2 * Copyright (c) 2011, 2013 Gaetano Catalli.
3 * Copyright (c) 2013 YAMAMOTO Takashi.
5 * Licensed under the Apache License, Version 2.0 (the "License");
6 * you may not use this file except in compliance with the License.
7 * You may obtain a copy of the License at:
9 * http://www.apache.org/licenses/LICENSE-2.0
11 * Unless required by applicable law or agreed to in writing, software
12 * distributed under the License is distributed on an "AS IS" BASIS,
13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 * See the License for the specific language governing permissions and
15 * limitations under the License.
20 #include "netdev-provider.h"
24 #include <sys/types.h>
26 #include <sys/ioctl.h>
27 #include <sys/socket.h>
28 #include <sys/sockio.h>
30 #include <pcap/pcap.h>
32 #include <net/if_dl.h>
33 #include <net/if_media.h>
34 #include <net/if_tap.h>
35 #include <netinet/in.h>
36 #ifdef HAVE_NET_IF_MIB_H
37 #include <net/if_mib.h>
42 #include <sys/sysctl.h>
43 #if defined(__NetBSD__)
44 #include <net/route.h>
45 #include <netinet/in.h>
46 #include <netinet/if_inarp.h>
50 #include "connectivity.h"
52 #include "dynamic-string.h"
53 #include "fatal-signal.h"
55 #include "openflow/openflow.h"
56 #include "ovs-thread.h"
58 #include "poll-loop.h"
61 #include "socket-util.h"
66 VLOG_DEFINE_THIS_MODULE(netdev_bsd);
69 struct netdev_rx_bsd {
72 /* Packet capture descriptor for a system network device.
73 * For a tap device this is NULL. */
76 /* Selectable file descriptor for the network device.
77 * This descriptor will be used for polling operations. */
84 /* Never changes after initialization. */
87 /* Protects all members below. */
88 struct ovs_mutex mutex;
90 unsigned int cache_valid;
93 uint8_t etheraddr[ETH_ADDR_LEN];
95 struct in_addr netmask;
100 int tap_fd; /* TAP character device, if any, otherwise -1. */
102 /* Used for sending packets on non-tap devices. */
109 VALID_IFINDEX = 1 << 0,
110 VALID_ETHERADDR = 1 << 1,
114 VALID_CARRIER = 1 << 5
117 #define PCAP_SNAPLEN 2048
121 * Notifier used to invalidate device informations in case of status change.
123 * It will be registered with a 'rtbsd_notifier_register()' when the first
124 * device will be created with the call of either 'netdev_bsd_tap_create()' or
125 * 'netdev_bsd_system_create()'.
127 * The callback associated with this notifier ('netdev_bsd_cache_cb()') will
128 * invalidate cached information about the device.
130 static struct rtbsd_notifier netdev_bsd_cache_notifier;
131 static int cache_notifier_refcount;
133 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
135 static void destroy_tap(int fd, const char *name);
136 static int get_flags(const struct netdev *, int *flagsp);
137 static int set_flags(const char *, int flags);
138 static int do_set_addr(struct netdev *netdev,
139 unsigned long ioctl_nr, const char *ioctl_name,
140 struct in_addr addr);
141 static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]);
142 static int set_etheraddr(const char *netdev_name, int hwaddr_family,
143 int hwaddr_len, const uint8_t[ETH_ADDR_LEN]);
144 static int get_ifindex(const struct netdev *, int *ifindexp);
146 static int ifr_get_flags(const struct ifreq *);
147 static void ifr_set_flags(struct ifreq *, int flags);
150 static int af_link_ioctl(unsigned long command, const void *arg);
153 static void netdev_bsd_run(void);
156 is_netdev_bsd_class(const struct netdev_class *netdev_class)
158 return netdev_class->run == netdev_bsd_run;
161 static struct netdev_bsd *
162 netdev_bsd_cast(const struct netdev *netdev)
164 ovs_assert(is_netdev_bsd_class(netdev_get_class(netdev)));
165 return CONTAINER_OF(netdev, struct netdev_bsd, up);
168 static struct netdev_rx_bsd *
169 netdev_rx_bsd_cast(const struct netdev_rx *rx)
171 ovs_assert(is_netdev_bsd_class(netdev_get_class(rx->netdev)));
172 return CONTAINER_OF(rx, struct netdev_rx_bsd, up);
176 netdev_get_kernel_name(const struct netdev *netdev)
178 return netdev_bsd_cast(netdev)->kernel_name;
182 * Perform periodic work needed by netdev. In BSD netdevs it checks for any
183 * interface status changes, and eventually calls all the user callbacks.
188 rtbsd_notifier_run();
192 * Arranges for poll_block() to wake up if the "run" member function needs to
196 netdev_bsd_wait(void)
198 rtbsd_notifier_wait();
201 /* Invalidate cache in case of interface status change. */
203 netdev_bsd_cache_cb(const struct rtbsd_change *change,
204 void *aux OVS_UNUSED)
206 struct netdev_bsd *dev;
209 struct netdev *base_dev = netdev_from_name(change->if_name);
212 const struct netdev_class *netdev_class =
213 netdev_get_class(base_dev);
215 if (is_netdev_bsd_class(netdev_class)) {
216 dev = netdev_bsd_cast(base_dev);
217 dev->cache_valid = 0;
218 seq_change(connectivity_seq_get());
220 netdev_close(base_dev);
224 * XXX the API is lacking, we should be able to iterate on the list of
225 * netdevs without having to store the info in a temp shash.
227 struct shash device_shash;
228 struct shash_node *node;
230 shash_init(&device_shash);
231 netdev_get_devices(&netdev_bsd_class, &device_shash);
232 SHASH_FOR_EACH (node, &device_shash) {
233 struct netdev *netdev = node->data;
234 dev = netdev_bsd_cast(netdev);
235 dev->cache_valid = 0;
236 seq_change(connectivity_seq_get());
237 netdev_close(netdev);
239 shash_destroy(&device_shash);
244 cache_notifier_ref(void)
248 if (!cache_notifier_refcount) {
249 ret = rtbsd_notifier_register(&netdev_bsd_cache_notifier,
250 netdev_bsd_cache_cb, NULL);
255 cache_notifier_refcount++;
260 cache_notifier_unref(void)
262 cache_notifier_refcount--;
263 if (cache_notifier_refcount == 0) {
264 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier);
269 static struct netdev *
270 netdev_bsd_alloc(void)
272 struct netdev_bsd *netdev = xzalloc(sizeof *netdev);
277 netdev_bsd_construct_system(struct netdev *netdev_)
279 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
280 enum netdev_flags flags;
283 error = cache_notifier_ref();
288 ovs_mutex_init(&netdev->mutex);
290 netdev->kernel_name = xstrdup(netdev_->name);
292 /* Verify that the netdev really exists by attempting to read its flags */
293 error = netdev_get_flags(netdev_, &flags);
294 if (error == ENXIO) {
295 free(netdev->kernel_name);
296 cache_notifier_unref();
304 netdev_bsd_construct_tap(struct netdev *netdev_)
306 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
307 const char *name = netdev_->name;
310 char *kernel_name = NULL;
312 error = cache_notifier_ref();
317 memset(&ifr, 0, sizeof(ifr));
319 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
320 * to retrieve the name of the tap device. */
321 ovs_mutex_init(&netdev->mutex);
322 netdev->tap_fd = open("/dev/tap", O_RDWR);
323 if (netdev->tap_fd < 0) {
325 VLOG_WARN("opening \"/dev/tap\" failed: %s", ovs_strerror(error));
326 goto error_unref_notifier;
329 /* Retrieve tap name (e.g. tap0) */
330 if (ioctl(netdev->tap_fd, TAPGIFNAME, &ifr) == -1) {
331 /* XXX Need to destroy the device? */
333 close(netdev->tap_fd);
334 goto error_unref_notifier;
337 /* Change the name of the tap device */
338 #if defined(SIOCSIFNAME)
339 ifr.ifr_data = (void *)name;
340 error = af_inet_ioctl(SIOCSIFNAME, &ifr);
342 destroy_tap(netdev->tap_fd, ifr.ifr_name);
343 goto error_unref_notifier;
345 kernel_name = xstrdup(name);
348 * NetBSD doesn't support inteface renaming.
350 VLOG_INFO("tap %s is created for bridge %s", ifr.ifr_name, name);
351 kernel_name = xstrdup(ifr.ifr_name);
354 /* set non-blocking. */
355 error = set_nonblocking(netdev->tap_fd);
357 destroy_tap(netdev->tap_fd, kernel_name);
358 goto error_unref_notifier;
362 ifr_set_flags(&ifr, IFF_UP);
363 strncpy(ifr.ifr_name, kernel_name, sizeof ifr.ifr_name);
364 error = af_inet_ioctl(SIOCSIFFLAGS, &ifr);
366 destroy_tap(netdev->tap_fd, kernel_name);
367 goto error_unref_notifier;
370 netdev->kernel_name = kernel_name;
374 error_unref_notifier:
375 ovs_mutex_destroy(&netdev->mutex);
376 cache_notifier_unref();
383 netdev_bsd_destruct(struct netdev *netdev_)
385 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
387 cache_notifier_unref();
389 if (netdev->tap_fd >= 0) {
390 destroy_tap(netdev->tap_fd, netdev_get_kernel_name(netdev_));
393 pcap_close(netdev->pcap);
395 free(netdev->kernel_name);
396 ovs_mutex_destroy(&netdev->mutex);
400 netdev_bsd_dealloc(struct netdev *netdev_)
402 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
408 netdev_bsd_open_pcap(const char *name, pcap_t **pcapp, int *fdp)
410 char errbuf[PCAP_ERRBUF_SIZE];
416 /* Open the pcap device. The device is opened in non-promiscuous mode
417 * because the interface flags are manually set by the caller. */
419 pcap = pcap_open_live(name, PCAP_SNAPLEN, 0, 1000, errbuf);
421 VLOG_ERR_RL(&rl, "%s: pcap_open_live failed: %s", name, errbuf);
425 if (errbuf[0] != '\0') {
426 VLOG_WARN_RL(&rl, "%s: pcap_open_live: %s", name, errbuf);
429 /* Get the underlying fd. */
430 fd = pcap_get_selectable_fd(pcap);
432 VLOG_WARN_RL(&rl, "%s: no selectable file descriptor", name);
437 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
438 * on the file descriptor returned by pcap_get_selectable_fd to achieve
439 * a real non-blocking behaviour.*/
440 error = pcap_setnonblock(pcap, 1, errbuf);
446 /* This call assure that reads return immediately upon packet
447 * reception. Otherwise, a read will block until either the kernel
448 * buffer becomes full or a timeout occurs. */
449 if (ioctl(fd, BIOCIMMEDIATE, &one) < 0 ) {
450 VLOG_ERR_RL(&rl, "ioctl(BIOCIMMEDIATE) on %s device failed: %s",
451 name, ovs_strerror(errno));
456 /* Capture only incoming packets. */
457 error = pcap_setdirection(pcap, PCAP_D_IN);
476 static struct netdev_rx *
477 netdev_bsd_rx_alloc(void)
479 struct netdev_rx_bsd *rx = xzalloc(sizeof *rx);
484 netdev_bsd_rx_construct(struct netdev_rx *rx_)
486 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
487 struct netdev *netdev_ = rx->up.netdev;
488 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
491 if (!strcmp(netdev_get_type(netdev_), "tap")) {
492 rx->pcap_handle = NULL;
493 rx->fd = netdev->tap_fd;
496 ovs_mutex_lock(&netdev->mutex);
497 error = netdev_bsd_open_pcap(netdev_get_kernel_name(netdev_),
498 &rx->pcap_handle, &rx->fd);
499 ovs_mutex_unlock(&netdev->mutex);
506 netdev_bsd_rx_destruct(struct netdev_rx *rx_)
508 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
510 if (rx->pcap_handle) {
511 pcap_close(rx->pcap_handle);
516 netdev_bsd_rx_dealloc(struct netdev_rx *rx_)
518 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
523 /* The recv callback of the netdev class returns the number of bytes of the
526 * This can be done by the pcap_next() function. Unfortunately pcap_next() does
527 * not make difference between a missing packet on the capture interface and
528 * an error during the file capture. We can use the pcap_dispatch() function
529 * instead, which is able to distinguish between errors and null packet.
531 * To make pcap_dispatch() returns the number of bytes read from the interface
532 * we need to define the following callback and argument.
541 * This callback will be executed on every captured packet.
543 * If the packet captured by pcap_dispatch() does not fit the pcap buffer,
544 * pcap returns a truncated packet and we follow this behavior.
546 * The argument args->retval is the packet size in bytes.
549 proc_pkt(u_char *args_, const struct pcap_pkthdr *hdr, const u_char *packet)
551 struct pcap_arg *args = (struct pcap_arg *)args_;
553 if (args->size < hdr->len) {
554 VLOG_WARN_RL(&rl, "packet truncated");
555 args->retval = args->size;
557 args->retval = hdr->len;
560 /* copy the packet to our buffer */
561 memcpy(args->data, packet, args->retval);
565 * This function attempts to receive a packet from the specified network
566 * device. It is assumed that the network device is a system device or a tap
567 * device opened as a system one. In this case the read operation is performed
571 netdev_rx_bsd_recv_pcap(struct netdev_rx_bsd *rx, void *data, size_t size)
576 /* prepare the pcap argument to store the packet */
581 ret = pcap_dispatch(rx->pcap_handle, 1, proc_pkt, (u_char *) &arg);
584 return arg.retval; /* arg.retval < 0 is handled in the caller */
587 if (errno == EINTR) {
597 * This function attempts to receive a packet from the specified network
598 * device. It is assumed that the network device is a tap device and
599 * 'rx->fd' is initialized with the tap file descriptor.
602 netdev_rx_bsd_recv_tap(struct netdev_rx_bsd *rx, void *data, size_t size)
605 ssize_t retval = read(rx->fd, data, size);
608 } else if (errno != EINTR) {
609 if (errno != EAGAIN) {
610 VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
611 ovs_strerror(errno), netdev_rx_get_name(&rx->up));
619 netdev_bsd_rx_recv(struct netdev_rx *rx_, void *data, size_t size)
621 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
623 return (rx->pcap_handle
624 ? netdev_rx_bsd_recv_pcap(rx, data, size)
625 : netdev_rx_bsd_recv_tap(rx, data, size));
629 * Registers with the poll loop to wake up from the next call to poll_block()
630 * when a packet is ready to be received with netdev_rx_recv() on 'rx'.
633 netdev_bsd_rx_wait(struct netdev_rx *rx_)
635 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
637 poll_fd_wait(rx->fd, POLLIN);
640 /* Discards all packets waiting to be received from 'rx'. */
642 netdev_bsd_rx_drain(struct netdev_rx *rx_)
645 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
647 strcpy(ifr.ifr_name, netdev_get_kernel_name(netdev_rx_get_netdev(rx_)));
648 if (ioctl(rx->fd, BIOCFLUSH, &ifr) == -1) {
649 VLOG_DBG_RL(&rl, "%s: ioctl(BIOCFLUSH) failed: %s",
650 netdev_rx_get_name(rx_), ovs_strerror(errno));
657 * Send a packet on the specified network device. The device could be either a
658 * system or a tap device.
661 netdev_bsd_send(struct netdev *netdev_, const void *data, size_t size)
663 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
664 const char *name = netdev_get_name(netdev_);
667 ovs_mutex_lock(&dev->mutex);
668 if (dev->tap_fd < 0 && !dev->pcap) {
669 error = netdev_bsd_open_pcap(name, &dev->pcap, &dev->fd);
676 if (dev->tap_fd >= 0) {
677 retval = write(dev->tap_fd, data, size);
679 retval = pcap_inject(dev->pcap, data, size);
682 if (errno == EINTR) {
686 if (error != EAGAIN) {
687 VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: "
688 "%s", name, ovs_strerror(error));
691 } else if (retval != size) {
692 VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%"PRIuSIZE"d bytes of "
693 "%"PRIuSIZE") on %s", retval, size, name);
700 ovs_mutex_unlock(&dev->mutex);
705 * Registers with the poll loop to wake up from the next call to poll_block()
706 * when the packet transmission queue has sufficient room to transmit a packet
707 * with netdev_send().
710 netdev_bsd_send_wait(struct netdev *netdev_)
712 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
714 ovs_mutex_lock(&dev->mutex);
715 if (dev->tap_fd >= 0) {
716 /* TAP device always accepts packets. */
717 poll_immediate_wake();
718 } else if (dev->pcap) {
719 poll_fd_wait(dev->fd, POLLOUT);
721 /* We haven't even tried to send a packet yet. */
722 poll_immediate_wake();
724 ovs_mutex_unlock(&dev->mutex);
728 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
729 * otherwise a positive errno value.
732 netdev_bsd_set_etheraddr(struct netdev *netdev_,
733 const uint8_t mac[ETH_ADDR_LEN])
735 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
738 ovs_mutex_lock(&netdev->mutex);
739 if (!(netdev->cache_valid & VALID_ETHERADDR)
740 || !eth_addr_equals(netdev->etheraddr, mac)) {
741 error = set_etheraddr(netdev_get_kernel_name(netdev_), AF_LINK,
744 netdev->cache_valid |= VALID_ETHERADDR;
745 memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN);
746 seq_change(connectivity_seq_get());
749 ovs_mutex_unlock(&netdev->mutex);
755 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
756 * free the returned buffer.
759 netdev_bsd_get_etheraddr(const struct netdev *netdev_,
760 uint8_t mac[ETH_ADDR_LEN])
762 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
765 ovs_mutex_lock(&netdev->mutex);
766 if (!(netdev->cache_valid & VALID_ETHERADDR)) {
767 error = get_etheraddr(netdev_get_kernel_name(netdev_),
770 netdev->cache_valid |= VALID_ETHERADDR;
774 memcpy(mac, netdev->etheraddr, ETH_ADDR_LEN);
776 ovs_mutex_unlock(&netdev->mutex);
782 * Returns the maximum size of transmitted (and received) packets on 'netdev',
783 * in bytes, not including the hardware header; thus, this is typically 1500
784 * bytes for Ethernet devices.
787 netdev_bsd_get_mtu(const struct netdev *netdev_, int *mtup)
789 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
792 ovs_mutex_lock(&netdev->mutex);
793 if (!(netdev->cache_valid & VALID_MTU)) {
796 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
797 SIOCGIFMTU, "SIOCGIFMTU");
799 netdev->mtu = ifr.ifr_mtu;
800 netdev->cache_valid |= VALID_MTU;
806 ovs_mutex_unlock(&netdev->mutex);
812 netdev_bsd_get_ifindex(const struct netdev *netdev_)
814 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
817 ovs_mutex_lock(&netdev->mutex);
818 error = get_ifindex(netdev_, &ifindex);
819 ovs_mutex_unlock(&netdev->mutex);
821 return error ? -error : ifindex;
825 netdev_bsd_get_carrier(const struct netdev *netdev_, bool *carrier)
827 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
830 ovs_mutex_lock(&netdev->mutex);
831 if (!(netdev->cache_valid & VALID_CARRIER)) {
832 struct ifmediareq ifmr;
834 memset(&ifmr, 0, sizeof(ifmr));
835 strncpy(ifmr.ifm_name, netdev_get_kernel_name(netdev_),
836 sizeof ifmr.ifm_name);
838 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
840 netdev->carrier = (ifmr.ifm_status & IFM_ACTIVE) == IFM_ACTIVE;
841 netdev->cache_valid |= VALID_CARRIER;
843 /* If the interface doesn't report whether the media is active,
844 * just assume it is active. */
845 if ((ifmr.ifm_status & IFM_AVALID) == 0) {
846 netdev->carrier = true;
849 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
850 netdev_get_name(netdev_), ovs_strerror(error));
854 *carrier = netdev->carrier;
856 ovs_mutex_unlock(&netdev->mutex);
862 convert_stats(struct netdev_stats *stats, const struct if_data *ifd)
865 * note: UINT64_MAX means unsupported
867 stats->rx_packets = ifd->ifi_ipackets;
868 stats->tx_packets = ifd->ifi_opackets;
869 stats->rx_bytes = ifd->ifi_obytes;
870 stats->tx_bytes = ifd->ifi_ibytes;
871 stats->rx_errors = ifd->ifi_ierrors;
872 stats->tx_errors = ifd->ifi_oerrors;
873 stats->rx_dropped = ifd->ifi_iqdrops;
874 stats->tx_dropped = UINT64_MAX;
875 stats->multicast = ifd->ifi_imcasts;
876 stats->collisions = ifd->ifi_collisions;
877 stats->rx_length_errors = UINT64_MAX;
878 stats->rx_over_errors = UINT64_MAX;
879 stats->rx_crc_errors = UINT64_MAX;
880 stats->rx_frame_errors = UINT64_MAX;
881 stats->rx_fifo_errors = UINT64_MAX;
882 stats->rx_missed_errors = UINT64_MAX;
883 stats->tx_aborted_errors = UINT64_MAX;
884 stats->tx_carrier_errors = UINT64_MAX;
885 stats->tx_fifo_errors = UINT64_MAX;
886 stats->tx_heartbeat_errors = UINT64_MAX;
887 stats->tx_window_errors = UINT64_MAX;
890 /* Retrieves current device stats for 'netdev'. */
892 netdev_bsd_get_stats(const struct netdev *netdev_, struct netdev_stats *stats)
894 #if defined(__FreeBSD__)
898 struct ifmibdata ifmd;
903 mib[2] = NETLINK_GENERIC;
904 mib[3] = IFMIB_SYSTEM;
905 mib[4] = IFMIB_IFCOUNT;
907 len = sizeof(if_count);
909 if (sysctl(mib, 5, &if_count, &len, (void *)0, 0) == -1) {
910 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
911 netdev_get_name(netdev_), ovs_strerror(errno));
915 mib[5] = IFDATA_GENERAL;
916 mib[3] = IFMIB_IFDATA;
918 for (i = 1; i <= if_count; i++) {
920 if (sysctl(mib, 6, &ifmd, &len, (void *)0, 0) == -1) {
921 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
922 netdev_get_name(netdev_), ovs_strerror(errno));
924 } else if (!strcmp(ifmd.ifmd_name, netdev_get_name(netdev_))) {
925 convert_stats(stats, &ifmd.ifmd_data);
931 #elif defined(__NetBSD__)
932 struct ifdatareq ifdr;
935 memset(&ifdr, 0, sizeof(ifdr));
936 strncpy(ifdr.ifdr_name, netdev_get_kernel_name(netdev_),
937 sizeof(ifdr.ifdr_name));
938 error = af_link_ioctl(SIOCGIFDATA, &ifdr);
940 convert_stats(stats, &ifdr.ifdr_data);
944 #error not implemented
949 netdev_bsd_parse_media(int media)
951 uint32_t supported = 0;
952 bool half_duplex = media & IFM_HDX ? true : false;
954 switch (IFM_SUBTYPE(media)) {
959 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
960 supported |= NETDEV_F_COPPER;
964 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
965 supported |= NETDEV_F_FIBER;
972 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
973 supported |= NETDEV_F_COPPER;
977 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
978 supported |= NETDEV_F_FIBER;
983 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
984 supported |= NETDEV_F_COPPER;
989 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
990 supported |= NETDEV_F_FIBER;
994 supported |= NETDEV_F_10GB_FD;
995 supported |= NETDEV_F_COPPER;
1000 supported |= NETDEV_F_10GB_FD;
1001 supported |= NETDEV_F_FIBER;
1008 if (IFM_SUBTYPE(media) == IFM_AUTO) {
1009 supported |= NETDEV_F_AUTONEG;
1012 if (media & IFM_ETH_FMASK) {
1013 supported |= NETDEV_F_PAUSE;
1021 * Stores the features supported by 'netdev' into each of '*current',
1022 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1023 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1024 * successful, otherwise a positive errno value. On failure, all of the
1025 * passed-in values are set to 0.
1028 netdev_bsd_get_features(const struct netdev *netdev,
1029 enum netdev_features *current, uint32_t *advertised,
1030 enum netdev_features *supported, uint32_t *peer)
1032 struct ifmediareq ifmr;
1038 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1040 memset(&ifmr, 0, sizeof(ifmr));
1041 strncpy(ifmr.ifm_name, netdev_get_name(netdev), sizeof ifmr.ifm_name);
1043 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1044 * number of supported modes, and a second with a buffer to retrieve
1046 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
1048 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1049 netdev_get_name(netdev), ovs_strerror(error));
1053 media_list = xcalloc(ifmr.ifm_count, sizeof(int));
1054 ifmr.ifm_ulist = media_list;
1056 if (IFM_TYPE(ifmr.ifm_current) != IFM_ETHER) {
1057 VLOG_DBG_RL(&rl, "%s: doesn't appear to be ethernet",
1058 netdev_get_name(netdev));
1063 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
1065 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1066 netdev_get_name(netdev), ovs_strerror(error));
1070 /* Current settings. */
1071 *current = netdev_bsd_parse_media(ifmr.ifm_active);
1073 /* Advertised features. */
1074 *advertised = netdev_bsd_parse_media(ifmr.ifm_current);
1076 /* Supported features. */
1078 for (i = 0; i < ifmr.ifm_count; i++) {
1079 *supported |= netdev_bsd_parse_media(ifmr.ifm_ulist[i]);
1082 /* Peer advertisements. */
1083 *peer = 0; /* XXX */
1092 * If 'netdev' has an assigned IPv4 address, sets '*in4' to that address and
1093 * '*netmask' to its netmask and returns true. Otherwise, returns false.
1096 netdev_bsd_get_in4(const struct netdev *netdev_, struct in_addr *in4,
1097 struct in_addr *netmask)
1099 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1102 ovs_mutex_lock(&netdev->mutex);
1103 if (!(netdev->cache_valid & VALID_IN4)) {
1106 ifr.ifr_addr.sa_family = AF_INET;
1107 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1108 SIOCGIFADDR, "SIOCGIFADDR");
1110 const struct sockaddr_in *sin;
1112 sin = (struct sockaddr_in *) &ifr.ifr_addr;
1113 netdev->in4 = sin->sin_addr;
1114 netdev->cache_valid |= VALID_IN4;
1115 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1116 SIOCGIFNETMASK, "SIOCGIFNETMASK");
1118 *netmask = sin->sin_addr;
1124 *netmask = netdev->netmask;
1126 ovs_mutex_unlock(&netdev->mutex);
1128 return error ? error : in4->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0;
1132 * Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
1133 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
1134 * positive errno value.
1137 netdev_bsd_set_in4(struct netdev *netdev_, struct in_addr addr,
1138 struct in_addr mask)
1140 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1143 ovs_mutex_lock(&netdev->mutex);
1144 error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", addr);
1146 if (addr.s_addr != INADDR_ANY) {
1147 error = do_set_addr(netdev_, SIOCSIFNETMASK,
1148 "SIOCSIFNETMASK", mask);
1150 netdev->cache_valid |= VALID_IN4;
1152 netdev->netmask = mask;
1155 seq_change(connectivity_seq_get());
1157 ovs_mutex_unlock(&netdev->mutex);
1163 netdev_bsd_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
1165 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1166 if (!(netdev->cache_valid & VALID_IN6)) {
1167 struct ifaddrs *ifa, *head;
1168 struct sockaddr_in6 *sin6;
1169 const char *netdev_name = netdev_get_name(netdev_);
1171 if (getifaddrs(&head) != 0) {
1172 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1173 ovs_strerror(errno));
1177 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1178 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1179 !strcmp(ifa->ifa_name, netdev_name)) {
1180 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1182 memcpy(&netdev->in6, &sin6->sin6_addr, sin6->sin6_len);
1183 netdev->cache_valid |= VALID_IN6;
1190 return EADDRNOTAVAIL;
1196 #if defined(__NetBSD__)
1198 netdev_bsd_kernel_name_to_ovs_name(const char *kernel_name)
1200 char *ovs_name = NULL;
1201 struct shash device_shash;
1202 struct shash_node *node;
1204 shash_init(&device_shash);
1205 netdev_get_devices(&netdev_tap_class, &device_shash);
1206 SHASH_FOR_EACH(node, &device_shash) {
1207 struct netdev *netdev = node->data;
1208 struct netdev_bsd * const dev = netdev_bsd_cast(netdev);
1210 if (!strcmp(dev->kernel_name, kernel_name)) {
1212 ovs_name = xstrdup(netdev_get_name(&dev->up));
1214 netdev_close(netdev);
1216 shash_destroy(&device_shash);
1218 return ovs_name ? ovs_name : xstrdup(kernel_name);
1223 netdev_bsd_get_next_hop(const struct in_addr *host OVS_UNUSED,
1224 struct in_addr *next_hop OVS_UNUSED,
1225 char **netdev_name OVS_UNUSED)
1227 #if defined(__NetBSD__)
1229 struct sockaddr_in sin;
1230 struct sockaddr_dl sdl;
1237 struct rt_msghdr *rtm = &buf.h;
1238 const pid_t pid = getpid();
1241 bool gateway = false;
1242 char *ifname = NULL;
1245 memset(next_hop, 0, sizeof(*next_hop));
1246 *netdev_name = NULL;
1248 memset(&sin, 0, sizeof(sin));
1249 sin.sin_len = sizeof(sin);
1250 sin.sin_family = AF_INET;
1252 sin.sin_addr = *host;
1254 memset(&sdl, 0, sizeof(sdl));
1255 sdl.sdl_len = sizeof(sdl);
1256 sdl.sdl_family = AF_LINK;
1258 s = socket(PF_ROUTE, SOCK_RAW, 0);
1259 memset(&buf, 0, sizeof(buf));
1260 rtm->rtm_flags = RTF_HOST|RTF_UP;
1261 rtm->rtm_version = RTM_VERSION;
1262 rtm->rtm_addrs = RTA_DST|RTA_IFP;
1263 cp = (void *)&buf.space;
1264 memcpy(cp, &sin, sizeof(sin));
1265 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sin);
1266 memcpy(cp, &sdl, sizeof(sdl));
1267 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sdl);
1268 rtm->rtm_msglen = cp - (char *)(void *)rtm;
1269 rtm->rtm_seq = ++seq;
1270 rtm->rtm_type = RTM_GET;
1272 write(s, rtm, rtm->rtm_msglen);
1273 memset(&buf, 0, sizeof(buf));
1275 ssz = read(s, &buf, sizeof(buf));
1276 } while (ssz > 0 && (rtm->rtm_seq != seq || rtm->rtm_pid != pid));
1277 saved_errno = errno;
1283 return EPIPE; /* XXX */
1285 cp = (void *)&buf.space;
1286 for (i = 1; i; i <<= 1) {
1287 if ((rtm->rtm_addrs & i) != 0) {
1288 const struct sockaddr *sa = (const void *)cp;
1290 if ((i == RTA_GATEWAY) && sa->sa_family == AF_INET) {
1291 const struct sockaddr_in * const sin =
1292 (const struct sockaddr_in *)sa;
1294 *next_hop = sin->sin_addr;
1297 if ((i == RTA_IFP) && sa->sa_family == AF_LINK) {
1298 const struct sockaddr_dl * const sdl =
1299 (const struct sockaddr_dl *)sa;
1302 kernel_name = xmemdup0(sdl->sdl_data, sdl->sdl_nlen);
1303 ifname = netdev_bsd_kernel_name_to_ovs_name(kernel_name);
1309 if (ifname == NULL) {
1315 *netdev_name = ifname;
1316 VLOG_DBG("host " IP_FMT " next-hop " IP_FMT " if %s",
1317 IP_ARGS(host->s_addr), IP_ARGS(next_hop->s_addr), *netdev_name);
1325 netdev_bsd_arp_lookup(const struct netdev *netdev OVS_UNUSED,
1326 ovs_be32 ip OVS_UNUSED,
1327 uint8_t mac[ETH_ADDR_LEN] OVS_UNUSED)
1329 #if defined(__NetBSD__)
1330 const struct rt_msghdr *rtm;
1343 mib[4] = NET_RT_FLAGS;
1344 mib[5] = RTF_LLINFO;
1345 if (sysctl(mib, 6, NULL, &needed, NULL, 0) == -1) {
1349 buf = xmalloc(needed);
1350 if (sysctl(mib, 6, buf, &needed, NULL, 0) == -1) {
1355 for (cp = buf; cp < ep; cp += rtm->rtm_msglen) {
1356 const struct sockaddr_inarp *sina;
1357 const struct sockaddr_dl *sdl;
1359 rtm = (const void *)cp;
1360 sina = (const void *)(rtm + 1);
1361 if (ip != sina->sin_addr.s_addr) {
1364 sdl = (const void *)
1365 ((const char *)(const void *)sina + RT_ROUNDUP(sina->sin_len));
1366 if (sdl->sdl_alen == ETH_ADDR_LEN) {
1367 memcpy(mac, &sdl->sdl_data[sdl->sdl_nlen], ETH_ADDR_LEN);
1382 make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
1384 struct sockaddr_in sin;
1385 memset(&sin, 0, sizeof sin);
1386 sin.sin_family = AF_INET;
1387 sin.sin_addr = addr;
1390 memset(sa, 0, sizeof *sa);
1391 memcpy(sa, &sin, sizeof sin);
1395 do_set_addr(struct netdev *netdev,
1396 unsigned long ioctl_nr, const char *ioctl_name,
1397 struct in_addr addr)
1400 make_in4_sockaddr(&ifr.ifr_addr, addr);
1401 return af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev), &ifr, ioctl_nr,
1406 nd_to_iff_flags(enum netdev_flags nd)
1409 if (nd & NETDEV_UP) {
1412 if (nd & NETDEV_PROMISC) {
1414 #if defined(IFF_PPROMISC)
1415 iff |= IFF_PPROMISC;
1418 if (nd & NETDEV_LOOPBACK) {
1419 iff |= IFF_LOOPBACK;
1425 iff_to_nd_flags(int iff)
1427 enum netdev_flags nd = 0;
1431 if (iff & IFF_PROMISC) {
1432 nd |= NETDEV_PROMISC;
1434 if (iff & IFF_LOOPBACK) {
1435 nd |= NETDEV_LOOPBACK;
1441 netdev_bsd_update_flags(struct netdev *netdev_, enum netdev_flags off,
1442 enum netdev_flags on, enum netdev_flags *old_flagsp)
1444 int old_flags, new_flags;
1447 error = get_flags(netdev_, &old_flags);
1449 *old_flagsp = iff_to_nd_flags(old_flags);
1450 new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on);
1451 if (new_flags != old_flags) {
1452 error = set_flags(netdev_get_kernel_name(netdev_), new_flags);
1453 seq_change(connectivity_seq_get());
1460 const struct netdev_class netdev_bsd_class = {
1467 netdev_bsd_construct_system,
1468 netdev_bsd_destruct,
1470 NULL, /* get_config */
1471 NULL, /* set_config */
1472 NULL, /* get_tunnel_config */
1475 netdev_bsd_send_wait,
1477 netdev_bsd_set_etheraddr,
1478 netdev_bsd_get_etheraddr,
1481 netdev_bsd_get_ifindex,
1482 netdev_bsd_get_carrier,
1483 NULL, /* get_carrier_resets */
1484 NULL, /* set_miimon_interval */
1485 netdev_bsd_get_stats,
1486 NULL, /* set_stats */
1488 netdev_bsd_get_features,
1489 NULL, /* set_advertisement */
1490 NULL, /* set_policing */
1491 NULL, /* get_qos_type */
1492 NULL, /* get_qos_capabilities */
1495 NULL, /* get_queue */
1496 NULL, /* set_queue */
1497 NULL, /* delete_queue */
1498 NULL, /* get_queue_stats */
1499 NULL, /* queue_dump_start */
1500 NULL, /* queue_dump_next */
1501 NULL, /* queue_dump_done */
1502 NULL, /* dump_queue_stats */
1507 NULL, /* add_router */
1508 netdev_bsd_get_next_hop,
1509 NULL, /* get_status */
1510 netdev_bsd_arp_lookup, /* arp_lookup */
1512 netdev_bsd_update_flags,
1514 netdev_bsd_rx_alloc,
1515 netdev_bsd_rx_construct,
1516 netdev_bsd_rx_destruct,
1517 netdev_bsd_rx_dealloc,
1520 netdev_bsd_rx_drain,
1523 const struct netdev_class netdev_tap_class = {
1530 netdev_bsd_construct_tap,
1531 netdev_bsd_destruct,
1533 NULL, /* get_config */
1534 NULL, /* set_config */
1535 NULL, /* get_tunnel_config */
1538 netdev_bsd_send_wait,
1540 netdev_bsd_set_etheraddr,
1541 netdev_bsd_get_etheraddr,
1544 netdev_bsd_get_ifindex,
1545 netdev_bsd_get_carrier,
1546 NULL, /* get_carrier_resets */
1547 NULL, /* set_miimon_interval */
1548 netdev_bsd_get_stats,
1549 NULL, /* set_stats */
1551 netdev_bsd_get_features,
1552 NULL, /* set_advertisement */
1553 NULL, /* set_policing */
1554 NULL, /* get_qos_type */
1555 NULL, /* get_qos_capabilities */
1558 NULL, /* get_queue */
1559 NULL, /* set_queue */
1560 NULL, /* delete_queue */
1561 NULL, /* get_queue_stats */
1562 NULL, /* queue_dump_start */
1563 NULL, /* queue_dump_next */
1564 NULL, /* queue_dump_done */
1565 NULL, /* dump_queue_stats */
1570 NULL, /* add_router */
1571 netdev_bsd_get_next_hop,
1572 NULL, /* get_status */
1573 netdev_bsd_arp_lookup, /* arp_lookup */
1575 netdev_bsd_update_flags,
1577 netdev_bsd_rx_alloc,
1578 netdev_bsd_rx_construct,
1579 netdev_bsd_rx_destruct,
1580 netdev_bsd_rx_dealloc,
1583 netdev_bsd_rx_drain,
1588 destroy_tap(int fd, const char *name)
1593 strcpy(ifr.ifr_name, name);
1594 /* XXX What to do if this call fails? */
1595 af_inet_ioctl(SIOCIFDESTROY, &ifr);
1599 get_flags(const struct netdev *netdev, int *flags)
1604 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev), &ifr,
1605 SIOCGIFFLAGS, "SIOCGIFFLAGS");
1607 *flags = ifr_get_flags(&ifr);
1613 set_flags(const char *name, int flags)
1617 ifr_set_flags(&ifr, flags);
1619 return af_inet_ifreq_ioctl(name, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS");
1623 get_ifindex(const struct netdev *netdev_, int *ifindexp)
1625 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1627 if (!(netdev->cache_valid & VALID_IFINDEX)) {
1628 int ifindex = if_nametoindex(netdev_get_name(netdev_));
1632 netdev->cache_valid |= VALID_IFINDEX;
1633 netdev->ifindex = ifindex;
1635 *ifindexp = netdev->ifindex;
1640 get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN])
1642 struct ifaddrs *head;
1643 struct ifaddrs *ifa;
1644 struct sockaddr_dl *sdl;
1646 if (getifaddrs(&head) != 0) {
1647 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1648 ovs_strerror(errno));
1652 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1653 if (ifa->ifa_addr->sa_family == AF_LINK) {
1654 if (!strcmp(ifa->ifa_name, netdev_name)) {
1655 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1657 memcpy(ea, LLADDR(sdl), sdl->sdl_alen);
1665 VLOG_ERR("could not find ethernet address for %s device", netdev_name);
1671 set_etheraddr(const char *netdev_name OVS_UNUSED, int hwaddr_family OVS_UNUSED,
1672 int hwaddr_len OVS_UNUSED,
1673 const uint8_t mac[ETH_ADDR_LEN] OVS_UNUSED)
1675 #if defined(__FreeBSD__)
1679 memset(&ifr, 0, sizeof ifr);
1680 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1681 ifr.ifr_addr.sa_family = hwaddr_family;
1682 ifr.ifr_addr.sa_len = hwaddr_len;
1683 memcpy(ifr.ifr_addr.sa_data, mac, hwaddr_len);
1684 error = af_inet_ioctl(SIOCSIFLLADDR, &ifr);
1686 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1687 netdev_name, ovs_strerror(error));
1691 #elif defined(__NetBSD__)
1692 struct if_laddrreq req;
1693 struct sockaddr_dl *sdl;
1694 struct sockaddr_storage oldaddr;
1698 * get the old address, add new one, and then remove old one.
1701 if (hwaddr_len != ETH_ADDR_LEN) {
1702 /* just to be safe about sockaddr storage size */
1705 memset(&req, 0, sizeof(req));
1706 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1707 req.addr.ss_len = sizeof(req.addr);
1708 req.addr.ss_family = hwaddr_family;
1709 sdl = (struct sockaddr_dl *)&req.addr;
1710 sdl->sdl_alen = hwaddr_len;
1712 error = af_link_ioctl(SIOCGLIFADDR, &req);
1716 if (!memcmp(&sdl->sdl_data[sdl->sdl_nlen], mac, hwaddr_len)) {
1721 memset(&req, 0, sizeof(req));
1722 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1723 req.flags = IFLR_ACTIVE;
1724 sdl = (struct sockaddr_dl *)&req.addr;
1725 sdl->sdl_len = offsetof(struct sockaddr_dl, sdl_data) + hwaddr_len;
1726 sdl->sdl_alen = hwaddr_len;
1727 sdl->sdl_family = hwaddr_family;
1728 memcpy(sdl->sdl_data, mac, hwaddr_len);
1729 error = af_link_ioctl(SIOCALIFADDR, &req);
1734 memset(&req, 0, sizeof(req));
1735 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1737 return af_link_ioctl(SIOCDLIFADDR, &req);
1739 #error not implemented
1744 ifr_get_flags(const struct ifreq *ifr)
1746 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1747 return (ifr->ifr_flagshigh << 16) | ifr->ifr_flags;
1749 return ifr->ifr_flags;
1754 ifr_set_flags(struct ifreq *ifr, int flags)
1756 ifr->ifr_flags = flags;
1757 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1758 ifr->ifr_flagshigh = flags >> 16;
1762 /* Calls ioctl() on an AF_LINK sock, passing the specified 'command' and
1763 * 'arg'. Returns 0 if successful, otherwise a positive errno value. */
1765 af_link_ioctl(unsigned long command, const void *arg)
1767 static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
1770 if (ovsthread_once_start(&once)) {
1771 sock = socket(AF_LINK, SOCK_DGRAM, 0);
1774 VLOG_ERR("failed to create link socket: %s", ovs_strerror(errno));
1776 ovsthread_once_done(&once);
1779 return (sock < 0 ? -sock
1780 : ioctl(sock, command, arg) == -1 ? errno