xref: /freebsd/sys/netinet/in_pcb.h (revision bdcbfde31e8e9b343f113a1956384bdf30d1ed62)
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 
37 #ifndef _NETINET_IN_PCB_H_
38 #define _NETINET_IN_PCB_H_
39 
40 #include <sys/queue.h>
41 #include <sys/epoch.h>
42 #include <sys/_lock.h>
43 #include <sys/_mutex.h>
44 #include <sys/_rwlock.h>
45 #include <sys/_smr.h>
46 #include <net/route.h>
47 
48 #ifdef _KERNEL
49 #include <sys/lock.h>
50 #include <sys/proc.h>
51 #include <sys/rwlock.h>
52 #include <sys/sysctl.h>
53 #include <net/vnet.h>
54 #include <vm/uma.h>
55 #endif
56 #include <sys/ck.h>
57 
58 /*
59  * struct inpcb is the common protocol control block structure used in most
60  * IP transport protocols.
61  *
62  * Pointers to local and foreign host table entries, local and foreign socket
63  * numbers, and pointers up (to a socket structure) and down (to a
64  * protocol-specific control block) are stored here.
65  */
66 CK_LIST_HEAD(inpcbhead, inpcb);
67 CK_LIST_HEAD(inpcbporthead, inpcbport);
68 CK_LIST_HEAD(inpcblbgrouphead, inpcblbgroup);
69 typedef	uint64_t	inp_gen_t;
70 
71 /*
72  * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet.
73  * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing
74  * the following structure.  This requires padding always be zeroed out,
75  * which is done right after inpcb allocation and stays through its lifetime.
76  */
77 struct in_addr_4in6 {
78 	u_int32_t	ia46_pad32[3];
79 	struct	in_addr	ia46_addr4;
80 };
81 
82 union in_dependaddr {
83 	struct in_addr_4in6 id46_addr;
84 	struct in6_addr	id6_addr;
85 };
86 
87 /*
88  * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553.  in_conninfo has
89  * some extra padding to accomplish this.
90  * NOTE 2: tcp_syncache.c uses first 5 32-bit words, which identify fport,
91  * lport, faddr to generate hash, so these fields shouldn't be moved.
92  */
93 struct in_endpoints {
94 	u_int16_t	ie_fport;		/* foreign port */
95 	u_int16_t	ie_lport;		/* local port */
96 	/* protocol dependent part, local and foreign addr */
97 	union in_dependaddr ie_dependfaddr;	/* foreign host table entry */
98 	union in_dependaddr ie_dependladdr;	/* local host table entry */
99 #define	ie_faddr	ie_dependfaddr.id46_addr.ia46_addr4
100 #define	ie_laddr	ie_dependladdr.id46_addr.ia46_addr4
101 #define	ie6_faddr	ie_dependfaddr.id6_addr
102 #define	ie6_laddr	ie_dependladdr.id6_addr
103 	u_int32_t	ie6_zoneid;		/* scope zone id */
104 };
105 
106 /*
107  * XXX The defines for inc_* are hacks and should be changed to direct
108  * references.
109  */
110 struct in_conninfo {
111 	u_int8_t	inc_flags;
112 	u_int8_t	inc_len;
113 	u_int16_t	inc_fibnum;	/* XXX was pad, 16 bits is plenty */
114 	/* protocol dependent part */
115 	struct	in_endpoints inc_ie;
116 };
117 
118 /*
119  * Flags for inc_flags.
120  */
121 #define	INC_ISIPV6	0x01
122 #define	INC_IPV6MINMTU	0x02
123 
124 #define	inc_fport	inc_ie.ie_fport
125 #define	inc_lport	inc_ie.ie_lport
126 #define	inc_faddr	inc_ie.ie_faddr
127 #define	inc_laddr	inc_ie.ie_laddr
128 #define	inc6_faddr	inc_ie.ie6_faddr
129 #define	inc6_laddr	inc_ie.ie6_laddr
130 #define	inc6_zoneid	inc_ie.ie6_zoneid
131 
132 #if defined(_KERNEL) || defined(_WANT_INPCB)
133 /*
134  * struct inpcb captures the network layer state for TCP, UDP, and raw IPv4 and
135  * IPv6 sockets.  In the case of TCP and UDP, further per-connection state is
136  * hung off of inp_ppcb most of the time.  Almost all fields of struct inpcb
137  * are static after creation or protected by a per-inpcb rwlock, inp_lock.
138  *
139  * A inpcb database is indexed by addresses/ports hash as well as list of
140  * all pcbs that belong to a certain proto. Database lookups or list traversals
141  * are be performed inside SMR section. Once desired PCB is found its own
142  * lock is to be obtained and SMR section exited.
143  *
144  * Key:
145  * (c) - Constant after initialization
146  * (e) - Protected by the SMR section
147  * (i) - Protected by the inpcb lock
148  * (p) - Protected by the pcbinfo lock for the inpcb
149  * (h) - Protected by the pcbhash lock for the inpcb
150  * (s) - Protected by another subsystem's locks
151  * (x) - Undefined locking
152  *
153  * A few other notes:
154  *
155  * When a read lock is held, stability of the field is guaranteed; to write
156  * to a field, a write lock must generally be held.
157  *
158  * netinet/netinet6-layer code should not assume that the inp_socket pointer
159  * is safe to dereference without inp_lock being held, there may be
160  * close(2)-related races.
161  *
162  * The inp_vflag field is overloaded, and would otherwise ideally be (c).
163  */
164 struct icmp6_filter;
165 struct inpcbpolicy;
166 struct m_snd_tag;
167 struct inpcb {
168 	/* Cache line #1 (amd64) */
169 	CK_LIST_ENTRY(inpcb) inp_hash_exact;	/* hash table linkage */
170 	CK_LIST_ENTRY(inpcb) inp_hash_wild;	/* hash table linkage */
171 	struct rwlock	inp_lock;
172 	/* Cache line #2 (amd64) */
173 #define	inp_start_zero	inp_refcount
174 #define	inp_zero_size	(sizeof(struct inpcb) - \
175 			    offsetof(struct inpcb, inp_start_zero))
176 	u_int	inp_refcount;		/* (i) refcount */
177 	int	inp_flags;		/* (i) generic IP/datagram flags */
178 	int	inp_flags2;		/* (i) generic IP/datagram flags #2*/
179 	uint8_t inp_numa_domain;	/* numa domain */
180 	void	*inp_ppcb;		/* (i) pointer to per-protocol pcb */
181 	struct	socket *inp_socket;	/* (i) back pointer to socket */
182 	struct	inpcbinfo *inp_pcbinfo;	/* (c) PCB list info */
183 	struct	ucred	*inp_cred;	/* (c) cache of socket cred */
184 	u_int32_t inp_flow;		/* (i) IPv6 flow information */
185 	u_char	inp_vflag;		/* (i) IP version flag (v4/v6) */
186 	u_char	inp_ip_ttl;		/* (i) time to live proto */
187 	u_char	inp_ip_p;		/* (c) protocol proto */
188 	u_char	inp_ip_minttl;		/* (i) minimum TTL or drop */
189 	uint32_t inp_flowid;		/* (x) flow id / queue id */
190 	smr_seq_t inp_smr;		/* (i) sequence number at disconnect */
191 	struct m_snd_tag *inp_snd_tag;	/* (i) send tag for outgoing mbufs */
192 	uint32_t inp_flowtype;		/* (x) M_HASHTYPE value */
193 
194 	/* Local and foreign ports, local and foreign addr. */
195 	struct	in_conninfo inp_inc;	/* (i,h) list for PCB's local port */
196 
197 	/* MAC and IPSEC policy information. */
198 	struct	label *inp_label;	/* (i) MAC label */
199 	struct	inpcbpolicy *inp_sp;    /* (s) for IPSEC */
200 
201 	/* Protocol-dependent part; options. */
202 	struct {
203 		u_char	inp_ip_tos;		/* (i) type of service proto */
204 		struct mbuf		*inp_options;	/* (i) IP options */
205 		struct ip_moptions	*inp_moptions;	/* (i) mcast options */
206 	};
207 	struct {
208 		/* (i) IP options */
209 		struct mbuf		*in6p_options;
210 		/* (i) IP6 options for outgoing packets */
211 		struct ip6_pktopts	*in6p_outputopts;
212 		/* (i) IP multicast options */
213 		struct ip6_moptions	*in6p_moptions;
214 		/* (i) ICMPv6 code type filter */
215 		struct icmp6_filter	*in6p_icmp6filt;
216 		/* (i) IPV6_CHECKSUM setsockopt */
217 		int	in6p_cksum;
218 		short	in6p_hops;
219 	};
220 	CK_LIST_ENTRY(inpcb) inp_portlist;	/* (r:e/w:h) port list */
221 	struct	inpcbport *inp_phd;	/* (r:e/w:h) head of this list */
222 	inp_gen_t	inp_gencnt;	/* (c) generation count */
223 	void		*spare_ptr;	/* Spare pointer. */
224 	rt_gen_t	inp_rt_cookie;	/* generation for route entry */
225 	union {				/* cached L3 information */
226 		struct route inp_route;
227 		struct route_in6 inp_route6;
228 	};
229 	CK_LIST_ENTRY(inpcb) inp_list;	/* (r:e/w:p) all PCBs for proto */
230 };
231 #endif	/* _KERNEL */
232 
233 #define	inp_fport	inp_inc.inc_fport
234 #define	inp_lport	inp_inc.inc_lport
235 #define	inp_faddr	inp_inc.inc_faddr
236 #define	inp_laddr	inp_inc.inc_laddr
237 
238 #define	in6p_faddr	inp_inc.inc6_faddr
239 #define	in6p_laddr	inp_inc.inc6_laddr
240 #define	in6p_zoneid	inp_inc.inc6_zoneid
241 
242 #define	inp_vnet	inp_pcbinfo->ipi_vnet
243 
244 /*
245  * The range of the generation count, as used in this implementation, is 9e19.
246  * We would have to create 300 billion connections per second for this number
247  * to roll over in a year.  This seems sufficiently unlikely that we simply
248  * don't concern ourselves with that possibility.
249  */
250 
251 /*
252  * Interface exported to userland by various protocols which use inpcbs.  Hack
253  * alert -- only define if struct xsocket is in scope.
254  * Fields prefixed with "xi_" are unique to this structure, and the rest
255  * match fields in the struct inpcb, to ease coding and porting.
256  *
257  * Legend:
258  * (s) - used by userland utilities in src
259  * (p) - used by utilities in ports
260  * (3) - is known to be used by third party software not in ports
261  * (n) - no known usage
262  */
263 #ifdef _SYS_SOCKETVAR_H_
264 struct xinpcb {
265 	ksize_t		xi_len;			/* length of this structure */
266 	struct xsocket	xi_socket;		/* (s,p) */
267 	struct in_conninfo inp_inc;		/* (s,p) */
268 	uint64_t	inp_gencnt;		/* (s,p) */
269 	kvaddr_t	inp_ppcb;		/* (s) netstat(1) */
270 	int64_t		inp_spare64[4];
271 	uint32_t	inp_flow;		/* (s) */
272 	uint32_t	inp_flowid;		/* (s) */
273 	uint32_t	inp_flowtype;		/* (s) */
274 	int32_t		inp_flags;		/* (s,p) */
275 	int32_t		inp_flags2;		/* (s) */
276 	uint32_t	inp_unused;
277 	int32_t		in6p_cksum;		/* (n) */
278 	int32_t		inp_spare32[4];
279 	uint16_t	in6p_hops;		/* (n) */
280 	uint8_t		inp_ip_tos;		/* (n) */
281 	int8_t		pad8;
282 	uint8_t		inp_vflag;		/* (s,p) */
283 	uint8_t		inp_ip_ttl;		/* (n) */
284 	uint8_t		inp_ip_p;		/* (n) */
285 	uint8_t		inp_ip_minttl;		/* (n) */
286 	int8_t		inp_spare8[4];
287 } __aligned(8);
288 
289 struct xinpgen {
290 	ksize_t	xig_len;	/* length of this structure */
291 	u_int		xig_count;	/* number of PCBs at this time */
292 	uint32_t	_xig_spare32;
293 	inp_gen_t	xig_gen;	/* generation count at this time */
294 	so_gen_t	xig_sogen;	/* socket generation count this time */
295 	uint64_t	_xig_spare64[4];
296 } __aligned(8);
297 
298 struct sockopt_parameters {
299 	struct in_conninfo sop_inc;
300 	uint64_t sop_id;
301 	int sop_level;
302 	int sop_optname;
303 	char sop_optval[];
304 };
305 
306 #ifdef	_KERNEL
307 int	sysctl_setsockopt(SYSCTL_HANDLER_ARGS, struct inpcbinfo *pcbinfo,
308 	    int (*ctloutput_set)(struct inpcb *, struct sockopt *));
309 void	in_pcbtoxinpcb(const struct inpcb *, struct xinpcb *);
310 #endif
311 #endif /* _SYS_SOCKETVAR_H_ */
312 
313 #ifdef _KERNEL
314 /*
315  * Per-VNET pcb database for each high-level protocol (UDP, TCP, ...) in both
316  * IPv4 and IPv6.
317  *
318  * The pcbs are protected with SMR section and thus all lists in inpcbinfo
319  * are CK-lists.  Locking is required to insert a pcb into database. Two
320  * locks are provided: one for the hash and one for the global list of pcbs,
321  * as well as overall count and generation count.
322  *
323  * Locking key:
324  *
325  * (c) Constant or nearly constant after initialisation
326  * (e) Protected by SMR section
327  * (g) Locked by ipi_lock
328  * (h) Locked by ipi_hash_lock
329  */
330 struct inpcbinfo {
331 	/*
332 	 * Global lock protecting inpcb list modification
333 	 */
334 	struct mtx		 ipi_lock;
335 	struct inpcbhead	 ipi_listhead;		/* (r:e/w:g) */
336 	u_int			 ipi_count;		/* (g) */
337 
338 	/*
339 	 * Generation count -- incremented each time a connection is allocated
340 	 * or freed.
341 	 */
342 	u_quad_t		 ipi_gencnt;		/* (g) */
343 
344 	/*
345 	 * Fields associated with port lookup and allocation.
346 	 */
347 	u_short			 ipi_lastport;		/* (h) */
348 	u_short			 ipi_lastlow;		/* (h) */
349 	u_short			 ipi_lasthi;		/* (h) */
350 
351 	/*
352 	 * UMA zone from which inpcbs are allocated for this protocol.
353 	 */
354 	uma_zone_t		 ipi_zone;		/* (c) */
355 	uma_zone_t		 ipi_portzone;		/* (c) */
356 	smr_t			 ipi_smr;		/* (c) */
357 
358 	/*
359 	 * Global hash of inpcbs, hashed by local and foreign addresses and
360 	 * port numbers.  The "exact" hash holds PCBs connected to a foreign
361 	 * address, and "wild" holds the rest.
362 	 */
363 	struct mtx		 ipi_hash_lock;
364 	struct inpcbhead 	*ipi_hash_exact;	/* (r:e/w:h) */
365 	struct inpcbhead 	*ipi_hash_wild;		/* (r:e/w:h) */
366 	u_long			 ipi_hashmask;		/* (c) */
367 
368 	/*
369 	 * Global hash of inpcbs, hashed by only local port number.
370 	 */
371 	struct inpcbporthead	*ipi_porthashbase;	/* (h) */
372 	u_long			 ipi_porthashmask;	/* (h) */
373 
374 	/*
375 	 * Load balance groups used for the SO_REUSEPORT_LB option,
376 	 * hashed by local port.
377 	 */
378 	struct	inpcblbgrouphead *ipi_lbgrouphashbase;	/* (r:e/w:h) */
379 	u_long			 ipi_lbgrouphashmask;	/* (h) */
380 
381 	/*
382 	 * Pointer to network stack instance
383 	 */
384 	struct vnet		*ipi_vnet;		/* (c) */
385 };
386 
387 /*
388  * Global allocation storage for each high-level protocol (UDP, TCP, ...).
389  * Each corresponding per-VNET inpcbinfo points into this one.
390  */
391 struct inpcbstorage {
392 	uma_zone_t	ips_zone;
393 	uma_zone_t	ips_portzone;
394 	uma_init	ips_pcbinit;
395 	size_t		ips_size;
396 	const char *	ips_zone_name;
397 	const char *	ips_portzone_name;
398 	const char *	ips_infolock_name;
399 	const char *	ips_hashlock_name;
400 };
401 
402 #define INPCBSTORAGE_DEFINE(prot, ppcb, lname, zname, iname, hname)	\
403 static int								\
404 prot##_inpcb_init(void *mem, int size __unused, int flags __unused)	\
405 {									\
406 	struct inpcb *inp = mem;					\
407 									\
408 	rw_init_flags(&inp->inp_lock, lname, RW_RECURSE | RW_DUPOK);	\
409 	return (0);							\
410 }									\
411 static struct inpcbstorage prot = {					\
412 	.ips_size = sizeof(struct ppcb),				\
413 	.ips_pcbinit = prot##_inpcb_init,				\
414 	.ips_zone_name = zname,						\
415 	.ips_portzone_name = zname " ports",				\
416 	.ips_infolock_name = iname,					\
417 	.ips_hashlock_name = hname,					\
418 };									\
419 SYSINIT(prot##_inpcbstorage_init, SI_SUB_PROTO_DOMAIN,			\
420     SI_ORDER_SECOND, in_pcbstorage_init, &prot);			\
421 SYSUNINIT(prot##_inpcbstorage_uninit, SI_SUB_PROTO_DOMAIN,		\
422     SI_ORDER_SECOND, in_pcbstorage_destroy, &prot)
423 
424 /*
425  * Load balance groups used for the SO_REUSEPORT_LB socket option. Each group
426  * (or unique address:port combination) can be re-used at most
427  * INPCBLBGROUP_SIZMAX (256) times. The inpcbs are stored in il_inp which
428  * is dynamically resized as processes bind/unbind to that specific group.
429  */
430 struct inpcblbgroup {
431 	CK_LIST_ENTRY(inpcblbgroup) il_list;
432 	struct epoch_context il_epoch_ctx;
433 	struct ucred	*il_cred;
434 	uint16_t	il_lport;			/* (c) */
435 	u_char		il_vflag;			/* (c) */
436 	uint8_t		il_numa_domain;
437 	uint32_t	il_pad2;
438 	union in_dependaddr il_dependladdr;		/* (c) */
439 #define	il_laddr	il_dependladdr.id46_addr.ia46_addr4
440 #define	il6_laddr	il_dependladdr.id6_addr
441 	uint32_t	il_inpsiz; /* max count in il_inp[] (h) */
442 	uint32_t	il_inpcnt; /* cur count in il_inp[] (h) */
443 	struct inpcb	*il_inp[];			/* (h) */
444 };
445 
446 #define INP_LOCK_DESTROY(inp)	rw_destroy(&(inp)->inp_lock)
447 #define INP_RLOCK(inp)		rw_rlock(&(inp)->inp_lock)
448 #define INP_WLOCK(inp)		rw_wlock(&(inp)->inp_lock)
449 #define INP_TRY_RLOCK(inp)	rw_try_rlock(&(inp)->inp_lock)
450 #define INP_TRY_WLOCK(inp)	rw_try_wlock(&(inp)->inp_lock)
451 #define INP_RUNLOCK(inp)	rw_runlock(&(inp)->inp_lock)
452 #define INP_WUNLOCK(inp)	rw_wunlock(&(inp)->inp_lock)
453 #define INP_UNLOCK(inp)		rw_unlock(&(inp)->inp_lock)
454 #define	INP_TRY_UPGRADE(inp)	rw_try_upgrade(&(inp)->inp_lock)
455 #define	INP_DOWNGRADE(inp)	rw_downgrade(&(inp)->inp_lock)
456 #define	INP_WLOCKED(inp)	rw_wowned(&(inp)->inp_lock)
457 #define	INP_LOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_LOCKED)
458 #define	INP_RLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_RLOCKED)
459 #define	INP_WLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_WLOCKED)
460 #define	INP_UNLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_UNLOCKED)
461 
462 /*
463  * These locking functions are for inpcb consumers outside of sys/netinet,
464  * more specifically, they were added for the benefit of TOE drivers. The
465  * macros are reserved for use by the stack.
466  */
467 void inp_wlock(struct inpcb *);
468 void inp_wunlock(struct inpcb *);
469 void inp_rlock(struct inpcb *);
470 void inp_runlock(struct inpcb *);
471 
472 #ifdef INVARIANT_SUPPORT
473 void inp_lock_assert(struct inpcb *);
474 void inp_unlock_assert(struct inpcb *);
475 #else
476 #define	inp_lock_assert(inp)	do {} while (0)
477 #define	inp_unlock_assert(inp)	do {} while (0)
478 #endif
479 
480 void	inp_apply_all(struct inpcbinfo *, void (*func)(struct inpcb *, void *),
481 	    void *arg);
482 int 	inp_ip_tos_get(const struct inpcb *inp);
483 void 	inp_ip_tos_set(struct inpcb *inp, int val);
484 struct socket *
485 	inp_inpcbtosocket(struct inpcb *inp);
486 struct tcpcb *
487 	inp_inpcbtotcpcb(struct inpcb *inp);
488 void 	inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
489 		uint32_t *faddr, uint16_t *fp);
490 
491 #endif /* _KERNEL */
492 
493 #define INP_INFO_WLOCK(ipi)	mtx_lock(&(ipi)->ipi_lock)
494 #define INP_INFO_WLOCKED(ipi)	mtx_owned(&(ipi)->ipi_lock)
495 #define INP_INFO_WUNLOCK(ipi)	mtx_unlock(&(ipi)->ipi_lock)
496 #define	INP_INFO_LOCK_ASSERT(ipi)	MPASS(SMR_ENTERED((ipi)->ipi_smr) || \
497 					mtx_owned(&(ipi)->ipi_lock))
498 #define INP_INFO_WLOCK_ASSERT(ipi)	mtx_assert(&(ipi)->ipi_lock, MA_OWNED)
499 #define INP_INFO_WUNLOCK_ASSERT(ipi)	\
500 				mtx_assert(&(ipi)->ipi_lock, MA_NOTOWNED)
501 
502 #define	INP_HASH_WLOCK(ipi)		mtx_lock(&(ipi)->ipi_hash_lock)
503 #define	INP_HASH_WUNLOCK(ipi)		mtx_unlock(&(ipi)->ipi_hash_lock)
504 #define	INP_HASH_LOCK_ASSERT(ipi)	MPASS(SMR_ENTERED((ipi)->ipi_smr) || \
505 					mtx_owned(&(ipi)->ipi_hash_lock))
506 #define	INP_HASH_WLOCK_ASSERT(ipi)	mtx_assert(&(ipi)->ipi_hash_lock, \
507 					MA_OWNED)
508 
509 /*
510  * Wildcard matching hash is not just a microoptimisation!  The hash for
511  * wildcard IPv4 and wildcard IPv6 must be the same, otherwise AF_INET6
512  * wildcard bound pcb won't be able to receive AF_INET connections, while:
513  * jenkins_hash(&zeroes, 1, s) != jenkins_hash(&zeroes, 4, s)
514  * See also comment above struct in_addr_4in6.
515  */
516 #define	IN_ADDR_JHASH32(addr)						\
517 	((addr)->s_addr == INADDR_ANY ? V_in_pcbhashseed :		\
518 	    jenkins_hash32((&(addr)->s_addr), 1, V_in_pcbhashseed))
519 #define	IN6_ADDR_JHASH32(addr)						\
520 	(memcmp((addr), &in6addr_any, sizeof(in6addr_any)) == 0 ?	\
521 	    V_in_pcbhashseed :						\
522 	    jenkins_hash32((addr)->__u6_addr.__u6_addr32,		\
523 	    nitems((addr)->__u6_addr.__u6_addr32), V_in_pcbhashseed))
524 
525 #define INP_PCBHASH(faddr, lport, fport, mask)				\
526 	((IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask))
527 #define	INP6_PCBHASH(faddr, lport, fport, mask)				\
528 	((IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask))
529 
530 #define	INP_PCBHASH_WILD(lport, mask)					\
531 	((V_in_pcbhashseed ^ ntohs(lport)) & (mask))
532 
533 #define	INP_PCBLBGROUP_PKTHASH(faddr, lport, fport)			\
534 	(IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport)))
535 #define	INP6_PCBLBGROUP_PKTHASH(faddr, lport, fport)			\
536 	(IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport)))
537 
538 #define INP_PCBPORTHASH(lport, mask)	(ntohs((lport)) & (mask))
539 
540 /*
541  * Flags for inp_vflags -- historically version flags only
542  */
543 #define	INP_IPV4	0x1
544 #define	INP_IPV6	0x2
545 #define	INP_IPV6PROTO	0x4		/* opened under IPv6 protocol */
546 
547 /*
548  * Flags for inp_flags.
549  */
550 #define	INP_RECVOPTS		0x00000001 /* receive incoming IP options */
551 #define	INP_RECVRETOPTS		0x00000002 /* receive IP options for reply */
552 #define	INP_RECVDSTADDR		0x00000004 /* receive IP dst address */
553 #define	INP_HDRINCL		0x00000008 /* user supplies entire IP header */
554 #define	INP_HIGHPORT		0x00000010 /* user wants "high" port binding */
555 #define	INP_LOWPORT		0x00000020 /* user wants "low" port binding */
556 #define	INP_ANONPORT		0x00000040 /* read by netstat(1) */
557 #define	INP_RECVIF		0x00000080 /* receive incoming interface */
558 #define	INP_MTUDISC		0x00000100 /* user can do MTU discovery */
559 /*	INP_FREED		0x00000200 private to in_pcb.c */
560 #define	INP_RECVTTL		0x00000400 /* receive incoming IP TTL */
561 #define	INP_DONTFRAG		0x00000800 /* don't fragment packet */
562 #define	INP_BINDANY		0x00001000 /* allow bind to any address */
563 #define	INP_INHASHLIST		0x00002000 /* in_pcbinshash() has been called */
564 #define	INP_RECVTOS		0x00004000 /* receive incoming IP TOS */
565 #define	IN6P_IPV6_V6ONLY	0x00008000 /* restrict AF_INET6 socket for v6 */
566 #define	IN6P_PKTINFO		0x00010000 /* receive IP6 dst and I/F */
567 #define	IN6P_HOPLIMIT		0x00020000 /* receive hoplimit */
568 #define	IN6P_HOPOPTS		0x00040000 /* receive hop-by-hop options */
569 #define	IN6P_DSTOPTS		0x00080000 /* receive dst options after rthdr */
570 #define	IN6P_RTHDR		0x00100000 /* receive routing header */
571 #define	IN6P_RTHDRDSTOPTS	0x00200000 /* receive dstoptions before rthdr */
572 #define	IN6P_TCLASS		0x00400000 /* receive traffic class value */
573 #define	IN6P_AUTOFLOWLABEL	0x00800000 /* attach flowlabel automatically */
574 /*	INP_INLBGROUP		0x01000000 private to in_pcb.c */
575 #define	INP_ONESBCAST		0x02000000 /* send all-ones broadcast */
576 #define	INP_DROPPED		0x04000000 /* protocol drop flag */
577 #define	INP_SOCKREF		0x08000000 /* strong socket reference */
578 #define	INP_RESERVED_0          0x10000000 /* reserved field */
579 #define	INP_RESERVED_1          0x20000000 /* reserved field */
580 #define	IN6P_RFC2292		0x40000000 /* used RFC2292 API on the socket */
581 #define	IN6P_MTU		0x80000000 /* receive path MTU */
582 
583 #define	INP_CONTROLOPTS		(INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\
584 				 INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\
585 				 IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\
586 				 IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\
587 				 IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\
588 				 IN6P_MTU)
589 
590 /*
591  * Flags for inp_flags2.
592  */
593 /*				0x00000001 */
594 /*				0x00000002 */
595 /*				0x00000004 */
596 /*				0x00000008 */
597 /*				0x00000010 */
598 /*				0x00000020 */
599 /*				0x00000040 */
600 /*				0x00000080 */
601 #define	INP_RECVFLOWID		0x00000100 /* populate recv datagram with flow info */
602 #define	INP_RECVRSSBUCKETID	0x00000200 /* populate recv datagram with bucket id */
603 #define	INP_RATE_LIMIT_CHANGED	0x00000400 /* rate limit needs attention */
604 #define	INP_ORIGDSTADDR		0x00000800 /* receive IP dst address/port */
605 /*				0x00001000 */
606 /*				0x00002000 */
607 /*				0x00004000 */
608 /*				0x00008000 */
609 /*				0x00010000 */
610 #define INP_2PCP_SET		0x00020000 /* If the Eth PCP should be set explicitly */
611 #define INP_2PCP_BIT0		0x00040000 /* Eth PCP Bit 0 */
612 #define INP_2PCP_BIT1		0x00080000 /* Eth PCP Bit 1 */
613 #define INP_2PCP_BIT2		0x00100000 /* Eth PCP Bit 2 */
614 #define INP_2PCP_BASE	INP_2PCP_BIT0
615 #define INP_2PCP_MASK	(INP_2PCP_BIT0 | INP_2PCP_BIT1 | INP_2PCP_BIT2)
616 #define INP_2PCP_SHIFT		18         /* shift PCP field in/out of inp_flags2 */
617 
618 /*
619  * Flags passed to in_pcblookup*(), inp_smr_lock() and inp_next().
620  */
621 typedef	enum {
622 	INPLOOKUP_WILDCARD = 0x00000001,	/* Allow wildcard sockets. */
623 	INPLOOKUP_RLOCKPCB = 0x00000002,	/* Return inpcb read-locked. */
624 	INPLOOKUP_WLOCKPCB = 0x00000004,	/* Return inpcb write-locked. */
625 } inp_lookup_t;
626 
627 #define	INPLOOKUP_MASK	(INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \
628 	    INPLOOKUP_WLOCKPCB)
629 #define	INPLOOKUP_LOCKMASK	(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)
630 
631 #define	sotoinpcb(so)	((struct inpcb *)(so)->so_pcb)
632 
633 #define	INP_SOCKAF(so) so->so_proto->pr_domain->dom_family
634 
635 #define	INP_CHECK_SOCKAF(so, af)	(INP_SOCKAF(so) == af)
636 
637 #ifdef _KERNEL
638 VNET_DECLARE(int, ipport_reservedhigh);
639 VNET_DECLARE(int, ipport_reservedlow);
640 VNET_DECLARE(int, ipport_lowfirstauto);
641 VNET_DECLARE(int, ipport_lowlastauto);
642 VNET_DECLARE(int, ipport_firstauto);
643 VNET_DECLARE(int, ipport_lastauto);
644 VNET_DECLARE(int, ipport_hifirstauto);
645 VNET_DECLARE(int, ipport_hilastauto);
646 VNET_DECLARE(int, ipport_randomized);
647 
648 #define	V_ipport_reservedhigh	VNET(ipport_reservedhigh)
649 #define	V_ipport_reservedlow	VNET(ipport_reservedlow)
650 #define	V_ipport_lowfirstauto	VNET(ipport_lowfirstauto)
651 #define	V_ipport_lowlastauto	VNET(ipport_lowlastauto)
652 #define	V_ipport_firstauto	VNET(ipport_firstauto)
653 #define	V_ipport_lastauto	VNET(ipport_lastauto)
654 #define	V_ipport_hifirstauto	VNET(ipport_hifirstauto)
655 #define	V_ipport_hilastauto	VNET(ipport_hilastauto)
656 #define	V_ipport_randomized	VNET(ipport_randomized)
657 
658 void	in_pcbinfo_init(struct inpcbinfo *, struct inpcbstorage *,
659 	    u_int, u_int);
660 void	in_pcbinfo_destroy(struct inpcbinfo *);
661 void	in_pcbstorage_init(void *);
662 void	in_pcbstorage_destroy(void *);
663 
664 void	in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *);
665 int	in_pcballoc(struct socket *, struct inpcbinfo *);
666 int	in_pcbbind(struct inpcb *, struct sockaddr_in *, struct ucred *);
667 int	in_pcbbind_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *,
668 	    u_short *, struct ucred *);
669 int	in_pcbconnect(struct inpcb *, struct sockaddr_in *, struct ucred *,
670 	    bool);
671 int	in_pcbconnect_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *,
672 	    u_short *, in_addr_t *, u_short *, struct ucred *);
673 void	in_pcbdetach(struct inpcb *);
674 void	in_pcbdisconnect(struct inpcb *);
675 void	in_pcbdrop(struct inpcb *);
676 void	in_pcbfree(struct inpcb *);
677 int	in_pcbinshash(struct inpcb *);
678 int	in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *,
679 	    struct ucred *);
680 int	in_pcblbgroup_numa(struct inpcb *, int arg);
681 struct inpcb *
682 	in_pcblookup(struct inpcbinfo *, struct in_addr, u_int,
683 	    struct in_addr, u_int, int, struct ifnet *);
684 struct inpcb *
685 	in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int,
686 	    struct in_addr, u_int, int, struct ifnet *, struct mbuf *);
687 void	in_pcbref(struct inpcb *);
688 void	in_pcbrehash(struct inpcb *);
689 void	in_pcbremhash_locked(struct inpcb *);
690 bool	in_pcbrele(struct inpcb *, inp_lookup_t);
691 bool	in_pcbrele_rlocked(struct inpcb *);
692 bool	in_pcbrele_wlocked(struct inpcb *);
693 
694 typedef bool inp_match_t(const struct inpcb *, void *);
695 struct inpcb_iterator {
696 	const struct inpcbinfo	*ipi;
697 	struct inpcb		*inp;
698 	inp_match_t		*match;
699 	void			*ctx;
700 	int			hash;
701 #define	INP_ALL_LIST		-1
702 	const inp_lookup_t	lock;
703 };
704 
705 /* Note: sparse initializers guarantee .inp = NULL. */
706 #define	INP_ITERATOR(_ipi, _lock, _match, _ctx)		\
707 	{						\
708 		.ipi = (_ipi),				\
709 		.lock = (_lock),			\
710 		.hash = INP_ALL_LIST,			\
711 		.match = (_match),			\
712 		.ctx = (_ctx),				\
713 	}
714 #define	INP_ALL_ITERATOR(_ipi, _lock)			\
715 	{						\
716 		.ipi = (_ipi),				\
717 		.lock = (_lock),			\
718 		.hash = INP_ALL_LIST,			\
719 	}
720 
721 struct inpcb *inp_next(struct inpcb_iterator *);
722 void	in_losing(struct inpcb *);
723 void	in_pcbsetsolabel(struct socket *so);
724 int	in_getpeeraddr(struct socket *so, struct sockaddr **nam);
725 int	in_getsockaddr(struct socket *so, struct sockaddr **nam);
726 struct sockaddr *
727 	in_sockaddr(in_port_t port, struct in_addr *addr);
728 void	in_pcbsosetlabel(struct socket *so);
729 #ifdef RATELIMIT
730 int
731 in_pcboutput_txrtlmt_locked(struct inpcb *, struct ifnet *,
732 	    struct mbuf *, uint32_t);
733 int	in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t,
734 	    uint32_t, struct m_snd_tag **);
735 void	in_pcbdetach_txrtlmt(struct inpcb *);
736 void    in_pcbdetach_tag(struct m_snd_tag *);
737 int	in_pcbmodify_txrtlmt(struct inpcb *, uint32_t);
738 int	in_pcbquery_txrtlmt(struct inpcb *, uint32_t *);
739 int	in_pcbquery_txrlevel(struct inpcb *, uint32_t *);
740 void	in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *);
741 void	in_pcboutput_eagain(struct inpcb *);
742 #endif
743 #endif /* _KERNEL */
744 
745 #endif /* !_NETINET_IN_PCB_H_ */
746