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"
54 #include "poll-loop.h"
55 #include "socket-util.h"
61 VLOG_DEFINE_THIS_MODULE(netdev_bsd);
64 struct netdev_rx_bsd {
67 /* Packet capture descriptor for a system network device.
68 * For a tap device this is NULL. */
71 /* Selectable file descriptor for the network device.
72 * This descriptor will be used for polling operations. */
76 static const struct netdev_rx_class netdev_rx_bsd_class;
80 unsigned int cache_valid;
81 unsigned int change_seq;
84 uint8_t etheraddr[ETH_ADDR_LEN];
86 struct in_addr netmask;
91 int tap_fd; /* TAP character device, if any, otherwise -1. */
93 /* Used for sending packets on non-tap devices. */
102 VALID_IFINDEX = 1 << 0,
103 VALID_ETHERADDR = 1 << 1,
107 VALID_CARRIER = 1 << 5
110 #if defined(__NetBSD__)
111 /* AF_LINK socket used for netdev_bsd_get_stats and set_etheraddr */
112 static int af_link_sock = -1;
113 #endif /* defined(__NetBSD__) */
115 #define PCAP_SNAPLEN 2048
119 * Notifier used to invalidate device informations in case of status change.
121 * It will be registered with a 'rtbsd_notifier_register()' when the first
122 * device will be created with the call of either 'netdev_bsd_tap_create()' or
123 * 'netdev_bsd_system_create()'.
125 * The callback associated with this notifier ('netdev_bsd_cache_cb()') will
126 * invalidate cached information about the device.
128 static struct rtbsd_notifier netdev_bsd_cache_notifier;
129 static int cache_notifier_refcount;
131 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
133 static void destroy_tap(int fd, const char *name);
134 static int get_flags(const struct netdev *, int *flagsp);
135 static int set_flags(const char *, int flags);
136 static int do_set_addr(struct netdev *netdev,
137 int ioctl_nr, const char *ioctl_name,
138 struct in_addr addr);
139 static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]);
140 static int set_etheraddr(const char *netdev_name, int hwaddr_family,
141 int hwaddr_len, const uint8_t[ETH_ADDR_LEN]);
142 static int get_ifindex(const struct netdev *, int *ifindexp);
144 static int ifr_get_flags(const struct ifreq *);
145 static void ifr_set_flags(struct ifreq *, int flags);
147 static int netdev_bsd_init(void);
150 is_netdev_bsd_class(const struct netdev_class *netdev_class)
152 return netdev_class->init == netdev_bsd_init;
155 static struct netdev_bsd *
156 netdev_bsd_cast(const struct netdev *netdev)
158 ovs_assert(is_netdev_bsd_class(netdev_get_class(netdev)));
159 return CONTAINER_OF(netdev, struct netdev_bsd, up);
162 static struct netdev_rx_bsd *
163 netdev_rx_bsd_cast(const struct netdev_rx *rx)
165 netdev_rx_assert_class(rx, &netdev_rx_bsd_class);
166 return CONTAINER_OF(rx, struct netdev_rx_bsd, up);
170 netdev_get_kernel_name(const struct netdev *netdev)
172 return netdev_bsd_cast(netdev)->kernel_name;
175 /* Initialize the AF_INET socket used for ioctl operations */
177 netdev_bsd_init(void)
179 #if defined(__NetBSD__)
180 static int status = -1;
182 if (status >= 0) { /* already initialized */
186 af_link_sock = socket(AF_LINK, SOCK_DGRAM, 0);
187 status = af_link_sock >= 0 ? 0 : errno;
189 VLOG_ERR("failed to create link socket: %s", ovs_strerror(status));
195 #endif /* defined(__NetBSD__) */
199 * Perform periodic work needed by netdev. In BSD netdevs it checks for any
200 * interface status changes, and eventually calls all the user callbacks.
205 rtbsd_notifier_run();
209 * Arranges for poll_block() to wake up if the "run" member function needs to
213 netdev_bsd_wait(void)
215 rtbsd_notifier_wait();
219 netdev_bsd_changed(struct netdev_bsd *dev)
222 if (!dev->change_seq) {
227 /* Invalidate cache in case of interface status change. */
229 netdev_bsd_cache_cb(const struct rtbsd_change *change,
230 void *aux OVS_UNUSED)
232 struct netdev_bsd *dev;
235 struct netdev *base_dev = netdev_from_name(change->if_name);
238 const struct netdev_class *netdev_class =
239 netdev_get_class(base_dev);
241 if (is_netdev_bsd_class(netdev_class)) {
242 dev = netdev_bsd_cast(base_dev);
243 dev->cache_valid = 0;
244 netdev_bsd_changed(dev);
246 netdev_close(base_dev);
250 * XXX the API is lacking, we should be able to iterate on the list of
251 * netdevs without having to store the info in a temp shash.
253 struct shash device_shash;
254 struct shash_node *node;
256 shash_init(&device_shash);
257 netdev_get_devices(&netdev_bsd_class, &device_shash);
258 SHASH_FOR_EACH (node, &device_shash) {
259 struct netdev *netdev = node->data;
260 dev = netdev_bsd_cast(netdev);
261 dev->cache_valid = 0;
262 netdev_bsd_changed(dev);
263 netdev_close(netdev);
265 shash_destroy(&device_shash);
270 cache_notifier_ref(void)
274 if (!cache_notifier_refcount) {
275 ret = rtbsd_notifier_register(&netdev_bsd_cache_notifier,
276 netdev_bsd_cache_cb, NULL);
281 cache_notifier_refcount++;
286 cache_notifier_unref(void)
288 cache_notifier_refcount--;
289 if (cache_notifier_refcount == 0) {
290 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier);
295 /* Allocate a netdev_bsd structure */
297 netdev_bsd_create_system(const struct netdev_class *class, const char *name,
298 struct netdev **netdevp)
300 struct netdev_bsd *netdev;
301 enum netdev_flags flags;
304 error = cache_notifier_ref();
309 netdev = xzalloc(sizeof *netdev);
310 netdev->change_seq = 1;
311 netdev_init(&netdev->up, name, class);
313 netdev->kernel_name = xstrdup(name);
315 /* Verify that the netdev really exists by attempting to read its flags */
316 error = netdev_get_flags(&netdev->up, &flags);
317 if (error == ENXIO) {
318 free(netdev->kernel_name);
319 netdev_uninit(&netdev->up, false);
321 cache_notifier_unref();
325 *netdevp = &netdev->up;
330 * Allocate a netdev_bsd structure with 'tap' class.
333 netdev_bsd_create_tap(const struct netdev_class *class, const char *name,
334 struct netdev **netdevp)
336 struct netdev_bsd *netdev = NULL;
339 char *kernel_name = NULL;
341 error = cache_notifier_ref();
346 /* allocate the device structure and set the internal flag */
347 netdev = xzalloc(sizeof *netdev);
349 memset(&ifr, 0, sizeof(ifr));
351 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
352 * to retrieve the name of the tap device. */
353 netdev->tap_fd = open("/dev/tap", O_RDWR);
354 netdev->change_seq = 1;
355 if (netdev->tap_fd < 0) {
357 VLOG_WARN("opening \"/dev/tap\" failed: %s", ovs_strerror(error));
358 goto error_unref_notifier;
361 /* Retrieve tap name (e.g. tap0) */
362 if (ioctl(netdev->tap_fd, TAPGIFNAME, &ifr) == -1) {
363 /* XXX Need to destroy the device? */
365 close(netdev->tap_fd);
366 goto error_unref_notifier;
369 /* Change the name of the tap device */
370 #if defined(SIOCSIFNAME)
371 ifr.ifr_data = (void *)name;
372 error = af_inet_ioctl(SIOCSIFNAME, &ifr);
374 destroy_tap(netdev->tap_fd, ifr.ifr_name);
375 goto error_unref_notifier;
377 kernel_name = xstrdup(name);
380 * NetBSD doesn't support inteface renaming.
382 VLOG_INFO("tap %s is created for bridge %s", ifr.ifr_name, name);
383 kernel_name = xstrdup(ifr.ifr_name);
386 /* set non-blocking. */
387 error = set_nonblocking(netdev->tap_fd);
389 destroy_tap(netdev->tap_fd, kernel_name);
390 goto error_unref_notifier;
394 ifr_set_flags(&ifr, IFF_UP);
395 strncpy(ifr.ifr_name, kernel_name, sizeof ifr.ifr_name);
396 error = af_inet_ioctl(SIOCSIFFLAGS, &ifr);
398 destroy_tap(netdev->tap_fd, kernel_name);
399 goto error_unref_notifier;
402 /* initialize the device structure and
403 * link the structure to its netdev */
404 netdev_init(&netdev->up, name, class);
405 netdev->kernel_name = kernel_name;
406 *netdevp = &netdev->up;
410 error_unref_notifier:
411 cache_notifier_unref();
419 netdev_bsd_destroy(struct netdev *netdev_)
421 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
423 cache_notifier_unref();
425 if (netdev->tap_fd >= 0) {
426 destroy_tap(netdev->tap_fd, netdev_get_kernel_name(netdev_));
429 pcap_close(netdev->pcap);
431 free(netdev->kernel_name);
436 netdev_bsd_open_pcap(const char *name, pcap_t **pcapp, int *fdp)
438 char errbuf[PCAP_ERRBUF_SIZE];
444 /* Open the pcap device. The device is opened in non-promiscuous mode
445 * because the interface flags are manually set by the caller. */
447 pcap = pcap_open_live(name, PCAP_SNAPLEN, 0, 1000, errbuf);
449 VLOG_ERR_RL(&rl, "%s: pcap_open_live failed: %s", name, errbuf);
453 if (errbuf[0] != '\0') {
454 VLOG_WARN_RL(&rl, "%s: pcap_open_live: %s", name, errbuf);
457 /* Get the underlying fd. */
458 fd = pcap_get_selectable_fd(pcap);
460 VLOG_WARN_RL(&rl, "%s: no selectable file descriptor", name);
465 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
466 * on the file descriptor returned by pcap_get_selectable_fd to achieve
467 * a real non-blocking behaviour.*/
468 error = pcap_setnonblock(pcap, 1, errbuf);
474 /* This call assure that reads return immediately upon packet
475 * reception. Otherwise, a read will block until either the kernel
476 * buffer becomes full or a timeout occurs. */
477 if (ioctl(fd, BIOCIMMEDIATE, &one) < 0 ) {
478 VLOG_ERR_RL(&rl, "ioctl(BIOCIMMEDIATE) on %s device failed: %s",
479 name, ovs_strerror(errno));
484 /* Capture only incoming packets. */
485 error = pcap_setdirection(pcap, PCAP_D_IN);
505 netdev_bsd_rx_open(struct netdev *netdev_, struct netdev_rx **rxp)
507 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
509 struct netdev_rx_bsd *rx;
513 if (!strcmp(netdev_get_type(netdev_), "tap")) {
517 int error = netdev_bsd_open_pcap(netdev_get_kernel_name(netdev_),
523 netdev_bsd_changed(netdev);
526 rx = xmalloc(sizeof *rx);
527 netdev_rx_init(&rx->up, netdev_, &netdev_rx_bsd_class);
528 rx->pcap_handle = pcap;
536 netdev_rx_bsd_destroy(struct netdev_rx *rx_)
538 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
540 if (rx->pcap_handle) {
541 pcap_close(rx->pcap_handle);
546 /* The recv callback of the netdev class returns the number of bytes of the
549 * This can be done by the pcap_next() function. Unfortunately pcap_next() does
550 * not make difference between a missing packet on the capture interface and
551 * an error during the file capture. We can use the pcap_dispatch() function
552 * instead, which is able to distinguish between errors and null packet.
554 * To make pcap_dispatch() returns the number of bytes read from the interface
555 * we need to define the following callback and argument.
564 * This callback will be executed on every captured packet.
566 * If the packet captured by pcap_dispatch() does not fit the pcap buffer,
567 * pcap returns a truncated packet and we follow this behavior.
569 * The argument args->retval is the packet size in bytes.
572 proc_pkt(u_char *args_, const struct pcap_pkthdr *hdr, const u_char *packet)
574 struct pcap_arg *args = (struct pcap_arg *)args_;
576 if (args->size < hdr->len) {
577 VLOG_WARN_RL(&rl, "packet truncated");
578 args->retval = args->size;
580 args->retval = hdr->len;
583 /* copy the packet to our buffer */
584 memcpy(args->data, packet, args->retval);
588 * This function attempts to receive a packet from the specified network
589 * device. It is assumed that the network device is a system device or a tap
590 * device opened as a system one. In this case the read operation is performed
594 netdev_rx_bsd_recv_pcap(struct netdev_rx_bsd *rx, void *data, size_t size)
599 /* prepare the pcap argument to store the packet */
604 ret = pcap_dispatch(rx->pcap_handle, 1, proc_pkt, (u_char *) &arg);
607 return arg.retval; /* arg.retval < 0 is handled in the caller */
610 if (errno == EINTR) {
620 * This function attempts to receive a packet from the specified network
621 * device. It is assumed that the network device is a tap device and
622 * 'rx->fd' is initialized with the tap file descriptor.
625 netdev_rx_bsd_recv_tap(struct netdev_rx_bsd *rx, void *data, size_t size)
628 ssize_t retval = read(rx->fd, data, size);
631 } else if (errno != EINTR) {
632 if (errno != EAGAIN) {
633 VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
634 ovs_strerror(errno), netdev_rx_get_name(&rx->up));
643 netdev_rx_bsd_recv(struct netdev_rx *rx_, void *data, size_t size)
645 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
647 return (rx->pcap_handle
648 ? netdev_rx_bsd_recv_pcap(rx, data, size)
649 : netdev_rx_bsd_recv_tap(rx, data, size));
653 * Registers with the poll loop to wake up from the next call to poll_block()
654 * when a packet is ready to be received with netdev_rx_recv() on 'rx'.
657 netdev_rx_bsd_wait(struct netdev_rx *rx_)
659 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
661 poll_fd_wait(rx->fd, POLLIN);
664 /* Discards all packets waiting to be received from 'rx'. */
666 netdev_rx_bsd_drain(struct netdev_rx *rx_)
669 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
671 strcpy(ifr.ifr_name, netdev_get_kernel_name(netdev_rx_get_netdev(rx_)));
672 if (ioctl(rx->fd, BIOCFLUSH, &ifr) == -1) {
673 VLOG_DBG_RL(&rl, "%s: ioctl(BIOCFLUSH) failed: %s",
674 netdev_rx_get_name(rx_), ovs_strerror(errno));
681 * Send a packet on the specified network device. The device could be either a
682 * system or a tap device.
685 netdev_bsd_send(struct netdev *netdev_, const void *data, size_t size)
687 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
688 const char *name = netdev_get_name(netdev_);
690 if (dev->tap_fd < 0 && !dev->pcap) {
691 int error = netdev_bsd_open_pcap(name, &dev->pcap, &dev->fd);
699 if (dev->tap_fd >= 0) {
700 retval = write(dev->tap_fd, data, size);
702 retval = pcap_inject(dev->pcap, data, size);
705 if (errno == EINTR) {
707 } else if (errno != EAGAIN) {
708 VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s",
709 name, ovs_strerror(errno));
712 } else if (retval != size) {
713 VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of "
714 "%zu) on %s", retval, size, name);
723 * Registers with the poll loop to wake up from the next call to poll_block()
724 * when the packet transmission queue has sufficient room to transmit a packet
725 * with netdev_send().
728 netdev_bsd_send_wait(struct netdev *netdev_)
730 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
732 if (dev->tap_fd >= 0) {
733 /* TAP device always accepts packets. */
734 poll_immediate_wake();
735 } else if (dev->pcap) {
736 poll_fd_wait(dev->fd, POLLOUT);
738 /* We haven't even tried to send a packet yet. */
739 poll_immediate_wake();
744 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
745 * otherwise a positive errno value.
748 netdev_bsd_set_etheraddr(struct netdev *netdev_,
749 const uint8_t mac[ETH_ADDR_LEN])
751 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
754 if (!(netdev->cache_valid & VALID_ETHERADDR)
755 || !eth_addr_equals(netdev->etheraddr, mac)) {
756 error = set_etheraddr(netdev_get_kernel_name(netdev_), AF_LINK,
759 netdev->cache_valid |= VALID_ETHERADDR;
760 memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN);
761 netdev_bsd_changed(netdev);
770 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
771 * free the returned buffer.
774 netdev_bsd_get_etheraddr(const struct netdev *netdev_,
775 uint8_t mac[ETH_ADDR_LEN])
777 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
779 if (!(netdev->cache_valid & VALID_ETHERADDR)) {
780 int error = get_etheraddr(netdev_get_kernel_name(netdev_),
785 netdev->cache_valid |= VALID_ETHERADDR;
787 memcpy(mac, netdev->etheraddr, ETH_ADDR_LEN);
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_);
802 if (!(netdev->cache_valid & VALID_MTU)) {
806 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
807 SIOCGIFMTU, "SIOCGIFMTU");
811 netdev->mtu = ifr.ifr_mtu;
812 netdev->cache_valid |= VALID_MTU;
820 netdev_bsd_get_ifindex(const struct netdev *netdev)
824 error = get_ifindex(netdev, &ifindex);
825 return error ? -error : ifindex;
829 netdev_bsd_get_carrier(const struct netdev *netdev_, bool *carrier)
831 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
833 if (!(netdev->cache_valid & VALID_CARRIER)) {
834 struct ifmediareq ifmr;
837 memset(&ifmr, 0, sizeof(ifmr));
838 strncpy(ifmr.ifm_name, netdev_get_kernel_name(netdev_),
839 sizeof ifmr.ifm_name);
841 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
843 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
844 netdev_get_name(netdev_), ovs_strerror(error));
848 netdev->carrier = (ifmr.ifm_status & IFM_ACTIVE) == IFM_ACTIVE;
849 netdev->cache_valid |= VALID_CARRIER;
851 /* If the interface doesn't report whether the media is active,
852 * just assume it is active. */
853 if ((ifmr.ifm_status & IFM_AVALID) == 0) {
854 netdev->carrier = true;
857 *carrier = netdev->carrier;
863 convert_stats(struct netdev_stats *stats, const struct if_data *ifd)
866 * note: UINT64_MAX means unsupported
868 stats->rx_packets = ifd->ifi_ipackets;
869 stats->tx_packets = ifd->ifi_opackets;
870 stats->rx_bytes = ifd->ifi_obytes;
871 stats->tx_bytes = ifd->ifi_ibytes;
872 stats->rx_errors = ifd->ifi_ierrors;
873 stats->tx_errors = ifd->ifi_oerrors;
874 stats->rx_dropped = ifd->ifi_iqdrops;
875 stats->tx_dropped = UINT64_MAX;
876 stats->multicast = ifd->ifi_imcasts;
877 stats->collisions = ifd->ifi_collisions;
878 stats->rx_length_errors = UINT64_MAX;
879 stats->rx_over_errors = UINT64_MAX;
880 stats->rx_crc_errors = UINT64_MAX;
881 stats->rx_frame_errors = UINT64_MAX;
882 stats->rx_fifo_errors = UINT64_MAX;
883 stats->rx_missed_errors = UINT64_MAX;
884 stats->tx_aborted_errors = UINT64_MAX;
885 stats->tx_carrier_errors = UINT64_MAX;
886 stats->tx_fifo_errors = UINT64_MAX;
887 stats->tx_heartbeat_errors = UINT64_MAX;
888 stats->tx_window_errors = UINT64_MAX;
891 /* Retrieves current device stats for 'netdev'. */
893 netdev_bsd_get_stats(const struct netdev *netdev_, struct netdev_stats *stats)
895 #if defined(__FreeBSD__)
899 struct ifmibdata ifmd;
904 mib[2] = NETLINK_GENERIC;
905 mib[3] = IFMIB_SYSTEM;
906 mib[4] = IFMIB_IFCOUNT;
908 len = sizeof(if_count);
910 if (sysctl(mib, 5, &if_count, &len, (void *)0, 0) == -1) {
911 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
912 netdev_get_name(netdev_), ovs_strerror(errno));
916 mib[5] = IFDATA_GENERAL;
917 mib[3] = IFMIB_IFDATA;
919 for (i = 1; i <= if_count; i++) {
921 if (sysctl(mib, 6, &ifmd, &len, (void *)0, 0) == -1) {
922 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
923 netdev_get_name(netdev_), ovs_strerror(errno));
925 } else if (!strcmp(ifmd.ifmd_name, netdev_get_name(netdev_))) {
926 convert_stats(stats, &ifmd.ifmd_data);
932 #elif defined(__NetBSD__)
933 struct ifdatareq ifdr;
937 memset(&ifdr, 0, sizeof(ifdr));
938 strncpy(ifdr.ifdr_name, netdev_get_kernel_name(netdev_),
939 sizeof(ifdr.ifdr_name));
940 ret = ioctl(af_link_sock, SIOCGIFDATA, &ifdr);
945 convert_stats(stats, &ifdr.ifdr_data);
948 #error not implemented
953 netdev_bsd_parse_media(int media)
955 uint32_t supported = 0;
956 bool half_duplex = media & IFM_HDX ? true : false;
958 switch (IFM_SUBTYPE(media)) {
963 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
964 supported |= NETDEV_F_COPPER;
968 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
969 supported |= NETDEV_F_FIBER;
976 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
977 supported |= NETDEV_F_COPPER;
981 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
982 supported |= NETDEV_F_FIBER;
987 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
988 supported |= NETDEV_F_COPPER;
993 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
994 supported |= NETDEV_F_FIBER;
998 supported |= NETDEV_F_10GB_FD;
999 supported |= NETDEV_F_COPPER;
1004 supported |= NETDEV_F_10GB_FD;
1005 supported |= NETDEV_F_FIBER;
1012 if (IFM_SUBTYPE(media) == IFM_AUTO) {
1013 supported |= NETDEV_F_AUTONEG;
1016 if (media & IFM_ETH_FMASK) {
1017 supported |= NETDEV_F_PAUSE;
1025 * Stores the features supported by 'netdev' into each of '*current',
1026 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1027 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1028 * successful, otherwise a positive errno value. On failure, all of the
1029 * passed-in values are set to 0.
1032 netdev_bsd_get_features(const struct netdev *netdev,
1033 enum netdev_features *current, uint32_t *advertised,
1034 enum netdev_features *supported, uint32_t *peer)
1036 struct ifmediareq ifmr;
1042 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1044 memset(&ifmr, 0, sizeof(ifmr));
1045 strncpy(ifmr.ifm_name, netdev_get_name(netdev), sizeof ifmr.ifm_name);
1047 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1048 * number of supported modes, and a second with a buffer to retrieve
1050 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
1052 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1053 netdev_get_name(netdev), ovs_strerror(error));
1057 media_list = xcalloc(ifmr.ifm_count, sizeof(int));
1058 ifmr.ifm_ulist = media_list;
1060 if (IFM_TYPE(ifmr.ifm_current) != IFM_ETHER) {
1061 VLOG_DBG_RL(&rl, "%s: doesn't appear to be ethernet",
1062 netdev_get_name(netdev));
1067 error = af_inet_ioctl(SIOCGIFMEDIA, &ifmr);
1069 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1070 netdev_get_name(netdev), ovs_strerror(error));
1074 /* Current settings. */
1075 *current = netdev_bsd_parse_media(ifmr.ifm_active);
1077 /* Advertised features. */
1078 *advertised = netdev_bsd_parse_media(ifmr.ifm_current);
1080 /* Supported features. */
1082 for (i = 0; i < ifmr.ifm_count; i++) {
1083 *supported |= netdev_bsd_parse_media(ifmr.ifm_ulist[i]);
1086 /* Peer advertisements. */
1087 *peer = 0; /* XXX */
1096 * If 'netdev' has an assigned IPv4 address, sets '*in4' to that address and
1097 * '*netmask' to its netmask and returns true. Otherwise, returns false.
1100 netdev_bsd_get_in4(const struct netdev *netdev_, struct in_addr *in4,
1101 struct in_addr *netmask)
1103 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1105 if (!(netdev->cache_valid & VALID_IN4)) {
1106 const struct sockaddr_in *sin;
1110 ifr.ifr_addr.sa_family = AF_INET;
1111 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1112 SIOCGIFADDR, "SIOCGIFADDR");
1117 sin = (struct sockaddr_in *) &ifr.ifr_addr;
1118 netdev->in4 = sin->sin_addr;
1119 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1120 SIOCGIFNETMASK, "SIOCGIFNETMASK");
1124 netdev->netmask = sin->sin_addr;
1125 netdev->cache_valid |= VALID_IN4;
1128 *netmask = netdev->netmask;
1130 return in4->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0;
1134 * Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
1135 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
1136 * positive errno value.
1139 netdev_bsd_set_in4(struct netdev *netdev_, struct in_addr addr,
1140 struct in_addr mask)
1142 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1145 error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", addr);
1147 if (addr.s_addr != INADDR_ANY) {
1148 error = do_set_addr(netdev_, SIOCSIFNETMASK,
1149 "SIOCSIFNETMASK", mask);
1151 netdev->cache_valid |= VALID_IN4;
1153 netdev->netmask = mask;
1156 netdev_bsd_changed(netdev);
1162 netdev_bsd_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
1164 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1165 if (!(netdev->cache_valid & VALID_IN6)) {
1166 struct ifaddrs *ifa, *head;
1167 struct sockaddr_in6 *sin6;
1168 const char *netdev_name = netdev_get_name(netdev_);
1170 if (getifaddrs(&head) != 0) {
1171 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1172 ovs_strerror(errno));
1176 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1177 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1178 !strcmp(ifa->ifa_name, netdev_name)) {
1179 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1181 memcpy(&netdev->in6, &sin6->sin6_addr, sin6->sin6_len);
1182 netdev->cache_valid |= VALID_IN6;
1189 return EADDRNOTAVAIL;
1195 #if defined(__NetBSD__)
1197 netdev_bsd_kernel_name_to_ovs_name(const char *kernel_name)
1199 char *ovs_name = NULL;
1200 struct shash device_shash;
1201 struct shash_node *node;
1203 shash_init(&device_shash);
1204 netdev_get_devices(&netdev_tap_class, &device_shash);
1205 SHASH_FOR_EACH(node, &device_shash) {
1206 struct netdev *netdev = node->data;
1207 struct netdev_bsd * const dev = netdev_bsd_cast(netdev);
1209 if (!strcmp(dev->kernel_name, kernel_name)) {
1211 ovs_name = xstrdup(netdev_get_name(&dev->up));
1213 netdev_close(netdev);
1215 shash_destroy(&device_shash);
1217 return ovs_name ? ovs_name : xstrdup(kernel_name);
1222 netdev_bsd_get_next_hop(const struct in_addr *host OVS_UNUSED,
1223 struct in_addr *next_hop OVS_UNUSED,
1224 char **netdev_name OVS_UNUSED)
1226 #if defined(__NetBSD__)
1228 struct sockaddr_in sin;
1229 struct sockaddr_dl sdl;
1236 struct rt_msghdr *rtm = &buf.h;
1237 const pid_t pid = getpid();
1240 bool gateway = false;
1241 char *ifname = NULL;
1244 memset(next_hop, 0, sizeof(*next_hop));
1245 *netdev_name = NULL;
1247 memset(&sin, 0, sizeof(sin));
1248 sin.sin_len = sizeof(sin);
1249 sin.sin_family = AF_INET;
1251 sin.sin_addr = *host;
1253 memset(&sdl, 0, sizeof(sdl));
1254 sdl.sdl_len = sizeof(sdl);
1255 sdl.sdl_family = AF_LINK;
1257 s = socket(PF_ROUTE, SOCK_RAW, 0);
1258 memset(&buf, 0, sizeof(buf));
1259 rtm->rtm_flags = RTF_HOST|RTF_UP;
1260 rtm->rtm_version = RTM_VERSION;
1261 rtm->rtm_addrs = RTA_DST|RTA_IFP;
1262 cp = (void *)&buf.space;
1263 memcpy(cp, &sin, sizeof(sin));
1264 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sin);
1265 memcpy(cp, &sdl, sizeof(sdl));
1266 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sdl);
1267 rtm->rtm_msglen = cp - (char *)(void *)rtm;
1268 rtm->rtm_seq = ++seq;
1269 rtm->rtm_type = RTM_GET;
1271 write(s, rtm, rtm->rtm_msglen);
1272 memset(&buf, 0, sizeof(buf));
1274 ssz = read(s, &buf, sizeof(buf));
1275 } while (ssz > 0 && (rtm->rtm_seq != seq || rtm->rtm_pid != pid));
1276 saved_errno = errno;
1282 return EPIPE; /* XXX */
1284 cp = (void *)&buf.space;
1285 for (i = 1; i; i <<= 1) {
1286 if ((rtm->rtm_addrs & i) != 0) {
1287 const struct sockaddr *sa = (const void *)cp;
1289 if ((i == RTA_GATEWAY) && sa->sa_family == AF_INET) {
1290 const struct sockaddr_in * const sin =
1291 (const struct sockaddr_in *)sa;
1293 *next_hop = sin->sin_addr;
1296 if ((i == RTA_IFP) && sa->sa_family == AF_LINK) {
1297 const struct sockaddr_dl * const sdl =
1298 (const struct sockaddr_dl *)sa;
1301 kernel_name = xmemdup0(sdl->sdl_data, sdl->sdl_nlen);
1302 ifname = netdev_bsd_kernel_name_to_ovs_name(kernel_name);
1308 if (ifname == NULL) {
1314 *netdev_name = ifname;
1315 VLOG_DBG("host " IP_FMT " next-hop " IP_FMT " if %s",
1316 IP_ARGS(host->s_addr), IP_ARGS(next_hop->s_addr), *netdev_name);
1324 make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
1326 struct sockaddr_in sin;
1327 memset(&sin, 0, sizeof sin);
1328 sin.sin_family = AF_INET;
1329 sin.sin_addr = addr;
1332 memset(sa, 0, sizeof *sa);
1333 memcpy(sa, &sin, sizeof sin);
1337 do_set_addr(struct netdev *netdev,
1338 int ioctl_nr, const char *ioctl_name, struct in_addr addr)
1341 make_in4_sockaddr(&ifr.ifr_addr, addr);
1342 return af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev), &ifr, ioctl_nr,
1347 nd_to_iff_flags(enum netdev_flags nd)
1350 if (nd & NETDEV_UP) {
1353 if (nd & NETDEV_PROMISC) {
1355 #if defined(IFF_PPROMISC)
1356 iff |= IFF_PPROMISC;
1363 iff_to_nd_flags(int iff)
1365 enum netdev_flags nd = 0;
1369 if (iff & IFF_PROMISC) {
1370 nd |= NETDEV_PROMISC;
1376 netdev_bsd_update_flags(struct netdev *netdev_, enum netdev_flags off,
1377 enum netdev_flags on, enum netdev_flags *old_flagsp)
1379 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1380 int old_flags, new_flags;
1383 error = get_flags(netdev_, &old_flags);
1385 *old_flagsp = iff_to_nd_flags(old_flags);
1386 new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on);
1387 if (new_flags != old_flags) {
1388 error = set_flags(netdev_get_kernel_name(netdev_), new_flags);
1389 netdev_bsd_changed(netdev);
1396 netdev_bsd_change_seq(const struct netdev *netdev)
1398 return netdev_bsd_cast(netdev)->change_seq;
1402 const struct netdev_class netdev_bsd_class = {
1408 netdev_bsd_create_system,
1410 NULL, /* get_config */
1411 NULL, /* set_config */
1412 NULL, /* get_tunnel_config */
1417 netdev_bsd_send_wait,
1419 netdev_bsd_set_etheraddr,
1420 netdev_bsd_get_etheraddr,
1423 netdev_bsd_get_ifindex,
1424 netdev_bsd_get_carrier,
1425 NULL, /* get_carrier_resets */
1426 NULL, /* set_miimon_interval */
1427 netdev_bsd_get_stats,
1428 NULL, /* set_stats */
1430 netdev_bsd_get_features,
1431 NULL, /* set_advertisement */
1432 NULL, /* set_policing */
1433 NULL, /* get_qos_type */
1434 NULL, /* get_qos_capabilities */
1437 NULL, /* get_queue */
1438 NULL, /* set_queue */
1439 NULL, /* delete_queue */
1440 NULL, /* get_queue_stats */
1441 NULL, /* dump_queue */
1442 NULL, /* dump_queue_stats */
1447 NULL, /* add_router */
1448 netdev_bsd_get_next_hop,
1449 NULL, /* get_status */
1450 NULL, /* arp_lookup */
1452 netdev_bsd_update_flags,
1454 netdev_bsd_change_seq
1457 const struct netdev_class netdev_tap_class = {
1463 netdev_bsd_create_tap,
1465 NULL, /* get_config */
1466 NULL, /* set_config */
1467 NULL, /* get_tunnel_config */
1472 netdev_bsd_send_wait,
1474 netdev_bsd_set_etheraddr,
1475 netdev_bsd_get_etheraddr,
1478 netdev_bsd_get_ifindex,
1479 netdev_bsd_get_carrier,
1480 NULL, /* get_carrier_resets */
1481 NULL, /* set_miimon_interval */
1482 netdev_bsd_get_stats,
1483 NULL, /* set_stats */
1485 netdev_bsd_get_features,
1486 NULL, /* set_advertisement */
1487 NULL, /* set_policing */
1488 NULL, /* get_qos_type */
1489 NULL, /* get_qos_capabilities */
1492 NULL, /* get_queue */
1493 NULL, /* set_queue */
1494 NULL, /* delete_queue */
1495 NULL, /* get_queue_stats */
1496 NULL, /* dump_queue */
1497 NULL, /* dump_queue_stats */
1502 NULL, /* add_router */
1503 netdev_bsd_get_next_hop,
1504 NULL, /* get_status */
1505 NULL, /* arp_lookup */
1507 netdev_bsd_update_flags,
1509 netdev_bsd_change_seq
1512 static const struct netdev_rx_class netdev_rx_bsd_class = {
1513 netdev_rx_bsd_destroy,
1516 netdev_rx_bsd_drain,
1521 destroy_tap(int fd, const char *name)
1526 strcpy(ifr.ifr_name, name);
1527 /* XXX What to do if this call fails? */
1528 af_inet_ioctl(SIOCIFDESTROY, &ifr);
1532 get_flags(const struct netdev *netdev, int *flags)
1537 error = af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev), &ifr,
1538 SIOCGIFFLAGS, "SIOCGIFFLAGS");
1540 *flags = ifr_get_flags(&ifr);
1546 set_flags(const char *name, int flags)
1550 ifr_set_flags(&ifr, flags);
1552 return af_inet_ifreq_ioctl(name, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS");
1556 get_ifindex(const struct netdev *netdev_, int *ifindexp)
1558 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1560 if (!(netdev->cache_valid & VALID_IFINDEX)) {
1561 int ifindex = if_nametoindex(netdev_get_name(netdev_));
1565 netdev->cache_valid |= VALID_IFINDEX;
1566 netdev->ifindex = ifindex;
1568 *ifindexp = netdev->ifindex;
1573 get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN])
1575 struct ifaddrs *head;
1576 struct ifaddrs *ifa;
1577 struct sockaddr_dl *sdl;
1579 if (getifaddrs(&head) != 0) {
1580 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1581 ovs_strerror(errno));
1585 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1586 if (ifa->ifa_addr->sa_family == AF_LINK) {
1587 if (!strcmp(ifa->ifa_name, netdev_name)) {
1588 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1590 memcpy(ea, LLADDR(sdl), sdl->sdl_alen);
1598 VLOG_ERR("could not find ethernet address for %s device", netdev_name);
1604 set_etheraddr(const char *netdev_name OVS_UNUSED, int hwaddr_family OVS_UNUSED,
1605 int hwaddr_len OVS_UNUSED,
1606 const uint8_t mac[ETH_ADDR_LEN] OVS_UNUSED)
1608 #if defined(__FreeBSD__)
1612 memset(&ifr, 0, sizeof ifr);
1613 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1614 ifr.ifr_addr.sa_family = hwaddr_family;
1615 ifr.ifr_addr.sa_len = hwaddr_len;
1616 memcpy(ifr.ifr_addr.sa_data, mac, hwaddr_len);
1617 error = af_inet_ioctl(SIOCSIFLLADDR, &ifr);
1619 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1620 netdev_name, ovs_strerror(error));
1624 #elif defined(__NetBSD__)
1625 struct if_laddrreq req;
1626 struct sockaddr_dl *sdl;
1627 struct sockaddr_storage oldaddr;
1631 * get the old address, add new one, and then remove old one.
1634 if (hwaddr_len != ETH_ADDR_LEN) {
1635 /* just to be safe about sockaddr storage size */
1638 memset(&req, 0, sizeof(req));
1639 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1640 req.addr.ss_len = sizeof(req.addr);
1641 req.addr.ss_family = hwaddr_family;
1642 sdl = (struct sockaddr_dl *)&req.addr;
1643 sdl->sdl_alen = hwaddr_len;
1644 ret = ioctl(af_link_sock, SIOCGLIFADDR, &req);
1648 if (!memcmp(&sdl->sdl_data[sdl->sdl_nlen], mac, hwaddr_len)) {
1653 memset(&req, 0, sizeof(req));
1654 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1655 req.flags = IFLR_ACTIVE;
1656 sdl = (struct sockaddr_dl *)&req.addr;
1657 sdl->sdl_len = offsetof(struct sockaddr_dl, sdl_data) + hwaddr_len;
1658 sdl->sdl_alen = hwaddr_len;
1659 sdl->sdl_family = hwaddr_family;
1660 memcpy(sdl->sdl_data, mac, hwaddr_len);
1661 ret = ioctl(af_link_sock, SIOCALIFADDR, &req);
1666 memset(&req, 0, sizeof(req));
1667 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1669 ret = ioctl(af_link_sock, SIOCDLIFADDR, &req);
1675 #error not implemented
1680 ifr_get_flags(const struct ifreq *ifr)
1682 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1683 return (ifr->ifr_flagshigh << 16) | ifr->ifr_flags;
1685 return ifr->ifr_flags;
1690 ifr_set_flags(struct ifreq *ifr, int flags)
1692 ifr->ifr_flags = flags;
1693 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1694 ifr->ifr_flagshigh = flags >> 16;