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);
261 * XXX the API is lacking, we should be able to iterate on the list of
262 * netdevs without having to store the info in a temp shash.
264 struct shash device_shash;
265 struct shash_node *node;
267 shash_init(&device_shash);
268 netdev_get_devices(&netdev_bsd_class, &device_shash);
269 SHASH_FOR_EACH (node, &device_shash) {
270 struct netdev *netdev = node->data;
271 dev = netdev_bsd_cast(netdev);
272 dev->cache_valid = 0;
273 netdev_bsd_changed(dev);
275 shash_destroy(&device_shash);
280 cache_notifier_ref(void)
284 if (!cache_notifier_refcount) {
285 ret = rtbsd_notifier_register(&netdev_bsd_cache_notifier,
286 netdev_bsd_cache_cb, NULL);
291 cache_notifier_refcount++;
296 cache_notifier_unref(void)
298 cache_notifier_refcount--;
299 if (cache_notifier_refcount == 0) {
300 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier);
305 /* Allocate a netdev_bsd structure */
307 netdev_bsd_create_system(const struct netdev_class *class, const char *name,
308 struct netdev **netdevp)
310 struct netdev_bsd *netdev;
311 enum netdev_flags flags;
314 error = cache_notifier_ref();
319 netdev = xzalloc(sizeof *netdev);
320 netdev->change_seq = 1;
321 netdev_init(&netdev->up, name, class);
323 netdev->kernel_name = xstrdup(name);
325 /* Verify that the netdev really exists by attempting to read its flags */
326 error = netdev_get_flags(&netdev->up, &flags);
327 if (error == ENXIO) {
328 free(netdev->kernel_name);
329 netdev_uninit(&netdev->up, false);
331 cache_notifier_unref();
335 *netdevp = &netdev->up;
340 * Allocate a netdev_bsd structure with 'tap' class.
343 netdev_bsd_create_tap(const struct netdev_class *class, const char *name,
344 struct netdev **netdevp)
346 struct netdev_bsd *netdev = NULL;
349 char *kernel_name = NULL;
351 error = cache_notifier_ref();
356 /* allocate the device structure and set the internal flag */
357 netdev = xzalloc(sizeof *netdev);
359 memset(&ifr, 0, sizeof(ifr));
361 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
362 * to retrieve the name of the tap device. */
363 netdev->tap_fd = open("/dev/tap", O_RDWR);
364 netdev->change_seq = 1;
365 if (netdev->tap_fd < 0) {
367 VLOG_WARN("opening \"/dev/tap\" failed: %s", ovs_strerror(error));
368 goto error_unref_notifier;
371 /* Retrieve tap name (e.g. tap0) */
372 if (ioctl(netdev->tap_fd, TAPGIFNAME, &ifr) == -1) {
373 /* XXX Need to destroy the device? */
375 close(netdev->tap_fd);
376 goto error_unref_notifier;
379 /* Change the name of the tap device */
380 #if defined(SIOCSIFNAME)
381 ifr.ifr_data = (void *)name;
382 if (ioctl(af_inet_sock, SIOCSIFNAME, &ifr) == -1) {
384 destroy_tap(netdev->tap_fd, ifr.ifr_name);
385 goto error_unref_notifier;
387 kernel_name = xstrdup(name);
390 * NetBSD doesn't support inteface renaming.
392 VLOG_INFO("tap %s is created for bridge %s", ifr.ifr_name, name);
393 kernel_name = xstrdup(ifr.ifr_name);
396 /* set non-blocking. */
397 error = set_nonblocking(netdev->tap_fd);
399 destroy_tap(netdev->tap_fd, kernel_name);
400 goto error_unref_notifier;
404 ifr_set_flags(&ifr, IFF_UP);
405 strncpy(ifr.ifr_name, kernel_name, sizeof ifr.ifr_name);
406 if (ioctl(af_inet_sock, SIOCSIFFLAGS, &ifr) == -1) {
408 destroy_tap(netdev->tap_fd, kernel_name);
409 goto error_unref_notifier;
412 /* initialize the device structure and
413 * link the structure to its netdev */
414 netdev_init(&netdev->up, name, class);
415 netdev->kernel_name = kernel_name;
416 *netdevp = &netdev->up;
420 error_unref_notifier:
421 cache_notifier_unref();
429 netdev_bsd_destroy(struct netdev *netdev_)
431 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
433 cache_notifier_unref();
435 if (netdev->tap_fd >= 0) {
436 destroy_tap(netdev->tap_fd, netdev_get_kernel_name(netdev_));
439 pcap_close(netdev->pcap);
441 free(netdev->kernel_name);
446 netdev_bsd_open_pcap(const char *name, pcap_t **pcapp, int *fdp)
448 char errbuf[PCAP_ERRBUF_SIZE];
454 /* Open the pcap device. The device is opened in non-promiscuous mode
455 * because the interface flags are manually set by the caller. */
457 pcap = pcap_open_live(name, PCAP_SNAPLEN, 0, 1000, errbuf);
459 VLOG_ERR_RL(&rl, "%s: pcap_open_live failed: %s", name, errbuf);
463 if (errbuf[0] != '\0') {
464 VLOG_WARN_RL(&rl, "%s: pcap_open_live: %s", name, errbuf);
467 /* Get the underlying fd. */
468 fd = pcap_get_selectable_fd(pcap);
470 VLOG_WARN_RL(&rl, "%s: no selectable file descriptor", name);
475 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
476 * on the file descriptor returned by pcap_get_selectable_fd to achieve
477 * a real non-blocking behaviour.*/
478 error = pcap_setnonblock(pcap, 1, errbuf);
484 /* This call assure that reads return immediately upon packet
485 * reception. Otherwise, a read will block until either the kernel
486 * buffer becomes full or a timeout occurs. */
487 if (ioctl(fd, BIOCIMMEDIATE, &one) < 0 ) {
488 VLOG_ERR_RL(&rl, "ioctl(BIOCIMMEDIATE) on %s device failed: %s",
489 name, ovs_strerror(errno));
494 /* Capture only incoming packets. */
495 error = pcap_setdirection(pcap, PCAP_D_IN);
515 netdev_bsd_rx_open(struct netdev *netdev_, struct netdev_rx **rxp)
517 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
519 struct netdev_rx_bsd *rx;
523 if (!strcmp(netdev_get_type(netdev_), "tap")) {
527 int error = netdev_bsd_open_pcap(netdev_get_kernel_name(netdev_),
533 netdev_bsd_changed(netdev);
536 rx = xmalloc(sizeof *rx);
537 netdev_rx_init(&rx->up, netdev_, &netdev_rx_bsd_class);
538 rx->pcap_handle = pcap;
546 netdev_rx_bsd_destroy(struct netdev_rx *rx_)
548 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
550 if (rx->pcap_handle) {
551 pcap_close(rx->pcap_handle);
556 /* The recv callback of the netdev class returns the number of bytes of the
559 * This can be done by the pcap_next() function. Unfortunately pcap_next() does
560 * not make difference between a missing packet on the capture interface and
561 * an error during the file capture. We can use the pcap_dispatch() function
562 * instead, which is able to distinguish between errors and null packet.
564 * To make pcap_dispatch() returns the number of bytes read from the interface
565 * we need to define the following callback and argument.
574 * This callback will be executed on every captured packet.
576 * If the packet captured by pcap_dispatch() does not fit the pcap buffer,
577 * pcap returns a truncated packet and we follow this behavior.
579 * The argument args->retval is the packet size in bytes.
582 proc_pkt(u_char *args_, const struct pcap_pkthdr *hdr, const u_char *packet)
584 struct pcap_arg *args = (struct pcap_arg *)args_;
586 if (args->size < hdr->len) {
587 VLOG_WARN_RL(&rl, "packet truncated");
588 args->retval = args->size;
590 args->retval = hdr->len;
593 /* copy the packet to our buffer */
594 memcpy(args->data, packet, args->retval);
598 * This function attempts to receive a packet from the specified network
599 * device. It is assumed that the network device is a system device or a tap
600 * device opened as a system one. In this case the read operation is performed
604 netdev_rx_bsd_recv_pcap(struct netdev_rx_bsd *rx, void *data, size_t size)
609 /* prepare the pcap argument to store the packet */
614 ret = pcap_dispatch(rx->pcap_handle, 1, proc_pkt, (u_char *) &arg);
617 return arg.retval; /* arg.retval < 0 is handled in the caller */
620 if (errno == EINTR) {
630 * This function attempts to receive a packet from the specified network
631 * device. It is assumed that the network device is a tap device and
632 * 'rx->fd' is initialized with the tap file descriptor.
635 netdev_rx_bsd_recv_tap(struct netdev_rx_bsd *rx, void *data, size_t size)
638 ssize_t retval = read(rx->fd, data, size);
641 } else if (errno != EINTR) {
642 if (errno != EAGAIN) {
643 VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
644 ovs_strerror(errno), netdev_rx_get_name(&rx->up));
653 netdev_rx_bsd_recv(struct netdev_rx *rx_, void *data, size_t size)
655 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
657 return (rx->pcap_handle
658 ? netdev_rx_bsd_recv_pcap(rx, data, size)
659 : netdev_rx_bsd_recv_tap(rx, data, size));
663 * Registers with the poll loop to wake up from the next call to poll_block()
664 * when a packet is ready to be received with netdev_rx_recv() on 'rx'.
667 netdev_rx_bsd_wait(struct netdev_rx *rx_)
669 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
671 poll_fd_wait(rx->fd, POLLIN);
674 /* Discards all packets waiting to be received from 'rx'. */
676 netdev_rx_bsd_drain(struct netdev_rx *rx_)
679 struct netdev_rx_bsd *rx = netdev_rx_bsd_cast(rx_);
681 strcpy(ifr.ifr_name, netdev_get_kernel_name(netdev_rx_get_netdev(rx_)));
682 if (ioctl(rx->fd, BIOCFLUSH, &ifr) == -1) {
683 VLOG_DBG_RL(&rl, "%s: ioctl(BIOCFLUSH) failed: %s",
684 netdev_rx_get_name(rx_), ovs_strerror(errno));
691 * Send a packet on the specified network device. The device could be either a
692 * system or a tap device.
695 netdev_bsd_send(struct netdev *netdev_, const void *data, size_t size)
697 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
698 const char *name = netdev_get_name(netdev_);
700 if (dev->tap_fd < 0 && !dev->pcap) {
701 int error = netdev_bsd_open_pcap(name, &dev->pcap, &dev->fd);
709 if (dev->tap_fd >= 0) {
710 retval = write(dev->tap_fd, data, size);
712 retval = pcap_inject(dev->pcap, data, size);
715 if (errno == EINTR) {
717 } else if (errno != EAGAIN) {
718 VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s",
719 name, ovs_strerror(errno));
722 } else if (retval != size) {
723 VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of "
724 "%zu) on %s", retval, size, name);
733 * Registers with the poll loop to wake up from the next call to poll_block()
734 * when the packet transmission queue has sufficient room to transmit a packet
735 * with netdev_send().
738 netdev_bsd_send_wait(struct netdev *netdev_)
740 struct netdev_bsd *dev = netdev_bsd_cast(netdev_);
742 if (dev->tap_fd >= 0) {
743 /* TAP device always accepts packets. */
744 poll_immediate_wake();
745 } else if (dev->pcap) {
746 poll_fd_wait(dev->fd, POLLOUT);
748 /* We haven't even tried to send a packet yet. */
749 poll_immediate_wake();
754 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
755 * otherwise a positive errno value.
758 netdev_bsd_set_etheraddr(struct netdev *netdev_,
759 const uint8_t mac[ETH_ADDR_LEN])
761 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
764 if (!(netdev->cache_valid & VALID_ETHERADDR)
765 || !eth_addr_equals(netdev->etheraddr, mac)) {
766 error = set_etheraddr(netdev_get_kernel_name(netdev_), AF_LINK,
769 netdev->cache_valid |= VALID_ETHERADDR;
770 memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN);
771 netdev_bsd_changed(netdev);
780 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
781 * free the returned buffer.
784 netdev_bsd_get_etheraddr(const struct netdev *netdev_,
785 uint8_t mac[ETH_ADDR_LEN])
787 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
789 if (!(netdev->cache_valid & VALID_ETHERADDR)) {
790 int error = get_etheraddr(netdev_get_kernel_name(netdev_),
795 netdev->cache_valid |= VALID_ETHERADDR;
797 memcpy(mac, netdev->etheraddr, ETH_ADDR_LEN);
803 * Returns the maximum size of transmitted (and received) packets on 'netdev',
804 * in bytes, not including the hardware header; thus, this is typically 1500
805 * bytes for Ethernet devices.
808 netdev_bsd_get_mtu(const struct netdev *netdev_, int *mtup)
810 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
812 if (!(netdev->cache_valid & VALID_MTU)) {
816 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev_), &ifr,
817 SIOCGIFMTU, "SIOCGIFMTU");
821 netdev->mtu = ifr.ifr_mtu;
822 netdev->cache_valid |= VALID_MTU;
830 netdev_bsd_get_ifindex(const struct netdev *netdev)
834 error = get_ifindex(netdev, &ifindex);
835 return error ? -error : ifindex;
839 netdev_bsd_get_carrier(const struct netdev *netdev_, bool *carrier)
841 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
843 if (!(netdev->cache_valid & VALID_CARRIER)) {
844 struct ifmediareq ifmr;
846 memset(&ifmr, 0, sizeof(ifmr));
847 strncpy(ifmr.ifm_name, netdev_get_kernel_name(netdev_),
848 sizeof ifmr.ifm_name);
850 if (ioctl(af_inet_sock, SIOCGIFMEDIA, &ifmr) == -1) {
851 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
852 netdev_get_name(netdev_), ovs_strerror(errno));
856 netdev->carrier = (ifmr.ifm_status & IFM_ACTIVE) == IFM_ACTIVE;
857 netdev->cache_valid |= VALID_CARRIER;
859 /* If the interface doesn't report whether the media is active,
860 * just assume it is active. */
861 if ((ifmr.ifm_status & IFM_AVALID) == 0) {
862 netdev->carrier = true;
865 *carrier = netdev->carrier;
871 convert_stats(struct netdev_stats *stats, const struct if_data *ifd)
874 * note: UINT64_MAX means unsupported
876 stats->rx_packets = ifd->ifi_ipackets;
877 stats->tx_packets = ifd->ifi_opackets;
878 stats->rx_bytes = ifd->ifi_obytes;
879 stats->tx_bytes = ifd->ifi_ibytes;
880 stats->rx_errors = ifd->ifi_ierrors;
881 stats->tx_errors = ifd->ifi_oerrors;
882 stats->rx_dropped = ifd->ifi_iqdrops;
883 stats->tx_dropped = UINT64_MAX;
884 stats->multicast = ifd->ifi_imcasts;
885 stats->collisions = ifd->ifi_collisions;
886 stats->rx_length_errors = UINT64_MAX;
887 stats->rx_over_errors = UINT64_MAX;
888 stats->rx_crc_errors = UINT64_MAX;
889 stats->rx_frame_errors = UINT64_MAX;
890 stats->rx_fifo_errors = UINT64_MAX;
891 stats->rx_missed_errors = UINT64_MAX;
892 stats->tx_aborted_errors = UINT64_MAX;
893 stats->tx_carrier_errors = UINT64_MAX;
894 stats->tx_fifo_errors = UINT64_MAX;
895 stats->tx_heartbeat_errors = UINT64_MAX;
896 stats->tx_window_errors = UINT64_MAX;
899 /* Retrieves current device stats for 'netdev'. */
901 netdev_bsd_get_stats(const struct netdev *netdev_, struct netdev_stats *stats)
903 #if defined(__FreeBSD__)
907 struct ifmibdata ifmd;
912 mib[2] = NETLINK_GENERIC;
913 mib[3] = IFMIB_SYSTEM;
914 mib[4] = IFMIB_IFCOUNT;
916 len = sizeof(if_count);
918 if (sysctl(mib, 5, &if_count, &len, (void *)0, 0) == -1) {
919 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
920 netdev_get_name(netdev_), ovs_strerror(errno));
924 mib[5] = IFDATA_GENERAL;
925 mib[3] = IFMIB_IFDATA;
927 for (i = 1; i <= if_count; i++) {
929 if (sysctl(mib, 6, &ifmd, &len, (void *)0, 0) == -1) {
930 VLOG_DBG_RL(&rl, "%s: sysctl failed: %s",
931 netdev_get_name(netdev_), ovs_strerror(errno));
933 } else if (!strcmp(ifmd.ifmd_name, netdev_get_name(netdev_))) {
934 convert_stats(stats, &ifmd.ifmd_data);
940 #elif defined(__NetBSD__)
941 struct ifdatareq ifdr;
945 memset(&ifdr, 0, sizeof(ifdr));
946 strncpy(ifdr.ifdr_name, netdev_get_kernel_name(netdev_),
947 sizeof(ifdr.ifdr_name));
948 ret = ioctl(af_link_sock, SIOCGIFDATA, &ifdr);
953 convert_stats(stats, &ifdr.ifdr_data);
956 #error not implemented
961 netdev_bsd_parse_media(int media)
963 uint32_t supported = 0;
964 bool half_duplex = media & IFM_HDX ? true : false;
966 switch (IFM_SUBTYPE(media)) {
971 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
972 supported |= NETDEV_F_COPPER;
976 supported |= half_duplex ? NETDEV_F_10MB_HD : NETDEV_F_10MB_FD;
977 supported |= NETDEV_F_FIBER;
984 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
985 supported |= NETDEV_F_COPPER;
989 supported |= half_duplex ? NETDEV_F_100MB_HD : NETDEV_F_100MB_FD;
990 supported |= NETDEV_F_FIBER;
995 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
996 supported |= NETDEV_F_COPPER;
1001 supported |= half_duplex ? NETDEV_F_1GB_HD : NETDEV_F_1GB_FD;
1002 supported |= NETDEV_F_FIBER;
1006 supported |= NETDEV_F_10GB_FD;
1007 supported |= NETDEV_F_COPPER;
1012 supported |= NETDEV_F_10GB_FD;
1013 supported |= NETDEV_F_FIBER;
1020 if (IFM_SUBTYPE(media) == IFM_AUTO) {
1021 supported |= NETDEV_F_AUTONEG;
1024 if (media & IFM_ETH_FMASK) {
1025 supported |= NETDEV_F_PAUSE;
1033 * Stores the features supported by 'netdev' into each of '*current',
1034 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1035 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1036 * successful, otherwise a positive errno value. On failure, all of the
1037 * passed-in values are set to 0.
1040 netdev_bsd_get_features(const struct netdev *netdev,
1041 enum netdev_features *current, uint32_t *advertised,
1042 enum netdev_features *supported, uint32_t *peer)
1044 struct ifmediareq ifmr;
1050 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1052 memset(&ifmr, 0, sizeof(ifmr));
1053 strncpy(ifmr.ifm_name, netdev_get_name(netdev), sizeof ifmr.ifm_name);
1055 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1056 * number of supported modes, and a second with a buffer to retrieve
1058 if (ioctl(af_inet_sock, SIOCGIFMEDIA, &ifmr) == -1) {
1059 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1060 netdev_get_name(netdev), ovs_strerror(errno));
1064 media_list = xcalloc(ifmr.ifm_count, sizeof(int));
1065 ifmr.ifm_ulist = media_list;
1067 if (IFM_TYPE(ifmr.ifm_current) != IFM_ETHER) {
1068 VLOG_DBG_RL(&rl, "%s: doesn't appear to be ethernet",
1069 netdev_get_name(netdev));
1074 if (ioctl(af_inet_sock, SIOCGIFMEDIA, &ifmr) == -1) {
1075 VLOG_DBG_RL(&rl, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1076 netdev_get_name(netdev), ovs_strerror(errno));
1081 /* Current settings. */
1082 *current = netdev_bsd_parse_media(ifmr.ifm_active);
1084 /* Advertised features. */
1085 *advertised = netdev_bsd_parse_media(ifmr.ifm_current);
1087 /* Supported features. */
1089 for (i = 0; i < ifmr.ifm_count; i++) {
1090 *supported |= netdev_bsd_parse_media(ifmr.ifm_ulist[i]);
1093 /* Peer advertisements. */
1094 *peer = 0; /* XXX */
1103 * If 'netdev' has an assigned IPv4 address, sets '*in4' to that address and
1104 * '*netmask' to its netmask and returns true. Otherwise, returns false.
1107 netdev_bsd_get_in4(const struct netdev *netdev_, struct in_addr *in4,
1108 struct in_addr *netmask)
1110 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1112 if (!(netdev->cache_valid & VALID_IN4)) {
1113 const struct sockaddr_in *sin;
1117 ifr.ifr_addr.sa_family = AF_INET;
1118 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1119 SIOCGIFADDR, "SIOCGIFADDR");
1124 sin = (struct sockaddr_in *) &ifr.ifr_addr;
1125 netdev->in4 = sin->sin_addr;
1126 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev_), &ifr,
1127 SIOCGIFNETMASK, "SIOCGIFNETMASK");
1131 netdev->netmask = sin->sin_addr;
1132 netdev->cache_valid |= VALID_IN4;
1135 *netmask = netdev->netmask;
1137 return in4->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0;
1141 * Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
1142 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
1143 * positive errno value.
1146 netdev_bsd_set_in4(struct netdev *netdev_, struct in_addr addr,
1147 struct in_addr mask)
1149 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1152 error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", addr);
1154 if (addr.s_addr != INADDR_ANY) {
1155 error = do_set_addr(netdev_, SIOCSIFNETMASK,
1156 "SIOCSIFNETMASK", mask);
1158 netdev->cache_valid |= VALID_IN4;
1160 netdev->netmask = mask;
1163 netdev_bsd_changed(netdev);
1169 netdev_bsd_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
1171 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1172 if (!(netdev->cache_valid & VALID_IN6)) {
1173 struct ifaddrs *ifa, *head;
1174 struct sockaddr_in6 *sin6;
1175 const char *netdev_name = netdev_get_name(netdev_);
1177 if (getifaddrs(&head) != 0) {
1178 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1179 ovs_strerror(errno));
1183 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1184 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1185 !strcmp(ifa->ifa_name, netdev_name)) {
1186 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1188 memcpy(&netdev->in6, &sin6->sin6_addr, sin6->sin6_len);
1189 netdev->cache_valid |= VALID_IN6;
1196 return EADDRNOTAVAIL;
1202 #if defined(__NetBSD__)
1203 static struct netdev *
1204 find_netdev_by_kernel_name(const char *kernel_name)
1206 struct shash device_shash;
1207 struct shash_node *node;
1209 shash_init(&device_shash);
1210 netdev_get_devices(&netdev_tap_class, &device_shash);
1211 SHASH_FOR_EACH(node, &device_shash) {
1212 struct netdev *netdev = node->data;
1213 struct netdev_bsd * const dev = netdev_bsd_cast(netdev);
1215 if (!strcmp(dev->kernel_name, kernel_name)) {
1216 shash_destroy(&device_shash);
1220 shash_destroy(&device_shash);
1225 netdev_bsd_convert_kernel_name_to_ovs_name(const char *kernel_name)
1227 const struct netdev * const netdev =
1228 find_netdev_by_kernel_name(kernel_name);
1230 if (netdev == NULL) {
1233 return netdev_get_name(netdev);
1238 netdev_bsd_get_next_hop(const struct in_addr *host OVS_UNUSED,
1239 struct in_addr *next_hop OVS_UNUSED,
1240 char **netdev_name OVS_UNUSED)
1242 #if defined(__NetBSD__)
1244 struct sockaddr_in sin;
1245 struct sockaddr_dl sdl;
1252 struct rt_msghdr *rtm = &buf.h;
1253 const pid_t pid = getpid();
1256 bool gateway = false;
1257 char *ifname = NULL;
1260 memset(next_hop, 0, sizeof(*next_hop));
1261 *netdev_name = NULL;
1263 memset(&sin, 0, sizeof(sin));
1264 sin.sin_len = sizeof(sin);
1265 sin.sin_family = AF_INET;
1267 sin.sin_addr = *host;
1269 memset(&sdl, 0, sizeof(sdl));
1270 sdl.sdl_len = sizeof(sdl);
1271 sdl.sdl_family = AF_LINK;
1273 s = socket(PF_ROUTE, SOCK_RAW, 0);
1274 memset(&buf, 0, sizeof(buf));
1275 rtm->rtm_flags = RTF_HOST|RTF_UP;
1276 rtm->rtm_version = RTM_VERSION;
1277 rtm->rtm_addrs = RTA_DST|RTA_IFP;
1278 cp = (void *)&buf.space;
1279 memcpy(cp, &sin, sizeof(sin));
1280 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sin);
1281 memcpy(cp, &sdl, sizeof(sdl));
1282 RT_ADVANCE(cp, (struct sockaddr *)(void *)&sdl);
1283 rtm->rtm_msglen = cp - (char *)(void *)rtm;
1284 rtm->rtm_seq = ++seq;
1285 rtm->rtm_type = RTM_GET;
1287 write(s, rtm, rtm->rtm_msglen);
1288 memset(&buf, 0, sizeof(buf));
1290 ssz = read(s, &buf, sizeof(buf));
1291 } while (ssz > 0 && (rtm->rtm_seq != seq || rtm->rtm_pid != pid));
1292 saved_errno = errno;
1298 return EPIPE; /* XXX */
1300 cp = (void *)&buf.space;
1301 for (i = 1; i; i <<= 1) {
1302 if ((rtm->rtm_addrs & i) != 0) {
1303 const struct sockaddr *sa = (const void *)cp;
1305 if ((i == RTA_GATEWAY) && sa->sa_family == AF_INET) {
1306 const struct sockaddr_in * const sin =
1307 (const struct sockaddr_in *)sa;
1309 *next_hop = sin->sin_addr;
1312 if ((i == RTA_IFP) && sa->sa_family == AF_LINK) {
1313 const struct sockaddr_dl * const sdl =
1314 (const struct sockaddr_dl *)sa;
1315 const size_t nlen = sdl->sdl_nlen;
1316 char * const kernel_name = xmalloc(nlen + 1);
1319 memcpy(kernel_name, sdl->sdl_data, nlen);
1320 kernel_name[nlen] = 0;
1321 name = netdev_bsd_convert_kernel_name_to_ovs_name(kernel_name);
1323 ifname = xstrdup(kernel_name);
1325 ifname = xstrdup(name);
1332 if (ifname == NULL) {
1338 *netdev_name = ifname;
1339 VLOG_DBG("host " IP_FMT " next-hop " IP_FMT " if %s",
1340 IP_ARGS(host->s_addr), IP_ARGS(next_hop->s_addr), *netdev_name);
1348 make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
1350 struct sockaddr_in sin;
1351 memset(&sin, 0, sizeof sin);
1352 sin.sin_family = AF_INET;
1353 sin.sin_addr = addr;
1356 memset(sa, 0, sizeof *sa);
1357 memcpy(sa, &sin, sizeof sin);
1361 do_set_addr(struct netdev *netdev,
1362 int ioctl_nr, const char *ioctl_name, struct in_addr addr)
1365 make_in4_sockaddr(&ifr.ifr_addr, addr);
1366 return netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev), &ifr, ioctl_nr,
1371 nd_to_iff_flags(enum netdev_flags nd)
1374 if (nd & NETDEV_UP) {
1377 if (nd & NETDEV_PROMISC) {
1379 #if defined(IFF_PPROMISC)
1380 iff |= IFF_PPROMISC;
1387 iff_to_nd_flags(int iff)
1389 enum netdev_flags nd = 0;
1393 if (iff & IFF_PROMISC) {
1394 nd |= NETDEV_PROMISC;
1400 netdev_bsd_update_flags(struct netdev *netdev_, enum netdev_flags off,
1401 enum netdev_flags on, enum netdev_flags *old_flagsp)
1403 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1404 int old_flags, new_flags;
1407 error = get_flags(netdev_, &old_flags);
1409 *old_flagsp = iff_to_nd_flags(old_flags);
1410 new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on);
1411 if (new_flags != old_flags) {
1412 error = set_flags(netdev_get_kernel_name(netdev_), new_flags);
1413 netdev_bsd_changed(netdev);
1420 netdev_bsd_change_seq(const struct netdev *netdev)
1422 return netdev_bsd_cast(netdev)->change_seq;
1426 const struct netdev_class netdev_bsd_class = {
1432 netdev_bsd_create_system,
1434 NULL, /* get_config */
1435 NULL, /* set_config */
1436 NULL, /* get_tunnel_config */
1441 netdev_bsd_send_wait,
1443 netdev_bsd_set_etheraddr,
1444 netdev_bsd_get_etheraddr,
1447 netdev_bsd_get_ifindex,
1448 netdev_bsd_get_carrier,
1449 NULL, /* get_carrier_resets */
1450 NULL, /* set_miimon_interval */
1451 netdev_bsd_get_stats,
1452 NULL, /* set_stats */
1454 netdev_bsd_get_features,
1455 NULL, /* set_advertisement */
1456 NULL, /* set_policing */
1457 NULL, /* get_qos_type */
1458 NULL, /* get_qos_capabilities */
1461 NULL, /* get_queue */
1462 NULL, /* set_queue */
1463 NULL, /* delete_queue */
1464 NULL, /* get_queue_stats */
1465 NULL, /* dump_queue */
1466 NULL, /* dump_queue_stats */
1471 NULL, /* add_router */
1472 netdev_bsd_get_next_hop,
1473 NULL, /* get_status */
1474 NULL, /* arp_lookup */
1476 netdev_bsd_update_flags,
1478 netdev_bsd_change_seq
1481 const struct netdev_class netdev_tap_class = {
1487 netdev_bsd_create_tap,
1489 NULL, /* get_config */
1490 NULL, /* set_config */
1491 NULL, /* get_tunnel_config */
1496 netdev_bsd_send_wait,
1498 netdev_bsd_set_etheraddr,
1499 netdev_bsd_get_etheraddr,
1502 netdev_bsd_get_ifindex,
1503 netdev_bsd_get_carrier,
1504 NULL, /* get_carrier_resets */
1505 NULL, /* set_miimon_interval */
1506 netdev_bsd_get_stats,
1507 NULL, /* set_stats */
1509 netdev_bsd_get_features,
1510 NULL, /* set_advertisement */
1511 NULL, /* set_policing */
1512 NULL, /* get_qos_type */
1513 NULL, /* get_qos_capabilities */
1516 NULL, /* get_queue */
1517 NULL, /* set_queue */
1518 NULL, /* delete_queue */
1519 NULL, /* get_queue_stats */
1520 NULL, /* dump_queue */
1521 NULL, /* dump_queue_stats */
1526 NULL, /* add_router */
1527 netdev_bsd_get_next_hop,
1528 NULL, /* get_status */
1529 NULL, /* arp_lookup */
1531 netdev_bsd_update_flags,
1533 netdev_bsd_change_seq
1536 static const struct netdev_rx_class netdev_rx_bsd_class = {
1537 netdev_rx_bsd_destroy,
1540 netdev_rx_bsd_drain,
1545 destroy_tap(int fd, const char *name)
1550 strcpy(ifr.ifr_name, name);
1551 /* XXX What to do if this call fails? */
1552 ioctl(af_inet_sock, SIOCIFDESTROY, &ifr);
1556 get_flags(const struct netdev *netdev, int *flags)
1561 error = netdev_bsd_do_ioctl(netdev_get_kernel_name(netdev), &ifr,
1562 SIOCGIFFLAGS, "SIOCGIFFLAGS");
1564 *flags = ifr_get_flags(&ifr);
1570 set_flags(const char *name, int flags)
1574 ifr_set_flags(&ifr, flags);
1576 return netdev_bsd_do_ioctl(name, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS");
1580 get_ifindex(const struct netdev *netdev_, int *ifindexp)
1582 struct netdev_bsd *netdev = netdev_bsd_cast(netdev_);
1584 if (!(netdev->cache_valid & VALID_IFINDEX)) {
1585 int ifindex = if_nametoindex(netdev_get_name(netdev_));
1589 netdev->cache_valid |= VALID_IFINDEX;
1590 netdev->ifindex = ifindex;
1592 *ifindexp = netdev->ifindex;
1597 get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN])
1599 struct ifaddrs *head;
1600 struct ifaddrs *ifa;
1601 struct sockaddr_dl *sdl;
1603 if (getifaddrs(&head) != 0) {
1604 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name,
1605 ovs_strerror(errno));
1609 for (ifa = head; ifa; ifa = ifa->ifa_next) {
1610 if (ifa->ifa_addr->sa_family == AF_LINK) {
1611 if (!strcmp(ifa->ifa_name, netdev_name)) {
1612 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1614 memcpy(ea, LLADDR(sdl), sdl->sdl_alen);
1622 VLOG_ERR("could not find ethernet address for %s device", netdev_name);
1628 set_etheraddr(const char *netdev_name OVS_UNUSED, int hwaddr_family OVS_UNUSED,
1629 int hwaddr_len OVS_UNUSED,
1630 const uint8_t mac[ETH_ADDR_LEN] OVS_UNUSED)
1632 #if defined(__FreeBSD__)
1635 memset(&ifr, 0, sizeof ifr);
1636 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1637 ifr.ifr_addr.sa_family = hwaddr_family;
1638 ifr.ifr_addr.sa_len = hwaddr_len;
1639 memcpy(ifr.ifr_addr.sa_data, mac, hwaddr_len);
1640 if (ioctl(af_inet_sock, SIOCSIFLLADDR, &ifr) < 0) {
1641 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1642 netdev_name, ovs_strerror(errno));
1646 #elif defined(__NetBSD__)
1647 struct if_laddrreq req;
1648 struct sockaddr_dl *sdl;
1649 struct sockaddr_storage oldaddr;
1653 * get the old address, add new one, and then remove old one.
1656 if (hwaddr_len != ETH_ADDR_LEN) {
1657 /* just to be safe about sockaddr storage size */
1660 memset(&req, 0, sizeof(req));
1661 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1662 req.addr.ss_len = sizeof(req.addr);
1663 req.addr.ss_family = hwaddr_family;
1664 sdl = (struct sockaddr_dl *)&req.addr;
1665 sdl->sdl_alen = hwaddr_len;
1666 ret = ioctl(af_link_sock, SIOCGLIFADDR, &req);
1670 if (!memcmp(&sdl->sdl_data[sdl->sdl_nlen], mac, hwaddr_len)) {
1675 memset(&req, 0, sizeof(req));
1676 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1677 req.flags = IFLR_ACTIVE;
1678 sdl = (struct sockaddr_dl *)&req.addr;
1679 sdl->sdl_len = offsetof(struct sockaddr_dl, sdl_data) + hwaddr_len;
1680 sdl->sdl_alen = hwaddr_len;
1681 sdl->sdl_family = hwaddr_family;
1682 memcpy(sdl->sdl_data, mac, hwaddr_len);
1683 ret = ioctl(af_link_sock, SIOCALIFADDR, &req);
1688 memset(&req, 0, sizeof(req));
1689 strncpy(req.iflr_name, netdev_name, sizeof(req.iflr_name));
1691 ret = ioctl(af_link_sock, SIOCDLIFADDR, &req);
1697 #error not implemented
1702 netdev_bsd_do_ioctl(const char *name, struct ifreq *ifr, unsigned long cmd,
1703 const char *cmd_name)
1705 strncpy(ifr->ifr_name, name, sizeof ifr->ifr_name);
1706 if (ioctl(af_inet_sock, cmd, ifr) == -1) {
1707 VLOG_DBG_RL(&rl, "%s: ioctl(%s) failed: %s", name, cmd_name,
1708 ovs_strerror(errno));
1715 ifr_get_flags(const struct ifreq *ifr)
1717 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1718 return (ifr->ifr_flagshigh << 16) | ifr->ifr_flags;
1720 return ifr->ifr_flags;
1725 ifr_set_flags(struct ifreq *ifr, int flags)
1727 ifr->ifr_flags = flags;
1728 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1729 ifr->ifr_flagshigh = flags >> 16;