FreeBSD/Linux Kernel Cross Reference
sys/net/if.c
1 /*-
2 * Copyright (c) 1980, 1986, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)if.c 8.5 (Berkeley) 1/9/95
30 * $FreeBSD$
31 */
32
33 #include "opt_compat.h"
34 #include "opt_bpf.h"
35 #include "opt_inet6.h"
36 #include "opt_inet.h"
37
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/conf.h>
41 #include <sys/malloc.h>
42 #include <sys/sbuf.h>
43 #include <sys/bus.h>
44 #include <sys/mbuf.h>
45 #include <sys/systm.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/protosw.h>
51 #include <sys/kernel.h>
52 #include <sys/lock.h>
53 #include <sys/refcount.h>
54 #include <sys/module.h>
55 #include <sys/rwlock.h>
56 #include <sys/sockio.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 #include <sys/sysent.h>
60 #include <sys/taskqueue.h>
61 #include <sys/domain.h>
62 #include <sys/jail.h>
63 #include <sys/priv.h>
64
65 #include <machine/stdarg.h>
66 #include <vm/uma.h>
67
68 #include <net/bpf.h>
69 #include <net/ethernet.h>
70 #include <net/if.h>
71 #include <net/if_arp.h>
72 #include <net/if_clone.h>
73 #include <net/if_dl.h>
74 #include <net/if_types.h>
75 #include <net/if_var.h>
76 #include <net/if_media.h>
77 #include <net/if_vlan_var.h>
78 #include <net/radix.h>
79 #include <net/route.h>
80 #include <net/vnet.h>
81
82 #if defined(INET) || defined(INET6)
83 #include <net/ethernet.h>
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/ip.h>
87 #include <netinet/ip_carp.h>
88 #ifdef INET
89 #include <netinet/if_ether.h>
90 #endif /* INET */
91 #ifdef INET6
92 #include <netinet6/in6_var.h>
93 #include <netinet6/in6_ifattach.h>
94 #endif /* INET6 */
95 #endif /* INET || INET6 */
96
97 #include <security/mac/mac_framework.h>
98
99 #ifdef COMPAT_FREEBSD32
100 #include <sys/mount.h>
101 #include <compat/freebsd32/freebsd32.h>
102
103 struct ifreq_buffer32 {
104 uint32_t length; /* (size_t) */
105 uint32_t buffer; /* (void *) */
106 };
107
108 /*
109 * Interface request structure used for socket
110 * ioctl's. All interface ioctl's must have parameter
111 * definitions which begin with ifr_name. The
112 * remainder may be interface specific.
113 */
114 struct ifreq32 {
115 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */
116 union {
117 struct sockaddr ifru_addr;
118 struct sockaddr ifru_dstaddr;
119 struct sockaddr ifru_broadaddr;
120 struct ifreq_buffer32 ifru_buffer;
121 short ifru_flags[2];
122 short ifru_index;
123 int ifru_jid;
124 int ifru_metric;
125 int ifru_mtu;
126 int ifru_phys;
127 int ifru_media;
128 uint32_t ifru_data;
129 int ifru_cap[2];
130 u_int ifru_fib;
131 u_char ifru_vlan_pcp;
132 } ifr_ifru;
133 };
134 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
135 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
136 __offsetof(struct ifreq32, ifr_ifru));
137
138 struct ifgroupreq32 {
139 char ifgr_name[IFNAMSIZ];
140 u_int ifgr_len;
141 union {
142 char ifgru_group[IFNAMSIZ];
143 uint32_t ifgru_groups;
144 } ifgr_ifgru;
145 };
146
147 struct ifmediareq32 {
148 char ifm_name[IFNAMSIZ];
149 int ifm_current;
150 int ifm_mask;
151 int ifm_status;
152 int ifm_active;
153 int ifm_count;
154 uint32_t ifm_ulist; /* (int *) */
155 };
156 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
157 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
158
159 #define _CASE_IOC_IFGROUPREQ_32(cmd) \
160 case _IOC_NEWTYPE((cmd), struct ifgroupreq32):
161 #else /* !COMPAT_FREEBSD32 */
162 #define _CASE_IOC_IFGROUPREQ_32(cmd)
163 #endif /* !COMPAT_FREEBSD32 */
164
165 #define CASE_IOC_IFGROUPREQ(cmd) \
166 _CASE_IOC_IFGROUPREQ_32(cmd) \
167 case (cmd)
168
169 union ifreq_union {
170 struct ifreq ifr;
171 #ifdef COMPAT_FREEBSD32
172 struct ifreq32 ifr32;
173 #endif
174 };
175
176 union ifgroupreq_union {
177 struct ifgroupreq ifgr;
178 #ifdef COMPAT_FREEBSD32
179 struct ifgroupreq32 ifgr32;
180 #endif
181 };
182
183 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
184 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
185
186 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
187 &ifqmaxlen, 0, "max send queue size");
188
189 /* Log link state change events */
190 static int log_link_state_change = 1;
191
192 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
193 &log_link_state_change, 0,
194 "log interface link state change events");
195
196 /* Log promiscuous mode change events */
197 static int log_promisc_mode_change = 1;
198
199 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
200 &log_promisc_mode_change, 1,
201 "log promiscuous mode change events");
202
203 /* Interface description */
204 static unsigned int ifdescr_maxlen = 1024;
205 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
206 &ifdescr_maxlen, 0,
207 "administrative maximum length for interface description");
208
209 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
210
211 /* global sx for non-critical path ifdescr */
212 static struct sx ifdescr_sx;
213 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
214
215 void (*bridge_linkstate_p)(struct ifnet *ifp);
216 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
217 void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
218 /* These are external hooks for CARP. */
219 void (*carp_linkstate_p)(struct ifnet *ifp);
220 void (*carp_demote_adj_p)(int, char *);
221 int (*carp_master_p)(struct ifaddr *);
222 #if defined(INET) || defined(INET6)
223 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
224 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
225 const struct sockaddr *sa);
226 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
227 int (*carp_attach_p)(struct ifaddr *, int);
228 void (*carp_detach_p)(struct ifaddr *, bool);
229 #endif
230 #ifdef INET
231 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **);
232 #endif
233 #ifdef INET6
234 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
235 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
236 const struct in6_addr *taddr);
237 #endif
238
239 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
240
241 /*
242 * XXX: Style; these should be sorted alphabetically, and unprototyped
243 * static functions should be prototyped. Currently they are sorted by
244 * declaration order.
245 */
246 static void if_attachdomain(void *);
247 static void if_attachdomain1(struct ifnet *);
248 static int ifconf(u_long, caddr_t);
249 static void if_freemulti(struct ifmultiaddr *);
250 static void if_grow(void);
251 static void if_input_default(struct ifnet *, struct mbuf *);
252 static int if_requestencap_default(struct ifnet *, struct if_encap_req *);
253 static void if_route(struct ifnet *, int flag, int fam);
254 static int if_setflag(struct ifnet *, int, int, int *, int);
255 static int if_transmit(struct ifnet *ifp, struct mbuf *m);
256 static void if_unroute(struct ifnet *, int flag, int fam);
257 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
258 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
259 static void do_link_state_change(void *, int);
260 static int if_getgroup(struct ifgroupreq *, struct ifnet *);
261 static int if_getgroupmembers(struct ifgroupreq *);
262 static void if_delgroups(struct ifnet *);
263 static void if_attach_internal(struct ifnet *, int, struct if_clone *);
264 static int if_detach_internal(struct ifnet *, int, struct if_clone **);
265 #ifdef VIMAGE
266 static void if_vmove(struct ifnet *, struct vnet *);
267 #endif
268
269 #ifdef INET6
270 /*
271 * XXX: declare here to avoid to include many inet6 related files..
272 * should be more generalized?
273 */
274 extern void nd6_setmtu(struct ifnet *);
275 #endif
276
277 /* ipsec helper hooks */
278 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
279 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
280
281 VNET_DEFINE(int, if_index);
282 int ifqmaxlen = IFQ_MAXLEN;
283 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */
284 VNET_DEFINE(struct ifgrouphead, ifg_head);
285
286 static VNET_DEFINE(int, if_indexlim) = 8;
287
288 /* Table of ifnet by index. */
289 VNET_DEFINE(struct ifnet **, ifindex_table);
290
291 #define V_if_indexlim VNET(if_indexlim)
292 #define V_ifindex_table VNET(ifindex_table)
293
294 /*
295 * The global network interface list (V_ifnet) and related state (such as
296 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
297 * an rwlock. Either may be acquired shared to stablize the list, but both
298 * must be acquired writable to modify the list. This model allows us to
299 * both stablize the interface list during interrupt thread processing, but
300 * also to stablize it over long-running ioctls, without introducing priority
301 * inversions and deadlocks.
302 */
303 struct rwlock ifnet_rwlock;
304 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
305 struct sx ifnet_sxlock;
306 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
307
308 /*
309 * The allocation of network interfaces is a rather non-atomic affair; we
310 * need to select an index before we are ready to expose the interface for
311 * use, so will use this pointer value to indicate reservation.
312 */
313 #define IFNET_HOLD (void *)(uintptr_t)(-1)
314
315 static if_com_alloc_t *if_com_alloc[256];
316 static if_com_free_t *if_com_free[256];
317
318 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
319 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
320 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
321
322 struct ifnet *
323 ifnet_byindex_locked(u_short idx)
324 {
325
326 if (idx > V_if_index)
327 return (NULL);
328 if (V_ifindex_table[idx] == IFNET_HOLD)
329 return (NULL);
330 return (V_ifindex_table[idx]);
331 }
332
333 struct ifnet *
334 ifnet_byindex(u_short idx)
335 {
336 struct ifnet *ifp;
337
338 IFNET_RLOCK_NOSLEEP();
339 ifp = ifnet_byindex_locked(idx);
340 IFNET_RUNLOCK_NOSLEEP();
341 return (ifp);
342 }
343
344 struct ifnet *
345 ifnet_byindex_ref(u_short idx)
346 {
347 struct ifnet *ifp;
348
349 IFNET_RLOCK_NOSLEEP();
350 ifp = ifnet_byindex_locked(idx);
351 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
352 IFNET_RUNLOCK_NOSLEEP();
353 return (NULL);
354 }
355 if_ref(ifp);
356 IFNET_RUNLOCK_NOSLEEP();
357 return (ifp);
358 }
359
360 /*
361 * Allocate an ifindex array entry; return 0 on success or an error on
362 * failure.
363 */
364 static u_short
365 ifindex_alloc(void)
366 {
367 u_short idx;
368
369 IFNET_WLOCK_ASSERT();
370 retry:
371 /*
372 * Try to find an empty slot below V_if_index. If we fail, take the
373 * next slot.
374 */
375 for (idx = 1; idx <= V_if_index; idx++) {
376 if (V_ifindex_table[idx] == NULL)
377 break;
378 }
379
380 /* Catch if_index overflow. */
381 if (idx >= V_if_indexlim) {
382 if_grow();
383 goto retry;
384 }
385 if (idx > V_if_index)
386 V_if_index = idx;
387 return (idx);
388 }
389
390 static void
391 ifindex_free_locked(u_short idx)
392 {
393
394 IFNET_WLOCK_ASSERT();
395
396 V_ifindex_table[idx] = NULL;
397 while (V_if_index > 0 &&
398 V_ifindex_table[V_if_index] == NULL)
399 V_if_index--;
400 }
401
402 static void
403 ifindex_free(u_short idx)
404 {
405
406 IFNET_WLOCK();
407 ifindex_free_locked(idx);
408 IFNET_WUNLOCK();
409 }
410
411 static void
412 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
413 {
414
415 IFNET_WLOCK_ASSERT();
416
417 V_ifindex_table[idx] = ifp;
418 }
419
420 static void
421 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
422 {
423
424 IFNET_WLOCK();
425 ifnet_setbyindex_locked(idx, ifp);
426 IFNET_WUNLOCK();
427 }
428
429 struct ifaddr *
430 ifaddr_byindex(u_short idx)
431 {
432 struct ifnet *ifp;
433 struct ifaddr *ifa = NULL;
434
435 IFNET_RLOCK_NOSLEEP();
436 ifp = ifnet_byindex_locked(idx);
437 if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
438 ifa_ref(ifa);
439 IFNET_RUNLOCK_NOSLEEP();
440 return (ifa);
441 }
442
443 /*
444 * Network interface utility routines.
445 *
446 * Routines with ifa_ifwith* names take sockaddr *'s as
447 * parameters.
448 */
449
450 static void
451 vnet_if_init(const void *unused __unused)
452 {
453
454 TAILQ_INIT(&V_ifnet);
455 TAILQ_INIT(&V_ifg_head);
456 IFNET_WLOCK();
457 if_grow(); /* create initial table */
458 IFNET_WUNLOCK();
459 vnet_if_clone_init();
460 }
461 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
462 NULL);
463
464 #ifdef VIMAGE
465 static void
466 vnet_if_uninit(const void *unused __unused)
467 {
468
469 VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
470 "not empty", __func__, __LINE__, &V_ifnet));
471 VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
472 "not empty", __func__, __LINE__, &V_ifg_head));
473
474 free((caddr_t)V_ifindex_table, M_IFNET);
475 }
476 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
477 vnet_if_uninit, NULL);
478
479 static void
480 vnet_if_return(const void *unused __unused)
481 {
482 struct ifnet *ifp, *nifp;
483
484 /* Return all inherited interfaces to their parent vnets. */
485 TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
486 if (ifp->if_home_vnet != ifp->if_vnet)
487 if_vmove(ifp, ifp->if_home_vnet);
488 }
489 }
490 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
491 vnet_if_return, NULL);
492 #endif
493
494 static void
495 if_grow(void)
496 {
497 int oldlim;
498 u_int n;
499 struct ifnet **e;
500
501 IFNET_WLOCK_ASSERT();
502 oldlim = V_if_indexlim;
503 IFNET_WUNLOCK();
504 n = (oldlim << 1) * sizeof(*e);
505 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
506 IFNET_WLOCK();
507 if (V_if_indexlim != oldlim) {
508 free(e, M_IFNET);
509 return;
510 }
511 if (V_ifindex_table != NULL) {
512 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
513 free((caddr_t)V_ifindex_table, M_IFNET);
514 }
515 V_if_indexlim <<= 1;
516 V_ifindex_table = e;
517 }
518
519 /*
520 * Allocate a struct ifnet and an index for an interface. A layer 2
521 * common structure will also be allocated if an allocation routine is
522 * registered for the passed type.
523 */
524 struct ifnet *
525 if_alloc(u_char type)
526 {
527 struct ifnet *ifp;
528 u_short idx;
529
530 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
531 IFNET_WLOCK();
532 idx = ifindex_alloc();
533 ifnet_setbyindex_locked(idx, IFNET_HOLD);
534 IFNET_WUNLOCK();
535 ifp->if_index = idx;
536 ifp->if_type = type;
537 ifp->if_alloctype = type;
538 #ifdef VIMAGE
539 ifp->if_vnet = curvnet;
540 #endif
541 if (if_com_alloc[type] != NULL) {
542 ifp->if_l2com = if_com_alloc[type](type, ifp);
543 if (ifp->if_l2com == NULL) {
544 free(ifp, M_IFNET);
545 ifindex_free(idx);
546 return (NULL);
547 }
548 }
549
550 IF_ADDR_LOCK_INIT(ifp);
551 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
552 ifp->if_afdata_initialized = 0;
553 IF_AFDATA_LOCK_INIT(ifp);
554 TAILQ_INIT(&ifp->if_addrhead);
555 TAILQ_INIT(&ifp->if_multiaddrs);
556 TAILQ_INIT(&ifp->if_groups);
557 #ifdef MAC
558 mac_ifnet_init(ifp);
559 #endif
560 ifq_init(&ifp->if_snd, ifp);
561
562 refcount_init(&ifp->if_refcount, 1); /* Index reference. */
563 for (int i = 0; i < IFCOUNTERS; i++)
564 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
565 ifp->if_get_counter = if_get_counter_default;
566 ifp->if_pcp = IFNET_PCP_NONE;
567 ifnet_setbyindex(ifp->if_index, ifp);
568 return (ifp);
569 }
570
571 /*
572 * Do the actual work of freeing a struct ifnet, and layer 2 common
573 * structure. This call is made when the last reference to an
574 * interface is released.
575 */
576 static void
577 if_free_internal(struct ifnet *ifp)
578 {
579
580 KASSERT((ifp->if_flags & IFF_DYING),
581 ("if_free_internal: interface not dying"));
582
583 if (if_com_free[ifp->if_alloctype] != NULL)
584 if_com_free[ifp->if_alloctype](ifp->if_l2com,
585 ifp->if_alloctype);
586
587 #ifdef MAC
588 mac_ifnet_destroy(ifp);
589 #endif /* MAC */
590 if (ifp->if_description != NULL)
591 free(ifp->if_description, M_IFDESCR);
592 IF_AFDATA_DESTROY(ifp);
593 IF_ADDR_LOCK_DESTROY(ifp);
594 ifq_delete(&ifp->if_snd);
595
596 for (int i = 0; i < IFCOUNTERS; i++)
597 counter_u64_free(ifp->if_counters[i]);
598
599 free(ifp, M_IFNET);
600 }
601
602 /*
603 * Deregister an interface and free the associated storage.
604 */
605 void
606 if_free(struct ifnet *ifp)
607 {
608
609 ifp->if_flags |= IFF_DYING; /* XXX: Locking */
610
611 CURVNET_SET_QUIET(ifp->if_vnet);
612 IFNET_WLOCK();
613 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
614 ("%s: freeing unallocated ifnet", ifp->if_xname));
615
616 ifindex_free_locked(ifp->if_index);
617 IFNET_WUNLOCK();
618
619 if (refcount_release(&ifp->if_refcount))
620 if_free_internal(ifp);
621 CURVNET_RESTORE();
622 }
623
624 /*
625 * Interfaces to keep an ifnet type-stable despite the possibility of the
626 * driver calling if_free(). If there are additional references, we defer
627 * freeing the underlying data structure.
628 */
629 void
630 if_ref(struct ifnet *ifp)
631 {
632
633 /* We don't assert the ifnet list lock here, but arguably should. */
634 refcount_acquire(&ifp->if_refcount);
635 }
636
637 void
638 if_rele(struct ifnet *ifp)
639 {
640
641 if (!refcount_release(&ifp->if_refcount))
642 return;
643 if_free_internal(ifp);
644 }
645
646 void
647 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
648 {
649
650 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
651
652 if (ifq->ifq_maxlen == 0)
653 ifq->ifq_maxlen = ifqmaxlen;
654
655 ifq->altq_type = 0;
656 ifq->altq_disc = NULL;
657 ifq->altq_flags &= ALTQF_CANTCHANGE;
658 ifq->altq_tbr = NULL;
659 ifq->altq_ifp = ifp;
660 }
661
662 void
663 ifq_delete(struct ifaltq *ifq)
664 {
665 mtx_destroy(&ifq->ifq_mtx);
666 }
667
668 /*
669 * Perform generic interface initialization tasks and attach the interface
670 * to the list of "active" interfaces. If vmove flag is set on entry
671 * to if_attach_internal(), perform only a limited subset of initialization
672 * tasks, given that we are moving from one vnet to another an ifnet which
673 * has already been fully initialized.
674 *
675 * Note that if_detach_internal() removes group membership unconditionally
676 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
677 * Thus, when if_vmove() is applied to a cloned interface, group membership
678 * is lost while a cloned one always joins a group whose name is
679 * ifc->ifc_name. To recover this after if_detach_internal() and
680 * if_attach_internal(), the cloner should be specified to
681 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal()
682 * attempts to join a group whose name is ifc->ifc_name.
683 *
684 * XXX:
685 * - The decision to return void and thus require this function to
686 * succeed is questionable.
687 * - We should probably do more sanity checking. For instance we don't
688 * do anything to insure if_xname is unique or non-empty.
689 */
690 void
691 if_attach(struct ifnet *ifp)
692 {
693
694 if_attach_internal(ifp, 0, NULL);
695 }
696
697 /*
698 * Compute the least common TSO limit.
699 */
700 void
701 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
702 {
703 /*
704 * 1) If there is no limit currently, take the limit from
705 * the network adapter.
706 *
707 * 2) If the network adapter has a limit below the current
708 * limit, apply it.
709 */
710 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
711 ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
712 pmax->tsomaxbytes = ifp->if_hw_tsomax;
713 }
714 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
715 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
716 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
717 }
718 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
719 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
720 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
721 }
722 }
723
724 /*
725 * Update TSO limit of a network adapter.
726 *
727 * Returns zero if no change. Else non-zero.
728 */
729 int
730 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
731 {
732 int retval = 0;
733 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
734 ifp->if_hw_tsomax = pmax->tsomaxbytes;
735 retval++;
736 }
737 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
738 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
739 retval++;
740 }
741 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
742 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
743 retval++;
744 }
745 return (retval);
746 }
747
748 static void
749 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
750 {
751 unsigned socksize, ifasize;
752 int namelen, masklen;
753 struct sockaddr_dl *sdl;
754 struct ifaddr *ifa;
755
756 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
757 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
758 ifp->if_xname);
759
760 #ifdef VIMAGE
761 ifp->if_vnet = curvnet;
762 if (ifp->if_home_vnet == NULL)
763 ifp->if_home_vnet = curvnet;
764 #endif
765
766 if_addgroup(ifp, IFG_ALL);
767
768 /* Restore group membership for cloned interfaces. */
769 if (vmove && ifc != NULL)
770 if_clone_addgroup(ifp, ifc);
771
772 getmicrotime(&ifp->if_lastchange);
773 ifp->if_epoch = time_uptime;
774
775 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
776 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
777 ("transmit and qflush must both either be set or both be NULL"));
778 if (ifp->if_transmit == NULL) {
779 ifp->if_transmit = if_transmit;
780 ifp->if_qflush = if_qflush;
781 }
782 if (ifp->if_input == NULL)
783 ifp->if_input = if_input_default;
784
785 if (ifp->if_requestencap == NULL)
786 ifp->if_requestencap = if_requestencap_default;
787
788 if (!vmove) {
789 #ifdef MAC
790 mac_ifnet_create(ifp);
791 #endif
792
793 /*
794 * Create a Link Level name for this device.
795 */
796 namelen = strlen(ifp->if_xname);
797 /*
798 * Always save enough space for any possiable name so we
799 * can do a rename in place later.
800 */
801 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
802 socksize = masklen + ifp->if_addrlen;
803 if (socksize < sizeof(*sdl))
804 socksize = sizeof(*sdl);
805 socksize = roundup2(socksize, sizeof(long));
806 ifasize = sizeof(*ifa) + 2 * socksize;
807 ifa = ifa_alloc(ifasize, M_WAITOK);
808 sdl = (struct sockaddr_dl *)(ifa + 1);
809 sdl->sdl_len = socksize;
810 sdl->sdl_family = AF_LINK;
811 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
812 sdl->sdl_nlen = namelen;
813 sdl->sdl_index = ifp->if_index;
814 sdl->sdl_type = ifp->if_type;
815 ifp->if_addr = ifa;
816 ifa->ifa_ifp = ifp;
817 ifa->ifa_rtrequest = link_rtrequest;
818 ifa->ifa_addr = (struct sockaddr *)sdl;
819 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
820 ifa->ifa_netmask = (struct sockaddr *)sdl;
821 sdl->sdl_len = masklen;
822 while (namelen != 0)
823 sdl->sdl_data[--namelen] = 0xff;
824 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
825 /* Reliably crash if used uninitialized. */
826 ifp->if_broadcastaddr = NULL;
827
828 if (ifp->if_type == IFT_ETHER) {
829 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
830 M_WAITOK | M_ZERO);
831 }
832
833 #if defined(INET) || defined(INET6)
834 /* Use defaults for TSO, if nothing is set */
835 if (ifp->if_hw_tsomax == 0 &&
836 ifp->if_hw_tsomaxsegcount == 0 &&
837 ifp->if_hw_tsomaxsegsize == 0) {
838 /*
839 * The TSO defaults needs to be such that an
840 * NFS mbuf list of 35 mbufs totalling just
841 * below 64K works and that a chain of mbufs
842 * can be defragged into at most 32 segments:
843 */
844 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
845 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
846 ifp->if_hw_tsomaxsegcount = 35;
847 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */
848
849 /* XXX some drivers set IFCAP_TSO after ethernet attach */
850 if (ifp->if_capabilities & IFCAP_TSO) {
851 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
852 ifp->if_hw_tsomax,
853 ifp->if_hw_tsomaxsegcount,
854 ifp->if_hw_tsomaxsegsize);
855 }
856 }
857 #endif
858 }
859 #ifdef VIMAGE
860 else {
861 /*
862 * Update the interface index in the link layer address
863 * of the interface.
864 */
865 for (ifa = ifp->if_addr; ifa != NULL;
866 ifa = TAILQ_NEXT(ifa, ifa_link)) {
867 if (ifa->ifa_addr->sa_family == AF_LINK) {
868 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
869 sdl->sdl_index = ifp->if_index;
870 }
871 }
872 }
873 #endif
874
875 IFNET_WLOCK();
876 TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
877 #ifdef VIMAGE
878 curvnet->vnet_ifcnt++;
879 #endif
880 IFNET_WUNLOCK();
881
882 if (domain_init_status >= 2)
883 if_attachdomain1(ifp);
884
885 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
886 if (IS_DEFAULT_VNET(curvnet))
887 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
888
889 /* Announce the interface. */
890 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
891 }
892
893 static void
894 if_attachdomain(void *dummy)
895 {
896 struct ifnet *ifp;
897
898 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
899 if_attachdomain1(ifp);
900 }
901 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
902 if_attachdomain, NULL);
903
904 static void
905 if_attachdomain1(struct ifnet *ifp)
906 {
907 struct domain *dp;
908
909 /*
910 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
911 * cannot lock ifp->if_afdata initialization, entirely.
912 */
913 IF_AFDATA_LOCK(ifp);
914 if (ifp->if_afdata_initialized >= domain_init_status) {
915 IF_AFDATA_UNLOCK(ifp);
916 log(LOG_WARNING, "%s called more than once on %s\n",
917 __func__, ifp->if_xname);
918 return;
919 }
920 ifp->if_afdata_initialized = domain_init_status;
921 IF_AFDATA_UNLOCK(ifp);
922
923 /* address family dependent data region */
924 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
925 for (dp = domains; dp; dp = dp->dom_next) {
926 if (dp->dom_ifattach)
927 ifp->if_afdata[dp->dom_family] =
928 (*dp->dom_ifattach)(ifp);
929 }
930 }
931
932 /*
933 * Remove any unicast or broadcast network addresses from an interface.
934 */
935 void
936 if_purgeaddrs(struct ifnet *ifp)
937 {
938 struct ifaddr *ifa, *next;
939
940 /* XXX cannot hold IF_ADDR_WLOCK over called functions. */
941 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
942 if (ifa->ifa_addr->sa_family == AF_LINK)
943 continue;
944 #ifdef INET
945 /* XXX: Ugly!! ad hoc just for INET */
946 if (ifa->ifa_addr->sa_family == AF_INET) {
947 struct ifaliasreq ifr;
948
949 bzero(&ifr, sizeof(ifr));
950 ifr.ifra_addr = *ifa->ifa_addr;
951 if (ifa->ifa_dstaddr)
952 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
953 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
954 NULL) == 0)
955 continue;
956 }
957 #endif /* INET */
958 #ifdef INET6
959 if (ifa->ifa_addr->sa_family == AF_INET6) {
960 in6_purgeaddr(ifa);
961 /* ifp_addrhead is already updated */
962 continue;
963 }
964 #endif /* INET6 */
965 IF_ADDR_WLOCK(ifp);
966 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
967 IF_ADDR_WUNLOCK(ifp);
968 ifa_free(ifa);
969 }
970 }
971
972 /*
973 * Remove any multicast network addresses from an interface when an ifnet
974 * is going away.
975 */
976 static void
977 if_purgemaddrs(struct ifnet *ifp)
978 {
979 struct ifmultiaddr *ifma;
980 struct ifmultiaddr *next;
981
982 IF_ADDR_WLOCK(ifp);
983 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
984 if_delmulti_locked(ifp, ifma, 1);
985 IF_ADDR_WUNLOCK(ifp);
986 }
987
988 /*
989 * Detach an interface, removing it from the list of "active" interfaces.
990 * If vmove flag is set on entry to if_detach_internal(), perform only a
991 * limited subset of cleanup tasks, given that we are moving an ifnet from
992 * one vnet to another, where it must be fully operational.
993 *
994 * XXXRW: There are some significant questions about event ordering, and
995 * how to prevent things from starting to use the interface during detach.
996 */
997 void
998 if_detach(struct ifnet *ifp)
999 {
1000
1001 CURVNET_SET_QUIET(ifp->if_vnet);
1002 if_detach_internal(ifp, 0, NULL);
1003 CURVNET_RESTORE();
1004 }
1005
1006 /*
1007 * The vmove flag, if set, indicates that we are called from a callpath
1008 * that is moving an interface to a different vnet instance.
1009 *
1010 * The shutdown flag, if set, indicates that we are called in the
1011 * process of shutting down a vnet instance. Currently only the
1012 * vnet_if_return SYSUNINIT function sets it. Note: we can be called
1013 * on a vnet instance shutdown without this flag being set, e.g., when
1014 * the cloned interfaces are destoyed as first thing of teardown.
1015 */
1016 static int
1017 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1018 {
1019 struct ifaddr *ifa;
1020 int i;
1021 struct domain *dp;
1022 struct ifnet *iter;
1023 int found = 0;
1024 #ifdef VIMAGE
1025 int shutdown;
1026
1027 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1028 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1029 #endif
1030 IFNET_WLOCK();
1031 TAILQ_FOREACH(iter, &V_ifnet, if_link)
1032 if (iter == ifp) {
1033 TAILQ_REMOVE(&V_ifnet, ifp, if_link);
1034 if (!vmove)
1035 ifp->if_flags |= IFF_DYING;
1036 found = 1;
1037 break;
1038 }
1039 IFNET_WUNLOCK();
1040 if (!found) {
1041 /*
1042 * While we would want to panic here, we cannot
1043 * guarantee that the interface is indeed still on
1044 * the list given we don't hold locks all the way.
1045 */
1046 return (ENOENT);
1047 #if 0
1048 if (vmove)
1049 panic("%s: ifp=%p not on the ifnet tailq %p",
1050 __func__, ifp, &V_ifnet);
1051 else
1052 return; /* XXX this should panic as well? */
1053 #endif
1054 }
1055
1056 /*
1057 * At this point we know the interface still was on the ifnet list
1058 * and we removed it so we are in a stable state.
1059 */
1060 #ifdef VIMAGE
1061 curvnet->vnet_ifcnt--;
1062 #endif
1063
1064 /*
1065 * In any case (destroy or vmove) detach us from the groups
1066 * and remove/wait for pending events on the taskq.
1067 * XXX-BZ in theory an interface could still enqueue a taskq change?
1068 */
1069 if_delgroups(ifp);
1070
1071 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1072
1073 /*
1074 * Check if this is a cloned interface or not. Must do even if
1075 * shutting down as a if_vmove_reclaim() would move the ifp and
1076 * the if_clone_addgroup() will have a corrupted string overwise
1077 * from a gibberish pointer.
1078 */
1079 if (vmove && ifcp != NULL)
1080 *ifcp = if_clone_findifc(ifp);
1081
1082 if_down(ifp);
1083
1084 #ifdef VIMAGE
1085 /*
1086 * On VNET shutdown abort here as the stack teardown will do all
1087 * the work top-down for us.
1088 */
1089 if (shutdown) {
1090 /* Give interface users the chance to clean up. */
1091 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1092
1093 /*
1094 * In case of a vmove we are done here without error.
1095 * If we would signal an error it would lead to the same
1096 * abort as if we did not find the ifnet anymore.
1097 * if_detach() calls us in void context and does not care
1098 * about an early abort notification, so life is splendid :)
1099 */
1100 goto finish_vnet_shutdown;
1101 }
1102 #endif
1103
1104 /*
1105 * At this point we are not tearing down a VNET and are either
1106 * going to destroy or vmove the interface and have to cleanup
1107 * accordingly.
1108 */
1109
1110 /*
1111 * Remove routes and flush queues.
1112 */
1113 #ifdef ALTQ
1114 if (ALTQ_IS_ENABLED(&ifp->if_snd))
1115 altq_disable(&ifp->if_snd);
1116 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1117 altq_detach(&ifp->if_snd);
1118 #endif
1119
1120 if_purgeaddrs(ifp);
1121
1122 #ifdef INET
1123 in_ifdetach(ifp);
1124 #endif
1125
1126 #ifdef INET6
1127 /*
1128 * Remove all IPv6 kernel structs related to ifp. This should be done
1129 * before removing routing entries below, since IPv6 interface direct
1130 * routes are expected to be removed by the IPv6-specific kernel API.
1131 * Otherwise, the kernel will detect some inconsistency and bark it.
1132 */
1133 in6_ifdetach(ifp);
1134 #endif
1135 if_purgemaddrs(ifp);
1136
1137 /* Announce that the interface is gone. */
1138 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1139 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1140 if (IS_DEFAULT_VNET(curvnet))
1141 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1142
1143 if (!vmove) {
1144 /*
1145 * Prevent further calls into the device driver via ifnet.
1146 */
1147 if_dead(ifp);
1148
1149 /*
1150 * Remove link ifaddr pointer and maybe decrement if_index.
1151 * Clean up all addresses.
1152 */
1153 free(ifp->if_hw_addr, M_IFADDR);
1154 ifp->if_hw_addr = NULL;
1155 ifp->if_addr = NULL;
1156
1157 /* We can now free link ifaddr. */
1158 IF_ADDR_WLOCK(ifp);
1159 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
1160 ifa = TAILQ_FIRST(&ifp->if_addrhead);
1161 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
1162 IF_ADDR_WUNLOCK(ifp);
1163 ifa_free(ifa);
1164 } else
1165 IF_ADDR_WUNLOCK(ifp);
1166 }
1167
1168 rt_flushifroutes(ifp);
1169
1170 #ifdef VIMAGE
1171 finish_vnet_shutdown:
1172 #endif
1173 /*
1174 * We cannot hold the lock over dom_ifdetach calls as they might
1175 * sleep, for example trying to drain a callout, thus open up the
1176 * theoretical race with re-attaching.
1177 */
1178 IF_AFDATA_LOCK(ifp);
1179 i = ifp->if_afdata_initialized;
1180 ifp->if_afdata_initialized = 0;
1181 IF_AFDATA_UNLOCK(ifp);
1182 for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1183 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1184 (*dp->dom_ifdetach)(ifp,
1185 ifp->if_afdata[dp->dom_family]);
1186 ifp->if_afdata[dp->dom_family] = NULL;
1187 }
1188 }
1189
1190 return (0);
1191 }
1192
1193 #ifdef VIMAGE
1194 /*
1195 * if_vmove() performs a limited version of if_detach() in current
1196 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1197 * An attempt is made to shrink if_index in current vnet, find an
1198 * unused if_index in target vnet and calls if_grow() if necessary,
1199 * and finally find an unused if_xname for the target vnet.
1200 */
1201 static void
1202 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1203 {
1204 struct if_clone *ifc;
1205 #ifdef DEV_BPF
1206 u_int bif_dlt, bif_hdrlen;
1207 #endif
1208 int rc;
1209
1210 #ifdef DEV_BPF
1211 /*
1212 * if_detach_internal() will call the eventhandler to notify
1213 * interface departure. That will detach if_bpf. We need to
1214 * safe the dlt and hdrlen so we can re-attach it later.
1215 */
1216 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1217 #endif
1218
1219 /*
1220 * Detach from current vnet, but preserve LLADDR info, do not
1221 * mark as dead etc. so that the ifnet can be reattached later.
1222 * If we cannot find it, we lost the race to someone else.
1223 */
1224 rc = if_detach_internal(ifp, 1, &ifc);
1225 if (rc != 0)
1226 return;
1227
1228 /*
1229 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1230 * the if_index for that vnet if possible.
1231 *
1232 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1233 * or we'd lock on one vnet and unlock on another.
1234 */
1235 IFNET_WLOCK();
1236 ifindex_free_locked(ifp->if_index);
1237 IFNET_WUNLOCK();
1238
1239 /*
1240 * Perform interface-specific reassignment tasks, if provided by
1241 * the driver.
1242 */
1243 if (ifp->if_reassign != NULL)
1244 ifp->if_reassign(ifp, new_vnet, NULL);
1245
1246 /*
1247 * Switch to the context of the target vnet.
1248 */
1249 CURVNET_SET_QUIET(new_vnet);
1250
1251 IFNET_WLOCK();
1252 ifp->if_index = ifindex_alloc();
1253 ifnet_setbyindex_locked(ifp->if_index, ifp);
1254 IFNET_WUNLOCK();
1255
1256 if_attach_internal(ifp, 1, ifc);
1257
1258 #ifdef DEV_BPF
1259 if (ifp->if_bpf == NULL)
1260 bpfattach(ifp, bif_dlt, bif_hdrlen);
1261 #endif
1262
1263 CURVNET_RESTORE();
1264 }
1265
1266 /*
1267 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1268 */
1269 static int
1270 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1271 {
1272 struct prison *pr;
1273 struct ifnet *difp;
1274 int shutdown;
1275
1276 /* Try to find the prison within our visibility. */
1277 sx_slock(&allprison_lock);
1278 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1279 sx_sunlock(&allprison_lock);
1280 if (pr == NULL)
1281 return (ENXIO);
1282 prison_hold_locked(pr);
1283 mtx_unlock(&pr->pr_mtx);
1284
1285 /* Do not try to move the iface from and to the same prison. */
1286 if (pr->pr_vnet == ifp->if_vnet) {
1287 prison_free(pr);
1288 return (EEXIST);
1289 }
1290
1291 /* Make sure the named iface does not exists in the dst. prison/vnet. */
1292 /* XXX Lock interfaces to avoid races. */
1293 CURVNET_SET_QUIET(pr->pr_vnet);
1294 difp = ifunit(ifname);
1295 if (difp != NULL) {
1296 CURVNET_RESTORE();
1297 prison_free(pr);
1298 return (EEXIST);
1299 }
1300
1301 /* Make sure the VNET is stable. */
1302 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1303 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1304 if (shutdown) {
1305 CURVNET_RESTORE();
1306 prison_free(pr);
1307 return (EBUSY);
1308 }
1309 CURVNET_RESTORE();
1310
1311 /* Move the interface into the child jail/vnet. */
1312 if_vmove(ifp, pr->pr_vnet);
1313
1314 /* Report the new if_xname back to the userland. */
1315 sprintf(ifname, "%s", ifp->if_xname);
1316
1317 prison_free(pr);
1318 return (0);
1319 }
1320
1321 static int
1322 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1323 {
1324 struct prison *pr;
1325 struct vnet *vnet_dst;
1326 struct ifnet *ifp;
1327 int shutdown;
1328
1329 /* Try to find the prison within our visibility. */
1330 sx_slock(&allprison_lock);
1331 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1332 sx_sunlock(&allprison_lock);
1333 if (pr == NULL)
1334 return (ENXIO);
1335 prison_hold_locked(pr);
1336 mtx_unlock(&pr->pr_mtx);
1337
1338 /* Make sure the named iface exists in the source prison/vnet. */
1339 CURVNET_SET(pr->pr_vnet);
1340 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1341 if (ifp == NULL) {
1342 CURVNET_RESTORE();
1343 prison_free(pr);
1344 return (ENXIO);
1345 }
1346
1347 /* Do not try to move the iface from and to the same prison. */
1348 vnet_dst = TD_TO_VNET(td);
1349 if (vnet_dst == ifp->if_vnet) {
1350 CURVNET_RESTORE();
1351 prison_free(pr);
1352 return (EEXIST);
1353 }
1354
1355 /* Make sure the VNET is stable. */
1356 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1357 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1358 if (shutdown) {
1359 CURVNET_RESTORE();
1360 prison_free(pr);
1361 return (EBUSY);
1362 }
1363
1364 /* Get interface back from child jail/vnet. */
1365 if_vmove(ifp, vnet_dst);
1366 CURVNET_RESTORE();
1367
1368 /* Report the new if_xname back to the userland. */
1369 sprintf(ifname, "%s", ifp->if_xname);
1370
1371 prison_free(pr);
1372 return (0);
1373 }
1374 #endif /* VIMAGE */
1375
1376 /*
1377 * Add a group to an interface
1378 */
1379 int
1380 if_addgroup(struct ifnet *ifp, const char *groupname)
1381 {
1382 struct ifg_list *ifgl;
1383 struct ifg_group *ifg = NULL;
1384 struct ifg_member *ifgm;
1385 int new = 0;
1386
1387 if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
1388 groupname[strlen(groupname) - 1] <= '9')
1389 return (EINVAL);
1390
1391 IFNET_WLOCK();
1392 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1393 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1394 IFNET_WUNLOCK();
1395 return (EEXIST);
1396 }
1397
1398 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1399 M_NOWAIT)) == NULL) {
1400 IFNET_WUNLOCK();
1401 return (ENOMEM);
1402 }
1403
1404 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1405 M_TEMP, M_NOWAIT)) == NULL) {
1406 free(ifgl, M_TEMP);
1407 IFNET_WUNLOCK();
1408 return (ENOMEM);
1409 }
1410
1411 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1412 if (!strcmp(ifg->ifg_group, groupname))
1413 break;
1414
1415 if (ifg == NULL) {
1416 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1417 M_TEMP, M_NOWAIT)) == NULL) {
1418 free(ifgl, M_TEMP);
1419 free(ifgm, M_TEMP);
1420 IFNET_WUNLOCK();
1421 return (ENOMEM);
1422 }
1423 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1424 ifg->ifg_refcnt = 0;
1425 TAILQ_INIT(&ifg->ifg_members);
1426 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1427 new = 1;
1428 }
1429
1430 ifg->ifg_refcnt++;
1431 ifgl->ifgl_group = ifg;
1432 ifgm->ifgm_ifp = ifp;
1433
1434 IF_ADDR_WLOCK(ifp);
1435 TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1436 TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1437 IF_ADDR_WUNLOCK(ifp);
1438
1439 IFNET_WUNLOCK();
1440
1441 if (new)
1442 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1443 EVENTHANDLER_INVOKE(group_change_event, groupname);
1444
1445 return (0);
1446 }
1447
1448 /*
1449 * Remove a group from an interface
1450 */
1451 int
1452 if_delgroup(struct ifnet *ifp, const char *groupname)
1453 {
1454 struct ifg_list *ifgl;
1455 struct ifg_member *ifgm;
1456
1457 IFNET_WLOCK();
1458 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1459 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1460 break;
1461 if (ifgl == NULL) {
1462 IFNET_WUNLOCK();
1463 return (ENOENT);
1464 }
1465
1466 IF_ADDR_WLOCK(ifp);
1467 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1468 IF_ADDR_WUNLOCK(ifp);
1469
1470 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1471 if (ifgm->ifgm_ifp == ifp)
1472 break;
1473
1474 if (ifgm != NULL) {
1475 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1476 free(ifgm, M_TEMP);
1477 }
1478
1479 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1480 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1481 IFNET_WUNLOCK();
1482 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1483 free(ifgl->ifgl_group, M_TEMP);
1484 } else
1485 IFNET_WUNLOCK();
1486
1487 free(ifgl, M_TEMP);
1488
1489 EVENTHANDLER_INVOKE(group_change_event, groupname);
1490
1491 return (0);
1492 }
1493
1494 /*
1495 * Remove an interface from all groups
1496 */
1497 static void
1498 if_delgroups(struct ifnet *ifp)
1499 {
1500 struct ifg_list *ifgl;
1501 struct ifg_member *ifgm;
1502 char groupname[IFNAMSIZ];
1503
1504 IFNET_WLOCK();
1505 while (!TAILQ_EMPTY(&ifp->if_groups)) {
1506 ifgl = TAILQ_FIRST(&ifp->if_groups);
1507
1508 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1509
1510 IF_ADDR_WLOCK(ifp);
1511 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1512 IF_ADDR_WUNLOCK(ifp);
1513
1514 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1515 if (ifgm->ifgm_ifp == ifp)
1516 break;
1517
1518 if (ifgm != NULL) {
1519 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1520 ifgm_next);
1521 free(ifgm, M_TEMP);
1522 }
1523
1524 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1525 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1526 IFNET_WUNLOCK();
1527 EVENTHANDLER_INVOKE(group_detach_event,
1528 ifgl->ifgl_group);
1529 free(ifgl->ifgl_group, M_TEMP);
1530 } else
1531 IFNET_WUNLOCK();
1532
1533 free(ifgl, M_TEMP);
1534
1535 EVENTHANDLER_INVOKE(group_change_event, groupname);
1536
1537 IFNET_WLOCK();
1538 }
1539 IFNET_WUNLOCK();
1540 }
1541
1542 static char *
1543 ifgr_group_get(void *ifgrp)
1544 {
1545 union ifgroupreq_union *ifgrup;
1546
1547 ifgrup = ifgrp;
1548 #ifdef COMPAT_FREEBSD32
1549 if (SV_CURPROC_FLAG(SV_ILP32))
1550 return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1551 #endif
1552 return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1553 }
1554
1555 static struct ifg_req *
1556 ifgr_groups_get(void *ifgrp)
1557 {
1558 union ifgroupreq_union *ifgrup;
1559
1560 ifgrup = ifgrp;
1561 #ifdef COMPAT_FREEBSD32
1562 if (SV_CURPROC_FLAG(SV_ILP32))
1563 return ((struct ifg_req *)(uintptr_t)
1564 ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1565 #endif
1566 return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1567 }
1568
1569 /*
1570 * Stores all groups from an interface in memory pointed to by ifgr.
1571 */
1572 static int
1573 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1574 {
1575 int len, error;
1576 struct ifg_list *ifgl;
1577 struct ifg_req ifgrq, *ifgp;
1578
1579 if (ifgr->ifgr_len == 0) {
1580 IF_ADDR_RLOCK(ifp);
1581 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1582 ifgr->ifgr_len += sizeof(struct ifg_req);
1583 IF_ADDR_RUNLOCK(ifp);
1584 return (0);
1585 }
1586
1587 len = ifgr->ifgr_len;
1588 ifgp = ifgr_groups_get(ifgr);
1589 /* XXX: wire */
1590 IF_ADDR_RLOCK(ifp);
1591 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1592 if (len < sizeof(ifgrq)) {
1593 IF_ADDR_RUNLOCK(ifp);
1594 return (EINVAL);
1595 }
1596 bzero(&ifgrq, sizeof ifgrq);
1597 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1598 sizeof(ifgrq.ifgrq_group));
1599 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1600 IF_ADDR_RUNLOCK(ifp);
1601 return (error);
1602 }
1603 len -= sizeof(ifgrq);
1604 ifgp++;
1605 }
1606 IF_ADDR_RUNLOCK(ifp);
1607
1608 return (0);
1609 }
1610
1611 /*
1612 * Stores all members of a group in memory pointed to by igfr
1613 */
1614 static int
1615 if_getgroupmembers(struct ifgroupreq *ifgr)
1616 {
1617 struct ifg_group *ifg;
1618 struct ifg_member *ifgm;
1619 struct ifg_req ifgrq, *ifgp;
1620 int len, error;
1621
1622 IFNET_RLOCK();
1623 TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1624 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1625 break;
1626 if (ifg == NULL) {
1627 IFNET_RUNLOCK();
1628 return (ENOENT);
1629 }
1630
1631 if (ifgr->ifgr_len == 0) {
1632 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1633 ifgr->ifgr_len += sizeof(ifgrq);
1634 IFNET_RUNLOCK();
1635 return (0);
1636 }
1637
1638 len = ifgr->ifgr_len;
1639 ifgp = ifgr_groups_get(ifgr);
1640 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1641 if (len < sizeof(ifgrq)) {
1642 IFNET_RUNLOCK();
1643 return (EINVAL);
1644 }
1645 bzero(&ifgrq, sizeof ifgrq);
1646 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1647 sizeof(ifgrq.ifgrq_member));
1648 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1649 IFNET_RUNLOCK();
1650 return (error);
1651 }
1652 len -= sizeof(ifgrq);
1653 ifgp++;
1654 }
1655 IFNET_RUNLOCK();
1656
1657 return (0);
1658 }
1659
1660 /*
1661 * Return counter values from counter(9)s stored in ifnet.
1662 */
1663 uint64_t
1664 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1665 {
1666
1667 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1668
1669 return (counter_u64_fetch(ifp->if_counters[cnt]));
1670 }
1671
1672 /*
1673 * Increase an ifnet counter. Usually used for counters shared
1674 * between the stack and a driver, but function supports them all.
1675 */
1676 void
1677 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1678 {
1679
1680 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1681
1682 counter_u64_add(ifp->if_counters[cnt], inc);
1683 }
1684
1685 /*
1686 * Copy data from ifnet to userland API structure if_data.
1687 */
1688 void
1689 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1690 {
1691
1692 ifd->ifi_type = ifp->if_type;
1693 ifd->ifi_physical = 0;
1694 ifd->ifi_addrlen = ifp->if_addrlen;
1695 ifd->ifi_hdrlen = ifp->if_hdrlen;
1696 ifd->ifi_link_state = ifp->if_link_state;
1697 ifd->ifi_vhid = 0;
1698 ifd->ifi_datalen = sizeof(struct if_data);
1699 ifd->ifi_mtu = ifp->if_mtu;
1700 ifd->ifi_metric = ifp->if_metric;
1701 ifd->ifi_baudrate = ifp->if_baudrate;
1702 ifd->ifi_hwassist = ifp->if_hwassist;
1703 ifd->ifi_epoch = ifp->if_epoch;
1704 ifd->ifi_lastchange = ifp->if_lastchange;
1705
1706 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1707 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1708 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1709 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1710 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1711 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1712 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1713 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1714 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1715 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1716 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1717 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1718 }
1719
1720 /*
1721 * Wrapper functions for struct ifnet address list locking macros. These are
1722 * used by kernel modules to avoid encoding programming interface or binary
1723 * interface assumptions that may be violated when kernel-internal locking
1724 * approaches change.
1725 */
1726 void
1727 if_addr_rlock(struct ifnet *ifp)
1728 {
1729
1730 IF_ADDR_RLOCK(ifp);
1731 }
1732
1733 void
1734 if_addr_runlock(struct ifnet *ifp)
1735 {
1736
1737 IF_ADDR_RUNLOCK(ifp);
1738 }
1739
1740 void
1741 if_maddr_rlock(if_t ifp)
1742 {
1743
1744 IF_ADDR_RLOCK((struct ifnet *)ifp);
1745 }
1746
1747 void
1748 if_maddr_runlock(if_t ifp)
1749 {
1750
1751 IF_ADDR_RUNLOCK((struct ifnet *)ifp);
1752 }
1753
1754 /*
1755 * Initialization, destruction and refcounting functions for ifaddrs.
1756 */
1757 struct ifaddr *
1758 ifa_alloc(size_t size, int flags)
1759 {
1760 struct ifaddr *ifa;
1761
1762 KASSERT(size >= sizeof(struct ifaddr),
1763 ("%s: invalid size %zu", __func__, size));
1764
1765 ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1766 if (ifa == NULL)
1767 return (NULL);
1768
1769 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1770 goto fail;
1771 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1772 goto fail;
1773 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1774 goto fail;
1775 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1776 goto fail;
1777
1778 refcount_init(&ifa->ifa_refcnt, 1);
1779
1780 return (ifa);
1781
1782 fail:
1783 /* free(NULL) is okay */
1784 counter_u64_free(ifa->ifa_opackets);
1785 counter_u64_free(ifa->ifa_ipackets);
1786 counter_u64_free(ifa->ifa_obytes);
1787 counter_u64_free(ifa->ifa_ibytes);
1788 free(ifa, M_IFADDR);
1789
1790 return (NULL);
1791 }
1792
1793 void
1794 ifa_ref(struct ifaddr *ifa)
1795 {
1796
1797 refcount_acquire(&ifa->ifa_refcnt);
1798 }
1799
1800 void
1801 ifa_free(struct ifaddr *ifa)
1802 {
1803
1804 if (refcount_release(&ifa->ifa_refcnt)) {
1805 counter_u64_free(ifa->ifa_opackets);
1806 counter_u64_free(ifa->ifa_ipackets);
1807 counter_u64_free(ifa->ifa_obytes);
1808 counter_u64_free(ifa->ifa_ibytes);
1809 free(ifa, M_IFADDR);
1810 }
1811 }
1812
1813 static int
1814 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1815 struct sockaddr *ia)
1816 {
1817 int error;
1818 struct rt_addrinfo info;
1819 struct sockaddr_dl null_sdl;
1820 struct ifnet *ifp;
1821
1822 ifp = ifa->ifa_ifp;
1823
1824 bzero(&info, sizeof(info));
1825 if (cmd != RTM_DELETE)
1826 info.rti_ifp = V_loif;
1827 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1828 info.rti_info[RTAX_DST] = ia;
1829 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1830 link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1831
1832 error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1833
1834 if (error == 0 ||
1835 (cmd == RTM_ADD && error == EEXIST) ||
1836 (cmd == RTM_DELETE && (error == ENOENT || error == ESRCH)))
1837 return (error);
1838
1839 log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
1840 __func__, otype, if_name(ifp), error);
1841
1842 return (error);
1843 }
1844
1845 int
1846 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1847 {
1848
1849 return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1850 }
1851
1852 int
1853 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1854 {
1855
1856 return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1857 }
1858
1859 int
1860 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1861 {
1862
1863 return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1864 }
1865
1866 /*
1867 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1868 * structs used to represent other address families, it is necessary
1869 * to perform a different comparison.
1870 */
1871
1872 #define sa_dl_equal(a1, a2) \
1873 ((((const struct sockaddr_dl *)(a1))->sdl_len == \
1874 ((const struct sockaddr_dl *)(a2))->sdl_len) && \
1875 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \
1876 CLLADDR((const struct sockaddr_dl *)(a2)), \
1877 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1878
1879 /*
1880 * Locate an interface based on a complete address.
1881 */
1882 /*ARGSUSED*/
1883 static struct ifaddr *
1884 ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1885 {
1886 struct ifnet *ifp;
1887 struct ifaddr *ifa;
1888
1889 IFNET_RLOCK_NOSLEEP();
1890 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1891 IF_ADDR_RLOCK(ifp);
1892 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1893 if (ifa->ifa_addr->sa_family != addr->sa_family)
1894 continue;
1895 if (sa_equal(addr, ifa->ifa_addr)) {
1896 if (getref)
1897 ifa_ref(ifa);
1898 IF_ADDR_RUNLOCK(ifp);
1899 goto done;
1900 }
1901 /* IP6 doesn't have broadcast */
1902 if ((ifp->if_flags & IFF_BROADCAST) &&
1903 ifa->ifa_broadaddr &&
1904 ifa->ifa_broadaddr->sa_len != 0 &&
1905 sa_equal(ifa->ifa_broadaddr, addr)) {
1906 if (getref)
1907 ifa_ref(ifa);
1908 IF_ADDR_RUNLOCK(ifp);
1909 goto done;
1910 }
1911 }
1912 IF_ADDR_RUNLOCK(ifp);
1913 }
1914 ifa = NULL;
1915 done:
1916 IFNET_RUNLOCK_NOSLEEP();
1917 return (ifa);
1918 }
1919
1920 struct ifaddr *
1921 ifa_ifwithaddr(const struct sockaddr *addr)
1922 {
1923
1924 return (ifa_ifwithaddr_internal(addr, 1));
1925 }
1926
1927 int
1928 ifa_ifwithaddr_check(const struct sockaddr *addr)
1929 {
1930
1931 return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1932 }
1933
1934 /*
1935 * Locate an interface based on the broadcast address.
1936 */
1937 /* ARGSUSED */
1938 struct ifaddr *
1939 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1940 {
1941 struct ifnet *ifp;
1942 struct ifaddr *ifa;
1943
1944 IFNET_RLOCK_NOSLEEP();
1945 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1946 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1947 continue;
1948 IF_ADDR_RLOCK(ifp);
1949 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1950 if (ifa->ifa_addr->sa_family != addr->sa_family)
1951 continue;
1952 if ((ifp->if_flags & IFF_BROADCAST) &&
1953 ifa->ifa_broadaddr &&
1954 ifa->ifa_broadaddr->sa_len != 0 &&
1955 sa_equal(ifa->ifa_broadaddr, addr)) {
1956 ifa_ref(ifa);
1957 IF_ADDR_RUNLOCK(ifp);
1958 goto done;
1959 }
1960 }
1961 IF_ADDR_RUNLOCK(ifp);
1962 }
1963 ifa = NULL;
1964 done:
1965 IFNET_RUNLOCK_NOSLEEP();
1966 return (ifa);
1967 }
1968
1969 /*
1970 * Locate the point to point interface with a given destination address.
1971 */
1972 /*ARGSUSED*/
1973 struct ifaddr *
1974 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1975 {
1976 struct ifnet *ifp;
1977 struct ifaddr *ifa;
1978
1979 IFNET_RLOCK_NOSLEEP();
1980 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1981 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1982 continue;
1983 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1984 continue;
1985 IF_ADDR_RLOCK(ifp);
1986 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1987 if (ifa->ifa_addr->sa_family != addr->sa_family)
1988 continue;
1989 if (ifa->ifa_dstaddr != NULL &&
1990 sa_equal(addr, ifa->ifa_dstaddr)) {
1991 ifa_ref(ifa);
1992 IF_ADDR_RUNLOCK(ifp);
1993 goto done;
1994 }
1995 }
1996 IF_ADDR_RUNLOCK(ifp);
1997 }
1998 ifa = NULL;
1999 done:
2000 IFNET_RUNLOCK_NOSLEEP();
2001 return (ifa);
2002 }
2003
2004 /*
2005 * Find an interface on a specific network. If many, choice
2006 * is most specific found.
2007 */
2008 struct ifaddr *
2009 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2010 {
2011 struct ifnet *ifp;
2012 struct ifaddr *ifa;
2013 struct ifaddr *ifa_maybe = NULL;
2014 u_int af = addr->sa_family;
2015 const char *addr_data = addr->sa_data, *cplim;
2016
2017 /*
2018 * AF_LINK addresses can be looked up directly by their index number,
2019 * so do that if we can.
2020 */
2021 if (af == AF_LINK) {
2022 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2023 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2024 return (ifaddr_byindex(sdl->sdl_index));
2025 }
2026
2027 /*
2028 * Scan though each interface, looking for ones that have addresses
2029 * in this address family and the requested fib. Maintain a reference
2030 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
2031 * kept it stable when we move onto the next interface.
2032 */
2033 IFNET_RLOCK_NOSLEEP();
2034 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2035 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2036 continue;
2037 IF_ADDR_RLOCK(ifp);
2038 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2039 const char *cp, *cp2, *cp3;
2040
2041 if (ifa->ifa_addr->sa_family != af)
2042 next: continue;
2043 if (af == AF_INET &&
2044 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2045 /*
2046 * This is a bit broken as it doesn't
2047 * take into account that the remote end may
2048 * be a single node in the network we are
2049 * looking for.
2050 * The trouble is that we don't know the
2051 * netmask for the remote end.
2052 */
2053 if (ifa->ifa_dstaddr != NULL &&
2054 sa_equal(addr, ifa->ifa_dstaddr)) {
2055 ifa_ref(ifa);
2056 IF_ADDR_RUNLOCK(ifp);
2057 goto done;
2058 }
2059 } else {
2060 /*
2061 * Scan all the bits in the ifa's address.
2062 * If a bit dissagrees with what we are
2063 * looking for, mask it with the netmask
2064 * to see if it really matters.
2065 * (A byte at a time)
2066 */
2067 if (ifa->ifa_netmask == 0)
2068 continue;
2069 cp = addr_data;
2070 cp2 = ifa->ifa_addr->sa_data;
2071 cp3 = ifa->ifa_netmask->sa_data;
2072 cplim = ifa->ifa_netmask->sa_len
2073 + (char *)ifa->ifa_netmask;
2074 while (cp3 < cplim)
2075 if ((*cp++ ^ *cp2++) & *cp3++)
2076 goto next; /* next address! */
2077 /*
2078 * If the netmask of what we just found
2079 * is more specific than what we had before
2080 * (if we had one), or if the virtual status
2081 * of new prefix is better than of the old one,
2082 * then remember the new one before continuing
2083 * to search for an even better one.
2084 */
2085 if (ifa_maybe == NULL ||
2086 ifa_preferred(ifa_maybe, ifa) ||
2087 rn_refines((caddr_t)ifa->ifa_netmask,
2088 (caddr_t)ifa_maybe->ifa_netmask)) {
2089 if (ifa_maybe != NULL)
2090 ifa_free(ifa_maybe);
2091 ifa_maybe = ifa;
2092 ifa_ref(ifa_maybe);
2093 }
2094 }
2095 }
2096 IF_ADDR_RUNLOCK(ifp);
2097 }
2098 ifa = ifa_maybe;
2099 ifa_maybe = NULL;
2100 done:
2101 IFNET_RUNLOCK_NOSLEEP();
2102 if (ifa_maybe != NULL)
2103 ifa_free(ifa_maybe);
2104 return (ifa);
2105 }
2106
2107 /*
2108 * Find an interface address specific to an interface best matching
2109 * a given address.
2110 */
2111 struct ifaddr *
2112 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2113 {
2114 struct ifaddr *ifa;
2115 const char *cp, *cp2, *cp3;
2116 char *cplim;
2117 struct ifaddr *ifa_maybe = NULL;
2118 u_int af = addr->sa_family;
2119
2120 if (af >= AF_MAX)
2121 return (NULL);
2122 IF_ADDR_RLOCK(ifp);
2123 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2124 if (ifa->ifa_addr->sa_family != af)
2125 continue;
2126 if (ifa_maybe == NULL)
2127 ifa_maybe = ifa;
2128 if (ifa->ifa_netmask == 0) {
2129 if (sa_equal(addr, ifa->ifa_addr) ||
2130 (ifa->ifa_dstaddr &&
2131 sa_equal(addr, ifa->ifa_dstaddr)))
2132 goto done;
2133 continue;
2134 }
2135 if (ifp->if_flags & IFF_POINTOPOINT) {
2136 if (sa_equal(addr, ifa->ifa_dstaddr))
2137 goto done;
2138 } else {
2139 cp = addr->sa_data;
2140 cp2 = ifa->ifa_addr->sa_data;
2141 cp3 = ifa->ifa_netmask->sa_data;
2142 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2143 for (; cp3 < cplim; cp3++)
2144 if ((*cp++ ^ *cp2++) & *cp3)
2145 break;
2146 if (cp3 == cplim)
2147 goto done;
2148 }
2149 }
2150 ifa = ifa_maybe;
2151 done:
2152 if (ifa != NULL)
2153 ifa_ref(ifa);
2154 IF_ADDR_RUNLOCK(ifp);
2155 return (ifa);
2156 }
2157
2158 /*
2159 * See whether new ifa is better than current one:
2160 * 1) A non-virtual one is preferred over virtual.
2161 * 2) A virtual in master state preferred over any other state.
2162 *
2163 * Used in several address selecting functions.
2164 */
2165 int
2166 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2167 {
2168
2169 return (cur->ifa_carp && (!next->ifa_carp ||
2170 ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2171 }
2172
2173 #include <net/if_llatbl.h>
2174
2175 /*
2176 * Default action when installing a route with a Link Level gateway.
2177 * Lookup an appropriate real ifa to point to.
2178 * This should be moved to /sys/net/link.c eventually.
2179 */
2180 static void
2181 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2182 {
2183 struct ifaddr *ifa, *oifa;
2184 struct sockaddr *dst;
2185 struct ifnet *ifp;
2186
2187 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2188 ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2189 return;
2190 ifa = ifaof_ifpforaddr(dst, ifp);
2191 if (ifa) {
2192 oifa = rt->rt_ifa;
2193 rt->rt_ifa = ifa;
2194 ifa_free(oifa);
2195 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2196 ifa->ifa_rtrequest(cmd, rt, info);
2197 }
2198 }
2199
2200 struct sockaddr_dl *
2201 link_alloc_sdl(size_t size, int flags)
2202 {
2203
2204 return (malloc(size, M_TEMP, flags));
2205 }
2206
2207 void
2208 link_free_sdl(struct sockaddr *sa)
2209 {
2210 free(sa, M_TEMP);
2211 }
2212
2213 /*
2214 * Fills in given sdl with interface basic info.
2215 * Returns pointer to filled sdl.
2216 */
2217 struct sockaddr_dl *
2218 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2219 {
2220 struct sockaddr_dl *sdl;
2221
2222 sdl = (struct sockaddr_dl *)paddr;
2223 memset(sdl, 0, sizeof(struct sockaddr_dl));
2224 sdl->sdl_len = sizeof(struct sockaddr_dl);
2225 sdl->sdl_family = AF_LINK;
2226 sdl->sdl_index = ifp->if_index;
2227 sdl->sdl_type = iftype;
2228
2229 return (sdl);
2230 }
2231
2232 /*
2233 * Mark an interface down and notify protocols of
2234 * the transition.
2235 */
2236 static void
2237 if_unroute(struct ifnet *ifp, int flag, int fam)
2238 {
2239 struct ifaddr *ifa;
2240
2241 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2242
2243 ifp->if_flags &= ~flag;
2244 getmicrotime(&ifp->if_lastchange);
2245 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2246 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2247 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2248 ifp->if_qflush(ifp);
2249
2250 if (ifp->if_carp)
2251 (*carp_linkstate_p)(ifp);
2252 rt_ifmsg(ifp);
2253 }
2254
2255 /*
2256 * Mark an interface up and notify protocols of
2257 * the transition.
2258 */
2259 static void
2260 if_route(struct ifnet *ifp, int flag, int fam)
2261 {
2262 struct ifaddr *ifa;
2263
2264 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2265
2266 ifp->if_flags |= flag;
2267 getmicrotime(&ifp->if_lastchange);
2268 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2269 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2270 pfctlinput(PRC_IFUP, ifa->ifa_addr);
2271 if (ifp->if_carp)
2272 (*carp_linkstate_p)(ifp);
2273 rt_ifmsg(ifp);
2274 #ifdef INET6
2275 in6_if_up(ifp);
2276 #endif
2277 }
2278
2279 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
2280 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
2281 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2282 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2283 int (*vlan_tag_p)(struct ifnet *, uint16_t *);
2284 int (*vlan_setcookie_p)(struct ifnet *, void *);
2285 void *(*vlan_cookie_p)(struct ifnet *);
2286
2287 /*
2288 * Handle a change in the interface link state. To avoid LORs
2289 * between driver lock and upper layer locks, as well as possible
2290 * recursions, we post event to taskqueue, and all job
2291 * is done in static do_link_state_change().
2292 */
2293 void
2294 if_link_state_change(struct ifnet *ifp, int link_state)
2295 {
2296 /* Return if state hasn't changed. */
2297 if (ifp->if_link_state == link_state)
2298 return;
2299
2300 ifp->if_link_state = link_state;
2301
2302 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2303 }
2304
2305 static void
2306 do_link_state_change(void *arg, int pending)
2307 {
2308 struct ifnet *ifp = (struct ifnet *)arg;
2309 int link_state = ifp->if_link_state;
2310 CURVNET_SET(ifp->if_vnet);
2311
2312 /* Notify that the link state has changed. */
2313 rt_ifmsg(ifp);
2314 if (ifp->if_vlantrunk != NULL)
2315 (*vlan_link_state_p)(ifp);
2316
2317 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2318 ifp->if_l2com != NULL)
2319 (*ng_ether_link_state_p)(ifp, link_state);
2320 if (ifp->if_carp)
2321 (*carp_linkstate_p)(ifp);
2322 if (ifp->if_bridge)
2323 (*bridge_linkstate_p)(ifp);
2324 if (ifp->if_lagg)
2325 (*lagg_linkstate_p)(ifp, link_state);
2326
2327 if (IS_DEFAULT_VNET(curvnet))
2328 devctl_notify("IFNET", ifp->if_xname,
2329 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2330 NULL);
2331 if (pending > 1)
2332 if_printf(ifp, "%d link states coalesced\n", pending);
2333 if (log_link_state_change)
2334 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
2335 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2336 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2337 CURVNET_RESTORE();
2338 }
2339
2340 /*
2341 * Mark an interface down and notify protocols of
2342 * the transition.
2343 */
2344 void
2345 if_down(struct ifnet *ifp)
2346 {
2347
2348 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2349 if_unroute(ifp, IFF_UP, AF_UNSPEC);
2350 }
2351
2352 /*
2353 * Mark an interface up and notify protocols of
2354 * the transition.
2355 */
2356 void
2357 if_up(struct ifnet *ifp)
2358 {
2359
2360 if_route(ifp, IFF_UP, AF_UNSPEC);
2361 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2362 }
2363
2364 /*
2365 * Flush an interface queue.
2366 */
2367 void
2368 if_qflush(struct ifnet *ifp)
2369 {
2370 struct mbuf *m, *n;
2371 struct ifaltq *ifq;
2372
2373 ifq = &ifp->if_snd;
2374 IFQ_LOCK(ifq);
2375 #ifdef ALTQ
2376 if (ALTQ_IS_ENABLED(ifq))
2377 ALTQ_PURGE(ifq);
2378 #endif
2379 n = ifq->ifq_head;
2380 while ((m = n) != NULL) {
2381 n = m->m_nextpkt;
2382 m_freem(m);
2383 }
2384 ifq->ifq_head = 0;
2385 ifq->ifq_tail = 0;
2386 ifq->ifq_len = 0;
2387 IFQ_UNLOCK(ifq);
2388 }
2389
2390 /*
2391 * Map interface name to interface structure pointer, with or without
2392 * returning a reference.
2393 */
2394 struct ifnet *
2395 ifunit_ref(const char *name)
2396 {
2397 struct ifnet *ifp;
2398
2399 IFNET_RLOCK_NOSLEEP();
2400 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2401 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2402 !(ifp->if_flags & IFF_DYING))
2403 break;
2404 }
2405 if (ifp != NULL)
2406 if_ref(ifp);
2407 IFNET_RUNLOCK_NOSLEEP();
2408 return (ifp);
2409 }
2410
2411 struct ifnet *
2412 ifunit(const char *name)
2413 {
2414 struct ifnet *ifp;
2415
2416 IFNET_RLOCK_NOSLEEP();
2417 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2418 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2419 break;
2420 }
2421 IFNET_RUNLOCK_NOSLEEP();
2422 return (ifp);
2423 }
2424
2425 void *
2426 ifr_buffer_get_buffer(void *data)
2427 {
2428 union ifreq_union *ifrup;
2429
2430 ifrup = data;
2431 #ifdef COMPAT_FREEBSD32
2432 if (SV_CURPROC_FLAG(SV_ILP32))
2433 return ((void *)(uintptr_t)
2434 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2435 #endif
2436 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2437 }
2438
2439 static void
2440 ifr_buffer_set_buffer_null(void *data)
2441 {
2442 union ifreq_union *ifrup;
2443
2444 ifrup = data;
2445 #ifdef COMPAT_FREEBSD32
2446 if (SV_CURPROC_FLAG(SV_ILP32))
2447 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2448 else
2449 #endif
2450 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2451 }
2452
2453 size_t
2454 ifr_buffer_get_length(void *data)
2455 {
2456 union ifreq_union *ifrup;
2457
2458 ifrup = data;
2459 #ifdef COMPAT_FREEBSD32
2460 if (SV_CURPROC_FLAG(SV_ILP32))
2461 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2462 #endif
2463 return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2464 }
2465
2466 static void
2467 ifr_buffer_set_length(void *data, size_t len)
2468 {
2469 union ifreq_union *ifrup;
2470
2471 ifrup = data;
2472 #ifdef COMPAT_FREEBSD32
2473 if (SV_CURPROC_FLAG(SV_ILP32))
2474 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2475 else
2476 #endif
2477 ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2478 }
2479
2480 void *
2481 ifr_data_get_ptr(void *ifrp)
2482 {
2483 union ifreq_union *ifrup;
2484
2485 ifrup = ifrp;
2486 #ifdef COMPAT_FREEBSD32
2487 if (SV_CURPROC_FLAG(SV_ILP32))
2488 return ((void *)(uintptr_t)
2489 ifrup->ifr32.ifr_ifru.ifru_data);
2490 #endif
2491 return (ifrup->ifr.ifr_ifru.ifru_data);
2492 }
2493
2494 /*
2495 * Hardware specific interface ioctls.
2496 */
2497 int
2498 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2499 {
2500 struct ifreq *ifr;
2501 int error = 0, do_ifup = 0;
2502 int new_flags, temp_flags;
2503 size_t namelen, onamelen;
2504 size_t descrlen;
2505 char *descrbuf, *odescrbuf;
2506 char new_name[IFNAMSIZ];
2507 struct ifaddr *ifa;
2508 struct sockaddr_dl *sdl;
2509
2510 ifr = (struct ifreq *)data;
2511 switch (cmd) {
2512 case SIOCGIFINDEX:
2513 ifr->ifr_index = ifp->if_index;
2514 break;
2515
2516 case SIOCGIFFLAGS:
2517 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2518 ifr->ifr_flags = temp_flags & 0xffff;
2519 ifr->ifr_flagshigh = temp_flags >> 16;
2520 break;
2521
2522 case SIOCGIFCAP:
2523 ifr->ifr_reqcap = ifp->if_capabilities;
2524 ifr->ifr_curcap = ifp->if_capenable;
2525 break;
2526
2527 #ifdef MAC
2528 case SIOCGIFMAC:
2529 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2530 break;
2531 #endif
2532
2533 case SIOCGIFMETRIC:
2534 ifr->ifr_metric = ifp->if_metric;
2535 break;
2536
2537 case SIOCGIFMTU:
2538 ifr->ifr_mtu = ifp->if_mtu;
2539 break;
2540
2541 case SIOCGIFPHYS:
2542 /* XXXGL: did this ever worked? */
2543 ifr->ifr_phys = 0;
2544 break;
2545
2546 case SIOCGIFDESCR:
2547 error = 0;
2548 sx_slock(&ifdescr_sx);
2549 if (ifp->if_description == NULL)
2550 error = ENOMSG;
2551 else {
2552 /* space for terminating nul */
2553 descrlen = strlen(ifp->if_description) + 1;
2554 if (ifr_buffer_get_length(ifr) < descrlen)
2555 ifr_buffer_set_buffer_null(ifr);
2556 else
2557 error = copyout(ifp->if_description,
2558 ifr_buffer_get_buffer(ifr), descrlen);
2559 ifr_buffer_set_length(ifr, descrlen);
2560 }
2561 sx_sunlock(&ifdescr_sx);
2562 break;
2563
2564 case SIOCSIFDESCR:
2565 error = priv_check(td, PRIV_NET_SETIFDESCR);
2566 if (error)
2567 return (error);
2568
2569 /*
2570 * Copy only (length-1) bytes to make sure that
2571 * if_description is always nul terminated. The
2572 * length parameter is supposed to count the
2573 * terminating nul in.
2574 */
2575 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2576 return (ENAMETOOLONG);
2577 else if (ifr_buffer_get_length(ifr) == 0)
2578 descrbuf = NULL;
2579 else {
2580 descrbuf = malloc(ifr_buffer_get_length(ifr),
2581 M_IFDESCR, M_WAITOK | M_ZERO);
2582 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2583 ifr_buffer_get_length(ifr) - 1);
2584 if (error) {
2585 free(descrbuf, M_IFDESCR);
2586 break;
2587 }
2588 }
2589
2590 sx_xlock(&ifdescr_sx);
2591 odescrbuf = ifp->if_description;
2592 ifp->if_description = descrbuf;
2593 sx_xunlock(&ifdescr_sx);
2594
2595 getmicrotime(&ifp->if_lastchange);
2596 free(odescrbuf, M_IFDESCR);
2597 break;
2598
2599 case SIOCGIFFIB:
2600 ifr->ifr_fib = ifp->if_fib;
2601 break;
2602
2603 case SIOCSIFFIB:
2604 error = priv_check(td, PRIV_NET_SETIFFIB);
2605 if (error)
2606 return (error);
2607 if (ifr->ifr_fib >= rt_numfibs)
2608 return (EINVAL);
2609
2610 ifp->if_fib = ifr->ifr_fib;
2611 break;
2612
2613 case SIOCSIFFLAGS:
2614 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2615 if (error)
2616 return (error);
2617 /*
2618 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2619 * check, so we don't need special handling here yet.
2620 */
2621 new_flags = (ifr->ifr_flags & 0xffff) |
2622 (ifr->ifr_flagshigh << 16);
2623 if (ifp->if_flags & IFF_UP &&
2624 (new_flags & IFF_UP) == 0) {
2625 if_down(ifp);
2626 } else if (new_flags & IFF_UP &&
2627 (ifp->if_flags & IFF_UP) == 0) {
2628 do_ifup = 1;
2629 }
2630 /* See if permanently promiscuous mode bit is about to flip */
2631 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2632 if (new_flags & IFF_PPROMISC)
2633 ifp->if_flags |= IFF_PROMISC;
2634 else if (ifp->if_pcount == 0)
2635 ifp->if_flags &= ~IFF_PROMISC;
2636 if (log_promisc_mode_change)
2637 log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2638 ifp->if_xname,
2639 ((new_flags & IFF_PPROMISC) ?
2640 "enabled" : "disabled"));
2641 }
2642 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2643 (new_flags &~ IFF_CANTCHANGE);
2644 if (ifp->if_ioctl) {
2645 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2646 }
2647 if (do_ifup)
2648 if_up(ifp);
2649 getmicrotime(&ifp->if_lastchange);
2650 break;
2651
2652 case SIOCSIFCAP:
2653 error = priv_check(td, PRIV_NET_SETIFCAP);
2654 if (error)
2655 return (error);
2656 if (ifp->if_ioctl == NULL)
2657 return (EOPNOTSUPP);
2658 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2659 return (EINVAL);
2660 error = (*ifp->if_ioctl)(ifp, cmd, data);
2661 if (error == 0)
2662 getmicrotime(&ifp->if_lastchange);
2663 break;
2664
2665 #ifdef MAC
2666 case SIOCSIFMAC:
2667 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2668 break;
2669 #endif
2670
2671 case SIOCSIFNAME:
2672 error = priv_check(td, PRIV_NET_SETIFNAME);
2673 if (error)
2674 return (error);
2675 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2676 NULL);
2677 if (error != 0)
2678 return (error);
2679 if (new_name[0] == '\0')
2680 return (EINVAL);
2681 if (new_name[IFNAMSIZ-1] != '\0') {
2682 new_name[IFNAMSIZ-1] = '\0';
2683 if (strlen(new_name) == IFNAMSIZ-1)
2684 return (EINVAL);
2685 }
2686 if (strcmp(new_name, ifp->if_xname) == 0)
2687 break;
2688 if (ifunit(new_name) != NULL)
2689 return (EEXIST);
2690
2691 /*
2692 * XXX: Locking. Nothing else seems to lock if_flags,
2693 * and there are numerous other races with the
2694 * ifunit() checks not being atomic with namespace
2695 * changes (renames, vmoves, if_attach, etc).
2696 */
2697 ifp->if_flags |= IFF_RENAMING;
2698
2699 /* Announce the departure of the interface. */
2700 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2701 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2702
2703 log(LOG_INFO, "%s: changing name to '%s'\n",
2704 ifp->if_xname, new_name);
2705
2706 IF_ADDR_WLOCK(ifp);
2707 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2708 ifa = ifp->if_addr;
2709 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2710 namelen = strlen(new_name);
2711 onamelen = sdl->sdl_nlen;
2712 /*
2713 * Move the address if needed. This is safe because we
2714 * allocate space for a name of length IFNAMSIZ when we
2715 * create this in if_attach().
2716 */
2717 if (namelen != onamelen) {
2718 bcopy(sdl->sdl_data + onamelen,
2719 sdl->sdl_data + namelen, sdl->sdl_alen);
2720 }
2721 bcopy(new_name, sdl->sdl_data, namelen);
2722 sdl->sdl_nlen = namelen;
2723 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2724 bzero(sdl->sdl_data, onamelen);
2725 while (namelen != 0)
2726 sdl->sdl_data[--namelen] = 0xff;
2727 IF_ADDR_WUNLOCK(ifp);
2728
2729 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2730 /* Announce the return of the interface. */
2731 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2732
2733 ifp->if_flags &= ~IFF_RENAMING;
2734 break;
2735
2736 #ifdef VIMAGE
2737 case SIOCSIFVNET:
2738 error = priv_check(td, PRIV_NET_SETIFVNET);
2739 if (error)
2740 return (error);
2741 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2742 break;
2743 #endif
2744
2745 case SIOCSIFMETRIC:
2746 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2747 if (error)
2748 return (error);
2749 ifp->if_metric = ifr->ifr_metric;
2750 getmicrotime(&ifp->if_lastchange);
2751 break;
2752
2753 case SIOCSIFPHYS:
2754 error = priv_check(td, PRIV_NET_SETIFPHYS);
2755 if (error)
2756 return (error);
2757 if (ifp->if_ioctl == NULL)
2758 return (EOPNOTSUPP);
2759 error = (*ifp->if_ioctl)(ifp, cmd, data);
2760 if (error == 0)
2761 getmicrotime(&ifp->if_lastchange);
2762 break;
2763
2764 case SIOCSIFMTU:
2765 {
2766 u_long oldmtu = ifp->if_mtu;
2767
2768 error = priv_check(td, PRIV_NET_SETIFMTU);
2769 if (error)
2770 return (error);
2771 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2772 return (EINVAL);
2773 if (ifp->if_ioctl == NULL)
2774 return (EOPNOTSUPP);
2775 error = (*ifp->if_ioctl)(ifp, cmd, data);
2776 if (error == 0) {
2777 getmicrotime(&ifp->if_lastchange);
2778 rt_ifmsg(ifp);
2779 }
2780 /*
2781 * If the link MTU changed, do network layer specific procedure.
2782 */
2783 if (ifp->if_mtu != oldmtu) {
2784 #ifdef INET6
2785 nd6_setmtu(ifp);
2786 #endif
2787 rt_updatemtu(ifp);
2788 }
2789 break;
2790 }
2791
2792 case SIOCADDMULTI:
2793 case SIOCDELMULTI:
2794 if (cmd == SIOCADDMULTI)
2795 error = priv_check(td, PRIV_NET_ADDMULTI);
2796 else
2797 error = priv_check(td, PRIV_NET_DELMULTI);
2798 if (error)
2799 return (error);
2800
2801 /* Don't allow group membership on non-multicast interfaces. */
2802 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2803 return (EOPNOTSUPP);
2804
2805 /* Don't let users screw up protocols' entries. */
2806 if (ifr->ifr_addr.sa_family != AF_LINK)
2807 return (EINVAL);
2808
2809 if (cmd == SIOCADDMULTI) {
2810 struct ifmultiaddr *ifma;
2811
2812 /*
2813 * Userland is only permitted to join groups once
2814 * via the if_addmulti() KPI, because it cannot hold
2815 * struct ifmultiaddr * between calls. It may also
2816 * lose a race while we check if the membership
2817 * already exists.
2818 */
2819 IF_ADDR_RLOCK(ifp);
2820 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2821 IF_ADDR_RUNLOCK(ifp);
2822 if (ifma != NULL)
2823 error = EADDRINUSE;
2824 else
2825 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2826 } else {
2827 error = if_delmulti(ifp, &ifr->ifr_addr);
2828 }
2829 if (error == 0)
2830 getmicrotime(&ifp->if_lastchange);
2831 break;
2832
2833 case SIOCSIFPHYADDR:
2834 case SIOCDIFPHYADDR:
2835 #ifdef INET6
2836 case SIOCSIFPHYADDR_IN6:
2837 #endif
2838 case SIOCSIFMEDIA:
2839 case SIOCSIFGENERIC:
2840 error = priv_check(td, PRIV_NET_HWIOCTL);
2841 if (error)
2842 return (error);
2843 if (ifp->if_ioctl == NULL)
2844 return (EOPNOTSUPP);
2845 error = (*ifp->if_ioctl)(ifp, cmd, data);
2846 if (error == 0)
2847 getmicrotime(&ifp->if_lastchange);
2848 break;
2849
2850 case SIOCGIFSTATUS:
2851 case SIOCGIFPSRCADDR:
2852 case SIOCGIFPDSTADDR:
2853 case SIOCGIFMEDIA:
2854 case SIOCGIFXMEDIA:
2855 case SIOCGIFGENERIC:
2856 case SIOCGIFRSSKEY:
2857 case SIOCGIFRSSHASH:
2858 case SIOCGIFDOWNREASON:
2859 if (ifp->if_ioctl == NULL)
2860 return (EOPNOTSUPP);
2861 error = (*ifp->if_ioctl)(ifp, cmd, data);
2862 break;
2863
2864 case SIOCSIFLLADDR:
2865 error = priv_check(td, PRIV_NET_SETLLADDR);
2866 if (error)
2867 return (error);
2868 error = if_setlladdr(ifp,
2869 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2870 break;
2871
2872 case SIOCGHWADDR:
2873 error = if_gethwaddr(ifp, ifr);
2874 break;
2875
2876 CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2877 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2878 if (error)
2879 return (error);
2880 if ((error = if_addgroup(ifp,
2881 ifgr_group_get((struct ifgroupreq *)data))))
2882 return (error);
2883 break;
2884
2885 CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2886 if ((error = if_getgroup((struct ifgroupreq *)data, ifp)))
2887 return (error);
2888 break;
2889
2890 CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2891 error = priv_check(td, PRIV_NET_DELIFGROUP);
2892 if (error)
2893 return (error);
2894 if ((error = if_delgroup(ifp,
2895 ifgr_group_get((struct ifgroupreq *)data))))
2896 return (error);
2897 break;
2898
2899 default:
2900 error = ENOIOCTL;
2901 break;
2902 }
2903 return (error);
2904 }
2905
2906 /* COMPAT_SVR4 */
2907 #define OSIOCGIFCONF _IOWR('i', 20, struct ifconf)
2908
2909 #ifdef COMPAT_FREEBSD32
2910 struct ifconf32 {
2911 int32_t ifc_len;
2912 union {
2913 uint32_t ifcu_buf;
2914 uint32_t ifcu_req;
2915 } ifc_ifcu;
2916 };
2917 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2918 #endif
2919
2920 #ifdef COMPAT_FREEBSD32
2921 static void
2922 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2923 {
2924 struct ifmediareq32 *ifmr32;
2925
2926 ifmr32 = (struct ifmediareq32 *)data;
2927 memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2928 sizeof(ifmr->ifm_name));
2929 ifmr->ifm_current = ifmr32->ifm_current;
2930 ifmr->ifm_mask = ifmr32->ifm_mask;
2931 ifmr->ifm_status = ifmr32->ifm_status;
2932 ifmr->ifm_active = ifmr32->ifm_active;
2933 ifmr->ifm_count = ifmr32->ifm_count;
2934 ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2935 }
2936
2937 static void
2938 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2939 {
2940 struct ifmediareq32 *ifmr32;
2941
2942 ifmr32 = (struct ifmediareq32 *)data;
2943 ifmr32->ifm_current = ifmr->ifm_current;
2944 ifmr32->ifm_mask = ifmr->ifm_mask;
2945 ifmr32->ifm_status = ifmr->ifm_status;
2946 ifmr32->ifm_active = ifmr->ifm_active;
2947 ifmr32->ifm_count = ifmr->ifm_count;
2948 }
2949 #endif
2950
2951 /*
2952 * Interface ioctls.
2953 */
2954 int
2955 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2956 {
2957 #ifdef COMPAT_FREEBSD32
2958 caddr_t saved_data;
2959 struct ifmediareq ifmr;
2960 #endif
2961 struct ifmediareq *ifmrp;
2962 struct ifnet *ifp;
2963 struct ifreq *ifr;
2964 int error;
2965 int oif_flags;
2966 #ifdef VIMAGE
2967 int shutdown;
2968 #endif
2969
2970 CURVNET_SET(so->so_vnet);
2971 #ifdef VIMAGE
2972 /* Make sure the VNET is stable. */
2973 shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2974 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2975 if (shutdown) {
2976 CURVNET_RESTORE();
2977 return (EBUSY);
2978 }
2979 #endif
2980
2981
2982 switch (cmd) {
2983 case SIOCGIFCONF:
2984 case OSIOCGIFCONF: /* COMPAT_SVR4 */
2985 error = ifconf(cmd, data);
2986 CURVNET_RESTORE();
2987 return (error);
2988
2989 #ifdef COMPAT_FREEBSD32
2990 case SIOCGIFCONF32:
2991 {
2992 struct ifconf32 *ifc32;
2993 struct ifconf ifc;
2994
2995 ifc32 = (struct ifconf32 *)data;
2996 ifc.ifc_len = ifc32->ifc_len;
2997 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2998
2999 error = ifconf(SIOCGIFCONF, (void *)&ifc);
3000 CURVNET_RESTORE();
3001 if (error == 0)
3002 ifc32->ifc_len = ifc.ifc_len;
3003 return (error);
3004 }
3005 #endif
3006 }
3007
3008 ifmrp = NULL;
3009 #ifdef COMPAT_FREEBSD32
3010 switch (cmd) {
3011 case SIOCGIFMEDIA32:
3012 case SIOCGIFXMEDIA32:
3013 ifmrp = &ifmr;
3014 ifmr_init(ifmrp, data);
3015 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3016 saved_data = data;
3017 data = (caddr_t)ifmrp;
3018 }
3019 #endif
3020
3021 ifr = (struct ifreq *)data;
3022 switch (cmd) {
3023 #ifdef VIMAGE
3024 case SIOCSIFRVNET:
3025 error = priv_check(td, PRIV_NET_SETIFVNET);
3026 if (error == 0)
3027 error = if_vmove_reclaim(td, ifr->ifr_name,
3028 ifr->ifr_jid);
3029 goto out_noref;
3030 #endif
3031 case SIOCIFCREATE:
3032 case SIOCIFCREATE2:
3033 error = priv_check(td, PRIV_NET_IFCREATE);
3034 if (error == 0)
3035 error = if_clone_create(ifr->ifr_name,
3036 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3037 ifr_data_get_ptr(ifr) : NULL);
3038 goto out_noref;
3039 case SIOCIFDESTROY:
3040 error = priv_check(td, PRIV_NET_IFDESTROY);
3041 if (error == 0)
3042 error = if_clone_destroy(ifr->ifr_name);
3043 goto out_noref;
3044
3045 case SIOCIFGCLONERS:
3046 error = if_clone_list((struct if_clonereq *)data);
3047 goto out_noref;
3048
3049 CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3050 error = if_getgroupmembers((struct ifgroupreq *)data);
3051 goto out_noref;
3052
3053 #if defined(INET) || defined(INET6)
3054 case SIOCSVH:
3055 case SIOCGVH:
3056 if (carp_ioctl_p == NULL)
3057 error = EPROTONOSUPPORT;
3058 else
3059 error = (*carp_ioctl_p)(ifr, cmd, td);
3060 goto out_noref;
3061 #endif
3062 }
3063
3064 ifp = ifunit_ref(ifr->ifr_name);
3065 if (ifp == NULL) {
3066 error = ENXIO;
3067 goto out_noref;
3068 }
3069
3070 error = ifhwioctl(cmd, ifp, data, td);
3071 if (error != ENOIOCTL)
3072 goto out_ref;
3073
3074 oif_flags = ifp->if_flags;
3075 if (so->so_proto == NULL) {
3076 error = EOPNOTSUPP;
3077 goto out_ref;
3078 }
3079
3080 /*
3081 * Pass the request on to the socket control method, and if the
3082 * latter returns EOPNOTSUPP, directly to the interface.
3083 *
3084 * Make an exception for the legacy SIOCSIF* requests. Drivers
3085 * trust SIOCSIFADDR et al to come from an already privileged
3086 * layer, and do not perform any credentials checks or input
3087 * validation.
3088 */
3089 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3090 ifp, td));
3091 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3092 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3093 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3094 error = (*ifp->if_ioctl)(ifp, cmd, data);
3095
3096 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3097 #ifdef INET6
3098 if (ifp->if_flags & IFF_UP)
3099 in6_if_up(ifp);
3100 #endif
3101 }
3102
3103 out_ref:
3104 if_rele(ifp);
3105 out_noref:
3106 #ifdef COMPAT_FREEBSD32
3107 if (ifmrp != NULL) {
3108 KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3109 ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3110 cmd));
3111 data = saved_data;
3112 ifmr_update(ifmrp, data);
3113 }
3114 #endif
3115 CURVNET_RESTORE();
3116 return (error);
3117 }
3118
3119 /*
3120 * The code common to handling reference counted flags,
3121 * e.g., in ifpromisc() and if_allmulti().
3122 * The "pflag" argument can specify a permanent mode flag to check,
3123 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3124 *
3125 * Only to be used on stack-owned flags, not driver-owned flags.
3126 */
3127 static int
3128 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3129 {
3130 struct ifreq ifr;
3131 int error;
3132 int oldflags, oldcount;
3133
3134 /* Sanity checks to catch programming errors */
3135 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3136 ("%s: setting driver-owned flag %d", __func__, flag));
3137
3138 if (onswitch)
3139 KASSERT(*refcount >= 0,
3140 ("%s: increment negative refcount %d for flag %d",
3141 __func__, *refcount, flag));
3142 else
3143 KASSERT(*refcount > 0,
3144 ("%s: decrement non-positive refcount %d for flag %d",
3145 __func__, *refcount, flag));
3146
3147 /* In case this mode is permanent, just touch refcount */
3148 if (ifp->if_flags & pflag) {
3149 *refcount += onswitch ? 1 : -1;
3150 return (0);
3151 }
3152
3153 /* Save ifnet parameters for if_ioctl() may fail */
3154 oldcount = *refcount;
3155 oldflags = ifp->if_flags;
3156
3157 /*
3158 * See if we aren't the only and touching refcount is enough.
3159 * Actually toggle interface flag if we are the first or last.
3160 */
3161 if (onswitch) {
3162 if ((*refcount)++)
3163 return (0);
3164 ifp->if_flags |= flag;
3165 } else {
3166 if (--(*refcount))
3167 return (0);
3168 ifp->if_flags &= ~flag;
3169 }
3170
3171 /* Call down the driver since we've changed interface flags */
3172 if (ifp->if_ioctl == NULL) {
3173 error = EOPNOTSUPP;
3174 goto recover;
3175 }
3176 ifr.ifr_flags = ifp->if_flags & 0xffff;
3177 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3178 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3179 if (error)
3180 goto recover;
3181 /* Notify userland that interface flags have changed */
3182 rt_ifmsg(ifp);
3183 return (0);
3184
3185 recover:
3186 /* Recover after driver error */
3187 *refcount = oldcount;
3188 ifp->if_flags = oldflags;
3189 return (error);
3190 }
3191
3192 /*
3193 * Set/clear promiscuous mode on interface ifp based on the truth value
3194 * of pswitch. The calls are reference counted so that only the first
3195 * "on" request actually has an effect, as does the final "off" request.
3196 * Results are undefined if the "off" and "on" requests are not matched.
3197 */
3198 int
3199 ifpromisc(struct ifnet *ifp, int pswitch)
3200 {
3201 int error;
3202 int oldflags = ifp->if_flags;
3203
3204 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3205 &ifp->if_pcount, pswitch);
3206 /* If promiscuous mode status has changed, log a message */
3207 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3208 log_promisc_mode_change)
3209 log(LOG_INFO, "%s: promiscuous mode %s\n",
3210 ifp->if_xname,
3211 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3212 return (error);
3213 }
3214
3215 /*
3216 * Return interface configuration
3217 * of system. List may be used
3218 * in later ioctl's (above) to get
3219 * other information.
3220 */
3221 /*ARGSUSED*/
3222 static int
3223 ifconf(u_long cmd, caddr_t data)
3224 {
3225 struct ifconf *ifc = (struct ifconf *)data;
3226 struct ifnet *ifp;
3227 struct ifaddr *ifa;
3228 struct ifreq ifr;
3229 struct sbuf *sb;
3230 int error, full = 0, valid_len, max_len;
3231
3232 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3233 max_len = MAXPHYS - 1;
3234
3235 /* Prevent hostile input from being able to crash the system */
3236 if (ifc->ifc_len <= 0)
3237 return (EINVAL);
3238
3239 again:
3240 if (ifc->ifc_len <= max_len) {
3241 max_len = ifc->ifc_len;
3242 full = 1;
3243 }
3244 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3245 max_len = 0;
3246 valid_len = 0;
3247
3248 IFNET_RLOCK();
3249 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3250 int addrs;
3251
3252 /*
3253 * Zero the ifr to make sure we don't disclose the contents
3254 * of the stack.
3255 */
3256 memset(&ifr, 0, sizeof(ifr));
3257
3258 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3259 >= sizeof(ifr.ifr_name)) {
3260 sbuf_delete(sb);
3261 IFNET_RUNLOCK();
3262 return (ENAMETOOLONG);
3263 }
3264
3265 addrs = 0;
3266 IF_ADDR_RLOCK(ifp);
3267 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3268 struct sockaddr *sa = ifa->ifa_addr;
3269
3270 if (prison_if(curthread->td_ucred, sa) != 0)
3271 continue;
3272 addrs++;
3273 /* COMPAT_SVR4 */
3274 if (cmd == OSIOCGIFCONF) {
3275 struct osockaddr *osa =
3276 (struct osockaddr *)&ifr.ifr_addr;
3277 ifr.ifr_addr = *sa;
3278 osa->sa_family = sa->sa_family;
3279 sbuf_bcat(sb, &ifr, sizeof(ifr));
3280 max_len += sizeof(ifr);
3281 } else
3282 if (sa->sa_len <= sizeof(*sa)) {
3283 if (sa->sa_len < sizeof(*sa)) {
3284 memset(&ifr.ifr_ifru.ifru_addr, 0,
3285 sizeof(ifr.ifr_ifru.ifru_addr));
3286 memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3287 sa->sa_len);
3288 } else
3289 ifr.ifr_ifru.ifru_addr = *sa;
3290 sbuf_bcat(sb, &ifr, sizeof(ifr));
3291 max_len += sizeof(ifr);
3292 } else {
3293 sbuf_bcat(sb, &ifr,
3294 offsetof(struct ifreq, ifr_addr));
3295 max_len += offsetof(struct ifreq, ifr_addr);
3296 sbuf_bcat(sb, sa, sa->sa_len);
3297 max_len += sa->sa_len;
3298 }
3299
3300 if (sbuf_error(sb) == 0)
3301 valid_len = sbuf_len(sb);
3302 }
3303 IF_ADDR_RUNLOCK(ifp);
3304 if (addrs == 0) {
3305 sbuf_bcat(sb, &ifr, sizeof(ifr));
3306 max_len += sizeof(ifr);
3307
3308 if (sbuf_error(sb) == 0)
3309 valid_len = sbuf_len(sb);
3310 }
3311 }
3312 IFNET_RUNLOCK();
3313
3314 /*
3315 * If we didn't allocate enough space (uncommon), try again. If
3316 * we have already allocated as much space as we are allowed,
3317 * return what we've got.
3318 */
3319 if (valid_len != max_len && !full) {
3320 sbuf_delete(sb);
3321 goto again;
3322 }
3323
3324 ifc->ifc_len = valid_len;
3325 sbuf_finish(sb);
3326 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3327 sbuf_delete(sb);
3328 return (error);
3329 }
3330
3331 /*
3332 * Just like ifpromisc(), but for all-multicast-reception mode.
3333 */
3334 int
3335 if_allmulti(struct ifnet *ifp, int onswitch)
3336 {
3337
3338 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3339 }
3340
3341 struct ifmultiaddr *
3342 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3343 {
3344 struct ifmultiaddr *ifma;
3345
3346 IF_ADDR_LOCK_ASSERT(ifp);
3347
3348 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3349 if (sa->sa_family == AF_LINK) {
3350 if (sa_dl_equal(ifma->ifma_addr, sa))
3351 break;
3352 } else {
3353 if (sa_equal(ifma->ifma_addr, sa))
3354 break;
3355 }
3356 }
3357
3358 return ifma;
3359 }
3360
3361 /*
3362 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
3363 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
3364 * the ifnet multicast address list here, so the caller must do that and
3365 * other setup work (such as notifying the device driver). The reference
3366 * count is initialized to 1.
3367 */
3368 static struct ifmultiaddr *
3369 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3370 int mflags)
3371 {
3372 struct ifmultiaddr *ifma;
3373 struct sockaddr *dupsa;
3374
3375 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3376 M_ZERO);
3377 if (ifma == NULL)
3378 return (NULL);
3379
3380 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3381 if (dupsa == NULL) {
3382 free(ifma, M_IFMADDR);
3383 return (NULL);
3384 }
3385 bcopy(sa, dupsa, sa->sa_len);
3386 ifma->ifma_addr = dupsa;
3387
3388 ifma->ifma_ifp = ifp;
3389 ifma->ifma_refcount = 1;
3390 ifma->ifma_protospec = NULL;
3391
3392 if (llsa == NULL) {
3393 ifma->ifma_lladdr = NULL;
3394 return (ifma);
3395 }
3396
3397 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3398 if (dupsa == NULL) {
3399 free(ifma->ifma_addr, M_IFMADDR);
3400 free(ifma, M_IFMADDR);
3401 return (NULL);
3402 }
3403 bcopy(llsa, dupsa, llsa->sa_len);
3404 ifma->ifma_lladdr = dupsa;
3405
3406 return (ifma);
3407 }
3408
3409 /*
3410 * if_freemulti: free ifmultiaddr structure and possibly attached related
3411 * addresses. The caller is responsible for implementing reference
3412 * counting, notifying the driver, handling routing messages, and releasing
3413 * any dependent link layer state.
3414 */
3415 static void
3416 if_freemulti(struct ifmultiaddr *ifma)
3417 {
3418
3419 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3420 ifma->ifma_refcount));
3421
3422 if (ifma->ifma_lladdr != NULL)
3423 free(ifma->ifma_lladdr, M_IFMADDR);
3424 free(ifma->ifma_addr, M_IFMADDR);
3425 free(ifma, M_IFMADDR);
3426 }
3427
3428 /*
3429 * Register an additional multicast address with a network interface.
3430 *
3431 * - If the address is already present, bump the reference count on the
3432 * address and return.
3433 * - If the address is not link-layer, look up a link layer address.
3434 * - Allocate address structures for one or both addresses, and attach to the
3435 * multicast address list on the interface. If automatically adding a link
3436 * layer address, the protocol address will own a reference to the link
3437 * layer address, to be freed when it is freed.
3438 * - Notify the network device driver of an addition to the multicast address
3439 * list.
3440 *
3441 * 'sa' points to caller-owned memory with the desired multicast address.
3442 *
3443 * 'retifma' will be used to return a pointer to the resulting multicast
3444 * address reference, if desired.
3445 */
3446 int
3447 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3448 struct ifmultiaddr **retifma)
3449 {
3450 struct ifmultiaddr *ifma, *ll_ifma;
3451 struct sockaddr *llsa;
3452 struct sockaddr_dl sdl;
3453 int error;
3454
3455 /*
3456 * If the address is already present, return a new reference to it;
3457 * otherwise, allocate storage and set up a new address.
3458 */
3459 IF_ADDR_WLOCK(ifp);
3460 ifma = if_findmulti(ifp, sa);
3461 if (ifma != NULL) {
3462 ifma->ifma_refcount++;
3463 if (retifma != NULL)
3464 *retifma = ifma;
3465 IF_ADDR_WUNLOCK(ifp);
3466 return (0);
3467 }
3468
3469 /*
3470 * The address isn't already present; resolve the protocol address
3471 * into a link layer address, and then look that up, bump its
3472 * refcount or allocate an ifma for that also.
3473 * Most link layer resolving functions returns address data which
3474 * fits inside default sockaddr_dl structure. However callback
3475 * can allocate another sockaddr structure, in that case we need to
3476 * free it later.
3477 */
3478 llsa = NULL;
3479 ll_ifma = NULL;
3480 if (ifp->if_resolvemulti != NULL) {
3481 /* Provide called function with buffer size information */
3482 sdl.sdl_len = sizeof(sdl);
3483 llsa = (struct sockaddr *)&sdl;
3484 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3485 if (error)
3486 goto unlock_out;
3487 }
3488
3489 /*
3490 * Allocate the new address. Don't hook it up yet, as we may also
3491 * need to allocate a link layer multicast address.
3492 */
3493 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3494 if (ifma == NULL) {
3495 error = ENOMEM;
3496 goto free_llsa_out;
3497 }
3498
3499 /*
3500 * If a link layer address is found, we'll need to see if it's
3501 * already present in the address list, or allocate is as well.
3502 * When this block finishes, the link layer address will be on the
3503 * list.
3504 */
3505 if (llsa != NULL) {
3506 ll_ifma = if_findmulti(ifp, llsa);
3507 if (ll_ifma == NULL) {
3508 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3509 if (ll_ifma == NULL) {
3510 --ifma->ifma_refcount;
3511 if_freemulti(ifma);
3512 error = ENOMEM;
3513 goto free_llsa_out;
3514 }
3515 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3516 ifma_link);
3517 } else
3518 ll_ifma->ifma_refcount++;
3519 ifma->ifma_llifma = ll_ifma;
3520 }
3521
3522 /*
3523 * We now have a new multicast address, ifma, and possibly a new or
3524 * referenced link layer address. Add the primary address to the
3525 * ifnet address list.
3526 */
3527 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3528
3529 if (retifma != NULL)
3530 *retifma = ifma;
3531
3532 /*
3533 * Must generate the message while holding the lock so that 'ifma'
3534 * pointer is still valid.
3535 */
3536 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3537 IF_ADDR_WUNLOCK(ifp);
3538
3539 /*
3540 * We are certain we have added something, so call down to the
3541 * interface to let them know about it.
3542 */
3543 if (ifp->if_ioctl != NULL) {
3544 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3545 }
3546
3547 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3548 link_free_sdl(llsa);
3549
3550 return (0);
3551
3552 free_llsa_out:
3553 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3554 link_free_sdl(llsa);
3555
3556 unlock_out:
3557 IF_ADDR_WUNLOCK(ifp);
3558 return (error);
3559 }
3560
3561 /*
3562 * Delete a multicast group membership by network-layer group address.
3563 *
3564 * Returns ENOENT if the entry could not be found. If ifp no longer
3565 * exists, results are undefined. This entry point should only be used
3566 * from subsystems which do appropriate locking to hold ifp for the
3567 * duration of the call.
3568 * Network-layer protocol domains must use if_delmulti_ifma().
3569 */
3570 int
3571 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3572 {
3573 struct ifmultiaddr *ifma;
3574 int lastref;
3575 #ifdef INVARIANTS
3576 struct ifnet *oifp;
3577
3578 IFNET_RLOCK_NOSLEEP();
3579 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3580 if (ifp == oifp)
3581 break;
3582 if (ifp != oifp)
3583 ifp = NULL;
3584 IFNET_RUNLOCK_NOSLEEP();
3585
3586 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3587 #endif
3588 if (ifp == NULL)
3589 return (ENOENT);
3590
3591 IF_ADDR_WLOCK(ifp);
3592 lastref = 0;
3593 ifma = if_findmulti(ifp, sa);
3594 if (ifma != NULL)
3595 lastref = if_delmulti_locked(ifp, ifma, 0);
3596 IF_ADDR_WUNLOCK(ifp);
3597
3598 if (ifma == NULL)
3599 return (ENOENT);
3600
3601 if (lastref && ifp->if_ioctl != NULL) {
3602 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3603 }
3604
3605 return (0);
3606 }
3607
3608 /*
3609 * Delete all multicast group membership for an interface.
3610 * Should be used to quickly flush all multicast filters.
3611 */
3612 void
3613 if_delallmulti(struct ifnet *ifp)
3614 {
3615 struct ifmultiaddr *ifma;
3616 struct ifmultiaddr *next;
3617
3618 IF_ADDR_WLOCK(ifp);
3619 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3620 if_delmulti_locked(ifp, ifma, 0);
3621 IF_ADDR_WUNLOCK(ifp);
3622 }
3623
3624 /*
3625 * Delete a multicast group membership by group membership pointer.
3626 * Network-layer protocol domains must use this routine.
3627 *
3628 * It is safe to call this routine if the ifp disappeared.
3629 */
3630 void
3631 if_delmulti_ifma(struct ifmultiaddr *ifma)
3632 {
3633 struct ifnet *ifp;
3634 int lastref;
3635
3636 ifp = ifma->ifma_ifp;
3637 #ifdef DIAGNOSTIC
3638 if (ifp == NULL) {
3639 printf("%s: ifma_ifp seems to be detached\n", __func__);
3640 } else {
3641 struct ifnet *oifp;
3642
3643 IFNET_RLOCK_NOSLEEP();
3644 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3645 if (ifp == oifp)
3646 break;
3647 if (ifp != oifp) {
3648 printf("%s: ifnet %p disappeared\n", __func__, ifp);
3649 ifp = NULL;
3650 }
3651 IFNET_RUNLOCK_NOSLEEP();
3652 }
3653 #endif
3654 /*
3655 * If and only if the ifnet instance exists: Acquire the address lock.
3656 */
3657 if (ifp != NULL)
3658 IF_ADDR_WLOCK(ifp);
3659
3660 lastref = if_delmulti_locked(ifp, ifma, 0);
3661
3662 if (ifp != NULL) {
3663 /*
3664 * If and only if the ifnet instance exists:
3665 * Release the address lock.
3666 * If the group was left: update the hardware hash filter.
3667 */
3668 IF_ADDR_WUNLOCK(ifp);
3669 if (lastref && ifp->if_ioctl != NULL) {
3670 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3671 }
3672 }
3673 }
3674
3675 /*
3676 * Perform deletion of network-layer and/or link-layer multicast address.
3677 *
3678 * Return 0 if the reference count was decremented.
3679 * Return 1 if the final reference was released, indicating that the
3680 * hardware hash filter should be reprogrammed.
3681 */
3682 static int
3683 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3684 {
3685 struct ifmultiaddr *ll_ifma;
3686
3687 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3688 KASSERT(ifma->ifma_ifp == ifp,
3689 ("%s: inconsistent ifp %p", __func__, ifp));
3690 IF_ADDR_WLOCK_ASSERT(ifp);
3691 }
3692
3693 ifp = ifma->ifma_ifp;
3694
3695 /*
3696 * If the ifnet is detaching, null out references to ifnet,
3697 * so that upper protocol layers will notice, and not attempt
3698 * to obtain locks for an ifnet which no longer exists. The
3699 * routing socket announcement must happen before the ifnet
3700 * instance is detached from the system.
3701 */
3702 if (detaching) {
3703 #ifdef DIAGNOSTIC
3704 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3705 #endif
3706 /*
3707 * ifp may already be nulled out if we are being reentered
3708 * to delete the ll_ifma.
3709 */
3710 if (ifp != NULL) {
3711 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3712 ifma->ifma_ifp = NULL;
3713 }
3714 }
3715
3716 if (--ifma->ifma_refcount > 0)
3717 return 0;
3718
3719 /*
3720 * If this ifma is a network-layer ifma, a link-layer ifma may
3721 * have been associated with it. Release it first if so.
3722 */
3723 ll_ifma = ifma->ifma_llifma;
3724 if (ll_ifma != NULL) {
3725 KASSERT(ifma->ifma_lladdr != NULL,
3726 ("%s: llifma w/o lladdr", __func__));
3727 if (detaching)
3728 ll_ifma->ifma_ifp = NULL; /* XXX */
3729 if (--ll_ifma->ifma_refcount == 0) {
3730 if (ifp != NULL) {
3731 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3732 ifma_link);
3733 }
3734 if_freemulti(ll_ifma);
3735 }
3736 }
3737
3738 if (ifp != NULL)
3739 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3740
3741 if_freemulti(ifma);
3742
3743 /*
3744 * The last reference to this instance of struct ifmultiaddr
3745 * was released; the hardware should be notified of this change.
3746 */
3747 return 1;
3748 }
3749
3750 /*
3751 * Set the link layer address on an interface.
3752 *
3753 * At this time we only support certain types of interfaces,
3754 * and we don't allow the length of the address to change.
3755 *
3756 * Set noinline to be dtrace-friendly
3757 */
3758 __noinline int
3759 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3760 {
3761 struct sockaddr_dl *sdl;
3762 struct ifaddr *ifa;
3763 struct ifreq ifr;
3764
3765 IF_ADDR_RLOCK(ifp);
3766 ifa = ifp->if_addr;
3767 if (ifa == NULL) {
3768 IF_ADDR_RUNLOCK(ifp);
3769 return (EINVAL);
3770 }
3771 ifa_ref(ifa);
3772 IF_ADDR_RUNLOCK(ifp);
3773 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3774 if (sdl == NULL) {
3775 ifa_free(ifa);
3776 return (EINVAL);
3777 }
3778 if (len != sdl->sdl_alen) { /* don't allow length to change */
3779 ifa_free(ifa);
3780 return (EINVAL);
3781 }
3782 switch (ifp->if_type) {
3783 case IFT_ETHER:
3784 case IFT_FDDI:
3785 case IFT_XETHER:
3786 case IFT_ISO88025:
3787 case IFT_L2VLAN:
3788 case IFT_BRIDGE:
3789 case IFT_ARCNET:
3790 case IFT_IEEE8023ADLAG:
3791 case IFT_IEEE80211:
3792 bcopy(lladdr, LLADDR(sdl), len);
3793 ifa_free(ifa);
3794 break;
3795 default:
3796 ifa_free(ifa);
3797 return (ENODEV);
3798 }
3799
3800 /*
3801 * If the interface is already up, we need
3802 * to re-init it in order to reprogram its
3803 * address filter.
3804 */
3805 if ((ifp->if_flags & IFF_UP) != 0) {
3806 if (ifp->if_ioctl) {
3807 ifp->if_flags &= ~IFF_UP;
3808 ifr.ifr_flags = ifp->if_flags & 0xffff;
3809 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3810 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3811 ifp->if_flags |= IFF_UP;
3812 ifr.ifr_flags = ifp->if_flags & 0xffff;
3813 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3814 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3815 }
3816 }
3817 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3818 return (0);
3819 }
3820
3821 /*
3822 * Compat function for handling basic encapsulation requests.
3823 * Not converted stacks (FDDI, IB, ..) supports traditional
3824 * output model: ARP (and other similar L2 protocols) are handled
3825 * inside output routine, arpresolve/nd6_resolve() returns MAC
3826 * address instead of full prepend.
3827 *
3828 * This function creates calculated header==MAC for IPv4/IPv6 and
3829 * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3830 * address families.
3831 */
3832 static int
3833 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3834 {
3835
3836 if (req->rtype != IFENCAP_LL)
3837 return (EOPNOTSUPP);
3838
3839 if (req->bufsize < req->lladdr_len)
3840 return (ENOMEM);
3841
3842 switch (req->family) {
3843 case AF_INET:
3844 case AF_INET6:
3845 break;
3846 default:
3847 return (EAFNOSUPPORT);
3848 }
3849
3850 /* Copy lladdr to storage as is */
3851 memmove(req->buf, req->lladdr, req->lladdr_len);
3852 req->bufsize = req->lladdr_len;
3853 req->lladdr_off = 0;
3854
3855 return (0);
3856 }
3857
3858 /*
3859 * Get the link layer address that was read from the hardware at attach.
3860 *
3861 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3862 * their component interfaces as IFT_IEEE8023ADLAG.
3863 */
3864 int
3865 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3866 {
3867
3868 if (ifp->if_hw_addr == NULL)
3869 return (ENODEV);
3870
3871 switch (ifp->if_type) {
3872 case IFT_ETHER:
3873 case IFT_IEEE8023ADLAG:
3874 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3875 return (0);
3876 default:
3877 return (ENODEV);
3878 }
3879 }
3880
3881 /*
3882 * The name argument must be a pointer to storage which will last as
3883 * long as the interface does. For physical devices, the result of
3884 * device_get_name(dev) is a good choice and for pseudo-devices a
3885 * static string works well.
3886 */
3887 void
3888 if_initname(struct ifnet *ifp, const char *name, int unit)
3889 {
3890 ifp->if_dname = name;
3891 ifp->if_dunit = unit;
3892 if (unit != IF_DUNIT_NONE)
3893 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3894 else
3895 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3896 }
3897
3898 int
3899 if_printf(struct ifnet *ifp, const char * fmt, ...)
3900 {
3901 va_list ap;
3902 int retval;
3903
3904 retval = printf("%s: ", ifp->if_xname);
3905 va_start(ap, fmt);
3906 retval += vprintf(fmt, ap);
3907 va_end(ap);
3908 return (retval);
3909 }
3910
3911 void
3912 if_start(struct ifnet *ifp)
3913 {
3914
3915 (*(ifp)->if_start)(ifp);
3916 }
3917
3918 /*
3919 * Backwards compatibility interface for drivers
3920 * that have not implemented it
3921 */
3922 static int
3923 if_transmit(struct ifnet *ifp, struct mbuf *m)
3924 {
3925 int error;
3926
3927 IFQ_HANDOFF(ifp, m, error);
3928 return (error);
3929 }
3930
3931 static void
3932 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
3933 {
3934
3935 m_freem(m);
3936 }
3937
3938 int
3939 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3940 {
3941 int active = 0;
3942
3943 IF_LOCK(ifq);
3944 if (_IF_QFULL(ifq)) {
3945 IF_UNLOCK(ifq);
3946 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
3947 m_freem(m);
3948 return (0);
3949 }
3950 if (ifp != NULL) {
3951 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
3952 if (m->m_flags & (M_BCAST|M_MCAST))
3953 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
3954 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3955 }
3956 _IF_ENQUEUE(ifq, m);
3957 IF_UNLOCK(ifq);
3958 if (ifp != NULL && !active)
3959 (*(ifp)->if_start)(ifp);
3960 return (1);
3961 }
3962
3963 void
3964 if_register_com_alloc(u_char type,
3965 if_com_alloc_t *a, if_com_free_t *f)
3966 {
3967
3968 KASSERT(if_com_alloc[type] == NULL,
3969 ("if_register_com_alloc: %d already registered", type));
3970 KASSERT(if_com_free[type] == NULL,
3971 ("if_register_com_alloc: %d free already registered", type));
3972
3973 if_com_alloc[type] = a;
3974 if_com_free[type] = f;
3975 }
3976
3977 void
3978 if_deregister_com_alloc(u_char type)
3979 {
3980
3981 KASSERT(if_com_alloc[type] != NULL,
3982 ("if_deregister_com_alloc: %d not registered", type));
3983 KASSERT(if_com_free[type] != NULL,
3984 ("if_deregister_com_alloc: %d free not registered", type));
3985 if_com_alloc[type] = NULL;
3986 if_com_free[type] = NULL;
3987 }
3988
3989 /* API for driver access to network stack owned ifnet.*/
3990 uint64_t
3991 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
3992 {
3993 uint64_t oldbrate;
3994
3995 oldbrate = ifp->if_baudrate;
3996 ifp->if_baudrate = baudrate;
3997 return (oldbrate);
3998 }
3999
4000 uint64_t
4001 if_getbaudrate(if_t ifp)
4002 {
4003
4004 return (((struct ifnet *)ifp)->if_baudrate);
4005 }
4006
4007 int
4008 if_setcapabilities(if_t ifp, int capabilities)
4009 {
4010 ((struct ifnet *)ifp)->if_capabilities = capabilities;
4011 return (0);
4012 }
4013
4014 int
4015 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4016 {
4017 ((struct ifnet *)ifp)->if_capabilities |= setbit;
4018 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4019
4020 return (0);
4021 }
4022
4023 int
4024 if_getcapabilities(if_t ifp)
4025 {
4026 return ((struct ifnet *)ifp)->if_capabilities;
4027 }
4028
4029 int
4030 if_setcapenable(if_t ifp, int capabilities)
4031 {
4032 ((struct ifnet *)ifp)->if_capenable = capabilities;
4033 return (0);
4034 }
4035
4036 int
4037 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4038 {
4039 if(setcap)
4040 ((struct ifnet *)ifp)->if_capenable |= setcap;
4041 if(clearcap)
4042 ((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4043
4044 return (0);
4045 }
4046
4047 const char *
4048 if_getdname(if_t ifp)
4049 {
4050 return ((struct ifnet *)ifp)->if_dname;
4051 }
4052
4053 int
4054 if_togglecapenable(if_t ifp, int togglecap)
4055 {
4056 ((struct ifnet *)ifp)->if_capenable ^= togglecap;
4057 return (0);
4058 }
4059
4060 int
4061 if_getcapenable(if_t ifp)
4062 {
4063 return ((struct ifnet *)ifp)->if_capenable;
4064 }
4065
4066 /*
4067 * This is largely undesirable because it ties ifnet to a device, but does
4068 * provide flexiblity for an embedded product vendor. Should be used with
4069 * the understanding that it violates the interface boundaries, and should be
4070 * a last resort only.
4071 */
4072 int
4073 if_setdev(if_t ifp, void *dev)
4074 {
4075 return (0);
4076 }
4077
4078 int
4079 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4080 {
4081 ((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4082 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4083
4084 return (0);
4085 }
4086
4087 int
4088 if_getdrvflags(if_t ifp)
4089 {
4090 return ((struct ifnet *)ifp)->if_drv_flags;
4091 }
4092
4093 int
4094 if_setdrvflags(if_t ifp, int flags)
4095 {
4096 ((struct ifnet *)ifp)->if_drv_flags = flags;
4097 return (0);
4098 }
4099
4100
4101 int
4102 if_setflags(if_t ifp, int flags)
4103 {
4104 ((struct ifnet *)ifp)->if_flags = flags;
4105 return (0);
4106 }
4107
4108 int
4109 if_setflagbits(if_t ifp, int set, int clear)
4110 {
4111 ((struct ifnet *)ifp)->if_flags |= set;
4112 ((struct ifnet *)ifp)->if_flags &= ~clear;
4113
4114 return (0);
4115 }
4116
4117 int
4118 if_getflags(if_t ifp)
4119 {
4120 return ((struct ifnet *)ifp)->if_flags;
4121 }
4122
4123 int
4124 if_clearhwassist(if_t ifp)
4125 {
4126 ((struct ifnet *)ifp)->if_hwassist = 0;
4127 return (0);
4128 }
4129
4130 int
4131 if_sethwassistbits(if_t ifp, int toset, int toclear)
4132 {
4133 ((struct ifnet *)ifp)->if_hwassist |= toset;
4134 ((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4135
4136 return (0);
4137 }
4138
4139 int
4140 if_sethwassist(if_t ifp, int hwassist_bit)
4141 {
4142 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4143 return (0);
4144 }
4145
4146 int
4147 if_gethwassist(if_t ifp)
4148 {
4149 return ((struct ifnet *)ifp)->if_hwassist;
4150 }
4151
4152 int
4153 if_setmtu(if_t ifp, int mtu)
4154 {
4155 ((struct ifnet *)ifp)->if_mtu = mtu;
4156 return (0);
4157 }
4158
4159 int
4160 if_getmtu(if_t ifp)
4161 {
4162 return ((struct ifnet *)ifp)->if_mtu;
4163 }
4164
4165 int
4166 if_getmtu_family(if_t ifp, int family)
4167 {
4168 struct domain *dp;
4169
4170 for (dp = domains; dp; dp = dp->dom_next) {
4171 if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4172 return (dp->dom_ifmtu((struct ifnet *)ifp));
4173 }
4174
4175 return (((struct ifnet *)ifp)->if_mtu);
4176 }
4177
4178 int
4179 if_setsoftc(if_t ifp, void *softc)
4180 {
4181 ((struct ifnet *)ifp)->if_softc = softc;
4182 return (0);
4183 }
4184
4185 void *
4186 if_getsoftc(if_t ifp)
4187 {
4188 return ((struct ifnet *)ifp)->if_softc;
4189 }
4190
4191 void
4192 if_setrcvif(struct mbuf *m, if_t ifp)
4193 {
4194 m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4195 }
4196
4197 void
4198 if_setvtag(struct mbuf *m, uint16_t tag)
4199 {
4200 m->m_pkthdr.ether_vtag = tag;
4201 }
4202
4203 uint16_t
4204 if_getvtag(struct mbuf *m)
4205 {
4206
4207 return (m->m_pkthdr.ether_vtag);
4208 }
4209
4210 int
4211 if_sendq_empty(if_t ifp)
4212 {
4213 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4214 }
4215
4216 struct ifaddr *
4217 if_getifaddr(if_t ifp)
4218 {
4219 return ((struct ifnet *)ifp)->if_addr;
4220 }
4221
4222 int
4223 if_getamcount(if_t ifp)
4224 {
4225 return ((struct ifnet *)ifp)->if_amcount;
4226 }
4227
4228
4229 int
4230 if_setsendqready(if_t ifp)
4231 {
4232 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4233 return (0);
4234 }
4235
4236 int
4237 if_setsendqlen(if_t ifp, int tx_desc_count)
4238 {
4239 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4240 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4241
4242 return (0);
4243 }
4244
4245 int
4246 if_vlantrunkinuse(if_t ifp)
4247 {
4248 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4249 }
4250
4251 int
4252 if_input(if_t ifp, struct mbuf* sendmp)
4253 {
4254 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4255 return (0);
4256
4257 }
4258
4259 /* XXX */
4260 #ifndef ETH_ADDR_LEN
4261 #define ETH_ADDR_LEN 6
4262 #endif
4263
4264 int
4265 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4266 {
4267 struct ifmultiaddr *ifma;
4268 uint8_t *lmta = (uint8_t *)mta;
4269 int mcnt = 0;
4270
4271 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4272 if (ifma->ifma_addr->sa_family != AF_LINK)
4273 continue;
4274
4275 if (mcnt == max)
4276 break;
4277
4278 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4279 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4280 mcnt++;
4281 }
4282 *cnt = mcnt;
4283
4284 return (0);
4285 }
4286
4287 int
4288 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4289 {
4290 int error;
4291
4292 if_maddr_rlock(ifp);
4293 error = if_setupmultiaddr(ifp, mta, cnt, max);
4294 if_maddr_runlock(ifp);
4295 return (error);
4296 }
4297
4298 int
4299 if_multiaddr_count(if_t ifp, int max)
4300 {
4301 struct ifmultiaddr *ifma;
4302 int count;
4303
4304 count = 0;
4305 if_maddr_rlock(ifp);
4306 TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4307 if (ifma->ifma_addr->sa_family != AF_LINK)
4308 continue;
4309 count++;
4310 if (count == max)
4311 break;
4312 }
4313 if_maddr_runlock(ifp);
4314 return (count);
4315 }
4316
4317 int
4318 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4319 {
4320 struct ifmultiaddr *ifma;
4321 int cnt = 0;
4322
4323 if_maddr_rlock(ifp);
4324 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4325 cnt += filter(arg, ifma, cnt);
4326 if_maddr_runlock(ifp);
4327 return (cnt);
4328 }
4329
4330 struct mbuf *
4331 if_dequeue(if_t ifp)
4332 {
4333 struct mbuf *m;
4334 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4335
4336 return (m);
4337 }
4338
4339 int
4340 if_sendq_prepend(if_t ifp, struct mbuf *m)
4341 {
4342 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4343 return (0);
4344 }
4345
4346 int
4347 if_setifheaderlen(if_t ifp, int len)
4348 {
4349 ((struct ifnet *)ifp)->if_hdrlen = len;
4350 return (0);
4351 }
4352
4353 caddr_t
4354 if_getlladdr(if_t ifp)
4355 {
4356 return (IF_LLADDR((struct ifnet *)ifp));
4357 }
4358
4359 void *
4360 if_gethandle(u_char type)
4361 {
4362 return (if_alloc(type));
4363 }
4364
4365 void
4366 if_bpfmtap(if_t ifh, struct mbuf *m)
4367 {
4368 struct ifnet *ifp = (struct ifnet *)ifh;
4369
4370 BPF_MTAP(ifp, m);
4371 }
4372
4373 void
4374 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4375 {
4376 struct ifnet *ifp = (struct ifnet *)ifh;
4377
4378 ETHER_BPF_MTAP(ifp, m);
4379 }
4380
4381 void
4382 if_vlancap(if_t ifh)
4383 {
4384 struct ifnet *ifp = (struct ifnet *)ifh;
4385 VLAN_CAPABILITIES(ifp);
4386 }
4387
4388 void
4389 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4390 {
4391 ((struct ifnet *)ifp)->if_init = init_fn;
4392 }
4393
4394 void
4395 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4396 {
4397 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4398 }
4399
4400 void
4401 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4402 {
4403 ((struct ifnet *)ifp)->if_start = (void *)start_fn;
4404 }
4405
4406 void
4407 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4408 {
4409 ((struct ifnet *)ifp)->if_transmit = start_fn;
4410 }
4411
4412 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4413 {
4414 ((struct ifnet *)ifp)->if_qflush = flush_fn;
4415
4416 }
4417
4418 void
4419 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4420 {
4421
4422 ifp->if_get_counter = fn;
4423 }
4424
4425 /* Revisit these - These are inline functions originally. */
4426 int
4427 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4428 {
4429 return drbr_inuse(ifh, br);
4430 }
4431
4432 struct mbuf*
4433 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4434 {
4435 return drbr_dequeue(ifh, br);
4436 }
4437
4438 int
4439 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4440 {
4441 return drbr_needs_enqueue(ifh, br);
4442 }
4443
4444 int
4445 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4446 {
4447 return drbr_enqueue(ifh, br, m);
4448
4449 }
Cache object: 9ee8e2d0d656f302198fad8021de4702
|