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];
90 int tap_fd; /* TAP character device, if any, otherwise -1. */
92 /* Used for sending packets on non-tap devices. */
101 VALID_IFINDEX = 1 << 0,
102 VALID_ETHERADDR = 1 << 1,
106 VALID_CARRIER = 1 << 5
109 /* An AF_INET socket (used for ioctl operations). */
110 static int af_inet_sock = -1;
112 #if defined(__NetBSD__)
113 /* AF_LINK socket used for netdev_bsd_get_stats and set_etheraddr */
114 static int af_link_sock = -1;
115 #endif /* defined(__NetBSD__) */
117 #define PCAP_SNAPLEN 2048
121 * Notifier used to invalidate device informations in case of status change.
123 * It will be registered with a 'rtbsd_notifier_register()' when the first
124 * device will be created with the call of either 'netdev_bsd_tap_create()' or
125 * 'netdev_bsd_system_create()'.
127 * The callback associated with this notifier ('netdev_bsd_cache_cb()') will
128 * invalidate cached information about the device.
130 static struct rtbsd_notifier netdev_bsd_cache_notifier;
131 static int cache_notifier_refcount;
133 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
135 static int netdev_bsd_do_ioctl(const char *, struct ifreq *, unsigned long cmd,
136 const char *cmd_name);
137 static void destroy_tap(int fd, const char *name);
138 static int get_flags(const struct netdev *, int *flagsp);
139 static int set_flags(const char *, int flags);
140 static int do_set_addr(struct netdev *netdev,
141 int ioctl_nr, const char *ioctl_name,
142 struct in_addr addr);
143 static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]);
144 static int set_etheraddr(const char *netdev_name, int hwaddr_family,
145 int hwaddr_len, const uint8_t[ETH_ADDR_LEN]);
146 static int get_ifindex(const struct netdev *, int *ifindexp);
148 static int ifr_get_flags(const struct ifreq *);
149 static void ifr_set_flags(struct ifreq *, int flags);
151 static int netdev_bsd_init(void);
154 is_netdev_bsd_class(const struct netdev_class *netdev_class)
156 return netdev_class->init == netdev_bsd_init;
159 static struct netdev_bsd *
160 netdev_bsd_cast(const struct netdev *netdev)
162 ovs_assert(is_netdev_bsd_class(netdev_get_class(netdev)));
163 return CONTAINER_OF(netdev, struct netdev_bsd, up);
166 static struct netdev_rx_bsd *
167 netdev_rx_bsd_cast(const struct netdev_rx *rx)
169 netdev_rx_assert_class(rx, &netdev_rx_bsd_class);
170 return CONTAINER_OF(rx, struct netdev_rx_bsd, up);
174 netdev_get_kernel_name(const struct netdev *netdev)
176 return netdev_bsd_cast(netdev)->kernel_name;
179 /* Initialize the AF_INET socket used for ioctl operations */
181 netdev_bsd_init(void)
183 static int status = -1;
185 if (status >= 0) { /* already initialized */
189 af_inet_sock = socket(AF_INET, SOCK_DGRAM, 0);
190 status = af_inet_sock >= 0 ? 0 : errno;
192 VLOG_ERR("failed to create inet socket: %s", ovs_strerror(status));
196 #if defined(__NetBSD__)
197 af_link_sock = socket(AF_LINK, SOCK_DGRAM, 0);
198 status = af_link_sock >= 0 ? 0 : errno;
200 VLOG_ERR("failed to create link socket: %s", ovs_strerror(status));
204 #endif /* defined(__NetBSD__) */
210 * Perform periodic work needed by netdev. In BSD netdevs it checks for any
211 * interface status changes, and eventually calls all the user callbacks.
216 rtbsd_notifier_run();
220 * Arranges for poll_block() to wake up if the "run" member function needs to
224 netdev_bsd_wait(void)
226 rtbsd_notifier_wait();
230 netdev_bsd_changed(struct netdev_bsd *dev)
233 if (!dev->change_seq) {
238 /* Invalidate cache in case of interface status change. */
240 netdev_bsd_cache_cb(const struct rtbsd_change *change,
241 void *aux OVS_UNUSED)
243 struct netdev_bsd *dev;
246 struct netdev *base_dev = netdev_from_name(change->if_name);
249 const struct netdev_class *netdev_class =
250 netdev_get_class(base_dev);
252 if (is_netdev_bsd_class(netdev_class)) {
253 dev = netdev_bsd_cast(base_dev);
254 dev->cache_valid = 0;
255 netdev_bsd_changed(dev);
260 * XXX the API is lacking, we should be able to iterate on the list of
261 * netdevs without having to store the info in a temp shash.
263 struct shash device_shash;
264 struct shash_node *node;
266 shash_init(&device_shash);
267 netdev_get_devices(&netdev_bsd_class, &device_shash);
268 SHASH_FOR_EACH (node, &device_shash) {
270 dev->cache_valid = 0;
271 netdev_bsd_changed(dev);
273 shash_destroy(&device_shash);
278 cache_notifier_ref(void)
282 if (!cache_notifier_refcount) {
283 ret = rtbsd_notifier_register(&netdev_bsd_cache_notifier,
284 netdev_bsd_cache_cb, NULL);
289 cache_notifier_refcount++;
294 cache_notifier_unref(void)
296 cache_notifier_refcount--;
297 if (cache_notifier_refcount == 0) {
298 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier);
303 /* Allocate a netdev_bsd structure */
305 netdev_bsd_create_system(const struct netdev_class *class, const char *name,
306 struct netdev **netdevp)
308 struct netdev_bsd *netdev;
309 enum netdev_flags flags;
312 error = cache_notifier_ref();
317 netdev = xzalloc(sizeof *netdev);
318 netdev->change_seq = 1;
319 netdev_init(&netdev->up, name, class);
321 netdev->kernel_name = xstrdup(name);
323 /* Verify that the netdev really exists by attempting to read its flags */
324 error = netdev_get_flags(&netdev->up, &flags);
325 if (error == ENXIO) {
326 free(netdev->kernel_name);
327 netdev_uninit(&netdev->up, false);
329 cache_notifier_unref();
333 *netdevp = &netdev->up;
338 * Allocate a netdev_bsd structure with 'tap' class.
341 netdev_bsd_create_tap(const struct netdev_class *class, const char *name,
342 struct netdev **netdevp)
344 struct netdev_bsd *netdev = NULL;
347 char *kernel_name = NULL;
349 error = cache_notifier_ref();
354 /* allocate the device structure and set the internal flag */
355 netdev = xzalloc(sizeof *netdev);
357 memset(&ifr, 0, sizeof(ifr));
359 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
360 * to retrieve the name of the tap device. */
361 netdev->tap_fd = open("/dev/tap", O_RDWR);
362 netdev->change_seq = 1;
363 if (netdev->tap_fd < 0) {
365 VLOG_WARN("opening \"/dev/tap\" failed: %s", ovs_strerror(error));
366 goto error_unref_notifier;
369 /* Retrieve tap name (e.g. tap0) */
370 if (ioctl(netdev->tap_fd, TAPGIFNAME, &ifr) == -1) {
371 /* XXX Need to destroy the device? */
373 close(netdev->tap_fd);
374 goto error_unref_notifier;
377 /* Change the name of the tap device */
378 #if defined(SIOCSIFNAME)
379 ifr.ifr_data = (void *)name;
380 if (ioctl(af_inet_sock, SIOCSIFNAME, &ifr) == -1) {
382 destroy_tap(netdev->tap_fd, ifr.ifr_name);
383 goto error_unref_notifier;
385 kernel_name = xstrdup(name);
388 * NetBSD doesn't support inteface renaming.
390 VLOG_INFO("tap %s is created for bridge %s", ifr.ifr_name, name);
391 kernel_name = xstrdup(ifr.ifr_name);
394 /* set non-blocking. */
395 error = set_nonblocking(netdev->tap_fd);
397 destroy_tap(netdev->tap_fd, kernel_name);
398 goto error_unref_notifier;
402 ifr_set_flags(&ifr, IFF_UP);
403 strncpy(ifr.ifr_name, kernel_name, sizeof ifr.ifr_name);
404 if (ioctl(af_inet_sock, SIOCSIFFLAGS, &ifr) == -1) {
406 destroy_tap(netdev->tap_fd, kernel_name);
407 goto error_unref_notifier;
410 /* initialize the device structure and
411 * link the structure to its netdev */
412 netdev_init(&netdev->up, name, class);
413 netdev->kernel_name = kernel_name;
414 *netdevp = &netdev->up;
418 error_unref_notifier:
419 cache_notifier_unref();
427 netdev_bsd_destroy(struct netdev *netdev_)
429 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
431 cache_notifier_unref();
433 if (netdev->tap_fd >= 0) {
434 destroy_tap(netdev->tap_fd, netdev_get_kernel_name(netdev_));
437 pcap_close(netdev->pcap);
439 free(netdev->kernel_name);
444 netdev_bsd_open_pcap(const char *name, pcap_t **pcapp, int *fdp)
446 char errbuf[PCAP_ERRBUF_SIZE];
452 /* Open the pcap device. The device is opened in non-promiscuous mode
453 * because the interface flags are manually set by the caller. */
455 pcap = pcap_open_live(name, PCAP_SNAPLEN, 0, 1000, errbuf);
457 VLOG_ERR_RL(&rl, "%s: pcap_open_live failed: %s", name, errbuf);
461 if (errbuf[0] != '\0') {
462 VLOG_WARN_RL(&rl, "%s: pcap_open_live: %s", name, errbuf);
465 /* Get the underlying fd. */
466 fd = pcap_get_selectable_fd(pcap);
468 VLOG_WARN_RL(&rl, "%s: no selectable file descriptor", name);
473 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
474 * on the file descriptor returned by pcap_get_selectable_fd to achieve
475 * a real non-blocking behaviour.*/
476 error = pcap_setnonblock(pcap, 1, errbuf);
482 /* This call assure that reads return immediately upon packet
483 * reception. Otherwise, a read will block until either the kernel
484 * buffer becomes full or a timeout occurs. */
485 if (ioctl(fd, BIOCIMMEDIATE, &one) < 0 ) {
486 VLOG_ERR_RL(&rl, "ioctl(BIOCIMMEDIATE) on %s device failed: %s",
487 name, ovs_strerror(errno));
492 /* Capture only incoming packets. */
493 error = pcap_setdirection(pcap, PCAP_D_IN);
513 netdev_bsd_rx_open(struct netdev *netdev_, struct netdev_rx **rxp)
515 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
517 struct netdev_rx_bsd *rx;
521 if (!strcmp(netdev_get_type(netdev_), "tap")) {
525 int error = netdev_bsd_open_pcap(netdev_get_kernel_name(netdev_),
531 netdev_bsd_changed(netdev);
534 rx = xmalloc(sizeof *rx);
535 netdev_rx_init(&rx->up, netdev_, &netdev_rx_bsd_class);
536 rx->pcap_handle = pcap;
544 netdev_rx_bsd_destroy(struct netdev_rx *rx_)
546 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
548 if (rx->pcap_handle) {
549 pcap_close(rx->pcap_handle);
554 /* The recv callback of the netdev class returns the number of bytes of the
557 * This can be done by the pcap_next() function. Unfortunately pcap_next() does
558 * not make difference between a missing packet on the capture interface and
559 * an error during the file capture. We can use the pcap_dispatch() function
560 * instead, which is able to distinguish between errors and null packet.
562 * To make pcap_dispatch() returns the number of bytes read from the interface
563 * we need to define the following callback and argument.
572 * This callback will be executed on every captured packet.
574 * If the packet captured by pcap_dispatch() does not fit the pcap buffer,
575 * pcap returns a truncated packet and we follow this behavior.
577 * The argument args->retval is the packet size in bytes.
580 proc_pkt(u_char *args_, const struct pcap_pkthdr *hdr, const u_char *packet)
582 struct pcap_arg *args = (struct pcap_arg *)args_;
584 if (args->size < hdr->len) {
585 VLOG_WARN_RL(&rl, "packet truncated");
586 args->retval = args->size;
588 args->retval = hdr->len;
591 /* copy the packet to our buffer */
592 memcpy(args->data, packet, args->retval);
596 * This function attempts to receive a packet from the specified network
597 * device. It is assumed that the network device is a system device or a tap
598 * device opened as a system one. In this case the read operation is performed
602 netdev_rx_bsd_recv_pcap(struct netdev_rx_bsd *rx, void *data, size_t size)
607 /* prepare the pcap argument to store the packet */
612 ret = pcap_dispatch(rx->pcap_handle, 1, proc_pkt, (u_char *) &arg);
615 return arg.retval; /* arg.retval < 0 is handled in the caller */
618 if (errno == EINTR) {
628 * This function attempts to receive a packet from the specified network
629 * device. It is assumed that the network device is a tap device and
630 * 'rx->fd' is initialized with the tap file descriptor.
633 netdev_rx_bsd_recv_tap(struct netdev_rx_bsd *rx, void *data, size_t size)
636 ssize_t retval = read(rx->fd, data, size);
639 } else if (errno != EINTR) {
640 if (errno != EAGAIN) {
641 VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
642 ovs_strerror(errno), netdev_rx_get_name(&rx->up));
651 netdev_rx_bsd_recv(struct netdev_rx *rx_, void *data, size_t size)
653 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
655 return (rx->pcap_handle
656 ? netdev_rx_bsd_recv_pcap(rx, data, size)
657 : netdev_rx_bsd_recv_tap(rx, data, size));
661 * Registers with the poll loop to wake up from the next call to poll_block()
662 * when a packet is ready to be received with netdev_rx_recv() on 'rx'.
665 netdev_rx_bsd_wait(struct netdev_rx *rx_)
667 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
669 poll_fd_wait(rx->fd, POLLIN);
672 /* Discards all packets waiting to be received from 'rx'. */
674 netdev_rx_bsd_drain(struct netdev_rx *rx_)
677 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
679 strcpy(ifr.ifr_name, netdev_get_kernel_name(netdev_rx_get_netdev(rx_)));
680 if (ioctl(rx->fd, BIOCFLUSH, &ifr) == -1) {
681 VLOG_DBG_RL(&rl, "%s: ioctl(BIOCFLUSH) failed: %s",
682 netdev_rx_get_name(rx_), ovs_strerror(errno));
689 * Send a packet on the specified network device. The device could be either a
690 * system or a tap device.
693 netdev_bsd_send(struct netdev *netdev_, const void *data, size_t size)
695 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
696 const char *name = netdev_get_name(netdev_);
698 if (dev->tap_fd < 0 && !dev->pcap) {
699 int error = netdev_bsd_open_pcap(name, &dev->pcap, &dev->fd);
707 if (dev->tap_fd >= 0) {
708 retval = write(dev->tap_fd, data, size);
710 retval = pcap_inject(dev->pcap, data, size);
713 if (errno == EINTR) {
715 } else if (errno != EAGAIN) {
716 VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s",
717 name, ovs_strerror(errno));
720 } else if (retval != size) {
721 VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of "
722 "%zu) on %s", retval, size, name);
731 * Registers with the poll loop to wake up from the next call to poll_block()
732 * when the packet transmission queue has sufficient room to transmit a packet
733 * with netdev_send().
736 netdev_bsd_send_wait(struct netdev *netdev_)
738 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
740 if (dev->tap_fd >= 0) {
741 /* TAP device always accepts packets. */
742 poll_immediate_wake();
743 } else if (dev->pcap) {
744 poll_fd_wait(dev->fd, POLLOUT);
746 /* We haven't even tried to send a packet yet. */
747 poll_immediate_wake();
752 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
753 * otherwise a positive errno value.
756 netdev_bsd_set_etheraddr(struct netdev *netdev_,
757 const uint8_t mac[ETH_ADDR_LEN])
759 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
762 if (!(netdev->cache_valid & VALID_ETHERADDR)
763 || !eth_addr_equals(netdev->etheraddr, mac)) {
764 error = set_etheraddr(netdev_get_kernel_name(netdev_), AF_LINK,
767 netdev->cache_valid |= VALID_ETHERADDR;
768 memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN);
769 netdev_bsd_changed(netdev);
778 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
779 * free the returned buffer.
782 netdev_bsd_get_etheraddr(const struct netdev *netdev_,
783 uint8_t mac[ETH_ADDR_LEN])
785 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
787 if (!(netdev->cache_valid & VALID_ETHERADDR)) {
788 int error = get_etheraddr(netdev_get_kernel_name(netdev_),
793 netdev->cache_valid |= VALID_ETHERADDR;
795 memcpy(mac, netdev->etheraddr, ETH_ADDR_LEN);
801 * Returns the maximum size of transmitted (and received) packets on 'netdev',
802 * in bytes, not including the hardware header; thus, this is typically 1500
803 * bytes for Ethernet devices.
806 netdev_bsd_get_mtu(const struct netdev *netdev_, int *mtup)
808 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
810 if (!(netdev->cache_valid & VALID_MTU)) {
814 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev_), &ifr,
815 SIOCGIFMTU, "SIOCGIFMTU");
819 netdev->mtu = ifr.ifr_mtu;
820 netdev->cache_valid |= VALID_MTU;
828 netdev_bsd_get_ifindex(const struct netdev *netdev)
832 error = get_ifindex(netdev, &ifindex);
833 return error ? -error : ifindex;
837 netdev_bsd_get_carrier(const struct netdev *netdev_, bool *carrier)
839 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
841 if (!(netdev->cache_valid & VALID_CARRIER)) {
842 struct ifmediareq ifmr;
844 memset(&ifmr, 0, sizeof(ifmr));
845 strncpy(ifmr.ifm_name, netdev_get_kernel_name(netdev_),
846 sizeof ifmr.ifm_name);
848 if (ioctl(af_inet_sock, SIOCGIFMEDIA, &ifmr) == -1) {
849 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
850 netdev_get_name(netdev_), ovs_strerror(errno));
854 netdev->carrier = (ifmr.ifm_status & IFM_ACTIVE) == IFM_ACTIVE;
855 netdev->cache_valid |= VALID_CARRIER;
857 /* If the interface doesn't report whether the media is active,
858 * just assume it is active. */
859 if ((ifmr.ifm_status & IFM_AVALID) == 0) {
860 netdev->carrier = true;
863 *carrier = netdev->carrier;
869 convert_stats(struct netdev_stats *stats, const struct if_data *ifd)
872 * note: UINT64_MAX means unsupported
874 stats->rx_packets = ifd->ifi_ipackets;
875 stats->tx_packets = ifd->ifi_opackets;
876 stats->rx_bytes = ifd->ifi_obytes;
877 stats->tx_bytes = ifd->ifi_ibytes;
878 stats->rx_errors = ifd->ifi_ierrors;
879 stats->tx_errors = ifd->ifi_oerrors;
880 stats->rx_dropped = ifd->ifi_iqdrops;
881 stats->tx_dropped = UINT64_MAX;
882 stats->multicast = ifd->ifi_imcasts;
883 stats->collisions = ifd->ifi_collisions;
884 stats->rx_length_errors = UINT64_MAX;
885 stats->rx_over_errors = UINT64_MAX;
886 stats->rx_crc_errors = UINT64_MAX;
887 stats->rx_frame_errors = UINT64_MAX;
888 stats->rx_fifo_errors = UINT64_MAX;
889 stats->rx_missed_errors = UINT64_MAX;
890 stats->tx_aborted_errors = UINT64_MAX;
891 stats->tx_carrier_errors = UINT64_MAX;
892 stats->tx_fifo_errors = UINT64_MAX;
893 stats->tx_heartbeat_errors = UINT64_MAX;
894 stats->tx_window_errors = UINT64_MAX;
897 /* Retrieves current device stats for 'netdev'. */
899 netdev_bsd_get_stats(const struct netdev *netdev_, struct netdev_stats *stats)
901 #if defined(__FreeBSD__)
905 struct ifmibdata ifmd;
910 mib[2] = NETLINK_GENERIC;
911 mib[3] = IFMIB_SYSTEM;
912 mib[4] = IFMIB_IFCOUNT;
914 len = sizeof(if_count);
916 if (sysctl(mib, 5, &if_count, &len, (void *)0, 0) == -1) {
917 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
918 netdev_get_name(netdev_), ovs_strerror(errno));
922 mib[5] = IFDATA_GENERAL;
923 mib[3] = IFMIB_IFDATA;
925 for (i = 1; i <= if_count; i++) {
927 if (sysctl(mib, 6, &ifmd, &len, (void *)0, 0) == -1) {
928 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
929 netdev_get_name(netdev_), ovs_strerror(errno));
931 } else if (!strcmp(ifmd.ifmd_name, netdev_get_name(netdev_))) {
932 convert_stats(stats, &ifmd.ifmd_data);
938 #elif defined(__NetBSD__)
939 struct ifdatareq ifdr;
943 memset(&ifdr, 0, sizeof(ifdr));
944 strncpy(ifdr.ifdr_name, netdev_get_kernel_name(netdev_),
945 sizeof(ifdr.ifdr_name));
946 ret = ioctl(af_link_sock, SIOCGIFDATA, &ifdr);
951 convert_stats(stats, &ifdr.ifdr_data);
954 #error not implemented
959 netdev_bsd_parse_media(int media)
961 uint32_t supported = 0;
962 bool half_duplex = media & IFM_HDX ? true : false;
964 switch (IFM_SUBTYPE(media)) {
969 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
970 supported |= NETDEV_F_COPPER;
974 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
975 supported |= NETDEV_F_FIBER;
982 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
983 supported |= NETDEV_F_COPPER;
987 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
988 supported |= NETDEV_F_FIBER;
993 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
994 supported |= NETDEV_F_COPPER;
999 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
1000 supported |= NETDEV_F_FIBER;
1004 supported |= NETDEV_F_10GB_FD;
1005 supported |= NETDEV_F_COPPER;
1010 supported |= NETDEV_F_10GB_FD;
1011 supported |= NETDEV_F_FIBER;
1018 if (IFM_SUBTYPE(media) == IFM_AUTO) {
1019 supported |= NETDEV_F_AUTONEG;
1022 if (media & IFM_ETH_FMASK) {
1023 supported |= NETDEV_F_PAUSE;
1031 * Stores the features supported by 'netdev' into each of '*current',
1032 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1033 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1034 * successful, otherwise a positive errno value. On failure, all of the
1035 * passed-in values are set to 0.
1038 netdev_bsd_get_features(const struct netdev *netdev,
1039 enum netdev_features *current, uint32_t *advertised,
1040 enum netdev_features *supported, uint32_t *peer)
1042 struct ifmediareq ifmr;
1048 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1050 memset(&ifmr, 0, sizeof(ifmr));
1051 strncpy(ifmr.ifm_name, netdev_get_name(netdev), sizeof ifmr.ifm_name);
1053 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1054 * number of supported modes, and a second with a buffer to retrieve
1056 if (ioctl(af_inet_sock, SIOCGIFMEDIA, &ifmr) == -1) {
1057 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1058 netdev_get_name(netdev), ovs_strerror(errno));
1062 media_list = xcalloc(ifmr.ifm_count, sizeof(int));
1063 ifmr.ifm_ulist = media_list;
1065 if (IFM_TYPE(ifmr.ifm_current) != IFM_ETHER) {
1066 VLOG_DBG_RL(&rl, "%s: doesn't appear to be ethernet",
1067 netdev_get_name(netdev));
1072 if (ioctl(af_inet_sock, SIOCGIFMEDIA, &ifmr) == -1) {
1073 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1074 netdev_get_name(netdev), ovs_strerror(errno));
1079 /* Current settings. */
1080 *current = netdev_bsd_parse_media(ifmr.ifm_active);
1082 /* Advertised features. */
1083 *advertised = netdev_bsd_parse_media(ifmr.ifm_current);
1085 /* Supported features. */
1087 for (i = 0; i < ifmr.ifm_count; i++) {
1088 *supported |= netdev_bsd_parse_media(ifmr.ifm_ulist[i]);
1091 /* Peer advertisements. */
1092 *peer = 0; /* XXX */
1101 * If 'netdev' has an assigned IPv4 address, sets '*in4' to that address (if
1102 * 'in4' is non-null) and returns true. Otherwise, returns false.
1105 netdev_bsd_get_in4(const struct netdev *netdev_, struct in_addr *in4,
1106 struct in_addr *netmask)
1108 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1110 if (!(netdev->cache_valid & VALID_IN4)) {
1111 const struct sockaddr_in *sin;
1115 ifr.ifr_addr.sa_family = AF_INET;
1116 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1117 SIOCGIFADDR, "SIOCGIFADDR");
1122 sin = (struct sockaddr_in *) &ifr.ifr_addr;
1123 netdev->in4 = sin->sin_addr;
1124 netdev->cache_valid |= VALID_IN4;
1125 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1126 SIOCGIFNETMASK, "SIOCGIFNETMASK");
1130 *netmask = ((struct sockaddr_in*)&ifr.ifr_addr)->sin_addr;
1134 return in4->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0;
1138 * Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
1139 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
1140 * positive errno value.
1143 netdev_bsd_set_in4(struct netdev *netdev_, struct in_addr addr,
1144 struct in_addr mask)
1146 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1149 error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", addr);
1151 netdev->cache_valid |= VALID_IN4;
1153 if (addr.s_addr != INADDR_ANY) {
1154 error = do_set_addr(netdev_, SIOCSIFNETMASK,
1155 "SIOCSIFNETMASK", mask);
1157 netdev_bsd_changed(netdev);
1163 netdev_bsd_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
1165 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1166 if (!(netdev->cache_valid & VALID_IN6)) {
1167 struct ifaddrs *ifa, *head;
1168 struct sockaddr_in6 *sin6;
1169 const char *netdev_name = netdev_get_name(netdev_);
1171 if (getifaddrs(&head) != 0) {
1172 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1173 ovs_strerror(errno));
1177 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1178 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1179 !strcmp(ifa->ifa_name, netdev_name)) {
1180 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1182 memcpy(&netdev->in6, &sin6->sin6_addr, sin6->sin6_len);
1183 netdev->cache_valid |= VALID_IN6;
1190 return EADDRNOTAVAIL;
1196 #if defined(__NetBSD__)
1197 static struct netdev *
1198 find_netdev_by_kernel_name(const char *kernel_name)
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_bsd * const dev = node->data;
1208 if (!strcmp(dev->kernel_name, kernel_name)) {
1209 shash_destroy(&device_shash);
1213 shash_destroy(&device_shash);
1218 netdev_bsd_convert_kernel_name_to_ovs_name(const char *kernel_name)
1220 const struct netdev * const netdev =
1221 find_netdev_by_kernel_name(kernel_name);
1223 if (netdev == NULL) {
1226 return netdev_get_name(netdev);
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;
1308 const size_t nlen = sdl->sdl_nlen;
1309 char * const kernel_name = xmalloc(nlen + 1);
1312 memcpy(kernel_name, sdl->sdl_data, nlen);
1313 kernel_name[nlen] = 0;
1314 name = netdev_bsd_convert_kernel_name_to_ovs_name(kernel_name);
1316 ifname = xstrdup(kernel_name);
1318 ifname = xstrdup(name);
1325 if (ifname == NULL) {
1331 *netdev_name = ifname;
1332 VLOG_DBG("host " IP_FMT " next-hop " IP_FMT " if %s",
1333 IP_ARGS(host->s_addr), IP_ARGS(next_hop->s_addr), *netdev_name);
1341 make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
1343 struct sockaddr_in sin;
1344 memset(&sin, 0, sizeof sin);
1345 sin.sin_family = AF_INET;
1346 sin.sin_addr = addr;
1349 memset(sa, 0, sizeof *sa);
1350 memcpy(sa, &sin, sizeof sin);
1354 do_set_addr(struct netdev *netdev,
1355 int ioctl_nr, const char *ioctl_name, struct in_addr addr)
1358 make_in4_sockaddr(&ifr.ifr_addr, addr);
1359 return netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev), &ifr, ioctl_nr,
1364 nd_to_iff_flags(enum netdev_flags nd)
1367 if (nd & NETDEV_UP) {
1370 if (nd & NETDEV_PROMISC) {
1372 #if defined(IFF_PPROMISC)
1373 iff |= IFF_PPROMISC;
1380 iff_to_nd_flags(int iff)
1382 enum netdev_flags nd = 0;
1386 if (iff & IFF_PROMISC) {
1387 nd |= NETDEV_PROMISC;
1393 netdev_bsd_update_flags(struct netdev *netdev_, enum netdev_flags off,
1394 enum netdev_flags on, enum netdev_flags *old_flagsp)
1396 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1397 int old_flags, new_flags;
1400 error = get_flags(netdev_, &old_flags);
1402 *old_flagsp = iff_to_nd_flags(old_flags);
1403 new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on);
1404 if (new_flags != old_flags) {
1405 error = set_flags(netdev_get_kernel_name(netdev_), new_flags);
1406 netdev_bsd_changed(netdev);
1413 netdev_bsd_change_seq(const struct netdev *netdev)
1415 return netdev_bsd_cast(netdev)->change_seq;
1419 const struct netdev_class netdev_bsd_class = {
1425 netdev_bsd_create_system,
1427 NULL, /* get_config */
1428 NULL, /* set_config */
1429 NULL, /* get_tunnel_config */
1434 netdev_bsd_send_wait,
1436 netdev_bsd_set_etheraddr,
1437 netdev_bsd_get_etheraddr,
1440 netdev_bsd_get_ifindex,
1441 netdev_bsd_get_carrier,
1442 NULL, /* get_carrier_resets */
1443 NULL, /* set_miimon_interval */
1444 netdev_bsd_get_stats,
1445 NULL, /* set_stats */
1447 netdev_bsd_get_features,
1448 NULL, /* set_advertisement */
1449 NULL, /* set_policing */
1450 NULL, /* get_qos_type */
1451 NULL, /* get_qos_capabilities */
1454 NULL, /* get_queue */
1455 NULL, /* set_queue */
1456 NULL, /* delete_queue */
1457 NULL, /* get_queue_stats */
1458 NULL, /* dump_queue */
1459 NULL, /* dump_queue_stats */
1464 NULL, /* add_router */
1465 netdev_bsd_get_next_hop,
1466 NULL, /* get_status */
1467 NULL, /* arp_lookup */
1469 netdev_bsd_update_flags,
1471 netdev_bsd_change_seq
1474 const struct netdev_class netdev_tap_class = {
1480 netdev_bsd_create_tap,
1482 NULL, /* get_config */
1483 NULL, /* set_config */
1484 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 NULL, /* arp_lookup */
1524 netdev_bsd_update_flags,
1526 netdev_bsd_change_seq
1529 static const struct netdev_rx_class netdev_rx_bsd_class = {
1530 netdev_rx_bsd_destroy,
1533 netdev_rx_bsd_drain,
1538 destroy_tap(int fd, const char *name)
1543 strcpy(ifr.ifr_name, name);
1544 /* XXX What to do if this call fails? */
1545 ioctl(af_inet_sock, SIOCIFDESTROY, &ifr);
1549 get_flags(const struct netdev *netdev, int *flags)
1554 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev), &ifr,
1555 SIOCGIFFLAGS, "SIOCGIFFLAGS");
1557 *flags = ifr_get_flags(&ifr);
1563 set_flags(const char *name, int flags)
1567 ifr_set_flags(&ifr, flags);
1569 return netdev_bsd_do_ioctl(name, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS");
1573 get_ifindex(const struct netdev *netdev_, int *ifindexp)
1575 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1577 if (!(netdev->cache_valid & VALID_IFINDEX)) {
1578 int ifindex = if_nametoindex(netdev_get_name(netdev_));
1582 netdev->cache_valid |= VALID_IFINDEX;
1583 netdev->ifindex = ifindex;
1585 *ifindexp = netdev->ifindex;
1590 get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN])
1592 struct ifaddrs *head;
1593 struct ifaddrs *ifa;
1594 struct sockaddr_dl *sdl;
1596 if (getifaddrs(&head) != 0) {
1597 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1598 ovs_strerror(errno));
1602 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1603 if (ifa->ifa_addr->sa_family == AF_LINK) {
1604 if (!strcmp(ifa->ifa_name, netdev_name)) {
1605 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1607 memcpy(ea, LLADDR(sdl), sdl->sdl_alen);
1615 VLOG_ERR("could not find ethernet address for %s device", netdev_name);
1621 set_etheraddr(const char *netdev_name OVS_UNUSED, int hwaddr_family OVS_UNUSED,
1622 int hwaddr_len OVS_UNUSED,
1623 const uint8_t mac[ETH_ADDR_LEN] OVS_UNUSED)
1625 #if defined(__FreeBSD__)
1628 memset(&ifr, 0, sizeof ifr);
1629 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1630 ifr.ifr_addr.sa_family = hwaddr_family;
1631 ifr.ifr_addr.sa_len = hwaddr_len;
1632 memcpy(ifr.ifr_addr.sa_data, mac, hwaddr_len);
1633 if (ioctl(af_inet_sock, SIOCSIFLLADDR, &ifr) < 0) {
1634 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1635 netdev_name, ovs_strerror(errno));
1639 #elif defined(__NetBSD__)
1640 struct if_laddrreq req;
1641 struct sockaddr_dl *sdl;
1642 struct sockaddr_storage oldaddr;
1646 * get the old address, add new one, and then remove old one.
1649 if (hwaddr_len != ETH_ADDR_LEN) {
1650 /* just to be safe about sockaddr storage size */
1653 memset(&req, 0, sizeof(req));
1654 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1655 req.addr.ss_len = sizeof(req.addr);
1656 req.addr.ss_family = hwaddr_family;
1657 sdl = (struct sockaddr_dl *)&req.addr;
1658 sdl->sdl_alen = hwaddr_len;
1659 ret = ioctl(af_link_sock, SIOCGLIFADDR, &req);
1663 if (!memcmp(&sdl->sdl_data[sdl->sdl_nlen], mac, hwaddr_len)) {
1668 memset(&req, 0, sizeof(req));
1669 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1670 req.flags = IFLR_ACTIVE;
1671 sdl = (struct sockaddr_dl *)&req.addr;
1672 sdl->sdl_len = offsetof(struct sockaddr_dl, sdl_data) + hwaddr_len;
1673 sdl->sdl_alen = hwaddr_len;
1674 sdl->sdl_family = hwaddr_family;
1675 memcpy(sdl->sdl_data, mac, hwaddr_len);
1676 ret = ioctl(af_link_sock, SIOCALIFADDR, &req);
1681 memset(&req, 0, sizeof(req));
1682 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1684 ret = ioctl(af_link_sock, SIOCDLIFADDR, &req);
1690 #error not implemented
1695 netdev_bsd_do_ioctl(const char *name, struct ifreq *ifr, unsigned long cmd,
1696 const char *cmd_name)
1698 strncpy(ifr->ifr_name, name, sizeof ifr->ifr_name);
1699 if (ioctl(af_inet_sock, cmd, ifr) == -1) {
1700 VLOG_DBG_RL(&rl, "%s: ioctl(%s) failed: %s", name, cmd_name,
1701 ovs_strerror(errno));
1708 ifr_get_flags(const struct ifreq *ifr)
1710 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1711 return (ifr->ifr_flagshigh << 16) | ifr->ifr_flags;
1713 return ifr->ifr_flags;
1718 ifr_set_flags(struct ifreq *ifr, int flags)
1720 ifr->ifr_flags = flags;
1721 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1722 ifr->ifr_flagshigh = flags >> 16;