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 #ifdef IFF_MULTI_QUEUE
61 #define RTE_PMD_TAP_MAX_QUEUES 16
63 #define RTE_PMD_TAP_MAX_QUEUES 1
66 static struct rte_vdev_driver pmd_tap_drv;
68 static const char *valid_arguments[] = {
76 static struct rte_eth_link pmd_link = {
77 .link_speed = ETH_SPEED_NUM_10G,
78 .link_duplex = ETH_LINK_FULL_DUPLEX,
79 .link_status = ETH_LINK_DOWN,
80 .link_autoneg = ETH_LINK_SPEED_AUTONEG
84 uint64_t opackets; /* Number of output packets */
85 uint64_t ipackets; /* Number of input packets */
86 uint64_t obytes; /* Number of bytes on output */
87 uint64_t ibytes; /* Number of bytes on input */
88 uint64_t errs; /* Number of error packets */
92 struct rte_mempool *mp; /* Mempool for RX packets */
93 uint16_t in_port; /* Port ID */
96 struct pkt_stats stats; /* Stats for this RX queue */
101 struct pkt_stats stats; /* Stats for this TX queue */
104 struct pmd_internals {
105 char name[RTE_ETH_NAME_MAX_LEN]; /* Internal Tap device name */
106 uint16_t nb_queues; /* Number of queues supported */
107 struct ether_addr eth_addr; /* Mac address of the device port */
109 int if_index; /* IF_INDEX for the port */
111 struct rx_queue rxq[RTE_PMD_TAP_MAX_QUEUES]; /* List of RX queues */
112 struct tx_queue txq[RTE_PMD_TAP_MAX_QUEUES]; /* List of TX queues */
115 /* Tun/Tap allocation routine
117 * name is the number of the interface to use, unless NULL to take the host
121 tun_alloc(struct pmd_internals *pmd, uint16_t qid)
124 #ifdef IFF_MULTI_QUEUE
125 unsigned int features;
129 memset(&ifr, 0, sizeof(struct ifreq));
131 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
132 strncpy(ifr.ifr_name, pmd->name, IFNAMSIZ);
134 RTE_LOG(DEBUG, PMD, "ifr_name '%s'\n", ifr.ifr_name);
136 fd = open(TUN_TAP_DEV_PATH, O_RDWR);
138 RTE_LOG(ERR, PMD, "Unable to create TAP interface");
142 #ifdef IFF_MULTI_QUEUE
143 /* Grab the TUN features to verify we can work multi-queue */
144 if (ioctl(fd, TUNGETFEATURES, &features) < 0) {
145 RTE_LOG(ERR, PMD, "TAP unable to get TUN/TAP features\n");
148 RTE_LOG(DEBUG, PMD, " TAP Features %08x\n", features);
150 if (features & IFF_MULTI_QUEUE) {
151 RTE_LOG(DEBUG, PMD, " Multi-queue support for %d queues\n",
152 RTE_PMD_TAP_MAX_QUEUES);
153 ifr.ifr_flags |= IFF_MULTI_QUEUE;
157 ifr.ifr_flags |= IFF_ONE_QUEUE;
158 RTE_LOG(DEBUG, PMD, " Single queue only support\n");
161 /* Set the TUN/TAP configuration and set the name if needed */
162 if (ioctl(fd, TUNSETIFF, (void *)&ifr) < 0) {
163 RTE_LOG(WARNING, PMD,
164 "Unable to set TUNSETIFF for %s\n",
170 /* Always set the file descriptor to non-blocking */
171 if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0) {
172 RTE_LOG(WARNING, PMD,
173 "Unable to set %s to nonblocking\n",
175 perror("F_SETFL, NONBLOCK");
180 if (ioctl(fd, SIOCGIFHWADDR, &ifr) == -1) {
181 RTE_LOG(ERR, PMD, "ioctl failed (SIOCGIFHWADDR) (%s)\n",
186 rte_memcpy(&pmd->eth_addr, ifr.ifr_hwaddr.sa_data, ETH_ALEN);
197 /* Callback to handle the rx burst of packets to the correct interface and
198 * file descriptor(s) in a multi-queue setup.
201 pmd_rx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
204 struct rte_mbuf *mbuf;
205 struct rx_queue *rxq = queue;
207 unsigned long num_rx_bytes = 0;
209 for (num_rx = 0; num_rx < nb_pkts; ) {
210 /* allocate the next mbuf */
211 mbuf = rte_pktmbuf_alloc(rxq->mp);
212 if (unlikely(!mbuf)) {
213 RTE_LOG(WARNING, PMD, "TAP unable to allocate mbuf\n");
217 len = read(rxq->fd, rte_pktmbuf_mtod(mbuf, char *),
218 rte_pktmbuf_tailroom(mbuf));
220 rte_pktmbuf_free(mbuf);
224 mbuf->data_len = len;
226 mbuf->port = rxq->in_port;
228 /* account for the receive frame */
229 bufs[num_rx++] = mbuf;
230 num_rx_bytes += mbuf->pkt_len;
232 rxq->stats.ipackets += num_rx;
233 rxq->stats.ibytes += num_rx_bytes;
238 /* Callback to handle sending packets from the tap interface
241 pmd_tx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
243 struct rte_mbuf *mbuf;
244 struct tx_queue *txq = queue;
247 unsigned long num_tx_bytes = 0;
250 if (unlikely(nb_pkts == 0))
253 pfd.events = POLLOUT;
255 for (i = 0; i < nb_pkts; i++) {
256 n = poll(&pfd, 1, 0);
261 if (pfd.revents & POLLOUT) {
262 /* copy the tx frame data */
264 n = write(pfd.fd, rte_pktmbuf_mtod(mbuf, void*),
265 rte_pktmbuf_pkt_len(mbuf));
270 num_tx_bytes += mbuf->pkt_len;
271 rte_pktmbuf_free(mbuf);
275 txq->stats.opackets += num_tx;
276 txq->stats.errs += nb_pkts - num_tx;
277 txq->stats.obytes += num_tx_bytes;
283 tap_link_set_flags(struct pmd_internals *pmd, short flags, int add)
289 * An AF_INET/DGRAM socket is needed for
290 * SIOCGIFFLAGS/SIOCSIFFLAGS, using fd won't work.
292 s = socket(AF_INET, SOCK_DGRAM, 0);
295 "Unable to get a socket to set flags: %s\n",
299 memset(&ifr, 0, sizeof(ifr));
300 strncpy(ifr.ifr_name, pmd->name, IFNAMSIZ);
301 err = ioctl(s, SIOCGIFFLAGS, &ifr);
303 RTE_LOG(WARNING, PMD, "Unable to get %s device flags: %s\n",
304 pmd->name, strerror(errno));
309 ifr.ifr_flags |= flags;
311 ifr.ifr_flags &= ~flags;
312 err = ioctl(s, SIOCSIFFLAGS, &ifr);
314 RTE_LOG(WARNING, PMD, "Unable to %s flags 0x%x: %s\n",
315 add ? "set" : "unset", flags, strerror(errno));
325 tap_link_set_down(struct rte_eth_dev *dev)
327 struct pmd_internals *pmd = dev->data->dev_private;
329 dev->data->dev_link.link_status = ETH_LINK_DOWN;
330 return tap_link_set_flags(pmd, IFF_UP | IFF_NOARP, 0);
334 tap_link_set_up(struct rte_eth_dev *dev)
336 struct pmd_internals *pmd = dev->data->dev_private;
338 dev->data->dev_link.link_status = ETH_LINK_UP;
339 return tap_link_set_flags(pmd, IFF_UP | IFF_NOARP, 1);
343 tap_dev_start(struct rte_eth_dev *dev)
345 return tap_link_set_up(dev);
348 /* This function gets called when the current port gets stopped.
351 tap_dev_stop(struct rte_eth_dev *dev)
354 struct pmd_internals *internals = dev->data->dev_private;
356 tap_link_set_down(dev);
358 for (i = 0; i < internals->nb_queues; i++)
359 if (internals->rxq[i].fd != -1)
360 close(internals->rxq[i].fd);
364 tap_dev_configure(struct rte_eth_dev *dev __rte_unused)
370 tap_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
372 struct pmd_internals *internals = dev->data->dev_private;
374 dev_info->if_index = internals->if_index;
375 dev_info->max_mac_addrs = 1;
376 dev_info->max_rx_pktlen = (uint32_t)ETHER_MAX_VLAN_FRAME_LEN;
377 dev_info->max_rx_queues = internals->nb_queues;
378 dev_info->max_tx_queues = internals->nb_queues;
379 dev_info->min_rx_bufsize = 0;
380 dev_info->pci_dev = NULL;
384 tap_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *tap_stats)
386 unsigned int i, imax;
387 unsigned long rx_total = 0, tx_total = 0, tx_err_total = 0;
388 unsigned long rx_bytes_total = 0, tx_bytes_total = 0;
389 const struct pmd_internals *pmd = dev->data->dev_private;
391 imax = (pmd->nb_queues < RTE_ETHDEV_QUEUE_STAT_CNTRS) ?
392 pmd->nb_queues : RTE_ETHDEV_QUEUE_STAT_CNTRS;
394 for (i = 0; i < imax; i++) {
395 tap_stats->q_ipackets[i] = pmd->rxq[i].stats.ipackets;
396 tap_stats->q_ibytes[i] = pmd->rxq[i].stats.ibytes;
397 rx_total += tap_stats->q_ipackets[i];
398 rx_bytes_total += tap_stats->q_ibytes[i];
400 tap_stats->q_opackets[i] = pmd->txq[i].stats.opackets;
401 tap_stats->q_errors[i] = pmd->txq[i].stats.errs;
402 tap_stats->q_obytes[i] = pmd->txq[i].stats.obytes;
403 tx_total += tap_stats->q_opackets[i];
404 tx_err_total += tap_stats->q_errors[i];
405 tx_bytes_total += tap_stats->q_obytes[i];
408 tap_stats->ipackets = rx_total;
409 tap_stats->ibytes = rx_bytes_total;
410 tap_stats->opackets = tx_total;
411 tap_stats->oerrors = tx_err_total;
412 tap_stats->obytes = tx_bytes_total;
416 tap_stats_reset(struct rte_eth_dev *dev)
419 struct pmd_internals *pmd = dev->data->dev_private;
421 for (i = 0; i < pmd->nb_queues; i++) {
422 pmd->rxq[i].stats.ipackets = 0;
423 pmd->rxq[i].stats.ibytes = 0;
425 pmd->txq[i].stats.opackets = 0;
426 pmd->txq[i].stats.errs = 0;
427 pmd->txq[i].stats.obytes = 0;
432 tap_dev_close(struct rte_eth_dev *dev __rte_unused)
437 tap_rx_queue_release(void *queue)
439 struct rx_queue *rxq = queue;
441 if (rxq && (rxq->fd > 0)) {
448 tap_tx_queue_release(void *queue)
450 struct tx_queue *txq = queue;
452 if (txq && (txq->fd > 0)) {
459 tap_link_update(struct rte_eth_dev *dev __rte_unused,
460 int wait_to_complete __rte_unused)
466 tap_promisc_enable(struct rte_eth_dev *dev)
468 struct pmd_internals *pmd = dev->data->dev_private;
470 dev->data->promiscuous = 1;
471 tap_link_set_flags(pmd, IFF_PROMISC, 1);
475 tap_promisc_disable(struct rte_eth_dev *dev)
477 struct pmd_internals *pmd = dev->data->dev_private;
479 dev->data->promiscuous = 0;
480 tap_link_set_flags(pmd, IFF_PROMISC, 0);
484 tap_allmulti_enable(struct rte_eth_dev *dev)
486 struct pmd_internals *pmd = dev->data->dev_private;
488 dev->data->all_multicast = 1;
489 tap_link_set_flags(pmd, IFF_ALLMULTI, 1);
493 tap_allmulti_disable(struct rte_eth_dev *dev)
495 struct pmd_internals *pmd = dev->data->dev_private;
497 dev->data->all_multicast = 0;
498 tap_link_set_flags(pmd, IFF_ALLMULTI, 0);
502 tap_setup_queue(struct rte_eth_dev *dev,
503 struct pmd_internals *internals,
506 struct pmd_internals *pmd = dev->data->dev_private;
507 struct rx_queue *rx = &internals->rxq[qid];
508 struct tx_queue *tx = &internals->txq[qid];
515 RTE_LOG(INFO, PMD, "Add queue to TAP %s for qid %d\n",
517 fd = tun_alloc(pmd, qid);
519 RTE_LOG(ERR, PMD, "tun_alloc(%s, %d) failed\n",
533 rx_setup_queue(struct rte_eth_dev *dev,
534 struct pmd_internals *internals,
537 dev->data->rx_queues[qid] = &internals->rxq[qid];
539 return tap_setup_queue(dev, internals, qid);
543 tx_setup_queue(struct rte_eth_dev *dev,
544 struct pmd_internals *internals,
547 dev->data->tx_queues[qid] = &internals->txq[qid];
549 return tap_setup_queue(dev, internals, qid);
553 tap_rx_queue_setup(struct rte_eth_dev *dev,
554 uint16_t rx_queue_id,
555 uint16_t nb_rx_desc __rte_unused,
556 unsigned int socket_id __rte_unused,
557 const struct rte_eth_rxconf *rx_conf __rte_unused,
558 struct rte_mempool *mp)
560 struct pmd_internals *internals = dev->data->dev_private;
564 if ((rx_queue_id >= internals->nb_queues) || !mp) {
565 RTE_LOG(WARNING, PMD,
566 "nb_queues %d too small or mempool NULL\n",
567 internals->nb_queues);
571 internals->rxq[rx_queue_id].mp = mp;
572 internals->rxq[rx_queue_id].in_port = dev->data->port_id;
574 /* Now get the space available for data in the mbuf */
575 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(mp) -
576 RTE_PKTMBUF_HEADROOM);
578 if (buf_size < ETH_FRAME_LEN) {
579 RTE_LOG(WARNING, PMD,
580 "%s: %d bytes will not fit in mbuf (%d bytes)\n",
581 dev->data->name, ETH_FRAME_LEN, buf_size);
585 fd = rx_setup_queue(dev, internals, rx_queue_id);
589 RTE_LOG(DEBUG, PMD, " RX TAP device name %s, qid %d on fd %d\n",
590 internals->name, rx_queue_id, internals->rxq[rx_queue_id].fd);
596 tap_tx_queue_setup(struct rte_eth_dev *dev,
597 uint16_t tx_queue_id,
598 uint16_t nb_tx_desc __rte_unused,
599 unsigned int socket_id __rte_unused,
600 const struct rte_eth_txconf *tx_conf __rte_unused)
602 struct pmd_internals *internals = dev->data->dev_private;
605 if (tx_queue_id >= internals->nb_queues)
608 ret = tx_setup_queue(dev, internals, tx_queue_id);
612 RTE_LOG(DEBUG, PMD, " TX TAP device name %s, qid %d on fd %d\n",
613 internals->name, tx_queue_id, internals->txq[tx_queue_id].fd);
618 static const struct eth_dev_ops ops = {
619 .dev_start = tap_dev_start,
620 .dev_stop = tap_dev_stop,
621 .dev_close = tap_dev_close,
622 .dev_configure = tap_dev_configure,
623 .dev_infos_get = tap_dev_info,
624 .rx_queue_setup = tap_rx_queue_setup,
625 .tx_queue_setup = tap_tx_queue_setup,
626 .rx_queue_release = tap_rx_queue_release,
627 .tx_queue_release = tap_tx_queue_release,
628 .link_update = tap_link_update,
629 .dev_set_link_up = tap_link_set_up,
630 .dev_set_link_down = tap_link_set_down,
631 .promiscuous_enable = tap_promisc_enable,
632 .promiscuous_disable = tap_promisc_disable,
633 .allmulticast_enable = tap_allmulti_enable,
634 .allmulticast_disable = tap_allmulti_disable,
635 .stats_get = tap_stats_get,
636 .stats_reset = tap_stats_reset,
640 eth_dev_tap_create(const char *name, char *tap_name)
642 int numa_node = rte_socket_id();
643 struct rte_eth_dev *dev = NULL;
644 struct pmd_internals *pmd = NULL;
645 struct rte_eth_dev_data *data = NULL;
648 RTE_LOG(DEBUG, PMD, " TAP device on numa %u\n", rte_socket_id());
650 data = rte_zmalloc_socket(tap_name, sizeof(*data), 0, numa_node);
652 RTE_LOG(ERR, PMD, "TAP Failed to allocate data\n");
656 pmd = rte_zmalloc_socket(tap_name, sizeof(*pmd), 0, numa_node);
658 RTE_LOG(ERR, PMD, "TAP Unable to allocate internal struct\n");
662 dev = rte_eth_dev_allocate(tap_name);
664 RTE_LOG(ERR, PMD, "TAP Unable to allocate device struct\n");
668 snprintf(pmd->name, sizeof(pmd->name), "%s", tap_name);
670 pmd->nb_queues = RTE_PMD_TAP_MAX_QUEUES;
672 /* Setup some default values */
673 data->dev_private = pmd;
674 data->port_id = dev->data->port_id;
675 data->dev_flags = RTE_ETH_DEV_DETACHABLE;
676 data->kdrv = RTE_KDRV_NONE;
677 data->drv_name = pmd_tap_drv.driver.name;
678 data->numa_node = numa_node;
680 data->dev_link = pmd_link;
681 data->mac_addrs = &pmd->eth_addr;
682 data->nb_rx_queues = pmd->nb_queues;
683 data->nb_tx_queues = pmd->nb_queues;
688 dev->rx_pkt_burst = pmd_rx_burst;
689 dev->tx_pkt_burst = pmd_tx_burst;
690 snprintf(dev->data->name, sizeof(dev->data->name), "%s", name);
692 /* Presetup the fds to -1 as being not valid */
693 for (i = 0; i < RTE_PMD_TAP_MAX_QUEUES; i++) {
701 RTE_LOG(DEBUG, PMD, "TAP Unable to initialize %s\n", name);
706 rte_eth_dev_release_port(dev);
712 set_interface_name(const char *key __rte_unused,
716 char *name = (char *)extra_args;
719 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s", value);
721 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s%d",
722 DEFAULT_TAP_NAME, (tap_unit - 1));
728 set_interface_speed(const char *key __rte_unused,
732 *(int *)extra_args = (value) ? atoi(value) : ETH_SPEED_NUM_10G;
737 /* Open a TAP interface device.
740 rte_pmd_tap_probe(const char *name, const char *params)
743 struct rte_kvargs *kvlist = NULL;
745 char tap_name[RTE_ETH_NAME_MAX_LEN];
747 speed = ETH_SPEED_NUM_10G;
748 snprintf(tap_name, sizeof(tap_name), "%s%d",
749 DEFAULT_TAP_NAME, tap_unit++);
751 if (params && (params[0] != '\0')) {
752 RTE_LOG(DEBUG, PMD, "paramaters (%s)\n", params);
754 kvlist = rte_kvargs_parse(params, valid_arguments);
756 if (rte_kvargs_count(kvlist, ETH_TAP_SPEED_ARG) == 1) {
757 ret = rte_kvargs_process(kvlist,
759 &set_interface_speed,
765 if (rte_kvargs_count(kvlist, ETH_TAP_IFACE_ARG) == 1) {
766 ret = rte_kvargs_process(kvlist,
775 pmd_link.link_speed = speed;
777 RTE_LOG(NOTICE, PMD, "Initializing pmd_tap for %s as %s\n",
780 ret = eth_dev_tap_create(name, tap_name);
784 RTE_LOG(ERR, PMD, "Failed to create pmd for %s as %s\n",
786 tap_unit--; /* Restore the unit number */
788 rte_kvargs_free(kvlist);
793 /* detach a TAP device.
796 rte_pmd_tap_remove(const char *name)
798 struct rte_eth_dev *eth_dev = NULL;
799 struct pmd_internals *internals;
802 RTE_LOG(DEBUG, PMD, "Closing TUN/TAP Ethernet device on numa %u\n",
805 /* find the ethdev entry */
806 eth_dev = rte_eth_dev_allocated(name);
810 internals = eth_dev->data->dev_private;
811 for (i = 0; i < internals->nb_queues; i++)
812 if (internals->rxq[i].fd != -1)
813 close(internals->rxq[i].fd);
815 rte_free(eth_dev->data->dev_private);
816 rte_free(eth_dev->data);
818 rte_eth_dev_release_port(eth_dev);
823 static struct rte_vdev_driver pmd_tap_drv = {
824 .probe = rte_pmd_tap_probe,
825 .remove = rte_pmd_tap_remove,
827 RTE_PMD_REGISTER_VDEV(net_tap, pmd_tap_drv);
828 RTE_PMD_REGISTER_ALIAS(net_tap, eth_tap);
829 RTE_PMD_REGISTER_PARAM_STRING(net_tap, "iface=<string>,speed=N");