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>
46 #include <arpa/inet.h>
48 #include <linux/if_tun.h>
49 #include <linux/if_ether.h>
52 /* Linux based path to the TUN device */
53 #define TUN_TAP_DEV_PATH "/dev/net/tun"
54 #define DEFAULT_TAP_NAME "dtap"
56 #define ETH_TAP_IFACE_ARG "iface"
57 #define ETH_TAP_SPEED_ARG "speed"
59 #ifdef IFF_MULTI_QUEUE
60 #define RTE_PMD_TAP_MAX_QUEUES 16
62 #define RTE_PMD_TAP_MAX_QUEUES 1
65 static struct rte_vdev_driver pmd_tap_drv;
67 static const char *valid_arguments[] = {
75 static struct rte_eth_link pmd_link = {
76 .link_speed = ETH_SPEED_NUM_10G,
77 .link_duplex = ETH_LINK_FULL_DUPLEX,
78 .link_status = ETH_LINK_DOWN,
79 .link_autoneg = ETH_LINK_SPEED_AUTONEG
83 uint64_t opackets; /* Number of output packets */
84 uint64_t ipackets; /* Number of input packets */
85 uint64_t obytes; /* Number of bytes on output */
86 uint64_t ibytes; /* Number of bytes on input */
87 uint64_t errs; /* Number of error packets */
91 struct rte_mempool *mp; /* Mempool for RX packets */
92 uint16_t in_port; /* Port ID */
95 struct pkt_stats stats; /* Stats for this RX queue */
100 struct pkt_stats stats; /* Stats for this TX queue */
103 struct pmd_internals {
104 char name[RTE_ETH_NAME_MAX_LEN]; /* Internal Tap device name */
105 uint16_t nb_queues; /* Number of queues supported */
106 struct ether_addr eth_addr; /* Mac address of the device port */
108 int if_index; /* IF_INDEX for the port */
110 struct rx_queue rxq[RTE_PMD_TAP_MAX_QUEUES]; /* List of RX queues */
111 struct tx_queue txq[RTE_PMD_TAP_MAX_QUEUES]; /* List of TX queues */
114 /* Tun/Tap allocation routine
116 * name is the number of the interface to use, unless NULL to take the host
120 tun_alloc(struct pmd_internals *pmd, uint16_t qid)
123 #ifdef IFF_MULTI_QUEUE
124 unsigned int features;
128 memset(&ifr, 0, sizeof(struct ifreq));
130 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
131 snprintf(ifr.ifr_name, IFNAMSIZ, "%s", pmd->name);
133 RTE_LOG(DEBUG, PMD, "ifr_name '%s'\n", ifr.ifr_name);
135 fd = open(TUN_TAP_DEV_PATH, O_RDWR);
137 RTE_LOG(ERR, PMD, "Unable to create TAP interface");
141 #ifdef IFF_MULTI_QUEUE
142 /* Grab the TUN features to verify we can work multi-queue */
143 if (ioctl(fd, TUNGETFEATURES, &features) < 0) {
144 RTE_LOG(ERR, PMD, "TAP unable to get TUN/TAP features\n");
147 RTE_LOG(DEBUG, PMD, " TAP Features %08x\n", features);
149 if (features & IFF_MULTI_QUEUE) {
150 RTE_LOG(DEBUG, PMD, " Multi-queue support for %d queues\n",
151 RTE_PMD_TAP_MAX_QUEUES);
152 ifr.ifr_flags |= IFF_MULTI_QUEUE;
156 ifr.ifr_flags |= IFF_ONE_QUEUE;
157 RTE_LOG(DEBUG, PMD, " Single queue only support\n");
160 /* Set the TUN/TAP configuration and set the name if needed */
161 if (ioctl(fd, TUNSETIFF, (void *)&ifr) < 0) {
162 RTE_LOG(WARNING, PMD,
163 "Unable to set TUNSETIFF for %s\n",
169 /* Always set the file descriptor to non-blocking */
170 if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0) {
171 RTE_LOG(WARNING, PMD,
172 "Unable to set %s to nonblocking\n",
174 perror("F_SETFL, NONBLOCK");
179 if (ioctl(fd, SIOCGIFHWADDR, &ifr) == -1) {
180 RTE_LOG(ERR, PMD, "ioctl failed (SIOCGIFHWADDR) (%s)\n",
185 rte_memcpy(&pmd->eth_addr, ifr.ifr_hwaddr.sa_data, ETH_ALEN);
196 /* Callback to handle the rx burst of packets to the correct interface and
197 * file descriptor(s) in a multi-queue setup.
200 pmd_rx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
203 struct rte_mbuf *mbuf;
204 struct rx_queue *rxq = queue;
206 unsigned long num_rx_bytes = 0;
208 for (num_rx = 0; num_rx < nb_pkts; ) {
209 /* allocate the next mbuf */
210 mbuf = rte_pktmbuf_alloc(rxq->mp);
211 if (unlikely(!mbuf)) {
212 RTE_LOG(WARNING, PMD, "TAP unable to allocate mbuf\n");
216 len = read(rxq->fd, rte_pktmbuf_mtod(mbuf, char *),
217 rte_pktmbuf_tailroom(mbuf));
219 rte_pktmbuf_free(mbuf);
223 mbuf->data_len = len;
225 mbuf->port = rxq->in_port;
227 /* account for the receive frame */
228 bufs[num_rx++] = mbuf;
229 num_rx_bytes += mbuf->pkt_len;
231 rxq->stats.ipackets += num_rx;
232 rxq->stats.ibytes += num_rx_bytes;
237 /* Callback to handle sending packets from the tap interface
240 pmd_tx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
242 struct rte_mbuf *mbuf;
243 struct tx_queue *txq = queue;
245 unsigned long num_tx_bytes = 0;
248 if (unlikely(nb_pkts == 0))
251 for (i = 0; i < nb_pkts; i++) {
252 /* copy the tx frame data */
255 rte_pktmbuf_mtod(mbuf, void *),
256 rte_pktmbuf_pkt_len(mbuf));
261 num_tx_bytes += mbuf->pkt_len;
262 rte_pktmbuf_free(mbuf);
265 txq->stats.opackets += num_tx;
266 txq->stats.errs += nb_pkts - num_tx;
267 txq->stats.obytes += num_tx_bytes;
273 tap_link_set_flags(struct pmd_internals *pmd, short flags, int add)
279 * An AF_INET/DGRAM socket is needed for
280 * SIOCGIFFLAGS/SIOCSIFFLAGS, using fd won't work.
282 s = socket(AF_INET, SOCK_DGRAM, 0);
285 "Unable to get a socket to set flags: %s\n",
289 memset(&ifr, 0, sizeof(ifr));
290 snprintf(ifr.ifr_name, IFNAMSIZ, "%s", pmd->name);
291 err = ioctl(s, SIOCGIFFLAGS, &ifr);
293 RTE_LOG(WARNING, PMD, "Unable to get %s device flags: %s\n",
294 pmd->name, strerror(errno));
299 ifr.ifr_flags |= flags;
301 ifr.ifr_flags &= ~flags;
302 err = ioctl(s, SIOCSIFFLAGS, &ifr);
304 RTE_LOG(WARNING, PMD, "Unable to %s flags 0x%x: %s\n",
305 add ? "set" : "unset", flags, strerror(errno));
315 tap_link_set_down(struct rte_eth_dev *dev)
317 struct pmd_internals *pmd = dev->data->dev_private;
319 dev->data->dev_link.link_status = ETH_LINK_DOWN;
320 return tap_link_set_flags(pmd, IFF_UP | IFF_NOARP, 0);
324 tap_link_set_up(struct rte_eth_dev *dev)
326 struct pmd_internals *pmd = dev->data->dev_private;
328 dev->data->dev_link.link_status = ETH_LINK_UP;
329 return tap_link_set_flags(pmd, IFF_UP | IFF_NOARP, 1);
333 tap_dev_start(struct rte_eth_dev *dev)
335 return tap_link_set_up(dev);
338 /* This function gets called when the current port gets stopped.
341 tap_dev_stop(struct rte_eth_dev *dev)
343 tap_link_set_down(dev);
347 tap_dev_configure(struct rte_eth_dev *dev __rte_unused)
353 tap_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
355 struct pmd_internals *internals = dev->data->dev_private;
357 dev_info->if_index = internals->if_index;
358 dev_info->max_mac_addrs = 1;
359 dev_info->max_rx_pktlen = (uint32_t)ETHER_MAX_VLAN_FRAME_LEN;
360 dev_info->max_rx_queues = internals->nb_queues;
361 dev_info->max_tx_queues = internals->nb_queues;
362 dev_info->min_rx_bufsize = 0;
363 dev_info->pci_dev = NULL;
367 tap_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *tap_stats)
369 unsigned int i, imax;
370 unsigned long rx_total = 0, tx_total = 0, tx_err_total = 0;
371 unsigned long rx_bytes_total = 0, tx_bytes_total = 0;
372 const struct pmd_internals *pmd = dev->data->dev_private;
374 imax = (pmd->nb_queues < RTE_ETHDEV_QUEUE_STAT_CNTRS) ?
375 pmd->nb_queues : RTE_ETHDEV_QUEUE_STAT_CNTRS;
377 for (i = 0; i < imax; i++) {
378 tap_stats->q_ipackets[i] = pmd->rxq[i].stats.ipackets;
379 tap_stats->q_ibytes[i] = pmd->rxq[i].stats.ibytes;
380 rx_total += tap_stats->q_ipackets[i];
381 rx_bytes_total += tap_stats->q_ibytes[i];
383 tap_stats->q_opackets[i] = pmd->txq[i].stats.opackets;
384 tap_stats->q_errors[i] = pmd->txq[i].stats.errs;
385 tap_stats->q_obytes[i] = pmd->txq[i].stats.obytes;
386 tx_total += tap_stats->q_opackets[i];
387 tx_err_total += tap_stats->q_errors[i];
388 tx_bytes_total += tap_stats->q_obytes[i];
391 tap_stats->ipackets = rx_total;
392 tap_stats->ibytes = rx_bytes_total;
393 tap_stats->opackets = tx_total;
394 tap_stats->oerrors = tx_err_total;
395 tap_stats->obytes = tx_bytes_total;
399 tap_stats_reset(struct rte_eth_dev *dev)
402 struct pmd_internals *pmd = dev->data->dev_private;
404 for (i = 0; i < pmd->nb_queues; i++) {
405 pmd->rxq[i].stats.ipackets = 0;
406 pmd->rxq[i].stats.ibytes = 0;
408 pmd->txq[i].stats.opackets = 0;
409 pmd->txq[i].stats.errs = 0;
410 pmd->txq[i].stats.obytes = 0;
415 tap_dev_close(struct rte_eth_dev *dev __rte_unused)
418 struct pmd_internals *internals = dev->data->dev_private;
420 tap_link_set_down(dev);
422 for (i = 0; i < internals->nb_queues; i++) {
423 if (internals->rxq[i].fd != -1)
424 close(internals->rxq[i].fd);
425 internals->rxq[i].fd = -1;
426 internals->txq[i].fd = -1;
431 tap_rx_queue_release(void *queue)
433 struct rx_queue *rxq = queue;
435 if (rxq && (rxq->fd > 0)) {
442 tap_tx_queue_release(void *queue)
444 struct tx_queue *txq = queue;
446 if (txq && (txq->fd > 0)) {
453 tap_link_update(struct rte_eth_dev *dev __rte_unused,
454 int wait_to_complete __rte_unused)
460 tap_promisc_enable(struct rte_eth_dev *dev)
462 struct pmd_internals *pmd = dev->data->dev_private;
464 dev->data->promiscuous = 1;
465 tap_link_set_flags(pmd, IFF_PROMISC, 1);
469 tap_promisc_disable(struct rte_eth_dev *dev)
471 struct pmd_internals *pmd = dev->data->dev_private;
473 dev->data->promiscuous = 0;
474 tap_link_set_flags(pmd, IFF_PROMISC, 0);
478 tap_allmulti_enable(struct rte_eth_dev *dev)
480 struct pmd_internals *pmd = dev->data->dev_private;
482 dev->data->all_multicast = 1;
483 tap_link_set_flags(pmd, IFF_ALLMULTI, 1);
487 tap_allmulti_disable(struct rte_eth_dev *dev)
489 struct pmd_internals *pmd = dev->data->dev_private;
491 dev->data->all_multicast = 0;
492 tap_link_set_flags(pmd, IFF_ALLMULTI, 0);
496 tap_setup_queue(struct rte_eth_dev *dev,
497 struct pmd_internals *internals,
500 struct pmd_internals *pmd = dev->data->dev_private;
501 struct rx_queue *rx = &internals->rxq[qid];
502 struct tx_queue *tx = &internals->txq[qid];
509 RTE_LOG(INFO, PMD, "Add queue to TAP %s for qid %d\n",
511 fd = tun_alloc(pmd, qid);
513 RTE_LOG(ERR, PMD, "tun_alloc(%s, %d) failed\n",
527 rx_setup_queue(struct rte_eth_dev *dev,
528 struct pmd_internals *internals,
531 dev->data->rx_queues[qid] = &internals->rxq[qid];
533 return tap_setup_queue(dev, internals, qid);
537 tx_setup_queue(struct rte_eth_dev *dev,
538 struct pmd_internals *internals,
541 dev->data->tx_queues[qid] = &internals->txq[qid];
543 return tap_setup_queue(dev, internals, qid);
547 tap_rx_queue_setup(struct rte_eth_dev *dev,
548 uint16_t rx_queue_id,
549 uint16_t nb_rx_desc __rte_unused,
550 unsigned int socket_id __rte_unused,
551 const struct rte_eth_rxconf *rx_conf __rte_unused,
552 struct rte_mempool *mp)
554 struct pmd_internals *internals = dev->data->dev_private;
558 if ((rx_queue_id >= internals->nb_queues) || !mp) {
559 RTE_LOG(WARNING, PMD,
560 "nb_queues %d too small or mempool NULL\n",
561 internals->nb_queues);
565 internals->rxq[rx_queue_id].mp = mp;
566 internals->rxq[rx_queue_id].in_port = dev->data->port_id;
568 /* Now get the space available for data in the mbuf */
569 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(mp) -
570 RTE_PKTMBUF_HEADROOM);
572 if (buf_size < ETH_FRAME_LEN) {
573 RTE_LOG(WARNING, PMD,
574 "%s: %d bytes will not fit in mbuf (%d bytes)\n",
575 dev->data->name, ETH_FRAME_LEN, buf_size);
579 fd = rx_setup_queue(dev, internals, rx_queue_id);
583 RTE_LOG(DEBUG, PMD, " RX TAP device name %s, qid %d on fd %d\n",
584 internals->name, rx_queue_id, internals->rxq[rx_queue_id].fd);
590 tap_tx_queue_setup(struct rte_eth_dev *dev,
591 uint16_t tx_queue_id,
592 uint16_t nb_tx_desc __rte_unused,
593 unsigned int socket_id __rte_unused,
594 const struct rte_eth_txconf *tx_conf __rte_unused)
596 struct pmd_internals *internals = dev->data->dev_private;
599 if (tx_queue_id >= internals->nb_queues)
602 ret = tx_setup_queue(dev, internals, tx_queue_id);
606 RTE_LOG(DEBUG, PMD, " TX TAP device name %s, qid %d on fd %d\n",
607 internals->name, tx_queue_id, internals->txq[tx_queue_id].fd);
612 static const struct eth_dev_ops ops = {
613 .dev_start = tap_dev_start,
614 .dev_stop = tap_dev_stop,
615 .dev_close = tap_dev_close,
616 .dev_configure = tap_dev_configure,
617 .dev_infos_get = tap_dev_info,
618 .rx_queue_setup = tap_rx_queue_setup,
619 .tx_queue_setup = tap_tx_queue_setup,
620 .rx_queue_release = tap_rx_queue_release,
621 .tx_queue_release = tap_tx_queue_release,
622 .link_update = tap_link_update,
623 .dev_set_link_up = tap_link_set_up,
624 .dev_set_link_down = tap_link_set_down,
625 .promiscuous_enable = tap_promisc_enable,
626 .promiscuous_disable = tap_promisc_disable,
627 .allmulticast_enable = tap_allmulti_enable,
628 .allmulticast_disable = tap_allmulti_disable,
629 .stats_get = tap_stats_get,
630 .stats_reset = tap_stats_reset,
634 eth_dev_tap_create(const char *name, char *tap_name)
636 int numa_node = rte_socket_id();
637 struct rte_eth_dev *dev = NULL;
638 struct pmd_internals *pmd = NULL;
639 struct rte_eth_dev_data *data = NULL;
642 RTE_LOG(DEBUG, PMD, " TAP device on numa %u\n", rte_socket_id());
644 data = rte_zmalloc_socket(tap_name, sizeof(*data), 0, numa_node);
646 RTE_LOG(ERR, PMD, "TAP Failed to allocate data\n");
650 pmd = rte_zmalloc_socket(tap_name, sizeof(*pmd), 0, numa_node);
652 RTE_LOG(ERR, PMD, "TAP Unable to allocate internal struct\n");
656 /* name in allocation and data->name must be consistent */
657 snprintf(data->name, sizeof(data->name), "%s", name);
658 dev = rte_eth_dev_allocate(name);
660 RTE_LOG(ERR, PMD, "TAP Unable to allocate device struct\n");
664 snprintf(pmd->name, sizeof(pmd->name), "%s", tap_name);
666 pmd->nb_queues = RTE_PMD_TAP_MAX_QUEUES;
668 /* Setup some default values */
669 data->dev_private = pmd;
670 data->port_id = dev->data->port_id;
671 data->dev_flags = RTE_ETH_DEV_DETACHABLE;
672 data->kdrv = RTE_KDRV_NONE;
673 data->drv_name = pmd_tap_drv.driver.name;
674 data->numa_node = numa_node;
676 data->dev_link = pmd_link;
677 data->mac_addrs = &pmd->eth_addr;
678 data->nb_rx_queues = pmd->nb_queues;
679 data->nb_tx_queues = pmd->nb_queues;
684 dev->rx_pkt_burst = pmd_rx_burst;
685 dev->tx_pkt_burst = pmd_tx_burst;
687 /* Presetup the fds to -1 as being not valid */
688 for (i = 0; i < RTE_PMD_TAP_MAX_QUEUES; i++) {
696 RTE_LOG(DEBUG, PMD, "TAP Unable to initialize %s\n", name);
701 rte_eth_dev_release_port(dev);
707 set_interface_name(const char *key __rte_unused,
711 char *name = (char *)extra_args;
714 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s", value);
716 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s%d",
717 DEFAULT_TAP_NAME, (tap_unit - 1));
723 set_interface_speed(const char *key __rte_unused,
727 *(int *)extra_args = (value) ? atoi(value) : ETH_SPEED_NUM_10G;
732 /* Open a TAP interface device.
735 rte_pmd_tap_probe(const char *name, const char *params)
738 struct rte_kvargs *kvlist = NULL;
740 char tap_name[RTE_ETH_NAME_MAX_LEN];
742 speed = ETH_SPEED_NUM_10G;
743 snprintf(tap_name, sizeof(tap_name), "%s%d",
744 DEFAULT_TAP_NAME, tap_unit++);
746 if (params && (params[0] != '\0')) {
747 RTE_LOG(DEBUG, PMD, "paramaters (%s)\n", params);
749 kvlist = rte_kvargs_parse(params, valid_arguments);
751 if (rte_kvargs_count(kvlist, ETH_TAP_SPEED_ARG) == 1) {
752 ret = rte_kvargs_process(kvlist,
754 &set_interface_speed,
760 if (rte_kvargs_count(kvlist, ETH_TAP_IFACE_ARG) == 1) {
761 ret = rte_kvargs_process(kvlist,
770 pmd_link.link_speed = speed;
772 RTE_LOG(NOTICE, PMD, "Initializing pmd_tap for %s as %s\n",
775 ret = eth_dev_tap_create(name, tap_name);
779 RTE_LOG(ERR, PMD, "Failed to create pmd for %s as %s\n",
781 tap_unit--; /* Restore the unit number */
783 rte_kvargs_free(kvlist);
788 /* detach a TAP device.
791 rte_pmd_tap_remove(const char *name)
793 struct rte_eth_dev *eth_dev = NULL;
794 struct pmd_internals *internals;
797 RTE_LOG(DEBUG, PMD, "Closing TUN/TAP Ethernet device on numa %u\n",
800 /* find the ethdev entry */
801 eth_dev = rte_eth_dev_allocated(name);
805 internals = eth_dev->data->dev_private;
806 for (i = 0; i < internals->nb_queues; i++)
807 if (internals->rxq[i].fd != -1)
808 close(internals->rxq[i].fd);
810 rte_free(eth_dev->data->dev_private);
811 rte_free(eth_dev->data);
813 rte_eth_dev_release_port(eth_dev);
818 static struct rte_vdev_driver pmd_tap_drv = {
819 .probe = rte_pmd_tap_probe,
820 .remove = rte_pmd_tap_remove,
822 RTE_PMD_REGISTER_VDEV(net_tap, pmd_tap_drv);
823 RTE_PMD_REGISTER_ALIAS(net_tap, eth_tap);
824 RTE_PMD_REGISTER_PARAM_STRING(net_tap, "iface=<string>,speed=N");