2 * Copyright (c) 2009, 2010 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)
70 /* Provider-specific netdev object. Netdev objects are devices that are
71 * created by the netdev library through a netdev_create() call. */
72 struct netdev_obj_linux {
73 struct netdev_obj netdev_obj;
75 int tap_fd; /* File descriptor for TAP device. */
81 /* File descriptors. For ordinary network devices, the two fds below are
82 * the same; for tap devices, they differ. */
83 int netdev_fd; /* Network device. */
84 int tap_fd; /* TAP character device, if any, otherwise the
87 struct netdev_linux_cache *cache;
91 VALID_IFINDEX = 1 << 0,
92 VALID_ETHERADDR = 1 << 1,
96 VALID_CARRIER = 1 << 5,
97 VALID_IS_INTERNAL = 1 << 6
100 /* Cached network device information. */
101 struct netdev_linux_cache {
102 struct shash_node *shash_node;
107 uint8_t etheraddr[ETH_ADDR_LEN];
108 struct in_addr address, netmask;
115 static struct shash cache_map = SHASH_INITIALIZER(&cache_map);
116 static struct rtnetlink_notifier netdev_linux_cache_notifier;
118 /* An AF_INET socket (used for ioctl operations). */
119 static int af_inet_sock = -1;
121 struct netdev_linux_notifier {
122 struct netdev_notifier notifier;
126 static struct shash netdev_linux_notifiers =
127 SHASH_INITIALIZER(&netdev_linux_notifiers);
128 static struct rtnetlink_notifier netdev_linux_poll_notifier;
130 /* This is set pretty low because we probably won't learn anything from the
131 * additional log messages. */
132 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
134 static int netdev_linux_do_ethtool(struct netdev *, struct ethtool_cmd *,
135 int cmd, const char *cmd_name);
136 static int netdev_linux_do_ioctl(const struct netdev *, struct ifreq *,
137 int cmd, const char *cmd_name);
138 static int netdev_linux_get_ipv4(const struct netdev *, struct in_addr *,
139 int cmd, const char *cmd_name);
140 static int get_flags(const struct netdev *, int *flagsp);
141 static int set_flags(struct netdev *, int flags);
142 static int do_get_ifindex(const char *netdev_name);
143 static int get_ifindex(const struct netdev *, int *ifindexp);
144 static int do_set_addr(struct netdev *netdev,
145 int ioctl_nr, const char *ioctl_name,
146 struct in_addr addr);
147 static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]);
148 static int set_etheraddr(const char *netdev_name, int hwaddr_family,
149 const uint8_t[ETH_ADDR_LEN]);
150 static int get_stats_via_netlink(int ifindex, struct netdev_stats *stats);
151 static int get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats);
153 static struct netdev_obj_linux *
154 netdev_obj_linux_cast(const struct netdev_obj *netdev_obj)
156 netdev_obj_assert_class(netdev_obj, &netdev_linux_class);
157 return CONTAINER_OF(netdev_obj, struct netdev_obj_linux, netdev_obj);
160 static struct netdev_linux *
161 netdev_linux_cast(const struct netdev *netdev)
163 netdev_assert_class(netdev, &netdev_linux_class);
164 return CONTAINER_OF(netdev, struct netdev_linux, netdev);
168 netdev_linux_init(void)
170 static int status = -1;
172 af_inet_sock = socket(AF_INET, SOCK_DGRAM, 0);
173 status = af_inet_sock >= 0 ? 0 : errno;
175 VLOG_ERR("failed to create inet socket: %s", strerror(status));
182 netdev_linux_run(void)
184 rtnetlink_notifier_run();
188 netdev_linux_wait(void)
190 rtnetlink_notifier_wait();
194 netdev_linux_cache_cb(const struct rtnetlink_change *change,
195 void *aux OVS_UNUSED)
197 struct netdev_linux_cache *cache;
199 cache = shash_find_data(&cache_map, change->ifname);
204 struct shash_node *node;
205 SHASH_FOR_EACH (node, &cache_map) {
212 /* Creates the netdev object of 'type' with 'name'. */
214 netdev_linux_create(const char *name, const char *type,
215 const struct shash *args, bool created)
217 struct netdev_obj_linux *netdev_obj;
218 static const char tap_dev[] = "/dev/net/tun";
222 if (!shash_is_empty(args)) {
223 VLOG_WARN("arguments for %s devices should be empty", type);
226 /* Create the name binding in the netdev library for this object. */
227 netdev_obj = xcalloc(1, sizeof *netdev_obj);
228 netdev_obj_init(&netdev_obj->netdev_obj, name, &netdev_linux_class,
230 netdev_obj->tap_fd = -1;
232 if (strcmp(type, "tap")) {
236 /* Open tap device. */
237 netdev_obj->tap_fd = open(tap_dev, O_RDWR);
238 if (netdev_obj->tap_fd < 0) {
240 VLOG_WARN("opening \"%s\" failed: %s", tap_dev, strerror(error));
244 /* Create tap device. */
245 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
246 strncpy(ifr.ifr_name, name, sizeof ifr.ifr_name);
247 if (ioctl(netdev_obj->tap_fd, TUNSETIFF, &ifr) == -1) {
248 VLOG_WARN("%s: creating tap device failed: %s", name,
254 /* Make non-blocking. */
255 error = set_nonblocking(netdev_obj->tap_fd);
263 netdev_destroy(name);
267 /* Destroys the netdev object 'netdev_obj_'. */
269 netdev_linux_destroy(struct netdev_obj *netdev_obj_)
271 struct netdev_obj_linux *netdev_obj = netdev_obj_linux_cast(netdev_obj_);
273 if (netdev_obj->tap_fd >= 0) {
274 close(netdev_obj->tap_fd);
282 netdev_linux_open(const char *name, int ethertype, struct netdev **netdevp)
284 struct netdev_linux *netdev;
285 enum netdev_flags flags;
288 /* Allocate network device. */
289 netdev = xcalloc(1, sizeof *netdev);
290 netdev_init(&netdev->netdev, name, &netdev_linux_class);
291 netdev->netdev_fd = -1;
293 netdev->cache = shash_find_data(&cache_map, name);
294 if (!netdev->cache) {
295 if (shash_is_empty(&cache_map)) {
296 int error = rtnetlink_notifier_register(
297 &netdev_linux_cache_notifier, netdev_linux_cache_cb, NULL);
299 netdev_close(&netdev->netdev);
303 netdev->cache = xmalloc(sizeof *netdev->cache);
304 netdev->cache->shash_node = shash_add(&cache_map, name,
306 netdev->cache->valid = 0;
307 netdev->cache->ref_cnt = 0;
309 netdev->cache->ref_cnt++;
311 if (!strcmp(netdev_get_type(&netdev->netdev), "tap")) {
312 static const char tap_dev[] = "/dev/net/tun";
315 /* Open tap device. */
316 netdev->tap_fd = open(tap_dev, O_RDWR);
317 if (netdev->tap_fd < 0) {
319 VLOG_WARN("opening \"%s\" failed: %s", tap_dev, strerror(error));
323 /* Create tap device. */
324 ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
325 strncpy(ifr.ifr_name, name, sizeof ifr.ifr_name);
326 if (ioctl(netdev->tap_fd, TUNSETIFF, &ifr) == -1) {
327 VLOG_WARN("%s: creating tap device failed: %s", name,
333 /* Make non-blocking. */
334 error = set_nonblocking(netdev->tap_fd);
340 error = netdev_get_flags(&netdev->netdev, &flags);
341 if (error == ENODEV) {
345 if (netdev->tap_fd >= 0 || ethertype != NETDEV_ETH_TYPE_NONE) {
346 struct sockaddr_ll sll;
350 /* Create file descriptor. */
351 protocol = (ethertype == NETDEV_ETH_TYPE_ANY ? ETH_P_ALL
352 : ethertype == NETDEV_ETH_TYPE_802_2 ? ETH_P_802_2
354 netdev->netdev_fd = socket(PF_PACKET, SOCK_RAW, htons(protocol));
355 if (netdev->netdev_fd < 0) {
359 if (netdev->tap_fd < 0) {
360 netdev->tap_fd = netdev->netdev_fd;
363 /* Set non-blocking mode. */
364 error = set_nonblocking(netdev->netdev_fd);
369 /* Get ethernet device index. */
370 error = get_ifindex(&netdev->netdev, &ifindex);
375 /* Bind to specific ethernet device. */
376 memset(&sll, 0, sizeof sll);
377 sll.sll_family = AF_PACKET;
378 sll.sll_ifindex = ifindex;
379 if (bind(netdev->netdev_fd,
380 (struct sockaddr *) &sll, sizeof sll) < 0) {
382 VLOG_ERR("bind to %s failed: %s", name, strerror(error));
386 /* Between the socket() and bind() calls above, the socket receives all
387 * packets of the requested type on all system interfaces. We do not
388 * want to receive that data, but there is no way to avoid it. So we
389 * must now drain out the receive queue. */
390 error = drain_rcvbuf(netdev->netdev_fd);
396 *netdevp = &netdev->netdev;
400 netdev_close(&netdev->netdev);
404 /* Closes and destroys 'netdev'. */
406 netdev_linux_close(struct netdev *netdev_)
408 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
410 if (netdev->cache && !--netdev->cache->ref_cnt) {
411 shash_delete(&cache_map, netdev->cache->shash_node);
414 if (shash_is_empty(&cache_map)) {
415 rtnetlink_notifier_unregister(&netdev_linux_cache_notifier);
418 if (netdev->netdev_fd >= 0) {
419 close(netdev->netdev_fd);
421 if (netdev->tap_fd >= 0 && netdev->netdev_fd != netdev->tap_fd) {
422 close(netdev->tap_fd);
427 /* Initializes 'svec' with a list of the names of all known network devices. */
429 netdev_linux_enumerate(struct svec *svec)
431 struct if_nameindex *names;
433 names = if_nameindex();
437 for (i = 0; names[i].if_name != NULL; i++) {
438 svec_add(svec, names[i].if_name);
440 if_freenameindex(names);
443 VLOG_WARN("could not obtain list of network device names: %s",
450 netdev_linux_recv(struct netdev *netdev_, void *data, size_t size)
452 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
454 if (netdev->tap_fd < 0) {
455 /* Device was opened with NETDEV_ETH_TYPE_NONE. */
460 ssize_t retval = read(netdev->tap_fd, data, size);
463 } else if (errno != EINTR) {
464 if (errno != EAGAIN) {
465 VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
466 strerror(errno), netdev_get_name(netdev_));
473 /* Registers with the poll loop to wake up from the next call to poll_block()
474 * when a packet is ready to be received with netdev_recv() on 'netdev'. */
476 netdev_linux_recv_wait(struct netdev *netdev_)
478 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
479 if (netdev->tap_fd >= 0) {
480 poll_fd_wait(netdev->tap_fd, POLLIN);
484 /* Discards all packets waiting to be received from 'netdev'. */
486 netdev_linux_drain(struct netdev *netdev_)
488 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
489 if (netdev->tap_fd < 0 && netdev->netdev_fd < 0) {
491 } else if (netdev->tap_fd != netdev->netdev_fd) {
493 int error = netdev_linux_do_ioctl(netdev_, &ifr,
494 SIOCGIFTXQLEN, "SIOCGIFTXQLEN");
498 drain_fd(netdev->tap_fd, ifr.ifr_qlen);
501 return drain_rcvbuf(netdev->netdev_fd);
505 /* Sends 'buffer' on 'netdev'. Returns 0 if successful, otherwise a positive
506 * errno value. Returns EAGAIN without blocking if the packet cannot be queued
507 * immediately. Returns EMSGSIZE if a partial packet was transmitted or if
508 * the packet is too big or too small to transmit on the device.
510 * The caller retains ownership of 'buffer' in all cases.
512 * The kernel maintains a packet transmission queue, so the caller is not
513 * expected to do additional queuing of packets. */
515 netdev_linux_send(struct netdev *netdev_, const void *data, size_t size)
517 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
519 /* XXX should support sending even if 'ethertype' was NETDEV_ETH_TYPE_NONE.
521 if (netdev->tap_fd < 0) {
526 ssize_t retval = write(netdev->tap_fd, data, size);
528 /* The Linux AF_PACKET implementation never blocks waiting for room
529 * for packets, instead returning ENOBUFS. Translate this into
530 * EAGAIN for the caller. */
531 if (errno == ENOBUFS) {
533 } else if (errno == EINTR) {
535 } else if (errno != EAGAIN) {
536 VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s",
537 netdev_get_name(netdev_), strerror(errno));
540 } else if (retval != size) {
541 VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%zd bytes of "
542 "%zu) on %s", retval, size, netdev_get_name(netdev_));
550 /* Registers with the poll loop to wake up from the next call to poll_block()
551 * when the packet transmission queue has sufficient room to transmit a packet
552 * with netdev_send().
554 * The kernel maintains a packet transmission queue, so the client is not
555 * expected to do additional queuing of packets. Thus, this function is
556 * unlikely to ever be used. It is included for completeness. */
558 netdev_linux_send_wait(struct netdev *netdev_)
560 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
561 if (netdev->tap_fd < 0 && netdev->netdev_fd < 0) {
563 } else if (netdev->tap_fd == netdev->netdev_fd) {
564 poll_fd_wait(netdev->tap_fd, POLLOUT);
566 /* TAP device always accepts packets.*/
567 poll_immediate_wake();
571 /* Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
572 * otherwise a positive errno value. */
574 netdev_linux_set_etheraddr(struct netdev *netdev_,
575 const uint8_t mac[ETH_ADDR_LEN])
577 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
580 if (!(netdev->cache->valid & VALID_ETHERADDR)
581 || !eth_addr_equals(netdev->cache->etheraddr, mac)) {
582 error = set_etheraddr(netdev_get_name(netdev_), ARPHRD_ETHER, mac);
584 netdev->cache->valid |= VALID_ETHERADDR;
585 memcpy(netdev->cache->etheraddr, mac, ETH_ADDR_LEN);
593 /* Returns a pointer to 'netdev''s MAC address. The caller must not modify or
594 * free the returned buffer. */
596 netdev_linux_get_etheraddr(const struct netdev *netdev_,
597 uint8_t mac[ETH_ADDR_LEN])
599 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
600 if (!(netdev->cache->valid & VALID_ETHERADDR)) {
601 int error = get_etheraddr(netdev_get_name(netdev_),
602 netdev->cache->etheraddr);
606 netdev->cache->valid |= VALID_ETHERADDR;
608 memcpy(mac, netdev->cache->etheraddr, ETH_ADDR_LEN);
612 /* Returns the maximum size of transmitted (and received) packets on 'netdev',
613 * in bytes, not including the hardware header; thus, this is typically 1500
614 * bytes for Ethernet devices. */
616 netdev_linux_get_mtu(const struct netdev *netdev_, int *mtup)
618 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
619 if (!(netdev->cache->valid & VALID_MTU)) {
623 error = netdev_linux_do_ioctl(netdev_, &ifr, SIOCGIFMTU, "SIOCGIFMTU");
627 netdev->cache->mtu = ifr.ifr_mtu;
628 netdev->cache->valid |= VALID_MTU;
630 *mtup = netdev->cache->mtu;
634 /* Returns the ifindex of 'netdev', if successful, as a positive number.
635 * On failure, returns a negative errno value. */
637 netdev_linux_get_ifindex(const struct netdev *netdev)
641 error = get_ifindex(netdev, &ifindex);
642 return error ? -error : ifindex;
646 netdev_linux_get_carrier(const struct netdev *netdev_, bool *carrier)
648 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
653 if (!(netdev->cache->valid & VALID_CARRIER)) {
657 fn = xasprintf("/sys/class/net/%s/carrier", netdev_get_name(netdev_));
658 fd = open(fn, O_RDONLY);
661 VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(error));
665 retval = read(fd, line, sizeof line);
668 if (error == EINVAL) {
669 /* This is the normal return value when we try to check carrier
670 * if the network device is not up. */
672 VLOG_WARN_RL(&rl, "%s: read failed: %s", fn, strerror(error));
675 } else if (retval == 0) {
677 VLOG_WARN_RL(&rl, "%s: unexpected end of file", fn);
681 if (line[0] != '0' && line[0] != '1') {
683 VLOG_WARN_RL(&rl, "%s: value is %c (expected 0 or 1)",
687 netdev->cache->carrier = line[0] != '0';
688 netdev->cache->valid |= VALID_CARRIER;
690 *carrier = netdev->cache->carrier;
701 /* Check whether we can we use RTM_GETLINK to get network device statistics.
702 * In pre-2.6.19 kernels, this was only available if wireless extensions were
705 check_for_working_netlink_stats(void)
707 /* Decide on the netdev_get_stats() implementation to use. Netlink is
708 * preferable, so if that works, we'll use it. */
709 int ifindex = do_get_ifindex("lo");
711 VLOG_WARN("failed to get ifindex for lo, "
712 "obtaining netdev stats from proc");
715 struct netdev_stats stats;
716 int error = get_stats_via_netlink(ifindex, &stats);
718 VLOG_DBG("obtaining netdev stats via rtnetlink");
721 VLOG_INFO("RTM_GETLINK failed (%s), obtaining netdev stats "
722 "via proc (you are probably running a pre-2.6.19 "
723 "kernel)", strerror(error));
729 /* Retrieves current device stats for 'netdev'.
731 * XXX All of the members of struct netdev_stats are 64 bits wide, but on
732 * 32-bit architectures the Linux network stats are only 32 bits. */
734 netdev_linux_get_stats(const struct netdev *netdev_, struct netdev_stats *stats)
736 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
737 static int use_netlink_stats = -1;
739 struct netdev_stats raw_stats;
740 struct netdev_stats *collect_stats = stats;
742 COVERAGE_INC(netdev_get_stats);
744 if (!(netdev->cache->valid & VALID_IS_INTERNAL)) {
745 netdev->cache->is_internal = (netdev->tap_fd != -1);
747 if (!netdev->cache->is_internal) {
748 struct ethtool_drvinfo drvinfo;
750 memset(&drvinfo, 0, sizeof drvinfo);
751 error = netdev_linux_do_ethtool(&netdev->netdev,
752 (struct ethtool_cmd *)&drvinfo,
757 netdev->cache->is_internal = !strcmp(drvinfo.driver,
762 netdev->cache->valid |= VALID_IS_INTERNAL;
765 if (netdev->cache->is_internal) {
766 collect_stats = &raw_stats;
769 if (use_netlink_stats < 0) {
770 use_netlink_stats = check_for_working_netlink_stats();
772 if (use_netlink_stats) {
775 error = get_ifindex(&netdev->netdev, &ifindex);
777 error = get_stats_via_netlink(ifindex, collect_stats);
780 error = get_stats_via_proc(netdev->netdev.name, collect_stats);
783 /* If this port is an internal port then the transmit and receive stats
784 * will appear to be swapped relative to the other ports since we are the
785 * one sending the data, not a remote computer. For consistency, we swap
787 if (netdev->cache->is_internal) {
788 stats->rx_packets = raw_stats.tx_packets;
789 stats->tx_packets = raw_stats.rx_packets;
790 stats->rx_bytes = raw_stats.tx_bytes;
791 stats->tx_bytes = raw_stats.rx_bytes;
792 stats->rx_errors = raw_stats.tx_errors;
793 stats->tx_errors = raw_stats.rx_errors;
794 stats->rx_dropped = raw_stats.tx_dropped;
795 stats->tx_dropped = raw_stats.rx_dropped;
796 stats->multicast = raw_stats.multicast;
797 stats->collisions = raw_stats.collisions;
798 stats->rx_length_errors = 0;
799 stats->rx_over_errors = 0;
800 stats->rx_crc_errors = 0;
801 stats->rx_frame_errors = 0;
802 stats->rx_fifo_errors = 0;
803 stats->rx_missed_errors = 0;
804 stats->tx_aborted_errors = 0;
805 stats->tx_carrier_errors = 0;
806 stats->tx_fifo_errors = 0;
807 stats->tx_heartbeat_errors = 0;
808 stats->tx_window_errors = 0;
814 /* Stores the features supported by 'netdev' into each of '*current',
815 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
816 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
817 * successful, otherwise a positive errno value. */
819 netdev_linux_get_features(struct netdev *netdev,
820 uint32_t *current, uint32_t *advertised,
821 uint32_t *supported, uint32_t *peer)
823 struct ethtool_cmd ecmd;
826 memset(&ecmd, 0, sizeof ecmd);
827 error = netdev_linux_do_ethtool(netdev, &ecmd,
828 ETHTOOL_GSET, "ETHTOOL_GSET");
833 /* Supported features. */
835 if (ecmd.supported & SUPPORTED_10baseT_Half) {
836 *supported |= OFPPF_10MB_HD;
838 if (ecmd.supported & SUPPORTED_10baseT_Full) {
839 *supported |= OFPPF_10MB_FD;
841 if (ecmd.supported & SUPPORTED_100baseT_Half) {
842 *supported |= OFPPF_100MB_HD;
844 if (ecmd.supported & SUPPORTED_100baseT_Full) {
845 *supported |= OFPPF_100MB_FD;
847 if (ecmd.supported & SUPPORTED_1000baseT_Half) {
848 *supported |= OFPPF_1GB_HD;
850 if (ecmd.supported & SUPPORTED_1000baseT_Full) {
851 *supported |= OFPPF_1GB_FD;
853 if (ecmd.supported & SUPPORTED_10000baseT_Full) {
854 *supported |= OFPPF_10GB_FD;
856 if (ecmd.supported & SUPPORTED_TP) {
857 *supported |= OFPPF_COPPER;
859 if (ecmd.supported & SUPPORTED_FIBRE) {
860 *supported |= OFPPF_FIBER;
862 if (ecmd.supported & SUPPORTED_Autoneg) {
863 *supported |= OFPPF_AUTONEG;
865 if (ecmd.supported & SUPPORTED_Pause) {
866 *supported |= OFPPF_PAUSE;
868 if (ecmd.supported & SUPPORTED_Asym_Pause) {
869 *supported |= OFPPF_PAUSE_ASYM;
872 /* Advertised features. */
874 if (ecmd.advertising & ADVERTISED_10baseT_Half) {
875 *advertised |= OFPPF_10MB_HD;
877 if (ecmd.advertising & ADVERTISED_10baseT_Full) {
878 *advertised |= OFPPF_10MB_FD;
880 if (ecmd.advertising & ADVERTISED_100baseT_Half) {
881 *advertised |= OFPPF_100MB_HD;
883 if (ecmd.advertising & ADVERTISED_100baseT_Full) {
884 *advertised |= OFPPF_100MB_FD;
886 if (ecmd.advertising & ADVERTISED_1000baseT_Half) {
887 *advertised |= OFPPF_1GB_HD;
889 if (ecmd.advertising & ADVERTISED_1000baseT_Full) {
890 *advertised |= OFPPF_1GB_FD;
892 if (ecmd.advertising & ADVERTISED_10000baseT_Full) {
893 *advertised |= OFPPF_10GB_FD;
895 if (ecmd.advertising & ADVERTISED_TP) {
896 *advertised |= OFPPF_COPPER;
898 if (ecmd.advertising & ADVERTISED_FIBRE) {
899 *advertised |= OFPPF_FIBER;
901 if (ecmd.advertising & ADVERTISED_Autoneg) {
902 *advertised |= OFPPF_AUTONEG;
904 if (ecmd.advertising & ADVERTISED_Pause) {
905 *advertised |= OFPPF_PAUSE;
907 if (ecmd.advertising & ADVERTISED_Asym_Pause) {
908 *advertised |= OFPPF_PAUSE_ASYM;
911 /* Current settings. */
912 if (ecmd.speed == SPEED_10) {
913 *current = ecmd.duplex ? OFPPF_10MB_FD : OFPPF_10MB_HD;
914 } else if (ecmd.speed == SPEED_100) {
915 *current = ecmd.duplex ? OFPPF_100MB_FD : OFPPF_100MB_HD;
916 } else if (ecmd.speed == SPEED_1000) {
917 *current = ecmd.duplex ? OFPPF_1GB_FD : OFPPF_1GB_HD;
918 } else if (ecmd.speed == SPEED_10000) {
919 *current = OFPPF_10GB_FD;
924 if (ecmd.port == PORT_TP) {
925 *current |= OFPPF_COPPER;
926 } else if (ecmd.port == PORT_FIBRE) {
927 *current |= OFPPF_FIBER;
931 *current |= OFPPF_AUTONEG;
934 /* Peer advertisements. */
940 /* Set the features advertised by 'netdev' to 'advertise'. */
942 netdev_linux_set_advertisements(struct netdev *netdev, uint32_t advertise)
944 struct ethtool_cmd ecmd;
947 memset(&ecmd, 0, sizeof ecmd);
948 error = netdev_linux_do_ethtool(netdev, &ecmd,
949 ETHTOOL_GSET, "ETHTOOL_GSET");
954 ecmd.advertising = 0;
955 if (advertise & OFPPF_10MB_HD) {
956 ecmd.advertising |= ADVERTISED_10baseT_Half;
958 if (advertise & OFPPF_10MB_FD) {
959 ecmd.advertising |= ADVERTISED_10baseT_Full;
961 if (advertise & OFPPF_100MB_HD) {
962 ecmd.advertising |= ADVERTISED_100baseT_Half;
964 if (advertise & OFPPF_100MB_FD) {
965 ecmd.advertising |= ADVERTISED_100baseT_Full;
967 if (advertise & OFPPF_1GB_HD) {
968 ecmd.advertising |= ADVERTISED_1000baseT_Half;
970 if (advertise & OFPPF_1GB_FD) {
971 ecmd.advertising |= ADVERTISED_1000baseT_Full;
973 if (advertise & OFPPF_10GB_FD) {
974 ecmd.advertising |= ADVERTISED_10000baseT_Full;
976 if (advertise & OFPPF_COPPER) {
977 ecmd.advertising |= ADVERTISED_TP;
979 if (advertise & OFPPF_FIBER) {
980 ecmd.advertising |= ADVERTISED_FIBRE;
982 if (advertise & OFPPF_AUTONEG) {
983 ecmd.advertising |= ADVERTISED_Autoneg;
985 if (advertise & OFPPF_PAUSE) {
986 ecmd.advertising |= ADVERTISED_Pause;
988 if (advertise & OFPPF_PAUSE_ASYM) {
989 ecmd.advertising |= ADVERTISED_Asym_Pause;
991 return netdev_linux_do_ethtool(netdev, &ecmd,
992 ETHTOOL_SSET, "ETHTOOL_SSET");
995 /* If 'netdev_name' is the name of a VLAN network device (e.g. one created with
996 * vconfig(8)), sets '*vlan_vid' to the VLAN VID associated with that device
997 * and returns 0. Otherwise returns a errno value (specifically ENOENT if
998 * 'netdev_name' is the name of a network device that is not a VLAN device) and
999 * sets '*vlan_vid' to -1. */
1001 netdev_linux_get_vlan_vid(const struct netdev *netdev, int *vlan_vid)
1003 const char *netdev_name = netdev_get_name(netdev);
1004 struct ds line = DS_EMPTY_INITIALIZER;
1005 FILE *stream = NULL;
1009 COVERAGE_INC(netdev_get_vlan_vid);
1010 fn = xasprintf("/proc/net/vlan/%s", netdev_name);
1011 stream = fopen(fn, "r");
1017 if (ds_get_line(&line, stream)) {
1018 if (ferror(stream)) {
1020 VLOG_ERR_RL(&rl, "error reading \"%s\": %s", fn, strerror(errno));
1023 VLOG_ERR_RL(&rl, "unexpected end of file reading \"%s\"", fn);
1028 if (!sscanf(ds_cstr(&line), "%*s VID: %d", vlan_vid)) {
1030 VLOG_ERR_RL(&rl, "parse error reading \"%s\" line 1: \"%s\"",
1031 fn, ds_cstr(&line));
1049 #define POLICE_ADD_CMD "/sbin/tc qdisc add dev %s handle ffff: ingress"
1050 #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"
1051 /* We redirect stderr to /dev/null because we often want to remove all
1052 * traffic control configuration on a port so its in a known state. If
1053 * this done when there is no such configuration, tc complains, so we just
1056 #define POLICE_DEL_CMD "/sbin/tc qdisc del dev %s handle ffff: ingress 2>/dev/null"
1058 /* Attempts to set input rate limiting (policing) policy. */
1060 netdev_linux_set_policing(struct netdev *netdev,
1061 uint32_t kbits_rate, uint32_t kbits_burst)
1063 const char *netdev_name = netdev_get_name(netdev);
1066 COVERAGE_INC(netdev_set_policing);
1069 /* Default to 10 kilobits if not specified. */
1073 /* xxx This should be more careful about only adding if it
1074 * xxx actually exists, as opposed to always deleting it. */
1075 snprintf(command, sizeof(command), POLICE_DEL_CMD, netdev_name);
1076 if (system(command) == -1) {
1077 VLOG_WARN_RL(&rl, "%s: problem removing policing", netdev_name);
1080 snprintf(command, sizeof(command), POLICE_ADD_CMD, netdev_name);
1081 if (system(command) != 0) {
1082 VLOG_WARN_RL(&rl, "%s: problem adding policing", netdev_name);
1086 snprintf(command, sizeof(command), POLICE_CONFIG_CMD, netdev_name,
1087 kbits_rate, kbits_burst);
1088 if (system(command) != 0) {
1089 VLOG_WARN_RL(&rl, "%s: problem configuring policing",
1094 snprintf(command, sizeof(command), POLICE_DEL_CMD, netdev_name);
1095 if (system(command) == -1) {
1096 VLOG_WARN_RL(&rl, "%s: problem removing policing", netdev_name);
1104 netdev_linux_get_in4(const struct netdev *netdev_,
1105 struct in_addr *address, struct in_addr *netmask)
1107 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
1108 if (!(netdev->cache->valid & VALID_IN4)) {
1111 error = netdev_linux_get_ipv4(netdev_, &netdev->cache->address,
1112 SIOCGIFADDR, "SIOCGIFADDR");
1117 error = netdev_linux_get_ipv4(netdev_, &netdev->cache->netmask,
1118 SIOCGIFNETMASK, "SIOCGIFNETMASK");
1123 netdev->cache->valid |= VALID_IN4;
1125 *address = netdev->cache->address;
1126 *netmask = netdev->cache->netmask;
1127 return address->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0;
1131 netdev_linux_set_in4(struct netdev *netdev_, struct in_addr address,
1132 struct in_addr netmask)
1134 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
1137 error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", address);
1139 netdev->cache->valid |= VALID_IN4;
1140 netdev->cache->address = address;
1141 netdev->cache->netmask = netmask;
1142 if (address.s_addr != INADDR_ANY) {
1143 error = do_set_addr(netdev_, SIOCSIFNETMASK,
1144 "SIOCSIFNETMASK", netmask);
1151 parse_if_inet6_line(const char *line,
1152 struct in6_addr *in6, char ifname[16 + 1])
1154 uint8_t *s6 = in6->s6_addr;
1155 #define X8 "%2"SCNx8
1157 " "X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8
1158 "%*x %*x %*x %*x %16s\n",
1159 &s6[0], &s6[1], &s6[2], &s6[3],
1160 &s6[4], &s6[5], &s6[6], &s6[7],
1161 &s6[8], &s6[9], &s6[10], &s6[11],
1162 &s6[12], &s6[13], &s6[14], &s6[15],
1166 /* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address (if
1167 * 'in6' is non-null) and returns true. Otherwise, returns false. */
1169 netdev_linux_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
1171 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
1172 if (!(netdev->cache->valid & VALID_IN6)) {
1176 netdev->cache->in6 = in6addr_any;
1178 file = fopen("/proc/net/if_inet6", "r");
1180 const char *name = netdev_get_name(netdev_);
1181 while (fgets(line, sizeof line, file)) {
1182 struct in6_addr in6;
1183 char ifname[16 + 1];
1184 if (parse_if_inet6_line(line, &in6, ifname)
1185 && !strcmp(name, ifname))
1187 netdev->cache->in6 = in6;
1193 netdev->cache->valid |= VALID_IN6;
1195 *in6 = netdev->cache->in6;
1200 make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
1202 struct sockaddr_in sin;
1203 memset(&sin, 0, sizeof sin);
1204 sin.sin_family = AF_INET;
1205 sin.sin_addr = addr;
1208 memset(sa, 0, sizeof *sa);
1209 memcpy(sa, &sin, sizeof sin);
1213 do_set_addr(struct netdev *netdev,
1214 int ioctl_nr, const char *ioctl_name, struct in_addr addr)
1217 strncpy(ifr.ifr_name, netdev->name, sizeof ifr.ifr_name);
1218 make_in4_sockaddr(&ifr.ifr_addr, addr);
1219 return netdev_linux_do_ioctl(netdev, &ifr, ioctl_nr, ioctl_name);
1222 /* Adds 'router' as a default IP gateway. */
1224 netdev_linux_add_router(struct netdev *netdev OVS_UNUSED, struct in_addr router)
1226 struct in_addr any = { INADDR_ANY };
1230 memset(&rt, 0, sizeof rt);
1231 make_in4_sockaddr(&rt.rt_dst, any);
1232 make_in4_sockaddr(&rt.rt_gateway, router);
1233 make_in4_sockaddr(&rt.rt_genmask, any);
1234 rt.rt_flags = RTF_UP | RTF_GATEWAY;
1235 COVERAGE_INC(netdev_add_router);
1236 error = ioctl(af_inet_sock, SIOCADDRT, &rt) < 0 ? errno : 0;
1238 VLOG_WARN("ioctl(SIOCADDRT): %s", strerror(error));
1244 netdev_linux_get_next_hop(const struct in_addr *host, struct in_addr *next_hop,
1247 static const char fn[] = "/proc/net/route";
1252 *netdev_name = NULL;
1253 stream = fopen(fn, "r");
1254 if (stream == NULL) {
1255 VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(errno));
1260 while (fgets(line, sizeof line, stream)) {
1263 uint32_t dest, gateway, mask;
1264 int refcnt, metric, mtu;
1265 unsigned int flags, use, window, irtt;
1268 "%16s %"SCNx32" %"SCNx32" %04X %d %u %d %"SCNx32
1270 iface, &dest, &gateway, &flags, &refcnt,
1271 &use, &metric, &mask, &mtu, &window, &irtt) != 11) {
1273 VLOG_WARN_RL(&rl, "%s: could not parse line %d: %s",
1277 if (!(flags & RTF_UP)) {
1278 /* Skip routes that aren't up. */
1282 /* The output of 'dest', 'mask', and 'gateway' were given in
1283 * network byte order, so we don't need need any endian
1284 * conversions here. */
1285 if ((dest & mask) == (host->s_addr & mask)) {
1287 /* The host is directly reachable. */
1288 next_hop->s_addr = 0;
1290 /* To reach the host, we must go through a gateway. */
1291 next_hop->s_addr = gateway;
1293 *netdev_name = xstrdup(iface);
1304 /* Looks up the ARP table entry for 'ip' on 'netdev'. If one exists and can be
1305 * successfully retrieved, it stores the corresponding MAC address in 'mac' and
1306 * returns 0. Otherwise, it returns a positive errno value; in particular,
1307 * ENXIO indicates that there is not ARP table entry for 'ip' on 'netdev'. */
1309 netdev_linux_arp_lookup(const struct netdev *netdev,
1310 uint32_t ip, uint8_t mac[ETH_ADDR_LEN])
1313 struct sockaddr_in *pa;
1316 memset(&r, 0, sizeof r);
1317 pa = (struct sockaddr_in *) &r.arp_pa;
1318 pa->sin_family = AF_INET;
1319 pa->sin_addr.s_addr = ip;
1321 r.arp_ha.sa_family = ARPHRD_ETHER;
1323 strncpy(r.arp_dev, netdev->name, sizeof r.arp_dev);
1324 COVERAGE_INC(netdev_arp_lookup);
1325 retval = ioctl(af_inet_sock, SIOCGARP, &r) < 0 ? errno : 0;
1327 memcpy(mac, r.arp_ha.sa_data, ETH_ADDR_LEN);
1328 } else if (retval != ENXIO) {
1329 VLOG_WARN_RL(&rl, "%s: could not look up ARP entry for "IP_FMT": %s",
1330 netdev->name, IP_ARGS(&ip), strerror(retval));
1336 nd_to_iff_flags(enum netdev_flags nd)
1339 if (nd & NETDEV_UP) {
1342 if (nd & NETDEV_PROMISC) {
1349 iff_to_nd_flags(int iff)
1351 enum netdev_flags nd = 0;
1355 if (iff & IFF_PROMISC) {
1356 nd |= NETDEV_PROMISC;
1362 netdev_linux_update_flags(struct netdev *netdev, enum netdev_flags off,
1363 enum netdev_flags on, enum netdev_flags *old_flagsp)
1365 int old_flags, new_flags;
1368 error = get_flags(netdev, &old_flags);
1370 *old_flagsp = iff_to_nd_flags(old_flags);
1371 new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on);
1372 if (new_flags != old_flags) {
1373 error = set_flags(netdev, new_flags);
1380 poll_notify(struct list *list)
1382 struct netdev_linux_notifier *notifier;
1383 LIST_FOR_EACH (notifier, struct netdev_linux_notifier, node, list) {
1384 struct netdev_notifier *n = ¬ifier->notifier;
1390 netdev_linux_poll_cb(const struct rtnetlink_change *change,
1391 void *aux OVS_UNUSED)
1394 struct list *list = shash_find_data(&netdev_linux_notifiers,
1400 struct shash_node *node;
1401 SHASH_FOR_EACH (node, &netdev_linux_notifiers) {
1402 poll_notify(node->data);
1408 netdev_linux_poll_add(struct netdev *netdev,
1409 void (*cb)(struct netdev_notifier *), void *aux,
1410 struct netdev_notifier **notifierp)
1412 const char *netdev_name = netdev_get_name(netdev);
1413 struct netdev_linux_notifier *notifier;
1416 if (shash_is_empty(&netdev_linux_notifiers)) {
1417 int error = rtnetlink_notifier_register(&netdev_linux_poll_notifier,
1418 netdev_linux_poll_cb, NULL);
1424 list = shash_find_data(&netdev_linux_notifiers, netdev_name);
1426 list = xmalloc(sizeof *list);
1428 shash_add(&netdev_linux_notifiers, netdev_name, list);
1431 notifier = xmalloc(sizeof *notifier);
1432 netdev_notifier_init(¬ifier->notifier, netdev, cb, aux);
1433 list_push_back(list, ¬ifier->node);
1434 *notifierp = ¬ifier->notifier;
1439 netdev_linux_poll_remove(struct netdev_notifier *notifier_)
1441 struct netdev_linux_notifier *notifier =
1442 CONTAINER_OF(notifier_, struct netdev_linux_notifier, notifier);
1445 /* Remove 'notifier' from its list. */
1446 list = list_remove(¬ifier->node);
1447 if (list_is_empty(list)) {
1448 /* The list is now empty. Remove it from the hash and free it. */
1449 const char *netdev_name = netdev_get_name(notifier->notifier.netdev);
1450 shash_delete(&netdev_linux_notifiers,
1451 shash_find(&netdev_linux_notifiers, netdev_name));
1456 /* If that was the last notifier, unregister. */
1457 if (shash_is_empty(&netdev_linux_notifiers)) {
1458 rtnetlink_notifier_unregister(&netdev_linux_poll_notifier);
1462 const struct netdev_class netdev_linux_class = {
1463 "system", /* type */
1469 netdev_linux_create,
1470 netdev_linux_destroy,
1471 NULL, /* reconfigure */
1476 netdev_linux_enumerate,
1479 netdev_linux_recv_wait,
1483 netdev_linux_send_wait,
1485 netdev_linux_set_etheraddr,
1486 netdev_linux_get_etheraddr,
1487 netdev_linux_get_mtu,
1488 netdev_linux_get_ifindex,
1489 netdev_linux_get_carrier,
1490 netdev_linux_get_stats,
1492 netdev_linux_get_features,
1493 netdev_linux_set_advertisements,
1494 netdev_linux_get_vlan_vid,
1495 netdev_linux_set_policing,
1497 netdev_linux_get_in4,
1498 netdev_linux_set_in4,
1499 netdev_linux_get_in6,
1500 netdev_linux_add_router,
1501 netdev_linux_get_next_hop,
1502 netdev_linux_arp_lookup,
1504 netdev_linux_update_flags,
1506 netdev_linux_poll_add,
1507 netdev_linux_poll_remove,
1510 const struct netdev_class netdev_tap_class = {
1517 netdev_linux_create,
1518 netdev_linux_destroy,
1519 NULL, /* reconfigure */
1524 netdev_linux_enumerate,
1527 netdev_linux_recv_wait,
1531 netdev_linux_send_wait,
1533 netdev_linux_set_etheraddr,
1534 netdev_linux_get_etheraddr,
1535 netdev_linux_get_mtu,
1536 netdev_linux_get_ifindex,
1537 netdev_linux_get_carrier,
1538 netdev_linux_get_stats,
1540 netdev_linux_get_features,
1541 netdev_linux_set_advertisements,
1542 netdev_linux_get_vlan_vid,
1543 netdev_linux_set_policing,
1545 netdev_linux_get_in4,
1546 netdev_linux_set_in4,
1547 netdev_linux_get_in6,
1548 netdev_linux_add_router,
1549 netdev_linux_get_next_hop,
1550 netdev_linux_arp_lookup,
1552 netdev_linux_update_flags,
1554 netdev_linux_poll_add,
1555 netdev_linux_poll_remove,
1559 get_stats_via_netlink(int ifindex, struct netdev_stats *stats)
1561 /* Policy for RTNLGRP_LINK messages.
1563 * There are *many* more fields in these messages, but currently we only
1564 * care about these fields. */
1565 static const struct nl_policy rtnlgrp_link_policy[] = {
1566 [IFLA_IFNAME] = { .type = NL_A_STRING, .optional = false },
1567 [IFLA_STATS] = { .type = NL_A_UNSPEC, .optional = true,
1568 .min_len = sizeof(struct rtnl_link_stats) },
1572 static struct nl_sock *rtnl_sock;
1573 struct ofpbuf request;
1574 struct ofpbuf *reply;
1575 struct ifinfomsg *ifi;
1576 const struct rtnl_link_stats *rtnl_stats;
1577 struct nlattr *attrs[ARRAY_SIZE(rtnlgrp_link_policy)];
1581 error = nl_sock_create(NETLINK_ROUTE, 0, 0, 0, &rtnl_sock);
1583 VLOG_ERR_RL(&rl, "failed to create rtnetlink socket: %s",
1589 ofpbuf_init(&request, 0);
1590 nl_msg_put_nlmsghdr(&request, rtnl_sock, sizeof *ifi,
1591 RTM_GETLINK, NLM_F_REQUEST);
1592 ifi = ofpbuf_put_zeros(&request, sizeof *ifi);
1593 ifi->ifi_family = PF_UNSPEC;
1594 ifi->ifi_index = ifindex;
1595 error = nl_sock_transact(rtnl_sock, &request, &reply);
1596 ofpbuf_uninit(&request);
1601 if (!nl_policy_parse(reply, NLMSG_HDRLEN + sizeof(struct ifinfomsg),
1602 rtnlgrp_link_policy,
1603 attrs, ARRAY_SIZE(rtnlgrp_link_policy))) {
1604 ofpbuf_delete(reply);
1608 if (!attrs[IFLA_STATS]) {
1609 VLOG_WARN_RL(&rl, "RTM_GETLINK reply lacks stats");
1610 ofpbuf_delete(reply);
1614 rtnl_stats = nl_attr_get(attrs[IFLA_STATS]);
1615 stats->rx_packets = rtnl_stats->rx_packets;
1616 stats->tx_packets = rtnl_stats->tx_packets;
1617 stats->rx_bytes = rtnl_stats->rx_bytes;
1618 stats->tx_bytes = rtnl_stats->tx_bytes;
1619 stats->rx_errors = rtnl_stats->rx_errors;
1620 stats->tx_errors = rtnl_stats->tx_errors;
1621 stats->rx_dropped = rtnl_stats->rx_dropped;
1622 stats->tx_dropped = rtnl_stats->tx_dropped;
1623 stats->multicast = rtnl_stats->multicast;
1624 stats->collisions = rtnl_stats->collisions;
1625 stats->rx_length_errors = rtnl_stats->rx_length_errors;
1626 stats->rx_over_errors = rtnl_stats->rx_over_errors;
1627 stats->rx_crc_errors = rtnl_stats->rx_crc_errors;
1628 stats->rx_frame_errors = rtnl_stats->rx_frame_errors;
1629 stats->rx_fifo_errors = rtnl_stats->rx_fifo_errors;
1630 stats->rx_missed_errors = rtnl_stats->rx_missed_errors;
1631 stats->tx_aborted_errors = rtnl_stats->tx_aborted_errors;
1632 stats->tx_carrier_errors = rtnl_stats->tx_carrier_errors;
1633 stats->tx_fifo_errors = rtnl_stats->tx_fifo_errors;
1634 stats->tx_heartbeat_errors = rtnl_stats->tx_heartbeat_errors;
1635 stats->tx_window_errors = rtnl_stats->tx_window_errors;
1637 ofpbuf_delete(reply);
1643 get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats)
1645 static const char fn[] = "/proc/net/dev";
1650 stream = fopen(fn, "r");
1652 VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(errno));
1657 while (fgets(line, sizeof line, stream)) {
1660 #define X64 "%"SCNu64
1663 X64 X64 X64 X64 X64 X64 X64 "%*u"
1664 X64 X64 X64 X64 X64 X64 X64 "%*u",
1670 &stats->rx_fifo_errors,
1671 &stats->rx_frame_errors,
1677 &stats->tx_fifo_errors,
1679 &stats->tx_carrier_errors) != 15) {
1680 VLOG_WARN_RL(&rl, "%s:%d: parse error", fn, ln);
1681 } else if (!strcmp(devname, netdev_name)) {
1682 stats->rx_length_errors = UINT64_MAX;
1683 stats->rx_over_errors = UINT64_MAX;
1684 stats->rx_crc_errors = UINT64_MAX;
1685 stats->rx_missed_errors = UINT64_MAX;
1686 stats->tx_aborted_errors = UINT64_MAX;
1687 stats->tx_heartbeat_errors = UINT64_MAX;
1688 stats->tx_window_errors = UINT64_MAX;
1694 VLOG_WARN_RL(&rl, "%s: no stats for %s", fn, netdev_name);
1700 get_flags(const struct netdev *netdev, int *flags)
1705 error = netdev_linux_do_ioctl(netdev, &ifr, SIOCGIFFLAGS, "SIOCGIFFLAGS");
1706 *flags = ifr.ifr_flags;
1711 set_flags(struct netdev *netdev, int flags)
1715 ifr.ifr_flags = flags;
1716 return netdev_linux_do_ioctl(netdev, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS");
1720 do_get_ifindex(const char *netdev_name)
1724 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1725 COVERAGE_INC(netdev_get_ifindex);
1726 if (ioctl(af_inet_sock, SIOCGIFINDEX, &ifr) < 0) {
1727 VLOG_WARN_RL(&rl, "ioctl(SIOCGIFINDEX) on %s device failed: %s",
1728 netdev_name, strerror(errno));
1731 return ifr.ifr_ifindex;
1735 get_ifindex(const struct netdev *netdev_, int *ifindexp)
1737 struct netdev_linux *netdev = netdev_linux_cast(netdev_);
1739 if (!(netdev->cache->valid & VALID_IFINDEX)) {
1740 int ifindex = do_get_ifindex(netdev_get_name(netdev_));
1744 netdev->cache->valid |= VALID_IFINDEX;
1745 netdev->cache->ifindex = ifindex;
1747 *ifindexp = netdev->cache->ifindex;
1752 get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN])
1757 memset(&ifr, 0, sizeof ifr);
1758 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1759 COVERAGE_INC(netdev_get_hwaddr);
1760 if (ioctl(af_inet_sock, SIOCGIFHWADDR, &ifr) < 0) {
1761 VLOG_ERR("ioctl(SIOCGIFHWADDR) on %s device failed: %s",
1762 netdev_name, strerror(errno));
1765 hwaddr_family = ifr.ifr_hwaddr.sa_family;
1766 if (hwaddr_family != AF_UNSPEC && hwaddr_family != ARPHRD_ETHER) {
1767 VLOG_WARN("%s device has unknown hardware address family %d",
1768 netdev_name, hwaddr_family);
1770 memcpy(ea, ifr.ifr_hwaddr.sa_data, ETH_ADDR_LEN);
1775 set_etheraddr(const char *netdev_name, int hwaddr_family,
1776 const uint8_t mac[ETH_ADDR_LEN])
1780 memset(&ifr, 0, sizeof ifr);
1781 strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
1782 ifr.ifr_hwaddr.sa_family = hwaddr_family;
1783 memcpy(ifr.ifr_hwaddr.sa_data, mac, ETH_ADDR_LEN);
1784 COVERAGE_INC(netdev_set_hwaddr);
1785 if (ioctl(af_inet_sock, SIOCSIFHWADDR, &ifr) < 0) {
1786 VLOG_ERR("ioctl(SIOCSIFHWADDR) on %s device failed: %s",
1787 netdev_name, strerror(errno));
1794 netdev_linux_do_ethtool(struct netdev *netdev, struct ethtool_cmd *ecmd,
1795 int cmd, const char *cmd_name)
1799 memset(&ifr, 0, sizeof ifr);
1800 strncpy(ifr.ifr_name, netdev->name, sizeof ifr.ifr_name);
1801 ifr.ifr_data = (caddr_t) ecmd;
1804 COVERAGE_INC(netdev_ethtool);
1805 if (ioctl(af_inet_sock, SIOCETHTOOL, &ifr) == 0) {
1808 if (errno != EOPNOTSUPP) {
1809 VLOG_WARN_RL(&rl, "ethtool command %s on network device %s "
1810 "failed: %s", cmd_name, netdev->name,
1813 /* The device doesn't support this operation. That's pretty
1814 * common, so there's no point in logging anything. */
1821 netdev_linux_do_ioctl(const struct netdev *netdev, struct ifreq *ifr,
1822 int cmd, const char *cmd_name)
1824 strncpy(ifr->ifr_name, netdev_get_name(netdev), sizeof ifr->ifr_name);
1825 if (ioctl(af_inet_sock, cmd, ifr) == -1) {
1826 VLOG_DBG_RL(&rl, "%s: ioctl(%s) failed: %s",
1827 netdev_get_name(netdev), cmd_name, strerror(errno));
1834 netdev_linux_get_ipv4(const struct netdev *netdev, struct in_addr *ip,
1835 int cmd, const char *cmd_name)
1840 ifr.ifr_addr.sa_family = AF_INET;
1841 error = netdev_linux_do_ioctl(netdev, &ifr, cmd, cmd_name);
1843 const struct sockaddr_in *sin = (struct sockaddr_in *) &ifr.ifr_addr;
1844 *ip = sin->sin_addr;