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.
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"
45 #include "netdev-provider.h"
47 #include "openflow/openflow.h"
49 #include "poll-loop.h"
50 #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 two 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. Again, private information
70 * such as file descriptor etc. are stored in our
71 * own struct netdev_bsd which includes a struct netdev.
73 * Both 'struct netdev' and 'struct netdev_dev' are referenced
74 * in containers which hold pointers to the data structures.
75 * We can reach our own struct netdev_XXX_bsd by putting a
76 * struct netdev_XXX within our own struct, and using CONTAINER_OF
77 * to access the parent structure.
82 int netdev_fd
; /* Selectable file descriptor for the network device.
83 This descriptor will be used for polling operations */
85 pcap_t
*pcap_handle
; /* Packet capture descriptor for a system network
89 struct netdev_dev_bsd
{
90 struct netdev_dev netdev_dev
;
91 unsigned int cache_valid
;
92 unsigned int change_seq
;
95 uint8_t etheraddr
[ETH_ADDR_LEN
];
102 int tap_fd
; /* TAP character device, if any */
107 VALID_IFINDEX
= 1 << 0,
108 VALID_ETHERADDR
= 1 << 1,
112 VALID_CARRIER
= 1 << 5
115 /* An AF_INET socket (used for ioctl operations). */
116 static int af_inet_sock
= -1;
118 #define PCAP_SNAPLEN 2048
122 * Notifier used to invalidate device informations in case of status change.
124 * It will be registered with a 'rtbsd_notifier_register()' when the first
125 * device will be created with the call of either 'netdev_bsd_tap_create()' or
126 * 'netdev_bsd_system_create()'.
128 * The callback associated with this notifier ('netdev_bsd_cache_cb()') will
129 * invalidate cached information about the device.
131 static struct rtbsd_notifier netdev_bsd_cache_notifier
;
132 static int cache_notifier_refcount
;
134 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 20);
136 static int netdev_bsd_do_ioctl(const struct netdev
*, struct ifreq
*,
137 unsigned long cmd
, const char *cmd_name
);
138 static void destroy_tap(int fd
, const char *name
);
139 static int get_flags(const struct netdev
*, int *flagsp
);
140 static int set_flags(struct netdev
*, int flags
);
141 static int do_set_addr(struct netdev
*netdev
,
142 int ioctl_nr
, const char *ioctl_name
,
143 struct in_addr addr
);
144 static int get_etheraddr(const char *netdev_name
, uint8_t ea
[ETH_ADDR_LEN
]);
145 static int set_etheraddr(const char *netdev_name
, int hwaddr_family
,
146 int hwaddr_len
, const uint8_t[ETH_ADDR_LEN
]);
147 static int get_ifindex(const struct netdev
*, int *ifindexp
);
149 static int netdev_bsd_init(void);
152 is_netdev_bsd_class(const struct netdev_class
*netdev_class
)
154 return netdev_class
->init
== netdev_bsd_init
;
157 static struct netdev_bsd
*
158 netdev_bsd_cast(const struct netdev
*netdev
)
160 assert(is_netdev_bsd_class(netdev_dev_get_class(netdev_get_dev(netdev
))));
161 return CONTAINER_OF(netdev
, struct netdev_bsd
, netdev
);
164 static struct netdev_dev_bsd
*
165 netdev_dev_bsd_cast(const struct netdev_dev
*netdev_dev
)
167 assert(is_netdev_bsd_class(netdev_dev_get_class(netdev_dev
)));
168 return CONTAINER_OF(netdev_dev
, struct netdev_dev_bsd
, netdev_dev
);
171 /* Initialize the AF_INET socket used for ioctl operations */
173 netdev_bsd_init(void)
175 static int status
= -1;
177 if (status
>= 0) { /* already initialized */
181 af_inet_sock
= socket(AF_INET
, SOCK_DGRAM
, 0);
182 status
= af_inet_sock
>= 0 ? 0 : errno
;
185 VLOG_ERR("failed to create inet socket: %s", strerror(status
));
192 * Perform periodic work needed by netdev. In BSD netdevs it checks for any
193 * interface status changes, and eventually calls all the user callbacks.
198 rtbsd_notifier_run();
202 * Arranges for poll_block() to wake up if the "run" member function needs to
206 netdev_bsd_wait(void)
208 rtbsd_notifier_wait();
212 netdev_dev_bsd_changed(struct netdev_dev_bsd
*dev
)
215 if (!dev
->change_seq
) {
220 /* Invalidate cache in case of interface status change. */
222 netdev_bsd_cache_cb(const struct rtbsd_change
*change
,
223 void *aux OVS_UNUSED
)
225 struct netdev_dev_bsd
*dev
;
228 struct netdev_dev
*base_dev
= netdev_dev_from_name(change
->if_name
);
231 const struct netdev_class
*netdev_class
=
232 netdev_dev_get_class(base_dev
);
234 if (is_netdev_bsd_class(netdev_class
)) {
235 dev
= netdev_dev_bsd_cast(base_dev
);
236 dev
->cache_valid
= 0;
237 netdev_dev_bsd_changed(dev
);
242 * XXX the API is lacking, we should be able to iterate on the list of
243 * netdevs without having to store the info in a temp shash.
245 struct shash device_shash
;
246 struct shash_node
*node
;
248 shash_init(&device_shash
);
249 netdev_dev_get_devices(&netdev_bsd_class
, &device_shash
);
250 SHASH_FOR_EACH (node
, &device_shash
) {
252 dev
->cache_valid
= 0;
253 netdev_dev_bsd_changed(dev
);
255 shash_destroy(&device_shash
);
260 cache_notifier_ref(void)
264 if (!cache_notifier_refcount
) {
265 ret
= rtbsd_notifier_register(&netdev_bsd_cache_notifier
,
266 netdev_bsd_cache_cb
, NULL
);
271 cache_notifier_refcount
++;
276 cache_notifier_unref(void)
278 cache_notifier_refcount
--;
279 if (cache_notifier_refcount
== 0) {
280 rtbsd_notifier_unregister(&netdev_bsd_cache_notifier
);
285 /* Allocate a netdev_dev_bsd structure */
287 netdev_bsd_create_system(const struct netdev_class
*class, const char *name
,
288 struct netdev_dev
**netdev_devp
)
290 struct netdev_dev_bsd
*netdev_dev
;
293 error
= cache_notifier_ref();
298 netdev_dev
= xzalloc(sizeof *netdev_dev
);
299 netdev_dev
->change_seq
= 1;
300 netdev_dev_init(&netdev_dev
->netdev_dev
, name
, class);
301 *netdev_devp
= &netdev_dev
->netdev_dev
;
307 * Allocate a netdev_dev_bsd structure with 'tap' class.
310 netdev_bsd_create_tap(const struct netdev_class
*class, const char *name
,
311 struct netdev_dev
**netdev_devp
)
313 struct netdev_dev_bsd
*netdev_dev
= NULL
;
317 error
= cache_notifier_ref();
322 /* allocate the device structure and set the internal flag */
323 netdev_dev
= xzalloc(sizeof *netdev_dev
);
325 memset(&ifr
, 0, sizeof(ifr
));
327 /* Create a tap device by opening /dev/tap. The TAPGIFNAME ioctl is used
328 * to retrieve the name of the tap device. */
329 netdev_dev
->tap_fd
= open("/dev/tap", O_RDWR
);
330 netdev_dev
->change_seq
= 1;
331 if (netdev_dev
->tap_fd
< 0) {
333 VLOG_WARN("opening \"/dev/tap\" failed: %s", strerror(error
));
334 goto error_undef_notifier
;
337 /* Retrieve tap name (e.g. tap0) */
338 if (ioctl(netdev_dev
->tap_fd
, TAPGIFNAME
, &ifr
) == -1) {
339 /* XXX Need to destroy the device? */
341 goto error_undef_notifier
;
344 /* Change the name of the tap device */
345 ifr
.ifr_data
= (void *)name
;
346 if (ioctl(af_inet_sock
, SIOCSIFNAME
, &ifr
) == -1) {
348 destroy_tap(netdev_dev
->tap_fd
, ifr
.ifr_name
);
349 goto error_undef_notifier
;
352 /* set non-blocking. */
353 error
= set_nonblocking(netdev_dev
->tap_fd
);
355 destroy_tap(netdev_dev
->tap_fd
, name
);
356 goto error_undef_notifier
;
360 ifr
.ifr_flags
= (uint16_t)IFF_UP
;
361 ifr
.ifr_flagshigh
= 0;
362 strncpy(ifr
.ifr_name
, name
, sizeof ifr
.ifr_name
);
363 if (ioctl(af_inet_sock
, SIOCSIFFLAGS
, &ifr
) == -1) {
365 destroy_tap(netdev_dev
->tap_fd
, name
);
366 goto error_undef_notifier
;
369 /* initialize the device structure and
370 * link the structure to its netdev */
371 netdev_dev_init(&netdev_dev
->netdev_dev
, name
, class);
372 *netdev_devp
= &netdev_dev
->netdev_dev
;
376 error_undef_notifier
:
377 cache_notifier_unref();
384 netdev_bsd_destroy(struct netdev_dev
*netdev_dev_
)
386 struct netdev_dev_bsd
*netdev_dev
= netdev_dev_bsd_cast(netdev_dev_
);
388 cache_notifier_unref();
390 if (netdev_dev
->tap_fd
>= 0 &&
391 !strcmp(netdev_dev_get_type(netdev_dev_
), "tap")) {
392 destroy_tap(netdev_dev
->tap_fd
, netdev_dev_get_name(netdev_dev_
));
399 netdev_bsd_open_system(struct netdev_dev
*netdev_dev_
, struct netdev
**netdevp
)
401 struct netdev_dev_bsd
*netdev_dev
= netdev_dev_bsd_cast(netdev_dev_
);
402 struct netdev_bsd
*netdev
;
404 enum netdev_flags flags
;
406 /* Allocate network device. */
407 netdev
= xcalloc(1, sizeof *netdev
);
408 netdev
->netdev_fd
= -1;
409 netdev_init(&netdev
->netdev
, netdev_dev_
);
411 /* Verify that the netdev really exists by attempting to read its flags */
412 error
= netdev_get_flags(&netdev
->netdev
, &flags
);
413 if (error
== ENXIO
) {
417 /* The first user that opens a tap port(from dpif_create_and_open()) will
418 * receive the file descriptor associated with the tap device. Instead, the
419 * following users will open the tap device as a normal 'system' device. */
420 if (!strcmp(netdev_dev_get_type(netdev_dev_
), "tap") &&
421 !netdev_dev
->tap_opened
) {
422 netdev_dev
->tap_opened
= true;
423 netdev
->netdev_fd
= netdev_dev
->tap_fd
;
426 *netdevp
= &netdev
->netdev
;
430 netdev_uninit(&netdev
->netdev
, true);
436 /* Close a 'netdev'. */
438 netdev_bsd_close(struct netdev
*netdev_
)
440 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
442 if (netdev
->netdev_fd
>= 0 && strcmp(netdev_get_type(netdev_
), "tap")) {
443 pcap_close(netdev
->pcap_handle
);
450 netdev_bsd_listen(struct netdev
*netdev_
)
452 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
453 char errbuf
[PCAP_ERRBUF_SIZE
];
458 if (netdev
->netdev_fd
>= 0) {
462 /* open the pcap device. The device is opened in non-promiscuous mode
463 * because the interface flags are manually set by the caller. */
465 netdev
->pcap_handle
= pcap_open_live(netdev_get_name(netdev_
), PCAP_SNAPLEN
,
467 if (netdev
->pcap_handle
== NULL
) {
468 VLOG_ERR("%s: pcap_open_live failed: %s",
469 netdev_get_name(netdev_
), errbuf
);
472 } else if (errbuf
[0] != '\0') {
473 VLOG_WARN("%s: pcap_open_live: %s",
474 netdev_get_name(netdev_
), errbuf
);
477 netdev_dev_bsd_changed(netdev_dev_bsd_cast(netdev_get_dev(netdev_
)));
479 /* initialize netdev->netdev_fd */
480 fd
= pcap_get_selectable_fd(netdev
->pcap_handle
);
486 /* Set non-blocking mode. Also the BIOCIMMEDIATE ioctl must be called
487 * on the file descriptor returned by pcap_get_selectable_fd to achieve
488 * a real non-blocking behaviour.*/
489 error
= pcap_setnonblock(netdev
->pcap_handle
, 1, errbuf
);
495 /* This call assure that reads return immediately upon packet reception.
496 * Otherwise, a read will block until either the kernel buffer becomes
497 * full or a timeout occurs. */
498 if(ioctl(fd
, BIOCIMMEDIATE
, &one
) < 0 ) {
499 VLOG_ERR("ioctl(BIOCIMMEDIATE) on %s device failed: %s",
500 netdev_get_name(netdev_
), strerror(errno
));
505 /* Capture only incoming packets */
506 error
= pcap_setdirection(netdev
->pcap_handle
, PCAP_D_IN
);
512 netdev
->netdev_fd
= fd
;
517 close(netdev
->netdev_fd
);
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
568 * on the 'netdev' pcap descriptor.
571 netdev_bsd_recv_system(struct netdev_bsd
*netdev
, void *data
, size_t size
)
576 if (netdev
->netdev_fd
< 0) {
580 /* prepare the pcap argument to store the packet */
585 ret
= pcap_dispatch(netdev
->pcap_handle
, 1, proc_pkt
, (u_char
*)&arg
);
588 return arg
.retval
; /* arg.retval < 0 is handled in the caller */
591 if (errno
== EINTR
) {
601 * This function attempts to receive a packet from the specified network
602 * device. It is assumed that the network device is a tap device and the
603 * 'netdev_fd' member of the 'netdev' structure is initialized with the tap
607 netdev_bsd_recv_tap(struct netdev_bsd
*netdev
, void *data
, size_t size
)
609 if (netdev
->netdev_fd
< 0) {
614 ssize_t retval
= read(netdev
->netdev_fd
, data
, size
);
617 } else if (errno
!= EINTR
) {
618 if (errno
!= EAGAIN
) {
619 VLOG_WARN_RL(&rl
, "error receiving Ethernet packet on %s: %s",
620 strerror(errno
), netdev
->netdev
.netdev_dev
->name
);
629 * According with the nature of the device a different function must be called.
630 * If the device is the bridge local port the 'netdev_bsd_recv_tap' function
631 * must be called, otherwise the 'netdev_bsd_recv_system' function is called.
633 * type!="tap" ---> system device.
634 * type=="tap" && netdev_fd == tap_fd ---> internal tap device
635 * type=="tap" && netdev_fd != tap_fd ---> internal tap device
640 netdev_bsd_recv(struct netdev
*netdev_
, void* data
, size_t size
)
642 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
643 struct netdev_dev_bsd
* netdev_dev
=
644 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
646 if (!strcmp(netdev_get_type(netdev_
), "tap") &&
647 netdev
->netdev_fd
== netdev_dev
->tap_fd
) {
648 return netdev_bsd_recv_tap(netdev
, data
, size
);
650 return netdev_bsd_recv_system(netdev
, data
, size
);
656 * Registers with the poll loop to wake up from the next call to poll_block()
657 * when a packet is ready to be received with netdev_recv() on 'netdev'.
660 netdev_bsd_recv_wait(struct netdev
*netdev_
)
662 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
664 if (netdev
->netdev_fd
>= 0) {
665 poll_fd_wait(netdev
->netdev_fd
, POLLIN
);
669 /* Discards all packets waiting to be received from 'netdev'. */
671 netdev_bsd_drain(struct netdev
*netdev_
)
674 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
676 strcpy(ifr
.ifr_name
, netdev_get_name(netdev_
));
677 if (ioctl(netdev
->netdev_fd
, BIOCFLUSH
, &ifr
) == -1) {
678 VLOG_DBG_RL(&rl
, "%s: ioctl(BIOCFLUSH) failed: %s",
679 netdev_get_name(netdev_
), strerror(errno
));
686 * Send a packet on the specified network device. The device could be either a
687 * system or a tap device.
690 netdev_bsd_send(struct netdev
*netdev_
, const void *data
, size_t size
)
692 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
693 struct netdev_dev_bsd
* netdev_dev
=
694 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
696 if (netdev
->netdev_fd
< 0) {
702 if (!strcmp(netdev_get_type(netdev_
), "tap") &&
703 netdev_dev
->tap_fd
== netdev
->netdev_fd
) {
704 retval
= write(netdev
->netdev_fd
, data
, size
);
706 retval
= pcap_inject(netdev
->pcap_handle
, data
, size
);
709 if (errno
== EINTR
) {
711 } else if (errno
!= EAGAIN
) {
712 VLOG_WARN_RL(&rl
, "error sending Ethernet packet on %s: %s",
713 netdev_get_name(netdev_
), strerror(errno
));
716 } else if (retval
!= size
) {
717 VLOG_WARN_RL(&rl
, "sent partial Ethernet packet (%zd bytes of "
718 "%zu) on %s", retval
, size
,
719 netdev_get_name(netdev_
));
728 * Registers with the poll loop to wake up from the next call to poll_block()
729 * when the packet transmission queue has sufficient room to transmit a packet
730 * with netdev_send().
733 netdev_bsd_send_wait(struct netdev
*netdev_
)
735 struct netdev_bsd
*netdev
= netdev_bsd_cast(netdev_
);
737 if (netdev
->netdev_fd
< 0) { /* Nothing to do. */
741 if (strcmp(netdev_get_type(netdev_
), "tap")) {
742 poll_fd_wait(netdev
->netdev_fd
, POLLOUT
);
744 /* TAP device always accepts packets. */
745 poll_immediate_wake();
750 * Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
751 * otherwise a positive errno value.
754 netdev_bsd_set_etheraddr(struct netdev
*netdev_
,
755 const uint8_t mac
[ETH_ADDR_LEN
])
757 struct netdev_dev_bsd
*netdev_dev
=
758 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
761 if (!(netdev_dev
->cache_valid
& VALID_ETHERADDR
)
762 || !eth_addr_equals(netdev_dev
->etheraddr
, mac
)) {
763 error
= set_etheraddr(netdev_get_name(netdev_
), AF_LINK
, ETH_ADDR_LEN
,
766 netdev_dev
->cache_valid
|= VALID_ETHERADDR
;
767 memcpy(netdev_dev
->etheraddr
, mac
, ETH_ADDR_LEN
);
768 netdev_dev_bsd_changed(netdev_dev
);
777 * Returns a pointer to 'netdev''s MAC address. The caller must not modify or
778 * free the returned buffer.
781 netdev_bsd_get_etheraddr(const struct netdev
*netdev_
,
782 uint8_t mac
[ETH_ADDR_LEN
])
784 struct netdev_dev_bsd
*netdev_dev
=
785 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
787 if (!(netdev_dev
->cache_valid
& VALID_ETHERADDR
)) {
788 int error
= get_etheraddr(netdev_get_name(netdev_
),
789 netdev_dev
->etheraddr
);
793 netdev_dev
->cache_valid
|= VALID_ETHERADDR
;
795 memcpy(mac
, netdev_dev
->etheraddr
, ETH_ADDR_LEN
);
801 * Returns the maximum size of transmitted (and received) packets on 'netdev',
802 * in bytes, not including the hardware header; thus, this is typically 1500
803 * bytes for Ethernet devices.
806 netdev_bsd_get_mtu(const struct netdev
*netdev_
, int *mtup
)
808 struct netdev_dev_bsd
*netdev_dev
=
809 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
811 if (!(netdev_dev
->cache_valid
& VALID_MTU
)) {
815 error
= netdev_bsd_do_ioctl(netdev_
, &ifr
, SIOCGIFMTU
, "SIOCGIFMTU");
819 netdev_dev
->mtu
= ifr
.ifr_mtu
;
820 netdev_dev
->cache_valid
|= VALID_MTU
;
823 *mtup
= netdev_dev
->mtu
;
828 netdev_bsd_get_ifindex(const struct netdev
*netdev
)
832 error
= get_ifindex(netdev
, &ifindex
);
833 return error
? -error
: ifindex
;
837 netdev_bsd_get_carrier(const struct netdev
*netdev_
, bool *carrier
)
839 struct netdev_dev_bsd
*netdev_dev
=
840 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
842 if (!(netdev_dev
->cache_valid
& VALID_CARRIER
)) {
843 struct ifmediareq ifmr
;
845 memset(&ifmr
, 0, sizeof(ifmr
));
846 strncpy(ifmr
.ifm_name
, netdev_get_name(netdev_
), sizeof ifmr
.ifm_name
);
848 if (ioctl(af_inet_sock
, SIOCGIFMEDIA
, &ifmr
) == -1) {
849 VLOG_DBG_RL(&rl
, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
850 netdev_get_name(netdev_
), strerror(errno
));
854 netdev_dev
->carrier
= (ifmr
.ifm_status
& IFM_ACTIVE
) == IFM_ACTIVE
;
855 netdev_dev
->cache_valid
|= VALID_CARRIER
;
857 /* If the interface doesn't report whether the media is active,
858 * just assume it is active. */
859 if ((ifmr
.ifm_status
& IFM_AVALID
) == 0) {
860 netdev_dev
->carrier
= true;
863 *carrier
= netdev_dev
->carrier
;
868 /* Retrieves current device stats for 'netdev'. */
870 netdev_bsd_get_stats(const struct netdev
*netdev_
, struct netdev_stats
*stats
)
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_
), 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_
), strerror(errno
));
901 } else if (!strcmp(ifmd
.ifmd_name
, netdev_get_name(netdev_
))) {
902 stats
->rx_packets
= ifmd
.ifmd_data
.ifi_ipackets
;
903 stats
->tx_packets
= ifmd
.ifmd_data
.ifi_opackets
;
904 stats
->rx_bytes
= ifmd
.ifmd_data
.ifi_ibytes
;
905 stats
->tx_bytes
= ifmd
.ifmd_data
.ifi_obytes
;
906 stats
->rx_errors
= ifmd
.ifmd_data
.ifi_ierrors
;
907 stats
->tx_errors
= ifmd
.ifmd_data
.ifi_oerrors
;
908 stats
->rx_dropped
= ifmd
.ifmd_data
.ifi_iqdrops
;
909 stats
->tx_dropped
= 0;
910 stats
->multicast
= ifmd
.ifmd_data
.ifi_imcasts
;
911 stats
->collisions
= ifmd
.ifmd_data
.ifi_collisions
;
913 stats
->rx_length_errors
= 0;
914 stats
->rx_over_errors
= 0;
915 stats
->rx_crc_errors
= 0;
916 stats
->rx_frame_errors
= 0;
917 stats
->rx_fifo_errors
= 0;
918 stats
->rx_missed_errors
= 0;
920 stats
->tx_aborted_errors
= 0;
921 stats
->tx_carrier_errors
= 0;
922 stats
->tx_fifo_errors
= 0;
923 stats
->tx_heartbeat_errors
= 0;
924 stats
->tx_window_errors
= 0;
933 netdev_bsd_parse_media(int media
)
935 uint32_t supported
= 0;
936 bool half_duplex
= media
& IFM_HDX
? true : false;
938 switch (IFM_SUBTYPE(media
)) {
943 supported
|= half_duplex
? NETDEV_F_10MB_HD
: NETDEV_F_10MB_FD
;
944 supported
|= NETDEV_F_COPPER
;
948 supported
|= half_duplex
? NETDEV_F_10MB_HD
: NETDEV_F_10MB_FD
;
949 supported
|= NETDEV_F_FIBER
;
956 supported
|= half_duplex
? NETDEV_F_100MB_HD
: NETDEV_F_100MB_FD
;
957 supported
|= NETDEV_F_COPPER
;
961 supported
|= half_duplex
? NETDEV_F_100MB_HD
: NETDEV_F_100MB_FD
;
962 supported
|= NETDEV_F_FIBER
;
967 supported
|= half_duplex
? NETDEV_F_1GB_HD
: NETDEV_F_1GB_FD
;
968 supported
|= NETDEV_F_COPPER
;
973 supported
|= half_duplex
? NETDEV_F_1GB_HD
: NETDEV_F_1GB_FD
;
974 supported
|= NETDEV_F_FIBER
;
978 supported
|= NETDEV_F_10GB_FD
;
979 supported
|= NETDEV_F_COPPER
;
984 supported
|= NETDEV_F_10GB_FD
;
985 supported
|= NETDEV_F_FIBER
;
992 if (IFM_SUBTYPE(media
) == IFM_AUTO
) {
993 supported
|= NETDEV_F_AUTONEG
;
996 if (media & IFM_ETH_FMASK) {
997 supported |= NETDEV_F_PAUSE;
1005 * Stores the features supported by 'netdev' into each of '*current',
1006 * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
1007 * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
1008 * successful, otherwise a positive errno value. On failure, all of the
1009 * passed-in values are set to 0.
1012 netdev_bsd_get_features(const struct netdev
*netdev
,
1013 enum netdev_features
*current
, uint32_t *advertised
,
1014 enum netdev_features
*supported
, uint32_t *peer
)
1016 struct ifmediareq ifmr
;
1022 /* XXX Look into SIOCGIFCAP instead of SIOCGIFMEDIA */
1024 memset(&ifmr
, 0, sizeof(ifmr
));
1025 strncpy(ifmr
.ifm_name
, netdev_get_name(netdev
), sizeof ifmr
.ifm_name
);
1027 /* We make two SIOCGIFMEDIA ioctl calls. The first to determine the
1028 * number of supported modes, and a second with a buffer to retrieve
1030 if (ioctl(af_inet_sock
, SIOCGIFMEDIA
, &ifmr
) == -1) {
1031 VLOG_DBG_RL(&rl
, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1032 netdev_get_name(netdev
), strerror(errno
));
1036 media_list
= xcalloc(ifmr
.ifm_count
, sizeof(int));
1037 ifmr
.ifm_ulist
= media_list
;
1039 if (!IFM_TYPE(ifmr
.ifm_current
) & IFM_ETHER
) {
1040 VLOG_DBG_RL(&rl
, "%s: doesn't appear to be ethernet",
1041 netdev_get_name(netdev
));
1046 if (ioctl(af_inet_sock
, SIOCGIFMEDIA
, &ifmr
) == -1) {
1047 VLOG_DBG_RL(&rl
, "%s: ioctl(SIOCGIFMEDIA) failed: %s",
1048 netdev_get_name(netdev
), strerror(errno
));
1053 /* Current settings. */
1054 *current
= netdev_bsd_parse_media(ifmr
.ifm_active
);
1056 /* Advertised features. */
1057 *advertised
= netdev_bsd_parse_media(ifmr
.ifm_current
);
1059 /* Supported features. */
1061 for (i
= 0; i
< ifmr
.ifm_count
; i
++) {
1062 *supported
|= netdev_bsd_parse_media(ifmr
.ifm_ulist
[i
]);
1065 /* Peer advertisements. */
1066 *peer
= 0; /* XXX */
1075 * If 'netdev' has an assigned IPv4 address, sets '*in4' to that address (if
1076 * 'in4' is non-null) and returns true. Otherwise, returns false.
1079 netdev_bsd_get_in4(const struct netdev
*netdev_
, struct in_addr
*in4
,
1080 struct in_addr
*netmask
)
1082 struct netdev_dev_bsd
*netdev_dev
=
1083 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
1085 if (!(netdev_dev
->cache_valid
& VALID_IN4
)) {
1086 const struct sockaddr_in
*sin
;
1090 ifr
.ifr_addr
.sa_family
= AF_INET
;
1091 error
= netdev_bsd_do_ioctl(netdev_
, &ifr
,
1092 SIOCGIFADDR
, "SIOCGIFADDR");
1097 sin
= (struct sockaddr_in
*) &ifr
.ifr_addr
;
1098 netdev_dev
->in4
= sin
->sin_addr
;
1099 netdev_dev
->cache_valid
|= VALID_IN4
;
1100 error
= netdev_bsd_do_ioctl(netdev_
, &ifr
,
1101 SIOCGIFNETMASK
, "SIOCGIFNETMASK");
1105 *netmask
= ((struct sockaddr_in
*)&ifr
.ifr_addr
)->sin_addr
;
1107 *in4
= netdev_dev
->in4
;
1109 return in4
->s_addr
== INADDR_ANY
? EADDRNOTAVAIL
: 0;
1113 * Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
1114 * 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
1115 * positive errno value.
1118 netdev_bsd_set_in4(struct netdev
*netdev_
, struct in_addr addr
,
1119 struct in_addr mask
)
1121 struct netdev_dev_bsd
*netdev_dev
=
1122 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
1125 error
= do_set_addr(netdev_
, SIOCSIFADDR
, "SIOCSIFADDR", addr
);
1127 netdev_dev
->cache_valid
|= VALID_IN4
;
1128 netdev_dev
->in4
= addr
;
1129 if (addr
.s_addr
!= INADDR_ANY
) {
1130 error
= do_set_addr(netdev_
, SIOCSIFNETMASK
,
1131 "SIOCSIFNETMASK", mask
);
1133 netdev_dev_bsd_changed(netdev_dev
);
1139 netdev_bsd_get_in6(const struct netdev
*netdev_
, struct in6_addr
*in6
)
1141 struct netdev_dev_bsd
*netdev_dev
=
1142 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
1143 if (!(netdev_dev
->cache_valid
& VALID_IN6
)) {
1144 struct ifaddrs
*ifa
, *head
;
1145 struct sockaddr_in6
*sin6
;
1146 const char *netdev_name
= netdev_get_name(netdev_
);
1148 if (getifaddrs(&head
) != 0) {
1149 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name
,
1154 for (ifa
= head
; ifa
; ifa
= ifa
->ifa_next
) {
1155 if (ifa
->ifa_addr
->sa_family
== AF_INET6
&&
1156 !strcmp(ifa
->ifa_name
, netdev_name
)) {
1157 sin6
= (struct sockaddr_in6
*)ifa
->ifa_addr
;
1159 memcpy(&netdev_dev
->in6
, &sin6
->sin6_addr
, sin6
->sin6_len
);
1160 netdev_dev
->cache_valid
|= VALID_IN6
;
1161 *in6
= netdev_dev
->in6
;
1167 return EADDRNOTAVAIL
;
1169 *in6
= netdev_dev
->in6
;
1174 make_in4_sockaddr(struct sockaddr
*sa
, struct in_addr addr
)
1176 struct sockaddr_in sin
;
1177 memset(&sin
, 0, sizeof sin
);
1178 sin
.sin_family
= AF_INET
;
1179 sin
.sin_addr
= addr
;
1182 memset(sa
, 0, sizeof *sa
);
1183 memcpy(sa
, &sin
, sizeof sin
);
1187 do_set_addr(struct netdev
*netdev
,
1188 int ioctl_nr
, const char *ioctl_name
, struct in_addr addr
)
1191 make_in4_sockaddr(&ifr
.ifr_addr
, addr
);
1192 return netdev_bsd_do_ioctl(netdev
, &ifr
, ioctl_nr
, ioctl_name
);
1196 nd_to_iff_flags(enum netdev_flags nd
)
1199 if (nd
& NETDEV_UP
) {
1202 if (nd
& NETDEV_PROMISC
) {
1204 iff
|= IFF_PPROMISC
;
1210 iff_to_nd_flags(int iff
)
1212 enum netdev_flags nd
= 0;
1216 if (iff
& IFF_PROMISC
) {
1217 nd
|= NETDEV_PROMISC
;
1223 netdev_bsd_update_flags(struct netdev
*netdev
, enum netdev_flags off
,
1224 enum netdev_flags on
, enum netdev_flags
*old_flagsp
)
1226 int old_flags
, new_flags
;
1229 error
= get_flags(netdev
, &old_flags
);
1231 *old_flagsp
= iff_to_nd_flags(old_flags
);
1232 new_flags
= (old_flags
& ~nd_to_iff_flags(off
)) | nd_to_iff_flags(on
);
1233 if (new_flags
!= old_flags
) {
1234 error
= set_flags(netdev
, new_flags
);
1235 netdev_dev_bsd_changed(netdev_dev_bsd_cast(netdev_get_dev(netdev
)));
1242 netdev_bsd_change_seq(const struct netdev
*netdev
)
1244 return netdev_dev_bsd_cast(netdev_get_dev(netdev
))->change_seq
;
1248 const struct netdev_class netdev_bsd_class
= {
1254 netdev_bsd_create_system
,
1256 NULL
, /* get_config */
1257 NULL
, /* set_config */
1258 netdev_bsd_open_system
,
1264 netdev_bsd_recv_wait
,
1268 netdev_bsd_send_wait
,
1270 netdev_bsd_set_etheraddr
,
1271 netdev_bsd_get_etheraddr
,
1274 netdev_bsd_get_ifindex
,
1275 netdev_bsd_get_carrier
,
1276 NULL
, /* get_carrier_resets */
1277 NULL
, /* set_miimon_interval */
1278 netdev_bsd_get_stats
,
1279 NULL
, /* set_stats */
1281 netdev_bsd_get_features
,
1282 NULL
, /* set_advertisement */
1283 NULL
, /* set_policing */
1284 NULL
, /* get_qos_type */
1285 NULL
, /* get_qos_capabilities */
1288 NULL
, /* get_queue */
1289 NULL
, /* set_queue */
1290 NULL
, /* delete_queue */
1291 NULL
, /* get_queue_stats */
1292 NULL
, /* dump_queue */
1293 NULL
, /* dump_queue_stats */
1298 NULL
, /* add_router */
1299 NULL
, /* get_next_hop */
1300 NULL
, /* get_drv_info */
1301 NULL
, /* arp_lookup */
1303 netdev_bsd_update_flags
,
1305 netdev_bsd_change_seq
1308 const struct netdev_class netdev_tap_class
= {
1314 netdev_bsd_create_tap
,
1316 NULL
, /* get_config */
1317 NULL
, /* set_config */
1318 netdev_bsd_open_system
,
1324 netdev_bsd_recv_wait
,
1328 netdev_bsd_send_wait
,
1330 netdev_bsd_set_etheraddr
,
1331 netdev_bsd_get_etheraddr
,
1334 netdev_bsd_get_ifindex
,
1335 netdev_bsd_get_carrier
,
1336 NULL
, /* get_carrier_resets */
1337 NULL
, /* set_miimon_interval */
1338 netdev_bsd_get_stats
,
1339 NULL
, /* set_stats */
1341 netdev_bsd_get_features
,
1342 NULL
, /* set_advertisement */
1343 NULL
, /* set_policing */
1344 NULL
, /* get_qos_type */
1345 NULL
, /* get_qos_capabilities */
1348 NULL
, /* get_queue */
1349 NULL
, /* set_queue */
1350 NULL
, /* delete_queue */
1351 NULL
, /* get_queue_stats */
1352 NULL
, /* dump_queue */
1353 NULL
, /* dump_queue_stats */
1358 NULL
, /* add_router */
1359 NULL
, /* get_next_hop */
1360 NULL
, /* get_drv_info */
1361 NULL
, /* arp_lookup */
1363 netdev_bsd_update_flags
,
1365 netdev_bsd_change_seq
1370 destroy_tap(int fd
, const char *name
)
1375 strcpy(ifr
.ifr_name
, name
);
1376 /* XXX What to do if this call fails? */
1377 ioctl(af_inet_sock
, SIOCIFDESTROY
, &ifr
);
1381 get_flags(const struct netdev
*netdev
, int *flags
)
1386 error
= netdev_bsd_do_ioctl(netdev
, &ifr
, SIOCGIFFLAGS
, "SIOCGIFFLAGS");
1388 *flags
= 0xFFFF0000 & (ifr
.ifr_flagshigh
<< 16);
1389 *flags
|= 0x0000FFFF & ifr
.ifr_flags
;
1395 set_flags(struct netdev
*netdev
, int flags
)
1399 ifr
.ifr_flags
= 0x0000FFFF & flags
;
1400 ifr
.ifr_flagshigh
= (0xFFFF0000 & flags
) >> 16;
1402 return netdev_bsd_do_ioctl(netdev
, &ifr
, SIOCSIFFLAGS
, "SIOCSIFFLAGS");
1406 get_ifindex(const struct netdev
*netdev_
, int *ifindexp
)
1408 struct netdev_dev_bsd
*netdev_dev
=
1409 netdev_dev_bsd_cast(netdev_get_dev(netdev_
));
1411 if (!(netdev_dev
->cache_valid
& VALID_IFINDEX
)) {
1412 int ifindex
= if_nametoindex(netdev_get_name(netdev_
));
1416 netdev_dev
->cache_valid
|= VALID_IFINDEX
;
1417 netdev_dev
->ifindex
= ifindex
;
1419 *ifindexp
= netdev_dev
->ifindex
;
1424 get_etheraddr(const char *netdev_name
, uint8_t ea
[ETH_ADDR_LEN
])
1426 struct ifaddrs
*head
;
1427 struct ifaddrs
*ifa
;
1428 struct sockaddr_dl
*sdl
;
1430 if (getifaddrs(&head
) != 0) {
1431 VLOG_ERR("getifaddrs on %s device failed: %s", netdev_name
,
1436 for (ifa
= head
; ifa
; ifa
= ifa
->ifa_next
) {
1437 if (ifa
->ifa_addr
->sa_family
== AF_LINK
) {
1438 if (!strcmp(ifa
->ifa_name
, netdev_name
)) {
1439 sdl
= (struct sockaddr_dl
*)ifa
->ifa_addr
;
1441 memcpy(ea
, LLADDR(sdl
), sdl
->sdl_alen
);
1449 VLOG_ERR("could not find ethernet address for %s device", netdev_name
);
1455 set_etheraddr(const char *netdev_name
, int hwaddr_family
,
1456 int hwaddr_len
, const uint8_t mac
[ETH_ADDR_LEN
])
1460 memset(&ifr
, 0, sizeof ifr
);
1461 strncpy(ifr
.ifr_name
, netdev_name
, sizeof ifr
.ifr_name
);
1462 ifr
.ifr_addr
.sa_family
= hwaddr_family
;
1463 ifr
.ifr_addr
.sa_len
= hwaddr_len
;
1464 memcpy(ifr
.ifr_addr
.sa_data
, mac
, hwaddr_len
);
1465 if (ioctl(af_inet_sock
, SIOCSIFLLADDR
, &ifr
) < 0) {
1466 VLOG_ERR("ioctl(SIOCSIFLLADDR) on %s device failed: %s",
1467 netdev_name
, strerror(errno
));
1474 netdev_bsd_do_ioctl(const struct netdev
*netdev
, struct ifreq
*ifr
,
1475 unsigned long cmd
, const char *cmd_name
)
1477 strncpy(ifr
->ifr_name
, netdev_get_name(netdev
), sizeof ifr
->ifr_name
);
1478 if (ioctl(af_inet_sock
, cmd
, ifr
) == -1) {
1479 VLOG_DBG_RL(&rl
, "%s: ioctl(%s) failed: %s",
1480 netdev_get_name(netdev
), cmd_name
, strerror(errno
));