Home | History | Annotate | Line # | Download | only in net
if.h revision 1.282
      1 /*	$NetBSD: if.h,v 1.282 2020/02/14 22:04:12 thorpej Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1999, 2000, 2001 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by William Studenmund and Jason R. Thorpe.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     29  * POSSIBILITY OF SUCH DAMAGE.
     30  */
     31 
     32 /*
     33  * Copyright (c) 1982, 1986, 1989, 1993
     34  *	The Regents of the University of California.  All rights reserved.
     35  *
     36  * Redistribution and use in source and binary forms, with or without
     37  * modification, are permitted provided that the following conditions
     38  * are met:
     39  * 1. Redistributions of source code must retain the above copyright
     40  *    notice, this list of conditions and the following disclaimer.
     41  * 2. Redistributions in binary form must reproduce the above copyright
     42  *    notice, this list of conditions and the following disclaimer in the
     43  *    documentation and/or other materials provided with the distribution.
     44  * 3. Neither the name of the University nor the names of its contributors
     45  *    may be used to endorse or promote products derived from this software
     46  *    without specific prior written permission.
     47  *
     48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     58  * SUCH DAMAGE.
     59  *
     60  *	@(#)if.h	8.3 (Berkeley) 2/9/95
     61  */
     62 
     63 #ifndef _NET_IF_H_
     64 #define _NET_IF_H_
     65 
     66 #if !defined(_KERNEL) && !defined(_STANDALONE)
     67 #include <stdbool.h>
     68 #endif
     69 
     70 #include <sys/featuretest.h>
     71 
     72 /*
     73  * Length of interface external name, including terminating '\0'.
     74  * Note: this is the same size as a generic device's external name.
     75  */
     76 #define IF_NAMESIZE 16
     77 
     78 /*
     79  * Length of interface description, including terminating '\0'.
     80  */
     81 #define	IFDESCRSIZE	64
     82 
     83 #if defined(_NETBSD_SOURCE)
     84 
     85 #include <sys/socket.h>
     86 #include <sys/queue.h>
     87 #include <sys/mutex.h>
     88 
     89 #include <net/dlt.h>
     90 #include <net/pfil.h>
     91 #ifdef _KERNEL
     92 #include <net/pktqueue.h>
     93 #include <sys/pslist.h>
     94 #include <sys/pserialize.h>
     95 #include <sys/psref.h>
     96 #include <sys/module_hook.h>
     97 #endif
     98 
     99 /*
    100  * Always include ALTQ glue here -- we use the ALTQ interface queue
    101  * structure even when ALTQ is not configured into the kernel so that
    102  * the size of struct ifnet does not changed based on the option.  The
    103  * ALTQ queue structure is API-compatible with the legacy ifqueue.
    104  */
    105 #include <altq/if_altq.h>
    106 
    107 /*
    108  * Structures defining a network interface, providing a packet
    109  * transport mechanism (ala level 0 of the PUP protocols).
    110  *
    111  * Each interface accepts output datagrams of a specified maximum
    112  * length, and provides higher level routines with input datagrams
    113  * received from its medium.
    114  *
    115  * Output occurs when the routine if_output is called, with four parameters:
    116  *	(*ifp->if_output)(ifp, m, dst, rt)
    117  * Here m is the mbuf chain to be sent and dst is the destination address.
    118  * The output routine encapsulates the supplied datagram if necessary,
    119  * and then transmits it on its medium.
    120  *
    121  * On input, each interface unwraps the data received by it, and either
    122  * places it on the input queue of a internetwork datagram routine
    123  * and posts the associated software interrupt, or passes the datagram to a raw
    124  * packet input routine.
    125  *
    126  * Routines exist for locating interfaces by their addresses
    127  * or for locating a interface on a certain network, as well as more general
    128  * routing and gateway routines maintaining information used to locate
    129  * interfaces.  These routines live in the files if.c and route.c
    130  */
    131 #include <sys/time.h>
    132 
    133 #if defined(_KERNEL_OPT)
    134 #include "opt_compat_netbsd.h"
    135 #include "opt_gateway.h"
    136 #endif
    137 
    138 struct mbuf;
    139 struct proc;
    140 struct rtentry;
    141 struct socket;
    142 struct ether_header;
    143 struct ifaddr;
    144 struct ifnet;
    145 struct rt_addrinfo;
    146 
    147 #define	IFNAMSIZ	IF_NAMESIZE
    148 
    149 /*
    150  * Structure describing a `cloning' interface.
    151  */
    152 struct if_clone {
    153 	LIST_ENTRY(if_clone) ifc_list;	/* on list of cloners */
    154 	const char *ifc_name;		/* name of device, e.g. `gif' */
    155 	size_t ifc_namelen;		/* length of name */
    156 
    157 	int	(*ifc_create)(struct if_clone *, int);
    158 	int	(*ifc_destroy)(struct ifnet *);
    159 };
    160 
    161 #define	IF_CLONE_INITIALIZER(name, create, destroy)			\
    162 	{ { NULL, NULL }, name, sizeof(name) - 1, create, destroy }
    163 
    164 /*
    165  * Structure used to query names of interface cloners.
    166  */
    167 struct if_clonereq {
    168 	int	ifcr_total;		/* total cloners (out) */
    169 	int	ifcr_count;		/* room for this many in user buffer */
    170 	char	*ifcr_buffer;		/* buffer for cloner names */
    171 };
    172 
    173 /*
    174  * Structure defining statistics and other data kept regarding a network
    175  * interface.
    176  *
    177  * Only used for exporting data from the interface.
    178  */
    179 struct if_data {
    180 	/* generic interface information */
    181 	u_char	ifi_type;		/* ethernet, tokenring, etc. */
    182 	u_char	ifi_addrlen;		/* media address length */
    183 	u_char	ifi_hdrlen;		/* media header length */
    184 	int	ifi_link_state;		/* current link state */
    185 	uint64_t ifi_mtu;		/* maximum transmission unit */
    186 	uint64_t ifi_metric;		/* routing metric (external only) */
    187 	uint64_t ifi_baudrate;		/* linespeed */
    188 	/* volatile statistics */
    189 	uint64_t ifi_ipackets;		/* packets received on interface */
    190 	uint64_t ifi_ierrors;		/* input errors on interface */
    191 	uint64_t ifi_opackets;		/* packets sent on interface */
    192 	uint64_t ifi_oerrors;		/* output errors on interface */
    193 	uint64_t ifi_collisions;	/* collisions on csma interfaces */
    194 	uint64_t ifi_ibytes;		/* total number of octets received */
    195 	uint64_t ifi_obytes;		/* total number of octets sent */
    196 	uint64_t ifi_imcasts;		/* packets received via multicast */
    197 	uint64_t ifi_omcasts;		/* packets sent via multicast */
    198 	uint64_t ifi_iqdrops;		/* dropped on input, this interface */
    199 	uint64_t ifi_noproto;		/* destined for unsupported protocol */
    200 	struct	timespec ifi_lastchange;/* last operational state change */
    201 };
    202 
    203 /*
    204  * Values for if_link_state.
    205  */
    206 #define	LINK_STATE_UNKNOWN	0	/* link invalid/unknown */
    207 #define	LINK_STATE_DOWN		1	/* link is down */
    208 #define	LINK_STATE_UP		2	/* link is up */
    209 
    210 /*
    211  * Structure defining a queue for a network interface.
    212  */
    213 struct ifqueue {
    214 	struct		mbuf *ifq_head;
    215 	struct		mbuf *ifq_tail;
    216 	int		ifq_len;
    217 	int		ifq_maxlen;
    218 	int		ifq_drops;
    219 	kmutex_t	*ifq_lock;
    220 };
    221 
    222 #ifdef _KERNEL
    223 #include <sys/percpu.h>
    224 #include <sys/callout.h>
    225 #include <sys/rwlock.h>
    226 #include <sys/workqueue.h>
    227 
    228 #endif /* _KERNEL */
    229 
    230 /*
    231  * Structure defining a queue for a network interface.
    232  *
    233  * (Would like to call this struct ``if'', but C isn't PL/1.)
    234  */
    235 TAILQ_HEAD(ifnet_head, ifnet);		/* the actual queue head */
    236 
    237 struct bridge_softc;
    238 struct bridge_iflist;
    239 struct callout;
    240 struct krwlock;
    241 struct if_percpuq;
    242 struct if_deferred_start;
    243 struct in6_multi;
    244 
    245 typedef unsigned short if_index_t;
    246 
    247 /*
    248  * Interface.  Field markings and the corresponding locks:
    249  *
    250  * i:	IFNET_LOCK (a.k.a., if_ioctl_lock)
    251  * q:	ifq_lock (struct ifaltq)
    252  * a:	if_afdata_lock
    253  * 6:	in6_multilock (global lock)
    254  * ::	unlocked, stable
    255  * ?:	unknown, maybe unsafe
    256  *
    257  * Lock order: IFNET_LOCK => in6_multilock => if_afdata_lock => ifq_lock
    258  *   Note that currently if_afdata_lock and ifq_lock aren't held
    259  *   at the same time, but define the order anyway.
    260  *
    261  * Lock order of IFNET_LOCK with other locks:
    262  *     softnet_lock => solock => IFNET_LOCK => ND6_LOCK, in_multilock
    263  */
    264 typedef struct ifnet {
    265 	void		*if_softc;	/* :: lower-level data for this if */
    266 	/* DEPRECATED. Keep it to avoid breaking kvm(3) users */
    267 	TAILQ_ENTRY(ifnet)
    268 			if_list;	/* i: all struct ifnets are chained */
    269 	TAILQ_HEAD(, ifaddr)
    270 			if_addrlist;	/* i: linked list of addresses per if */
    271 	char		if_xname[IFNAMSIZ];
    272 					/* :: external name (name + unit) */
    273 	int		if_pcount;	/* i: number of promiscuous listeners */
    274 	struct bpf_if	*if_bpf;	/* :: packet filter structure */
    275 	if_index_t	if_index;	/* :: numeric abbreviation for this if */
    276 	short		if_timer;	/* ?: time 'til if_slowtimo called */
    277 	unsigned short	if_flags;	/* i: up/down, broadcast, etc. */
    278 	short		if_extflags;	/* :: if_output MP-safe, etc. */
    279 	u_char		if_type;	/* :: ethernet, tokenring, etc. */
    280 	u_char		if_addrlen;	/* :: media address length */
    281 	u_char		if_hdrlen;	/* :: media header length */
    282 	/* XXX audit :? fields here. */
    283 	int		if_link_state;	/* :? current link state */
    284 	uint64_t	if_mtu;		/* :? maximum transmission unit */
    285 	uint64_t	if_metric;	/* :? routing metric (external only) */
    286 	uint64_t	if_baudrate;	/* :? linespeed */
    287 	struct timespec	if_lastchange;	/* :? last operational state change */
    288 #ifdef _KERNEL
    289 	percpu_t	*if_stats;	/* :: statistics */
    290 #else
    291 	void		*if_stats;	/* opaque to user-space */
    292 #endif /* _KERNEL */
    293 	/*
    294 	 * Procedure handles.  If you add more of these, don't forget the
    295 	 * corresponding NULL stub in if.c.
    296 	 */
    297 	int		(*if_output)	/* :: output routine (enqueue) */
    298 			    (struct ifnet *, struct mbuf *, const struct sockaddr *,
    299 			     const struct rtentry *);
    300 	void		(*_if_input)	/* :: input routine (from h/w driver) */
    301 			    (struct ifnet *, struct mbuf *);
    302 	void		(*if_start)	/* :: initiate output routine */
    303 			    (struct ifnet *);
    304 	int		(*if_transmit)	/* :: output routine, must be MP-safe */
    305 			    (struct ifnet *, struct mbuf *);
    306 	int		(*if_ioctl)	/* :: ioctl routine */
    307 			    (struct ifnet *, u_long, void *);
    308 	int		(*if_init)	/* :: init routine */
    309 			    (struct ifnet *);
    310 	void		(*if_stop)	/* :: stop routine */
    311 			    (struct ifnet *, int);
    312 	void		(*if_slowtimo)	/* :: timer routine */
    313 			    (struct ifnet *);
    314 #define	if_watchdog	if_slowtimo
    315 	void		(*if_drain)	/* :: routine to release resources */
    316 			    (struct ifnet *);
    317 	struct ifaltq	if_snd;		/* q: output queue (includes altq) */
    318 	struct ifaddr	*if_dl;		/* i: identity of this interface. */
    319 	const struct sockaddr_dl
    320 			*if_sadl;	/* i: pointer to sockaddr_dl of if_dl */
    321 	/*
    322 	 * May be NULL.  If not NULL, it is the address assigned
    323 	 * to the interface by the manufacturer, so it very likely
    324 	 * to be unique.  It MUST NOT be deleted.  It is highly
    325 	 * suitable for deriving the EUI64 for the interface.
    326 	 */
    327 	struct ifaddr	*if_hwdl;	/* i: h/w identity */
    328 	const uint8_t	*if_broadcastaddr;
    329 					/* :: linklevel broadcast bytestring */
    330 	struct bridge_softc
    331 			*if_bridge;	/* i: bridge glue */
    332 	struct bridge_iflist
    333 			*if_bridgeif;	/* i: shortcut to interface list entry */
    334 	int		if_dlt;		/* :: data link type (<net/dlt.h>) */
    335 	pfil_head_t *	if_pfil;	/* :: filtering point */
    336 	uint64_t	if_capabilities;
    337 					/* i: interface capabilities */
    338 	uint64_t	if_capenable;	/* i: capabilities enabled */
    339 	union {
    340 		void *		carp_s;	/* carp structure (used by !carp ifs) */
    341 		struct ifnet	*carp_d;/* ptr to carpdev (used by carp ifs) */
    342 	}		if_carp_ptr;	/* ?: */
    343 #define if_carp		if_carp_ptr.carp_s
    344 #define if_carpdev	if_carp_ptr.carp_d
    345 	/*
    346 	 * These are pre-computed based on an interfaces enabled
    347 	 * capabilities, for speed elsewhere.
    348 	 */
    349 	int		if_csum_flags_tx;
    350 					/* i: M_CSUM_* flags for Tx */
    351 	int		if_csum_flags_rx;
    352 					/* i: M_CSUM_* flags for Rx */
    353 
    354 	void		*if_afdata[AF_MAX];
    355 					/* a: */
    356 	struct mowner	*if_mowner;	/* ?: who owns mbufs for this interface */
    357 
    358 	void		*if_agrprivate;	/* ?: used only when #if NAGR > 0 */
    359 	void		*if_npf_private;/* ?: associated NPF context */
    360 
    361 	/*
    362 	 * pf specific data, used only when #if NPF > 0.
    363 	 */
    364 	void		*if_pf_kif;	/* ?: pf interface abstraction */
    365 	void		*if_pf_groups;	/* ?: pf interface groups */
    366 	/*
    367 	 * During an ifnet's lifetime, it has only one if_index, but
    368 	 * and if_index is not sufficient to identify an ifnet
    369 	 * because during the lifetime of the system, many ifnets may occupy a
    370 	 * given if_index.  Let us tell different ifnets at the same
    371 	 * if_index apart by their if_index_gen, a unique number that each ifnet
    372 	 * is assigned when it if_attach()s.  Now, the kernel can use the
    373 	 * pair (if_index, if_index_gen) as a weak reference to an ifnet.
    374 	 */
    375 	uint64_t	if_index_gen;	/* :: generation number for the ifnet
    376 					 * at if_index: if two ifnets' index
    377 					 * and generation number are both the
    378 					 * same, they are the same ifnet.
    379 					 */
    380 	struct sysctllog
    381 			*if_sysctl_log;	/* :: */
    382 	int		(*if_initaddr)  /* :: */
    383 			    (struct ifnet *, struct ifaddr *, bool);
    384 	int		(*if_mcastop)	/* :: */
    385 			    (struct ifnet *, const unsigned long,
    386 			    const struct sockaddr *);
    387 	int		(*if_setflags)	/* :: */
    388 			    (struct ifnet *, const u_short);
    389 	kmutex_t	*if_ioctl_lock;	/* :: */
    390 	char		*if_description;	/* i: interface description */
    391 #ifdef _KERNEL /* XXX kvm(3) */
    392 	struct callout	*if_slowtimo_ch;/* :: */
    393 	struct krwlock	*if_afdata_lock;/* :: */
    394 	struct if_percpuq
    395 			*if_percpuq;	/* :: we should remove it in the future */
    396 	struct work	if_link_work;	/* q: linkage on link state work queue */
    397 	uint16_t	if_link_queue;	/* q: masked link state change queue */
    398 					/* q: is link state work scheduled? */
    399 	bool		if_link_scheduled;
    400 					/* q: can link state work be scheduled? */
    401 	bool		if_link_cansched;
    402 	struct pslist_entry
    403 			if_pslist_entry;/* i: */
    404 	struct psref_target
    405 			if_psref;	/* :: */
    406 	struct pslist_head
    407 			if_addr_pslist;	/* i: */
    408 	struct if_deferred_start
    409 			*if_deferred_start;
    410 					/* :: */
    411 	/* XXX should be protocol independent */
    412 	LIST_HEAD(, in6_multi)
    413 			if_multiaddrs;	/* 6: */
    414 #endif
    415 } ifnet_t;
    416 
    417 #include <net/if_stats.h>
    418 
    419 #define	if_name(ifp)	((ifp)->if_xname)
    420 
    421 #define	IFF_UP		0x0001		/* interface is up */
    422 #define	IFF_BROADCAST	0x0002		/* broadcast address valid */
    423 #define	IFF_DEBUG	0x0004		/* turn on debugging */
    424 #define	IFF_LOOPBACK	0x0008		/* is a loopback net */
    425 #define	IFF_POINTOPOINT	0x0010		/* interface is point-to-point link */
    426 /*			0x0020		   was IFF_NOTRAILERS */
    427 #define	IFF_RUNNING	0x0040		/* resources allocated */
    428 #define	IFF_NOARP	0x0080		/* no address resolution protocol */
    429 #define	IFF_PROMISC	0x0100		/* receive all packets */
    430 #define	IFF_ALLMULTI	0x0200		/* receive all multicast packets */
    431 #define	IFF_OACTIVE	0x0400		/* transmission in progress */
    432 #define	IFF_SIMPLEX	0x0800		/* can't hear own transmissions */
    433 #define	IFF_LINK0	0x1000		/* per link layer defined bit */
    434 #define	IFF_LINK1	0x2000		/* per link layer defined bit */
    435 #define	IFF_LINK2	0x4000		/* per link layer defined bit */
    436 #define	IFF_MULTICAST	0x8000		/* supports multicast */
    437 
    438 #define	IFEF_MPSAFE			__BIT(0)	/* handlers can run in parallel (see below) */
    439 #define	IFEF_NO_LINK_STATE_CHANGE	__BIT(1)	/* doesn't use link state interrupts */
    440 
    441 /*
    442  * The guidelines for converting an interface to IFEF_MPSAFE are as follows
    443  *
    444  * Enabling IFEF_MPSAFE on an interface suppresses taking KERNEL_LOCK when
    445  * calling the following handlers:
    446  * - if_start
    447  *   - Note that if_transmit is always called without KERNEL_LOCK
    448  * - if_output
    449  * - if_ioctl
    450  * - if_init
    451  * - if_stop
    452  *
    453  * This means that an interface with IFEF_MPSAFE must make the above handlers
    454  * MP-safe or take KERNEL_LOCK by itself inside handlers that aren't MP-safe
    455  * yet.
    456  *
    457  * There are some additional restrictions to access member variables of struct
    458  * ifnet:
    459  * - if_flags
    460  *   - Must be updated with holding IFNET_LOCK
    461  *   - You cannot use the flag in Tx/Rx paths anymore because there is no
    462  *     synchronization on the flag except for IFNET_LOCK
    463  *   - Note that IFNET_LOCK can't be taken in softint because it's known
    464  *     that it causes a deadlock
    465  *     - Some synchronization mechanisms such as pserialize_perform are called
    466  *       with IFNET_LOCK and also require context switches on every CPUs
    467  *       that mean softints finish so trying to take IFNET_LOCK in softint
    468  *       might block on IFNET_LOCK and prevent such synchronization mechanisms
    469  *       from being completed
    470  *     - Currently the deadlock occurs only if NET_MPSAFE is enabled, however,
    471  *       we should deal with the restriction because NET_MPSAFE will be enabled
    472  *       by default in the future
    473  * - if_watchdog and if_timer
    474  *   - The watchdog framework works only for non-IFEF_MPSAFE interfaces
    475  *     that rely on KERNEL_LOCK
    476  *   - Interfaces with IFEF_MPSAFE have to provide its own watchdog mechanism
    477  *     if needed
    478  *     - Keep if_watchdog NULL when calling if_attach
    479  */
    480 
    481 #ifdef _KERNEL
    482 static __inline bool
    483 if_is_mpsafe(struct ifnet *ifp)
    484 {
    485 
    486 	return ((ifp->if_extflags & IFEF_MPSAFE) != 0);
    487 }
    488 
    489 static __inline int
    490 if_output_lock(struct ifnet *cifp, struct ifnet *ifp, struct mbuf *m,
    491     const struct sockaddr *dst, const struct rtentry *rt)
    492 {
    493 
    494 	if (if_is_mpsafe(cifp)) {
    495 		return (*cifp->if_output)(ifp, m, dst, rt);
    496 	} else {
    497 		int ret;
    498 
    499 		KERNEL_LOCK(1, NULL);
    500 		ret = (*cifp->if_output)(ifp, m, dst, rt);
    501 		KERNEL_UNLOCK_ONE(NULL);
    502 		return ret;
    503 	}
    504 }
    505 
    506 static __inline void
    507 if_start_lock(struct ifnet *ifp)
    508 {
    509 
    510 	if (if_is_mpsafe(ifp)) {
    511 		(*ifp->if_start)(ifp);
    512 	} else {
    513 		KERNEL_LOCK(1, NULL);
    514 		(*ifp->if_start)(ifp);
    515 		KERNEL_UNLOCK_ONE(NULL);
    516 	}
    517 }
    518 
    519 static __inline bool
    520 if_is_link_state_changeable(struct ifnet *ifp)
    521 {
    522 
    523 	return ((ifp->if_extflags & IFEF_NO_LINK_STATE_CHANGE) == 0);
    524 }
    525 
    526 #define KERNEL_LOCK_IF_IFP_MPSAFE(ifp)					\
    527 	do { if (if_is_mpsafe(ifp)) { KERNEL_LOCK(1, NULL); } } while (0)
    528 #define KERNEL_UNLOCK_IF_IFP_MPSAFE(ifp)				\
    529 	do { if (if_is_mpsafe(ifp)) { KERNEL_UNLOCK_ONE(NULL); } } while (0)
    530 
    531 #define KERNEL_LOCK_UNLESS_IFP_MPSAFE(ifp)				\
    532 	do { if (!if_is_mpsafe(ifp)) { KERNEL_LOCK(1, NULL); } } while (0)
    533 #define KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(ifp)				\
    534 	do { if (!if_is_mpsafe(ifp)) { KERNEL_UNLOCK_ONE(NULL); } } while (0)
    535 
    536 #ifdef _KERNEL_OPT
    537 #include "opt_net_mpsafe.h"
    538 #endif
    539 
    540 /* XXX explore a better place to define */
    541 #ifdef NET_MPSAFE
    542 
    543 #define KERNEL_LOCK_UNLESS_NET_MPSAFE()		do { } while (0)
    544 #define KERNEL_UNLOCK_UNLESS_NET_MPSAFE()	do { } while (0)
    545 
    546 #define SOFTNET_LOCK_UNLESS_NET_MPSAFE()	do { } while (0)
    547 #define SOFTNET_UNLOCK_UNLESS_NET_MPSAFE()	do { } while (0)
    548 
    549 #define SOFTNET_LOCK_IF_NET_MPSAFE()					\
    550 	do { mutex_enter(softnet_lock); } while (0)
    551 #define SOFTNET_UNLOCK_IF_NET_MPSAFE()					\
    552 	do { mutex_exit(softnet_lock); } while (0)
    553 
    554 #else /* NET_MPSAFE */
    555 
    556 #define KERNEL_LOCK_UNLESS_NET_MPSAFE()					\
    557 	do { KERNEL_LOCK(1, NULL); } while (0)
    558 #define KERNEL_UNLOCK_UNLESS_NET_MPSAFE()				\
    559 	do { KERNEL_UNLOCK_ONE(NULL); } while (0)
    560 
    561 #define SOFTNET_LOCK_UNLESS_NET_MPSAFE()				\
    562 	do { mutex_enter(softnet_lock); } while (0)
    563 #define SOFTNET_UNLOCK_UNLESS_NET_MPSAFE()				\
    564 	do { mutex_exit(softnet_lock); } while (0)
    565 
    566 #define SOFTNET_LOCK_IF_NET_MPSAFE()		do { } while (0)
    567 #define SOFTNET_UNLOCK_IF_NET_MPSAFE()		do { } while (0)
    568 
    569 #endif /* NET_MPSAFE */
    570 
    571 #define SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE()				\
    572 	do {								\
    573 		SOFTNET_LOCK_UNLESS_NET_MPSAFE();			\
    574 		KERNEL_LOCK_UNLESS_NET_MPSAFE();			\
    575 	} while (0)
    576 
    577 #define SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE()			\
    578 	do {								\
    579 		KERNEL_UNLOCK_UNLESS_NET_MPSAFE();			\
    580 		SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();			\
    581 	} while (0)
    582 
    583 #endif /* _KERNEL */
    584 
    585 #define	IFFBITS \
    586     "\020\1UP\2BROADCAST\3DEBUG\4LOOPBACK\5POINTOPOINT" \
    587     "\7RUNNING\10NOARP\11PROMISC\12ALLMULTI\13OACTIVE\14SIMPLEX" \
    588     "\15LINK0\16LINK1\17LINK2\20MULTICAST"
    589 
    590 /* flags set internally only: */
    591 #define	IFF_CANTCHANGE \
    592 	(IFF_BROADCAST|IFF_POINTOPOINT|IFF_RUNNING|IFF_OACTIVE|\
    593 	    IFF_SIMPLEX|IFF_MULTICAST|IFF_ALLMULTI|IFF_PROMISC)
    594 
    595 /*
    596  * Some convenience macros used for setting ifi_baudrate.
    597  */
    598 #define	IF_Kbps(x)	((x) * 1000ULL)			/* kilobits/sec. */
    599 #define	IF_Mbps(x)	(IF_Kbps((x) * 1000ULL))	/* megabits/sec. */
    600 #define	IF_Gbps(x)	(IF_Mbps((x) * 1000ULL))	/* gigabits/sec. */
    601 
    602 /* Capabilities that interfaces can advertise. */
    603 					/* 0x01 .. 0x40 were previously used */
    604 #define	IFCAP_TSOv4		0x00080	/* can do TCPv4 segmentation offload */
    605 #define	IFCAP_CSUM_IPv4_Rx	0x00100	/* can do IPv4 header checksums (Rx) */
    606 #define	IFCAP_CSUM_IPv4_Tx	0x00200	/* can do IPv4 header checksums (Tx) */
    607 #define	IFCAP_CSUM_TCPv4_Rx	0x00400	/* can do IPv4/TCP checksums (Rx) */
    608 #define	IFCAP_CSUM_TCPv4_Tx	0x00800	/* can do IPv4/TCP checksums (Tx) */
    609 #define	IFCAP_CSUM_UDPv4_Rx	0x01000	/* can do IPv4/UDP checksums (Rx) */
    610 #define	IFCAP_CSUM_UDPv4_Tx	0x02000	/* can do IPv4/UDP checksums (Tx) */
    611 #define	IFCAP_CSUM_TCPv6_Rx	0x04000	/* can do IPv6/TCP checksums (Rx) */
    612 #define	IFCAP_CSUM_TCPv6_Tx	0x08000	/* can do IPv6/TCP checksums (Tx) */
    613 #define	IFCAP_CSUM_UDPv6_Rx	0x10000	/* can do IPv6/UDP checksums (Rx) */
    614 #define	IFCAP_CSUM_UDPv6_Tx	0x20000	/* can do IPv6/UDP checksums (Tx) */
    615 #define	IFCAP_TSOv6		0x40000	/* can do TCPv6 segmentation offload */
    616 #define	IFCAP_LRO		0x80000	/* can do Large Receive Offload */
    617 #define	IFCAP_MASK		0xfff80 /* currently valid capabilities */
    618 
    619 #define	IFCAPBITS		\
    620 	"\020"			\
    621 	"\10TSO4"		\
    622 	"\11IP4CSUM_Rx"		\
    623 	"\12IP4CSUM_Tx"		\
    624 	"\13TCP4CSUM_Rx"	\
    625 	"\14TCP4CSUM_Tx"	\
    626 	"\15UDP4CSUM_Rx"	\
    627 	"\16UDP4CSUM_Tx"	\
    628 	"\17TCP6CSUM_Rx"	\
    629 	"\20TCP6CSUM_Tx"	\
    630 	"\21UDP6CSUM_Rx"	\
    631 	"\22UDP6CSUM_Tx"	\
    632 	"\23TSO6"		\
    633 	"\24LRO"		\
    634 
    635 #define	IF_AFDATA_LOCK_INIT(ifp)	\
    636 	do {(ifp)->if_afdata_lock = rw_obj_alloc();} while (0)
    637 
    638 #define	IF_AFDATA_LOCK_DESTROY(ifp)	rw_obj_free((ifp)->if_afdata_lock)
    639 
    640 #define	IF_AFDATA_WLOCK(ifp)	rw_enter((ifp)->if_afdata_lock, RW_WRITER)
    641 #define	IF_AFDATA_RLOCK(ifp)	rw_enter((ifp)->if_afdata_lock, RW_READER)
    642 #define	IF_AFDATA_WUNLOCK(ifp)	rw_exit((ifp)->if_afdata_lock)
    643 #define	IF_AFDATA_RUNLOCK(ifp)	rw_exit((ifp)->if_afdata_lock)
    644 #define	IF_AFDATA_LOCK(ifp)	IF_AFDATA_WLOCK(ifp)
    645 #define	IF_AFDATA_UNLOCK(ifp)	IF_AFDATA_WUNLOCK(ifp)
    646 #define	IF_AFDATA_TRYLOCK(ifp)	rw_tryenter((ifp)->if_afdata_lock, RW_WRITER)
    647 
    648 #define	IF_AFDATA_LOCK_ASSERT(ifp)	\
    649 	KASSERT(rw_lock_held((ifp)->if_afdata_lock))
    650 #define	IF_AFDATA_RLOCK_ASSERT(ifp)	\
    651 	KASSERT(rw_read_held((ifp)->if_afdata_lock))
    652 #define	IF_AFDATA_WLOCK_ASSERT(ifp)	\
    653 	KASSERT(rw_write_held((ifp)->if_afdata_lock))
    654 
    655 /*
    656  * Output queues (ifp->if_snd) and internetwork datagram level (pup level 1)
    657  * input routines have queues of messages stored on ifqueue structures
    658  * (defined above).  Entries are added to and deleted from these structures
    659  * by these macros, which should be called with ipl raised to splnet().
    660  */
    661 #define	IF_QFULL(ifq)		((ifq)->ifq_len >= (ifq)->ifq_maxlen)
    662 #define	IF_DROP(ifq)		((ifq)->ifq_drops++)
    663 #define	IF_ENQUEUE(ifq, m) do { \
    664 	(m)->m_nextpkt = 0; \
    665 	if ((ifq)->ifq_tail == 0) \
    666 		(ifq)->ifq_head = m; \
    667 	else \
    668 		(ifq)->ifq_tail->m_nextpkt = m; \
    669 	(ifq)->ifq_tail = m; \
    670 	(ifq)->ifq_len++; \
    671 } while (/*CONSTCOND*/0)
    672 #define	IF_PREPEND(ifq, m) do { \
    673 	(m)->m_nextpkt = (ifq)->ifq_head; \
    674 	if ((ifq)->ifq_tail == 0) \
    675 		(ifq)->ifq_tail = (m); \
    676 	(ifq)->ifq_head = (m); \
    677 	(ifq)->ifq_len++; \
    678 } while (/*CONSTCOND*/0)
    679 #define	IF_DEQUEUE(ifq, m) do { \
    680 	(m) = (ifq)->ifq_head; \
    681 	if (m) { \
    682 		if (((ifq)->ifq_head = (m)->m_nextpkt) == 0) \
    683 			(ifq)->ifq_tail = 0; \
    684 		(m)->m_nextpkt = 0; \
    685 		(ifq)->ifq_len--; \
    686 	} \
    687 } while (/*CONSTCOND*/0)
    688 #define	IF_POLL(ifq, m)		((m) = (ifq)->ifq_head)
    689 #define	IF_PURGE(ifq)							\
    690 do {									\
    691 	struct mbuf *__m0;						\
    692 									\
    693 	for (;;) {							\
    694 		IF_DEQUEUE((ifq), __m0);				\
    695 		if (__m0 == NULL)					\
    696 			break;						\
    697 		else							\
    698 			m_freem(__m0);					\
    699 	}								\
    700 } while (/*CONSTCOND*/ 0)
    701 #define	IF_IS_EMPTY(ifq)	((ifq)->ifq_len == 0)
    702 
    703 #ifndef IFQ_MAXLEN
    704 #define	IFQ_MAXLEN	256
    705 #endif
    706 #define	IFNET_SLOWHZ	1		/* granularity is 1 second */
    707 
    708 /*
    709  * Structure defining statistics and other data kept regarding an address
    710  * on a network interface.
    711  */
    712 struct ifaddr_data {
    713 	int64_t	ifad_inbytes;
    714 	int64_t	ifad_outbytes;
    715 };
    716 
    717 /*
    718  * The ifaddr structure contains information about one address
    719  * of an interface.  They are maintained by the different address families,
    720  * are allocated and attached when an address is set, and are linked
    721  * together so all addresses for an interface can be located.
    722  */
    723 struct ifaddr {
    724 	struct	sockaddr *ifa_addr;	/* address of interface */
    725 	struct	sockaddr *ifa_dstaddr;	/* other end of p-to-p link */
    726 #define	ifa_broadaddr	ifa_dstaddr	/* broadcast address interface */
    727 	struct	sockaddr *ifa_netmask;	/* used to determine subnet */
    728 	struct	ifnet *ifa_ifp;		/* back-pointer to interface */
    729 	TAILQ_ENTRY(ifaddr) ifa_list;	/* list of addresses for interface */
    730 	struct	ifaddr_data	ifa_data;	/* statistics on the address */
    731 	void	(*ifa_rtrequest)	/* check or clean routes (+ or -)'d */
    732 		        (int, struct rtentry *, const struct rt_addrinfo *);
    733 	u_int	ifa_flags;		/* mostly rt_flags for cloning */
    734 	int	ifa_refcnt;		/* count of references */
    735 	int	ifa_metric;		/* cost of going out this interface */
    736 	struct ifaddr	*(*ifa_getifa)(struct ifaddr *,
    737 			               const struct sockaddr *);
    738 	uint32_t	*ifa_seqno;
    739 	int16_t	ifa_preference;	/* preference level for this address */
    740 #ifdef _KERNEL
    741 	struct pslist_entry     ifa_pslist_entry;
    742 	struct psref_target	ifa_psref;
    743 #endif
    744 };
    745 #define	IFA_ROUTE	RTF_UP	/* (0x01) route installed */
    746 #define	IFA_DESTROYING	0x2
    747 
    748 /*
    749  * Message format for use in obtaining information about interfaces from
    750  * sysctl and the routing socket.  We need to force 64-bit alignment if we
    751  * aren't using compatiblity definitons.
    752  */
    753 #if !defined(_KERNEL) || !defined(COMPAT_RTSOCK)
    754 #define	__align64	__aligned(sizeof(uint64_t))
    755 #else
    756 #define	__align64
    757 #endif
    758 struct if_msghdr {
    759 	u_short	ifm_msglen __align64;
    760 				/* to skip over non-understood messages */
    761 	u_char	ifm_version;	/* future binary compatibility */
    762 	u_char	ifm_type;	/* message type */
    763 	int	ifm_addrs;	/* like rtm_addrs */
    764 	int	ifm_flags;	/* value of if_flags */
    765 	u_short	ifm_index;	/* index for associated ifp */
    766 	struct	if_data ifm_data __align64;
    767 				/* statistics and other data about if */
    768 };
    769 
    770 /*
    771  * Message format for use in obtaining information about interface addresses
    772  * from sysctl and the routing socket.
    773  */
    774 struct ifa_msghdr {
    775 	u_short	ifam_msglen __align64;
    776 				/* to skip over non-understood messages */
    777 	u_char	ifam_version;	/* future binary compatibility */
    778 	u_char	ifam_type;	/* message type */
    779 	u_short	ifam_index;	/* index for associated ifp */
    780 	int	ifam_flags;	/* value of ifa_flags */
    781 	int	ifam_addrs;	/* like rtm_addrs */
    782 	pid_t	ifam_pid;	/* identify sender */
    783 	int	ifam_addrflags;	/* family specific address flags */
    784 	int	ifam_metric;	/* value of ifa_metric */
    785 };
    786 
    787 /*
    788  * Message format announcing the arrival or departure of a network interface.
    789  */
    790 struct if_announcemsghdr {
    791 	u_short	ifan_msglen __align64;
    792 				/* to skip over non-understood messages */
    793 	u_char	ifan_version;	/* future binary compatibility */
    794 	u_char	ifan_type;	/* message type */
    795 	u_short	ifan_index;	/* index for associated ifp */
    796 	char	ifan_name[IFNAMSIZ]; /* if name, e.g. "en0" */
    797 	u_short	ifan_what;	/* what type of announcement */
    798 };
    799 
    800 #define	IFAN_ARRIVAL	0	/* interface arrival */
    801 #define	IFAN_DEPARTURE	1	/* interface departure */
    802 
    803 #undef __align64
    804 
    805 /*
    806  * Interface request structure used for socket
    807  * ioctl's.  All interface ioctl's must have parameter
    808  * definitions which begin with ifr_name.  The
    809  * remainder may be interface specific.
    810  */
    811 struct	ifreq {
    812 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
    813 	union {
    814 		struct	sockaddr ifru_addr;
    815 		struct	sockaddr ifru_dstaddr;
    816 		struct	sockaddr ifru_broadaddr;
    817 		struct	sockaddr_storage ifru_space;
    818 		short	ifru_flags;
    819 		int	ifru_addrflags;
    820 		int	ifru_metric;
    821 		int	ifru_mtu;
    822 		int	ifru_dlt;
    823 		u_int	ifru_value;
    824 		void *	ifru_data;
    825 		struct {
    826 			uint32_t	b_buflen;
    827 			void		*b_buf;
    828 		} ifru_b;
    829 	} ifr_ifru;
    830 #define	ifr_addr	ifr_ifru.ifru_addr	/* address */
    831 #define	ifr_dstaddr	ifr_ifru.ifru_dstaddr	/* other end of p-to-p link */
    832 #define	ifr_broadaddr	ifr_ifru.ifru_broadaddr	/* broadcast address */
    833 #define	ifr_space	ifr_ifru.ifru_space	/* sockaddr_storage */
    834 #define	ifr_flags	ifr_ifru.ifru_flags	/* flags */
    835 #define	ifr_addrflags	ifr_ifru.ifru_addrflags	/* addr flags */
    836 #define	ifr_metric	ifr_ifru.ifru_metric	/* metric */
    837 #define	ifr_mtu		ifr_ifru.ifru_mtu	/* mtu */
    838 #define	ifr_dlt		ifr_ifru.ifru_dlt	/* data link type (DLT_*) */
    839 #define	ifr_value	ifr_ifru.ifru_value	/* generic value */
    840 #define	ifr_media	ifr_ifru.ifru_metric	/* media options (overload) */
    841 #define	ifr_data	ifr_ifru.ifru_data	/* for use by interface
    842 						 * XXX deprecated
    843 						 */
    844 #define	ifr_buf		ifr_ifru.ifru_b.b_buf	/* new interface ioctls */
    845 #define	ifr_buflen	ifr_ifru.ifru_b.b_buflen
    846 #define	ifr_index	ifr_ifru.ifru_value	/* interface index, BSD */
    847 #define	ifr_ifindex	ifr_index		/* interface index, linux */
    848 };
    849 
    850 #ifdef _KERNEL
    851 #define	ifreq_setdstaddr	ifreq_setaddr
    852 #define	ifreq_setbroadaddr	ifreq_setaddr
    853 #define	ifreq_getdstaddr	ifreq_getaddr
    854 #define	ifreq_getbroadaddr	ifreq_getaddr
    855 
    856 static __inline const struct sockaddr *
    857 /*ARGSUSED*/
    858 ifreq_getaddr(u_long cmd, const struct ifreq *ifr)
    859 {
    860 	return &ifr->ifr_addr;
    861 }
    862 #endif /* _KERNEL */
    863 
    864 struct ifcapreq {
    865 	char		ifcr_name[IFNAMSIZ];	/* if name, e.g. "en0" */
    866 	uint64_t	ifcr_capabilities;	/* supported capabiliites */
    867 	uint64_t	ifcr_capenable;		/* capabilities enabled */
    868 };
    869 
    870 struct ifaliasreq {
    871 	char	ifra_name[IFNAMSIZ];		/* if name, e.g. "en0" */
    872 	struct	sockaddr ifra_addr;
    873 	struct	sockaddr ifra_dstaddr;
    874 #define	ifra_broadaddr	ifra_dstaddr
    875 	struct	sockaddr ifra_mask;
    876 };
    877 
    878 struct ifdatareq {
    879 	char	ifdr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
    880 	struct	if_data ifdr_data;
    881 };
    882 
    883 struct ifmediareq {
    884 	char	ifm_name[IFNAMSIZ];	/* if name, e.g. "en0" */
    885 	int	ifm_current;		/* IFMWD: current media options */
    886 	int	ifm_mask;		/* IFMWD: don't care mask */
    887 	int	ifm_status;		/* media status */
    888 	int	ifm_active;		/* IFMWD: active options */
    889 	int	ifm_count;		/* # entries in ifm_ulist
    890 					   array */
    891 	int	*ifm_ulist;		/* array of ifmedia word */
    892 };
    893 
    894 
    895 struct  ifdrv {
    896 	char		ifd_name[IFNAMSIZ];	/* if name, e.g. "en0" */
    897 	unsigned long	ifd_cmd;
    898 	size_t		ifd_len;
    899 	void		*ifd_data;
    900 };
    901 #define IFLINKSTR_QUERYLEN	0x01
    902 #define IFLINKSTR_UNSET		0x02
    903 
    904 /*
    905  * Structure used in SIOCGIFCONF request.
    906  * Used to retrieve interface configuration
    907  * for machine (useful for programs which
    908  * must know all networks accessible).
    909  */
    910 struct	ifconf {
    911 	int	ifc_len;		/* size of associated buffer */
    912 	union {
    913 		void *	ifcu_buf;
    914 		struct	ifreq *ifcu_req;
    915 	} ifc_ifcu;
    916 #define	ifc_buf	ifc_ifcu.ifcu_buf	/* buffer address */
    917 #define	ifc_req	ifc_ifcu.ifcu_req	/* array of structures returned */
    918 };
    919 
    920 /*
    921  * Structure for SIOC[AGD]LIFADDR
    922  */
    923 struct if_laddrreq {
    924 	char iflr_name[IFNAMSIZ];
    925 	unsigned int flags;
    926 #define IFLR_PREFIX	0x8000	/* in: prefix given  out: kernel fills id */
    927 #define IFLR_ACTIVE	0x4000	/* in/out: link-layer address activation */
    928 #define IFLR_FACTORY	0x2000	/* in/out: factory link-layer address */
    929 	unsigned int prefixlen;		/* in/out */
    930 	struct sockaddr_storage addr;	/* in/out */
    931 	struct sockaddr_storage dstaddr; /* out */
    932 };
    933 
    934 /*
    935  * Structure for SIOC[SG]IFADDRPREF
    936  */
    937 struct if_addrprefreq {
    938 	char			ifap_name[IFNAMSIZ];
    939 	int16_t			ifap_preference;	/* in/out */
    940 	struct sockaddr_storage	ifap_addr;		/* in/out */
    941 };
    942 
    943 #include <net/if_arp.h>
    944 
    945 #endif /* _NETBSD_SOURCE */
    946 
    947 #ifdef _KERNEL
    948 #ifdef ALTQ
    949 #define IFQ_ENQUEUE(ifq, m, err)					\
    950 do {									\
    951 	mutex_enter((ifq)->ifq_lock);					\
    952 	if (ALTQ_IS_ENABLED(ifq))					\
    953 		ALTQ_ENQUEUE((ifq), (m), (err));			\
    954 	else {								\
    955 		if (IF_QFULL(ifq)) {					\
    956 			m_freem(m);					\
    957 			(err) = ENOBUFS;				\
    958 		} else {						\
    959 			IF_ENQUEUE((ifq), (m));				\
    960 			(err) = 0;					\
    961 		}							\
    962 	}								\
    963 	if ((err))							\
    964 		(ifq)->ifq_drops++;					\
    965 	mutex_exit((ifq)->ifq_lock);					\
    966 } while (/*CONSTCOND*/ 0)
    967 
    968 #define IFQ_DEQUEUE(ifq, m)						\
    969 do {									\
    970 	mutex_enter((ifq)->ifq_lock);					\
    971 	if (TBR_IS_ENABLED(ifq))					\
    972 		(m) = tbr_dequeue((ifq), ALTDQ_REMOVE);			\
    973 	else if (ALTQ_IS_ENABLED(ifq))					\
    974 		ALTQ_DEQUEUE((ifq), (m));				\
    975 	else								\
    976 		IF_DEQUEUE((ifq), (m));					\
    977 	mutex_exit((ifq)->ifq_lock);					\
    978 } while (/*CONSTCOND*/ 0)
    979 
    980 #define	IFQ_POLL(ifq, m)						\
    981 do {									\
    982 	mutex_enter((ifq)->ifq_lock);					\
    983 	if (TBR_IS_ENABLED(ifq))					\
    984 		(m) = tbr_dequeue((ifq), ALTDQ_POLL);			\
    985 	else if (ALTQ_IS_ENABLED(ifq))					\
    986 		ALTQ_POLL((ifq), (m));					\
    987 	else								\
    988 		IF_POLL((ifq), (m));					\
    989 	mutex_exit((ifq)->ifq_lock);					\
    990 } while (/*CONSTCOND*/ 0)
    991 
    992 #define	IFQ_PURGE(ifq)							\
    993 do {									\
    994 	mutex_enter((ifq)->ifq_lock);					\
    995 	if (ALTQ_IS_ENABLED(ifq))					\
    996 		ALTQ_PURGE(ifq);					\
    997 	else								\
    998 		IF_PURGE(ifq);						\
    999 	mutex_exit((ifq)->ifq_lock);					\
   1000 } while (/*CONSTCOND*/ 0)
   1001 
   1002 #define	IFQ_SET_READY(ifq)						\
   1003 do {									\
   1004 	(ifq)->altq_flags |= ALTQF_READY;				\
   1005 } while (/*CONSTCOND*/ 0)
   1006 
   1007 #define	IFQ_CLASSIFY(ifq, m, af)					\
   1008 do {									\
   1009 	KASSERT(((m)->m_flags & M_PKTHDR) != 0);			\
   1010 	mutex_enter((ifq)->ifq_lock);					\
   1011 	if (ALTQ_IS_ENABLED(ifq)) {					\
   1012 		if (ALTQ_NEEDS_CLASSIFY(ifq))				\
   1013 			(m)->m_pkthdr.pattr_class = (*(ifq)->altq_classify) \
   1014 				((ifq)->altq_clfier, (m), (af));	\
   1015 		(m)->m_pkthdr.pattr_af = (af);				\
   1016 		(m)->m_pkthdr.pattr_hdr = mtod((m), void *);		\
   1017 	}								\
   1018 	mutex_exit((ifq)->ifq_lock);					\
   1019 } while (/*CONSTCOND*/ 0)
   1020 #else /* ! ALTQ */
   1021 #define	IFQ_ENQUEUE(ifq, m, err)					\
   1022 do {									\
   1023 	mutex_enter((ifq)->ifq_lock);					\
   1024 	if (IF_QFULL(ifq)) {						\
   1025 		m_freem(m);						\
   1026 		(err) = ENOBUFS;					\
   1027 	} else {							\
   1028 		IF_ENQUEUE((ifq), (m));					\
   1029 		(err) = 0;						\
   1030 	}								\
   1031 	if (err)							\
   1032 		(ifq)->ifq_drops++;					\
   1033 	mutex_exit((ifq)->ifq_lock);					\
   1034 } while (/*CONSTCOND*/ 0)
   1035 
   1036 #define	IFQ_DEQUEUE(ifq, m)						\
   1037 do {									\
   1038 	mutex_enter((ifq)->ifq_lock);					\
   1039 	IF_DEQUEUE((ifq), (m));						\
   1040 	mutex_exit((ifq)->ifq_lock);					\
   1041 } while (/*CONSTCOND*/ 0)
   1042 
   1043 #define	IFQ_POLL(ifq, m)						\
   1044 do {									\
   1045 	mutex_enter((ifq)->ifq_lock);					\
   1046 	IF_POLL((ifq), (m));						\
   1047 	mutex_exit((ifq)->ifq_lock);					\
   1048 } while (/*CONSTCOND*/ 0)
   1049 
   1050 #define	IFQ_PURGE(ifq)							\
   1051 do {									\
   1052 	mutex_enter((ifq)->ifq_lock);					\
   1053 	IF_PURGE(ifq);							\
   1054 	mutex_exit((ifq)->ifq_lock);					\
   1055 } while (/*CONSTCOND*/ 0)
   1056 
   1057 #define	IFQ_SET_READY(ifq)	/* nothing */
   1058 
   1059 #define	IFQ_CLASSIFY(ifq, m, af) /* nothing */
   1060 
   1061 #endif /* ALTQ */
   1062 
   1063 #define IFQ_LOCK_INIT(ifq)	(ifq)->ifq_lock =			\
   1064 	    mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET)
   1065 #define IFQ_LOCK_DESTROY(ifq)	mutex_obj_free((ifq)->ifq_lock)
   1066 #define IFQ_LOCK(ifq)		mutex_enter((ifq)->ifq_lock)
   1067 #define IFQ_UNLOCK(ifq)		mutex_exit((ifq)->ifq_lock)
   1068 
   1069 #define	IFQ_IS_EMPTY(ifq)		IF_IS_EMPTY(ifq)
   1070 #define	IFQ_INC_LEN(ifq)		((ifq)->ifq_len++)
   1071 #define	IFQ_DEC_LEN(ifq)		(--(ifq)->ifq_len)
   1072 #define	IFQ_INC_DROPS(ifq)		((ifq)->ifq_drops++)
   1073 #define	IFQ_SET_MAXLEN(ifq, len)	((ifq)->ifq_maxlen = (len))
   1074 
   1075 #include <sys/mallocvar.h>
   1076 MALLOC_DECLARE(M_IFADDR);
   1077 MALLOC_DECLARE(M_IFMADDR);
   1078 
   1079 int ifreq_setaddr(u_long, struct ifreq *, const struct sockaddr *);
   1080 
   1081 struct ifnet *if_alloc(u_char);
   1082 void if_free(struct ifnet *);
   1083 void if_initname(struct ifnet *, const char *, int);
   1084 struct ifaddr *if_dl_create(const struct ifnet *, const struct sockaddr_dl **);
   1085 void if_activate_sadl(struct ifnet *, struct ifaddr *,
   1086     const struct sockaddr_dl *);
   1087 void	if_set_sadl(struct ifnet *, const void *, u_char, bool);
   1088 void	if_alloc_sadl(struct ifnet *);
   1089 void	if_free_sadl(struct ifnet *, int);
   1090 int	if_initialize(struct ifnet *);
   1091 void	if_register(struct ifnet *);
   1092 int	if_attach(struct ifnet *); /* Deprecated. Use if_initialize and if_register */
   1093 void	if_attachdomain(void);
   1094 void	if_deactivate(struct ifnet *);
   1095 bool	if_is_deactivated(const struct ifnet *);
   1096 void	if_export_if_data(struct ifnet *, struct if_data *, bool);
   1097 void	if_purgeaddrs(struct ifnet *, int, void (*)(struct ifaddr *));
   1098 void	if_detach(struct ifnet *);
   1099 void	if_down(struct ifnet *);
   1100 void	if_down_locked(struct ifnet *);
   1101 void	if_link_state_change(struct ifnet *, int);
   1102 void	if_up(struct ifnet *);
   1103 void	ifinit(void);
   1104 void	ifinit1(void);
   1105 void	ifinit_post(void);
   1106 int	ifaddrpref_ioctl(struct socket *, u_long, void *, struct ifnet *);
   1107 extern int (*ifioctl)(struct socket *, u_long, void *, struct lwp *);
   1108 int	ifioctl_common(struct ifnet *, u_long, void *);
   1109 int	ifpromisc(struct ifnet *, int);
   1110 int	ifpromisc_locked(struct ifnet *, int);
   1111 int	if_addr_init(ifnet_t *, struct ifaddr *, bool);
   1112 int	if_do_dad(struct ifnet *);
   1113 int	if_mcast_op(ifnet_t *, const unsigned long, const struct sockaddr *);
   1114 int	if_flags_set(struct ifnet *, const u_short);
   1115 int	if_clone_list(int, char *, int *);
   1116 
   1117 struct	ifnet *ifunit(const char *);
   1118 struct	ifnet *if_get(const char *, struct psref *);
   1119 ifnet_t *if_byindex(u_int);
   1120 ifnet_t *_if_byindex(u_int);
   1121 ifnet_t *if_get_byindex(u_int, struct psref *);
   1122 ifnet_t *if_get_bylla(const void *, unsigned char, struct psref *);
   1123 void	if_put(const struct ifnet *, struct psref *);
   1124 void	if_acquire(struct ifnet *, struct psref *);
   1125 #define	if_release	if_put
   1126 
   1127 int if_tunnel_check_nesting(struct ifnet *, struct mbuf *, int);
   1128 percpu_t *if_tunnel_alloc_ro_percpu(void);
   1129 void if_tunnel_free_ro_percpu(percpu_t *);
   1130 void if_tunnel_ro_percpu_rtcache_free(percpu_t *);
   1131 
   1132 struct tunnel_ro {
   1133 	struct route *tr_ro;
   1134 	kmutex_t *tr_lock;
   1135 };
   1136 
   1137 static inline void
   1138 if_tunnel_get_ro(percpu_t *ro_percpu, struct route **ro, kmutex_t **lock)
   1139 {
   1140 	struct tunnel_ro *tro;
   1141 
   1142 	tro = percpu_getref(ro_percpu);
   1143 	*ro = tro->tr_ro;
   1144 	*lock = tro->tr_lock;
   1145 	mutex_enter(*lock);
   1146 }
   1147 
   1148 static inline void
   1149 if_tunnel_put_ro(percpu_t *ro_percpu, kmutex_t *lock)
   1150 {
   1151 
   1152 	mutex_exit(lock);
   1153 	percpu_putref(ro_percpu);
   1154 }
   1155 
   1156 static __inline if_index_t
   1157 if_get_index(const struct ifnet *ifp)
   1158 {
   1159 
   1160 	return ifp != NULL ? ifp->if_index : 0;
   1161 }
   1162 
   1163 bool	if_held(struct ifnet *);
   1164 
   1165 void	if_input(struct ifnet *, struct mbuf *);
   1166 
   1167 struct if_percpuq *
   1168 	if_percpuq_create(struct ifnet *);
   1169 void	if_percpuq_destroy(struct if_percpuq *);
   1170 void
   1171 	if_percpuq_enqueue(struct if_percpuq *, struct mbuf *);
   1172 
   1173 void	if_deferred_start_init(struct ifnet *, void (*)(struct ifnet *));
   1174 void	if_schedule_deferred_start(struct ifnet *);
   1175 
   1176 void ifa_insert(struct ifnet *, struct ifaddr *);
   1177 void ifa_remove(struct ifnet *, struct ifaddr *);
   1178 
   1179 void	ifa_psref_init(struct ifaddr *);
   1180 void	ifa_acquire(struct ifaddr *, struct psref *);
   1181 void	ifa_release(struct ifaddr *, struct psref *);
   1182 bool	ifa_held(struct ifaddr *);
   1183 bool	ifa_is_destroying(struct ifaddr *);
   1184 
   1185 void	ifaref(struct ifaddr *);
   1186 void	ifafree(struct ifaddr *);
   1187 
   1188 struct	ifaddr *ifa_ifwithaddr(const struct sockaddr *);
   1189 struct	ifaddr *ifa_ifwithaddr_psref(const struct sockaddr *, struct psref *);
   1190 struct	ifaddr *ifa_ifwithaf(int);
   1191 struct	ifaddr *ifa_ifwithdstaddr(const struct sockaddr *);
   1192 struct	ifaddr *ifa_ifwithdstaddr_psref(const struct sockaddr *,
   1193 	    struct psref *);
   1194 struct	ifaddr *ifa_ifwithnet(const struct sockaddr *);
   1195 struct	ifaddr *ifa_ifwithnet_psref(const struct sockaddr *, struct psref *);
   1196 struct	ifaddr *ifa_ifwithladdr(const struct sockaddr *);
   1197 struct	ifaddr *ifa_ifwithladdr_psref(const struct sockaddr *, struct psref *);
   1198 struct	ifaddr *ifaof_ifpforaddr(const struct sockaddr *, struct ifnet *);
   1199 struct	ifaddr *ifaof_ifpforaddr_psref(const struct sockaddr *, struct ifnet *,
   1200 	    struct psref *);
   1201 void	link_rtrequest(int, struct rtentry *, const struct rt_addrinfo *);
   1202 void	p2p_rtrequest(int, struct rtentry *, const struct rt_addrinfo *);
   1203 
   1204 void	if_clone_attach(struct if_clone *);
   1205 void	if_clone_detach(struct if_clone *);
   1206 
   1207 int	if_transmit_lock(struct ifnet *, struct mbuf *);
   1208 
   1209 int	ifq_enqueue(struct ifnet *, struct mbuf *);
   1210 int	ifq_enqueue2(struct ifnet *, struct ifqueue *, struct mbuf *);
   1211 
   1212 int	loioctl(struct ifnet *, u_long, void *);
   1213 void	loopattach(int);
   1214 void	loopinit(void);
   1215 int	looutput(struct ifnet *,
   1216 	   struct mbuf *, const struct sockaddr *, const struct rtentry *);
   1217 
   1218 /*
   1219  * These are exported because they're an easy way to tell if
   1220  * an interface is going away without having to burn a flag.
   1221  */
   1222 int	if_nulloutput(struct ifnet *, struct mbuf *,
   1223 	    const struct sockaddr *, const struct rtentry *);
   1224 void	if_nullinput(struct ifnet *, struct mbuf *);
   1225 void	if_nullstart(struct ifnet *);
   1226 int	if_nulltransmit(struct ifnet *, struct mbuf *);
   1227 int	if_nullioctl(struct ifnet *, u_long, void *);
   1228 int	if_nullinit(struct ifnet *);
   1229 void	if_nullstop(struct ifnet *, int);
   1230 void	if_nullslowtimo(struct ifnet *);
   1231 #define	if_nullwatchdog	if_nullslowtimo
   1232 void	if_nulldrain(struct ifnet *);
   1233 #else
   1234 struct if_nameindex {
   1235 	unsigned int	if_index;	/* 1, 2, ... */
   1236 	char		*if_name;	/* null terminated name: "le0", ... */
   1237 };
   1238 
   1239 #include <sys/cdefs.h>
   1240 __BEGIN_DECLS
   1241 unsigned int if_nametoindex(const char *);
   1242 char *	if_indextoname(unsigned int, char *);
   1243 struct	if_nameindex * if_nameindex(void);
   1244 void	if_freenameindex(struct if_nameindex *);
   1245 __END_DECLS
   1246 #endif /* _KERNEL */ /* XXX really ALTQ? */
   1247 
   1248 #ifdef _KERNEL
   1249 
   1250 #define	IFADDR_FIRST(__ifp)		TAILQ_FIRST(&(__ifp)->if_addrlist)
   1251 #define	IFADDR_NEXT(__ifa)		TAILQ_NEXT((__ifa), ifa_list)
   1252 #define	IFADDR_FOREACH(__ifa, __ifp)	TAILQ_FOREACH(__ifa, \
   1253 					    &(__ifp)->if_addrlist, ifa_list)
   1254 #define	IFADDR_FOREACH_SAFE(__ifa, __ifp, __nifa) \
   1255 					    TAILQ_FOREACH_SAFE(__ifa, \
   1256 					    &(__ifp)->if_addrlist, ifa_list, __nifa)
   1257 #define	IFADDR_EMPTY(__ifp)		TAILQ_EMPTY(&(__ifp)->if_addrlist)
   1258 
   1259 #define IFADDR_ENTRY_INIT(__ifa)					\
   1260 	PSLIST_ENTRY_INIT((__ifa), ifa_pslist_entry)
   1261 #define IFADDR_ENTRY_DESTROY(__ifa)					\
   1262 	PSLIST_ENTRY_DESTROY((__ifa), ifa_pslist_entry)
   1263 #define IFADDR_READER_EMPTY(__ifp)					\
   1264 	(PSLIST_READER_FIRST(&(__ifp)->if_addr_pslist, struct ifaddr,	\
   1265 	                     ifa_pslist_entry) == NULL)
   1266 #define IFADDR_READER_FIRST(__ifp)					\
   1267 	PSLIST_READER_FIRST(&(__ifp)->if_addr_pslist, struct ifaddr,	\
   1268 	                    ifa_pslist_entry)
   1269 #define IFADDR_READER_NEXT(__ifa)					\
   1270 	PSLIST_READER_NEXT((__ifa), struct ifaddr, ifa_pslist_entry)
   1271 #define IFADDR_READER_FOREACH(__ifa, __ifp)				\
   1272 	PSLIST_READER_FOREACH((__ifa), &(__ifp)->if_addr_pslist, struct ifaddr,\
   1273 	                      ifa_pslist_entry)
   1274 #define IFADDR_WRITER_INSERT_HEAD(__ifp, __ifa)				\
   1275 	PSLIST_WRITER_INSERT_HEAD(&(__ifp)->if_addr_pslist, (__ifa),	\
   1276 	                          ifa_pslist_entry)
   1277 #define IFADDR_WRITER_REMOVE(__ifa)					\
   1278 	PSLIST_WRITER_REMOVE((__ifa), ifa_pslist_entry)
   1279 #define IFADDR_WRITER_FOREACH(__ifa, __ifp)				\
   1280 	PSLIST_WRITER_FOREACH((__ifa), &(__ifp)->if_addr_pslist, struct ifaddr,\
   1281 	                      ifa_pslist_entry)
   1282 #define IFADDR_WRITER_NEXT(__ifp)					\
   1283 	PSLIST_WRITER_NEXT((__ifp), struct ifaddr, ifa_pslist_entry)
   1284 #define IFADDR_WRITER_INSERT_AFTER(__ifp, __new)			\
   1285 	PSLIST_WRITER_INSERT_AFTER((__ifp), (__new), ifa_pslist_entry)
   1286 #define IFADDR_WRITER_EMPTY(__ifp)					\
   1287 	(PSLIST_WRITER_FIRST(&(__ifp)->if_addr_pslist, struct ifaddr,	\
   1288 	                     ifa_pslist_entry) == NULL)
   1289 #define IFADDR_WRITER_INSERT_TAIL(__ifp, __new)				\
   1290 	do {								\
   1291 		if (IFADDR_WRITER_EMPTY(__ifp)) {			\
   1292 			IFADDR_WRITER_INSERT_HEAD((__ifp), (__new));	\
   1293 		} else {						\
   1294 			struct ifaddr *__ifa;				\
   1295 			IFADDR_WRITER_FOREACH(__ifa, (__ifp)) {		\
   1296 				if (IFADDR_WRITER_NEXT(__ifa) == NULL) {\
   1297 					IFADDR_WRITER_INSERT_AFTER(__ifa,\
   1298 					    (__new));			\
   1299 					break;				\
   1300 				}					\
   1301 			}						\
   1302 		}							\
   1303 	} while (0)
   1304 
   1305 #define	IFNET_GLOBAL_LOCK()			mutex_enter(&ifnet_mtx)
   1306 #define	IFNET_GLOBAL_UNLOCK()			mutex_exit(&ifnet_mtx)
   1307 #define	IFNET_GLOBAL_LOCKED()			mutex_owned(&ifnet_mtx)
   1308 
   1309 #define IFNET_READER_EMPTY() \
   1310 	(PSLIST_READER_FIRST(&ifnet_pslist, struct ifnet, if_pslist_entry) == NULL)
   1311 #define IFNET_READER_FIRST() \
   1312 	PSLIST_READER_FIRST(&ifnet_pslist, struct ifnet, if_pslist_entry)
   1313 #define IFNET_READER_NEXT(__ifp) \
   1314 	PSLIST_READER_NEXT((__ifp), struct ifnet, if_pslist_entry)
   1315 #define IFNET_READER_FOREACH(__ifp) \
   1316 	PSLIST_READER_FOREACH((__ifp), &ifnet_pslist, struct ifnet, \
   1317 	                      if_pslist_entry)
   1318 #define IFNET_WRITER_INSERT_HEAD(__ifp) \
   1319 	PSLIST_WRITER_INSERT_HEAD(&ifnet_pslist, (__ifp), if_pslist_entry)
   1320 #define IFNET_WRITER_REMOVE(__ifp) \
   1321 	PSLIST_WRITER_REMOVE((__ifp), if_pslist_entry)
   1322 #define IFNET_WRITER_FOREACH(__ifp) \
   1323 	PSLIST_WRITER_FOREACH((__ifp), &ifnet_pslist, struct ifnet, \
   1324 	                      if_pslist_entry)
   1325 #define IFNET_WRITER_NEXT(__ifp) \
   1326 	PSLIST_WRITER_NEXT((__ifp), struct ifnet, if_pslist_entry)
   1327 #define IFNET_WRITER_INSERT_AFTER(__ifp, __new) \
   1328 	PSLIST_WRITER_INSERT_AFTER((__ifp), (__new), if_pslist_entry)
   1329 #define IFNET_WRITER_EMPTY() \
   1330 	(PSLIST_WRITER_FIRST(&ifnet_pslist, struct ifnet, if_pslist_entry) == NULL)
   1331 #define IFNET_WRITER_INSERT_TAIL(__new)					\
   1332 	do {								\
   1333 		if (IFNET_WRITER_EMPTY()) {				\
   1334 			IFNET_WRITER_INSERT_HEAD(__new);		\
   1335 		} else {						\
   1336 			struct ifnet *__ifp;				\
   1337 			IFNET_WRITER_FOREACH(__ifp) {			\
   1338 				if (IFNET_WRITER_NEXT(__ifp) == NULL) {	\
   1339 					IFNET_WRITER_INSERT_AFTER(__ifp,\
   1340 					    (__new));			\
   1341 					break;				\
   1342 				}					\
   1343 			}						\
   1344 		}							\
   1345 	} while (0)
   1346 
   1347 #define IFNET_LOCK(ifp)		mutex_enter((ifp)->if_ioctl_lock)
   1348 #define IFNET_UNLOCK(ifp)	mutex_exit((ifp)->if_ioctl_lock)
   1349 #define IFNET_LOCKED(ifp)	mutex_owned((ifp)->if_ioctl_lock)
   1350 
   1351 #define IFNET_ASSERT_UNLOCKED(ifp)	\
   1352 	KDASSERT(mutex_ownable((ifp)->if_ioctl_lock))
   1353 
   1354 extern struct pslist_head ifnet_pslist;
   1355 extern kmutex_t ifnet_mtx;
   1356 
   1357 extern struct ifnet *lo0ifp;
   1358 
   1359 /*
   1360  * ifq sysctl support
   1361  */
   1362 int	sysctl_ifq(int *name, u_int namelen, void *oldp,
   1363 		       size_t *oldlenp, void *newp, size_t newlen,
   1364 		       struct ifqueue *ifq);
   1365 /* symbolic names for terminal (per-protocol) CTL_IFQ_ nodes */
   1366 #define IFQCTL_LEN	1
   1367 #define IFQCTL_MAXLEN	2
   1368 #define IFQCTL_PEAK	3
   1369 #define IFQCTL_DROPS	4
   1370 
   1371 /*
   1372  * Hook for if_vlan - needed by if_agr
   1373  */
   1374 MODULE_HOOK(if_vlan_vlan_input_hook, void, (struct ifnet *, struct mbuf *));
   1375 
   1376 #endif /* _KERNEL */
   1377 
   1378 #endif /* !_NET_IF_H_ */
   1379