xref: /freebsd/sys/net/if.c (revision 1ba655149ed0447f93e997a60330d9b962d061f2)
1  /*-
2   * SPDX-License-Identifier: BSD-3-Clause
3   *
4   * Copyright (c) 2010 Bjoern A. Zeeb <bz@FreeBSD.org>
5   * Copyright (c) 1980, 1986, 1993
6   *	The Regents of the University of California.  All rights reserved.
7   *
8   * Redistribution and use in source and binary forms, with or without
9   * modification, are permitted provided that the following conditions
10   * are met:
11   * 1. Redistributions of source code must retain the above copyright
12   *    notice, this list of conditions and the following disclaimer.
13   * 2. Redistributions in binary form must reproduce the above copyright
14   *    notice, this list of conditions and the following disclaimer in the
15   *    documentation and/or other materials provided with the distribution.
16   * 3. Neither the name of the University nor the names of its contributors
17   *    may be used to endorse or promote products derived from this software
18   *    without specific prior written permission.
19   *
20   * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21   * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22   * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23   * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24   * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25   * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26   * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27   * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28   * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29   * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30   * SUCH DAMAGE.
31   */
32  
33  #include "opt_bpf.h"
34  #include "opt_inet6.h"
35  #include "opt_inet.h"
36  #include "opt_ddb.h"
37  
38  #include <sys/param.h>
39  #include <sys/capsicum.h>
40  #include <sys/conf.h>
41  #include <sys/eventhandler.h>
42  #include <sys/malloc.h>
43  #include <sys/domainset.h>
44  #include <sys/sbuf.h>
45  #include <sys/bus.h>
46  #include <sys/epoch.h>
47  #include <sys/mbuf.h>
48  #include <sys/systm.h>
49  #include <sys/priv.h>
50  #include <sys/proc.h>
51  #include <sys/socket.h>
52  #include <sys/socketvar.h>
53  #include <sys/protosw.h>
54  #include <sys/kernel.h>
55  #include <sys/lock.h>
56  #include <sys/refcount.h>
57  #include <sys/module.h>
58  #include <sys/nv.h>
59  #include <sys/rwlock.h>
60  #include <sys/sockio.h>
61  #include <sys/syslog.h>
62  #include <sys/sysctl.h>
63  #include <sys/sysent.h>
64  #include <sys/taskqueue.h>
65  #include <sys/domain.h>
66  #include <sys/jail.h>
67  #include <sys/priv.h>
68  
69  #ifdef DDB
70  #include <ddb/ddb.h>
71  #endif
72  
73  #include <machine/stdarg.h>
74  #include <vm/uma.h>
75  
76  #include <net/bpf.h>
77  #include <net/ethernet.h>
78  #include <net/if.h>
79  #include <net/if_arp.h>
80  #include <net/if_clone.h>
81  #include <net/if_dl.h>
82  #include <net/if_strings.h>
83  #include <net/if_types.h>
84  #include <net/if_var.h>
85  #include <net/if_media.h>
86  #include <net/if_mib.h>
87  #include <net/if_private.h>
88  #include <net/if_vlan_var.h>
89  #include <net/radix.h>
90  #include <net/route.h>
91  #include <net/route/route_ctl.h>
92  #include <net/vnet.h>
93  
94  #if defined(INET) || defined(INET6)
95  #include <net/ethernet.h>
96  #include <netinet/in.h>
97  #include <netinet/in_var.h>
98  #include <netinet/ip.h>
99  #include <netinet/ip_carp.h>
100  #ifdef INET
101  #include <net/debugnet.h>
102  #include <netinet/if_ether.h>
103  #endif /* INET */
104  #ifdef INET6
105  #include <netinet6/in6_var.h>
106  #include <netinet6/in6_ifattach.h>
107  #endif /* INET6 */
108  #endif /* INET || INET6 */
109  
110  #include <security/mac/mac_framework.h>
111  
112  /*
113   * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
114   * and ifr_ifru when it is used in SIOCGIFCONF.
115   */
116  _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
117      offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
118  
119  __read_mostly epoch_t net_epoch_preempt;
120  #ifdef COMPAT_FREEBSD32
121  #include <sys/mount.h>
122  #include <compat/freebsd32/freebsd32.h>
123  
124  struct ifreq_buffer32 {
125  	uint32_t	length;		/* (size_t) */
126  	uint32_t	buffer;		/* (void *) */
127  };
128  
129  /*
130   * Interface request structure used for socket
131   * ioctl's.  All interface ioctl's must have parameter
132   * definitions which begin with ifr_name.  The
133   * remainder may be interface specific.
134   */
135  struct ifreq32 {
136  	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
137  	union {
138  		struct sockaddr	ifru_addr;
139  		struct sockaddr	ifru_dstaddr;
140  		struct sockaddr	ifru_broadaddr;
141  		struct ifreq_buffer32 ifru_buffer;
142  		short		ifru_flags[2];
143  		short		ifru_index;
144  		int		ifru_jid;
145  		int		ifru_metric;
146  		int		ifru_mtu;
147  		int		ifru_phys;
148  		int		ifru_media;
149  		uint32_t	ifru_data;
150  		int		ifru_cap[2];
151  		u_int		ifru_fib;
152  		u_char		ifru_vlan_pcp;
153  	} ifr_ifru;
154  };
155  CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
156  CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
157      __offsetof(struct ifreq32, ifr_ifru));
158  
159  struct ifconf32 {
160  	int32_t	ifc_len;
161  	union {
162  		uint32_t	ifcu_buf;
163  		uint32_t	ifcu_req;
164  	} ifc_ifcu;
165  };
166  #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
167  
168  struct ifdrv32 {
169  	char		ifd_name[IFNAMSIZ];
170  	uint32_t	ifd_cmd;
171  	uint32_t	ifd_len;
172  	uint32_t	ifd_data;
173  };
174  #define SIOCSDRVSPEC32	_IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32)
175  #define SIOCGDRVSPEC32	_IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32)
176  
177  struct ifgroupreq32 {
178  	char	ifgr_name[IFNAMSIZ];
179  	u_int	ifgr_len;
180  	union {
181  		char		ifgru_group[IFNAMSIZ];
182  		uint32_t	ifgru_groups;
183  	} ifgr_ifgru;
184  };
185  #define	SIOCAIFGROUP32	_IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32)
186  #define	SIOCGIFGROUP32	_IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32)
187  #define	SIOCDIFGROUP32	_IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32)
188  #define	SIOCGIFGMEMB32	_IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32)
189  
190  struct ifmediareq32 {
191  	char		ifm_name[IFNAMSIZ];
192  	int		ifm_current;
193  	int		ifm_mask;
194  	int		ifm_status;
195  	int		ifm_active;
196  	int		ifm_count;
197  	uint32_t	ifm_ulist;	/* (int *) */
198  };
199  #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
200  #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
201  #endif /* COMPAT_FREEBSD32 */
202  
203  union ifreq_union {
204  	struct ifreq	ifr;
205  #ifdef COMPAT_FREEBSD32
206  	struct ifreq32	ifr32;
207  #endif
208  };
209  
210  SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
211      "Link layers");
212  SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
213      "Generic link-management");
214  
215  SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
216      &ifqmaxlen, 0, "max send queue size");
217  
218  /* Log link state change events */
219  static int log_link_state_change = 1;
220  
221  SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
222  	&log_link_state_change, 0,
223  	"log interface link state change events");
224  
225  /* Log promiscuous mode change events */
226  static int log_promisc_mode_change = 1;
227  
228  SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
229  	&log_promisc_mode_change, 1,
230  	"log promiscuous mode change events");
231  
232  /* Interface description */
233  static unsigned int ifdescr_maxlen = 1024;
234  SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
235  	&ifdescr_maxlen, 0,
236  	"administrative maximum length for interface description");
237  
238  static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
239  
240  /* global sx for non-critical path ifdescr */
241  static struct sx ifdescr_sx;
242  SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
243  
244  void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
245  void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
246  /* These are external hooks for CARP. */
247  void	(*carp_linkstate_p)(struct ifnet *ifp);
248  void	(*carp_demote_adj_p)(int, char *);
249  int	(*carp_master_p)(struct ifaddr *);
250  #if defined(INET) || defined(INET6)
251  int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
252  int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
253      const struct sockaddr *sa);
254  int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
255  int	(*carp_attach_p)(struct ifaddr *, int);
256  void	(*carp_detach_p)(struct ifaddr *, bool);
257  #endif
258  #ifdef INET
259  int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
260  #endif
261  #ifdef INET6
262  struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
263  caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
264      const struct in6_addr *taddr);
265  #endif
266  
267  struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
268  
269  /*
270   * XXX: Style; these should be sorted alphabetically, and unprototyped
271   * static functions should be prototyped. Currently they are sorted by
272   * declaration order.
273   */
274  static void	if_attachdomain(void *);
275  static void	if_attachdomain1(struct ifnet *);
276  static int	ifconf(u_long, caddr_t);
277  static void	if_input_default(struct ifnet *, struct mbuf *);
278  static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
279  static int	if_setflag(struct ifnet *, int, int, int *, int);
280  static int	if_transmit_default(struct ifnet *ifp, struct mbuf *m);
281  static void	if_unroute(struct ifnet *, int flag, int fam);
282  static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
283  static void	do_link_state_change(void *, int);
284  static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
285  static int	if_getgroupmembers(struct ifgroupreq *);
286  static void	if_delgroups(struct ifnet *);
287  static void	if_attach_internal(struct ifnet *, bool);
288  static void	if_detach_internal(struct ifnet *, bool);
289  static void	if_siocaddmulti(void *, int);
290  static void	if_link_ifnet(struct ifnet *);
291  static bool	if_unlink_ifnet(struct ifnet *, bool);
292  #ifdef VIMAGE
293  static void	if_vmove(struct ifnet *, struct vnet *);
294  #endif
295  
296  #ifdef INET6
297  /*
298   * XXX: declare here to avoid to include many inet6 related files..
299   * should be more generalized?
300   */
301  extern void	nd6_setmtu(struct ifnet *);
302  #endif
303  
304  /* ipsec helper hooks */
305  VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
306  VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
307  
308  int	ifqmaxlen = IFQ_MAXLEN;
309  VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
310  VNET_DEFINE(struct ifgrouphead, ifg_head);
311  
312  /* Table of ifnet by index. */
313  static int if_index;
314  static int if_indexlim = 8;
315  static struct ifindex_entry {
316  	struct ifnet	*ife_ifnet;
317  	uint16_t	ife_gencnt;
318  } *ifindex_table;
319  
320  SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system,
321      CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
322      "Variables global to all interfaces");
323  static int
sysctl_ifcount(SYSCTL_HANDLER_ARGS)324  sysctl_ifcount(SYSCTL_HANDLER_ARGS)
325  {
326  	int rv = 0;
327  
328  	IFNET_RLOCK();
329  	for (int i = 1; i <= if_index; i++)
330  		if (ifindex_table[i].ife_ifnet != NULL &&
331  		    ifindex_table[i].ife_ifnet->if_vnet == curvnet)
332  			rv = i;
333  	IFNET_RUNLOCK();
334  
335  	return (sysctl_handle_int(oidp, &rv, 0, req));
336  }
337  SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount,
338      CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I",
339      "Maximum known interface index");
340  
341  /*
342   * The global network interface list (V_ifnet) and related state (such as
343   * if_index, if_indexlim, and ifindex_table) are protected by an sxlock.
344   * This may be acquired to stabilise the list, or we may rely on NET_EPOCH.
345   */
346  struct sx ifnet_sxlock;
347  SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
348  
349  struct sx ifnet_detach_sxlock;
350  SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx",
351      SX_RECURSE);
352  
353  #ifdef VIMAGE
354  #define	VNET_IS_SHUTTING_DOWN(_vnet)					\
355      ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE)
356  #endif
357  
358  static	if_com_alloc_t *if_com_alloc[256];
359  static	if_com_free_t *if_com_free[256];
360  
361  static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
362  MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
363  MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
364  
365  struct ifnet *
ifnet_byindex(u_int idx)366  ifnet_byindex(u_int idx)
367  {
368  	struct ifnet *ifp;
369  
370  	NET_EPOCH_ASSERT();
371  
372  	if (__predict_false(idx > if_index))
373  		return (NULL);
374  
375  	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
376  
377  	if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet)
378  		ifp = NULL;
379  
380  	return (ifp);
381  }
382  
383  struct ifnet *
ifnet_byindex_ref(u_int idx)384  ifnet_byindex_ref(u_int idx)
385  {
386  	struct ifnet *ifp;
387  
388  	ifp = ifnet_byindex(idx);
389  	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
390  		return (NULL);
391  	if (!if_try_ref(ifp))
392  		return (NULL);
393  	return (ifp);
394  }
395  
396  struct ifnet *
ifnet_byindexgen(uint16_t idx,uint16_t gen)397  ifnet_byindexgen(uint16_t idx, uint16_t gen)
398  {
399  	struct ifnet *ifp;
400  
401  	NET_EPOCH_ASSERT();
402  
403  	if (__predict_false(idx > if_index))
404  		return (NULL);
405  
406  	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
407  
408  	if (ifindex_table[idx].ife_gencnt == gen)
409  		return (ifp);
410  	else
411  		return (NULL);
412  }
413  
414  /*
415   * Network interface utility routines.
416   *
417   * Routines with ifa_ifwith* names take sockaddr *'s as
418   * parameters.
419   */
420  
421  static void
if_init_idxtable(void * arg __unused)422  if_init_idxtable(void *arg __unused)
423  {
424  
425  	ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table),
426  	    M_IFNET, M_WAITOK | M_ZERO);
427  }
428  SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init_idxtable, NULL);
429  
430  static void
vnet_if_init(const void * unused __unused)431  vnet_if_init(const void *unused __unused)
432  {
433  
434  	CK_STAILQ_INIT(&V_ifnet);
435  	CK_STAILQ_INIT(&V_ifg_head);
436  }
437  VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
438      NULL);
439  
440  static void
if_link_ifnet(struct ifnet * ifp)441  if_link_ifnet(struct ifnet *ifp)
442  {
443  
444  	IFNET_WLOCK();
445  	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
446  #ifdef VIMAGE
447  	curvnet->vnet_ifcnt++;
448  #endif
449  	IFNET_WUNLOCK();
450  }
451  
452  static bool
if_unlink_ifnet(struct ifnet * ifp,bool vmove)453  if_unlink_ifnet(struct ifnet *ifp, bool vmove)
454  {
455  	struct ifnet *iter;
456  	int found = 0;
457  
458  	IFNET_WLOCK();
459  	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
460  		if (iter == ifp) {
461  			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
462  			if (!vmove)
463  				ifp->if_flags |= IFF_DYING;
464  			found = 1;
465  			break;
466  		}
467  #ifdef VIMAGE
468  	curvnet->vnet_ifcnt--;
469  #endif
470  	IFNET_WUNLOCK();
471  
472  	return (found);
473  }
474  
475  #ifdef VIMAGE
476  static void
vnet_if_return(const void * unused __unused)477  vnet_if_return(const void *unused __unused)
478  {
479  	struct ifnet *ifp, *nifp;
480  	struct ifnet **pending;
481  	int found __diagused;
482  	int i;
483  
484  	i = 0;
485  
486  	/*
487  	 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd
488  	 * enter NET_EPOCH, but that's not possible, because if_vmove() calls
489  	 * if_detach_internal(), which waits for NET_EPOCH callbacks to
490  	 * complete. We can't do that from within NET_EPOCH.
491  	 *
492  	 * However, we can also use the IFNET_xLOCK, which is the V_ifnet
493  	 * read/write lock. We cannot hold the lock as we call if_vmove()
494  	 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib
495  	 * ctx lock.
496  	 */
497  	IFNET_WLOCK();
498  
499  	pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt,
500  	    M_IFNET, M_WAITOK | M_ZERO);
501  
502  	/* Return all inherited interfaces to their parent vnets. */
503  	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
504  		if (ifp->if_home_vnet != ifp->if_vnet) {
505  			found = if_unlink_ifnet(ifp, true);
506  			MPASS(found);
507  
508  			pending[i++] = ifp;
509  		}
510  	}
511  	IFNET_WUNLOCK();
512  
513  	for (int j = 0; j < i; j++) {
514  		sx_xlock(&ifnet_detach_sxlock);
515  		if_vmove(pending[j], pending[j]->if_home_vnet);
516  		sx_xunlock(&ifnet_detach_sxlock);
517  	}
518  
519  	free(pending, M_IFNET);
520  }
521  VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
522      vnet_if_return, NULL);
523  #endif
524  
525  /*
526   * Allocate a struct ifnet and an index for an interface.  A layer 2
527   * common structure will also be allocated if an allocation routine is
528   * registered for the passed type.
529   */
530  static struct ifnet *
if_alloc_domain(u_char type,int numa_domain)531  if_alloc_domain(u_char type, int numa_domain)
532  {
533  	struct ifnet *ifp;
534  	u_short idx;
535  
536  	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
537  	if (numa_domain == IF_NODOM)
538  		ifp = malloc(sizeof(struct ifnet), M_IFNET,
539  		    M_WAITOK | M_ZERO);
540  	else
541  		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
542  		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
543  	ifp->if_type = type;
544  	ifp->if_alloctype = type;
545  	ifp->if_numa_domain = numa_domain;
546  #ifdef VIMAGE
547  	ifp->if_vnet = curvnet;
548  #endif
549  	if (if_com_alloc[type] != NULL) {
550  		ifp->if_l2com = if_com_alloc[type](type, ifp);
551  		KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__,
552  		    type));
553  	}
554  
555  	IF_ADDR_LOCK_INIT(ifp);
556  	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
557  	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
558  	ifp->if_afdata_initialized = 0;
559  	IF_AFDATA_LOCK_INIT(ifp);
560  	CK_STAILQ_INIT(&ifp->if_addrhead);
561  	CK_STAILQ_INIT(&ifp->if_multiaddrs);
562  	CK_STAILQ_INIT(&ifp->if_groups);
563  #ifdef MAC
564  	mac_ifnet_init(ifp);
565  #endif
566  	ifq_init(&ifp->if_snd, ifp);
567  
568  	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
569  	for (int i = 0; i < IFCOUNTERS; i++)
570  		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
571  	ifp->if_get_counter = if_get_counter_default;
572  	ifp->if_pcp = IFNET_PCP_NONE;
573  
574  	/* Allocate an ifindex array entry. */
575  	IFNET_WLOCK();
576  	/*
577  	 * Try to find an empty slot below if_index.  If we fail, take the
578  	 * next slot.
579  	 */
580  	for (idx = 1; idx <= if_index; idx++) {
581  		if (ifindex_table[idx].ife_ifnet == NULL)
582  			break;
583  	}
584  
585  	/* Catch if_index overflow. */
586  	if (idx >= if_indexlim) {
587  		struct ifindex_entry *new, *old;
588  		int newlim;
589  
590  		newlim = if_indexlim * 2;
591  		new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO);
592  		memcpy(new, ifindex_table, if_indexlim * sizeof(*new));
593  		old = ifindex_table;
594  		ck_pr_store_ptr(&ifindex_table, new);
595  		if_indexlim = newlim;
596  		NET_EPOCH_WAIT();
597  		free(old, M_IFNET);
598  	}
599  	if (idx > if_index)
600  		if_index = idx;
601  
602  	ifp->if_index = idx;
603  	ifp->if_idxgen = ifindex_table[idx].ife_gencnt;
604  	ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp);
605  	IFNET_WUNLOCK();
606  
607  	return (ifp);
608  }
609  
610  struct ifnet *
if_alloc_dev(u_char type,device_t dev)611  if_alloc_dev(u_char type, device_t dev)
612  {
613  	int numa_domain;
614  
615  	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
616  		return (if_alloc_domain(type, IF_NODOM));
617  	return (if_alloc_domain(type, numa_domain));
618  }
619  
620  struct ifnet *
if_alloc(u_char type)621  if_alloc(u_char type)
622  {
623  
624  	return (if_alloc_domain(type, IF_NODOM));
625  }
626  /*
627   * Do the actual work of freeing a struct ifnet, and layer 2 common
628   * structure.  This call is made when the network epoch guarantees
629   * us that nobody holds a pointer to the interface.
630   */
631  static void
if_free_deferred(epoch_context_t ctx)632  if_free_deferred(epoch_context_t ctx)
633  {
634  	struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
635  
636  	KASSERT((ifp->if_flags & IFF_DYING),
637  	    ("%s: interface not dying", __func__));
638  
639  	if (if_com_free[ifp->if_alloctype] != NULL)
640  		if_com_free[ifp->if_alloctype](ifp->if_l2com,
641  		    ifp->if_alloctype);
642  
643  #ifdef MAC
644  	mac_ifnet_destroy(ifp);
645  #endif /* MAC */
646  	IF_AFDATA_DESTROY(ifp);
647  	IF_ADDR_LOCK_DESTROY(ifp);
648  	ifq_delete(&ifp->if_snd);
649  
650  	for (int i = 0; i < IFCOUNTERS; i++)
651  		counter_u64_free(ifp->if_counters[i]);
652  
653  	if_freedescr(ifp->if_description);
654  	free(ifp->if_hw_addr, M_IFADDR);
655  	free(ifp, M_IFNET);
656  }
657  
658  /*
659   * Deregister an interface and free the associated storage.
660   */
661  void
if_free(struct ifnet * ifp)662  if_free(struct ifnet *ifp)
663  {
664  
665  	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
666  
667  	/*
668  	 * XXXGL: An interface index is really an alias to ifp pointer.
669  	 * Why would we clear the alias now, and not in the deferred
670  	 * context?  Indeed there is nothing wrong with some network
671  	 * thread obtaining ifp via ifnet_byindex() inside the network
672  	 * epoch and then dereferencing ifp while we perform if_free(),
673  	 * and after if_free() finished, too.
674  	 *
675  	 * This early index freeing was important back when ifindex was
676  	 * virtualized and interface would outlive the vnet.
677  	 */
678  	IFNET_WLOCK();
679  	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
680  	ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL);
681  	ifindex_table[ifp->if_index].ife_gencnt++;
682  	while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL)
683  		if_index--;
684  	IFNET_WUNLOCK();
685  
686  	if (refcount_release(&ifp->if_refcount))
687  		NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
688  }
689  
690  /*
691   * Interfaces to keep an ifnet type-stable despite the possibility of the
692   * driver calling if_free().  If there are additional references, we defer
693   * freeing the underlying data structure.
694   */
695  void
if_ref(struct ifnet * ifp)696  if_ref(struct ifnet *ifp)
697  {
698  	u_int old __diagused;
699  
700  	/* We don't assert the ifnet list lock here, but arguably should. */
701  	old = refcount_acquire(&ifp->if_refcount);
702  	KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp));
703  }
704  
705  bool
if_try_ref(struct ifnet * ifp)706  if_try_ref(struct ifnet *ifp)
707  {
708  	NET_EPOCH_ASSERT();
709  	return (refcount_acquire_if_not_zero(&ifp->if_refcount));
710  }
711  
712  void
if_rele(struct ifnet * ifp)713  if_rele(struct ifnet *ifp)
714  {
715  
716  	if (!refcount_release(&ifp->if_refcount))
717  		return;
718  	NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
719  }
720  
721  void
ifq_init(struct ifaltq * ifq,struct ifnet * ifp)722  ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
723  {
724  
725  	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
726  
727  	if (ifq->ifq_maxlen == 0)
728  		ifq->ifq_maxlen = ifqmaxlen;
729  
730  	ifq->altq_type = 0;
731  	ifq->altq_disc = NULL;
732  	ifq->altq_flags &= ALTQF_CANTCHANGE;
733  	ifq->altq_tbr  = NULL;
734  	ifq->altq_ifp  = ifp;
735  }
736  
737  void
ifq_delete(struct ifaltq * ifq)738  ifq_delete(struct ifaltq *ifq)
739  {
740  	mtx_destroy(&ifq->ifq_mtx);
741  }
742  
743  /*
744   * Perform generic interface initialization tasks and attach the interface
745   * to the list of "active" interfaces.  If vmove flag is set on entry
746   * to if_attach_internal(), perform only a limited subset of initialization
747   * tasks, given that we are moving from one vnet to another an ifnet which
748   * has already been fully initialized.
749   *
750   * Note that if_detach_internal() removes group membership unconditionally
751   * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
752   * Thus, when if_vmove() is applied to a cloned interface, group membership
753   * is lost while a cloned one always joins a group whose name is
754   * ifc->ifc_name.  To recover this after if_detach_internal() and
755   * if_attach_internal(), the cloner should be specified to
756   * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
757   * attempts to join a group whose name is ifc->ifc_name.
758   *
759   * XXX:
760   *  - The decision to return void and thus require this function to
761   *    succeed is questionable.
762   *  - We should probably do more sanity checking.  For instance we don't
763   *    do anything to insure if_xname is unique or non-empty.
764   */
765  void
if_attach(struct ifnet * ifp)766  if_attach(struct ifnet *ifp)
767  {
768  
769  	if_attach_internal(ifp, false);
770  }
771  
772  /*
773   * Compute the least common TSO limit.
774   */
775  void
if_hw_tsomax_common(if_t ifp,struct ifnet_hw_tsomax * pmax)776  if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
777  {
778  	/*
779  	 * 1) If there is no limit currently, take the limit from
780  	 * the network adapter.
781  	 *
782  	 * 2) If the network adapter has a limit below the current
783  	 * limit, apply it.
784  	 */
785  	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
786  	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
787  		pmax->tsomaxbytes = ifp->if_hw_tsomax;
788  	}
789  	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
790  	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
791  		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
792  	}
793  	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
794  	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
795  		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
796  	}
797  }
798  
799  /*
800   * Update TSO limit of a network adapter.
801   *
802   * Returns zero if no change. Else non-zero.
803   */
804  int
if_hw_tsomax_update(if_t ifp,struct ifnet_hw_tsomax * pmax)805  if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
806  {
807  	int retval = 0;
808  	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
809  		ifp->if_hw_tsomax = pmax->tsomaxbytes;
810  		retval++;
811  	}
812  	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
813  		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
814  		retval++;
815  	}
816  	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
817  		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
818  		retval++;
819  	}
820  	return (retval);
821  }
822  
823  static void
if_attach_internal(struct ifnet * ifp,bool vmove)824  if_attach_internal(struct ifnet *ifp, bool vmove)
825  {
826  	unsigned socksize, ifasize;
827  	int namelen, masklen;
828  	struct sockaddr_dl *sdl;
829  	struct ifaddr *ifa;
830  
831  	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
832  
833  #ifdef VIMAGE
834  	CURVNET_ASSERT_SET();
835  	ifp->if_vnet = curvnet;
836  	if (ifp->if_home_vnet == NULL)
837  		ifp->if_home_vnet = curvnet;
838  #endif
839  
840  	if_addgroup(ifp, IFG_ALL);
841  
842  #ifdef VIMAGE
843  	/* Restore group membership for cloned interface. */
844  	if (vmove)
845  		if_clone_restoregroup(ifp);
846  #endif
847  
848  	getmicrotime(&ifp->if_lastchange);
849  	ifp->if_epoch = time_uptime;
850  
851  	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
852  	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
853  	    ("transmit and qflush must both either be set or both be NULL"));
854  	if (ifp->if_transmit == NULL) {
855  		ifp->if_transmit = if_transmit_default;
856  		ifp->if_qflush = if_qflush;
857  	}
858  	if (ifp->if_input == NULL)
859  		ifp->if_input = if_input_default;
860  
861  	if (ifp->if_requestencap == NULL)
862  		ifp->if_requestencap = if_requestencap_default;
863  
864  	if (!vmove) {
865  #ifdef MAC
866  		mac_ifnet_create(ifp);
867  #endif
868  
869  		/*
870  		 * Create a Link Level name for this device.
871  		 */
872  		namelen = strlen(ifp->if_xname);
873  		/*
874  		 * Always save enough space for any possiable name so we
875  		 * can do a rename in place later.
876  		 */
877  		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
878  		socksize = masklen + ifp->if_addrlen;
879  		if (socksize < sizeof(*sdl))
880  			socksize = sizeof(*sdl);
881  		socksize = roundup2(socksize, sizeof(long));
882  		ifasize = sizeof(*ifa) + 2 * socksize;
883  		ifa = ifa_alloc(ifasize, M_WAITOK);
884  		sdl = (struct sockaddr_dl *)(ifa + 1);
885  		sdl->sdl_len = socksize;
886  		sdl->sdl_family = AF_LINK;
887  		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
888  		sdl->sdl_nlen = namelen;
889  		sdl->sdl_index = ifp->if_index;
890  		sdl->sdl_type = ifp->if_type;
891  		ifp->if_addr = ifa;
892  		ifa->ifa_ifp = ifp;
893  		ifa->ifa_addr = (struct sockaddr *)sdl;
894  		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
895  		ifa->ifa_netmask = (struct sockaddr *)sdl;
896  		sdl->sdl_len = masklen;
897  		while (namelen != 0)
898  			sdl->sdl_data[--namelen] = 0xff;
899  		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
900  		/* Reliably crash if used uninitialized. */
901  		ifp->if_broadcastaddr = NULL;
902  
903  		if (ifp->if_type == IFT_ETHER) {
904  			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
905  			    M_WAITOK | M_ZERO);
906  		}
907  
908  #if defined(INET) || defined(INET6)
909  		/* Use defaults for TSO, if nothing is set */
910  		if (ifp->if_hw_tsomax == 0 &&
911  		    ifp->if_hw_tsomaxsegcount == 0 &&
912  		    ifp->if_hw_tsomaxsegsize == 0) {
913  			/*
914  			 * The TSO defaults needs to be such that an
915  			 * NFS mbuf list of 35 mbufs totalling just
916  			 * below 64K works and that a chain of mbufs
917  			 * can be defragged into at most 32 segments:
918  			 */
919  			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
920  			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
921  			ifp->if_hw_tsomaxsegcount = 35;
922  			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
923  
924  			/* XXX some drivers set IFCAP_TSO after ethernet attach */
925  			if (ifp->if_capabilities & IFCAP_TSO) {
926  				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
927  				    ifp->if_hw_tsomax,
928  				    ifp->if_hw_tsomaxsegcount,
929  				    ifp->if_hw_tsomaxsegsize);
930  			}
931  		}
932  #endif
933  	}
934  #ifdef VIMAGE
935  	else {
936  		/*
937  		 * Update the interface index in the link layer address
938  		 * of the interface.
939  		 */
940  		for (ifa = ifp->if_addr; ifa != NULL;
941  		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
942  			if (ifa->ifa_addr->sa_family == AF_LINK) {
943  				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
944  				sdl->sdl_index = ifp->if_index;
945  			}
946  		}
947  	}
948  #endif
949  
950  	if_link_ifnet(ifp);
951  
952  	if (domain_init_status >= 2)
953  		if_attachdomain1(ifp);
954  
955  	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
956  	if (IS_DEFAULT_VNET(curvnet))
957  		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
958  }
959  
960  static void
if_epochalloc(void * dummy __unused)961  if_epochalloc(void *dummy __unused)
962  {
963  
964  	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
965  }
966  SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
967  
968  static void
if_attachdomain(void * dummy)969  if_attachdomain(void *dummy)
970  {
971  	struct ifnet *ifp;
972  
973  	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
974  		if_attachdomain1(ifp);
975  }
976  SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
977      if_attachdomain, NULL);
978  
979  static void
if_attachdomain1(struct ifnet * ifp)980  if_attachdomain1(struct ifnet *ifp)
981  {
982  	struct domain *dp;
983  
984  	/*
985  	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
986  	 * cannot lock ifp->if_afdata initialization, entirely.
987  	 */
988  	IF_AFDATA_LOCK(ifp);
989  	if (ifp->if_afdata_initialized >= domain_init_status) {
990  		IF_AFDATA_UNLOCK(ifp);
991  		log(LOG_WARNING, "%s called more than once on %s\n",
992  		    __func__, ifp->if_xname);
993  		return;
994  	}
995  	ifp->if_afdata_initialized = domain_init_status;
996  	IF_AFDATA_UNLOCK(ifp);
997  
998  	/* address family dependent data region */
999  	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
1000  	SLIST_FOREACH(dp, &domains, dom_next) {
1001  		if (dp->dom_ifattach)
1002  			ifp->if_afdata[dp->dom_family] =
1003  			    (*dp->dom_ifattach)(ifp);
1004  	}
1005  }
1006  
1007  /*
1008   * Remove any unicast or broadcast network addresses from an interface.
1009   */
1010  void
if_purgeaddrs(struct ifnet * ifp)1011  if_purgeaddrs(struct ifnet *ifp)
1012  {
1013  	struct ifaddr *ifa;
1014  
1015  #ifdef INET6
1016  	/*
1017  	 * Need to leave multicast addresses of proxy NDP llentries
1018  	 * before in6_purgeifaddr() because the llentries are keys
1019  	 * for in6_multi objects of proxy NDP entries.
1020  	 * in6_purgeifaddr()s clean up llentries including proxy NDPs
1021  	 * then we would lose the keys if they are called earlier.
1022  	 */
1023  	in6_purge_proxy_ndp(ifp);
1024  #endif
1025  	while (1) {
1026  		struct epoch_tracker et;
1027  
1028  		NET_EPOCH_ENTER(et);
1029  		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1030  			if (ifa->ifa_addr->sa_family != AF_LINK)
1031  				break;
1032  		}
1033  		NET_EPOCH_EXIT(et);
1034  
1035  		if (ifa == NULL)
1036  			break;
1037  #ifdef INET
1038  		/* XXX: Ugly!! ad hoc just for INET */
1039  		if (ifa->ifa_addr->sa_family == AF_INET) {
1040  			struct ifreq ifr;
1041  
1042  			bzero(&ifr, sizeof(ifr));
1043  			ifr.ifr_addr = *ifa->ifa_addr;
1044  			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1045  			    NULL) == 0)
1046  				continue;
1047  		}
1048  #endif /* INET */
1049  #ifdef INET6
1050  		if (ifa->ifa_addr->sa_family == AF_INET6) {
1051  			in6_purgeifaddr((struct in6_ifaddr *)ifa);
1052  			/* ifp_addrhead is already updated */
1053  			continue;
1054  		}
1055  #endif /* INET6 */
1056  		IF_ADDR_WLOCK(ifp);
1057  		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1058  		IF_ADDR_WUNLOCK(ifp);
1059  		ifa_free(ifa);
1060  	}
1061  }
1062  
1063  /*
1064   * Remove any multicast network addresses from an interface when an ifnet
1065   * is going away.
1066   */
1067  static void
if_purgemaddrs(struct ifnet * ifp)1068  if_purgemaddrs(struct ifnet *ifp)
1069  {
1070  	struct ifmultiaddr *ifma;
1071  
1072  	IF_ADDR_WLOCK(ifp);
1073  	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1074  		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1075  		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1076  		if_delmulti_locked(ifp, ifma, 1);
1077  	}
1078  	IF_ADDR_WUNLOCK(ifp);
1079  }
1080  
1081  /*
1082   * Detach an interface, removing it from the list of "active" interfaces.
1083   * If vmove flag is set on entry to if_detach_internal(), perform only a
1084   * limited subset of cleanup tasks, given that we are moving an ifnet from
1085   * one vnet to another, where it must be fully operational.
1086   *
1087   * XXXRW: There are some significant questions about event ordering, and
1088   * how to prevent things from starting to use the interface during detach.
1089   */
1090  void
if_detach(struct ifnet * ifp)1091  if_detach(struct ifnet *ifp)
1092  {
1093  	bool found;
1094  
1095  	CURVNET_SET_QUIET(ifp->if_vnet);
1096  	found = if_unlink_ifnet(ifp, false);
1097  	if (found) {
1098  		sx_xlock(&ifnet_detach_sxlock);
1099  		if_detach_internal(ifp, false);
1100  		sx_xunlock(&ifnet_detach_sxlock);
1101  	}
1102  	CURVNET_RESTORE();
1103  }
1104  
1105  /*
1106   * The vmove flag, if set, indicates that we are called from a callpath
1107   * that is moving an interface to a different vnet instance.
1108   *
1109   * The shutdown flag, if set, indicates that we are called in the
1110   * process of shutting down a vnet instance.  Currently only the
1111   * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1112   * on a vnet instance shutdown without this flag being set, e.g., when
1113   * the cloned interfaces are destoyed as first thing of teardown.
1114   */
1115  static void
if_detach_internal(struct ifnet * ifp,bool vmove)1116  if_detach_internal(struct ifnet *ifp, bool vmove)
1117  {
1118  	struct ifaddr *ifa;
1119  	int i;
1120  	struct domain *dp;
1121  #ifdef VIMAGE
1122  	bool shutdown;
1123  
1124  	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1125  #endif
1126  
1127  	sx_assert(&ifnet_detach_sxlock, SX_XLOCKED);
1128  
1129  	/*
1130  	 * At this point we know the interface still was on the ifnet list
1131  	 * and we removed it so we are in a stable state.
1132  	 */
1133  	NET_EPOCH_WAIT();
1134  
1135  	/*
1136  	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1137  	 * fixes issues about late destruction of multicast options
1138  	 * which lead to leave group calls, which in turn access the
1139  	 * belonging ifnet structure:
1140  	 */
1141  	NET_EPOCH_DRAIN_CALLBACKS();
1142  
1143  	/*
1144  	 * In any case (destroy or vmove) detach us from the groups
1145  	 * and remove/wait for pending events on the taskq.
1146  	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1147  	 */
1148  	if_delgroups(ifp);
1149  
1150  	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1151  	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1152  
1153  	if_down(ifp);
1154  
1155  #ifdef VIMAGE
1156  	/*
1157  	 * On VNET shutdown abort here as the stack teardown will do all
1158  	 * the work top-down for us.
1159  	 */
1160  	if (shutdown) {
1161  		/* Give interface users the chance to clean up. */
1162  		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1163  
1164  		/*
1165  		 * In case of a vmove we are done here without error.
1166  		 * If we would signal an error it would lead to the same
1167  		 * abort as if we did not find the ifnet anymore.
1168  		 * if_detach() calls us in void context and does not care
1169  		 * about an early abort notification, so life is splendid :)
1170  		 */
1171  		goto finish_vnet_shutdown;
1172  	}
1173  #endif
1174  
1175  	/*
1176  	 * At this point we are not tearing down a VNET and are either
1177  	 * going to destroy or vmove the interface and have to cleanup
1178  	 * accordingly.
1179  	 */
1180  
1181  	/*
1182  	 * Remove routes and flush queues.
1183  	 */
1184  #ifdef ALTQ
1185  	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1186  		altq_disable(&ifp->if_snd);
1187  	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1188  		altq_detach(&ifp->if_snd);
1189  #endif
1190  
1191  	if_purgeaddrs(ifp);
1192  
1193  #ifdef INET
1194  	in_ifdetach(ifp);
1195  #endif
1196  
1197  #ifdef INET6
1198  	/*
1199  	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1200  	 * before removing routing entries below, since IPv6 interface direct
1201  	 * routes are expected to be removed by the IPv6-specific kernel API.
1202  	 * Otherwise, the kernel will detect some inconsistency and bark it.
1203  	 */
1204  	in6_ifdetach(ifp);
1205  #endif
1206  	if_purgemaddrs(ifp);
1207  
1208  	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1209  	if (IS_DEFAULT_VNET(curvnet))
1210  		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1211  
1212  	if (!vmove) {
1213  		/*
1214  		 * Prevent further calls into the device driver via ifnet.
1215  		 */
1216  		if_dead(ifp);
1217  
1218  		/*
1219  		 * Clean up all addresses.
1220  		 */
1221  		IF_ADDR_WLOCK(ifp);
1222  		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1223  			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1224  			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1225  			IF_ADDR_WUNLOCK(ifp);
1226  			ifa_free(ifa);
1227  		} else
1228  			IF_ADDR_WUNLOCK(ifp);
1229  	}
1230  
1231  	rt_flushifroutes(ifp);
1232  
1233  #ifdef VIMAGE
1234  finish_vnet_shutdown:
1235  #endif
1236  	/*
1237  	 * We cannot hold the lock over dom_ifdetach calls as they might
1238  	 * sleep, for example trying to drain a callout, thus open up the
1239  	 * theoretical race with re-attaching.
1240  	 */
1241  	IF_AFDATA_LOCK(ifp);
1242  	i = ifp->if_afdata_initialized;
1243  	ifp->if_afdata_initialized = 0;
1244  	IF_AFDATA_UNLOCK(ifp);
1245  	if (i == 0)
1246  		return;
1247  	SLIST_FOREACH(dp, &domains, dom_next) {
1248  		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1249  			(*dp->dom_ifdetach)(ifp,
1250  			    ifp->if_afdata[dp->dom_family]);
1251  			ifp->if_afdata[dp->dom_family] = NULL;
1252  		}
1253  	}
1254  }
1255  
1256  #ifdef VIMAGE
1257  /*
1258   * if_vmove() performs a limited version of if_detach() in current
1259   * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1260   */
1261  static void
if_vmove(struct ifnet * ifp,struct vnet * new_vnet)1262  if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1263  {
1264  #ifdef DEV_BPF
1265  	/*
1266  	 * Detach BPF file descriptors from its interface.
1267  	 */
1268  	bpf_ifdetach(ifp);
1269  #endif
1270  
1271  	/*
1272  	 * Detach from current vnet, but preserve LLADDR info, do not
1273  	 * mark as dead etc. so that the ifnet can be reattached later.
1274  	 */
1275  	if_detach_internal(ifp, true);
1276  
1277  	/*
1278  	 * Perform interface-specific reassignment tasks, if provided by
1279  	 * the driver.
1280  	 */
1281  	if (ifp->if_reassign != NULL)
1282  		ifp->if_reassign(ifp, new_vnet, NULL);
1283  
1284  	/*
1285  	 * Switch to the context of the target vnet.
1286  	 */
1287  	CURVNET_SET_QUIET(new_vnet);
1288  	if_attach_internal(ifp, true);
1289  	CURVNET_RESTORE();
1290  }
1291  
1292  /*
1293   * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1294   */
1295  static int
if_vmove_loan(struct thread * td,struct ifnet * ifp,char * ifname,int jid)1296  if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1297  {
1298  	struct prison *pr;
1299  	struct ifnet *difp;
1300  	bool found;
1301  	bool shutdown;
1302  
1303  	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
1304  
1305  	/* Try to find the prison within our visibility. */
1306  	sx_slock(&allprison_lock);
1307  	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1308  	sx_sunlock(&allprison_lock);
1309  	if (pr == NULL)
1310  		return (ENXIO);
1311  	prison_hold_locked(pr);
1312  	mtx_unlock(&pr->pr_mtx);
1313  
1314  	/* Do not try to move the iface from and to the same prison. */
1315  	if (pr->pr_vnet == ifp->if_vnet) {
1316  		prison_free(pr);
1317  		return (EEXIST);
1318  	}
1319  
1320  	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1321  	/* XXX Lock interfaces to avoid races. */
1322  	CURVNET_SET_QUIET(pr->pr_vnet);
1323  	difp = ifunit(ifname);
1324  	CURVNET_RESTORE();
1325  	if (difp != NULL) {
1326  		prison_free(pr);
1327  		return (EEXIST);
1328  	}
1329  	sx_xlock(&ifnet_detach_sxlock);
1330  
1331  	/* Make sure the VNET is stable. */
1332  	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1333  	if (shutdown) {
1334  		sx_xunlock(&ifnet_detach_sxlock);
1335  		prison_free(pr);
1336  		return (EBUSY);
1337  	}
1338  
1339  	found = if_unlink_ifnet(ifp, true);
1340  	if (! found) {
1341  		sx_xunlock(&ifnet_detach_sxlock);
1342  		prison_free(pr);
1343  		return (ENODEV);
1344  	}
1345  
1346  	/* Move the interface into the child jail/vnet. */
1347  	if_vmove(ifp, pr->pr_vnet);
1348  
1349  	/* Report the new if_xname back to the userland. */
1350  	sprintf(ifname, "%s", ifp->if_xname);
1351  
1352  	sx_xunlock(&ifnet_detach_sxlock);
1353  
1354  	prison_free(pr);
1355  	return (0);
1356  }
1357  
1358  static int
if_vmove_reclaim(struct thread * td,char * ifname,int jid)1359  if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1360  {
1361  	struct prison *pr;
1362  	struct vnet *vnet_dst;
1363  	struct ifnet *ifp;
1364  	int found __diagused;
1365   	bool shutdown;
1366  
1367  	/* Try to find the prison within our visibility. */
1368  	sx_slock(&allprison_lock);
1369  	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1370  	sx_sunlock(&allprison_lock);
1371  	if (pr == NULL)
1372  		return (ENXIO);
1373  	prison_hold_locked(pr);
1374  	mtx_unlock(&pr->pr_mtx);
1375  
1376  	/* Make sure the named iface exists in the source prison/vnet. */
1377  	CURVNET_SET(pr->pr_vnet);
1378  	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1379  	if (ifp == NULL) {
1380  		CURVNET_RESTORE();
1381  		prison_free(pr);
1382  		return (ENXIO);
1383  	}
1384  
1385  	/* Do not try to move the iface from and to the same prison. */
1386  	vnet_dst = TD_TO_VNET(td);
1387  	if (vnet_dst == ifp->if_vnet) {
1388  		CURVNET_RESTORE();
1389  		prison_free(pr);
1390  		return (EEXIST);
1391  	}
1392  
1393  	/* Make sure the VNET is stable. */
1394  	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1395  	if (shutdown) {
1396  		CURVNET_RESTORE();
1397  		prison_free(pr);
1398  		return (EBUSY);
1399  	}
1400  
1401  	/* Get interface back from child jail/vnet. */
1402  	found = if_unlink_ifnet(ifp, true);
1403  	MPASS(found);
1404  	sx_xlock(&ifnet_detach_sxlock);
1405  	if_vmove(ifp, vnet_dst);
1406  	sx_xunlock(&ifnet_detach_sxlock);
1407  	CURVNET_RESTORE();
1408  
1409  	/* Report the new if_xname back to the userland. */
1410  	sprintf(ifname, "%s", ifp->if_xname);
1411  
1412  	prison_free(pr);
1413  	return (0);
1414  }
1415  #endif /* VIMAGE */
1416  
1417  /*
1418   * Add a group to an interface
1419   */
1420  int
if_addgroup(struct ifnet * ifp,const char * groupname)1421  if_addgroup(struct ifnet *ifp, const char *groupname)
1422  {
1423  	struct ifg_list		*ifgl;
1424  	struct ifg_group	*ifg = NULL;
1425  	struct ifg_member	*ifgm;
1426  	int 			 new = 0;
1427  
1428  	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1429  	    groupname[strlen(groupname) - 1] <= '9')
1430  		return (EINVAL);
1431  
1432  	IFNET_WLOCK();
1433  	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1434  		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1435  			IFNET_WUNLOCK();
1436  			return (EEXIST);
1437  		}
1438  
1439  	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1440  	    	IFNET_WUNLOCK();
1441  		return (ENOMEM);
1442  	}
1443  
1444  	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1445  		free(ifgl, M_TEMP);
1446  		IFNET_WUNLOCK();
1447  		return (ENOMEM);
1448  	}
1449  
1450  	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1451  		if (!strcmp(ifg->ifg_group, groupname))
1452  			break;
1453  
1454  	if (ifg == NULL) {
1455  		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1456  			free(ifgl, M_TEMP);
1457  			free(ifgm, M_TEMP);
1458  			IFNET_WUNLOCK();
1459  			return (ENOMEM);
1460  		}
1461  		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1462  		ifg->ifg_refcnt = 0;
1463  		CK_STAILQ_INIT(&ifg->ifg_members);
1464  		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1465  		new = 1;
1466  	}
1467  
1468  	ifg->ifg_refcnt++;
1469  	ifgl->ifgl_group = ifg;
1470  	ifgm->ifgm_ifp = ifp;
1471  
1472  	IF_ADDR_WLOCK(ifp);
1473  	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1474  	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1475  	IF_ADDR_WUNLOCK(ifp);
1476  
1477  	IFNET_WUNLOCK();
1478  
1479  	if (new)
1480  		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1481  	EVENTHANDLER_INVOKE(group_change_event, groupname);
1482  
1483  	return (0);
1484  }
1485  
1486  /*
1487   * Helper function to remove a group out of an interface.  Expects the global
1488   * ifnet lock to be write-locked, and drops it before returning.
1489   */
1490  static void
_if_delgroup_locked(struct ifnet * ifp,struct ifg_list * ifgl,const char * groupname)1491  _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1492      const char *groupname)
1493  {
1494  	struct ifg_member *ifgm;
1495  	bool freeifgl;
1496  
1497  	IFNET_WLOCK_ASSERT();
1498  
1499  	IF_ADDR_WLOCK(ifp);
1500  	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1501  	IF_ADDR_WUNLOCK(ifp);
1502  
1503  	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1504  		if (ifgm->ifgm_ifp == ifp) {
1505  			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1506  			    ifg_member, ifgm_next);
1507  			break;
1508  		}
1509  	}
1510  
1511  	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1512  		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1513  		    ifg_next);
1514  		freeifgl = true;
1515  	} else {
1516  		freeifgl = false;
1517  	}
1518  	IFNET_WUNLOCK();
1519  
1520  	NET_EPOCH_WAIT();
1521  	EVENTHANDLER_INVOKE(group_change_event, groupname);
1522  	if (freeifgl) {
1523  		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1524  		free(ifgl->ifgl_group, M_TEMP);
1525  	}
1526  	free(ifgm, M_TEMP);
1527  	free(ifgl, M_TEMP);
1528  }
1529  
1530  /*
1531   * Remove a group from an interface
1532   */
1533  int
if_delgroup(struct ifnet * ifp,const char * groupname)1534  if_delgroup(struct ifnet *ifp, const char *groupname)
1535  {
1536  	struct ifg_list *ifgl;
1537  
1538  	IFNET_WLOCK();
1539  	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1540  		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1541  			break;
1542  	if (ifgl == NULL) {
1543  		IFNET_WUNLOCK();
1544  		return (ENOENT);
1545  	}
1546  
1547  	_if_delgroup_locked(ifp, ifgl, groupname);
1548  
1549  	return (0);
1550  }
1551  
1552  /*
1553   * Remove an interface from all groups
1554   */
1555  static void
if_delgroups(struct ifnet * ifp)1556  if_delgroups(struct ifnet *ifp)
1557  {
1558  	struct ifg_list *ifgl;
1559  	char groupname[IFNAMSIZ];
1560  
1561  	IFNET_WLOCK();
1562  	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1563  		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1564  		_if_delgroup_locked(ifp, ifgl, groupname);
1565  		IFNET_WLOCK();
1566  	}
1567  	IFNET_WUNLOCK();
1568  }
1569  
1570  /*
1571   * Stores all groups from an interface in memory pointed to by ifgr.
1572   */
1573  static int
if_getgroup(struct ifgroupreq * ifgr,struct ifnet * ifp)1574  if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1575  {
1576  	int			 len, error;
1577  	struct ifg_list		*ifgl;
1578  	struct ifg_req		 ifgrq, *ifgp;
1579  
1580  	NET_EPOCH_ASSERT();
1581  
1582  	if (ifgr->ifgr_len == 0) {
1583  		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1584  			ifgr->ifgr_len += sizeof(struct ifg_req);
1585  		return (0);
1586  	}
1587  
1588  	len = ifgr->ifgr_len;
1589  	ifgp = ifgr->ifgr_groups;
1590  	/* XXX: wire */
1591  	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1592  		if (len < sizeof(ifgrq))
1593  			return (EINVAL);
1594  		bzero(&ifgrq, sizeof ifgrq);
1595  		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1596  		    sizeof(ifgrq.ifgrq_group));
1597  		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1598  			return (error);
1599  		len -= sizeof(ifgrq);
1600  		ifgp++;
1601  	}
1602  
1603  	return (0);
1604  }
1605  
1606  /*
1607   * Stores all members of a group in memory pointed to by igfr
1608   */
1609  static int
if_getgroupmembers(struct ifgroupreq * ifgr)1610  if_getgroupmembers(struct ifgroupreq *ifgr)
1611  {
1612  	struct ifg_group	*ifg;
1613  	struct ifg_member	*ifgm;
1614  	struct ifg_req		 ifgrq, *ifgp;
1615  	int			 len, error;
1616  
1617  	IFNET_RLOCK();
1618  	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1619  		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1620  			break;
1621  	if (ifg == NULL) {
1622  		IFNET_RUNLOCK();
1623  		return (ENOENT);
1624  	}
1625  
1626  	if (ifgr->ifgr_len == 0) {
1627  		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1628  			ifgr->ifgr_len += sizeof(ifgrq);
1629  		IFNET_RUNLOCK();
1630  		return (0);
1631  	}
1632  
1633  	len = ifgr->ifgr_len;
1634  	ifgp = ifgr->ifgr_groups;
1635  	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1636  		if (len < sizeof(ifgrq)) {
1637  			IFNET_RUNLOCK();
1638  			return (EINVAL);
1639  		}
1640  		bzero(&ifgrq, sizeof ifgrq);
1641  		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1642  		    sizeof(ifgrq.ifgrq_member));
1643  		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1644  			IFNET_RUNLOCK();
1645  			return (error);
1646  		}
1647  		len -= sizeof(ifgrq);
1648  		ifgp++;
1649  	}
1650  	IFNET_RUNLOCK();
1651  
1652  	return (0);
1653  }
1654  
1655  /*
1656   * Return counter values from counter(9)s stored in ifnet.
1657   */
1658  uint64_t
if_get_counter_default(struct ifnet * ifp,ift_counter cnt)1659  if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1660  {
1661  
1662  	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1663  
1664  	return (counter_u64_fetch(ifp->if_counters[cnt]));
1665  }
1666  
1667  /*
1668   * Increase an ifnet counter. Usually used for counters shared
1669   * between the stack and a driver, but function supports them all.
1670   */
1671  void
if_inc_counter(struct ifnet * ifp,ift_counter cnt,int64_t inc)1672  if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1673  {
1674  
1675  	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1676  
1677  	counter_u64_add(ifp->if_counters[cnt], inc);
1678  }
1679  
1680  /*
1681   * Copy data from ifnet to userland API structure if_data.
1682   */
1683  void
if_data_copy(struct ifnet * ifp,struct if_data * ifd)1684  if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1685  {
1686  
1687  	ifd->ifi_type = ifp->if_type;
1688  	ifd->ifi_physical = 0;
1689  	ifd->ifi_addrlen = ifp->if_addrlen;
1690  	ifd->ifi_hdrlen = ifp->if_hdrlen;
1691  	ifd->ifi_link_state = ifp->if_link_state;
1692  	ifd->ifi_vhid = 0;
1693  	ifd->ifi_datalen = sizeof(struct if_data);
1694  	ifd->ifi_mtu = ifp->if_mtu;
1695  	ifd->ifi_metric = ifp->if_metric;
1696  	ifd->ifi_baudrate = ifp->if_baudrate;
1697  	ifd->ifi_hwassist = ifp->if_hwassist;
1698  	ifd->ifi_epoch = ifp->if_epoch;
1699  	ifd->ifi_lastchange = ifp->if_lastchange;
1700  
1701  	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1702  	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1703  	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1704  	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1705  	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1706  	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1707  	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1708  	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1709  	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1710  	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1711  	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1712  	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1713  }
1714  
1715  /*
1716   * Initialization, destruction and refcounting functions for ifaddrs.
1717   */
1718  struct ifaddr *
ifa_alloc(size_t size,int flags)1719  ifa_alloc(size_t size, int flags)
1720  {
1721  	struct ifaddr *ifa;
1722  
1723  	KASSERT(size >= sizeof(struct ifaddr),
1724  	    ("%s: invalid size %zu", __func__, size));
1725  
1726  	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1727  	if (ifa == NULL)
1728  		return (NULL);
1729  
1730  	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1731  		goto fail;
1732  	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1733  		goto fail;
1734  	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1735  		goto fail;
1736  	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1737  		goto fail;
1738  
1739  	refcount_init(&ifa->ifa_refcnt, 1);
1740  
1741  	return (ifa);
1742  
1743  fail:
1744  	/* free(NULL) is okay */
1745  	counter_u64_free(ifa->ifa_opackets);
1746  	counter_u64_free(ifa->ifa_ipackets);
1747  	counter_u64_free(ifa->ifa_obytes);
1748  	counter_u64_free(ifa->ifa_ibytes);
1749  	free(ifa, M_IFADDR);
1750  
1751  	return (NULL);
1752  }
1753  
1754  void
ifa_ref(struct ifaddr * ifa)1755  ifa_ref(struct ifaddr *ifa)
1756  {
1757  	u_int old __diagused;
1758  
1759  	old = refcount_acquire(&ifa->ifa_refcnt);
1760  	KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa));
1761  }
1762  
1763  int
ifa_try_ref(struct ifaddr * ifa)1764  ifa_try_ref(struct ifaddr *ifa)
1765  {
1766  
1767  	NET_EPOCH_ASSERT();
1768  	return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt));
1769  }
1770  
1771  static void
ifa_destroy(epoch_context_t ctx)1772  ifa_destroy(epoch_context_t ctx)
1773  {
1774  	struct ifaddr *ifa;
1775  
1776  	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1777  	counter_u64_free(ifa->ifa_opackets);
1778  	counter_u64_free(ifa->ifa_ipackets);
1779  	counter_u64_free(ifa->ifa_obytes);
1780  	counter_u64_free(ifa->ifa_ibytes);
1781  	free(ifa, M_IFADDR);
1782  }
1783  
1784  void
ifa_free(struct ifaddr * ifa)1785  ifa_free(struct ifaddr *ifa)
1786  {
1787  
1788  	if (refcount_release(&ifa->ifa_refcnt))
1789  		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1790  }
1791  
1792  /*
1793   * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1794   * structs used to represent other address families, it is necessary
1795   * to perform a different comparison.
1796   */
1797  static bool
sa_dl_equal(const struct sockaddr * a,const struct sockaddr * b)1798  sa_dl_equal(const struct sockaddr *a, const struct sockaddr *b)
1799  {
1800  	const struct sockaddr_dl *sdl1 = (const struct sockaddr_dl *)a;
1801  	const struct sockaddr_dl *sdl2 = (const struct sockaddr_dl *)b;
1802  
1803  	return (sdl1->sdl_len == sdl2->sdl_len &&
1804  	    bcmp(sdl1->sdl_data + sdl1->sdl_nlen,
1805  	    sdl2->sdl_data + sdl2->sdl_nlen, sdl1->sdl_alen) == 0);
1806  }
1807  
1808  /*
1809   * Locate an interface based on a complete address.
1810   */
1811  /*ARGSUSED*/
1812  struct ifaddr *
ifa_ifwithaddr(const struct sockaddr * addr)1813  ifa_ifwithaddr(const struct sockaddr *addr)
1814  {
1815  	struct ifnet *ifp;
1816  	struct ifaddr *ifa;
1817  
1818  	NET_EPOCH_ASSERT();
1819  
1820  	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1821  		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1822  			if (ifa->ifa_addr->sa_family != addr->sa_family)
1823  				continue;
1824  			if (sa_equal(addr, ifa->ifa_addr)) {
1825  				goto done;
1826  			}
1827  			/* IP6 doesn't have broadcast */
1828  			if ((ifp->if_flags & IFF_BROADCAST) &&
1829  			    ifa->ifa_broadaddr &&
1830  			    ifa->ifa_broadaddr->sa_len != 0 &&
1831  			    sa_equal(ifa->ifa_broadaddr, addr)) {
1832  				goto done;
1833  			}
1834  		}
1835  	}
1836  	ifa = NULL;
1837  done:
1838  	return (ifa);
1839  }
1840  
1841  int
ifa_ifwithaddr_check(const struct sockaddr * addr)1842  ifa_ifwithaddr_check(const struct sockaddr *addr)
1843  {
1844  	struct epoch_tracker et;
1845  	int rc;
1846  
1847  	NET_EPOCH_ENTER(et);
1848  	rc = (ifa_ifwithaddr(addr) != NULL);
1849  	NET_EPOCH_EXIT(et);
1850  	return (rc);
1851  }
1852  
1853  /*
1854   * Locate an interface based on the broadcast address.
1855   */
1856  /* ARGSUSED */
1857  struct ifaddr *
ifa_ifwithbroadaddr(const struct sockaddr * addr,int fibnum)1858  ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1859  {
1860  	struct ifnet *ifp;
1861  	struct ifaddr *ifa;
1862  
1863  	NET_EPOCH_ASSERT();
1864  	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1865  		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1866  			continue;
1867  		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1868  			if (ifa->ifa_addr->sa_family != addr->sa_family)
1869  				continue;
1870  			if ((ifp->if_flags & IFF_BROADCAST) &&
1871  			    ifa->ifa_broadaddr &&
1872  			    ifa->ifa_broadaddr->sa_len != 0 &&
1873  			    sa_equal(ifa->ifa_broadaddr, addr)) {
1874  				goto done;
1875  			}
1876  		}
1877  	}
1878  	ifa = NULL;
1879  done:
1880  	return (ifa);
1881  }
1882  
1883  /*
1884   * Locate the point to point interface with a given destination address.
1885   */
1886  /*ARGSUSED*/
1887  struct ifaddr *
ifa_ifwithdstaddr(const struct sockaddr * addr,int fibnum)1888  ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1889  {
1890  	struct ifnet *ifp;
1891  	struct ifaddr *ifa;
1892  
1893  	NET_EPOCH_ASSERT();
1894  	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1895  		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1896  			continue;
1897  		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1898  			continue;
1899  		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1900  			if (ifa->ifa_addr->sa_family != addr->sa_family)
1901  				continue;
1902  			if (ifa->ifa_dstaddr != NULL &&
1903  			    sa_equal(addr, ifa->ifa_dstaddr)) {
1904  				goto done;
1905  			}
1906  		}
1907  	}
1908  	ifa = NULL;
1909  done:
1910  	return (ifa);
1911  }
1912  
1913  /*
1914   * Find an interface on a specific network.  If many, choice
1915   * is most specific found.
1916   */
1917  struct ifaddr *
ifa_ifwithnet(const struct sockaddr * addr,int ignore_ptp,int fibnum)1918  ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1919  {
1920  	struct ifnet *ifp;
1921  	struct ifaddr *ifa;
1922  	struct ifaddr *ifa_maybe = NULL;
1923  	u_int af = addr->sa_family;
1924  	const char *addr_data = addr->sa_data, *cplim;
1925  
1926  	NET_EPOCH_ASSERT();
1927  	/*
1928  	 * AF_LINK addresses can be looked up directly by their index number,
1929  	 * so do that if we can.
1930  	 */
1931  	if (af == AF_LINK) {
1932  		ifp = ifnet_byindex(
1933  		    ((const struct sockaddr_dl *)addr)->sdl_index);
1934  		return (ifp ? ifp->if_addr : NULL);
1935  	}
1936  
1937  	/*
1938  	 * Scan though each interface, looking for ones that have addresses
1939  	 * in this address family and the requested fib.
1940  	 */
1941  	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1942  		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1943  			continue;
1944  		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1945  			const char *cp, *cp2, *cp3;
1946  
1947  			if (ifa->ifa_addr->sa_family != af)
1948  next:				continue;
1949  			if (af == AF_INET &&
1950  			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1951  				/*
1952  				 * This is a bit broken as it doesn't
1953  				 * take into account that the remote end may
1954  				 * be a single node in the network we are
1955  				 * looking for.
1956  				 * The trouble is that we don't know the
1957  				 * netmask for the remote end.
1958  				 */
1959  				if (ifa->ifa_dstaddr != NULL &&
1960  				    sa_equal(addr, ifa->ifa_dstaddr)) {
1961  					goto done;
1962  				}
1963  			} else {
1964  				/*
1965  				 * Scan all the bits in the ifa's address.
1966  				 * If a bit dissagrees with what we are
1967  				 * looking for, mask it with the netmask
1968  				 * to see if it really matters.
1969  				 * (A byte at a time)
1970  				 */
1971  				if (ifa->ifa_netmask == 0)
1972  					continue;
1973  				cp = addr_data;
1974  				cp2 = ifa->ifa_addr->sa_data;
1975  				cp3 = ifa->ifa_netmask->sa_data;
1976  				cplim = ifa->ifa_netmask->sa_len
1977  					+ (char *)ifa->ifa_netmask;
1978  				while (cp3 < cplim)
1979  					if ((*cp++ ^ *cp2++) & *cp3++)
1980  						goto next; /* next address! */
1981  				/*
1982  				 * If the netmask of what we just found
1983  				 * is more specific than what we had before
1984  				 * (if we had one), or if the virtual status
1985  				 * of new prefix is better than of the old one,
1986  				 * then remember the new one before continuing
1987  				 * to search for an even better one.
1988  				 */
1989  				if (ifa_maybe == NULL ||
1990  				    ifa_preferred(ifa_maybe, ifa) ||
1991  				    rn_refines((caddr_t)ifa->ifa_netmask,
1992  				    (caddr_t)ifa_maybe->ifa_netmask)) {
1993  					ifa_maybe = ifa;
1994  				}
1995  			}
1996  		}
1997  	}
1998  	ifa = ifa_maybe;
1999  	ifa_maybe = NULL;
2000  done:
2001  	return (ifa);
2002  }
2003  
2004  /*
2005   * Find an interface address specific to an interface best matching
2006   * a given address.
2007   */
2008  struct ifaddr *
ifaof_ifpforaddr(const struct sockaddr * addr,struct ifnet * ifp)2009  ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2010  {
2011  	struct ifaddr *ifa;
2012  	const char *cp, *cp2, *cp3;
2013  	char *cplim;
2014  	struct ifaddr *ifa_maybe = NULL;
2015  	u_int af = addr->sa_family;
2016  
2017  	if (af >= AF_MAX)
2018  		return (NULL);
2019  
2020  	NET_EPOCH_ASSERT();
2021  	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2022  		if (ifa->ifa_addr->sa_family != af)
2023  			continue;
2024  		if (ifa_maybe == NULL)
2025  			ifa_maybe = ifa;
2026  		if (ifa->ifa_netmask == 0) {
2027  			if (sa_equal(addr, ifa->ifa_addr) ||
2028  			    (ifa->ifa_dstaddr &&
2029  			    sa_equal(addr, ifa->ifa_dstaddr)))
2030  				goto done;
2031  			continue;
2032  		}
2033  		if (ifp->if_flags & IFF_POINTOPOINT) {
2034  			if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr))
2035  				goto done;
2036  		} else {
2037  			cp = addr->sa_data;
2038  			cp2 = ifa->ifa_addr->sa_data;
2039  			cp3 = ifa->ifa_netmask->sa_data;
2040  			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2041  			for (; cp3 < cplim; cp3++)
2042  				if ((*cp++ ^ *cp2++) & *cp3)
2043  					break;
2044  			if (cp3 == cplim)
2045  				goto done;
2046  		}
2047  	}
2048  	ifa = ifa_maybe;
2049  done:
2050  	return (ifa);
2051  }
2052  
2053  /*
2054   * See whether new ifa is better than current one:
2055   * 1) A non-virtual one is preferred over virtual.
2056   * 2) A virtual in master state preferred over any other state.
2057   *
2058   * Used in several address selecting functions.
2059   */
2060  int
ifa_preferred(struct ifaddr * cur,struct ifaddr * next)2061  ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2062  {
2063  
2064  	return (cur->ifa_carp && (!next->ifa_carp ||
2065  	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2066  }
2067  
2068  struct sockaddr_dl *
link_alloc_sdl(size_t size,int flags)2069  link_alloc_sdl(size_t size, int flags)
2070  {
2071  
2072  	return (malloc(size, M_TEMP, flags));
2073  }
2074  
2075  void
link_free_sdl(struct sockaddr * sa)2076  link_free_sdl(struct sockaddr *sa)
2077  {
2078  	free(sa, M_TEMP);
2079  }
2080  
2081  /*
2082   * Fills in given sdl with interface basic info.
2083   * Returns pointer to filled sdl.
2084   */
2085  struct sockaddr_dl *
link_init_sdl(struct ifnet * ifp,struct sockaddr * paddr,u_char iftype)2086  link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2087  {
2088  	struct sockaddr_dl *sdl;
2089  
2090  	sdl = (struct sockaddr_dl *)paddr;
2091  	memset(sdl, 0, sizeof(struct sockaddr_dl));
2092  	sdl->sdl_len = sizeof(struct sockaddr_dl);
2093  	sdl->sdl_family = AF_LINK;
2094  	sdl->sdl_index = ifp->if_index;
2095  	sdl->sdl_type = iftype;
2096  
2097  	return (sdl);
2098  }
2099  
2100  /*
2101   * Mark an interface down and notify protocols of
2102   * the transition.
2103   */
2104  static void
if_unroute(struct ifnet * ifp,int flag,int fam)2105  if_unroute(struct ifnet *ifp, int flag, int fam)
2106  {
2107  
2108  	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2109  
2110  	ifp->if_flags &= ~flag;
2111  	getmicrotime(&ifp->if_lastchange);
2112  	ifp->if_qflush(ifp);
2113  
2114  	if (ifp->if_carp)
2115  		(*carp_linkstate_p)(ifp);
2116  	rt_ifmsg(ifp, IFF_UP);
2117  }
2118  
2119  void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2120  void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2121  struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2122  struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2123  int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2124  int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2125  int	(*vlan_setcookie_p)(struct ifnet *, void *);
2126  void	*(*vlan_cookie_p)(struct ifnet *);
2127  
2128  /*
2129   * Handle a change in the interface link state. To avoid LORs
2130   * between driver lock and upper layer locks, as well as possible
2131   * recursions, we post event to taskqueue, and all job
2132   * is done in static do_link_state_change().
2133   */
2134  void
if_link_state_change(struct ifnet * ifp,int link_state)2135  if_link_state_change(struct ifnet *ifp, int link_state)
2136  {
2137  	/* Return if state hasn't changed. */
2138  	if (ifp->if_link_state == link_state)
2139  		return;
2140  
2141  	ifp->if_link_state = link_state;
2142  
2143  	/* XXXGL: reference ifp? */
2144  	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2145  }
2146  
2147  static void
do_link_state_change(void * arg,int pending)2148  do_link_state_change(void *arg, int pending)
2149  {
2150  	struct ifnet *ifp;
2151  	int link_state;
2152  
2153  	ifp = arg;
2154  	link_state = ifp->if_link_state;
2155  
2156  	CURVNET_SET(ifp->if_vnet);
2157  	rt_ifmsg(ifp, 0);
2158  	if (ifp->if_vlantrunk != NULL)
2159  		(*vlan_link_state_p)(ifp);
2160  
2161  	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2162  	    ifp->if_l2com != NULL)
2163  		(*ng_ether_link_state_p)(ifp, link_state);
2164  	if (ifp->if_carp)
2165  		(*carp_linkstate_p)(ifp);
2166  	if (ifp->if_bridge)
2167  		ifp->if_bridge_linkstate(ifp);
2168  	if (ifp->if_lagg)
2169  		(*lagg_linkstate_p)(ifp, link_state);
2170  
2171  	if (IS_DEFAULT_VNET(curvnet))
2172  		devctl_notify("IFNET", ifp->if_xname,
2173  		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2174  		    NULL);
2175  	if (pending > 1)
2176  		if_printf(ifp, "%d link states coalesced\n", pending);
2177  	if (log_link_state_change)
2178  		if_printf(ifp, "link state changed to %s\n",
2179  		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2180  	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2181  	CURVNET_RESTORE();
2182  }
2183  
2184  /*
2185   * Mark an interface down and notify protocols of
2186   * the transition.
2187   */
2188  void
if_down(struct ifnet * ifp)2189  if_down(struct ifnet *ifp)
2190  {
2191  
2192  	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2193  	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2194  }
2195  
2196  /*
2197   * Mark an interface up and notify protocols of
2198   * the transition.
2199   */
2200  void
if_up(struct ifnet * ifp)2201  if_up(struct ifnet *ifp)
2202  {
2203  
2204  	ifp->if_flags |= IFF_UP;
2205  	getmicrotime(&ifp->if_lastchange);
2206  	if (ifp->if_carp)
2207  		(*carp_linkstate_p)(ifp);
2208  	rt_ifmsg(ifp, IFF_UP);
2209  	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2210  }
2211  
2212  /*
2213   * Flush an interface queue.
2214   */
2215  void
if_qflush(struct ifnet * ifp)2216  if_qflush(struct ifnet *ifp)
2217  {
2218  	struct mbuf *m, *n;
2219  	struct ifaltq *ifq;
2220  
2221  	ifq = &ifp->if_snd;
2222  	IFQ_LOCK(ifq);
2223  #ifdef ALTQ
2224  	if (ALTQ_IS_ENABLED(ifq))
2225  		ALTQ_PURGE(ifq);
2226  #endif
2227  	n = ifq->ifq_head;
2228  	while ((m = n) != NULL) {
2229  		n = m->m_nextpkt;
2230  		m_freem(m);
2231  	}
2232  	ifq->ifq_head = 0;
2233  	ifq->ifq_tail = 0;
2234  	ifq->ifq_len = 0;
2235  	IFQ_UNLOCK(ifq);
2236  }
2237  
2238  /*
2239   * Map interface name to interface structure pointer, with or without
2240   * returning a reference.
2241   */
2242  struct ifnet *
ifunit_ref(const char * name)2243  ifunit_ref(const char *name)
2244  {
2245  	struct epoch_tracker et;
2246  	struct ifnet *ifp;
2247  
2248  	NET_EPOCH_ENTER(et);
2249  	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2250  		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2251  		    !(ifp->if_flags & IFF_DYING))
2252  			break;
2253  	}
2254  	if (ifp != NULL) {
2255  		if_ref(ifp);
2256  		MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
2257  	}
2258  
2259  	NET_EPOCH_EXIT(et);
2260  	return (ifp);
2261  }
2262  
2263  struct ifnet *
ifunit(const char * name)2264  ifunit(const char *name)
2265  {
2266  	struct epoch_tracker et;
2267  	struct ifnet *ifp;
2268  
2269  	NET_EPOCH_ENTER(et);
2270  	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2271  		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2272  			break;
2273  	}
2274  	NET_EPOCH_EXIT(et);
2275  	return (ifp);
2276  }
2277  
2278  void *
ifr_buffer_get_buffer(void * data)2279  ifr_buffer_get_buffer(void *data)
2280  {
2281  	union ifreq_union *ifrup;
2282  
2283  	ifrup = data;
2284  #ifdef COMPAT_FREEBSD32
2285  	if (SV_CURPROC_FLAG(SV_ILP32))
2286  		return ((void *)(uintptr_t)
2287  		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2288  #endif
2289  	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2290  }
2291  
2292  static void
ifr_buffer_set_buffer_null(void * data)2293  ifr_buffer_set_buffer_null(void *data)
2294  {
2295  	union ifreq_union *ifrup;
2296  
2297  	ifrup = data;
2298  #ifdef COMPAT_FREEBSD32
2299  	if (SV_CURPROC_FLAG(SV_ILP32))
2300  		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2301  	else
2302  #endif
2303  		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2304  }
2305  
2306  size_t
ifr_buffer_get_length(void * data)2307  ifr_buffer_get_length(void *data)
2308  {
2309  	union ifreq_union *ifrup;
2310  
2311  	ifrup = data;
2312  #ifdef COMPAT_FREEBSD32
2313  	if (SV_CURPROC_FLAG(SV_ILP32))
2314  		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2315  #endif
2316  	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2317  }
2318  
2319  static void
ifr_buffer_set_length(void * data,size_t len)2320  ifr_buffer_set_length(void *data, size_t len)
2321  {
2322  	union ifreq_union *ifrup;
2323  
2324  	ifrup = data;
2325  #ifdef COMPAT_FREEBSD32
2326  	if (SV_CURPROC_FLAG(SV_ILP32))
2327  		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2328  	else
2329  #endif
2330  		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2331  }
2332  
2333  void *
ifr_data_get_ptr(void * ifrp)2334  ifr_data_get_ptr(void *ifrp)
2335  {
2336  	union ifreq_union *ifrup;
2337  
2338  	ifrup = ifrp;
2339  #ifdef COMPAT_FREEBSD32
2340  	if (SV_CURPROC_FLAG(SV_ILP32))
2341  		return ((void *)(uintptr_t)
2342  		    ifrup->ifr32.ifr_ifru.ifru_data);
2343  #endif
2344  		return (ifrup->ifr.ifr_ifru.ifru_data);
2345  }
2346  
2347  struct ifcap_nv_bit_name {
2348  	uint64_t cap_bit;
2349  	const char *cap_name;
2350  };
2351  #define CAPNV(x) {.cap_bit = IFCAP_##x, \
2352      .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) }
2353  const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = {
2354  	CAPNV(RXCSUM),
2355  	CAPNV(TXCSUM),
2356  	CAPNV(NETCONS),
2357  	CAPNV(VLAN_MTU),
2358  	CAPNV(VLAN_HWTAGGING),
2359  	CAPNV(JUMBO_MTU),
2360  	CAPNV(POLLING),
2361  	CAPNV(VLAN_HWCSUM),
2362  	CAPNV(TSO4),
2363  	CAPNV(TSO6),
2364  	CAPNV(LRO),
2365  	CAPNV(WOL_UCAST),
2366  	CAPNV(WOL_MCAST),
2367  	CAPNV(WOL_MAGIC),
2368  	CAPNV(TOE4),
2369  	CAPNV(TOE6),
2370  	CAPNV(VLAN_HWFILTER),
2371  	CAPNV(VLAN_HWTSO),
2372  	CAPNV(LINKSTATE),
2373  	CAPNV(NETMAP),
2374  	CAPNV(RXCSUM_IPV6),
2375  	CAPNV(TXCSUM_IPV6),
2376  	CAPNV(HWSTATS),
2377  	CAPNV(TXRTLMT),
2378  	CAPNV(HWRXTSTMP),
2379  	CAPNV(MEXTPG),
2380  	CAPNV(TXTLS4),
2381  	CAPNV(TXTLS6),
2382  	CAPNV(VXLAN_HWCSUM),
2383  	CAPNV(VXLAN_HWTSO),
2384  	CAPNV(TXTLS_RTLMT),
2385  	{0, NULL}
2386  };
2387  #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \
2388      .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) }
2389  const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = {
2390  	CAP2NV(RXTLS4),
2391  	CAP2NV(RXTLS6),
2392  	CAP2NV(IPSEC_OFFLOAD),
2393  	{0, NULL}
2394  };
2395  #undef CAPNV
2396  #undef CAP2NV
2397  
2398  int
if_capnv_to_capint(const nvlist_t * nv,int * old_cap,const struct ifcap_nv_bit_name * nn,bool all)2399  if_capnv_to_capint(const nvlist_t *nv, int *old_cap,
2400      const struct ifcap_nv_bit_name *nn, bool all)
2401  {
2402  	int i, res;
2403  
2404  	res = 0;
2405  	for (i = 0; nn[i].cap_name != NULL; i++) {
2406  		if (nvlist_exists_bool(nv, nn[i].cap_name)) {
2407  			if (all || nvlist_get_bool(nv, nn[i].cap_name))
2408  				res |= nn[i].cap_bit;
2409  		} else {
2410  			res |= *old_cap & nn[i].cap_bit;
2411  		}
2412  	}
2413  	return (res);
2414  }
2415  
2416  void
if_capint_to_capnv(nvlist_t * nv,const struct ifcap_nv_bit_name * nn,int ifr_cap,int ifr_req)2417  if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn,
2418      int ifr_cap, int ifr_req)
2419  {
2420  	int i;
2421  
2422  	for (i = 0; nn[i].cap_name != NULL; i++) {
2423  		if ((nn[i].cap_bit & ifr_cap) != 0) {
2424  			nvlist_add_bool(nv, nn[i].cap_name,
2425  			    (nn[i].cap_bit & ifr_req) != 0);
2426  		}
2427  	}
2428  }
2429  
2430  /*
2431   * Hardware specific interface ioctls.
2432   */
2433  int
ifhwioctl(u_long cmd,struct ifnet * ifp,caddr_t data,struct thread * td)2434  ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2435  {
2436  	struct ifreq *ifr;
2437  	int error = 0, do_ifup = 0;
2438  	int new_flags, temp_flags;
2439  	size_t descrlen, nvbuflen;
2440  	char *descrbuf;
2441  	char new_name[IFNAMSIZ];
2442  	void *buf;
2443  	nvlist_t *nvcap;
2444  	struct siocsifcapnv_driver_data drv_ioctl_data;
2445  
2446  	ifr = (struct ifreq *)data;
2447  	switch (cmd) {
2448  	case SIOCGIFINDEX:
2449  		ifr->ifr_index = ifp->if_index;
2450  		break;
2451  
2452  	case SIOCGIFFLAGS:
2453  		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2454  		ifr->ifr_flags = temp_flags & 0xffff;
2455  		ifr->ifr_flagshigh = temp_flags >> 16;
2456  		break;
2457  
2458  	case SIOCGIFCAP:
2459  		ifr->ifr_reqcap = ifp->if_capabilities;
2460  		ifr->ifr_curcap = ifp->if_capenable;
2461  		break;
2462  
2463  	case SIOCGIFCAPNV:
2464  		if ((ifp->if_capabilities & IFCAP_NV) == 0) {
2465  			error = EINVAL;
2466  			break;
2467  		}
2468  		buf = NULL;
2469  		nvcap = nvlist_create(0);
2470  		for (;;) {
2471  			if_capint_to_capnv(nvcap, ifcap_nv_bit_names,
2472  			    ifp->if_capabilities, ifp->if_capenable);
2473  			if_capint_to_capnv(nvcap, ifcap2_nv_bit_names,
2474  			    ifp->if_capabilities2, ifp->if_capenable2);
2475  			error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV,
2476  			    __DECONST(caddr_t, nvcap));
2477  			if (error != 0) {
2478  				if_printf(ifp,
2479  			    "SIOCGIFCAPNV driver mistake: nvlist error %d\n",
2480  				    error);
2481  				break;
2482  			}
2483  			buf = nvlist_pack(nvcap, &nvbuflen);
2484  			if (buf == NULL) {
2485  				error = nvlist_error(nvcap);
2486  				if (error == 0)
2487  					error = EDOOFUS;
2488  				break;
2489  			}
2490  			if (nvbuflen > ifr->ifr_cap_nv.buf_length) {
2491  				ifr->ifr_cap_nv.length = nvbuflen;
2492  				ifr->ifr_cap_nv.buffer = NULL;
2493  				error = EFBIG;
2494  				break;
2495  			}
2496  			ifr->ifr_cap_nv.length = nvbuflen;
2497  			error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen);
2498  			break;
2499  		}
2500  		free(buf, M_NVLIST);
2501  		nvlist_destroy(nvcap);
2502  		break;
2503  
2504  	case SIOCGIFDATA:
2505  	{
2506  		struct if_data ifd;
2507  
2508  		/* Ensure uninitialised padding is not leaked. */
2509  		memset(&ifd, 0, sizeof(ifd));
2510  
2511  		if_data_copy(ifp, &ifd);
2512  		error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd));
2513  		break;
2514  	}
2515  
2516  #ifdef MAC
2517  	case SIOCGIFMAC:
2518  		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2519  		break;
2520  #endif
2521  
2522  	case SIOCGIFMETRIC:
2523  		ifr->ifr_metric = ifp->if_metric;
2524  		break;
2525  
2526  	case SIOCGIFMTU:
2527  		ifr->ifr_mtu = ifp->if_mtu;
2528  		break;
2529  
2530  	case SIOCGIFPHYS:
2531  		/* XXXGL: did this ever worked? */
2532  		ifr->ifr_phys = 0;
2533  		break;
2534  
2535  	case SIOCGIFDESCR:
2536  		error = 0;
2537  		sx_slock(&ifdescr_sx);
2538  		if (ifp->if_description == NULL)
2539  			error = ENOMSG;
2540  		else {
2541  			/* space for terminating nul */
2542  			descrlen = strlen(ifp->if_description) + 1;
2543  			if (ifr_buffer_get_length(ifr) < descrlen)
2544  				ifr_buffer_set_buffer_null(ifr);
2545  			else
2546  				error = copyout(ifp->if_description,
2547  				    ifr_buffer_get_buffer(ifr), descrlen);
2548  			ifr_buffer_set_length(ifr, descrlen);
2549  		}
2550  		sx_sunlock(&ifdescr_sx);
2551  		break;
2552  
2553  	case SIOCSIFDESCR:
2554  		error = priv_check(td, PRIV_NET_SETIFDESCR);
2555  		if (error)
2556  			return (error);
2557  
2558  		/*
2559  		 * Copy only (length-1) bytes to make sure that
2560  		 * if_description is always nul terminated.  The
2561  		 * length parameter is supposed to count the
2562  		 * terminating nul in.
2563  		 */
2564  		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2565  			return (ENAMETOOLONG);
2566  		else if (ifr_buffer_get_length(ifr) == 0)
2567  			descrbuf = NULL;
2568  		else {
2569  			descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK);
2570  			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2571  			    ifr_buffer_get_length(ifr) - 1);
2572  			if (error) {
2573  				if_freedescr(descrbuf);
2574  				break;
2575  			}
2576  		}
2577  
2578  		if_setdescr(ifp, descrbuf);
2579  		getmicrotime(&ifp->if_lastchange);
2580  		break;
2581  
2582  	case SIOCGIFFIB:
2583  		ifr->ifr_fib = ifp->if_fib;
2584  		break;
2585  
2586  	case SIOCSIFFIB:
2587  		error = priv_check(td, PRIV_NET_SETIFFIB);
2588  		if (error)
2589  			return (error);
2590  		if (ifr->ifr_fib >= rt_numfibs)
2591  			return (EINVAL);
2592  
2593  		ifp->if_fib = ifr->ifr_fib;
2594  		break;
2595  
2596  	case SIOCSIFFLAGS:
2597  		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2598  		if (error)
2599  			return (error);
2600  		/*
2601  		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2602  		 * check, so we don't need special handling here yet.
2603  		 */
2604  		new_flags = (ifr->ifr_flags & 0xffff) |
2605  		    (ifr->ifr_flagshigh << 16);
2606  		if (ifp->if_flags & IFF_UP &&
2607  		    (new_flags & IFF_UP) == 0) {
2608  			if_down(ifp);
2609  		} else if (new_flags & IFF_UP &&
2610  		    (ifp->if_flags & IFF_UP) == 0) {
2611  			do_ifup = 1;
2612  		}
2613  
2614  		/*
2615  		 * See if the promiscuous mode or allmulti bits are about to
2616  		 * flip.  They require special handling because in-kernel
2617  		 * consumers may indepdently toggle them.
2618  		 */
2619  		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2620  			if (new_flags & IFF_PPROMISC)
2621  				ifp->if_flags |= IFF_PROMISC;
2622  			else if (ifp->if_pcount == 0)
2623  				ifp->if_flags &= ~IFF_PROMISC;
2624  			if (log_promisc_mode_change)
2625                                  if_printf(ifp, "permanently promiscuous mode %s\n",
2626                                      ((new_flags & IFF_PPROMISC) ?
2627                                       "enabled" : "disabled"));
2628  		}
2629  		if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) {
2630  			if (new_flags & IFF_PALLMULTI)
2631  				ifp->if_flags |= IFF_ALLMULTI;
2632  			else if (ifp->if_amcount == 0)
2633  				ifp->if_flags &= ~IFF_ALLMULTI;
2634  		}
2635  		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2636  			(new_flags &~ IFF_CANTCHANGE);
2637  		if (ifp->if_ioctl) {
2638  			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2639  		}
2640  		if (do_ifup)
2641  			if_up(ifp);
2642  		getmicrotime(&ifp->if_lastchange);
2643  		break;
2644  
2645  	case SIOCSIFCAP:
2646  		error = priv_check(td, PRIV_NET_SETIFCAP);
2647  		if (error != 0)
2648  			return (error);
2649  		if (ifp->if_ioctl == NULL)
2650  			return (EOPNOTSUPP);
2651  		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2652  			return (EINVAL);
2653  		error = (*ifp->if_ioctl)(ifp, cmd, data);
2654  		if (error == 0)
2655  			getmicrotime(&ifp->if_lastchange);
2656  		break;
2657  
2658  	case SIOCSIFCAPNV:
2659  		error = priv_check(td, PRIV_NET_SETIFCAP);
2660  		if (error != 0)
2661  			return (error);
2662  		if (ifp->if_ioctl == NULL)
2663  			return (EOPNOTSUPP);
2664  		if ((ifp->if_capabilities & IFCAP_NV) == 0)
2665  			return (EINVAL);
2666  		if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE)
2667  			return (EINVAL);
2668  		nvcap = NULL;
2669  		buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK);
2670  		for (;;) {
2671  			error = copyin(ifr->ifr_cap_nv.buffer, buf,
2672  			    ifr->ifr_cap_nv.length);
2673  			if (error != 0)
2674  				break;
2675  			nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0);
2676  			if (nvcap == NULL) {
2677  				error = EINVAL;
2678  				break;
2679  			}
2680  			drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap,
2681  			    &ifp->if_capenable, ifcap_nv_bit_names, false);
2682  			if ((drv_ioctl_data.reqcap &
2683  			    ~ifp->if_capabilities) != 0) {
2684  				error = EINVAL;
2685  				break;
2686  			}
2687  			drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap,
2688  			    &ifp->if_capenable2, ifcap2_nv_bit_names, false);
2689  			if ((drv_ioctl_data.reqcap2 &
2690  			    ~ifp->if_capabilities2) != 0) {
2691  				error = EINVAL;
2692  				break;
2693  			}
2694  			drv_ioctl_data.nvcap = nvcap;
2695  			error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV,
2696  			    (caddr_t)&drv_ioctl_data);
2697  			break;
2698  		}
2699  		nvlist_destroy(nvcap);
2700  		free(buf, M_TEMP);
2701  		if (error == 0)
2702  			getmicrotime(&ifp->if_lastchange);
2703  		break;
2704  
2705  #ifdef MAC
2706  	case SIOCSIFMAC:
2707  		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2708  		break;
2709  #endif
2710  
2711  	case SIOCSIFNAME:
2712  		error = priv_check(td, PRIV_NET_SETIFNAME);
2713  		if (error)
2714  			return (error);
2715  		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2716  		    NULL);
2717  		if (error != 0)
2718  			return (error);
2719  		error = if_rename(ifp, new_name);
2720  		break;
2721  
2722  #ifdef VIMAGE
2723  	case SIOCSIFVNET:
2724  		error = priv_check(td, PRIV_NET_SETIFVNET);
2725  		if (error)
2726  			return (error);
2727  		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2728  		break;
2729  #endif
2730  
2731  	case SIOCSIFMETRIC:
2732  		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2733  		if (error)
2734  			return (error);
2735  		ifp->if_metric = ifr->ifr_metric;
2736  		getmicrotime(&ifp->if_lastchange);
2737  		break;
2738  
2739  	case SIOCSIFPHYS:
2740  		error = priv_check(td, PRIV_NET_SETIFPHYS);
2741  		if (error)
2742  			return (error);
2743  		if (ifp->if_ioctl == NULL)
2744  			return (EOPNOTSUPP);
2745  		error = (*ifp->if_ioctl)(ifp, cmd, data);
2746  		if (error == 0)
2747  			getmicrotime(&ifp->if_lastchange);
2748  		break;
2749  
2750  	case SIOCSIFMTU:
2751  	{
2752  		u_long oldmtu = ifp->if_mtu;
2753  
2754  		error = priv_check(td, PRIV_NET_SETIFMTU);
2755  		if (error)
2756  			return (error);
2757  		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2758  			return (EINVAL);
2759  		if (ifp->if_ioctl == NULL)
2760  			return (EOPNOTSUPP);
2761  		/* Disallow MTU changes on bridge member interfaces. */
2762  		if (ifp->if_bridge)
2763  			return (EOPNOTSUPP);
2764  		error = (*ifp->if_ioctl)(ifp, cmd, data);
2765  		if (error == 0) {
2766  			getmicrotime(&ifp->if_lastchange);
2767  			rt_ifmsg(ifp, 0);
2768  #ifdef INET
2769  			DEBUGNET_NOTIFY_MTU(ifp);
2770  #endif
2771  		}
2772  		/*
2773  		 * If the link MTU changed, do network layer specific procedure.
2774  		 */
2775  		if (ifp->if_mtu != oldmtu)
2776  			if_notifymtu(ifp);
2777  		break;
2778  	}
2779  
2780  	case SIOCADDMULTI:
2781  	case SIOCDELMULTI:
2782  		if (cmd == SIOCADDMULTI)
2783  			error = priv_check(td, PRIV_NET_ADDMULTI);
2784  		else
2785  			error = priv_check(td, PRIV_NET_DELMULTI);
2786  		if (error)
2787  			return (error);
2788  
2789  		/* Don't allow group membership on non-multicast interfaces. */
2790  		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2791  			return (EOPNOTSUPP);
2792  
2793  		/* Don't let users screw up protocols' entries. */
2794  		if (ifr->ifr_addr.sa_family != AF_LINK)
2795  			return (EINVAL);
2796  
2797  		if (cmd == SIOCADDMULTI) {
2798  			struct epoch_tracker et;
2799  			struct ifmultiaddr *ifma;
2800  
2801  			/*
2802  			 * Userland is only permitted to join groups once
2803  			 * via the if_addmulti() KPI, because it cannot hold
2804  			 * struct ifmultiaddr * between calls. It may also
2805  			 * lose a race while we check if the membership
2806  			 * already exists.
2807  			 */
2808  			NET_EPOCH_ENTER(et);
2809  			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2810  			NET_EPOCH_EXIT(et);
2811  			if (ifma != NULL)
2812  				error = EADDRINUSE;
2813  			else
2814  				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2815  		} else {
2816  			error = if_delmulti(ifp, &ifr->ifr_addr);
2817  		}
2818  		if (error == 0)
2819  			getmicrotime(&ifp->if_lastchange);
2820  		break;
2821  
2822  	case SIOCSIFPHYADDR:
2823  	case SIOCDIFPHYADDR:
2824  #ifdef INET6
2825  	case SIOCSIFPHYADDR_IN6:
2826  #endif
2827  	case SIOCSIFMEDIA:
2828  	case SIOCSIFGENERIC:
2829  		error = priv_check(td, PRIV_NET_HWIOCTL);
2830  		if (error)
2831  			return (error);
2832  		if (ifp->if_ioctl == NULL)
2833  			return (EOPNOTSUPP);
2834  		error = (*ifp->if_ioctl)(ifp, cmd, data);
2835  		if (error == 0)
2836  			getmicrotime(&ifp->if_lastchange);
2837  		break;
2838  
2839  	case SIOCGIFSTATUS:
2840  	case SIOCGIFPSRCADDR:
2841  	case SIOCGIFPDSTADDR:
2842  	case SIOCGIFMEDIA:
2843  	case SIOCGIFXMEDIA:
2844  	case SIOCGIFGENERIC:
2845  	case SIOCGIFRSSKEY:
2846  	case SIOCGIFRSSHASH:
2847  	case SIOCGIFDOWNREASON:
2848  		if (ifp->if_ioctl == NULL)
2849  			return (EOPNOTSUPP);
2850  		error = (*ifp->if_ioctl)(ifp, cmd, data);
2851  		break;
2852  
2853  	case SIOCSIFLLADDR:
2854  		error = priv_check(td, PRIV_NET_SETLLADDR);
2855  		if (error)
2856  			return (error);
2857  		error = if_setlladdr(ifp,
2858  		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2859  		break;
2860  
2861  	case SIOCGHWADDR:
2862  		error = if_gethwaddr(ifp, ifr);
2863  		break;
2864  
2865  	case SIOCAIFGROUP:
2866  		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2867  		if (error)
2868  			return (error);
2869  		error = if_addgroup(ifp,
2870  		    ((struct ifgroupreq *)data)->ifgr_group);
2871  		if (error != 0)
2872  			return (error);
2873  		break;
2874  
2875  	case SIOCGIFGROUP:
2876  	{
2877  		struct epoch_tracker et;
2878  
2879  		NET_EPOCH_ENTER(et);
2880  		error = if_getgroup((struct ifgroupreq *)data, ifp);
2881  		NET_EPOCH_EXIT(et);
2882  		break;
2883  	}
2884  
2885  	case SIOCDIFGROUP:
2886  		error = priv_check(td, PRIV_NET_DELIFGROUP);
2887  		if (error)
2888  			return (error);
2889  		error = if_delgroup(ifp,
2890  		    ((struct ifgroupreq *)data)->ifgr_group);
2891  		if (error != 0)
2892  			return (error);
2893  		break;
2894  
2895  	default:
2896  		error = ENOIOCTL;
2897  		break;
2898  	}
2899  	return (error);
2900  }
2901  
2902  /*
2903   * Interface ioctls.
2904   */
2905  int
ifioctl(struct socket * so,u_long cmd,caddr_t data,struct thread * td)2906  ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2907  {
2908  #ifdef COMPAT_FREEBSD32
2909  	union {
2910  		struct ifconf ifc;
2911  		struct ifdrv ifd;
2912  		struct ifgroupreq ifgr;
2913  		struct ifmediareq ifmr;
2914  	} thunk;
2915  	u_long saved_cmd;
2916  	struct ifconf32 *ifc32;
2917  	struct ifdrv32 *ifd32;
2918  	struct ifgroupreq32 *ifgr32;
2919  	struct ifmediareq32 *ifmr32;
2920  #endif
2921  	struct ifnet *ifp;
2922  	struct ifreq *ifr;
2923  	int error;
2924  	int oif_flags;
2925  #ifdef VIMAGE
2926  	bool shutdown;
2927  #endif
2928  
2929  	CURVNET_SET(so->so_vnet);
2930  #ifdef VIMAGE
2931  	/* Make sure the VNET is stable. */
2932  	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2933  	if (shutdown) {
2934  		CURVNET_RESTORE();
2935  		return (EBUSY);
2936  	}
2937  #endif
2938  
2939  #ifdef COMPAT_FREEBSD32
2940  	saved_cmd = cmd;
2941  	switch (cmd) {
2942  	case SIOCGIFCONF32:
2943  		ifc32 = (struct ifconf32 *)data;
2944  		thunk.ifc.ifc_len = ifc32->ifc_len;
2945  		thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2946  		data = (caddr_t)&thunk.ifc;
2947  		cmd = SIOCGIFCONF;
2948  		break;
2949  	case SIOCGDRVSPEC32:
2950  	case SIOCSDRVSPEC32:
2951  		ifd32 = (struct ifdrv32 *)data;
2952  		memcpy(thunk.ifd.ifd_name, ifd32->ifd_name,
2953  		    sizeof(thunk.ifd.ifd_name));
2954  		thunk.ifd.ifd_cmd = ifd32->ifd_cmd;
2955  		thunk.ifd.ifd_len = ifd32->ifd_len;
2956  		thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data);
2957  		data = (caddr_t)&thunk.ifd;
2958  		cmd = _IOC_NEWTYPE(cmd, struct ifdrv);
2959  		break;
2960  	case SIOCAIFGROUP32:
2961  	case SIOCGIFGROUP32:
2962  	case SIOCDIFGROUP32:
2963  	case SIOCGIFGMEMB32:
2964  		ifgr32 = (struct ifgroupreq32 *)data;
2965  		memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name,
2966  		    sizeof(thunk.ifgr.ifgr_name));
2967  		thunk.ifgr.ifgr_len = ifgr32->ifgr_len;
2968  		switch (cmd) {
2969  		case SIOCAIFGROUP32:
2970  		case SIOCDIFGROUP32:
2971  			memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group,
2972  			    sizeof(thunk.ifgr.ifgr_group));
2973  			break;
2974  		case SIOCGIFGROUP32:
2975  		case SIOCGIFGMEMB32:
2976  			thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups);
2977  			break;
2978  		}
2979  		data = (caddr_t)&thunk.ifgr;
2980  		cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq);
2981  		break;
2982  	case SIOCGIFMEDIA32:
2983  	case SIOCGIFXMEDIA32:
2984  		ifmr32 = (struct ifmediareq32 *)data;
2985  		memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name,
2986  		    sizeof(thunk.ifmr.ifm_name));
2987  		thunk.ifmr.ifm_current = ifmr32->ifm_current;
2988  		thunk.ifmr.ifm_mask = ifmr32->ifm_mask;
2989  		thunk.ifmr.ifm_status = ifmr32->ifm_status;
2990  		thunk.ifmr.ifm_active = ifmr32->ifm_active;
2991  		thunk.ifmr.ifm_count = ifmr32->ifm_count;
2992  		thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist);
2993  		data = (caddr_t)&thunk.ifmr;
2994  		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
2995  		break;
2996  	}
2997  #endif
2998  
2999  	switch (cmd) {
3000  	case SIOCGIFCONF:
3001  		error = ifconf(cmd, data);
3002  		goto out_noref;
3003  	}
3004  
3005  	ifr = (struct ifreq *)data;
3006  	switch (cmd) {
3007  #ifdef VIMAGE
3008  	case SIOCSIFRVNET:
3009  		error = priv_check(td, PRIV_NET_SETIFVNET);
3010  		if (error == 0)
3011  			error = if_vmove_reclaim(td, ifr->ifr_name,
3012  			    ifr->ifr_jid);
3013  		goto out_noref;
3014  #endif
3015  	case SIOCIFCREATE:
3016  	case SIOCIFCREATE2:
3017  		error = priv_check(td, PRIV_NET_IFCREATE);
3018  		if (error == 0)
3019  			error = if_clone_create(ifr->ifr_name,
3020  			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3021  			    ifr_data_get_ptr(ifr) : NULL);
3022  		goto out_noref;
3023  	case SIOCIFDESTROY:
3024  		error = priv_check(td, PRIV_NET_IFDESTROY);
3025  
3026  		if (error == 0) {
3027  			sx_xlock(&ifnet_detach_sxlock);
3028  			error = if_clone_destroy(ifr->ifr_name);
3029  			sx_xunlock(&ifnet_detach_sxlock);
3030  		}
3031  		goto out_noref;
3032  
3033  	case SIOCIFGCLONERS:
3034  		error = if_clone_list((struct if_clonereq *)data);
3035  		goto out_noref;
3036  
3037  	case SIOCGIFGMEMB:
3038  		error = if_getgroupmembers((struct ifgroupreq *)data);
3039  		goto out_noref;
3040  
3041  #if defined(INET) || defined(INET6)
3042  	case SIOCSVH:
3043  	case SIOCGVH:
3044  		if (carp_ioctl_p == NULL)
3045  			error = EPROTONOSUPPORT;
3046  		else
3047  			error = (*carp_ioctl_p)(ifr, cmd, td);
3048  		goto out_noref;
3049  #endif
3050  	}
3051  
3052  	ifp = ifunit_ref(ifr->ifr_name);
3053  	if (ifp == NULL) {
3054  		error = ENXIO;
3055  		goto out_noref;
3056  	}
3057  
3058  	error = ifhwioctl(cmd, ifp, data, td);
3059  	if (error != ENOIOCTL)
3060  		goto out_ref;
3061  
3062  	oif_flags = ifp->if_flags;
3063  	if (so->so_proto == NULL) {
3064  		error = EOPNOTSUPP;
3065  		goto out_ref;
3066  	}
3067  
3068  	/*
3069  	 * Pass the request on to the socket control method, and if the
3070  	 * latter returns EOPNOTSUPP, directly to the interface.
3071  	 *
3072  	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3073  	 * trust SIOCSIFADDR et al to come from an already privileged
3074  	 * layer, and do not perform any credentials checks or input
3075  	 * validation.
3076  	 */
3077  	error = so->so_proto->pr_control(so, cmd, data, ifp, td);
3078  	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3079  	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3080  	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3081  		error = (*ifp->if_ioctl)(ifp, cmd, data);
3082  
3083  	if (!(oif_flags & IFF_UP) && (ifp->if_flags & IFF_UP))
3084  		if_up(ifp);
3085  out_ref:
3086  	if_rele(ifp);
3087  out_noref:
3088  	CURVNET_RESTORE();
3089  #ifdef COMPAT_FREEBSD32
3090  	if (error != 0)
3091  		return (error);
3092  	switch (saved_cmd) {
3093  	case SIOCGIFCONF32:
3094  		ifc32->ifc_len = thunk.ifc.ifc_len;
3095  		break;
3096  	case SIOCGDRVSPEC32:
3097  		/*
3098  		 * SIOCGDRVSPEC is IOWR, but nothing actually touches
3099  		 * the struct so just assert that ifd_len (the only
3100  		 * field it might make sense to update) hasn't
3101  		 * changed.
3102  		 */
3103  		KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len,
3104  		    ("ifd_len was updated %u -> %zu", ifd32->ifd_len,
3105  			thunk.ifd.ifd_len));
3106  		break;
3107  	case SIOCGIFGROUP32:
3108  	case SIOCGIFGMEMB32:
3109  		ifgr32->ifgr_len = thunk.ifgr.ifgr_len;
3110  		break;
3111  	case SIOCGIFMEDIA32:
3112  	case SIOCGIFXMEDIA32:
3113  		ifmr32->ifm_current = thunk.ifmr.ifm_current;
3114  		ifmr32->ifm_mask = thunk.ifmr.ifm_mask;
3115  		ifmr32->ifm_status = thunk.ifmr.ifm_status;
3116  		ifmr32->ifm_active = thunk.ifmr.ifm_active;
3117  		ifmr32->ifm_count = thunk.ifmr.ifm_count;
3118  		break;
3119  	}
3120  #endif
3121  	return (error);
3122  }
3123  
3124  int
if_rename(struct ifnet * ifp,char * new_name)3125  if_rename(struct ifnet *ifp, char *new_name)
3126  {
3127  	struct ifaddr *ifa;
3128  	struct sockaddr_dl *sdl;
3129  	size_t namelen, onamelen;
3130  	char old_name[IFNAMSIZ];
3131  	char strbuf[IFNAMSIZ + 8];
3132  
3133  	if (new_name[0] == '\0')
3134  		return (EINVAL);
3135  	if (strcmp(new_name, ifp->if_xname) == 0)
3136  		return (0);
3137  	if (ifunit(new_name) != NULL)
3138  		return (EEXIST);
3139  
3140  	/*
3141  	 * XXX: Locking.  Nothing else seems to lock if_flags,
3142  	 * and there are numerous other races with the
3143  	 * ifunit() checks not being atomic with namespace
3144  	 * changes (renames, vmoves, if_attach, etc).
3145  	 */
3146  	ifp->if_flags |= IFF_RENAMING;
3147  
3148  	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
3149  
3150  	if_printf(ifp, "changing name to '%s'\n", new_name);
3151  
3152  	IF_ADDR_WLOCK(ifp);
3153  	strlcpy(old_name, ifp->if_xname, sizeof(old_name));
3154  	strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
3155  	ifa = ifp->if_addr;
3156  	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3157  	namelen = strlen(new_name);
3158  	onamelen = sdl->sdl_nlen;
3159  	/*
3160  	 * Move the address if needed.  This is safe because we
3161  	 * allocate space for a name of length IFNAMSIZ when we
3162  	 * create this in if_attach().
3163  	 */
3164  	if (namelen != onamelen) {
3165  		bcopy(sdl->sdl_data + onamelen,
3166  		    sdl->sdl_data + namelen, sdl->sdl_alen);
3167  	}
3168  	bcopy(new_name, sdl->sdl_data, namelen);
3169  	sdl->sdl_nlen = namelen;
3170  	sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
3171  	bzero(sdl->sdl_data, onamelen);
3172  	while (namelen != 0)
3173  		sdl->sdl_data[--namelen] = 0xff;
3174  	IF_ADDR_WUNLOCK(ifp);
3175  
3176  	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
3177  
3178  	ifp->if_flags &= ~IFF_RENAMING;
3179  
3180  	snprintf(strbuf, sizeof(strbuf), "name=%s", new_name);
3181  	devctl_notify("IFNET", old_name, "RENAME", strbuf);
3182  
3183  	return (0);
3184  }
3185  
3186  /*
3187   * The code common to handling reference counted flags,
3188   * e.g., in ifpromisc() and if_allmulti().
3189   * The "pflag" argument can specify a permanent mode flag to check,
3190   * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3191   *
3192   * Only to be used on stack-owned flags, not driver-owned flags.
3193   */
3194  static int
if_setflag(struct ifnet * ifp,int flag,int pflag,int * refcount,int onswitch)3195  if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3196  {
3197  	struct ifreq ifr;
3198  	int error;
3199  	int oldflags, oldcount;
3200  
3201  	/* Sanity checks to catch programming errors */
3202  	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3203  	    ("%s: setting driver-owned flag %d", __func__, flag));
3204  
3205  	if (onswitch)
3206  		KASSERT(*refcount >= 0,
3207  		    ("%s: increment negative refcount %d for flag %d",
3208  		    __func__, *refcount, flag));
3209  	else
3210  		KASSERT(*refcount > 0,
3211  		    ("%s: decrement non-positive refcount %d for flag %d",
3212  		    __func__, *refcount, flag));
3213  
3214  	/* In case this mode is permanent, just touch refcount */
3215  	if (ifp->if_flags & pflag) {
3216  		*refcount += onswitch ? 1 : -1;
3217  		return (0);
3218  	}
3219  
3220  	/* Save ifnet parameters for if_ioctl() may fail */
3221  	oldcount = *refcount;
3222  	oldflags = ifp->if_flags;
3223  
3224  	/*
3225  	 * See if we aren't the only and touching refcount is enough.
3226  	 * Actually toggle interface flag if we are the first or last.
3227  	 */
3228  	if (onswitch) {
3229  		if ((*refcount)++)
3230  			return (0);
3231  		ifp->if_flags |= flag;
3232  	} else {
3233  		if (--(*refcount))
3234  			return (0);
3235  		ifp->if_flags &= ~flag;
3236  	}
3237  
3238  	/* Call down the driver since we've changed interface flags */
3239  	if (ifp->if_ioctl == NULL) {
3240  		error = EOPNOTSUPP;
3241  		goto recover;
3242  	}
3243  	ifr.ifr_flags = ifp->if_flags & 0xffff;
3244  	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3245  	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3246  	if (error)
3247  		goto recover;
3248  	/* Notify userland that interface flags have changed */
3249  	rt_ifmsg(ifp, flag);
3250  	return (0);
3251  
3252  recover:
3253  	/* Recover after driver error */
3254  	*refcount = oldcount;
3255  	ifp->if_flags = oldflags;
3256  	return (error);
3257  }
3258  
3259  /*
3260   * Set/clear promiscuous mode on interface ifp based on the truth value
3261   * of pswitch.  The calls are reference counted so that only the first
3262   * "on" request actually has an effect, as does the final "off" request.
3263   * Results are undefined if the "off" and "on" requests are not matched.
3264   */
3265  int
ifpromisc(struct ifnet * ifp,int pswitch)3266  ifpromisc(struct ifnet *ifp, int pswitch)
3267  {
3268  	int error;
3269  	int oldflags = ifp->if_flags;
3270  
3271  	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3272  			   &ifp->if_pcount, pswitch);
3273  	/* If promiscuous mode status has changed, log a message */
3274  	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3275              log_promisc_mode_change)
3276  		if_printf(ifp, "promiscuous mode %s\n",
3277  		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3278  	return (error);
3279  }
3280  
3281  /*
3282   * Return interface configuration
3283   * of system.  List may be used
3284   * in later ioctl's (above) to get
3285   * other information.
3286   */
3287  /*ARGSUSED*/
3288  static int
ifconf(u_long cmd,caddr_t data)3289  ifconf(u_long cmd, caddr_t data)
3290  {
3291  	struct ifconf *ifc = (struct ifconf *)data;
3292  	struct ifnet *ifp;
3293  	struct ifaddr *ifa;
3294  	struct ifreq ifr;
3295  	struct sbuf *sb;
3296  	int error, full = 0, valid_len, max_len;
3297  
3298  	/* Limit initial buffer size to maxphys to avoid DoS from userspace. */
3299  	max_len = maxphys - 1;
3300  
3301  	/* Prevent hostile input from being able to crash the system */
3302  	if (ifc->ifc_len <= 0)
3303  		return (EINVAL);
3304  
3305  again:
3306  	if (ifc->ifc_len <= max_len) {
3307  		max_len = ifc->ifc_len;
3308  		full = 1;
3309  	}
3310  	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3311  	max_len = 0;
3312  	valid_len = 0;
3313  
3314  	IFNET_RLOCK();
3315  	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3316  		struct epoch_tracker et;
3317  		int addrs;
3318  
3319  		/*
3320  		 * Zero the ifr to make sure we don't disclose the contents
3321  		 * of the stack.
3322  		 */
3323  		memset(&ifr, 0, sizeof(ifr));
3324  
3325  		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3326  		    >= sizeof(ifr.ifr_name)) {
3327  			sbuf_delete(sb);
3328  			IFNET_RUNLOCK();
3329  			return (ENAMETOOLONG);
3330  		}
3331  
3332  		addrs = 0;
3333  		NET_EPOCH_ENTER(et);
3334  		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3335  			struct sockaddr *sa = ifa->ifa_addr;
3336  
3337  			if (prison_if(curthread->td_ucred, sa) != 0)
3338  				continue;
3339  			addrs++;
3340  			if (sa->sa_len <= sizeof(*sa)) {
3341  				if (sa->sa_len < sizeof(*sa)) {
3342  					memset(&ifr.ifr_ifru.ifru_addr, 0,
3343  					    sizeof(ifr.ifr_ifru.ifru_addr));
3344  					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3345  					    sa->sa_len);
3346  				} else
3347  					ifr.ifr_ifru.ifru_addr = *sa;
3348  				sbuf_bcat(sb, &ifr, sizeof(ifr));
3349  				max_len += sizeof(ifr);
3350  			} else {
3351  				sbuf_bcat(sb, &ifr,
3352  				    offsetof(struct ifreq, ifr_addr));
3353  				max_len += offsetof(struct ifreq, ifr_addr);
3354  				sbuf_bcat(sb, sa, sa->sa_len);
3355  				max_len += sa->sa_len;
3356  			}
3357  
3358  			if (sbuf_error(sb) == 0)
3359  				valid_len = sbuf_len(sb);
3360  		}
3361  		NET_EPOCH_EXIT(et);
3362  		if (addrs == 0) {
3363  			sbuf_bcat(sb, &ifr, sizeof(ifr));
3364  			max_len += sizeof(ifr);
3365  
3366  			if (sbuf_error(sb) == 0)
3367  				valid_len = sbuf_len(sb);
3368  		}
3369  	}
3370  	IFNET_RUNLOCK();
3371  
3372  	/*
3373  	 * If we didn't allocate enough space (uncommon), try again.  If
3374  	 * we have already allocated as much space as we are allowed,
3375  	 * return what we've got.
3376  	 */
3377  	if (valid_len != max_len && !full) {
3378  		sbuf_delete(sb);
3379  		goto again;
3380  	}
3381  
3382  	ifc->ifc_len = valid_len;
3383  	sbuf_finish(sb);
3384  	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3385  	sbuf_delete(sb);
3386  	return (error);
3387  }
3388  
3389  /*
3390   * Just like ifpromisc(), but for all-multicast-reception mode.
3391   */
3392  int
if_allmulti(struct ifnet * ifp,int onswitch)3393  if_allmulti(struct ifnet *ifp, int onswitch)
3394  {
3395  
3396  	return (if_setflag(ifp, IFF_ALLMULTI, IFF_PALLMULTI, &ifp->if_amcount,
3397  	    onswitch));
3398  }
3399  
3400  struct ifmultiaddr *
if_findmulti(struct ifnet * ifp,const struct sockaddr * sa)3401  if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3402  {
3403  	struct ifmultiaddr *ifma;
3404  
3405  	IF_ADDR_LOCK_ASSERT(ifp);
3406  
3407  	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3408  		if (sa->sa_family == AF_LINK) {
3409  			if (sa_dl_equal(ifma->ifma_addr, sa))
3410  				break;
3411  		} else {
3412  			if (sa_equal(ifma->ifma_addr, sa))
3413  				break;
3414  		}
3415  	}
3416  
3417  	return ifma;
3418  }
3419  
3420  /*
3421   * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3422   * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3423   * the ifnet multicast address list here, so the caller must do that and
3424   * other setup work (such as notifying the device driver).  The reference
3425   * count is initialized to 1.
3426   */
3427  static struct ifmultiaddr *
if_allocmulti(struct ifnet * ifp,struct sockaddr * sa,struct sockaddr * llsa,int mflags)3428  if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3429      int mflags)
3430  {
3431  	struct ifmultiaddr *ifma;
3432  	struct sockaddr *dupsa;
3433  
3434  	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3435  	    M_ZERO);
3436  	if (ifma == NULL)
3437  		return (NULL);
3438  
3439  	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3440  	if (dupsa == NULL) {
3441  		free(ifma, M_IFMADDR);
3442  		return (NULL);
3443  	}
3444  	bcopy(sa, dupsa, sa->sa_len);
3445  	ifma->ifma_addr = dupsa;
3446  
3447  	ifma->ifma_ifp = ifp;
3448  	ifma->ifma_refcount = 1;
3449  	ifma->ifma_protospec = NULL;
3450  
3451  	if (llsa == NULL) {
3452  		ifma->ifma_lladdr = NULL;
3453  		return (ifma);
3454  	}
3455  
3456  	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3457  	if (dupsa == NULL) {
3458  		free(ifma->ifma_addr, M_IFMADDR);
3459  		free(ifma, M_IFMADDR);
3460  		return (NULL);
3461  	}
3462  	bcopy(llsa, dupsa, llsa->sa_len);
3463  	ifma->ifma_lladdr = dupsa;
3464  
3465  	return (ifma);
3466  }
3467  
3468  /*
3469   * if_freemulti: free ifmultiaddr structure and possibly attached related
3470   * addresses.  The caller is responsible for implementing reference
3471   * counting, notifying the driver, handling routing messages, and releasing
3472   * any dependent link layer state.
3473   */
3474  #ifdef MCAST_VERBOSE
3475  extern void kdb_backtrace(void);
3476  #endif
3477  static void
if_freemulti_internal(struct ifmultiaddr * ifma)3478  if_freemulti_internal(struct ifmultiaddr *ifma)
3479  {
3480  
3481  	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3482  	    ifma->ifma_refcount));
3483  
3484  	if (ifma->ifma_lladdr != NULL)
3485  		free(ifma->ifma_lladdr, M_IFMADDR);
3486  #ifdef MCAST_VERBOSE
3487  	kdb_backtrace();
3488  	printf("%s freeing ifma: %p\n", __func__, ifma);
3489  #endif
3490  	free(ifma->ifma_addr, M_IFMADDR);
3491  	free(ifma, M_IFMADDR);
3492  }
3493  
3494  static void
if_destroymulti(epoch_context_t ctx)3495  if_destroymulti(epoch_context_t ctx)
3496  {
3497  	struct ifmultiaddr *ifma;
3498  
3499  	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3500  	if_freemulti_internal(ifma);
3501  }
3502  
3503  void
if_freemulti(struct ifmultiaddr * ifma)3504  if_freemulti(struct ifmultiaddr *ifma)
3505  {
3506  	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3507  	    ifma->ifma_refcount));
3508  
3509  	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3510  }
3511  
3512  /*
3513   * Register an additional multicast address with a network interface.
3514   *
3515   * - If the address is already present, bump the reference count on the
3516   *   address and return.
3517   * - If the address is not link-layer, look up a link layer address.
3518   * - Allocate address structures for one or both addresses, and attach to the
3519   *   multicast address list on the interface.  If automatically adding a link
3520   *   layer address, the protocol address will own a reference to the link
3521   *   layer address, to be freed when it is freed.
3522   * - Notify the network device driver of an addition to the multicast address
3523   *   list.
3524   *
3525   * 'sa' points to caller-owned memory with the desired multicast address.
3526   *
3527   * 'retifma' will be used to return a pointer to the resulting multicast
3528   * address reference, if desired.
3529   */
3530  int
if_addmulti(struct ifnet * ifp,struct sockaddr * sa,struct ifmultiaddr ** retifma)3531  if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3532      struct ifmultiaddr **retifma)
3533  {
3534  	struct ifmultiaddr *ifma, *ll_ifma;
3535  	struct sockaddr *llsa;
3536  	struct sockaddr_dl sdl;
3537  	int error;
3538  
3539  #ifdef INET
3540  	IN_MULTI_LIST_UNLOCK_ASSERT();
3541  #endif
3542  #ifdef INET6
3543  	IN6_MULTI_LIST_UNLOCK_ASSERT();
3544  #endif
3545  	/*
3546  	 * If the address is already present, return a new reference to it;
3547  	 * otherwise, allocate storage and set up a new address.
3548  	 */
3549  	IF_ADDR_WLOCK(ifp);
3550  	ifma = if_findmulti(ifp, sa);
3551  	if (ifma != NULL) {
3552  		ifma->ifma_refcount++;
3553  		if (retifma != NULL)
3554  			*retifma = ifma;
3555  		IF_ADDR_WUNLOCK(ifp);
3556  		return (0);
3557  	}
3558  
3559  	/*
3560  	 * The address isn't already present; resolve the protocol address
3561  	 * into a link layer address, and then look that up, bump its
3562  	 * refcount or allocate an ifma for that also.
3563  	 * Most link layer resolving functions returns address data which
3564  	 * fits inside default sockaddr_dl structure. However callback
3565  	 * can allocate another sockaddr structure, in that case we need to
3566  	 * free it later.
3567  	 */
3568  	llsa = NULL;
3569  	ll_ifma = NULL;
3570  	if (ifp->if_resolvemulti != NULL) {
3571  		/* Provide called function with buffer size information */
3572  		sdl.sdl_len = sizeof(sdl);
3573  		llsa = (struct sockaddr *)&sdl;
3574  		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3575  		if (error)
3576  			goto unlock_out;
3577  	}
3578  
3579  	/*
3580  	 * Allocate the new address.  Don't hook it up yet, as we may also
3581  	 * need to allocate a link layer multicast address.
3582  	 */
3583  	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3584  	if (ifma == NULL) {
3585  		error = ENOMEM;
3586  		goto free_llsa_out;
3587  	}
3588  
3589  	/*
3590  	 * If a link layer address is found, we'll need to see if it's
3591  	 * already present in the address list, or allocate is as well.
3592  	 * When this block finishes, the link layer address will be on the
3593  	 * list.
3594  	 */
3595  	if (llsa != NULL) {
3596  		ll_ifma = if_findmulti(ifp, llsa);
3597  		if (ll_ifma == NULL) {
3598  			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3599  			if (ll_ifma == NULL) {
3600  				--ifma->ifma_refcount;
3601  				if_freemulti(ifma);
3602  				error = ENOMEM;
3603  				goto free_llsa_out;
3604  			}
3605  			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3606  			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3607  			    ifma_link);
3608  		} else
3609  			ll_ifma->ifma_refcount++;
3610  		ifma->ifma_llifma = ll_ifma;
3611  	}
3612  
3613  	/*
3614  	 * We now have a new multicast address, ifma, and possibly a new or
3615  	 * referenced link layer address.  Add the primary address to the
3616  	 * ifnet address list.
3617  	 */
3618  	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3619  	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3620  
3621  	if (retifma != NULL)
3622  		*retifma = ifma;
3623  
3624  	/*
3625  	 * Must generate the message while holding the lock so that 'ifma'
3626  	 * pointer is still valid.
3627  	 */
3628  	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3629  	IF_ADDR_WUNLOCK(ifp);
3630  
3631  	/*
3632  	 * We are certain we have added something, so call down to the
3633  	 * interface to let them know about it.
3634  	 */
3635  	if (ifp->if_ioctl != NULL) {
3636  		if (THREAD_CAN_SLEEP())
3637  			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3638  		else
3639  			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3640  	}
3641  
3642  	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3643  		link_free_sdl(llsa);
3644  
3645  	return (0);
3646  
3647  free_llsa_out:
3648  	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3649  		link_free_sdl(llsa);
3650  
3651  unlock_out:
3652  	IF_ADDR_WUNLOCK(ifp);
3653  	return (error);
3654  }
3655  
3656  static void
if_siocaddmulti(void * arg,int pending)3657  if_siocaddmulti(void *arg, int pending)
3658  {
3659  	struct ifnet *ifp;
3660  
3661  	ifp = arg;
3662  #ifdef DIAGNOSTIC
3663  	if (pending > 1)
3664  		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3665  #endif
3666  	CURVNET_SET(ifp->if_vnet);
3667  	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3668  	CURVNET_RESTORE();
3669  }
3670  
3671  /*
3672   * Delete a multicast group membership by network-layer group address.
3673   *
3674   * Returns ENOENT if the entry could not be found. If ifp no longer
3675   * exists, results are undefined. This entry point should only be used
3676   * from subsystems which do appropriate locking to hold ifp for the
3677   * duration of the call.
3678   * Network-layer protocol domains must use if_delmulti_ifma().
3679   */
3680  int
if_delmulti(struct ifnet * ifp,struct sockaddr * sa)3681  if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3682  {
3683  	struct ifmultiaddr *ifma;
3684  	int lastref;
3685  
3686  	KASSERT(ifp, ("%s: NULL ifp", __func__));
3687  
3688  	IF_ADDR_WLOCK(ifp);
3689  	lastref = 0;
3690  	ifma = if_findmulti(ifp, sa);
3691  	if (ifma != NULL)
3692  		lastref = if_delmulti_locked(ifp, ifma, 0);
3693  	IF_ADDR_WUNLOCK(ifp);
3694  
3695  	if (ifma == NULL)
3696  		return (ENOENT);
3697  
3698  	if (lastref && ifp->if_ioctl != NULL) {
3699  		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3700  	}
3701  
3702  	return (0);
3703  }
3704  
3705  /*
3706   * Delete all multicast group membership for an interface.
3707   * Should be used to quickly flush all multicast filters.
3708   */
3709  void
if_delallmulti(struct ifnet * ifp)3710  if_delallmulti(struct ifnet *ifp)
3711  {
3712  	struct ifmultiaddr *ifma;
3713  	struct ifmultiaddr *next;
3714  
3715  	IF_ADDR_WLOCK(ifp);
3716  	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3717  		if_delmulti_locked(ifp, ifma, 0);
3718  	IF_ADDR_WUNLOCK(ifp);
3719  }
3720  
3721  void
if_delmulti_ifma(struct ifmultiaddr * ifma)3722  if_delmulti_ifma(struct ifmultiaddr *ifma)
3723  {
3724  	if_delmulti_ifma_flags(ifma, 0);
3725  }
3726  
3727  /*
3728   * Delete a multicast group membership by group membership pointer.
3729   * Network-layer protocol domains must use this routine.
3730   *
3731   * It is safe to call this routine if the ifp disappeared.
3732   */
3733  void
if_delmulti_ifma_flags(struct ifmultiaddr * ifma,int flags)3734  if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3735  {
3736  	struct ifnet *ifp;
3737  	int lastref;
3738  	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3739  #ifdef INET
3740  	IN_MULTI_LIST_UNLOCK_ASSERT();
3741  #endif
3742  	ifp = ifma->ifma_ifp;
3743  #ifdef DIAGNOSTIC
3744  	if (ifp == NULL) {
3745  		printf("%s: ifma_ifp seems to be detached\n", __func__);
3746  	} else {
3747  		struct epoch_tracker et;
3748  		struct ifnet *oifp;
3749  
3750  		NET_EPOCH_ENTER(et);
3751  		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3752  			if (ifp == oifp)
3753  				break;
3754  		NET_EPOCH_EXIT(et);
3755  		if (ifp != oifp)
3756  			ifp = NULL;
3757  	}
3758  #endif
3759  	/*
3760  	 * If and only if the ifnet instance exists: Acquire the address lock.
3761  	 */
3762  	if (ifp != NULL)
3763  		IF_ADDR_WLOCK(ifp);
3764  
3765  	lastref = if_delmulti_locked(ifp, ifma, flags);
3766  
3767  	if (ifp != NULL) {
3768  		/*
3769  		 * If and only if the ifnet instance exists:
3770  		 *  Release the address lock.
3771  		 *  If the group was left: update the hardware hash filter.
3772  		 */
3773  		IF_ADDR_WUNLOCK(ifp);
3774  		if (lastref && ifp->if_ioctl != NULL) {
3775  			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3776  		}
3777  	}
3778  }
3779  
3780  /*
3781   * Perform deletion of network-layer and/or link-layer multicast address.
3782   *
3783   * Return 0 if the reference count was decremented.
3784   * Return 1 if the final reference was released, indicating that the
3785   * hardware hash filter should be reprogrammed.
3786   */
3787  static int
if_delmulti_locked(struct ifnet * ifp,struct ifmultiaddr * ifma,int detaching)3788  if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3789  {
3790  	struct ifmultiaddr *ll_ifma;
3791  
3792  	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3793  		KASSERT(ifma->ifma_ifp == ifp,
3794  		    ("%s: inconsistent ifp %p", __func__, ifp));
3795  		IF_ADDR_WLOCK_ASSERT(ifp);
3796  	}
3797  
3798  	ifp = ifma->ifma_ifp;
3799  	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3800  
3801  	/*
3802  	 * If the ifnet is detaching, null out references to ifnet,
3803  	 * so that upper protocol layers will notice, and not attempt
3804  	 * to obtain locks for an ifnet which no longer exists. The
3805  	 * routing socket announcement must happen before the ifnet
3806  	 * instance is detached from the system.
3807  	 */
3808  	if (detaching) {
3809  #ifdef DIAGNOSTIC
3810  		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3811  #endif
3812  		/*
3813  		 * ifp may already be nulled out if we are being reentered
3814  		 * to delete the ll_ifma.
3815  		 */
3816  		if (ifp != NULL) {
3817  			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3818  			ifma->ifma_ifp = NULL;
3819  		}
3820  	}
3821  
3822  	if (--ifma->ifma_refcount > 0)
3823  		return 0;
3824  
3825  	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3826  		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3827  		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3828  	}
3829  	/*
3830  	 * If this ifma is a network-layer ifma, a link-layer ifma may
3831  	 * have been associated with it. Release it first if so.
3832  	 */
3833  	ll_ifma = ifma->ifma_llifma;
3834  	if (ll_ifma != NULL) {
3835  		KASSERT(ifma->ifma_lladdr != NULL,
3836  		    ("%s: llifma w/o lladdr", __func__));
3837  		if (detaching)
3838  			ll_ifma->ifma_ifp = NULL;	/* XXX */
3839  		if (--ll_ifma->ifma_refcount == 0) {
3840  			if (ifp != NULL) {
3841  				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3842  					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3843  						ifma_link);
3844  					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3845  				}
3846  			}
3847  			if_freemulti(ll_ifma);
3848  		}
3849  	}
3850  #ifdef INVARIANTS
3851  	if (ifp) {
3852  		struct ifmultiaddr *ifmatmp;
3853  
3854  		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3855  			MPASS(ifma != ifmatmp);
3856  	}
3857  #endif
3858  	if_freemulti(ifma);
3859  	/*
3860  	 * The last reference to this instance of struct ifmultiaddr
3861  	 * was released; the hardware should be notified of this change.
3862  	 */
3863  	return 1;
3864  }
3865  
3866  /*
3867   * Set the link layer address on an interface.
3868   *
3869   * At this time we only support certain types of interfaces,
3870   * and we don't allow the length of the address to change.
3871   *
3872   * Set noinline to be dtrace-friendly
3873   */
3874  __noinline int
if_setlladdr(struct ifnet * ifp,const u_char * lladdr,int len)3875  if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3876  {
3877  	struct sockaddr_dl *sdl;
3878  	struct ifaddr *ifa;
3879  	struct ifreq ifr;
3880  
3881  	ifa = ifp->if_addr;
3882  	if (ifa == NULL)
3883  		return (EINVAL);
3884  
3885  	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3886  	if (sdl == NULL)
3887  		return (EINVAL);
3888  
3889  	if (len != sdl->sdl_alen)	/* don't allow length to change */
3890  		return (EINVAL);
3891  
3892  	switch (ifp->if_type) {
3893  	case IFT_ETHER:
3894  	case IFT_XETHER:
3895  	case IFT_L2VLAN:
3896  	case IFT_BRIDGE:
3897  	case IFT_IEEE8023ADLAG:
3898  		bcopy(lladdr, LLADDR(sdl), len);
3899  		break;
3900  	default:
3901  		return (ENODEV);
3902  	}
3903  
3904  	/*
3905  	 * If the interface is already up, we need
3906  	 * to re-init it in order to reprogram its
3907  	 * address filter.
3908  	 */
3909  	if ((ifp->if_flags & IFF_UP) != 0) {
3910  		if (ifp->if_ioctl) {
3911  			ifp->if_flags &= ~IFF_UP;
3912  			ifr.ifr_flags = ifp->if_flags & 0xffff;
3913  			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3914  			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3915  			ifp->if_flags |= IFF_UP;
3916  			ifr.ifr_flags = ifp->if_flags & 0xffff;
3917  			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3918  			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3919  		}
3920  	}
3921  	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3922  
3923  	return (0);
3924  }
3925  
3926  /*
3927   * Compat function for handling basic encapsulation requests.
3928   * Not converted stacks (FDDI, IB, ..) supports traditional
3929   * output model: ARP (and other similar L2 protocols) are handled
3930   * inside output routine, arpresolve/nd6_resolve() returns MAC
3931   * address instead of full prepend.
3932   *
3933   * This function creates calculated header==MAC for IPv4/IPv6 and
3934   * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3935   * address families.
3936   */
3937  static int
if_requestencap_default(struct ifnet * ifp,struct if_encap_req * req)3938  if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3939  {
3940  	if (req->rtype != IFENCAP_LL)
3941  		return (EOPNOTSUPP);
3942  
3943  	if (req->bufsize < req->lladdr_len)
3944  		return (ENOMEM);
3945  
3946  	switch (req->family) {
3947  	case AF_INET:
3948  	case AF_INET6:
3949  		break;
3950  	default:
3951  		return (EAFNOSUPPORT);
3952  	}
3953  
3954  	/* Copy lladdr to storage as is */
3955  	memmove(req->buf, req->lladdr, req->lladdr_len);
3956  	req->bufsize = req->lladdr_len;
3957  	req->lladdr_off = 0;
3958  
3959  	return (0);
3960  }
3961  
3962  /*
3963   * Tunnel interfaces can nest, also they may cause infinite recursion
3964   * calls when misconfigured. We'll prevent this by detecting loops.
3965   * High nesting level may cause stack exhaustion. We'll prevent this
3966   * by introducing upper limit.
3967   *
3968   * Return 0, if tunnel nesting count is equal or less than limit.
3969   */
3970  int
if_tunnel_check_nesting(struct ifnet * ifp,struct mbuf * m,uint32_t cookie,int limit)3971  if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3972      int limit)
3973  {
3974  	struct m_tag *mtag;
3975  	int count;
3976  
3977  	count = 1;
3978  	mtag = NULL;
3979  	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3980  		if (*(struct ifnet **)(mtag + 1) == ifp) {
3981  			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3982  			return (EIO);
3983  		}
3984  		count++;
3985  	}
3986  	if (count > limit) {
3987  		log(LOG_NOTICE,
3988  		    "%s: if_output recursively called too many times(%d)\n",
3989  		    if_name(ifp), count);
3990  		return (EIO);
3991  	}
3992  	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3993  	if (mtag == NULL)
3994  		return (ENOMEM);
3995  	*(struct ifnet **)(mtag + 1) = ifp;
3996  	m_tag_prepend(m, mtag);
3997  	return (0);
3998  }
3999  
4000  /*
4001   * Get the link layer address that was read from the hardware at attach.
4002   *
4003   * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
4004   * their component interfaces as IFT_IEEE8023ADLAG.
4005   */
4006  int
if_gethwaddr(struct ifnet * ifp,struct ifreq * ifr)4007  if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
4008  {
4009  	if (ifp->if_hw_addr == NULL)
4010  		return (ENODEV);
4011  
4012  	switch (ifp->if_type) {
4013  	case IFT_ETHER:
4014  	case IFT_IEEE8023ADLAG:
4015  		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
4016  		return (0);
4017  	default:
4018  		return (ENODEV);
4019  	}
4020  }
4021  
4022  /*
4023   * The name argument must be a pointer to storage which will last as
4024   * long as the interface does.  For physical devices, the result of
4025   * device_get_name(dev) is a good choice and for pseudo-devices a
4026   * static string works well.
4027   */
4028  void
if_initname(struct ifnet * ifp,const char * name,int unit)4029  if_initname(struct ifnet *ifp, const char *name, int unit)
4030  {
4031  	ifp->if_dname = name;
4032  	ifp->if_dunit = unit;
4033  	if (unit != IF_DUNIT_NONE)
4034  		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
4035  	else
4036  		strlcpy(ifp->if_xname, name, IFNAMSIZ);
4037  }
4038  
4039  static int
if_vlog(struct ifnet * ifp,int pri,const char * fmt,va_list ap)4040  if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap)
4041  {
4042  	char if_fmt[256];
4043  
4044  	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
4045  	vlog(pri, if_fmt, ap);
4046  	return (0);
4047  }
4048  
4049  
4050  int
if_printf(struct ifnet * ifp,const char * fmt,...)4051  if_printf(struct ifnet *ifp, const char *fmt, ...)
4052  {
4053  	va_list ap;
4054  
4055  	va_start(ap, fmt);
4056  	if_vlog(ifp, LOG_INFO, fmt, ap);
4057  	va_end(ap);
4058  	return (0);
4059  }
4060  
4061  int
if_log(struct ifnet * ifp,int pri,const char * fmt,...)4062  if_log(struct ifnet *ifp, int pri, const char *fmt, ...)
4063  {
4064  	va_list ap;
4065  
4066  	va_start(ap, fmt);
4067  	if_vlog(ifp, pri, fmt, ap);
4068  	va_end(ap);
4069  	return (0);
4070  }
4071  
4072  void
if_start(struct ifnet * ifp)4073  if_start(struct ifnet *ifp)
4074  {
4075  
4076  	(*(ifp)->if_start)(ifp);
4077  }
4078  
4079  /*
4080   * Backwards compatibility interface for drivers
4081   * that have not implemented it
4082   */
4083  static int
if_transmit_default(struct ifnet * ifp,struct mbuf * m)4084  if_transmit_default(struct ifnet *ifp, struct mbuf *m)
4085  {
4086  	int error;
4087  
4088  	IFQ_HANDOFF(ifp, m, error);
4089  	return (error);
4090  }
4091  
4092  static void
if_input_default(struct ifnet * ifp __unused,struct mbuf * m)4093  if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
4094  {
4095  	m_freem(m);
4096  }
4097  
4098  int
if_handoff(struct ifqueue * ifq,struct mbuf * m,struct ifnet * ifp,int adjust)4099  if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4100  {
4101  	int active = 0;
4102  
4103  	IF_LOCK(ifq);
4104  	if (_IF_QFULL(ifq)) {
4105  		IF_UNLOCK(ifq);
4106  		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4107  		m_freem(m);
4108  		return (0);
4109  	}
4110  	if (ifp != NULL) {
4111  		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4112  		if (m->m_flags & (M_BCAST|M_MCAST))
4113  			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4114  		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4115  	}
4116  	_IF_ENQUEUE(ifq, m);
4117  	IF_UNLOCK(ifq);
4118  	if (ifp != NULL && !active)
4119  		(*(ifp)->if_start)(ifp);
4120  	return (1);
4121  }
4122  
4123  void
if_register_com_alloc(u_char type,if_com_alloc_t * a,if_com_free_t * f)4124  if_register_com_alloc(u_char type,
4125      if_com_alloc_t *a, if_com_free_t *f)
4126  {
4127  
4128  	KASSERT(if_com_alloc[type] == NULL,
4129  	    ("if_register_com_alloc: %d already registered", type));
4130  	KASSERT(if_com_free[type] == NULL,
4131  	    ("if_register_com_alloc: %d free already registered", type));
4132  
4133  	if_com_alloc[type] = a;
4134  	if_com_free[type] = f;
4135  }
4136  
4137  void
if_deregister_com_alloc(u_char type)4138  if_deregister_com_alloc(u_char type)
4139  {
4140  
4141  	KASSERT(if_com_alloc[type] != NULL,
4142  	    ("if_deregister_com_alloc: %d not registered", type));
4143  	KASSERT(if_com_free[type] != NULL,
4144  	    ("if_deregister_com_alloc: %d free not registered", type));
4145  
4146  	/*
4147  	 * Ensure all pending EPOCH(9) callbacks have been executed. This
4148  	 * fixes issues about late invocation of if_destroy(), which leads
4149  	 * to memory leak from if_com_alloc[type] allocated if_l2com.
4150  	 */
4151  	NET_EPOCH_DRAIN_CALLBACKS();
4152  
4153  	if_com_alloc[type] = NULL;
4154  	if_com_free[type] = NULL;
4155  }
4156  
4157  /* API for driver access to network stack owned ifnet.*/
4158  uint64_t
if_setbaudrate(struct ifnet * ifp,uint64_t baudrate)4159  if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4160  {
4161  	uint64_t oldbrate;
4162  
4163  	oldbrate = ifp->if_baudrate;
4164  	ifp->if_baudrate = baudrate;
4165  	return (oldbrate);
4166  }
4167  
4168  uint64_t
if_getbaudrate(const if_t ifp)4169  if_getbaudrate(const if_t ifp)
4170  {
4171  	return (ifp->if_baudrate);
4172  }
4173  
4174  int
if_setcapabilities(if_t ifp,int capabilities)4175  if_setcapabilities(if_t ifp, int capabilities)
4176  {
4177  	ifp->if_capabilities = capabilities;
4178  	return (0);
4179  }
4180  
4181  int
if_setcapabilitiesbit(if_t ifp,int setbit,int clearbit)4182  if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4183  {
4184  	ifp->if_capabilities &= ~clearbit;
4185  	ifp->if_capabilities |= setbit;
4186  	return (0);
4187  }
4188  
4189  int
if_getcapabilities(const if_t ifp)4190  if_getcapabilities(const if_t ifp)
4191  {
4192  	return (ifp->if_capabilities);
4193  }
4194  
4195  int
if_setcapenable(if_t ifp,int capabilities)4196  if_setcapenable(if_t ifp, int capabilities)
4197  {
4198  	ifp->if_capenable = capabilities;
4199  	return (0);
4200  }
4201  
4202  int
if_setcapenablebit(if_t ifp,int setcap,int clearcap)4203  if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4204  {
4205  	ifp->if_capenable &= ~clearcap;
4206  	ifp->if_capenable |= setcap;
4207  	return (0);
4208  }
4209  
4210  int
if_setcapabilities2(if_t ifp,int capabilities)4211  if_setcapabilities2(if_t ifp, int capabilities)
4212  {
4213  	ifp->if_capabilities2 = capabilities;
4214  	return (0);
4215  }
4216  
4217  int
if_setcapabilities2bit(if_t ifp,int setbit,int clearbit)4218  if_setcapabilities2bit(if_t ifp, int setbit, int clearbit)
4219  {
4220  	ifp->if_capabilities2 &= ~clearbit;
4221  	ifp->if_capabilities2 |= setbit;
4222  	return (0);
4223  }
4224  
4225  int
if_getcapabilities2(const if_t ifp)4226  if_getcapabilities2(const if_t ifp)
4227  {
4228  	return (ifp->if_capabilities2);
4229  }
4230  
4231  int
if_setcapenable2(if_t ifp,int capabilities2)4232  if_setcapenable2(if_t ifp, int capabilities2)
4233  {
4234  	ifp->if_capenable2 = capabilities2;
4235  	return (0);
4236  }
4237  
4238  int
if_setcapenable2bit(if_t ifp,int setcap,int clearcap)4239  if_setcapenable2bit(if_t ifp, int setcap, int clearcap)
4240  {
4241  	ifp->if_capenable2 &= ~clearcap;
4242  	ifp->if_capenable2 |= setcap;
4243  	return (0);
4244  }
4245  
4246  const char *
if_getdname(const if_t ifp)4247  if_getdname(const if_t ifp)
4248  {
4249  	return (ifp->if_dname);
4250  }
4251  
4252  void
if_setdname(if_t ifp,const char * dname)4253  if_setdname(if_t ifp, const char *dname)
4254  {
4255  	ifp->if_dname = dname;
4256  }
4257  
4258  const char *
if_name(if_t ifp)4259  if_name(if_t ifp)
4260  {
4261  	return (ifp->if_xname);
4262  }
4263  
4264  int
if_setname(if_t ifp,const char * name)4265  if_setname(if_t ifp, const char *name)
4266  {
4267  	if (strlen(name) > sizeof(ifp->if_xname) - 1)
4268  		return (ENAMETOOLONG);
4269  	strcpy(ifp->if_xname, name);
4270  
4271  	return (0);
4272  }
4273  
4274  int
if_togglecapenable(if_t ifp,int togglecap)4275  if_togglecapenable(if_t ifp, int togglecap)
4276  {
4277  	ifp->if_capenable ^= togglecap;
4278  	return (0);
4279  }
4280  
4281  int
if_getcapenable(const if_t ifp)4282  if_getcapenable(const if_t ifp)
4283  {
4284  	return (ifp->if_capenable);
4285  }
4286  
4287  int
if_togglecapenable2(if_t ifp,int togglecap)4288  if_togglecapenable2(if_t ifp, int togglecap)
4289  {
4290  	ifp->if_capenable2 ^= togglecap;
4291  	return (0);
4292  }
4293  
4294  int
if_getcapenable2(const if_t ifp)4295  if_getcapenable2(const if_t ifp)
4296  {
4297  	return (ifp->if_capenable2);
4298  }
4299  
4300  int
if_getdunit(const if_t ifp)4301  if_getdunit(const if_t ifp)
4302  {
4303  	return (ifp->if_dunit);
4304  }
4305  
4306  int
if_getindex(const if_t ifp)4307  if_getindex(const if_t ifp)
4308  {
4309  	return (ifp->if_index);
4310  }
4311  
4312  int
if_getidxgen(const if_t ifp)4313  if_getidxgen(const if_t ifp)
4314  {
4315  	return (ifp->if_idxgen);
4316  }
4317  
4318  const char *
if_getdescr(if_t ifp)4319  if_getdescr(if_t ifp)
4320  {
4321  	return (ifp->if_description);
4322  }
4323  
4324  void
if_setdescr(if_t ifp,char * descrbuf)4325  if_setdescr(if_t ifp, char *descrbuf)
4326  {
4327  	sx_xlock(&ifdescr_sx);
4328  	char *odescrbuf = ifp->if_description;
4329  	ifp->if_description = descrbuf;
4330  	sx_xunlock(&ifdescr_sx);
4331  
4332  	if_freedescr(odescrbuf);
4333  }
4334  
4335  char *
if_allocdescr(size_t sz,int malloc_flag)4336  if_allocdescr(size_t sz, int malloc_flag)
4337  {
4338  	malloc_flag &= (M_WAITOK | M_NOWAIT);
4339  	return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag));
4340  }
4341  
4342  void
if_freedescr(char * descrbuf)4343  if_freedescr(char *descrbuf)
4344  {
4345  	free(descrbuf, M_IFDESCR);
4346  }
4347  
4348  int
if_getalloctype(const if_t ifp)4349  if_getalloctype(const if_t ifp)
4350  {
4351  	return (ifp->if_alloctype);
4352  }
4353  
4354  void
if_setlastchange(if_t ifp)4355  if_setlastchange(if_t ifp)
4356  {
4357  	getmicrotime(&ifp->if_lastchange);
4358  }
4359  
4360  /*
4361   * This is largely undesirable because it ties ifnet to a device, but does
4362   * provide flexiblity for an embedded product vendor. Should be used with
4363   * the understanding that it violates the interface boundaries, and should be
4364   * a last resort only.
4365   */
4366  int
if_setdev(if_t ifp,void * dev)4367  if_setdev(if_t ifp, void *dev)
4368  {
4369  	return (0);
4370  }
4371  
4372  int
if_setdrvflagbits(if_t ifp,int set_flags,int clear_flags)4373  if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4374  {
4375  	ifp->if_drv_flags &= ~clear_flags;
4376  	ifp->if_drv_flags |= set_flags;
4377  
4378  	return (0);
4379  }
4380  
4381  int
if_getdrvflags(const if_t ifp)4382  if_getdrvflags(const if_t ifp)
4383  {
4384  	return (ifp->if_drv_flags);
4385  }
4386  
4387  int
if_setdrvflags(if_t ifp,int flags)4388  if_setdrvflags(if_t ifp, int flags)
4389  {
4390  	ifp->if_drv_flags = flags;
4391  	return (0);
4392  }
4393  
4394  int
if_setflags(if_t ifp,int flags)4395  if_setflags(if_t ifp, int flags)
4396  {
4397  	ifp->if_flags = flags;
4398  	return (0);
4399  }
4400  
4401  int
if_setflagbits(if_t ifp,int set,int clear)4402  if_setflagbits(if_t ifp, int set, int clear)
4403  {
4404  	ifp->if_flags &= ~clear;
4405  	ifp->if_flags |= set;
4406  	return (0);
4407  }
4408  
4409  int
if_getflags(const if_t ifp)4410  if_getflags(const if_t ifp)
4411  {
4412  	return (ifp->if_flags);
4413  }
4414  
4415  int
if_clearhwassist(if_t ifp)4416  if_clearhwassist(if_t ifp)
4417  {
4418  	ifp->if_hwassist = 0;
4419  	return (0);
4420  }
4421  
4422  int
if_sethwassistbits(if_t ifp,int toset,int toclear)4423  if_sethwassistbits(if_t ifp, int toset, int toclear)
4424  {
4425  	ifp->if_hwassist &= ~toclear;
4426  	ifp->if_hwassist |= toset;
4427  
4428  	return (0);
4429  }
4430  
4431  int
if_sethwassist(if_t ifp,int hwassist_bit)4432  if_sethwassist(if_t ifp, int hwassist_bit)
4433  {
4434  	ifp->if_hwassist = hwassist_bit;
4435  	return (0);
4436  }
4437  
4438  int
if_gethwassist(const if_t ifp)4439  if_gethwassist(const if_t ifp)
4440  {
4441  	return (ifp->if_hwassist);
4442  }
4443  
4444  int
if_togglehwassist(if_t ifp,int toggle_bits)4445  if_togglehwassist(if_t ifp, int toggle_bits)
4446  {
4447  	ifp->if_hwassist ^= toggle_bits;
4448  	return (0);
4449  }
4450  
4451  int
if_setmtu(if_t ifp,int mtu)4452  if_setmtu(if_t ifp, int mtu)
4453  {
4454  	ifp->if_mtu = mtu;
4455  	return (0);
4456  }
4457  
4458  void
if_notifymtu(if_t ifp)4459  if_notifymtu(if_t ifp)
4460  {
4461  #ifdef INET6
4462  	nd6_setmtu(ifp);
4463  #endif
4464  	rt_updatemtu(ifp);
4465  }
4466  
4467  int
if_getmtu(const if_t ifp)4468  if_getmtu(const if_t ifp)
4469  {
4470  	return (ifp->if_mtu);
4471  }
4472  
4473  int
if_getmtu_family(const if_t ifp,int family)4474  if_getmtu_family(const if_t ifp, int family)
4475  {
4476  	struct domain *dp;
4477  
4478  	SLIST_FOREACH(dp, &domains, dom_next) {
4479  		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4480  			return (dp->dom_ifmtu(ifp));
4481  	}
4482  
4483  	return (ifp->if_mtu);
4484  }
4485  
4486  /*
4487   * Methods for drivers to access interface unicast and multicast
4488   * link level addresses.  Driver shall not know 'struct ifaddr' neither
4489   * 'struct ifmultiaddr'.
4490   */
4491  u_int
if_lladdr_count(if_t ifp)4492  if_lladdr_count(if_t ifp)
4493  {
4494  	struct epoch_tracker et;
4495  	struct ifaddr *ifa;
4496  	u_int count;
4497  
4498  	count = 0;
4499  	NET_EPOCH_ENTER(et);
4500  	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4501  		if (ifa->ifa_addr->sa_family == AF_LINK)
4502  			count++;
4503  	NET_EPOCH_EXIT(et);
4504  
4505  	return (count);
4506  }
4507  
4508  int
if_foreach(if_foreach_cb_t cb,void * cb_arg)4509  if_foreach(if_foreach_cb_t cb, void *cb_arg)
4510  {
4511  	if_t ifp;
4512  	int error;
4513  
4514  	NET_EPOCH_ASSERT();
4515  	MPASS(cb);
4516  
4517  	error = 0;
4518  	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4519  		error = cb(ifp, cb_arg);
4520  		if (error != 0)
4521  			break;
4522  	}
4523  
4524  	return (error);
4525  }
4526  
4527  /*
4528   * Iterates over the list of interfaces, permitting callback function @cb to sleep.
4529   * Stops iteration if @cb returns non-zero error code.
4530   * Returns the last error code from @cb.
4531   * @match_cb: optional match callback limiting the iteration to only matched interfaces
4532   * @match_arg: argument to pass to @match_cb
4533   * @cb: iteration callback
4534   * @cb_arg: argument to pass to @cb
4535   */
4536  int
if_foreach_sleep(if_foreach_match_t match_cb,void * match_arg,if_foreach_cb_t cb,void * cb_arg)4537  if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb,
4538      void *cb_arg)
4539  {
4540  	int match_count = 0, array_size = 16; /* 128 bytes for malloc */
4541  	struct ifnet **match_array = NULL;
4542  	int error = 0;
4543  
4544  	MPASS(cb);
4545  
4546  	while (true) {
4547  		struct ifnet **new_array;
4548  		int new_size = array_size;
4549  		struct epoch_tracker et;
4550  		struct ifnet *ifp;
4551  
4552  		while (new_size < match_count)
4553  			new_size *= 2;
4554  		new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK);
4555  		if (match_array != NULL)
4556  			memcpy(new_array, match_array, array_size * sizeof(void *));
4557  		free(match_array, M_TEMP);
4558  		match_array = new_array;
4559  		array_size = new_size;
4560  
4561  		match_count = 0;
4562  		NET_EPOCH_ENTER(et);
4563  		CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4564  			if (match_cb != NULL && !match_cb(ifp, match_arg))
4565  				continue;
4566  			if (match_count < array_size) {
4567  				if (if_try_ref(ifp))
4568  					match_array[match_count++] = ifp;
4569  			} else
4570  				match_count++;
4571  		}
4572  		NET_EPOCH_EXIT(et);
4573  
4574  		if (match_count > array_size) {
4575  			for (int i = 0; i < array_size; i++)
4576  				if_rele(match_array[i]);
4577  			continue;
4578  		} else {
4579  			for (int i = 0; i < match_count; i++) {
4580  				if (error == 0)
4581  					error = cb(match_array[i], cb_arg);
4582  				if_rele(match_array[i]);
4583  			}
4584  			free(match_array, M_TEMP);
4585  			break;
4586  		}
4587  	}
4588  
4589  	return (error);
4590  }
4591  
4592  
4593  /*
4594   * Uses just 1 pointer of the 4 available in the public struct.
4595   */
4596  if_t
if_iter_start(struct if_iter * iter)4597  if_iter_start(struct if_iter *iter)
4598  {
4599  	if_t ifp;
4600  
4601  	NET_EPOCH_ASSERT();
4602  
4603  	bzero(iter, sizeof(*iter));
4604  	ifp = CK_STAILQ_FIRST(&V_ifnet);
4605  	if (ifp != NULL)
4606  		iter->context[0] = CK_STAILQ_NEXT(ifp, if_link);
4607  	else
4608  		iter->context[0] = NULL;
4609  	return (ifp);
4610  }
4611  
4612  if_t
if_iter_next(struct if_iter * iter)4613  if_iter_next(struct if_iter *iter)
4614  {
4615  	if_t cur_ifp = iter->context[0];
4616  
4617  	if (cur_ifp != NULL)
4618  		iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link);
4619  	return (cur_ifp);
4620  }
4621  
4622  void
if_iter_finish(struct if_iter * iter)4623  if_iter_finish(struct if_iter *iter)
4624  {
4625  	/* Nothing to do here for now. */
4626  }
4627  
4628  u_int
if_foreach_lladdr(if_t ifp,iflladdr_cb_t cb,void * cb_arg)4629  if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4630  {
4631  	struct epoch_tracker et;
4632  	struct ifaddr *ifa;
4633  	u_int count;
4634  
4635  	MPASS(cb);
4636  
4637  	count = 0;
4638  	NET_EPOCH_ENTER(et);
4639  	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4640  		if (ifa->ifa_addr->sa_family != AF_LINK)
4641  			continue;
4642  		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4643  		    count);
4644  	}
4645  	NET_EPOCH_EXIT(et);
4646  
4647  	return (count);
4648  }
4649  
4650  u_int
if_llmaddr_count(if_t ifp)4651  if_llmaddr_count(if_t ifp)
4652  {
4653  	struct epoch_tracker et;
4654  	struct ifmultiaddr *ifma;
4655  	int count;
4656  
4657  	count = 0;
4658  	NET_EPOCH_ENTER(et);
4659  	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4660  		if (ifma->ifma_addr->sa_family == AF_LINK)
4661  			count++;
4662  	NET_EPOCH_EXIT(et);
4663  
4664  	return (count);
4665  }
4666  
4667  bool
if_maddr_empty(if_t ifp)4668  if_maddr_empty(if_t ifp)
4669  {
4670  
4671  	return (CK_STAILQ_EMPTY(&ifp->if_multiaddrs));
4672  }
4673  
4674  u_int
if_foreach_llmaddr(if_t ifp,iflladdr_cb_t cb,void * cb_arg)4675  if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4676  {
4677  	struct epoch_tracker et;
4678  	struct ifmultiaddr *ifma;
4679  	u_int count;
4680  
4681  	MPASS(cb);
4682  
4683  	count = 0;
4684  	NET_EPOCH_ENTER(et);
4685  	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4686  		if (ifma->ifma_addr->sa_family != AF_LINK)
4687  			continue;
4688  		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4689  		    count);
4690  	}
4691  	NET_EPOCH_EXIT(et);
4692  
4693  	return (count);
4694  }
4695  
4696  u_int
if_foreach_addr_type(if_t ifp,int type,if_addr_cb_t cb,void * cb_arg)4697  if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg)
4698  {
4699  	struct epoch_tracker et;
4700  	struct ifaddr *ifa;
4701  	u_int count;
4702  
4703  	MPASS(cb);
4704  
4705  	count = 0;
4706  	NET_EPOCH_ENTER(et);
4707  	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4708  		if (ifa->ifa_addr->sa_family != type)
4709  			continue;
4710  		count += (*cb)(cb_arg, ifa, count);
4711  	}
4712  	NET_EPOCH_EXIT(et);
4713  
4714  	return (count);
4715  }
4716  
4717  struct ifaddr *
ifa_iter_start(if_t ifp,struct ifa_iter * iter)4718  ifa_iter_start(if_t ifp, struct ifa_iter *iter)
4719  {
4720  	struct ifaddr *ifa;
4721  
4722  	NET_EPOCH_ASSERT();
4723  
4724  	bzero(iter, sizeof(*iter));
4725  	ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
4726  	if (ifa != NULL)
4727  		iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link);
4728  	else
4729  		iter->context[0] = NULL;
4730  	return (ifa);
4731  }
4732  
4733  struct ifaddr *
ifa_iter_next(struct ifa_iter * iter)4734  ifa_iter_next(struct ifa_iter *iter)
4735  {
4736  	struct ifaddr *ifa = iter->context[0];
4737  
4738  	if (ifa != NULL)
4739  		iter->context[0] = CK_STAILQ_NEXT(ifa, ifa_link);
4740  	return (ifa);
4741  }
4742  
4743  void
ifa_iter_finish(struct ifa_iter * iter)4744  ifa_iter_finish(struct ifa_iter *iter)
4745  {
4746  	/* Nothing to do here for now. */
4747  }
4748  
4749  int
if_setsoftc(if_t ifp,void * softc)4750  if_setsoftc(if_t ifp, void *softc)
4751  {
4752  	ifp->if_softc = softc;
4753  	return (0);
4754  }
4755  
4756  void *
if_getsoftc(const if_t ifp)4757  if_getsoftc(const if_t ifp)
4758  {
4759  	return (ifp->if_softc);
4760  }
4761  
4762  void
if_setrcvif(struct mbuf * m,if_t ifp)4763  if_setrcvif(struct mbuf *m, if_t ifp)
4764  {
4765  
4766  	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4767  	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4768  }
4769  
4770  void
if_setvtag(struct mbuf * m,uint16_t tag)4771  if_setvtag(struct mbuf *m, uint16_t tag)
4772  {
4773  	m->m_pkthdr.ether_vtag = tag;
4774  }
4775  
4776  uint16_t
if_getvtag(struct mbuf * m)4777  if_getvtag(struct mbuf *m)
4778  {
4779  	return (m->m_pkthdr.ether_vtag);
4780  }
4781  
4782  int
if_sendq_empty(if_t ifp)4783  if_sendq_empty(if_t ifp)
4784  {
4785  	return (IFQ_DRV_IS_EMPTY(&ifp->if_snd));
4786  }
4787  
4788  struct ifaddr *
if_getifaddr(const if_t ifp)4789  if_getifaddr(const if_t ifp)
4790  {
4791  	return (ifp->if_addr);
4792  }
4793  
4794  int
if_setsendqready(if_t ifp)4795  if_setsendqready(if_t ifp)
4796  {
4797  	IFQ_SET_READY(&ifp->if_snd);
4798  	return (0);
4799  }
4800  
4801  int
if_setsendqlen(if_t ifp,int tx_desc_count)4802  if_setsendqlen(if_t ifp, int tx_desc_count)
4803  {
4804  	IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count);
4805  	ifp->if_snd.ifq_drv_maxlen = tx_desc_count;
4806  	return (0);
4807  }
4808  
4809  void
if_setnetmapadapter(if_t ifp,struct netmap_adapter * na)4810  if_setnetmapadapter(if_t ifp, struct netmap_adapter *na)
4811  {
4812  	ifp->if_netmap = na;
4813  }
4814  
4815  struct netmap_adapter *
if_getnetmapadapter(if_t ifp)4816  if_getnetmapadapter(if_t ifp)
4817  {
4818  	return (ifp->if_netmap);
4819  }
4820  
4821  int
if_vlantrunkinuse(if_t ifp)4822  if_vlantrunkinuse(if_t ifp)
4823  {
4824  	return (ifp->if_vlantrunk != NULL);
4825  }
4826  
4827  void
if_init(if_t ifp,void * ctx)4828  if_init(if_t ifp, void *ctx)
4829  {
4830  	(*ifp->if_init)(ctx);
4831  }
4832  
4833  void
if_input(if_t ifp,struct mbuf * sendmp)4834  if_input(if_t ifp, struct mbuf* sendmp)
4835  {
4836  	(*ifp->if_input)(ifp, sendmp);
4837  }
4838  
4839  int
if_transmit(if_t ifp,struct mbuf * m)4840  if_transmit(if_t ifp, struct mbuf *m)
4841  {
4842  	return ((*ifp->if_transmit)(ifp, m));
4843  }
4844  
4845  int
if_resolvemulti(if_t ifp,struct sockaddr ** srcs,struct sockaddr * dst)4846  if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst)
4847  {
4848  	if (ifp->if_resolvemulti == NULL)
4849  		return (EOPNOTSUPP);
4850  
4851  	return (ifp->if_resolvemulti(ifp, srcs, dst));
4852  }
4853  
4854  int
if_ioctl(if_t ifp,u_long cmd,void * data)4855  if_ioctl(if_t ifp, u_long cmd, void *data)
4856  {
4857  	if (ifp->if_ioctl == NULL)
4858  		return (EOPNOTSUPP);
4859  
4860  	return (ifp->if_ioctl(ifp, cmd, data));
4861  }
4862  
4863  struct mbuf *
if_dequeue(if_t ifp)4864  if_dequeue(if_t ifp)
4865  {
4866  	struct mbuf *m;
4867  
4868  	IFQ_DRV_DEQUEUE(&ifp->if_snd, m);
4869  	return (m);
4870  }
4871  
4872  int
if_sendq_prepend(if_t ifp,struct mbuf * m)4873  if_sendq_prepend(if_t ifp, struct mbuf *m)
4874  {
4875  	IFQ_DRV_PREPEND(&ifp->if_snd, m);
4876  	return (0);
4877  }
4878  
4879  int
if_setifheaderlen(if_t ifp,int len)4880  if_setifheaderlen(if_t ifp, int len)
4881  {
4882  	ifp->if_hdrlen = len;
4883  	return (0);
4884  }
4885  
4886  char *
if_getlladdr(const if_t ifp)4887  if_getlladdr(const if_t ifp)
4888  {
4889  	return (IF_LLADDR(ifp));
4890  }
4891  
4892  void *
if_gethandle(u_char type)4893  if_gethandle(u_char type)
4894  {
4895  	return (if_alloc(type));
4896  }
4897  
4898  void
if_vlancap(if_t ifp)4899  if_vlancap(if_t ifp)
4900  {
4901  	VLAN_CAPABILITIES(ifp);
4902  }
4903  
4904  int
if_sethwtsomax(if_t ifp,u_int if_hw_tsomax)4905  if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4906  {
4907  	ifp->if_hw_tsomax = if_hw_tsomax;
4908          return (0);
4909  }
4910  
4911  int
if_sethwtsomaxsegcount(if_t ifp,u_int if_hw_tsomaxsegcount)4912  if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4913  {
4914  	ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4915          return (0);
4916  }
4917  
4918  int
if_sethwtsomaxsegsize(if_t ifp,u_int if_hw_tsomaxsegsize)4919  if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4920  {
4921  	ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4922          return (0);
4923  }
4924  
4925  u_int
if_gethwtsomax(const if_t ifp)4926  if_gethwtsomax(const if_t ifp)
4927  {
4928  	return (ifp->if_hw_tsomax);
4929  }
4930  
4931  u_int
if_gethwtsomaxsegcount(const if_t ifp)4932  if_gethwtsomaxsegcount(const if_t ifp)
4933  {
4934  	return (ifp->if_hw_tsomaxsegcount);
4935  }
4936  
4937  u_int
if_gethwtsomaxsegsize(const if_t ifp)4938  if_gethwtsomaxsegsize(const if_t ifp)
4939  {
4940  	return (ifp->if_hw_tsomaxsegsize);
4941  }
4942  
4943  void
if_setinitfn(if_t ifp,if_init_fn_t init_fn)4944  if_setinitfn(if_t ifp, if_init_fn_t init_fn)
4945  {
4946  	ifp->if_init = init_fn;
4947  }
4948  
4949  void
if_setinputfn(if_t ifp,if_input_fn_t input_fn)4950  if_setinputfn(if_t ifp, if_input_fn_t input_fn)
4951  {
4952  	ifp->if_input = input_fn;
4953  }
4954  
4955  if_input_fn_t
if_getinputfn(if_t ifp)4956  if_getinputfn(if_t ifp)
4957  {
4958  	return (ifp->if_input);
4959  }
4960  
4961  void
if_setioctlfn(if_t ifp,if_ioctl_fn_t ioctl_fn)4962  if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn)
4963  {
4964  	ifp->if_ioctl = ioctl_fn;
4965  }
4966  
4967  void
if_setoutputfn(if_t ifp,if_output_fn_t output_fn)4968  if_setoutputfn(if_t ifp, if_output_fn_t output_fn)
4969  {
4970  	ifp->if_output = output_fn;
4971  }
4972  
4973  void
if_setstartfn(if_t ifp,if_start_fn_t start_fn)4974  if_setstartfn(if_t ifp, if_start_fn_t start_fn)
4975  {
4976  	ifp->if_start = start_fn;
4977  }
4978  
4979  if_start_fn_t
if_getstartfn(if_t ifp)4980  if_getstartfn(if_t ifp)
4981  {
4982  	return (ifp->if_start);
4983  }
4984  
4985  void
if_settransmitfn(if_t ifp,if_transmit_fn_t start_fn)4986  if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4987  {
4988  	ifp->if_transmit = start_fn;
4989  }
4990  
4991  if_transmit_fn_t
if_gettransmitfn(if_t ifp)4992  if_gettransmitfn(if_t ifp)
4993  {
4994  	return (ifp->if_transmit);
4995  }
4996  
4997  void
if_setqflushfn(if_t ifp,if_qflush_fn_t flush_fn)4998  if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4999  {
5000  	ifp->if_qflush = flush_fn;
5001  }
5002  
5003  void
if_setsndtagallocfn(if_t ifp,if_snd_tag_alloc_t alloc_fn)5004  if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn)
5005  {
5006  	ifp->if_snd_tag_alloc = alloc_fn;
5007  }
5008  
5009  int
if_snd_tag_alloc(if_t ifp,union if_snd_tag_alloc_params * params,struct m_snd_tag ** mstp)5010  if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params,
5011      struct m_snd_tag **mstp)
5012  {
5013  	if (ifp->if_snd_tag_alloc == NULL)
5014  		return (EOPNOTSUPP);
5015  	return (ifp->if_snd_tag_alloc(ifp, params, mstp));
5016  }
5017  
5018  void
if_setgetcounterfn(if_t ifp,if_get_counter_t fn)5019  if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
5020  {
5021  	ifp->if_get_counter = fn;
5022  }
5023  
5024  void
if_setreassignfn(if_t ifp,if_reassign_fn_t fn)5025  if_setreassignfn(if_t ifp, if_reassign_fn_t fn)
5026  {
5027  	ifp->if_reassign = fn;
5028  }
5029  
5030  void
if_setratelimitqueryfn(if_t ifp,if_ratelimit_query_t fn)5031  if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn)
5032  {
5033  	ifp->if_ratelimit_query = fn;
5034  }
5035  
5036  void
if_setdebugnet_methods(if_t ifp,struct debugnet_methods * m)5037  if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m)
5038  {
5039  	ifp->if_debugnet_methods = m;
5040  }
5041  
5042  struct label *
if_getmaclabel(if_t ifp)5043  if_getmaclabel(if_t ifp)
5044  {
5045  	return (ifp->if_label);
5046  }
5047  
5048  void
if_setmaclabel(if_t ifp,struct label * label)5049  if_setmaclabel(if_t ifp, struct label *label)
5050  {
5051  	ifp->if_label = label;
5052  }
5053  
5054  int
if_gettype(if_t ifp)5055  if_gettype(if_t ifp)
5056  {
5057  	return (ifp->if_type);
5058  }
5059  
5060  void *
if_getllsoftc(if_t ifp)5061  if_getllsoftc(if_t ifp)
5062  {
5063  	return (ifp->if_llsoftc);
5064  }
5065  
5066  void
if_setllsoftc(if_t ifp,void * llsoftc)5067  if_setllsoftc(if_t ifp, void *llsoftc)
5068  {
5069  	ifp->if_llsoftc = llsoftc;
5070  };
5071  
5072  int
if_getlinkstate(if_t ifp)5073  if_getlinkstate(if_t ifp)
5074  {
5075  	return (ifp->if_link_state);
5076  }
5077  
5078  const uint8_t *
if_getbroadcastaddr(if_t ifp)5079  if_getbroadcastaddr(if_t ifp)
5080  {
5081  	return (ifp->if_broadcastaddr);
5082  }
5083  
5084  void
if_setbroadcastaddr(if_t ifp,const uint8_t * addr)5085  if_setbroadcastaddr(if_t ifp, const uint8_t *addr)
5086  {
5087  	ifp->if_broadcastaddr = addr;
5088  }
5089  
5090  int
if_getnumadomain(if_t ifp)5091  if_getnumadomain(if_t ifp)
5092  {
5093  	return (ifp->if_numa_domain);
5094  }
5095  
5096  uint64_t
if_getcounter(if_t ifp,ift_counter counter)5097  if_getcounter(if_t ifp, ift_counter counter)
5098  {
5099  	return (ifp->if_get_counter(ifp, counter));
5100  }
5101  
5102  bool
if_altq_is_enabled(if_t ifp)5103  if_altq_is_enabled(if_t ifp)
5104  {
5105  	return (ALTQ_IS_ENABLED(&ifp->if_snd));
5106  }
5107  
5108  struct vnet *
if_getvnet(if_t ifp)5109  if_getvnet(if_t ifp)
5110  {
5111  	return (ifp->if_vnet);
5112  }
5113  
5114  void *
if_getafdata(if_t ifp,int af)5115  if_getafdata(if_t ifp, int af)
5116  {
5117  	return (ifp->if_afdata[af]);
5118  }
5119  
5120  u_int
if_getfib(if_t ifp)5121  if_getfib(if_t ifp)
5122  {
5123  	return (ifp->if_fib);
5124  }
5125  
5126  uint8_t
if_getaddrlen(if_t ifp)5127  if_getaddrlen(if_t ifp)
5128  {
5129  	return (ifp->if_addrlen);
5130  }
5131  
5132  struct bpf_if *
if_getbpf(if_t ifp)5133  if_getbpf(if_t ifp)
5134  {
5135  	return (ifp->if_bpf);
5136  }
5137  
5138  struct ifvlantrunk *
if_getvlantrunk(if_t ifp)5139  if_getvlantrunk(if_t ifp)
5140  {
5141  	return (ifp->if_vlantrunk);
5142  }
5143  
5144  uint8_t
if_getpcp(if_t ifp)5145  if_getpcp(if_t ifp)
5146  {
5147  	return (ifp->if_pcp);
5148  }
5149  
5150  void *
if_getl2com(if_t ifp)5151  if_getl2com(if_t ifp)
5152  {
5153  	return (ifp->if_l2com);
5154  }
5155  
5156  void
if_setipsec_accel_methods(if_t ifp,const struct if_ipsec_accel_methods * m)5157  if_setipsec_accel_methods(if_t ifp, const struct if_ipsec_accel_methods *m)
5158  {
5159  	ifp->if_ipsec_accel_m = m;
5160  }
5161  
5162  #ifdef DDB
5163  static void
if_show_ifnet(struct ifnet * ifp)5164  if_show_ifnet(struct ifnet *ifp)
5165  {
5166  	if (ifp == NULL)
5167  		return;
5168  	db_printf("%s:\n", ifp->if_xname);
5169  #define	IF_DB_PRINTF(f, e)	db_printf("   %s = " f "\n", #e, ifp->e);
5170  	IF_DB_PRINTF("%s", if_dname);
5171  	IF_DB_PRINTF("%d", if_dunit);
5172  	IF_DB_PRINTF("%s", if_description);
5173  	IF_DB_PRINTF("%u", if_index);
5174  	IF_DB_PRINTF("%d", if_idxgen);
5175  	IF_DB_PRINTF("%u", if_refcount);
5176  	IF_DB_PRINTF("%p", if_softc);
5177  	IF_DB_PRINTF("%p", if_l2com);
5178  	IF_DB_PRINTF("%p", if_llsoftc);
5179  	IF_DB_PRINTF("%d", if_amcount);
5180  	IF_DB_PRINTF("%p", if_addr);
5181  	IF_DB_PRINTF("%p", if_broadcastaddr);
5182  	IF_DB_PRINTF("%p", if_afdata);
5183  	IF_DB_PRINTF("%d", if_afdata_initialized);
5184  	IF_DB_PRINTF("%u", if_fib);
5185  	IF_DB_PRINTF("%p", if_vnet);
5186  	IF_DB_PRINTF("%p", if_home_vnet);
5187  	IF_DB_PRINTF("%p", if_vlantrunk);
5188  	IF_DB_PRINTF("%p", if_bpf);
5189  	IF_DB_PRINTF("%u", if_pcount);
5190  	IF_DB_PRINTF("%p", if_bridge);
5191  	IF_DB_PRINTF("%p", if_lagg);
5192  	IF_DB_PRINTF("%p", if_pf_kif);
5193  	IF_DB_PRINTF("%p", if_carp);
5194  	IF_DB_PRINTF("%p", if_label);
5195  	IF_DB_PRINTF("%p", if_netmap);
5196  	IF_DB_PRINTF("0x%08x", if_flags);
5197  	IF_DB_PRINTF("0x%08x", if_drv_flags);
5198  	IF_DB_PRINTF("0x%08x", if_capabilities);
5199  	IF_DB_PRINTF("0x%08x", if_capenable);
5200  	IF_DB_PRINTF("%p", if_snd.ifq_head);
5201  	IF_DB_PRINTF("%p", if_snd.ifq_tail);
5202  	IF_DB_PRINTF("%d", if_snd.ifq_len);
5203  	IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
5204  	IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
5205  	IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
5206  	IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
5207  	IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
5208  	IF_DB_PRINTF("%d", if_snd.altq_type);
5209  	IF_DB_PRINTF("%x", if_snd.altq_flags);
5210  #undef IF_DB_PRINTF
5211  }
5212  
DB_SHOW_COMMAND(ifnet,db_show_ifnet)5213  DB_SHOW_COMMAND(ifnet, db_show_ifnet)
5214  {
5215  	if (!have_addr) {
5216  		db_printf("usage: show ifnet <struct ifnet *>\n");
5217  		return;
5218  	}
5219  
5220  	if_show_ifnet((struct ifnet *)addr);
5221  }
5222  
DB_SHOW_ALL_COMMAND(ifnets,db_show_all_ifnets)5223  DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
5224  {
5225  	struct ifnet *ifp;
5226  	u_short idx;
5227  
5228  	for (idx = 1; idx <= if_index; idx++) {
5229  		ifp = ifindex_table[idx].ife_ifnet;
5230  		if (ifp == NULL)
5231  			continue;
5232  		db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
5233  		if (db_pager_quit)
5234  			break;
5235  	}
5236  }
5237  #endif	/* DDB */
5238