2 * Copyright (c) 2009 Nicira Networks.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
21 #include <arpa/inet.h>
23 #include <linux/if_tun.h>
24 #include <linux/types.h>
25 #include <linux/ethtool.h>
26 #include <linux/rtnetlink.h>
27 #include <linux/sockios.h>
28 #include <linux/version.h>
29 #include <sys/types.h>
30 #include <sys/ioctl.h>
31 #include <sys/socket.h>
32 #include <netpacket/packet.h>
33 #include <net/ethernet.h>
35 #include <net/if_arp.h>
36 #include <net/if_packet.h>
37 #include <net/route.h>
38 #include <netinet/in.h>
45 #include "dynamic-string.h"
46 #include "fatal-signal.h"
47 #include "netdev-provider.h"
50 #include "openflow/openflow.h"
52 #include "poll-loop.h"
53 #include "rtnetlink.h"
54 #include "socket-util.h"
58 #define THIS_MODULE VLM_netdev_linux
61 /* These were introduced in Linux 2.6.14, so they might be missing if we have
63 #ifndef ADVERTISED_Pause
64 #define ADVERTISED_Pause (1 << 13)
66 #ifndef ADVERTISED_Asym_Pause
67 #define ADVERTISED_Asym_Pause (1 << 14)
73 /* File descriptors. For ordinary network devices, the two fds below are
74 * the same; for tap devices, they differ. */
75 int netdev_fd; /* Network device. */
76 int tap_fd; /* TAP character device, if any, otherwise the
79 struct netdev_linux_cache *cache;
83 VALID_IFINDEX = 1 << 0,
84 VALID_ETHERADDR = 1 << 1,
88 VALID_CARRIER = 1 << 5,
89 VALID_IS_INTERNAL = 1 << 6
92 /* Cached network device information. */
93 struct netdev_linux_cache {
94 struct shash_node *shash_node;
99 uint8_t etheraddr[ETH_ADDR_LEN];
100 struct in_addr address, netmask;
107 static struct shash cache_map = SHASH_INITIALIZER(&cache_map);
108 static struct rtnetlink_notifier netdev_linux_cache_notifier;
110 /* An AF_INET socket (used for ioctl operations). */
111 static int af_inet_sock = -1;
113 struct netdev_linux_notifier {
114 struct netdev_notifier notifier;
118 static struct shash netdev_linux_notifiers =
119 SHASH_INITIALIZER(&netdev_linux_notifiers);
120 static struct rtnetlink_notifier netdev_linux_poll_notifier;
122 /* This is set pretty low because we probably won't learn anything from the
123 * additional log messages. */
124 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
126 static int netdev_linux_do_ethtool(struct netdev *, struct ethtool_cmd *,
127 int cmd, const char *cmd_name);
128 static int netdev_linux_do_ioctl(const struct netdev *, struct ifreq *,
129 int cmd, const char *cmd_name);
130 static int netdev_linux_get_ipv4(const struct netdev *, struct in_addr *,
131 int cmd, const char *cmd_name);
132 static int get_flags(const struct netdev *, int *flagsp);
133 static int set_flags(struct netdev *, int flags);
134 static int do_get_ifindex(const char *netdev_name);
135 static int get_ifindex(const struct netdev *, int *ifindexp);
136 static int do_set_addr(struct netdev *netdev,
137 int ioctl_nr, const char *ioctl_name,
138 struct in_addr addr);
139 static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]);
140 static int set_etheraddr(const char *netdev_name, int hwaddr_family,
141 const uint8_t[ETH_ADDR_LEN]);
142 static int get_stats_via_netlink(int ifindex, struct netdev_stats *stats);
143 static int get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats);
145 static struct netdev_linux *
146 netdev_linux_cast(const struct netdev *netdev)
148 netdev_assert_class(netdev, &netdev_linux_class);
149 return CONTAINER_OF(netdev, struct netdev_linux, netdev);
153 netdev_linux_init(void)
155 static int status = -1;
157 af_inet_sock = socket(AF_INET, SOCK_DGRAM, 0);
158 status = af_inet_sock >= 0 ? 0 : errno;
160 VLOG_ERR("failed to create inet socket: %s", strerror(status));
167 netdev_linux_run(void)
169 rtnetlink_notifier_run();
173 netdev_linux_wait(void)
175 rtnetlink_notifier_wait();
179 netdev_linux_cache_cb(const struct rtnetlink_change *change,
182 struct netdev_linux_cache *cache;
184 cache = shash_find_data(&cache_map, change->ifname);
189 struct shash_node *node;
190 SHASH_FOR_EACH (node, &cache_map) {
198 netdev_linux_open(const char *name, char *suffix, int ethertype,
199 struct netdev **netdevp)
201 struct netdev_linux *netdev;
202 enum netdev_flags flags;
205 /* Allocate network device. */
206 netdev = xcalloc(1, sizeof *netdev);
207 netdev_init(&netdev->netdev, suffix, &netdev_linux_class);
208 netdev->netdev_fd = -1;
210 netdev->cache = shash_find_data(&cache_map, suffix);
211 if (!netdev->cache) {
212 if (shash_is_empty(&cache_map)) {
213 int error = rtnetlink_notifier_register(
214 &netdev_linux_cache_notifier, netdev_linux_cache_cb, NULL);
216 netdev_close(&netdev->netdev);
220 netdev->cache = xmalloc(sizeof *netdev->cache);
221 netdev->cache->shash_node = shash_add(&cache_map, suffix,
223 netdev->cache->valid = 0;
224 netdev->cache->ref_cnt = 0;
226 netdev->cache->ref_cnt++;
228 if (!strncmp(name, "tap:", 4)) {
229 static const char tap_dev[] = "/dev/net/tun";
232 /* Open tap device. */
233 netdev->tap_fd = open(tap_dev, O_RDWR);
234 if (netdev->tap_fd < 0) {
236 VLOG_WARN("opening \"%s\" failed: %s", tap_dev, strerror(error));
240 /* Create tap device. */
241 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
242 strncpy(ifr.ifr_name, suffix, sizeof ifr.ifr_name);
243 if (ioctl(netdev->tap_fd, TUNSETIFF, &ifr) == -1) {
244 VLOG_WARN("%s: creating tap device failed: %s", suffix,
250 /* Make non-blocking. */
251 error = set_nonblocking(netdev->tap_fd);
257 error = netdev_get_flags(&netdev->netdev, &flags);
258 if (error == ENODEV) {
262 if (netdev->tap_fd >= 0 || ethertype != NETDEV_ETH_TYPE_NONE) {
263 struct sockaddr_ll sll;
267 /* Create file descriptor. */
268 protocol = (ethertype == NETDEV_ETH_TYPE_ANY ? ETH_P_ALL
269 : ethertype == NETDEV_ETH_TYPE_802_2 ? ETH_P_802_2
271 netdev->netdev_fd = socket(PF_PACKET, SOCK_RAW, htons(protocol));
272 if (netdev->netdev_fd < 0) {
276 if (netdev->tap_fd < 0) {
277 netdev->tap_fd = netdev->netdev_fd;
280 /* Set non-blocking mode. */
281 error = set_nonblocking(netdev->netdev_fd);
286 /* Get ethernet device index. */
287 error = get_ifindex(&netdev->netdev, &ifindex);
292 /* Bind to specific ethernet device. */
293 memset(&sll, 0, sizeof sll);
294 sll.sll_family = AF_PACKET;
295 sll.sll_ifindex = ifindex;
296 if (bind(netdev->netdev_fd,
297 (struct sockaddr *) &sll, sizeof sll) < 0) {
299 VLOG_ERR("bind to %s failed: %s", suffix, strerror(error));
303 /* Between the socket() and bind() calls above, the socket receives all
304 * packets of the requested type on all system interfaces. We do not
305 * want to receive that data, but there is no way to avoid it. So we
306 * must now drain out the receive queue. */
307 error = drain_rcvbuf(netdev->netdev_fd);
313 *netdevp = &netdev->netdev;
317 netdev_close(&netdev->netdev);
321 /* Closes and destroys 'netdev'. */
323 netdev_linux_close(struct netdev *netdev_)
325 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
327 if (netdev->cache && !--netdev->cache->ref_cnt) {
328 shash_delete(&cache_map, netdev->cache->shash_node);
331 if (shash_is_empty(&cache_map)) {
332 rtnetlink_notifier_unregister(&netdev_linux_cache_notifier);
335 if (netdev->netdev_fd >= 0) {
336 close(netdev->netdev_fd);
338 if (netdev->tap_fd >= 0 && netdev->netdev_fd != netdev->tap_fd) {
339 close(netdev->tap_fd);
344 /* Initializes 'svec' with a list of the names of all known network devices. */
346 netdev_linux_enumerate(struct svec *svec)
348 struct if_nameindex *names;
350 names = if_nameindex();
354 for (i = 0; names[i].if_name != NULL; i++) {
355 svec_add(svec, names[i].if_name);
357 if_freenameindex(names);
360 VLOG_WARN("could not obtain list of network device names: %s",
367 netdev_linux_recv(struct netdev *netdev_, void *data, size_t size)
369 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
371 if (netdev->tap_fd < 0) {
372 /* Device was opened with NETDEV_ETH_TYPE_NONE. */
377 ssize_t retval = read(netdev->tap_fd, data, size);
380 } else if (errno != EINTR) {
381 if (errno != EAGAIN) {
382 VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
383 strerror(errno), netdev_get_name(netdev_));
390 /* Registers with the poll loop to wake up from the next call to poll_block()
391 * when a packet is ready to be received with netdev_recv() on 'netdev'. */
393 netdev_linux_recv_wait(struct netdev *netdev_)
395 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
396 if (netdev->tap_fd >= 0) {
397 poll_fd_wait(netdev->tap_fd, POLLIN);
401 /* Discards all packets waiting to be received from 'netdev'. */
403 netdev_linux_drain(struct netdev *netdev_)
405 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
406 if (netdev->tap_fd < 0 && netdev->netdev_fd < 0) {
408 } else if (netdev->tap_fd != netdev->netdev_fd) {
410 int error = netdev_linux_do_ioctl(netdev_, &ifr,
411 SIOCGIFTXQLEN, "SIOCGIFTXQLEN");
415 drain_fd(netdev->tap_fd, ifr.ifr_qlen);
418 return drain_rcvbuf(netdev->netdev_fd);
422 /* Sends 'buffer' on 'netdev'. Returns 0 if successful, otherwise a positive
423 * errno value. Returns EAGAIN without blocking if the packet cannot be queued
424 * immediately. Returns EMSGSIZE if a partial packet was transmitted or if
425 * the packet is too big or too small to transmit on the device.
427 * The caller retains ownership of 'buffer' in all cases.
429 * The kernel maintains a packet transmission queue, so the caller is not
430 * expected to do additional queuing of packets. */
432 netdev_linux_send(struct netdev *netdev_, const void *data, size_t size)
434 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
436 /* XXX should support sending even if 'ethertype' was NETDEV_ETH_TYPE_NONE.
438 if (netdev->tap_fd < 0) {
443 ssize_t retval = write(netdev->tap_fd, data, size);
445 /* The Linux AF_PACKET implementation never blocks waiting for room
446 * for packets, instead returning ENOBUFS. Translate this into
447 * EAGAIN for the caller. */
448 if (errno == ENOBUFS) {
450 } else if (errno == EINTR) {
452 } else if (errno != EAGAIN) {
453 VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s",
454 netdev_get_name(netdev_), strerror(errno));
457 } else if (retval != size) {
458 VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of "
459 "%zu) on %s", retval, size, netdev_get_name(netdev_));
467 /* Registers with the poll loop to wake up from the next call to poll_block()
468 * when the packet transmission queue has sufficient room to transmit a packet
469 * with netdev_send().
471 * The kernel maintains a packet transmission queue, so the client is not
472 * expected to do additional queuing of packets. Thus, this function is
473 * unlikely to ever be used. It is included for completeness. */
475 netdev_linux_send_wait(struct netdev *netdev_)
477 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
478 if (netdev->tap_fd < 0 && netdev->netdev_fd < 0) {
480 } else if (netdev->tap_fd == netdev->netdev_fd) {
481 poll_fd_wait(netdev->tap_fd, POLLOUT);
483 /* TAP device always accepts packets.*/
484 poll_immediate_wake();
488 /* Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
489 * otherwise a positive errno value. */
491 netdev_linux_set_etheraddr(struct netdev *netdev_,
492 const uint8_t mac[ETH_ADDR_LEN])
494 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
497 if (!(netdev->cache->valid & VALID_ETHERADDR)
498 || !eth_addr_equals(netdev->cache->etheraddr, mac)) {
499 error = set_etheraddr(netdev_get_name(netdev_), ARPHRD_ETHER, mac);
501 netdev->cache->valid |= VALID_ETHERADDR;
502 memcpy(netdev->cache->etheraddr, mac, ETH_ADDR_LEN);
510 /* Returns a pointer to 'netdev''s MAC address. The caller must not modify or
511 * free the returned buffer. */
513 netdev_linux_get_etheraddr(const struct netdev *netdev_,
514 uint8_t mac[ETH_ADDR_LEN])
516 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
517 if (!(netdev->cache->valid & VALID_ETHERADDR)) {
518 int error = get_etheraddr(netdev_get_name(netdev_),
519 netdev->cache->etheraddr);
523 netdev->cache->valid |= VALID_ETHERADDR;
525 memcpy(mac, netdev->cache->etheraddr, ETH_ADDR_LEN);
529 /* Returns the maximum size of transmitted (and received) packets on 'netdev',
530 * in bytes, not including the hardware header; thus, this is typically 1500
531 * bytes for Ethernet devices. */
533 netdev_linux_get_mtu(const struct netdev *netdev_, int *mtup)
535 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
536 if (!(netdev->cache->valid & VALID_MTU)) {
540 error = netdev_linux_do_ioctl(netdev_, &ifr, SIOCGIFMTU, "SIOCGIFMTU");
544 netdev->cache->mtu = ifr.ifr_mtu;
545 netdev->cache->valid |= VALID_MTU;
547 *mtup = netdev->cache->mtu;
551 /* Returns the ifindex of 'netdev', if successful, as a positive number.
552 * On failure, returns a negative errno value. */
554 netdev_linux_get_ifindex(const struct netdev *netdev)
558 error = get_ifindex(netdev, &ifindex);
559 return error ? -error : ifindex;
563 netdev_linux_get_carrier(const struct netdev *netdev_, bool *carrier)
565 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
570 if (!(netdev->cache->valid & VALID_CARRIER)) {
574 fn = xasprintf("/sys/class/net/%s/carrier", netdev_get_name(netdev_));
575 fd = open(fn, O_RDONLY);
578 VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(error));
582 retval = read(fd, line, sizeof line);
585 if (error == EINVAL) {
586 /* This is the normal return value when we try to check carrier
587 * if the network device is not up. */
589 VLOG_WARN_RL(&rl, "%s: read failed: %s", fn, strerror(error));
592 } else if (retval == 0) {
594 VLOG_WARN_RL(&rl, "%s: unexpected end of file", fn);
598 if (line[0] != '0' && line[0] != '1') {
600 VLOG_WARN_RL(&rl, "%s: value is %c (expected 0 or 1)",
604 netdev->cache->carrier = line[0] != '0';
605 netdev->cache->valid |= VALID_CARRIER;
607 *carrier = netdev->cache->carrier;
618 /* Check whether we can we use RTM_GETLINK to get network device statistics.
619 * In pre-2.6.19 kernels, this was only available if wireless extensions were
622 check_for_working_netlink_stats(void)
624 /* Decide on the netdev_get_stats() implementation to use. Netlink is
625 * preferable, so if that works, we'll use it. */
626 int ifindex = do_get_ifindex("lo");
628 VLOG_WARN("failed to get ifindex for lo, "
629 "obtaining netdev stats from proc");
632 struct netdev_stats stats;
633 int error = get_stats_via_netlink(ifindex, &stats);
635 VLOG_DBG("obtaining netdev stats via rtnetlink");
638 VLOG_INFO("RTM_GETLINK failed (%s), obtaining netdev stats "
639 "via proc (you are probably running a pre-2.6.19 "
640 "kernel)", strerror(error));
646 /* Retrieves current device stats for 'netdev'.
648 * XXX All of the members of struct netdev_stats are 64 bits wide, but on
649 * 32-bit architectures the Linux network stats are only 32 bits. */
651 netdev_linux_get_stats(const struct netdev *netdev_, struct netdev_stats *stats)
653 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
654 static int use_netlink_stats = -1;
656 struct netdev_stats raw_stats;
657 struct netdev_stats *collect_stats = stats;
659 COVERAGE_INC(netdev_get_stats);
661 if (!(netdev->cache->valid & VALID_IS_INTERNAL)) {
662 netdev->cache->is_internal = (netdev->tap_fd != -1);
664 if (!netdev->cache->is_internal) {
665 struct ethtool_drvinfo drvinfo;
667 memset(&drvinfo, 0, sizeof drvinfo);
668 error = netdev_linux_do_ethtool(&netdev->netdev,
669 (struct ethtool_cmd *)&drvinfo,
674 netdev->cache->is_internal = !strcmp(drvinfo.driver,
679 netdev->cache->valid |= VALID_IS_INTERNAL;
682 if (netdev->cache->is_internal) {
683 collect_stats = &raw_stats;
686 if (use_netlink_stats < 0) {
687 use_netlink_stats = check_for_working_netlink_stats();
689 if (use_netlink_stats) {
692 error = get_ifindex(&netdev->netdev, &ifindex);
694 error = get_stats_via_netlink(ifindex, collect_stats);
697 error = get_stats_via_proc(netdev->netdev.name, collect_stats);
700 /* If this port is an internal port then the transmit and receive stats
701 * will appear to be swapped relative to the other ports since we are the
702 * one sending the data, not a remote computer. For consistency, we swap
704 if (netdev->cache->is_internal) {
705 stats->rx_packets = raw_stats.tx_packets;
706 stats->tx_packets = raw_stats.rx_packets;
707 stats->rx_bytes = raw_stats.tx_bytes;
708 stats->tx_bytes = raw_stats.rx_bytes;
709 stats->rx_errors = raw_stats.tx_errors;
710 stats->tx_errors = raw_stats.rx_errors;
711 stats->rx_dropped = raw_stats.tx_dropped;
712 stats->tx_dropped = raw_stats.rx_dropped;
713 stats->multicast = raw_stats.multicast;
714 stats->collisions = raw_stats.collisions;
715 stats->rx_length_errors = 0;
716 stats->rx_over_errors = 0;
717 stats->rx_crc_errors = 0;
718 stats->rx_frame_errors = 0;
719 stats->rx_fifo_errors = 0;
720 stats->rx_missed_errors = 0;
721 stats->tx_aborted_errors = 0;
722 stats->tx_carrier_errors = 0;
723 stats->tx_fifo_errors = 0;
724 stats->tx_heartbeat_errors = 0;
725 stats->tx_window_errors = 0;
731 /* Stores the features supported by 'netdev' into each of '*current',
732 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
733 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
734 * successful, otherwise a positive errno value. */
736 netdev_linux_get_features(struct netdev *netdev,
737 uint32_t *current, uint32_t *advertised,
738 uint32_t *supported, uint32_t *peer)
740 struct ethtool_cmd ecmd;
743 memset(&ecmd, 0, sizeof ecmd);
744 error = netdev_linux_do_ethtool(netdev, &ecmd,
745 ETHTOOL_GSET, "ETHTOOL_GSET");
750 /* Supported features. */
752 if (ecmd.supported & SUPPORTED_10baseT_Half) {
753 *supported |= OFPPF_10MB_HD;
755 if (ecmd.supported & SUPPORTED_10baseT_Full) {
756 *supported |= OFPPF_10MB_FD;
758 if (ecmd.supported & SUPPORTED_100baseT_Half) {
759 *supported |= OFPPF_100MB_HD;
761 if (ecmd.supported & SUPPORTED_100baseT_Full) {
762 *supported |= OFPPF_100MB_FD;
764 if (ecmd.supported & SUPPORTED_1000baseT_Half) {
765 *supported |= OFPPF_1GB_HD;
767 if (ecmd.supported & SUPPORTED_1000baseT_Full) {
768 *supported |= OFPPF_1GB_FD;
770 if (ecmd.supported & SUPPORTED_10000baseT_Full) {
771 *supported |= OFPPF_10GB_FD;
773 if (ecmd.supported & SUPPORTED_TP) {
774 *supported |= OFPPF_COPPER;
776 if (ecmd.supported & SUPPORTED_FIBRE) {
777 *supported |= OFPPF_FIBER;
779 if (ecmd.supported & SUPPORTED_Autoneg) {
780 *supported |= OFPPF_AUTONEG;
782 if (ecmd.supported & SUPPORTED_Pause) {
783 *supported |= OFPPF_PAUSE;
785 if (ecmd.supported & SUPPORTED_Asym_Pause) {
786 *supported |= OFPPF_PAUSE_ASYM;
789 /* Advertised features. */
791 if (ecmd.advertising & ADVERTISED_10baseT_Half) {
792 *advertised |= OFPPF_10MB_HD;
794 if (ecmd.advertising & ADVERTISED_10baseT_Full) {
795 *advertised |= OFPPF_10MB_FD;
797 if (ecmd.advertising & ADVERTISED_100baseT_Half) {
798 *advertised |= OFPPF_100MB_HD;
800 if (ecmd.advertising & ADVERTISED_100baseT_Full) {
801 *advertised |= OFPPF_100MB_FD;
803 if (ecmd.advertising & ADVERTISED_1000baseT_Half) {
804 *advertised |= OFPPF_1GB_HD;
806 if (ecmd.advertising & ADVERTISED_1000baseT_Full) {
807 *advertised |= OFPPF_1GB_FD;
809 if (ecmd.advertising & ADVERTISED_10000baseT_Full) {
810 *advertised |= OFPPF_10GB_FD;
812 if (ecmd.advertising & ADVERTISED_TP) {
813 *advertised |= OFPPF_COPPER;
815 if (ecmd.advertising & ADVERTISED_FIBRE) {
816 *advertised |= OFPPF_FIBER;
818 if (ecmd.advertising & ADVERTISED_Autoneg) {
819 *advertised |= OFPPF_AUTONEG;
821 if (ecmd.advertising & ADVERTISED_Pause) {
822 *advertised |= OFPPF_PAUSE;
824 if (ecmd.advertising & ADVERTISED_Asym_Pause) {
825 *advertised |= OFPPF_PAUSE_ASYM;
828 /* Current settings. */
829 if (ecmd.speed == SPEED_10) {
830 *current = ecmd.duplex ? OFPPF_10MB_FD : OFPPF_10MB_HD;
831 } else if (ecmd.speed == SPEED_100) {
832 *current = ecmd.duplex ? OFPPF_100MB_FD : OFPPF_100MB_HD;
833 } else if (ecmd.speed == SPEED_1000) {
834 *current = ecmd.duplex ? OFPPF_1GB_FD : OFPPF_1GB_HD;
835 } else if (ecmd.speed == SPEED_10000) {
836 *current = OFPPF_10GB_FD;
841 if (ecmd.port == PORT_TP) {
842 *current |= OFPPF_COPPER;
843 } else if (ecmd.port == PORT_FIBRE) {
844 *current |= OFPPF_FIBER;
848 *current |= OFPPF_AUTONEG;
851 /* Peer advertisements. */
857 /* Set the features advertised by 'netdev' to 'advertise'. */
859 netdev_linux_set_advertisements(struct netdev *netdev, uint32_t advertise)
861 struct ethtool_cmd ecmd;
864 memset(&ecmd, 0, sizeof ecmd);
865 error = netdev_linux_do_ethtool(netdev, &ecmd,
866 ETHTOOL_GSET, "ETHTOOL_GSET");
871 ecmd.advertising = 0;
872 if (advertise & OFPPF_10MB_HD) {
873 ecmd.advertising |= ADVERTISED_10baseT_Half;
875 if (advertise & OFPPF_10MB_FD) {
876 ecmd.advertising |= ADVERTISED_10baseT_Full;
878 if (advertise & OFPPF_100MB_HD) {
879 ecmd.advertising |= ADVERTISED_100baseT_Half;
881 if (advertise & OFPPF_100MB_FD) {
882 ecmd.advertising |= ADVERTISED_100baseT_Full;
884 if (advertise & OFPPF_1GB_HD) {
885 ecmd.advertising |= ADVERTISED_1000baseT_Half;
887 if (advertise & OFPPF_1GB_FD) {
888 ecmd.advertising |= ADVERTISED_1000baseT_Full;
890 if (advertise & OFPPF_10GB_FD) {
891 ecmd.advertising |= ADVERTISED_10000baseT_Full;
893 if (advertise & OFPPF_COPPER) {
894 ecmd.advertising |= ADVERTISED_TP;
896 if (advertise & OFPPF_FIBER) {
897 ecmd.advertising |= ADVERTISED_FIBRE;
899 if (advertise & OFPPF_AUTONEG) {
900 ecmd.advertising |= ADVERTISED_Autoneg;
902 if (advertise & OFPPF_PAUSE) {
903 ecmd.advertising |= ADVERTISED_Pause;
905 if (advertise & OFPPF_PAUSE_ASYM) {
906 ecmd.advertising |= ADVERTISED_Asym_Pause;
908 return netdev_linux_do_ethtool(netdev, &ecmd,
909 ETHTOOL_SSET, "ETHTOOL_SSET");
912 /* If 'netdev_name' is the name of a VLAN network device (e.g. one created with
913 * vconfig(8)), sets '*vlan_vid' to the VLAN VID associated with that device
914 * and returns 0. Otherwise returns a errno value (specifically ENOENT if
915 * 'netdev_name' is the name of a network device that is not a VLAN device) and
916 * sets '*vlan_vid' to -1. */
918 netdev_linux_get_vlan_vid(const struct netdev *netdev, int *vlan_vid)
920 const char *netdev_name = netdev_get_name(netdev);
921 struct ds line = DS_EMPTY_INITIALIZER;
926 COVERAGE_INC(netdev_get_vlan_vid);
927 fn = xasprintf("/proc/net/vlan/%s", netdev_name);
928 stream = fopen(fn, "r");
934 if (ds_get_line(&line, stream)) {
935 if (ferror(stream)) {
937 VLOG_ERR_RL(&rl, "error reading \"%s\": %s", fn, strerror(errno));
940 VLOG_ERR_RL(&rl, "unexpected end of file reading \"%s\"", fn);
945 if (!sscanf(ds_cstr(&line), "%*s VID: %d", vlan_vid)) {
947 VLOG_ERR_RL(&rl, "parse error reading \"%s\" line 1: \"%s\"",
966 #define POLICE_ADD_CMD "/sbin/tc qdisc add dev %s handle ffff: ingress"
967 #define POLICE_CONFIG_CMD "/sbin/tc filter add dev %s parent ffff: protocol ip prio 50 u32 match ip src 0.0.0.0/0 police rate %dkbit burst %dk mtu 65535 drop flowid :1"
968 /* We redirect stderr to /dev/null because we often want to remove all
969 * traffic control configuration on a port so its in a known state. If
970 * this done when there is no such configuration, tc complains, so we just
973 #define POLICE_DEL_CMD "/sbin/tc qdisc del dev %s handle ffff: ingress 2>/dev/null"
975 /* Attempts to set input rate limiting (policing) policy. */
977 netdev_linux_set_policing(struct netdev *netdev,
978 uint32_t kbits_rate, uint32_t kbits_burst)
980 const char *netdev_name = netdev_get_name(netdev);
983 COVERAGE_INC(netdev_set_policing);
986 /* Default to 10 kilobits if not specified. */
990 /* xxx This should be more careful about only adding if it
991 * xxx actually exists, as opposed to always deleting it. */
992 snprintf(command, sizeof(command), POLICE_DEL_CMD, netdev_name);
993 if (system(command) == -1) {
994 VLOG_WARN_RL(&rl, "%s: problem removing policing", netdev_name);
997 snprintf(command, sizeof(command), POLICE_ADD_CMD, netdev_name);
998 if (system(command) != 0) {
999 VLOG_WARN_RL(&rl, "%s: problem adding policing", netdev_name);
1003 snprintf(command, sizeof(command), POLICE_CONFIG_CMD, netdev_name,
1004 kbits_rate, kbits_burst);
1005 if (system(command) != 0) {
1006 VLOG_WARN_RL(&rl, "%s: problem configuring policing",
1011 snprintf(command, sizeof(command), POLICE_DEL_CMD, netdev_name);
1012 if (system(command) == -1) {
1013 VLOG_WARN_RL(&rl, "%s: problem removing policing", netdev_name);
1021 netdev_linux_get_in4(const struct netdev *netdev_,
1022 struct in_addr *address, struct in_addr *netmask)
1024 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
1025 if (!(netdev->cache->valid & VALID_IN4)) {
1028 error = netdev_linux_get_ipv4(netdev_, &netdev->cache->address,
1029 SIOCGIFADDR, "SIOCGIFADDR");
1034 error = netdev_linux_get_ipv4(netdev_, &netdev->cache->netmask,
1035 SIOCGIFNETMASK, "SIOCGIFNETMASK");
1040 netdev->cache->valid |= VALID_IN4;
1042 *address = netdev->cache->address;
1043 *netmask = netdev->cache->netmask;
1044 return address->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0;
1048 netdev_linux_set_in4(struct netdev *netdev_, struct in_addr address,
1049 struct in_addr netmask)
1051 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
1054 error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", address);
1056 netdev->cache->valid |= VALID_IN4;
1057 netdev->cache->address = address;
1058 netdev->cache->netmask = netmask;
1059 if (address.s_addr != INADDR_ANY) {
1060 error = do_set_addr(netdev_, SIOCSIFNETMASK,
1061 "SIOCSIFNETMASK", netmask);
1068 parse_if_inet6_line(const char *line,
1069 struct in6_addr *in6, char ifname[16 + 1])
1071 uint8_t *s6 = in6->s6_addr;
1072 #define X8 "%2"SCNx8
1074 " "X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8
1075 "%*x %*x %*x %*x %16s\n",
1076 &s6[0], &s6[1], &s6[2], &s6[3],
1077 &s6[4], &s6[5], &s6[6], &s6[7],
1078 &s6[8], &s6[9], &s6[10], &s6[11],
1079 &s6[12], &s6[13], &s6[14], &s6[15],
1083 /* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address (if
1084 * 'in6' is non-null) and returns true. Otherwise, returns false. */
1086 netdev_linux_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
1088 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
1089 if (!(netdev->cache->valid & VALID_IN6)) {
1093 netdev->cache->in6 = in6addr_any;
1095 file = fopen("/proc/net/if_inet6", "r");
1097 const char *name = netdev_get_name(netdev_);
1098 while (fgets(line, sizeof line, file)) {
1099 struct in6_addr in6;
1100 char ifname[16 + 1];
1101 if (parse_if_inet6_line(line, &in6, ifname)
1102 && !strcmp(name, ifname))
1104 netdev->cache->in6 = in6;
1110 netdev->cache->valid |= VALID_IN6;
1112 *in6 = netdev->cache->in6;
1117 make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
1119 struct sockaddr_in sin;
1120 memset(&sin, 0, sizeof sin);
1121 sin.sin_family = AF_INET;
1122 sin.sin_addr = addr;
1125 memset(sa, 0, sizeof *sa);
1126 memcpy(sa, &sin, sizeof sin);
1130 do_set_addr(struct netdev *netdev,
1131 int ioctl_nr, const char *ioctl_name, struct in_addr addr)
1134 strncpy(ifr.ifr_name, netdev->name, sizeof ifr.ifr_name);
1135 make_in4_sockaddr(&ifr.ifr_addr, addr);
1136 return netdev_linux_do_ioctl(netdev, &ifr, ioctl_nr, ioctl_name);
1139 /* Adds 'router' as a default IP gateway. */
1141 netdev_linux_add_router(struct netdev *netdev UNUSED, struct in_addr router)
1143 struct in_addr any = { INADDR_ANY };
1147 memset(&rt, 0, sizeof rt);
1148 make_in4_sockaddr(&rt.rt_dst, any);
1149 make_in4_sockaddr(&rt.rt_gateway, router);
1150 make_in4_sockaddr(&rt.rt_genmask, any);
1151 rt.rt_flags = RTF_UP | RTF_GATEWAY;
1152 COVERAGE_INC(netdev_add_router);
1153 error = ioctl(af_inet_sock, SIOCADDRT, &rt) < 0 ? errno : 0;
1155 VLOG_WARN("ioctl(SIOCADDRT): %s", strerror(error));
1161 netdev_linux_get_next_hop(const struct in_addr *host, struct in_addr *next_hop,
1164 static const char fn[] = "/proc/net/route";
1169 *netdev_name = NULL;
1170 stream = fopen(fn, "r");
1171 if (stream == NULL) {
1172 VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(errno));
1177 while (fgets(line, sizeof line, stream)) {
1180 uint32_t dest, gateway, mask;
1181 int refcnt, metric, mtu;
1182 unsigned int flags, use, window, irtt;
1185 "%16s %"SCNx32" %"SCNx32" %04X %d %u %d %"SCNx32
1187 iface, &dest, &gateway, &flags, &refcnt,
1188 &use, &metric, &mask, &mtu, &window, &irtt) != 11) {
1190 VLOG_WARN_RL(&rl, "%s: could not parse line %d: %s",
1194 if (!(flags & RTF_UP)) {
1195 /* Skip routes that aren't up. */
1199 /* The output of 'dest', 'mask', and 'gateway' were given in
1200 * network byte order, so we don't need need any endian
1201 * conversions here. */
1202 if ((dest & mask) == (host->s_addr & mask)) {
1204 /* The host is directly reachable. */
1205 next_hop->s_addr = 0;
1207 /* To reach the host, we must go through a gateway. */
1208 next_hop->s_addr = gateway;
1210 *netdev_name = xstrdup(iface);
1221 /* Looks up the ARP table entry for 'ip' on 'netdev'. If one exists and can be
1222 * successfully retrieved, it stores the corresponding MAC address in 'mac' and
1223 * returns 0. Otherwise, it returns a positive errno value; in particular,
1224 * ENXIO indicates that there is not ARP table entry for 'ip' on 'netdev'. */
1226 netdev_linux_arp_lookup(const struct netdev *netdev,
1227 uint32_t ip, uint8_t mac[ETH_ADDR_LEN])
1230 struct sockaddr_in *pa;
1233 memset(&r, 0, sizeof r);
1234 pa = (struct sockaddr_in *) &r.arp_pa;
1235 pa->sin_family = AF_INET;
1236 pa->sin_addr.s_addr = ip;
1238 r.arp_ha.sa_family = ARPHRD_ETHER;
1240 strncpy(r.arp_dev, netdev->name, sizeof r.arp_dev);
1241 COVERAGE_INC(netdev_arp_lookup);
1242 retval = ioctl(af_inet_sock, SIOCGARP, &r) < 0 ? errno : 0;
1244 memcpy(mac, r.arp_ha.sa_data, ETH_ADDR_LEN);
1245 } else if (retval != ENXIO) {
1246 VLOG_WARN_RL(&rl, "%s: could not look up ARP entry for "IP_FMT": %s",
1247 netdev->name, IP_ARGS(&ip), strerror(retval));
1253 nd_to_iff_flags(enum netdev_flags nd)
1256 if (nd & NETDEV_UP) {
1259 if (nd & NETDEV_PROMISC) {
1266 iff_to_nd_flags(int iff)
1268 enum netdev_flags nd = 0;
1272 if (iff & IFF_PROMISC) {
1273 nd |= NETDEV_PROMISC;
1279 netdev_linux_update_flags(struct netdev *netdev, enum netdev_flags off,
1280 enum netdev_flags on, enum netdev_flags *old_flagsp)
1282 int old_flags, new_flags;
1285 error = get_flags(netdev, &old_flags);
1287 *old_flagsp = iff_to_nd_flags(old_flags);
1288 new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on);
1289 if (new_flags != old_flags) {
1290 error = set_flags(netdev, new_flags);
1297 poll_notify(struct list *list)
1299 struct netdev_linux_notifier *notifier;
1300 LIST_FOR_EACH (notifier, struct netdev_linux_notifier, node, list) {
1301 struct netdev_notifier *n = ¬ifier->notifier;
1307 netdev_linux_poll_cb(const struct rtnetlink_change *change,
1311 struct list *list = shash_find_data(&netdev_linux_notifiers,
1317 struct shash_node *node;
1318 SHASH_FOR_EACH (node, &netdev_linux_notifiers) {
1319 poll_notify(node->data);
1325 netdev_linux_poll_add(struct netdev *netdev,
1326 void (*cb)(struct netdev_notifier *), void *aux,
1327 struct netdev_notifier **notifierp)
1329 const char *netdev_name = netdev_get_name(netdev);
1330 struct netdev_linux_notifier *notifier;
1333 if (shash_is_empty(&netdev_linux_notifiers)) {
1334 int error = rtnetlink_notifier_register(&netdev_linux_poll_notifier,
1335 netdev_linux_poll_cb, NULL);
1341 list = shash_find_data(&netdev_linux_notifiers, netdev_name);
1343 list = xmalloc(sizeof *list);
1345 shash_add(&netdev_linux_notifiers, netdev_name, list);
1348 notifier = xmalloc(sizeof *notifier);
1349 netdev_notifier_init(¬ifier->notifier, netdev, cb, aux);
1350 list_push_back(list, ¬ifier->node);
1351 *notifierp = ¬ifier->notifier;
1356 netdev_linux_poll_remove(struct netdev_notifier *notifier_)
1358 struct netdev_linux_notifier *notifier =
1359 CONTAINER_OF(notifier_, struct netdev_linux_notifier, notifier);
1362 /* Remove 'notifier' from its list. */
1363 list = list_remove(¬ifier->node);
1364 if (list_is_empty(list)) {
1365 /* The list is now empty. Remove it from the hash and free it. */
1366 const char *netdev_name = netdev_get_name(notifier->notifier.netdev);
1367 shash_delete(&netdev_linux_notifiers,
1368 shash_find(&netdev_linux_notifiers, netdev_name));
1373 /* If that was the last notifier, unregister. */
1374 if (shash_is_empty(&netdev_linux_notifiers)) {
1375 rtnetlink_notifier_unregister(&netdev_linux_poll_notifier);
1379 const struct netdev_class netdev_linux_class = {
1390 netdev_linux_enumerate,
1393 netdev_linux_recv_wait,
1397 netdev_linux_send_wait,
1399 netdev_linux_set_etheraddr,
1400 netdev_linux_get_etheraddr,
1401 netdev_linux_get_mtu,
1402 netdev_linux_get_ifindex,
1403 netdev_linux_get_carrier,
1404 netdev_linux_get_stats,
1406 netdev_linux_get_features,
1407 netdev_linux_set_advertisements,
1408 netdev_linux_get_vlan_vid,
1409 netdev_linux_set_policing,
1411 netdev_linux_get_in4,
1412 netdev_linux_set_in4,
1413 netdev_linux_get_in6,
1414 netdev_linux_add_router,
1415 netdev_linux_get_next_hop,
1416 netdev_linux_arp_lookup,
1418 netdev_linux_update_flags,
1420 netdev_linux_poll_add,
1421 netdev_linux_poll_remove,
1424 const struct netdev_class netdev_tap_class = {
1435 netdev_linux_enumerate,
1438 netdev_linux_recv_wait,
1442 netdev_linux_send_wait,
1444 netdev_linux_set_etheraddr,
1445 netdev_linux_get_etheraddr,
1446 netdev_linux_get_mtu,
1447 netdev_linux_get_ifindex,
1448 netdev_linux_get_carrier,
1449 netdev_linux_get_stats,
1451 netdev_linux_get_features,
1452 netdev_linux_set_advertisements,
1453 netdev_linux_get_vlan_vid,
1454 netdev_linux_set_policing,
1456 netdev_linux_get_in4,
1457 netdev_linux_set_in4,
1458 netdev_linux_get_in6,
1459 netdev_linux_add_router,
1460 netdev_linux_get_next_hop,
1461 netdev_linux_arp_lookup,
1463 netdev_linux_update_flags,
1465 netdev_linux_poll_add,
1466 netdev_linux_poll_remove,
1470 get_stats_via_netlink(int ifindex, struct netdev_stats *stats)
1472 /* Policy for RTNLGRP_LINK messages.
1474 * There are *many* more fields in these messages, but currently we only
1475 * care about these fields. */
1476 static const struct nl_policy rtnlgrp_link_policy[] = {
1477 [IFLA_IFNAME] = { .type = NL_A_STRING, .optional = false },
1478 [IFLA_STATS] = { .type = NL_A_UNSPEC, .optional = true,
1479 .min_len = sizeof(struct rtnl_link_stats) },
1483 static struct nl_sock *rtnl_sock;
1484 struct ofpbuf request;
1485 struct ofpbuf *reply;
1486 struct ifinfomsg *ifi;
1487 const struct rtnl_link_stats *rtnl_stats;
1488 struct nlattr *attrs[ARRAY_SIZE(rtnlgrp_link_policy)];
1492 error = nl_sock_create(NETLINK_ROUTE, 0, 0, 0, &rtnl_sock);
1494 VLOG_ERR_RL(&rl, "failed to create rtnetlink socket: %s",
1500 ofpbuf_init(&request, 0);
1501 nl_msg_put_nlmsghdr(&request, rtnl_sock, sizeof *ifi,
1502 RTM_GETLINK, NLM_F_REQUEST);
1503 ifi = ofpbuf_put_zeros(&request, sizeof *ifi);
1504 ifi->ifi_family = PF_UNSPEC;
1505 ifi->ifi_index = ifindex;
1506 error = nl_sock_transact(rtnl_sock, &request, &reply);
1507 ofpbuf_uninit(&request);
1512 if (!nl_policy_parse(reply, NLMSG_HDRLEN + sizeof(struct ifinfomsg),
1513 rtnlgrp_link_policy,
1514 attrs, ARRAY_SIZE(rtnlgrp_link_policy))) {
1515 ofpbuf_delete(reply);
1519 if (!attrs[IFLA_STATS]) {
1520 VLOG_WARN_RL(&rl, "RTM_GETLINK reply lacks stats");
1521 ofpbuf_delete(reply);
1525 rtnl_stats = nl_attr_get(attrs[IFLA_STATS]);
1526 stats->rx_packets = rtnl_stats->rx_packets;
1527 stats->tx_packets = rtnl_stats->tx_packets;
1528 stats->rx_bytes = rtnl_stats->rx_bytes;
1529 stats->tx_bytes = rtnl_stats->tx_bytes;
1530 stats->rx_errors = rtnl_stats->rx_errors;
1531 stats->tx_errors = rtnl_stats->tx_errors;
1532 stats->rx_dropped = rtnl_stats->rx_dropped;
1533 stats->tx_dropped = rtnl_stats->tx_dropped;
1534 stats->multicast = rtnl_stats->multicast;
1535 stats->collisions = rtnl_stats->collisions;
1536 stats->rx_length_errors = rtnl_stats->rx_length_errors;
1537 stats->rx_over_errors = rtnl_stats->rx_over_errors;
1538 stats->rx_crc_errors = rtnl_stats->rx_crc_errors;
1539 stats->rx_frame_errors = rtnl_stats->rx_frame_errors;
1540 stats->rx_fifo_errors = rtnl_stats->rx_fifo_errors;
1541 stats->rx_missed_errors = rtnl_stats->rx_missed_errors;
1542 stats->tx_aborted_errors = rtnl_stats->tx_aborted_errors;
1543 stats->tx_carrier_errors = rtnl_stats->tx_carrier_errors;
1544 stats->tx_fifo_errors = rtnl_stats->tx_fifo_errors;
1545 stats->tx_heartbeat_errors = rtnl_stats->tx_heartbeat_errors;
1546 stats->tx_window_errors = rtnl_stats->tx_window_errors;
1548 ofpbuf_delete(reply);
1554 get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats)
1556 static const char fn[] = "/proc/net/dev";
1561 stream = fopen(fn, "r");
1563 VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(errno));
1568 while (fgets(line, sizeof line, stream)) {
1571 #define X64 "%"SCNu64
1574 X64 X64 X64 X64 X64 X64 X64 "%*u"
1575 X64 X64 X64 X64 X64 X64 X64 "%*u",
1581 &stats->rx_fifo_errors,
1582 &stats->rx_frame_errors,
1588 &stats->tx_fifo_errors,
1590 &stats->tx_carrier_errors) != 15) {
1591 VLOG_WARN_RL(&rl, "%s:%d: parse error", fn, ln);
1592 } else if (!strcmp(devname, netdev_name)) {
1593 stats->rx_length_errors = UINT64_MAX;
1594 stats->rx_over_errors = UINT64_MAX;
1595 stats->rx_crc_errors = UINT64_MAX;
1596 stats->rx_missed_errors = UINT64_MAX;
1597 stats->tx_aborted_errors = UINT64_MAX;
1598 stats->tx_heartbeat_errors = UINT64_MAX;
1599 stats->tx_window_errors = UINT64_MAX;
1605 VLOG_WARN_RL(&rl, "%s: no stats for %s", fn, netdev_name);
1611 get_flags(const struct netdev *netdev, int *flags)
1616 error = netdev_linux_do_ioctl(netdev, &ifr, SIOCGIFFLAGS, "SIOCGIFFLAGS");
1617 *flags = ifr.ifr_flags;
1622 set_flags(struct netdev *netdev, int flags)
1626 ifr.ifr_flags = flags;
1627 return netdev_linux_do_ioctl(netdev, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS");
1631 do_get_ifindex(const char *netdev_name)
1635 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1636 COVERAGE_INC(netdev_get_ifindex);
1637 if (ioctl(af_inet_sock, SIOCGIFINDEX, &ifr) < 0) {
1638 VLOG_WARN_RL(&rl, "ioctl(SIOCGIFINDEX) on %s device failed: %s",
1639 netdev_name, strerror(errno));
1642 return ifr.ifr_ifindex;
1646 get_ifindex(const struct netdev *netdev_, int *ifindexp)
1648 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
1650 if (!(netdev->cache->valid & VALID_IFINDEX)) {
1651 int ifindex = do_get_ifindex(netdev_get_name(netdev_));
1655 netdev->cache->valid |= VALID_IFINDEX;
1656 netdev->cache->ifindex = ifindex;
1658 *ifindexp = netdev->cache->ifindex;
1663 get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN])
1668 memset(&ifr, 0, sizeof ifr);
1669 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1670 COVERAGE_INC(netdev_get_hwaddr);
1671 if (ioctl(af_inet_sock, SIOCGIFHWADDR, &ifr) < 0) {
1672 VLOG_ERR("ioctl(SIOCGIFHWADDR) on %s device failed: %s",
1673 netdev_name, strerror(errno));
1676 hwaddr_family = ifr.ifr_hwaddr.sa_family;
1677 if (hwaddr_family != AF_UNSPEC && hwaddr_family != ARPHRD_ETHER) {
1678 VLOG_WARN("%s device has unknown hardware address family %d",
1679 netdev_name, hwaddr_family);
1681 memcpy(ea, ifr.ifr_hwaddr.sa_data, ETH_ADDR_LEN);
1686 set_etheraddr(const char *netdev_name, int hwaddr_family,
1687 const uint8_t mac[ETH_ADDR_LEN])
1691 memset(&ifr, 0, sizeof ifr);
1692 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1693 ifr.ifr_hwaddr.sa_family = hwaddr_family;
1694 memcpy(ifr.ifr_hwaddr.sa_data, mac, ETH_ADDR_LEN);
1695 COVERAGE_INC(netdev_set_hwaddr);
1696 if (ioctl(af_inet_sock, SIOCSIFHWADDR, &ifr) < 0) {
1697 VLOG_ERR("ioctl(SIOCSIFHWADDR) on %s device failed: %s",
1698 netdev_name, strerror(errno));
1705 netdev_linux_do_ethtool(struct netdev *netdev, struct ethtool_cmd *ecmd,
1706 int cmd, const char *cmd_name)
1710 memset(&ifr, 0, sizeof ifr);
1711 strncpy(ifr.ifr_name, netdev->name, sizeof ifr.ifr_name);
1712 ifr.ifr_data = (caddr_t) ecmd;
1715 COVERAGE_INC(netdev_ethtool);
1716 if (ioctl(af_inet_sock, SIOCETHTOOL, &ifr) == 0) {
1719 if (errno != EOPNOTSUPP) {
1720 VLOG_WARN_RL(&rl, "ethtool command %s on network device %s "
1721 "failed: %s", cmd_name, netdev->name,
1724 /* The device doesn't support this operation. That's pretty
1725 * common, so there's no point in logging anything. */
1732 netdev_linux_do_ioctl(const struct netdev *netdev, struct ifreq *ifr,
1733 int cmd, const char *cmd_name)
1735 strncpy(ifr->ifr_name, netdev_get_name(netdev), sizeof ifr->ifr_name);
1736 if (ioctl(af_inet_sock, cmd, ifr) == -1) {
1737 VLOG_DBG_RL(&rl, "%s: ioctl(%s) failed: %s",
1738 netdev_get_name(netdev), cmd_name, strerror(errno));
1745 netdev_linux_get_ipv4(const struct netdev *netdev, struct in_addr *ip,
1746 int cmd, const char *cmd_name)
1751 ifr.ifr_addr.sa_family = AF_INET;
1752 error = netdev_linux_do_ioctl(netdev, &ifr, cmd, cmd_name);
1754 const struct sockaddr_in *sin = (struct sockaddr_in *) &ifr.ifr_addr;
1755 *ip = sin->sin_addr;