xref: /freebsd/sys/netinet/in_pcb.h (revision ca94f92c23fd09b28ac3398657ae2ae9367bcdf5)
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  * located in a larger protocol specific structure that embeds inpcb in it.
137  * Almost all fields of struct inpcb are static after creation or protected by
138  * a per-inpcb rwlock, inp_lock.
139  *
140  * A inpcb database is indexed by addresses/ports hash as well as list of
141  * all pcbs that belong to a certain proto. Database lookups or list traversals
142  * are be performed inside SMR section. Once desired PCB is found its own
143  * lock is to be obtained and SMR section exited.
144  *
145  * Key:
146  * (c) - Constant after initialization
147  * (e) - Protected by the SMR section
148  * (i) - Protected by the inpcb lock
149  * (p) - Protected by the pcbinfo lock for the inpcb
150  * (h) - Protected by the pcbhash lock for the inpcb
151  * (s) - Protected by another subsystem's locks
152  * (x) - Undefined locking
153  *
154  * A few other notes:
155  *
156  * When a read lock is held, stability of the field is guaranteed; to write
157  * to a field, a write lock must generally be held.
158  *
159  * netinet/netinet6-layer code should not assume that the inp_socket pointer
160  * is safe to dereference without inp_lock being held, there may be
161  * close(2)-related races.
162  *
163  * The inp_vflag field is overloaded, and would otherwise ideally be (c).
164  */
165 struct icmp6_filter;
166 struct inpcbpolicy;
167 struct m_snd_tag;
168 struct inpcb {
169 	/* Cache line #1 (amd64) */
170 	union {
171 		CK_LIST_ENTRY(inpcb) inp_hash_exact;	/* hash table linkage */
172 		LIST_ENTRY(inpcb) inp_lbgroup_list;	/* lb group list */
173 	};
174 	CK_LIST_ENTRY(inpcb) inp_hash_wild;	/* hash table linkage */
175 	struct rwlock	inp_lock;
176 	/* Cache line #2 (amd64) */
177 #define	inp_start_zero	inp_refcount
178 #define	inp_zero_size	(sizeof(struct inpcb) - \
179 			    offsetof(struct inpcb, inp_start_zero))
180 	u_int	inp_refcount;		/* (i) refcount */
181 	int	inp_flags;		/* (i) generic IP/datagram flags */
182 	int	inp_flags2;		/* (i) generic IP/datagram flags #2*/
183 	uint8_t inp_numa_domain;	/* numa domain */
184 	struct	socket *inp_socket;	/* (i) back pointer to socket */
185 	struct	inpcbinfo *inp_pcbinfo;	/* (c) PCB list info */
186 	struct	ucred	*inp_cred;	/* (c) cache of socket cred */
187 	u_int32_t inp_flow;		/* (i) IPv6 flow information */
188 	u_char	inp_vflag;		/* (i) IP version flag (v4/v6) */
189 	u_char	inp_ip_ttl;		/* (i) time to live proto */
190 	u_char	inp_ip_p;		/* (c) protocol proto */
191 	u_char	inp_ip_minttl;		/* (i) minimum TTL or drop */
192 	uint32_t inp_flowid;		/* (x) flow id / queue id */
193 	smr_seq_t inp_smr;		/* (i) sequence number at disconnect */
194 	struct m_snd_tag *inp_snd_tag;	/* (i) send tag for outgoing mbufs */
195 	uint32_t inp_flowtype;		/* (x) M_HASHTYPE value */
196 
197 	/* Local and foreign ports, local and foreign addr. */
198 	struct	in_conninfo inp_inc;	/* (i,h) list for PCB's local port */
199 
200 	/* MAC and IPSEC policy information. */
201 	struct	label *inp_label;	/* (i) MAC label */
202 	struct	inpcbpolicy *inp_sp;    /* (s) for IPSEC */
203 
204 	/* Protocol-dependent part; options. */
205 	struct {
206 		u_char	inp_ip_tos;		/* (i) type of service proto */
207 		struct mbuf		*inp_options;	/* (i) IP options */
208 		struct ip_moptions	*inp_moptions;	/* (i) mcast options */
209 	};
210 	struct {
211 		/* (i) IP options */
212 		struct mbuf		*in6p_options;
213 		/* (i) IP6 options for outgoing packets */
214 		struct ip6_pktopts	*in6p_outputopts;
215 		/* (i) IP multicast options */
216 		struct ip6_moptions	*in6p_moptions;
217 		/* (i) ICMPv6 code type filter */
218 		struct icmp6_filter	*in6p_icmp6filt;
219 		/* (i) IPV6_CHECKSUM setsockopt */
220 		int	in6p_cksum;
221 		short	in6p_hops;
222 	};
223 	CK_LIST_ENTRY(inpcb) inp_portlist;	/* (r:e/w:h) port list */
224 	struct	inpcbport *inp_phd;	/* (r:e/w:h) head of this list */
225 	inp_gen_t	inp_gencnt;	/* (c) generation count */
226 	void		*spare_ptr;	/* Spare pointer. */
227 	rt_gen_t	inp_rt_cookie;	/* generation for route entry */
228 	union {				/* cached L3 information */
229 		struct route inp_route;
230 		struct route_in6 inp_route6;
231 	};
232 	CK_LIST_ENTRY(inpcb) inp_list;	/* (r:e/w:p) all PCBs for proto */
233 };
234 #endif	/* _KERNEL */
235 
236 #define	inp_fport	inp_inc.inc_fport
237 #define	inp_lport	inp_inc.inc_lport
238 #define	inp_faddr	inp_inc.inc_faddr
239 #define	inp_laddr	inp_inc.inc_laddr
240 
241 #define	in6p_faddr	inp_inc.inc6_faddr
242 #define	in6p_laddr	inp_inc.inc6_laddr
243 #define	in6p_zoneid	inp_inc.inc6_zoneid
244 
245 #define	inp_vnet	inp_pcbinfo->ipi_vnet
246 
247 /*
248  * The range of the generation count, as used in this implementation, is 9e19.
249  * We would have to create 300 billion connections per second for this number
250  * to roll over in a year.  This seems sufficiently unlikely that we simply
251  * don't concern ourselves with that possibility.
252  */
253 
254 /*
255  * Interface exported to userland by various protocols which use inpcbs.  Hack
256  * alert -- only define if struct xsocket is in scope.
257  * Fields prefixed with "xi_" are unique to this structure, and the rest
258  * match fields in the struct inpcb, to ease coding and porting.
259  *
260  * Legend:
261  * (s) - used by userland utilities in src
262  * (p) - used by utilities in ports
263  * (3) - is known to be used by third party software not in ports
264  * (n) - no known usage
265  */
266 #ifdef _SYS_SOCKETVAR_H_
267 struct xinpcb {
268 	ksize_t		xi_len;			/* length of this structure */
269 	struct xsocket	xi_socket;		/* (s,p) */
270 	struct in_conninfo inp_inc;		/* (s,p) */
271 	uint64_t	inp_gencnt;		/* (s,p) */
272 	int64_t		inp_spare64[5];
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 
291 struct 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 
300 struct 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
309 int	sysctl_setsockopt(SYSCTL_HANDLER_ARGS, struct inpcbinfo *pcbinfo,
310 	    int (*ctloutput_set)(struct inpcb *, struct sockopt *));
311 void	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  */
332 struct 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  */
393 struct 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)	\
405 static int								\
406 prot##_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 }									\
413 static 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 };									\
421 SYSINIT(prot##_inpcbstorage_init, SI_SUB_PROTO_DOMAIN,			\
422     SI_ORDER_SECOND, in_pcbstorage_init, &prot);			\
423 SYSUNINIT(prot##_inpcbstorage_uninit, SI_SUB_PROTO_DOMAIN,		\
424     SI_ORDER_SECOND, in_pcbstorage_destroy, &prot)
425 
426 #define INP_LOCK_DESTROY(inp)	rw_destroy(&(inp)->inp_lock)
427 #define INP_RLOCK(inp)		rw_rlock(&(inp)->inp_lock)
428 #define INP_WLOCK(inp)		rw_wlock(&(inp)->inp_lock)
429 #define INP_TRY_RLOCK(inp)	rw_try_rlock(&(inp)->inp_lock)
430 #define INP_TRY_WLOCK(inp)	rw_try_wlock(&(inp)->inp_lock)
431 #define INP_RUNLOCK(inp)	rw_runlock(&(inp)->inp_lock)
432 #define INP_WUNLOCK(inp)	rw_wunlock(&(inp)->inp_lock)
433 #define INP_UNLOCK(inp)		rw_unlock(&(inp)->inp_lock)
434 #define	INP_TRY_UPGRADE(inp)	rw_try_upgrade(&(inp)->inp_lock)
435 #define	INP_DOWNGRADE(inp)	rw_downgrade(&(inp)->inp_lock)
436 #define	INP_WLOCKED(inp)	rw_wowned(&(inp)->inp_lock)
437 #define	INP_LOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_LOCKED)
438 #define	INP_RLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_RLOCKED)
439 #define	INP_WLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_WLOCKED)
440 #define	INP_UNLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_UNLOCKED)
441 
442 /*
443  * These locking functions are for inpcb consumers outside of sys/netinet,
444  * more specifically, they were added for the benefit of TOE drivers. The
445  * macros are reserved for use by the stack.
446  */
447 void inp_wlock(struct inpcb *);
448 void inp_wunlock(struct inpcb *);
449 void inp_rlock(struct inpcb *);
450 void inp_runlock(struct inpcb *);
451 
452 #ifdef INVARIANT_SUPPORT
453 void inp_lock_assert(struct inpcb *);
454 void inp_unlock_assert(struct inpcb *);
455 #else
456 #define	inp_lock_assert(inp)	do {} while (0)
457 #define	inp_unlock_assert(inp)	do {} while (0)
458 #endif
459 
460 void	inp_apply_all(struct inpcbinfo *, void (*func)(struct inpcb *, void *),
461 	    void *arg);
462 struct socket *
463 	inp_inpcbtosocket(struct inpcb *inp);
464 void 	inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
465 		uint32_t *faddr, uint16_t *fp);
466 
467 #endif /* _KERNEL */
468 
469 #define INP_INFO_WLOCK(ipi)	mtx_lock(&(ipi)->ipi_lock)
470 #define INP_INFO_WLOCKED(ipi)	mtx_owned(&(ipi)->ipi_lock)
471 #define INP_INFO_WUNLOCK(ipi)	mtx_unlock(&(ipi)->ipi_lock)
472 #define	INP_INFO_LOCK_ASSERT(ipi)	MPASS(SMR_ENTERED((ipi)->ipi_smr) || \
473 					mtx_owned(&(ipi)->ipi_lock))
474 #define INP_INFO_WLOCK_ASSERT(ipi)	mtx_assert(&(ipi)->ipi_lock, MA_OWNED)
475 #define INP_INFO_WUNLOCK_ASSERT(ipi)	\
476 				mtx_assert(&(ipi)->ipi_lock, MA_NOTOWNED)
477 
478 #define	INP_HASH_WLOCK(ipi)		mtx_lock(&(ipi)->ipi_hash_lock)
479 #define	INP_HASH_WUNLOCK(ipi)		mtx_unlock(&(ipi)->ipi_hash_lock)
480 #define	INP_HASH_LOCK_ASSERT(ipi)	MPASS(SMR_ENTERED((ipi)->ipi_smr) || \
481 					mtx_owned(&(ipi)->ipi_hash_lock))
482 #define	INP_HASH_WLOCK_ASSERT(ipi)	mtx_assert(&(ipi)->ipi_hash_lock, \
483 					MA_OWNED)
484 
485 /*
486  * Wildcard matching hash is not just a microoptimisation!  The hash for
487  * wildcard IPv4 and wildcard IPv6 must be the same, otherwise AF_INET6
488  * wildcard bound pcb won't be able to receive AF_INET connections, while:
489  * jenkins_hash(&zeroes, 1, s) != jenkins_hash(&zeroes, 4, s)
490  * See also comment above struct in_addr_4in6.
491  */
492 #define	IN_ADDR_JHASH32(addr)						\
493 	((addr)->s_addr == INADDR_ANY ? V_in_pcbhashseed :		\
494 	    jenkins_hash32((&(addr)->s_addr), 1, V_in_pcbhashseed))
495 #define	IN6_ADDR_JHASH32(addr)						\
496 	(memcmp((addr), &in6addr_any, sizeof(in6addr_any)) == 0 ?	\
497 	    V_in_pcbhashseed :						\
498 	    jenkins_hash32((addr)->__u6_addr.__u6_addr32,		\
499 	    nitems((addr)->__u6_addr.__u6_addr32), V_in_pcbhashseed))
500 
501 #define INP_PCBHASH(faddr, lport, fport, mask)				\
502 	((IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask))
503 #define	INP6_PCBHASH(faddr, lport, fport, mask)				\
504 	((IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask))
505 
506 #define	INP_PCBHASH_WILD(lport, mask)					\
507 	((V_in_pcbhashseed ^ ntohs(lport)) & (mask))
508 
509 #define	INP_PCBLBGROUP_PKTHASH(faddr, lport, fport)			\
510 	(IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport)))
511 #define	INP6_PCBLBGROUP_PKTHASH(faddr, lport, fport)			\
512 	(IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport)))
513 
514 #define INP_PCBPORTHASH(lport, mask)	(ntohs((lport)) & (mask))
515 
516 /*
517  * Flags for inp_vflags -- historically version flags only
518  */
519 #define	INP_IPV4	0x1
520 #define	INP_IPV6	0x2
521 #define	INP_IPV6PROTO	0x4		/* opened under IPv6 protocol */
522 
523 /*
524  * Flags for inp_flags.
525  */
526 #define	INP_RECVOPTS		0x00000001 /* receive incoming IP options */
527 #define	INP_RECVRETOPTS		0x00000002 /* receive IP options for reply */
528 #define	INP_RECVDSTADDR		0x00000004 /* receive IP dst address */
529 #define	INP_HDRINCL		0x00000008 /* user supplies entire IP header */
530 #define	INP_HIGHPORT		0x00000010 /* user wants "high" port binding */
531 #define	INP_LOWPORT		0x00000020 /* user wants "low" port binding */
532 #define	INP_ANONPORT		0x00000040 /* read by netstat(1) */
533 #define	INP_RECVIF		0x00000080 /* receive incoming interface */
534 #define	INP_MTUDISC		0x00000100 /* user can do MTU discovery */
535 /*	INP_FREED		0x00000200 private to in_pcb.c */
536 #define	INP_RECVTTL		0x00000400 /* receive incoming IP TTL */
537 #define	INP_DONTFRAG		0x00000800 /* don't fragment packet */
538 #define	INP_BINDANY		0x00001000 /* allow bind to any address */
539 #define	INP_INHASHLIST		0x00002000 /* in_pcbinshash() has been called */
540 #define	INP_RECVTOS		0x00004000 /* receive incoming IP TOS */
541 #define	IN6P_IPV6_V6ONLY	0x00008000 /* restrict AF_INET6 socket for v6 */
542 #define	IN6P_PKTINFO		0x00010000 /* receive IP6 dst and I/F */
543 #define	IN6P_HOPLIMIT		0x00020000 /* receive hoplimit */
544 #define	IN6P_HOPOPTS		0x00040000 /* receive hop-by-hop options */
545 #define	IN6P_DSTOPTS		0x00080000 /* receive dst options after rthdr */
546 #define	IN6P_RTHDR		0x00100000 /* receive routing header */
547 #define	IN6P_RTHDRDSTOPTS	0x00200000 /* receive dstoptions before rthdr */
548 #define	IN6P_TCLASS		0x00400000 /* receive traffic class value */
549 #define	IN6P_AUTOFLOWLABEL	0x00800000 /* attach flowlabel automatically */
550 /*	INP_INLBGROUP		0x01000000 private to in_pcb.c */
551 #define	INP_ONESBCAST		0x02000000 /* send all-ones broadcast */
552 #define	INP_DROPPED		0x04000000 /* protocol drop flag */
553 #define	INP_SOCKREF		0x08000000 /* strong socket reference */
554 #define	INP_RESERVED_0          0x10000000 /* reserved field */
555 #define	INP_BOUNDFIB		0x20000000 /* Bound to a specific FIB. */
556 #define	IN6P_RFC2292		0x40000000 /* used RFC2292 API on the socket */
557 #define	IN6P_MTU		0x80000000 /* receive path MTU */
558 
559 #define	INP_CONTROLOPTS		(INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\
560 				 INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\
561 				 IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\
562 				 IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\
563 				 IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\
564 				 IN6P_MTU)
565 
566 /*
567  * Flags for inp_flags2.
568  */
569 /*				0x00000001 */
570 /*				0x00000002 */
571 /*				0x00000004 */
572 /*				0x00000008 */
573 /*				0x00000010 */
574 /*				0x00000020 */
575 /*				0x00000040 */
576 /*				0x00000080 */
577 #define	INP_RECVFLOWID		0x00000100 /* populate recv datagram with flow info */
578 #define	INP_RECVRSSBUCKETID	0x00000200 /* populate recv datagram with bucket id */
579 #define	INP_RATE_LIMIT_CHANGED	0x00000400 /* rate limit needs attention */
580 #define	INP_ORIGDSTADDR		0x00000800 /* receive IP dst address/port */
581 /*				0x00001000 */
582 /*				0x00002000 */
583 /*				0x00004000 */
584 /*				0x00008000 */
585 /*				0x00010000 */
586 #define INP_2PCP_SET		0x00020000 /* If the Eth PCP should be set explicitly */
587 #define INP_2PCP_BIT0		0x00040000 /* Eth PCP Bit 0 */
588 #define INP_2PCP_BIT1		0x00080000 /* Eth PCP Bit 1 */
589 #define INP_2PCP_BIT2		0x00100000 /* Eth PCP Bit 2 */
590 #define INP_2PCP_BASE	INP_2PCP_BIT0
591 #define INP_2PCP_MASK	(INP_2PCP_BIT0 | INP_2PCP_BIT1 | INP_2PCP_BIT2)
592 #define INP_2PCP_SHIFT		18         /* shift PCP field in/out of inp_flags2 */
593 
594 /*
595  * Flags passed to in_pcblookup*(), inp_smr_lock() and inp_next().
596  */
597 typedef	enum {
598 	INPLOOKUP_WILDCARD = 0x00000001,	/* Allow wildcard sockets. */
599 	INPLOOKUP_RLOCKPCB = 0x00000002,	/* Return inpcb read-locked. */
600 	INPLOOKUP_WLOCKPCB = 0x00000004,	/* Return inpcb write-locked. */
601 	INPLOOKUP_FIB = 0x00000008,		/* inp must be from same FIB. */
602 } inp_lookup_t;
603 
604 #define	INPLOOKUP_MASK	(INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \
605 	    INPLOOKUP_WLOCKPCB | INPLOOKUP_FIB)
606 #define	INPLOOKUP_LOCKMASK	(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)
607 
608 #define	sotoinpcb(so)	((struct inpcb *)(so)->so_pcb)
609 
610 #define	INP_SOCKAF(so) so->so_proto->pr_domain->dom_family
611 
612 #define	INP_CHECK_SOCKAF(so, af)	(INP_SOCKAF(so) == af)
613 
614 #ifdef _KERNEL
615 VNET_DECLARE(int, ipport_reservedhigh);
616 VNET_DECLARE(int, ipport_reservedlow);
617 VNET_DECLARE(int, ipport_lowfirstauto);
618 VNET_DECLARE(int, ipport_lowlastauto);
619 VNET_DECLARE(int, ipport_firstauto);
620 VNET_DECLARE(int, ipport_lastauto);
621 VNET_DECLARE(int, ipport_hifirstauto);
622 VNET_DECLARE(int, ipport_hilastauto);
623 VNET_DECLARE(int, ipport_randomized);
624 
625 #define	V_ipport_reservedhigh	VNET(ipport_reservedhigh)
626 #define	V_ipport_reservedlow	VNET(ipport_reservedlow)
627 #define	V_ipport_lowfirstauto	VNET(ipport_lowfirstauto)
628 #define	V_ipport_lowlastauto	VNET(ipport_lowlastauto)
629 #define	V_ipport_firstauto	VNET(ipport_firstauto)
630 #define	V_ipport_lastauto	VNET(ipport_lastauto)
631 #define	V_ipport_hifirstauto	VNET(ipport_hifirstauto)
632 #define	V_ipport_hilastauto	VNET(ipport_hilastauto)
633 #define	V_ipport_randomized	VNET(ipport_randomized)
634 
635 void	in_pcbinfo_init(struct inpcbinfo *, struct inpcbstorage *,
636 	    u_int, u_int);
637 void	in_pcbinfo_destroy(struct inpcbinfo *);
638 void	in_pcbstorage_init(void *);
639 void	in_pcbstorage_destroy(void *);
640 
641 void	in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *);
642 int	in_pcballoc(struct socket *, struct inpcbinfo *);
643 #define	INPBIND_FIB	0x0001	/* bind to the PCB's FIB only */
644 int	in_pcbbind(struct inpcb *, struct sockaddr_in *, int, struct ucred *);
645 int	in_pcbbind_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *,
646 	    u_short *, int, struct ucred *);
647 int	in_pcbconnect(struct inpcb *, struct sockaddr_in *, struct ucred *);
648 int	in_pcbconnect_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *,
649 	    u_short *, in_addr_t *, u_short *, struct ucred *);
650 void	in_pcbdisconnect(struct inpcb *);
651 void	in_pcbdrop(struct inpcb *);
652 void	in_pcbfree(struct inpcb *);
653 int	in_pcbinshash(struct inpcb *);
654 int	in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *,
655 	    struct ucred *);
656 int	in_pcblbgroup_numa(struct inpcb *, int arg);
657 void	in_pcblisten(struct inpcb *);
658 struct inpcb *
659 	in_pcblookup(struct inpcbinfo *, struct in_addr, u_int,
660 	    struct in_addr, u_int, int, struct ifnet *);
661 struct inpcb *
662 	in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int,
663 	    struct in_addr, u_int, int, struct ifnet *, struct mbuf *);
664 void	in_pcbref(struct inpcb *);
665 void	in_pcbrehash(struct inpcb *);
666 void	in_pcbremhash_locked(struct inpcb *);
667 bool	in_pcbrele(struct inpcb *, inp_lookup_t);
668 bool	in_pcbrele_rlocked(struct inpcb *);
669 bool	in_pcbrele_wlocked(struct inpcb *);
670 
671 typedef bool inp_match_t(const struct inpcb *, void *);
672 struct inpcb_iterator {
673 	const struct inpcbinfo	*ipi;
674 	struct inpcb		*inp;
675 	inp_match_t		*match;
676 	void			*ctx;
677 	int			hash;
678 #define	INP_ALL_LIST		-1
679 	const inp_lookup_t	lock;
680 };
681 
682 /* Note: sparse initializers guarantee .inp = NULL. */
683 #define	INP_ITERATOR(_ipi, _lock, _match, _ctx)		\
684 	{						\
685 		.ipi = (_ipi),				\
686 		.lock = (_lock),			\
687 		.hash = INP_ALL_LIST,			\
688 		.match = (_match),			\
689 		.ctx = (_ctx),				\
690 	}
691 #define	INP_ALL_ITERATOR(_ipi, _lock)			\
692 	{						\
693 		.ipi = (_ipi),				\
694 		.lock = (_lock),			\
695 		.hash = INP_ALL_LIST,			\
696 	}
697 
698 struct inpcb *inp_next(struct inpcb_iterator *);
699 void	in_losing(struct inpcb *);
700 void	in_pcbsetsolabel(struct socket *so);
701 int	in_getpeeraddr(struct socket *, struct sockaddr *sa);
702 int	in_getsockaddr(struct socket *, struct sockaddr *sa);
703 void	in_pcbsosetlabel(struct socket *so);
704 #ifdef RATELIMIT
705 int
706 in_pcboutput_txrtlmt_locked(struct inpcb *, struct ifnet *,
707 	    struct mbuf *, uint32_t);
708 int	in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t,
709 	    uint32_t, struct m_snd_tag **);
710 void	in_pcbdetach_txrtlmt(struct inpcb *);
711 void    in_pcbdetach_tag(struct m_snd_tag *);
712 int	in_pcbmodify_txrtlmt(struct inpcb *, uint32_t);
713 int	in_pcbquery_txrtlmt(struct inpcb *, uint32_t *);
714 int	in_pcbquery_txrlevel(struct inpcb *, uint32_t *);
715 void	in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *);
716 void	in_pcboutput_eagain(struct inpcb *);
717 #endif
718 #endif /* _KERNEL */
719 
720 #endif /* !_NETINET_IN_PCB_H_ */
721