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.
35 #include <rte_ethdev.h>
36 #include <rte_malloc.h>
38 #include <rte_kvargs.h>
40 #include <sys/types.h>
42 #include <sys/socket.h>
43 #include <sys/ioctl.h>
47 #include <arpa/inet.h>
49 #include <linux/if_tun.h>
50 #include <linux/if_ether.h>
53 /* Linux based path to the TUN device */
54 #define TUN_TAP_DEV_PATH "/dev/net/tun"
55 #define DEFAULT_TAP_NAME "dtap"
57 #define ETH_TAP_IFACE_ARG "iface"
58 #define ETH_TAP_SPEED_ARG "speed"
60 #define RTE_PMD_TAP_MAX_QUEUES 16
62 static struct rte_vdev_driver pmd_tap_drv;
64 static const char *valid_arguments[] = {
72 static struct rte_eth_link pmd_link = {
73 .link_speed = ETH_SPEED_NUM_10G,
74 .link_duplex = ETH_LINK_FULL_DUPLEX,
75 .link_status = ETH_LINK_DOWN,
76 .link_autoneg = ETH_LINK_SPEED_AUTONEG
80 uint64_t opackets; /* Number of output packets */
81 uint64_t ipackets; /* Number of input packets */
82 uint64_t obytes; /* Number of bytes on output */
83 uint64_t ibytes; /* Number of bytes on input */
84 uint64_t errs; /* Number of error packets */
88 struct rte_mempool *mp; /* Mempool for RX packets */
89 uint16_t in_port; /* Port ID */
92 struct pkt_stats stats; /* Stats for this RX queue */
97 struct pkt_stats stats; /* Stats for this TX queue */
100 struct pmd_internals {
101 char name[RTE_ETH_NAME_MAX_LEN]; /* Internal Tap device name */
102 uint16_t nb_queues; /* Number of queues supported */
103 struct ether_addr eth_addr; /* Mac address of the device port */
105 int if_index; /* IF_INDEX for the port */
106 int fds[RTE_PMD_TAP_MAX_QUEUES]; /* List of all file descriptors */
108 struct rx_queue rxq[RTE_PMD_TAP_MAX_QUEUES]; /* List of RX queues */
109 struct tx_queue txq[RTE_PMD_TAP_MAX_QUEUES]; /* List of TX queues */
112 /* Tun/Tap allocation routine
114 * name is the number of the interface to use, unless NULL to take the host
118 tun_alloc(char *name)
121 unsigned int features;
124 memset(&ifr, 0, sizeof(struct ifreq));
126 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
128 strncpy(ifr.ifr_name, name, IFNAMSIZ);
130 fd = open(TUN_TAP_DEV_PATH, O_RDWR);
132 RTE_LOG(ERR, PMD, "Unable to create TAP interface");
136 /* Grab the TUN features to verify we can work */
137 if (ioctl(fd, TUNGETFEATURES, &features) < 0) {
138 RTE_LOG(ERR, PMD, "Unable to get TUN/TAP features\n");
141 RTE_LOG(DEBUG, PMD, "TUN/TAP Features %08x\n", features);
143 #ifdef IFF_MULTI_QUEUE
144 if (!(features & IFF_MULTI_QUEUE) && (RTE_PMD_TAP_MAX_QUEUES > 1)) {
145 RTE_LOG(DEBUG, PMD, "TUN/TAP device only one queue\n");
147 } else if ((features & IFF_ONE_QUEUE) &&
148 (RTE_PMD_TAP_MAX_QUEUES == 1)) {
149 ifr.ifr_flags |= IFF_ONE_QUEUE;
150 RTE_LOG(DEBUG, PMD, "Single queue only support\n");
152 ifr.ifr_flags |= IFF_MULTI_QUEUE;
153 RTE_LOG(DEBUG, PMD, "Multi-queue support for %d queues\n",
154 RTE_PMD_TAP_MAX_QUEUES);
157 if (RTE_PMD_TAP_MAX_QUEUES > 1) {
158 RTE_LOG(DEBUG, PMD, "TUN/TAP device only one queue\n");
161 ifr.ifr_flags |= IFF_ONE_QUEUE;
162 RTE_LOG(DEBUG, PMD, "Single queue only support\n");
166 /* Set the TUN/TAP configuration and get the name if needed */
167 if (ioctl(fd, TUNSETIFF, (void *)&ifr) < 0) {
168 RTE_LOG(ERR, PMD, "Unable to set TUNSETIFF for %s\n",
174 /* Always set the file descriptor to non-blocking */
175 if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0) {
176 RTE_LOG(ERR, PMD, "Unable to set to nonblocking\n");
177 perror("F_SETFL, NONBLOCK");
181 /* If the name is different that new name as default */
182 if (name && strcmp(name, ifr.ifr_name))
183 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s", ifr.ifr_name);
193 /* Callback to handle the rx burst of packets to the correct interface and
194 * file descriptor(s) in a multi-queue setup.
197 pmd_rx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
200 struct rte_mbuf *mbuf;
201 struct rx_queue *rxq = queue;
203 unsigned long num_rx_bytes = 0;
205 for (num_rx = 0; num_rx < nb_pkts; ) {
206 /* allocate the next mbuf */
207 mbuf = rte_pktmbuf_alloc(rxq->mp);
208 if (unlikely(!mbuf)) {
209 RTE_LOG(WARNING, PMD, "Unable to allocate mbuf\n");
213 len = read(rxq->fd, rte_pktmbuf_mtod(mbuf, char *),
214 rte_pktmbuf_tailroom(mbuf));
216 rte_pktmbuf_free(mbuf);
220 mbuf->data_len = len;
222 mbuf->port = rxq->in_port;
224 /* account for the receive frame */
225 bufs[num_rx++] = mbuf;
226 num_rx_bytes += mbuf->pkt_len;
228 rxq->stats.ipackets += num_rx;
229 rxq->stats.ibytes += num_rx_bytes;
234 /* Callback to handle sending packets from the tap interface
237 pmd_tx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
239 struct rte_mbuf *mbuf;
240 struct tx_queue *txq = queue;
243 unsigned long num_tx_bytes = 0;
246 if (unlikely(nb_pkts == 0))
249 pfd.events = POLLOUT;
251 for (i = 0; i < nb_pkts; i++) {
252 n = poll(&pfd, 1, 0);
257 if (pfd.revents & POLLOUT) {
258 /* copy the tx frame data */
260 n = write(pfd.fd, rte_pktmbuf_mtod(mbuf, void*),
261 rte_pktmbuf_pkt_len(mbuf));
266 num_tx_bytes += mbuf->pkt_len;
267 rte_pktmbuf_free(mbuf);
271 txq->stats.opackets += num_tx;
272 txq->stats.errs += nb_pkts - num_tx;
273 txq->stats.obytes += num_tx_bytes;
279 tap_dev_start(struct rte_eth_dev *dev)
281 /* Force the Link up */
282 dev->data->dev_link.link_status = ETH_LINK_UP;
287 /* This function gets called when the current port gets stopped.
290 tap_dev_stop(struct rte_eth_dev *dev)
293 struct pmd_internals *internals = dev->data->dev_private;
295 for (i = 0; i < internals->nb_queues; i++)
296 if (internals->fds[i] != -1)
297 close(internals->fds[i]);
299 dev->data->dev_link.link_status = ETH_LINK_DOWN;
303 tap_dev_configure(struct rte_eth_dev *dev __rte_unused)
309 tap_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
311 struct pmd_internals *internals = dev->data->dev_private;
313 dev_info->if_index = internals->if_index;
314 dev_info->max_mac_addrs = 1;
315 dev_info->max_rx_pktlen = (uint32_t)ETHER_MAX_VLAN_FRAME_LEN;
316 dev_info->max_rx_queues = internals->nb_queues;
317 dev_info->max_tx_queues = internals->nb_queues;
318 dev_info->min_rx_bufsize = 0;
319 dev_info->pci_dev = NULL;
323 tap_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *tap_stats)
325 unsigned int i, imax;
326 unsigned long rx_total = 0, tx_total = 0, tx_err_total = 0;
327 unsigned long rx_bytes_total = 0, tx_bytes_total = 0;
328 const struct pmd_internals *pmd = dev->data->dev_private;
330 imax = (pmd->nb_queues < RTE_ETHDEV_QUEUE_STAT_CNTRS) ?
331 pmd->nb_queues : RTE_ETHDEV_QUEUE_STAT_CNTRS;
333 for (i = 0; i < imax; i++) {
334 tap_stats->q_ipackets[i] = pmd->rxq[i].stats.ipackets;
335 tap_stats->q_ibytes[i] = pmd->rxq[i].stats.ibytes;
336 rx_total += tap_stats->q_ipackets[i];
337 rx_bytes_total += tap_stats->q_ibytes[i];
340 for (i = 0; i < imax; i++) {
341 tap_stats->q_opackets[i] = pmd->txq[i].stats.opackets;
342 tap_stats->q_errors[i] = pmd->txq[i].stats.errs;
343 tap_stats->q_obytes[i] = pmd->txq[i].stats.obytes;
344 tx_total += tap_stats->q_opackets[i];
345 tx_err_total += tap_stats->q_errors[i];
346 tx_bytes_total += tap_stats->q_obytes[i];
349 tap_stats->ipackets = rx_total;
350 tap_stats->ibytes = rx_bytes_total;
351 tap_stats->opackets = tx_total;
352 tap_stats->oerrors = tx_err_total;
353 tap_stats->obytes = tx_bytes_total;
357 tap_stats_reset(struct rte_eth_dev *dev)
360 struct pmd_internals *pmd = dev->data->dev_private;
362 for (i = 0; i < pmd->nb_queues; i++) {
363 pmd->rxq[i].stats.ipackets = 0;
364 pmd->rxq[i].stats.ibytes = 0;
367 for (i = 0; i < pmd->nb_queues; i++) {
368 pmd->txq[i].stats.opackets = 0;
369 pmd->txq[i].stats.errs = 0;
370 pmd->txq[i].stats.obytes = 0;
375 tap_dev_close(struct rte_eth_dev *dev __rte_unused)
380 tap_rx_queue_release(void *queue)
382 struct rx_queue *rxq = queue;
384 if (rxq && (rxq->fd > 0)) {
391 tap_tx_queue_release(void *queue)
393 struct tx_queue *txq = queue;
395 if (txq && (txq->fd > 0)) {
402 tap_link_update(struct rte_eth_dev *dev __rte_unused,
403 int wait_to_complete __rte_unused)
409 tap_setup_queue(struct rte_eth_dev *dev,
410 struct pmd_internals *internals,
413 struct pmd_internals *pmd = dev->data->dev_private;
414 struct rx_queue *rx = &internals->rxq[qid];
415 struct tx_queue *tx = &internals->txq[qid];
422 RTE_LOG(INFO, PMD, "Add queue to TAP %s for qid %d\n",
424 fd = tun_alloc(pmd->name);
426 RTE_LOG(ERR, PMD, "tun_alloc(%s) failed\n",
440 rx_setup_queue(struct rte_eth_dev *dev,
441 struct pmd_internals *internals,
444 dev->data->rx_queues[qid] = &internals->rxq[qid];
446 return tap_setup_queue(dev, internals, qid);
450 tx_setup_queue(struct rte_eth_dev *dev,
451 struct pmd_internals *internals,
454 dev->data->tx_queues[qid] = &internals->txq[qid];
456 return tap_setup_queue(dev, internals, qid);
460 tap_rx_queue_setup(struct rte_eth_dev *dev,
461 uint16_t rx_queue_id,
462 uint16_t nb_rx_desc __rte_unused,
463 unsigned int socket_id __rte_unused,
464 const struct rte_eth_rxconf *rx_conf __rte_unused,
465 struct rte_mempool *mp)
467 struct pmd_internals *internals = dev->data->dev_private;
471 if ((rx_queue_id >= internals->nb_queues) || !mp) {
472 RTE_LOG(ERR, PMD, "nb_queues %d mp %p\n",
473 internals->nb_queues, mp);
477 internals->rxq[rx_queue_id].mp = mp;
478 internals->rxq[rx_queue_id].in_port = dev->data->port_id;
480 /* Now get the space available for data in the mbuf */
481 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(mp) -
482 RTE_PKTMBUF_HEADROOM);
484 if (buf_size < ETH_FRAME_LEN) {
486 "%s: %d bytes will not fit in mbuf (%d bytes)\n",
487 dev->data->name, ETH_FRAME_LEN, buf_size);
491 fd = rx_setup_queue(dev, internals, rx_queue_id);
495 internals->fds[rx_queue_id] = fd;
496 RTE_LOG(INFO, PMD, "RX TAP device name %s, qid %d on fd %d\n",
497 internals->name, rx_queue_id, internals->rxq[rx_queue_id].fd);
503 tap_tx_queue_setup(struct rte_eth_dev *dev,
504 uint16_t tx_queue_id,
505 uint16_t nb_tx_desc __rte_unused,
506 unsigned int socket_id __rte_unused,
507 const struct rte_eth_txconf *tx_conf __rte_unused)
509 struct pmd_internals *internals = dev->data->dev_private;
512 if (tx_queue_id >= internals->nb_queues)
515 ret = tx_setup_queue(dev, internals, tx_queue_id);
519 RTE_LOG(INFO, PMD, "TX TAP device name %s, qid %d on fd %d\n",
520 internals->name, tx_queue_id, internals->txq[tx_queue_id].fd);
525 static const struct eth_dev_ops ops = {
526 .dev_start = tap_dev_start,
527 .dev_stop = tap_dev_stop,
528 .dev_close = tap_dev_close,
529 .dev_configure = tap_dev_configure,
530 .dev_infos_get = tap_dev_info,
531 .rx_queue_setup = tap_rx_queue_setup,
532 .tx_queue_setup = tap_tx_queue_setup,
533 .rx_queue_release = tap_rx_queue_release,
534 .tx_queue_release = tap_tx_queue_release,
535 .link_update = tap_link_update,
536 .stats_get = tap_stats_get,
537 .stats_reset = tap_stats_reset,
541 pmd_mac_address(int fd, struct rte_eth_dev *dev, struct ether_addr *addr)
545 if ((fd <= 0) || !dev || !addr)
548 memset(&ifr, 0, sizeof(ifr));
550 if (ioctl(fd, SIOCGIFHWADDR, &ifr) == -1) {
551 RTE_LOG(ERR, PMD, "ioctl failed (SIOCGIFHWADDR) (%s)\n",
556 /* Set the host based MAC address to this special MAC format */
557 ifr.ifr_hwaddr.sa_data[0] = 'T';
558 ifr.ifr_hwaddr.sa_data[1] = 'a';
559 ifr.ifr_hwaddr.sa_data[2] = 'p';
560 ifr.ifr_hwaddr.sa_data[3] = '-';
561 ifr.ifr_hwaddr.sa_data[4] = dev->data->port_id;
562 ifr.ifr_hwaddr.sa_data[5] = dev->data->numa_node;
563 if (ioctl(fd, SIOCSIFHWADDR, &ifr) == -1) {
564 RTE_LOG(ERR, PMD, "%s: ioctl failed (SIOCSIFHWADDR) (%s)\n",
565 dev->data->name, ifr.ifr_name);
569 /* Set the local application MAC address, needs to be different then
570 * the host based MAC address.
572 ifr.ifr_hwaddr.sa_data[0] = 'd';
573 ifr.ifr_hwaddr.sa_data[1] = 'n';
574 ifr.ifr_hwaddr.sa_data[2] = 'e';
575 ifr.ifr_hwaddr.sa_data[3] = 't';
576 ifr.ifr_hwaddr.sa_data[4] = dev->data->port_id;
577 ifr.ifr_hwaddr.sa_data[5] = dev->data->numa_node;
578 rte_memcpy(addr, ifr.ifr_hwaddr.sa_data, ETH_ALEN);
584 eth_dev_tap_create(const char *name, char *tap_name)
586 int numa_node = rte_socket_id();
587 struct rte_eth_dev *dev = NULL;
588 struct pmd_internals *pmd = NULL;
589 struct rte_eth_dev_data *data = NULL;
593 "%s: Create TAP Ethernet device with %d queues on numa %u\n",
594 name, RTE_PMD_TAP_MAX_QUEUES, rte_socket_id());
596 data = rte_zmalloc_socket(tap_name, sizeof(*data), 0, numa_node);
598 RTE_LOG(ERR, PMD, "Failed to allocate data\n");
602 pmd = rte_zmalloc_socket(tap_name, sizeof(*pmd), 0, numa_node);
604 RTE_LOG(ERR, PMD, "Unable to allocate internal struct\n");
608 /* Use the name and not the tap_name */
609 dev = rte_eth_dev_allocate(tap_name);
611 RTE_LOG(ERR, PMD, "Unable to allocate device struct\n");
615 snprintf(pmd->name, sizeof(pmd->name), "%s", tap_name);
617 pmd->nb_queues = RTE_PMD_TAP_MAX_QUEUES;
619 /* Setup some default values */
620 data->dev_private = pmd;
621 data->port_id = dev->data->port_id;
622 data->dev_flags = RTE_ETH_DEV_DETACHABLE;
623 data->kdrv = RTE_KDRV_NONE;
624 data->drv_name = pmd_tap_drv.driver.name;
625 data->numa_node = numa_node;
627 data->dev_link = pmd_link;
628 data->mac_addrs = &pmd->eth_addr;
629 data->nb_rx_queues = pmd->nb_queues;
630 data->nb_tx_queues = pmd->nb_queues;
635 dev->rx_pkt_burst = pmd_rx_burst;
636 dev->tx_pkt_burst = pmd_tx_burst;
637 snprintf(dev->data->name, sizeof(dev->data->name), "%s", name);
639 /* Create the first Tap device */
640 fd = tun_alloc(tap_name);
642 RTE_LOG(ERR, PMD, "tun_alloc() failed\n");
646 /* Presetup the fds to -1 as being not working */
647 for (i = 1; i < RTE_PMD_TAP_MAX_QUEUES; i++) {
653 /* Take the TUN/TAP fd and place in the first location */
658 if (pmd_mac_address(fd, dev, &pmd->eth_addr) < 0) {
659 RTE_LOG(ERR, PMD, "Unable to get MAC address\n");
666 RTE_PMD_DEBUG_TRACE("Unable to initialize %s\n", name);
671 rte_eth_dev_release_port(dev);
677 set_interface_name(const char *key __rte_unused,
681 char *name = (char *)extra_args;
684 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s", value);
686 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s%d",
687 DEFAULT_TAP_NAME, (tap_unit - 1));
693 set_interface_speed(const char *key __rte_unused,
697 *(int *)extra_args = (value) ? atoi(value) : ETH_SPEED_NUM_10G;
702 /* Open a TAP interface device.
705 rte_pmd_tap_probe(const char *name, const char *params)
708 struct rte_kvargs *kvlist = NULL;
710 char tap_name[RTE_ETH_NAME_MAX_LEN];
712 speed = ETH_SPEED_NUM_10G;
713 snprintf(tap_name, sizeof(tap_name), "%s%d",
714 DEFAULT_TAP_NAME, tap_unit++);
716 RTE_LOG(INFO, PMD, "Initializing pmd_tap for %s as %s\n",
719 if (params && (params[0] != '\0')) {
720 RTE_LOG(INFO, PMD, "paramaters (%s)\n", params);
722 kvlist = rte_kvargs_parse(params, valid_arguments);
724 if (rte_kvargs_count(kvlist, ETH_TAP_SPEED_ARG) == 1) {
725 ret = rte_kvargs_process(kvlist,
727 &set_interface_speed,
733 if (rte_kvargs_count(kvlist, ETH_TAP_IFACE_ARG) == 1) {
734 ret = rte_kvargs_process(kvlist,
743 pmd_link.link_speed = speed;
745 ret = eth_dev_tap_create(name, tap_name);
749 RTE_LOG(INFO, PMD, "Failed to create pmd for %s as %s\n",
751 tap_unit--; /* Restore the unit number */
753 rte_kvargs_free(kvlist);
758 /* detach a TAP device.
761 rte_pmd_tap_remove(const char *name)
763 struct rte_eth_dev *eth_dev = NULL;
764 struct pmd_internals *internals;
767 RTE_LOG(INFO, PMD, "Closing TUN/TAP Ethernet device on numa %u\n",
770 /* find the ethdev entry */
771 eth_dev = rte_eth_dev_allocated(name);
775 internals = eth_dev->data->dev_private;
776 for (i = 0; i < internals->nb_queues; i++)
777 if (internals->fds[i] != -1)
778 close(internals->fds[i]);
780 rte_free(eth_dev->data->dev_private);
781 rte_free(eth_dev->data);
783 rte_eth_dev_release_port(eth_dev);
788 static struct rte_vdev_driver pmd_tap_drv = {
789 .probe = rte_pmd_tap_probe,
790 .remove = rte_pmd_tap_remove,
792 RTE_PMD_REGISTER_VDEV(net_tap, pmd_tap_drv);
793 RTE_PMD_REGISTER_ALIAS(net_tap, eth_tap);
794 RTE_PMD_REGISTER_PARAM_STRING(net_tap, "iface=<string>,speed=N");