Home | History | Annotate | Line # | Download | only in net
if.h revision 1.302
      1 /*	$NetBSD: if.h,v 1.302 2022/09/18 16:58:54 martin 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 #include <sys/hook.h>
     89 
     90 #include <net/dlt.h>
     91 #include <net/pfil.h>
     92 #ifdef _KERNEL
     93 #include <net/pktqueue.h>
     94 #include <sys/pslist.h>
     95 #include <sys/pserialize.h>
     96 #include <sys/psref.h>
     97 #include <sys/module_hook.h>
     98 #endif
     99 
    100 /*
    101  * Always include ALTQ glue here -- we use the ALTQ interface queue
    102  * structure even when ALTQ is not configured into the kernel so that
    103  * the size of struct ifnet does not changed based on the option.  The
    104  * ALTQ queue structure is API-compatible with the legacy ifqueue.
    105  */
    106 #include <altq/if_altq.h>
    107 
    108 /*
    109  * Structures defining a network interface, providing a packet
    110  * transport mechanism (ala level 0 of the PUP protocols).
    111  *
    112  * Each interface accepts output datagrams of a specified maximum
    113  * length, and provides higher level routines with input datagrams
    114  * received from its medium.
    115  *
    116  * Output occurs when the routine if_output is called, with four parameters:
    117  *	(*ifp->if_output)(ifp, m, dst, rt)
    118  * Here m is the mbuf chain to be sent and dst is the destination address.
    119  * The output routine encapsulates the supplied datagram if necessary,
    120  * and then transmits it on its medium.
    121  *
    122  * On input, each interface unwraps the data received by it, and either
    123  * places it on the input queue of a internetwork datagram routine
    124  * and posts the associated software interrupt, or passes the datagram to a raw
    125  * packet input routine.
    126  *
    127  * Routines exist for locating interfaces by their addresses
    128  * or for locating a interface on a certain network, as well as more general
    129  * routing and gateway routines maintaining information used to locate
    130  * interfaces.  These routines live in the files if.c and route.c
    131  */
    132 #include <sys/time.h>
    133 
    134 #if defined(_KERNEL_OPT)
    135 #include "opt_compat_netbsd.h"
    136 #include "opt_gateway.h"
    137 #endif
    138 
    139 struct mbuf;
    140 struct proc;
    141 struct rtentry;
    142 struct socket;
    143 struct ether_header;
    144 struct ifaddr;
    145 struct ifnet;
    146 struct rt_addrinfo;
    147 
    148 #define	IFNAMSIZ	IF_NAMESIZE
    149 
    150 /*
    151  * Structure describing a `cloning' interface.
    152  */
    153 struct if_clone {
    154 	LIST_ENTRY(if_clone) ifc_list;	/* on list of cloners */
    155 	const char *ifc_name;		/* name of device, e.g. `gif' */
    156 	size_t ifc_namelen;		/* length of name */
    157 
    158 	int	(*ifc_create)(struct if_clone *, int);
    159 	int	(*ifc_destroy)(struct ifnet *);
    160 };
    161 
    162 #define	IF_CLONE_INITIALIZER(name, create, destroy)			\
    163 	{ { NULL, NULL }, name, sizeof(name) - 1, create, destroy }
    164 
    165 /*
    166  * Structure used to query names of interface cloners.
    167  */
    168 struct if_clonereq {
    169 	int	ifcr_total;		/* total cloners (out) */
    170 	int	ifcr_count;		/* room for this many in user buffer */
    171 	char	*ifcr_buffer;		/* buffer for cloner names */
    172 };
    173 
    174 /*
    175  * Structure defining statistics and other data kept regarding a network
    176  * interface.
    177  *
    178  * Only used for exporting data from the interface.
    179  */
    180 struct if_data {
    181 	/* generic interface information */
    182 	u_char	ifi_type;		/* ethernet, tokenring, etc. */
    183 	u_char	ifi_addrlen;		/* media address length */
    184 	u_char	ifi_hdrlen;		/* media header length */
    185 	int	ifi_link_state;		/* current link state */
    186 	uint64_t ifi_mtu;		/* maximum transmission unit */
    187 	uint64_t ifi_metric;		/* routing metric (external only) */
    188 	uint64_t ifi_baudrate;		/* linespeed */
    189 	/* volatile statistics */
    190 	uint64_t ifi_ipackets;		/* packets received on interface */
    191 	uint64_t ifi_ierrors;		/* input errors on interface */
    192 	uint64_t ifi_opackets;		/* packets sent on interface */
    193 	uint64_t ifi_oerrors;		/* output errors on interface */
    194 	uint64_t ifi_collisions;	/* collisions on csma interfaces */
    195 	uint64_t ifi_ibytes;		/* total number of octets received */
    196 	uint64_t ifi_obytes;		/* total number of octets sent */
    197 	uint64_t ifi_imcasts;		/* packets received via multicast */
    198 	uint64_t ifi_omcasts;		/* packets sent via multicast */
    199 	uint64_t ifi_iqdrops;		/* dropped on input, this interface */
    200 	uint64_t ifi_noproto;		/* destined for unsupported protocol */
    201 	struct	timespec ifi_lastchange;/* last operational state change */
    202 };
    203 
    204 /*
    205  * Values for if_link_state.
    206  */
    207 #define	LINK_STATE_UNKNOWN	0	/* link invalid/unknown */
    208 #define	LINK_STATE_DOWN		1	/* link is down */
    209 #define	LINK_STATE_UP		2	/* link is up */
    210 
    211 /*
    212  * Status bit descriptions for the various interface types.
    213  */
    214 struct if_status_description {
    215 	unsigned char	ifs_type;
    216 	unsigned char	ifs_state;
    217 	const char	*ifs_string;
    218 };
    219 
    220 #define LINK_STATE_DESC_MATCH(_ifs, _t, _s)				\
    221 	(((_ifs)->ifs_type == (_t) || (_ifs)->ifs_type == 0) &&		\
    222 	    (_ifs)->ifs_state == (_s))
    223 
    224 #define LINK_STATE_DESCRIPTIONS {					\
    225 	{ IFT_ETHER, LINK_STATE_DOWN, "no carrier" },			\
    226 	{ IFT_IEEE80211, LINK_STATE_DOWN, "no network" },		\
    227 	{ IFT_PPP, LINK_STATE_DOWN, "no carrier" },			\
    228 	{ IFT_CARP, LINK_STATE_DOWN, "backup" },			\
    229 	{ IFT_CARP, LINK_STATE_UP, "master" },				\
    230 	{ 0, LINK_STATE_UP, "active" },					\
    231 	{ 0, LINK_STATE_UNKNOWN, "unknown" },				\
    232 	{ 0, LINK_STATE_DOWN, "down" },					\
    233 	{ 0, 0, NULL }							\
    234 }
    235 
    236 /*
    237  * Structure defining a queue for a network interface.
    238  */
    239 struct ifqueue {
    240 	struct		mbuf *ifq_head;
    241 	struct		mbuf *ifq_tail;
    242 	int		ifq_len;
    243 	int		ifq_maxlen;
    244 	int		ifq_drops;
    245 	kmutex_t	*ifq_lock;
    246 };
    247 
    248 #ifdef _KERNEL
    249 #include <sys/percpu.h>
    250 #include <sys/callout.h>
    251 #include <sys/rwlock.h>
    252 #include <sys/workqueue.h>
    253 
    254 #endif /* _KERNEL */
    255 
    256 /*
    257  * Structure defining a queue for a network interface.
    258  *
    259  * (Would like to call this struct ``if'', but C isn't PL/1.)
    260  */
    261 TAILQ_HEAD(ifnet_head, ifnet);		/* the actual queue head */
    262 
    263 struct bridge_softc;
    264 struct bridge_iflist;
    265 struct callout;
    266 struct krwlock;
    267 struct if_percpuq;
    268 struct if_deferred_start;
    269 struct in6_multi;
    270 
    271 typedef unsigned short if_index_t;
    272 
    273 /*
    274  * Interface.  Field markings and the corresponding locks:
    275  *
    276  * i:	IFNET_LOCK (a.k.a., if_ioctl_lock)
    277  * q:	ifq_lock (struct ifaltq)
    278  * a:	if_afdata_lock
    279  * 6:	in6_multilock (global lock)
    280  * ::	unlocked, stable
    281  * ?:	unknown, maybe unsafe
    282  *
    283  * Lock order: IFNET_LOCK => in6_multilock => if_afdata_lock => ifq_lock
    284  *   Note that currently if_afdata_lock and ifq_lock aren't held
    285  *   at the same time, but define the order anyway.
    286  *
    287  * Lock order of IFNET_LOCK with other locks:
    288  *     softnet_lock => solock => IFNET_LOCK => ND6_LOCK, in_multilock
    289  */
    290 typedef struct ifnet {
    291 	void		*if_softc;	/* :: lower-level data for this if */
    292 	/* DEPRECATED. Keep it to avoid breaking kvm(3) users */
    293 	TAILQ_ENTRY(ifnet)
    294 			if_list;	/* i: all struct ifnets are chained */
    295 	TAILQ_HEAD(, ifaddr)
    296 			if_addrlist;	/* i: linked list of addresses per if */
    297 	char		if_xname[IFNAMSIZ];
    298 					/* :: external name (name + unit) */
    299 	int		if_pcount;	/* i: number of promiscuous listeners */
    300 	struct bpf_if	*if_bpf;	/* :: packet filter structure */
    301 	if_index_t	if_index;	/* :: numeric abbreviation for this if */
    302 	short		if_timer;	/* ?: time 'til if_slowtimo called */
    303 	unsigned short	if_flags;	/* i: up/down, broadcast, etc. */
    304 	short		if_extflags;	/* :: if_output MP-safe, etc. */
    305 	u_char		if_type;	/* :: ethernet, tokenring, etc. */
    306 	u_char		if_addrlen;	/* :: media address length */
    307 	u_char		if_hdrlen;	/* :: media header length */
    308 	/* XXX audit :? fields here. */
    309 	int		if_link_state;	/* :? current link state */
    310 	uint64_t	if_mtu;		/* :? maximum transmission unit */
    311 	uint64_t	if_metric;	/* :? routing metric (external only) */
    312 	uint64_t	if_baudrate;	/* :? linespeed */
    313 	struct timespec	if_lastchange;	/* :? last operational state change */
    314 #ifdef _KERNEL
    315 	percpu_t	*if_stats;	/* :: statistics */
    316 #else
    317 	void		*if_stats;	/* opaque to user-space */
    318 #endif /* _KERNEL */
    319 	/*
    320 	 * Procedure handles.  If you add more of these, don't forget the
    321 	 * corresponding NULL stub in if.c.
    322 	 */
    323 	int		(*if_output)	/* :: output routine (enqueue) */
    324 			    (struct ifnet *, struct mbuf *, const struct sockaddr *,
    325 			     const struct rtentry *);
    326 	void		(*_if_input)	/* :: input routine (from h/w driver) */
    327 			    (struct ifnet *, struct mbuf *);
    328 	void		(*if_start)	/* :: initiate output routine */
    329 			    (struct ifnet *);
    330 	int		(*if_transmit)	/* :: output routine, must be MP-safe */
    331 			    (struct ifnet *, struct mbuf *);
    332 	int		(*if_ioctl)	/* :: ioctl routine */
    333 			    (struct ifnet *, u_long, void *);
    334 	int		(*if_init)	/* :: init routine */
    335 			    (struct ifnet *);
    336 	void		(*if_stop)	/* :: stop routine */
    337 			    (struct ifnet *, int);
    338 	void		(*if_slowtimo)	/* :: timer routine */
    339 			    (struct ifnet *);
    340 #define	if_watchdog	if_slowtimo
    341 	void		(*if_drain)	/* :: routine to release resources */
    342 			    (struct ifnet *);
    343 	void		(*if_bpf_mtap)	/* :: bpf routine */
    344 			    (struct bpf_if *, struct mbuf *, u_int);
    345 	struct ifaltq	if_snd;		/* q: output queue (includes altq) */
    346 	struct ifaddr	*if_dl;		/* i: identity of this interface. */
    347 	const struct sockaddr_dl
    348 			*if_sadl;	/* i: pointer to sockaddr_dl of if_dl */
    349 	/*
    350 	 * May be NULL.  If not NULL, it is the address assigned
    351 	 * to the interface by the manufacturer, so it very likely
    352 	 * to be unique.  It MUST NOT be deleted.  It is highly
    353 	 * suitable for deriving the EUI64 for the interface.
    354 	 */
    355 	struct ifaddr	*if_hwdl;	/* i: h/w identity */
    356 	const uint8_t	*if_broadcastaddr;
    357 					/* :: linklevel broadcast bytestring */
    358 	struct bridge_softc
    359 			*if_bridge;	/* i: bridge glue */
    360 	struct bridge_iflist
    361 			*if_bridgeif;	/* i: shortcut to interface list entry */
    362 	int		if_dlt;		/* :: data link type (<net/dlt.h>) */
    363 	pfil_head_t *	if_pfil;	/* :: filtering point */
    364 	uint64_t	if_capabilities;
    365 					/* i: interface capabilities */
    366 	uint64_t	if_capenable;	/* i: capabilities enabled */
    367 	union {
    368 		void *		carp_s;	/* carp structure (used by !carp ifs) */
    369 		struct ifnet	*carp_d;/* ptr to carpdev (used by carp ifs) */
    370 	}		if_carp_ptr;	/* ?: */
    371 #define if_carp		if_carp_ptr.carp_s
    372 #define if_carpdev	if_carp_ptr.carp_d
    373 	/*
    374 	 * These are pre-computed based on an interfaces enabled
    375 	 * capabilities, for speed elsewhere.
    376 	 */
    377 	int		if_csum_flags_tx;
    378 					/* i: M_CSUM_* flags for Tx */
    379 	int		if_csum_flags_rx;
    380 					/* i: M_CSUM_* flags for Rx */
    381 
    382 	void		*if_afdata[AF_MAX];
    383 					/* a: */
    384 	struct mowner	*if_mowner;	/* ?: who owns mbufs for this interface */
    385 
    386 	void		*if_lagg;	/* :: lagg or agr structure */
    387 	void		*if_npf_private;/* ?: associated NPF context */
    388 
    389 	/*
    390 	 * pf specific data, used only when #if NPF > 0.
    391 	 */
    392 	void		*if_pf_kif;	/* ?: pf interface abstraction */
    393 	void		*if_pf_groups;	/* ?: pf interface groups */
    394 	/*
    395 	 * During an ifnet's lifetime, it has only one if_index, but
    396 	 * an if_index is not sufficient to identify an ifnet
    397 	 * because during the lifetime of the system, many ifnets may occupy a
    398 	 * given if_index.  Let us tell different ifnets at the same
    399 	 * if_index apart by their if_index_gen, a unique number that each ifnet
    400 	 * is assigned when it if_attach()s.  Now, the kernel can use the
    401 	 * pair (if_index, if_index_gen) as a weak reference to an ifnet.
    402 	 */
    403 	uint64_t	if_index_gen;	/* :: generation number for the ifnet
    404 					 * at if_index: if two ifnets' index
    405 					 * and generation number are both the
    406 					 * same, they are the same ifnet.
    407 					 */
    408 	struct sysctllog
    409 			*if_sysctl_log;	/* :: */
    410 	int		(*if_initaddr)  /* :: */
    411 			    (struct ifnet *, struct ifaddr *, bool);
    412 	int		(*if_setflags)	/* :: */
    413 			    (struct ifnet *, const u_short);
    414 	kmutex_t	*if_ioctl_lock;	/* :: */
    415 	char		*if_description;	/* i: interface description */
    416 #ifdef _KERNEL /* XXX kvm(3) */
    417 	struct if_slowtimo_data *if_slowtimo_data; /* :: */
    418 	struct krwlock	*if_afdata_lock;/* :: */
    419 	struct if_percpuq
    420 			*if_percpuq;	/* :: we should remove it in the future */
    421 	struct work	if_link_work;	/* q: linkage on link state work queue */
    422 	uint16_t	if_link_queue;	/* q: masked link state change queue */
    423 					/* q: is link state work scheduled? */
    424 	bool		if_link_scheduled;
    425 	struct pslist_entry
    426 			if_pslist_entry;/* i: */
    427 	struct psref_target
    428 			if_psref;	/* :: */
    429 	struct pslist_head
    430 			if_addr_pslist;	/* i: */
    431 	struct if_deferred_start
    432 			*if_deferred_start;
    433 					/* :: */
    434 	/* XXX should be protocol independent */
    435 	LIST_HEAD(, in6_multi)
    436 			if_multiaddrs;	/* 6: */
    437 	khook_list_t	*if_linkstate_hooks;	/* :: */
    438 #endif
    439 } ifnet_t;
    440 
    441 #include <net/if_stats.h>
    442 
    443 #define	if_name(ifp)	((ifp)->if_xname)
    444 
    445 #define	IFF_UP		0x0001		/* interface is up */
    446 #define	IFF_BROADCAST	0x0002		/* broadcast address valid */
    447 #define	IFF_DEBUG	0x0004		/* turn on debugging */
    448 #define	IFF_LOOPBACK	0x0008		/* is a loopback net */
    449 #define	IFF_POINTOPOINT	0x0010		/* interface is point-to-point link */
    450 /*			0x0020		   was IFF_NOTRAILERS */
    451 #define	IFF_RUNNING	0x0040		/* resources allocated */
    452 #define	IFF_NOARP	0x0080		/* no address resolution protocol */
    453 #define	IFF_PROMISC	0x0100		/* receive all packets */
    454 #define	IFF_ALLMULTI	0x0200		/* receive all multicast packets */
    455 #define	IFF_OACTIVE	0x0400		/* transmission in progress */
    456 #define	IFF_SIMPLEX	0x0800		/* can't hear own transmissions */
    457 #define	IFF_LINK0	0x1000		/* per link layer defined bit */
    458 #define	IFF_LINK1	0x2000		/* per link layer defined bit */
    459 #define	IFF_LINK2	0x4000		/* per link layer defined bit */
    460 #define	IFF_MULTICAST	0x8000		/* supports multicast */
    461 
    462 #define	IFEF_MPSAFE			__BIT(0)	/* handlers can run in parallel (see below) */
    463 
    464 /*
    465  * The guidelines for converting an interface to IFEF_MPSAFE are as follows
    466  *
    467  * Enabling IFEF_MPSAFE on an interface suppresses taking KERNEL_LOCK when
    468  * calling the following handlers:
    469  * - if_start
    470  *   - Note that if_transmit is always called without KERNEL_LOCK
    471  * - if_output
    472  * - if_ioctl
    473  * - if_init
    474  * - if_stop
    475  *
    476  * This means that an interface with IFEF_MPSAFE must make the above handlers
    477  * MP-safe or take KERNEL_LOCK by itself inside handlers that aren't MP-safe
    478  * yet.
    479  *
    480  * There are some additional restrictions to access member variables of struct
    481  * ifnet:
    482  * - if_flags
    483  *   - Must be updated with holding IFNET_LOCK
    484  *   - You cannot use the flag in Tx/Rx paths anymore because there is no
    485  *     synchronization on the flag except for IFNET_LOCK
    486  *   - Note that IFNET_LOCK can't be taken in softint because it's known
    487  *     that it causes a deadlock
    488  *     - Some synchronization mechanisms such as pserialize_perform are called
    489  *       with IFNET_LOCK and also require context switches on every CPUs
    490  *       that mean softints finish so trying to take IFNET_LOCK in softint
    491  *       might block on IFNET_LOCK and prevent such synchronization mechanisms
    492  *       from being completed
    493  *     - Currently the deadlock occurs only if NET_MPSAFE is enabled, however,
    494  *       we should deal with the restriction because NET_MPSAFE will be enabled
    495  *       by default in the future
    496  * - if_watchdog and if_timer
    497  *   - The watchdog framework works only for non-IFEF_MPSAFE interfaces
    498  *     that rely on KERNEL_LOCK
    499  *   - Interfaces with IFEF_MPSAFE have to provide its own watchdog mechanism
    500  *     if needed
    501  *     - Keep if_watchdog NULL when calling if_attach
    502  */
    503 
    504 #ifdef _KERNEL
    505 static __inline bool
    506 if_is_mpsafe(struct ifnet *ifp)
    507 {
    508 
    509 	return ((ifp->if_extflags & IFEF_MPSAFE) != 0);
    510 }
    511 
    512 static __inline int
    513 if_output_lock(struct ifnet *cifp, struct ifnet *ifp, struct mbuf *m,
    514     const struct sockaddr *dst, const struct rtentry *rt)
    515 {
    516 
    517 	if (if_is_mpsafe(cifp)) {
    518 		return (*cifp->if_output)(ifp, m, dst, rt);
    519 	} else {
    520 		int ret;
    521 
    522 		KERNEL_LOCK(1, NULL);
    523 		ret = (*cifp->if_output)(ifp, m, dst, rt);
    524 		KERNEL_UNLOCK_ONE(NULL);
    525 		return ret;
    526 	}
    527 }
    528 
    529 static __inline void
    530 if_start_lock(struct ifnet *ifp)
    531 {
    532 
    533 	if (if_is_mpsafe(ifp)) {
    534 		(*ifp->if_start)(ifp);
    535 	} else {
    536 		KERNEL_LOCK(1, NULL);
    537 		(*ifp->if_start)(ifp);
    538 		KERNEL_UNLOCK_ONE(NULL);
    539 	}
    540 }
    541 
    542 #define KERNEL_LOCK_IF_IFP_MPSAFE(ifp)					\
    543 	do { if (if_is_mpsafe(ifp)) { KERNEL_LOCK(1, NULL); } } while (0)
    544 #define KERNEL_UNLOCK_IF_IFP_MPSAFE(ifp)				\
    545 	do { if (if_is_mpsafe(ifp)) { KERNEL_UNLOCK_ONE(NULL); } } while (0)
    546 
    547 #define KERNEL_LOCK_UNLESS_IFP_MPSAFE(ifp)				\
    548 	do { if (!if_is_mpsafe(ifp)) { KERNEL_LOCK(1, NULL); } } while (0)
    549 #define KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(ifp)				\
    550 	do { if (!if_is_mpsafe(ifp)) { KERNEL_UNLOCK_ONE(NULL); } } while (0)
    551 
    552 #ifdef _KERNEL_OPT
    553 #include "opt_net_mpsafe.h"
    554 #endif
    555 
    556 /* XXX explore a better place to define */
    557 #ifdef NET_MPSAFE
    558 
    559 #define KERNEL_LOCK_UNLESS_NET_MPSAFE()		do { } while (0)
    560 #define KERNEL_UNLOCK_UNLESS_NET_MPSAFE()	do { } while (0)
    561 
    562 #define SOFTNET_LOCK_UNLESS_NET_MPSAFE()	do { } while (0)
    563 #define SOFTNET_UNLOCK_UNLESS_NET_MPSAFE()	do { } while (0)
    564 
    565 #define SOFTNET_LOCK_IF_NET_MPSAFE()					\
    566 	do { mutex_enter(softnet_lock); } while (0)
    567 #define SOFTNET_UNLOCK_IF_NET_MPSAFE()					\
    568 	do { mutex_exit(softnet_lock); } while (0)
    569 
    570 #else /* NET_MPSAFE */
    571 
    572 #define KERNEL_LOCK_UNLESS_NET_MPSAFE()					\
    573 	do { KERNEL_LOCK(1, NULL); } while (0)
    574 #define KERNEL_UNLOCK_UNLESS_NET_MPSAFE()				\
    575 	do { KERNEL_UNLOCK_ONE(NULL); } while (0)
    576 
    577 #define SOFTNET_LOCK_UNLESS_NET_MPSAFE()				\
    578 	do { mutex_enter(softnet_lock); } while (0)
    579 #define SOFTNET_UNLOCK_UNLESS_NET_MPSAFE()				\
    580 	do { mutex_exit(softnet_lock); } while (0)
    581 
    582 #define SOFTNET_LOCK_IF_NET_MPSAFE()		do { } while (0)
    583 #define SOFTNET_UNLOCK_IF_NET_MPSAFE()		do { } while (0)
    584 
    585 #endif /* NET_MPSAFE */
    586 
    587 #define SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE()				\
    588 	do {								\
    589 		SOFTNET_LOCK_UNLESS_NET_MPSAFE();			\
    590 		KERNEL_LOCK_UNLESS_NET_MPSAFE();			\
    591 	} while (0)
    592 
    593 #define SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE()			\
    594 	do {								\
    595 		KERNEL_UNLOCK_UNLESS_NET_MPSAFE();			\
    596 		SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();			\
    597 	} while (0)
    598 
    599 #endif /* _KERNEL */
    600 
    601 #define	IFFBITS \
    602     "\020\1UP\2BROADCAST\3DEBUG\4LOOPBACK\5POINTOPOINT" \
    603     "\7RUNNING\10NOARP\11PROMISC\12ALLMULTI\13OACTIVE\14SIMPLEX" \
    604     "\15LINK0\16LINK1\17LINK2\20MULTICAST"
    605 
    606 /* flags set internally only: */
    607 #define	IFF_CANTCHANGE \
    608 	(IFF_BROADCAST|IFF_POINTOPOINT|IFF_RUNNING|IFF_OACTIVE|\
    609 	    IFF_SIMPLEX|IFF_MULTICAST|IFF_ALLMULTI|IFF_PROMISC)
    610 
    611 /*
    612  * Some convenience macros used for setting ifi_baudrate.
    613  */
    614 #define	IF_Kbps(x)	((x) * 1000ULL)			/* kilobits/sec. */
    615 #define	IF_Mbps(x)	(IF_Kbps((x) * 1000ULL))	/* megabits/sec. */
    616 #define	IF_Gbps(x)	(IF_Mbps((x) * 1000ULL))	/* gigabits/sec. */
    617 
    618 /* Capabilities that interfaces can advertise. */
    619 					/* 0x01 .. 0x40 were previously used */
    620 #define	IFCAP_TSOv4		0x00080	/* can do TCPv4 segmentation offload */
    621 #define	IFCAP_CSUM_IPv4_Rx	0x00100	/* can do IPv4 header checksums (Rx) */
    622 #define	IFCAP_CSUM_IPv4_Tx	0x00200	/* can do IPv4 header checksums (Tx) */
    623 #define	IFCAP_CSUM_TCPv4_Rx	0x00400	/* can do IPv4/TCP checksums (Rx) */
    624 #define	IFCAP_CSUM_TCPv4_Tx	0x00800	/* can do IPv4/TCP checksums (Tx) */
    625 #define	IFCAP_CSUM_UDPv4_Rx	0x01000	/* can do IPv4/UDP checksums (Rx) */
    626 #define	IFCAP_CSUM_UDPv4_Tx	0x02000	/* can do IPv4/UDP checksums (Tx) */
    627 #define	IFCAP_CSUM_TCPv6_Rx	0x04000	/* can do IPv6/TCP checksums (Rx) */
    628 #define	IFCAP_CSUM_TCPv6_Tx	0x08000	/* can do IPv6/TCP checksums (Tx) */
    629 #define	IFCAP_CSUM_UDPv6_Rx	0x10000	/* can do IPv6/UDP checksums (Rx) */
    630 #define	IFCAP_CSUM_UDPv6_Tx	0x20000	/* can do IPv6/UDP checksums (Tx) */
    631 #define	IFCAP_TSOv6		0x40000	/* can do TCPv6 segmentation offload */
    632 #define	IFCAP_LRO		0x80000	/* can do Large Receive Offload */
    633 #define	IFCAP_MASK		0xfff80 /* currently valid capabilities */
    634 
    635 #define	IFCAPBITS		\
    636 	"\020"			\
    637 	"\10TSO4"		\
    638 	"\11IP4CSUM_Rx"		\
    639 	"\12IP4CSUM_Tx"		\
    640 	"\13TCP4CSUM_Rx"	\
    641 	"\14TCP4CSUM_Tx"	\
    642 	"\15UDP4CSUM_Rx"	\
    643 	"\16UDP4CSUM_Tx"	\
    644 	"\17TCP6CSUM_Rx"	\
    645 	"\20TCP6CSUM_Tx"	\
    646 	"\21UDP6CSUM_Rx"	\
    647 	"\22UDP6CSUM_Tx"	\
    648 	"\23TSO6"		\
    649 	"\24LRO"		\
    650 
    651 #define	IF_AFDATA_LOCK_INIT(ifp)	\
    652 	do {(ifp)->if_afdata_lock = rw_obj_alloc();} while (0)
    653 
    654 #define	IF_AFDATA_LOCK_DESTROY(ifp)	rw_obj_free((ifp)->if_afdata_lock)
    655 
    656 #define	IF_AFDATA_WLOCK(ifp)	rw_enter((ifp)->if_afdata_lock, RW_WRITER)
    657 #define	IF_AFDATA_RLOCK(ifp)	rw_enter((ifp)->if_afdata_lock, RW_READER)
    658 #define	IF_AFDATA_WUNLOCK(ifp)	rw_exit((ifp)->if_afdata_lock)
    659 #define	IF_AFDATA_RUNLOCK(ifp)	rw_exit((ifp)->if_afdata_lock)
    660 #define	IF_AFDATA_LOCK(ifp)	IF_AFDATA_WLOCK(ifp)
    661 #define	IF_AFDATA_UNLOCK(ifp)	IF_AFDATA_WUNLOCK(ifp)
    662 #define	IF_AFDATA_TRYLOCK(ifp)	rw_tryenter((ifp)->if_afdata_lock, RW_WRITER)
    663 
    664 #define	IF_AFDATA_LOCK_ASSERT(ifp)	\
    665 	KASSERT(rw_lock_held((ifp)->if_afdata_lock))
    666 #define	IF_AFDATA_RLOCK_ASSERT(ifp)	\
    667 	KASSERT(rw_read_held((ifp)->if_afdata_lock))
    668 #define	IF_AFDATA_WLOCK_ASSERT(ifp)	\
    669 	KASSERT(rw_write_held((ifp)->if_afdata_lock))
    670 
    671 /*
    672  * Output queues (ifp->if_snd) and internetwork datagram level (pup level 1)
    673  * input routines have queues of messages stored on ifqueue structures
    674  * (defined above).  Entries are added to and deleted from these structures
    675  * by these macros, which should be called with ipl raised to splnet().
    676  */
    677 #define	IF_QFULL(ifq)		((ifq)->ifq_len >= (ifq)->ifq_maxlen)
    678 #define	IF_DROP(ifq)		((ifq)->ifq_drops++)
    679 #define	IF_ENQUEUE(ifq, m) do { \
    680 	(m)->m_nextpkt = 0; \
    681 	if ((ifq)->ifq_tail == 0) \
    682 		(ifq)->ifq_head = m; \
    683 	else \
    684 		(ifq)->ifq_tail->m_nextpkt = m; \
    685 	(ifq)->ifq_tail = m; \
    686 	(ifq)->ifq_len++; \
    687 } while (/*CONSTCOND*/0)
    688 #define	IF_PREPEND(ifq, m) do { \
    689 	(m)->m_nextpkt = (ifq)->ifq_head; \
    690 	if ((ifq)->ifq_tail == 0) \
    691 		(ifq)->ifq_tail = (m); \
    692 	(ifq)->ifq_head = (m); \
    693 	(ifq)->ifq_len++; \
    694 } while (/*CONSTCOND*/0)
    695 #define	IF_DEQUEUE(ifq, m) do { \
    696 	(m) = (ifq)->ifq_head; \
    697 	if (m) { \
    698 		if (((ifq)->ifq_head = (m)->m_nextpkt) == 0) \
    699 			(ifq)->ifq_tail = 0; \
    700 		(m)->m_nextpkt = 0; \
    701 		(ifq)->ifq_len--; \
    702 	} \
    703 } while (/*CONSTCOND*/0)
    704 #define	IF_POLL(ifq, m)		((m) = (ifq)->ifq_head)
    705 #define	IF_PURGE(ifq)							\
    706 do {									\
    707 	struct mbuf *__m0;						\
    708 									\
    709 	for (;;) {							\
    710 		IF_DEQUEUE((ifq), __m0);				\
    711 		if (__m0 == NULL)					\
    712 			break;						\
    713 		else							\
    714 			m_freem(__m0);					\
    715 	}								\
    716 } while (/*CONSTCOND*/ 0)
    717 #define	IF_IS_EMPTY(ifq)	((ifq)->ifq_len == 0)
    718 
    719 #ifndef IFQ_MAXLEN
    720 #define	IFQ_MAXLEN	256
    721 #endif
    722 #define	IFNET_SLOWHZ	1		/* granularity is 1 second */
    723 
    724 /*
    725  * Structure defining statistics and other data kept regarding an address
    726  * on a network interface.
    727  */
    728 struct ifaddr_data {
    729 	int64_t	ifad_inbytes;
    730 	int64_t	ifad_outbytes;
    731 };
    732 
    733 /*
    734  * The ifaddr structure contains information about one address
    735  * of an interface.  They are maintained by the different address families,
    736  * are allocated and attached when an address is set, and are linked
    737  * together so all addresses for an interface can be located.
    738  */
    739 struct ifaddr {
    740 	struct	sockaddr *ifa_addr;	/* address of interface */
    741 	struct	sockaddr *ifa_dstaddr;	/* other end of p-to-p link */
    742 #define	ifa_broadaddr	ifa_dstaddr	/* broadcast address interface */
    743 	struct	sockaddr *ifa_netmask;	/* used to determine subnet */
    744 	struct	ifnet *ifa_ifp;		/* back-pointer to interface */
    745 	TAILQ_ENTRY(ifaddr) ifa_list;	/* list of addresses for interface */
    746 	struct	ifaddr_data	ifa_data;	/* statistics on the address */
    747 	void	(*ifa_rtrequest)	/* check or clean routes (+ or -)'d */
    748 		        (int, struct rtentry *, const struct rt_addrinfo *);
    749 	u_int	ifa_flags;		/* mostly rt_flags for cloning */
    750 	int	ifa_refcnt;		/* count of references */
    751 	int	ifa_metric;		/* cost of going out this interface */
    752 	struct ifaddr	*(*ifa_getifa)(struct ifaddr *,
    753 			               const struct sockaddr *);
    754 	uint32_t	*ifa_seqno;
    755 	int16_t	ifa_preference;	/* preference level for this address */
    756 #ifdef _KERNEL
    757 	struct pslist_entry     ifa_pslist_entry;
    758 	struct psref_target	ifa_psref;
    759 #endif
    760 };
    761 #define	IFA_ROUTE	RTF_UP	/* (0x01) route installed */
    762 #define	IFA_DESTROYING	0x2
    763 
    764 /*
    765  * Message format for use in obtaining information about interfaces from
    766  * sysctl and the routing socket.  We need to force 64-bit alignment if we
    767  * aren't using compatibility definitions.
    768  */
    769 #if !defined(_KERNEL) || !defined(COMPAT_RTSOCK)
    770 #define	__align64	__aligned(sizeof(uint64_t))
    771 #else
    772 #define	__align64
    773 #endif
    774 struct if_msghdr {
    775 	u_short	ifm_msglen __align64;
    776 				/* to skip over non-understood messages */
    777 	u_char	ifm_version;	/* future binary compatibility */
    778 	u_char	ifm_type;	/* message type */
    779 	int	ifm_addrs;	/* like rtm_addrs */
    780 	int	ifm_flags;	/* value of if_flags */
    781 	u_short	ifm_index;	/* index for associated ifp */
    782 	struct	if_data ifm_data __align64;
    783 				/* statistics and other data about if */
    784 };
    785 
    786 /*
    787  * Message format for use in obtaining information about interface addresses
    788  * from sysctl and the routing socket.
    789  */
    790 struct ifa_msghdr {
    791 	u_short	ifam_msglen __align64;
    792 				/* to skip over non-understood messages */
    793 	u_char	ifam_version;	/* future binary compatibility */
    794 	u_char	ifam_type;	/* message type */
    795 	u_short	ifam_index;	/* index for associated ifp */
    796 	int	ifam_flags;	/* value of ifa_flags */
    797 	int	ifam_addrs;	/* like rtm_addrs */
    798 	pid_t	ifam_pid;	/* identify sender */
    799 	int	ifam_addrflags;	/* family specific address flags */
    800 	int	ifam_metric;	/* value of ifa_metric */
    801 };
    802 
    803 /*
    804  * Message format announcing the arrival or departure of a network interface.
    805  */
    806 struct if_announcemsghdr {
    807 	u_short	ifan_msglen __align64;
    808 				/* to skip over non-understood messages */
    809 	u_char	ifan_version;	/* future binary compatibility */
    810 	u_char	ifan_type;	/* message type */
    811 	u_short	ifan_index;	/* index for associated ifp */
    812 	char	ifan_name[IFNAMSIZ]; /* if name, e.g. "en0" */
    813 	u_short	ifan_what;	/* what type of announcement */
    814 };
    815 
    816 #define	IFAN_ARRIVAL	0	/* interface arrival */
    817 #define	IFAN_DEPARTURE	1	/* interface departure */
    818 
    819 #undef __align64
    820 
    821 /*
    822  * Interface request structure used for socket
    823  * ioctl's.  All interface ioctl's must have parameter
    824  * definitions which begin with ifr_name.  The
    825  * remainder may be interface specific.
    826  */
    827 struct	ifreq {
    828 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
    829 	union {
    830 		struct	sockaddr ifru_addr;
    831 		struct	sockaddr ifru_dstaddr;
    832 		struct	sockaddr ifru_broadaddr;
    833 		struct	sockaddr_storage ifru_space;
    834 		short	ifru_flags;
    835 		int	ifru_addrflags;
    836 		int	ifru_metric;
    837 		int	ifru_mtu;
    838 		int	ifru_dlt;
    839 		u_int	ifru_value;
    840 		void *	ifru_data;
    841 		struct {
    842 			uint32_t	b_buflen;
    843 			void		*b_buf;
    844 		} ifru_b;
    845 	} ifr_ifru;
    846 #define	ifr_addr	ifr_ifru.ifru_addr	/* address */
    847 #define	ifr_dstaddr	ifr_ifru.ifru_dstaddr	/* other end of p-to-p link */
    848 #define	ifr_broadaddr	ifr_ifru.ifru_broadaddr	/* broadcast address */
    849 #define	ifr_space	ifr_ifru.ifru_space	/* sockaddr_storage */
    850 #define	ifr_flags	ifr_ifru.ifru_flags	/* flags */
    851 #define	ifr_addrflags	ifr_ifru.ifru_addrflags	/* addr flags */
    852 #define	ifr_metric	ifr_ifru.ifru_metric	/* metric */
    853 #define	ifr_mtu		ifr_ifru.ifru_mtu	/* mtu */
    854 #define	ifr_dlt		ifr_ifru.ifru_dlt	/* data link type (DLT_*) */
    855 #define	ifr_value	ifr_ifru.ifru_value	/* generic value */
    856 #define	ifr_media	ifr_ifru.ifru_metric	/* media options (overload) */
    857 #define	ifr_data	ifr_ifru.ifru_data	/* for use by interface
    858 						 * XXX deprecated
    859 						 */
    860 #define	ifr_buf		ifr_ifru.ifru_b.b_buf	/* new interface ioctls */
    861 #define	ifr_buflen	ifr_ifru.ifru_b.b_buflen
    862 #define	ifr_index	ifr_ifru.ifru_value	/* interface index, BSD */
    863 #define	ifr_ifindex	ifr_index		/* interface index, linux */
    864 };
    865 
    866 #ifdef _KERNEL
    867 #define	ifreq_setdstaddr	ifreq_setaddr
    868 #define	ifreq_setbroadaddr	ifreq_setaddr
    869 #define	ifreq_getdstaddr	ifreq_getaddr
    870 #define	ifreq_getbroadaddr	ifreq_getaddr
    871 
    872 static __inline const struct sockaddr *
    873 /*ARGSUSED*/
    874 ifreq_getaddr(u_long cmd, const struct ifreq *ifr)
    875 {
    876 	return &ifr->ifr_addr;
    877 }
    878 #endif /* _KERNEL */
    879 
    880 struct ifcapreq {
    881 	char		ifcr_name[IFNAMSIZ];	/* if name, e.g. "en0" */
    882 	uint64_t	ifcr_capabilities;	/* supported capabiliites */
    883 	uint64_t	ifcr_capenable;		/* capabilities enabled */
    884 };
    885 
    886 struct ifaliasreq {
    887 	char	ifra_name[IFNAMSIZ];		/* if name, e.g. "en0" */
    888 	struct	sockaddr ifra_addr;
    889 	struct	sockaddr ifra_dstaddr;
    890 #define	ifra_broadaddr	ifra_dstaddr
    891 	struct	sockaddr ifra_mask;
    892 };
    893 
    894 struct ifdatareq {
    895 	char	ifdr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
    896 	struct	if_data ifdr_data;
    897 };
    898 
    899 struct ifmediareq {
    900 	char	ifm_name[IFNAMSIZ];	/* if name, e.g. "en0" */
    901 	int	ifm_current;		/* IFMWD: current media options */
    902 	int	ifm_mask;		/* IFMWD: don't care mask */
    903 	int	ifm_status;		/* media status */
    904 	int	ifm_active;		/* IFMWD: active options */
    905 	int	ifm_count;		/* # entries in ifm_ulist
    906 					   array */
    907 	int	*ifm_ulist;		/* array of ifmedia word */
    908 };
    909 
    910 
    911 struct  ifdrv {
    912 	char		ifd_name[IFNAMSIZ];	/* if name, e.g. "en0" */
    913 	unsigned long	ifd_cmd;
    914 	size_t		ifd_len;
    915 	void		*ifd_data;
    916 };
    917 #define IFLINKSTR_QUERYLEN	0x01
    918 #define IFLINKSTR_UNSET		0x02
    919 
    920 /*
    921  * Structure used in SIOCGIFCONF request.
    922  * Used to retrieve interface configuration
    923  * for machine (useful for programs which
    924  * must know all networks accessible).
    925  */
    926 struct	ifconf {
    927 	int	ifc_len;		/* size of associated buffer */
    928 	union {
    929 		void *	ifcu_buf;
    930 		struct	ifreq *ifcu_req;
    931 	} ifc_ifcu;
    932 #define	ifc_buf	ifc_ifcu.ifcu_buf	/* buffer address */
    933 #define	ifc_req	ifc_ifcu.ifcu_req	/* array of structures returned */
    934 };
    935 
    936 /*
    937  * Structure for SIOC[AGD]LIFADDR
    938  */
    939 struct if_laddrreq {
    940 	char iflr_name[IFNAMSIZ];
    941 	unsigned int flags;
    942 #define IFLR_PREFIX	0x8000	/* in: prefix given  out: kernel fills id */
    943 #define IFLR_ACTIVE	0x4000	/* in/out: link-layer address activation */
    944 #define IFLR_FACTORY	0x2000	/* in/out: factory link-layer address */
    945 	unsigned int prefixlen;		/* in/out */
    946 	struct sockaddr_storage addr;	/* in/out */
    947 	struct sockaddr_storage dstaddr; /* out */
    948 };
    949 
    950 /*
    951  * Structure for SIOC[SG]IFADDRPREF
    952  */
    953 struct if_addrprefreq {
    954 	char			ifap_name[IFNAMSIZ];
    955 	int16_t			ifap_preference;	/* in/out */
    956 	struct sockaddr_storage	ifap_addr;		/* in/out */
    957 };
    958 
    959 #include <net/if_arp.h>
    960 
    961 #endif /* _NETBSD_SOURCE */
    962 
    963 #ifdef _KERNEL
    964 #ifdef ALTQ
    965 #define IFQ_ENQUEUE(ifq, m, err)					\
    966 do {									\
    967 	mutex_enter((ifq)->ifq_lock);					\
    968 	if (ALTQ_IS_ENABLED(ifq))					\
    969 		ALTQ_ENQUEUE((ifq), (m), (err));			\
    970 	else {								\
    971 		if (IF_QFULL(ifq)) {					\
    972 			m_freem(m);					\
    973 			(err) = ENOBUFS;				\
    974 		} else {						\
    975 			IF_ENQUEUE((ifq), (m));				\
    976 			(err) = 0;					\
    977 		}							\
    978 	}								\
    979 	if ((err))							\
    980 		(ifq)->ifq_drops++;					\
    981 	mutex_exit((ifq)->ifq_lock);					\
    982 } while (/*CONSTCOND*/ 0)
    983 
    984 #define IFQ_DEQUEUE(ifq, m)						\
    985 do {									\
    986 	mutex_enter((ifq)->ifq_lock);					\
    987 	if (TBR_IS_ENABLED(ifq))					\
    988 		(m) = tbr_dequeue((ifq), ALTDQ_REMOVE);			\
    989 	else if (ALTQ_IS_ENABLED(ifq))					\
    990 		ALTQ_DEQUEUE((ifq), (m));				\
    991 	else								\
    992 		IF_DEQUEUE((ifq), (m));					\
    993 	mutex_exit((ifq)->ifq_lock);					\
    994 } while (/*CONSTCOND*/ 0)
    995 
    996 #define	IFQ_POLL(ifq, m)						\
    997 do {									\
    998 	mutex_enter((ifq)->ifq_lock);					\
    999 	if (TBR_IS_ENABLED(ifq))					\
   1000 		(m) = tbr_dequeue((ifq), ALTDQ_POLL);			\
   1001 	else if (ALTQ_IS_ENABLED(ifq))					\
   1002 		ALTQ_POLL((ifq), (m));					\
   1003 	else								\
   1004 		IF_POLL((ifq), (m));					\
   1005 	mutex_exit((ifq)->ifq_lock);					\
   1006 } while (/*CONSTCOND*/ 0)
   1007 
   1008 #define	IFQ_PURGE(ifq)							\
   1009 do {									\
   1010 	mutex_enter((ifq)->ifq_lock);					\
   1011 	if (ALTQ_IS_ENABLED(ifq))					\
   1012 		ALTQ_PURGE(ifq);					\
   1013 	else								\
   1014 		IF_PURGE(ifq);						\
   1015 	mutex_exit((ifq)->ifq_lock);					\
   1016 } while (/*CONSTCOND*/ 0)
   1017 
   1018 #define	IFQ_SET_READY(ifq)						\
   1019 do {									\
   1020 	(ifq)->altq_flags |= ALTQF_READY;				\
   1021 } while (/*CONSTCOND*/ 0)
   1022 
   1023 #define	IFQ_CLASSIFY(ifq, m, af)					\
   1024 do {									\
   1025 	KASSERT(((m)->m_flags & M_PKTHDR) != 0);			\
   1026 	mutex_enter((ifq)->ifq_lock);					\
   1027 	if (ALTQ_IS_ENABLED(ifq)) {					\
   1028 		if (ALTQ_NEEDS_CLASSIFY(ifq))				\
   1029 			(m)->m_pkthdr.pattr_class = (*(ifq)->altq_classify) \
   1030 				((ifq)->altq_clfier, (m), (af));	\
   1031 		(m)->m_pkthdr.pattr_af = (af);				\
   1032 		(m)->m_pkthdr.pattr_hdr = mtod((m), void *);		\
   1033 	}								\
   1034 	mutex_exit((ifq)->ifq_lock);					\
   1035 } while (/*CONSTCOND*/ 0)
   1036 #else /* ! ALTQ */
   1037 #define	IFQ_ENQUEUE(ifq, m, err)					\
   1038 do {									\
   1039 	mutex_enter((ifq)->ifq_lock);					\
   1040 	if (IF_QFULL(ifq)) {						\
   1041 		m_freem(m);						\
   1042 		(err) = ENOBUFS;					\
   1043 	} else {							\
   1044 		IF_ENQUEUE((ifq), (m));					\
   1045 		(err) = 0;						\
   1046 	}								\
   1047 	if (err)							\
   1048 		(ifq)->ifq_drops++;					\
   1049 	mutex_exit((ifq)->ifq_lock);					\
   1050 } while (/*CONSTCOND*/ 0)
   1051 
   1052 #define	IFQ_DEQUEUE(ifq, m)						\
   1053 do {									\
   1054 	mutex_enter((ifq)->ifq_lock);					\
   1055 	IF_DEQUEUE((ifq), (m));						\
   1056 	mutex_exit((ifq)->ifq_lock);					\
   1057 } while (/*CONSTCOND*/ 0)
   1058 
   1059 #define	IFQ_POLL(ifq, m)						\
   1060 do {									\
   1061 	mutex_enter((ifq)->ifq_lock);					\
   1062 	IF_POLL((ifq), (m));						\
   1063 	mutex_exit((ifq)->ifq_lock);					\
   1064 } while (/*CONSTCOND*/ 0)
   1065 
   1066 #define	IFQ_PURGE(ifq)							\
   1067 do {									\
   1068 	mutex_enter((ifq)->ifq_lock);					\
   1069 	IF_PURGE(ifq);							\
   1070 	mutex_exit((ifq)->ifq_lock);					\
   1071 } while (/*CONSTCOND*/ 0)
   1072 
   1073 #define	IFQ_SET_READY(ifq)	/* nothing */
   1074 
   1075 #define	IFQ_CLASSIFY(ifq, m, af) /* nothing */
   1076 
   1077 #endif /* ALTQ */
   1078 
   1079 #define IFQ_LOCK_INIT(ifq)	(ifq)->ifq_lock =			\
   1080 	    mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET)
   1081 #define IFQ_LOCK_DESTROY(ifq)	mutex_obj_free((ifq)->ifq_lock)
   1082 #define IFQ_LOCK(ifq)		mutex_enter((ifq)->ifq_lock)
   1083 #define IFQ_UNLOCK(ifq)		mutex_exit((ifq)->ifq_lock)
   1084 
   1085 #define	IFQ_IS_EMPTY(ifq)		IF_IS_EMPTY(ifq)
   1086 #define	IFQ_INC_LEN(ifq)		((ifq)->ifq_len++)
   1087 #define	IFQ_DEC_LEN(ifq)		(--(ifq)->ifq_len)
   1088 #define	IFQ_INC_DROPS(ifq)		((ifq)->ifq_drops++)
   1089 #define	IFQ_SET_MAXLEN(ifq, len)	((ifq)->ifq_maxlen = (len))
   1090 
   1091 #include <sys/mallocvar.h>
   1092 MALLOC_DECLARE(M_IFADDR);
   1093 MALLOC_DECLARE(M_IFMADDR);
   1094 
   1095 int ifreq_setaddr(u_long, struct ifreq *, const struct sockaddr *);
   1096 
   1097 struct ifnet *if_alloc(u_char);
   1098 void if_free(struct ifnet *);
   1099 void if_initname(struct ifnet *, const char *, int);
   1100 struct ifaddr *if_dl_create(const struct ifnet *, const struct sockaddr_dl **);
   1101 void if_activate_sadl(struct ifnet *, struct ifaddr *,
   1102     const struct sockaddr_dl *);
   1103 void	if_set_sadl(struct ifnet *, const void *, u_char, bool);
   1104 void	if_alloc_sadl(struct ifnet *);
   1105 void	if_free_sadl(struct ifnet *, int);
   1106 void	if_initialize(struct ifnet *);
   1107 void	if_register(struct ifnet *);
   1108 void	if_attach(struct ifnet *); /* Deprecated. Use if_initialize and if_register */
   1109 void	if_attachdomain(void);
   1110 void	if_deactivate(struct ifnet *);
   1111 bool	if_is_deactivated(const struct ifnet *);
   1112 void	if_export_if_data(struct ifnet *, struct if_data *, bool);
   1113 void	if_purgeaddrs(struct ifnet *, int, void (*)(struct ifaddr *));
   1114 void	if_detach(struct ifnet *);
   1115 void	if_down(struct ifnet *);
   1116 void	if_down_locked(struct ifnet *);
   1117 void	if_link_state_change(struct ifnet *, int);
   1118 void	if_domain_link_state_change(struct ifnet *, int);
   1119 void	if_up(struct ifnet *);
   1120 void	ifinit(void);
   1121 void	ifinit1(void);
   1122 void	ifinit_post(void);
   1123 int	ifaddrpref_ioctl(struct socket *, u_long, void *, struct ifnet *);
   1124 extern int (*ifioctl)(struct socket *, u_long, void *, struct lwp *);
   1125 int	ifioctl_common(struct ifnet *, u_long, void *);
   1126 int	ifpromisc(struct ifnet *, int);
   1127 int	ifpromisc_locked(struct ifnet *, int);
   1128 int	if_addr_init(ifnet_t *, struct ifaddr *, bool);
   1129 int	if_do_dad(struct ifnet *);
   1130 int	if_mcast_op(ifnet_t *, const unsigned long, const struct sockaddr *);
   1131 int	if_flags_set(struct ifnet *, const u_short);
   1132 int	if_clone_list(int, char *, int *);
   1133 
   1134 int	if_ioctl(struct ifnet *, u_long, void *);
   1135 int	if_init(struct ifnet *);
   1136 void	if_stop(struct ifnet *, int);
   1137 
   1138 struct	ifnet *ifunit(const char *);
   1139 struct	ifnet *if_get(const char *, struct psref *);
   1140 ifnet_t *if_byindex(u_int);
   1141 ifnet_t *_if_byindex(u_int);
   1142 ifnet_t *if_get_byindex(u_int, struct psref *);
   1143 ifnet_t *if_get_bylla(const void *, unsigned char, struct psref *);
   1144 void	if_put(const struct ifnet *, struct psref *);
   1145 void	if_acquire(struct ifnet *, struct psref *);
   1146 #define	if_release	if_put
   1147 
   1148 int if_tunnel_check_nesting(struct ifnet *, struct mbuf *, int);
   1149 percpu_t *if_tunnel_alloc_ro_percpu(void);
   1150 void if_tunnel_free_ro_percpu(percpu_t *);
   1151 void if_tunnel_ro_percpu_rtcache_free(percpu_t *);
   1152 
   1153 struct tunnel_ro {
   1154 	struct route *tr_ro;
   1155 	kmutex_t *tr_lock;
   1156 };
   1157 
   1158 static inline void
   1159 if_tunnel_get_ro(percpu_t *ro_percpu, struct route **ro, kmutex_t **lock)
   1160 {
   1161 	struct tunnel_ro *tro;
   1162 
   1163 	tro = percpu_getref(ro_percpu);
   1164 	*ro = tro->tr_ro;
   1165 	*lock = tro->tr_lock;
   1166 	mutex_enter(*lock);
   1167 }
   1168 
   1169 static inline void
   1170 if_tunnel_put_ro(percpu_t *ro_percpu, kmutex_t *lock)
   1171 {
   1172 
   1173 	mutex_exit(lock);
   1174 	percpu_putref(ro_percpu);
   1175 }
   1176 
   1177 static __inline if_index_t
   1178 if_get_index(const struct ifnet *ifp)
   1179 {
   1180 
   1181 	return ifp != NULL ? ifp->if_index : 0;
   1182 }
   1183 
   1184 bool	if_held(struct ifnet *);
   1185 
   1186 void	if_input(struct ifnet *, struct mbuf *);
   1187 
   1188 struct if_percpuq *
   1189 	if_percpuq_create(struct ifnet *);
   1190 void	if_percpuq_destroy(struct if_percpuq *);
   1191 void
   1192 	if_percpuq_enqueue(struct if_percpuq *, struct mbuf *);
   1193 
   1194 void	if_deferred_start_init(struct ifnet *, void (*)(struct ifnet *));
   1195 void	if_schedule_deferred_start(struct ifnet *);
   1196 
   1197 void ifa_insert(struct ifnet *, struct ifaddr *);
   1198 void ifa_remove(struct ifnet *, struct ifaddr *);
   1199 
   1200 void	ifa_psref_init(struct ifaddr *);
   1201 void	ifa_acquire(struct ifaddr *, struct psref *);
   1202 void	ifa_release(struct ifaddr *, struct psref *);
   1203 bool	ifa_held(struct ifaddr *);
   1204 bool	ifa_is_destroying(struct ifaddr *);
   1205 
   1206 void	ifaref(struct ifaddr *);
   1207 void	ifafree(struct ifaddr *);
   1208 
   1209 struct	ifaddr *ifa_ifwithaddr(const struct sockaddr *);
   1210 struct	ifaddr *ifa_ifwithaddr_psref(const struct sockaddr *, struct psref *);
   1211 struct	ifaddr *ifa_ifwithaf(int);
   1212 struct	ifaddr *ifa_ifwithdstaddr(const struct sockaddr *);
   1213 struct	ifaddr *ifa_ifwithdstaddr_psref(const struct sockaddr *,
   1214 	    struct psref *);
   1215 struct	ifaddr *ifa_ifwithnet(const struct sockaddr *);
   1216 struct	ifaddr *ifa_ifwithnet_psref(const struct sockaddr *, struct psref *);
   1217 struct	ifaddr *ifa_ifwithladdr(const struct sockaddr *);
   1218 struct	ifaddr *ifa_ifwithladdr_psref(const struct sockaddr *, struct psref *);
   1219 struct	ifaddr *ifaof_ifpforaddr(const struct sockaddr *, struct ifnet *);
   1220 struct	ifaddr *ifaof_ifpforaddr_psref(const struct sockaddr *, struct ifnet *,
   1221 	    struct psref *);
   1222 void	link_rtrequest(int, struct rtentry *, const struct rt_addrinfo *);
   1223 void	p2p_rtrequest(int, struct rtentry *, const struct rt_addrinfo *);
   1224 
   1225 void	if_clone_attach(struct if_clone *);
   1226 void	if_clone_detach(struct if_clone *);
   1227 
   1228 int	if_transmit_lock(struct ifnet *, struct mbuf *);
   1229 
   1230 int	ifq_enqueue(struct ifnet *, struct mbuf *);
   1231 int	ifq_enqueue2(struct ifnet *, struct ifqueue *, struct mbuf *);
   1232 
   1233 int	loioctl(struct ifnet *, u_long, void *);
   1234 void	loopattach(int);
   1235 void	loopinit(void);
   1236 int	looutput(struct ifnet *,
   1237 	   struct mbuf *, const struct sockaddr *, const struct rtentry *);
   1238 
   1239 void *	if_linkstate_change_establish(struct ifnet *,
   1240 	    void (*)(void *), void *);
   1241 void	if_linkstate_change_disestablish(struct ifnet *,
   1242 	    void *, kmutex_t *);
   1243 
   1244 /*
   1245  * These are exported because they're an easy way to tell if
   1246  * an interface is going away without having to burn a flag.
   1247  */
   1248 int	if_nulloutput(struct ifnet *, struct mbuf *,
   1249 	    const struct sockaddr *, const struct rtentry *);
   1250 void	if_nullinput(struct ifnet *, struct mbuf *);
   1251 void	if_nullstart(struct ifnet *);
   1252 int	if_nulltransmit(struct ifnet *, struct mbuf *);
   1253 int	if_nullioctl(struct ifnet *, u_long, void *);
   1254 int	if_nullinit(struct ifnet *);
   1255 void	if_nullstop(struct ifnet *, int);
   1256 void	if_nullslowtimo(struct ifnet *);
   1257 #define	if_nullwatchdog	if_nullslowtimo
   1258 void	if_nulldrain(struct ifnet *);
   1259 #else
   1260 struct if_nameindex {
   1261 	unsigned int	if_index;	/* 1, 2, ... */
   1262 	char		*if_name;	/* null terminated name: "le0", ... */
   1263 };
   1264 
   1265 #include <sys/cdefs.h>
   1266 __BEGIN_DECLS
   1267 unsigned int if_nametoindex(const char *);
   1268 char *	if_indextoname(unsigned int, char *);
   1269 struct	if_nameindex * if_nameindex(void);
   1270 void	if_freenameindex(struct if_nameindex *);
   1271 __END_DECLS
   1272 #endif /* _KERNEL */ /* XXX really ALTQ? */
   1273 
   1274 #ifdef _KERNEL
   1275 
   1276 #define	IFADDR_FIRST(__ifp)		TAILQ_FIRST(&(__ifp)->if_addrlist)
   1277 #define	IFADDR_NEXT(__ifa)		TAILQ_NEXT((__ifa), ifa_list)
   1278 #define	IFADDR_FOREACH(__ifa, __ifp)	TAILQ_FOREACH(__ifa, \
   1279 					    &(__ifp)->if_addrlist, ifa_list)
   1280 #define	IFADDR_FOREACH_SAFE(__ifa, __ifp, __nifa) \
   1281 					    TAILQ_FOREACH_SAFE(__ifa, \
   1282 					    &(__ifp)->if_addrlist, ifa_list, __nifa)
   1283 #define	IFADDR_EMPTY(__ifp)		TAILQ_EMPTY(&(__ifp)->if_addrlist)
   1284 
   1285 #define IFADDR_ENTRY_INIT(__ifa)					\
   1286 	PSLIST_ENTRY_INIT((__ifa), ifa_pslist_entry)
   1287 #define IFADDR_ENTRY_DESTROY(__ifa)					\
   1288 	PSLIST_ENTRY_DESTROY((__ifa), ifa_pslist_entry)
   1289 #define IFADDR_READER_EMPTY(__ifp)					\
   1290 	(PSLIST_READER_FIRST(&(__ifp)->if_addr_pslist, struct ifaddr,	\
   1291 	                     ifa_pslist_entry) == NULL)
   1292 #define IFADDR_READER_FIRST(__ifp)					\
   1293 	PSLIST_READER_FIRST(&(__ifp)->if_addr_pslist, struct ifaddr,	\
   1294 	                    ifa_pslist_entry)
   1295 #define IFADDR_READER_NEXT(__ifa)					\
   1296 	PSLIST_READER_NEXT((__ifa), struct ifaddr, ifa_pslist_entry)
   1297 #define IFADDR_READER_FOREACH(__ifa, __ifp)				\
   1298 	PSLIST_READER_FOREACH((__ifa), &(__ifp)->if_addr_pslist, struct ifaddr,\
   1299 	                      ifa_pslist_entry)
   1300 #define IFADDR_WRITER_INSERT_HEAD(__ifp, __ifa)				\
   1301 	PSLIST_WRITER_INSERT_HEAD(&(__ifp)->if_addr_pslist, (__ifa),	\
   1302 	                          ifa_pslist_entry)
   1303 #define IFADDR_WRITER_REMOVE(__ifa)					\
   1304 	PSLIST_WRITER_REMOVE((__ifa), ifa_pslist_entry)
   1305 #define IFADDR_WRITER_FOREACH(__ifa, __ifp)				\
   1306 	PSLIST_WRITER_FOREACH((__ifa), &(__ifp)->if_addr_pslist, struct ifaddr,\
   1307 	                      ifa_pslist_entry)
   1308 #define IFADDR_WRITER_NEXT(__ifp)					\
   1309 	PSLIST_WRITER_NEXT((__ifp), struct ifaddr, ifa_pslist_entry)
   1310 #define IFADDR_WRITER_INSERT_AFTER(__ifp, __new)			\
   1311 	PSLIST_WRITER_INSERT_AFTER((__ifp), (__new), ifa_pslist_entry)
   1312 #define IFADDR_WRITER_EMPTY(__ifp)					\
   1313 	(PSLIST_WRITER_FIRST(&(__ifp)->if_addr_pslist, struct ifaddr,	\
   1314 	                     ifa_pslist_entry) == NULL)
   1315 #define IFADDR_WRITER_INSERT_TAIL(__ifp, __new)				\
   1316 	do {								\
   1317 		if (IFADDR_WRITER_EMPTY(__ifp)) {			\
   1318 			IFADDR_WRITER_INSERT_HEAD((__ifp), (__new));	\
   1319 		} else {						\
   1320 			struct ifaddr *__ifa;				\
   1321 			IFADDR_WRITER_FOREACH(__ifa, (__ifp)) {		\
   1322 				if (IFADDR_WRITER_NEXT(__ifa) == NULL) {\
   1323 					IFADDR_WRITER_INSERT_AFTER(__ifa,\
   1324 					    (__new));			\
   1325 					break;				\
   1326 				}					\
   1327 			}						\
   1328 		}							\
   1329 	} while (0)
   1330 
   1331 #define	IFNET_GLOBAL_LOCK()			mutex_enter(&ifnet_mtx)
   1332 #define	IFNET_GLOBAL_UNLOCK()			mutex_exit(&ifnet_mtx)
   1333 #define	IFNET_GLOBAL_LOCKED()			mutex_owned(&ifnet_mtx)
   1334 
   1335 #define IFNET_READER_EMPTY() \
   1336 	(PSLIST_READER_FIRST(&ifnet_pslist, struct ifnet, if_pslist_entry) == NULL)
   1337 #define IFNET_READER_FIRST() \
   1338 	PSLIST_READER_FIRST(&ifnet_pslist, struct ifnet, if_pslist_entry)
   1339 #define IFNET_READER_NEXT(__ifp) \
   1340 	PSLIST_READER_NEXT((__ifp), struct ifnet, if_pslist_entry)
   1341 #define IFNET_READER_FOREACH(__ifp) \
   1342 	PSLIST_READER_FOREACH((__ifp), &ifnet_pslist, struct ifnet, \
   1343 	                      if_pslist_entry)
   1344 #define IFNET_WRITER_INSERT_HEAD(__ifp) \
   1345 	PSLIST_WRITER_INSERT_HEAD(&ifnet_pslist, (__ifp), if_pslist_entry)
   1346 #define IFNET_WRITER_REMOVE(__ifp) \
   1347 	PSLIST_WRITER_REMOVE((__ifp), if_pslist_entry)
   1348 #define IFNET_WRITER_FOREACH(__ifp) \
   1349 	PSLIST_WRITER_FOREACH((__ifp), &ifnet_pslist, struct ifnet, \
   1350 	                      if_pslist_entry)
   1351 #define IFNET_WRITER_NEXT(__ifp) \
   1352 	PSLIST_WRITER_NEXT((__ifp), struct ifnet, if_pslist_entry)
   1353 #define IFNET_WRITER_INSERT_AFTER(__ifp, __new) \
   1354 	PSLIST_WRITER_INSERT_AFTER((__ifp), (__new), if_pslist_entry)
   1355 #define IFNET_WRITER_EMPTY() \
   1356 	(PSLIST_WRITER_FIRST(&ifnet_pslist, struct ifnet, if_pslist_entry) == NULL)
   1357 #define IFNET_WRITER_INSERT_TAIL(__new)					\
   1358 	do {								\
   1359 		if (IFNET_WRITER_EMPTY()) {				\
   1360 			IFNET_WRITER_INSERT_HEAD(__new);		\
   1361 		} else {						\
   1362 			struct ifnet *__ifp;				\
   1363 			IFNET_WRITER_FOREACH(__ifp) {			\
   1364 				if (IFNET_WRITER_NEXT(__ifp) == NULL) {	\
   1365 					IFNET_WRITER_INSERT_AFTER(__ifp,\
   1366 					    (__new));			\
   1367 					break;				\
   1368 				}					\
   1369 			}						\
   1370 		}							\
   1371 	} while (0)
   1372 
   1373 #define IFNET_LOCK(ifp)		mutex_enter((ifp)->if_ioctl_lock)
   1374 #define IFNET_UNLOCK(ifp)	mutex_exit((ifp)->if_ioctl_lock)
   1375 #define IFNET_LOCKED(ifp)	mutex_owned((ifp)->if_ioctl_lock)
   1376 
   1377 #define IFNET_ASSERT_UNLOCKED(ifp)	\
   1378 	KDASSERT(mutex_ownable((ifp)->if_ioctl_lock))
   1379 
   1380 extern struct pslist_head ifnet_pslist;
   1381 extern kmutex_t ifnet_mtx;
   1382 
   1383 extern struct ifnet *lo0ifp;
   1384 
   1385 /*
   1386  * ifq sysctl support
   1387  */
   1388 int	sysctl_ifq(int *name, u_int namelen, void *oldp,
   1389 		       size_t *oldlenp, void *newp, size_t newlen,
   1390 		       struct ifqueue *ifq);
   1391 /* symbolic names for terminal (per-protocol) CTL_IFQ_ nodes */
   1392 #define IFQCTL_LEN	1
   1393 #define IFQCTL_MAXLEN	2
   1394 #define IFQCTL_PEAK	3
   1395 #define IFQCTL_DROPS	4
   1396 
   1397 /*
   1398  * Hook for if_vlan - needed by if_agr
   1399  */
   1400 MODULE_HOOK(if_vlan_vlan_input_hook,
   1401     struct mbuf *, (struct ifnet *, struct mbuf *));
   1402 
   1403 #endif /* _KERNEL */
   1404 
   1405 #endif /* !_NET_IF_H_ */
   1406