1/*-
  2 * SPDX-License-Identifier: BSD-3-Clause
  3 *
  4 * Copyright (c) 1982, 1986, 1990, 1993
  5 *	The Regents of the University of California.
  6 * Copyright (c) 2010-2011 Juniper Networks, Inc.
  7 * All rights reserved.
  8 *
  9 * Portions of this software were developed by Robert N. M. Watson under
 10 * contract to Juniper Networks, Inc.
 11 *
 12 * Redistribution and use in source and binary forms, with or without
 13 * modification, are permitted provided that the following conditions
 14 * are met:
 15 * 1. Redistributions of source code must retain the above copyright
 16 *    notice, this list of conditions and the following disclaimer.
 17 * 2. Redistributions in binary form must reproduce the above copyright
 18 *    notice, this list of conditions and the following disclaimer in the
 19 *    documentation and/or other materials provided with the distribution.
 20 * 3. Neither the name of the University nor the names of its contributors
 21 *    may be used to endorse or promote products derived from this software
 22 *    without specific prior written permission.
 23 *
 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 27 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 34 * SUCH DAMAGE.
 35 *
 36 *	@(#)in_pcb.h	8.1 (Berkeley) 6/10/93
 37 */
 38
 39#ifndef _NETINET_IN_PCB_H_
 40#define _NETINET_IN_PCB_H_
 41
 42#include <sys/queue.h>
 43#include <sys/epoch.h>
 44#include <sys/_lock.h>
 45#include <sys/_mutex.h>
 46#include <sys/_rwlock.h>
 47#include <sys/_smr.h>
 48#include <net/route.h>
 49
 50#ifdef _KERNEL
 51#include <sys/lock.h>
 52#include <sys/proc.h>
 53#include <sys/rwlock.h>
 54#include <sys/sysctl.h>
 55#include <net/vnet.h>
 56#include <vm/uma.h>
 57#endif
 58#include <sys/ck.h>
 59
 60/*
 61 * struct inpcb is the common protocol control block structure used in most
 62 * IP transport protocols.
 63 *
 64 * Pointers to local and foreign host table entries, local and foreign socket
 65 * numbers, and pointers up (to a socket structure) and down (to a
 66 * protocol-specific control block) are stored here.
 67 */
 68CK_LIST_HEAD(inpcbhead, inpcb);
 69CK_LIST_HEAD(inpcbporthead, inpcbport);
 70CK_LIST_HEAD(inpcblbgrouphead, inpcblbgroup);
 71typedef	uint64_t	inp_gen_t;
 72
 73/*
 74 * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet.
 75 * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing
 76 * the following structure.  This requires padding always be zeroed out,
 77 * which is done right after inpcb allocation and stays through its lifetime.
 78 */
 79struct in_addr_4in6 {
 80	u_int32_t	ia46_pad32[3];
 81	struct	in_addr	ia46_addr4;
 82};
 83
 84union in_dependaddr {
 85	struct in_addr_4in6 id46_addr;
 86	struct in6_addr	id6_addr;
 87};
 88
 89/*
 90 * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553.  in_conninfo has
 91 * some extra padding to accomplish this.
 92 * NOTE 2: tcp_syncache.c uses first 5 32-bit words, which identify fport,
 93 * lport, faddr to generate hash, so these fields shouldn't be moved.
 94 */
 95struct in_endpoints {
 96	u_int16_t	ie_fport;		/* foreign port */
 97	u_int16_t	ie_lport;		/* local port */
 98	/* protocol dependent part, local and foreign addr */
 99	union in_dependaddr ie_dependfaddr;	/* foreign host table entry */
100	union in_dependaddr ie_dependladdr;	/* local host table entry */
101#define	ie_faddr	ie_dependfaddr.id46_addr.ia46_addr4
102#define	ie_laddr	ie_dependladdr.id46_addr.ia46_addr4
103#define	ie6_faddr	ie_dependfaddr.id6_addr
104#define	ie6_laddr	ie_dependladdr.id6_addr
105	u_int32_t	ie6_zoneid;		/* scope zone id */
106};
107
108/*
109 * XXX The defines for inc_* are hacks and should be changed to direct
110 * references.
111 */
112struct in_conninfo {
113	u_int8_t	inc_flags;
114	u_int8_t	inc_len;
115	u_int16_t	inc_fibnum;	/* XXX was pad, 16 bits is plenty */
116	/* protocol dependent part */
117	struct	in_endpoints inc_ie;
118};
119
120/*
121 * Flags for inc_flags.
122 */
123#define	INC_ISIPV6	0x01
124#define	INC_IPV6MINMTU	0x02
125
126#define	inc_fport	inc_ie.ie_fport
127#define	inc_lport	inc_ie.ie_lport
128#define	inc_faddr	inc_ie.ie_faddr
129#define	inc_laddr	inc_ie.ie_laddr
130#define	inc6_faddr	inc_ie.ie6_faddr
131#define	inc6_laddr	inc_ie.ie6_laddr
132#define	inc6_zoneid	inc_ie.ie6_zoneid
133
134#if defined(_KERNEL) || defined(_WANT_INPCB)
135/*
136 * struct inpcb captures the network layer state for TCP, UDP, and raw IPv4 and
137 * IPv6 sockets.  In the case of TCP and UDP, further per-connection state is
138 * hung off of inp_ppcb most of the time.  Almost all fields of struct inpcb
139 * are static after creation or protected by a per-inpcb rwlock, inp_lock.
140 *
141 * A inpcb database is indexed by addresses/ports hash as well as list of
142 * all pcbs that belong to a certain proto. Database lookups or list traversals
143 * are be performed inside SMR section. Once desired PCB is found its own
144 * lock is to be obtained and SMR section exited.
145 *
146 * Key:
147 * (c) - Constant after initialization
148 * (e) - Protected by the SMR section
149 * (i) - Protected by the inpcb lock
150 * (p) - Protected by the pcbinfo lock for the inpcb
151 * (h) - Protected by the pcbhash lock for the inpcb
152 * (s) - Protected by another subsystem's locks
153 * (x) - Undefined locking
154 *
155 * A few other notes:
156 *
157 * When a read lock is held, stability of the field is guaranteed; to write
158 * to a field, a write lock must generally be held.
159 *
160 * netinet/netinet6-layer code should not assume that the inp_socket pointer
161 * is safe to dereference without inp_lock being held, there may be
162 * close(2)-related races.
163 *
164 * The inp_vflag field is overloaded, and would otherwise ideally be (c).
165 */
166struct icmp6_filter;
167struct inpcbpolicy;
168struct m_snd_tag;
169struct inpcb {
170	/* Cache line #1 (amd64) */
171	CK_LIST_ENTRY(inpcb) inp_hash_exact;	/* hash table linkage */
172	CK_LIST_ENTRY(inpcb) inp_hash_wild;	/* hash table linkage */
173	struct rwlock	inp_lock;
174	/* Cache line #2 (amd64) */
175#define	inp_start_zero	inp_refcount
176#define	inp_zero_size	(sizeof(struct inpcb) - \
177			    offsetof(struct inpcb, inp_start_zero))
178	u_int	inp_refcount;		/* (i) refcount */
179	int	inp_flags;		/* (i) generic IP/datagram flags */
180	int	inp_flags2;		/* (i) generic IP/datagram flags #2*/
181	uint8_t inp_numa_domain;	/* numa domain */
182	void	*inp_ppcb;		/* (i) pointer to per-protocol pcb */
183	struct	socket *inp_socket;	/* (i) back pointer to socket */
184	struct	inpcbinfo *inp_pcbinfo;	/* (c) PCB list info */
185	struct	ucred	*inp_cred;	/* (c) cache of socket cred */
186	u_int32_t inp_flow;		/* (i) IPv6 flow information */
187	u_char	inp_vflag;		/* (i) IP version flag (v4/v6) */
188	u_char	inp_ip_ttl;		/* (i) time to live proto */
189	u_char	inp_ip_p;		/* (c) protocol proto */
190	u_char	inp_ip_minttl;		/* (i) minimum TTL or drop */
191	uint32_t inp_flowid;		/* (x) flow id / queue id */
192	smr_seq_t inp_smr;		/* (i) sequence number at disconnect */
193	struct m_snd_tag *inp_snd_tag;	/* (i) send tag for outgoing mbufs */
194	uint32_t inp_flowtype;		/* (x) M_HASHTYPE value */
195
196	/* Local and foreign ports, local and foreign addr. */
197	struct	in_conninfo inp_inc;	/* (i,h) list for PCB's local port */
198
199	/* MAC and IPSEC policy information. */
200	struct	label *inp_label;	/* (i) MAC label */
201	struct	inpcbpolicy *inp_sp;    /* (s) for IPSEC */
202
203	/* Protocol-dependent part; options. */
204	struct {
205		u_char	inp_ip_tos;		/* (i) type of service proto */
206		struct mbuf		*inp_options;	/* (i) IP options */
207		struct ip_moptions	*inp_moptions;	/* (i) mcast options */
208	};
209	struct {
210		/* (i) IP options */
211		struct mbuf		*in6p_options;
212		/* (i) IP6 options for outgoing packets */
213		struct ip6_pktopts	*in6p_outputopts;
214		/* (i) IP multicast options */
215		struct ip6_moptions	*in6p_moptions;
216		/* (i) ICMPv6 code type filter */
217		struct icmp6_filter	*in6p_icmp6filt;
218		/* (i) IPV6_CHECKSUM setsockopt */
219		int	in6p_cksum;
220		short	in6p_hops;
221	};
222	CK_LIST_ENTRY(inpcb) inp_portlist;	/* (r:e/w:h) port list */
223	struct	inpcbport *inp_phd;	/* (r:e/w:h) head of this list */
224	inp_gen_t	inp_gencnt;	/* (c) generation count */
225	void		*spare_ptr;	/* Spare pointer. */
226	rt_gen_t	inp_rt_cookie;	/* generation for route entry */
227	union {				/* cached L3 information */
228		struct route inp_route;
229		struct route_in6 inp_route6;
230	};
231	CK_LIST_ENTRY(inpcb) inp_list;	/* (r:e/w:p) all PCBs for proto */
232};
233#endif	/* _KERNEL */
234
235#define	inp_fport	inp_inc.inc_fport
236#define	inp_lport	inp_inc.inc_lport
237#define	inp_faddr	inp_inc.inc_faddr
238#define	inp_laddr	inp_inc.inc_laddr
239
240#define	in6p_faddr	inp_inc.inc6_faddr
241#define	in6p_laddr	inp_inc.inc6_laddr
242#define	in6p_zoneid	inp_inc.inc6_zoneid
243
244#define	inp_vnet	inp_pcbinfo->ipi_vnet
245
246/*
247 * The range of the generation count, as used in this implementation, is 9e19.
248 * We would have to create 300 billion connections per second for this number
249 * to roll over in a year.  This seems sufficiently unlikely that we simply
250 * don't concern ourselves with that possibility.
251 */
252
253/*
254 * Interface exported to userland by various protocols which use inpcbs.  Hack
255 * alert -- only define if struct xsocket is in scope.
256 * Fields prefixed with "xi_" are unique to this structure, and the rest
257 * match fields in the struct inpcb, to ease coding and porting.
258 *
259 * Legend:
260 * (s) - used by userland utilities in src
261 * (p) - used by utilities in ports
262 * (3) - is known to be used by third party software not in ports
263 * (n) - no known usage
264 */
265#ifdef _SYS_SOCKETVAR_H_
266struct xinpcb {
267	ksize_t		xi_len;			/* length of this structure */
268	struct xsocket	xi_socket;		/* (s,p) */
269	struct in_conninfo inp_inc;		/* (s,p) */
270	uint64_t	inp_gencnt;		/* (s,p) */
271	kvaddr_t	inp_ppcb;		/* (s) netstat(1) */
272	int64_t		inp_spare64[4];
273	uint32_t	inp_flow;		/* (s) */
274	uint32_t	inp_flowid;		/* (s) */
275	uint32_t	inp_flowtype;		/* (s) */
276	int32_t		inp_flags;		/* (s,p) */
277	int32_t		inp_flags2;		/* (s) */
278	uint32_t	inp_unused;
279	int32_t		in6p_cksum;		/* (n) */
280	int32_t		inp_spare32[4];
281	uint16_t	in6p_hops;		/* (n) */
282	uint8_t		inp_ip_tos;		/* (n) */
283	int8_t		pad8;
284	uint8_t		inp_vflag;		/* (s,p) */
285	uint8_t		inp_ip_ttl;		/* (n) */
286	uint8_t		inp_ip_p;		/* (n) */
287	uint8_t		inp_ip_minttl;		/* (n) */
288	int8_t		inp_spare8[4];
289} __aligned(8);
290
291struct xinpgen {
292	ksize_t	xig_len;	/* length of this structure */
293	u_int		xig_count;	/* number of PCBs at this time */
294	uint32_t	_xig_spare32;
295	inp_gen_t	xig_gen;	/* generation count at this time */
296	so_gen_t	xig_sogen;	/* socket generation count this time */
297	uint64_t	_xig_spare64[4];
298} __aligned(8);
299
300struct sockopt_parameters {
301	struct in_conninfo sop_inc;
302	uint64_t sop_id;
303	int sop_level;
304	int sop_optname;
305	char sop_optval[];
306};
307
308#ifdef	_KERNEL
309int	sysctl_setsockopt(SYSCTL_HANDLER_ARGS, struct inpcbinfo *pcbinfo,
310	    int (*ctloutput_set)(struct inpcb *, struct sockopt *));
311void	in_pcbtoxinpcb(const struct inpcb *, struct xinpcb *);
312#endif
313#endif /* _SYS_SOCKETVAR_H_ */
314
315#ifdef _KERNEL
316/*
317 * Per-VNET pcb database for each high-level protocol (UDP, TCP, ...) in both
318 * IPv4 and IPv6.
319 *
320 * The pcbs are protected with SMR section and thus all lists in inpcbinfo
321 * are CK-lists.  Locking is required to insert a pcb into database. Two
322 * locks are provided: one for the hash and one for the global list of pcbs,
323 * as well as overall count and generation count.
324 *
325 * Locking key:
326 *
327 * (c) Constant or nearly constant after initialisation
328 * (e) Protected by SMR section
329 * (g) Locked by ipi_lock
330 * (h) Locked by ipi_hash_lock
331 */
332struct inpcbinfo {
333	/*
334	 * Global lock protecting inpcb list modification
335	 */
336	struct mtx		 ipi_lock;
337	struct inpcbhead	 ipi_listhead;		/* (r:e/w:g) */
338	u_int			 ipi_count;		/* (g) */
339
340	/*
341	 * Generation count -- incremented each time a connection is allocated
342	 * or freed.
343	 */
344	u_quad_t		 ipi_gencnt;		/* (g) */
345
346	/*
347	 * Fields associated with port lookup and allocation.
348	 */
349	u_short			 ipi_lastport;		/* (h) */
350	u_short			 ipi_lastlow;		/* (h) */
351	u_short			 ipi_lasthi;		/* (h) */
352
353	/*
354	 * UMA zone from which inpcbs are allocated for this protocol.
355	 */
356	uma_zone_t		 ipi_zone;		/* (c) */
357	uma_zone_t		 ipi_portzone;		/* (c) */
358	smr_t			 ipi_smr;		/* (c) */
359
360	/*
361	 * Global hash of inpcbs, hashed by local and foreign addresses and
362	 * port numbers.  The "exact" hash holds PCBs connected to a foreign
363	 * address, and "wild" holds the rest.
364	 */
365	struct mtx		 ipi_hash_lock;
366	struct inpcbhead 	*ipi_hash_exact;	/* (r:e/w:h) */
367	struct inpcbhead 	*ipi_hash_wild;		/* (r:e/w:h) */
368	u_long			 ipi_hashmask;		/* (c) */
369
370	/*
371	 * Global hash of inpcbs, hashed by only local port number.
372	 */
373	struct inpcbporthead	*ipi_porthashbase;	/* (h) */
374	u_long			 ipi_porthashmask;	/* (h) */
375
376	/*
377	 * Load balance groups used for the SO_REUSEPORT_LB option,
378	 * hashed by local port.
379	 */
380	struct	inpcblbgrouphead *ipi_lbgrouphashbase;	/* (r:e/w:h) */
381	u_long			 ipi_lbgrouphashmask;	/* (h) */
382
383	/*
384	 * Pointer to network stack instance
385	 */
386	struct vnet		*ipi_vnet;		/* (c) */
387};
388
389/*
390 * Global allocation storage for each high-level protocol (UDP, TCP, ...).
391 * Each corresponding per-VNET inpcbinfo points into this one.
392 */
393struct inpcbstorage {
394	uma_zone_t	ips_zone;
395	uma_zone_t	ips_portzone;
396	uma_init	ips_pcbinit;
397	size_t		ips_size;
398	const char *	ips_zone_name;
399	const char *	ips_portzone_name;
400	const char *	ips_infolock_name;
401	const char *	ips_hashlock_name;
402};
403
404#define INPCBSTORAGE_DEFINE(prot, ppcb, lname, zname, iname, hname)	\
405static int								\
406prot##_inpcb_init(void *mem, int size __unused, int flags __unused)	\
407{									\
408	struct inpcb *inp = mem;					\
409									\
410	rw_init_flags(&inp->inp_lock, lname, RW_RECURSE | RW_DUPOK);	\
411	return (0);							\
412}									\
413static struct inpcbstorage prot = {					\
414	.ips_size = sizeof(struct ppcb),				\
415	.ips_pcbinit = prot##_inpcb_init,				\
416	.ips_zone_name = zname,						\
417	.ips_portzone_name = zname " ports",				\
418	.ips_infolock_name = iname,					\
419	.ips_hashlock_name = hname,					\
420};									\
421SYSINIT(prot##_inpcbstorage_init, SI_SUB_PROTO_DOMAIN,			\
422    SI_ORDER_SECOND, in_pcbstorage_init, &prot);			\
423SYSUNINIT(prot##_inpcbstorage_uninit, SI_SUB_PROTO_DOMAIN,		\
424    SI_ORDER_SECOND, in_pcbstorage_destroy, &prot)
425
426/*
427 * Load balance groups used for the SO_REUSEPORT_LB socket option. Each group
428 * (or unique address:port combination) can be re-used at most
429 * INPCBLBGROUP_SIZMAX (256) times. The inpcbs are stored in il_inp which
430 * is dynamically resized as processes bind/unbind to that specific group.
431 */
432struct inpcblbgroup {
433	CK_LIST_ENTRY(inpcblbgroup) il_list;
434	struct epoch_context il_epoch_ctx;
435	struct ucred	*il_cred;
436	uint16_t	il_lport;			/* (c) */
437	u_char		il_vflag;			/* (c) */
438	uint8_t		il_numa_domain;
439	uint32_t	il_pad2;
440	union in_dependaddr il_dependladdr;		/* (c) */
441#define	il_laddr	il_dependladdr.id46_addr.ia46_addr4
442#define	il6_laddr	il_dependladdr.id6_addr
443	uint32_t	il_inpsiz; /* max count in il_inp[] (h) */
444	uint32_t	il_inpcnt; /* cur count in il_inp[] (h) */
445	struct inpcb	*il_inp[];			/* (h) */
446};
447
448#define INP_LOCK_DESTROY(inp)	rw_destroy(&(inp)->inp_lock)
449#define INP_RLOCK(inp)		rw_rlock(&(inp)->inp_lock)
450#define INP_WLOCK(inp)		rw_wlock(&(inp)->inp_lock)
451#define INP_TRY_RLOCK(inp)	rw_try_rlock(&(inp)->inp_lock)
452#define INP_TRY_WLOCK(inp)	rw_try_wlock(&(inp)->inp_lock)
453#define INP_RUNLOCK(inp)	rw_runlock(&(inp)->inp_lock)
454#define INP_WUNLOCK(inp)	rw_wunlock(&(inp)->inp_lock)
455#define INP_UNLOCK(inp)		rw_unlock(&(inp)->inp_lock)
456#define	INP_TRY_UPGRADE(inp)	rw_try_upgrade(&(inp)->inp_lock)
457#define	INP_DOWNGRADE(inp)	rw_downgrade(&(inp)->inp_lock)
458#define	INP_WLOCKED(inp)	rw_wowned(&(inp)->inp_lock)
459#define	INP_LOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_LOCKED)
460#define	INP_RLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_RLOCKED)
461#define	INP_WLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_WLOCKED)
462#define	INP_UNLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_UNLOCKED)
463
464/*
465 * These locking functions are for inpcb consumers outside of sys/netinet,
466 * more specifically, they were added for the benefit of TOE drivers. The
467 * macros are reserved for use by the stack.
468 */
469void inp_wlock(struct inpcb *);
470void inp_wunlock(struct inpcb *);
471void inp_rlock(struct inpcb *);
472void inp_runlock(struct inpcb *);
473
474#ifdef INVARIANT_SUPPORT
475void inp_lock_assert(struct inpcb *);
476void inp_unlock_assert(struct inpcb *);
477#else
478#define	inp_lock_assert(inp)	do {} while (0)
479#define	inp_unlock_assert(inp)	do {} while (0)
480#endif
481
482void	inp_apply_all(struct inpcbinfo *, void (*func)(struct inpcb *, void *),
483	    void *arg);
484int 	inp_ip_tos_get(const struct inpcb *inp);
485void 	inp_ip_tos_set(struct inpcb *inp, int val);
486struct socket *
487	inp_inpcbtosocket(struct inpcb *inp);
488struct tcpcb *
489	inp_inpcbtotcpcb(struct inpcb *inp);
490void 	inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
491		uint32_t *faddr, uint16_t *fp);
492
493#endif /* _KERNEL */
494
495#define INP_INFO_WLOCK(ipi)	mtx_lock(&(ipi)->ipi_lock)
496#define INP_INFO_WLOCKED(ipi)	mtx_owned(&(ipi)->ipi_lock)
497#define INP_INFO_WUNLOCK(ipi)	mtx_unlock(&(ipi)->ipi_lock)
498#define	INP_INFO_LOCK_ASSERT(ipi)	MPASS(SMR_ENTERED((ipi)->ipi_smr) || \
499					mtx_owned(&(ipi)->ipi_lock))
500#define INP_INFO_WLOCK_ASSERT(ipi)	mtx_assert(&(ipi)->ipi_lock, MA_OWNED)
501#define INP_INFO_WUNLOCK_ASSERT(ipi)	\
502				mtx_assert(&(ipi)->ipi_lock, MA_NOTOWNED)
503
504#define	INP_HASH_WLOCK(ipi)		mtx_lock(&(ipi)->ipi_hash_lock)
505#define	INP_HASH_WUNLOCK(ipi)		mtx_unlock(&(ipi)->ipi_hash_lock)
506#define	INP_HASH_LOCK_ASSERT(ipi)	MPASS(SMR_ENTERED((ipi)->ipi_smr) || \
507					mtx_owned(&(ipi)->ipi_hash_lock))
508#define	INP_HASH_WLOCK_ASSERT(ipi)	mtx_assert(&(ipi)->ipi_hash_lock, \
509					MA_OWNED)
510
511/*
512 * Wildcard matching hash is not just a microoptimisation!  The hash for
513 * wildcard IPv4 and wildcard IPv6 must be the same, otherwise AF_INET6
514 * wildcard bound pcb won't be able to receive AF_INET connections, while:
515 * jenkins_hash(&zeroes, 1, s) != jenkins_hash(&zeroes, 4, s)
516 * See also comment above struct in_addr_4in6.
517 */
518#define	IN_ADDR_JHASH32(addr)						\
519	((addr)->s_addr == INADDR_ANY ? V_in_pcbhashseed :		\
520	    jenkins_hash32((&(addr)->s_addr), 1, V_in_pcbhashseed))
521#define	IN6_ADDR_JHASH32(addr)						\
522	(memcmp((addr), &in6addr_any, sizeof(in6addr_any)) == 0 ?	\
523	    V_in_pcbhashseed :						\
524	    jenkins_hash32((addr)->__u6_addr.__u6_addr32,		\
525	    nitems((addr)->__u6_addr.__u6_addr32), V_in_pcbhashseed))
526
527#define INP_PCBHASH(faddr, lport, fport, mask)				\
528	((IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask))
529#define	INP6_PCBHASH(faddr, lport, fport, mask)				\
530	((IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask))
531
532#define	INP_PCBHASH_WILD(lport, mask)					\
533	((V_in_pcbhashseed ^ ntohs(lport)) & (mask))
534
535#define	INP_PCBLBGROUP_PKTHASH(faddr, lport, fport)			\
536	(IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport)))
537#define	INP6_PCBLBGROUP_PKTHASH(faddr, lport, fport)			\
538	(IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport)))
539
540#define INP_PCBPORTHASH(lport, mask)	(ntohs((lport)) & (mask))
541
542/*
543 * Flags for inp_vflags -- historically version flags only
544 */
545#define	INP_IPV4	0x1
546#define	INP_IPV6	0x2
547#define	INP_IPV6PROTO	0x4		/* opened under IPv6 protocol */
548
549/*
550 * Flags for inp_flags.
551 */
552#define	INP_RECVOPTS		0x00000001 /* receive incoming IP options */
553#define	INP_RECVRETOPTS		0x00000002 /* receive IP options for reply */
554#define	INP_RECVDSTADDR		0x00000004 /* receive IP dst address */
555#define	INP_HDRINCL		0x00000008 /* user supplies entire IP header */
556#define	INP_HIGHPORT		0x00000010 /* user wants "high" port binding */
557#define	INP_LOWPORT		0x00000020 /* user wants "low" port binding */
558#define	INP_ANONPORT		0x00000040 /* read by netstat(1) */
559#define	INP_RECVIF		0x00000080 /* receive incoming interface */
560#define	INP_MTUDISC		0x00000100 /* user can do MTU discovery */
561/*	INP_FREED		0x00000200 private to in_pcb.c */
562#define	INP_RECVTTL		0x00000400 /* receive incoming IP TTL */
563#define	INP_DONTFRAG		0x00000800 /* don't fragment packet */
564#define	INP_BINDANY		0x00001000 /* allow bind to any address */
565#define	INP_INHASHLIST		0x00002000 /* in_pcbinshash() has been called */
566#define	INP_RECVTOS		0x00004000 /* receive incoming IP TOS */
567#define	IN6P_IPV6_V6ONLY	0x00008000 /* restrict AF_INET6 socket for v6 */
568#define	IN6P_PKTINFO		0x00010000 /* receive IP6 dst and I/F */
569#define	IN6P_HOPLIMIT		0x00020000 /* receive hoplimit */
570#define	IN6P_HOPOPTS		0x00040000 /* receive hop-by-hop options */
571#define	IN6P_DSTOPTS		0x00080000 /* receive dst options after rthdr */
572#define	IN6P_RTHDR		0x00100000 /* receive routing header */
573#define	IN6P_RTHDRDSTOPTS	0x00200000 /* receive dstoptions before rthdr */
574#define	IN6P_TCLASS		0x00400000 /* receive traffic class value */
575#define	IN6P_AUTOFLOWLABEL	0x00800000 /* attach flowlabel automatically */
576/*	INP_INLBGROUP		0x01000000 private to in_pcb.c */
577#define	INP_ONESBCAST		0x02000000 /* send all-ones broadcast */
578#define	INP_DROPPED		0x04000000 /* protocol drop flag */
579#define	INP_SOCKREF		0x08000000 /* strong socket reference */
580#define	INP_RESERVED_0          0x10000000 /* reserved field */
581#define	INP_RESERVED_1          0x20000000 /* reserved field */
582#define	IN6P_RFC2292		0x40000000 /* used RFC2292 API on the socket */
583#define	IN6P_MTU		0x80000000 /* receive path MTU */
584
585#define	INP_CONTROLOPTS		(INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\
586				 INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\
587				 IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\
588				 IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\
589				 IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\
590				 IN6P_MTU)
591
592/*
593 * Flags for inp_flags2.
594 */
595/*				0x00000001 */
596/*				0x00000002 */
597/*				0x00000004 */
598/*				0x00000008 */
599/*				0x00000010 */
600/*				0x00000020 */
601/*				0x00000040 */
602/*				0x00000080 */
603#define	INP_RECVFLOWID		0x00000100 /* populate recv datagram with flow info */
604#define	INP_RECVRSSBUCKETID	0x00000200 /* populate recv datagram with bucket id */
605#define	INP_RATE_LIMIT_CHANGED	0x00000400 /* rate limit needs attention */
606#define	INP_ORIGDSTADDR		0x00000800 /* receive IP dst address/port */
607/*				0x00001000 */
608/*				0x00002000 */
609/*				0x00004000 */
610/*				0x00008000 */
611/*				0x00010000 */
612#define INP_2PCP_SET		0x00020000 /* If the Eth PCP should be set explicitly */
613#define INP_2PCP_BIT0		0x00040000 /* Eth PCP Bit 0 */
614#define INP_2PCP_BIT1		0x00080000 /* Eth PCP Bit 1 */
615#define INP_2PCP_BIT2		0x00100000 /* Eth PCP Bit 2 */
616#define INP_2PCP_BASE	INP_2PCP_BIT0
617#define INP_2PCP_MASK	(INP_2PCP_BIT0 | INP_2PCP_BIT1 | INP_2PCP_BIT2)
618#define INP_2PCP_SHIFT		18         /* shift PCP field in/out of inp_flags2 */
619
620/*
621 * Flags passed to in_pcblookup*(), inp_smr_lock() and inp_next().
622 */
623typedef	enum {
624	INPLOOKUP_WILDCARD = 0x00000001,	/* Allow wildcard sockets. */
625	INPLOOKUP_RLOCKPCB = 0x00000002,	/* Return inpcb read-locked. */
626	INPLOOKUP_WLOCKPCB = 0x00000004,	/* Return inpcb write-locked. */
627} inp_lookup_t;
628
629#define	INPLOOKUP_MASK	(INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \
630	    INPLOOKUP_WLOCKPCB)
631#define	INPLOOKUP_LOCKMASK	(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)
632
633#define	sotoinpcb(so)	((struct inpcb *)(so)->so_pcb)
634
635#define	INP_SOCKAF(so) so->so_proto->pr_domain->dom_family
636
637#define	INP_CHECK_SOCKAF(so, af)	(INP_SOCKAF(so) == af)
638
639#ifdef _KERNEL
640VNET_DECLARE(int, ipport_reservedhigh);
641VNET_DECLARE(int, ipport_reservedlow);
642VNET_DECLARE(int, ipport_lowfirstauto);
643VNET_DECLARE(int, ipport_lowlastauto);
644VNET_DECLARE(int, ipport_firstauto);
645VNET_DECLARE(int, ipport_lastauto);
646VNET_DECLARE(int, ipport_hifirstauto);
647VNET_DECLARE(int, ipport_hilastauto);
648VNET_DECLARE(int, ipport_randomized);
649
650#define	V_ipport_reservedhigh	VNET(ipport_reservedhigh)
651#define	V_ipport_reservedlow	VNET(ipport_reservedlow)
652#define	V_ipport_lowfirstauto	VNET(ipport_lowfirstauto)
653#define	V_ipport_lowlastauto	VNET(ipport_lowlastauto)
654#define	V_ipport_firstauto	VNET(ipport_firstauto)
655#define	V_ipport_lastauto	VNET(ipport_lastauto)
656#define	V_ipport_hifirstauto	VNET(ipport_hifirstauto)
657#define	V_ipport_hilastauto	VNET(ipport_hilastauto)
658#define	V_ipport_randomized	VNET(ipport_randomized)
659
660void	in_pcbinfo_init(struct inpcbinfo *, struct inpcbstorage *,
661	    u_int, u_int);
662void	in_pcbinfo_destroy(struct inpcbinfo *);
663void	in_pcbstorage_init(void *);
664void	in_pcbstorage_destroy(void *);
665
666void	in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *);
667int	in_pcballoc(struct socket *, struct inpcbinfo *);
668int	in_pcbbind(struct inpcb *, struct sockaddr_in *, struct ucred *);
669int	in_pcbbind_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *,
670	    u_short *, struct ucred *);
671int	in_pcbconnect(struct inpcb *, struct sockaddr_in *, struct ucred *,
672	    bool);
673int	in_pcbconnect_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *,
674	    u_short *, in_addr_t *, u_short *, struct ucred *);
675void	in_pcbdisconnect(struct inpcb *);
676void	in_pcbdrop(struct inpcb *);
677void	in_pcbfree(struct inpcb *);
678int	in_pcbinshash(struct inpcb *);
679int	in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *,
680	    struct ucred *);
681int	in_pcblbgroup_numa(struct inpcb *, int arg);
682struct inpcb *
683	in_pcblookup(struct inpcbinfo *, struct in_addr, u_int,
684	    struct in_addr, u_int, int, struct ifnet *);
685struct inpcb *
686	in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int,
687	    struct in_addr, u_int, int, struct ifnet *, struct mbuf *);
688void	in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr,
689	    int, struct inpcb *(*)(struct inpcb *, int));
690void	in_pcbref(struct inpcb *);
691void	in_pcbrehash(struct inpcb *);
692void	in_pcbremhash_locked(struct inpcb *);
693bool	in_pcbrele(struct inpcb *, inp_lookup_t);
694bool	in_pcbrele_rlocked(struct inpcb *);
695bool	in_pcbrele_wlocked(struct inpcb *);
696
697typedef bool inp_match_t(const struct inpcb *, void *);
698struct inpcb_iterator {
699	const struct inpcbinfo	*ipi;
700	struct inpcb		*inp;
701	inp_match_t		*match;
702	void			*ctx;
703	int			hash;
704#define	INP_ALL_LIST		-1
705	const inp_lookup_t	lock;
706};
707
708/* Note: sparse initializers guarantee .inp = NULL. */
709#define	INP_ITERATOR(_ipi, _lock, _match, _ctx)		\
710	{						\
711		.ipi = (_ipi),				\
712		.lock = (_lock),			\
713		.hash = INP_ALL_LIST,			\
714		.match = (_match),			\
715		.ctx = (_ctx),				\
716	}
717#define	INP_ALL_ITERATOR(_ipi, _lock)			\
718	{						\
719		.ipi = (_ipi),				\
720		.lock = (_lock),			\
721		.hash = INP_ALL_LIST,			\
722	}
723
724struct inpcb *inp_next(struct inpcb_iterator *);
725void	in_losing(struct inpcb *);
726void	in_pcbsetsolabel(struct socket *so);
727int	in_getpeeraddr(struct socket *so, struct sockaddr **nam);
728int	in_getsockaddr(struct socket *so, struct sockaddr **nam);
729struct sockaddr *
730	in_sockaddr(in_port_t port, struct in_addr *addr);
731void	in_pcbsosetlabel(struct socket *so);
732#ifdef RATELIMIT
733int
734in_pcboutput_txrtlmt_locked(struct inpcb *, struct ifnet *,
735	    struct mbuf *, uint32_t);
736int	in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t,
737	    uint32_t, struct m_snd_tag **);
738void	in_pcbdetach_txrtlmt(struct inpcb *);
739void    in_pcbdetach_tag(struct m_snd_tag *);
740int	in_pcbmodify_txrtlmt(struct inpcb *, uint32_t);
741int	in_pcbquery_txrtlmt(struct inpcb *, uint32_t *);
742int	in_pcbquery_txrlevel(struct inpcb *, uint32_t *);
743void	in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *);
744void	in_pcboutput_eagain(struct inpcb *);
745#endif
746#endif /* _KERNEL */
747
748#endif /* !_NETINET_IN_PCB_H_ */