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)
353 tap_link_set_down(dev);
357 tap_dev_configure(struct rte_eth_dev *dev __rte_unused)
363 tap_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
365 struct pmd_internals *internals = dev->data->dev_private;
367 dev_info->if_index = internals->if_index;
368 dev_info->max_mac_addrs = 1;
369 dev_info->max_rx_pktlen = (uint32_t)ETHER_MAX_VLAN_FRAME_LEN;
370 dev_info->max_rx_queues = internals->nb_queues;
371 dev_info->max_tx_queues = internals->nb_queues;
372 dev_info->min_rx_bufsize = 0;
373 dev_info->pci_dev = NULL;
377 tap_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *tap_stats)
379 unsigned int i, imax;
380 unsigned long rx_total = 0, tx_total = 0, tx_err_total = 0;
381 unsigned long rx_bytes_total = 0, tx_bytes_total = 0;
382 const struct pmd_internals *pmd = dev->data->dev_private;
384 imax = (pmd->nb_queues < RTE_ETHDEV_QUEUE_STAT_CNTRS) ?
385 pmd->nb_queues : RTE_ETHDEV_QUEUE_STAT_CNTRS;
387 for (i = 0; i < imax; i++) {
388 tap_stats->q_ipackets[i] = pmd->rxq[i].stats.ipackets;
389 tap_stats->q_ibytes[i] = pmd->rxq[i].stats.ibytes;
390 rx_total += tap_stats->q_ipackets[i];
391 rx_bytes_total += tap_stats->q_ibytes[i];
393 tap_stats->q_opackets[i] = pmd->txq[i].stats.opackets;
394 tap_stats->q_errors[i] = pmd->txq[i].stats.errs;
395 tap_stats->q_obytes[i] = pmd->txq[i].stats.obytes;
396 tx_total += tap_stats->q_opackets[i];
397 tx_err_total += tap_stats->q_errors[i];
398 tx_bytes_total += tap_stats->q_obytes[i];
401 tap_stats->ipackets = rx_total;
402 tap_stats->ibytes = rx_bytes_total;
403 tap_stats->opackets = tx_total;
404 tap_stats->oerrors = tx_err_total;
405 tap_stats->obytes = tx_bytes_total;
409 tap_stats_reset(struct rte_eth_dev *dev)
412 struct pmd_internals *pmd = dev->data->dev_private;
414 for (i = 0; i < pmd->nb_queues; i++) {
415 pmd->rxq[i].stats.ipackets = 0;
416 pmd->rxq[i].stats.ibytes = 0;
418 pmd->txq[i].stats.opackets = 0;
419 pmd->txq[i].stats.errs = 0;
420 pmd->txq[i].stats.obytes = 0;
425 tap_dev_close(struct rte_eth_dev *dev __rte_unused)
428 struct pmd_internals *internals = dev->data->dev_private;
430 tap_link_set_down(dev);
432 for (i = 0; i < internals->nb_queues; i++) {
433 if (internals->rxq[i].fd != -1)
434 close(internals->rxq[i].fd);
435 internals->rxq[i].fd = -1;
436 internals->txq[i].fd = -1;
441 tap_rx_queue_release(void *queue)
443 struct rx_queue *rxq = queue;
445 if (rxq && (rxq->fd > 0)) {
452 tap_tx_queue_release(void *queue)
454 struct tx_queue *txq = queue;
456 if (txq && (txq->fd > 0)) {
463 tap_link_update(struct rte_eth_dev *dev __rte_unused,
464 int wait_to_complete __rte_unused)
470 tap_promisc_enable(struct rte_eth_dev *dev)
472 struct pmd_internals *pmd = dev->data->dev_private;
474 dev->data->promiscuous = 1;
475 tap_link_set_flags(pmd, IFF_PROMISC, 1);
479 tap_promisc_disable(struct rte_eth_dev *dev)
481 struct pmd_internals *pmd = dev->data->dev_private;
483 dev->data->promiscuous = 0;
484 tap_link_set_flags(pmd, IFF_PROMISC, 0);
488 tap_allmulti_enable(struct rte_eth_dev *dev)
490 struct pmd_internals *pmd = dev->data->dev_private;
492 dev->data->all_multicast = 1;
493 tap_link_set_flags(pmd, IFF_ALLMULTI, 1);
497 tap_allmulti_disable(struct rte_eth_dev *dev)
499 struct pmd_internals *pmd = dev->data->dev_private;
501 dev->data->all_multicast = 0;
502 tap_link_set_flags(pmd, IFF_ALLMULTI, 0);
506 tap_setup_queue(struct rte_eth_dev *dev,
507 struct pmd_internals *internals,
510 struct pmd_internals *pmd = dev->data->dev_private;
511 struct rx_queue *rx = &internals->rxq[qid];
512 struct tx_queue *tx = &internals->txq[qid];
519 RTE_LOG(INFO, PMD, "Add queue to TAP %s for qid %d\n",
521 fd = tun_alloc(pmd, qid);
523 RTE_LOG(ERR, PMD, "tun_alloc(%s, %d) failed\n",
537 rx_setup_queue(struct rte_eth_dev *dev,
538 struct pmd_internals *internals,
541 dev->data->rx_queues[qid] = &internals->rxq[qid];
543 return tap_setup_queue(dev, internals, qid);
547 tx_setup_queue(struct rte_eth_dev *dev,
548 struct pmd_internals *internals,
551 dev->data->tx_queues[qid] = &internals->txq[qid];
553 return tap_setup_queue(dev, internals, qid);
557 tap_rx_queue_setup(struct rte_eth_dev *dev,
558 uint16_t rx_queue_id,
559 uint16_t nb_rx_desc __rte_unused,
560 unsigned int socket_id __rte_unused,
561 const struct rte_eth_rxconf *rx_conf __rte_unused,
562 struct rte_mempool *mp)
564 struct pmd_internals *internals = dev->data->dev_private;
568 if ((rx_queue_id >= internals->nb_queues) || !mp) {
569 RTE_LOG(WARNING, PMD,
570 "nb_queues %d too small or mempool NULL\n",
571 internals->nb_queues);
575 internals->rxq[rx_queue_id].mp = mp;
576 internals->rxq[rx_queue_id].in_port = dev->data->port_id;
578 /* Now get the space available for data in the mbuf */
579 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(mp) -
580 RTE_PKTMBUF_HEADROOM);
582 if (buf_size < ETH_FRAME_LEN) {
583 RTE_LOG(WARNING, PMD,
584 "%s: %d bytes will not fit in mbuf (%d bytes)\n",
585 dev->data->name, ETH_FRAME_LEN, buf_size);
589 fd = rx_setup_queue(dev, internals, rx_queue_id);
593 RTE_LOG(DEBUG, PMD, " RX TAP device name %s, qid %d on fd %d\n",
594 internals->name, rx_queue_id, internals->rxq[rx_queue_id].fd);
600 tap_tx_queue_setup(struct rte_eth_dev *dev,
601 uint16_t tx_queue_id,
602 uint16_t nb_tx_desc __rte_unused,
603 unsigned int socket_id __rte_unused,
604 const struct rte_eth_txconf *tx_conf __rte_unused)
606 struct pmd_internals *internals = dev->data->dev_private;
609 if (tx_queue_id >= internals->nb_queues)
612 ret = tx_setup_queue(dev, internals, tx_queue_id);
616 RTE_LOG(DEBUG, PMD, " TX TAP device name %s, qid %d on fd %d\n",
617 internals->name, tx_queue_id, internals->txq[tx_queue_id].fd);
622 static const struct eth_dev_ops ops = {
623 .dev_start = tap_dev_start,
624 .dev_stop = tap_dev_stop,
625 .dev_close = tap_dev_close,
626 .dev_configure = tap_dev_configure,
627 .dev_infos_get = tap_dev_info,
628 .rx_queue_setup = tap_rx_queue_setup,
629 .tx_queue_setup = tap_tx_queue_setup,
630 .rx_queue_release = tap_rx_queue_release,
631 .tx_queue_release = tap_tx_queue_release,
632 .link_update = tap_link_update,
633 .dev_set_link_up = tap_link_set_up,
634 .dev_set_link_down = tap_link_set_down,
635 .promiscuous_enable = tap_promisc_enable,
636 .promiscuous_disable = tap_promisc_disable,
637 .allmulticast_enable = tap_allmulti_enable,
638 .allmulticast_disable = tap_allmulti_disable,
639 .stats_get = tap_stats_get,
640 .stats_reset = tap_stats_reset,
644 eth_dev_tap_create(const char *name, char *tap_name)
646 int numa_node = rte_socket_id();
647 struct rte_eth_dev *dev = NULL;
648 struct pmd_internals *pmd = NULL;
649 struct rte_eth_dev_data *data = NULL;
652 RTE_LOG(DEBUG, PMD, " TAP device on numa %u\n", rte_socket_id());
654 data = rte_zmalloc_socket(tap_name, sizeof(*data), 0, numa_node);
656 RTE_LOG(ERR, PMD, "TAP Failed to allocate data\n");
660 pmd = rte_zmalloc_socket(tap_name, sizeof(*pmd), 0, numa_node);
662 RTE_LOG(ERR, PMD, "TAP Unable to allocate internal struct\n");
666 dev = rte_eth_dev_allocate(tap_name);
668 RTE_LOG(ERR, PMD, "TAP Unable to allocate device struct\n");
672 snprintf(pmd->name, sizeof(pmd->name), "%s", tap_name);
674 pmd->nb_queues = RTE_PMD_TAP_MAX_QUEUES;
676 /* Setup some default values */
677 data->dev_private = pmd;
678 data->port_id = dev->data->port_id;
679 data->dev_flags = RTE_ETH_DEV_DETACHABLE;
680 data->kdrv = RTE_KDRV_NONE;
681 data->drv_name = pmd_tap_drv.driver.name;
682 data->numa_node = numa_node;
684 data->dev_link = pmd_link;
685 data->mac_addrs = &pmd->eth_addr;
686 data->nb_rx_queues = pmd->nb_queues;
687 data->nb_tx_queues = pmd->nb_queues;
692 dev->rx_pkt_burst = pmd_rx_burst;
693 dev->tx_pkt_burst = pmd_tx_burst;
694 snprintf(dev->data->name, sizeof(dev->data->name), "%s", name);
696 /* Presetup the fds to -1 as being not valid */
697 for (i = 0; i < RTE_PMD_TAP_MAX_QUEUES; i++) {
705 RTE_LOG(DEBUG, PMD, "TAP Unable to initialize %s\n", name);
710 rte_eth_dev_release_port(dev);
716 set_interface_name(const char *key __rte_unused,
720 char *name = (char *)extra_args;
723 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s", value);
725 snprintf(name, RTE_ETH_NAME_MAX_LEN - 1, "%s%d",
726 DEFAULT_TAP_NAME, (tap_unit - 1));
732 set_interface_speed(const char *key __rte_unused,
736 *(int *)extra_args = (value) ? atoi(value) : ETH_SPEED_NUM_10G;
741 /* Open a TAP interface device.
744 rte_pmd_tap_probe(const char *name, const char *params)
747 struct rte_kvargs *kvlist = NULL;
749 char tap_name[RTE_ETH_NAME_MAX_LEN];
751 speed = ETH_SPEED_NUM_10G;
752 snprintf(tap_name, sizeof(tap_name), "%s%d",
753 DEFAULT_TAP_NAME, tap_unit++);
755 if (params && (params[0] != '\0')) {
756 RTE_LOG(DEBUG, PMD, "paramaters (%s)\n", params);
758 kvlist = rte_kvargs_parse(params, valid_arguments);
760 if (rte_kvargs_count(kvlist, ETH_TAP_SPEED_ARG) == 1) {
761 ret = rte_kvargs_process(kvlist,
763 &set_interface_speed,
769 if (rte_kvargs_count(kvlist, ETH_TAP_IFACE_ARG) == 1) {
770 ret = rte_kvargs_process(kvlist,
779 pmd_link.link_speed = speed;
781 RTE_LOG(NOTICE, PMD, "Initializing pmd_tap for %s as %s\n",
784 ret = eth_dev_tap_create(name, tap_name);
788 RTE_LOG(ERR, PMD, "Failed to create pmd for %s as %s\n",
790 tap_unit--; /* Restore the unit number */
792 rte_kvargs_free(kvlist);
797 /* detach a TAP device.
800 rte_pmd_tap_remove(const char *name)
802 struct rte_eth_dev *eth_dev = NULL;
803 struct pmd_internals *internals;
806 RTE_LOG(DEBUG, PMD, "Closing TUN/TAP Ethernet device on numa %u\n",
809 /* find the ethdev entry */
810 eth_dev = rte_eth_dev_allocated(name);
814 internals = eth_dev->data->dev_private;
815 for (i = 0; i < internals->nb_queues; i++)
816 if (internals->rxq[i].fd != -1)
817 close(internals->rxq[i].fd);
819 rte_free(eth_dev->data->dev_private);
820 rte_free(eth_dev->data);
822 rte_eth_dev_release_port(eth_dev);
827 static struct rte_vdev_driver pmd_tap_drv = {
828 .probe = rte_pmd_tap_probe,
829 .remove = rte_pmd_tap_remove,
831 RTE_PMD_REGISTER_VDEV(net_tap, pmd_tap_drv);
832 RTE_PMD_REGISTER_ALIAS(net_tap, eth_tap);
833 RTE_PMD_REGISTER_PARAM_STRING(net_tap, "iface=<string>,speed=N");