev->data->mtu;
871 data->dev_flags = RTE_ETH_DEV_DETACHABLE;
872 data->kdrv = RTE_KDRV_NONE;
873 data->drv_name = pmd_tap_drv.driver.name;
874 data->numa_node = numa_node;
876 data->dev_link = pmd_link;
877 data->mac_addrs = &pmd->eth_addr;
878 data->nb_rx_queues = pmd->nb_queues;
879 data->nb_tx_queues = pmd->nb_queues;
884 dev->rx_pkt_burst = pmd_rx_burst;
885 dev->tx_pkt_burst = pmd_tx_burst;
887 /* Presetup the fds to -1 as being not valid */
888 for (i = 0; i < RTE_PMD_TAP_MAX_QUEUES; i++) {
896 RTE_LOG(DEBUG, PMD, "TAP Unable to initialize %s\n", name);
901 rte_eth_dev_release_port(dev);
907 set_interface_name(const char *key __rte_unused,
911 char *name = (char *)extra_args;
914 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s", value);
916 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s%d",
917 DEFAULT_TAP_NAME, (tap_unit - 1));
923 set_interface_speed(const char *key __rte_unused,
927 *(int *)extra_args = (value) ? atoi(value) : ETH_SPEED_NUM_10G;
932 /* Open a TAP interface device.
935 rte_pmd_tap_probe(const char *name, const char *params)
938 struct rte_kvargs *kvlist = NULL;
940 char tap_name[RTE_ETH_NAME_MAX_LEN];
942 speed = ETH_SPEED_NUM_10G;
943 snprintf(tap_name, sizeof(tap_name), "%s%d",
944 DEFAULT_TAP_NAME, tap_unit++);
946 if (params && (params[0] != '\0')) {
947 RTE_LOG(DEBUG, PMD, "paramaters (%s)\n", params);
949 kvlist = rte_kvargs_parse(params, valid_arguments);
951 if (rte_kvargs_count(kvlist, ETH_TAP_SPEED_ARG) == 1) {
952 ret = rte_kvargs_process(kvlist,
954 &set_interface_speed,
960 if (rte_kvargs_count(kvlist, ETH_TAP_IFACE_ARG) == 1) {
961 ret = rte_kvargs_process(kvlist,
970 pmd_link.link_speed = speed;
972 RTE_LOG(NOTICE, PMD, "Initializing pmd_tap for %s as %s\n",
975 ret = eth_dev_tap_create(name, tap_name);
979 RTE_LOG(ERR, PMD, "Failed to create pmd for %s as %s\n",
981 tap_unit--; /* Restore the unit number */
983 rte_kvargs_free(kvlist);
988 /* detach a TAP device.
991 rte_pmd_tap_remove(const char *name)
993 struct rte_eth_dev *eth_dev = NULL;
994 struct pmd_internals *internals;
997 RTE_LOG(DEBUG, PMD, "Closing TUN/TAP Ethernet device on numa %u\n",
1000 /* find the ethdev entry */
1001 eth_dev = rte_eth_dev_allocated(name);
1005 internals = eth_dev->data->dev_private;
1006 for (i = 0; i < internals->nb_queues; i++)
1007 if (internals->rxq[i].fd != -1)
1008 close(internals->rxq[i].fd);
1010 close(internals->ioctl_sock);
1011 rte_free(eth_dev->data->dev_private);
1012 rte_free(eth_dev->data);
1014 rte_eth_dev_release_port(eth_dev);
1019 static struct rte_vdev_driver pmd_tap_drv = {
1020 .probe = rte_pmd_tap_probe,
1021 .remove = rte_pmd_tap_remove,
1023 RTE_PMD_REGISTER_VDEV(net_tap, pmd_tap_drv);
1024 RTE_PMD_REGISTER_ALIAS(net_tap, eth_tap);
1025 RTE_PMD_REGISTER_PARAM_STRING(net_tap, "iface=<string>,speed=N");