Home | History | Annotate | Line # | Download | only in net
if.h revision 1.288
      1 /*	$NetBSD: if.h,v 1.288 2020/09/27 19:16:28 roy 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  * Status bit descriptions for the various interface types.
    212  */
    213 struct if_status_description {
    214 	unsigned char	ifs_type;
    215 	unsigned char	ifs_state;
    216 	const char	*ifs_string;
    217 };
    218 
    219 #define LINK_STATE_DESC_MATCH(_ifs, _t, _s)				\
    220 	(((_ifs)->ifs_type == (_t) || (_ifs)->ifs_type == 0) &&		\
    221 	    (_ifs)->ifs_state == (_s))
    222 
    223 #define LINK_STATE_DESCRIPTIONS {					\
    224 	{ IFT_ETHER, LINK_STATE_DOWN, "no carrier" },			\
    225 	{ IFT_IEEE80211, LINK_STATE_DOWN, "no network" },		\
    226 	{ IFT_PPP, LINK_STATE_DOWN, "no carrier" },			\
    227 	{ IFT_CARP, LINK_STATE_DOWN, "backup" },			\
    228 	{ IFT_CARP, LINK_STATE_UP, "master" },				\
    229 	{ 0, LINK_STATE_UP, "active" },					\
    230 	{ 0, LINK_STATE_UNKNOWN, "unknown" },				\
    231 	{ 0, LINK_STATE_DOWN, "down" },					\
    232 	{ 0, 0, NULL }							\
    233 }
    234 
    235 /*
    236  * Structure defining a queue for a network interface.
    237  */
    238 struct ifqueue {
    239 	struct		mbuf *ifq_head;
    240 	struct		mbuf *ifq_tail;
    241 	int		ifq_len;
    242 	int		ifq_maxlen;
    243 	int		ifq_drops;
    244 	kmutex_t	*ifq_lock;
    245 };
    246 
    247 #ifdef _KERNEL
    248 #include <sys/percpu.h>
    249 #include <sys/callout.h>
    250 #include <sys/rwlock.h>
    251 #include <sys/workqueue.h>
    252 
    253 #endif /* _KERNEL */
    254 
    255 /*
    256  * Structure defining a queue for a network interface.
    257  *
    258  * (Would like to call this struct ``if'', but C isn't PL/1.)
    259  */
    260 TAILQ_HEAD(ifnet_head, ifnet);		/* the actual queue head */
    261 
    262 struct bridge_softc;
    263 struct bridge_iflist;
    264 struct callout;
    265 struct krwlock;
    266 struct if_percpuq;
    267 struct if_deferred_start;
    268 struct in6_multi;
    269 
    270 typedef unsigned short if_index_t;
    271 
    272 /*
    273  * Interface.  Field markings and the corresponding locks:
    274  *
    275  * i:	IFNET_LOCK (a.k.a., if_ioctl_lock)
    276  * q:	ifq_lock (struct ifaltq)
    277  * a:	if_afdata_lock
    278  * 6:	in6_multilock (global lock)
    279  * ::	unlocked, stable
    280  * ?:	unknown, maybe unsafe
    281  *
    282  * Lock order: IFNET_LOCK => in6_multilock => if_afdata_lock => ifq_lock
    283  *   Note that currently if_afdata_lock and ifq_lock aren't held
    284  *   at the same time, but define the order anyway.
    285  *
    286  * Lock order of IFNET_LOCK with other locks:
    287  *     softnet_lock => solock => IFNET_LOCK => ND6_LOCK, in_multilock
    288  */
    289 typedef struct ifnet {
    290 	void		*if_softc;	/* :: lower-level data for this if */
    291 	/* DEPRECATED. Keep it to avoid breaking kvm(3) users */
    292 	TAILQ_ENTRY(ifnet)
    293 			if_list;	/* i: all struct ifnets are chained */
    294 	TAILQ_HEAD(, ifaddr)
    295 			if_addrlist;	/* i: linked list of addresses per if */
    296 	char		if_xname[IFNAMSIZ];
    297 					/* :: external name (name + unit) */
    298 	int		if_pcount;	/* i: number of promiscuous listeners */
    299 	struct bpf_if	*if_bpf;	/* :: packet filter structure */
    300 	if_index_t	if_index;	/* :: numeric abbreviation for this if */
    301 	short		if_timer;	/* ?: time 'til if_slowtimo called */
    302 	unsigned short	if_flags;	/* i: up/down, broadcast, etc. */
    303 	short		if_extflags;	/* :: if_output MP-safe, etc. */
    304 	u_char		if_type;	/* :: ethernet, tokenring, etc. */
    305 	u_char		if_addrlen;	/* :: media address length */
    306 	u_char		if_hdrlen;	/* :: media header length */
    307 	/* XXX audit :? fields here. */
    308 	int		if_link_state;	/* :? current link state */
    309 	uint64_t	if_mtu;		/* :? maximum transmission unit */
    310 	uint64_t	if_metric;	/* :? routing metric (external only) */
    311 	uint64_t	if_baudrate;	/* :? linespeed */
    312 	struct timespec	if_lastchange;	/* :? last operational state change */
    313 #ifdef _KERNEL
    314 	percpu_t	*if_stats;	/* :: statistics */
    315 #else
    316 	void		*if_stats;	/* opaque to user-space */
    317 #endif /* _KERNEL */
    318 	/*
    319 	 * Procedure handles.  If you add more of these, don't forget the
    320 	 * corresponding NULL stub in if.c.
    321 	 */
    322 	int		(*if_output)	/* :: output routine (enqueue) */
    323 			    (struct ifnet *, struct mbuf *, const struct sockaddr *,
    324 			     const struct rtentry *);
    325 	void		(*_if_input)	/* :: input routine (from h/w driver) */
    326 			    (struct ifnet *, struct mbuf *);
    327 	void		(*if_start)	/* :: initiate output routine */
    328 			    (struct ifnet *);
    329 	int		(*if_transmit)	/* :: output routine, must be MP-safe */
    330 			    (struct ifnet *, struct mbuf *);
    331 	int		(*if_ioctl)	/* :: ioctl routine */
    332 			    (struct ifnet *, u_long, void *);
    333 	int		(*if_init)	/* :: init routine */
    334 			    (struct ifnet *);
    335 	void		(*if_stop)	/* :: stop routine */
    336 			    (struct ifnet *, int);
    337 	void		(*if_slowtimo)	/* :: timer routine */
    338 			    (struct ifnet *);
    339 #define	if_watchdog	if_slowtimo
    340 	void		(*if_drain)	/* :: routine to release resources */
    341 			    (struct ifnet *);
    342 	struct ifaltq	if_snd;		/* q: output queue (includes altq) */
    343 	struct ifaddr	*if_dl;		/* i: identity of this interface. */
    344 	const struct sockaddr_dl
    345 			*if_sadl;	/* i: pointer to sockaddr_dl of if_dl */
    346 	/*
    347 	 * May be NULL.  If not NULL, it is the address assigned
    348 	 * to the interface by the manufacturer, so it very likely
    349 	 * to be unique.  It MUST NOT be deleted.  It is highly
    350 	 * suitable for deriving the EUI64 for the interface.
    351 	 */
    352 	struct ifaddr	*if_hwdl;	/* i: h/w identity */
    353 	const uint8_t	*if_broadcastaddr;
    354 					/* :: linklevel broadcast bytestring */
    355 	struct bridge_softc
    356 			*if_bridge;	/* i: bridge glue */
    357 	struct bridge_iflist
    358 			*if_bridgeif;	/* i: shortcut to interface list entry */
    359 	int		if_dlt;		/* :: data link type (<net/dlt.h>) */
    360 	pfil_head_t *	if_pfil;	/* :: filtering point */
    361 	uint64_t	if_capabilities;
    362 					/* i: interface capabilities */
    363 	uint64_t	if_capenable;	/* i: capabilities enabled */
    364 	union {
    365 		void *		carp_s;	/* carp structure (used by !carp ifs) */
    366 		struct ifnet	*carp_d;/* ptr to carpdev (used by carp ifs) */
    367 	}		if_carp_ptr;	/* ?: */
    368 #define if_carp		if_carp_ptr.carp_s
    369 #define if_carpdev	if_carp_ptr.carp_d
    370 	/*
    371 	 * These are pre-computed based on an interfaces enabled
    372 	 * capabilities, for speed elsewhere.
    373 	 */
    374 	int		if_csum_flags_tx;
    375 					/* i: M_CSUM_* flags for Tx */
    376 	int		if_csum_flags_rx;
    377 					/* i: M_CSUM_* flags for Rx */
    378 
    379 	void		*if_afdata[AF_MAX];
    380 					/* a: */
    381 	struct mowner	*if_mowner;	/* ?: who owns mbufs for this interface */
    382 
    383 	void		*if_agrprivate;	/* ?: used only when #if NAGR > 0 */
    384 	void		*if_npf_private;/* ?: associated NPF context */
    385 
    386 	/*
    387 	 * pf specific data, used only when #if NPF > 0.
    388 	 */
    389 	void		*if_pf_kif;	/* ?: pf interface abstraction */
    390 	void		*if_pf_groups;	/* ?: pf interface groups */
    391 	/*
    392 	 * During an ifnet's lifetime, it has only one if_index, but
    393 	 * and if_index is not sufficient to identify an ifnet
    394 	 * because during the lifetime of the system, many ifnets may occupy a
    395 	 * given if_index.  Let us tell different ifnets at the same
    396 	 * if_index apart by their if_index_gen, a unique number that each ifnet
    397 	 * is assigned when it if_attach()s.  Now, the kernel can use the
    398 	 * pair (if_index, if_index_gen) as a weak reference to an ifnet.
    399 	 */
    400 	uint64_t	if_index_gen;	/* :: generation number for the ifnet
    401 					 * at if_index: if two ifnets' index
    402 					 * and generation number are both the
    403 					 * same, they are the same ifnet.
    404 					 */
    405 	struct sysctllog
    406 			*if_sysctl_log;	/* :: */
    407 	int		(*if_initaddr)  /* :: */
    408 			    (struct ifnet *, struct ifaddr *, bool);
    409 	int		(*if_setflags)	/* :: */
    410 			    (struct ifnet *, const u_short);
    411 	kmutex_t	*if_ioctl_lock;	/* :: */
    412 	char		*if_description;	/* i: interface description */
    413 #ifdef _KERNEL /* XXX kvm(3) */
    414 	struct callout	*if_slowtimo_ch;/* :: */
    415 	struct krwlock	*if_afdata_lock;/* :: */
    416 	struct if_percpuq
    417 			*if_percpuq;	/* :: we should remove it in the future */
    418 	struct work	if_link_work;	/* q: linkage on link state work queue */
    419 	uint16_t	if_link_queue;	/* q: masked link state change queue */
    420 					/* q: is link state work scheduled? */
    421 	bool		if_link_scheduled;
    422 	void		(*if_link_state_changed)(struct ifnet *, int);
    423 	struct pslist_entry
    424 			if_pslist_entry;/* i: */
    425 	struct psref_target
    426 			if_psref;	/* :: */
    427 	struct pslist_head
    428 			if_addr_pslist;	/* i: */
    429 	struct if_deferred_start
    430 			*if_deferred_start;
    431 					/* :: */
    432 	/* XXX should be protocol independent */
    433 	LIST_HEAD(, in6_multi)
    434 			if_multiaddrs;	/* 6: */
    435 #endif
    436 } ifnet_t;
    437 
    438 #include <net/if_stats.h>
    439 
    440 #define	if_name(ifp)	((ifp)->if_xname)
    441 
    442 #define	IFF_UP		0x0001		/* interface is up */
    443 #define	IFF_BROADCAST	0x0002		/* broadcast address valid */
    444 #define	IFF_DEBUG	0x0004		/* turn on debugging */
    445 #define	IFF_LOOPBACK	0x0008		/* is a loopback net */
    446 #define	IFF_POINTOPOINT	0x0010		/* interface is point-to-point link */
    447 /*			0x0020		   was IFF_NOTRAILERS */
    448 #define	IFF_RUNNING	0x0040		/* resources allocated */
    449 #define	IFF_NOARP	0x0080		/* no address resolution protocol */
    450 #define	IFF_PROMISC	0x0100		/* receive all packets */
    451 #define	IFF_ALLMULTI	0x0200		/* receive all multicast packets */
    452 #define	IFF_OACTIVE	0x0400		/* transmission in progress */
    453 #define	IFF_SIMPLEX	0x0800		/* can't hear own transmissions */
    454 #define	IFF_LINK0	0x1000		/* per link layer defined bit */
    455 #define	IFF_LINK1	0x2000		/* per link layer defined bit */
    456 #define	IFF_LINK2	0x4000		/* per link layer defined bit */
    457 #define	IFF_MULTICAST	0x8000		/* supports multicast */
    458 
    459 #define	IFEF_MPSAFE			__BIT(0)	/* handlers can run in parallel (see below) */
    460 #define	IFEF_NO_LINK_STATE_CHANGE	__BIT(1)	/* doesn't use link state interrupts */
    461 
    462 /*
    463  * The guidelines for converting an interface to IFEF_MPSAFE are as follows
    464  *
    465  * Enabling IFEF_MPSAFE on an interface suppresses taking KERNEL_LOCK when
    466  * calling the following handlers:
    467  * - if_start
    468  *   - Note that if_transmit is always called without KERNEL_LOCK
    469  * - if_output
    470  * - if_ioctl
    471  * - if_init
    472  * - if_stop
    473  *
    474  * This means that an interface with IFEF_MPSAFE must make the above handlers
    475  * MP-safe or take KERNEL_LOCK by itself inside handlers that aren't MP-safe
    476  * yet.
    477  *
    478  * There are some additional restrictions to access member variables of struct
    479  * ifnet:
    480  * - if_flags
    481  *   - Must be updated with holding IFNET_LOCK
    482  *   - You cannot use the flag in Tx/Rx paths anymore because there is no
    483  *     synchronization on the flag except for IFNET_LOCK
    484  *   - Note that IFNET_LOCK can't be taken in softint because it's known
    485  *     that it causes a deadlock
    486  *     - Some synchronization mechanisms such as pserialize_perform are called
    487  *       with IFNET_LOCK and also require context switches on every CPUs
    488  *       that mean softints finish so trying to take IFNET_LOCK in softint
    489  *       might block on IFNET_LOCK and prevent such synchronization mechanisms
    490  *       from being completed
    491  *     - Currently the deadlock occurs only if NET_MPSAFE is enabled, however,
    492  *       we should deal with the restriction because NET_MPSAFE will be enabled
    493  *       by default in the future
    494  * - if_watchdog and if_timer
    495  *   - The watchdog framework works only for non-IFEF_MPSAFE interfaces
    496  *     that rely on KERNEL_LOCK
    497  *   - Interfaces with IFEF_MPSAFE have to provide its own watchdog mechanism
    498  *     if needed
    499  *     - Keep if_watchdog NULL when calling if_attach
    500  */
    501 
    502 #ifdef _KERNEL
    503 static __inline bool
    504 if_is_mpsafe(struct ifnet *ifp)
    505 {
    506 
    507 	return ((ifp->if_extflags & IFEF_MPSAFE) != 0);
    508 }
    509 
    510 static __inline int
    511 if_output_lock(struct ifnet *cifp, struct ifnet *ifp, struct mbuf *m,
    512     const struct sockaddr *dst, const struct rtentry *rt)
    513 {
    514 
    515 	if (if_is_mpsafe(cifp)) {
    516 		return (*cifp->if_output)(ifp, m, dst, rt);
    517 	} else {
    518 		int ret;
    519 
    520 		KERNEL_LOCK(1, NULL);
    521 		ret = (*cifp->if_output)(ifp, m, dst, rt);
    522 		KERNEL_UNLOCK_ONE(NULL);
    523 		return ret;
    524 	}
    525 }
    526 
    527 static __inline void
    528 if_start_lock(struct ifnet *ifp)
    529 {
    530 
    531 	if (if_is_mpsafe(ifp)) {
    532 		(*ifp->if_start)(ifp);
    533 	} else {
    534 		KERNEL_LOCK(1, NULL);
    535 		(*ifp->if_start)(ifp);
    536 		KERNEL_UNLOCK_ONE(NULL);
    537 	}
    538 }
    539 
    540 static __inline bool
    541 if_is_link_state_changeable(struct ifnet *ifp)
    542 {
    543 
    544 	return ((ifp->if_extflags & IFEF_NO_LINK_STATE_CHANGE) == 0);
    545 }
    546 
    547 #define KERNEL_LOCK_IF_IFP_MPSAFE(ifp)					\
    548 	do { if (if_is_mpsafe(ifp)) { KERNEL_LOCK(1, NULL); } } while (0)
    549 #define KERNEL_UNLOCK_IF_IFP_MPSAFE(ifp)				\
    550 	do { if (if_is_mpsafe(ifp)) { KERNEL_UNLOCK_ONE(NULL); } } while (0)
    551 
    552 #define KERNEL_LOCK_UNLESS_IFP_MPSAFE(ifp)				\
    553 	do { if (!if_is_mpsafe(ifp)) { KERNEL_LOCK(1, NULL); } } while (0)
    554 #define KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(ifp)				\
    555 	do { if (!if_is_mpsafe(ifp)) { KERNEL_UNLOCK_ONE(NULL); } } while (0)
    556 
    557 #ifdef _KERNEL_OPT
    558 #include "opt_net_mpsafe.h"
    559 #endif
    560 
    561 /* XXX explore a better place to define */
    562 #ifdef NET_MPSAFE
    563 
    564 #define KERNEL_LOCK_UNLESS_NET_MPSAFE()		do { } while (0)
    565 #define KERNEL_UNLOCK_UNLESS_NET_MPSAFE()	do { } while (0)
    566 
    567 #define SOFTNET_LOCK_UNLESS_NET_MPSAFE()	do { } while (0)
    568 #define SOFTNET_UNLOCK_UNLESS_NET_MPSAFE()	do { } while (0)
    569 
    570 #define SOFTNET_LOCK_IF_NET_MPSAFE()					\
    571 	do { mutex_enter(softnet_lock); } while (0)
    572 #define SOFTNET_UNLOCK_IF_NET_MPSAFE()					\
    573 	do { mutex_exit(softnet_lock); } while (0)
    574 
    575 #else /* NET_MPSAFE */
    576 
    577 #define KERNEL_LOCK_UNLESS_NET_MPSAFE()					\
    578 	do { KERNEL_LOCK(1, NULL); } while (0)
    579 #define KERNEL_UNLOCK_UNLESS_NET_MPSAFE()				\
    580 	do { KERNEL_UNLOCK_ONE(NULL); } while (0)
    581 
    582 #define SOFTNET_LOCK_UNLESS_NET_MPSAFE()				\
    583 	do { mutex_enter(softnet_lock); } while (0)
    584 #define SOFTNET_UNLOCK_UNLESS_NET_MPSAFE()				\
    585 	do { mutex_exit(softnet_lock); } while (0)
    586 
    587 #define SOFTNET_LOCK_IF_NET_MPSAFE()		do { } while (0)
    588 #define SOFTNET_UNLOCK_IF_NET_MPSAFE()		do { } while (0)
    589 
    590 #endif /* NET_MPSAFE */
    591 
    592 #define SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE()				\
    593 	do {								\
    594 		SOFTNET_LOCK_UNLESS_NET_MPSAFE();			\
    595 		KERNEL_LOCK_UNLESS_NET_MPSAFE();			\
    596 	} while (0)
    597 
    598 #define SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE()			\
    599 	do {								\
    600 		KERNEL_UNLOCK_UNLESS_NET_MPSAFE();			\
    601 		SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();			\
    602 	} while (0)
    603 
    604 #endif /* _KERNEL */
    605 
    606 #define	IFFBITS \
    607     "\020\1UP\2BROADCAST\3DEBUG\4LOOPBACK\5POINTOPOINT" \
    608     "\7RUNNING\10NOARP\11PROMISC\12ALLMULTI\13OACTIVE\14SIMPLEX" \
    609     "\15LINK0\16LINK1\17LINK2\20MULTICAST"
    610 
    611 /* flags set internally only: */
    612 #define	IFF_CANTCHANGE \
    613 	(IFF_BROADCAST|IFF_POINTOPOINT|IFF_RUNNING|IFF_OACTIVE|\
    614 	    IFF_SIMPLEX|IFF_MULTICAST|IFF_ALLMULTI|IFF_PROMISC)
    615 
    616 /*
    617  * Some convenience macros used for setting ifi_baudrate.
    618  */
    619 #define	IF_Kbps(x)	((x) * 1000ULL)			/* kilobits/sec. */
    620 #define	IF_Mbps(x)	(IF_Kbps((x) * 1000ULL))	/* megabits/sec. */
    621 #define	IF_Gbps(x)	(IF_Mbps((x) * 1000ULL))	/* gigabits/sec. */
    622 
    623 /* Capabilities that interfaces can advertise. */
    624 					/* 0x01 .. 0x40 were previously used */
    625 #define	IFCAP_TSOv4		0x00080	/* can do TCPv4 segmentation offload */
    626 #define	IFCAP_CSUM_IPv4_Rx	0x00100	/* can do IPv4 header checksums (Rx) */
    627 #define	IFCAP_CSUM_IPv4_Tx	0x00200	/* can do IPv4 header checksums (Tx) */
    628 #define	IFCAP_CSUM_TCPv4_Rx	0x00400	/* can do IPv4/TCP checksums (Rx) */
    629 #define	IFCAP_CSUM_TCPv4_Tx	0x00800	/* can do IPv4/TCP checksums (Tx) */
    630 #define	IFCAP_CSUM_UDPv4_Rx	0x01000	/* can do IPv4/UDP checksums (Rx) */
    631 #define	IFCAP_CSUM_UDPv4_Tx	0x02000	/* can do IPv4/UDP checksums (Tx) */
    632 #define	IFCAP_CSUM_TCPv6_Rx	0x04000	/* can do IPv6/TCP checksums (Rx) */
    633 #define	IFCAP_CSUM_TCPv6_Tx	0x08000	/* can do IPv6/TCP checksums (Tx) */
    634 #define	IFCAP_CSUM_UDPv6_Rx	0x10000	/* can do IPv6/UDP checksums (Rx) */
    635 #define	IFCAP_CSUM_UDPv6_Tx	0x20000	/* can do IPv6/UDP checksums (Tx) */
    636 #define	IFCAP_TSOv6		0x40000	/* can do TCPv6 segmentation offload */
    637 #define	IFCAP_LRO		0x80000	/* can do Large Receive Offload */
    638 #define	IFCAP_MASK		0xfff80 /* currently valid capabilities */
    639 
    640 #define	IFCAPBITS		\
    641 	"\020"			\
    642 	"\10TSO4"		\
    643 	"\11IP4CSUM_Rx"		\
    644 	"\12IP4CSUM_Tx"		\
    645 	"\13TCP4CSUM_Rx"	\
    646 	"\14TCP4CSUM_Tx"	\
    647 	"\15UDP4CSUM_Rx"	\
    648 	"\16UDP4CSUM_Tx"	\
    649 	"\17TCP6CSUM_Rx"	\
    650 	"\20TCP6CSUM_Tx"	\
    651 	"\21UDP6CSUM_Rx"	\
    652 	"\22UDP6CSUM_Tx"	\
    653 	"\23TSO6"		\
    654 	"\24LRO"		\
    655 
    656 #define	IF_AFDATA_LOCK_INIT(ifp)	\
    657 	do {(ifp)->if_afdata_lock = rw_obj_alloc();} while (0)
    658 
    659 #define	IF_AFDATA_LOCK_DESTROY(ifp)	rw_obj_free((ifp)->if_afdata_lock)
    660 
    661 #define	IF_AFDATA_WLOCK(ifp)	rw_enter((ifp)->if_afdata_lock, RW_WRITER)
    662 #define	IF_AFDATA_RLOCK(ifp)	rw_enter((ifp)->if_afdata_lock, RW_READER)
    663 #define	IF_AFDATA_WUNLOCK(ifp)	rw_exit((ifp)->if_afdata_lock)
    664 #define	IF_AFDATA_RUNLOCK(ifp)	rw_exit((ifp)->if_afdata_lock)
    665 #define	IF_AFDATA_LOCK(ifp)	IF_AFDATA_WLOCK(ifp)
    666 #define	IF_AFDATA_UNLOCK(ifp)	IF_AFDATA_WUNLOCK(ifp)
    667 #define	IF_AFDATA_TRYLOCK(ifp)	rw_tryenter((ifp)->if_afdata_lock, RW_WRITER)
    668 
    669 #define	IF_AFDATA_LOCK_ASSERT(ifp)	\
    670 	KASSERT(rw_lock_held((ifp)->if_afdata_lock))
    671 #define	IF_AFDATA_RLOCK_ASSERT(ifp)	\
    672 	KASSERT(rw_read_held((ifp)->if_afdata_lock))
    673 #define	IF_AFDATA_WLOCK_ASSERT(ifp)	\
    674 	KASSERT(rw_write_held((ifp)->if_afdata_lock))
    675 
    676 /*
    677  * Output queues (ifp->if_snd) and internetwork datagram level (pup level 1)
    678  * input routines have queues of messages stored on ifqueue structures
    679  * (defined above).  Entries are added to and deleted from these structures
    680  * by these macros, which should be called with ipl raised to splnet().
    681  */
    682 #define	IF_QFULL(ifq)		((ifq)->ifq_len >= (ifq)->ifq_maxlen)
    683 #define	IF_DROP(ifq)		((ifq)->ifq_drops++)
    684 #define	IF_ENQUEUE(ifq, m) do { \
    685 	(m)->m_nextpkt = 0; \
    686 	if ((ifq)->ifq_tail == 0) \
    687 		(ifq)->ifq_head = m; \
    688 	else \
    689 		(ifq)->ifq_tail->m_nextpkt = m; \
    690 	(ifq)->ifq_tail = m; \
    691 	(ifq)->ifq_len++; \
    692 } while (/*CONSTCOND*/0)
    693 #define	IF_PREPEND(ifq, m) do { \
    694 	(m)->m_nextpkt = (ifq)->ifq_head; \
    695 	if ((ifq)->ifq_tail == 0) \
    696 		(ifq)->ifq_tail = (m); \
    697 	(ifq)->ifq_head = (m); \
    698 	(ifq)->ifq_len++; \
    699 } while (/*CONSTCOND*/0)
    700 #define	IF_DEQUEUE(ifq, m) do { \
    701 	(m) = (ifq)->ifq_head; \
    702 	if (m) { \
    703 		if (((ifq)->ifq_head = (m)->m_nextpkt) == 0) \
    704 			(ifq)->ifq_tail = 0; \
    705 		(m)->m_nextpkt = 0; \
    706 		(ifq)->ifq_len--; \
    707 	} \
    708 } while (/*CONSTCOND*/0)
    709 #define	IF_POLL(ifq, m)		((m) = (ifq)->ifq_head)
    710 #define	IF_PURGE(ifq)							\
    711 do {									\
    712 	struct mbuf *__m0;						\
    713 									\
    714 	for (;;) {							\
    715 		IF_DEQUEUE((ifq), __m0);				\
    716 		if (__m0 == NULL)					\
    717 			break;						\
    718 		else							\
    719 			m_freem(__m0);					\
    720 	}								\
    721 } while (/*CONSTCOND*/ 0)
    722 #define	IF_IS_EMPTY(ifq)	((ifq)->ifq_len == 0)
    723 
    724 #ifndef IFQ_MAXLEN
    725 #define	IFQ_MAXLEN	256
    726 #endif
    727 #define	IFNET_SLOWHZ	1		/* granularity is 1 second */
    728 
    729 /*
    730  * Structure defining statistics and other data kept regarding an address
    731  * on a network interface.
    732  */
    733 struct ifaddr_data {
    734 	int64_t	ifad_inbytes;
    735 	int64_t	ifad_outbytes;
    736 };
    737 
    738 /*
    739  * The ifaddr structure contains information about one address
    740  * of an interface.  They are maintained by the different address families,
    741  * are allocated and attached when an address is set, and are linked
    742  * together so all addresses for an interface can be located.
    743  */
    744 struct ifaddr {
    745 	struct	sockaddr *ifa_addr;	/* address of interface */
    746 	struct	sockaddr *ifa_dstaddr;	/* other end of p-to-p link */
    747 #define	ifa_broadaddr	ifa_dstaddr	/* broadcast address interface */
    748 	struct	sockaddr *ifa_netmask;	/* used to determine subnet */
    749 	struct	ifnet *ifa_ifp;		/* back-pointer to interface */
    750 	TAILQ_ENTRY(ifaddr) ifa_list;	/* list of addresses for interface */
    751 	struct	ifaddr_data	ifa_data;	/* statistics on the address */
    752 	void	(*ifa_rtrequest)	/* check or clean routes (+ or -)'d */
    753 		        (int, struct rtentry *, const struct rt_addrinfo *);
    754 	u_int	ifa_flags;		/* mostly rt_flags for cloning */
    755 	int	ifa_refcnt;		/* count of references */
    756 	int	ifa_metric;		/* cost of going out this interface */
    757 	struct ifaddr	*(*ifa_getifa)(struct ifaddr *,
    758 			               const struct sockaddr *);
    759 	uint32_t	*ifa_seqno;
    760 	int16_t	ifa_preference;	/* preference level for this address */
    761 #ifdef _KERNEL
    762 	struct pslist_entry     ifa_pslist_entry;
    763 	struct psref_target	ifa_psref;
    764 #endif
    765 };
    766 #define	IFA_ROUTE	RTF_UP	/* (0x01) route installed */
    767 #define	IFA_DESTROYING	0x2
    768 
    769 /*
    770  * Message format for use in obtaining information about interfaces from
    771  * sysctl and the routing socket.  We need to force 64-bit alignment if we
    772  * aren't using compatiblity definitons.
    773  */
    774 #if !defined(_KERNEL) || !defined(COMPAT_RTSOCK)
    775 #define	__align64	__aligned(sizeof(uint64_t))
    776 #else
    777 #define	__align64
    778 #endif
    779 struct if_msghdr {
    780 	u_short	ifm_msglen __align64;
    781 				/* to skip over non-understood messages */
    782 	u_char	ifm_version;	/* future binary compatibility */
    783 	u_char	ifm_type;	/* message type */
    784 	int	ifm_addrs;	/* like rtm_addrs */
    785 	int	ifm_flags;	/* value of if_flags */
    786 	u_short	ifm_index;	/* index for associated ifp */
    787 	struct	if_data ifm_data __align64;
    788 				/* statistics and other data about if */
    789 };
    790 
    791 /*
    792  * Message format for use in obtaining information about interface addresses
    793  * from sysctl and the routing socket.
    794  */
    795 struct ifa_msghdr {
    796 	u_short	ifam_msglen __align64;
    797 				/* to skip over non-understood messages */
    798 	u_char	ifam_version;	/* future binary compatibility */
    799 	u_char	ifam_type;	/* message type */
    800 	u_short	ifam_index;	/* index for associated ifp */
    801 	int	ifam_flags;	/* value of ifa_flags */
    802 	int	ifam_addrs;	/* like rtm_addrs */
    803 	pid_t	ifam_pid;	/* identify sender */
    804 	int	ifam_addrflags;	/* family specific address flags */
    805 	int	ifam_metric;	/* value of ifa_metric */
    806 };
    807 
    808 /*
    809  * Message format announcing the arrival or departure of a network interface.
    810  */
    811 struct if_announcemsghdr {
    812 	u_short	ifan_msglen __align64;
    813 				/* to skip over non-understood messages */
    814 	u_char	ifan_version;	/* future binary compatibility */
    815 	u_char	ifan_type;	/* message type */
    816 	u_short	ifan_index;	/* index for associated ifp */
    817 	char	ifan_name[IFNAMSIZ]; /* if name, e.g. "en0" */
    818 	u_short	ifan_what;	/* what type of announcement */
    819 };
    820 
    821 #define	IFAN_ARRIVAL	0	/* interface arrival */
    822 #define	IFAN_DEPARTURE	1	/* interface departure */
    823 
    824 #undef __align64
    825 
    826 /*
    827  * Interface request structure used for socket
    828  * ioctl's.  All interface ioctl's must have parameter
    829  * definitions which begin with ifr_name.  The
    830  * remainder may be interface specific.
    831  */
    832 struct	ifreq {
    833 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
    834 	union {
    835 		struct	sockaddr ifru_addr;
    836 		struct	sockaddr ifru_dstaddr;
    837 		struct	sockaddr ifru_broadaddr;
    838 		struct	sockaddr_storage ifru_space;
    839 		short	ifru_flags;
    840 		int	ifru_addrflags;
    841 		int	ifru_metric;
    842 		int	ifru_mtu;
    843 		int	ifru_dlt;
    844 		u_int	ifru_value;
    845 		void *	ifru_data;
    846 		struct {
    847 			uint32_t	b_buflen;
    848 			void		*b_buf;
    849 		} ifru_b;
    850 	} ifr_ifru;
    851 #define	ifr_addr	ifr_ifru.ifru_addr	/* address */
    852 #define	ifr_dstaddr	ifr_ifru.ifru_dstaddr	/* other end of p-to-p link */
    853 #define	ifr_broadaddr	ifr_ifru.ifru_broadaddr	/* broadcast address */
    854 #define	ifr_space	ifr_ifru.ifru_space	/* sockaddr_storage */
    855 #define	ifr_flags	ifr_ifru.ifru_flags	/* flags */
    856 #define	ifr_addrflags	ifr_ifru.ifru_addrflags	/* addr flags */
    857 #define	ifr_metric	ifr_ifru.ifru_metric	/* metric */
    858 #define	ifr_mtu		ifr_ifru.ifru_mtu	/* mtu */
    859 #define	ifr_dlt		ifr_ifru.ifru_dlt	/* data link type (DLT_*) */
    860 #define	ifr_value	ifr_ifru.ifru_value	/* generic value */
    861 #define	ifr_media	ifr_ifru.ifru_metric	/* media options (overload) */
    862 #define	ifr_data	ifr_ifru.ifru_data	/* for use by interface
    863 						 * XXX deprecated
    864 						 */
    865 #define	ifr_buf		ifr_ifru.ifru_b.b_buf	/* new interface ioctls */
    866 #define	ifr_buflen	ifr_ifru.ifru_b.b_buflen
    867 #define	ifr_index	ifr_ifru.ifru_value	/* interface index, BSD */
    868 #define	ifr_ifindex	ifr_index		/* interface index, linux */
    869 };
    870 
    871 #ifdef _KERNEL
    872 #define	ifreq_setdstaddr	ifreq_setaddr
    873 #define	ifreq_setbroadaddr	ifreq_setaddr
    874 #define	ifreq_getdstaddr	ifreq_getaddr
    875 #define	ifreq_getbroadaddr	ifreq_getaddr
    876 
    877 static __inline const struct sockaddr *
    878 /*ARGSUSED*/
    879 ifreq_getaddr(u_long cmd, const struct ifreq *ifr)
    880 {
    881 	return &ifr->ifr_addr;
    882 }
    883 #endif /* _KERNEL */
    884 
    885 struct ifcapreq {
    886 	char		ifcr_name[IFNAMSIZ];	/* if name, e.g. "en0" */
    887 	uint64_t	ifcr_capabilities;	/* supported capabiliites */
    888 	uint64_t	ifcr_capenable;		/* capabilities enabled */
    889 };
    890 
    891 struct ifaliasreq {
    892 	char	ifra_name[IFNAMSIZ];		/* if name, e.g. "en0" */
    893 	struct	sockaddr ifra_addr;
    894 	struct	sockaddr ifra_dstaddr;
    895 #define	ifra_broadaddr	ifra_dstaddr
    896 	struct	sockaddr ifra_mask;
    897 };
    898 
    899 struct ifdatareq {
    900 	char	ifdr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
    901 	struct	if_data ifdr_data;
    902 };
    903 
    904 struct ifmediareq {
    905 	char	ifm_name[IFNAMSIZ];	/* if name, e.g. "en0" */
    906 	int	ifm_current;		/* IFMWD: current media options */
    907 	int	ifm_mask;		/* IFMWD: don't care mask */
    908 	int	ifm_status;		/* media status */
    909 	int	ifm_active;		/* IFMWD: active options */
    910 	int	ifm_count;		/* # entries in ifm_ulist
    911 					   array */
    912 	int	*ifm_ulist;		/* array of ifmedia word */
    913 };
    914 
    915 
    916 struct  ifdrv {
    917 	char		ifd_name[IFNAMSIZ];	/* if name, e.g. "en0" */
    918 	unsigned long	ifd_cmd;
    919 	size_t		ifd_len;
    920 	void		*ifd_data;
    921 };
    922 #define IFLINKSTR_QUERYLEN	0x01
    923 #define IFLINKSTR_UNSET		0x02
    924 
    925 /*
    926  * Structure used in SIOCGIFCONF request.
    927  * Used to retrieve interface configuration
    928  * for machine (useful for programs which
    929  * must know all networks accessible).
    930  */
    931 struct	ifconf {
    932 	int	ifc_len;		/* size of associated buffer */
    933 	union {
    934 		void *	ifcu_buf;
    935 		struct	ifreq *ifcu_req;
    936 	} ifc_ifcu;
    937 #define	ifc_buf	ifc_ifcu.ifcu_buf	/* buffer address */
    938 #define	ifc_req	ifc_ifcu.ifcu_req	/* array of structures returned */
    939 };
    940 
    941 /*
    942  * Structure for SIOC[AGD]LIFADDR
    943  */
    944 struct if_laddrreq {
    945 	char iflr_name[IFNAMSIZ];
    946 	unsigned int flags;
    947 #define IFLR_PREFIX	0x8000	/* in: prefix given  out: kernel fills id */
    948 #define IFLR_ACTIVE	0x4000	/* in/out: link-layer address activation */
    949 #define IFLR_FACTORY	0x2000	/* in/out: factory link-layer address */
    950 	unsigned int prefixlen;		/* in/out */
    951 	struct sockaddr_storage addr;	/* in/out */
    952 	struct sockaddr_storage dstaddr; /* out */
    953 };
    954 
    955 /*
    956  * Structure for SIOC[SG]IFADDRPREF
    957  */
    958 struct if_addrprefreq {
    959 	char			ifap_name[IFNAMSIZ];
    960 	int16_t			ifap_preference;	/* in/out */
    961 	struct sockaddr_storage	ifap_addr;		/* in/out */
    962 };
    963 
    964 #include <net/if_arp.h>
    965 
    966 #endif /* _NETBSD_SOURCE */
    967 
    968 #ifdef _KERNEL
    969 #ifdef ALTQ
    970 #define IFQ_ENQUEUE(ifq, m, err)					\
    971 do {									\
    972 	mutex_enter((ifq)->ifq_lock);					\
    973 	if (ALTQ_IS_ENABLED(ifq))					\
    974 		ALTQ_ENQUEUE((ifq), (m), (err));			\
    975 	else {								\
    976 		if (IF_QFULL(ifq)) {					\
    977 			m_freem(m);					\
    978 			(err) = ENOBUFS;				\
    979 		} else {						\
    980 			IF_ENQUEUE((ifq), (m));				\
    981 			(err) = 0;					\
    982 		}							\
    983 	}								\
    984 	if ((err))							\
    985 		(ifq)->ifq_drops++;					\
    986 	mutex_exit((ifq)->ifq_lock);					\
    987 } while (/*CONSTCOND*/ 0)
    988 
    989 #define IFQ_DEQUEUE(ifq, m)						\
    990 do {									\
    991 	mutex_enter((ifq)->ifq_lock);					\
    992 	if (TBR_IS_ENABLED(ifq))					\
    993 		(m) = tbr_dequeue((ifq), ALTDQ_REMOVE);			\
    994 	else if (ALTQ_IS_ENABLED(ifq))					\
    995 		ALTQ_DEQUEUE((ifq), (m));				\
    996 	else								\
    997 		IF_DEQUEUE((ifq), (m));					\
    998 	mutex_exit((ifq)->ifq_lock);					\
    999 } while (/*CONSTCOND*/ 0)
   1000 
   1001 #define	IFQ_POLL(ifq, m)						\
   1002 do {									\
   1003 	mutex_enter((ifq)->ifq_lock);					\
   1004 	if (TBR_IS_ENABLED(ifq))					\
   1005 		(m) = tbr_dequeue((ifq), ALTDQ_POLL);			\
   1006 	else if (ALTQ_IS_ENABLED(ifq))					\
   1007 		ALTQ_POLL((ifq), (m));					\
   1008 	else								\
   1009 		IF_POLL((ifq), (m));					\
   1010 	mutex_exit((ifq)->ifq_lock);					\
   1011 } while (/*CONSTCOND*/ 0)
   1012 
   1013 #define	IFQ_PURGE(ifq)							\
   1014 do {									\
   1015 	mutex_enter((ifq)->ifq_lock);					\
   1016 	if (ALTQ_IS_ENABLED(ifq))					\
   1017 		ALTQ_PURGE(ifq);					\
   1018 	else								\
   1019 		IF_PURGE(ifq);						\
   1020 	mutex_exit((ifq)->ifq_lock);					\
   1021 } while (/*CONSTCOND*/ 0)
   1022 
   1023 #define	IFQ_SET_READY(ifq)						\
   1024 do {									\
   1025 	(ifq)->altq_flags |= ALTQF_READY;				\
   1026 } while (/*CONSTCOND*/ 0)
   1027 
   1028 #define	IFQ_CLASSIFY(ifq, m, af)					\
   1029 do {									\
   1030 	KASSERT(((m)->m_flags & M_PKTHDR) != 0);			\
   1031 	mutex_enter((ifq)->ifq_lock);					\
   1032 	if (ALTQ_IS_ENABLED(ifq)) {					\
   1033 		if (ALTQ_NEEDS_CLASSIFY(ifq))				\
   1034 			(m)->m_pkthdr.pattr_class = (*(ifq)->altq_classify) \
   1035 				((ifq)->altq_clfier, (m), (af));	\
   1036 		(m)->m_pkthdr.pattr_af = (af);				\
   1037 		(m)->m_pkthdr.pattr_hdr = mtod((m), void *);		\
   1038 	}								\
   1039 	mutex_exit((ifq)->ifq_lock);					\
   1040 } while (/*CONSTCOND*/ 0)
   1041 #else /* ! ALTQ */
   1042 #define	IFQ_ENQUEUE(ifq, m, err)					\
   1043 do {									\
   1044 	mutex_enter((ifq)->ifq_lock);					\
   1045 	if (IF_QFULL(ifq)) {						\
   1046 		m_freem(m);						\
   1047 		(err) = ENOBUFS;					\
   1048 	} else {							\
   1049 		IF_ENQUEUE((ifq), (m));					\
   1050 		(err) = 0;						\
   1051 	}								\
   1052 	if (err)							\
   1053 		(ifq)->ifq_drops++;					\
   1054 	mutex_exit((ifq)->ifq_lock);					\
   1055 } while (/*CONSTCOND*/ 0)
   1056 
   1057 #define	IFQ_DEQUEUE(ifq, m)						\
   1058 do {									\
   1059 	mutex_enter((ifq)->ifq_lock);					\
   1060 	IF_DEQUEUE((ifq), (m));						\
   1061 	mutex_exit((ifq)->ifq_lock);					\
   1062 } while (/*CONSTCOND*/ 0)
   1063 
   1064 #define	IFQ_POLL(ifq, m)						\
   1065 do {									\
   1066 	mutex_enter((ifq)->ifq_lock);					\
   1067 	IF_POLL((ifq), (m));						\
   1068 	mutex_exit((ifq)->ifq_lock);					\
   1069 } while (/*CONSTCOND*/ 0)
   1070 
   1071 #define	IFQ_PURGE(ifq)							\
   1072 do {									\
   1073 	mutex_enter((ifq)->ifq_lock);					\
   1074 	IF_PURGE(ifq);							\
   1075 	mutex_exit((ifq)->ifq_lock);					\
   1076 } while (/*CONSTCOND*/ 0)
   1077 
   1078 #define	IFQ_SET_READY(ifq)	/* nothing */
   1079 
   1080 #define	IFQ_CLASSIFY(ifq, m, af) /* nothing */
   1081 
   1082 #endif /* ALTQ */
   1083 
   1084 #define IFQ_LOCK_INIT(ifq)	(ifq)->ifq_lock =			\
   1085 	    mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET)
   1086 #define IFQ_LOCK_DESTROY(ifq)	mutex_obj_free((ifq)->ifq_lock)
   1087 #define IFQ_LOCK(ifq)		mutex_enter((ifq)->ifq_lock)
   1088 #define IFQ_UNLOCK(ifq)		mutex_exit((ifq)->ifq_lock)
   1089 
   1090 #define	IFQ_IS_EMPTY(ifq)		IF_IS_EMPTY(ifq)
   1091 #define	IFQ_INC_LEN(ifq)		((ifq)->ifq_len++)
   1092 #define	IFQ_DEC_LEN(ifq)		(--(ifq)->ifq_len)
   1093 #define	IFQ_INC_DROPS(ifq)		((ifq)->ifq_drops++)
   1094 #define	IFQ_SET_MAXLEN(ifq, len)	((ifq)->ifq_maxlen = (len))
   1095 
   1096 #define	IFQ_ENQUEUE_ISR(ifq, m, isr)					\
   1097 do {									\
   1098 	IFQ_LOCK(inq);							\
   1099 	if (IF_QFULL(inq)) {						\
   1100 		IF_DROP(inq);						\
   1101 		IFQ_UNLOCK(inq);					\
   1102 		m_freem(m);						\
   1103 	} else {							\
   1104 		IF_ENQUEUE(inq, m);					\
   1105 		IFQ_UNLOCK(inq);					\
   1106 		schednetisr(isr);					\
   1107 	}								\
   1108 } while (/*CONSTCOND*/ 0)
   1109 
   1110 #include <sys/mallocvar.h>
   1111 MALLOC_DECLARE(M_IFADDR);
   1112 MALLOC_DECLARE(M_IFMADDR);
   1113 
   1114 int ifreq_setaddr(u_long, struct ifreq *, const struct sockaddr *);
   1115 
   1116 struct ifnet *if_alloc(u_char);
   1117 void if_free(struct ifnet *);
   1118 void if_initname(struct ifnet *, const char *, int);
   1119 struct ifaddr *if_dl_create(const struct ifnet *, const struct sockaddr_dl **);
   1120 void if_activate_sadl(struct ifnet *, struct ifaddr *,
   1121     const struct sockaddr_dl *);
   1122 void	if_set_sadl(struct ifnet *, const void *, u_char, bool);
   1123 void	if_alloc_sadl(struct ifnet *);
   1124 void	if_free_sadl(struct ifnet *, int);
   1125 int	if_initialize(struct ifnet *);
   1126 void	if_register(struct ifnet *);
   1127 int	if_attach(struct ifnet *); /* Deprecated. Use if_initialize and if_register */
   1128 void	if_attachdomain(void);
   1129 void	if_deactivate(struct ifnet *);
   1130 bool	if_is_deactivated(const struct ifnet *);
   1131 void	if_export_if_data(struct ifnet *, struct if_data *, bool);
   1132 void	if_purgeaddrs(struct ifnet *, int, void (*)(struct ifaddr *));
   1133 void	if_detach(struct ifnet *);
   1134 void	if_down(struct ifnet *);
   1135 void	if_down_locked(struct ifnet *);
   1136 void	if_link_state_change(struct ifnet *, int);
   1137 void	if_domain_link_state_change(struct ifnet *, int);
   1138 void	if_up(struct ifnet *);
   1139 void	ifinit(void);
   1140 void	ifinit1(void);
   1141 void	ifinit_post(void);
   1142 int	ifaddrpref_ioctl(struct socket *, u_long, void *, struct ifnet *);
   1143 extern int (*ifioctl)(struct socket *, u_long, void *, struct lwp *);
   1144 int	ifioctl_common(struct ifnet *, u_long, void *);
   1145 int	ifpromisc(struct ifnet *, int);
   1146 int	ifpromisc_locked(struct ifnet *, int);
   1147 int	if_addr_init(ifnet_t *, struct ifaddr *, bool);
   1148 int	if_do_dad(struct ifnet *);
   1149 int	if_mcast_op(ifnet_t *, const unsigned long, const struct sockaddr *);
   1150 int	if_flags_set(struct ifnet *, const u_short);
   1151 int	if_clone_list(int, char *, int *);
   1152 
   1153 struct	ifnet *ifunit(const char *);
   1154 struct	ifnet *if_get(const char *, struct psref *);
   1155 ifnet_t *if_byindex(u_int);
   1156 ifnet_t *_if_byindex(u_int);
   1157 ifnet_t *if_get_byindex(u_int, struct psref *);
   1158 ifnet_t *if_get_bylla(const void *, unsigned char, struct psref *);
   1159 void	if_put(const struct ifnet *, struct psref *);
   1160 void	if_acquire(struct ifnet *, struct psref *);
   1161 #define	if_release	if_put
   1162 
   1163 int if_tunnel_check_nesting(struct ifnet *, struct mbuf *, int);
   1164 percpu_t *if_tunnel_alloc_ro_percpu(void);
   1165 void if_tunnel_free_ro_percpu(percpu_t *);
   1166 void if_tunnel_ro_percpu_rtcache_free(percpu_t *);
   1167 
   1168 struct tunnel_ro {
   1169 	struct route *tr_ro;
   1170 	kmutex_t *tr_lock;
   1171 };
   1172 
   1173 static inline void
   1174 if_tunnel_get_ro(percpu_t *ro_percpu, struct route **ro, kmutex_t **lock)
   1175 {
   1176 	struct tunnel_ro *tro;
   1177 
   1178 	tro = percpu_getref(ro_percpu);
   1179 	*ro = tro->tr_ro;
   1180 	*lock = tro->tr_lock;
   1181 	mutex_enter(*lock);
   1182 }
   1183 
   1184 static inline void
   1185 if_tunnel_put_ro(percpu_t *ro_percpu, kmutex_t *lock)
   1186 {
   1187 
   1188 	mutex_exit(lock);
   1189 	percpu_putref(ro_percpu);
   1190 }
   1191 
   1192 static __inline if_index_t
   1193 if_get_index(const struct ifnet *ifp)
   1194 {
   1195 
   1196 	return ifp != NULL ? ifp->if_index : 0;
   1197 }
   1198 
   1199 bool	if_held(struct ifnet *);
   1200 
   1201 void	if_input(struct ifnet *, struct mbuf *);
   1202 
   1203 struct if_percpuq *
   1204 	if_percpuq_create(struct ifnet *);
   1205 void	if_percpuq_destroy(struct if_percpuq *);
   1206 void
   1207 	if_percpuq_enqueue(struct if_percpuq *, struct mbuf *);
   1208 
   1209 void	if_deferred_start_init(struct ifnet *, void (*)(struct ifnet *));
   1210 void	if_schedule_deferred_start(struct ifnet *);
   1211 
   1212 void ifa_insert(struct ifnet *, struct ifaddr *);
   1213 void ifa_remove(struct ifnet *, struct ifaddr *);
   1214 
   1215 void	ifa_psref_init(struct ifaddr *);
   1216 void	ifa_acquire(struct ifaddr *, struct psref *);
   1217 void	ifa_release(struct ifaddr *, struct psref *);
   1218 bool	ifa_held(struct ifaddr *);
   1219 bool	ifa_is_destroying(struct ifaddr *);
   1220 
   1221 void	ifaref(struct ifaddr *);
   1222 void	ifafree(struct ifaddr *);
   1223 
   1224 struct	ifaddr *ifa_ifwithaddr(const struct sockaddr *);
   1225 struct	ifaddr *ifa_ifwithaddr_psref(const struct sockaddr *, struct psref *);
   1226 struct	ifaddr *ifa_ifwithaf(int);
   1227 struct	ifaddr *ifa_ifwithdstaddr(const struct sockaddr *);
   1228 struct	ifaddr *ifa_ifwithdstaddr_psref(const struct sockaddr *,
   1229 	    struct psref *);
   1230 struct	ifaddr *ifa_ifwithnet(const struct sockaddr *);
   1231 struct	ifaddr *ifa_ifwithnet_psref(const struct sockaddr *, struct psref *);
   1232 struct	ifaddr *ifa_ifwithladdr(const struct sockaddr *);
   1233 struct	ifaddr *ifa_ifwithladdr_psref(const struct sockaddr *, struct psref *);
   1234 struct	ifaddr *ifaof_ifpforaddr(const struct sockaddr *, struct ifnet *);
   1235 struct	ifaddr *ifaof_ifpforaddr_psref(const struct sockaddr *, struct ifnet *,
   1236 	    struct psref *);
   1237 void	link_rtrequest(int, struct rtentry *, const struct rt_addrinfo *);
   1238 void	p2p_rtrequest(int, struct rtentry *, const struct rt_addrinfo *);
   1239 
   1240 void	if_clone_attach(struct if_clone *);
   1241 void	if_clone_detach(struct if_clone *);
   1242 
   1243 int	if_transmit_lock(struct ifnet *, struct mbuf *);
   1244 
   1245 int	ifq_enqueue(struct ifnet *, struct mbuf *);
   1246 int	ifq_enqueue2(struct ifnet *, struct ifqueue *, struct mbuf *);
   1247 
   1248 int	loioctl(struct ifnet *, u_long, void *);
   1249 void	loopattach(int);
   1250 void	loopinit(void);
   1251 int	looutput(struct ifnet *,
   1252 	   struct mbuf *, const struct sockaddr *, const struct rtentry *);
   1253 
   1254 /*
   1255  * These are exported because they're an easy way to tell if
   1256  * an interface is going away without having to burn a flag.
   1257  */
   1258 int	if_nulloutput(struct ifnet *, struct mbuf *,
   1259 	    const struct sockaddr *, const struct rtentry *);
   1260 void	if_nullinput(struct ifnet *, struct mbuf *);
   1261 void	if_nullstart(struct ifnet *);
   1262 int	if_nulltransmit(struct ifnet *, struct mbuf *);
   1263 int	if_nullioctl(struct ifnet *, u_long, void *);
   1264 int	if_nullinit(struct ifnet *);
   1265 void	if_nullstop(struct ifnet *, int);
   1266 void	if_nullslowtimo(struct ifnet *);
   1267 #define	if_nullwatchdog	if_nullslowtimo
   1268 void	if_nulldrain(struct ifnet *);
   1269 #else
   1270 struct if_nameindex {
   1271 	unsigned int	if_index;	/* 1, 2, ... */
   1272 	char		*if_name;	/* null terminated name: "le0", ... */
   1273 };
   1274 
   1275 #include <sys/cdefs.h>
   1276 __BEGIN_DECLS
   1277 unsigned int if_nametoindex(const char *);
   1278 char *	if_indextoname(unsigned int, char *);
   1279 struct	if_nameindex * if_nameindex(void);
   1280 void	if_freenameindex(struct if_nameindex *);
   1281 __END_DECLS
   1282 #endif /* _KERNEL */ /* XXX really ALTQ? */
   1283 
   1284 #ifdef _KERNEL
   1285 
   1286 #define	IFADDR_FIRST(__ifp)		TAILQ_FIRST(&(__ifp)->if_addrlist)
   1287 #define	IFADDR_NEXT(__ifa)		TAILQ_NEXT((__ifa), ifa_list)
   1288 #define	IFADDR_FOREACH(__ifa, __ifp)	TAILQ_FOREACH(__ifa, \
   1289 					    &(__ifp)->if_addrlist, ifa_list)
   1290 #define	IFADDR_FOREACH_SAFE(__ifa, __ifp, __nifa) \
   1291 					    TAILQ_FOREACH_SAFE(__ifa, \
   1292 					    &(__ifp)->if_addrlist, ifa_list, __nifa)
   1293 #define	IFADDR_EMPTY(__ifp)		TAILQ_EMPTY(&(__ifp)->if_addrlist)
   1294 
   1295 #define IFADDR_ENTRY_INIT(__ifa)					\
   1296 	PSLIST_ENTRY_INIT((__ifa), ifa_pslist_entry)
   1297 #define IFADDR_ENTRY_DESTROY(__ifa)					\
   1298 	PSLIST_ENTRY_DESTROY((__ifa), ifa_pslist_entry)
   1299 #define IFADDR_READER_EMPTY(__ifp)					\
   1300 	(PSLIST_READER_FIRST(&(__ifp)->if_addr_pslist, struct ifaddr,	\
   1301 	                     ifa_pslist_entry) == NULL)
   1302 #define IFADDR_READER_FIRST(__ifp)					\
   1303 	PSLIST_READER_FIRST(&(__ifp)->if_addr_pslist, struct ifaddr,	\
   1304 	                    ifa_pslist_entry)
   1305 #define IFADDR_READER_NEXT(__ifa)					\
   1306 	PSLIST_READER_NEXT((__ifa), struct ifaddr, ifa_pslist_entry)
   1307 #define IFADDR_READER_FOREACH(__ifa, __ifp)				\
   1308 	PSLIST_READER_FOREACH((__ifa), &(__ifp)->if_addr_pslist, struct ifaddr,\
   1309 	                      ifa_pslist_entry)
   1310 #define IFADDR_WRITER_INSERT_HEAD(__ifp, __ifa)				\
   1311 	PSLIST_WRITER_INSERT_HEAD(&(__ifp)->if_addr_pslist, (__ifa),	\
   1312 	                          ifa_pslist_entry)
   1313 #define IFADDR_WRITER_REMOVE(__ifa)					\
   1314 	PSLIST_WRITER_REMOVE((__ifa), ifa_pslist_entry)
   1315 #define IFADDR_WRITER_FOREACH(__ifa, __ifp)				\
   1316 	PSLIST_WRITER_FOREACH((__ifa), &(__ifp)->if_addr_pslist, struct ifaddr,\
   1317 	                      ifa_pslist_entry)
   1318 #define IFADDR_WRITER_NEXT(__ifp)					\
   1319 	PSLIST_WRITER_NEXT((__ifp), struct ifaddr, ifa_pslist_entry)
   1320 #define IFADDR_WRITER_INSERT_AFTER(__ifp, __new)			\
   1321 	PSLIST_WRITER_INSERT_AFTER((__ifp), (__new), ifa_pslist_entry)
   1322 #define IFADDR_WRITER_EMPTY(__ifp)					\
   1323 	(PSLIST_WRITER_FIRST(&(__ifp)->if_addr_pslist, struct ifaddr,	\
   1324 	                     ifa_pslist_entry) == NULL)
   1325 #define IFADDR_WRITER_INSERT_TAIL(__ifp, __new)				\
   1326 	do {								\
   1327 		if (IFADDR_WRITER_EMPTY(__ifp)) {			\
   1328 			IFADDR_WRITER_INSERT_HEAD((__ifp), (__new));	\
   1329 		} else {						\
   1330 			struct ifaddr *__ifa;				\
   1331 			IFADDR_WRITER_FOREACH(__ifa, (__ifp)) {		\
   1332 				if (IFADDR_WRITER_NEXT(__ifa) == NULL) {\
   1333 					IFADDR_WRITER_INSERT_AFTER(__ifa,\
   1334 					    (__new));			\
   1335 					break;				\
   1336 				}					\
   1337 			}						\
   1338 		}							\
   1339 	} while (0)
   1340 
   1341 #define	IFNET_GLOBAL_LOCK()			mutex_enter(&ifnet_mtx)
   1342 #define	IFNET_GLOBAL_UNLOCK()			mutex_exit(&ifnet_mtx)
   1343 #define	IFNET_GLOBAL_LOCKED()			mutex_owned(&ifnet_mtx)
   1344 
   1345 #define IFNET_READER_EMPTY() \
   1346 	(PSLIST_READER_FIRST(&ifnet_pslist, struct ifnet, if_pslist_entry) == NULL)
   1347 #define IFNET_READER_FIRST() \
   1348 	PSLIST_READER_FIRST(&ifnet_pslist, struct ifnet, if_pslist_entry)
   1349 #define IFNET_READER_NEXT(__ifp) \
   1350 	PSLIST_READER_NEXT((__ifp), struct ifnet, if_pslist_entry)
   1351 #define IFNET_READER_FOREACH(__ifp) \
   1352 	PSLIST_READER_FOREACH((__ifp), &ifnet_pslist, struct ifnet, \
   1353 	                      if_pslist_entry)
   1354 #define IFNET_WRITER_INSERT_HEAD(__ifp) \
   1355 	PSLIST_WRITER_INSERT_HEAD(&ifnet_pslist, (__ifp), if_pslist_entry)
   1356 #define IFNET_WRITER_REMOVE(__ifp) \
   1357 	PSLIST_WRITER_REMOVE((__ifp), if_pslist_entry)
   1358 #define IFNET_WRITER_FOREACH(__ifp) \
   1359 	PSLIST_WRITER_FOREACH((__ifp), &ifnet_pslist, struct ifnet, \
   1360 	                      if_pslist_entry)
   1361 #define IFNET_WRITER_NEXT(__ifp) \
   1362 	PSLIST_WRITER_NEXT((__ifp), struct ifnet, if_pslist_entry)
   1363 #define IFNET_WRITER_INSERT_AFTER(__ifp, __new) \
   1364 	PSLIST_WRITER_INSERT_AFTER((__ifp), (__new), if_pslist_entry)
   1365 #define IFNET_WRITER_EMPTY() \
   1366 	(PSLIST_WRITER_FIRST(&ifnet_pslist, struct ifnet, if_pslist_entry) == NULL)
   1367 #define IFNET_WRITER_INSERT_TAIL(__new)					\
   1368 	do {								\
   1369 		if (IFNET_WRITER_EMPTY()) {				\
   1370 			IFNET_WRITER_INSERT_HEAD(__new);		\
   1371 		} else {						\
   1372 			struct ifnet *__ifp;				\
   1373 			IFNET_WRITER_FOREACH(__ifp) {			\
   1374 				if (IFNET_WRITER_NEXT(__ifp) == NULL) {	\
   1375 					IFNET_WRITER_INSERT_AFTER(__ifp,\
   1376 					    (__new));			\
   1377 					break;				\
   1378 				}					\
   1379 			}						\
   1380 		}							\
   1381 	} while (0)
   1382 
   1383 #define IFNET_LOCK(ifp)		mutex_enter((ifp)->if_ioctl_lock)
   1384 #define IFNET_UNLOCK(ifp)	mutex_exit((ifp)->if_ioctl_lock)
   1385 #define IFNET_LOCKED(ifp)	mutex_owned((ifp)->if_ioctl_lock)
   1386 
   1387 #define IFNET_ASSERT_UNLOCKED(ifp)	\
   1388 	KDASSERT(mutex_ownable((ifp)->if_ioctl_lock))
   1389 
   1390 extern struct pslist_head ifnet_pslist;
   1391 extern kmutex_t ifnet_mtx;
   1392 
   1393 extern struct ifnet *lo0ifp;
   1394 
   1395 /*
   1396  * ifq sysctl support
   1397  */
   1398 int	sysctl_ifq(int *name, u_int namelen, void *oldp,
   1399 		       size_t *oldlenp, void *newp, size_t newlen,
   1400 		       struct ifqueue *ifq);
   1401 /* symbolic names for terminal (per-protocol) CTL_IFQ_ nodes */
   1402 #define IFQCTL_LEN	1
   1403 #define IFQCTL_MAXLEN	2
   1404 #define IFQCTL_PEAK	3
   1405 #define IFQCTL_DROPS	4
   1406 
   1407 /*
   1408  * Hook for if_vlan - needed by if_agr
   1409  */
   1410 MODULE_HOOK(if_vlan_vlan_input_hook, void, (struct ifnet *, struct mbuf *));
   1411 
   1412 #endif /* _KERNEL */
   1413 
   1414 #endif /* !_NET_IF_H_ */
   1415