2 * Copyright (c) 2011 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>
49 #include "dynamic-string.h"
50 #include "fatal-signal.h"
52 #include "openflow/openflow.h"
53 #include "ovs-thread.h"
55 #include "poll-loop.h"
56 #include "socket-util.h"
62 VLOG_DEFINE_THIS_MODULE(netdev_bsd);
65 struct netdev_rx_bsd {
68 /* Packet capture descriptor for a system network device.
69 * For a tap device this is NULL. */
72 /* Selectable file descriptor for the network device.
73 * This descriptor will be used for polling operations. */
80 /* Never changes after initialization. */
83 /* Protects all members below. */
84 struct ovs_mutex mutex;
86 unsigned int cache_valid;
87 unsigned int change_seq;
90 uint8_t etheraddr[ETH_ADDR_LEN];
92 struct in_addr netmask;
97 int tap_fd; /* TAP character device, if any, otherwise -1. */
99 /* Used for sending packets on non-tap devices. */
106 VALID_IFINDEX = 1 << 0,
107 VALID_ETHERADDR = 1 << 1,
111 VALID_CARRIER = 1 << 5
114 #define PCAP_SNAPLEN 2048
118 * Notifier used to invalidate device informations in case of status change.
120 * It will be registered with a 'rtbsd_notifier_register()' when the first
121 * device will be created with the call of either 'netdev_bsd_tap_create()' or
122 * 'netdev_bsd_system_create()'.
124 * The callback associated with this notifier ('netdev_bsd_cache_cb()') will
125 * invalidate cached information about the device.
127 static struct rtbsd_notifier netdev_bsd_cache_notifier;
128 static int cache_notifier_refcount;
130 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
132 static void destroy_tap(int fd, const char *name);
133 static int get_flags(const struct netdev *, int *flagsp);
134 static int set_flags(const char *, int flags);
135 static int do_set_addr(struct netdev *netdev,
136 int ioctl_nr, const char *ioctl_name,
137 struct in_addr addr);
138 static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]);
139 static int set_etheraddr(const char *netdev_name, int hwaddr_family,
140 int hwaddr_len, const uint8_t[ETH_ADDR_LEN]);
141 static int get_ifindex(const struct netdev *, int *ifindexp);
143 static int ifr_get_flags(const struct ifreq *);
144 static void ifr_set_flags(struct ifreq *, int flags);
147 static int af_link_ioctl(int command, const void *arg);
150 static void netdev_bsd_run(void);
153 is_netdev_bsd_class(const struct netdev_class *netdev_class)
155 return netdev_class->run == netdev_bsd_run;
158 static struct netdev_bsd *
159 netdev_bsd_cast(const struct netdev *netdev)
161 ovs_assert(is_netdev_bsd_class(netdev_get_class(netdev)));
162 return CONTAINER_OF(netdev, struct netdev_bsd, up);
165 static struct netdev_rx_bsd *
166 netdev_rx_bsd_cast(const struct netdev_rx *rx)
168 ovs_assert(is_netdev_bsd_class(netdev_get_class(rx->netdev)));
169 return CONTAINER_OF(rx, struct netdev_rx_bsd, up);
173 netdev_get_kernel_name(const struct netdev *netdev)
175 return netdev_bsd_cast(netdev)->kernel_name;
179 * Perform periodic work needed by netdev. In BSD netdevs it checks for any
180 * interface status changes, and eventually calls all the user callbacks.
185 rtbsd_notifier_run();
189 * Arranges for poll_block() to wake up if the "run" member function needs to
193 netdev_bsd_wait(void)
195 rtbsd_notifier_wait();
199 netdev_bsd_changed(struct netdev_bsd *dev)
202 if (!dev->change_seq) {
207 /* Invalidate cache in case of interface status change. */
209 netdev_bsd_cache_cb(const struct rtbsd_change *change,
210 void *aux OVS_UNUSED)
212 struct netdev_bsd *dev;
215 struct netdev *base_dev = netdev_from_name(change->if_name);
218 const struct netdev_class *netdev_class =
219 netdev_get_class(base_dev);
221 if (is_netdev_bsd_class(netdev_class)) {
222 dev = netdev_bsd_cast(base_dev);
223 dev->cache_valid = 0;
224 netdev_bsd_changed(dev);
226 netdev_close(base_dev);
230 * XXX the API is lacking, we should be able to iterate on the list of
231 * netdevs without having to store the info in a temp shash.
233 struct shash device_shash;
234 struct shash_node *node;
236 shash_init(&device_shash);
237 netdev_get_devices(&netdev_bsd_class, &device_shash);
238 SHASH_FOR_EACH (node, &device_shash) {
239 struct netdev *netdev = node->data;
240 dev = netdev_bsd_cast(netdev);
241 dev->cache_valid = 0;
242 netdev_bsd_changed(dev);
243 netdev_close(netdev);
245 shash_destroy(&device_shash);
250 cache_notifier_ref(void)
254 if (!cache_notifier_refcount) {
255 ret = rtbsd_notifier_register(&netdev_bsd_cache_notifier,
256 netdev_bsd_cache_cb, NULL);
261 cache_notifier_refcount++;
266 cache_notifier_unref(void)
268 cache_notifier_refcount--;
269 if (cache_notifier_refcount == 0) {
270 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier);
275 static struct netdev *
276 netdev_bsd_alloc(void)
278 struct netdev_bsd *netdev = xzalloc(sizeof *netdev);
283 netdev_bsd_construct_system(struct netdev *netdev_)
285 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
286 enum netdev_flags flags;
289 error = cache_notifier_ref();
294 ovs_mutex_init(&netdev->mutex, PTHREAD_MUTEX_NORMAL);
295 netdev->change_seq = 1;
297 netdev->kernel_name = xstrdup(netdev_->name);
299 /* Verify that the netdev really exists by attempting to read its flags */
300 error = netdev_get_flags(netdev_, &flags);
301 if (error == ENXIO) {
302 free(netdev->kernel_name);
303 cache_notifier_unref();
311 netdev_bsd_construct_tap(struct netdev *netdev_)
313 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
314 const char *name = netdev_->name;
317 char *kernel_name = NULL;
319 error = cache_notifier_ref();
324 memset(&ifr, 0, sizeof(ifr));
326 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
327 * to retrieve the name of the tap device. */
328 ovs_mutex_init(&netdev->mutex, PTHREAD_MUTEX_NORMAL);
329 netdev->tap_fd = open("/dev/tap", O_RDWR);
330 netdev->change_seq = 1;
331 if (netdev->tap_fd < 0) {
333 VLOG_WARN("opening \"/dev/tap\" failed: %s", ovs_strerror(error));
334 goto error_unref_notifier;
337 /* Retrieve tap name (e.g. tap0) */
338 if (ioctl(netdev->tap_fd, TAPGIFNAME, &ifr) == -1) {
339 /* XXX Need to destroy the device? */
341 close(netdev->tap_fd);
342 goto error_unref_notifier;
345 /* Change the name of the tap device */
346 #if defined(SIOCSIFNAME)
347 ifr.ifr_data = (void *)name;
348 error = af_inet_ioctl(SIOCSIFNAME, &ifr);
350 destroy_tap(netdev->tap_fd, ifr.ifr_name);
351 goto error_unref_notifier;
353 kernel_name = xstrdup(name);
356 * NetBSD doesn't support inteface renaming.
358 VLOG_INFO("tap %s is created for bridge %s", ifr.ifr_name, name);
359 kernel_name = xstrdup(ifr.ifr_name);
362 /* set non-blocking. */
363 error = set_nonblocking(netdev->tap_fd);
365 destroy_tap(netdev->tap_fd, kernel_name);
366 goto error_unref_notifier;
370 ifr_set_flags(&ifr, IFF_UP);
371 strncpy(ifr.ifr_name, kernel_name, sizeof ifr.ifr_name);
372 error = af_inet_ioctl(SIOCSIFFLAGS, &ifr);
374 destroy_tap(netdev->tap_fd, kernel_name);
375 goto error_unref_notifier;
378 netdev->kernel_name = kernel_name;
382 error_unref_notifier:
383 ovs_mutex_destroy(&netdev->mutex);
384 cache_notifier_unref();
391 netdev_bsd_destruct(struct netdev *netdev_)
393 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
395 cache_notifier_unref();
397 if (netdev->tap_fd >= 0) {
398 destroy_tap(netdev->tap_fd, netdev_get_kernel_name(netdev_));
401 pcap_close(netdev->pcap);
403 free(netdev->kernel_name);
404 ovs_mutex_destroy(&netdev->mutex);
408 netdev_bsd_dealloc(struct netdev *netdev_)
410 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
416 netdev_bsd_open_pcap(const char *name, pcap_t **pcapp, int *fdp)
418 char errbuf[PCAP_ERRBUF_SIZE];
424 /* Open the pcap device. The device is opened in non-promiscuous mode
425 * because the interface flags are manually set by the caller. */
427 pcap = pcap_open_live(name, PCAP_SNAPLEN, 0, 1000, errbuf);
429 VLOG_ERR_RL(&rl, "%s: pcap_open_live failed: %s", name, errbuf);
433 if (errbuf[0] != '\0') {
434 VLOG_WARN_RL(&rl, "%s: pcap_open_live: %s", name, errbuf);
437 /* Get the underlying fd. */
438 fd = pcap_get_selectable_fd(pcap);
440 VLOG_WARN_RL(&rl, "%s: no selectable file descriptor", name);
445 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
446 * on the file descriptor returned by pcap_get_selectable_fd to achieve
447 * a real non-blocking behaviour.*/
448 error = pcap_setnonblock(pcap, 1, errbuf);
454 /* This call assure that reads return immediately upon packet
455 * reception. Otherwise, a read will block until either the kernel
456 * buffer becomes full or a timeout occurs. */
457 if (ioctl(fd, BIOCIMMEDIATE, &one) < 0 ) {
458 VLOG_ERR_RL(&rl, "ioctl(BIOCIMMEDIATE) on %s device failed: %s",
459 name, ovs_strerror(errno));
464 /* Capture only incoming packets. */
465 error = pcap_setdirection(pcap, PCAP_D_IN);
484 static struct netdev_rx *
485 netdev_bsd_rx_alloc(void)
487 struct netdev_rx_bsd *rx = xzalloc(sizeof *rx);
492 netdev_bsd_rx_construct(struct netdev_rx *rx_)
494 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
495 struct netdev *netdev_ = rx->up.netdev;
496 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
499 if (!strcmp(netdev_get_type(netdev_), "tap")) {
500 rx->pcap_handle = NULL;
501 rx->fd = netdev->tap_fd;
504 ovs_mutex_lock(&netdev->mutex);
505 error = netdev_bsd_open_pcap(netdev_get_kernel_name(netdev_),
506 &rx->pcap_handle, &rx->fd);
508 netdev_bsd_changed(netdev);
510 ovs_mutex_unlock(&netdev->mutex);
517 netdev_bsd_rx_destruct(struct netdev_rx *rx_)
519 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
521 if (rx->pcap_handle) {
522 pcap_close(rx->pcap_handle);
527 netdev_bsd_rx_dealloc(struct netdev_rx *rx_)
529 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
534 /* The recv callback of the netdev class returns the number of bytes of the
537 * This can be done by the pcap_next() function. Unfortunately pcap_next() does
538 * not make difference between a missing packet on the capture interface and
539 * an error during the file capture. We can use the pcap_dispatch() function
540 * instead, which is able to distinguish between errors and null packet.
542 * To make pcap_dispatch() returns the number of bytes read from the interface
543 * we need to define the following callback and argument.
552 * This callback will be executed on every captured packet.
554 * If the packet captured by pcap_dispatch() does not fit the pcap buffer,
555 * pcap returns a truncated packet and we follow this behavior.
557 * The argument args->retval is the packet size in bytes.
560 proc_pkt(u_char *args_, const struct pcap_pkthdr *hdr, const u_char *packet)
562 struct pcap_arg *args = (struct pcap_arg *)args_;
564 if (args->size < hdr->len) {
565 VLOG_WARN_RL(&rl, "packet truncated");
566 args->retval = args->size;
568 args->retval = hdr->len;
571 /* copy the packet to our buffer */
572 memcpy(args->data, packet, args->retval);
576 * This function attempts to receive a packet from the specified network
577 * device. It is assumed that the network device is a system device or a tap
578 * device opened as a system one. In this case the read operation is performed
582 netdev_rx_bsd_recv_pcap(struct netdev_rx_bsd *rx, void *data, size_t size)
587 /* prepare the pcap argument to store the packet */
592 ret = pcap_dispatch(rx->pcap_handle, 1, proc_pkt, (u_char *) &arg);
595 return arg.retval; /* arg.retval < 0 is handled in the caller */
598 if (errno == EINTR) {
608 * This function attempts to receive a packet from the specified network
609 * device. It is assumed that the network device is a tap device and
610 * 'rx->fd' is initialized with the tap file descriptor.
613 netdev_rx_bsd_recv_tap(struct netdev_rx_bsd *rx, void *data, size_t size)
616 ssize_t retval = read(rx->fd, data, size);
619 } else if (errno != EINTR) {
620 if (errno != EAGAIN) {
621 VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
622 ovs_strerror(errno), netdev_rx_get_name(&rx->up));
630 netdev_bsd_rx_recv(struct netdev_rx *rx_, void *data, size_t size)
632 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
634 return (rx->pcap_handle
635 ? netdev_rx_bsd_recv_pcap(rx, data, size)
636 : netdev_rx_bsd_recv_tap(rx, data, size));
640 * Registers with the poll loop to wake up from the next call to poll_block()
641 * when a packet is ready to be received with netdev_rx_recv() on 'rx'.
644 netdev_bsd_rx_wait(struct netdev_rx *rx_)
646 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
648 poll_fd_wait(rx->fd, POLLIN);
651 /* Discards all packets waiting to be received from 'rx'. */
653 netdev_bsd_rx_drain(struct netdev_rx *rx_)
656 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
658 strcpy(ifr.ifr_name, netdev_get_kernel_name(netdev_rx_get_netdev(rx_)));
659 if (ioctl(rx->fd, BIOCFLUSH, &ifr) == -1) {
660 VLOG_DBG_RL(&rl, "%s: ioctl(BIOCFLUSH) failed: %s",
661 netdev_rx_get_name(rx_), ovs_strerror(errno));
668 * Send a packet on the specified network device. The device could be either a
669 * system or a tap device.
672 netdev_bsd_send(struct netdev *netdev_, const void *data, size_t size)
674 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
675 const char *name = netdev_get_name(netdev_);
678 ovs_mutex_lock(&dev->mutex);
679 if (dev->tap_fd < 0 && !dev->pcap) {
680 error = netdev_bsd_open_pcap(name, &dev->pcap, &dev->fd);
687 if (dev->tap_fd >= 0) {
688 retval = write(dev->tap_fd, data, size);
690 retval = pcap_inject(dev->pcap, data, size);
693 if (errno == EINTR) {
697 if (error != EAGAIN) {
698 VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: "
699 "%s", name, ovs_strerror(error));
702 } else if (retval != size) {
703 VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of "
704 "%zu) on %s", retval, size, name);
711 ovs_mutex_unlock(&dev->mutex);
716 * Registers with the poll loop to wake up from the next call to poll_block()
717 * when the packet transmission queue has sufficient room to transmit a packet
718 * with netdev_send().
721 netdev_bsd_send_wait(struct netdev *netdev_)
723 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
725 ovs_mutex_lock(&dev->mutex);
726 if (dev->tap_fd >= 0) {
727 /* TAP device always accepts packets. */
728 poll_immediate_wake();
729 } else if (dev->pcap) {
730 poll_fd_wait(dev->fd, POLLOUT);
732 /* We haven't even tried to send a packet yet. */
733 poll_immediate_wake();
735 ovs_mutex_unlock(&dev->mutex);
739 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
740 * otherwise a positive errno value.
743 netdev_bsd_set_etheraddr(struct netdev *netdev_,
744 const uint8_t mac[ETH_ADDR_LEN])
746 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
749 ovs_mutex_lock(&netdev->mutex);
750 if (!(netdev->cache_valid & VALID_ETHERADDR)
751 || !eth_addr_equals(netdev->etheraddr, mac)) {
752 error = set_etheraddr(netdev_get_kernel_name(netdev_), AF_LINK,
755 netdev->cache_valid |= VALID_ETHERADDR;
756 memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN);
757 netdev_bsd_changed(netdev);
760 ovs_mutex_unlock(&netdev->mutex);
766 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
767 * free the returned buffer.
770 netdev_bsd_get_etheraddr(const struct netdev *netdev_,
771 uint8_t mac[ETH_ADDR_LEN])
773 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
776 ovs_mutex_lock(&netdev->mutex);
777 if (!(netdev->cache_valid & VALID_ETHERADDR)) {
778 error = get_etheraddr(netdev_get_kernel_name(netdev_),
781 netdev->cache_valid |= VALID_ETHERADDR;
785 memcpy(mac, netdev->etheraddr, ETH_ADDR_LEN);
787 ovs_mutex_unlock(&netdev->mutex);
793 * Returns the maximum size of transmitted (and received) packets on 'netdev',
794 * in bytes, not including the hardware header; thus, this is typically 1500
795 * bytes for Ethernet devices.
798 netdev_bsd_get_mtu(const struct netdev *netdev_, int *mtup)
800 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
803 ovs_mutex_lock(&netdev->mutex);
804 if (!(netdev->cache_valid & VALID_MTU)) {
807 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
808 SIOCGIFMTU, "SIOCGIFMTU");
810 netdev->mtu = ifr.ifr_mtu;
811 netdev->cache_valid |= VALID_MTU;
817 ovs_mutex_unlock(&netdev->mutex);
823 netdev_bsd_get_ifindex(const struct netdev *netdev_)
825 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
828 ovs_mutex_lock(&netdev->mutex);
829 error = get_ifindex(netdev_, &ifindex);
830 ovs_mutex_unlock(&netdev->mutex);
832 return error ? -error : ifindex;
836 netdev_bsd_get_carrier(const struct netdev *netdev_, bool *carrier)
838 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
841 ovs_mutex_lock(&netdev->mutex);
842 if (!(netdev->cache_valid & VALID_CARRIER)) {
843 struct ifmediareq ifmr;
845 memset(&ifmr, 0, sizeof(ifmr));
846 strncpy(ifmr.ifm_name, netdev_get_kernel_name(netdev_),
847 sizeof ifmr.ifm_name);
849 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
851 netdev->carrier = (ifmr.ifm_status & IFM_ACTIVE) == IFM_ACTIVE;
852 netdev->cache_valid |= VALID_CARRIER;
854 /* If the interface doesn't report whether the media is active,
855 * just assume it is active. */
856 if ((ifmr.ifm_status & IFM_AVALID) == 0) {
857 netdev->carrier = true;
860 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
861 netdev_get_name(netdev_), ovs_strerror(error));
865 *carrier = netdev->carrier;
867 ovs_mutex_unlock(&netdev->mutex);
873 convert_stats(struct netdev_stats *stats, const struct if_data *ifd)
876 * note: UINT64_MAX means unsupported
878 stats->rx_packets = ifd->ifi_ipackets;
879 stats->tx_packets = ifd->ifi_opackets;
880 stats->rx_bytes = ifd->ifi_obytes;
881 stats->tx_bytes = ifd->ifi_ibytes;
882 stats->rx_errors = ifd->ifi_ierrors;
883 stats->tx_errors = ifd->ifi_oerrors;
884 stats->rx_dropped = ifd->ifi_iqdrops;
885 stats->tx_dropped = UINT64_MAX;
886 stats->multicast = ifd->ifi_imcasts;
887 stats->collisions = ifd->ifi_collisions;
888 stats->rx_length_errors = UINT64_MAX;
889 stats->rx_over_errors = UINT64_MAX;
890 stats->rx_crc_errors = UINT64_MAX;
891 stats->rx_frame_errors = UINT64_MAX;
892 stats->rx_fifo_errors = UINT64_MAX;
893 stats->rx_missed_errors = UINT64_MAX;
894 stats->tx_aborted_errors = UINT64_MAX;
895 stats->tx_carrier_errors = UINT64_MAX;
896 stats->tx_fifo_errors = UINT64_MAX;
897 stats->tx_heartbeat_errors = UINT64_MAX;
898 stats->tx_window_errors = UINT64_MAX;
901 /* Retrieves current device stats for 'netdev'. */
903 netdev_bsd_get_stats(const struct netdev *netdev_, struct netdev_stats *stats)
905 #if defined(__FreeBSD__)
909 struct ifmibdata ifmd;
914 mib[2] = NETLINK_GENERIC;
915 mib[3] = IFMIB_SYSTEM;
916 mib[4] = IFMIB_IFCOUNT;
918 len = sizeof(if_count);
920 if (sysctl(mib, 5, &if_count, &len, (void *)0, 0) == -1) {
921 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
922 netdev_get_name(netdev_), ovs_strerror(errno));
926 mib[5] = IFDATA_GENERAL;
927 mib[3] = IFMIB_IFDATA;
929 for (i = 1; i <= if_count; i++) {
931 if (sysctl(mib, 6, &ifmd, &len, (void *)0, 0) == -1) {
932 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
933 netdev_get_name(netdev_), ovs_strerror(errno));
935 } else if (!strcmp(ifmd.ifmd_name, netdev_get_name(netdev_))) {
936 convert_stats(stats, &ifmd.ifmd_data);
942 #elif defined(__NetBSD__)
943 struct ifdatareq ifdr;
946 memset(&ifdr, 0, sizeof(ifdr));
947 strncpy(ifdr.ifdr_name, netdev_get_kernel_name(netdev_),
948 sizeof(ifdr.ifdr_name));
949 error = af_link_ioctl(SIOCGIFDATA, &ifdr);
951 convert_stats(stats, &ifdr.ifdr_data);
955 #error not implemented
960 netdev_bsd_parse_media(int media)
962 uint32_t supported = 0;
963 bool half_duplex = media & IFM_HDX ? true : false;
965 switch (IFM_SUBTYPE(media)) {
970 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
971 supported |= NETDEV_F_COPPER;
975 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
976 supported |= NETDEV_F_FIBER;
983 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
984 supported |= NETDEV_F_COPPER;
988 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
989 supported |= NETDEV_F_FIBER;
994 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
995 supported |= NETDEV_F_COPPER;
1000 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
1001 supported |= NETDEV_F_FIBER;
1005 supported |= NETDEV_F_10GB_FD;
1006 supported |= NETDEV_F_COPPER;
1011 supported |= NETDEV_F_10GB_FD;
1012 supported |= NETDEV_F_FIBER;
1019 if (IFM_SUBTYPE(media) == IFM_AUTO) {
1020 supported |= NETDEV_F_AUTONEG;
1023 if (media & IFM_ETH_FMASK) {
1024 supported |= NETDEV_F_PAUSE;
1032 * Stores the features supported by 'netdev' into each of '*current',
1033 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1034 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1035 * successful, otherwise a positive errno value. On failure, all of the
1036 * passed-in values are set to 0.
1039 netdev_bsd_get_features(const struct netdev *netdev,
1040 enum netdev_features *current, uint32_t *advertised,
1041 enum netdev_features *supported, uint32_t *peer)
1043 struct ifmediareq ifmr;
1049 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1051 memset(&ifmr, 0, sizeof(ifmr));
1052 strncpy(ifmr.ifm_name, netdev_get_name(netdev), sizeof ifmr.ifm_name);
1054 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1055 * number of supported modes, and a second with a buffer to retrieve
1057 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
1059 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1060 netdev_get_name(netdev), ovs_strerror(error));
1064 media_list = xcalloc(ifmr.ifm_count, sizeof(int));
1065 ifmr.ifm_ulist = media_list;
1067 if (IFM_TYPE(ifmr.ifm_current) != IFM_ETHER) {
1068 VLOG_DBG_RL(&rl, "%s: doesn't appear to be ethernet",
1069 netdev_get_name(netdev));
1074 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
1076 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1077 netdev_get_name(netdev), ovs_strerror(error));
1081 /* Current settings. */
1082 *current = netdev_bsd_parse_media(ifmr.ifm_active);
1084 /* Advertised features. */
1085 *advertised = netdev_bsd_parse_media(ifmr.ifm_current);
1087 /* Supported features. */
1089 for (i = 0; i < ifmr.ifm_count; i++) {
1090 *supported |= netdev_bsd_parse_media(ifmr.ifm_ulist[i]);
1093 /* Peer advertisements. */
1094 *peer = 0; /* XXX */
1103 * If 'netdev' has an assigned IPv4 address, sets '*in4' to that address and
1104 * '*netmask' to its netmask and returns true. Otherwise, returns false.
1107 netdev_bsd_get_in4(const struct netdev *netdev_, struct in_addr *in4,
1108 struct in_addr *netmask)
1110 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1113 ovs_mutex_lock(&netdev->mutex);
1114 if (!(netdev->cache_valid & VALID_IN4)) {
1117 ifr.ifr_addr.sa_family = AF_INET;
1118 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1119 SIOCGIFADDR, "SIOCGIFADDR");
1121 const struct sockaddr_in *sin;
1123 sin = (struct sockaddr_in *) &ifr.ifr_addr;
1124 netdev->in4 = sin->sin_addr;
1125 netdev->cache_valid |= VALID_IN4;
1126 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1127 SIOCGIFNETMASK, "SIOCGIFNETMASK");
1129 *netmask = sin->sin_addr;
1135 *netmask = netdev->netmask;
1137 ovs_mutex_unlock(&netdev->mutex);
1139 return error ? error : in4->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0;
1143 * Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
1144 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
1145 * positive errno value.
1148 netdev_bsd_set_in4(struct netdev *netdev_, struct in_addr addr,
1149 struct in_addr mask)
1151 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1154 ovs_mutex_lock(&netdev->mutex);
1155 error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", addr);
1157 if (addr.s_addr != INADDR_ANY) {
1158 error = do_set_addr(netdev_, SIOCSIFNETMASK,
1159 "SIOCSIFNETMASK", mask);
1161 netdev->cache_valid |= VALID_IN4;
1163 netdev->netmask = mask;
1166 netdev_bsd_changed(netdev);
1168 ovs_mutex_unlock(&netdev->mutex);
1174 netdev_bsd_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
1176 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1177 if (!(netdev->cache_valid & VALID_IN6)) {
1178 struct ifaddrs *ifa, *head;
1179 struct sockaddr_in6 *sin6;
1180 const char *netdev_name = netdev_get_name(netdev_);
1182 if (getifaddrs(&head) != 0) {
1183 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1184 ovs_strerror(errno));
1188 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1189 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1190 !strcmp(ifa->ifa_name, netdev_name)) {
1191 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1193 memcpy(&netdev->in6, &sin6->sin6_addr, sin6->sin6_len);
1194 netdev->cache_valid |= VALID_IN6;
1201 return EADDRNOTAVAIL;
1207 #if defined(__NetBSD__)
1209 netdev_bsd_kernel_name_to_ovs_name(const char *kernel_name)
1211 char *ovs_name = NULL;
1212 struct shash device_shash;
1213 struct shash_node *node;
1215 shash_init(&device_shash);
1216 netdev_get_devices(&netdev_tap_class, &device_shash);
1217 SHASH_FOR_EACH(node, &device_shash) {
1218 struct netdev *netdev = node->data;
1219 struct netdev_bsd * const dev = netdev_bsd_cast(netdev);
1221 if (!strcmp(dev->kernel_name, kernel_name)) {
1223 ovs_name = xstrdup(netdev_get_name(&dev->up));
1225 netdev_close(netdev);
1227 shash_destroy(&device_shash);
1229 return ovs_name ? ovs_name : xstrdup(kernel_name);
1234 netdev_bsd_get_next_hop(const struct in_addr *host OVS_UNUSED,
1235 struct in_addr *next_hop OVS_UNUSED,
1236 char **netdev_name OVS_UNUSED)
1238 #if defined(__NetBSD__)
1240 struct sockaddr_in sin;
1241 struct sockaddr_dl sdl;
1248 struct rt_msghdr *rtm = &buf.h;
1249 const pid_t pid = getpid();
1252 bool gateway = false;
1253 char *ifname = NULL;
1256 memset(next_hop, 0, sizeof(*next_hop));
1257 *netdev_name = NULL;
1259 memset(&sin, 0, sizeof(sin));
1260 sin.sin_len = sizeof(sin);
1261 sin.sin_family = AF_INET;
1263 sin.sin_addr = *host;
1265 memset(&sdl, 0, sizeof(sdl));
1266 sdl.sdl_len = sizeof(sdl);
1267 sdl.sdl_family = AF_LINK;
1269 s = socket(PF_ROUTE, SOCK_RAW, 0);
1270 memset(&buf, 0, sizeof(buf));
1271 rtm->rtm_flags = RTF_HOST|RTF_UP;
1272 rtm->rtm_version = RTM_VERSION;
1273 rtm->rtm_addrs = RTA_DST|RTA_IFP;
1274 cp = (void *)&buf.space;
1275 memcpy(cp, &sin, sizeof(sin));
1276 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sin);
1277 memcpy(cp, &sdl, sizeof(sdl));
1278 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sdl);
1279 rtm->rtm_msglen = cp - (char *)(void *)rtm;
1280 rtm->rtm_seq = ++seq;
1281 rtm->rtm_type = RTM_GET;
1283 write(s, rtm, rtm->rtm_msglen);
1284 memset(&buf, 0, sizeof(buf));
1286 ssz = read(s, &buf, sizeof(buf));
1287 } while (ssz > 0 && (rtm->rtm_seq != seq || rtm->rtm_pid != pid));
1288 saved_errno = errno;
1294 return EPIPE; /* XXX */
1296 cp = (void *)&buf.space;
1297 for (i = 1; i; i <<= 1) {
1298 if ((rtm->rtm_addrs & i) != 0) {
1299 const struct sockaddr *sa = (const void *)cp;
1301 if ((i == RTA_GATEWAY) && sa->sa_family == AF_INET) {
1302 const struct sockaddr_in * const sin =
1303 (const struct sockaddr_in *)sa;
1305 *next_hop = sin->sin_addr;
1308 if ((i == RTA_IFP) && sa->sa_family == AF_LINK) {
1309 const struct sockaddr_dl * const sdl =
1310 (const struct sockaddr_dl *)sa;
1313 kernel_name = xmemdup0(sdl->sdl_data, sdl->sdl_nlen);
1314 ifname = netdev_bsd_kernel_name_to_ovs_name(kernel_name);
1320 if (ifname == NULL) {
1326 *netdev_name = ifname;
1327 VLOG_DBG("host " IP_FMT " next-hop " IP_FMT " if %s",
1328 IP_ARGS(host->s_addr), IP_ARGS(next_hop->s_addr), *netdev_name);
1336 make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
1338 struct sockaddr_in sin;
1339 memset(&sin, 0, sizeof sin);
1340 sin.sin_family = AF_INET;
1341 sin.sin_addr = addr;
1344 memset(sa, 0, sizeof *sa);
1345 memcpy(sa, &sin, sizeof sin);
1349 do_set_addr(struct netdev *netdev,
1350 int ioctl_nr, const char *ioctl_name, struct in_addr addr)
1353 make_in4_sockaddr(&ifr.ifr_addr, addr);
1354 return af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev), &ifr, ioctl_nr,
1359 nd_to_iff_flags(enum netdev_flags nd)
1362 if (nd & NETDEV_UP) {
1365 if (nd & NETDEV_PROMISC) {
1367 #if defined(IFF_PPROMISC)
1368 iff |= IFF_PPROMISC;
1375 iff_to_nd_flags(int iff)
1377 enum netdev_flags nd = 0;
1381 if (iff & IFF_PROMISC) {
1382 nd |= NETDEV_PROMISC;
1388 netdev_bsd_update_flags(struct netdev *netdev_, enum netdev_flags off,
1389 enum netdev_flags on, enum netdev_flags *old_flagsp)
1391 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1392 int old_flags, new_flags;
1395 error = get_flags(netdev_, &old_flags);
1397 *old_flagsp = iff_to_nd_flags(old_flags);
1398 new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on);
1399 if (new_flags != old_flags) {
1400 error = set_flags(netdev_get_kernel_name(netdev_), new_flags);
1401 netdev_bsd_changed(netdev);
1408 netdev_bsd_change_seq(const struct netdev *netdev)
1410 return netdev_bsd_cast(netdev)->change_seq;
1414 const struct netdev_class netdev_bsd_class = {
1421 netdev_bsd_construct_system,
1422 netdev_bsd_destruct,
1424 NULL, /* get_config */
1425 NULL, /* set_config */
1426 NULL, /* get_tunnel_config */
1429 netdev_bsd_send_wait,
1431 netdev_bsd_set_etheraddr,
1432 netdev_bsd_get_etheraddr,
1435 netdev_bsd_get_ifindex,
1436 netdev_bsd_get_carrier,
1437 NULL, /* get_carrier_resets */
1438 NULL, /* set_miimon_interval */
1439 netdev_bsd_get_stats,
1440 NULL, /* set_stats */
1442 netdev_bsd_get_features,
1443 NULL, /* set_advertisement */
1444 NULL, /* set_policing */
1445 NULL, /* get_qos_type */
1446 NULL, /* get_qos_capabilities */
1449 NULL, /* get_queue */
1450 NULL, /* set_queue */
1451 NULL, /* delete_queue */
1452 NULL, /* get_queue_stats */
1453 NULL, /* dump_queue */
1454 NULL, /* dump_queue_stats */
1459 NULL, /* add_router */
1460 netdev_bsd_get_next_hop,
1461 NULL, /* get_status */
1462 NULL, /* arp_lookup */
1464 netdev_bsd_update_flags,
1466 netdev_bsd_change_seq,
1468 netdev_bsd_rx_alloc,
1469 netdev_bsd_rx_construct,
1470 netdev_bsd_rx_destruct,
1471 netdev_bsd_rx_dealloc,
1474 netdev_bsd_rx_drain,
1477 const struct netdev_class netdev_tap_class = {
1484 netdev_bsd_construct_tap,
1485 netdev_bsd_destruct,
1487 NULL, /* get_config */
1488 NULL, /* set_config */
1489 NULL, /* get_tunnel_config */
1492 netdev_bsd_send_wait,
1494 netdev_bsd_set_etheraddr,
1495 netdev_bsd_get_etheraddr,
1498 netdev_bsd_get_ifindex,
1499 netdev_bsd_get_carrier,
1500 NULL, /* get_carrier_resets */
1501 NULL, /* set_miimon_interval */
1502 netdev_bsd_get_stats,
1503 NULL, /* set_stats */
1505 netdev_bsd_get_features,
1506 NULL, /* set_advertisement */
1507 NULL, /* set_policing */
1508 NULL, /* get_qos_type */
1509 NULL, /* get_qos_capabilities */
1512 NULL, /* get_queue */
1513 NULL, /* set_queue */
1514 NULL, /* delete_queue */
1515 NULL, /* get_queue_stats */
1516 NULL, /* dump_queue */
1517 NULL, /* dump_queue_stats */
1522 NULL, /* add_router */
1523 netdev_bsd_get_next_hop,
1524 NULL, /* get_status */
1525 NULL, /* arp_lookup */
1527 netdev_bsd_update_flags,
1529 netdev_bsd_change_seq,
1531 netdev_bsd_rx_alloc,
1532 netdev_bsd_rx_construct,
1533 netdev_bsd_rx_destruct,
1534 netdev_bsd_rx_dealloc,
1537 netdev_bsd_rx_drain,
1542 destroy_tap(int fd, const char *name)
1547 strcpy(ifr.ifr_name, name);
1548 /* XXX What to do if this call fails? */
1549 af_inet_ioctl(SIOCIFDESTROY, &ifr);
1553 get_flags(const struct netdev *netdev, int *flags)
1558 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev), &ifr,
1559 SIOCGIFFLAGS, "SIOCGIFFLAGS");
1561 *flags = ifr_get_flags(&ifr);
1567 set_flags(const char *name, int flags)
1571 ifr_set_flags(&ifr, flags);
1573 return af_inet_ifreq_ioctl(name, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS");
1577 get_ifindex(const struct netdev *netdev_, int *ifindexp)
1579 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1581 if (!(netdev->cache_valid & VALID_IFINDEX)) {
1582 int ifindex = if_nametoindex(netdev_get_name(netdev_));
1586 netdev->cache_valid |= VALID_IFINDEX;
1587 netdev->ifindex = ifindex;
1589 *ifindexp = netdev->ifindex;
1594 get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN])
1596 struct ifaddrs *head;
1597 struct ifaddrs *ifa;
1598 struct sockaddr_dl *sdl;
1600 if (getifaddrs(&head) != 0) {
1601 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1602 ovs_strerror(errno));
1606 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1607 if (ifa->ifa_addr->sa_family == AF_LINK) {
1608 if (!strcmp(ifa->ifa_name, netdev_name)) {
1609 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1611 memcpy(ea, LLADDR(sdl), sdl->sdl_alen);
1619 VLOG_ERR("could not find ethernet address for %s device", netdev_name);
1625 set_etheraddr(const char *netdev_name OVS_UNUSED, int hwaddr_family OVS_UNUSED,
1626 int hwaddr_len OVS_UNUSED,
1627 const uint8_t mac[ETH_ADDR_LEN] OVS_UNUSED)
1629 #if defined(__FreeBSD__)
1633 memset(&ifr, 0, sizeof ifr);
1634 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1635 ifr.ifr_addr.sa_family = hwaddr_family;
1636 ifr.ifr_addr.sa_len = hwaddr_len;
1637 memcpy(ifr.ifr_addr.sa_data, mac, hwaddr_len);
1638 error = af_inet_ioctl(SIOCSIFLLADDR, &ifr);
1640 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1641 netdev_name, ovs_strerror(error));
1645 #elif defined(__NetBSD__)
1646 struct if_laddrreq req;
1647 struct sockaddr_dl *sdl;
1648 struct sockaddr_storage oldaddr;
1652 * get the old address, add new one, and then remove old one.
1655 if (hwaddr_len != ETH_ADDR_LEN) {
1656 /* just to be safe about sockaddr storage size */
1659 memset(&req, 0, sizeof(req));
1660 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1661 req.addr.ss_len = sizeof(req.addr);
1662 req.addr.ss_family = hwaddr_family;
1663 sdl = (struct sockaddr_dl *)&req.addr;
1664 sdl->sdl_alen = hwaddr_len;
1666 error = af_link_ioctl(SIOCGLIFADDR, &req);
1670 if (!memcmp(&sdl->sdl_data[sdl->sdl_nlen], mac, hwaddr_len)) {
1675 memset(&req, 0, sizeof(req));
1676 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1677 req.flags = IFLR_ACTIVE;
1678 sdl = (struct sockaddr_dl *)&req.addr;
1679 sdl->sdl_len = offsetof(struct sockaddr_dl, sdl_data) + hwaddr_len;
1680 sdl->sdl_alen = hwaddr_len;
1681 sdl->sdl_family = hwaddr_family;
1682 memcpy(sdl->sdl_data, mac, hwaddr_len);
1683 error = af_link_ioctl(SIOCALIFADDR, &req);
1688 memset(&req, 0, sizeof(req));
1689 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1691 return af_link_ioctl(SIOCDLIFADDR, &req);
1693 #error not implemented
1698 ifr_get_flags(const struct ifreq *ifr)
1700 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1701 return (ifr->ifr_flagshigh << 16) | ifr->ifr_flags;
1703 return ifr->ifr_flags;
1708 ifr_set_flags(struct ifreq *ifr, int flags)
1710 ifr->ifr_flags = flags;
1711 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1712 ifr->ifr_flagshigh = flags >> 16;
1716 /* Calls ioctl() on an AF_LINK sock, passing the specified 'command' and
1717 * 'arg'. Returns 0 if successful, otherwise a positive errno value. */
1719 af_link_ioctl(int command, const void *arg)
1721 static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
1724 if (ovsthread_once_start(&once)) {
1725 sock = socket(AF_LINK, SOCK_DGRAM, 0);
1728 VLOG_ERR("failed to create link socket: %s", ovs_strerror(errno));
1730 ovsthread_once_done(&once);
1733 return (sock < 0 ? -sock
1734 : ioctl(sock, command, arg) == -1 ? errno