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"
53 #include "ovs-thread.h"
55 #include "poll-loop.h"
56 #include "socket-util.h"
62 VLOG_DEFINE_THIS_MODULE(netdev_bsd
);
65 struct netdev_rx_bsd
{
68 /* Packet capture descriptor for a system network device.
69 * For a tap device this is NULL. */
72 /* Selectable file descriptor for the network device.
73 * This descriptor will be used for polling operations. */
79 unsigned int cache_valid
;
80 unsigned int change_seq
;
83 uint8_t etheraddr
[ETH_ADDR_LEN
];
85 struct in_addr netmask
;
90 int tap_fd
; /* TAP character device, if any, otherwise -1. */
92 /* Used for sending packets on non-tap devices. */
101 VALID_IFINDEX
= 1 << 0,
102 VALID_ETHERADDR
= 1 << 1,
106 VALID_CARRIER
= 1 << 5
109 #define PCAP_SNAPLEN 2048
113 * Notifier used to invalidate device informations in case of status change.
115 * It will be registered with a 'rtbsd_notifier_register()' when the first
116 * device will be created with the call of either 'netdev_bsd_tap_create()' or
117 * 'netdev_bsd_system_create()'.
119 * The callback associated with this notifier ('netdev_bsd_cache_cb()') will
120 * invalidate cached information about the device.
122 static struct rtbsd_notifier netdev_bsd_cache_notifier
;
123 static int cache_notifier_refcount
;
125 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 20);
127 static void destroy_tap(int fd
, const char *name
);
128 static int get_flags(const struct netdev
*, int *flagsp
);
129 static int set_flags(const char *, int flags
);
130 static int do_set_addr(struct netdev
*netdev
,
131 int ioctl_nr
, const char *ioctl_name
,
132 struct in_addr addr
);
133 static int get_etheraddr(const char *netdev_name
, uint8_t ea
[ETH_ADDR_LEN
]);
134 static int set_etheraddr(const char *netdev_name
, int hwaddr_family
,
135 int hwaddr_len
, const uint8_t[ETH_ADDR_LEN
]);
136 static int get_ifindex(const struct netdev
*, int *ifindexp
);
138 static int ifr_get_flags(const struct ifreq
*);
139 static void ifr_set_flags(struct ifreq
*, int flags
);
142 static int af_link_ioctl(int command
, const void *arg
);
145 static void netdev_bsd_run(void);
148 is_netdev_bsd_class(const struct netdev_class
*netdev_class
)
150 return netdev_class
->run
== netdev_bsd_run
;
153 static struct netdev_bsd
*
154 netdev_bsd_cast(const struct netdev
*netdev
)
156 ovs_assert(is_netdev_bsd_class(netdev_get_class(netdev
)));
157 return CONTAINER_OF(netdev
, struct netdev_bsd
, up
);
160 static struct netdev_rx_bsd
*
161 netdev_rx_bsd_cast(const struct netdev_rx
*rx
)
163 ovs_assert(is_netdev_bsd_class(netdev_get_class(rx
->netdev
)));
164 return CONTAINER_OF(rx
, struct netdev_rx_bsd
, up
);
168 netdev_get_kernel_name(const struct netdev
*netdev
)
170 return netdev_bsd_cast(netdev
)->kernel_name
;
174 * Perform periodic work needed by netdev. In BSD netdevs it checks for any
175 * interface status changes, and eventually calls all the user callbacks.
180 rtbsd_notifier_run();
184 * Arranges for poll_block() to wake up if the "run" member function needs to
188 netdev_bsd_wait(void)
190 rtbsd_notifier_wait();
194 netdev_bsd_changed(struct netdev_bsd
*dev
)
197 if (!dev
->change_seq
) {
202 /* Invalidate cache in case of interface status change. */
204 netdev_bsd_cache_cb(const struct rtbsd_change
*change
,
205 void *aux OVS_UNUSED
)
207 struct netdev_bsd
*dev
;
210 struct netdev
*base_dev
= netdev_from_name(change
->if_name
);
213 const struct netdev_class
*netdev_class
=
214 netdev_get_class(base_dev
);
216 if (is_netdev_bsd_class(netdev_class
)) {
217 dev
= netdev_bsd_cast(base_dev
);
218 dev
->cache_valid
= 0;
219 netdev_bsd_changed(dev
);
221 netdev_close(base_dev
);
225 * XXX the API is lacking, we should be able to iterate on the list of
226 * netdevs without having to store the info in a temp shash.
228 struct shash device_shash
;
229 struct shash_node
*node
;
231 shash_init(&device_shash
);
232 netdev_get_devices(&netdev_bsd_class
, &device_shash
);
233 SHASH_FOR_EACH (node
, &device_shash
) {
234 struct netdev
*netdev
= node
->data
;
235 dev
= netdev_bsd_cast(netdev
);
236 dev
->cache_valid
= 0;
237 netdev_bsd_changed(dev
);
238 netdev_close(netdev
);
240 shash_destroy(&device_shash
);
245 cache_notifier_ref(void)
249 if (!cache_notifier_refcount
) {
250 ret
= rtbsd_notifier_register(&netdev_bsd_cache_notifier
,
251 netdev_bsd_cache_cb
, NULL
);
256 cache_notifier_refcount
++;
261 cache_notifier_unref(void)
263 cache_notifier_refcount
--;
264 if (cache_notifier_refcount
== 0) {
265 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier
);
270 static struct netdev
*
271 netdev_bsd_alloc(void)
273 struct netdev_bsd
*netdev
= xzalloc(sizeof *netdev
);
278 netdev_bsd_construct_system(struct netdev
*netdev_
)
280 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
281 enum netdev_flags flags
;
284 error
= cache_notifier_ref();
289 netdev
->change_seq
= 1;
291 netdev
->kernel_name
= xstrdup(netdev_
->name
);
293 /* Verify that the netdev really exists by attempting to read its flags */
294 error
= netdev_get_flags(netdev_
, &flags
);
295 if (error
== ENXIO
) {
296 free(netdev
->kernel_name
);
297 cache_notifier_unref();
305 netdev_bsd_construct_tap(struct netdev
*netdev_
)
307 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
308 const char *name
= netdev_
->name
;
311 char *kernel_name
= NULL
;
313 error
= cache_notifier_ref();
318 memset(&ifr
, 0, sizeof(ifr
));
320 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
321 * to retrieve the name of the tap device. */
322 netdev
->tap_fd
= open("/dev/tap", O_RDWR
);
323 netdev
->change_seq
= 1;
324 if (netdev
->tap_fd
< 0) {
326 VLOG_WARN("opening \"/dev/tap\" failed: %s", ovs_strerror(error
));
327 goto error_unref_notifier
;
330 /* Retrieve tap name (e.g. tap0) */
331 if (ioctl(netdev
->tap_fd
, TAPGIFNAME
, &ifr
) == -1) {
332 /* XXX Need to destroy the device? */
334 close(netdev
->tap_fd
);
335 goto error_unref_notifier
;
338 /* Change the name of the tap device */
339 #if defined(SIOCSIFNAME)
340 ifr
.ifr_data
= (void *)name
;
341 error
= af_inet_ioctl(SIOCSIFNAME
, &ifr
);
343 destroy_tap(netdev
->tap_fd
, ifr
.ifr_name
);
344 goto error_unref_notifier
;
346 kernel_name
= xstrdup(name
);
349 * NetBSD doesn't support inteface renaming.
351 VLOG_INFO("tap %s is created for bridge %s", ifr
.ifr_name
, name
);
352 kernel_name
= xstrdup(ifr
.ifr_name
);
355 /* set non-blocking. */
356 error
= set_nonblocking(netdev
->tap_fd
);
358 destroy_tap(netdev
->tap_fd
, kernel_name
);
359 goto error_unref_notifier
;
363 ifr_set_flags(&ifr
, IFF_UP
);
364 strncpy(ifr
.ifr_name
, kernel_name
, sizeof ifr
.ifr_name
);
365 error
= af_inet_ioctl(SIOCSIFFLAGS
, &ifr
);
367 destroy_tap(netdev
->tap_fd
, kernel_name
);
368 goto error_unref_notifier
;
371 netdev
->kernel_name
= kernel_name
;
375 error_unref_notifier
:
376 cache_notifier_unref();
383 netdev_bsd_destruct(struct netdev
*netdev_
)
385 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
387 cache_notifier_unref();
389 if (netdev
->tap_fd
>= 0) {
390 destroy_tap(netdev
->tap_fd
, netdev_get_kernel_name(netdev_
));
393 pcap_close(netdev
->pcap
);
395 free(netdev
->kernel_name
);
399 netdev_bsd_dealloc(struct netdev
*netdev_
)
401 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
407 netdev_bsd_open_pcap(const char *name
, pcap_t
**pcapp
, int *fdp
)
409 char errbuf
[PCAP_ERRBUF_SIZE
];
415 /* Open the pcap device. The device is opened in non-promiscuous mode
416 * because the interface flags are manually set by the caller. */
418 pcap
= pcap_open_live(name
, PCAP_SNAPLEN
, 0, 1000, errbuf
);
420 VLOG_ERR_RL(&rl
, "%s: pcap_open_live failed: %s", name
, errbuf
);
424 if (errbuf
[0] != '\0') {
425 VLOG_WARN_RL(&rl
, "%s: pcap_open_live: %s", name
, errbuf
);
428 /* Get the underlying fd. */
429 fd
= pcap_get_selectable_fd(pcap
);
431 VLOG_WARN_RL(&rl
, "%s: no selectable file descriptor", name
);
436 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
437 * on the file descriptor returned by pcap_get_selectable_fd to achieve
438 * a real non-blocking behaviour.*/
439 error
= pcap_setnonblock(pcap
, 1, errbuf
);
445 /* This call assure that reads return immediately upon packet
446 * reception. Otherwise, a read will block until either the kernel
447 * buffer becomes full or a timeout occurs. */
448 if (ioctl(fd
, BIOCIMMEDIATE
, &one
) < 0 ) {
449 VLOG_ERR_RL(&rl
, "ioctl(BIOCIMMEDIATE) on %s device failed: %s",
450 name
, ovs_strerror(errno
));
455 /* Capture only incoming packets. */
456 error
= pcap_setdirection(pcap
, PCAP_D_IN
);
475 static struct netdev_rx
*
476 netdev_bsd_rx_alloc(void)
478 struct netdev_rx_bsd
*rx
= xzalloc(sizeof *rx
);
483 netdev_bsd_rx_construct(struct netdev_rx
*rx_
)
485 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
486 struct netdev
*netdev_
= rx
->up
.netdev
;
487 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
489 if (!strcmp(netdev_get_type(netdev_
), "tap")) {
490 rx
->pcap_handle
= NULL
;
491 rx
->fd
= netdev
->tap_fd
;
493 int error
= netdev_bsd_open_pcap(netdev_get_kernel_name(netdev_
),
494 &rx
->pcap_handle
, &rx
->fd
);
499 netdev_bsd_changed(netdev
);
506 netdev_bsd_rx_destruct(struct netdev_rx
*rx_
)
508 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
510 if (rx
->pcap_handle
) {
511 pcap_close(rx
->pcap_handle
);
516 netdev_bsd_rx_dealloc(struct netdev_rx
*rx_
)
518 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
523 /* The recv callback of the netdev class returns the number of bytes of the
526 * This can be done by the pcap_next() function. Unfortunately pcap_next() does
527 * not make difference between a missing packet on the capture interface and
528 * an error during the file capture. We can use the pcap_dispatch() function
529 * instead, which is able to distinguish between errors and null packet.
531 * To make pcap_dispatch() returns the number of bytes read from the interface
532 * we need to define the following callback and argument.
541 * This callback will be executed on every captured packet.
543 * If the packet captured by pcap_dispatch() does not fit the pcap buffer,
544 * pcap returns a truncated packet and we follow this behavior.
546 * The argument args->retval is the packet size in bytes.
549 proc_pkt(u_char
*args_
, const struct pcap_pkthdr
*hdr
, const u_char
*packet
)
551 struct pcap_arg
*args
= (struct pcap_arg
*)args_
;
553 if (args
->size
< hdr
->len
) {
554 VLOG_WARN_RL(&rl
, "packet truncated");
555 args
->retval
= args
->size
;
557 args
->retval
= hdr
->len
;
560 /* copy the packet to our buffer */
561 memcpy(args
->data
, packet
, args
->retval
);
565 * This function attempts to receive a packet from the specified network
566 * device. It is assumed that the network device is a system device or a tap
567 * device opened as a system one. In this case the read operation is performed
571 netdev_rx_bsd_recv_pcap(struct netdev_rx_bsd
*rx
, void *data
, size_t size
)
576 /* prepare the pcap argument to store the packet */
581 ret
= pcap_dispatch(rx
->pcap_handle
, 1, proc_pkt
, (u_char
*) &arg
);
584 return arg
.retval
; /* arg.retval < 0 is handled in the caller */
587 if (errno
== EINTR
) {
597 * This function attempts to receive a packet from the specified network
598 * device. It is assumed that the network device is a tap device and
599 * 'rx->fd' is initialized with the tap file descriptor.
602 netdev_rx_bsd_recv_tap(struct netdev_rx_bsd
*rx
, void *data
, size_t size
)
605 ssize_t retval
= read(rx
->fd
, data
, size
);
608 } else if (errno
!= EINTR
) {
609 if (errno
!= EAGAIN
) {
610 VLOG_WARN_RL(&rl
, "error receiving Ethernet packet on %s: %s",
611 ovs_strerror(errno
), netdev_rx_get_name(&rx
->up
));
619 netdev_bsd_rx_recv(struct netdev_rx
*rx_
, void *data
, size_t size
)
621 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
623 return (rx
->pcap_handle
624 ? netdev_rx_bsd_recv_pcap(rx
, data
, size
)
625 : netdev_rx_bsd_recv_tap(rx
, data
, size
));
629 * Registers with the poll loop to wake up from the next call to poll_block()
630 * when a packet is ready to be received with netdev_rx_recv() on 'rx'.
633 netdev_bsd_rx_wait(struct netdev_rx
*rx_
)
635 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
637 poll_fd_wait(rx
->fd
, POLLIN
);
640 /* Discards all packets waiting to be received from 'rx'. */
642 netdev_bsd_rx_drain(struct netdev_rx
*rx_
)
645 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
647 strcpy(ifr
.ifr_name
, netdev_get_kernel_name(netdev_rx_get_netdev(rx_
)));
648 if (ioctl(rx
->fd
, BIOCFLUSH
, &ifr
) == -1) {
649 VLOG_DBG_RL(&rl
, "%s: ioctl(BIOCFLUSH) failed: %s",
650 netdev_rx_get_name(rx_
), ovs_strerror(errno
));
657 * Send a packet on the specified network device. The device could be either a
658 * system or a tap device.
661 netdev_bsd_send(struct netdev
*netdev_
, const void *data
, size_t size
)
663 struct netdev_bsd
*dev
= netdev_bsd_cast(netdev_
);
664 const char *name
= netdev_get_name(netdev_
);
666 if (dev
->tap_fd
< 0 && !dev
->pcap
) {
667 int error
= netdev_bsd_open_pcap(name
, &dev
->pcap
, &dev
->fd
);
675 if (dev
->tap_fd
>= 0) {
676 retval
= write(dev
->tap_fd
, data
, size
);
678 retval
= pcap_inject(dev
->pcap
, data
, size
);
681 if (errno
== EINTR
) {
683 } else if (errno
!= EAGAIN
) {
684 VLOG_WARN_RL(&rl
, "error sending Ethernet packet on %s: %s",
685 name
, ovs_strerror(errno
));
688 } else if (retval
!= size
) {
689 VLOG_WARN_RL(&rl
, "sent partial Ethernet packet (%zd bytes of "
690 "%zu) on %s", retval
, size
, name
);
699 * Registers with the poll loop to wake up from the next call to poll_block()
700 * when the packet transmission queue has sufficient room to transmit a packet
701 * with netdev_send().
704 netdev_bsd_send_wait(struct netdev
*netdev_
)
706 struct netdev_bsd
*dev
= netdev_bsd_cast(netdev_
);
708 if (dev
->tap_fd
>= 0) {
709 /* TAP device always accepts packets. */
710 poll_immediate_wake();
711 } else if (dev
->pcap
) {
712 poll_fd_wait(dev
->fd
, POLLOUT
);
714 /* We haven't even tried to send a packet yet. */
715 poll_immediate_wake();
720 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
721 * otherwise a positive errno value.
724 netdev_bsd_set_etheraddr(struct netdev
*netdev_
,
725 const uint8_t mac
[ETH_ADDR_LEN
])
727 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
730 if (!(netdev
->cache_valid
& VALID_ETHERADDR
)
731 || !eth_addr_equals(netdev
->etheraddr
, mac
)) {
732 error
= set_etheraddr(netdev_get_kernel_name(netdev_
), AF_LINK
,
735 netdev
->cache_valid
|= VALID_ETHERADDR
;
736 memcpy(netdev
->etheraddr
, mac
, ETH_ADDR_LEN
);
737 netdev_bsd_changed(netdev
);
746 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
747 * free the returned buffer.
750 netdev_bsd_get_etheraddr(const struct netdev
*netdev_
,
751 uint8_t mac
[ETH_ADDR_LEN
])
753 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
755 if (!(netdev
->cache_valid
& VALID_ETHERADDR
)) {
756 int error
= get_etheraddr(netdev_get_kernel_name(netdev_
),
761 netdev
->cache_valid
|= VALID_ETHERADDR
;
763 memcpy(mac
, netdev
->etheraddr
, ETH_ADDR_LEN
);
769 * Returns the maximum size of transmitted (and received) packets on 'netdev',
770 * in bytes, not including the hardware header; thus, this is typically 1500
771 * bytes for Ethernet devices.
774 netdev_bsd_get_mtu(const struct netdev
*netdev_
, int *mtup
)
776 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
778 if (!(netdev
->cache_valid
& VALID_MTU
)) {
782 error
= af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_
), &ifr
,
783 SIOCGIFMTU
, "SIOCGIFMTU");
787 netdev
->mtu
= ifr
.ifr_mtu
;
788 netdev
->cache_valid
|= VALID_MTU
;
796 netdev_bsd_get_ifindex(const struct netdev
*netdev
)
800 error
= get_ifindex(netdev
, &ifindex
);
801 return error
? -error
: ifindex
;
805 netdev_bsd_get_carrier(const struct netdev
*netdev_
, bool *carrier
)
807 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
809 if (!(netdev
->cache_valid
& VALID_CARRIER
)) {
810 struct ifmediareq ifmr
;
813 memset(&ifmr
, 0, sizeof(ifmr
));
814 strncpy(ifmr
.ifm_name
, netdev_get_kernel_name(netdev_
),
815 sizeof ifmr
.ifm_name
);
817 error
= af_inet_ioctl(SIOCGIFMEDIA
, &ifmr
);
819 VLOG_DBG_RL(&rl
, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
820 netdev_get_name(netdev_
), ovs_strerror(error
));
824 netdev
->carrier
= (ifmr
.ifm_status
& IFM_ACTIVE
) == IFM_ACTIVE
;
825 netdev
->cache_valid
|= VALID_CARRIER
;
827 /* If the interface doesn't report whether the media is active,
828 * just assume it is active. */
829 if ((ifmr
.ifm_status
& IFM_AVALID
) == 0) {
830 netdev
->carrier
= true;
833 *carrier
= netdev
->carrier
;
839 convert_stats(struct netdev_stats
*stats
, const struct if_data
*ifd
)
842 * note: UINT64_MAX means unsupported
844 stats
->rx_packets
= ifd
->ifi_ipackets
;
845 stats
->tx_packets
= ifd
->ifi_opackets
;
846 stats
->rx_bytes
= ifd
->ifi_obytes
;
847 stats
->tx_bytes
= ifd
->ifi_ibytes
;
848 stats
->rx_errors
= ifd
->ifi_ierrors
;
849 stats
->tx_errors
= ifd
->ifi_oerrors
;
850 stats
->rx_dropped
= ifd
->ifi_iqdrops
;
851 stats
->tx_dropped
= UINT64_MAX
;
852 stats
->multicast
= ifd
->ifi_imcasts
;
853 stats
->collisions
= ifd
->ifi_collisions
;
854 stats
->rx_length_errors
= UINT64_MAX
;
855 stats
->rx_over_errors
= UINT64_MAX
;
856 stats
->rx_crc_errors
= UINT64_MAX
;
857 stats
->rx_frame_errors
= UINT64_MAX
;
858 stats
->rx_fifo_errors
= UINT64_MAX
;
859 stats
->rx_missed_errors
= UINT64_MAX
;
860 stats
->tx_aborted_errors
= UINT64_MAX
;
861 stats
->tx_carrier_errors
= UINT64_MAX
;
862 stats
->tx_fifo_errors
= UINT64_MAX
;
863 stats
->tx_heartbeat_errors
= UINT64_MAX
;
864 stats
->tx_window_errors
= UINT64_MAX
;
867 /* Retrieves current device stats for 'netdev'. */
869 netdev_bsd_get_stats(const struct netdev
*netdev_
, struct netdev_stats
*stats
)
871 #if defined(__FreeBSD__)
875 struct ifmibdata ifmd
;
880 mib
[2] = NETLINK_GENERIC
;
881 mib
[3] = IFMIB_SYSTEM
;
882 mib
[4] = IFMIB_IFCOUNT
;
884 len
= sizeof(if_count
);
886 if (sysctl(mib
, 5, &if_count
, &len
, (void *)0, 0) == -1) {
887 VLOG_DBG_RL(&rl
, "%s: sysctl failed: %s",
888 netdev_get_name(netdev_
), ovs_strerror(errno
));
892 mib
[5] = IFDATA_GENERAL
;
893 mib
[3] = IFMIB_IFDATA
;
895 for (i
= 1; i
<= if_count
; i
++) {
897 if (sysctl(mib
, 6, &ifmd
, &len
, (void *)0, 0) == -1) {
898 VLOG_DBG_RL(&rl
, "%s: sysctl failed: %s",
899 netdev_get_name(netdev_
), ovs_strerror(errno
));
901 } else if (!strcmp(ifmd
.ifmd_name
, netdev_get_name(netdev_
))) {
902 convert_stats(stats
, &ifmd
.ifmd_data
);
908 #elif defined(__NetBSD__)
909 struct ifdatareq ifdr
;
912 memset(&ifdr
, 0, sizeof(ifdr
));
913 strncpy(ifdr
.ifdr_name
, netdev_get_kernel_name(netdev_
),
914 sizeof(ifdr
.ifdr_name
));
915 error
= af_link_ioctl(SIOCGIFDATA
, &ifdr
);
917 convert_stats(stats
, &ifdr
.ifdr_data
);
921 #error not implemented
926 netdev_bsd_parse_media(int media
)
928 uint32_t supported
= 0;
929 bool half_duplex
= media
& IFM_HDX
? true : false;
931 switch (IFM_SUBTYPE(media
)) {
936 supported
|= half_duplex
? NETDEV_F_10MB_HD
: NETDEV_F_10MB_FD
;
937 supported
|= NETDEV_F_COPPER
;
941 supported
|= half_duplex
? NETDEV_F_10MB_HD
: NETDEV_F_10MB_FD
;
942 supported
|= NETDEV_F_FIBER
;
949 supported
|= half_duplex
? NETDEV_F_100MB_HD
: NETDEV_F_100MB_FD
;
950 supported
|= NETDEV_F_COPPER
;
954 supported
|= half_duplex
? NETDEV_F_100MB_HD
: NETDEV_F_100MB_FD
;
955 supported
|= NETDEV_F_FIBER
;
960 supported
|= half_duplex
? NETDEV_F_1GB_HD
: NETDEV_F_1GB_FD
;
961 supported
|= NETDEV_F_COPPER
;
966 supported
|= half_duplex
? NETDEV_F_1GB_HD
: NETDEV_F_1GB_FD
;
967 supported
|= NETDEV_F_FIBER
;
971 supported
|= NETDEV_F_10GB_FD
;
972 supported
|= NETDEV_F_COPPER
;
977 supported
|= NETDEV_F_10GB_FD
;
978 supported
|= NETDEV_F_FIBER
;
985 if (IFM_SUBTYPE(media
) == IFM_AUTO
) {
986 supported
|= NETDEV_F_AUTONEG
;
989 if (media & IFM_ETH_FMASK) {
990 supported |= NETDEV_F_PAUSE;
998 * Stores the features supported by 'netdev' into each of '*current',
999 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1000 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1001 * successful, otherwise a positive errno value. On failure, all of the
1002 * passed-in values are set to 0.
1005 netdev_bsd_get_features(const struct netdev
*netdev
,
1006 enum netdev_features
*current
, uint32_t *advertised
,
1007 enum netdev_features
*supported
, uint32_t *peer
)
1009 struct ifmediareq ifmr
;
1015 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1017 memset(&ifmr
, 0, sizeof(ifmr
));
1018 strncpy(ifmr
.ifm_name
, netdev_get_name(netdev
), sizeof ifmr
.ifm_name
);
1020 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1021 * number of supported modes, and a second with a buffer to retrieve
1023 error
= af_inet_ioctl(SIOCGIFMEDIA
, &ifmr
);
1025 VLOG_DBG_RL(&rl
, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1026 netdev_get_name(netdev
), ovs_strerror(error
));
1030 media_list
= xcalloc(ifmr
.ifm_count
, sizeof(int));
1031 ifmr
.ifm_ulist
= media_list
;
1033 if (IFM_TYPE(ifmr
.ifm_current
) != IFM_ETHER
) {
1034 VLOG_DBG_RL(&rl
, "%s: doesn't appear to be ethernet",
1035 netdev_get_name(netdev
));
1040 error
= af_inet_ioctl(SIOCGIFMEDIA
, &ifmr
);
1042 VLOG_DBG_RL(&rl
, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1043 netdev_get_name(netdev
), ovs_strerror(error
));
1047 /* Current settings. */
1048 *current
= netdev_bsd_parse_media(ifmr
.ifm_active
);
1050 /* Advertised features. */
1051 *advertised
= netdev_bsd_parse_media(ifmr
.ifm_current
);
1053 /* Supported features. */
1055 for (i
= 0; i
< ifmr
.ifm_count
; i
++) {
1056 *supported
|= netdev_bsd_parse_media(ifmr
.ifm_ulist
[i
]);
1059 /* Peer advertisements. */
1060 *peer
= 0; /* XXX */
1069 * If 'netdev' has an assigned IPv4 address, sets '*in4' to that address and
1070 * '*netmask' to its netmask and returns true. Otherwise, returns false.
1073 netdev_bsd_get_in4(const struct netdev
*netdev_
, struct in_addr
*in4
,
1074 struct in_addr
*netmask
)
1076 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
1078 if (!(netdev
->cache_valid
& VALID_IN4
)) {
1079 const struct sockaddr_in
*sin
;
1083 ifr
.ifr_addr
.sa_family
= AF_INET
;
1084 error
= af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_
), &ifr
,
1085 SIOCGIFADDR
, "SIOCGIFADDR");
1090 sin
= (struct sockaddr_in
*) &ifr
.ifr_addr
;
1091 netdev
->in4
= sin
->sin_addr
;
1092 error
= af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev_
), &ifr
,
1093 SIOCGIFNETMASK
, "SIOCGIFNETMASK");
1097 netdev
->netmask
= sin
->sin_addr
;
1098 netdev
->cache_valid
|= VALID_IN4
;
1101 *netmask
= netdev
->netmask
;
1103 return in4
->s_addr
== INADDR_ANY
? EADDRNOTAVAIL
: 0;
1107 * Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
1108 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
1109 * positive errno value.
1112 netdev_bsd_set_in4(struct netdev
*netdev_
, struct in_addr addr
,
1113 struct in_addr mask
)
1115 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
1118 error
= do_set_addr(netdev_
, SIOCSIFADDR
, "SIOCSIFADDR", addr
);
1120 if (addr
.s_addr
!= INADDR_ANY
) {
1121 error
= do_set_addr(netdev_
, SIOCSIFNETMASK
,
1122 "SIOCSIFNETMASK", mask
);
1124 netdev
->cache_valid
|= VALID_IN4
;
1126 netdev
->netmask
= mask
;
1129 netdev_bsd_changed(netdev
);
1135 netdev_bsd_get_in6(const struct netdev
*netdev_
, struct in6_addr
*in6
)
1137 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
1138 if (!(netdev
->cache_valid
& VALID_IN6
)) {
1139 struct ifaddrs
*ifa
, *head
;
1140 struct sockaddr_in6
*sin6
;
1141 const char *netdev_name
= netdev_get_name(netdev_
);
1143 if (getifaddrs(&head
) != 0) {
1144 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name
,
1145 ovs_strerror(errno
));
1149 for (ifa
= head
; ifa
; ifa
= ifa
->ifa_next
) {
1150 if (ifa
->ifa_addr
->sa_family
== AF_INET6
&&
1151 !strcmp(ifa
->ifa_name
, netdev_name
)) {
1152 sin6
= (struct sockaddr_in6
*)ifa
->ifa_addr
;
1154 memcpy(&netdev
->in6
, &sin6
->sin6_addr
, sin6
->sin6_len
);
1155 netdev
->cache_valid
|= VALID_IN6
;
1162 return EADDRNOTAVAIL
;
1168 #if defined(__NetBSD__)
1170 netdev_bsd_kernel_name_to_ovs_name(const char *kernel_name
)
1172 char *ovs_name
= NULL
;
1173 struct shash device_shash
;
1174 struct shash_node
*node
;
1176 shash_init(&device_shash
);
1177 netdev_get_devices(&netdev_tap_class
, &device_shash
);
1178 SHASH_FOR_EACH(node
, &device_shash
) {
1179 struct netdev
*netdev
= node
->data
;
1180 struct netdev_bsd
* const dev
= netdev_bsd_cast(netdev
);
1182 if (!strcmp(dev
->kernel_name
, kernel_name
)) {
1184 ovs_name
= xstrdup(netdev_get_name(&dev
->up
));
1186 netdev_close(netdev
);
1188 shash_destroy(&device_shash
);
1190 return ovs_name
? ovs_name
: xstrdup(kernel_name
);
1195 netdev_bsd_get_next_hop(const struct in_addr
*host OVS_UNUSED
,
1196 struct in_addr
*next_hop OVS_UNUSED
,
1197 char **netdev_name OVS_UNUSED
)
1199 #if defined(__NetBSD__)
1201 struct sockaddr_in sin
;
1202 struct sockaddr_dl sdl
;
1209 struct rt_msghdr
*rtm
= &buf
.h
;
1210 const pid_t pid
= getpid();
1213 bool gateway
= false;
1214 char *ifname
= NULL
;
1217 memset(next_hop
, 0, sizeof(*next_hop
));
1218 *netdev_name
= NULL
;
1220 memset(&sin
, 0, sizeof(sin
));
1221 sin
.sin_len
= sizeof(sin
);
1222 sin
.sin_family
= AF_INET
;
1224 sin
.sin_addr
= *host
;
1226 memset(&sdl
, 0, sizeof(sdl
));
1227 sdl
.sdl_len
= sizeof(sdl
);
1228 sdl
.sdl_family
= AF_LINK
;
1230 s
= socket(PF_ROUTE
, SOCK_RAW
, 0);
1231 memset(&buf
, 0, sizeof(buf
));
1232 rtm
->rtm_flags
= RTF_HOST
|RTF_UP
;
1233 rtm
->rtm_version
= RTM_VERSION
;
1234 rtm
->rtm_addrs
= RTA_DST
|RTA_IFP
;
1235 cp
= (void *)&buf
.space
;
1236 memcpy(cp
, &sin
, sizeof(sin
));
1237 RT_ADVANCE(cp
, (struct sockaddr
*)(void *)&sin
);
1238 memcpy(cp
, &sdl
, sizeof(sdl
));
1239 RT_ADVANCE(cp
, (struct sockaddr
*)(void *)&sdl
);
1240 rtm
->rtm_msglen
= cp
- (char *)(void *)rtm
;
1241 rtm
->rtm_seq
= ++seq
;
1242 rtm
->rtm_type
= RTM_GET
;
1244 write(s
, rtm
, rtm
->rtm_msglen
);
1245 memset(&buf
, 0, sizeof(buf
));
1247 ssz
= read(s
, &buf
, sizeof(buf
));
1248 } while (ssz
> 0 && (rtm
->rtm_seq
!= seq
|| rtm
->rtm_pid
!= pid
));
1249 saved_errno
= errno
;
1255 return EPIPE
; /* XXX */
1257 cp
= (void *)&buf
.space
;
1258 for (i
= 1; i
; i
<<= 1) {
1259 if ((rtm
->rtm_addrs
& i
) != 0) {
1260 const struct sockaddr
*sa
= (const void *)cp
;
1262 if ((i
== RTA_GATEWAY
) && sa
->sa_family
== AF_INET
) {
1263 const struct sockaddr_in
* const sin
=
1264 (const struct sockaddr_in
*)sa
;
1266 *next_hop
= sin
->sin_addr
;
1269 if ((i
== RTA_IFP
) && sa
->sa_family
== AF_LINK
) {
1270 const struct sockaddr_dl
* const sdl
=
1271 (const struct sockaddr_dl
*)sa
;
1274 kernel_name
= xmemdup0(sdl
->sdl_data
, sdl
->sdl_nlen
);
1275 ifname
= netdev_bsd_kernel_name_to_ovs_name(kernel_name
);
1281 if (ifname
== NULL
) {
1287 *netdev_name
= ifname
;
1288 VLOG_DBG("host " IP_FMT
" next-hop " IP_FMT
" if %s",
1289 IP_ARGS(host
->s_addr
), IP_ARGS(next_hop
->s_addr
), *netdev_name
);
1297 make_in4_sockaddr(struct sockaddr
*sa
, struct in_addr addr
)
1299 struct sockaddr_in sin
;
1300 memset(&sin
, 0, sizeof sin
);
1301 sin
.sin_family
= AF_INET
;
1302 sin
.sin_addr
= addr
;
1305 memset(sa
, 0, sizeof *sa
);
1306 memcpy(sa
, &sin
, sizeof sin
);
1310 do_set_addr(struct netdev
*netdev
,
1311 int ioctl_nr
, const char *ioctl_name
, struct in_addr addr
)
1314 make_in4_sockaddr(&ifr
.ifr_addr
, addr
);
1315 return af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev
), &ifr
, ioctl_nr
,
1320 nd_to_iff_flags(enum netdev_flags nd
)
1323 if (nd
& NETDEV_UP
) {
1326 if (nd
& NETDEV_PROMISC
) {
1328 #if defined(IFF_PPROMISC)
1329 iff
|= IFF_PPROMISC
;
1336 iff_to_nd_flags(int iff
)
1338 enum netdev_flags nd
= 0;
1342 if (iff
& IFF_PROMISC
) {
1343 nd
|= NETDEV_PROMISC
;
1349 netdev_bsd_update_flags(struct netdev
*netdev_
, enum netdev_flags off
,
1350 enum netdev_flags on
, enum netdev_flags
*old_flagsp
)
1352 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
1353 int old_flags
, new_flags
;
1356 error
= get_flags(netdev_
, &old_flags
);
1358 *old_flagsp
= iff_to_nd_flags(old_flags
);
1359 new_flags
= (old_flags
& ~nd_to_iff_flags(off
)) | nd_to_iff_flags(on
);
1360 if (new_flags
!= old_flags
) {
1361 error
= set_flags(netdev_get_kernel_name(netdev_
), new_flags
);
1362 netdev_bsd_changed(netdev
);
1369 netdev_bsd_change_seq(const struct netdev
*netdev
)
1371 return netdev_bsd_cast(netdev
)->change_seq
;
1375 const struct netdev_class netdev_bsd_class
= {
1382 netdev_bsd_construct_system
,
1383 netdev_bsd_destruct
,
1385 NULL
, /* get_config */
1386 NULL
, /* set_config */
1387 NULL
, /* get_tunnel_config */
1390 netdev_bsd_send_wait
,
1392 netdev_bsd_set_etheraddr
,
1393 netdev_bsd_get_etheraddr
,
1396 netdev_bsd_get_ifindex
,
1397 netdev_bsd_get_carrier
,
1398 NULL
, /* get_carrier_resets */
1399 NULL
, /* set_miimon_interval */
1400 netdev_bsd_get_stats
,
1401 NULL
, /* set_stats */
1403 netdev_bsd_get_features
,
1404 NULL
, /* set_advertisement */
1405 NULL
, /* set_policing */
1406 NULL
, /* get_qos_type */
1407 NULL
, /* get_qos_capabilities */
1410 NULL
, /* get_queue */
1411 NULL
, /* set_queue */
1412 NULL
, /* delete_queue */
1413 NULL
, /* get_queue_stats */
1414 NULL
, /* dump_queue */
1415 NULL
, /* dump_queue_stats */
1420 NULL
, /* add_router */
1421 netdev_bsd_get_next_hop
,
1422 NULL
, /* get_status */
1423 NULL
, /* arp_lookup */
1425 netdev_bsd_update_flags
,
1427 netdev_bsd_change_seq
,
1429 netdev_bsd_rx_alloc
,
1430 netdev_bsd_rx_construct
,
1431 netdev_bsd_rx_destruct
,
1432 netdev_bsd_rx_dealloc
,
1435 netdev_bsd_rx_drain
,
1438 const struct netdev_class netdev_tap_class
= {
1445 netdev_bsd_construct_tap
,
1446 netdev_bsd_destruct
,
1448 NULL
, /* get_config */
1449 NULL
, /* set_config */
1450 NULL
, /* get_tunnel_config */
1453 netdev_bsd_send_wait
,
1455 netdev_bsd_set_etheraddr
,
1456 netdev_bsd_get_etheraddr
,
1459 netdev_bsd_get_ifindex
,
1460 netdev_bsd_get_carrier
,
1461 NULL
, /* get_carrier_resets */
1462 NULL
, /* set_miimon_interval */
1463 netdev_bsd_get_stats
,
1464 NULL
, /* set_stats */
1466 netdev_bsd_get_features
,
1467 NULL
, /* set_advertisement */
1468 NULL
, /* set_policing */
1469 NULL
, /* get_qos_type */
1470 NULL
, /* get_qos_capabilities */
1473 NULL
, /* get_queue */
1474 NULL
, /* set_queue */
1475 NULL
, /* delete_queue */
1476 NULL
, /* get_queue_stats */
1477 NULL
, /* dump_queue */
1478 NULL
, /* dump_queue_stats */
1483 NULL
, /* add_router */
1484 netdev_bsd_get_next_hop
,
1485 NULL
, /* get_status */
1486 NULL
, /* arp_lookup */
1488 netdev_bsd_update_flags
,
1490 netdev_bsd_change_seq
,
1492 netdev_bsd_rx_alloc
,
1493 netdev_bsd_rx_construct
,
1494 netdev_bsd_rx_destruct
,
1495 netdev_bsd_rx_dealloc
,
1498 netdev_bsd_rx_drain
,
1503 destroy_tap(int fd
, const char *name
)
1508 strcpy(ifr
.ifr_name
, name
);
1509 /* XXX What to do if this call fails? */
1510 af_inet_ioctl(SIOCIFDESTROY
, &ifr
);
1514 get_flags(const struct netdev
*netdev
, int *flags
)
1519 error
= af_inet_ifreq_ioctl(netdev_get_kernel_name(netdev
), &ifr
,
1520 SIOCGIFFLAGS
, "SIOCGIFFLAGS");
1522 *flags
= ifr_get_flags(&ifr
);
1528 set_flags(const char *name
, int flags
)
1532 ifr_set_flags(&ifr
, flags
);
1534 return af_inet_ifreq_ioctl(name
, &ifr
, SIOCSIFFLAGS
, "SIOCSIFFLAGS");
1538 get_ifindex(const struct netdev
*netdev_
, int *ifindexp
)
1540 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
1542 if (!(netdev
->cache_valid
& VALID_IFINDEX
)) {
1543 int ifindex
= if_nametoindex(netdev_get_name(netdev_
));
1547 netdev
->cache_valid
|= VALID_IFINDEX
;
1548 netdev
->ifindex
= ifindex
;
1550 *ifindexp
= netdev
->ifindex
;
1555 get_etheraddr(const char *netdev_name
, uint8_t ea
[ETH_ADDR_LEN
])
1557 struct ifaddrs
*head
;
1558 struct ifaddrs
*ifa
;
1559 struct sockaddr_dl
*sdl
;
1561 if (getifaddrs(&head
) != 0) {
1562 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name
,
1563 ovs_strerror(errno
));
1567 for (ifa
= head
; ifa
; ifa
= ifa
->ifa_next
) {
1568 if (ifa
->ifa_addr
->sa_family
== AF_LINK
) {
1569 if (!strcmp(ifa
->ifa_name
, netdev_name
)) {
1570 sdl
= (struct sockaddr_dl
*)ifa
->ifa_addr
;
1572 memcpy(ea
, LLADDR(sdl
), sdl
->sdl_alen
);
1580 VLOG_ERR("could not find ethernet address for %s device", netdev_name
);
1586 set_etheraddr(const char *netdev_name OVS_UNUSED
, int hwaddr_family OVS_UNUSED
,
1587 int hwaddr_len OVS_UNUSED
,
1588 const uint8_t mac
[ETH_ADDR_LEN
] OVS_UNUSED
)
1590 #if defined(__FreeBSD__)
1594 memset(&ifr
, 0, sizeof ifr
);
1595 strncpy(ifr
.ifr_name
, netdev_name
, sizeof ifr
.ifr_name
);
1596 ifr
.ifr_addr
.sa_family
= hwaddr_family
;
1597 ifr
.ifr_addr
.sa_len
= hwaddr_len
;
1598 memcpy(ifr
.ifr_addr
.sa_data
, mac
, hwaddr_len
);
1599 error
= af_inet_ioctl(SIOCSIFLLADDR
, &ifr
);
1601 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1602 netdev_name
, ovs_strerror(error
));
1606 #elif defined(__NetBSD__)
1607 struct if_laddrreq req
;
1608 struct sockaddr_dl
*sdl
;
1609 struct sockaddr_storage oldaddr
;
1613 * get the old address, add new one, and then remove old one.
1616 if (hwaddr_len
!= ETH_ADDR_LEN
) {
1617 /* just to be safe about sockaddr storage size */
1620 memset(&req
, 0, sizeof(req
));
1621 strncpy(req
.iflr_name
, netdev_name
, sizeof(req
.iflr_name
));
1622 req
.addr
.ss_len
= sizeof(req
.addr
);
1623 req
.addr
.ss_family
= hwaddr_family
;
1624 sdl
= (struct sockaddr_dl
*)&req
.addr
;
1625 sdl
->sdl_alen
= hwaddr_len
;
1627 error
= af_link_ioctl(SIOCGLIFADDR
, &req
);
1631 if (!memcmp(&sdl
->sdl_data
[sdl
->sdl_nlen
], mac
, hwaddr_len
)) {
1636 memset(&req
, 0, sizeof(req
));
1637 strncpy(req
.iflr_name
, netdev_name
, sizeof(req
.iflr_name
));
1638 req
.flags
= IFLR_ACTIVE
;
1639 sdl
= (struct sockaddr_dl
*)&req
.addr
;
1640 sdl
->sdl_len
= offsetof(struct sockaddr_dl
, sdl_data
) + hwaddr_len
;
1641 sdl
->sdl_alen
= hwaddr_len
;
1642 sdl
->sdl_family
= hwaddr_family
;
1643 memcpy(sdl
->sdl_data
, mac
, hwaddr_len
);
1644 error
= af_link_ioctl(SIOCALIFADDR
, &req
);
1649 memset(&req
, 0, sizeof(req
));
1650 strncpy(req
.iflr_name
, netdev_name
, sizeof(req
.iflr_name
));
1652 return af_link_ioctl(SIOCDLIFADDR
, &req
);
1654 #error not implemented
1659 ifr_get_flags(const struct ifreq
*ifr
)
1661 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1662 return (ifr
->ifr_flagshigh
<< 16) | ifr
->ifr_flags
;
1664 return ifr
->ifr_flags
;
1669 ifr_set_flags(struct ifreq
*ifr
, int flags
)
1671 ifr
->ifr_flags
= flags
;
1672 #ifdef HAVE_STRUCT_IFREQ_IFR_FLAGSHIGH
1673 ifr
->ifr_flagshigh
= flags
>> 16;
1677 /* Calls ioctl() on an AF_LINK sock, passing the specified 'command' and
1678 * 'arg'. Returns 0 if successful, otherwise a positive errno value. */
1680 af_link_ioctl(int command
, const void *arg
)
1682 static struct ovsthread_once once
= OVSTHREAD_ONCE_INITIALIZER
;
1685 if (ovsthread_once_start(&once
)) {
1686 sock
= socket(AF_LINK
, SOCK_DGRAM
, 0);
1689 VLOG_ERR("failed to create link socket: %s", ovs_strerror(errno
));
1691 ovsthread_once_done(&once
);
1694 return (sock
< 0 ? -sock
1695 : ioctl(sock
, command
, arg
) == -1 ? errno