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>
51 #include "dynamic-string.h"
52 #include "fatal-signal.h"
54 #include "openflow/openflow.h"
55 #include "ovs-thread.h"
57 #include "poll-loop.h"
58 #include "socket-util.h"
64 VLOG_DEFINE_THIS_MODULE(netdev_bsd);
67 struct netdev_rx_bsd {
70 /* Packet capture descriptor for a system network device.
71 * For a tap device this is NULL. */
74 /* Selectable file descriptor for the network device.
75 * This descriptor will be used for polling operations. */
82 /* Never changes after initialization. */
85 /* Protects all members below. */
86 struct ovs_mutex mutex;
88 unsigned int cache_valid;
89 unsigned int change_seq;
92 uint8_t etheraddr[ETH_ADDR_LEN];
94 struct in_addr netmask;
99 int tap_fd; /* TAP character device, if any, otherwise -1. */
101 /* Used for sending packets on non-tap devices. */
108 VALID_IFINDEX = 1 << 0,
109 VALID_ETHERADDR = 1 << 1,
113 VALID_CARRIER = 1 << 5
116 #define PCAP_SNAPLEN 2048
120 * Notifier used to invalidate device informations in case of status change.
122 * It will be registered with a 'rtbsd_notifier_register()' when the first
123 * device will be created with the call of either 'netdev_bsd_tap_create()' or
124 * 'netdev_bsd_system_create()'.
126 * The callback associated with this notifier ('netdev_bsd_cache_cb()') will
127 * invalidate cached information about the device.
129 static struct rtbsd_notifier netdev_bsd_cache_notifier;
130 static int cache_notifier_refcount;
132 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
134 static void destroy_tap(int fd, const char *name);
135 static int get_flags(const struct netdev *, int *flagsp);
136 static int set_flags(const char *, int flags);
137 static int do_set_addr(struct netdev *netdev,
138 unsigned long ioctl_nr, const char *ioctl_name,
139 struct in_addr addr);
140 static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]);
141 static int set_etheraddr(const char *netdev_name, int hwaddr_family,
142 int hwaddr_len, const uint8_t[ETH_ADDR_LEN]);
143 static int get_ifindex(const struct netdev *, int *ifindexp);
145 static int ifr_get_flags(const struct ifreq *);
146 static void ifr_set_flags(struct ifreq *, int flags);
149 static int af_link_ioctl(unsigned long command, const void *arg);
152 static void netdev_bsd_run(void);
155 is_netdev_bsd_class(const struct netdev_class *netdev_class)
157 return netdev_class->run == netdev_bsd_run;
160 static struct netdev_bsd *
161 netdev_bsd_cast(const struct netdev *netdev)
163 ovs_assert(is_netdev_bsd_class(netdev_get_class(netdev)));
164 return CONTAINER_OF(netdev, struct netdev_bsd, up);
167 static struct netdev_rx_bsd *
168 netdev_rx_bsd_cast(const struct netdev_rx *rx)
170 ovs_assert(is_netdev_bsd_class(netdev_get_class(rx->netdev)));
171 return CONTAINER_OF(rx, struct netdev_rx_bsd, up);
175 netdev_get_kernel_name(const struct netdev *netdev)
177 return netdev_bsd_cast(netdev)->kernel_name;
181 * Perform periodic work needed by netdev. In BSD netdevs it checks for any
182 * interface status changes, and eventually calls all the user callbacks.
187 rtbsd_notifier_run();
191 * Arranges for poll_block() to wake up if the "run" member function needs to
195 netdev_bsd_wait(void)
197 rtbsd_notifier_wait();
201 netdev_bsd_changed(struct netdev_bsd *dev)
204 if (!dev->change_seq) {
209 /* Invalidate cache in case of interface status change. */
211 netdev_bsd_cache_cb(const struct rtbsd_change *change,
212 void *aux OVS_UNUSED)
214 struct netdev_bsd *dev;
217 struct netdev *base_dev = netdev_from_name(change->if_name);
220 const struct netdev_class *netdev_class =
221 netdev_get_class(base_dev);
223 if (is_netdev_bsd_class(netdev_class)) {
224 dev = netdev_bsd_cast(base_dev);
225 dev->cache_valid = 0;
226 netdev_bsd_changed(dev);
228 netdev_close(base_dev);
232 * XXX the API is lacking, we should be able to iterate on the list of
233 * netdevs without having to store the info in a temp shash.
235 struct shash device_shash;
236 struct shash_node *node;
238 shash_init(&device_shash);
239 netdev_get_devices(&netdev_bsd_class, &device_shash);
240 SHASH_FOR_EACH (node, &device_shash) {
241 struct netdev *netdev = node->data;
242 dev = netdev_bsd_cast(netdev);
243 dev->cache_valid = 0;
244 netdev_bsd_changed(dev);
245 netdev_close(netdev);
247 shash_destroy(&device_shash);
252 cache_notifier_ref(void)
256 if (!cache_notifier_refcount) {
257 ret = rtbsd_notifier_register(&netdev_bsd_cache_notifier,
258 netdev_bsd_cache_cb, NULL);
263 cache_notifier_refcount++;
268 cache_notifier_unref(void)
270 cache_notifier_refcount--;
271 if (cache_notifier_refcount == 0) {
272 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier);
277 static struct netdev *
278 netdev_bsd_alloc(void)
280 struct netdev_bsd *netdev = xzalloc(sizeof *netdev);
285 netdev_bsd_construct_system(struct netdev *netdev_)
287 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
288 enum netdev_flags flags;
291 error = cache_notifier_ref();
296 ovs_mutex_init(&netdev->mutex);
297 netdev->change_seq = 1;
299 netdev->kernel_name = xstrdup(netdev_->name);
301 /* Verify that the netdev really exists by attempting to read its flags */
302 error = netdev_get_flags(netdev_, &flags);
303 if (error == ENXIO) {
304 free(netdev->kernel_name);
305 cache_notifier_unref();
313 netdev_bsd_construct_tap(struct netdev *netdev_)
315 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
316 const char *name = netdev_->name;
319 char *kernel_name = NULL;
321 error = cache_notifier_ref();
326 memset(&ifr, 0, sizeof(ifr));
328 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
329 * to retrieve the name of the tap device. */
330 ovs_mutex_init(&netdev->mutex);
331 netdev->tap_fd = open("/dev/tap", O_RDWR);
332 netdev->change_seq = 1;
333 if (netdev->tap_fd < 0) {
335 VLOG_WARN("opening \"/dev/tap\" failed: %s", ovs_strerror(error));
336 goto error_unref_notifier;
339 /* Retrieve tap name (e.g. tap0) */
340 if (ioctl(netdev->tap_fd, TAPGIFNAME, &ifr) == -1) {
341 /* XXX Need to destroy the device? */
343 close(netdev->tap_fd);
344 goto error_unref_notifier;
347 /* Change the name of the tap device */
348 #if defined(SIOCSIFNAME)
349 ifr.ifr_data = (void *)name;
350 error = af_inet_ioctl(SIOCSIFNAME, &ifr);
352 destroy_tap(netdev->tap_fd, ifr.ifr_name);
353 goto error_unref_notifier;
355 kernel_name = xstrdup(name);
358 * NetBSD doesn't support inteface renaming.
360 VLOG_INFO("tap %s is created for bridge %s", ifr.ifr_name, name);
361 kernel_name = xstrdup(ifr.ifr_name);
364 /* set non-blocking. */
365 error = set_nonblocking(netdev->tap_fd);
367 destroy_tap(netdev->tap_fd, kernel_name);
368 goto error_unref_notifier;
372 ifr_set_flags(&ifr, IFF_UP);
373 strncpy(ifr.ifr_name, kernel_name, sizeof ifr.ifr_name);
374 error = af_inet_ioctl(SIOCSIFFLAGS, &ifr);
376 destroy_tap(netdev->tap_fd, kernel_name);
377 goto error_unref_notifier;
380 netdev->kernel_name = kernel_name;
384 error_unref_notifier:
385 ovs_mutex_destroy(&netdev->mutex);
386 cache_notifier_unref();
393 netdev_bsd_destruct(struct netdev *netdev_)
395 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
397 cache_notifier_unref();
399 if (netdev->tap_fd >= 0) {
400 destroy_tap(netdev->tap_fd, netdev_get_kernel_name(netdev_));
403 pcap_close(netdev->pcap);
405 free(netdev->kernel_name);
406 ovs_mutex_destroy(&netdev->mutex);
410 netdev_bsd_dealloc(struct netdev *netdev_)
412 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
418 netdev_bsd_open_pcap(const char *name, pcap_t **pcapp, int *fdp)
420 char errbuf[PCAP_ERRBUF_SIZE];
426 /* Open the pcap device. The device is opened in non-promiscuous mode
427 * because the interface flags are manually set by the caller. */
429 pcap = pcap_open_live(name, PCAP_SNAPLEN, 0, 1000, errbuf);
431 VLOG_ERR_RL(&rl, "%s: pcap_open_live failed: %s", name, errbuf);
435 if (errbuf[0] != '\0') {
436 VLOG_WARN_RL(&rl, "%s: pcap_open_live: %s", name, errbuf);
439 /* Get the underlying fd. */
440 fd = pcap_get_selectable_fd(pcap);
442 VLOG_WARN_RL(&rl, "%s: no selectable file descriptor", name);
447 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
448 * on the file descriptor returned by pcap_get_selectable_fd to achieve
449 * a real non-blocking behaviour.*/
450 error = pcap_setnonblock(pcap, 1, errbuf);
456 /* This call assure that reads return immediately upon packet
457 * reception. Otherwise, a read will block until either the kernel
458 * buffer becomes full or a timeout occurs. */
459 if (ioctl(fd, BIOCIMMEDIATE, &one) < 0 ) {
460 VLOG_ERR_RL(&rl, "ioctl(BIOCIMMEDIATE) on %s device failed: %s",
461 name, ovs_strerror(errno));
466 /* Capture only incoming packets. */
467 error = pcap_setdirection(pcap, PCAP_D_IN);
486 static struct netdev_rx *
487 netdev_bsd_rx_alloc(void)
489 struct netdev_rx_bsd *rx = xzalloc(sizeof *rx);
494 netdev_bsd_rx_construct(struct netdev_rx *rx_)
496 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
497 struct netdev *netdev_ = rx->up.netdev;
498 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
501 if (!strcmp(netdev_get_type(netdev_), "tap")) {
502 rx->pcap_handle = NULL;
503 rx->fd = netdev->tap_fd;
506 ovs_mutex_lock(&netdev->mutex);
507 error = netdev_bsd_open_pcap(netdev_get_kernel_name(netdev_),
508 &rx->pcap_handle, &rx->fd);
510 netdev_bsd_changed(netdev);
512 ovs_mutex_unlock(&netdev->mutex);
519 netdev_bsd_rx_destruct(struct netdev_rx *rx_)
521 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
523 if (rx->pcap_handle) {
524 pcap_close(rx->pcap_handle);
529 netdev_bsd_rx_dealloc(struct netdev_rx *rx_)
531 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
536 /* The recv callback of the netdev class returns the number of bytes of the
539 * This can be done by the pcap_next() function. Unfortunately pcap_next() does
540 * not make difference between a missing packet on the capture interface and
541 * an error during the file capture. We can use the pcap_dispatch() function
542 * instead, which is able to distinguish between errors and null packet.
544 * To make pcap_dispatch() returns the number of bytes read from the interface
545 * we need to define the following callback and argument.
554 * This callback will be executed on every captured packet.
556 * If the packet captured by pcap_dispatch() does not fit the pcap buffer,
557 * pcap returns a truncated packet and we follow this behavior.
559 * The argument args->retval is the packet size in bytes.
562 proc_pkt(u_char *args_, const struct pcap_pkthdr *hdr, const u_char *packet)
564 struct pcap_arg *args = (struct pcap_arg *)args_;
566 if (args->size < hdr->len) {
567 VLOG_WARN_RL(&rl, "packet truncated");
568 args->retval = args->size;
570 args->retval = hdr->len;
573 /* copy the packet to our buffer */
574 memcpy(args->data, packet, args->retval);
578 * This function attempts to receive a packet from the specified network
579 * device. It is assumed that the network device is a system device or a tap
580 * device opened as a system one. In this case the read operation is performed
584 netdev_rx_bsd_recv_pcap(struct netdev_rx_bsd *rx, void *data, size_t size)
589 /* prepare the pcap argument to store the packet */
594 ret = pcap_dispatch(rx->pcap_handle, 1, proc_pkt, (u_char *) &arg);
597 return arg.retval; /* arg.retval < 0 is handled in the caller */
600 if (errno == EINTR) {
610 * This function attempts to receive a packet from the specified network
611 * device. It is assumed that the network device is a tap device and
612 * 'rx->fd' is initialized with the tap file descriptor.
615 netdev_rx_bsd_recv_tap(struct netdev_rx_bsd *rx, void *data, size_t size)
618 ssize_t retval = read(rx->fd, data, size);
621 } else if (errno != EINTR) {
622 if (errno != EAGAIN) {
623 VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
624 ovs_strerror(errno), netdev_rx_get_name(&rx->up));
632 netdev_bsd_rx_recv(struct netdev_rx *rx_, void *data, size_t size)
634 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
636 return (rx->pcap_handle
637 ? netdev_rx_bsd_recv_pcap(rx, data, size)
638 : netdev_rx_bsd_recv_tap(rx, data, size));
642 * Registers with the poll loop to wake up from the next call to poll_block()
643 * when a packet is ready to be received with netdev_rx_recv() on 'rx'.
646 netdev_bsd_rx_wait(struct netdev_rx *rx_)
648 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
650 poll_fd_wait(rx->fd, POLLIN);
653 /* Discards all packets waiting to be received from 'rx'. */
655 netdev_bsd_rx_drain(struct netdev_rx *rx_)
658 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
660 strcpy(ifr.ifr_name, netdev_get_kernel_name(netdev_rx_get_netdev(rx_)));
661 if (ioctl(rx->fd, BIOCFLUSH, &ifr) == -1) {
662 VLOG_DBG_RL(&rl, "%s: ioctl(BIOCFLUSH) failed: %s",
663 netdev_rx_get_name(rx_), ovs_strerror(errno));
670 * Send a packet on the specified network device. The device could be either a
671 * system or a tap device.
674 netdev_bsd_send(struct netdev *netdev_, const void *data, size_t size)
676 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
677 const char *name = netdev_get_name(netdev_);
680 ovs_mutex_lock(&dev->mutex);
681 if (dev->tap_fd < 0 && !dev->pcap) {
682 error = netdev_bsd_open_pcap(name, &dev->pcap, &dev->fd);
689 if (dev->tap_fd >= 0) {
690 retval = write(dev->tap_fd, data, size);
692 retval = pcap_inject(dev->pcap, data, size);
695 if (errno == EINTR) {
699 if (error != EAGAIN) {
700 VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: "
701 "%s", name, ovs_strerror(error));
704 } else if (retval != size) {
705 VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of "
706 "%zu) on %s", retval, size, name);
713 ovs_mutex_unlock(&dev->mutex);
718 * Registers with the poll loop to wake up from the next call to poll_block()
719 * when the packet transmission queue has sufficient room to transmit a packet
720 * with netdev_send().
723 netdev_bsd_send_wait(struct netdev *netdev_)
725 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
727 ovs_mutex_lock(&dev->mutex);
728 if (dev->tap_fd >= 0) {
729 /* TAP device always accepts packets. */
730 poll_immediate_wake();
731 } else if (dev->pcap) {
732 poll_fd_wait(dev->fd, POLLOUT);
734 /* We haven't even tried to send a packet yet. */
735 poll_immediate_wake();
737 ovs_mutex_unlock(&dev->mutex);
741 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
742 * otherwise a positive errno value.
745 netdev_bsd_set_etheraddr(struct netdev *netdev_,
746 const uint8_t mac[ETH_ADDR_LEN])
748 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
751 ovs_mutex_lock(&netdev->mutex);
752 if (!(netdev->cache_valid & VALID_ETHERADDR)
753 || !eth_addr_equals(netdev->etheraddr, mac)) {
754 error = set_etheraddr(netdev_get_kernel_name(netdev_), AF_LINK,
757 netdev->cache_valid |= VALID_ETHERADDR;
758 memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN);
759 netdev_bsd_changed(netdev);
762 ovs_mutex_unlock(&netdev->mutex);
768 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
769 * free the returned buffer.
772 netdev_bsd_get_etheraddr(const struct netdev *netdev_,
773 uint8_t mac[ETH_ADDR_LEN])
775 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
778 ovs_mutex_lock(&netdev->mutex);
779 if (!(netdev->cache_valid & VALID_ETHERADDR)) {
780 error = get_etheraddr(netdev_get_kernel_name(netdev_),
783 netdev->cache_valid |= VALID_ETHERADDR;
787 memcpy(mac, netdev->etheraddr, ETH_ADDR_LEN);
789 ovs_mutex_unlock(&netdev->mutex);
795 * Returns the maximum size of transmitted (and received) packets on 'netdev',
796 * in bytes, not including the hardware header; thus, this is typically 1500
797 * bytes for Ethernet devices.
800 netdev_bsd_get_mtu(const struct netdev *netdev_, int *mtup)
802 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
805 ovs_mutex_lock(&netdev->mutex);
806 if (!(netdev->cache_valid & VALID_MTU)) {
809 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
810 SIOCGIFMTU, "SIOCGIFMTU");
812 netdev->mtu = ifr.ifr_mtu;
813 netdev->cache_valid |= VALID_MTU;
819 ovs_mutex_unlock(&netdev->mutex);
825 netdev_bsd_get_ifindex(const struct netdev *netdev_)
827 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
830 ovs_mutex_lock(&netdev->mutex);
831 error = get_ifindex(netdev_, &ifindex);
832 ovs_mutex_unlock(&netdev->mutex);
834 return error ? -error : ifindex;
838 netdev_bsd_get_carrier(const struct netdev *netdev_, bool *carrier)
840 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
843 ovs_mutex_lock(&netdev->mutex);
844 if (!(netdev->cache_valid & VALID_CARRIER)) {
845 struct ifmediareq ifmr;
847 memset(&ifmr, 0, sizeof(ifmr));
848 strncpy(ifmr.ifm_name, netdev_get_kernel_name(netdev_),
849 sizeof ifmr.ifm_name);
851 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
853 netdev->carrier = (ifmr.ifm_status & IFM_ACTIVE) == IFM_ACTIVE;
854 netdev->cache_valid |= VALID_CARRIER;
856 /* If the interface doesn't report whether the media is active,
857 * just assume it is active. */
858 if ((ifmr.ifm_status & IFM_AVALID) == 0) {
859 netdev->carrier = true;
862 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
863 netdev_get_name(netdev_), ovs_strerror(error));
867 *carrier = netdev->carrier;
869 ovs_mutex_unlock(&netdev->mutex);
875 convert_stats(struct netdev_stats *stats, const struct if_data *ifd)
878 * note: UINT64_MAX means unsupported
880 stats->rx_packets = ifd->ifi_ipackets;
881 stats->tx_packets = ifd->ifi_opackets;
882 stats->rx_bytes = ifd->ifi_obytes;
883 stats->tx_bytes = ifd->ifi_ibytes;
884 stats->rx_errors = ifd->ifi_ierrors;
885 stats->tx_errors = ifd->ifi_oerrors;
886 stats->rx_dropped = ifd->ifi_iqdrops;
887 stats->tx_dropped = UINT64_MAX;
888 stats->multicast = ifd->ifi_imcasts;
889 stats->collisions = ifd->ifi_collisions;
890 stats->rx_length_errors = UINT64_MAX;
891 stats->rx_over_errors = UINT64_MAX;
892 stats->rx_crc_errors = UINT64_MAX;
893 stats->rx_frame_errors = UINT64_MAX;
894 stats->rx_fifo_errors = UINT64_MAX;
895 stats->rx_missed_errors = UINT64_MAX;
896 stats->tx_aborted_errors = UINT64_MAX;
897 stats->tx_carrier_errors = UINT64_MAX;
898 stats->tx_fifo_errors = UINT64_MAX;
899 stats->tx_heartbeat_errors = UINT64_MAX;
900 stats->tx_window_errors = UINT64_MAX;
903 /* Retrieves current device stats for 'netdev'. */
905 netdev_bsd_get_stats(const struct netdev *netdev_, struct netdev_stats *stats)
907 #if defined(__FreeBSD__)
911 struct ifmibdata ifmd;
916 mib[2] = NETLINK_GENERIC;
917 mib[3] = IFMIB_SYSTEM;
918 mib[4] = IFMIB_IFCOUNT;
920 len = sizeof(if_count);
922 if (sysctl(mib, 5, &if_count, &len, (void *)0, 0) == -1) {
923 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
924 netdev_get_name(netdev_), ovs_strerror(errno));
928 mib[5] = IFDATA_GENERAL;
929 mib[3] = IFMIB_IFDATA;
931 for (i = 1; i <= if_count; i++) {
933 if (sysctl(mib, 6, &ifmd, &len, (void *)0, 0) == -1) {
934 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
935 netdev_get_name(netdev_), ovs_strerror(errno));
937 } else if (!strcmp(ifmd.ifmd_name, netdev_get_name(netdev_))) {
938 convert_stats(stats, &ifmd.ifmd_data);
944 #elif defined(__NetBSD__)
945 struct ifdatareq ifdr;
948 memset(&ifdr, 0, sizeof(ifdr));
949 strncpy(ifdr.ifdr_name, netdev_get_kernel_name(netdev_),
950 sizeof(ifdr.ifdr_name));
951 error = af_link_ioctl(SIOCGIFDATA, &ifdr);
953 convert_stats(stats, &ifdr.ifdr_data);
957 #error not implemented
962 netdev_bsd_parse_media(int media)
964 uint32_t supported = 0;
965 bool half_duplex = media & IFM_HDX ? true : false;
967 switch (IFM_SUBTYPE(media)) {
972 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
973 supported |= NETDEV_F_COPPER;
977 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
978 supported |= NETDEV_F_FIBER;
985 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
986 supported |= NETDEV_F_COPPER;
990 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
991 supported |= NETDEV_F_FIBER;
996 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
997 supported |= NETDEV_F_COPPER;
1002 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
1003 supported |= NETDEV_F_FIBER;
1007 supported |= NETDEV_F_10GB_FD;
1008 supported |= NETDEV_F_COPPER;
1013 supported |= NETDEV_F_10GB_FD;
1014 supported |= NETDEV_F_FIBER;
1021 if (IFM_SUBTYPE(media) == IFM_AUTO) {
1022 supported |= NETDEV_F_AUTONEG;
1025 if (media & IFM_ETH_FMASK) {
1026 supported |= NETDEV_F_PAUSE;
1034 * Stores the features supported by 'netdev' into each of '*current',
1035 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1036 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1037 * successful, otherwise a positive errno value. On failure, all of the
1038 * passed-in values are set to 0.
1041 netdev_bsd_get_features(const struct netdev *netdev,
1042 enum netdev_features *current, uint32_t *advertised,
1043 enum netdev_features *supported, uint32_t *peer)
1045 struct ifmediareq ifmr;
1051 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1053 memset(&ifmr, 0, sizeof(ifmr));
1054 strncpy(ifmr.ifm_name, netdev_get_name(netdev), sizeof ifmr.ifm_name);
1056 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1057 * number of supported modes, and a second with a buffer to retrieve
1059 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
1061 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1062 netdev_get_name(netdev), ovs_strerror(error));
1066 media_list = xcalloc(ifmr.ifm_count, sizeof(int));
1067 ifmr.ifm_ulist = media_list;
1069 if (IFM_TYPE(ifmr.ifm_current) != IFM_ETHER) {
1070 VLOG_DBG_RL(&rl, "%s: doesn't appear to be ethernet",
1071 netdev_get_name(netdev));
1076 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
1078 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1079 netdev_get_name(netdev), ovs_strerror(error));
1083 /* Current settings. */
1084 *current = netdev_bsd_parse_media(ifmr.ifm_active);
1086 /* Advertised features. */
1087 *advertised = netdev_bsd_parse_media(ifmr.ifm_current);
1089 /* Supported features. */
1091 for (i = 0; i < ifmr.ifm_count; i++) {
1092 *supported |= netdev_bsd_parse_media(ifmr.ifm_ulist[i]);
1095 /* Peer advertisements. */
1096 *peer = 0; /* XXX */
1105 * If 'netdev' has an assigned IPv4 address, sets '*in4' to that address and
1106 * '*netmask' to its netmask and returns true. Otherwise, returns false.
1109 netdev_bsd_get_in4(const struct netdev *netdev_, struct in_addr *in4,
1110 struct in_addr *netmask)
1112 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1115 ovs_mutex_lock(&netdev->mutex);
1116 if (!(netdev->cache_valid & VALID_IN4)) {
1119 ifr.ifr_addr.sa_family = AF_INET;
1120 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1121 SIOCGIFADDR, "SIOCGIFADDR");
1123 const struct sockaddr_in *sin;
1125 sin = (struct sockaddr_in *) &ifr.ifr_addr;
1126 netdev->in4 = sin->sin_addr;
1127 netdev->cache_valid |= VALID_IN4;
1128 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1129 SIOCGIFNETMASK, "SIOCGIFNETMASK");
1131 *netmask = sin->sin_addr;
1137 *netmask = netdev->netmask;
1139 ovs_mutex_unlock(&netdev->mutex);
1141 return error ? error : in4->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0;
1145 * Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
1146 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
1147 * positive errno value.
1150 netdev_bsd_set_in4(struct netdev *netdev_, struct in_addr addr,
1151 struct in_addr mask)
1153 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1156 ovs_mutex_lock(&netdev->mutex);
1157 error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", addr);
1159 if (addr.s_addr != INADDR_ANY) {
1160 error = do_set_addr(netdev_, SIOCSIFNETMASK,
1161 "SIOCSIFNETMASK", mask);
1163 netdev->cache_valid |= VALID_IN4;
1165 netdev->netmask = mask;
1168 netdev_bsd_changed(netdev);
1170 ovs_mutex_unlock(&netdev->mutex);
1176 netdev_bsd_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
1178 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1179 if (!(netdev->cache_valid & VALID_IN6)) {
1180 struct ifaddrs *ifa, *head;
1181 struct sockaddr_in6 *sin6;
1182 const char *netdev_name = netdev_get_name(netdev_);
1184 if (getifaddrs(&head) != 0) {
1185 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1186 ovs_strerror(errno));
1190 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1191 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1192 !strcmp(ifa->ifa_name, netdev_name)) {
1193 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1195 memcpy(&netdev->in6, &sin6->sin6_addr, sin6->sin6_len);
1196 netdev->cache_valid |= VALID_IN6;
1203 return EADDRNOTAVAIL;
1209 #if defined(__NetBSD__)
1211 netdev_bsd_kernel_name_to_ovs_name(const char *kernel_name)
1213 char *ovs_name = NULL;
1214 struct shash device_shash;
1215 struct shash_node *node;
1217 shash_init(&device_shash);
1218 netdev_get_devices(&netdev_tap_class, &device_shash);
1219 SHASH_FOR_EACH(node, &device_shash) {
1220 struct netdev *netdev = node->data;
1221 struct netdev_bsd * const dev = netdev_bsd_cast(netdev);
1223 if (!strcmp(dev->kernel_name, kernel_name)) {
1225 ovs_name = xstrdup(netdev_get_name(&dev->up));
1227 netdev_close(netdev);
1229 shash_destroy(&device_shash);
1231 return ovs_name ? ovs_name : xstrdup(kernel_name);
1236 netdev_bsd_get_next_hop(const struct in_addr *host OVS_UNUSED,
1237 struct in_addr *next_hop OVS_UNUSED,
1238 char **netdev_name OVS_UNUSED)
1240 #if defined(__NetBSD__)
1242 struct sockaddr_in sin;
1243 struct sockaddr_dl sdl;
1250 struct rt_msghdr *rtm = &buf.h;
1251 const pid_t pid = getpid();
1254 bool gateway = false;
1255 char *ifname = NULL;
1258 memset(next_hop, 0, sizeof(*next_hop));
1259 *netdev_name = NULL;
1261 memset(&sin, 0, sizeof(sin));
1262 sin.sin_len = sizeof(sin);
1263 sin.sin_family = AF_INET;
1265 sin.sin_addr = *host;
1267 memset(&sdl, 0, sizeof(sdl));
1268 sdl.sdl_len = sizeof(sdl);
1269 sdl.sdl_family = AF_LINK;
1271 s = socket(PF_ROUTE, SOCK_RAW, 0);
1272 memset(&buf, 0, sizeof(buf));
1273 rtm->rtm_flags = RTF_HOST|RTF_UP;
1274 rtm->rtm_version = RTM_VERSION;
1275 rtm->rtm_addrs = RTA_DST|RTA_IFP;
1276 cp = (void *)&buf.space;
1277 memcpy(cp, &sin, sizeof(sin));
1278 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sin);
1279 memcpy(cp, &sdl, sizeof(sdl));
1280 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sdl);
1281 rtm->rtm_msglen = cp - (char *)(void *)rtm;
1282 rtm->rtm_seq = ++seq;
1283 rtm->rtm_type = RTM_GET;
1285 write(s, rtm, rtm->rtm_msglen);
1286 memset(&buf, 0, sizeof(buf));
1288 ssz = read(s, &buf, sizeof(buf));
1289 } while (ssz > 0 && (rtm->rtm_seq != seq || rtm->rtm_pid != pid));
1290 saved_errno = errno;
1296 return EPIPE; /* XXX */
1298 cp = (void *)&buf.space;
1299 for (i = 1; i; i <<= 1) {
1300 if ((rtm->rtm_addrs & i) != 0) {
1301 const struct sockaddr *sa = (const void *)cp;
1303 if ((i == RTA_GATEWAY) && sa->sa_family == AF_INET) {
1304 const struct sockaddr_in * const sin =
1305 (const struct sockaddr_in *)sa;
1307 *next_hop = sin->sin_addr;
1310 if ((i == RTA_IFP) && sa->sa_family == AF_LINK) {
1311 const struct sockaddr_dl * const sdl =
1312 (const struct sockaddr_dl *)sa;
1315 kernel_name = xmemdup0(sdl->sdl_data, sdl->sdl_nlen);
1316 ifname = netdev_bsd_kernel_name_to_ovs_name(kernel_name);
1322 if (ifname == NULL) {
1328 *netdev_name = ifname;
1329 VLOG_DBG("host " IP_FMT " next-hop " IP_FMT " if %s",
1330 IP_ARGS(host->s_addr), IP_ARGS(next_hop->s_addr), *netdev_name);
1338 netdev_bsd_arp_lookup(const struct netdev *netdev OVS_UNUSED,
1339 ovs_be32 ip OVS_UNUSED,
1340 uint8_t mac[ETH_ADDR_LEN] OVS_UNUSED)
1342 #if defined(__NetBSD__)
1343 const struct rt_msghdr *rtm;
1356 mib[4] = NET_RT_FLAGS;
1357 mib[5] = RTF_LLINFO;
1358 if (sysctl(mib, 6, NULL, &needed, NULL, 0) == -1) {
1362 buf = xmalloc(needed);
1363 if (sysctl(mib, 6, buf, &needed, NULL, 0) == -1) {
1368 for (cp = buf; cp < ep; cp += rtm->rtm_msglen) {
1369 const struct sockaddr_inarp *sina;
1370 const struct sockaddr_dl *sdl;
1372 rtm = (const void *)cp;
1373 sina = (const void *)(rtm + 1);
1374 if (ip != sina->sin_addr.s_addr) {
1377 sdl = (const void *)
1378 ((const char *)(const void *)sina + RT_ROUNDUP(sina->sin_len));
1379 if (sdl->sdl_alen == ETH_ADDR_LEN) {
1380 memcpy(mac, &sdl->sdl_data[sdl->sdl_nlen], ETH_ADDR_LEN);
1395 make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
1397 struct sockaddr_in sin;
1398 memset(&sin, 0, sizeof sin);
1399 sin.sin_family = AF_INET;
1400 sin.sin_addr = addr;
1403 memset(sa, 0, sizeof *sa);
1404 memcpy(sa, &sin, sizeof sin);
1408 do_set_addr(struct netdev *netdev,
1409 unsigned long ioctl_nr, const char *ioctl_name,
1410 struct in_addr addr)
1413 make_in4_sockaddr(&ifr.ifr_addr, addr);
1414 return af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev), &ifr, ioctl_nr,
1419 nd_to_iff_flags(enum netdev_flags nd)
1422 if (nd & NETDEV_UP) {
1425 if (nd & NETDEV_PROMISC) {
1427 #if defined(IFF_PPROMISC)
1428 iff |= IFF_PPROMISC;
1435 iff_to_nd_flags(int iff)
1437 enum netdev_flags nd = 0;
1441 if (iff & IFF_PROMISC) {
1442 nd |= NETDEV_PROMISC;
1448 netdev_bsd_update_flags(struct netdev *netdev_, enum netdev_flags off,
1449 enum netdev_flags on, enum netdev_flags *old_flagsp)
1451 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1452 int old_flags, new_flags;
1455 error = get_flags(netdev_, &old_flags);
1457 *old_flagsp = iff_to_nd_flags(old_flags);
1458 new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on);
1459 if (new_flags != old_flags) {
1460 error = set_flags(netdev_get_kernel_name(netdev_), new_flags);
1461 netdev_bsd_changed(netdev);
1468 netdev_bsd_change_seq(const struct netdev *netdev)
1470 return netdev_bsd_cast(netdev)->change_seq;
1474 const struct netdev_class netdev_bsd_class = {
1481 netdev_bsd_construct_system,
1482 netdev_bsd_destruct,
1484 NULL, /* get_config */
1485 NULL, /* set_config */
1486 NULL, /* get_tunnel_config */
1489 netdev_bsd_send_wait,
1491 netdev_bsd_set_etheraddr,
1492 netdev_bsd_get_etheraddr,
1495 netdev_bsd_get_ifindex,
1496 netdev_bsd_get_carrier,
1497 NULL, /* get_carrier_resets */
1498 NULL, /* set_miimon_interval */
1499 netdev_bsd_get_stats,
1500 NULL, /* set_stats */
1502 netdev_bsd_get_features,
1503 NULL, /* set_advertisement */
1504 NULL, /* set_policing */
1505 NULL, /* get_qos_type */
1506 NULL, /* get_qos_capabilities */
1509 NULL, /* get_queue */
1510 NULL, /* set_queue */
1511 NULL, /* delete_queue */
1512 NULL, /* get_queue_stats */
1513 NULL, /* dump_queue */
1514 NULL, /* dump_queue_stats */
1519 NULL, /* add_router */
1520 netdev_bsd_get_next_hop,
1521 NULL, /* get_status */
1522 netdev_bsd_arp_lookup, /* arp_lookup */
1524 netdev_bsd_update_flags,
1526 netdev_bsd_change_seq,
1528 netdev_bsd_rx_alloc,
1529 netdev_bsd_rx_construct,
1530 netdev_bsd_rx_destruct,
1531 netdev_bsd_rx_dealloc,
1534 netdev_bsd_rx_drain,
1537 const struct netdev_class netdev_tap_class = {
1544 netdev_bsd_construct_tap,
1545 netdev_bsd_destruct,
1547 NULL, /* get_config */
1548 NULL, /* set_config */
1549 NULL, /* get_tunnel_config */
1552 netdev_bsd_send_wait,
1554 netdev_bsd_set_etheraddr,
1555 netdev_bsd_get_etheraddr,
1558 netdev_bsd_get_ifindex,
1559 netdev_bsd_get_carrier,
1560 NULL, /* get_carrier_resets */
1561 NULL, /* set_miimon_interval */
1562 netdev_bsd_get_stats,
1563 NULL, /* set_stats */
1565 netdev_bsd_get_features,
1566 NULL, /* set_advertisement */
1567 NULL, /* set_policing */
1568 NULL, /* get_qos_type */
1569 NULL, /* get_qos_capabilities */
1572 NULL, /* get_queue */
1573 NULL, /* set_queue */
1574 NULL, /* delete_queue */
1575 NULL, /* get_queue_stats */
1576 NULL, /* dump_queue */
1577 NULL, /* dump_queue_stats */
1582 NULL, /* add_router */
1583 netdev_bsd_get_next_hop,
1584 NULL, /* get_status */
1585 netdev_bsd_arp_lookup, /* arp_lookup */
1587 netdev_bsd_update_flags,
1589 netdev_bsd_change_seq,
1591 netdev_bsd_rx_alloc,
1592 netdev_bsd_rx_construct,
1593 netdev_bsd_rx_destruct,
1594 netdev_bsd_rx_dealloc,
1597 netdev_bsd_rx_drain,
1602 destroy_tap(int fd, const char *name)
1607 strcpy(ifr.ifr_name, name);
1608 /* XXX What to do if this call fails? */
1609 af_inet_ioctl(SIOCIFDESTROY, &ifr);
1613 get_flags(const struct netdev *netdev, int *flags)
1618 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev), &ifr,
1619 SIOCGIFFLAGS, "SIOCGIFFLAGS");
1621 *flags = ifr_get_flags(&ifr);
1627 set_flags(const char *name, int flags)
1631 ifr_set_flags(&ifr, flags);
1633 return af_inet_ifreq_ioctl(name, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS");
1637 get_ifindex(const struct netdev *netdev_, int *ifindexp)
1639 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1641 if (!(netdev->cache_valid & VALID_IFINDEX)) {
1642 int ifindex = if_nametoindex(netdev_get_name(netdev_));
1646 netdev->cache_valid |= VALID_IFINDEX;
1647 netdev->ifindex = ifindex;
1649 *ifindexp = netdev->ifindex;
1654 get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN])
1656 struct ifaddrs *head;
1657 struct ifaddrs *ifa;
1658 struct sockaddr_dl *sdl;
1660 if (getifaddrs(&head) != 0) {
1661 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1662 ovs_strerror(errno));
1666 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1667 if (ifa->ifa_addr->sa_family == AF_LINK) {
1668 if (!strcmp(ifa->ifa_name, netdev_name)) {
1669 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1671 memcpy(ea, LLADDR(sdl), sdl->sdl_alen);
1679 VLOG_ERR("could not find ethernet address for %s device", netdev_name);
1685 set_etheraddr(const char *netdev_name OVS_UNUSED, int hwaddr_family OVS_UNUSED,
1686 int hwaddr_len OVS_UNUSED,
1687 const uint8_t mac[ETH_ADDR_LEN] OVS_UNUSED)
1689 #if defined(__FreeBSD__)
1693 memset(&ifr, 0, sizeof ifr);
1694 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1695 ifr.ifr_addr.sa_family = hwaddr_family;
1696 ifr.ifr_addr.sa_len = hwaddr_len;
1697 memcpy(ifr.ifr_addr.sa_data, mac, hwaddr_len);
1698 error = af_inet_ioctl(SIOCSIFLLADDR, &ifr);
1700 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1701 netdev_name, ovs_strerror(error));
1705 #elif defined(__NetBSD__)
1706 struct if_laddrreq req;
1707 struct sockaddr_dl *sdl;
1708 struct sockaddr_storage oldaddr;
1712 * get the old address, add new one, and then remove old one.
1715 if (hwaddr_len != ETH_ADDR_LEN) {
1716 /* just to be safe about sockaddr storage size */
1719 memset(&req, 0, sizeof(req));
1720 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1721 req.addr.ss_len = sizeof(req.addr);
1722 req.addr.ss_family = hwaddr_family;
1723 sdl = (struct sockaddr_dl *)&req.addr;
1724 sdl->sdl_alen = hwaddr_len;
1726 error = af_link_ioctl(SIOCGLIFADDR, &req);
1730 if (!memcmp(&sdl->sdl_data[sdl->sdl_nlen], mac, hwaddr_len)) {
1735 memset(&req, 0, sizeof(req));
1736 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1737 req.flags = IFLR_ACTIVE;
1738 sdl = (struct sockaddr_dl *)&req.addr;
1739 sdl->sdl_len = offsetof(struct sockaddr_dl, sdl_data) + hwaddr_len;
1740 sdl->sdl_alen = hwaddr_len;
1741 sdl->sdl_family = hwaddr_family;
1742 memcpy(sdl->sdl_data, mac, hwaddr_len);
1743 error = af_link_ioctl(SIOCALIFADDR, &req);
1748 memset(&req, 0, sizeof(req));
1749 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1751 return af_link_ioctl(SIOCDLIFADDR, &req);
1753 #error not implemented
1758 ifr_get_flags(const struct ifreq *ifr)
1760 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1761 return (ifr->ifr_flagshigh << 16) | ifr->ifr_flags;
1763 return ifr->ifr_flags;
1768 ifr_set_flags(struct ifreq *ifr, int flags)
1770 ifr->ifr_flags = flags;
1771 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1772 ifr->ifr_flagshigh = flags >> 16;
1776 /* Calls ioctl() on an AF_LINK sock, passing the specified 'command' and
1777 * 'arg'. Returns 0 if successful, otherwise a positive errno value. */
1779 af_link_ioctl(unsigned long command, const void *arg)
1781 static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
1784 if (ovsthread_once_start(&once)) {
1785 sock = socket(AF_LINK, SOCK_DGRAM, 0);
1788 VLOG_ERR("failed to create link socket: %s", ovs_strerror(errno));
1790 ovsthread_once_done(&once);
1793 return (sock < 0 ? -sock
1794 : ioctl(sock, command, arg) == -1 ? errno