2 * Copyright (c) 1980, 1986, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)if.c 8.3 (Berkeley) 1/4/94
34 * $FreeBSD: src/sys/net/if.c,v 1.185 2004/03/13 02:35:03 brooks Exp $
35 * $DragonFly: src/sys/net/if.c,v 1.84 2008/11/15 11:58:16 sephe Exp $
38 #include "opt_compat.h"
39 #include "opt_inet6.h"
41 #include "opt_polling.h"
43 #include <sys/param.h>
44 #include <sys/malloc.h>
46 #include <sys/systm.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/socketops.h>
53 #include <sys/protosw.h>
54 #include <sys/kernel.h>
56 #include <sys/sockio.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 #include <sys/domain.h>
60 #include <sys/thread.h>
61 #include <sys/thread2.h>
62 #include <sys/serialize.h>
63 #include <sys/msgport2.h>
67 #include <net/if_arp.h>
68 #include <net/if_dl.h>
69 #include <net/if_types.h>
70 #include <net/if_var.h>
71 #include <net/ifq_var.h>
72 #include <net/radix.h>
73 #include <net/route.h>
74 #include <net/if_clone.h>
75 #include <net/netisr.h>
76 #include <net/netmsg2.h>
78 #include <machine/atomic.h>
79 #include <machine/stdarg.h>
80 #include <machine/smp.h>
82 #if defined(INET) || defined(INET6)
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/if_ether.h>
88 #include <netinet6/in6_var.h>
89 #include <netinet6/in6_ifattach.h>
93 #if defined(COMPAT_43)
94 #include <emulation/43bsd/43bsd_socket.h>
95 #endif /* COMPAT_43 */
97 struct netmsg_ifaddr
{
105 * System initialization
107 static void if_attachdomain(void *);
108 static void if_attachdomain1(struct ifnet
*);
109 static int ifconf(u_long
, caddr_t
, struct ucred
*);
110 static void ifinit(void *);
111 static void ifnetinit(void *);
112 static void if_slowtimo(void *);
113 static void link_rtrequest(int, struct rtentry
*, struct rt_addrinfo
*);
114 static int if_rtdel(struct radix_node
*, void *);
118 * XXX: declare here to avoid to include many inet6 related files..
119 * should be more generalized?
121 extern void nd6_setmtu(struct ifnet
*);
124 SYSCTL_NODE(_net
, PF_LINK
, link
, CTLFLAG_RW
, 0, "Link layers");
125 SYSCTL_NODE(_net_link
, 0, generic
, CTLFLAG_RW
, 0, "Generic link-management");
127 SYSINIT(interfaces
, SI_SUB_PROTO_IF
, SI_ORDER_FIRST
, ifinit
, NULL
)
128 /* Must be after netisr_init */
129 SYSINIT(ifnet
, SI_SUB_PRE_DRIVERS
, SI_ORDER_SECOND
, ifnetinit
, NULL
)
131 MALLOC_DEFINE(M_IFADDR
, "ifaddr", "interface address");
132 MALLOC_DEFINE(M_IFMADDR
, "ether_multi", "link-level multicast address");
134 int ifqmaxlen
= IFQ_MAXLEN
;
135 struct ifnethead ifnet
= TAILQ_HEAD_INITIALIZER(ifnet
);
137 /* In ifq_dispatch(), try to do direct ifnet.if_start first */
138 static int ifq_dispatch_schedonly
= 0;
139 SYSCTL_INT(_net_link_generic
, OID_AUTO
, ifq_dispatch_schedonly
, CTLFLAG_RW
,
140 &ifq_dispatch_schedonly
, 0, "");
142 /* In ifq_dispatch(), schedule ifnet.if_start without checking ifnet.if_snd */
143 static int ifq_dispatch_schednochk
= 0;
144 SYSCTL_INT(_net_link_generic
, OID_AUTO
, ifq_dispatch_schednochk
, CTLFLAG_RW
,
145 &ifq_dispatch_schednochk
, 0, "");
147 /* In if_devstart(), try to do direct ifnet.if_start first */
148 static int if_devstart_schedonly
= 0;
149 SYSCTL_INT(_net_link_generic
, OID_AUTO
, if_devstart_schedonly
, CTLFLAG_RW
,
150 &if_devstart_schedonly
, 0, "");
152 /* In if_devstart(), schedule ifnet.if_start without checking ifnet.if_snd */
153 static int if_devstart_schednochk
= 0;
154 SYSCTL_INT(_net_link_generic
, OID_AUTO
, if_devstart_schednochk
, CTLFLAG_RW
,
155 &if_devstart_schednochk
, 0, "");
158 /* Schedule ifnet.if_start on the current CPU */
159 static int if_start_oncpu_sched
= 0;
160 SYSCTL_INT(_net_link_generic
, OID_AUTO
, if_start_oncpu_sched
, CTLFLAG_RW
,
161 &if_start_oncpu_sched
, 0, "");
164 struct callout if_slowtimo_timer
;
167 struct ifnet
**ifindex2ifnet
= NULL
;
168 static struct thread ifnet_threads
[MAXCPU
];
169 static int ifnet_mpsafe_thread
= NETMSG_SERVICE_MPSAFE
;
171 #define IFQ_KTR_STRING "ifq=%p"
172 #define IFQ_KTR_ARG_SIZE (sizeof(void *))
174 #define KTR_IFQ KTR_ALL
176 KTR_INFO_MASTER(ifq
);
177 KTR_INFO(KTR_IFQ
, ifq
, enqueue
, 0, IFQ_KTR_STRING
, IFQ_KTR_ARG_SIZE
);
178 KTR_INFO(KTR_IFQ
, ifq
, dequeue
, 1, IFQ_KTR_STRING
, IFQ_KTR_ARG_SIZE
);
179 #define logifq(name, arg) KTR_LOG(ifq_ ## name, arg)
181 #define IF_START_KTR_STRING "ifp=%p"
182 #define IF_START_KTR_ARG_SIZE (sizeof(void *))
184 #define KTR_IF_START KTR_ALL
186 KTR_INFO_MASTER(if_start
);
187 KTR_INFO(KTR_IF_START
, if_start
, run
, 0,
188 IF_START_KTR_STRING
, IF_START_KTR_ARG_SIZE
);
189 KTR_INFO(KTR_IF_START
, if_start
, sched
, 1,
190 IF_START_KTR_STRING
, IF_START_KTR_ARG_SIZE
);
191 KTR_INFO(KTR_IF_START
, if_start
, avoid
, 2,
192 IF_START_KTR_STRING
, IF_START_KTR_ARG_SIZE
);
193 KTR_INFO(KTR_IF_START
, if_start
, contend_sched
, 3,
194 IF_START_KTR_STRING
, IF_START_KTR_ARG_SIZE
);
195 KTR_INFO(KTR_IF_START
, if_start
, chase_sched
, 4,
196 IF_START_KTR_STRING
, IF_START_KTR_ARG_SIZE
);
197 #define logifstart(name, arg) KTR_LOG(if_start_ ## name, arg)
200 * Network interface utility routines.
202 * Routines with ifa_ifwith* names take sockaddr *'s as
211 callout_init(&if_slowtimo_timer
);
214 TAILQ_FOREACH(ifp
, &ifnet
, if_link
) {
215 if (ifp
->if_snd
.ifq_maxlen
== 0) {
216 if_printf(ifp
, "XXX: driver didn't set ifq_maxlen\n");
217 ifp
->if_snd
.ifq_maxlen
= ifqmaxlen
;
226 if_start_cpuid(struct ifnet
*ifp
)
228 return ifp
->if_cpuid
;
231 #ifdef DEVICE_POLLING
233 if_start_cpuid_poll(struct ifnet
*ifp
)
235 int poll_cpuid
= ifp
->if_poll_cpuid
;
240 return ifp
->if_cpuid
;
245 if_start_ipifunc(void *arg
)
247 struct ifnet
*ifp
= arg
;
248 struct lwkt_msg
*lmsg
= &ifp
->if_start_nmsg
[mycpuid
].nm_lmsg
;
251 if (lmsg
->ms_flags
& MSGF_DONE
)
252 lwkt_sendmsg(ifnet_portfn(mycpuid
), lmsg
);
257 * Schedule ifnet.if_start on ifnet's CPU
260 if_start_schedule(struct ifnet
*ifp
)
265 if (if_start_oncpu_sched
)
268 cpu
= ifp
->if_start_cpuid(ifp
);
271 lwkt_send_ipiq(globaldata_find(cpu
), if_start_ipifunc
, ifp
);
274 if_start_ipifunc(ifp
);
279 * This function will release ifnet.if_start interlock,
280 * if ifnet.if_start does not need to be scheduled
283 if_start_need_schedule(struct ifaltq
*ifq
, int running
)
285 if (!running
|| ifq_is_empty(ifq
)
287 || ifq
->altq_tbr
!= NULL
292 * ifnet.if_start interlock is released, if:
293 * 1) Hardware can not take any packets, due to
294 * o interface is marked down
295 * o hardware queue is full (IFF_OACTIVE)
296 * Under the second situation, hardware interrupt
297 * or polling(4) will call/schedule ifnet.if_start
298 * when hardware queue is ready
299 * 2) There is not packet in the ifnet.if_snd.
300 * Further ifq_dispatch or ifq_handoff will call/
301 * schedule ifnet.if_start
302 * 3) TBR is used and it does not allow further
304 * TBR callout will call ifnet.if_start
306 if (!running
|| !ifq_data_ready(ifq
)) {
307 ifq
->altq_started
= 0;
317 if_start_dispatch(struct netmsg
*nmsg
)
319 struct lwkt_msg
*lmsg
= &nmsg
->nm_lmsg
;
320 struct ifnet
*ifp
= lmsg
->u
.ms_resultp
;
321 struct ifaltq
*ifq
= &ifp
->if_snd
;
325 lwkt_replymsg(lmsg
, 0); /* reply ASAP */
329 if (!if_start_oncpu_sched
&& mycpuid
!= ifp
->if_start_cpuid(ifp
)) {
331 * If the ifnet is still up, we need to
332 * chase its CPU change.
334 if (ifp
->if_flags
& IFF_UP
) {
335 logifstart(chase_sched
, ifp
);
336 if_start_schedule(ifp
);
344 if (ifp
->if_flags
& IFF_UP
) {
345 lwkt_serialize_enter(ifp
->if_serializer
); /* XXX try? */
346 if ((ifp
->if_flags
& IFF_OACTIVE
) == 0) {
347 logifstart(run
, ifp
);
350 (IFF_OACTIVE
| IFF_RUNNING
)) == IFF_RUNNING
)
353 lwkt_serialize_exit(ifp
->if_serializer
);
358 if (if_start_need_schedule(ifq
, running
)) {
360 if (lmsg
->ms_flags
& MSGF_DONE
) { /* XXX necessary? */
361 logifstart(sched
, ifp
);
362 lwkt_sendmsg(ifnet_portfn(mycpuid
), lmsg
);
368 /* Device driver ifnet.if_start helper function */
370 if_devstart(struct ifnet
*ifp
)
372 struct ifaltq
*ifq
= &ifp
->if_snd
;
375 ASSERT_SERIALIZED(ifp
->if_serializer
);
378 if (ifq
->altq_started
|| !ifq_data_ready(ifq
)) {
379 logifstart(avoid
, ifp
);
383 ifq
->altq_started
= 1;
386 if (if_devstart_schedonly
) {
388 * Always schedule ifnet.if_start on ifnet's CPU,
389 * short circuit the rest of this function.
391 logifstart(sched
, ifp
);
392 if_start_schedule(ifp
);
396 logifstart(run
, ifp
);
399 if ((ifp
->if_flags
& (IFF_OACTIVE
| IFF_RUNNING
)) == IFF_RUNNING
)
402 if (if_devstart_schednochk
|| if_start_need_schedule(ifq
, running
)) {
404 * More data need to be transmitted, ifnet.if_start is
405 * scheduled on ifnet's CPU, and we keep going.
406 * NOTE: ifnet.if_start interlock is not released.
408 logifstart(sched
, ifp
);
409 if_start_schedule(ifp
);
414 * Attach an interface to the list of "active" interfaces.
416 * The serializer is optional. If non-NULL access to the interface
420 if_attach(struct ifnet
*ifp
, lwkt_serialize_t serializer
)
422 unsigned socksize
, ifasize
;
423 int namelen
, masklen
;
424 struct sockaddr_dl
*sdl
;
429 static int if_indexlim
= 8;
432 * The serializer can be passed in from the device, allowing the
433 * same serializer to be used for both the interrupt interlock and
434 * the device queue. If not specified, the netif structure will
435 * use an embedded serializer.
437 if (serializer
== NULL
) {
438 serializer
= &ifp
->if_default_serializer
;
439 lwkt_serialize_init(serializer
);
441 ifp
->if_serializer
= serializer
;
443 ifp
->if_start_cpuid
= if_start_cpuid
;
446 #ifdef DEVICE_POLLING
447 /* Device is not in polling mode by default */
448 ifp
->if_poll_cpuid
= -1;
449 if (ifp
->if_poll
!= NULL
)
450 ifp
->if_start_cpuid
= if_start_cpuid_poll
;
453 ifp
->if_start_nmsg
= kmalloc(ncpus
* sizeof(struct netmsg
),
454 M_LWKTMSG
, M_WAITOK
);
455 for (i
= 0; i
< ncpus
; ++i
) {
456 netmsg_init(&ifp
->if_start_nmsg
[i
], &netisr_adone_rport
, 0,
458 ifp
->if_start_nmsg
[i
].nm_lmsg
.u
.ms_resultp
= ifp
;
461 TAILQ_INSERT_TAIL(&ifnet
, ifp
, if_link
);
462 ifp
->if_index
= ++if_index
;
466 * The old code would work if the interface passed a pre-existing
467 * chain of ifaddrs to this code. We don't trust our callers to
468 * properly initialize the tailq, however, so we no longer allow
469 * this unlikely case.
471 ifp
->if_addrheads
= kmalloc(ncpus
* sizeof(struct ifaddrhead
),
472 M_IFADDR
, M_WAITOK
| M_ZERO
);
473 for (i
= 0; i
< ncpus
; ++i
)
474 TAILQ_INIT(&ifp
->if_addrheads
[i
]);
476 TAILQ_INIT(&ifp
->if_prefixhead
);
477 LIST_INIT(&ifp
->if_multiaddrs
);
478 getmicrotime(&ifp
->if_lastchange
);
479 if (ifindex2ifnet
== NULL
|| if_index
>= if_indexlim
) {
485 /* grow ifindex2ifnet */
486 n
= if_indexlim
* sizeof(*q
);
487 q
= kmalloc(n
, M_IFADDR
, M_WAITOK
| M_ZERO
);
489 bcopy(ifindex2ifnet
, q
, n
/2);
490 kfree(ifindex2ifnet
, M_IFADDR
);
495 ifindex2ifnet
[if_index
] = ifp
;
498 * create a Link Level name for this device
500 namelen
= strlen(ifp
->if_xname
);
501 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
502 masklen
= _offsetof(struct sockaddr_dl
, sdl_data
[0]) + namelen
;
503 socksize
= masklen
+ ifp
->if_addrlen
;
504 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
505 if (socksize
< sizeof(*sdl
))
506 socksize
= sizeof(*sdl
);
507 socksize
= ROUNDUP(socksize
);
508 ifasize
= sizeof(struct ifaddr
) + 2 * socksize
;
509 ifa
= ifa_create(ifasize
, M_WAITOK
);
510 sdl
= (struct sockaddr_dl
*)(ifa
+ 1);
511 sdl
->sdl_len
= socksize
;
512 sdl
->sdl_family
= AF_LINK
;
513 bcopy(ifp
->if_xname
, sdl
->sdl_data
, namelen
);
514 sdl
->sdl_nlen
= namelen
;
515 sdl
->sdl_index
= ifp
->if_index
;
516 sdl
->sdl_type
= ifp
->if_type
;
517 ifp
->if_lladdr
= ifa
;
519 ifa
->ifa_rtrequest
= link_rtrequest
;
520 ifa
->ifa_addr
= (struct sockaddr
*)sdl
;
521 sdl
= (struct sockaddr_dl
*)(socksize
+ (caddr_t
)sdl
);
522 ifa
->ifa_netmask
= (struct sockaddr
*)sdl
;
523 sdl
->sdl_len
= masklen
;
525 sdl
->sdl_data
[--namelen
] = 0xff;
526 ifa_iflink(ifa
, ifp
, 0 /* Insert head */);
528 EVENTHANDLER_INVOKE(ifnet_attach_event
, ifp
);
529 devctl_notify("IFNET", ifp
->if_xname
, "ATTACH", NULL
);
533 ifq
->altq_disc
= NULL
;
534 ifq
->altq_flags
&= ALTQF_CANTCHANGE
;
535 ifq
->altq_tbr
= NULL
;
537 ifq
->altq_started
= 0;
538 ifq
->altq_prepended
= NULL
;
540 ifq_set_classic(ifq
);
542 if (!SLIST_EMPTY(&domains
))
543 if_attachdomain1(ifp
);
545 /* Announce the interface. */
546 rt_ifannouncemsg(ifp
, IFAN_ARRIVAL
);
550 if_attachdomain(void *dummy
)
555 TAILQ_FOREACH(ifp
, &ifnet
, if_list
)
556 if_attachdomain1(ifp
);
559 SYSINIT(domainifattach
, SI_SUB_PROTO_IFATTACHDOMAIN
, SI_ORDER_FIRST
,
560 if_attachdomain
, NULL
);
563 if_attachdomain1(struct ifnet
*ifp
)
569 /* address family dependent data region */
570 bzero(ifp
->if_afdata
, sizeof(ifp
->if_afdata
));
571 SLIST_FOREACH(dp
, &domains
, dom_next
)
572 if (dp
->dom_ifattach
)
573 ifp
->if_afdata
[dp
->dom_family
] =
574 (*dp
->dom_ifattach
)(ifp
);
579 * Purge all addresses whose type is _not_ AF_LINK
582 if_purgeaddrs_nolink(struct ifnet
*ifp
)
584 struct ifaddr_container
*ifac
, *next
;
586 TAILQ_FOREACH_MUTABLE(ifac
, &ifp
->if_addrheads
[mycpuid
],
588 struct ifaddr
*ifa
= ifac
->ifa
;
590 /* Leave link ifaddr as it is */
591 if (ifa
->ifa_addr
->sa_family
== AF_LINK
)
594 /* XXX: Ugly!! ad hoc just for INET */
595 if (ifa
->ifa_addr
&& ifa
->ifa_addr
->sa_family
== AF_INET
) {
596 struct ifaliasreq ifr
;
597 #ifdef IFADDR_DEBUG_VERBOSE
600 kprintf("purge in4 addr %p: ", ifa
);
601 for (i
= 0; i
< ncpus
; ++i
)
602 kprintf("%d ", ifa
->ifa_containers
[i
].ifa_refcnt
);
606 bzero(&ifr
, sizeof ifr
);
607 ifr
.ifra_addr
= *ifa
->ifa_addr
;
608 if (ifa
->ifa_dstaddr
)
609 ifr
.ifra_broadaddr
= *ifa
->ifa_dstaddr
;
610 if (in_control(NULL
, SIOCDIFADDR
, (caddr_t
)&ifr
, ifp
,
616 if (ifa
->ifa_addr
&& ifa
->ifa_addr
->sa_family
== AF_INET6
) {
617 #ifdef IFADDR_DEBUG_VERBOSE
620 kprintf("purge in6 addr %p: ", ifa
);
621 for (i
= 0; i
< ncpus
; ++i
)
622 kprintf("%d ", ifa
->ifa_containers
[i
].ifa_refcnt
);
627 /* ifp_addrhead is already updated */
631 ifa_ifunlink(ifa
, ifp
);
637 * Detach an interface, removing it from the
638 * list of "active" interfaces.
641 if_detach(struct ifnet
*ifp
)
643 struct radix_node_head
*rnh
;
648 EVENTHANDLER_INVOKE(ifnet_detach_event
, ifp
);
651 * Remove routes and flush queues.
654 #ifdef DEVICE_POLLING
655 if (ifp
->if_flags
& IFF_POLLING
)
656 ether_poll_deregister(ifp
);
660 if (ifq_is_enabled(&ifp
->if_snd
))
661 altq_disable(&ifp
->if_snd
);
662 if (ifq_is_attached(&ifp
->if_snd
))
663 altq_detach(&ifp
->if_snd
);
666 * Clean up all addresses.
668 ifp
->if_lladdr
= NULL
;
670 if_purgeaddrs_nolink(ifp
);
671 if (!TAILQ_EMPTY(&ifp
->if_addrheads
[mycpuid
])) {
674 ifa
= TAILQ_FIRST(&ifp
->if_addrheads
[mycpuid
])->ifa
;
675 KASSERT(ifa
->ifa_addr
->sa_family
== AF_LINK
,
676 ("non-link ifaddr is left on if_addrheads"));
678 ifa_ifunlink(ifa
, ifp
);
680 KASSERT(TAILQ_EMPTY(&ifp
->if_addrheads
[mycpuid
]),
681 ("there are still ifaddrs left on if_addrheads"));
686 * Remove all IPv4 kernel structures related to ifp.
693 * Remove all IPv6 kernel structs related to ifp. This should be done
694 * before removing routing entries below, since IPv6 interface direct
695 * routes are expected to be removed by the IPv6-specific kernel API.
696 * Otherwise, the kernel will detect some inconsistency and bark it.
702 * Delete all remaining routes using this interface
703 * Unfortuneatly the only way to do this is to slog through
704 * the entire routing table looking for routes which point
705 * to this interface...oh well...
708 for (cpu
= 0; cpu
< ncpus2
; cpu
++) {
709 lwkt_migratecpu(cpu
);
710 for (i
= 1; i
<= AF_MAX
; i
++) {
711 if ((rnh
= rt_tables
[cpu
][i
]) == NULL
)
713 rnh
->rnh_walktree(rnh
, if_rtdel
, ifp
);
716 lwkt_migratecpu(origcpu
);
718 /* Announce that the interface is gone. */
719 rt_ifannouncemsg(ifp
, IFAN_DEPARTURE
);
720 devctl_notify("IFNET", ifp
->if_xname
, "DETACH", NULL
);
722 SLIST_FOREACH(dp
, &domains
, dom_next
)
723 if (dp
->dom_ifdetach
&& ifp
->if_afdata
[dp
->dom_family
])
724 (*dp
->dom_ifdetach
)(ifp
,
725 ifp
->if_afdata
[dp
->dom_family
]);
728 * Remove interface from ifindex2ifp[] and maybe decrement if_index.
730 ifindex2ifnet
[ifp
->if_index
] = NULL
;
731 while (if_index
> 0 && ifindex2ifnet
[if_index
] == NULL
)
734 TAILQ_REMOVE(&ifnet
, ifp
, if_link
);
735 kfree(ifp
->if_addrheads
, M_IFADDR
);
736 kfree(ifp
->if_start_nmsg
, M_LWKTMSG
);
741 * Delete Routes for a Network Interface
743 * Called for each routing entry via the rnh->rnh_walktree() call above
744 * to delete all route entries referencing a detaching network interface.
747 * rn pointer to node in the routing table
748 * arg argument passed to rnh->rnh_walktree() - detaching interface
752 * errno failed - reason indicated
756 if_rtdel(struct radix_node
*rn
, void *arg
)
758 struct rtentry
*rt
= (struct rtentry
*)rn
;
759 struct ifnet
*ifp
= arg
;
762 if (rt
->rt_ifp
== ifp
) {
765 * Protect (sorta) against walktree recursion problems
768 if (!(rt
->rt_flags
& RTF_UP
))
771 err
= rtrequest(RTM_DELETE
, rt_key(rt
), rt
->rt_gateway
,
772 rt_mask(rt
), rt
->rt_flags
,
773 (struct rtentry
**) NULL
);
775 log(LOG_WARNING
, "if_rtdel: error %d\n", err
);
783 * Locate an interface based on a complete address.
786 ifa_ifwithaddr(struct sockaddr
*addr
)
790 TAILQ_FOREACH(ifp
, &ifnet
, if_link
) {
791 struct ifaddr_container
*ifac
;
793 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
794 struct ifaddr
*ifa
= ifac
->ifa
;
796 if (ifa
->ifa_addr
->sa_family
!= addr
->sa_family
)
798 if (sa_equal(addr
, ifa
->ifa_addr
))
800 if ((ifp
->if_flags
& IFF_BROADCAST
) &&
801 ifa
->ifa_broadaddr
&&
802 /* IPv6 doesn't have broadcast */
803 ifa
->ifa_broadaddr
->sa_len
!= 0 &&
804 sa_equal(ifa
->ifa_broadaddr
, addr
))
811 * Locate the point to point interface with a given destination address.
814 ifa_ifwithdstaddr(struct sockaddr
*addr
)
818 TAILQ_FOREACH(ifp
, &ifnet
, if_link
) {
819 struct ifaddr_container
*ifac
;
821 if (!(ifp
->if_flags
& IFF_POINTOPOINT
))
824 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
825 struct ifaddr
*ifa
= ifac
->ifa
;
827 if (ifa
->ifa_addr
->sa_family
!= addr
->sa_family
)
829 if (ifa
->ifa_dstaddr
&&
830 sa_equal(addr
, ifa
->ifa_dstaddr
))
838 * Find an interface on a specific network. If many, choice
839 * is most specific found.
842 ifa_ifwithnet(struct sockaddr
*addr
)
845 struct ifaddr
*ifa_maybe
= NULL
;
846 u_int af
= addr
->sa_family
;
847 char *addr_data
= addr
->sa_data
, *cplim
;
850 * AF_LINK addresses can be looked up directly by their index number,
851 * so do that if we can.
854 struct sockaddr_dl
*sdl
= (struct sockaddr_dl
*)addr
;
856 if (sdl
->sdl_index
&& sdl
->sdl_index
<= if_index
)
857 return (ifindex2ifnet
[sdl
->sdl_index
]->if_lladdr
);
861 * Scan though each interface, looking for ones that have
862 * addresses in this address family.
864 TAILQ_FOREACH(ifp
, &ifnet
, if_link
) {
865 struct ifaddr_container
*ifac
;
867 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
868 struct ifaddr
*ifa
= ifac
->ifa
;
869 char *cp
, *cp2
, *cp3
;
871 if (ifa
->ifa_addr
->sa_family
!= af
)
873 if (af
== AF_INET
&& ifp
->if_flags
& IFF_POINTOPOINT
) {
875 * This is a bit broken as it doesn't
876 * take into account that the remote end may
877 * be a single node in the network we are
879 * The trouble is that we don't know the
880 * netmask for the remote end.
882 if (ifa
->ifa_dstaddr
!= NULL
&&
883 sa_equal(addr
, ifa
->ifa_dstaddr
))
887 * if we have a special address handler,
888 * then use it instead of the generic one.
890 if (ifa
->ifa_claim_addr
) {
891 if ((*ifa
->ifa_claim_addr
)(ifa
, addr
)) {
899 * Scan all the bits in the ifa's address.
900 * If a bit dissagrees with what we are
901 * looking for, mask it with the netmask
902 * to see if it really matters.
905 if (ifa
->ifa_netmask
== 0)
908 cp2
= ifa
->ifa_addr
->sa_data
;
909 cp3
= ifa
->ifa_netmask
->sa_data
;
910 cplim
= ifa
->ifa_netmask
->sa_len
+
911 (char *)ifa
->ifa_netmask
;
913 if ((*cp
++ ^ *cp2
++) & *cp3
++)
914 goto next
; /* next address! */
916 * If the netmask of what we just found
917 * is more specific than what we had before
918 * (if we had one) then remember the new one
919 * before continuing to search
920 * for an even better one.
922 if (ifa_maybe
== 0 ||
923 rn_refines((char *)ifa
->ifa_netmask
,
924 (char *)ifa_maybe
->ifa_netmask
))
933 * Find an interface address specific to an interface best matching
937 ifaof_ifpforaddr(struct sockaddr
*addr
, struct ifnet
*ifp
)
939 struct ifaddr_container
*ifac
;
940 char *cp
, *cp2
, *cp3
;
942 struct ifaddr
*ifa_maybe
= 0;
943 u_int af
= addr
->sa_family
;
947 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
948 struct ifaddr
*ifa
= ifac
->ifa
;
950 if (ifa
->ifa_addr
->sa_family
!= af
)
954 if (ifa
->ifa_netmask
== NULL
) {
955 if (sa_equal(addr
, ifa
->ifa_addr
) ||
956 (ifa
->ifa_dstaddr
!= NULL
&&
957 sa_equal(addr
, ifa
->ifa_dstaddr
)))
961 if (ifp
->if_flags
& IFF_POINTOPOINT
) {
962 if (sa_equal(addr
, ifa
->ifa_dstaddr
))
966 cp2
= ifa
->ifa_addr
->sa_data
;
967 cp3
= ifa
->ifa_netmask
->sa_data
;
968 cplim
= ifa
->ifa_netmask
->sa_len
+ (char *)ifa
->ifa_netmask
;
969 for (; cp3
< cplim
; cp3
++)
970 if ((*cp
++ ^ *cp2
++) & *cp3
)
980 * Default action when installing a route with a Link Level gateway.
981 * Lookup an appropriate real ifa to point to.
982 * This should be moved to /sys/net/link.c eventually.
985 link_rtrequest(int cmd
, struct rtentry
*rt
, struct rt_addrinfo
*info
)
988 struct sockaddr
*dst
;
991 if (cmd
!= RTM_ADD
|| (ifa
= rt
->rt_ifa
) == NULL
||
992 (ifp
= ifa
->ifa_ifp
) == NULL
|| (dst
= rt_key(rt
)) == NULL
)
994 ifa
= ifaof_ifpforaddr(dst
, ifp
);
999 if (ifa
->ifa_rtrequest
&& ifa
->ifa_rtrequest
!= link_rtrequest
)
1000 ifa
->ifa_rtrequest(cmd
, rt
, info
);
1005 * Mark an interface down and notify protocols of
1007 * NOTE: must be called at splnet or eqivalent.
1010 if_unroute(struct ifnet
*ifp
, int flag
, int fam
)
1012 struct ifaddr_container
*ifac
;
1014 ifp
->if_flags
&= ~flag
;
1015 getmicrotime(&ifp
->if_lastchange
);
1016 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
1017 struct ifaddr
*ifa
= ifac
->ifa
;
1019 if (fam
== PF_UNSPEC
|| (fam
== ifa
->ifa_addr
->sa_family
))
1020 kpfctlinput(PRC_IFDOWN
, ifa
->ifa_addr
);
1022 ifq_purge(&ifp
->if_snd
);
1027 * Mark an interface up and notify protocols of
1029 * NOTE: must be called at splnet or eqivalent.
1032 if_route(struct ifnet
*ifp
, int flag
, int fam
)
1034 struct ifaddr_container
*ifac
;
1036 ifq_purge(&ifp
->if_snd
);
1037 ifp
->if_flags
|= flag
;
1038 getmicrotime(&ifp
->if_lastchange
);
1039 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
1040 struct ifaddr
*ifa
= ifac
->ifa
;
1042 if (fam
== PF_UNSPEC
|| (fam
== ifa
->ifa_addr
->sa_family
))
1043 kpfctlinput(PRC_IFUP
, ifa
->ifa_addr
);
1052 * Mark an interface down and notify protocols of the transition. An
1053 * interface going down is also considered to be a synchronizing event.
1054 * We must ensure that all packet processing related to the interface
1055 * has completed before we return so e.g. the caller can free the ifnet
1056 * structure that the mbufs may be referencing.
1058 * NOTE: must be called at splnet or eqivalent.
1061 if_down(struct ifnet
*ifp
)
1063 if_unroute(ifp
, IFF_UP
, AF_UNSPEC
);
1064 netmsg_service_sync();
1068 * Mark an interface up and notify protocols of
1070 * NOTE: must be called at splnet or eqivalent.
1073 if_up(struct ifnet
*ifp
)
1075 if_route(ifp
, IFF_UP
, AF_UNSPEC
);
1079 * Process a link state change.
1080 * NOTE: must be called at splsoftnet or equivalent.
1083 if_link_state_change(struct ifnet
*ifp
)
1085 int link_state
= ifp
->if_link_state
;
1088 devctl_notify("IFNET", ifp
->if_xname
,
1089 (link_state
== LINK_STATE_UP
) ? "LINK_UP" : "LINK_DOWN", NULL
);
1093 * Handle interface watchdog timer routines. Called
1094 * from softclock, we decrement timers (if set) and
1095 * call the appropriate interface routine on expiration.
1098 if_slowtimo(void *arg
)
1104 TAILQ_FOREACH(ifp
, &ifnet
, if_link
) {
1105 if (ifp
->if_timer
== 0 || --ifp
->if_timer
)
1107 if (ifp
->if_watchdog
) {
1108 if (lwkt_serialize_try(ifp
->if_serializer
)) {
1109 (*ifp
->if_watchdog
)(ifp
);
1110 lwkt_serialize_exit(ifp
->if_serializer
);
1112 /* try again next timeout */
1120 callout_reset(&if_slowtimo_timer
, hz
/ IFNET_SLOWHZ
, if_slowtimo
, NULL
);
1124 * Map interface name to
1125 * interface structure pointer.
1128 ifunit(const char *name
)
1133 * Search all the interfaces for this name/number
1136 TAILQ_FOREACH(ifp
, &ifnet
, if_link
) {
1137 if (strncmp(ifp
->if_xname
, name
, IFNAMSIZ
) == 0)
1145 * Map interface name in a sockaddr_dl to
1146 * interface structure pointer.
1149 if_withname(struct sockaddr
*sa
)
1151 char ifname
[IFNAMSIZ
+1];
1152 struct sockaddr_dl
*sdl
= (struct sockaddr_dl
*)sa
;
1154 if ( (sa
->sa_family
!= AF_LINK
) || (sdl
->sdl_nlen
== 0) ||
1155 (sdl
->sdl_nlen
> IFNAMSIZ
) )
1159 * ifunit wants a null-terminated name. It may not be null-terminated
1160 * in the sockaddr. We don't want to change the caller's sockaddr,
1161 * and there might not be room to put the trailing null anyway, so we
1162 * make a local copy that we know we can null terminate safely.
1165 bcopy(sdl
->sdl_data
, ifname
, sdl
->sdl_nlen
);
1166 ifname
[sdl
->sdl_nlen
] = '\0';
1167 return ifunit(ifname
);
1175 ifioctl(struct socket
*so
, u_long cmd
, caddr_t data
, struct ucred
*cred
)
1183 size_t namelen
, onamelen
;
1184 char new_name
[IFNAMSIZ
];
1186 struct sockaddr_dl
*sdl
;
1192 return (ifconf(cmd
, data
, cred
));
1194 ifr
= (struct ifreq
*)data
;
1199 if ((error
= priv_check_cred(cred
, PRIV_ROOT
, 0)) != 0)
1201 return ((cmd
== SIOCIFCREATE
) ?
1202 if_clone_create(ifr
->ifr_name
, sizeof(ifr
->ifr_name
)) :
1203 if_clone_destroy(ifr
->ifr_name
));
1205 case SIOCIFGCLONERS
:
1206 return (if_clone_list((struct if_clonereq
*)data
));
1209 ifp
= ifunit(ifr
->ifr_name
);
1215 ifr
->ifr_flags
= ifp
->if_flags
;
1216 ifr
->ifr_flagshigh
= ifp
->if_flags
>> 16;
1220 ifr
->ifr_reqcap
= ifp
->if_capabilities
;
1221 ifr
->ifr_curcap
= ifp
->if_capenable
;
1225 ifr
->ifr_metric
= ifp
->if_metric
;
1229 ifr
->ifr_mtu
= ifp
->if_mtu
;
1233 ifr
->ifr_phys
= ifp
->if_physical
;
1236 case SIOCGIFPOLLCPU
:
1237 #ifdef DEVICE_POLLING
1238 ifr
->ifr_pollcpu
= ifp
->if_poll_cpuid
;
1240 ifr
->ifr_pollcpu
= -1;
1244 case SIOCSIFPOLLCPU
:
1245 #ifdef DEVICE_POLLING
1246 if ((ifp
->if_flags
& IFF_POLLING
) == 0)
1247 ether_pollcpu_register(ifp
, ifr
->ifr_pollcpu
);
1252 error
= priv_check_cred(cred
, PRIV_ROOT
, 0);
1255 new_flags
= (ifr
->ifr_flags
& 0xffff) |
1256 (ifr
->ifr_flagshigh
<< 16);
1257 if (ifp
->if_flags
& IFF_SMART
) {
1258 /* Smart drivers twiddle their own routes */
1259 } else if (ifp
->if_flags
& IFF_UP
&&
1260 (new_flags
& IFF_UP
) == 0) {
1264 } else if (new_flags
& IFF_UP
&&
1265 (ifp
->if_flags
& IFF_UP
) == 0) {
1271 #ifdef DEVICE_POLLING
1272 if ((new_flags
^ ifp
->if_flags
) & IFF_POLLING
) {
1273 if (new_flags
& IFF_POLLING
) {
1274 ether_poll_register(ifp
);
1276 ether_poll_deregister(ifp
);
1281 ifp
->if_flags
= (ifp
->if_flags
& IFF_CANTCHANGE
) |
1282 (new_flags
&~ IFF_CANTCHANGE
);
1283 if (new_flags
& IFF_PPROMISC
) {
1284 /* Permanently promiscuous mode requested */
1285 ifp
->if_flags
|= IFF_PROMISC
;
1286 } else if (ifp
->if_pcount
== 0) {
1287 ifp
->if_flags
&= ~IFF_PROMISC
;
1289 if (ifp
->if_ioctl
) {
1290 lwkt_serialize_enter(ifp
->if_serializer
);
1291 ifp
->if_ioctl(ifp
, cmd
, data
, cred
);
1292 lwkt_serialize_exit(ifp
->if_serializer
);
1294 getmicrotime(&ifp
->if_lastchange
);
1298 error
= priv_check_cred(cred
, PRIV_ROOT
, 0);
1301 if (ifr
->ifr_reqcap
& ~ifp
->if_capabilities
)
1303 lwkt_serialize_enter(ifp
->if_serializer
);
1304 ifp
->if_ioctl(ifp
, cmd
, data
, cred
);
1305 lwkt_serialize_exit(ifp
->if_serializer
);
1309 error
= priv_check_cred(cred
, PRIV_ROOT
, 0);
1312 error
= copyinstr(ifr
->ifr_data
, new_name
, IFNAMSIZ
, NULL
);
1315 if (new_name
[0] == '\0')
1317 if (ifunit(new_name
) != NULL
)
1320 EVENTHANDLER_INVOKE(ifnet_detach_event
, ifp
);
1322 /* Announce the departure of the interface. */
1323 rt_ifannouncemsg(ifp
, IFAN_DEPARTURE
);
1325 strlcpy(ifp
->if_xname
, new_name
, sizeof(ifp
->if_xname
));
1326 ifa
= TAILQ_FIRST(&ifp
->if_addrheads
[mycpuid
])->ifa
;
1327 /* XXX IFA_LOCK(ifa); */
1328 sdl
= (struct sockaddr_dl
*)ifa
->ifa_addr
;
1329 namelen
= strlen(new_name
);
1330 onamelen
= sdl
->sdl_nlen
;
1332 * Move the address if needed. This is safe because we
1333 * allocate space for a name of length IFNAMSIZ when we
1334 * create this in if_attach().
1336 if (namelen
!= onamelen
) {
1337 bcopy(sdl
->sdl_data
+ onamelen
,
1338 sdl
->sdl_data
+ namelen
, sdl
->sdl_alen
);
1340 bcopy(new_name
, sdl
->sdl_data
, namelen
);
1341 sdl
->sdl_nlen
= namelen
;
1342 sdl
= (struct sockaddr_dl
*)ifa
->ifa_netmask
;
1343 bzero(sdl
->sdl_data
, onamelen
);
1344 while (namelen
!= 0)
1345 sdl
->sdl_data
[--namelen
] = 0xff;
1346 /* XXX IFA_UNLOCK(ifa) */
1348 EVENTHANDLER_INVOKE(ifnet_attach_event
, ifp
);
1350 /* Announce the return of the interface. */
1351 rt_ifannouncemsg(ifp
, IFAN_ARRIVAL
);
1355 error
= priv_check_cred(cred
, PRIV_ROOT
, 0);
1358 ifp
->if_metric
= ifr
->ifr_metric
;
1359 getmicrotime(&ifp
->if_lastchange
);
1363 error
= priv_check_cred(cred
, PRIV_ROOT
, 0);
1368 lwkt_serialize_enter(ifp
->if_serializer
);
1369 error
= ifp
->if_ioctl(ifp
, cmd
, data
, cred
);
1370 lwkt_serialize_exit(ifp
->if_serializer
);
1372 getmicrotime(&ifp
->if_lastchange
);
1377 u_long oldmtu
= ifp
->if_mtu
;
1379 error
= priv_check_cred(cred
, PRIV_ROOT
, 0);
1382 if (ifp
->if_ioctl
== NULL
)
1383 return (EOPNOTSUPP
);
1384 if (ifr
->ifr_mtu
< IF_MINMTU
|| ifr
->ifr_mtu
> IF_MAXMTU
)
1386 lwkt_serialize_enter(ifp
->if_serializer
);
1387 error
= ifp
->if_ioctl(ifp
, cmd
, data
, cred
);
1388 lwkt_serialize_exit(ifp
->if_serializer
);
1390 getmicrotime(&ifp
->if_lastchange
);
1394 * If the link MTU changed, do network layer specific procedure.
1396 if (ifp
->if_mtu
!= oldmtu
) {
1406 error
= priv_check_cred(cred
, PRIV_ROOT
, 0);
1410 /* Don't allow group membership on non-multicast interfaces. */
1411 if ((ifp
->if_flags
& IFF_MULTICAST
) == 0)
1414 /* Don't let users screw up protocols' entries. */
1415 if (ifr
->ifr_addr
.sa_family
!= AF_LINK
)
1418 if (cmd
== SIOCADDMULTI
) {
1419 struct ifmultiaddr
*ifma
;
1420 error
= if_addmulti(ifp
, &ifr
->ifr_addr
, &ifma
);
1422 error
= if_delmulti(ifp
, &ifr
->ifr_addr
);
1425 getmicrotime(&ifp
->if_lastchange
);
1428 case SIOCSIFPHYADDR
:
1429 case SIOCDIFPHYADDR
:
1431 case SIOCSIFPHYADDR_IN6
:
1433 case SIOCSLIFPHYADDR
:
1435 case SIOCSIFGENERIC
:
1436 error
= priv_check_cred(cred
, PRIV_ROOT
, 0);
1439 if (ifp
->if_ioctl
== 0)
1440 return (EOPNOTSUPP
);
1441 lwkt_serialize_enter(ifp
->if_serializer
);
1442 error
= ifp
->if_ioctl(ifp
, cmd
, data
, cred
);
1443 lwkt_serialize_exit(ifp
->if_serializer
);
1445 getmicrotime(&ifp
->if_lastchange
);
1449 ifs
= (struct ifstat
*)data
;
1450 ifs
->ascii
[0] = '\0';
1452 case SIOCGIFPSRCADDR
:
1453 case SIOCGIFPDSTADDR
:
1454 case SIOCGLIFPHYADDR
:
1456 case SIOCGIFGENERIC
:
1457 if (ifp
->if_ioctl
== NULL
)
1458 return (EOPNOTSUPP
);
1459 lwkt_serialize_enter(ifp
->if_serializer
);
1460 error
= ifp
->if_ioctl(ifp
, cmd
, data
, cred
);
1461 lwkt_serialize_exit(ifp
->if_serializer
);
1465 error
= priv_check_cred(cred
, PRIV_ROOT
, 0);
1468 return if_setlladdr(ifp
,
1469 ifr
->ifr_addr
.sa_data
, ifr
->ifr_addr
.sa_len
);
1472 oif_flags
= ifp
->if_flags
;
1473 if (so
->so_proto
== 0)
1474 return (EOPNOTSUPP
);
1476 error
= so_pru_control(so
, cmd
, data
, ifp
);
1483 case SIOCSIFDSTADDR
:
1485 case SIOCSIFBRDADDR
:
1486 case SIOCSIFNETMASK
:
1487 #if BYTE_ORDER != BIG_ENDIAN
1488 if (ifr
->ifr_addr
.sa_family
== 0 &&
1489 ifr
->ifr_addr
.sa_len
< 16) {
1490 ifr
->ifr_addr
.sa_family
= ifr
->ifr_addr
.sa_len
;
1491 ifr
->ifr_addr
.sa_len
= 16;
1494 if (ifr
->ifr_addr
.sa_len
== 0)
1495 ifr
->ifr_addr
.sa_len
= 16;
1503 case OSIOCGIFDSTADDR
:
1504 cmd
= SIOCGIFDSTADDR
;
1507 case OSIOCGIFBRDADDR
:
1508 cmd
= SIOCGIFBRDADDR
;
1511 case OSIOCGIFNETMASK
:
1512 cmd
= SIOCGIFNETMASK
;
1514 error
= so_pru_control(so
, cmd
, data
, ifp
);
1518 case OSIOCGIFDSTADDR
:
1519 case OSIOCGIFBRDADDR
:
1520 case OSIOCGIFNETMASK
:
1521 *(u_short
*)&ifr
->ifr_addr
= ifr
->ifr_addr
.sa_family
;
1525 #endif /* COMPAT_43 */
1527 if ((oif_flags
^ ifp
->if_flags
) & IFF_UP
) {
1529 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1530 if (ifp
->if_flags
& IFF_UP
) {
1544 * Set/clear promiscuous mode on interface ifp based on the truth value
1545 * of pswitch. The calls are reference counted so that only the first
1546 * "on" request actually has an effect, as does the final "off" request.
1547 * Results are undefined if the "off" and "on" requests are not matched.
1550 ifpromisc(struct ifnet
*ifp
, int pswitch
)
1556 oldflags
= ifp
->if_flags
;
1557 if (ifp
->if_flags
& IFF_PPROMISC
) {
1558 /* Do nothing if device is in permanently promiscuous mode */
1559 ifp
->if_pcount
+= pswitch
? 1 : -1;
1564 * If the device is not configured up, we cannot put it in
1567 if ((ifp
->if_flags
& IFF_UP
) == 0)
1569 if (ifp
->if_pcount
++ != 0)
1571 ifp
->if_flags
|= IFF_PROMISC
;
1572 log(LOG_INFO
, "%s: promiscuous mode enabled\n",
1575 if (--ifp
->if_pcount
> 0)
1577 ifp
->if_flags
&= ~IFF_PROMISC
;
1578 log(LOG_INFO
, "%s: promiscuous mode disabled\n",
1581 ifr
.ifr_flags
= ifp
->if_flags
;
1582 ifr
.ifr_flagshigh
= ifp
->if_flags
>> 16;
1583 lwkt_serialize_enter(ifp
->if_serializer
);
1584 error
= ifp
->if_ioctl(ifp
, SIOCSIFFLAGS
, (caddr_t
)&ifr
,
1585 (struct ucred
*)NULL
);
1586 lwkt_serialize_exit(ifp
->if_serializer
);
1590 ifp
->if_flags
= oldflags
;
1595 * Return interface configuration
1596 * of system. List may be used
1597 * in later ioctl's (above) to get
1598 * other information.
1601 ifconf(u_long cmd
, caddr_t data
, struct ucred
*cred
)
1603 struct ifconf
*ifc
= (struct ifconf
*)data
;
1605 struct sockaddr
*sa
;
1606 struct ifreq ifr
, *ifrp
;
1607 int space
= ifc
->ifc_len
, error
= 0;
1609 ifrp
= ifc
->ifc_req
;
1610 TAILQ_FOREACH(ifp
, &ifnet
, if_link
) {
1611 struct ifaddr_container
*ifac
;
1614 if (space
<= sizeof ifr
)
1618 * Zero the stack declared structure first to prevent
1619 * memory disclosure.
1621 bzero(&ifr
, sizeof(ifr
));
1622 if (strlcpy(ifr
.ifr_name
, ifp
->if_xname
, sizeof(ifr
.ifr_name
))
1623 >= sizeof(ifr
.ifr_name
)) {
1624 error
= ENAMETOOLONG
;
1629 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
1630 struct ifaddr
*ifa
= ifac
->ifa
;
1632 if (space
<= sizeof ifr
)
1635 if (cred
->cr_prison
&&
1636 prison_if(cred
, sa
))
1640 if (cmd
== OSIOCGIFCONF
) {
1641 struct osockaddr
*osa
=
1642 (struct osockaddr
*)&ifr
.ifr_addr
;
1644 osa
->sa_family
= sa
->sa_family
;
1645 error
= copyout(&ifr
, ifrp
, sizeof ifr
);
1649 if (sa
->sa_len
<= sizeof(*sa
)) {
1651 error
= copyout(&ifr
, ifrp
, sizeof ifr
);
1654 if (space
< (sizeof ifr
) + sa
->sa_len
-
1657 space
-= sa
->sa_len
- sizeof(*sa
);
1658 error
= copyout(&ifr
, ifrp
,
1659 sizeof ifr
.ifr_name
);
1661 error
= copyout(sa
, &ifrp
->ifr_addr
,
1663 ifrp
= (struct ifreq
*)
1664 (sa
->sa_len
+ (caddr_t
)&ifrp
->ifr_addr
);
1668 space
-= sizeof ifr
;
1673 bzero(&ifr
.ifr_addr
, sizeof ifr
.ifr_addr
);
1674 error
= copyout(&ifr
, ifrp
, sizeof ifr
);
1677 space
-= sizeof ifr
;
1681 ifc
->ifc_len
-= space
;
1686 * Just like if_promisc(), but for all-multicast-reception mode.
1689 if_allmulti(struct ifnet
*ifp
, int onswitch
)
1697 if (ifp
->if_amcount
++ == 0) {
1698 ifp
->if_flags
|= IFF_ALLMULTI
;
1699 ifr
.ifr_flags
= ifp
->if_flags
;
1700 ifr
.ifr_flagshigh
= ifp
->if_flags
>> 16;
1701 lwkt_serialize_enter(ifp
->if_serializer
);
1702 error
= ifp
->if_ioctl(ifp
, SIOCSIFFLAGS
, (caddr_t
)&ifr
,
1703 (struct ucred
*)NULL
);
1704 lwkt_serialize_exit(ifp
->if_serializer
);
1707 if (ifp
->if_amcount
> 1) {
1710 ifp
->if_amcount
= 0;
1711 ifp
->if_flags
&= ~IFF_ALLMULTI
;
1712 ifr
.ifr_flags
= ifp
->if_flags
;
1713 ifr
.ifr_flagshigh
= ifp
->if_flags
>> 16;
1714 lwkt_serialize_enter(ifp
->if_serializer
);
1715 error
= ifp
->if_ioctl(ifp
, SIOCSIFFLAGS
, (caddr_t
)&ifr
,
1716 (struct ucred
*)NULL
);
1717 lwkt_serialize_exit(ifp
->if_serializer
);
1729 * Add a multicast listenership to the interface in question.
1730 * The link layer provides a routine which converts
1734 struct ifnet
*ifp
, /* interface to manipulate */
1735 struct sockaddr
*sa
, /* address to add */
1736 struct ifmultiaddr
**retifma
)
1738 struct sockaddr
*llsa
, *dupsa
;
1740 struct ifmultiaddr
*ifma
;
1743 * If the matching multicast address already exists
1744 * then don't add a new one, just add a reference
1746 LIST_FOREACH(ifma
, &ifp
->if_multiaddrs
, ifma_link
) {
1747 if (sa_equal(sa
, ifma
->ifma_addr
)) {
1748 ifma
->ifma_refcount
++;
1756 * Give the link layer a chance to accept/reject it, and also
1757 * find out which AF_LINK address this maps to, if it isn't one
1760 if (ifp
->if_resolvemulti
) {
1761 lwkt_serialize_enter(ifp
->if_serializer
);
1762 error
= ifp
->if_resolvemulti(ifp
, &llsa
, sa
);
1763 lwkt_serialize_exit(ifp
->if_serializer
);
1770 MALLOC(ifma
, struct ifmultiaddr
*, sizeof *ifma
, M_IFMADDR
, M_WAITOK
);
1771 MALLOC(dupsa
, struct sockaddr
*, sa
->sa_len
, M_IFMADDR
, M_WAITOK
);
1772 bcopy(sa
, dupsa
, sa
->sa_len
);
1774 ifma
->ifma_addr
= dupsa
;
1775 ifma
->ifma_lladdr
= llsa
;
1776 ifma
->ifma_ifp
= ifp
;
1777 ifma
->ifma_refcount
= 1;
1778 ifma
->ifma_protospec
= 0;
1779 rt_newmaddrmsg(RTM_NEWMADDR
, ifma
);
1782 * Some network interfaces can scan the address list at
1783 * interrupt time; lock them out.
1786 LIST_INSERT_HEAD(&ifp
->if_multiaddrs
, ifma
, ifma_link
);
1791 LIST_FOREACH(ifma
, &ifp
->if_multiaddrs
, ifma_link
) {
1792 if (sa_equal(ifma
->ifma_addr
, llsa
))
1796 ifma
->ifma_refcount
++;
1798 MALLOC(ifma
, struct ifmultiaddr
*, sizeof *ifma
,
1799 M_IFMADDR
, M_WAITOK
);
1800 MALLOC(dupsa
, struct sockaddr
*, llsa
->sa_len
,
1801 M_IFMADDR
, M_WAITOK
);
1802 bcopy(llsa
, dupsa
, llsa
->sa_len
);
1803 ifma
->ifma_addr
= dupsa
;
1804 ifma
->ifma_ifp
= ifp
;
1805 ifma
->ifma_refcount
= 1;
1807 LIST_INSERT_HEAD(&ifp
->if_multiaddrs
, ifma
, ifma_link
);
1812 * We are certain we have added something, so call down to the
1813 * interface to let them know about it.
1816 lwkt_serialize_enter(ifp
->if_serializer
);
1817 ifp
->if_ioctl(ifp
, SIOCADDMULTI
, 0, (struct ucred
*)NULL
);
1818 lwkt_serialize_exit(ifp
->if_serializer
);
1825 * Remove a reference to a multicast address on this interface. Yell
1826 * if the request does not match an existing membership.
1829 if_delmulti(struct ifnet
*ifp
, struct sockaddr
*sa
)
1831 struct ifmultiaddr
*ifma
;
1833 LIST_FOREACH(ifma
, &ifp
->if_multiaddrs
, ifma_link
)
1834 if (sa_equal(sa
, ifma
->ifma_addr
))
1839 if (ifma
->ifma_refcount
> 1) {
1840 ifma
->ifma_refcount
--;
1844 rt_newmaddrmsg(RTM_DELMADDR
, ifma
);
1845 sa
= ifma
->ifma_lladdr
;
1847 LIST_REMOVE(ifma
, ifma_link
);
1849 * Make sure the interface driver is notified
1850 * in the case of a link layer mcast group being left.
1852 if (ifma
->ifma_addr
->sa_family
== AF_LINK
&& sa
== 0) {
1853 lwkt_serialize_enter(ifp
->if_serializer
);
1854 ifp
->if_ioctl(ifp
, SIOCDELMULTI
, 0, (struct ucred
*)NULL
);
1855 lwkt_serialize_exit(ifp
->if_serializer
);
1858 kfree(ifma
->ifma_addr
, M_IFMADDR
);
1859 kfree(ifma
, M_IFMADDR
);
1864 * Now look for the link-layer address which corresponds to
1865 * this network address. It had been squirreled away in
1866 * ifma->ifma_lladdr for this purpose (so we don't have
1867 * to call ifp->if_resolvemulti() again), and we saved that
1868 * value in sa above. If some nasty deleted the
1869 * link-layer address out from underneath us, we can deal because
1870 * the address we stored was is not the same as the one which was
1871 * in the record for the link-layer address. (So we don't complain
1874 LIST_FOREACH(ifma
, &ifp
->if_multiaddrs
, ifma_link
)
1875 if (sa_equal(sa
, ifma
->ifma_addr
))
1880 if (ifma
->ifma_refcount
> 1) {
1881 ifma
->ifma_refcount
--;
1886 lwkt_serialize_enter(ifp
->if_serializer
);
1887 LIST_REMOVE(ifma
, ifma_link
);
1888 ifp
->if_ioctl(ifp
, SIOCDELMULTI
, 0, (struct ucred
*)NULL
);
1889 lwkt_serialize_exit(ifp
->if_serializer
);
1891 kfree(ifma
->ifma_addr
, M_IFMADDR
);
1892 kfree(sa
, M_IFMADDR
);
1893 kfree(ifma
, M_IFMADDR
);
1899 * Set the link layer address on an interface.
1901 * At this time we only support certain types of interfaces,
1902 * and we don't allow the length of the address to change.
1905 if_setlladdr(struct ifnet
*ifp
, const u_char
*lladdr
, int len
)
1907 struct sockaddr_dl
*sdl
;
1910 sdl
= IF_LLSOCKADDR(ifp
);
1913 if (len
!= sdl
->sdl_alen
) /* don't allow length to change */
1915 switch (ifp
->if_type
) {
1916 case IFT_ETHER
: /* these types use struct arpcom */
1919 bcopy(lladdr
, ((struct arpcom
*)ifp
->if_softc
)->ac_enaddr
, len
);
1920 bcopy(lladdr
, LLADDR(sdl
), len
);
1926 * If the interface is already up, we need
1927 * to re-init it in order to reprogram its
1930 lwkt_serialize_enter(ifp
->if_serializer
);
1931 if ((ifp
->if_flags
& IFF_UP
) != 0) {
1932 struct ifaddr_container
*ifac
;
1934 ifp
->if_flags
&= ~IFF_UP
;
1935 ifr
.ifr_flags
= ifp
->if_flags
;
1936 ifr
.ifr_flagshigh
= ifp
->if_flags
>> 16;
1937 ifp
->if_ioctl(ifp
, SIOCSIFFLAGS
, (caddr_t
)&ifr
,
1938 (struct ucred
*)NULL
);
1939 ifp
->if_flags
|= IFF_UP
;
1940 ifr
.ifr_flags
= ifp
->if_flags
;
1941 ifr
.ifr_flagshigh
= ifp
->if_flags
>> 16;
1942 ifp
->if_ioctl(ifp
, SIOCSIFFLAGS
, (caddr_t
)&ifr
,
1943 (struct ucred
*)NULL
);
1946 * Also send gratuitous ARPs to notify other nodes about
1947 * the address change.
1949 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
1950 struct ifaddr
*ifa
= ifac
->ifa
;
1952 if (ifa
->ifa_addr
!= NULL
&&
1953 ifa
->ifa_addr
->sa_family
== AF_INET
)
1954 arp_ifinit(ifp
, ifa
);
1958 lwkt_serialize_exit(ifp
->if_serializer
);
1962 struct ifmultiaddr
*
1963 ifmaof_ifpforaddr(struct sockaddr
*sa
, struct ifnet
*ifp
)
1965 struct ifmultiaddr
*ifma
;
1967 LIST_FOREACH(ifma
, &ifp
->if_multiaddrs
, ifma_link
)
1968 if (sa_equal(ifma
->ifma_addr
, sa
))
1975 * This function locates the first real ethernet MAC from a network
1976 * card and loads it into node, returning 0 on success or ENOENT if
1977 * no suitable interfaces were found. It is used by the uuid code to
1978 * generate a unique 6-byte number.
1981 if_getanyethermac(uint16_t *node
, int minlen
)
1984 struct sockaddr_dl
*sdl
;
1986 TAILQ_FOREACH(ifp
, &ifnet
, if_link
) {
1987 if (ifp
->if_type
!= IFT_ETHER
)
1989 sdl
= IF_LLSOCKADDR(ifp
);
1990 if (sdl
->sdl_alen
< minlen
)
1992 bcopy(((struct arpcom
*)ifp
->if_softc
)->ac_enaddr
, node
,
2000 * The name argument must be a pointer to storage which will last as
2001 * long as the interface does. For physical devices, the result of
2002 * device_get_name(dev) is a good choice and for pseudo-devices a
2003 * static string works well.
2006 if_initname(struct ifnet
*ifp
, const char *name
, int unit
)
2008 ifp
->if_dname
= name
;
2009 ifp
->if_dunit
= unit
;
2010 if (unit
!= IF_DUNIT_NONE
)
2011 ksnprintf(ifp
->if_xname
, IFNAMSIZ
, "%s%d", name
, unit
);
2013 strlcpy(ifp
->if_xname
, name
, IFNAMSIZ
);
2017 if_printf(struct ifnet
*ifp
, const char *fmt
, ...)
2022 retval
= kprintf("%s: ", ifp
->if_xname
);
2023 __va_start(ap
, fmt
);
2024 retval
+= kvprintf(fmt
, ap
);
2030 ifq_set_classic(struct ifaltq
*ifq
)
2032 ifq
->altq_enqueue
= ifq_classic_enqueue
;
2033 ifq
->altq_dequeue
= ifq_classic_dequeue
;
2034 ifq
->altq_request
= ifq_classic_request
;
2038 ifq_classic_enqueue(struct ifaltq
*ifq
, struct mbuf
*m
,
2039 struct altq_pktattr
*pa __unused
)
2041 logifq(enqueue
, ifq
);
2042 if (IF_QFULL(ifq
)) {
2052 ifq_classic_dequeue(struct ifaltq
*ifq
, struct mbuf
*mpolled
, int op
)
2061 logifq(dequeue
, ifq
);
2065 panic("unsupported ALTQ dequeue op: %d", op
);
2067 KKASSERT(mpolled
== NULL
|| mpolled
== m
);
2072 ifq_classic_request(struct ifaltq
*ifq
, int req
, void *arg
)
2079 panic("unsupported ALTQ request: %d", req
);
2085 ifq_dispatch(struct ifnet
*ifp
, struct mbuf
*m
, struct altq_pktattr
*pa
)
2087 struct ifaltq
*ifq
= &ifp
->if_snd
;
2088 int running
= 0, error
, start
= 0;
2090 ASSERT_NOT_SERIALIZED(ifp
->if_serializer
);
2093 error
= ifq_enqueue_locked(ifq
, m
, pa
);
2098 if (!ifq
->altq_started
) {
2100 * Hold the interlock of ifnet.if_start
2102 ifq
->altq_started
= 1;
2107 ifp
->if_obytes
+= m
->m_pkthdr
.len
;
2108 if (m
->m_flags
& M_MCAST
)
2112 logifstart(avoid
, ifp
);
2116 if (ifq_dispatch_schedonly
) {
2118 * Always schedule ifnet.if_start on ifnet's CPU,
2119 * short circuit the rest of this function.
2121 logifstart(sched
, ifp
);
2122 if_start_schedule(ifp
);
2127 * Try to do direct ifnet.if_start first, if there is
2128 * contention on ifnet's serializer, ifnet.if_start will
2129 * be scheduled on ifnet's CPU.
2131 if (!lwkt_serialize_try(ifp
->if_serializer
)) {
2133 * ifnet serializer contention happened,
2134 * ifnet.if_start is scheduled on ifnet's
2135 * CPU, and we keep going.
2137 logifstart(contend_sched
, ifp
);
2138 if_start_schedule(ifp
);
2142 if ((ifp
->if_flags
& IFF_OACTIVE
) == 0) {
2143 logifstart(run
, ifp
);
2145 if ((ifp
->if_flags
&
2146 (IFF_OACTIVE
| IFF_RUNNING
)) == IFF_RUNNING
)
2150 lwkt_serialize_exit(ifp
->if_serializer
);
2152 if (ifq_dispatch_schednochk
|| if_start_need_schedule(ifq
, running
)) {
2154 * More data need to be transmitted, ifnet.if_start is
2155 * scheduled on ifnet's CPU, and we keep going.
2156 * NOTE: ifnet.if_start interlock is not released.
2158 logifstart(sched
, ifp
);
2159 if_start_schedule(ifp
);
2165 ifa_create(int size
, int flags
)
2170 KASSERT(size
>= sizeof(*ifa
), ("ifaddr size too small\n"));
2172 ifa
= kmalloc(size
, M_IFADDR
, flags
| M_ZERO
);
2176 ifa
->ifa_containers
= kmalloc(ncpus
* sizeof(struct ifaddr_container
),
2177 M_IFADDR
, M_WAITOK
| M_ZERO
);
2178 ifa
->ifa_ncnt
= ncpus
;
2179 for (i
= 0; i
< ncpus
; ++i
) {
2180 struct ifaddr_container
*ifac
= &ifa
->ifa_containers
[i
];
2182 ifac
->ifa_magic
= IFA_CONTAINER_MAGIC
;
2184 ifac
->ifa_refcnt
= 1;
2187 kprintf("alloc ifa %p %d\n", ifa
, size
);
2193 ifac_free(struct ifaddr_container
*ifac
, int cpu_id
)
2195 struct ifaddr
*ifa
= ifac
->ifa
;
2197 KKASSERT(ifac
->ifa_magic
== IFA_CONTAINER_MAGIC
);
2198 KKASSERT(ifac
->ifa_refcnt
== 0);
2199 KASSERT(ifac
->ifa_listmask
== 0,
2200 ("ifa is still on %#x lists\n", ifac
->ifa_listmask
));
2202 ifac
->ifa_magic
= IFA_CONTAINER_DEAD
;
2204 #ifdef IFADDR_DEBUG_VERBOSE
2205 kprintf("try free ifa %p cpu_id %d\n", ifac
->ifa
, cpu_id
);
2208 KASSERT(ifa
->ifa_ncnt
> 0 && ifa
->ifa_ncnt
<= ncpus
,
2209 ("invalid # of ifac, %d\n", ifa
->ifa_ncnt
));
2210 if (atomic_fetchadd_int(&ifa
->ifa_ncnt
, -1) == 1) {
2212 kprintf("free ifa %p\n", ifa
);
2214 kfree(ifa
->ifa_containers
, M_IFADDR
);
2215 kfree(ifa
, M_IFADDR
);
2220 ifa_iflink_dispatch(struct netmsg
*nmsg
)
2222 struct netmsg_ifaddr
*msg
= (struct netmsg_ifaddr
*)nmsg
;
2223 struct ifaddr
*ifa
= msg
->ifa
;
2224 struct ifnet
*ifp
= msg
->ifp
;
2226 struct ifaddr_container
*ifac
;
2230 ifac
= &ifa
->ifa_containers
[cpu
];
2231 ASSERT_IFAC_VALID(ifac
);
2232 KASSERT((ifac
->ifa_listmask
& IFA_LIST_IFADDRHEAD
) == 0,
2233 ("ifaddr is on if_addrheads\n"));
2235 ifac
->ifa_listmask
|= IFA_LIST_IFADDRHEAD
;
2237 TAILQ_INSERT_TAIL(&ifp
->if_addrheads
[cpu
], ifac
, ifa_link
);
2239 TAILQ_INSERT_HEAD(&ifp
->if_addrheads
[cpu
], ifac
, ifa_link
);
2243 ifa_forwardmsg(&nmsg
->nm_lmsg
, cpu
+ 1);
2247 ifa_iflink(struct ifaddr
*ifa
, struct ifnet
*ifp
, int tail
)
2249 struct netmsg_ifaddr msg
;
2251 netmsg_init(&msg
.netmsg
, &curthread
->td_msgport
, 0,
2252 ifa_iflink_dispatch
);
2257 ifa_domsg(&msg
.netmsg
.nm_lmsg
, 0);
2261 ifa_ifunlink_dispatch(struct netmsg
*nmsg
)
2263 struct netmsg_ifaddr
*msg
= (struct netmsg_ifaddr
*)nmsg
;
2264 struct ifaddr
*ifa
= msg
->ifa
;
2265 struct ifnet
*ifp
= msg
->ifp
;
2267 struct ifaddr_container
*ifac
;
2271 ifac
= &ifa
->ifa_containers
[cpu
];
2272 ASSERT_IFAC_VALID(ifac
);
2273 KASSERT(ifac
->ifa_listmask
& IFA_LIST_IFADDRHEAD
,
2274 ("ifaddr is not on if_addrhead\n"));
2276 TAILQ_REMOVE(&ifp
->if_addrheads
[cpu
], ifac
, ifa_link
);
2277 ifac
->ifa_listmask
&= ~IFA_LIST_IFADDRHEAD
;
2281 ifa_forwardmsg(&nmsg
->nm_lmsg
, cpu
+ 1);
2285 ifa_ifunlink(struct ifaddr
*ifa
, struct ifnet
*ifp
)
2287 struct netmsg_ifaddr msg
;
2289 netmsg_init(&msg
.netmsg
, &curthread
->td_msgport
, 0,
2290 ifa_ifunlink_dispatch
);
2294 ifa_domsg(&msg
.netmsg
.nm_lmsg
, 0);
2298 ifa_destroy_dispatch(struct netmsg
*nmsg
)
2300 struct netmsg_ifaddr
*msg
= (struct netmsg_ifaddr
*)nmsg
;
2303 ifa_forwardmsg(&nmsg
->nm_lmsg
, mycpuid
+ 1);
2307 ifa_destroy(struct ifaddr
*ifa
)
2309 struct netmsg_ifaddr msg
;
2311 netmsg_init(&msg
.netmsg
, &curthread
->td_msgport
, 0,
2312 ifa_destroy_dispatch
);
2315 ifa_domsg(&msg
.netmsg
.nm_lmsg
, 0);
2319 ifnet_portfn(int cpu
)
2321 return &ifnet_threads
[cpu
].td_msgport
;
2325 ifnet_forwardmsg(struct lwkt_msg
*lmsg
, int next_cpu
)
2327 KKASSERT(next_cpu
> mycpuid
&& next_cpu
<= ncpus
);
2329 if (next_cpu
< ncpus
)
2330 lwkt_forwardmsg(ifnet_portfn(next_cpu
), lmsg
);
2332 lwkt_replymsg(lmsg
, 0);
2336 ifnet_domsg(struct lwkt_msg
*lmsg
, int cpu
)
2338 KKASSERT(cpu
< ncpus
);
2339 return lwkt_domsg(ifnet_portfn(cpu
), lmsg
, 0);
2343 ifnet_sendmsg(struct lwkt_msg
*lmsg
, int cpu
)
2345 KKASSERT(cpu
< ncpus
);
2346 lwkt_sendmsg(ifnet_portfn(cpu
), lmsg
);
2350 ifnetinit(void *dummy __unused
)
2354 for (i
= 0; i
< ncpus
; ++i
) {
2355 struct thread
*thr
= &ifnet_threads
[i
];
2357 lwkt_create(netmsg_service_loop
, &ifnet_mpsafe_thread
, NULL
,
2358 thr
, TDF_NETWORK
| TDF_MPSAFE
, i
, "ifnet %d", i
);
2359 netmsg_service_port_init(&thr
->td_msgport
);