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 /* An AF_INET socket (used for ioctl operations). */
111 static int af_inet_sock = -1;
113 #if defined(__NetBSD__)
114 /* AF_LINK socket used for netdev_bsd_get_stats and set_etheraddr */
115 static int af_link_sock = -1;
116 #endif /* defined(__NetBSD__) */
118 #define PCAP_SNAPLEN 2048
122 * Notifier used to invalidate device informations in case of status change.
124 * It will be registered with a 'rtbsd_notifier_register()' when the first
125 * device will be created with the call of either 'netdev_bsd_tap_create()' or
126 * 'netdev_bsd_system_create()'.
128 * The callback associated with this notifier ('netdev_bsd_cache_cb()') will
129 * invalidate cached information about the device.
131 static struct rtbsd_notifier netdev_bsd_cache_notifier;
132 static int cache_notifier_refcount;
134 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
136 static int netdev_bsd_do_ioctl(const char *, struct ifreq *, unsigned long cmd,
137 const char *cmd_name);
138 static void destroy_tap(int fd, const char *name);
139 static int get_flags(const struct netdev *, int *flagsp);
140 static int set_flags(const char *, int flags);
141 static int do_set_addr(struct netdev *netdev,
142 int ioctl_nr, const char *ioctl_name,
143 struct in_addr addr);
144 static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]);
145 static int set_etheraddr(const char *netdev_name, int hwaddr_family,
146 int hwaddr_len, const uint8_t[ETH_ADDR_LEN]);
147 static int get_ifindex(const struct netdev *, int *ifindexp);
149 static int ifr_get_flags(const struct ifreq *);
150 static void ifr_set_flags(struct ifreq *, int flags);
152 static int netdev_bsd_init(void);
155 is_netdev_bsd_class(const struct netdev_class *netdev_class)
157 return netdev_class->init == netdev_bsd_init;
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 netdev_rx_assert_class(rx, &netdev_rx_bsd_class);
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;
180 /* Initialize the AF_INET socket used for ioctl operations */
182 netdev_bsd_init(void)
184 static int status = -1;
186 if (status >= 0) { /* already initialized */
190 af_inet_sock = socket(AF_INET, SOCK_DGRAM, 0);
191 status = af_inet_sock >= 0 ? 0 : errno;
193 VLOG_ERR("failed to create inet socket: %s", ovs_strerror(status));
197 #if defined(__NetBSD__)
198 af_link_sock = socket(AF_LINK, SOCK_DGRAM, 0);
199 status = af_link_sock >= 0 ? 0 : errno;
201 VLOG_ERR("failed to create link socket: %s", ovs_strerror(status));
205 #endif /* defined(__NetBSD__) */
211 * Perform periodic work needed by netdev. In BSD netdevs it checks for any
212 * interface status changes, and eventually calls all the user callbacks.
217 rtbsd_notifier_run();
221 * Arranges for poll_block() to wake up if the "run" member function needs to
225 netdev_bsd_wait(void)
227 rtbsd_notifier_wait();
231 netdev_bsd_changed(struct netdev_bsd *dev)
234 if (!dev->change_seq) {
239 /* Invalidate cache in case of interface status change. */
241 netdev_bsd_cache_cb(const struct rtbsd_change *change,
242 void *aux OVS_UNUSED)
244 struct netdev_bsd *dev;
247 struct netdev *base_dev = netdev_from_name(change->if_name);
250 const struct netdev_class *netdev_class =
251 netdev_get_class(base_dev);
253 if (is_netdev_bsd_class(netdev_class)) {
254 dev = netdev_bsd_cast(base_dev);
255 dev->cache_valid = 0;
256 netdev_bsd_changed(dev);
258 netdev_close(base_dev);
262 * XXX the API is lacking, we should be able to iterate on the list of
263 * netdevs without having to store the info in a temp shash.
265 struct shash device_shash;
266 struct shash_node *node;
268 shash_init(&device_shash);
269 netdev_get_devices(&netdev_bsd_class, &device_shash);
270 SHASH_FOR_EACH (node, &device_shash) {
271 struct netdev *netdev = node->data;
272 dev = netdev_bsd_cast(netdev);
273 dev->cache_valid = 0;
274 netdev_bsd_changed(dev);
275 netdev_close(netdev);
277 shash_destroy(&device_shash);
282 cache_notifier_ref(void)
286 if (!cache_notifier_refcount) {
287 ret = rtbsd_notifier_register(&netdev_bsd_cache_notifier,
288 netdev_bsd_cache_cb, NULL);
293 cache_notifier_refcount++;
298 cache_notifier_unref(void)
300 cache_notifier_refcount--;
301 if (cache_notifier_refcount == 0) {
302 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier);
307 /* Allocate a netdev_bsd structure */
309 netdev_bsd_create_system(const struct netdev_class *class, const char *name,
310 struct netdev **netdevp)
312 struct netdev_bsd *netdev;
313 enum netdev_flags flags;
316 error = cache_notifier_ref();
321 netdev = xzalloc(sizeof *netdev);
322 netdev->change_seq = 1;
323 netdev_init(&netdev->up, name, class);
325 netdev->kernel_name = xstrdup(name);
327 /* Verify that the netdev really exists by attempting to read its flags */
328 error = netdev_get_flags(&netdev->up, &flags);
329 if (error == ENXIO) {
330 free(netdev->kernel_name);
331 netdev_uninit(&netdev->up, false);
333 cache_notifier_unref();
337 *netdevp = &netdev->up;
342 * Allocate a netdev_bsd structure with 'tap' class.
345 netdev_bsd_create_tap(const struct netdev_class *class, const char *name,
346 struct netdev **netdevp)
348 struct netdev_bsd *netdev = NULL;
351 char *kernel_name = NULL;
353 error = cache_notifier_ref();
358 /* allocate the device structure and set the internal flag */
359 netdev = xzalloc(sizeof *netdev);
361 memset(&ifr, 0, sizeof(ifr));
363 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
364 * to retrieve the name of the tap device. */
365 netdev->tap_fd = open("/dev/tap", O_RDWR);
366 netdev->change_seq = 1;
367 if (netdev->tap_fd < 0) {
369 VLOG_WARN("opening \"/dev/tap\" failed: %s", ovs_strerror(error));
370 goto error_unref_notifier;
373 /* Retrieve tap name (e.g. tap0) */
374 if (ioctl(netdev->tap_fd, TAPGIFNAME, &ifr) == -1) {
375 /* XXX Need to destroy the device? */
377 close(netdev->tap_fd);
378 goto error_unref_notifier;
381 /* Change the name of the tap device */
382 #if defined(SIOCSIFNAME)
383 ifr.ifr_data = (void *)name;
384 if (ioctl(af_inet_sock, SIOCSIFNAME, &ifr) == -1) {
386 destroy_tap(netdev->tap_fd, ifr.ifr_name);
387 goto error_unref_notifier;
389 kernel_name = xstrdup(name);
392 * NetBSD doesn't support inteface renaming.
394 VLOG_INFO("tap %s is created for bridge %s", ifr.ifr_name, name);
395 kernel_name = xstrdup(ifr.ifr_name);
398 /* set non-blocking. */
399 error = set_nonblocking(netdev->tap_fd);
401 destroy_tap(netdev->tap_fd, kernel_name);
402 goto error_unref_notifier;
406 ifr_set_flags(&ifr, IFF_UP);
407 strncpy(ifr.ifr_name, kernel_name, sizeof ifr.ifr_name);
408 if (ioctl(af_inet_sock, SIOCSIFFLAGS, &ifr) == -1) {
410 destroy_tap(netdev->tap_fd, kernel_name);
411 goto error_unref_notifier;
414 /* initialize the device structure and
415 * link the structure to its netdev */
416 netdev_init(&netdev->up, name, class);
417 netdev->kernel_name = kernel_name;
418 *netdevp = &netdev->up;
422 error_unref_notifier:
423 cache_notifier_unref();
431 netdev_bsd_destroy(struct netdev *netdev_)
433 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
435 cache_notifier_unref();
437 if (netdev->tap_fd >= 0) {
438 destroy_tap(netdev->tap_fd, netdev_get_kernel_name(netdev_));
441 pcap_close(netdev->pcap);
443 free(netdev->kernel_name);
448 netdev_bsd_open_pcap(const char *name, pcap_t **pcapp, int *fdp)
450 char errbuf[PCAP_ERRBUF_SIZE];
456 /* Open the pcap device. The device is opened in non-promiscuous mode
457 * because the interface flags are manually set by the caller. */
459 pcap = pcap_open_live(name, PCAP_SNAPLEN, 0, 1000, errbuf);
461 VLOG_ERR_RL(&rl, "%s: pcap_open_live failed: %s", name, errbuf);
465 if (errbuf[0] != '\0') {
466 VLOG_WARN_RL(&rl, "%s: pcap_open_live: %s", name, errbuf);
469 /* Get the underlying fd. */
470 fd = pcap_get_selectable_fd(pcap);
472 VLOG_WARN_RL(&rl, "%s: no selectable file descriptor", name);
477 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
478 * on the file descriptor returned by pcap_get_selectable_fd to achieve
479 * a real non-blocking behaviour.*/
480 error = pcap_setnonblock(pcap, 1, errbuf);
486 /* This call assure that reads return immediately upon packet
487 * reception. Otherwise, a read will block until either the kernel
488 * buffer becomes full or a timeout occurs. */
489 if (ioctl(fd, BIOCIMMEDIATE, &one) < 0 ) {
490 VLOG_ERR_RL(&rl, "ioctl(BIOCIMMEDIATE) on %s device failed: %s",
491 name, ovs_strerror(errno));
496 /* Capture only incoming packets. */
497 error = pcap_setdirection(pcap, PCAP_D_IN);
517 netdev_bsd_rx_open(struct netdev *netdev_, struct netdev_rx **rxp)
519 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
521 struct netdev_rx_bsd *rx;
525 if (!strcmp(netdev_get_type(netdev_), "tap")) {
529 int error = netdev_bsd_open_pcap(netdev_get_kernel_name(netdev_),
535 netdev_bsd_changed(netdev);
538 rx = xmalloc(sizeof *rx);
539 netdev_rx_init(&rx->up, netdev_, &netdev_rx_bsd_class);
540 rx->pcap_handle = pcap;
548 netdev_rx_bsd_destroy(struct netdev_rx *rx_)
550 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
552 if (rx->pcap_handle) {
553 pcap_close(rx->pcap_handle);
558 /* The recv callback of the netdev class returns the number of bytes of the
561 * This can be done by the pcap_next() function. Unfortunately pcap_next() does
562 * not make difference between a missing packet on the capture interface and
563 * an error during the file capture. We can use the pcap_dispatch() function
564 * instead, which is able to distinguish between errors and null packet.
566 * To make pcap_dispatch() returns the number of bytes read from the interface
567 * we need to define the following callback and argument.
576 * This callback will be executed on every captured packet.
578 * If the packet captured by pcap_dispatch() does not fit the pcap buffer,
579 * pcap returns a truncated packet and we follow this behavior.
581 * The argument args->retval is the packet size in bytes.
584 proc_pkt(u_char *args_, const struct pcap_pkthdr *hdr, const u_char *packet)
586 struct pcap_arg *args = (struct pcap_arg *)args_;
588 if (args->size < hdr->len) {
589 VLOG_WARN_RL(&rl, "packet truncated");
590 args->retval = args->size;
592 args->retval = hdr->len;
595 /* copy the packet to our buffer */
596 memcpy(args->data, packet, args->retval);
600 * This function attempts to receive a packet from the specified network
601 * device. It is assumed that the network device is a system device or a tap
602 * device opened as a system one. In this case the read operation is performed
606 netdev_rx_bsd_recv_pcap(struct netdev_rx_bsd *rx, void *data, size_t size)
611 /* prepare the pcap argument to store the packet */
616 ret = pcap_dispatch(rx->pcap_handle, 1, proc_pkt, (u_char *) &arg);
619 return arg.retval; /* arg.retval < 0 is handled in the caller */
622 if (errno == EINTR) {
632 * This function attempts to receive a packet from the specified network
633 * device. It is assumed that the network device is a tap device and
634 * 'rx->fd' is initialized with the tap file descriptor.
637 netdev_rx_bsd_recv_tap(struct netdev_rx_bsd *rx, void *data, size_t size)
640 ssize_t retval = read(rx->fd, data, size);
643 } else if (errno != EINTR) {
644 if (errno != EAGAIN) {
645 VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
646 ovs_strerror(errno), netdev_rx_get_name(&rx->up));
655 netdev_rx_bsd_recv(struct netdev_rx *rx_, void *data, size_t size)
657 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
659 return (rx->pcap_handle
660 ? netdev_rx_bsd_recv_pcap(rx, data, size)
661 : netdev_rx_bsd_recv_tap(rx, data, size));
665 * Registers with the poll loop to wake up from the next call to poll_block()
666 * when a packet is ready to be received with netdev_rx_recv() on 'rx'.
669 netdev_rx_bsd_wait(struct netdev_rx *rx_)
671 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
673 poll_fd_wait(rx->fd, POLLIN);
676 /* Discards all packets waiting to be received from 'rx'. */
678 netdev_rx_bsd_drain(struct netdev_rx *rx_)
681 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
683 strcpy(ifr.ifr_name, netdev_get_kernel_name(netdev_rx_get_netdev(rx_)));
684 if (ioctl(rx->fd, BIOCFLUSH, &ifr) == -1) {
685 VLOG_DBG_RL(&rl, "%s: ioctl(BIOCFLUSH) failed: %s",
686 netdev_rx_get_name(rx_), ovs_strerror(errno));
693 * Send a packet on the specified network device. The device could be either a
694 * system or a tap device.
697 netdev_bsd_send(struct netdev *netdev_, const void *data, size_t size)
699 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
700 const char *name = netdev_get_name(netdev_);
702 if (dev->tap_fd < 0 && !dev->pcap) {
703 int error = netdev_bsd_open_pcap(name, &dev->pcap, &dev->fd);
711 if (dev->tap_fd >= 0) {
712 retval = write(dev->tap_fd, data, size);
714 retval = pcap_inject(dev->pcap, data, size);
717 if (errno == EINTR) {
719 } else if (errno != EAGAIN) {
720 VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s",
721 name, ovs_strerror(errno));
724 } else if (retval != size) {
725 VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of "
726 "%zu) on %s", retval, size, name);
735 * Registers with the poll loop to wake up from the next call to poll_block()
736 * when the packet transmission queue has sufficient room to transmit a packet
737 * with netdev_send().
740 netdev_bsd_send_wait(struct netdev *netdev_)
742 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
744 if (dev->tap_fd >= 0) {
745 /* TAP device always accepts packets. */
746 poll_immediate_wake();
747 } else if (dev->pcap) {
748 poll_fd_wait(dev->fd, POLLOUT);
750 /* We haven't even tried to send a packet yet. */
751 poll_immediate_wake();
756 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
757 * otherwise a positive errno value.
760 netdev_bsd_set_etheraddr(struct netdev *netdev_,
761 const uint8_t mac[ETH_ADDR_LEN])
763 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
766 if (!(netdev->cache_valid & VALID_ETHERADDR)
767 || !eth_addr_equals(netdev->etheraddr, mac)) {
768 error = set_etheraddr(netdev_get_kernel_name(netdev_), AF_LINK,
771 netdev->cache_valid |= VALID_ETHERADDR;
772 memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN);
773 netdev_bsd_changed(netdev);
782 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
783 * free the returned buffer.
786 netdev_bsd_get_etheraddr(const struct netdev *netdev_,
787 uint8_t mac[ETH_ADDR_LEN])
789 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
791 if (!(netdev->cache_valid & VALID_ETHERADDR)) {
792 int error = get_etheraddr(netdev_get_kernel_name(netdev_),
797 netdev->cache_valid |= VALID_ETHERADDR;
799 memcpy(mac, netdev->etheraddr, ETH_ADDR_LEN);
805 * Returns the maximum size of transmitted (and received) packets on 'netdev',
806 * in bytes, not including the hardware header; thus, this is typically 1500
807 * bytes for Ethernet devices.
810 netdev_bsd_get_mtu(const struct netdev *netdev_, int *mtup)
812 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
814 if (!(netdev->cache_valid & VALID_MTU)) {
818 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev_), &ifr,
819 SIOCGIFMTU, "SIOCGIFMTU");
823 netdev->mtu = ifr.ifr_mtu;
824 netdev->cache_valid |= VALID_MTU;
832 netdev_bsd_get_ifindex(const struct netdev *netdev)
836 error = get_ifindex(netdev, &ifindex);
837 return error ? -error : ifindex;
841 netdev_bsd_get_carrier(const struct netdev *netdev_, bool *carrier)
843 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
845 if (!(netdev->cache_valid & VALID_CARRIER)) {
846 struct ifmediareq ifmr;
848 memset(&ifmr, 0, sizeof(ifmr));
849 strncpy(ifmr.ifm_name, netdev_get_kernel_name(netdev_),
850 sizeof ifmr.ifm_name);
852 if (ioctl(af_inet_sock, SIOCGIFMEDIA, &ifmr) == -1) {
853 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
854 netdev_get_name(netdev_), ovs_strerror(errno));
858 netdev->carrier = (ifmr.ifm_status & IFM_ACTIVE) == IFM_ACTIVE;
859 netdev->cache_valid |= VALID_CARRIER;
861 /* If the interface doesn't report whether the media is active,
862 * just assume it is active. */
863 if ((ifmr.ifm_status & IFM_AVALID) == 0) {
864 netdev->carrier = true;
867 *carrier = netdev->carrier;
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;
947 memset(&ifdr, 0, sizeof(ifdr));
948 strncpy(ifdr.ifdr_name, netdev_get_kernel_name(netdev_),
949 sizeof(ifdr.ifdr_name));
950 ret = ioctl(af_link_sock, SIOCGIFDATA, &ifdr);
955 convert_stats(stats, &ifdr.ifdr_data);
958 #error not implemented
963 netdev_bsd_parse_media(int media)
965 uint32_t supported = 0;
966 bool half_duplex = media & IFM_HDX ? true : false;
968 switch (IFM_SUBTYPE(media)) {
973 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
974 supported |= NETDEV_F_COPPER;
978 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
979 supported |= NETDEV_F_FIBER;
986 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
987 supported |= NETDEV_F_COPPER;
991 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
992 supported |= NETDEV_F_FIBER;
997 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
998 supported |= NETDEV_F_COPPER;
1003 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
1004 supported |= NETDEV_F_FIBER;
1008 supported |= NETDEV_F_10GB_FD;
1009 supported |= NETDEV_F_COPPER;
1014 supported |= NETDEV_F_10GB_FD;
1015 supported |= NETDEV_F_FIBER;
1022 if (IFM_SUBTYPE(media) == IFM_AUTO) {
1023 supported |= NETDEV_F_AUTONEG;
1026 if (media & IFM_ETH_FMASK) {
1027 supported |= NETDEV_F_PAUSE;
1035 * Stores the features supported by 'netdev' into each of '*current',
1036 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1037 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1038 * successful, otherwise a positive errno value. On failure, all of the
1039 * passed-in values are set to 0.
1042 netdev_bsd_get_features(const struct netdev *netdev,
1043 enum netdev_features *current, uint32_t *advertised,
1044 enum netdev_features *supported, uint32_t *peer)
1046 struct ifmediareq ifmr;
1052 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1054 memset(&ifmr, 0, sizeof(ifmr));
1055 strncpy(ifmr.ifm_name, netdev_get_name(netdev), sizeof ifmr.ifm_name);
1057 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1058 * number of supported modes, and a second with a buffer to retrieve
1060 if (ioctl(af_inet_sock, SIOCGIFMEDIA, &ifmr) == -1) {
1061 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1062 netdev_get_name(netdev), ovs_strerror(errno));
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 if (ioctl(af_inet_sock, SIOCGIFMEDIA, &ifmr) == -1) {
1077 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1078 netdev_get_name(netdev), ovs_strerror(errno));
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_);
1114 if (!(netdev->cache_valid & VALID_IN4)) {
1115 const struct sockaddr_in *sin;
1119 ifr.ifr_addr.sa_family = AF_INET;
1120 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1121 SIOCGIFADDR, "SIOCGIFADDR");
1126 sin = (struct sockaddr_in *) &ifr.ifr_addr;
1127 netdev->in4 = sin->sin_addr;
1128 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1129 SIOCGIFNETMASK, "SIOCGIFNETMASK");
1133 netdev->netmask = sin->sin_addr;
1134 netdev->cache_valid |= VALID_IN4;
1137 *netmask = netdev->netmask;
1139 return 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 error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", addr);
1156 if (addr.s_addr != INADDR_ANY) {
1157 error = do_set_addr(netdev_, SIOCSIFNETMASK,
1158 "SIOCSIFNETMASK", mask);
1160 netdev->cache_valid |= VALID_IN4;
1162 netdev->netmask = mask;
1165 netdev_bsd_changed(netdev);
1171 netdev_bsd_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
1173 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1174 if (!(netdev->cache_valid & VALID_IN6)) {
1175 struct ifaddrs *ifa, *head;
1176 struct sockaddr_in6 *sin6;
1177 const char *netdev_name = netdev_get_name(netdev_);
1179 if (getifaddrs(&head) != 0) {
1180 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1181 ovs_strerror(errno));
1185 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1186 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1187 !strcmp(ifa->ifa_name, netdev_name)) {
1188 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1190 memcpy(&netdev->in6, &sin6->sin6_addr, sin6->sin6_len);
1191 netdev->cache_valid |= VALID_IN6;
1198 return EADDRNOTAVAIL;
1204 #if defined(__NetBSD__)
1206 netdev_bsd_kernel_name_to_ovs_name(const char *kernel_name)
1208 char *ovs_name = NULL;
1209 struct shash device_shash;
1210 struct shash_node *node;
1212 shash_init(&device_shash);
1213 netdev_get_devices(&netdev_tap_class, &device_shash);
1214 SHASH_FOR_EACH(node, &device_shash) {
1215 struct netdev *netdev = node->data;
1216 struct netdev_bsd * const dev = netdev_bsd_cast(netdev);
1218 if (!strcmp(dev->kernel_name, kernel_name)) {
1220 ovs_name = xstrdup(netdev_get_name(&dev->up));
1222 netdev_close(netdev);
1224 shash_destroy(&device_shash);
1226 return ovs_name ? ovs_name : xstrdup(kernel_name);
1231 netdev_bsd_get_next_hop(const struct in_addr *host OVS_UNUSED,
1232 struct in_addr *next_hop OVS_UNUSED,
1233 char **netdev_name OVS_UNUSED)
1235 #if defined(__NetBSD__)
1237 struct sockaddr_in sin;
1238 struct sockaddr_dl sdl;
1245 struct rt_msghdr *rtm = &buf.h;
1246 const pid_t pid = getpid();
1249 bool gateway = false;
1250 char *ifname = NULL;
1253 memset(next_hop, 0, sizeof(*next_hop));
1254 *netdev_name = NULL;
1256 memset(&sin, 0, sizeof(sin));
1257 sin.sin_len = sizeof(sin);
1258 sin.sin_family = AF_INET;
1260 sin.sin_addr = *host;
1262 memset(&sdl, 0, sizeof(sdl));
1263 sdl.sdl_len = sizeof(sdl);
1264 sdl.sdl_family = AF_LINK;
1266 s = socket(PF_ROUTE, SOCK_RAW, 0);
1267 memset(&buf, 0, sizeof(buf));
1268 rtm->rtm_flags = RTF_HOST|RTF_UP;
1269 rtm->rtm_version = RTM_VERSION;
1270 rtm->rtm_addrs = RTA_DST|RTA_IFP;
1271 cp = (void *)&buf.space;
1272 memcpy(cp, &sin, sizeof(sin));
1273 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sin);
1274 memcpy(cp, &sdl, sizeof(sdl));
1275 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sdl);
1276 rtm->rtm_msglen = cp - (char *)(void *)rtm;
1277 rtm->rtm_seq = ++seq;
1278 rtm->rtm_type = RTM_GET;
1280 write(s, rtm, rtm->rtm_msglen);
1281 memset(&buf, 0, sizeof(buf));
1283 ssz = read(s, &buf, sizeof(buf));
1284 } while (ssz > 0 && (rtm->rtm_seq != seq || rtm->rtm_pid != pid));
1285 saved_errno = errno;
1291 return EPIPE; /* XXX */
1293 cp = (void *)&buf.space;
1294 for (i = 1; i; i <<= 1) {
1295 if ((rtm->rtm_addrs & i) != 0) {
1296 const struct sockaddr *sa = (const void *)cp;
1298 if ((i == RTA_GATEWAY) && sa->sa_family == AF_INET) {
1299 const struct sockaddr_in * const sin =
1300 (const struct sockaddr_in *)sa;
1302 *next_hop = sin->sin_addr;
1305 if ((i == RTA_IFP) && sa->sa_family == AF_LINK) {
1306 const struct sockaddr_dl * const sdl =
1307 (const struct sockaddr_dl *)sa;
1310 kernel_name = xmemdup0(sdl->sdl_data, sdl->sdl_nlen);
1311 ifname = netdev_bsd_kernel_name_to_ovs_name(kernel_name);
1317 if (ifname == NULL) {
1323 *netdev_name = ifname;
1324 VLOG_DBG("host " IP_FMT " next-hop " IP_FMT " if %s",
1325 IP_ARGS(host->s_addr), IP_ARGS(next_hop->s_addr), *netdev_name);
1333 make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
1335 struct sockaddr_in sin;
1336 memset(&sin, 0, sizeof sin);
1337 sin.sin_family = AF_INET;
1338 sin.sin_addr = addr;
1341 memset(sa, 0, sizeof *sa);
1342 memcpy(sa, &sin, sizeof sin);
1346 do_set_addr(struct netdev *netdev,
1347 int ioctl_nr, const char *ioctl_name, struct in_addr addr)
1350 make_in4_sockaddr(&ifr.ifr_addr, addr);
1351 return netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev), &ifr, ioctl_nr,
1356 nd_to_iff_flags(enum netdev_flags nd)
1359 if (nd & NETDEV_UP) {
1362 if (nd & NETDEV_PROMISC) {
1364 #if defined(IFF_PPROMISC)
1365 iff |= IFF_PPROMISC;
1372 iff_to_nd_flags(int iff)
1374 enum netdev_flags nd = 0;
1378 if (iff & IFF_PROMISC) {
1379 nd |= NETDEV_PROMISC;
1385 netdev_bsd_update_flags(struct netdev *netdev_, enum netdev_flags off,
1386 enum netdev_flags on, enum netdev_flags *old_flagsp)
1388 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1389 int old_flags, new_flags;
1392 error = get_flags(netdev_, &old_flags);
1394 *old_flagsp = iff_to_nd_flags(old_flags);
1395 new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on);
1396 if (new_flags != old_flags) {
1397 error = set_flags(netdev_get_kernel_name(netdev_), new_flags);
1398 netdev_bsd_changed(netdev);
1405 netdev_bsd_change_seq(const struct netdev *netdev)
1407 return netdev_bsd_cast(netdev)->change_seq;
1411 const struct netdev_class netdev_bsd_class = {
1417 netdev_bsd_create_system,
1419 NULL, /* get_config */
1420 NULL, /* set_config */
1421 NULL, /* get_tunnel_config */
1426 netdev_bsd_send_wait,
1428 netdev_bsd_set_etheraddr,
1429 netdev_bsd_get_etheraddr,
1432 netdev_bsd_get_ifindex,
1433 netdev_bsd_get_carrier,
1434 NULL, /* get_carrier_resets */
1435 NULL, /* set_miimon_interval */
1436 netdev_bsd_get_stats,
1437 NULL, /* set_stats */
1439 netdev_bsd_get_features,
1440 NULL, /* set_advertisement */
1441 NULL, /* set_policing */
1442 NULL, /* get_qos_type */
1443 NULL, /* get_qos_capabilities */
1446 NULL, /* get_queue */
1447 NULL, /* set_queue */
1448 NULL, /* delete_queue */
1449 NULL, /* get_queue_stats */
1450 NULL, /* dump_queue */
1451 NULL, /* dump_queue_stats */
1456 NULL, /* add_router */
1457 netdev_bsd_get_next_hop,
1458 NULL, /* get_status */
1459 NULL, /* arp_lookup */
1461 netdev_bsd_update_flags,
1463 netdev_bsd_change_seq
1466 const struct netdev_class netdev_tap_class = {
1472 netdev_bsd_create_tap,
1474 NULL, /* get_config */
1475 NULL, /* set_config */
1476 NULL, /* get_tunnel_config */
1481 netdev_bsd_send_wait,
1483 netdev_bsd_set_etheraddr,
1484 netdev_bsd_get_etheraddr,
1487 netdev_bsd_get_ifindex,
1488 netdev_bsd_get_carrier,
1489 NULL, /* get_carrier_resets */
1490 NULL, /* set_miimon_interval */
1491 netdev_bsd_get_stats,
1492 NULL, /* set_stats */
1494 netdev_bsd_get_features,
1495 NULL, /* set_advertisement */
1496 NULL, /* set_policing */
1497 NULL, /* get_qos_type */
1498 NULL, /* get_qos_capabilities */
1501 NULL, /* get_queue */
1502 NULL, /* set_queue */
1503 NULL, /* delete_queue */
1504 NULL, /* get_queue_stats */
1505 NULL, /* dump_queue */
1506 NULL, /* dump_queue_stats */
1511 NULL, /* add_router */
1512 netdev_bsd_get_next_hop,
1513 NULL, /* get_status */
1514 NULL, /* arp_lookup */
1516 netdev_bsd_update_flags,
1518 netdev_bsd_change_seq
1521 static const struct netdev_rx_class netdev_rx_bsd_class = {
1522 netdev_rx_bsd_destroy,
1525 netdev_rx_bsd_drain,
1530 destroy_tap(int fd, const char *name)
1535 strcpy(ifr.ifr_name, name);
1536 /* XXX What to do if this call fails? */
1537 ioctl(af_inet_sock, SIOCIFDESTROY, &ifr);
1541 get_flags(const struct netdev *netdev, int *flags)
1546 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev), &ifr,
1547 SIOCGIFFLAGS, "SIOCGIFFLAGS");
1549 *flags = ifr_get_flags(&ifr);
1555 set_flags(const char *name, int flags)
1559 ifr_set_flags(&ifr, flags);
1561 return netdev_bsd_do_ioctl(name, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS");
1565 get_ifindex(const struct netdev *netdev_, int *ifindexp)
1567 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1569 if (!(netdev->cache_valid & VALID_IFINDEX)) {
1570 int ifindex = if_nametoindex(netdev_get_name(netdev_));
1574 netdev->cache_valid |= VALID_IFINDEX;
1575 netdev->ifindex = ifindex;
1577 *ifindexp = netdev->ifindex;
1582 get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN])
1584 struct ifaddrs *head;
1585 struct ifaddrs *ifa;
1586 struct sockaddr_dl *sdl;
1588 if (getifaddrs(&head) != 0) {
1589 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1590 ovs_strerror(errno));
1594 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1595 if (ifa->ifa_addr->sa_family == AF_LINK) {
1596 if (!strcmp(ifa->ifa_name, netdev_name)) {
1597 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1599 memcpy(ea, LLADDR(sdl), sdl->sdl_alen);
1607 VLOG_ERR("could not find ethernet address for %s device", netdev_name);
1613 set_etheraddr(const char *netdev_name OVS_UNUSED, int hwaddr_family OVS_UNUSED,
1614 int hwaddr_len OVS_UNUSED,
1615 const uint8_t mac[ETH_ADDR_LEN] OVS_UNUSED)
1617 #if defined(__FreeBSD__)
1620 memset(&ifr, 0, sizeof ifr);
1621 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1622 ifr.ifr_addr.sa_family = hwaddr_family;
1623 ifr.ifr_addr.sa_len = hwaddr_len;
1624 memcpy(ifr.ifr_addr.sa_data, mac, hwaddr_len);
1625 if (ioctl(af_inet_sock, SIOCSIFLLADDR, &ifr) < 0) {
1626 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1627 netdev_name, ovs_strerror(errno));
1631 #elif defined(__NetBSD__)
1632 struct if_laddrreq req;
1633 struct sockaddr_dl *sdl;
1634 struct sockaddr_storage oldaddr;
1638 * get the old address, add new one, and then remove old one.
1641 if (hwaddr_len != ETH_ADDR_LEN) {
1642 /* just to be safe about sockaddr storage size */
1645 memset(&req, 0, sizeof(req));
1646 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1647 req.addr.ss_len = sizeof(req.addr);
1648 req.addr.ss_family = hwaddr_family;
1649 sdl = (struct sockaddr_dl *)&req.addr;
1650 sdl->sdl_alen = hwaddr_len;
1651 ret = ioctl(af_link_sock, SIOCGLIFADDR, &req);
1655 if (!memcmp(&sdl->sdl_data[sdl->sdl_nlen], mac, hwaddr_len)) {
1660 memset(&req, 0, sizeof(req));
1661 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1662 req.flags = IFLR_ACTIVE;
1663 sdl = (struct sockaddr_dl *)&req.addr;
1664 sdl->sdl_len = offsetof(struct sockaddr_dl, sdl_data) + hwaddr_len;
1665 sdl->sdl_alen = hwaddr_len;
1666 sdl->sdl_family = hwaddr_family;
1667 memcpy(sdl->sdl_data, mac, hwaddr_len);
1668 ret = ioctl(af_link_sock, SIOCALIFADDR, &req);
1673 memset(&req, 0, sizeof(req));
1674 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1676 ret = ioctl(af_link_sock, SIOCDLIFADDR, &req);
1682 #error not implemented
1687 netdev_bsd_do_ioctl(const char *name, struct ifreq *ifr, unsigned long cmd,
1688 const char *cmd_name)
1690 strncpy(ifr->ifr_name, name, sizeof ifr->ifr_name);
1691 if (ioctl(af_inet_sock, cmd, ifr) == -1) {
1692 VLOG_DBG_RL(&rl, "%s: ioctl(%s) failed: %s", name, cmd_name,
1693 ovs_strerror(errno));
1700 ifr_get_flags(const struct ifreq *ifr)
1702 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1703 return (ifr->ifr_flagshigh << 16) | ifr->ifr_flags;
1705 return ifr->ifr_flags;
1710 ifr_set_flags(struct ifreq *ifr, int flags)
1712 ifr->ifr_flags = flags;
1713 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1714 ifr->ifr_flagshigh = flags >> 16;