FreeBSD/Linux Kernel Cross Reference
sys/net/if.c
1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1980, 1986, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * @(#)if.c 8.5 (Berkeley) 1/9/95
32 * $FreeBSD: releng/12.0/sys/net/if.c 339859 2018-10-29 13:17:41Z hselasky $
33 */
34
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/epoch.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/refcount.h>
55 #include <sys/module.h>
56 #include <sys/rwlock.h>
57 #include <sys/sockio.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 #include <sys/sysent.h>
61 #include <sys/taskqueue.h>
62 #include <sys/domain.h>
63 #include <sys/jail.h>
64 #include <sys/priv.h>
65
66 #include <machine/stdarg.h>
67 #include <vm/uma.h>
68
69 #include <net/bpf.h>
70 #include <net/ethernet.h>
71 #include <net/if.h>
72 #include <net/if_arp.h>
73 #include <net/if_clone.h>
74 #include <net/if_dl.h>
75 #include <net/if_types.h>
76 #include <net/if_var.h>
77 #include <net/if_media.h>
78 #include <net/if_vlan_var.h>
79 #include <net/radix.h>
80 #include <net/route.h>
81 #include <net/vnet.h>
82
83 #if defined(INET) || defined(INET6)
84 #include <net/ethernet.h>
85 #include <netinet/in.h>
86 #include <netinet/in_var.h>
87 #include <netinet/ip.h>
88 #include <netinet/ip_carp.h>
89 #ifdef INET
90 #include <netinet/if_ether.h>
91 #include <netinet/netdump/netdump.h>
92 #endif /* INET */
93 #ifdef INET6
94 #include <netinet6/in6_var.h>
95 #include <netinet6/in6_ifattach.h>
96 #endif /* INET6 */
97 #endif /* INET || INET6 */
98
99 #include <security/mac/mac_framework.h>
100
101 /*
102 * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
103 * and ifr_ifru when it is used in SIOCGIFCONF.
104 */
105 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
106 offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
107
108 __read_mostly epoch_t net_epoch_preempt;
109 __read_mostly epoch_t net_epoch;
110 #ifdef COMPAT_FREEBSD32
111 #include <sys/mount.h>
112 #include <compat/freebsd32/freebsd32.h>
113
114 struct ifreq_buffer32 {
115 uint32_t length; /* (size_t) */
116 uint32_t buffer; /* (void *) */
117 };
118
119 /*
120 * Interface request structure used for socket
121 * ioctl's. All interface ioctl's must have parameter
122 * definitions which begin with ifr_name. The
123 * remainder may be interface specific.
124 */
125 struct ifreq32 {
126 char ifr_name[IFNAMSIZ]; /* if name, e.g. "en0" */
127 union {
128 struct sockaddr ifru_addr;
129 struct sockaddr ifru_dstaddr;
130 struct sockaddr ifru_broadaddr;
131 struct ifreq_buffer32 ifru_buffer;
132 short ifru_flags[2];
133 short ifru_index;
134 int ifru_jid;
135 int ifru_metric;
136 int ifru_mtu;
137 int ifru_phys;
138 int ifru_media;
139 uint32_t ifru_data;
140 int ifru_cap[2];
141 u_int ifru_fib;
142 u_char ifru_vlan_pcp;
143 } ifr_ifru;
144 };
145 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
146 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
147 __offsetof(struct ifreq32, ifr_ifru));
148
149 struct ifgroupreq32 {
150 char ifgr_name[IFNAMSIZ];
151 u_int ifgr_len;
152 union {
153 char ifgru_group[IFNAMSIZ];
154 uint32_t ifgru_groups;
155 } ifgr_ifgru;
156 };
157
158 struct ifmediareq32 {
159 char ifm_name[IFNAMSIZ];
160 int ifm_current;
161 int ifm_mask;
162 int ifm_status;
163 int ifm_active;
164 int ifm_count;
165 uint32_t ifm_ulist; /* (int *) */
166 };
167 #define SIOCGIFMEDIA32 _IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
168 #define SIOCGIFXMEDIA32 _IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
169
170 #define _CASE_IOC_IFGROUPREQ_32(cmd) \
171 case _IOC_NEWTYPE((cmd), struct ifgroupreq32):
172 #else /* !COMPAT_FREEBSD32 */
173 #define _CASE_IOC_IFGROUPREQ_32(cmd)
174 #endif /* !COMPAT_FREEBSD32 */
175
176 #define CASE_IOC_IFGROUPREQ(cmd) \
177 _CASE_IOC_IFGROUPREQ_32(cmd) \
178 case (cmd)
179
180 union ifreq_union {
181 struct ifreq ifr;
182 #ifdef COMPAT_FREEBSD32
183 struct ifreq32 ifr32;
184 #endif
185 };
186
187 union ifgroupreq_union {
188 struct ifgroupreq ifgr;
189 #ifdef COMPAT_FREEBSD32
190 struct ifgroupreq32 ifgr32;
191 #endif
192 };
193
194 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
195 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
196
197 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
198 &ifqmaxlen, 0, "max send queue size");
199
200 /* Log link state change events */
201 static int log_link_state_change = 1;
202
203 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
204 &log_link_state_change, 0,
205 "log interface link state change events");
206
207 /* Log promiscuous mode change events */
208 static int log_promisc_mode_change = 1;
209
210 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
211 &log_promisc_mode_change, 1,
212 "log promiscuous mode change events");
213
214 /* Interface description */
215 static unsigned int ifdescr_maxlen = 1024;
216 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
217 &ifdescr_maxlen, 0,
218 "administrative maximum length for interface description");
219
220 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
221
222 /* global sx for non-critical path ifdescr */
223 static struct sx ifdescr_sx;
224 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
225
226 void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
227 void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
228 /* These are external hooks for CARP. */
229 void (*carp_linkstate_p)(struct ifnet *ifp);
230 void (*carp_demote_adj_p)(int, char *);
231 int (*carp_master_p)(struct ifaddr *);
232 #if defined(INET) || defined(INET6)
233 int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
234 int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
235 const struct sockaddr *sa);
236 int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
237 int (*carp_attach_p)(struct ifaddr *, int);
238 void (*carp_detach_p)(struct ifaddr *, bool);
239 #endif
240 #ifdef INET
241 int (*carp_iamatch_p)(struct ifaddr *, uint8_t **);
242 #endif
243 #ifdef INET6
244 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
245 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
246 const struct in6_addr *taddr);
247 #endif
248
249 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
250
251 /*
252 * XXX: Style; these should be sorted alphabetically, and unprototyped
253 * static functions should be prototyped. Currently they are sorted by
254 * declaration order.
255 */
256 static void if_attachdomain(void *);
257 static void if_attachdomain1(struct ifnet *);
258 static int ifconf(u_long, caddr_t);
259 static void *if_grow(void);
260 static void if_input_default(struct ifnet *, struct mbuf *);
261 static int if_requestencap_default(struct ifnet *, struct if_encap_req *);
262 static void if_route(struct ifnet *, int flag, int fam);
263 static int if_setflag(struct ifnet *, int, int, int *, int);
264 static int if_transmit(struct ifnet *ifp, struct mbuf *m);
265 static void if_unroute(struct ifnet *, int flag, int fam);
266 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
267 static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
268 static void do_link_state_change(void *, int);
269 static int if_getgroup(struct ifgroupreq *, struct ifnet *);
270 static int if_getgroupmembers(struct ifgroupreq *);
271 static void if_delgroups(struct ifnet *);
272 static void if_attach_internal(struct ifnet *, int, struct if_clone *);
273 static int if_detach_internal(struct ifnet *, int, struct if_clone **);
274 #ifdef VIMAGE
275 static void if_vmove(struct ifnet *, struct vnet *);
276 #endif
277
278 #ifdef INET6
279 /*
280 * XXX: declare here to avoid to include many inet6 related files..
281 * should be more generalized?
282 */
283 extern void nd6_setmtu(struct ifnet *);
284 #endif
285
286 /* ipsec helper hooks */
287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
289
290 VNET_DEFINE(int, if_index);
291 int ifqmaxlen = IFQ_MAXLEN;
292 VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */
293 VNET_DEFINE(struct ifgrouphead, ifg_head);
294
295 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
296
297 /* Table of ifnet by index. */
298 VNET_DEFINE(struct ifnet **, ifindex_table);
299
300 #define V_if_indexlim VNET(if_indexlim)
301 #define V_ifindex_table VNET(ifindex_table)
302
303 /*
304 * The global network interface list (V_ifnet) and related state (such as
305 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
306 * an rwlock. Either may be acquired shared to stablize the list, but both
307 * must be acquired writable to modify the list. This model allows us to
308 * both stablize the interface list during interrupt thread processing, but
309 * also to stablize it over long-running ioctls, without introducing priority
310 * inversions and deadlocks.
311 */
312 struct rwlock ifnet_rwlock;
313 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
314 struct sx ifnet_sxlock;
315 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
316
317 /*
318 * The allocation of network interfaces is a rather non-atomic affair; we
319 * need to select an index before we are ready to expose the interface for
320 * use, so will use this pointer value to indicate reservation.
321 */
322 #define IFNET_HOLD (void *)(uintptr_t)(-1)
323
324 static if_com_alloc_t *if_com_alloc[256];
325 static if_com_free_t *if_com_free[256];
326
327 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
328 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
329 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
330
331 struct ifnet *
332 ifnet_byindex_locked(u_short idx)
333 {
334
335 if (idx > V_if_index)
336 return (NULL);
337 if (V_ifindex_table[idx] == IFNET_HOLD)
338 return (NULL);
339 return (V_ifindex_table[idx]);
340 }
341
342 struct ifnet *
343 ifnet_byindex(u_short idx)
344 {
345 struct ifnet *ifp;
346
347 ifp = ifnet_byindex_locked(idx);
348 return (ifp);
349 }
350
351 struct ifnet *
352 ifnet_byindex_ref(u_short idx)
353 {
354 struct ifnet *ifp;
355
356 IFNET_RLOCK_NOSLEEP();
357 ifp = ifnet_byindex_locked(idx);
358 if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
359 IFNET_RUNLOCK_NOSLEEP();
360 return (NULL);
361 }
362 if_ref(ifp);
363 IFNET_RUNLOCK_NOSLEEP();
364 return (ifp);
365 }
366
367 /*
368 * Allocate an ifindex array entry; return 0 on success or an error on
369 * failure.
370 */
371 static u_short
372 ifindex_alloc(void **old)
373 {
374 u_short idx;
375
376 IFNET_WLOCK_ASSERT();
377 /*
378 * Try to find an empty slot below V_if_index. If we fail, take the
379 * next slot.
380 */
381 for (idx = 1; idx <= V_if_index; idx++) {
382 if (V_ifindex_table[idx] == NULL)
383 break;
384 }
385
386 /* Catch if_index overflow. */
387 if (idx >= V_if_indexlim) {
388 *old = if_grow();
389 return (USHRT_MAX);
390 }
391 if (idx > V_if_index)
392 V_if_index = idx;
393 return (idx);
394 }
395
396 static void
397 ifindex_free_locked(u_short idx)
398 {
399
400 IFNET_WLOCK_ASSERT();
401
402 V_ifindex_table[idx] = NULL;
403 while (V_if_index > 0 &&
404 V_ifindex_table[V_if_index] == NULL)
405 V_if_index--;
406 }
407
408 static void
409 ifindex_free(u_short idx)
410 {
411
412 IFNET_WLOCK();
413 ifindex_free_locked(idx);
414 IFNET_WUNLOCK();
415 }
416
417 static void
418 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
419 {
420
421 V_ifindex_table[idx] = ifp;
422 }
423
424 struct ifaddr *
425 ifaddr_byindex(u_short idx)
426 {
427 struct ifnet *ifp;
428 struct ifaddr *ifa = NULL;
429
430 IFNET_RLOCK_NOSLEEP();
431 ifp = ifnet_byindex_locked(idx);
432 if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
433 ifa_ref(ifa);
434 IFNET_RUNLOCK_NOSLEEP();
435 return (ifa);
436 }
437
438 /*
439 * Network interface utility routines.
440 *
441 * Routines with ifa_ifwith* names take sockaddr *'s as
442 * parameters.
443 */
444
445 static void
446 vnet_if_init(const void *unused __unused)
447 {
448 void *old;
449
450 CK_STAILQ_INIT(&V_ifnet);
451 CK_STAILQ_INIT(&V_ifg_head);
452 IFNET_WLOCK();
453 old = if_grow(); /* create initial table */
454 IFNET_WUNLOCK();
455 epoch_wait_preempt(net_epoch_preempt);
456 free(old, M_IFNET);
457 vnet_if_clone_init();
458 }
459 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
460 NULL);
461
462 #ifdef VIMAGE
463 static void
464 vnet_if_uninit(const void *unused __unused)
465 {
466
467 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
468 "not empty", __func__, __LINE__, &V_ifnet));
469 VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
470 "not empty", __func__, __LINE__, &V_ifg_head));
471
472 free((caddr_t)V_ifindex_table, M_IFNET);
473 }
474 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
475 vnet_if_uninit, NULL);
476
477 static void
478 vnet_if_return(const void *unused __unused)
479 {
480 struct ifnet *ifp, *nifp;
481
482 /* Return all inherited interfaces to their parent vnets. */
483 CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
484 if (ifp->if_home_vnet != ifp->if_vnet)
485 if_vmove(ifp, ifp->if_home_vnet);
486 }
487 }
488 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
489 vnet_if_return, NULL);
490 #endif
491
492
493 static void *
494 if_grow(void)
495 {
496 int oldlim;
497 u_int n;
498 struct ifnet **e;
499 void *old;
500
501 old = NULL;
502 IFNET_WLOCK_ASSERT();
503 oldlim = V_if_indexlim;
504 IFNET_WUNLOCK();
505 n = (oldlim << 1) * sizeof(*e);
506 e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
507 IFNET_WLOCK();
508 if (V_if_indexlim != oldlim) {
509 free(e, M_IFNET);
510 return (NULL);
511 }
512 if (V_ifindex_table != NULL) {
513 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
514 old = V_ifindex_table;
515 }
516 V_if_indexlim <<= 1;
517 V_ifindex_table = e;
518 return (old);
519 }
520
521 /*
522 * Allocate a struct ifnet and an index for an interface. A layer 2
523 * common structure will also be allocated if an allocation routine is
524 * registered for the passed type.
525 */
526 struct ifnet *
527 if_alloc(u_char type)
528 {
529 struct ifnet *ifp;
530 u_short idx;
531 void *old;
532
533 ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
534 restart:
535 IFNET_WLOCK();
536 idx = ifindex_alloc(&old);
537 if (__predict_false(idx == USHRT_MAX)) {
538 IFNET_WUNLOCK();
539 epoch_wait_preempt(net_epoch_preempt);
540 free(old, M_IFNET);
541 goto restart;
542 }
543 ifnet_setbyindex(idx, IFNET_HOLD);
544 IFNET_WUNLOCK();
545 ifp->if_index = idx;
546 ifp->if_type = type;
547 ifp->if_alloctype = type;
548 #ifdef VIMAGE
549 ifp->if_vnet = curvnet;
550 #endif
551 if (if_com_alloc[type] != NULL) {
552 ifp->if_l2com = if_com_alloc[type](type, ifp);
553 if (ifp->if_l2com == NULL) {
554 free(ifp, M_IFNET);
555 ifindex_free(idx);
556 return (NULL);
557 }
558 }
559
560 IF_ADDR_LOCK_INIT(ifp);
561 TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
562 ifp->if_afdata_initialized = 0;
563 IF_AFDATA_LOCK_INIT(ifp);
564 CK_STAILQ_INIT(&ifp->if_addrhead);
565 CK_STAILQ_INIT(&ifp->if_multiaddrs);
566 CK_STAILQ_INIT(&ifp->if_groups);
567 #ifdef MAC
568 mac_ifnet_init(ifp);
569 #endif
570 ifq_init(&ifp->if_snd, ifp);
571
572 refcount_init(&ifp->if_refcount, 1); /* Index reference. */
573 for (int i = 0; i < IFCOUNTERS; i++)
574 ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
575 ifp->if_get_counter = if_get_counter_default;
576 ifp->if_pcp = IFNET_PCP_NONE;
577 ifnet_setbyindex(ifp->if_index, ifp);
578 return (ifp);
579 }
580
581 /*
582 * Do the actual work of freeing a struct ifnet, and layer 2 common
583 * structure. This call is made when the last reference to an
584 * interface is released.
585 */
586 static void
587 if_free_internal(struct ifnet *ifp)
588 {
589
590 KASSERT((ifp->if_flags & IFF_DYING),
591 ("if_free_internal: interface not dying"));
592
593 if (if_com_free[ifp->if_alloctype] != NULL)
594 if_com_free[ifp->if_alloctype](ifp->if_l2com,
595 ifp->if_alloctype);
596
597 #ifdef MAC
598 mac_ifnet_destroy(ifp);
599 #endif /* MAC */
600 if (ifp->if_description != NULL)
601 free(ifp->if_description, M_IFDESCR);
602 IF_AFDATA_DESTROY(ifp);
603 IF_ADDR_LOCK_DESTROY(ifp);
604 ifq_delete(&ifp->if_snd);
605
606 for (int i = 0; i < IFCOUNTERS; i++)
607 counter_u64_free(ifp->if_counters[i]);
608
609 free(ifp, M_IFNET);
610 }
611
612 static void
613 if_destroy(epoch_context_t ctx)
614 {
615 struct ifnet *ifp;
616
617 ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
618 if_free_internal(ifp);
619 }
620
621 /*
622 * Deregister an interface and free the associated storage.
623 */
624 void
625 if_free(struct ifnet *ifp)
626 {
627
628 ifp->if_flags |= IFF_DYING; /* XXX: Locking */
629
630 CURVNET_SET_QUIET(ifp->if_vnet);
631 IFNET_WLOCK();
632 KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
633 ("%s: freeing unallocated ifnet", ifp->if_xname));
634
635 ifindex_free_locked(ifp->if_index);
636 IFNET_WUNLOCK();
637
638 if (refcount_release(&ifp->if_refcount))
639 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
640 CURVNET_RESTORE();
641 }
642
643 /*
644 * Interfaces to keep an ifnet type-stable despite the possibility of the
645 * driver calling if_free(). If there are additional references, we defer
646 * freeing the underlying data structure.
647 */
648 void
649 if_ref(struct ifnet *ifp)
650 {
651
652 /* We don't assert the ifnet list lock here, but arguably should. */
653 refcount_acquire(&ifp->if_refcount);
654 }
655
656 void
657 if_rele(struct ifnet *ifp)
658 {
659
660 if (!refcount_release(&ifp->if_refcount))
661 return;
662 epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
663 }
664
665 void
666 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
667 {
668
669 mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
670
671 if (ifq->ifq_maxlen == 0)
672 ifq->ifq_maxlen = ifqmaxlen;
673
674 ifq->altq_type = 0;
675 ifq->altq_disc = NULL;
676 ifq->altq_flags &= ALTQF_CANTCHANGE;
677 ifq->altq_tbr = NULL;
678 ifq->altq_ifp = ifp;
679 }
680
681 void
682 ifq_delete(struct ifaltq *ifq)
683 {
684 mtx_destroy(&ifq->ifq_mtx);
685 }
686
687 /*
688 * Perform generic interface initialization tasks and attach the interface
689 * to the list of "active" interfaces. If vmove flag is set on entry
690 * to if_attach_internal(), perform only a limited subset of initialization
691 * tasks, given that we are moving from one vnet to another an ifnet which
692 * has already been fully initialized.
693 *
694 * Note that if_detach_internal() removes group membership unconditionally
695 * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
696 * Thus, when if_vmove() is applied to a cloned interface, group membership
697 * is lost while a cloned one always joins a group whose name is
698 * ifc->ifc_name. To recover this after if_detach_internal() and
699 * if_attach_internal(), the cloner should be specified to
700 * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal()
701 * attempts to join a group whose name is ifc->ifc_name.
702 *
703 * XXX:
704 * - The decision to return void and thus require this function to
705 * succeed is questionable.
706 * - We should probably do more sanity checking. For instance we don't
707 * do anything to insure if_xname is unique or non-empty.
708 */
709 void
710 if_attach(struct ifnet *ifp)
711 {
712
713 if_attach_internal(ifp, 0, NULL);
714 }
715
716 /*
717 * Compute the least common TSO limit.
718 */
719 void
720 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
721 {
722 /*
723 * 1) If there is no limit currently, take the limit from
724 * the network adapter.
725 *
726 * 2) If the network adapter has a limit below the current
727 * limit, apply it.
728 */
729 if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
730 ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
731 pmax->tsomaxbytes = ifp->if_hw_tsomax;
732 }
733 if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
734 ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
735 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
736 }
737 if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
738 ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
739 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
740 }
741 }
742
743 /*
744 * Update TSO limit of a network adapter.
745 *
746 * Returns zero if no change. Else non-zero.
747 */
748 int
749 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
750 {
751 int retval = 0;
752 if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
753 ifp->if_hw_tsomax = pmax->tsomaxbytes;
754 retval++;
755 }
756 if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
757 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
758 retval++;
759 }
760 if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
761 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
762 retval++;
763 }
764 return (retval);
765 }
766
767 static void
768 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
769 {
770 unsigned socksize, ifasize;
771 int namelen, masklen;
772 struct sockaddr_dl *sdl;
773 struct ifaddr *ifa;
774
775 if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
776 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
777 ifp->if_xname);
778
779 #ifdef VIMAGE
780 ifp->if_vnet = curvnet;
781 if (ifp->if_home_vnet == NULL)
782 ifp->if_home_vnet = curvnet;
783 #endif
784
785 if_addgroup(ifp, IFG_ALL);
786
787 /* Restore group membership for cloned interfaces. */
788 if (vmove && ifc != NULL)
789 if_clone_addgroup(ifp, ifc);
790
791 getmicrotime(&ifp->if_lastchange);
792 ifp->if_epoch = time_uptime;
793
794 KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
795 (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
796 ("transmit and qflush must both either be set or both be NULL"));
797 if (ifp->if_transmit == NULL) {
798 ifp->if_transmit = if_transmit;
799 ifp->if_qflush = if_qflush;
800 }
801 if (ifp->if_input == NULL)
802 ifp->if_input = if_input_default;
803
804 if (ifp->if_requestencap == NULL)
805 ifp->if_requestencap = if_requestencap_default;
806
807 if (!vmove) {
808 #ifdef MAC
809 mac_ifnet_create(ifp);
810 #endif
811
812 /*
813 * Create a Link Level name for this device.
814 */
815 namelen = strlen(ifp->if_xname);
816 /*
817 * Always save enough space for any possiable name so we
818 * can do a rename in place later.
819 */
820 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
821 socksize = masklen + ifp->if_addrlen;
822 if (socksize < sizeof(*sdl))
823 socksize = sizeof(*sdl);
824 socksize = roundup2(socksize, sizeof(long));
825 ifasize = sizeof(*ifa) + 2 * socksize;
826 ifa = ifa_alloc(ifasize, M_WAITOK);
827 sdl = (struct sockaddr_dl *)(ifa + 1);
828 sdl->sdl_len = socksize;
829 sdl->sdl_family = AF_LINK;
830 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
831 sdl->sdl_nlen = namelen;
832 sdl->sdl_index = ifp->if_index;
833 sdl->sdl_type = ifp->if_type;
834 ifp->if_addr = ifa;
835 ifa->ifa_ifp = ifp;
836 ifa->ifa_rtrequest = link_rtrequest;
837 ifa->ifa_addr = (struct sockaddr *)sdl;
838 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
839 ifa->ifa_netmask = (struct sockaddr *)sdl;
840 sdl->sdl_len = masklen;
841 while (namelen != 0)
842 sdl->sdl_data[--namelen] = 0xff;
843 CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
844 /* Reliably crash if used uninitialized. */
845 ifp->if_broadcastaddr = NULL;
846
847 if (ifp->if_type == IFT_ETHER) {
848 ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
849 M_WAITOK | M_ZERO);
850 }
851
852 #if defined(INET) || defined(INET6)
853 /* Use defaults for TSO, if nothing is set */
854 if (ifp->if_hw_tsomax == 0 &&
855 ifp->if_hw_tsomaxsegcount == 0 &&
856 ifp->if_hw_tsomaxsegsize == 0) {
857 /*
858 * The TSO defaults needs to be such that an
859 * NFS mbuf list of 35 mbufs totalling just
860 * below 64K works and that a chain of mbufs
861 * can be defragged into at most 32 segments:
862 */
863 ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
864 (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
865 ifp->if_hw_tsomaxsegcount = 35;
866 ifp->if_hw_tsomaxsegsize = 2048; /* 2K */
867
868 /* XXX some drivers set IFCAP_TSO after ethernet attach */
869 if (ifp->if_capabilities & IFCAP_TSO) {
870 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
871 ifp->if_hw_tsomax,
872 ifp->if_hw_tsomaxsegcount,
873 ifp->if_hw_tsomaxsegsize);
874 }
875 }
876 #endif
877 }
878 #ifdef VIMAGE
879 else {
880 /*
881 * Update the interface index in the link layer address
882 * of the interface.
883 */
884 for (ifa = ifp->if_addr; ifa != NULL;
885 ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
886 if (ifa->ifa_addr->sa_family == AF_LINK) {
887 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
888 sdl->sdl_index = ifp->if_index;
889 }
890 }
891 }
892 #endif
893
894 IFNET_WLOCK();
895 CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
896 #ifdef VIMAGE
897 curvnet->vnet_ifcnt++;
898 #endif
899 IFNET_WUNLOCK();
900
901 if (domain_init_status >= 2)
902 if_attachdomain1(ifp);
903
904 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
905 if (IS_DEFAULT_VNET(curvnet))
906 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
907
908 /* Announce the interface. */
909 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
910 }
911
912 static void
913 if_epochalloc(void *dummy __unused)
914 {
915
916 net_epoch_preempt = epoch_alloc(EPOCH_PREEMPT);
917 net_epoch = epoch_alloc(0);
918 }
919 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY,
920 if_epochalloc, NULL);
921
922 static void
923 if_attachdomain(void *dummy)
924 {
925 struct ifnet *ifp;
926
927 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
928 if_attachdomain1(ifp);
929 }
930 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
931 if_attachdomain, NULL);
932
933 static void
934 if_attachdomain1(struct ifnet *ifp)
935 {
936 struct domain *dp;
937
938 /*
939 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
940 * cannot lock ifp->if_afdata initialization, entirely.
941 */
942 IF_AFDATA_LOCK(ifp);
943 if (ifp->if_afdata_initialized >= domain_init_status) {
944 IF_AFDATA_UNLOCK(ifp);
945 log(LOG_WARNING, "%s called more than once on %s\n",
946 __func__, ifp->if_xname);
947 return;
948 }
949 ifp->if_afdata_initialized = domain_init_status;
950 IF_AFDATA_UNLOCK(ifp);
951
952 /* address family dependent data region */
953 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
954 for (dp = domains; dp; dp = dp->dom_next) {
955 if (dp->dom_ifattach)
956 ifp->if_afdata[dp->dom_family] =
957 (*dp->dom_ifattach)(ifp);
958 }
959 }
960
961 /*
962 * Remove any unicast or broadcast network addresses from an interface.
963 */
964 void
965 if_purgeaddrs(struct ifnet *ifp)
966 {
967 struct ifaddr *ifa;
968
969 while (1) {
970 NET_EPOCH_ENTER();
971 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
972 if (ifa->ifa_addr->sa_family != AF_LINK)
973 break;
974 }
975 NET_EPOCH_EXIT();
976
977 if (ifa == NULL)
978 break;
979 #ifdef INET
980 /* XXX: Ugly!! ad hoc just for INET */
981 if (ifa->ifa_addr->sa_family == AF_INET) {
982 struct ifaliasreq ifr;
983
984 bzero(&ifr, sizeof(ifr));
985 ifr.ifra_addr = *ifa->ifa_addr;
986 if (ifa->ifa_dstaddr)
987 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
988 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
989 NULL) == 0)
990 continue;
991 }
992 #endif /* INET */
993 #ifdef INET6
994 if (ifa->ifa_addr->sa_family == AF_INET6) {
995 in6_purgeaddr(ifa);
996 /* ifp_addrhead is already updated */
997 continue;
998 }
999 #endif /* INET6 */
1000 IF_ADDR_WLOCK(ifp);
1001 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1002 IF_ADDR_WUNLOCK(ifp);
1003 ifa_free(ifa);
1004 }
1005 }
1006
1007 /*
1008 * Remove any multicast network addresses from an interface when an ifnet
1009 * is going away.
1010 */
1011 static void
1012 if_purgemaddrs(struct ifnet *ifp)
1013 {
1014 struct ifmultiaddr *ifma;
1015
1016 IF_ADDR_WLOCK(ifp);
1017 while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1018 ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1019 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1020 if_delmulti_locked(ifp, ifma, 1);
1021 }
1022 IF_ADDR_WUNLOCK(ifp);
1023 }
1024
1025 /*
1026 * Detach an interface, removing it from the list of "active" interfaces.
1027 * If vmove flag is set on entry to if_detach_internal(), perform only a
1028 * limited subset of cleanup tasks, given that we are moving an ifnet from
1029 * one vnet to another, where it must be fully operational.
1030 *
1031 * XXXRW: There are some significant questions about event ordering, and
1032 * how to prevent things from starting to use the interface during detach.
1033 */
1034 void
1035 if_detach(struct ifnet *ifp)
1036 {
1037
1038 CURVNET_SET_QUIET(ifp->if_vnet);
1039 if_detach_internal(ifp, 0, NULL);
1040 CURVNET_RESTORE();
1041 }
1042
1043 /*
1044 * The vmove flag, if set, indicates that we are called from a callpath
1045 * that is moving an interface to a different vnet instance.
1046 *
1047 * The shutdown flag, if set, indicates that we are called in the
1048 * process of shutting down a vnet instance. Currently only the
1049 * vnet_if_return SYSUNINIT function sets it. Note: we can be called
1050 * on a vnet instance shutdown without this flag being set, e.g., when
1051 * the cloned interfaces are destoyed as first thing of teardown.
1052 */
1053 static int
1054 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1055 {
1056 struct ifaddr *ifa;
1057 int i;
1058 struct domain *dp;
1059 struct ifnet *iter;
1060 int found = 0;
1061 #ifdef VIMAGE
1062 int shutdown;
1063
1064 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1065 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1066 #endif
1067 IFNET_WLOCK();
1068 CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
1069 if (iter == ifp) {
1070 CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
1071 found = 1;
1072 break;
1073 }
1074 IFNET_WUNLOCK();
1075 if (!found) {
1076 /*
1077 * While we would want to panic here, we cannot
1078 * guarantee that the interface is indeed still on
1079 * the list given we don't hold locks all the way.
1080 */
1081 return (ENOENT);
1082 #if 0
1083 if (vmove)
1084 panic("%s: ifp=%p not on the ifnet tailq %p",
1085 __func__, ifp, &V_ifnet);
1086 else
1087 return; /* XXX this should panic as well? */
1088 #endif
1089 }
1090
1091 /*
1092 * At this point we know the interface still was on the ifnet list
1093 * and we removed it so we are in a stable state.
1094 */
1095 #ifdef VIMAGE
1096 curvnet->vnet_ifcnt--;
1097 #endif
1098 epoch_wait_preempt(net_epoch_preempt);
1099 /*
1100 * In any case (destroy or vmove) detach us from the groups
1101 * and remove/wait for pending events on the taskq.
1102 * XXX-BZ in theory an interface could still enqueue a taskq change?
1103 */
1104 if_delgroups(ifp);
1105
1106 taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1107
1108 /*
1109 * Check if this is a cloned interface or not. Must do even if
1110 * shutting down as a if_vmove_reclaim() would move the ifp and
1111 * the if_clone_addgroup() will have a corrupted string overwise
1112 * from a gibberish pointer.
1113 */
1114 if (vmove && ifcp != NULL)
1115 *ifcp = if_clone_findifc(ifp);
1116
1117 if_down(ifp);
1118
1119 #ifdef VIMAGE
1120 /*
1121 * On VNET shutdown abort here as the stack teardown will do all
1122 * the work top-down for us.
1123 */
1124 if (shutdown) {
1125 /*
1126 * In case of a vmove we are done here without error.
1127 * If we would signal an error it would lead to the same
1128 * abort as if we did not find the ifnet anymore.
1129 * if_detach() calls us in void context and does not care
1130 * about an early abort notification, so life is splendid :)
1131 */
1132 goto finish_vnet_shutdown;
1133 }
1134 #endif
1135
1136 /*
1137 * At this point we are not tearing down a VNET and are either
1138 * going to destroy or vmove the interface and have to cleanup
1139 * accordingly.
1140 */
1141
1142 /*
1143 * Remove routes and flush queues.
1144 */
1145 #ifdef ALTQ
1146 if (ALTQ_IS_ENABLED(&ifp->if_snd))
1147 altq_disable(&ifp->if_snd);
1148 if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1149 altq_detach(&ifp->if_snd);
1150 #endif
1151
1152 if_purgeaddrs(ifp);
1153
1154 #ifdef INET
1155 in_ifdetach(ifp);
1156 #endif
1157
1158 #ifdef INET6
1159 /*
1160 * Remove all IPv6 kernel structs related to ifp. This should be done
1161 * before removing routing entries below, since IPv6 interface direct
1162 * routes are expected to be removed by the IPv6-specific kernel API.
1163 * Otherwise, the kernel will detect some inconsistency and bark it.
1164 */
1165 in6_ifdetach(ifp);
1166 #endif
1167 if_purgemaddrs(ifp);
1168
1169 /* Announce that the interface is gone. */
1170 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1171 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1172 if (IS_DEFAULT_VNET(curvnet))
1173 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1174
1175 if (!vmove) {
1176 /*
1177 * Prevent further calls into the device driver via ifnet.
1178 */
1179 if_dead(ifp);
1180
1181 /*
1182 * Remove link ifaddr pointer and maybe decrement if_index.
1183 * Clean up all addresses.
1184 */
1185 free(ifp->if_hw_addr, M_IFADDR);
1186 ifp->if_hw_addr = NULL;
1187 ifp->if_addr = NULL;
1188
1189 /* We can now free link ifaddr. */
1190 IF_ADDR_WLOCK(ifp);
1191 if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1192 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1193 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1194 IF_ADDR_WUNLOCK(ifp);
1195 ifa_free(ifa);
1196 } else
1197 IF_ADDR_WUNLOCK(ifp);
1198 }
1199
1200 rt_flushifroutes(ifp);
1201
1202 #ifdef VIMAGE
1203 finish_vnet_shutdown:
1204 #endif
1205 /*
1206 * We cannot hold the lock over dom_ifdetach calls as they might
1207 * sleep, for example trying to drain a callout, thus open up the
1208 * theoretical race with re-attaching.
1209 */
1210 IF_AFDATA_LOCK(ifp);
1211 i = ifp->if_afdata_initialized;
1212 ifp->if_afdata_initialized = 0;
1213 IF_AFDATA_UNLOCK(ifp);
1214 for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1215 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1216 (*dp->dom_ifdetach)(ifp,
1217 ifp->if_afdata[dp->dom_family]);
1218 ifp->if_afdata[dp->dom_family] = NULL;
1219 }
1220 }
1221
1222 return (0);
1223 }
1224
1225 #ifdef VIMAGE
1226 /*
1227 * if_vmove() performs a limited version of if_detach() in current
1228 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1229 * An attempt is made to shrink if_index in current vnet, find an
1230 * unused if_index in target vnet and calls if_grow() if necessary,
1231 * and finally find an unused if_xname for the target vnet.
1232 */
1233 static void
1234 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1235 {
1236 struct if_clone *ifc;
1237 u_int bif_dlt, bif_hdrlen;
1238 void *old;
1239 int rc;
1240
1241 /*
1242 * if_detach_internal() will call the eventhandler to notify
1243 * interface departure. That will detach if_bpf. We need to
1244 * safe the dlt and hdrlen so we can re-attach it later.
1245 */
1246 bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1247
1248 /*
1249 * Detach from current vnet, but preserve LLADDR info, do not
1250 * mark as dead etc. so that the ifnet can be reattached later.
1251 * If we cannot find it, we lost the race to someone else.
1252 */
1253 rc = if_detach_internal(ifp, 1, &ifc);
1254 if (rc != 0)
1255 return;
1256
1257 /*
1258 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1259 * the if_index for that vnet if possible.
1260 *
1261 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1262 * or we'd lock on one vnet and unlock on another.
1263 */
1264 IFNET_WLOCK();
1265 ifindex_free_locked(ifp->if_index);
1266 IFNET_WUNLOCK();
1267
1268 /*
1269 * Perform interface-specific reassignment tasks, if provided by
1270 * the driver.
1271 */
1272 if (ifp->if_reassign != NULL)
1273 ifp->if_reassign(ifp, new_vnet, NULL);
1274
1275 /*
1276 * Switch to the context of the target vnet.
1277 */
1278 CURVNET_SET_QUIET(new_vnet);
1279 restart:
1280 IFNET_WLOCK();
1281 ifp->if_index = ifindex_alloc(&old);
1282 if (__predict_false(ifp->if_index == USHRT_MAX)) {
1283 IFNET_WUNLOCK();
1284 epoch_wait_preempt(net_epoch_preempt);
1285 free(old, M_IFNET);
1286 goto restart;
1287 }
1288 ifnet_setbyindex(ifp->if_index, ifp);
1289 IFNET_WUNLOCK();
1290
1291 if_attach_internal(ifp, 1, ifc);
1292
1293 if (ifp->if_bpf == NULL)
1294 bpfattach(ifp, bif_dlt, bif_hdrlen);
1295
1296 CURVNET_RESTORE();
1297 }
1298
1299 /*
1300 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1301 */
1302 static int
1303 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1304 {
1305 struct prison *pr;
1306 struct ifnet *difp;
1307 int shutdown;
1308
1309 /* Try to find the prison within our visibility. */
1310 sx_slock(&allprison_lock);
1311 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1312 sx_sunlock(&allprison_lock);
1313 if (pr == NULL)
1314 return (ENXIO);
1315 prison_hold_locked(pr);
1316 mtx_unlock(&pr->pr_mtx);
1317
1318 /* Do not try to move the iface from and to the same prison. */
1319 if (pr->pr_vnet == ifp->if_vnet) {
1320 prison_free(pr);
1321 return (EEXIST);
1322 }
1323
1324 /* Make sure the named iface does not exists in the dst. prison/vnet. */
1325 /* XXX Lock interfaces to avoid races. */
1326 CURVNET_SET_QUIET(pr->pr_vnet);
1327 difp = ifunit(ifname);
1328 if (difp != NULL) {
1329 CURVNET_RESTORE();
1330 prison_free(pr);
1331 return (EEXIST);
1332 }
1333
1334 /* Make sure the VNET is stable. */
1335 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1336 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1337 if (shutdown) {
1338 CURVNET_RESTORE();
1339 prison_free(pr);
1340 return (EBUSY);
1341 }
1342 CURVNET_RESTORE();
1343
1344 /* Move the interface into the child jail/vnet. */
1345 if_vmove(ifp, pr->pr_vnet);
1346
1347 /* Report the new if_xname back to the userland. */
1348 sprintf(ifname, "%s", ifp->if_xname);
1349
1350 prison_free(pr);
1351 return (0);
1352 }
1353
1354 static int
1355 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1356 {
1357 struct prison *pr;
1358 struct vnet *vnet_dst;
1359 struct ifnet *ifp;
1360 int shutdown;
1361
1362 /* Try to find the prison within our visibility. */
1363 sx_slock(&allprison_lock);
1364 pr = prison_find_child(td->td_ucred->cr_prison, jid);
1365 sx_sunlock(&allprison_lock);
1366 if (pr == NULL)
1367 return (ENXIO);
1368 prison_hold_locked(pr);
1369 mtx_unlock(&pr->pr_mtx);
1370
1371 /* Make sure the named iface exists in the source prison/vnet. */
1372 CURVNET_SET(pr->pr_vnet);
1373 ifp = ifunit(ifname); /* XXX Lock to avoid races. */
1374 if (ifp == NULL) {
1375 CURVNET_RESTORE();
1376 prison_free(pr);
1377 return (ENXIO);
1378 }
1379
1380 /* Do not try to move the iface from and to the same prison. */
1381 vnet_dst = TD_TO_VNET(td);
1382 if (vnet_dst == ifp->if_vnet) {
1383 CURVNET_RESTORE();
1384 prison_free(pr);
1385 return (EEXIST);
1386 }
1387
1388 /* Make sure the VNET is stable. */
1389 shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1390 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1391 if (shutdown) {
1392 CURVNET_RESTORE();
1393 prison_free(pr);
1394 return (EBUSY);
1395 }
1396
1397 /* Get interface back from child jail/vnet. */
1398 if_vmove(ifp, vnet_dst);
1399 CURVNET_RESTORE();
1400
1401 /* Report the new if_xname back to the userland. */
1402 sprintf(ifname, "%s", ifp->if_xname);
1403
1404 prison_free(pr);
1405 return (0);
1406 }
1407 #endif /* VIMAGE */
1408
1409 /*
1410 * Add a group to an interface
1411 */
1412 int
1413 if_addgroup(struct ifnet *ifp, const char *groupname)
1414 {
1415 struct ifg_list *ifgl;
1416 struct ifg_group *ifg = NULL;
1417 struct ifg_member *ifgm;
1418 int new = 0;
1419
1420 if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
1421 groupname[strlen(groupname) - 1] <= '9')
1422 return (EINVAL);
1423
1424 IFNET_WLOCK();
1425 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1426 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1427 IFNET_WUNLOCK();
1428 return (EEXIST);
1429 }
1430
1431 if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1432 M_NOWAIT)) == NULL) {
1433 IFNET_WUNLOCK();
1434 return (ENOMEM);
1435 }
1436
1437 if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1438 M_TEMP, M_NOWAIT)) == NULL) {
1439 free(ifgl, M_TEMP);
1440 IFNET_WUNLOCK();
1441 return (ENOMEM);
1442 }
1443
1444 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1445 if (!strcmp(ifg->ifg_group, groupname))
1446 break;
1447
1448 if (ifg == NULL) {
1449 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1450 M_TEMP, M_NOWAIT)) == NULL) {
1451 free(ifgl, M_TEMP);
1452 free(ifgm, M_TEMP);
1453 IFNET_WUNLOCK();
1454 return (ENOMEM);
1455 }
1456 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1457 ifg->ifg_refcnt = 0;
1458 CK_STAILQ_INIT(&ifg->ifg_members);
1459 CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1460 new = 1;
1461 }
1462
1463 ifg->ifg_refcnt++;
1464 ifgl->ifgl_group = ifg;
1465 ifgm->ifgm_ifp = ifp;
1466
1467 IF_ADDR_WLOCK(ifp);
1468 CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1469 CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1470 IF_ADDR_WUNLOCK(ifp);
1471
1472 IFNET_WUNLOCK();
1473
1474 if (new)
1475 EVENTHANDLER_INVOKE(group_attach_event, ifg);
1476 EVENTHANDLER_INVOKE(group_change_event, groupname);
1477
1478 return (0);
1479 }
1480
1481 /*
1482 * Remove a group from an interface
1483 */
1484 int
1485 if_delgroup(struct ifnet *ifp, const char *groupname)
1486 {
1487 struct ifg_list *ifgl;
1488 struct ifg_member *ifgm;
1489 int freeifgl;
1490
1491 IFNET_WLOCK();
1492 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1493 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1494 break;
1495 if (ifgl == NULL) {
1496 IFNET_WUNLOCK();
1497 return (ENOENT);
1498 }
1499
1500 freeifgl = 0;
1501 IF_ADDR_WLOCK(ifp);
1502 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1503 IF_ADDR_WUNLOCK(ifp);
1504
1505 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1506 if (ifgm->ifgm_ifp == ifp)
1507 break;
1508
1509 if (ifgm != NULL)
1510 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next);
1511
1512 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1513 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1514 freeifgl = 1;
1515 }
1516 IFNET_WUNLOCK();
1517
1518 epoch_wait_preempt(net_epoch_preempt);
1519 if (freeifgl) {
1520 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1521 free(ifgl->ifgl_group, M_TEMP);
1522 }
1523 free(ifgm, M_TEMP);
1524 free(ifgl, M_TEMP);
1525
1526 EVENTHANDLER_INVOKE(group_change_event, groupname);
1527
1528 return (0);
1529 }
1530
1531 /*
1532 * Remove an interface from all groups
1533 */
1534 static void
1535 if_delgroups(struct ifnet *ifp)
1536 {
1537 struct ifg_list *ifgl;
1538 struct ifg_member *ifgm;
1539 char groupname[IFNAMSIZ];
1540 int ifglfree;
1541
1542 IFNET_WLOCK();
1543 while (!CK_STAILQ_EMPTY(&ifp->if_groups)) {
1544 ifgl = CK_STAILQ_FIRST(&ifp->if_groups);
1545
1546 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1547
1548 IF_ADDR_WLOCK(ifp);
1549 CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1550 IF_ADDR_WUNLOCK(ifp);
1551
1552 CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1553 if (ifgm->ifgm_ifp == ifp)
1554 break;
1555
1556 if (ifgm != NULL)
1557 CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member,
1558 ifgm_next);
1559 ifglfree = 0;
1560 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1561 CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1562 ifglfree = 1;
1563 }
1564
1565 IFNET_WUNLOCK();
1566 epoch_wait_preempt(net_epoch_preempt);
1567 free(ifgm, M_TEMP);
1568 if (ifglfree) {
1569 EVENTHANDLER_INVOKE(group_detach_event,
1570 ifgl->ifgl_group);
1571 free(ifgl->ifgl_group, M_TEMP);
1572 }
1573 EVENTHANDLER_INVOKE(group_change_event, groupname);
1574
1575 IFNET_WLOCK();
1576 }
1577 IFNET_WUNLOCK();
1578 }
1579
1580 static char *
1581 ifgr_group_get(void *ifgrp)
1582 {
1583 union ifgroupreq_union *ifgrup;
1584
1585 ifgrup = ifgrp;
1586 #ifdef COMPAT_FREEBSD32
1587 if (SV_CURPROC_FLAG(SV_ILP32))
1588 return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1589 #endif
1590 return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1591 }
1592
1593 static struct ifg_req *
1594 ifgr_groups_get(void *ifgrp)
1595 {
1596 union ifgroupreq_union *ifgrup;
1597
1598 ifgrup = ifgrp;
1599 #ifdef COMPAT_FREEBSD32
1600 if (SV_CURPROC_FLAG(SV_ILP32))
1601 return ((struct ifg_req *)(uintptr_t)
1602 ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1603 #endif
1604 return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1605 }
1606
1607 /*
1608 * Stores all groups from an interface in memory pointed to by ifgr.
1609 */
1610 static int
1611 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1612 {
1613 int len, error;
1614 struct ifg_list *ifgl;
1615 struct ifg_req ifgrq, *ifgp;
1616
1617 if (ifgr->ifgr_len == 0) {
1618 IF_ADDR_RLOCK(ifp);
1619 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1620 ifgr->ifgr_len += sizeof(struct ifg_req);
1621 IF_ADDR_RUNLOCK(ifp);
1622 return (0);
1623 }
1624
1625 len = ifgr->ifgr_len;
1626 ifgp = ifgr_groups_get(ifgr);
1627 /* XXX: wire */
1628 IF_ADDR_RLOCK(ifp);
1629 CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1630 if (len < sizeof(ifgrq)) {
1631 IF_ADDR_RUNLOCK(ifp);
1632 return (EINVAL);
1633 }
1634 bzero(&ifgrq, sizeof ifgrq);
1635 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1636 sizeof(ifgrq.ifgrq_group));
1637 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1638 IF_ADDR_RUNLOCK(ifp);
1639 return (error);
1640 }
1641 len -= sizeof(ifgrq);
1642 ifgp++;
1643 }
1644 IF_ADDR_RUNLOCK(ifp);
1645
1646 return (0);
1647 }
1648
1649 /*
1650 * Stores all members of a group in memory pointed to by igfr
1651 */
1652 static int
1653 if_getgroupmembers(struct ifgroupreq *ifgr)
1654 {
1655 struct ifg_group *ifg;
1656 struct ifg_member *ifgm;
1657 struct ifg_req ifgrq, *ifgp;
1658 int len, error;
1659
1660 IFNET_RLOCK();
1661 CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1662 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1663 break;
1664 if (ifg == NULL) {
1665 IFNET_RUNLOCK();
1666 return (ENOENT);
1667 }
1668
1669 if (ifgr->ifgr_len == 0) {
1670 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1671 ifgr->ifgr_len += sizeof(ifgrq);
1672 IFNET_RUNLOCK();
1673 return (0);
1674 }
1675
1676 len = ifgr->ifgr_len;
1677 ifgp = ifgr_groups_get(ifgr);
1678 CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1679 if (len < sizeof(ifgrq)) {
1680 IFNET_RUNLOCK();
1681 return (EINVAL);
1682 }
1683 bzero(&ifgrq, sizeof ifgrq);
1684 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1685 sizeof(ifgrq.ifgrq_member));
1686 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1687 IFNET_RUNLOCK();
1688 return (error);
1689 }
1690 len -= sizeof(ifgrq);
1691 ifgp++;
1692 }
1693 IFNET_RUNLOCK();
1694
1695 return (0);
1696 }
1697
1698 /*
1699 * Return counter values from counter(9)s stored in ifnet.
1700 */
1701 uint64_t
1702 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1703 {
1704
1705 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1706
1707 return (counter_u64_fetch(ifp->if_counters[cnt]));
1708 }
1709
1710 /*
1711 * Increase an ifnet counter. Usually used for counters shared
1712 * between the stack and a driver, but function supports them all.
1713 */
1714 void
1715 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1716 {
1717
1718 KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1719
1720 counter_u64_add(ifp->if_counters[cnt], inc);
1721 }
1722
1723 /*
1724 * Copy data from ifnet to userland API structure if_data.
1725 */
1726 void
1727 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1728 {
1729
1730 ifd->ifi_type = ifp->if_type;
1731 ifd->ifi_physical = 0;
1732 ifd->ifi_addrlen = ifp->if_addrlen;
1733 ifd->ifi_hdrlen = ifp->if_hdrlen;
1734 ifd->ifi_link_state = ifp->if_link_state;
1735 ifd->ifi_vhid = 0;
1736 ifd->ifi_datalen = sizeof(struct if_data);
1737 ifd->ifi_mtu = ifp->if_mtu;
1738 ifd->ifi_metric = ifp->if_metric;
1739 ifd->ifi_baudrate = ifp->if_baudrate;
1740 ifd->ifi_hwassist = ifp->if_hwassist;
1741 ifd->ifi_epoch = ifp->if_epoch;
1742 ifd->ifi_lastchange = ifp->if_lastchange;
1743
1744 ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1745 ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1746 ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1747 ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1748 ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1749 ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1750 ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1751 ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1752 ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1753 ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1754 ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1755 ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1756 }
1757
1758 /*
1759 * Wrapper functions for struct ifnet address list locking macros. These are
1760 * used by kernel modules to avoid encoding programming interface or binary
1761 * interface assumptions that may be violated when kernel-internal locking
1762 * approaches change.
1763 */
1764 void
1765 if_addr_rlock(struct ifnet *ifp)
1766 {
1767 MPASS(*(uint64_t *)&ifp->if_addr_et == 0);
1768 epoch_enter_preempt(net_epoch_preempt, &ifp->if_addr_et);
1769 }
1770
1771 void
1772 if_addr_runlock(struct ifnet *ifp)
1773 {
1774 epoch_exit_preempt(net_epoch_preempt, &ifp->if_addr_et);
1775 #ifdef INVARIANTS
1776 bzero(&ifp->if_addr_et, sizeof(struct epoch_tracker));
1777 #endif
1778 }
1779
1780 void
1781 if_maddr_rlock(if_t ifp)
1782 {
1783
1784 MPASS(*(uint64_t *)&ifp->if_maddr_et == 0);
1785 epoch_enter_preempt(net_epoch_preempt, &ifp->if_maddr_et);
1786 }
1787
1788 void
1789 if_maddr_runlock(if_t ifp)
1790 {
1791
1792 epoch_exit_preempt(net_epoch_preempt, &ifp->if_maddr_et);
1793 #ifdef INVARIANTS
1794 bzero(&ifp->if_maddr_et, sizeof(struct epoch_tracker));
1795 #endif
1796 }
1797
1798 /*
1799 * Initialization, destruction and refcounting functions for ifaddrs.
1800 */
1801 struct ifaddr *
1802 ifa_alloc(size_t size, int flags)
1803 {
1804 struct ifaddr *ifa;
1805
1806 KASSERT(size >= sizeof(struct ifaddr),
1807 ("%s: invalid size %zu", __func__, size));
1808
1809 ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1810 if (ifa == NULL)
1811 return (NULL);
1812
1813 if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1814 goto fail;
1815 if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1816 goto fail;
1817 if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1818 goto fail;
1819 if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1820 goto fail;
1821
1822 refcount_init(&ifa->ifa_refcnt, 1);
1823
1824 return (ifa);
1825
1826 fail:
1827 /* free(NULL) is okay */
1828 counter_u64_free(ifa->ifa_opackets);
1829 counter_u64_free(ifa->ifa_ipackets);
1830 counter_u64_free(ifa->ifa_obytes);
1831 counter_u64_free(ifa->ifa_ibytes);
1832 free(ifa, M_IFADDR);
1833
1834 return (NULL);
1835 }
1836
1837 void
1838 ifa_ref(struct ifaddr *ifa)
1839 {
1840
1841 refcount_acquire(&ifa->ifa_refcnt);
1842 }
1843
1844 static void
1845 ifa_destroy(epoch_context_t ctx)
1846 {
1847 struct ifaddr *ifa;
1848
1849 ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1850 counter_u64_free(ifa->ifa_opackets);
1851 counter_u64_free(ifa->ifa_ipackets);
1852 counter_u64_free(ifa->ifa_obytes);
1853 counter_u64_free(ifa->ifa_ibytes);
1854 free(ifa, M_IFADDR);
1855 }
1856
1857 void
1858 ifa_free(struct ifaddr *ifa)
1859 {
1860
1861 if (refcount_release(&ifa->ifa_refcnt))
1862 epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy);
1863 }
1864
1865
1866 static int
1867 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1868 struct sockaddr *ia)
1869 {
1870 int error;
1871 struct rt_addrinfo info;
1872 struct sockaddr_dl null_sdl;
1873 struct ifnet *ifp;
1874
1875 ifp = ifa->ifa_ifp;
1876
1877 bzero(&info, sizeof(info));
1878 if (cmd != RTM_DELETE)
1879 info.rti_ifp = V_loif;
1880 info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1881 info.rti_info[RTAX_DST] = ia;
1882 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1883 link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1884
1885 error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1886
1887 if (error != 0 &&
1888 !(cmd == RTM_ADD && error == EEXIST) &&
1889 !(cmd == RTM_DELETE && error == ENOENT))
1890 if_printf(ifp, "%s failed: %d\n", otype, error);
1891
1892 return (error);
1893 }
1894
1895 int
1896 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1897 {
1898
1899 return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1900 }
1901
1902 int
1903 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1904 {
1905
1906 return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1907 }
1908
1909 int
1910 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1911 {
1912
1913 return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1914 }
1915
1916 /*
1917 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1918 * structs used to represent other address families, it is necessary
1919 * to perform a different comparison.
1920 */
1921
1922 #define sa_dl_equal(a1, a2) \
1923 ((((const struct sockaddr_dl *)(a1))->sdl_len == \
1924 ((const struct sockaddr_dl *)(a2))->sdl_len) && \
1925 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \
1926 CLLADDR((const struct sockaddr_dl *)(a2)), \
1927 ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1928
1929 /*
1930 * Locate an interface based on a complete address.
1931 */
1932 /*ARGSUSED*/
1933 struct ifaddr *
1934 ifa_ifwithaddr(const struct sockaddr *addr)
1935 {
1936 struct ifnet *ifp;
1937 struct ifaddr *ifa;
1938
1939 MPASS(in_epoch(net_epoch_preempt));
1940 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1941 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1942 if (ifa->ifa_addr->sa_family != addr->sa_family)
1943 continue;
1944 if (sa_equal(addr, ifa->ifa_addr)) {
1945 goto done;
1946 }
1947 /* IP6 doesn't have broadcast */
1948 if ((ifp->if_flags & IFF_BROADCAST) &&
1949 ifa->ifa_broadaddr &&
1950 ifa->ifa_broadaddr->sa_len != 0 &&
1951 sa_equal(ifa->ifa_broadaddr, addr)) {
1952 goto done;
1953 }
1954 }
1955 }
1956 ifa = NULL;
1957 done:
1958 return (ifa);
1959 }
1960
1961 int
1962 ifa_ifwithaddr_check(const struct sockaddr *addr)
1963 {
1964 int rc;
1965
1966 NET_EPOCH_ENTER();
1967 rc = (ifa_ifwithaddr(addr) != NULL);
1968 NET_EPOCH_EXIT();
1969 return (rc);
1970 }
1971
1972 /*
1973 * Locate an interface based on the broadcast address.
1974 */
1975 /* ARGSUSED */
1976 struct ifaddr *
1977 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1978 {
1979 struct ifnet *ifp;
1980 struct ifaddr *ifa;
1981
1982 MPASS(in_epoch(net_epoch_preempt));
1983 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1984 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1985 continue;
1986 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1987 if (ifa->ifa_addr->sa_family != addr->sa_family)
1988 continue;
1989 if ((ifp->if_flags & IFF_BROADCAST) &&
1990 ifa->ifa_broadaddr &&
1991 ifa->ifa_broadaddr->sa_len != 0 &&
1992 sa_equal(ifa->ifa_broadaddr, addr)) {
1993 goto done;
1994 }
1995 }
1996 }
1997 ifa = NULL;
1998 done:
1999 return (ifa);
2000 }
2001
2002 /*
2003 * Locate the point to point interface with a given destination address.
2004 */
2005 /*ARGSUSED*/
2006 struct ifaddr *
2007 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
2008 {
2009 struct ifnet *ifp;
2010 struct ifaddr *ifa;
2011
2012 MPASS(in_epoch(net_epoch_preempt));
2013 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2014 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2015 continue;
2016 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2017 continue;
2018 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2019 if (ifa->ifa_addr->sa_family != addr->sa_family)
2020 continue;
2021 if (ifa->ifa_dstaddr != NULL &&
2022 sa_equal(addr, ifa->ifa_dstaddr)) {
2023 goto done;
2024 }
2025 }
2026 }
2027 ifa = NULL;
2028 done:
2029 return (ifa);
2030 }
2031
2032 /*
2033 * Find an interface on a specific network. If many, choice
2034 * is most specific found.
2035 */
2036 struct ifaddr *
2037 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2038 {
2039 struct ifnet *ifp;
2040 struct ifaddr *ifa;
2041 struct ifaddr *ifa_maybe = NULL;
2042 u_int af = addr->sa_family;
2043 const char *addr_data = addr->sa_data, *cplim;
2044
2045 MPASS(in_epoch(net_epoch_preempt));
2046 /*
2047 * AF_LINK addresses can be looked up directly by their index number,
2048 * so do that if we can.
2049 */
2050 if (af == AF_LINK) {
2051 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2052 if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2053 return (ifaddr_byindex(sdl->sdl_index));
2054 }
2055
2056 /*
2057 * Scan though each interface, looking for ones that have addresses
2058 * in this address family and the requested fib. Maintain a reference
2059 * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
2060 * kept it stable when we move onto the next interface.
2061 */
2062 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2063 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2064 continue;
2065 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2066 const char *cp, *cp2, *cp3;
2067
2068 if (ifa->ifa_addr->sa_family != af)
2069 next: continue;
2070 if (af == AF_INET &&
2071 ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2072 /*
2073 * This is a bit broken as it doesn't
2074 * take into account that the remote end may
2075 * be a single node in the network we are
2076 * looking for.
2077 * The trouble is that we don't know the
2078 * netmask for the remote end.
2079 */
2080 if (ifa->ifa_dstaddr != NULL &&
2081 sa_equal(addr, ifa->ifa_dstaddr)) {
2082 goto done;
2083 }
2084 } else {
2085 /*
2086 * Scan all the bits in the ifa's address.
2087 * If a bit dissagrees with what we are
2088 * looking for, mask it with the netmask
2089 * to see if it really matters.
2090 * (A byte at a time)
2091 */
2092 if (ifa->ifa_netmask == 0)
2093 continue;
2094 cp = addr_data;
2095 cp2 = ifa->ifa_addr->sa_data;
2096 cp3 = ifa->ifa_netmask->sa_data;
2097 cplim = ifa->ifa_netmask->sa_len
2098 + (char *)ifa->ifa_netmask;
2099 while (cp3 < cplim)
2100 if ((*cp++ ^ *cp2++) & *cp3++)
2101 goto next; /* next address! */
2102 /*
2103 * If the netmask of what we just found
2104 * is more specific than what we had before
2105 * (if we had one), or if the virtual status
2106 * of new prefix is better than of the old one,
2107 * then remember the new one before continuing
2108 * to search for an even better one.
2109 */
2110 if (ifa_maybe == NULL ||
2111 ifa_preferred(ifa_maybe, ifa) ||
2112 rn_refines((caddr_t)ifa->ifa_netmask,
2113 (caddr_t)ifa_maybe->ifa_netmask)) {
2114 ifa_maybe = ifa;
2115 }
2116 }
2117 }
2118 }
2119 ifa = ifa_maybe;
2120 ifa_maybe = NULL;
2121 done:
2122 return (ifa);
2123 }
2124
2125 /*
2126 * Find an interface address specific to an interface best matching
2127 * a given address.
2128 */
2129 struct ifaddr *
2130 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2131 {
2132 struct ifaddr *ifa;
2133 const char *cp, *cp2, *cp3;
2134 char *cplim;
2135 struct ifaddr *ifa_maybe = NULL;
2136 u_int af = addr->sa_family;
2137
2138 if (af >= AF_MAX)
2139 return (NULL);
2140
2141 MPASS(in_epoch(net_epoch_preempt));
2142 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2143 if (ifa->ifa_addr->sa_family != af)
2144 continue;
2145 if (ifa_maybe == NULL)
2146 ifa_maybe = ifa;
2147 if (ifa->ifa_netmask == 0) {
2148 if (sa_equal(addr, ifa->ifa_addr) ||
2149 (ifa->ifa_dstaddr &&
2150 sa_equal(addr, ifa->ifa_dstaddr)))
2151 goto done;
2152 continue;
2153 }
2154 if (ifp->if_flags & IFF_POINTOPOINT) {
2155 if (sa_equal(addr, ifa->ifa_dstaddr))
2156 goto done;
2157 } else {
2158 cp = addr->sa_data;
2159 cp2 = ifa->ifa_addr->sa_data;
2160 cp3 = ifa->ifa_netmask->sa_data;
2161 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2162 for (; cp3 < cplim; cp3++)
2163 if ((*cp++ ^ *cp2++) & *cp3)
2164 break;
2165 if (cp3 == cplim)
2166 goto done;
2167 }
2168 }
2169 ifa = ifa_maybe;
2170 done:
2171 return (ifa);
2172 }
2173
2174 /*
2175 * See whether new ifa is better than current one:
2176 * 1) A non-virtual one is preferred over virtual.
2177 * 2) A virtual in master state preferred over any other state.
2178 *
2179 * Used in several address selecting functions.
2180 */
2181 int
2182 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2183 {
2184
2185 return (cur->ifa_carp && (!next->ifa_carp ||
2186 ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2187 }
2188
2189 #include <net/if_llatbl.h>
2190
2191 /*
2192 * Default action when installing a route with a Link Level gateway.
2193 * Lookup an appropriate real ifa to point to.
2194 * This should be moved to /sys/net/link.c eventually.
2195 */
2196 static void
2197 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
2198 {
2199 struct ifaddr *ifa, *oifa;
2200 struct sockaddr *dst;
2201 struct ifnet *ifp;
2202
2203 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
2204 ((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
2205 return;
2206 NET_EPOCH_ENTER();
2207 ifa = ifaof_ifpforaddr(dst, ifp);
2208 if (ifa) {
2209 oifa = rt->rt_ifa;
2210 if (oifa != ifa) {
2211 ifa_free(oifa);
2212 ifa_ref(ifa);
2213 }
2214 rt->rt_ifa = ifa;
2215 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
2216 ifa->ifa_rtrequest(cmd, rt, info);
2217 }
2218 NET_EPOCH_EXIT();
2219 }
2220
2221 struct sockaddr_dl *
2222 link_alloc_sdl(size_t size, int flags)
2223 {
2224
2225 return (malloc(size, M_TEMP, flags));
2226 }
2227
2228 void
2229 link_free_sdl(struct sockaddr *sa)
2230 {
2231 free(sa, M_TEMP);
2232 }
2233
2234 /*
2235 * Fills in given sdl with interface basic info.
2236 * Returns pointer to filled sdl.
2237 */
2238 struct sockaddr_dl *
2239 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2240 {
2241 struct sockaddr_dl *sdl;
2242
2243 sdl = (struct sockaddr_dl *)paddr;
2244 memset(sdl, 0, sizeof(struct sockaddr_dl));
2245 sdl->sdl_len = sizeof(struct sockaddr_dl);
2246 sdl->sdl_family = AF_LINK;
2247 sdl->sdl_index = ifp->if_index;
2248 sdl->sdl_type = iftype;
2249
2250 return (sdl);
2251 }
2252
2253 /*
2254 * Mark an interface down and notify protocols of
2255 * the transition.
2256 */
2257 static void
2258 if_unroute(struct ifnet *ifp, int flag, int fam)
2259 {
2260 struct ifaddr *ifa;
2261
2262 KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2263
2264 ifp->if_flags &= ~flag;
2265 getmicrotime(&ifp->if_lastchange);
2266 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2267 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2268 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2269 ifp->if_qflush(ifp);
2270
2271 if (ifp->if_carp)
2272 (*carp_linkstate_p)(ifp);
2273 rt_ifmsg(ifp);
2274 }
2275
2276 /*
2277 * Mark an interface up and notify protocols of
2278 * the transition.
2279 */
2280 static void
2281 if_route(struct ifnet *ifp, int flag, int fam)
2282 {
2283 struct ifaddr *ifa;
2284
2285 KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2286
2287 ifp->if_flags |= flag;
2288 getmicrotime(&ifp->if_lastchange);
2289 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2290 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2291 pfctlinput(PRC_IFUP, ifa->ifa_addr);
2292 if (ifp->if_carp)
2293 (*carp_linkstate_p)(ifp);
2294 rt_ifmsg(ifp);
2295 #ifdef INET6
2296 in6_if_up(ifp);
2297 #endif
2298 }
2299
2300 void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
2301 void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
2302 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2303 struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2304 int (*vlan_tag_p)(struct ifnet *, uint16_t *);
2305 int (*vlan_pcp_p)(struct ifnet *, uint16_t *);
2306 int (*vlan_setcookie_p)(struct ifnet *, void *);
2307 void *(*vlan_cookie_p)(struct ifnet *);
2308
2309 /*
2310 * Handle a change in the interface link state. To avoid LORs
2311 * between driver lock and upper layer locks, as well as possible
2312 * recursions, we post event to taskqueue, and all job
2313 * is done in static do_link_state_change().
2314 */
2315 void
2316 if_link_state_change(struct ifnet *ifp, int link_state)
2317 {
2318 /* Return if state hasn't changed. */
2319 if (ifp->if_link_state == link_state)
2320 return;
2321
2322 ifp->if_link_state = link_state;
2323
2324 taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2325 }
2326
2327 static void
2328 do_link_state_change(void *arg, int pending)
2329 {
2330 struct ifnet *ifp = (struct ifnet *)arg;
2331 int link_state = ifp->if_link_state;
2332 CURVNET_SET(ifp->if_vnet);
2333
2334 /* Notify that the link state has changed. */
2335 rt_ifmsg(ifp);
2336 if (ifp->if_vlantrunk != NULL)
2337 (*vlan_link_state_p)(ifp);
2338
2339 if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2340 ifp->if_l2com != NULL)
2341 (*ng_ether_link_state_p)(ifp, link_state);
2342 if (ifp->if_carp)
2343 (*carp_linkstate_p)(ifp);
2344 if (ifp->if_bridge)
2345 ifp->if_bridge_linkstate(ifp);
2346 if (ifp->if_lagg)
2347 (*lagg_linkstate_p)(ifp, link_state);
2348
2349 if (IS_DEFAULT_VNET(curvnet))
2350 devctl_notify("IFNET", ifp->if_xname,
2351 (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2352 NULL);
2353 if (pending > 1)
2354 if_printf(ifp, "%d link states coalesced\n", pending);
2355 if (log_link_state_change)
2356 if_printf(ifp, "link state changed to %s\n",
2357 (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2358 EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2359 CURVNET_RESTORE();
2360 }
2361
2362 /*
2363 * Mark an interface down and notify protocols of
2364 * the transition.
2365 */
2366 void
2367 if_down(struct ifnet *ifp)
2368 {
2369
2370 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2371 if_unroute(ifp, IFF_UP, AF_UNSPEC);
2372 }
2373
2374 /*
2375 * Mark an interface up and notify protocols of
2376 * the transition.
2377 */
2378 void
2379 if_up(struct ifnet *ifp)
2380 {
2381
2382 if_route(ifp, IFF_UP, AF_UNSPEC);
2383 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2384 }
2385
2386 /*
2387 * Flush an interface queue.
2388 */
2389 void
2390 if_qflush(struct ifnet *ifp)
2391 {
2392 struct mbuf *m, *n;
2393 struct ifaltq *ifq;
2394
2395 ifq = &ifp->if_snd;
2396 IFQ_LOCK(ifq);
2397 #ifdef ALTQ
2398 if (ALTQ_IS_ENABLED(ifq))
2399 ALTQ_PURGE(ifq);
2400 #endif
2401 n = ifq->ifq_head;
2402 while ((m = n) != NULL) {
2403 n = m->m_nextpkt;
2404 m_freem(m);
2405 }
2406 ifq->ifq_head = 0;
2407 ifq->ifq_tail = 0;
2408 ifq->ifq_len = 0;
2409 IFQ_UNLOCK(ifq);
2410 }
2411
2412 /*
2413 * Map interface name to interface structure pointer, with or without
2414 * returning a reference.
2415 */
2416 struct ifnet *
2417 ifunit_ref(const char *name)
2418 {
2419 struct ifnet *ifp;
2420
2421 IFNET_RLOCK_NOSLEEP();
2422 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2423 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2424 !(ifp->if_flags & IFF_DYING))
2425 break;
2426 }
2427 if (ifp != NULL)
2428 if_ref(ifp);
2429 IFNET_RUNLOCK_NOSLEEP();
2430 return (ifp);
2431 }
2432
2433 struct ifnet *
2434 ifunit(const char *name)
2435 {
2436 struct ifnet *ifp;
2437
2438 IFNET_RLOCK_NOSLEEP();
2439 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2440 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2441 break;
2442 }
2443 IFNET_RUNLOCK_NOSLEEP();
2444 return (ifp);
2445 }
2446
2447 static void *
2448 ifr_buffer_get_buffer(void *data)
2449 {
2450 union ifreq_union *ifrup;
2451
2452 ifrup = data;
2453 #ifdef COMPAT_FREEBSD32
2454 if (SV_CURPROC_FLAG(SV_ILP32))
2455 return ((void *)(uintptr_t)
2456 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2457 #endif
2458 return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2459 }
2460
2461 static void
2462 ifr_buffer_set_buffer_null(void *data)
2463 {
2464 union ifreq_union *ifrup;
2465
2466 ifrup = data;
2467 #ifdef COMPAT_FREEBSD32
2468 if (SV_CURPROC_FLAG(SV_ILP32))
2469 ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2470 else
2471 #endif
2472 ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2473 }
2474
2475 static size_t
2476 ifr_buffer_get_length(void *data)
2477 {
2478 union ifreq_union *ifrup;
2479
2480 ifrup = data;
2481 #ifdef COMPAT_FREEBSD32
2482 if (SV_CURPROC_FLAG(SV_ILP32))
2483 return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2484 #endif
2485 return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2486 }
2487
2488 static void
2489 ifr_buffer_set_length(void *data, size_t len)
2490 {
2491 union ifreq_union *ifrup;
2492
2493 ifrup = data;
2494 #ifdef COMPAT_FREEBSD32
2495 if (SV_CURPROC_FLAG(SV_ILP32))
2496 ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2497 else
2498 #endif
2499 ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2500 }
2501
2502 void *
2503 ifr_data_get_ptr(void *ifrp)
2504 {
2505 union ifreq_union *ifrup;
2506
2507 ifrup = ifrp;
2508 #ifdef COMPAT_FREEBSD32
2509 if (SV_CURPROC_FLAG(SV_ILP32))
2510 return ((void *)(uintptr_t)
2511 ifrup->ifr32.ifr_ifru.ifru_data);
2512 #endif
2513 return (ifrup->ifr.ifr_ifru.ifru_data);
2514 }
2515
2516 /*
2517 * Hardware specific interface ioctls.
2518 */
2519 int
2520 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2521 {
2522 struct ifreq *ifr;
2523 int error = 0, do_ifup = 0;
2524 int new_flags, temp_flags;
2525 size_t namelen, onamelen;
2526 size_t descrlen;
2527 char *descrbuf, *odescrbuf;
2528 char new_name[IFNAMSIZ];
2529 struct ifaddr *ifa;
2530 struct sockaddr_dl *sdl;
2531
2532 ifr = (struct ifreq *)data;
2533 switch (cmd) {
2534 case SIOCGIFINDEX:
2535 ifr->ifr_index = ifp->if_index;
2536 break;
2537
2538 case SIOCGIFFLAGS:
2539 temp_flags = ifp->if_flags | ifp->if_drv_flags;
2540 ifr->ifr_flags = temp_flags & 0xffff;
2541 ifr->ifr_flagshigh = temp_flags >> 16;
2542 break;
2543
2544 case SIOCGIFCAP:
2545 ifr->ifr_reqcap = ifp->if_capabilities;
2546 ifr->ifr_curcap = ifp->if_capenable;
2547 break;
2548
2549 #ifdef MAC
2550 case SIOCGIFMAC:
2551 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2552 break;
2553 #endif
2554
2555 case SIOCGIFMETRIC:
2556 ifr->ifr_metric = ifp->if_metric;
2557 break;
2558
2559 case SIOCGIFMTU:
2560 ifr->ifr_mtu = ifp->if_mtu;
2561 break;
2562
2563 case SIOCGIFPHYS:
2564 /* XXXGL: did this ever worked? */
2565 ifr->ifr_phys = 0;
2566 break;
2567
2568 case SIOCGIFDESCR:
2569 error = 0;
2570 sx_slock(&ifdescr_sx);
2571 if (ifp->if_description == NULL)
2572 error = ENOMSG;
2573 else {
2574 /* space for terminating nul */
2575 descrlen = strlen(ifp->if_description) + 1;
2576 if (ifr_buffer_get_length(ifr) < descrlen)
2577 ifr_buffer_set_buffer_null(ifr);
2578 else
2579 error = copyout(ifp->if_description,
2580 ifr_buffer_get_buffer(ifr), descrlen);
2581 ifr_buffer_set_length(ifr, descrlen);
2582 }
2583 sx_sunlock(&ifdescr_sx);
2584 break;
2585
2586 case SIOCSIFDESCR:
2587 error = priv_check(td, PRIV_NET_SETIFDESCR);
2588 if (error)
2589 return (error);
2590
2591 /*
2592 * Copy only (length-1) bytes to make sure that
2593 * if_description is always nul terminated. The
2594 * length parameter is supposed to count the
2595 * terminating nul in.
2596 */
2597 if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2598 return (ENAMETOOLONG);
2599 else if (ifr_buffer_get_length(ifr) == 0)
2600 descrbuf = NULL;
2601 else {
2602 descrbuf = malloc(ifr_buffer_get_length(ifr),
2603 M_IFDESCR, M_WAITOK | M_ZERO);
2604 error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2605 ifr_buffer_get_length(ifr) - 1);
2606 if (error) {
2607 free(descrbuf, M_IFDESCR);
2608 break;
2609 }
2610 }
2611
2612 sx_xlock(&ifdescr_sx);
2613 odescrbuf = ifp->if_description;
2614 ifp->if_description = descrbuf;
2615 sx_xunlock(&ifdescr_sx);
2616
2617 getmicrotime(&ifp->if_lastchange);
2618 free(odescrbuf, M_IFDESCR);
2619 break;
2620
2621 case SIOCGIFFIB:
2622 ifr->ifr_fib = ifp->if_fib;
2623 break;
2624
2625 case SIOCSIFFIB:
2626 error = priv_check(td, PRIV_NET_SETIFFIB);
2627 if (error)
2628 return (error);
2629 if (ifr->ifr_fib >= rt_numfibs)
2630 return (EINVAL);
2631
2632 ifp->if_fib = ifr->ifr_fib;
2633 break;
2634
2635 case SIOCSIFFLAGS:
2636 error = priv_check(td, PRIV_NET_SETIFFLAGS);
2637 if (error)
2638 return (error);
2639 /*
2640 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2641 * check, so we don't need special handling here yet.
2642 */
2643 new_flags = (ifr->ifr_flags & 0xffff) |
2644 (ifr->ifr_flagshigh << 16);
2645 if (ifp->if_flags & IFF_UP &&
2646 (new_flags & IFF_UP) == 0) {
2647 if_down(ifp);
2648 } else if (new_flags & IFF_UP &&
2649 (ifp->if_flags & IFF_UP) == 0) {
2650 do_ifup = 1;
2651 }
2652 /* See if permanently promiscuous mode bit is about to flip */
2653 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2654 if (new_flags & IFF_PPROMISC)
2655 ifp->if_flags |= IFF_PROMISC;
2656 else if (ifp->if_pcount == 0)
2657 ifp->if_flags &= ~IFF_PROMISC;
2658 if (log_promisc_mode_change)
2659 if_printf(ifp, "permanently promiscuous mode %s\n",
2660 ((new_flags & IFF_PPROMISC) ?
2661 "enabled" : "disabled"));
2662 }
2663 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2664 (new_flags &~ IFF_CANTCHANGE);
2665 if (ifp->if_ioctl) {
2666 (void) (*ifp->if_ioctl)(ifp, cmd, data);
2667 }
2668 if (do_ifup)
2669 if_up(ifp);
2670 getmicrotime(&ifp->if_lastchange);
2671 break;
2672
2673 case SIOCSIFCAP:
2674 error = priv_check(td, PRIV_NET_SETIFCAP);
2675 if (error)
2676 return (error);
2677 if (ifp->if_ioctl == NULL)
2678 return (EOPNOTSUPP);
2679 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2680 return (EINVAL);
2681 error = (*ifp->if_ioctl)(ifp, cmd, data);
2682 if (error == 0)
2683 getmicrotime(&ifp->if_lastchange);
2684 break;
2685
2686 #ifdef MAC
2687 case SIOCSIFMAC:
2688 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2689 break;
2690 #endif
2691
2692 case SIOCSIFNAME:
2693 error = priv_check(td, PRIV_NET_SETIFNAME);
2694 if (error)
2695 return (error);
2696 error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2697 NULL);
2698 if (error != 0)
2699 return (error);
2700 if (new_name[0] == '\0')
2701 return (EINVAL);
2702 if (new_name[IFNAMSIZ-1] != '\0') {
2703 new_name[IFNAMSIZ-1] = '\0';
2704 if (strlen(new_name) == IFNAMSIZ-1)
2705 return (EINVAL);
2706 }
2707 if (ifunit(new_name) != NULL)
2708 return (EEXIST);
2709
2710 /*
2711 * XXX: Locking. Nothing else seems to lock if_flags,
2712 * and there are numerous other races with the
2713 * ifunit() checks not being atomic with namespace
2714 * changes (renames, vmoves, if_attach, etc).
2715 */
2716 ifp->if_flags |= IFF_RENAMING;
2717
2718 /* Announce the departure of the interface. */
2719 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2720 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2721
2722 if_printf(ifp, "changing name to '%s'\n", new_name);
2723
2724 IF_ADDR_WLOCK(ifp);
2725 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2726 ifa = ifp->if_addr;
2727 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2728 namelen = strlen(new_name);
2729 onamelen = sdl->sdl_nlen;
2730 /*
2731 * Move the address if needed. This is safe because we
2732 * allocate space for a name of length IFNAMSIZ when we
2733 * create this in if_attach().
2734 */
2735 if (namelen != onamelen) {
2736 bcopy(sdl->sdl_data + onamelen,
2737 sdl->sdl_data + namelen, sdl->sdl_alen);
2738 }
2739 bcopy(new_name, sdl->sdl_data, namelen);
2740 sdl->sdl_nlen = namelen;
2741 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2742 bzero(sdl->sdl_data, onamelen);
2743 while (namelen != 0)
2744 sdl->sdl_data[--namelen] = 0xff;
2745 IF_ADDR_WUNLOCK(ifp);
2746
2747 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2748 /* Announce the return of the interface. */
2749 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2750
2751 ifp->if_flags &= ~IFF_RENAMING;
2752 break;
2753
2754 #ifdef VIMAGE
2755 case SIOCSIFVNET:
2756 error = priv_check(td, PRIV_NET_SETIFVNET);
2757 if (error)
2758 return (error);
2759 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2760 break;
2761 #endif
2762
2763 case SIOCSIFMETRIC:
2764 error = priv_check(td, PRIV_NET_SETIFMETRIC);
2765 if (error)
2766 return (error);
2767 ifp->if_metric = ifr->ifr_metric;
2768 getmicrotime(&ifp->if_lastchange);
2769 break;
2770
2771 case SIOCSIFPHYS:
2772 error = priv_check(td, PRIV_NET_SETIFPHYS);
2773 if (error)
2774 return (error);
2775 if (ifp->if_ioctl == NULL)
2776 return (EOPNOTSUPP);
2777 error = (*ifp->if_ioctl)(ifp, cmd, data);
2778 if (error == 0)
2779 getmicrotime(&ifp->if_lastchange);
2780 break;
2781
2782 case SIOCSIFMTU:
2783 {
2784 u_long oldmtu = ifp->if_mtu;
2785
2786 error = priv_check(td, PRIV_NET_SETIFMTU);
2787 if (error)
2788 return (error);
2789 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2790 return (EINVAL);
2791 if (ifp->if_ioctl == NULL)
2792 return (EOPNOTSUPP);
2793 error = (*ifp->if_ioctl)(ifp, cmd, data);
2794 if (error == 0) {
2795 getmicrotime(&ifp->if_lastchange);
2796 rt_ifmsg(ifp);
2797 #ifdef INET
2798 NETDUMP_REINIT(ifp);
2799 #endif
2800 }
2801 /*
2802 * If the link MTU changed, do network layer specific procedure.
2803 */
2804 if (ifp->if_mtu != oldmtu) {
2805 #ifdef INET6
2806 nd6_setmtu(ifp);
2807 #endif
2808 rt_updatemtu(ifp);
2809 }
2810 break;
2811 }
2812
2813 case SIOCADDMULTI:
2814 case SIOCDELMULTI:
2815 if (cmd == SIOCADDMULTI)
2816 error = priv_check(td, PRIV_NET_ADDMULTI);
2817 else
2818 error = priv_check(td, PRIV_NET_DELMULTI);
2819 if (error)
2820 return (error);
2821
2822 /* Don't allow group membership on non-multicast interfaces. */
2823 if ((ifp->if_flags & IFF_MULTICAST) == 0)
2824 return (EOPNOTSUPP);
2825
2826 /* Don't let users screw up protocols' entries. */
2827 if (ifr->ifr_addr.sa_family != AF_LINK)
2828 return (EINVAL);
2829
2830 if (cmd == SIOCADDMULTI) {
2831 struct ifmultiaddr *ifma;
2832
2833 /*
2834 * Userland is only permitted to join groups once
2835 * via the if_addmulti() KPI, because it cannot hold
2836 * struct ifmultiaddr * between calls. It may also
2837 * lose a race while we check if the membership
2838 * already exists.
2839 */
2840 IF_ADDR_RLOCK(ifp);
2841 ifma = if_findmulti(ifp, &ifr->ifr_addr);
2842 IF_ADDR_RUNLOCK(ifp);
2843 if (ifma != NULL)
2844 error = EADDRINUSE;
2845 else
2846 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2847 } else {
2848 error = if_delmulti(ifp, &ifr->ifr_addr);
2849 }
2850 if (error == 0)
2851 getmicrotime(&ifp->if_lastchange);
2852 break;
2853
2854 case SIOCSIFPHYADDR:
2855 case SIOCDIFPHYADDR:
2856 #ifdef INET6
2857 case SIOCSIFPHYADDR_IN6:
2858 #endif
2859 case SIOCSIFMEDIA:
2860 case SIOCSIFGENERIC:
2861 error = priv_check(td, PRIV_NET_HWIOCTL);
2862 if (error)
2863 return (error);
2864 if (ifp->if_ioctl == NULL)
2865 return (EOPNOTSUPP);
2866 error = (*ifp->if_ioctl)(ifp, cmd, data);
2867 if (error == 0)
2868 getmicrotime(&ifp->if_lastchange);
2869 break;
2870
2871 case SIOCGIFSTATUS:
2872 case SIOCGIFPSRCADDR:
2873 case SIOCGIFPDSTADDR:
2874 case SIOCGIFMEDIA:
2875 case SIOCGIFXMEDIA:
2876 case SIOCGIFGENERIC:
2877 case SIOCGIFRSSKEY:
2878 case SIOCGIFRSSHASH:
2879 if (ifp->if_ioctl == NULL)
2880 return (EOPNOTSUPP);
2881 error = (*ifp->if_ioctl)(ifp, cmd, data);
2882 break;
2883
2884 case SIOCSIFLLADDR:
2885 error = priv_check(td, PRIV_NET_SETLLADDR);
2886 if (error)
2887 return (error);
2888 error = if_setlladdr(ifp,
2889 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2890 break;
2891
2892 case SIOCGHWADDR:
2893 error = if_gethwaddr(ifp, ifr);
2894 break;
2895
2896 CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2897 error = priv_check(td, PRIV_NET_ADDIFGROUP);
2898 if (error)
2899 return (error);
2900 if ((error = if_addgroup(ifp,
2901 ifgr_group_get((struct ifgroupreq *)data))))
2902 return (error);
2903 break;
2904
2905 CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2906 if ((error = if_getgroup((struct ifgroupreq *)data, ifp)))
2907 return (error);
2908 break;
2909
2910 CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2911 error = priv_check(td, PRIV_NET_DELIFGROUP);
2912 if (error)
2913 return (error);
2914 if ((error = if_delgroup(ifp,
2915 ifgr_group_get((struct ifgroupreq *)data))))
2916 return (error);
2917 break;
2918
2919 default:
2920 error = ENOIOCTL;
2921 break;
2922 }
2923 return (error);
2924 }
2925
2926 #ifdef COMPAT_FREEBSD32
2927 struct ifconf32 {
2928 int32_t ifc_len;
2929 union {
2930 uint32_t ifcu_buf;
2931 uint32_t ifcu_req;
2932 } ifc_ifcu;
2933 };
2934 #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
2935 #endif
2936
2937 #ifdef COMPAT_FREEBSD32
2938 static void
2939 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2940 {
2941 struct ifmediareq32 *ifmr32;
2942
2943 ifmr32 = (struct ifmediareq32 *)data;
2944 memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2945 sizeof(ifmr->ifm_name));
2946 ifmr->ifm_current = ifmr32->ifm_current;
2947 ifmr->ifm_mask = ifmr32->ifm_mask;
2948 ifmr->ifm_status = ifmr32->ifm_status;
2949 ifmr->ifm_active = ifmr32->ifm_active;
2950 ifmr->ifm_count = ifmr32->ifm_count;
2951 ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2952 }
2953
2954 static void
2955 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2956 {
2957 struct ifmediareq32 *ifmr32;
2958
2959 ifmr32 = (struct ifmediareq32 *)data;
2960 ifmr32->ifm_current = ifmr->ifm_current;
2961 ifmr32->ifm_mask = ifmr->ifm_mask;
2962 ifmr32->ifm_status = ifmr->ifm_status;
2963 ifmr32->ifm_active = ifmr->ifm_active;
2964 ifmr32->ifm_count = ifmr->ifm_count;
2965 }
2966 #endif
2967
2968 /*
2969 * Interface ioctls.
2970 */
2971 int
2972 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2973 {
2974 #ifdef COMPAT_FREEBSD32
2975 caddr_t saved_data = NULL;
2976 struct ifmediareq ifmr;
2977 struct ifmediareq *ifmrp;
2978 #endif
2979 struct ifnet *ifp;
2980 struct ifreq *ifr;
2981 int error;
2982 int oif_flags;
2983 #ifdef VIMAGE
2984 int shutdown;
2985 #endif
2986
2987 CURVNET_SET(so->so_vnet);
2988 #ifdef VIMAGE
2989 /* Make sure the VNET is stable. */
2990 shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2991 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2992 if (shutdown) {
2993 CURVNET_RESTORE();
2994 return (EBUSY);
2995 }
2996 #endif
2997
2998
2999 switch (cmd) {
3000 case SIOCGIFCONF:
3001 error = ifconf(cmd, data);
3002 CURVNET_RESTORE();
3003 return (error);
3004
3005 #ifdef COMPAT_FREEBSD32
3006 case SIOCGIFCONF32:
3007 {
3008 struct ifconf32 *ifc32;
3009 struct ifconf ifc;
3010
3011 ifc32 = (struct ifconf32 *)data;
3012 ifc.ifc_len = ifc32->ifc_len;
3013 ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
3014
3015 error = ifconf(SIOCGIFCONF, (void *)&ifc);
3016 CURVNET_RESTORE();
3017 if (error == 0)
3018 ifc32->ifc_len = ifc.ifc_len;
3019 return (error);
3020 }
3021 #endif
3022 }
3023
3024 #ifdef COMPAT_FREEBSD32
3025 ifmrp = NULL;
3026 switch (cmd) {
3027 case SIOCGIFMEDIA32:
3028 case SIOCGIFXMEDIA32:
3029 ifmrp = &ifmr;
3030 ifmr_init(ifmrp, data);
3031 cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3032 saved_data = data;
3033 data = (caddr_t)ifmrp;
3034 }
3035 #endif
3036
3037 ifr = (struct ifreq *)data;
3038 switch (cmd) {
3039 #ifdef VIMAGE
3040 case SIOCSIFRVNET:
3041 error = priv_check(td, PRIV_NET_SETIFVNET);
3042 if (error == 0)
3043 error = if_vmove_reclaim(td, ifr->ifr_name,
3044 ifr->ifr_jid);
3045 goto out_noref;
3046 #endif
3047 case SIOCIFCREATE:
3048 case SIOCIFCREATE2:
3049 error = priv_check(td, PRIV_NET_IFCREATE);
3050 if (error == 0)
3051 error = if_clone_create(ifr->ifr_name,
3052 sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3053 ifr_data_get_ptr(ifr) : NULL);
3054 goto out_noref;
3055 case SIOCIFDESTROY:
3056 error = priv_check(td, PRIV_NET_IFDESTROY);
3057 if (error == 0)
3058 error = if_clone_destroy(ifr->ifr_name);
3059 goto out_noref;
3060
3061 case SIOCIFGCLONERS:
3062 error = if_clone_list((struct if_clonereq *)data);
3063 goto out_noref;
3064
3065 CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3066 error = if_getgroupmembers((struct ifgroupreq *)data);
3067 goto out_noref;
3068
3069 #if defined(INET) || defined(INET6)
3070 case SIOCSVH:
3071 case SIOCGVH:
3072 if (carp_ioctl_p == NULL)
3073 error = EPROTONOSUPPORT;
3074 else
3075 error = (*carp_ioctl_p)(ifr, cmd, td);
3076 goto out_noref;
3077 #endif
3078 }
3079
3080 ifp = ifunit_ref(ifr->ifr_name);
3081 if (ifp == NULL) {
3082 error = ENXIO;
3083 goto out_noref;
3084 }
3085
3086 error = ifhwioctl(cmd, ifp, data, td);
3087 if (error != ENOIOCTL)
3088 goto out_ref;
3089
3090 oif_flags = ifp->if_flags;
3091 if (so->so_proto == NULL) {
3092 error = EOPNOTSUPP;
3093 goto out_ref;
3094 }
3095
3096 /*
3097 * Pass the request on to the socket control method, and if the
3098 * latter returns EOPNOTSUPP, directly to the interface.
3099 *
3100 * Make an exception for the legacy SIOCSIF* requests. Drivers
3101 * trust SIOCSIFADDR et al to come from an already privileged
3102 * layer, and do not perform any credentials checks or input
3103 * validation.
3104 */
3105 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3106 ifp, td));
3107 if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3108 cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3109 cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3110 error = (*ifp->if_ioctl)(ifp, cmd, data);
3111
3112 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3113 #ifdef INET6
3114 if (ifp->if_flags & IFF_UP)
3115 in6_if_up(ifp);
3116 #endif
3117 }
3118
3119 out_ref:
3120 if_rele(ifp);
3121 out_noref:
3122 #ifdef COMPAT_FREEBSD32
3123 if (ifmrp != NULL) {
3124 KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3125 ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3126 cmd));
3127 data = saved_data;
3128 ifmr_update(ifmrp, data);
3129 }
3130 #endif
3131 CURVNET_RESTORE();
3132 return (error);
3133 }
3134
3135 /*
3136 * The code common to handling reference counted flags,
3137 * e.g., in ifpromisc() and if_allmulti().
3138 * The "pflag" argument can specify a permanent mode flag to check,
3139 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3140 *
3141 * Only to be used on stack-owned flags, not driver-owned flags.
3142 */
3143 static int
3144 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3145 {
3146 struct ifreq ifr;
3147 int error;
3148 int oldflags, oldcount;
3149
3150 /* Sanity checks to catch programming errors */
3151 KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3152 ("%s: setting driver-owned flag %d", __func__, flag));
3153
3154 if (onswitch)
3155 KASSERT(*refcount >= 0,
3156 ("%s: increment negative refcount %d for flag %d",
3157 __func__, *refcount, flag));
3158 else
3159 KASSERT(*refcount > 0,
3160 ("%s: decrement non-positive refcount %d for flag %d",
3161 __func__, *refcount, flag));
3162
3163 /* In case this mode is permanent, just touch refcount */
3164 if (ifp->if_flags & pflag) {
3165 *refcount += onswitch ? 1 : -1;
3166 return (0);
3167 }
3168
3169 /* Save ifnet parameters for if_ioctl() may fail */
3170 oldcount = *refcount;
3171 oldflags = ifp->if_flags;
3172
3173 /*
3174 * See if we aren't the only and touching refcount is enough.
3175 * Actually toggle interface flag if we are the first or last.
3176 */
3177 if (onswitch) {
3178 if ((*refcount)++)
3179 return (0);
3180 ifp->if_flags |= flag;
3181 } else {
3182 if (--(*refcount))
3183 return (0);
3184 ifp->if_flags &= ~flag;
3185 }
3186
3187 /* Call down the driver since we've changed interface flags */
3188 if (ifp->if_ioctl == NULL) {
3189 error = EOPNOTSUPP;
3190 goto recover;
3191 }
3192 ifr.ifr_flags = ifp->if_flags & 0xffff;
3193 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3194 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3195 if (error)
3196 goto recover;
3197 /* Notify userland that interface flags have changed */
3198 rt_ifmsg(ifp);
3199 return (0);
3200
3201 recover:
3202 /* Recover after driver error */
3203 *refcount = oldcount;
3204 ifp->if_flags = oldflags;
3205 return (error);
3206 }
3207
3208 /*
3209 * Set/clear promiscuous mode on interface ifp based on the truth value
3210 * of pswitch. The calls are reference counted so that only the first
3211 * "on" request actually has an effect, as does the final "off" request.
3212 * Results are undefined if the "off" and "on" requests are not matched.
3213 */
3214 int
3215 ifpromisc(struct ifnet *ifp, int pswitch)
3216 {
3217 int error;
3218 int oldflags = ifp->if_flags;
3219
3220 error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3221 &ifp->if_pcount, pswitch);
3222 /* If promiscuous mode status has changed, log a message */
3223 if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3224 log_promisc_mode_change)
3225 if_printf(ifp, "promiscuous mode %s\n",
3226 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3227 return (error);
3228 }
3229
3230 /*
3231 * Return interface configuration
3232 * of system. List may be used
3233 * in later ioctl's (above) to get
3234 * other information.
3235 */
3236 /*ARGSUSED*/
3237 static int
3238 ifconf(u_long cmd, caddr_t data)
3239 {
3240 struct ifconf *ifc = (struct ifconf *)data;
3241 struct ifnet *ifp;
3242 struct ifaddr *ifa;
3243 struct ifreq ifr;
3244 struct sbuf *sb;
3245 int error, full = 0, valid_len, max_len;
3246
3247 /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3248 max_len = MAXPHYS - 1;
3249
3250 /* Prevent hostile input from being able to crash the system */
3251 if (ifc->ifc_len <= 0)
3252 return (EINVAL);
3253
3254 again:
3255 if (ifc->ifc_len <= max_len) {
3256 max_len = ifc->ifc_len;
3257 full = 1;
3258 }
3259 sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3260 max_len = 0;
3261 valid_len = 0;
3262
3263 IFNET_RLOCK();
3264 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3265 int addrs;
3266
3267 /*
3268 * Zero the ifr to make sure we don't disclose the contents
3269 * of the stack.
3270 */
3271 memset(&ifr, 0, sizeof(ifr));
3272
3273 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3274 >= sizeof(ifr.ifr_name)) {
3275 sbuf_delete(sb);
3276 IFNET_RUNLOCK();
3277 return (ENAMETOOLONG);
3278 }
3279
3280 addrs = 0;
3281 IF_ADDR_RLOCK(ifp);
3282 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3283 struct sockaddr *sa = ifa->ifa_addr;
3284
3285 if (prison_if(curthread->td_ucred, sa) != 0)
3286 continue;
3287 addrs++;
3288 if (sa->sa_len <= sizeof(*sa)) {
3289 if (sa->sa_len < sizeof(*sa)) {
3290 memset(&ifr.ifr_ifru.ifru_addr, 0,
3291 sizeof(ifr.ifr_ifru.ifru_addr));
3292 memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3293 sa->sa_len);
3294 } else
3295 ifr.ifr_ifru.ifru_addr = *sa;
3296 sbuf_bcat(sb, &ifr, sizeof(ifr));
3297 max_len += sizeof(ifr);
3298 } else {
3299 sbuf_bcat(sb, &ifr,
3300 offsetof(struct ifreq, ifr_addr));
3301 max_len += offsetof(struct ifreq, ifr_addr);
3302 sbuf_bcat(sb, sa, sa->sa_len);
3303 max_len += sa->sa_len;
3304 }
3305
3306 if (sbuf_error(sb) == 0)
3307 valid_len = sbuf_len(sb);
3308 }
3309 IF_ADDR_RUNLOCK(ifp);
3310 if (addrs == 0) {
3311 sbuf_bcat(sb, &ifr, sizeof(ifr));
3312 max_len += sizeof(ifr);
3313
3314 if (sbuf_error(sb) == 0)
3315 valid_len = sbuf_len(sb);
3316 }
3317 }
3318 IFNET_RUNLOCK();
3319
3320 /*
3321 * If we didn't allocate enough space (uncommon), try again. If
3322 * we have already allocated as much space as we are allowed,
3323 * return what we've got.
3324 */
3325 if (valid_len != max_len && !full) {
3326 sbuf_delete(sb);
3327 goto again;
3328 }
3329
3330 ifc->ifc_len = valid_len;
3331 sbuf_finish(sb);
3332 error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3333 sbuf_delete(sb);
3334 return (error);
3335 }
3336
3337 /*
3338 * Just like ifpromisc(), but for all-multicast-reception mode.
3339 */
3340 int
3341 if_allmulti(struct ifnet *ifp, int onswitch)
3342 {
3343
3344 return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3345 }
3346
3347 struct ifmultiaddr *
3348 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3349 {
3350 struct ifmultiaddr *ifma;
3351
3352 IF_ADDR_LOCK_ASSERT(ifp);
3353
3354 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3355 if (sa->sa_family == AF_LINK) {
3356 if (sa_dl_equal(ifma->ifma_addr, sa))
3357 break;
3358 } else {
3359 if (sa_equal(ifma->ifma_addr, sa))
3360 break;
3361 }
3362 }
3363
3364 return ifma;
3365 }
3366
3367 /*
3368 * Allocate a new ifmultiaddr and initialize based on passed arguments. We
3369 * make copies of passed sockaddrs. The ifmultiaddr will not be added to
3370 * the ifnet multicast address list here, so the caller must do that and
3371 * other setup work (such as notifying the device driver). The reference
3372 * count is initialized to 1.
3373 */
3374 static struct ifmultiaddr *
3375 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3376 int mflags)
3377 {
3378 struct ifmultiaddr *ifma;
3379 struct sockaddr *dupsa;
3380
3381 ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3382 M_ZERO);
3383 if (ifma == NULL)
3384 return (NULL);
3385
3386 dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3387 if (dupsa == NULL) {
3388 free(ifma, M_IFMADDR);
3389 return (NULL);
3390 }
3391 bcopy(sa, dupsa, sa->sa_len);
3392 ifma->ifma_addr = dupsa;
3393
3394 ifma->ifma_ifp = ifp;
3395 ifma->ifma_refcount = 1;
3396 ifma->ifma_protospec = NULL;
3397
3398 if (llsa == NULL) {
3399 ifma->ifma_lladdr = NULL;
3400 return (ifma);
3401 }
3402
3403 dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3404 if (dupsa == NULL) {
3405 free(ifma->ifma_addr, M_IFMADDR);
3406 free(ifma, M_IFMADDR);
3407 return (NULL);
3408 }
3409 bcopy(llsa, dupsa, llsa->sa_len);
3410 ifma->ifma_lladdr = dupsa;
3411
3412 return (ifma);
3413 }
3414
3415 /*
3416 * if_freemulti: free ifmultiaddr structure and possibly attached related
3417 * addresses. The caller is responsible for implementing reference
3418 * counting, notifying the driver, handling routing messages, and releasing
3419 * any dependent link layer state.
3420 */
3421 #ifdef MCAST_VERBOSE
3422 extern void kdb_backtrace(void);
3423 #endif
3424 static void
3425 if_freemulti_internal(struct ifmultiaddr *ifma)
3426 {
3427
3428 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3429 ifma->ifma_refcount));
3430
3431 if (ifma->ifma_lladdr != NULL)
3432 free(ifma->ifma_lladdr, M_IFMADDR);
3433 #ifdef MCAST_VERBOSE
3434 kdb_backtrace();
3435 printf("%s freeing ifma: %p\n", __func__, ifma);
3436 #endif
3437 free(ifma->ifma_addr, M_IFMADDR);
3438 free(ifma, M_IFMADDR);
3439 }
3440
3441 static void
3442 if_destroymulti(epoch_context_t ctx)
3443 {
3444 struct ifmultiaddr *ifma;
3445
3446 ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3447 if_freemulti_internal(ifma);
3448 }
3449
3450 void
3451 if_freemulti(struct ifmultiaddr *ifma)
3452 {
3453 KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3454 ifma->ifma_refcount));
3455
3456 epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti);
3457 }
3458
3459
3460 /*
3461 * Register an additional multicast address with a network interface.
3462 *
3463 * - If the address is already present, bump the reference count on the
3464 * address and return.
3465 * - If the address is not link-layer, look up a link layer address.
3466 * - Allocate address structures for one or both addresses, and attach to the
3467 * multicast address list on the interface. If automatically adding a link
3468 * layer address, the protocol address will own a reference to the link
3469 * layer address, to be freed when it is freed.
3470 * - Notify the network device driver of an addition to the multicast address
3471 * list.
3472 *
3473 * 'sa' points to caller-owned memory with the desired multicast address.
3474 *
3475 * 'retifma' will be used to return a pointer to the resulting multicast
3476 * address reference, if desired.
3477 */
3478 int
3479 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3480 struct ifmultiaddr **retifma)
3481 {
3482 struct ifmultiaddr *ifma, *ll_ifma;
3483 struct sockaddr *llsa;
3484 struct sockaddr_dl sdl;
3485 int error;
3486
3487 #ifdef INET
3488 IN_MULTI_LIST_UNLOCK_ASSERT();
3489 #endif
3490 #ifdef INET6
3491 IN6_MULTI_LIST_UNLOCK_ASSERT();
3492 #endif
3493 /*
3494 * If the address is already present, return a new reference to it;
3495 * otherwise, allocate storage and set up a new address.
3496 */
3497 IF_ADDR_WLOCK(ifp);
3498 ifma = if_findmulti(ifp, sa);
3499 if (ifma != NULL) {
3500 ifma->ifma_refcount++;
3501 if (retifma != NULL)
3502 *retifma = ifma;
3503 IF_ADDR_WUNLOCK(ifp);
3504 return (0);
3505 }
3506
3507 /*
3508 * The address isn't already present; resolve the protocol address
3509 * into a link layer address, and then look that up, bump its
3510 * refcount or allocate an ifma for that also.
3511 * Most link layer resolving functions returns address data which
3512 * fits inside default sockaddr_dl structure. However callback
3513 * can allocate another sockaddr structure, in that case we need to
3514 * free it later.
3515 */
3516 llsa = NULL;
3517 ll_ifma = NULL;
3518 if (ifp->if_resolvemulti != NULL) {
3519 /* Provide called function with buffer size information */
3520 sdl.sdl_len = sizeof(sdl);
3521 llsa = (struct sockaddr *)&sdl;
3522 error = ifp->if_resolvemulti(ifp, &llsa, sa);
3523 if (error)
3524 goto unlock_out;
3525 }
3526
3527 /*
3528 * Allocate the new address. Don't hook it up yet, as we may also
3529 * need to allocate a link layer multicast address.
3530 */
3531 ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3532 if (ifma == NULL) {
3533 error = ENOMEM;
3534 goto free_llsa_out;
3535 }
3536
3537 /*
3538 * If a link layer address is found, we'll need to see if it's
3539 * already present in the address list, or allocate is as well.
3540 * When this block finishes, the link layer address will be on the
3541 * list.
3542 */
3543 if (llsa != NULL) {
3544 ll_ifma = if_findmulti(ifp, llsa);
3545 if (ll_ifma == NULL) {
3546 ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3547 if (ll_ifma == NULL) {
3548 --ifma->ifma_refcount;
3549 if_freemulti(ifma);
3550 error = ENOMEM;
3551 goto free_llsa_out;
3552 }
3553 ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3554 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3555 ifma_link);
3556 } else
3557 ll_ifma->ifma_refcount++;
3558 ifma->ifma_llifma = ll_ifma;
3559 }
3560
3561 /*
3562 * We now have a new multicast address, ifma, and possibly a new or
3563 * referenced link layer address. Add the primary address to the
3564 * ifnet address list.
3565 */
3566 ifma->ifma_flags |= IFMA_F_ENQUEUED;
3567 CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3568
3569 if (retifma != NULL)
3570 *retifma = ifma;
3571
3572 /*
3573 * Must generate the message while holding the lock so that 'ifma'
3574 * pointer is still valid.
3575 */
3576 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3577 IF_ADDR_WUNLOCK(ifp);
3578
3579 /*
3580 * We are certain we have added something, so call down to the
3581 * interface to let them know about it.
3582 */
3583 if (ifp->if_ioctl != NULL) {
3584 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3585 }
3586
3587 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3588 link_free_sdl(llsa);
3589
3590 return (0);
3591
3592 free_llsa_out:
3593 if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3594 link_free_sdl(llsa);
3595
3596 unlock_out:
3597 IF_ADDR_WUNLOCK(ifp);
3598 return (error);
3599 }
3600
3601 /*
3602 * Delete a multicast group membership by network-layer group address.
3603 *
3604 * Returns ENOENT if the entry could not be found. If ifp no longer
3605 * exists, results are undefined. This entry point should only be used
3606 * from subsystems which do appropriate locking to hold ifp for the
3607 * duration of the call.
3608 * Network-layer protocol domains must use if_delmulti_ifma().
3609 */
3610 int
3611 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3612 {
3613 struct ifmultiaddr *ifma;
3614 int lastref;
3615 #ifdef INVARIANTS
3616 struct ifnet *oifp;
3617
3618 IFNET_RLOCK_NOSLEEP();
3619 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3620 if (ifp == oifp)
3621 break;
3622 if (ifp != oifp)
3623 ifp = NULL;
3624 IFNET_RUNLOCK_NOSLEEP();
3625
3626 KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3627 #endif
3628 if (ifp == NULL)
3629 return (ENOENT);
3630
3631 IF_ADDR_WLOCK(ifp);
3632 lastref = 0;
3633 ifma = if_findmulti(ifp, sa);
3634 if (ifma != NULL)
3635 lastref = if_delmulti_locked(ifp, ifma, 0);
3636 IF_ADDR_WUNLOCK(ifp);
3637
3638 if (ifma == NULL)
3639 return (ENOENT);
3640
3641 if (lastref && ifp->if_ioctl != NULL) {
3642 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3643 }
3644
3645 return (0);
3646 }
3647
3648 /*
3649 * Delete all multicast group membership for an interface.
3650 * Should be used to quickly flush all multicast filters.
3651 */
3652 void
3653 if_delallmulti(struct ifnet *ifp)
3654 {
3655 struct ifmultiaddr *ifma;
3656 struct ifmultiaddr *next;
3657
3658 IF_ADDR_WLOCK(ifp);
3659 CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3660 if_delmulti_locked(ifp, ifma, 0);
3661 IF_ADDR_WUNLOCK(ifp);
3662 }
3663
3664 void
3665 if_delmulti_ifma(struct ifmultiaddr *ifma)
3666 {
3667 if_delmulti_ifma_flags(ifma, 0);
3668 }
3669
3670 /*
3671 * Delete a multicast group membership by group membership pointer.
3672 * Network-layer protocol domains must use this routine.
3673 *
3674 * It is safe to call this routine if the ifp disappeared.
3675 */
3676 void
3677 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3678 {
3679 struct ifnet *ifp;
3680 int lastref;
3681 MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3682 #ifdef INET
3683 IN_MULTI_LIST_UNLOCK_ASSERT();
3684 #endif
3685 ifp = ifma->ifma_ifp;
3686 #ifdef DIAGNOSTIC
3687 if (ifp == NULL) {
3688 printf("%s: ifma_ifp seems to be detached\n", __func__);
3689 } else {
3690 struct ifnet *oifp;
3691
3692 IFNET_RLOCK_NOSLEEP();
3693 CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3694 if (ifp == oifp)
3695 break;
3696 if (ifp != oifp)
3697 ifp = NULL;
3698 IFNET_RUNLOCK_NOSLEEP();
3699 }
3700 #endif
3701 /*
3702 * If and only if the ifnet instance exists: Acquire the address lock.
3703 */
3704 if (ifp != NULL)
3705 IF_ADDR_WLOCK(ifp);
3706
3707 lastref = if_delmulti_locked(ifp, ifma, flags);
3708
3709 if (ifp != NULL) {
3710 /*
3711 * If and only if the ifnet instance exists:
3712 * Release the address lock.
3713 * If the group was left: update the hardware hash filter.
3714 */
3715 IF_ADDR_WUNLOCK(ifp);
3716 if (lastref && ifp->if_ioctl != NULL) {
3717 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3718 }
3719 }
3720 }
3721
3722 /*
3723 * Perform deletion of network-layer and/or link-layer multicast address.
3724 *
3725 * Return 0 if the reference count was decremented.
3726 * Return 1 if the final reference was released, indicating that the
3727 * hardware hash filter should be reprogrammed.
3728 */
3729 static int
3730 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3731 {
3732 struct ifmultiaddr *ll_ifma;
3733
3734 if (ifp != NULL && ifma->ifma_ifp != NULL) {
3735 KASSERT(ifma->ifma_ifp == ifp,
3736 ("%s: inconsistent ifp %p", __func__, ifp));
3737 IF_ADDR_WLOCK_ASSERT(ifp);
3738 }
3739
3740 ifp = ifma->ifma_ifp;
3741 MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3742
3743 /*
3744 * If the ifnet is detaching, null out references to ifnet,
3745 * so that upper protocol layers will notice, and not attempt
3746 * to obtain locks for an ifnet which no longer exists. The
3747 * routing socket announcement must happen before the ifnet
3748 * instance is detached from the system.
3749 */
3750 if (detaching) {
3751 #ifdef DIAGNOSTIC
3752 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3753 #endif
3754 /*
3755 * ifp may already be nulled out if we are being reentered
3756 * to delete the ll_ifma.
3757 */
3758 if (ifp != NULL) {
3759 rt_newmaddrmsg(RTM_DELMADDR, ifma);
3760 ifma->ifma_ifp = NULL;
3761 }
3762 }
3763
3764 if (--ifma->ifma_refcount > 0)
3765 return 0;
3766
3767 if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3768 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3769 ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3770 }
3771 /*
3772 * If this ifma is a network-layer ifma, a link-layer ifma may
3773 * have been associated with it. Release it first if so.
3774 */
3775 ll_ifma = ifma->ifma_llifma;
3776 if (ll_ifma != NULL) {
3777 KASSERT(ifma->ifma_lladdr != NULL,
3778 ("%s: llifma w/o lladdr", __func__));
3779 if (detaching)
3780 ll_ifma->ifma_ifp = NULL; /* XXX */
3781 if (--ll_ifma->ifma_refcount == 0) {
3782 if (ifp != NULL) {
3783 if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3784 CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3785 ifma_link);
3786 ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3787 }
3788 }
3789 if_freemulti(ll_ifma);
3790 }
3791 }
3792 #ifdef INVARIANTS
3793 if (ifp) {
3794 struct ifmultiaddr *ifmatmp;
3795
3796 CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3797 MPASS(ifma != ifmatmp);
3798 }
3799 #endif
3800 if_freemulti(ifma);
3801 /*
3802 * The last reference to this instance of struct ifmultiaddr
3803 * was released; the hardware should be notified of this change.
3804 */
3805 return 1;
3806 }
3807
3808 /*
3809 * Set the link layer address on an interface.
3810 *
3811 * At this time we only support certain types of interfaces,
3812 * and we don't allow the length of the address to change.
3813 *
3814 * Set noinline to be dtrace-friendly
3815 */
3816 __noinline int
3817 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3818 {
3819 struct sockaddr_dl *sdl;
3820 struct ifaddr *ifa;
3821 struct ifreq ifr;
3822 int rc;
3823
3824 rc = 0;
3825 NET_EPOCH_ENTER();
3826 ifa = ifp->if_addr;
3827 if (ifa == NULL) {
3828 rc = EINVAL;
3829 goto out;
3830 }
3831
3832 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3833 if (sdl == NULL) {
3834 rc = EINVAL;
3835 goto out;
3836 }
3837 if (len != sdl->sdl_alen) { /* don't allow length to change */
3838 rc = EINVAL;
3839 goto out;
3840 }
3841 switch (ifp->if_type) {
3842 case IFT_ETHER:
3843 case IFT_XETHER:
3844 case IFT_L2VLAN:
3845 case IFT_BRIDGE:
3846 case IFT_IEEE8023ADLAG:
3847 bcopy(lladdr, LLADDR(sdl), len);
3848 break;
3849 default:
3850 rc = ENODEV;
3851 goto out;
3852 }
3853
3854 /*
3855 * If the interface is already up, we need
3856 * to re-init it in order to reprogram its
3857 * address filter.
3858 */
3859 NET_EPOCH_EXIT();
3860 if ((ifp->if_flags & IFF_UP) != 0) {
3861 if (ifp->if_ioctl) {
3862 ifp->if_flags &= ~IFF_UP;
3863 ifr.ifr_flags = ifp->if_flags & 0xffff;
3864 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3865 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3866 ifp->if_flags |= IFF_UP;
3867 ifr.ifr_flags = ifp->if_flags & 0xffff;
3868 ifr.ifr_flagshigh = ifp->if_flags >> 16;
3869 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3870 }
3871 }
3872 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3873 return (0);
3874 out:
3875 NET_EPOCH_EXIT();
3876 return (rc);
3877 }
3878
3879 /*
3880 * Compat function for handling basic encapsulation requests.
3881 * Not converted stacks (FDDI, IB, ..) supports traditional
3882 * output model: ARP (and other similar L2 protocols) are handled
3883 * inside output routine, arpresolve/nd6_resolve() returns MAC
3884 * address instead of full prepend.
3885 *
3886 * This function creates calculated header==MAC for IPv4/IPv6 and
3887 * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3888 * address families.
3889 */
3890 static int
3891 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3892 {
3893
3894 if (req->rtype != IFENCAP_LL)
3895 return (EOPNOTSUPP);
3896
3897 if (req->bufsize < req->lladdr_len)
3898 return (ENOMEM);
3899
3900 switch (req->family) {
3901 case AF_INET:
3902 case AF_INET6:
3903 break;
3904 default:
3905 return (EAFNOSUPPORT);
3906 }
3907
3908 /* Copy lladdr to storage as is */
3909 memmove(req->buf, req->lladdr, req->lladdr_len);
3910 req->bufsize = req->lladdr_len;
3911 req->lladdr_off = 0;
3912
3913 return (0);
3914 }
3915
3916 /*
3917 * Tunnel interfaces can nest, also they may cause infinite recursion
3918 * calls when misconfigured. We'll prevent this by detecting loops.
3919 * High nesting level may cause stack exhaustion. We'll prevent this
3920 * by introducing upper limit.
3921 *
3922 * Return 0, if tunnel nesting count is equal or less than limit.
3923 */
3924 int
3925 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3926 int limit)
3927 {
3928 struct m_tag *mtag;
3929 int count;
3930
3931 count = 1;
3932 mtag = NULL;
3933 while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3934 if (*(struct ifnet **)(mtag + 1) == ifp) {
3935 log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3936 return (EIO);
3937 }
3938 count++;
3939 }
3940 if (count > limit) {
3941 log(LOG_NOTICE,
3942 "%s: if_output recursively called too many times(%d)\n",
3943 if_name(ifp), count);
3944 return (EIO);
3945 }
3946 mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3947 if (mtag == NULL)
3948 return (ENOMEM);
3949 *(struct ifnet **)(mtag + 1) = ifp;
3950 m_tag_prepend(m, mtag);
3951 return (0);
3952 }
3953
3954 /*
3955 * Get the link layer address that was read from the hardware at attach.
3956 *
3957 * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3958 * their component interfaces as IFT_IEEE8023ADLAG.
3959 */
3960 int
3961 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3962 {
3963
3964 if (ifp->if_hw_addr == NULL)
3965 return (ENODEV);
3966
3967 switch (ifp->if_type) {
3968 case IFT_ETHER:
3969 case IFT_IEEE8023ADLAG:
3970 bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3971 return (0);
3972 default:
3973 return (ENODEV);
3974 }
3975 }
3976
3977 /*
3978 * The name argument must be a pointer to storage which will last as
3979 * long as the interface does. For physical devices, the result of
3980 * device_get_name(dev) is a good choice and for pseudo-devices a
3981 * static string works well.
3982 */
3983 void
3984 if_initname(struct ifnet *ifp, const char *name, int unit)
3985 {
3986 ifp->if_dname = name;
3987 ifp->if_dunit = unit;
3988 if (unit != IF_DUNIT_NONE)
3989 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3990 else
3991 strlcpy(ifp->if_xname, name, IFNAMSIZ);
3992 }
3993
3994 int
3995 if_printf(struct ifnet *ifp, const char *fmt, ...)
3996 {
3997 char if_fmt[256];
3998 va_list ap;
3999
4000 snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
4001 va_start(ap, fmt);
4002 vlog(LOG_INFO, if_fmt, ap);
4003 va_end(ap);
4004 return (0);
4005 }
4006
4007 void
4008 if_start(struct ifnet *ifp)
4009 {
4010
4011 (*(ifp)->if_start)(ifp);
4012 }
4013
4014 /*
4015 * Backwards compatibility interface for drivers
4016 * that have not implemented it
4017 */
4018 static int
4019 if_transmit(struct ifnet *ifp, struct mbuf *m)
4020 {
4021 int error;
4022
4023 IFQ_HANDOFF(ifp, m, error);
4024 return (error);
4025 }
4026
4027 static void
4028 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
4029 {
4030
4031 m_freem(m);
4032 }
4033
4034 int
4035 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4036 {
4037 int active = 0;
4038
4039 IF_LOCK(ifq);
4040 if (_IF_QFULL(ifq)) {
4041 IF_UNLOCK(ifq);
4042 if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4043 m_freem(m);
4044 return (0);
4045 }
4046 if (ifp != NULL) {
4047 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4048 if (m->m_flags & (M_BCAST|M_MCAST))
4049 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4050 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4051 }
4052 _IF_ENQUEUE(ifq, m);
4053 IF_UNLOCK(ifq);
4054 if (ifp != NULL && !active)
4055 (*(ifp)->if_start)(ifp);
4056 return (1);
4057 }
4058
4059 void
4060 if_register_com_alloc(u_char type,
4061 if_com_alloc_t *a, if_com_free_t *f)
4062 {
4063
4064 KASSERT(if_com_alloc[type] == NULL,
4065 ("if_register_com_alloc: %d already registered", type));
4066 KASSERT(if_com_free[type] == NULL,
4067 ("if_register_com_alloc: %d free already registered", type));
4068
4069 if_com_alloc[type] = a;
4070 if_com_free[type] = f;
4071 }
4072
4073 void
4074 if_deregister_com_alloc(u_char type)
4075 {
4076
4077 KASSERT(if_com_alloc[type] != NULL,
4078 ("if_deregister_com_alloc: %d not registered", type));
4079 KASSERT(if_com_free[type] != NULL,
4080 ("if_deregister_com_alloc: %d free not registered", type));
4081 if_com_alloc[type] = NULL;
4082 if_com_free[type] = NULL;
4083 }
4084
4085 /* API for driver access to network stack owned ifnet.*/
4086 uint64_t
4087 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4088 {
4089 uint64_t oldbrate;
4090
4091 oldbrate = ifp->if_baudrate;
4092 ifp->if_baudrate = baudrate;
4093 return (oldbrate);
4094 }
4095
4096 uint64_t
4097 if_getbaudrate(if_t ifp)
4098 {
4099
4100 return (((struct ifnet *)ifp)->if_baudrate);
4101 }
4102
4103 int
4104 if_setcapabilities(if_t ifp, int capabilities)
4105 {
4106 ((struct ifnet *)ifp)->if_capabilities = capabilities;
4107 return (0);
4108 }
4109
4110 int
4111 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4112 {
4113 ((struct ifnet *)ifp)->if_capabilities |= setbit;
4114 ((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4115
4116 return (0);
4117 }
4118
4119 int
4120 if_getcapabilities(if_t ifp)
4121 {
4122 return ((struct ifnet *)ifp)->if_capabilities;
4123 }
4124
4125 int
4126 if_setcapenable(if_t ifp, int capabilities)
4127 {
4128 ((struct ifnet *)ifp)->if_capenable = capabilities;
4129 return (0);
4130 }
4131
4132 int
4133 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4134 {
4135 if(setcap)
4136 ((struct ifnet *)ifp)->if_capenable |= setcap;
4137 if(clearcap)
4138 ((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4139
4140 return (0);
4141 }
4142
4143 const char *
4144 if_getdname(if_t ifp)
4145 {
4146 return ((struct ifnet *)ifp)->if_dname;
4147 }
4148
4149 int
4150 if_togglecapenable(if_t ifp, int togglecap)
4151 {
4152 ((struct ifnet *)ifp)->if_capenable ^= togglecap;
4153 return (0);
4154 }
4155
4156 int
4157 if_getcapenable(if_t ifp)
4158 {
4159 return ((struct ifnet *)ifp)->if_capenable;
4160 }
4161
4162 /*
4163 * This is largely undesirable because it ties ifnet to a device, but does
4164 * provide flexiblity for an embedded product vendor. Should be used with
4165 * the understanding that it violates the interface boundaries, and should be
4166 * a last resort only.
4167 */
4168 int
4169 if_setdev(if_t ifp, void *dev)
4170 {
4171 return (0);
4172 }
4173
4174 int
4175 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4176 {
4177 ((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4178 ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4179
4180 return (0);
4181 }
4182
4183 int
4184 if_getdrvflags(if_t ifp)
4185 {
4186 return ((struct ifnet *)ifp)->if_drv_flags;
4187 }
4188
4189 int
4190 if_setdrvflags(if_t ifp, int flags)
4191 {
4192 ((struct ifnet *)ifp)->if_drv_flags = flags;
4193 return (0);
4194 }
4195
4196
4197 int
4198 if_setflags(if_t ifp, int flags)
4199 {
4200 ((struct ifnet *)ifp)->if_flags = flags;
4201 return (0);
4202 }
4203
4204 int
4205 if_setflagbits(if_t ifp, int set, int clear)
4206 {
4207 ((struct ifnet *)ifp)->if_flags |= set;
4208 ((struct ifnet *)ifp)->if_flags &= ~clear;
4209
4210 return (0);
4211 }
4212
4213 int
4214 if_getflags(if_t ifp)
4215 {
4216 return ((struct ifnet *)ifp)->if_flags;
4217 }
4218
4219 int
4220 if_clearhwassist(if_t ifp)
4221 {
4222 ((struct ifnet *)ifp)->if_hwassist = 0;
4223 return (0);
4224 }
4225
4226 int
4227 if_sethwassistbits(if_t ifp, int toset, int toclear)
4228 {
4229 ((struct ifnet *)ifp)->if_hwassist |= toset;
4230 ((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4231
4232 return (0);
4233 }
4234
4235 int
4236 if_sethwassist(if_t ifp, int hwassist_bit)
4237 {
4238 ((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4239 return (0);
4240 }
4241
4242 int
4243 if_gethwassist(if_t ifp)
4244 {
4245 return ((struct ifnet *)ifp)->if_hwassist;
4246 }
4247
4248 int
4249 if_setmtu(if_t ifp, int mtu)
4250 {
4251 ((struct ifnet *)ifp)->if_mtu = mtu;
4252 return (0);
4253 }
4254
4255 int
4256 if_getmtu(if_t ifp)
4257 {
4258 return ((struct ifnet *)ifp)->if_mtu;
4259 }
4260
4261 int
4262 if_getmtu_family(if_t ifp, int family)
4263 {
4264 struct domain *dp;
4265
4266 for (dp = domains; dp; dp = dp->dom_next) {
4267 if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4268 return (dp->dom_ifmtu((struct ifnet *)ifp));
4269 }
4270
4271 return (((struct ifnet *)ifp)->if_mtu);
4272 }
4273
4274 int
4275 if_setsoftc(if_t ifp, void *softc)
4276 {
4277 ((struct ifnet *)ifp)->if_softc = softc;
4278 return (0);
4279 }
4280
4281 void *
4282 if_getsoftc(if_t ifp)
4283 {
4284 return ((struct ifnet *)ifp)->if_softc;
4285 }
4286
4287 void
4288 if_setrcvif(struct mbuf *m, if_t ifp)
4289 {
4290 m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4291 }
4292
4293 void
4294 if_setvtag(struct mbuf *m, uint16_t tag)
4295 {
4296 m->m_pkthdr.ether_vtag = tag;
4297 }
4298
4299 uint16_t
4300 if_getvtag(struct mbuf *m)
4301 {
4302
4303 return (m->m_pkthdr.ether_vtag);
4304 }
4305
4306 int
4307 if_sendq_empty(if_t ifp)
4308 {
4309 return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4310 }
4311
4312 struct ifaddr *
4313 if_getifaddr(if_t ifp)
4314 {
4315 return ((struct ifnet *)ifp)->if_addr;
4316 }
4317
4318 int
4319 if_getamcount(if_t ifp)
4320 {
4321 return ((struct ifnet *)ifp)->if_amcount;
4322 }
4323
4324
4325 int
4326 if_setsendqready(if_t ifp)
4327 {
4328 IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4329 return (0);
4330 }
4331
4332 int
4333 if_setsendqlen(if_t ifp, int tx_desc_count)
4334 {
4335 IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4336 ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4337
4338 return (0);
4339 }
4340
4341 int
4342 if_vlantrunkinuse(if_t ifp)
4343 {
4344 return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4345 }
4346
4347 int
4348 if_input(if_t ifp, struct mbuf* sendmp)
4349 {
4350 (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4351 return (0);
4352
4353 }
4354
4355 /* XXX */
4356 #ifndef ETH_ADDR_LEN
4357 #define ETH_ADDR_LEN 6
4358 #endif
4359
4360 int
4361 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4362 {
4363 struct ifmultiaddr *ifma;
4364 uint8_t *lmta = (uint8_t *)mta;
4365 int mcnt = 0;
4366
4367 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4368 if (ifma->ifma_addr->sa_family != AF_LINK)
4369 continue;
4370
4371 if (mcnt == max)
4372 break;
4373
4374 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4375 &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4376 mcnt++;
4377 }
4378 *cnt = mcnt;
4379
4380 return (0);
4381 }
4382
4383 int
4384 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4385 {
4386 int error;
4387
4388 if_maddr_rlock(ifp);
4389 error = if_setupmultiaddr(ifp, mta, cnt, max);
4390 if_maddr_runlock(ifp);
4391 return (error);
4392 }
4393
4394 int
4395 if_multiaddr_count(if_t ifp, int max)
4396 {
4397 struct ifmultiaddr *ifma;
4398 int count;
4399
4400 count = 0;
4401 if_maddr_rlock(ifp);
4402 CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4403 if (ifma->ifma_addr->sa_family != AF_LINK)
4404 continue;
4405 count++;
4406 if (count == max)
4407 break;
4408 }
4409 if_maddr_runlock(ifp);
4410 return (count);
4411 }
4412
4413 int
4414 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4415 {
4416 struct ifmultiaddr *ifma;
4417 int cnt = 0;
4418
4419 if_maddr_rlock(ifp);
4420 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4421 cnt += filter(arg, ifma, cnt);
4422 if_maddr_runlock(ifp);
4423 return (cnt);
4424 }
4425
4426 struct mbuf *
4427 if_dequeue(if_t ifp)
4428 {
4429 struct mbuf *m;
4430 IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4431
4432 return (m);
4433 }
4434
4435 int
4436 if_sendq_prepend(if_t ifp, struct mbuf *m)
4437 {
4438 IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4439 return (0);
4440 }
4441
4442 int
4443 if_setifheaderlen(if_t ifp, int len)
4444 {
4445 ((struct ifnet *)ifp)->if_hdrlen = len;
4446 return (0);
4447 }
4448
4449 caddr_t
4450 if_getlladdr(if_t ifp)
4451 {
4452 return (IF_LLADDR((struct ifnet *)ifp));
4453 }
4454
4455 void *
4456 if_gethandle(u_char type)
4457 {
4458 return (if_alloc(type));
4459 }
4460
4461 void
4462 if_bpfmtap(if_t ifh, struct mbuf *m)
4463 {
4464 struct ifnet *ifp = (struct ifnet *)ifh;
4465
4466 BPF_MTAP(ifp, m);
4467 }
4468
4469 void
4470 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4471 {
4472 struct ifnet *ifp = (struct ifnet *)ifh;
4473
4474 ETHER_BPF_MTAP(ifp, m);
4475 }
4476
4477 void
4478 if_vlancap(if_t ifh)
4479 {
4480 struct ifnet *ifp = (struct ifnet *)ifh;
4481 VLAN_CAPABILITIES(ifp);
4482 }
4483
4484 int
4485 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4486 {
4487
4488 ((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4489 return (0);
4490 }
4491
4492 int
4493 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4494 {
4495
4496 ((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4497 return (0);
4498 }
4499
4500 int
4501 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4502 {
4503
4504 ((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4505 return (0);
4506 }
4507
4508 u_int
4509 if_gethwtsomax(if_t ifp)
4510 {
4511
4512 return (((struct ifnet *)ifp)->if_hw_tsomax);
4513 }
4514
4515 u_int
4516 if_gethwtsomaxsegcount(if_t ifp)
4517 {
4518
4519 return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4520 }
4521
4522 u_int
4523 if_gethwtsomaxsegsize(if_t ifp)
4524 {
4525
4526 return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4527 }
4528
4529 void
4530 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4531 {
4532 ((struct ifnet *)ifp)->if_init = init_fn;
4533 }
4534
4535 void
4536 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4537 {
4538 ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4539 }
4540
4541 void
4542 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4543 {
4544 ((struct ifnet *)ifp)->if_start = (void *)start_fn;
4545 }
4546
4547 void
4548 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4549 {
4550 ((struct ifnet *)ifp)->if_transmit = start_fn;
4551 }
4552
4553 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4554 {
4555 ((struct ifnet *)ifp)->if_qflush = flush_fn;
4556
4557 }
4558
4559 void
4560 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4561 {
4562
4563 ifp->if_get_counter = fn;
4564 }
4565
4566 /* Revisit these - These are inline functions originally. */
4567 int
4568 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4569 {
4570 return drbr_inuse(ifh, br);
4571 }
4572
4573 struct mbuf*
4574 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4575 {
4576 return drbr_dequeue(ifh, br);
4577 }
4578
4579 int
4580 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4581 {
4582 return drbr_needs_enqueue(ifh, br);
4583 }
4584
4585 int
4586 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4587 {
4588 return drbr_enqueue(ifh, br, m);
4589
4590 }
Cache object: 578466fa8bd610cf3a0a8c5ea6ec552a
|