2 * Copyright (c) 2011 Gaetano Catalli.
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.
19 #include "netdev-provider.h"
23 #include <sys/types.h>
25 #include <sys/ioctl.h>
26 #include <sys/socket.h>
27 #include <sys/sockio.h>
29 #include <pcap/pcap.h>
31 #include <net/if_dl.h>
32 #include <net/if_media.h>
33 #include <net/if_tap.h>
34 #include <netinet/in.h>
35 #include <net/if_mib.h>
39 #include <sys/sysctl.h>
43 #include "dynamic-string.h"
44 #include "fatal-signal.h"
46 #include "openflow/openflow.h"
48 #include "poll-loop.h"
49 #include "socket-util.h"
55 VLOG_DEFINE_THIS_MODULE(netdev_bsd
);
59 * This file implements objects to access interfaces.
60 * Externally, interfaces are represented by three structures:
61 * + struct netdev_dev, representing a network device,
62 * containing e.g. name and a refcount;
63 * We can have private variables by embedding the
64 * struct netdev_dev into our own structure
65 * (e.g. netdev_dev_bsd)
67 * + struct netdev, representing an instance of an open netdev_dev.
68 * The structure contains a pointer to the 'struct netdev'
69 * representing the device.
71 * + struct netdev_rx, which represents a netdev open to capture received
72 * packets. Again, private information such as file descriptor etc. are
73 * stored in our own struct netdev_rx_bsd which includes a struct
76 * 'struct netdev', 'struct netdev_dev', and 'struct netdev_rx' are referenced
77 * in containers which hold pointers to the data structures. We can reach our
78 * own struct netdev_XXX_bsd by putting a struct netdev_XXX within our own
79 * struct, and using CONTAINER_OF to access the parent structure.
85 struct netdev_rx_bsd
{
88 /* Packet capture descriptor for a system network device.
89 * For a tap device this is NULL. */
92 /* Selectable file descriptor for the network device.
93 * This descriptor will be used for polling operations. */
97 static const struct netdev_rx_class netdev_rx_bsd_class
;
99 struct netdev_dev_bsd
{
100 struct netdev_dev netdev_dev
;
101 unsigned int cache_valid
;
102 unsigned int change_seq
;
105 uint8_t etheraddr
[ETH_ADDR_LEN
];
111 int tap_fd
; /* TAP character device, if any, otherwise -1. */
113 /* Used for sending packets on non-tap devices. */
120 VALID_IFINDEX
= 1 << 0,
121 VALID_ETHERADDR
= 1 << 1,
125 VALID_CARRIER
= 1 << 5
128 /* An AF_INET socket (used for ioctl operations). */
129 static int af_inet_sock
= -1;
131 #define PCAP_SNAPLEN 2048
135 * Notifier used to invalidate device informations in case of status change.
137 * It will be registered with a 'rtbsd_notifier_register()' when the first
138 * device will be created with the call of either 'netdev_bsd_tap_create()' or
139 * 'netdev_bsd_system_create()'.
141 * The callback associated with this notifier ('netdev_bsd_cache_cb()') will
142 * invalidate cached information about the device.
144 static struct rtbsd_notifier netdev_bsd_cache_notifier
;
145 static int cache_notifier_refcount
;
147 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 20);
149 static int netdev_bsd_do_ioctl(const char *, struct ifreq
*, unsigned long cmd
,
150 const char *cmd_name
);
151 static void destroy_tap(int fd
, const char *name
);
152 static int get_flags(const struct netdev_dev
*, int *flagsp
);
153 static int set_flags(const char *, int flags
);
154 static int do_set_addr(struct netdev
*netdev
,
155 int ioctl_nr
, const char *ioctl_name
,
156 struct in_addr addr
);
157 static int get_etheraddr(const char *netdev_name
, uint8_t ea
[ETH_ADDR_LEN
]);
158 static int set_etheraddr(const char *netdev_name
, int hwaddr_family
,
159 int hwaddr_len
, const uint8_t[ETH_ADDR_LEN
]);
160 static int get_ifindex(const struct netdev
*, int *ifindexp
);
162 static int netdev_bsd_init(void);
165 is_netdev_bsd_class(const struct netdev_class
*netdev_class
)
167 return netdev_class
->init
== netdev_bsd_init
;
170 static struct netdev_bsd
*
171 netdev_bsd_cast(const struct netdev
*netdev
)
173 ovs_assert(is_netdev_bsd_class(netdev_dev_get_class(
174 netdev_get_dev(netdev
))));
175 return CONTAINER_OF(netdev
, struct netdev_bsd
, netdev
);
178 static struct netdev_dev_bsd
*
179 netdev_dev_bsd_cast(const struct netdev_dev
*netdev_dev
)
181 ovs_assert(is_netdev_bsd_class(netdev_dev_get_class(netdev_dev
)));
182 return CONTAINER_OF(netdev_dev
, struct netdev_dev_bsd
, netdev_dev
);
185 static struct netdev_rx_bsd
*
186 netdev_rx_bsd_cast(const struct netdev_rx
*rx
)
188 netdev_rx_assert_class(rx
, &netdev_rx_bsd_class
);
189 return CONTAINER_OF(rx
, struct netdev_rx_bsd
, up
);
192 /* Initialize the AF_INET socket used for ioctl operations */
194 netdev_bsd_init(void)
196 static int status
= -1;
198 if (status
>= 0) { /* already initialized */
202 af_inet_sock
= socket(AF_INET
, SOCK_DGRAM
, 0);
203 status
= af_inet_sock
>= 0 ? 0 : errno
;
206 VLOG_ERR("failed to create inet socket: %s", strerror(status
));
213 * Perform periodic work needed by netdev. In BSD netdevs it checks for any
214 * interface status changes, and eventually calls all the user callbacks.
219 rtbsd_notifier_run();
223 * Arranges for poll_block() to wake up if the "run" member function needs to
227 netdev_bsd_wait(void)
229 rtbsd_notifier_wait();
233 netdev_dev_bsd_changed(struct netdev_dev_bsd
*dev
)
236 if (!dev
->change_seq
) {
241 /* Invalidate cache in case of interface status change. */
243 netdev_bsd_cache_cb(const struct rtbsd_change
*change
,
244 void *aux OVS_UNUSED
)
246 struct netdev_dev_bsd
*dev
;
249 struct netdev_dev
*base_dev
= netdev_dev_from_name(change
->if_name
);
252 const struct netdev_class
*netdev_class
=
253 netdev_dev_get_class(base_dev
);
255 if (is_netdev_bsd_class(netdev_class
)) {
256 dev
= netdev_dev_bsd_cast(base_dev
);
257 dev
->cache_valid
= 0;
258 netdev_dev_bsd_changed(dev
);
263 * XXX the API is lacking, we should be able to iterate on the list of
264 * netdevs without having to store the info in a temp shash.
266 struct shash device_shash
;
267 struct shash_node
*node
;
269 shash_init(&device_shash
);
270 netdev_dev_get_devices(&netdev_bsd_class
, &device_shash
);
271 SHASH_FOR_EACH (node
, &device_shash
) {
273 dev
->cache_valid
= 0;
274 netdev_dev_bsd_changed(dev
);
276 shash_destroy(&device_shash
);
281 cache_notifier_ref(void)
285 if (!cache_notifier_refcount
) {
286 ret
= rtbsd_notifier_register(&netdev_bsd_cache_notifier
,
287 netdev_bsd_cache_cb
, NULL
);
292 cache_notifier_refcount
++;
297 cache_notifier_unref(void)
299 cache_notifier_refcount
--;
300 if (cache_notifier_refcount
== 0) {
301 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier
);
306 /* Allocate a netdev_dev_bsd structure */
308 netdev_bsd_create_system(const struct netdev_class
*class, const char *name
,
309 struct netdev_dev
**netdev_devp
)
311 struct netdev_dev_bsd
*netdev_dev
;
314 error
= cache_notifier_ref();
319 netdev_dev
= xzalloc(sizeof *netdev_dev
);
320 netdev_dev
->change_seq
= 1;
321 netdev_dev_init(&netdev_dev
->netdev_dev
, name
, class);
322 netdev_dev
->tap_fd
= -1;
323 *netdev_devp
= &netdev_dev
->netdev_dev
;
329 * Allocate a netdev_dev_bsd structure with 'tap' class.
332 netdev_bsd_create_tap(const struct netdev_class
*class, const char *name
,
333 struct netdev_dev
**netdev_devp
)
335 struct netdev_dev_bsd
*netdev_dev
= NULL
;
339 error
= cache_notifier_ref();
344 /* allocate the device structure and set the internal flag */
345 netdev_dev
= xzalloc(sizeof *netdev_dev
);
347 memset(&ifr
, 0, sizeof(ifr
));
349 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
350 * to retrieve the name of the tap device. */
351 netdev_dev
->tap_fd
= open("/dev/tap", O_RDWR
);
352 netdev_dev
->change_seq
= 1;
353 if (netdev_dev
->tap_fd
< 0) {
355 VLOG_WARN("opening \"/dev/tap\" failed: %s", strerror(error
));
356 goto error_undef_notifier
;
359 /* Retrieve tap name (e.g. tap0) */
360 if (ioctl(netdev_dev
->tap_fd
, TAPGIFNAME
, &ifr
) == -1) {
361 /* XXX Need to destroy the device? */
363 goto error_undef_notifier
;
366 /* Change the name of the tap device */
367 ifr
.ifr_data
= (void *)name
;
368 if (ioctl(af_inet_sock
, SIOCSIFNAME
, &ifr
) == -1) {
370 destroy_tap(netdev_dev
->tap_fd
, ifr
.ifr_name
);
371 goto error_undef_notifier
;
374 /* set non-blocking. */
375 error
= set_nonblocking(netdev_dev
->tap_fd
);
377 destroy_tap(netdev_dev
->tap_fd
, name
);
378 goto error_undef_notifier
;
382 ifr
.ifr_flags
= (uint16_t)IFF_UP
;
383 ifr
.ifr_flagshigh
= 0;
384 strncpy(ifr
.ifr_name
, name
, sizeof ifr
.ifr_name
);
385 if (ioctl(af_inet_sock
, SIOCSIFFLAGS
, &ifr
) == -1) {
387 destroy_tap(netdev_dev
->tap_fd
, name
);
388 goto error_undef_notifier
;
391 /* initialize the device structure and
392 * link the structure to its netdev */
393 netdev_dev_init(&netdev_dev
->netdev_dev
, name
, class);
394 *netdev_devp
= &netdev_dev
->netdev_dev
;
398 error_undef_notifier
:
399 cache_notifier_unref();
406 netdev_bsd_destroy(struct netdev_dev
*netdev_dev_
)
408 struct netdev_dev_bsd
*netdev_dev
= netdev_dev_bsd_cast(netdev_dev_
);
410 cache_notifier_unref();
412 if (netdev_dev
->tap_fd
>= 0) {
413 destroy_tap(netdev_dev
->tap_fd
, netdev_dev_get_name(netdev_dev_
));
415 if (netdev_dev
->pcap
) {
416 pcap_close(netdev_dev
->pcap
);
423 netdev_bsd_open_system(struct netdev_dev
*netdev_dev_
, struct netdev
**netdevp
)
425 struct netdev_bsd
*netdev
;
427 enum netdev_flags flags
;
429 /* Allocate network device. */
430 netdev
= xcalloc(1, sizeof *netdev
);
431 netdev_init(&netdev
->netdev
, netdev_dev_
);
433 /* Verify that the netdev really exists by attempting to read its flags */
434 error
= netdev_get_flags(&netdev
->netdev
, &flags
);
435 if (error
== ENXIO
) {
439 *netdevp
= &netdev
->netdev
;
443 netdev_uninit(&netdev
->netdev
, true);
449 /* Close a 'netdev'. */
451 netdev_bsd_close(struct netdev
*netdev_
)
453 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
459 netdev_bsd_open_pcap(const char *name
, pcap_t
**pcapp
, int *fdp
)
461 char errbuf
[PCAP_ERRBUF_SIZE
];
467 /* Open the pcap device. The device is opened in non-promiscuous mode
468 * because the interface flags are manually set by the caller. */
470 pcap
= pcap_open_live(name
, PCAP_SNAPLEN
, 0, 1000, errbuf
);
472 VLOG_ERR_RL(&rl
, "%s: pcap_open_live failed: %s", name
, errbuf
);
476 if (errbuf
[0] != '\0') {
477 VLOG_WARN_RL(&rl
, "%s: pcap_open_live: %s", name
, errbuf
);
480 /* Get the underlying fd. */
481 fd
= pcap_get_selectable_fd(pcap
);
483 VLOG_WARN_RL(&rl
, "%s: no selectable file descriptor", name
);
488 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
489 * on the file descriptor returned by pcap_get_selectable_fd to achieve
490 * a real non-blocking behaviour.*/
491 error
= pcap_setnonblock(pcap
, 1, errbuf
);
497 /* This call assure that reads return immediately upon packet
498 * reception. Otherwise, a read will block until either the kernel
499 * buffer becomes full or a timeout occurs. */
500 if (ioctl(fd
, BIOCIMMEDIATE
, &one
) < 0 ) {
501 VLOG_ERR_RL(&rl
, "ioctl(BIOCIMMEDIATE) on %s device failed: %s",
502 name
, strerror(errno
));
507 /* Capture only incoming packets. */
508 error
= pcap_setdirection(pcap
, PCAP_D_IN
);
528 netdev_bsd_rx_open(struct netdev
*netdev_
, struct netdev_rx
**rxp
)
530 struct netdev_dev_bsd
*netdev_dev
=
531 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
533 struct netdev_rx_bsd
*rx
;
537 if (!strcmp(netdev_get_type(netdev_
), "tap")) {
539 fd
= netdev_dev
->tap_fd
;
541 int error
= netdev_bsd_open_pcap(netdev_get_name(netdev_
), &pcap
, &fd
);
546 netdev_dev_bsd_changed(netdev_dev
);
549 rx
= xmalloc(sizeof *rx
);
550 netdev_rx_init(&rx
->up
, netdev_get_dev(netdev_
), &netdev_rx_bsd_class
);
551 rx
->pcap_handle
= pcap
;
559 netdev_rx_bsd_destroy(struct netdev_rx
*rx_
)
561 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
563 if (rx
->pcap_handle
) {
564 pcap_close(rx
->pcap_handle
);
569 /* The recv callback of the netdev class returns the number of bytes of the
572 * This can be done by the pcap_next() function. Unfortunately pcap_next() does
573 * not make difference between a missing packet on the capture interface and
574 * an error during the file capture. We can use the pcap_dispatch() function
575 * instead, which is able to distinguish between errors and null packet.
577 * To make pcap_dispatch() returns the number of bytes read from the interface
578 * we need to define the following callback and argument.
587 * This callback will be executed on every captured packet.
589 * If the packet captured by pcap_dispatch() does not fit the pcap buffer,
590 * pcap returns a truncated packet and we follow this behavior.
592 * The argument args->retval is the packet size in bytes.
595 proc_pkt(u_char
*args_
, const struct pcap_pkthdr
*hdr
, const u_char
*packet
)
597 struct pcap_arg
*args
= (struct pcap_arg
*)args_
;
599 if (args
->size
< hdr
->len
) {
600 VLOG_WARN_RL(&rl
, "packet truncated");
601 args
->retval
= args
->size
;
603 args
->retval
= hdr
->len
;
606 /* copy the packet to our buffer */
607 memcpy(args
->data
, packet
, args
->retval
);
611 * This function attempts to receive a packet from the specified network
612 * device. It is assumed that the network device is a system device or a tap
613 * device opened as a system one. In this case the read operation is performed
617 netdev_rx_bsd_recv_pcap(struct netdev_rx_bsd
*rx
, void *data
, size_t size
)
622 /* prepare the pcap argument to store the packet */
627 ret
= pcap_dispatch(rx
->pcap_handle
, 1, proc_pkt
, (u_char
*) &arg
);
630 return arg
.retval
; /* arg.retval < 0 is handled in the caller */
633 if (errno
== EINTR
) {
643 * This function attempts to receive a packet from the specified network
644 * device. It is assumed that the network device is a tap device and
645 * 'rx->fd' is initialized with the tap file descriptor.
648 netdev_rx_bsd_recv_tap(struct netdev_rx_bsd
*rx
, void *data
, size_t size
)
651 ssize_t retval
= read(rx
->fd
, data
, size
);
654 } else if (errno
!= EINTR
) {
655 if (errno
!= EAGAIN
) {
656 VLOG_WARN_RL(&rl
, "error receiving Ethernet packet on %s: %s",
657 strerror(errno
), netdev_rx_get_name(&rx
->up
));
666 netdev_rx_bsd_recv(struct netdev_rx
*rx_
, void *data
, size_t size
)
668 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
670 return (rx
->pcap_handle
671 ? netdev_rx_bsd_recv_pcap(rx
, data
, size
)
672 : netdev_rx_bsd_recv_tap(rx
, data
, size
));
676 * Registers with the poll loop to wake up from the next call to poll_block()
677 * when a packet is ready to be received with netdev_rx_recv() on 'rx'.
680 netdev_rx_bsd_wait(struct netdev_rx
*rx_
)
682 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
684 poll_fd_wait(rx
->fd
, POLLIN
);
687 /* Discards all packets waiting to be received from 'rx'. */
689 netdev_rx_bsd_drain(struct netdev_rx
*rx_
)
692 struct netdev_rx_bsd
*rx
= netdev_rx_bsd_cast(rx_
);
694 strcpy(ifr
.ifr_name
, netdev_rx_get_name(rx_
));
695 if (ioctl(rx
->fd
, BIOCFLUSH
, &ifr
) == -1) {
696 VLOG_DBG_RL(&rl
, "%s: ioctl(BIOCFLUSH) failed: %s",
697 netdev_rx_get_name(rx_
), strerror(errno
));
704 * Send a packet on the specified network device. The device could be either a
705 * system or a tap device.
708 netdev_bsd_send(struct netdev
*netdev_
, const void *data
, size_t size
)
710 struct netdev_dev_bsd
*dev
= netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
711 const char *name
= netdev_get_name(netdev_
);
713 if (dev
->tap_fd
< 0 && !dev
->pcap
) {
714 int error
= netdev_bsd_open_pcap(name
, &dev
->pcap
, &dev
->fd
);
722 if (dev
->tap_fd
>= 0) {
723 retval
= write(dev
->tap_fd
, data
, size
);
725 retval
= pcap_inject(dev
->pcap
, data
, size
);
728 if (errno
== EINTR
) {
730 } else if (errno
!= EAGAIN
) {
731 VLOG_WARN_RL(&rl
, "error sending Ethernet packet on %s: %s",
732 name
, strerror(errno
));
735 } else if (retval
!= size
) {
736 VLOG_WARN_RL(&rl
, "sent partial Ethernet packet (%zd bytes of "
737 "%zu) on %s", retval
, size
, name
);
746 * Registers with the poll loop to wake up from the next call to poll_block()
747 * when the packet transmission queue has sufficient room to transmit a packet
748 * with netdev_send().
751 netdev_bsd_send_wait(struct netdev
*netdev_
)
753 struct netdev_dev_bsd
*dev
= netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
755 if (dev
->tap_fd
>= 0) {
756 /* TAP device always accepts packets. */
757 poll_immediate_wake();
758 } else if (dev
->pcap
) {
759 poll_fd_wait(dev
->fd
, POLLOUT
);
761 /* We haven't even tried to send a packet yet. */
762 poll_immediate_wake();
767 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
768 * otherwise a positive errno value.
771 netdev_bsd_set_etheraddr(struct netdev
*netdev_
,
772 const uint8_t mac
[ETH_ADDR_LEN
])
774 struct netdev_dev_bsd
*netdev_dev
=
775 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
778 if (!(netdev_dev
->cache_valid
& VALID_ETHERADDR
)
779 || !eth_addr_equals(netdev_dev
->etheraddr
, mac
)) {
780 error
= set_etheraddr(netdev_get_name(netdev_
), AF_LINK
, ETH_ADDR_LEN
,
783 netdev_dev
->cache_valid
|= VALID_ETHERADDR
;
784 memcpy(netdev_dev
->etheraddr
, mac
, ETH_ADDR_LEN
);
785 netdev_dev_bsd_changed(netdev_dev
);
794 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
795 * free the returned buffer.
798 netdev_bsd_get_etheraddr(const struct netdev
*netdev_
,
799 uint8_t mac
[ETH_ADDR_LEN
])
801 struct netdev_dev_bsd
*netdev_dev
=
802 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
804 if (!(netdev_dev
->cache_valid
& VALID_ETHERADDR
)) {
805 int error
= get_etheraddr(netdev_get_name(netdev_
),
806 netdev_dev
->etheraddr
);
810 netdev_dev
->cache_valid
|= VALID_ETHERADDR
;
812 memcpy(mac
, netdev_dev
->etheraddr
, ETH_ADDR_LEN
);
818 * Returns the maximum size of transmitted (and received) packets on 'netdev',
819 * in bytes, not including the hardware header; thus, this is typically 1500
820 * bytes for Ethernet devices.
823 netdev_bsd_get_mtu(const struct netdev
*netdev_
, int *mtup
)
825 struct netdev_dev_bsd
*netdev_dev
=
826 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
828 if (!(netdev_dev
->cache_valid
& VALID_MTU
)) {
832 error
= netdev_bsd_do_ioctl(netdev_get_name(netdev_
), &ifr
, SIOCGIFMTU
,
837 netdev_dev
->mtu
= ifr
.ifr_mtu
;
838 netdev_dev
->cache_valid
|= VALID_MTU
;
841 *mtup
= netdev_dev
->mtu
;
846 netdev_bsd_get_ifindex(const struct netdev
*netdev
)
850 error
= get_ifindex(netdev
, &ifindex
);
851 return error
? -error
: ifindex
;
855 netdev_bsd_get_carrier(const struct netdev
*netdev_
, bool *carrier
)
857 struct netdev_dev_bsd
*netdev_dev
=
858 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
860 if (!(netdev_dev
->cache_valid
& VALID_CARRIER
)) {
861 struct ifmediareq ifmr
;
863 memset(&ifmr
, 0, sizeof(ifmr
));
864 strncpy(ifmr
.ifm_name
, netdev_get_name(netdev_
), sizeof ifmr
.ifm_name
);
866 if (ioctl(af_inet_sock
, SIOCGIFMEDIA
, &ifmr
) == -1) {
867 VLOG_DBG_RL(&rl
, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
868 netdev_get_name(netdev_
), strerror(errno
));
872 netdev_dev
->carrier
= (ifmr
.ifm_status
& IFM_ACTIVE
) == IFM_ACTIVE
;
873 netdev_dev
->cache_valid
|= VALID_CARRIER
;
875 /* If the interface doesn't report whether the media is active,
876 * just assume it is active. */
877 if ((ifmr
.ifm_status
& IFM_AVALID
) == 0) {
878 netdev_dev
->carrier
= true;
881 *carrier
= netdev_dev
->carrier
;
886 /* Retrieves current device stats for 'netdev'. */
888 netdev_bsd_get_stats(const struct netdev
*netdev_
, struct netdev_stats
*stats
)
893 struct ifmibdata ifmd
;
898 mib
[2] = NETLINK_GENERIC
;
899 mib
[3] = IFMIB_SYSTEM
;
900 mib
[4] = IFMIB_IFCOUNT
;
902 len
= sizeof(if_count
);
904 if (sysctl(mib
, 5, &if_count
, &len
, (void *)0, 0) == -1) {
905 VLOG_DBG_RL(&rl
, "%s: sysctl failed: %s",
906 netdev_get_name(netdev_
), strerror(errno
));
910 mib
[5] = IFDATA_GENERAL
;
911 mib
[3] = IFMIB_IFDATA
;
913 for (i
= 1; i
<= if_count
; i
++) {
915 if (sysctl(mib
, 6, &ifmd
, &len
, (void *)0, 0) == -1) {
916 VLOG_DBG_RL(&rl
, "%s: sysctl failed: %s",
917 netdev_get_name(netdev_
), strerror(errno
));
919 } else if (!strcmp(ifmd
.ifmd_name
, netdev_get_name(netdev_
))) {
920 stats
->rx_packets
= ifmd
.ifmd_data
.ifi_ipackets
;
921 stats
->tx_packets
= ifmd
.ifmd_data
.ifi_opackets
;
922 stats
->rx_bytes
= ifmd
.ifmd_data
.ifi_ibytes
;
923 stats
->tx_bytes
= ifmd
.ifmd_data
.ifi_obytes
;
924 stats
->rx_errors
= ifmd
.ifmd_data
.ifi_ierrors
;
925 stats
->tx_errors
= ifmd
.ifmd_data
.ifi_oerrors
;
926 stats
->rx_dropped
= ifmd
.ifmd_data
.ifi_iqdrops
;
927 stats
->tx_dropped
= UINT64_MAX
;
928 stats
->multicast
= ifmd
.ifmd_data
.ifi_imcasts
;
929 stats
->collisions
= ifmd
.ifmd_data
.ifi_collisions
;
931 stats
->rx_length_errors
= UINT64_MAX
;
932 stats
->rx_over_errors
= UINT64_MAX
;
933 stats
->rx_crc_errors
= UINT64_MAX
;
934 stats
->rx_frame_errors
= UINT64_MAX
;
935 stats
->rx_fifo_errors
= UINT64_MAX
;
936 stats
->rx_missed_errors
= UINT64_MAX
;
938 stats
->tx_aborted_errors
= UINT64_MAX
;
939 stats
->tx_carrier_errors
= UINT64_MAX
;
940 stats
->tx_fifo_errors
= UINT64_MAX
;
941 stats
->tx_heartbeat_errors
= UINT64_MAX
;
942 stats
->tx_window_errors
= UINT64_MAX
;
951 netdev_bsd_parse_media(int media
)
953 uint32_t supported
= 0;
954 bool half_duplex
= media
& IFM_HDX
? true : false;
956 switch (IFM_SUBTYPE(media
)) {
961 supported
|= half_duplex
? NETDEV_F_10MB_HD
: NETDEV_F_10MB_FD
;
962 supported
|= NETDEV_F_COPPER
;
966 supported
|= half_duplex
? NETDEV_F_10MB_HD
: NETDEV_F_10MB_FD
;
967 supported
|= NETDEV_F_FIBER
;
974 supported
|= half_duplex
? NETDEV_F_100MB_HD
: NETDEV_F_100MB_FD
;
975 supported
|= NETDEV_F_COPPER
;
979 supported
|= half_duplex
? NETDEV_F_100MB_HD
: NETDEV_F_100MB_FD
;
980 supported
|= NETDEV_F_FIBER
;
985 supported
|= half_duplex
? NETDEV_F_1GB_HD
: NETDEV_F_1GB_FD
;
986 supported
|= NETDEV_F_COPPER
;
991 supported
|= half_duplex
? NETDEV_F_1GB_HD
: NETDEV_F_1GB_FD
;
992 supported
|= NETDEV_F_FIBER
;
996 supported
|= NETDEV_F_10GB_FD
;
997 supported
|= NETDEV_F_COPPER
;
1002 supported
|= NETDEV_F_10GB_FD
;
1003 supported
|= NETDEV_F_FIBER
;
1010 if (IFM_SUBTYPE(media
) == IFM_AUTO
) {
1011 supported
|= NETDEV_F_AUTONEG
;
1014 if (media & IFM_ETH_FMASK) {
1015 supported |= NETDEV_F_PAUSE;
1023 * Stores the features supported by 'netdev' into each of '*current',
1024 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1025 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1026 * successful, otherwise a positive errno value. On failure, all of the
1027 * passed-in values are set to 0.
1030 netdev_bsd_get_features(const struct netdev
*netdev
,
1031 enum netdev_features
*current
, uint32_t *advertised
,
1032 enum netdev_features
*supported
, uint32_t *peer
)
1034 struct ifmediareq ifmr
;
1040 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1042 memset(&ifmr
, 0, sizeof(ifmr
));
1043 strncpy(ifmr
.ifm_name
, netdev_get_name(netdev
), sizeof ifmr
.ifm_name
);
1045 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1046 * number of supported modes, and a second with a buffer to retrieve
1048 if (ioctl(af_inet_sock
, SIOCGIFMEDIA
, &ifmr
) == -1) {
1049 VLOG_DBG_RL(&rl
, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1050 netdev_get_name(netdev
), strerror(errno
));
1054 media_list
= xcalloc(ifmr
.ifm_count
, sizeof(int));
1055 ifmr
.ifm_ulist
= media_list
;
1057 if (IFM_TYPE(ifmr
.ifm_current
) != IFM_ETHER
) {
1058 VLOG_DBG_RL(&rl
, "%s: doesn't appear to be ethernet",
1059 netdev_get_name(netdev
));
1064 if (ioctl(af_inet_sock
, SIOCGIFMEDIA
, &ifmr
) == -1) {
1065 VLOG_DBG_RL(&rl
, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1066 netdev_get_name(netdev
), strerror(errno
));
1071 /* Current settings. */
1072 *current
= netdev_bsd_parse_media(ifmr
.ifm_active
);
1074 /* Advertised features. */
1075 *advertised
= netdev_bsd_parse_media(ifmr
.ifm_current
);
1077 /* Supported features. */
1079 for (i
= 0; i
< ifmr
.ifm_count
; i
++) {
1080 *supported
|= netdev_bsd_parse_media(ifmr
.ifm_ulist
[i
]);
1083 /* Peer advertisements. */
1084 *peer
= 0; /* XXX */
1093 * If 'netdev' has an assigned IPv4 address, sets '*in4' to that address (if
1094 * 'in4' is non-null) and returns true. Otherwise, returns false.
1097 netdev_bsd_get_in4(const struct netdev
*netdev_
, struct in_addr
*in4
,
1098 struct in_addr
*netmask
)
1100 struct netdev_dev_bsd
*netdev_dev
=
1101 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
1103 if (!(netdev_dev
->cache_valid
& VALID_IN4
)) {
1104 const struct sockaddr_in
*sin
;
1108 ifr
.ifr_addr
.sa_family
= AF_INET
;
1109 error
= netdev_bsd_do_ioctl(netdev_get_name(netdev_
), &ifr
,
1110 SIOCGIFADDR
, "SIOCGIFADDR");
1115 sin
= (struct sockaddr_in
*) &ifr
.ifr_addr
;
1116 netdev_dev
->in4
= sin
->sin_addr
;
1117 netdev_dev
->cache_valid
|= VALID_IN4
;
1118 error
= netdev_bsd_do_ioctl(netdev_get_name(netdev_
), &ifr
,
1119 SIOCGIFNETMASK
, "SIOCGIFNETMASK");
1123 *netmask
= ((struct sockaddr_in
*)&ifr
.ifr_addr
)->sin_addr
;
1125 *in4
= netdev_dev
->in4
;
1127 return in4
->s_addr
== INADDR_ANY
? EADDRNOTAVAIL
: 0;
1131 * Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
1132 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
1133 * positive errno value.
1136 netdev_bsd_set_in4(struct netdev
*netdev_
, struct in_addr addr
,
1137 struct in_addr mask
)
1139 struct netdev_dev_bsd
*netdev_dev
=
1140 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
1143 error
= do_set_addr(netdev_
, SIOCSIFADDR
, "SIOCSIFADDR", addr
);
1145 netdev_dev
->cache_valid
|= VALID_IN4
;
1146 netdev_dev
->in4
= addr
;
1147 if (addr
.s_addr
!= INADDR_ANY
) {
1148 error
= do_set_addr(netdev_
, SIOCSIFNETMASK
,
1149 "SIOCSIFNETMASK", mask
);
1151 netdev_dev_bsd_changed(netdev_dev
);
1157 netdev_bsd_get_in6(const struct netdev
*netdev_
, struct in6_addr
*in6
)
1159 struct netdev_dev_bsd
*netdev_dev
=
1160 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
1161 if (!(netdev_dev
->cache_valid
& VALID_IN6
)) {
1162 struct ifaddrs
*ifa
, *head
;
1163 struct sockaddr_in6
*sin6
;
1164 const char *netdev_name
= netdev_get_name(netdev_
);
1166 if (getifaddrs(&head
) != 0) {
1167 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name
,
1172 for (ifa
= head
; ifa
; ifa
= ifa
->ifa_next
) {
1173 if (ifa
->ifa_addr
->sa_family
== AF_INET6
&&
1174 !strcmp(ifa
->ifa_name
, netdev_name
)) {
1175 sin6
= (struct sockaddr_in6
*)ifa
->ifa_addr
;
1177 memcpy(&netdev_dev
->in6
, &sin6
->sin6_addr
, sin6
->sin6_len
);
1178 netdev_dev
->cache_valid
|= VALID_IN6
;
1179 *in6
= netdev_dev
->in6
;
1185 return EADDRNOTAVAIL
;
1187 *in6
= netdev_dev
->in6
;
1192 make_in4_sockaddr(struct sockaddr
*sa
, struct in_addr addr
)
1194 struct sockaddr_in sin
;
1195 memset(&sin
, 0, sizeof sin
);
1196 sin
.sin_family
= AF_INET
;
1197 sin
.sin_addr
= addr
;
1200 memset(sa
, 0, sizeof *sa
);
1201 memcpy(sa
, &sin
, sizeof sin
);
1205 do_set_addr(struct netdev
*netdev
,
1206 int ioctl_nr
, const char *ioctl_name
, struct in_addr addr
)
1209 make_in4_sockaddr(&ifr
.ifr_addr
, addr
);
1210 return netdev_bsd_do_ioctl(netdev
, &ifr
, ioctl_nr
, ioctl_name
);
1214 nd_to_iff_flags(enum netdev_flags nd
)
1217 if (nd
& NETDEV_UP
) {
1220 if (nd
& NETDEV_PROMISC
) {
1222 iff
|= IFF_PPROMISC
;
1228 iff_to_nd_flags(int iff
)
1230 enum netdev_flags nd
= 0;
1234 if (iff
& IFF_PROMISC
) {
1235 nd
|= NETDEV_PROMISC
;
1241 netdev_bsd_update_flags(struct netdev_dev
*dev_
, enum netdev_flags off
,
1242 enum netdev_flags on
, enum netdev_flags
*old_flagsp
)
1244 struct netdev_dev_bsd
*netdev_dev
;
1245 int old_flags
, new_flags
;
1248 netdev_dev
= netdev_dev_bsd_cast(dev_
);
1249 error
= get_flags(dev_
, &old_flags
);
1251 *old_flagsp
= iff_to_nd_flags(old_flags
);
1252 new_flags
= (old_flags
& ~nd_to_iff_flags(off
)) | nd_to_iff_flags(on
);
1253 if (new_flags
!= old_flags
) {
1254 error
= set_flags(netdev_dev_get_name(dev_
), new_flags
);
1255 netdev_dev_bsd_changed(netdev_dev
);
1262 netdev_bsd_change_seq(const struct netdev
*netdev
)
1264 return netdev_dev_bsd_cast(netdev_get_dev(netdev
))->change_seq
;
1268 const struct netdev_class netdev_bsd_class
= {
1274 netdev_bsd_create_system
,
1276 NULL
, /* get_config */
1277 NULL
, /* set_config */
1278 NULL
, /* get_tunnel_config */
1279 netdev_bsd_open_system
,
1285 netdev_bsd_send_wait
,
1287 netdev_bsd_set_etheraddr
,
1288 netdev_bsd_get_etheraddr
,
1291 netdev_bsd_get_ifindex
,
1292 netdev_bsd_get_carrier
,
1293 NULL
, /* get_carrier_resets */
1294 NULL
, /* set_miimon_interval */
1295 netdev_bsd_get_stats
,
1296 NULL
, /* set_stats */
1298 netdev_bsd_get_features
,
1299 NULL
, /* set_advertisement */
1300 NULL
, /* set_policing */
1301 NULL
, /* get_qos_type */
1302 NULL
, /* get_qos_capabilities */
1305 NULL
, /* get_queue */
1306 NULL
, /* set_queue */
1307 NULL
, /* delete_queue */
1308 NULL
, /* get_queue_stats */
1309 NULL
, /* dump_queue */
1310 NULL
, /* dump_queue_stats */
1315 NULL
, /* add_router */
1316 NULL
, /* get_next_hop */
1317 NULL
, /* get_status */
1318 NULL
, /* arp_lookup */
1320 netdev_bsd_update_flags
,
1322 netdev_bsd_change_seq
1325 const struct netdev_class netdev_tap_class
= {
1331 netdev_bsd_create_tap
,
1333 NULL
, /* get_config */
1334 NULL
, /* set_config */
1335 NULL
, /* get_tunnel_config */
1336 netdev_bsd_open_system
,
1342 netdev_bsd_send_wait
,
1344 netdev_bsd_set_etheraddr
,
1345 netdev_bsd_get_etheraddr
,
1348 netdev_bsd_get_ifindex
,
1349 netdev_bsd_get_carrier
,
1350 NULL
, /* get_carrier_resets */
1351 NULL
, /* set_miimon_interval */
1352 netdev_bsd_get_stats
,
1353 NULL
, /* set_stats */
1355 netdev_bsd_get_features
,
1356 NULL
, /* set_advertisement */
1357 NULL
, /* set_policing */
1358 NULL
, /* get_qos_type */
1359 NULL
, /* get_qos_capabilities */
1362 NULL
, /* get_queue */
1363 NULL
, /* set_queue */
1364 NULL
, /* delete_queue */
1365 NULL
, /* get_queue_stats */
1366 NULL
, /* dump_queue */
1367 NULL
, /* dump_queue_stats */
1372 NULL
, /* add_router */
1373 NULL
, /* get_next_hop */
1374 NULL
, /* get_status */
1375 NULL
, /* arp_lookup */
1377 netdev_bsd_update_flags
,
1379 netdev_bsd_change_seq
1382 static const struct netdev_rx_class netdev_rx_bsd_class
= {
1383 netdev_rx_bsd_destroy
,
1386 netdev_rx_bsd_drain
,
1391 destroy_tap(int fd
, const char *name
)
1396 strcpy(ifr
.ifr_name
, name
);
1397 /* XXX What to do if this call fails? */
1398 ioctl(af_inet_sock
, SIOCIFDESTROY
, &ifr
);
1402 get_flags(const struct netdev_dev
*dev
, int *flags
)
1407 error
= netdev_bsd_do_ioctl(dev
->name
, &ifr
, SIOCGIFFLAGS
, "SIOCGIFFLAGS");
1409 *flags
= 0xFFFF0000 & (ifr
.ifr_flagshigh
<< 16);
1410 *flags
|= 0x0000FFFF & ifr
.ifr_flags
;
1416 set_flags(const char *name
, int flags
)
1420 ifr
.ifr_flags
= 0x0000FFFF & flags
;
1421 ifr
.ifr_flagshigh
= (0xFFFF0000 & flags
) >> 16;
1423 return netdev_bsd_do_ioctl(name
, &ifr
, SIOCSIFFLAGS
, "SIOCSIFFLAGS");
1427 get_ifindex(const struct netdev
*netdev_
, int *ifindexp
)
1429 struct netdev_dev_bsd
*netdev_dev
=
1430 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
1432 if (!(netdev_dev
->cache_valid
& VALID_IFINDEX
)) {
1433 int ifindex
= if_nametoindex(netdev_get_name(netdev_
));
1437 netdev_dev
->cache_valid
|= VALID_IFINDEX
;
1438 netdev_dev
->ifindex
= ifindex
;
1440 *ifindexp
= netdev_dev
->ifindex
;
1445 get_etheraddr(const char *netdev_name
, uint8_t ea
[ETH_ADDR_LEN
])
1447 struct ifaddrs
*head
;
1448 struct ifaddrs
*ifa
;
1449 struct sockaddr_dl
*sdl
;
1451 if (getifaddrs(&head
) != 0) {
1452 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name
,
1457 for (ifa
= head
; ifa
; ifa
= ifa
->ifa_next
) {
1458 if (ifa
->ifa_addr
->sa_family
== AF_LINK
) {
1459 if (!strcmp(ifa
->ifa_name
, netdev_name
)) {
1460 sdl
= (struct sockaddr_dl
*)ifa
->ifa_addr
;
1462 memcpy(ea
, LLADDR(sdl
), sdl
->sdl_alen
);
1470 VLOG_ERR("could not find ethernet address for %s device", netdev_name
);
1476 set_etheraddr(const char *netdev_name
, int hwaddr_family
,
1477 int hwaddr_len
, const uint8_t mac
[ETH_ADDR_LEN
])
1481 memset(&ifr
, 0, sizeof ifr
);
1482 strncpy(ifr
.ifr_name
, netdev_name
, sizeof ifr
.ifr_name
);
1483 ifr
.ifr_addr
.sa_family
= hwaddr_family
;
1484 ifr
.ifr_addr
.sa_len
= hwaddr_len
;
1485 memcpy(ifr
.ifr_addr
.sa_data
, mac
, hwaddr_len
);
1486 if (ioctl(af_inet_sock
, SIOCSIFLLADDR
, &ifr
) < 0) {
1487 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1488 netdev_name
, strerror(errno
));
1495 netdev_bsd_do_ioctl(const char *name
, struct ifreq
*ifr
, unsigned long cmd
,
1496 const char *cmd_name
)
1498 strncpy(ifr
->ifr_name
, name
, sizeof ifr
->ifr_name
);
1499 if (ioctl(af_inet_sock
, cmd
, ifr
) == -1) {
1500 VLOG_DBG_RL(&rl
, "%s: ioctl(%s) failed: %s", name
, cmd_name
,